xref: /openbmc/qemu/qemu-img.c (revision 39f6049e)
1 /*
2  * QEMU disk image utility
3  *
4  * Copyright (c) 2003-2008 Fabrice Bellard
5  *
6  * Permission is hereby granted, free of charge, to any person obtaining a copy
7  * of this software and associated documentation files (the "Software"), to deal
8  * in the Software without restriction, including without limitation the rights
9  * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
10  * copies of the Software, and to permit persons to whom the Software is
11  * furnished to do so, subject to the following conditions:
12  *
13  * The above copyright notice and this permission notice shall be included in
14  * all copies or substantial portions of the Software.
15  *
16  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
17  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
18  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
19  * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
20  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
21  * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
22  * THE SOFTWARE.
23  */
24 
25 #include "qemu/osdep.h"
26 #include <getopt.h>
27 
28 #include "qemu-common.h"
29 #include "qemu-version.h"
30 #include "qapi/error.h"
31 #include "qapi/qapi-commands-block-core.h"
32 #include "qapi/qapi-visit-block-core.h"
33 #include "qapi/qobject-output-visitor.h"
34 #include "qapi/qmp/qjson.h"
35 #include "qapi/qmp/qdict.h"
36 #include "qemu/cutils.h"
37 #include "qemu/config-file.h"
38 #include "qemu/option.h"
39 #include "qemu/error-report.h"
40 #include "qemu/log.h"
41 #include "qemu/main-loop.h"
42 #include "qemu/module.h"
43 #include "qemu/sockets.h"
44 #include "qemu/units.h"
45 #include "qom/object_interfaces.h"
46 #include "sysemu/block-backend.h"
47 #include "block/block_int.h"
48 #include "block/blockjob.h"
49 #include "block/qapi.h"
50 #include "crypto/init.h"
51 #include "trace/control.h"
52 #include "qemu/throttle.h"
53 #include "block/throttle-groups.h"
54 
55 #define QEMU_IMG_VERSION "qemu-img version " QEMU_FULL_VERSION \
56                           "\n" QEMU_COPYRIGHT "\n"
57 
58 typedef struct img_cmd_t {
59     const char *name;
60     int (*handler)(int argc, char **argv);
61 } img_cmd_t;
62 
63 enum {
64     OPTION_OUTPUT = 256,
65     OPTION_BACKING_CHAIN = 257,
66     OPTION_OBJECT = 258,
67     OPTION_IMAGE_OPTS = 259,
68     OPTION_PATTERN = 260,
69     OPTION_FLUSH_INTERVAL = 261,
70     OPTION_NO_DRAIN = 262,
71     OPTION_TARGET_IMAGE_OPTS = 263,
72     OPTION_SIZE = 264,
73     OPTION_PREALLOCATION = 265,
74     OPTION_SHRINK = 266,
75     OPTION_SALVAGE = 267,
76     OPTION_TARGET_IS_ZERO = 268,
77     OPTION_ADD = 269,
78     OPTION_REMOVE = 270,
79     OPTION_CLEAR = 271,
80     OPTION_ENABLE = 272,
81     OPTION_DISABLE = 273,
82     OPTION_MERGE = 274,
83     OPTION_BITMAPS = 275,
84     OPTION_FORCE = 276,
85 };
86 
87 typedef enum OutputFormat {
88     OFORMAT_JSON,
89     OFORMAT_HUMAN,
90 } OutputFormat;
91 
92 /* Default to cache=writeback as data integrity is not important for qemu-img */
93 #define BDRV_DEFAULT_CACHE "writeback"
94 
95 static void format_print(void *opaque, const char *name)
96 {
97     printf(" %s", name);
98 }
99 
100 static void QEMU_NORETURN GCC_FMT_ATTR(1, 2) error_exit(const char *fmt, ...)
101 {
102     va_list ap;
103 
104     va_start(ap, fmt);
105     error_vreport(fmt, ap);
106     va_end(ap);
107 
108     error_printf("Try 'qemu-img --help' for more information\n");
109     exit(EXIT_FAILURE);
110 }
111 
112 static void QEMU_NORETURN missing_argument(const char *option)
113 {
114     error_exit("missing argument for option '%s'", option);
115 }
116 
117 static void QEMU_NORETURN unrecognized_option(const char *option)
118 {
119     error_exit("unrecognized option '%s'", option);
120 }
121 
122 /* Please keep in synch with docs/tools/qemu-img.rst */
123 static void QEMU_NORETURN help(void)
124 {
125     const char *help_msg =
126            QEMU_IMG_VERSION
127            "usage: qemu-img [standard options] command [command options]\n"
128            "QEMU disk image utility\n"
129            "\n"
130            "    '-h', '--help'       display this help and exit\n"
131            "    '-V', '--version'    output version information and exit\n"
132            "    '-T', '--trace'      [[enable=]<pattern>][,events=<file>][,file=<file>]\n"
133            "                         specify tracing options\n"
134            "\n"
135            "Command syntax:\n"
136 #define DEF(option, callback, arg_string)        \
137            "  " arg_string "\n"
138 #include "qemu-img-cmds.h"
139 #undef DEF
140            "\n"
141            "Command parameters:\n"
142            "  'filename' is a disk image filename\n"
143            "  'objectdef' is a QEMU user creatable object definition. See the qemu(1)\n"
144            "    manual page for a description of the object properties. The most common\n"
145            "    object type is a 'secret', which is used to supply passwords and/or\n"
146            "    encryption keys.\n"
147            "  'fmt' is the disk image format. It is guessed automatically in most cases\n"
148            "  'cache' is the cache mode used to write the output disk image, the valid\n"
149            "    options are: 'none', 'writeback' (default, except for convert), 'writethrough',\n"
150            "    'directsync' and 'unsafe' (default for convert)\n"
151            "  'src_cache' is the cache mode used to read input disk images, the valid\n"
152            "    options are the same as for the 'cache' option\n"
153            "  'size' is the disk image size in bytes. Optional suffixes\n"
154            "    'k' or 'K' (kilobyte, 1024), 'M' (megabyte, 1024k), 'G' (gigabyte, 1024M),\n"
155            "    'T' (terabyte, 1024G), 'P' (petabyte, 1024T) and 'E' (exabyte, 1024P)  are\n"
156            "    supported. 'b' is ignored.\n"
157            "  'output_filename' is the destination disk image filename\n"
158            "  'output_fmt' is the destination format\n"
159            "  'options' is a comma separated list of format specific options in a\n"
160            "    name=value format. Use -o ? for an overview of the options supported by the\n"
161            "    used format\n"
162            "  'snapshot_param' is param used for internal snapshot, format\n"
163            "    is 'snapshot.id=[ID],snapshot.name=[NAME]', or\n"
164            "    '[ID_OR_NAME]'\n"
165            "  '-c' indicates that target image must be compressed (qcow format only)\n"
166            "  '-u' allows unsafe backing chains. For rebasing, it is assumed that old and\n"
167            "       new backing file match exactly. The image doesn't need a working\n"
168            "       backing file before rebasing in this case (useful for renaming the\n"
169            "       backing file). For image creation, allow creating without attempting\n"
170            "       to open the backing file.\n"
171            "  '-h' with or without a command shows this help and lists the supported formats\n"
172            "  '-p' show progress of command (only certain commands)\n"
173            "  '-q' use Quiet mode - do not print any output (except errors)\n"
174            "  '-S' indicates the consecutive number of bytes (defaults to 4k) that must\n"
175            "       contain only zeros for qemu-img to create a sparse image during\n"
176            "       conversion. If the number of bytes is 0, the source will not be scanned for\n"
177            "       unallocated or zero sectors, and the destination image will always be\n"
178            "       fully allocated\n"
179            "  '--output' takes the format in which the output must be done (human or json)\n"
180            "  '-n' skips the target volume creation (useful if the volume is created\n"
181            "       prior to running qemu-img)\n"
182            "\n"
183            "Parameters to bitmap subcommand:\n"
184            "  'bitmap' is the name of the bitmap to manipulate, through one or more\n"
185            "       actions from '--add', '--remove', '--clear', '--enable', '--disable',\n"
186            "       or '--merge source'\n"
187            "  '-g granularity' sets the granularity for '--add' actions\n"
188            "  '-b source' and '-F src_fmt' tell '--merge' actions to find the source\n"
189            "       bitmaps from an alternative file\n"
190            "\n"
191            "Parameters to check subcommand:\n"
192            "  '-r' tries to repair any inconsistencies that are found during the check.\n"
193            "       '-r leaks' repairs only cluster leaks, whereas '-r all' fixes all\n"
194            "       kinds of errors, with a higher risk of choosing the wrong fix or\n"
195            "       hiding corruption that has already occurred.\n"
196            "\n"
197            "Parameters to convert subcommand:\n"
198            "  '--bitmaps' copies all top-level persistent bitmaps to destination\n"
199            "  '-m' specifies how many coroutines work in parallel during the convert\n"
200            "       process (defaults to 8)\n"
201            "  '-W' allow to write to the target out of order rather than sequential\n"
202            "\n"
203            "Parameters to snapshot subcommand:\n"
204            "  'snapshot' is the name of the snapshot to create, apply or delete\n"
205            "  '-a' applies a snapshot (revert disk to saved state)\n"
206            "  '-c' creates a snapshot\n"
207            "  '-d' deletes a snapshot\n"
208            "  '-l' lists all snapshots in the given image\n"
209            "\n"
210            "Parameters to compare subcommand:\n"
211            "  '-f' first image format\n"
212            "  '-F' second image format\n"
213            "  '-s' run in Strict mode - fail on different image size or sector allocation\n"
214            "\n"
215            "Parameters to dd subcommand:\n"
216            "  'bs=BYTES' read and write up to BYTES bytes at a time "
217            "(default: 512)\n"
218            "  'count=N' copy only N input blocks\n"
219            "  'if=FILE' read from FILE\n"
220            "  'of=FILE' write to FILE\n"
221            "  'skip=N' skip N bs-sized blocks at the start of input\n";
222 
223     printf("%s\nSupported formats:", help_msg);
224     bdrv_iterate_format(format_print, NULL, false);
225     printf("\n\n" QEMU_HELP_BOTTOM "\n");
226     exit(EXIT_SUCCESS);
227 }
228 
229 /*
230  * Is @optarg safe for accumulate_options()?
231  * It is when multiple of them can be joined together separated by ','.
232  * To make that work, @optarg must not start with ',' (or else a
233  * separating ',' preceding it gets escaped), and it must not end with
234  * an odd number of ',' (or else a separating ',' following it gets
235  * escaped), or be empty (or else a separating ',' preceding it can
236  * escape a separating ',' following it).
237  *
238  */
239 static bool is_valid_option_list(const char *optarg)
240 {
241     size_t len = strlen(optarg);
242     size_t i;
243 
244     if (!optarg[0] || optarg[0] == ',') {
245         return false;
246     }
247 
248     for (i = len; i > 0 && optarg[i - 1] == ','; i--) {
249     }
250     if ((len - i) % 2) {
251         return false;
252     }
253 
254     return true;
255 }
256 
257 static int accumulate_options(char **options, char *optarg)
258 {
259     char *new_options;
260 
261     if (!is_valid_option_list(optarg)) {
262         error_report("Invalid option list: %s", optarg);
263         return -1;
264     }
265 
266     if (!*options) {
267         *options = g_strdup(optarg);
268     } else {
269         new_options = g_strdup_printf("%s,%s", *options, optarg);
270         g_free(*options);
271         *options = new_options;
272     }
273     return 0;
274 }
275 
276 static QemuOptsList qemu_source_opts = {
277     .name = "source",
278     .implied_opt_name = "file",
279     .head = QTAILQ_HEAD_INITIALIZER(qemu_source_opts.head),
280     .desc = {
281         { }
282     },
283 };
284 
285 static int GCC_FMT_ATTR(2, 3) qprintf(bool quiet, const char *fmt, ...)
286 {
287     int ret = 0;
288     if (!quiet) {
289         va_list args;
290         va_start(args, fmt);
291         ret = vprintf(fmt, args);
292         va_end(args);
293     }
294     return ret;
295 }
296 
297 
298 static int print_block_option_help(const char *filename, const char *fmt)
299 {
300     BlockDriver *drv, *proto_drv;
301     QemuOptsList *create_opts = NULL;
302     Error *local_err = NULL;
303 
304     /* Find driver and parse its options */
305     drv = bdrv_find_format(fmt);
306     if (!drv) {
307         error_report("Unknown file format '%s'", fmt);
308         return 1;
309     }
310 
311     if (!drv->create_opts) {
312         error_report("Format driver '%s' does not support image creation", fmt);
313         return 1;
314     }
315 
316     create_opts = qemu_opts_append(create_opts, drv->create_opts);
317     if (filename) {
318         proto_drv = bdrv_find_protocol(filename, true, &local_err);
319         if (!proto_drv) {
320             error_report_err(local_err);
321             qemu_opts_free(create_opts);
322             return 1;
323         }
324         if (!proto_drv->create_opts) {
325             error_report("Protocol driver '%s' does not support image creation",
326                          proto_drv->format_name);
327             qemu_opts_free(create_opts);
328             return 1;
329         }
330         create_opts = qemu_opts_append(create_opts, proto_drv->create_opts);
331     }
332 
333     if (filename) {
334         printf("Supported options:\n");
335     } else {
336         printf("Supported %s options:\n", fmt);
337     }
338     qemu_opts_print_help(create_opts, false);
339     qemu_opts_free(create_opts);
340 
341     if (!filename) {
342         printf("\n"
343                "The protocol level may support further options.\n"
344                "Specify the target filename to include those options.\n");
345     }
346 
347     return 0;
348 }
349 
350 
351 static BlockBackend *img_open_opts(const char *optstr,
352                                    QemuOpts *opts, int flags, bool writethrough,
353                                    bool quiet, bool force_share)
354 {
355     QDict *options;
356     Error *local_err = NULL;
357     BlockBackend *blk;
358     options = qemu_opts_to_qdict(opts, NULL);
359     if (force_share) {
360         if (qdict_haskey(options, BDRV_OPT_FORCE_SHARE)
361             && strcmp(qdict_get_str(options, BDRV_OPT_FORCE_SHARE), "on")) {
362             error_report("--force-share/-U conflicts with image options");
363             qobject_unref(options);
364             return NULL;
365         }
366         qdict_put_str(options, BDRV_OPT_FORCE_SHARE, "on");
367     }
368     blk = blk_new_open(NULL, NULL, options, flags, &local_err);
369     if (!blk) {
370         error_reportf_err(local_err, "Could not open '%s': ", optstr);
371         return NULL;
372     }
373     blk_set_enable_write_cache(blk, !writethrough);
374 
375     return blk;
376 }
377 
378 static BlockBackend *img_open_file(const char *filename,
379                                    QDict *options,
380                                    const char *fmt, int flags,
381                                    bool writethrough, bool quiet,
382                                    bool force_share)
383 {
384     BlockBackend *blk;
385     Error *local_err = NULL;
386 
387     if (!options) {
388         options = qdict_new();
389     }
390     if (fmt) {
391         qdict_put_str(options, "driver", fmt);
392     }
393 
394     if (force_share) {
395         qdict_put_bool(options, BDRV_OPT_FORCE_SHARE, true);
396     }
397     blk = blk_new_open(filename, NULL, options, flags, &local_err);
398     if (!blk) {
399         error_reportf_err(local_err, "Could not open '%s': ", filename);
400         return NULL;
401     }
402     blk_set_enable_write_cache(blk, !writethrough);
403 
404     return blk;
405 }
406 
407 
408 static int img_add_key_secrets(void *opaque,
409                                const char *name, const char *value,
410                                Error **errp)
411 {
412     QDict *options = opaque;
413 
414     if (g_str_has_suffix(name, "key-secret")) {
415         qdict_put_str(options, name, value);
416     }
417 
418     return 0;
419 }
420 
421 
422 static BlockBackend *img_open(bool image_opts,
423                               const char *filename,
424                               const char *fmt, int flags, bool writethrough,
425                               bool quiet, bool force_share)
426 {
427     BlockBackend *blk;
428     if (image_opts) {
429         QemuOpts *opts;
430         if (fmt) {
431             error_report("--image-opts and --format are mutually exclusive");
432             return NULL;
433         }
434         opts = qemu_opts_parse_noisily(qemu_find_opts("source"),
435                                        filename, true);
436         if (!opts) {
437             return NULL;
438         }
439         blk = img_open_opts(filename, opts, flags, writethrough, quiet,
440                             force_share);
441     } else {
442         blk = img_open_file(filename, NULL, fmt, flags, writethrough, quiet,
443                             force_share);
444     }
445     return blk;
446 }
447 
448 
449 static int add_old_style_options(const char *fmt, QemuOpts *opts,
450                                  const char *base_filename,
451                                  const char *base_fmt)
452 {
453     if (base_filename) {
454         if (!qemu_opt_set(opts, BLOCK_OPT_BACKING_FILE, base_filename,
455                           NULL)) {
456             error_report("Backing file not supported for file format '%s'",
457                          fmt);
458             return -1;
459         }
460     }
461     if (base_fmt) {
462         if (!qemu_opt_set(opts, BLOCK_OPT_BACKING_FMT, base_fmt, NULL)) {
463             error_report("Backing file format not supported for file "
464                          "format '%s'", fmt);
465             return -1;
466         }
467     }
468     return 0;
469 }
470 
471 static int64_t cvtnum_full(const char *name, const char *value, int64_t min,
472                            int64_t max)
473 {
474     int err;
475     uint64_t res;
476 
477     err = qemu_strtosz(value, NULL, &res);
478     if (err < 0 && err != -ERANGE) {
479         error_report("Invalid %s specified. You may use "
480                      "k, M, G, T, P or E suffixes for", name);
481         error_report("kilobytes, megabytes, gigabytes, terabytes, "
482                      "petabytes and exabytes.");
483         return err;
484     }
485     if (err == -ERANGE || res > max || res < min) {
486         error_report("Invalid %s specified. Must be between %" PRId64
487                      " and %" PRId64 ".", name, min, max);
488         return -ERANGE;
489     }
490     return res;
491 }
492 
493 static int64_t cvtnum(const char *name, const char *value)
494 {
495     return cvtnum_full(name, value, 0, INT64_MAX);
496 }
497 
498 static int img_create(int argc, char **argv)
499 {
500     int c;
501     uint64_t img_size = -1;
502     const char *fmt = "raw";
503     const char *base_fmt = NULL;
504     const char *filename;
505     const char *base_filename = NULL;
506     char *options = NULL;
507     Error *local_err = NULL;
508     bool quiet = false;
509     int flags = 0;
510 
511     for(;;) {
512         static const struct option long_options[] = {
513             {"help", no_argument, 0, 'h'},
514             {"object", required_argument, 0, OPTION_OBJECT},
515             {0, 0, 0, 0}
516         };
517         c = getopt_long(argc, argv, ":F:b:f:ho:qu",
518                         long_options, NULL);
519         if (c == -1) {
520             break;
521         }
522         switch(c) {
523         case ':':
524             missing_argument(argv[optind - 1]);
525             break;
526         case '?':
527             unrecognized_option(argv[optind - 1]);
528             break;
529         case 'h':
530             help();
531             break;
532         case 'F':
533             base_fmt = optarg;
534             break;
535         case 'b':
536             base_filename = optarg;
537             break;
538         case 'f':
539             fmt = optarg;
540             break;
541         case 'o':
542             if (accumulate_options(&options, optarg) < 0) {
543                 goto fail;
544             }
545             break;
546         case 'q':
547             quiet = true;
548             break;
549         case 'u':
550             flags |= BDRV_O_NO_BACKING;
551             break;
552         case OPTION_OBJECT:
553             user_creatable_process_cmdline(optarg);
554             break;
555         }
556     }
557 
558     /* Get the filename */
559     filename = (optind < argc) ? argv[optind] : NULL;
560     if (options && has_help_option(options)) {
561         g_free(options);
562         return print_block_option_help(filename, fmt);
563     }
564 
565     if (optind >= argc) {
566         error_exit("Expecting image file name");
567     }
568     optind++;
569 
570     /* Get image size, if specified */
571     if (optind < argc) {
572         int64_t sval;
573 
574         sval = cvtnum("image size", argv[optind++]);
575         if (sval < 0) {
576             goto fail;
577         }
578         img_size = (uint64_t)sval;
579     }
580     if (optind != argc) {
581         error_exit("Unexpected argument: %s", argv[optind]);
582     }
583 
584     bdrv_img_create(filename, fmt, base_filename, base_fmt,
585                     options, img_size, flags, quiet, &local_err);
586     if (local_err) {
587         error_reportf_err(local_err, "%s: ", filename);
588         goto fail;
589     }
590 
591     g_free(options);
592     return 0;
593 
594 fail:
595     g_free(options);
596     return 1;
597 }
598 
599 static void dump_json_image_check(ImageCheck *check, bool quiet)
600 {
601     GString *str;
602     QObject *obj;
603     Visitor *v = qobject_output_visitor_new(&obj);
604 
605     visit_type_ImageCheck(v, NULL, &check, &error_abort);
606     visit_complete(v, &obj);
607     str = qobject_to_json_pretty(obj, true);
608     assert(str != NULL);
609     qprintf(quiet, "%s\n", str->str);
610     qobject_unref(obj);
611     visit_free(v);
612     g_string_free(str, true);
613 }
614 
615 static void dump_human_image_check(ImageCheck *check, bool quiet)
616 {
617     if (!(check->corruptions || check->leaks || check->check_errors)) {
618         qprintf(quiet, "No errors were found on the image.\n");
619     } else {
620         if (check->corruptions) {
621             qprintf(quiet, "\n%" PRId64 " errors were found on the image.\n"
622                     "Data may be corrupted, or further writes to the image "
623                     "may corrupt it.\n",
624                     check->corruptions);
625         }
626 
627         if (check->leaks) {
628             qprintf(quiet,
629                     "\n%" PRId64 " leaked clusters were found on the image.\n"
630                     "This means waste of disk space, but no harm to data.\n",
631                     check->leaks);
632         }
633 
634         if (check->check_errors) {
635             qprintf(quiet,
636                     "\n%" PRId64
637                     " internal errors have occurred during the check.\n",
638                     check->check_errors);
639         }
640     }
641 
642     if (check->total_clusters != 0 && check->allocated_clusters != 0) {
643         qprintf(quiet, "%" PRId64 "/%" PRId64 " = %0.2f%% allocated, "
644                 "%0.2f%% fragmented, %0.2f%% compressed clusters\n",
645                 check->allocated_clusters, check->total_clusters,
646                 check->allocated_clusters * 100.0 / check->total_clusters,
647                 check->fragmented_clusters * 100.0 / check->allocated_clusters,
648                 check->compressed_clusters * 100.0 /
649                 check->allocated_clusters);
650     }
651 
652     if (check->image_end_offset) {
653         qprintf(quiet,
654                 "Image end offset: %" PRId64 "\n", check->image_end_offset);
655     }
656 }
657 
658 static int collect_image_check(BlockDriverState *bs,
659                    ImageCheck *check,
660                    const char *filename,
661                    const char *fmt,
662                    int fix)
663 {
664     int ret;
665     BdrvCheckResult result;
666 
667     ret = bdrv_check(bs, &result, fix);
668     if (ret < 0) {
669         return ret;
670     }
671 
672     check->filename                 = g_strdup(filename);
673     check->format                   = g_strdup(bdrv_get_format_name(bs));
674     check->check_errors             = result.check_errors;
675     check->corruptions              = result.corruptions;
676     check->has_corruptions          = result.corruptions != 0;
677     check->leaks                    = result.leaks;
678     check->has_leaks                = result.leaks != 0;
679     check->corruptions_fixed        = result.corruptions_fixed;
680     check->has_corruptions_fixed    = result.corruptions_fixed != 0;
681     check->leaks_fixed              = result.leaks_fixed;
682     check->has_leaks_fixed          = result.leaks_fixed != 0;
683     check->image_end_offset         = result.image_end_offset;
684     check->has_image_end_offset     = result.image_end_offset != 0;
685     check->total_clusters           = result.bfi.total_clusters;
686     check->has_total_clusters       = result.bfi.total_clusters != 0;
687     check->allocated_clusters       = result.bfi.allocated_clusters;
688     check->has_allocated_clusters   = result.bfi.allocated_clusters != 0;
689     check->fragmented_clusters      = result.bfi.fragmented_clusters;
690     check->has_fragmented_clusters  = result.bfi.fragmented_clusters != 0;
691     check->compressed_clusters      = result.bfi.compressed_clusters;
692     check->has_compressed_clusters  = result.bfi.compressed_clusters != 0;
693 
694     return 0;
695 }
696 
697 /*
698  * Checks an image for consistency. Exit codes:
699  *
700  *  0 - Check completed, image is good
701  *  1 - Check not completed because of internal errors
702  *  2 - Check completed, image is corrupted
703  *  3 - Check completed, image has leaked clusters, but is good otherwise
704  * 63 - Checks are not supported by the image format
705  */
706 static int img_check(int argc, char **argv)
707 {
708     int c, ret;
709     OutputFormat output_format = OFORMAT_HUMAN;
710     const char *filename, *fmt, *output, *cache;
711     BlockBackend *blk;
712     BlockDriverState *bs;
713     int fix = 0;
714     int flags = BDRV_O_CHECK;
715     bool writethrough;
716     ImageCheck *check;
717     bool quiet = false;
718     bool image_opts = false;
719     bool force_share = false;
720 
721     fmt = NULL;
722     output = NULL;
723     cache = BDRV_DEFAULT_CACHE;
724 
725     for(;;) {
726         int option_index = 0;
727         static const struct option long_options[] = {
728             {"help", no_argument, 0, 'h'},
729             {"format", required_argument, 0, 'f'},
730             {"repair", required_argument, 0, 'r'},
731             {"output", required_argument, 0, OPTION_OUTPUT},
732             {"object", required_argument, 0, OPTION_OBJECT},
733             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
734             {"force-share", no_argument, 0, 'U'},
735             {0, 0, 0, 0}
736         };
737         c = getopt_long(argc, argv, ":hf:r:T:qU",
738                         long_options, &option_index);
739         if (c == -1) {
740             break;
741         }
742         switch(c) {
743         case ':':
744             missing_argument(argv[optind - 1]);
745             break;
746         case '?':
747             unrecognized_option(argv[optind - 1]);
748             break;
749         case 'h':
750             help();
751             break;
752         case 'f':
753             fmt = optarg;
754             break;
755         case 'r':
756             flags |= BDRV_O_RDWR;
757 
758             if (!strcmp(optarg, "leaks")) {
759                 fix = BDRV_FIX_LEAKS;
760             } else if (!strcmp(optarg, "all")) {
761                 fix = BDRV_FIX_LEAKS | BDRV_FIX_ERRORS;
762             } else {
763                 error_exit("Unknown option value for -r "
764                            "(expecting 'leaks' or 'all'): %s", optarg);
765             }
766             break;
767         case OPTION_OUTPUT:
768             output = optarg;
769             break;
770         case 'T':
771             cache = optarg;
772             break;
773         case 'q':
774             quiet = true;
775             break;
776         case 'U':
777             force_share = true;
778             break;
779         case OPTION_OBJECT:
780             user_creatable_process_cmdline(optarg);
781             break;
782         case OPTION_IMAGE_OPTS:
783             image_opts = true;
784             break;
785         }
786     }
787     if (optind != argc - 1) {
788         error_exit("Expecting one image file name");
789     }
790     filename = argv[optind++];
791 
792     if (output && !strcmp(output, "json")) {
793         output_format = OFORMAT_JSON;
794     } else if (output && !strcmp(output, "human")) {
795         output_format = OFORMAT_HUMAN;
796     } else if (output) {
797         error_report("--output must be used with human or json as argument.");
798         return 1;
799     }
800 
801     ret = bdrv_parse_cache_mode(cache, &flags, &writethrough);
802     if (ret < 0) {
803         error_report("Invalid source cache option: %s", cache);
804         return 1;
805     }
806 
807     blk = img_open(image_opts, filename, fmt, flags, writethrough, quiet,
808                    force_share);
809     if (!blk) {
810         return 1;
811     }
812     bs = blk_bs(blk);
813 
814     check = g_new0(ImageCheck, 1);
815     ret = collect_image_check(bs, check, filename, fmt, fix);
816 
817     if (ret == -ENOTSUP) {
818         error_report("This image format does not support checks");
819         ret = 63;
820         goto fail;
821     }
822 
823     if (check->corruptions_fixed || check->leaks_fixed) {
824         int corruptions_fixed, leaks_fixed;
825         bool has_leaks_fixed, has_corruptions_fixed;
826 
827         leaks_fixed         = check->leaks_fixed;
828         has_leaks_fixed     = check->has_leaks_fixed;
829         corruptions_fixed   = check->corruptions_fixed;
830         has_corruptions_fixed = check->has_corruptions_fixed;
831 
832         if (output_format == OFORMAT_HUMAN) {
833             qprintf(quiet,
834                     "The following inconsistencies were found and repaired:\n\n"
835                     "    %" PRId64 " leaked clusters\n"
836                     "    %" PRId64 " corruptions\n\n"
837                     "Double checking the fixed image now...\n",
838                     check->leaks_fixed,
839                     check->corruptions_fixed);
840         }
841 
842         qapi_free_ImageCheck(check);
843         check = g_new0(ImageCheck, 1);
844         ret = collect_image_check(bs, check, filename, fmt, 0);
845 
846         check->leaks_fixed          = leaks_fixed;
847         check->has_leaks_fixed      = has_leaks_fixed;
848         check->corruptions_fixed    = corruptions_fixed;
849         check->has_corruptions_fixed = has_corruptions_fixed;
850     }
851 
852     if (!ret) {
853         switch (output_format) {
854         case OFORMAT_HUMAN:
855             dump_human_image_check(check, quiet);
856             break;
857         case OFORMAT_JSON:
858             dump_json_image_check(check, quiet);
859             break;
860         }
861     }
862 
863     if (ret || check->check_errors) {
864         if (ret) {
865             error_report("Check failed: %s", strerror(-ret));
866         } else {
867             error_report("Check failed");
868         }
869         ret = 1;
870         goto fail;
871     }
872 
873     if (check->corruptions) {
874         ret = 2;
875     } else if (check->leaks) {
876         ret = 3;
877     } else {
878         ret = 0;
879     }
880 
881 fail:
882     qapi_free_ImageCheck(check);
883     blk_unref(blk);
884     return ret;
885 }
886 
887 typedef struct CommonBlockJobCBInfo {
888     BlockDriverState *bs;
889     Error **errp;
890 } CommonBlockJobCBInfo;
891 
892 static void common_block_job_cb(void *opaque, int ret)
893 {
894     CommonBlockJobCBInfo *cbi = opaque;
895 
896     if (ret < 0) {
897         error_setg_errno(cbi->errp, -ret, "Block job failed");
898     }
899 }
900 
901 static void run_block_job(BlockJob *job, Error **errp)
902 {
903     uint64_t progress_current, progress_total;
904     AioContext *aio_context = blk_get_aio_context(job->blk);
905     int ret = 0;
906 
907     aio_context_acquire(aio_context);
908     job_ref(&job->job);
909     do {
910         float progress = 0.0f;
911         aio_poll(aio_context, true);
912 
913         progress_get_snapshot(&job->job.progress, &progress_current,
914                               &progress_total);
915         if (progress_total) {
916             progress = (float)progress_current / progress_total * 100.f;
917         }
918         qemu_progress_print(progress, 0);
919     } while (!job_is_ready(&job->job) && !job_is_completed(&job->job));
920 
921     if (!job_is_completed(&job->job)) {
922         ret = job_complete_sync(&job->job, errp);
923     } else {
924         ret = job->job.ret;
925     }
926     job_unref(&job->job);
927     aio_context_release(aio_context);
928 
929     /* publish completion progress only when success */
930     if (!ret) {
931         qemu_progress_print(100.f, 0);
932     }
933 }
934 
935 static int img_commit(int argc, char **argv)
936 {
937     int c, ret, flags;
938     const char *filename, *fmt, *cache, *base;
939     BlockBackend *blk;
940     BlockDriverState *bs, *base_bs;
941     BlockJob *job;
942     bool progress = false, quiet = false, drop = false;
943     bool writethrough;
944     Error *local_err = NULL;
945     CommonBlockJobCBInfo cbi;
946     bool image_opts = false;
947     AioContext *aio_context;
948     int64_t rate_limit = 0;
949 
950     fmt = NULL;
951     cache = BDRV_DEFAULT_CACHE;
952     base = NULL;
953     for(;;) {
954         static const struct option long_options[] = {
955             {"help", no_argument, 0, 'h'},
956             {"object", required_argument, 0, OPTION_OBJECT},
957             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
958             {0, 0, 0, 0}
959         };
960         c = getopt_long(argc, argv, ":f:ht:b:dpqr:",
961                         long_options, NULL);
962         if (c == -1) {
963             break;
964         }
965         switch(c) {
966         case ':':
967             missing_argument(argv[optind - 1]);
968             break;
969         case '?':
970             unrecognized_option(argv[optind - 1]);
971             break;
972         case 'h':
973             help();
974             break;
975         case 'f':
976             fmt = optarg;
977             break;
978         case 't':
979             cache = optarg;
980             break;
981         case 'b':
982             base = optarg;
983             /* -b implies -d */
984             drop = true;
985             break;
986         case 'd':
987             drop = true;
988             break;
989         case 'p':
990             progress = true;
991             break;
992         case 'q':
993             quiet = true;
994             break;
995         case 'r':
996             rate_limit = cvtnum("rate limit", optarg);
997             if (rate_limit < 0) {
998                 return 1;
999             }
1000             break;
1001         case OPTION_OBJECT:
1002             user_creatable_process_cmdline(optarg);
1003             break;
1004         case OPTION_IMAGE_OPTS:
1005             image_opts = true;
1006             break;
1007         }
1008     }
1009 
1010     /* Progress is not shown in Quiet mode */
1011     if (quiet) {
1012         progress = false;
1013     }
1014 
1015     if (optind != argc - 1) {
1016         error_exit("Expecting one image file name");
1017     }
1018     filename = argv[optind++];
1019 
1020     flags = BDRV_O_RDWR | BDRV_O_UNMAP;
1021     ret = bdrv_parse_cache_mode(cache, &flags, &writethrough);
1022     if (ret < 0) {
1023         error_report("Invalid cache option: %s", cache);
1024         return 1;
1025     }
1026 
1027     blk = img_open(image_opts, filename, fmt, flags, writethrough, quiet,
1028                    false);
1029     if (!blk) {
1030         return 1;
1031     }
1032     bs = blk_bs(blk);
1033 
1034     qemu_progress_init(progress, 1.f);
1035     qemu_progress_print(0.f, 100);
1036 
1037     if (base) {
1038         base_bs = bdrv_find_backing_image(bs, base);
1039         if (!base_bs) {
1040             error_setg(&local_err,
1041                        "Did not find '%s' in the backing chain of '%s'",
1042                        base, filename);
1043             goto done;
1044         }
1045     } else {
1046         /* This is different from QMP, which by default uses the deepest file in
1047          * the backing chain (i.e., the very base); however, the traditional
1048          * behavior of qemu-img commit is using the immediate backing file. */
1049         base_bs = bdrv_backing_chain_next(bs);
1050         if (!base_bs) {
1051             error_setg(&local_err, "Image does not have a backing file");
1052             goto done;
1053         }
1054     }
1055 
1056     cbi = (CommonBlockJobCBInfo){
1057         .errp = &local_err,
1058         .bs   = bs,
1059     };
1060 
1061     aio_context = bdrv_get_aio_context(bs);
1062     aio_context_acquire(aio_context);
1063     commit_active_start("commit", bs, base_bs, JOB_DEFAULT, rate_limit,
1064                         BLOCKDEV_ON_ERROR_REPORT, NULL, common_block_job_cb,
1065                         &cbi, false, &local_err);
1066     aio_context_release(aio_context);
1067     if (local_err) {
1068         goto done;
1069     }
1070 
1071     /* When the block job completes, the BlockBackend reference will point to
1072      * the old backing file. In order to avoid that the top image is already
1073      * deleted, so we can still empty it afterwards, increment the reference
1074      * counter here preemptively. */
1075     if (!drop) {
1076         bdrv_ref(bs);
1077     }
1078 
1079     job = block_job_get("commit");
1080     assert(job);
1081     run_block_job(job, &local_err);
1082     if (local_err) {
1083         goto unref_backing;
1084     }
1085 
1086     if (!drop) {
1087         BlockBackend *old_backing_blk;
1088 
1089         old_backing_blk = blk_new_with_bs(bs, BLK_PERM_WRITE, BLK_PERM_ALL,
1090                                           &local_err);
1091         if (!old_backing_blk) {
1092             goto unref_backing;
1093         }
1094         ret = blk_make_empty(old_backing_blk, &local_err);
1095         blk_unref(old_backing_blk);
1096         if (ret == -ENOTSUP) {
1097             error_free(local_err);
1098             local_err = NULL;
1099         } else if (ret < 0) {
1100             goto unref_backing;
1101         }
1102     }
1103 
1104 unref_backing:
1105     if (!drop) {
1106         bdrv_unref(bs);
1107     }
1108 
1109 done:
1110     qemu_progress_end();
1111 
1112     blk_unref(blk);
1113 
1114     if (local_err) {
1115         error_report_err(local_err);
1116         return 1;
1117     }
1118 
1119     qprintf(quiet, "Image committed.\n");
1120     return 0;
1121 }
1122 
1123 /*
1124  * Returns -1 if 'buf' contains only zeroes, otherwise the byte index
1125  * of the first sector boundary within buf where the sector contains a
1126  * non-zero byte.  This function is robust to a buffer that is not
1127  * sector-aligned.
1128  */
1129 static int64_t find_nonzero(const uint8_t *buf, int64_t n)
1130 {
1131     int64_t i;
1132     int64_t end = QEMU_ALIGN_DOWN(n, BDRV_SECTOR_SIZE);
1133 
1134     for (i = 0; i < end; i += BDRV_SECTOR_SIZE) {
1135         if (!buffer_is_zero(buf + i, BDRV_SECTOR_SIZE)) {
1136             return i;
1137         }
1138     }
1139     if (i < n && !buffer_is_zero(buf + i, n - end)) {
1140         return i;
1141     }
1142     return -1;
1143 }
1144 
1145 /*
1146  * Returns true iff the first sector pointed to by 'buf' contains at least
1147  * a non-NUL byte.
1148  *
1149  * 'pnum' is set to the number of sectors (including and immediately following
1150  * the first one) that are known to be in the same allocated/unallocated state.
1151  * The function will try to align the end offset to alignment boundaries so
1152  * that the request will at least end aligned and consecutive requests will
1153  * also start at an aligned offset.
1154  */
1155 static int is_allocated_sectors(const uint8_t *buf, int n, int *pnum,
1156                                 int64_t sector_num, int alignment)
1157 {
1158     bool is_zero;
1159     int i, tail;
1160 
1161     if (n <= 0) {
1162         *pnum = 0;
1163         return 0;
1164     }
1165     is_zero = buffer_is_zero(buf, BDRV_SECTOR_SIZE);
1166     for(i = 1; i < n; i++) {
1167         buf += BDRV_SECTOR_SIZE;
1168         if (is_zero != buffer_is_zero(buf, BDRV_SECTOR_SIZE)) {
1169             break;
1170         }
1171     }
1172 
1173     tail = (sector_num + i) & (alignment - 1);
1174     if (tail) {
1175         if (is_zero && i <= tail) {
1176             /* treat unallocated areas which only consist
1177              * of a small tail as allocated. */
1178             is_zero = false;
1179         }
1180         if (!is_zero) {
1181             /* align up end offset of allocated areas. */
1182             i += alignment - tail;
1183             i = MIN(i, n);
1184         } else {
1185             /* align down end offset of zero areas. */
1186             i -= tail;
1187         }
1188     }
1189     *pnum = i;
1190     return !is_zero;
1191 }
1192 
1193 /*
1194  * Like is_allocated_sectors, but if the buffer starts with a used sector,
1195  * up to 'min' consecutive sectors containing zeros are ignored. This avoids
1196  * breaking up write requests for only small sparse areas.
1197  */
1198 static int is_allocated_sectors_min(const uint8_t *buf, int n, int *pnum,
1199     int min, int64_t sector_num, int alignment)
1200 {
1201     int ret;
1202     int num_checked, num_used;
1203 
1204     if (n < min) {
1205         min = n;
1206     }
1207 
1208     ret = is_allocated_sectors(buf, n, pnum, sector_num, alignment);
1209     if (!ret) {
1210         return ret;
1211     }
1212 
1213     num_used = *pnum;
1214     buf += BDRV_SECTOR_SIZE * *pnum;
1215     n -= *pnum;
1216     sector_num += *pnum;
1217     num_checked = num_used;
1218 
1219     while (n > 0) {
1220         ret = is_allocated_sectors(buf, n, pnum, sector_num, alignment);
1221 
1222         buf += BDRV_SECTOR_SIZE * *pnum;
1223         n -= *pnum;
1224         sector_num += *pnum;
1225         num_checked += *pnum;
1226         if (ret) {
1227             num_used = num_checked;
1228         } else if (*pnum >= min) {
1229             break;
1230         }
1231     }
1232 
1233     *pnum = num_used;
1234     return 1;
1235 }
1236 
1237 /*
1238  * Compares two buffers sector by sector. Returns 0 if the first
1239  * sector of each buffer matches, non-zero otherwise.
1240  *
1241  * pnum is set to the sector-aligned size of the buffer prefix that
1242  * has the same matching status as the first sector.
1243  */
1244 static int compare_buffers(const uint8_t *buf1, const uint8_t *buf2,
1245                            int64_t bytes, int64_t *pnum)
1246 {
1247     bool res;
1248     int64_t i = MIN(bytes, BDRV_SECTOR_SIZE);
1249 
1250     assert(bytes > 0);
1251 
1252     res = !!memcmp(buf1, buf2, i);
1253     while (i < bytes) {
1254         int64_t len = MIN(bytes - i, BDRV_SECTOR_SIZE);
1255 
1256         if (!!memcmp(buf1 + i, buf2 + i, len) != res) {
1257             break;
1258         }
1259         i += len;
1260     }
1261 
1262     *pnum = i;
1263     return res;
1264 }
1265 
1266 #define IO_BUF_SIZE (2 * MiB)
1267 
1268 /*
1269  * Check if passed sectors are empty (not allocated or contain only 0 bytes)
1270  *
1271  * Intended for use by 'qemu-img compare': Returns 0 in case sectors are
1272  * filled with 0, 1 if sectors contain non-zero data (this is a comparison
1273  * failure), and 4 on error (the exit status for read errors), after emitting
1274  * an error message.
1275  *
1276  * @param blk:  BlockBackend for the image
1277  * @param offset: Starting offset to check
1278  * @param bytes: Number of bytes to check
1279  * @param filename: Name of disk file we are checking (logging purpose)
1280  * @param buffer: Allocated buffer for storing read data
1281  * @param quiet: Flag for quiet mode
1282  */
1283 static int check_empty_sectors(BlockBackend *blk, int64_t offset,
1284                                int64_t bytes, const char *filename,
1285                                uint8_t *buffer, bool quiet)
1286 {
1287     int ret = 0;
1288     int64_t idx;
1289 
1290     ret = blk_pread(blk, offset, buffer, bytes);
1291     if (ret < 0) {
1292         error_report("Error while reading offset %" PRId64 " of %s: %s",
1293                      offset, filename, strerror(-ret));
1294         return 4;
1295     }
1296     idx = find_nonzero(buffer, bytes);
1297     if (idx >= 0) {
1298         qprintf(quiet, "Content mismatch at offset %" PRId64 "!\n",
1299                 offset + idx);
1300         return 1;
1301     }
1302 
1303     return 0;
1304 }
1305 
1306 /*
1307  * Compares two images. Exit codes:
1308  *
1309  * 0 - Images are identical or the requested help was printed
1310  * 1 - Images differ
1311  * >1 - Error occurred
1312  */
1313 static int img_compare(int argc, char **argv)
1314 {
1315     const char *fmt1 = NULL, *fmt2 = NULL, *cache, *filename1, *filename2;
1316     BlockBackend *blk1, *blk2;
1317     BlockDriverState *bs1, *bs2;
1318     int64_t total_size1, total_size2;
1319     uint8_t *buf1 = NULL, *buf2 = NULL;
1320     int64_t pnum1, pnum2;
1321     int allocated1, allocated2;
1322     int ret = 0; /* return value - 0 Ident, 1 Different, >1 Error */
1323     bool progress = false, quiet = false, strict = false;
1324     int flags;
1325     bool writethrough;
1326     int64_t total_size;
1327     int64_t offset = 0;
1328     int64_t chunk;
1329     int c;
1330     uint64_t progress_base;
1331     bool image_opts = false;
1332     bool force_share = false;
1333 
1334     cache = BDRV_DEFAULT_CACHE;
1335     for (;;) {
1336         static const struct option long_options[] = {
1337             {"help", no_argument, 0, 'h'},
1338             {"object", required_argument, 0, OPTION_OBJECT},
1339             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
1340             {"force-share", no_argument, 0, 'U'},
1341             {0, 0, 0, 0}
1342         };
1343         c = getopt_long(argc, argv, ":hf:F:T:pqsU",
1344                         long_options, NULL);
1345         if (c == -1) {
1346             break;
1347         }
1348         switch (c) {
1349         case ':':
1350             missing_argument(argv[optind - 1]);
1351             break;
1352         case '?':
1353             unrecognized_option(argv[optind - 1]);
1354             break;
1355         case 'h':
1356             help();
1357             break;
1358         case 'f':
1359             fmt1 = optarg;
1360             break;
1361         case 'F':
1362             fmt2 = optarg;
1363             break;
1364         case 'T':
1365             cache = optarg;
1366             break;
1367         case 'p':
1368             progress = true;
1369             break;
1370         case 'q':
1371             quiet = true;
1372             break;
1373         case 's':
1374             strict = true;
1375             break;
1376         case 'U':
1377             force_share = true;
1378             break;
1379         case OPTION_OBJECT:
1380             {
1381                 Error *local_err = NULL;
1382 
1383                 if (!user_creatable_add_from_str(optarg, &local_err)) {
1384                     if (local_err) {
1385                         error_report_err(local_err);
1386                         exit(2);
1387                     } else {
1388                         /* Help was printed */
1389                         exit(EXIT_SUCCESS);
1390                     }
1391                 }
1392                 break;
1393             }
1394         case OPTION_IMAGE_OPTS:
1395             image_opts = true;
1396             break;
1397         }
1398     }
1399 
1400     /* Progress is not shown in Quiet mode */
1401     if (quiet) {
1402         progress = false;
1403     }
1404 
1405 
1406     if (optind != argc - 2) {
1407         error_exit("Expecting two image file names");
1408     }
1409     filename1 = argv[optind++];
1410     filename2 = argv[optind++];
1411 
1412     /* Initialize before goto out */
1413     qemu_progress_init(progress, 2.0);
1414 
1415     flags = 0;
1416     ret = bdrv_parse_cache_mode(cache, &flags, &writethrough);
1417     if (ret < 0) {
1418         error_report("Invalid source cache option: %s", cache);
1419         ret = 2;
1420         goto out3;
1421     }
1422 
1423     blk1 = img_open(image_opts, filename1, fmt1, flags, writethrough, quiet,
1424                     force_share);
1425     if (!blk1) {
1426         ret = 2;
1427         goto out3;
1428     }
1429 
1430     blk2 = img_open(image_opts, filename2, fmt2, flags, writethrough, quiet,
1431                     force_share);
1432     if (!blk2) {
1433         ret = 2;
1434         goto out2;
1435     }
1436     bs1 = blk_bs(blk1);
1437     bs2 = blk_bs(blk2);
1438 
1439     buf1 = blk_blockalign(blk1, IO_BUF_SIZE);
1440     buf2 = blk_blockalign(blk2, IO_BUF_SIZE);
1441     total_size1 = blk_getlength(blk1);
1442     if (total_size1 < 0) {
1443         error_report("Can't get size of %s: %s",
1444                      filename1, strerror(-total_size1));
1445         ret = 4;
1446         goto out;
1447     }
1448     total_size2 = blk_getlength(blk2);
1449     if (total_size2 < 0) {
1450         error_report("Can't get size of %s: %s",
1451                      filename2, strerror(-total_size2));
1452         ret = 4;
1453         goto out;
1454     }
1455     total_size = MIN(total_size1, total_size2);
1456     progress_base = MAX(total_size1, total_size2);
1457 
1458     qemu_progress_print(0, 100);
1459 
1460     if (strict && total_size1 != total_size2) {
1461         ret = 1;
1462         qprintf(quiet, "Strict mode: Image size mismatch!\n");
1463         goto out;
1464     }
1465 
1466     while (offset < total_size) {
1467         int status1, status2;
1468 
1469         status1 = bdrv_block_status_above(bs1, NULL, offset,
1470                                           total_size1 - offset, &pnum1, NULL,
1471                                           NULL);
1472         if (status1 < 0) {
1473             ret = 3;
1474             error_report("Sector allocation test failed for %s", filename1);
1475             goto out;
1476         }
1477         allocated1 = status1 & BDRV_BLOCK_ALLOCATED;
1478 
1479         status2 = bdrv_block_status_above(bs2, NULL, offset,
1480                                           total_size2 - offset, &pnum2, NULL,
1481                                           NULL);
1482         if (status2 < 0) {
1483             ret = 3;
1484             error_report("Sector allocation test failed for %s", filename2);
1485             goto out;
1486         }
1487         allocated2 = status2 & BDRV_BLOCK_ALLOCATED;
1488 
1489         assert(pnum1 && pnum2);
1490         chunk = MIN(pnum1, pnum2);
1491 
1492         if (strict) {
1493             if (status1 != status2) {
1494                 ret = 1;
1495                 qprintf(quiet, "Strict mode: Offset %" PRId64
1496                         " block status mismatch!\n", offset);
1497                 goto out;
1498             }
1499         }
1500         if ((status1 & BDRV_BLOCK_ZERO) && (status2 & BDRV_BLOCK_ZERO)) {
1501             /* nothing to do */
1502         } else if (allocated1 == allocated2) {
1503             if (allocated1) {
1504                 int64_t pnum;
1505 
1506                 chunk = MIN(chunk, IO_BUF_SIZE);
1507                 ret = blk_pread(blk1, offset, buf1, chunk);
1508                 if (ret < 0) {
1509                     error_report("Error while reading offset %" PRId64
1510                                  " of %s: %s",
1511                                  offset, filename1, strerror(-ret));
1512                     ret = 4;
1513                     goto out;
1514                 }
1515                 ret = blk_pread(blk2, offset, buf2, chunk);
1516                 if (ret < 0) {
1517                     error_report("Error while reading offset %" PRId64
1518                                  " of %s: %s",
1519                                  offset, filename2, strerror(-ret));
1520                     ret = 4;
1521                     goto out;
1522                 }
1523                 ret = compare_buffers(buf1, buf2, chunk, &pnum);
1524                 if (ret || pnum != chunk) {
1525                     qprintf(quiet, "Content mismatch at offset %" PRId64 "!\n",
1526                             offset + (ret ? 0 : pnum));
1527                     ret = 1;
1528                     goto out;
1529                 }
1530             }
1531         } else {
1532             chunk = MIN(chunk, IO_BUF_SIZE);
1533             if (allocated1) {
1534                 ret = check_empty_sectors(blk1, offset, chunk,
1535                                           filename1, buf1, quiet);
1536             } else {
1537                 ret = check_empty_sectors(blk2, offset, chunk,
1538                                           filename2, buf1, quiet);
1539             }
1540             if (ret) {
1541                 goto out;
1542             }
1543         }
1544         offset += chunk;
1545         qemu_progress_print(((float) chunk / progress_base) * 100, 100);
1546     }
1547 
1548     if (total_size1 != total_size2) {
1549         BlockBackend *blk_over;
1550         const char *filename_over;
1551 
1552         qprintf(quiet, "Warning: Image size mismatch!\n");
1553         if (total_size1 > total_size2) {
1554             blk_over = blk1;
1555             filename_over = filename1;
1556         } else {
1557             blk_over = blk2;
1558             filename_over = filename2;
1559         }
1560 
1561         while (offset < progress_base) {
1562             ret = bdrv_block_status_above(blk_bs(blk_over), NULL, offset,
1563                                           progress_base - offset, &chunk,
1564                                           NULL, NULL);
1565             if (ret < 0) {
1566                 ret = 3;
1567                 error_report("Sector allocation test failed for %s",
1568                              filename_over);
1569                 goto out;
1570 
1571             }
1572             if (ret & BDRV_BLOCK_ALLOCATED && !(ret & BDRV_BLOCK_ZERO)) {
1573                 chunk = MIN(chunk, IO_BUF_SIZE);
1574                 ret = check_empty_sectors(blk_over, offset, chunk,
1575                                           filename_over, buf1, quiet);
1576                 if (ret) {
1577                     goto out;
1578                 }
1579             }
1580             offset += chunk;
1581             qemu_progress_print(((float) chunk / progress_base) * 100, 100);
1582         }
1583     }
1584 
1585     qprintf(quiet, "Images are identical.\n");
1586     ret = 0;
1587 
1588 out:
1589     qemu_vfree(buf1);
1590     qemu_vfree(buf2);
1591     blk_unref(blk2);
1592 out2:
1593     blk_unref(blk1);
1594 out3:
1595     qemu_progress_end();
1596     return ret;
1597 }
1598 
1599 /* Convenience wrapper around qmp_block_dirty_bitmap_merge */
1600 static void do_dirty_bitmap_merge(const char *dst_node, const char *dst_name,
1601                                   const char *src_node, const char *src_name,
1602                                   Error **errp)
1603 {
1604     BlockDirtyBitmapMergeSource *merge_src;
1605     BlockDirtyBitmapMergeSourceList *list = NULL;
1606 
1607     merge_src = g_new0(BlockDirtyBitmapMergeSource, 1);
1608     merge_src->type = QTYPE_QDICT;
1609     merge_src->u.external.node = g_strdup(src_node);
1610     merge_src->u.external.name = g_strdup(src_name);
1611     QAPI_LIST_PREPEND(list, merge_src);
1612     qmp_block_dirty_bitmap_merge(dst_node, dst_name, list, errp);
1613     qapi_free_BlockDirtyBitmapMergeSourceList(list);
1614 }
1615 
1616 enum ImgConvertBlockStatus {
1617     BLK_DATA,
1618     BLK_ZERO,
1619     BLK_BACKING_FILE,
1620 };
1621 
1622 #define MAX_COROUTINES 16
1623 #define CONVERT_THROTTLE_GROUP "img_convert"
1624 
1625 typedef struct ImgConvertState {
1626     BlockBackend **src;
1627     int64_t *src_sectors;
1628     int *src_alignment;
1629     int src_num;
1630     int64_t total_sectors;
1631     int64_t allocated_sectors;
1632     int64_t allocated_done;
1633     int64_t sector_num;
1634     int64_t wr_offs;
1635     enum ImgConvertBlockStatus status;
1636     int64_t sector_next_status;
1637     BlockBackend *target;
1638     bool has_zero_init;
1639     bool compressed;
1640     bool target_is_new;
1641     bool target_has_backing;
1642     int64_t target_backing_sectors; /* negative if unknown */
1643     bool wr_in_order;
1644     bool copy_range;
1645     bool salvage;
1646     bool quiet;
1647     int min_sparse;
1648     int alignment;
1649     size_t cluster_sectors;
1650     size_t buf_sectors;
1651     long num_coroutines;
1652     int running_coroutines;
1653     Coroutine *co[MAX_COROUTINES];
1654     int64_t wait_sector_num[MAX_COROUTINES];
1655     CoMutex lock;
1656     int ret;
1657 } ImgConvertState;
1658 
1659 static void convert_select_part(ImgConvertState *s, int64_t sector_num,
1660                                 int *src_cur, int64_t *src_cur_offset)
1661 {
1662     *src_cur = 0;
1663     *src_cur_offset = 0;
1664     while (sector_num - *src_cur_offset >= s->src_sectors[*src_cur]) {
1665         *src_cur_offset += s->src_sectors[*src_cur];
1666         (*src_cur)++;
1667         assert(*src_cur < s->src_num);
1668     }
1669 }
1670 
1671 static int convert_iteration_sectors(ImgConvertState *s, int64_t sector_num)
1672 {
1673     int64_t src_cur_offset;
1674     int ret, n, src_cur;
1675     bool post_backing_zero = false;
1676 
1677     convert_select_part(s, sector_num, &src_cur, &src_cur_offset);
1678 
1679     assert(s->total_sectors > sector_num);
1680     n = MIN(s->total_sectors - sector_num, BDRV_REQUEST_MAX_SECTORS);
1681 
1682     if (s->target_backing_sectors >= 0) {
1683         if (sector_num >= s->target_backing_sectors) {
1684             post_backing_zero = true;
1685         } else if (sector_num + n > s->target_backing_sectors) {
1686             /* Split requests around target_backing_sectors (because
1687              * starting from there, zeros are handled differently) */
1688             n = s->target_backing_sectors - sector_num;
1689         }
1690     }
1691 
1692     if (s->sector_next_status <= sector_num) {
1693         uint64_t offset = (sector_num - src_cur_offset) * BDRV_SECTOR_SIZE;
1694         int64_t count;
1695         int tail;
1696         BlockDriverState *src_bs = blk_bs(s->src[src_cur]);
1697         BlockDriverState *base;
1698 
1699         if (s->target_has_backing) {
1700             base = bdrv_cow_bs(bdrv_skip_filters(src_bs));
1701         } else {
1702             base = NULL;
1703         }
1704 
1705         do {
1706             count = n * BDRV_SECTOR_SIZE;
1707 
1708             ret = bdrv_block_status_above(src_bs, base, offset, count, &count,
1709                                           NULL, NULL);
1710 
1711             if (ret < 0) {
1712                 if (s->salvage) {
1713                     if (n == 1) {
1714                         if (!s->quiet) {
1715                             warn_report("error while reading block status at "
1716                                         "offset %" PRIu64 ": %s", offset,
1717                                         strerror(-ret));
1718                         }
1719                         /* Just try to read the data, then */
1720                         ret = BDRV_BLOCK_DATA;
1721                         count = BDRV_SECTOR_SIZE;
1722                     } else {
1723                         /* Retry on a shorter range */
1724                         n = DIV_ROUND_UP(n, 4);
1725                     }
1726                 } else {
1727                     error_report("error while reading block status at offset "
1728                                  "%" PRIu64 ": %s", offset, strerror(-ret));
1729                     return ret;
1730                 }
1731             }
1732         } while (ret < 0);
1733 
1734         n = DIV_ROUND_UP(count, BDRV_SECTOR_SIZE);
1735 
1736         /*
1737          * Avoid that s->sector_next_status becomes unaligned to the source
1738          * request alignment and/or cluster size to avoid unnecessary read
1739          * cycles.
1740          */
1741         tail = (sector_num - src_cur_offset + n) % s->src_alignment[src_cur];
1742         if (n > tail) {
1743             n -= tail;
1744         }
1745 
1746         if (ret & BDRV_BLOCK_ZERO) {
1747             s->status = post_backing_zero ? BLK_BACKING_FILE : BLK_ZERO;
1748         } else if (ret & BDRV_BLOCK_DATA) {
1749             s->status = BLK_DATA;
1750         } else {
1751             s->status = s->target_has_backing ? BLK_BACKING_FILE : BLK_DATA;
1752         }
1753 
1754         s->sector_next_status = sector_num + n;
1755     }
1756 
1757     n = MIN(n, s->sector_next_status - sector_num);
1758     if (s->status == BLK_DATA) {
1759         n = MIN(n, s->buf_sectors);
1760     }
1761 
1762     /* We need to write complete clusters for compressed images, so if an
1763      * unallocated area is shorter than that, we must consider the whole
1764      * cluster allocated. */
1765     if (s->compressed) {
1766         if (n < s->cluster_sectors) {
1767             n = MIN(s->cluster_sectors, s->total_sectors - sector_num);
1768             s->status = BLK_DATA;
1769         } else {
1770             n = QEMU_ALIGN_DOWN(n, s->cluster_sectors);
1771         }
1772     }
1773 
1774     return n;
1775 }
1776 
1777 static int coroutine_fn convert_co_read(ImgConvertState *s, int64_t sector_num,
1778                                         int nb_sectors, uint8_t *buf)
1779 {
1780     uint64_t single_read_until = 0;
1781     int n, ret;
1782 
1783     assert(nb_sectors <= s->buf_sectors);
1784     while (nb_sectors > 0) {
1785         BlockBackend *blk;
1786         int src_cur;
1787         int64_t bs_sectors, src_cur_offset;
1788         uint64_t offset;
1789 
1790         /* In the case of compression with multiple source files, we can get a
1791          * nb_sectors that spreads into the next part. So we must be able to
1792          * read across multiple BDSes for one convert_read() call. */
1793         convert_select_part(s, sector_num, &src_cur, &src_cur_offset);
1794         blk = s->src[src_cur];
1795         bs_sectors = s->src_sectors[src_cur];
1796 
1797         offset = (sector_num - src_cur_offset) << BDRV_SECTOR_BITS;
1798 
1799         n = MIN(nb_sectors, bs_sectors - (sector_num - src_cur_offset));
1800         if (single_read_until > offset) {
1801             n = 1;
1802         }
1803 
1804         ret = blk_co_pread(blk, offset, n << BDRV_SECTOR_BITS, buf, 0);
1805         if (ret < 0) {
1806             if (s->salvage) {
1807                 if (n > 1) {
1808                     single_read_until = offset + (n << BDRV_SECTOR_BITS);
1809                     continue;
1810                 } else {
1811                     if (!s->quiet) {
1812                         warn_report("error while reading offset %" PRIu64
1813                                     ": %s", offset, strerror(-ret));
1814                     }
1815                     memset(buf, 0, BDRV_SECTOR_SIZE);
1816                 }
1817             } else {
1818                 return ret;
1819             }
1820         }
1821 
1822         sector_num += n;
1823         nb_sectors -= n;
1824         buf += n * BDRV_SECTOR_SIZE;
1825     }
1826 
1827     return 0;
1828 }
1829 
1830 
1831 static int coroutine_fn convert_co_write(ImgConvertState *s, int64_t sector_num,
1832                                          int nb_sectors, uint8_t *buf,
1833                                          enum ImgConvertBlockStatus status)
1834 {
1835     int ret;
1836 
1837     while (nb_sectors > 0) {
1838         int n = nb_sectors;
1839         BdrvRequestFlags flags = s->compressed ? BDRV_REQ_WRITE_COMPRESSED : 0;
1840 
1841         switch (status) {
1842         case BLK_BACKING_FILE:
1843             /* If we have a backing file, leave clusters unallocated that are
1844              * unallocated in the source image, so that the backing file is
1845              * visible at the respective offset. */
1846             assert(s->target_has_backing);
1847             break;
1848 
1849         case BLK_DATA:
1850             /* If we're told to keep the target fully allocated (-S 0) or there
1851              * is real non-zero data, we must write it. Otherwise we can treat
1852              * it as zero sectors.
1853              * Compressed clusters need to be written as a whole, so in that
1854              * case we can only save the write if the buffer is completely
1855              * zeroed. */
1856             if (!s->min_sparse ||
1857                 (!s->compressed &&
1858                  is_allocated_sectors_min(buf, n, &n, s->min_sparse,
1859                                           sector_num, s->alignment)) ||
1860                 (s->compressed &&
1861                  !buffer_is_zero(buf, n * BDRV_SECTOR_SIZE)))
1862             {
1863                 ret = blk_co_pwrite(s->target, sector_num << BDRV_SECTOR_BITS,
1864                                     n << BDRV_SECTOR_BITS, buf, flags);
1865                 if (ret < 0) {
1866                     return ret;
1867                 }
1868                 break;
1869             }
1870             /* fall-through */
1871 
1872         case BLK_ZERO:
1873             if (s->has_zero_init) {
1874                 assert(!s->target_has_backing);
1875                 break;
1876             }
1877             ret = blk_co_pwrite_zeroes(s->target,
1878                                        sector_num << BDRV_SECTOR_BITS,
1879                                        n << BDRV_SECTOR_BITS,
1880                                        BDRV_REQ_MAY_UNMAP);
1881             if (ret < 0) {
1882                 return ret;
1883             }
1884             break;
1885         }
1886 
1887         sector_num += n;
1888         nb_sectors -= n;
1889         buf += n * BDRV_SECTOR_SIZE;
1890     }
1891 
1892     return 0;
1893 }
1894 
1895 static int coroutine_fn convert_co_copy_range(ImgConvertState *s, int64_t sector_num,
1896                                               int nb_sectors)
1897 {
1898     int n, ret;
1899 
1900     while (nb_sectors > 0) {
1901         BlockBackend *blk;
1902         int src_cur;
1903         int64_t bs_sectors, src_cur_offset;
1904         int64_t offset;
1905 
1906         convert_select_part(s, sector_num, &src_cur, &src_cur_offset);
1907         offset = (sector_num - src_cur_offset) << BDRV_SECTOR_BITS;
1908         blk = s->src[src_cur];
1909         bs_sectors = s->src_sectors[src_cur];
1910 
1911         n = MIN(nb_sectors, bs_sectors - (sector_num - src_cur_offset));
1912 
1913         ret = blk_co_copy_range(blk, offset, s->target,
1914                                 sector_num << BDRV_SECTOR_BITS,
1915                                 n << BDRV_SECTOR_BITS, 0, 0);
1916         if (ret < 0) {
1917             return ret;
1918         }
1919 
1920         sector_num += n;
1921         nb_sectors -= n;
1922     }
1923     return 0;
1924 }
1925 
1926 static void coroutine_fn convert_co_do_copy(void *opaque)
1927 {
1928     ImgConvertState *s = opaque;
1929     uint8_t *buf = NULL;
1930     int ret, i;
1931     int index = -1;
1932 
1933     for (i = 0; i < s->num_coroutines; i++) {
1934         if (s->co[i] == qemu_coroutine_self()) {
1935             index = i;
1936             break;
1937         }
1938     }
1939     assert(index >= 0);
1940 
1941     s->running_coroutines++;
1942     buf = blk_blockalign(s->target, s->buf_sectors * BDRV_SECTOR_SIZE);
1943 
1944     while (1) {
1945         int n;
1946         int64_t sector_num;
1947         enum ImgConvertBlockStatus status;
1948         bool copy_range;
1949 
1950         qemu_co_mutex_lock(&s->lock);
1951         if (s->ret != -EINPROGRESS || s->sector_num >= s->total_sectors) {
1952             qemu_co_mutex_unlock(&s->lock);
1953             break;
1954         }
1955         n = convert_iteration_sectors(s, s->sector_num);
1956         if (n < 0) {
1957             qemu_co_mutex_unlock(&s->lock);
1958             s->ret = n;
1959             break;
1960         }
1961         /* save current sector and allocation status to local variables */
1962         sector_num = s->sector_num;
1963         status = s->status;
1964         if (!s->min_sparse && s->status == BLK_ZERO) {
1965             n = MIN(n, s->buf_sectors);
1966         }
1967         /* increment global sector counter so that other coroutines can
1968          * already continue reading beyond this request */
1969         s->sector_num += n;
1970         qemu_co_mutex_unlock(&s->lock);
1971 
1972         if (status == BLK_DATA || (!s->min_sparse && status == BLK_ZERO)) {
1973             s->allocated_done += n;
1974             qemu_progress_print(100.0 * s->allocated_done /
1975                                         s->allocated_sectors, 0);
1976         }
1977 
1978 retry:
1979         copy_range = s->copy_range && s->status == BLK_DATA;
1980         if (status == BLK_DATA && !copy_range) {
1981             ret = convert_co_read(s, sector_num, n, buf);
1982             if (ret < 0) {
1983                 error_report("error while reading at byte %lld: %s",
1984                              sector_num * BDRV_SECTOR_SIZE, strerror(-ret));
1985                 s->ret = ret;
1986             }
1987         } else if (!s->min_sparse && status == BLK_ZERO) {
1988             status = BLK_DATA;
1989             memset(buf, 0x00, n * BDRV_SECTOR_SIZE);
1990         }
1991 
1992         if (s->wr_in_order) {
1993             /* keep writes in order */
1994             while (s->wr_offs != sector_num && s->ret == -EINPROGRESS) {
1995                 s->wait_sector_num[index] = sector_num;
1996                 qemu_coroutine_yield();
1997             }
1998             s->wait_sector_num[index] = -1;
1999         }
2000 
2001         if (s->ret == -EINPROGRESS) {
2002             if (copy_range) {
2003                 ret = convert_co_copy_range(s, sector_num, n);
2004                 if (ret) {
2005                     s->copy_range = false;
2006                     goto retry;
2007                 }
2008             } else {
2009                 ret = convert_co_write(s, sector_num, n, buf, status);
2010             }
2011             if (ret < 0) {
2012                 error_report("error while writing at byte %lld: %s",
2013                              sector_num * BDRV_SECTOR_SIZE, strerror(-ret));
2014                 s->ret = ret;
2015             }
2016         }
2017 
2018         if (s->wr_in_order) {
2019             /* reenter the coroutine that might have waited
2020              * for this write to complete */
2021             s->wr_offs = sector_num + n;
2022             for (i = 0; i < s->num_coroutines; i++) {
2023                 if (s->co[i] && s->wait_sector_num[i] == s->wr_offs) {
2024                     /*
2025                      * A -> B -> A cannot occur because A has
2026                      * s->wait_sector_num[i] == -1 during A -> B.  Therefore
2027                      * B will never enter A during this time window.
2028                      */
2029                     qemu_coroutine_enter(s->co[i]);
2030                     break;
2031                 }
2032             }
2033         }
2034     }
2035 
2036     qemu_vfree(buf);
2037     s->co[index] = NULL;
2038     s->running_coroutines--;
2039     if (!s->running_coroutines && s->ret == -EINPROGRESS) {
2040         /* the convert job finished successfully */
2041         s->ret = 0;
2042     }
2043 }
2044 
2045 static int convert_do_copy(ImgConvertState *s)
2046 {
2047     int ret, i, n;
2048     int64_t sector_num = 0;
2049 
2050     /* Check whether we have zero initialisation or can get it efficiently */
2051     if (!s->has_zero_init && s->target_is_new && s->min_sparse &&
2052         !s->target_has_backing) {
2053         s->has_zero_init = bdrv_has_zero_init(blk_bs(s->target));
2054     }
2055 
2056     /* Allocate buffer for copied data. For compressed images, only one cluster
2057      * can be copied at a time. */
2058     if (s->compressed) {
2059         if (s->cluster_sectors <= 0 || s->cluster_sectors > s->buf_sectors) {
2060             error_report("invalid cluster size");
2061             return -EINVAL;
2062         }
2063         s->buf_sectors = s->cluster_sectors;
2064     }
2065 
2066     while (sector_num < s->total_sectors) {
2067         n = convert_iteration_sectors(s, sector_num);
2068         if (n < 0) {
2069             return n;
2070         }
2071         if (s->status == BLK_DATA || (!s->min_sparse && s->status == BLK_ZERO))
2072         {
2073             s->allocated_sectors += n;
2074         }
2075         sector_num += n;
2076     }
2077 
2078     /* Do the copy */
2079     s->sector_next_status = 0;
2080     s->ret = -EINPROGRESS;
2081 
2082     qemu_co_mutex_init(&s->lock);
2083     for (i = 0; i < s->num_coroutines; i++) {
2084         s->co[i] = qemu_coroutine_create(convert_co_do_copy, s);
2085         s->wait_sector_num[i] = -1;
2086         qemu_coroutine_enter(s->co[i]);
2087     }
2088 
2089     while (s->running_coroutines) {
2090         main_loop_wait(false);
2091     }
2092 
2093     if (s->compressed && !s->ret) {
2094         /* signal EOF to align */
2095         ret = blk_pwrite_compressed(s->target, 0, NULL, 0);
2096         if (ret < 0) {
2097             return ret;
2098         }
2099     }
2100 
2101     return s->ret;
2102 }
2103 
2104 static int convert_copy_bitmaps(BlockDriverState *src, BlockDriverState *dst)
2105 {
2106     BdrvDirtyBitmap *bm;
2107     Error *err = NULL;
2108 
2109     FOR_EACH_DIRTY_BITMAP(src, bm) {
2110         const char *name;
2111 
2112         if (!bdrv_dirty_bitmap_get_persistence(bm)) {
2113             continue;
2114         }
2115         name = bdrv_dirty_bitmap_name(bm);
2116         qmp_block_dirty_bitmap_add(dst->node_name, name,
2117                                    true, bdrv_dirty_bitmap_granularity(bm),
2118                                    true, true,
2119                                    true, !bdrv_dirty_bitmap_enabled(bm),
2120                                    &err);
2121         if (err) {
2122             error_reportf_err(err, "Failed to create bitmap %s: ", name);
2123             return -1;
2124         }
2125 
2126         do_dirty_bitmap_merge(dst->node_name, name, src->node_name, name,
2127                               &err);
2128         if (err) {
2129             error_reportf_err(err, "Failed to populate bitmap %s: ", name);
2130             return -1;
2131         }
2132     }
2133 
2134     return 0;
2135 }
2136 
2137 #define MAX_BUF_SECTORS 32768
2138 
2139 static void set_rate_limit(BlockBackend *blk, int64_t rate_limit)
2140 {
2141     ThrottleConfig cfg;
2142 
2143     throttle_config_init(&cfg);
2144     cfg.buckets[THROTTLE_BPS_WRITE].avg = rate_limit;
2145 
2146     blk_io_limits_enable(blk, CONVERT_THROTTLE_GROUP);
2147     blk_set_io_limits(blk, &cfg);
2148 }
2149 
2150 static int img_convert(int argc, char **argv)
2151 {
2152     int c, bs_i, flags, src_flags = BDRV_O_NO_SHARE;
2153     const char *fmt = NULL, *out_fmt = NULL, *cache = "unsafe",
2154                *src_cache = BDRV_DEFAULT_CACHE, *out_baseimg = NULL,
2155                *out_filename, *out_baseimg_param, *snapshot_name = NULL;
2156     BlockDriver *drv = NULL, *proto_drv = NULL;
2157     BlockDriverInfo bdi;
2158     BlockDriverState *out_bs;
2159     QemuOpts *opts = NULL, *sn_opts = NULL;
2160     QemuOptsList *create_opts = NULL;
2161     QDict *open_opts = NULL;
2162     char *options = NULL;
2163     Error *local_err = NULL;
2164     bool writethrough, src_writethrough, image_opts = false,
2165          skip_create = false, progress = false, tgt_image_opts = false;
2166     int64_t ret = -EINVAL;
2167     bool force_share = false;
2168     bool explict_min_sparse = false;
2169     bool bitmaps = false;
2170     int64_t rate_limit = 0;
2171 
2172     ImgConvertState s = (ImgConvertState) {
2173         /* Need at least 4k of zeros for sparse detection */
2174         .min_sparse         = 8,
2175         .copy_range         = false,
2176         .buf_sectors        = IO_BUF_SIZE / BDRV_SECTOR_SIZE,
2177         .wr_in_order        = true,
2178         .num_coroutines     = 8,
2179     };
2180 
2181     for(;;) {
2182         static const struct option long_options[] = {
2183             {"help", no_argument, 0, 'h'},
2184             {"object", required_argument, 0, OPTION_OBJECT},
2185             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
2186             {"force-share", no_argument, 0, 'U'},
2187             {"target-image-opts", no_argument, 0, OPTION_TARGET_IMAGE_OPTS},
2188             {"salvage", no_argument, 0, OPTION_SALVAGE},
2189             {"target-is-zero", no_argument, 0, OPTION_TARGET_IS_ZERO},
2190             {"bitmaps", no_argument, 0, OPTION_BITMAPS},
2191             {0, 0, 0, 0}
2192         };
2193         c = getopt_long(argc, argv, ":hf:O:B:Cco:l:S:pt:T:qnm:WUr:",
2194                         long_options, NULL);
2195         if (c == -1) {
2196             break;
2197         }
2198         switch(c) {
2199         case ':':
2200             missing_argument(argv[optind - 1]);
2201             break;
2202         case '?':
2203             unrecognized_option(argv[optind - 1]);
2204             break;
2205         case 'h':
2206             help();
2207             break;
2208         case 'f':
2209             fmt = optarg;
2210             break;
2211         case 'O':
2212             out_fmt = optarg;
2213             break;
2214         case 'B':
2215             out_baseimg = optarg;
2216             break;
2217         case 'C':
2218             s.copy_range = true;
2219             break;
2220         case 'c':
2221             s.compressed = true;
2222             break;
2223         case 'o':
2224             if (accumulate_options(&options, optarg) < 0) {
2225                 goto fail_getopt;
2226             }
2227             break;
2228         case 'l':
2229             if (strstart(optarg, SNAPSHOT_OPT_BASE, NULL)) {
2230                 sn_opts = qemu_opts_parse_noisily(&internal_snapshot_opts,
2231                                                   optarg, false);
2232                 if (!sn_opts) {
2233                     error_report("Failed in parsing snapshot param '%s'",
2234                                  optarg);
2235                     goto fail_getopt;
2236                 }
2237             } else {
2238                 snapshot_name = optarg;
2239             }
2240             break;
2241         case 'S':
2242         {
2243             int64_t sval;
2244 
2245             sval = cvtnum("buffer size for sparse output", optarg);
2246             if (sval < 0) {
2247                 goto fail_getopt;
2248             } else if (!QEMU_IS_ALIGNED(sval, BDRV_SECTOR_SIZE) ||
2249                 sval / BDRV_SECTOR_SIZE > MAX_BUF_SECTORS) {
2250                 error_report("Invalid buffer size for sparse output specified. "
2251                     "Valid sizes are multiples of %llu up to %llu. Select "
2252                     "0 to disable sparse detection (fully allocates output).",
2253                     BDRV_SECTOR_SIZE, MAX_BUF_SECTORS * BDRV_SECTOR_SIZE);
2254                 goto fail_getopt;
2255             }
2256 
2257             s.min_sparse = sval / BDRV_SECTOR_SIZE;
2258             explict_min_sparse = true;
2259             break;
2260         }
2261         case 'p':
2262             progress = true;
2263             break;
2264         case 't':
2265             cache = optarg;
2266             break;
2267         case 'T':
2268             src_cache = optarg;
2269             break;
2270         case 'q':
2271             s.quiet = true;
2272             break;
2273         case 'n':
2274             skip_create = true;
2275             break;
2276         case 'm':
2277             if (qemu_strtol(optarg, NULL, 0, &s.num_coroutines) ||
2278                 s.num_coroutines < 1 || s.num_coroutines > MAX_COROUTINES) {
2279                 error_report("Invalid number of coroutines. Allowed number of"
2280                              " coroutines is between 1 and %d", MAX_COROUTINES);
2281                 goto fail_getopt;
2282             }
2283             break;
2284         case 'W':
2285             s.wr_in_order = false;
2286             break;
2287         case 'U':
2288             force_share = true;
2289             break;
2290         case 'r':
2291             rate_limit = cvtnum("rate limit", optarg);
2292             if (rate_limit < 0) {
2293                 goto fail_getopt;
2294             }
2295             break;
2296         case OPTION_OBJECT:
2297             user_creatable_process_cmdline(optarg);
2298             break;
2299         case OPTION_IMAGE_OPTS:
2300             image_opts = true;
2301             break;
2302         case OPTION_SALVAGE:
2303             s.salvage = true;
2304             break;
2305         case OPTION_TARGET_IMAGE_OPTS:
2306             tgt_image_opts = true;
2307             break;
2308         case OPTION_TARGET_IS_ZERO:
2309             /*
2310              * The user asserting that the target is blank has the
2311              * same effect as the target driver supporting zero
2312              * initialisation.
2313              */
2314             s.has_zero_init = true;
2315             break;
2316         case OPTION_BITMAPS:
2317             bitmaps = true;
2318             break;
2319         }
2320     }
2321 
2322     if (!out_fmt && !tgt_image_opts) {
2323         out_fmt = "raw";
2324     }
2325 
2326     if (s.compressed && s.copy_range) {
2327         error_report("Cannot enable copy offloading when -c is used");
2328         goto fail_getopt;
2329     }
2330 
2331     if (explict_min_sparse && s.copy_range) {
2332         error_report("Cannot enable copy offloading when -S is used");
2333         goto fail_getopt;
2334     }
2335 
2336     if (s.copy_range && s.salvage) {
2337         error_report("Cannot use copy offloading in salvaging mode");
2338         goto fail_getopt;
2339     }
2340 
2341     if (tgt_image_opts && !skip_create) {
2342         error_report("--target-image-opts requires use of -n flag");
2343         goto fail_getopt;
2344     }
2345 
2346     if (skip_create && options) {
2347         error_report("-o has no effect when skipping image creation");
2348         goto fail_getopt;
2349     }
2350 
2351     if (s.has_zero_init && !skip_create) {
2352         error_report("--target-is-zero requires use of -n flag");
2353         goto fail_getopt;
2354     }
2355 
2356     s.src_num = argc - optind - 1;
2357     out_filename = s.src_num >= 1 ? argv[argc - 1] : NULL;
2358 
2359     if (options && has_help_option(options)) {
2360         if (out_fmt) {
2361             ret = print_block_option_help(out_filename, out_fmt);
2362             goto fail_getopt;
2363         } else {
2364             error_report("Option help requires a format be specified");
2365             goto fail_getopt;
2366         }
2367     }
2368 
2369     if (s.src_num < 1) {
2370         error_report("Must specify image file name");
2371         goto fail_getopt;
2372     }
2373 
2374     /* ret is still -EINVAL until here */
2375     ret = bdrv_parse_cache_mode(src_cache, &src_flags, &src_writethrough);
2376     if (ret < 0) {
2377         error_report("Invalid source cache option: %s", src_cache);
2378         goto fail_getopt;
2379     }
2380 
2381     /* Initialize before goto out */
2382     if (s.quiet) {
2383         progress = false;
2384     }
2385     qemu_progress_init(progress, 1.0);
2386     qemu_progress_print(0, 100);
2387 
2388     s.src = g_new0(BlockBackend *, s.src_num);
2389     s.src_sectors = g_new(int64_t, s.src_num);
2390     s.src_alignment = g_new(int, s.src_num);
2391 
2392     for (bs_i = 0; bs_i < s.src_num; bs_i++) {
2393         BlockDriverState *src_bs;
2394         s.src[bs_i] = img_open(image_opts, argv[optind + bs_i],
2395                                fmt, src_flags, src_writethrough, s.quiet,
2396                                force_share);
2397         if (!s.src[bs_i]) {
2398             ret = -1;
2399             goto out;
2400         }
2401         s.src_sectors[bs_i] = blk_nb_sectors(s.src[bs_i]);
2402         if (s.src_sectors[bs_i] < 0) {
2403             error_report("Could not get size of %s: %s",
2404                          argv[optind + bs_i], strerror(-s.src_sectors[bs_i]));
2405             ret = -1;
2406             goto out;
2407         }
2408         src_bs = blk_bs(s.src[bs_i]);
2409         s.src_alignment[bs_i] = DIV_ROUND_UP(src_bs->bl.request_alignment,
2410                                              BDRV_SECTOR_SIZE);
2411         if (!bdrv_get_info(src_bs, &bdi)) {
2412             s.src_alignment[bs_i] = MAX(s.src_alignment[bs_i],
2413                                         bdi.cluster_size / BDRV_SECTOR_SIZE);
2414         }
2415         s.total_sectors += s.src_sectors[bs_i];
2416     }
2417 
2418     if (sn_opts) {
2419         bdrv_snapshot_load_tmp(blk_bs(s.src[0]),
2420                                qemu_opt_get(sn_opts, SNAPSHOT_OPT_ID),
2421                                qemu_opt_get(sn_opts, SNAPSHOT_OPT_NAME),
2422                                &local_err);
2423     } else if (snapshot_name != NULL) {
2424         if (s.src_num > 1) {
2425             error_report("No support for concatenating multiple snapshot");
2426             ret = -1;
2427             goto out;
2428         }
2429 
2430         bdrv_snapshot_load_tmp_by_id_or_name(blk_bs(s.src[0]), snapshot_name,
2431                                              &local_err);
2432     }
2433     if (local_err) {
2434         error_reportf_err(local_err, "Failed to load snapshot: ");
2435         ret = -1;
2436         goto out;
2437     }
2438 
2439     if (!skip_create) {
2440         /* Find driver and parse its options */
2441         drv = bdrv_find_format(out_fmt);
2442         if (!drv) {
2443             error_report("Unknown file format '%s'", out_fmt);
2444             ret = -1;
2445             goto out;
2446         }
2447 
2448         proto_drv = bdrv_find_protocol(out_filename, true, &local_err);
2449         if (!proto_drv) {
2450             error_report_err(local_err);
2451             ret = -1;
2452             goto out;
2453         }
2454 
2455         if (!drv->create_opts) {
2456             error_report("Format driver '%s' does not support image creation",
2457                          drv->format_name);
2458             ret = -1;
2459             goto out;
2460         }
2461 
2462         if (!proto_drv->create_opts) {
2463             error_report("Protocol driver '%s' does not support image creation",
2464                          proto_drv->format_name);
2465             ret = -1;
2466             goto out;
2467         }
2468 
2469         create_opts = qemu_opts_append(create_opts, drv->create_opts);
2470         create_opts = qemu_opts_append(create_opts, proto_drv->create_opts);
2471 
2472         opts = qemu_opts_create(create_opts, NULL, 0, &error_abort);
2473         if (options) {
2474             if (!qemu_opts_do_parse(opts, options, NULL, &local_err)) {
2475                 error_report_err(local_err);
2476                 ret = -1;
2477                 goto out;
2478             }
2479         }
2480 
2481         qemu_opt_set_number(opts, BLOCK_OPT_SIZE,
2482                             s.total_sectors * BDRV_SECTOR_SIZE, &error_abort);
2483         ret = add_old_style_options(out_fmt, opts, out_baseimg, NULL);
2484         if (ret < 0) {
2485             goto out;
2486         }
2487     }
2488 
2489     /* Get backing file name if -o backing_file was used */
2490     out_baseimg_param = qemu_opt_get(opts, BLOCK_OPT_BACKING_FILE);
2491     if (out_baseimg_param) {
2492         out_baseimg = out_baseimg_param;
2493     }
2494     s.target_has_backing = (bool) out_baseimg;
2495 
2496     if (s.has_zero_init && s.target_has_backing) {
2497         error_report("Cannot use --target-is-zero when the destination "
2498                      "image has a backing file");
2499         goto out;
2500     }
2501 
2502     if (s.src_num > 1 && out_baseimg) {
2503         error_report("Having a backing file for the target makes no sense when "
2504                      "concatenating multiple input images");
2505         ret = -1;
2506         goto out;
2507     }
2508 
2509     if (out_baseimg_param) {
2510         if (!qemu_opt_get(opts, BLOCK_OPT_BACKING_FMT)) {
2511             error_report("Use of backing file requires explicit "
2512                          "backing format");
2513             ret = -1;
2514             goto out;
2515         }
2516     }
2517 
2518     /* Check if compression is supported */
2519     if (s.compressed) {
2520         bool encryption =
2521             qemu_opt_get_bool(opts, BLOCK_OPT_ENCRYPT, false);
2522         const char *encryptfmt =
2523             qemu_opt_get(opts, BLOCK_OPT_ENCRYPT_FORMAT);
2524         const char *preallocation =
2525             qemu_opt_get(opts, BLOCK_OPT_PREALLOC);
2526 
2527         if (drv && !block_driver_can_compress(drv)) {
2528             error_report("Compression not supported for this file format");
2529             ret = -1;
2530             goto out;
2531         }
2532 
2533         if (encryption || encryptfmt) {
2534             error_report("Compression and encryption not supported at "
2535                          "the same time");
2536             ret = -1;
2537             goto out;
2538         }
2539 
2540         if (preallocation
2541             && strcmp(preallocation, "off"))
2542         {
2543             error_report("Compression and preallocation not supported at "
2544                          "the same time");
2545             ret = -1;
2546             goto out;
2547         }
2548     }
2549 
2550     /* Determine if bitmaps need copying */
2551     if (bitmaps) {
2552         if (s.src_num > 1) {
2553             error_report("Copying bitmaps only possible with single source");
2554             ret = -1;
2555             goto out;
2556         }
2557         if (!bdrv_supports_persistent_dirty_bitmap(blk_bs(s.src[0]))) {
2558             error_report("Source lacks bitmap support");
2559             ret = -1;
2560             goto out;
2561         }
2562     }
2563 
2564     /*
2565      * The later open call will need any decryption secrets, and
2566      * bdrv_create() will purge "opts", so extract them now before
2567      * they are lost.
2568      */
2569     if (!skip_create) {
2570         open_opts = qdict_new();
2571         qemu_opt_foreach(opts, img_add_key_secrets, open_opts, &error_abort);
2572 
2573         /* Create the new image */
2574         ret = bdrv_create(drv, out_filename, opts, &local_err);
2575         if (ret < 0) {
2576             error_reportf_err(local_err, "%s: error while converting %s: ",
2577                               out_filename, out_fmt);
2578             goto out;
2579         }
2580     }
2581 
2582     s.target_is_new = !skip_create;
2583 
2584     flags = s.min_sparse ? (BDRV_O_RDWR | BDRV_O_UNMAP) : BDRV_O_RDWR;
2585     ret = bdrv_parse_cache_mode(cache, &flags, &writethrough);
2586     if (ret < 0) {
2587         error_report("Invalid cache option: %s", cache);
2588         goto out;
2589     }
2590 
2591     if (skip_create) {
2592         s.target = img_open(tgt_image_opts, out_filename, out_fmt,
2593                             flags, writethrough, s.quiet, false);
2594     } else {
2595         /* TODO ultimately we should allow --target-image-opts
2596          * to be used even when -n is not given.
2597          * That has to wait for bdrv_create to be improved
2598          * to allow filenames in option syntax
2599          */
2600         s.target = img_open_file(out_filename, open_opts, out_fmt,
2601                                  flags, writethrough, s.quiet, false);
2602         open_opts = NULL; /* blk_new_open will have freed it */
2603     }
2604     if (!s.target) {
2605         ret = -1;
2606         goto out;
2607     }
2608     out_bs = blk_bs(s.target);
2609 
2610     if (bitmaps && !bdrv_supports_persistent_dirty_bitmap(out_bs)) {
2611         error_report("Format driver '%s' does not support bitmaps",
2612                      out_bs->drv->format_name);
2613         ret = -1;
2614         goto out;
2615     }
2616 
2617     if (s.compressed && !block_driver_can_compress(out_bs->drv)) {
2618         error_report("Compression not supported for this file format");
2619         ret = -1;
2620         goto out;
2621     }
2622 
2623     /* increase bufsectors from the default 4096 (2M) if opt_transfer
2624      * or discard_alignment of the out_bs is greater. Limit to
2625      * MAX_BUF_SECTORS as maximum which is currently 32768 (16MB). */
2626     s.buf_sectors = MIN(MAX_BUF_SECTORS,
2627                         MAX(s.buf_sectors,
2628                             MAX(out_bs->bl.opt_transfer >> BDRV_SECTOR_BITS,
2629                                 out_bs->bl.pdiscard_alignment >>
2630                                 BDRV_SECTOR_BITS)));
2631 
2632     /* try to align the write requests to the destination to avoid unnecessary
2633      * RMW cycles. */
2634     s.alignment = MAX(pow2floor(s.min_sparse),
2635                       DIV_ROUND_UP(out_bs->bl.request_alignment,
2636                                    BDRV_SECTOR_SIZE));
2637     assert(is_power_of_2(s.alignment));
2638 
2639     if (skip_create) {
2640         int64_t output_sectors = blk_nb_sectors(s.target);
2641         if (output_sectors < 0) {
2642             error_report("unable to get output image length: %s",
2643                          strerror(-output_sectors));
2644             ret = -1;
2645             goto out;
2646         } else if (output_sectors < s.total_sectors) {
2647             error_report("output file is smaller than input file");
2648             ret = -1;
2649             goto out;
2650         }
2651     }
2652 
2653     if (s.target_has_backing && s.target_is_new) {
2654         /* Errors are treated as "backing length unknown" (which means
2655          * s.target_backing_sectors has to be negative, which it will
2656          * be automatically).  The backing file length is used only
2657          * for optimizations, so such a case is not fatal. */
2658         s.target_backing_sectors =
2659             bdrv_nb_sectors(bdrv_backing_chain_next(out_bs));
2660     } else {
2661         s.target_backing_sectors = -1;
2662     }
2663 
2664     ret = bdrv_get_info(out_bs, &bdi);
2665     if (ret < 0) {
2666         if (s.compressed) {
2667             error_report("could not get block driver info");
2668             goto out;
2669         }
2670     } else {
2671         s.compressed = s.compressed || bdi.needs_compressed_writes;
2672         s.cluster_sectors = bdi.cluster_size / BDRV_SECTOR_SIZE;
2673     }
2674 
2675     if (rate_limit) {
2676         set_rate_limit(s.target, rate_limit);
2677     }
2678 
2679     ret = convert_do_copy(&s);
2680 
2681     /* Now copy the bitmaps */
2682     if (bitmaps && ret == 0) {
2683         ret = convert_copy_bitmaps(blk_bs(s.src[0]), out_bs);
2684     }
2685 
2686 out:
2687     if (!ret) {
2688         qemu_progress_print(100, 0);
2689     }
2690     qemu_progress_end();
2691     qemu_opts_del(opts);
2692     qemu_opts_free(create_opts);
2693     qobject_unref(open_opts);
2694     blk_unref(s.target);
2695     if (s.src) {
2696         for (bs_i = 0; bs_i < s.src_num; bs_i++) {
2697             blk_unref(s.src[bs_i]);
2698         }
2699         g_free(s.src);
2700     }
2701     g_free(s.src_sectors);
2702     g_free(s.src_alignment);
2703 fail_getopt:
2704     qemu_opts_del(sn_opts);
2705     g_free(options);
2706 
2707     return !!ret;
2708 }
2709 
2710 
2711 static void dump_snapshots(BlockDriverState *bs)
2712 {
2713     QEMUSnapshotInfo *sn_tab, *sn;
2714     int nb_sns, i;
2715 
2716     nb_sns = bdrv_snapshot_list(bs, &sn_tab);
2717     if (nb_sns <= 0)
2718         return;
2719     printf("Snapshot list:\n");
2720     bdrv_snapshot_dump(NULL);
2721     printf("\n");
2722     for(i = 0; i < nb_sns; i++) {
2723         sn = &sn_tab[i];
2724         bdrv_snapshot_dump(sn);
2725         printf("\n");
2726     }
2727     g_free(sn_tab);
2728 }
2729 
2730 static void dump_json_image_info_list(ImageInfoList *list)
2731 {
2732     GString *str;
2733     QObject *obj;
2734     Visitor *v = qobject_output_visitor_new(&obj);
2735 
2736     visit_type_ImageInfoList(v, NULL, &list, &error_abort);
2737     visit_complete(v, &obj);
2738     str = qobject_to_json_pretty(obj, true);
2739     assert(str != NULL);
2740     printf("%s\n", str->str);
2741     qobject_unref(obj);
2742     visit_free(v);
2743     g_string_free(str, true);
2744 }
2745 
2746 static void dump_json_image_info(ImageInfo *info)
2747 {
2748     GString *str;
2749     QObject *obj;
2750     Visitor *v = qobject_output_visitor_new(&obj);
2751 
2752     visit_type_ImageInfo(v, NULL, &info, &error_abort);
2753     visit_complete(v, &obj);
2754     str = qobject_to_json_pretty(obj, true);
2755     assert(str != NULL);
2756     printf("%s\n", str->str);
2757     qobject_unref(obj);
2758     visit_free(v);
2759     g_string_free(str, true);
2760 }
2761 
2762 static void dump_human_image_info_list(ImageInfoList *list)
2763 {
2764     ImageInfoList *elem;
2765     bool delim = false;
2766 
2767     for (elem = list; elem; elem = elem->next) {
2768         if (delim) {
2769             printf("\n");
2770         }
2771         delim = true;
2772 
2773         bdrv_image_info_dump(elem->value);
2774     }
2775 }
2776 
2777 static gboolean str_equal_func(gconstpointer a, gconstpointer b)
2778 {
2779     return strcmp(a, b) == 0;
2780 }
2781 
2782 /**
2783  * Open an image file chain and return an ImageInfoList
2784  *
2785  * @filename: topmost image filename
2786  * @fmt: topmost image format (may be NULL to autodetect)
2787  * @chain: true  - enumerate entire backing file chain
2788  *         false - only topmost image file
2789  *
2790  * Returns a list of ImageInfo objects or NULL if there was an error opening an
2791  * image file.  If there was an error a message will have been printed to
2792  * stderr.
2793  */
2794 static ImageInfoList *collect_image_info_list(bool image_opts,
2795                                               const char *filename,
2796                                               const char *fmt,
2797                                               bool chain, bool force_share)
2798 {
2799     ImageInfoList *head = NULL;
2800     ImageInfoList **tail = &head;
2801     GHashTable *filenames;
2802     Error *err = NULL;
2803 
2804     filenames = g_hash_table_new_full(g_str_hash, str_equal_func, NULL, NULL);
2805 
2806     while (filename) {
2807         BlockBackend *blk;
2808         BlockDriverState *bs;
2809         ImageInfo *info;
2810 
2811         if (g_hash_table_lookup_extended(filenames, filename, NULL, NULL)) {
2812             error_report("Backing file '%s' creates an infinite loop.",
2813                          filename);
2814             goto err;
2815         }
2816         g_hash_table_insert(filenames, (gpointer)filename, NULL);
2817 
2818         blk = img_open(image_opts, filename, fmt,
2819                        BDRV_O_NO_BACKING | BDRV_O_NO_IO, false, false,
2820                        force_share);
2821         if (!blk) {
2822             goto err;
2823         }
2824         bs = blk_bs(blk);
2825 
2826         bdrv_query_image_info(bs, &info, &err);
2827         if (err) {
2828             error_report_err(err);
2829             blk_unref(blk);
2830             goto err;
2831         }
2832 
2833         QAPI_LIST_APPEND(tail, info);
2834 
2835         blk_unref(blk);
2836 
2837         /* Clear parameters that only apply to the topmost image */
2838         filename = fmt = NULL;
2839         image_opts = false;
2840 
2841         if (chain) {
2842             if (info->has_full_backing_filename) {
2843                 filename = info->full_backing_filename;
2844             } else if (info->has_backing_filename) {
2845                 error_report("Could not determine absolute backing filename,"
2846                              " but backing filename '%s' present",
2847                              info->backing_filename);
2848                 goto err;
2849             }
2850             if (info->has_backing_filename_format) {
2851                 fmt = info->backing_filename_format;
2852             }
2853         }
2854     }
2855     g_hash_table_destroy(filenames);
2856     return head;
2857 
2858 err:
2859     qapi_free_ImageInfoList(head);
2860     g_hash_table_destroy(filenames);
2861     return NULL;
2862 }
2863 
2864 static int img_info(int argc, char **argv)
2865 {
2866     int c;
2867     OutputFormat output_format = OFORMAT_HUMAN;
2868     bool chain = false;
2869     const char *filename, *fmt, *output;
2870     ImageInfoList *list;
2871     bool image_opts = false;
2872     bool force_share = false;
2873 
2874     fmt = NULL;
2875     output = NULL;
2876     for(;;) {
2877         int option_index = 0;
2878         static const struct option long_options[] = {
2879             {"help", no_argument, 0, 'h'},
2880             {"format", required_argument, 0, 'f'},
2881             {"output", required_argument, 0, OPTION_OUTPUT},
2882             {"backing-chain", no_argument, 0, OPTION_BACKING_CHAIN},
2883             {"object", required_argument, 0, OPTION_OBJECT},
2884             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
2885             {"force-share", no_argument, 0, 'U'},
2886             {0, 0, 0, 0}
2887         };
2888         c = getopt_long(argc, argv, ":f:hU",
2889                         long_options, &option_index);
2890         if (c == -1) {
2891             break;
2892         }
2893         switch(c) {
2894         case ':':
2895             missing_argument(argv[optind - 1]);
2896             break;
2897         case '?':
2898             unrecognized_option(argv[optind - 1]);
2899             break;
2900         case 'h':
2901             help();
2902             break;
2903         case 'f':
2904             fmt = optarg;
2905             break;
2906         case 'U':
2907             force_share = true;
2908             break;
2909         case OPTION_OUTPUT:
2910             output = optarg;
2911             break;
2912         case OPTION_BACKING_CHAIN:
2913             chain = true;
2914             break;
2915         case OPTION_OBJECT:
2916             user_creatable_process_cmdline(optarg);
2917             break;
2918         case OPTION_IMAGE_OPTS:
2919             image_opts = true;
2920             break;
2921         }
2922     }
2923     if (optind != argc - 1) {
2924         error_exit("Expecting one image file name");
2925     }
2926     filename = argv[optind++];
2927 
2928     if (output && !strcmp(output, "json")) {
2929         output_format = OFORMAT_JSON;
2930     } else if (output && !strcmp(output, "human")) {
2931         output_format = OFORMAT_HUMAN;
2932     } else if (output) {
2933         error_report("--output must be used with human or json as argument.");
2934         return 1;
2935     }
2936 
2937     list = collect_image_info_list(image_opts, filename, fmt, chain,
2938                                    force_share);
2939     if (!list) {
2940         return 1;
2941     }
2942 
2943     switch (output_format) {
2944     case OFORMAT_HUMAN:
2945         dump_human_image_info_list(list);
2946         break;
2947     case OFORMAT_JSON:
2948         if (chain) {
2949             dump_json_image_info_list(list);
2950         } else {
2951             dump_json_image_info(list->value);
2952         }
2953         break;
2954     }
2955 
2956     qapi_free_ImageInfoList(list);
2957     return 0;
2958 }
2959 
2960 static int dump_map_entry(OutputFormat output_format, MapEntry *e,
2961                           MapEntry *next)
2962 {
2963     switch (output_format) {
2964     case OFORMAT_HUMAN:
2965         if (e->data && !e->has_offset) {
2966             error_report("File contains external, encrypted or compressed clusters.");
2967             return -1;
2968         }
2969         if (e->data && !e->zero) {
2970             printf("%#-16"PRIx64"%#-16"PRIx64"%#-16"PRIx64"%s\n",
2971                    e->start, e->length,
2972                    e->has_offset ? e->offset : 0,
2973                    e->has_filename ? e->filename : "");
2974         }
2975         /* This format ignores the distinction between 0, ZERO and ZERO|DATA.
2976          * Modify the flags here to allow more coalescing.
2977          */
2978         if (next && (!next->data || next->zero)) {
2979             next->data = false;
2980             next->zero = true;
2981         }
2982         break;
2983     case OFORMAT_JSON:
2984         printf("{ \"start\": %"PRId64", \"length\": %"PRId64","
2985                " \"depth\": %"PRId64", \"present\": %s, \"zero\": %s,"
2986                " \"data\": %s", e->start, e->length, e->depth,
2987                e->present ? "true" : "false",
2988                e->zero ? "true" : "false",
2989                e->data ? "true" : "false");
2990         if (e->has_offset) {
2991             printf(", \"offset\": %"PRId64"", e->offset);
2992         }
2993         putchar('}');
2994 
2995         if (next) {
2996             puts(",");
2997         }
2998         break;
2999     }
3000     return 0;
3001 }
3002 
3003 static int get_block_status(BlockDriverState *bs, int64_t offset,
3004                             int64_t bytes, MapEntry *e)
3005 {
3006     int ret;
3007     int depth;
3008     BlockDriverState *file;
3009     bool has_offset;
3010     int64_t map;
3011     char *filename = NULL;
3012 
3013     /* As an optimization, we could cache the current range of unallocated
3014      * clusters in each file of the chain, and avoid querying the same
3015      * range repeatedly.
3016      */
3017 
3018     depth = 0;
3019     for (;;) {
3020         bs = bdrv_skip_filters(bs);
3021         ret = bdrv_block_status(bs, offset, bytes, &bytes, &map, &file);
3022         if (ret < 0) {
3023             return ret;
3024         }
3025         assert(bytes);
3026         if (ret & (BDRV_BLOCK_ZERO|BDRV_BLOCK_DATA)) {
3027             break;
3028         }
3029         bs = bdrv_cow_bs(bs);
3030         if (bs == NULL) {
3031             ret = 0;
3032             break;
3033         }
3034 
3035         depth++;
3036     }
3037 
3038     has_offset = !!(ret & BDRV_BLOCK_OFFSET_VALID);
3039 
3040     if (file && has_offset) {
3041         bdrv_refresh_filename(file);
3042         filename = file->filename;
3043     }
3044 
3045     *e = (MapEntry) {
3046         .start = offset,
3047         .length = bytes,
3048         .data = !!(ret & BDRV_BLOCK_DATA),
3049         .zero = !!(ret & BDRV_BLOCK_ZERO),
3050         .offset = map,
3051         .has_offset = has_offset,
3052         .depth = depth,
3053         .present = !!(ret & BDRV_BLOCK_ALLOCATED),
3054         .has_filename = filename,
3055         .filename = filename,
3056     };
3057 
3058     return 0;
3059 }
3060 
3061 static inline bool entry_mergeable(const MapEntry *curr, const MapEntry *next)
3062 {
3063     if (curr->length == 0) {
3064         return false;
3065     }
3066     if (curr->zero != next->zero ||
3067         curr->data != next->data ||
3068         curr->depth != next->depth ||
3069         curr->present != next->present ||
3070         curr->has_filename != next->has_filename ||
3071         curr->has_offset != next->has_offset) {
3072         return false;
3073     }
3074     if (curr->has_filename && strcmp(curr->filename, next->filename)) {
3075         return false;
3076     }
3077     if (curr->has_offset && curr->offset + curr->length != next->offset) {
3078         return false;
3079     }
3080     return true;
3081 }
3082 
3083 static int img_map(int argc, char **argv)
3084 {
3085     int c;
3086     OutputFormat output_format = OFORMAT_HUMAN;
3087     BlockBackend *blk;
3088     BlockDriverState *bs;
3089     const char *filename, *fmt, *output;
3090     int64_t length;
3091     MapEntry curr = { .length = 0 }, next;
3092     int ret = 0;
3093     bool image_opts = false;
3094     bool force_share = false;
3095     int64_t start_offset = 0;
3096     int64_t max_length = -1;
3097 
3098     fmt = NULL;
3099     output = NULL;
3100     for (;;) {
3101         int option_index = 0;
3102         static const struct option long_options[] = {
3103             {"help", no_argument, 0, 'h'},
3104             {"format", required_argument, 0, 'f'},
3105             {"output", required_argument, 0, OPTION_OUTPUT},
3106             {"object", required_argument, 0, OPTION_OBJECT},
3107             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
3108             {"force-share", no_argument, 0, 'U'},
3109             {"start-offset", required_argument, 0, 's'},
3110             {"max-length", required_argument, 0, 'l'},
3111             {0, 0, 0, 0}
3112         };
3113         c = getopt_long(argc, argv, ":f:s:l:hU",
3114                         long_options, &option_index);
3115         if (c == -1) {
3116             break;
3117         }
3118         switch (c) {
3119         case ':':
3120             missing_argument(argv[optind - 1]);
3121             break;
3122         case '?':
3123             unrecognized_option(argv[optind - 1]);
3124             break;
3125         case 'h':
3126             help();
3127             break;
3128         case 'f':
3129             fmt = optarg;
3130             break;
3131         case 'U':
3132             force_share = true;
3133             break;
3134         case OPTION_OUTPUT:
3135             output = optarg;
3136             break;
3137         case 's':
3138             start_offset = cvtnum("start offset", optarg);
3139             if (start_offset < 0) {
3140                 return 1;
3141             }
3142             break;
3143         case 'l':
3144             max_length = cvtnum("max length", optarg);
3145             if (max_length < 0) {
3146                 return 1;
3147             }
3148             break;
3149         case OPTION_OBJECT:
3150             user_creatable_process_cmdline(optarg);
3151             break;
3152         case OPTION_IMAGE_OPTS:
3153             image_opts = true;
3154             break;
3155         }
3156     }
3157     if (optind != argc - 1) {
3158         error_exit("Expecting one image file name");
3159     }
3160     filename = argv[optind];
3161 
3162     if (output && !strcmp(output, "json")) {
3163         output_format = OFORMAT_JSON;
3164     } else if (output && !strcmp(output, "human")) {
3165         output_format = OFORMAT_HUMAN;
3166     } else if (output) {
3167         error_report("--output must be used with human or json as argument.");
3168         return 1;
3169     }
3170 
3171     blk = img_open(image_opts, filename, fmt, 0, false, false, force_share);
3172     if (!blk) {
3173         return 1;
3174     }
3175     bs = blk_bs(blk);
3176 
3177     if (output_format == OFORMAT_HUMAN) {
3178         printf("%-16s%-16s%-16s%s\n", "Offset", "Length", "Mapped to", "File");
3179     } else if (output_format == OFORMAT_JSON) {
3180         putchar('[');
3181     }
3182 
3183     length = blk_getlength(blk);
3184     if (length < 0) {
3185         error_report("Failed to get size for '%s'", filename);
3186         return 1;
3187     }
3188     if (max_length != -1) {
3189         length = MIN(start_offset + max_length, length);
3190     }
3191 
3192     curr.start = start_offset;
3193     while (curr.start + curr.length < length) {
3194         int64_t offset = curr.start + curr.length;
3195         int64_t n = length - offset;
3196 
3197         ret = get_block_status(bs, offset, n, &next);
3198         if (ret < 0) {
3199             error_report("Could not read file metadata: %s", strerror(-ret));
3200             goto out;
3201         }
3202 
3203         if (entry_mergeable(&curr, &next)) {
3204             curr.length += next.length;
3205             continue;
3206         }
3207 
3208         if (curr.length > 0) {
3209             ret = dump_map_entry(output_format, &curr, &next);
3210             if (ret < 0) {
3211                 goto out;
3212             }
3213         }
3214         curr = next;
3215     }
3216 
3217     ret = dump_map_entry(output_format, &curr, NULL);
3218     if (output_format == OFORMAT_JSON) {
3219         puts("]");
3220     }
3221 
3222 out:
3223     blk_unref(blk);
3224     return ret < 0;
3225 }
3226 
3227 #define SNAPSHOT_LIST   1
3228 #define SNAPSHOT_CREATE 2
3229 #define SNAPSHOT_APPLY  3
3230 #define SNAPSHOT_DELETE 4
3231 
3232 static int img_snapshot(int argc, char **argv)
3233 {
3234     BlockBackend *blk;
3235     BlockDriverState *bs;
3236     QEMUSnapshotInfo sn;
3237     char *filename, *snapshot_name = NULL;
3238     int c, ret = 0, bdrv_oflags;
3239     int action = 0;
3240     qemu_timeval tv;
3241     bool quiet = false;
3242     Error *err = NULL;
3243     bool image_opts = false;
3244     bool force_share = false;
3245 
3246     bdrv_oflags = BDRV_O_RDWR;
3247     /* Parse commandline parameters */
3248     for(;;) {
3249         static const struct option long_options[] = {
3250             {"help", no_argument, 0, 'h'},
3251             {"object", required_argument, 0, OPTION_OBJECT},
3252             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
3253             {"force-share", no_argument, 0, 'U'},
3254             {0, 0, 0, 0}
3255         };
3256         c = getopt_long(argc, argv, ":la:c:d:hqU",
3257                         long_options, NULL);
3258         if (c == -1) {
3259             break;
3260         }
3261         switch(c) {
3262         case ':':
3263             missing_argument(argv[optind - 1]);
3264             break;
3265         case '?':
3266             unrecognized_option(argv[optind - 1]);
3267             break;
3268         case 'h':
3269             help();
3270             return 0;
3271         case 'l':
3272             if (action) {
3273                 error_exit("Cannot mix '-l', '-a', '-c', '-d'");
3274                 return 0;
3275             }
3276             action = SNAPSHOT_LIST;
3277             bdrv_oflags &= ~BDRV_O_RDWR; /* no need for RW */
3278             break;
3279         case 'a':
3280             if (action) {
3281                 error_exit("Cannot mix '-l', '-a', '-c', '-d'");
3282                 return 0;
3283             }
3284             action = SNAPSHOT_APPLY;
3285             snapshot_name = optarg;
3286             break;
3287         case 'c':
3288             if (action) {
3289                 error_exit("Cannot mix '-l', '-a', '-c', '-d'");
3290                 return 0;
3291             }
3292             action = SNAPSHOT_CREATE;
3293             snapshot_name = optarg;
3294             break;
3295         case 'd':
3296             if (action) {
3297                 error_exit("Cannot mix '-l', '-a', '-c', '-d'");
3298                 return 0;
3299             }
3300             action = SNAPSHOT_DELETE;
3301             snapshot_name = optarg;
3302             break;
3303         case 'q':
3304             quiet = true;
3305             break;
3306         case 'U':
3307             force_share = true;
3308             break;
3309         case OPTION_OBJECT:
3310             user_creatable_process_cmdline(optarg);
3311             break;
3312         case OPTION_IMAGE_OPTS:
3313             image_opts = true;
3314             break;
3315         }
3316     }
3317 
3318     if (optind != argc - 1) {
3319         error_exit("Expecting one image file name");
3320     }
3321     filename = argv[optind++];
3322 
3323     /* Open the image */
3324     blk = img_open(image_opts, filename, NULL, bdrv_oflags, false, quiet,
3325                    force_share);
3326     if (!blk) {
3327         return 1;
3328     }
3329     bs = blk_bs(blk);
3330 
3331     /* Perform the requested action */
3332     switch(action) {
3333     case SNAPSHOT_LIST:
3334         dump_snapshots(bs);
3335         break;
3336 
3337     case SNAPSHOT_CREATE:
3338         memset(&sn, 0, sizeof(sn));
3339         pstrcpy(sn.name, sizeof(sn.name), snapshot_name);
3340 
3341         qemu_gettimeofday(&tv);
3342         sn.date_sec = tv.tv_sec;
3343         sn.date_nsec = tv.tv_usec * 1000;
3344 
3345         ret = bdrv_snapshot_create(bs, &sn);
3346         if (ret) {
3347             error_report("Could not create snapshot '%s': %d (%s)",
3348                 snapshot_name, ret, strerror(-ret));
3349         }
3350         break;
3351 
3352     case SNAPSHOT_APPLY:
3353         ret = bdrv_snapshot_goto(bs, snapshot_name, &err);
3354         if (ret) {
3355             error_reportf_err(err, "Could not apply snapshot '%s': ",
3356                               snapshot_name);
3357         }
3358         break;
3359 
3360     case SNAPSHOT_DELETE:
3361         ret = bdrv_snapshot_find(bs, &sn, snapshot_name);
3362         if (ret < 0) {
3363             error_report("Could not delete snapshot '%s': snapshot not "
3364                          "found", snapshot_name);
3365             ret = 1;
3366         } else {
3367             ret = bdrv_snapshot_delete(bs, sn.id_str, sn.name, &err);
3368             if (ret < 0) {
3369                 error_reportf_err(err, "Could not delete snapshot '%s': ",
3370                                   snapshot_name);
3371                 ret = 1;
3372             }
3373         }
3374         break;
3375     }
3376 
3377     /* Cleanup */
3378     blk_unref(blk);
3379     if (ret) {
3380         return 1;
3381     }
3382     return 0;
3383 }
3384 
3385 static int img_rebase(int argc, char **argv)
3386 {
3387     BlockBackend *blk = NULL, *blk_old_backing = NULL, *blk_new_backing = NULL;
3388     uint8_t *buf_old = NULL;
3389     uint8_t *buf_new = NULL;
3390     BlockDriverState *bs = NULL, *prefix_chain_bs = NULL;
3391     BlockDriverState *unfiltered_bs;
3392     char *filename;
3393     const char *fmt, *cache, *src_cache, *out_basefmt, *out_baseimg;
3394     int c, flags, src_flags, ret;
3395     bool writethrough, src_writethrough;
3396     int unsafe = 0;
3397     bool force_share = false;
3398     int progress = 0;
3399     bool quiet = false;
3400     Error *local_err = NULL;
3401     bool image_opts = false;
3402 
3403     /* Parse commandline parameters */
3404     fmt = NULL;
3405     cache = BDRV_DEFAULT_CACHE;
3406     src_cache = BDRV_DEFAULT_CACHE;
3407     out_baseimg = NULL;
3408     out_basefmt = NULL;
3409     for(;;) {
3410         static const struct option long_options[] = {
3411             {"help", no_argument, 0, 'h'},
3412             {"object", required_argument, 0, OPTION_OBJECT},
3413             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
3414             {"force-share", no_argument, 0, 'U'},
3415             {0, 0, 0, 0}
3416         };
3417         c = getopt_long(argc, argv, ":hf:F:b:upt:T:qU",
3418                         long_options, NULL);
3419         if (c == -1) {
3420             break;
3421         }
3422         switch(c) {
3423         case ':':
3424             missing_argument(argv[optind - 1]);
3425             break;
3426         case '?':
3427             unrecognized_option(argv[optind - 1]);
3428             break;
3429         case 'h':
3430             help();
3431             return 0;
3432         case 'f':
3433             fmt = optarg;
3434             break;
3435         case 'F':
3436             out_basefmt = optarg;
3437             break;
3438         case 'b':
3439             out_baseimg = optarg;
3440             break;
3441         case 'u':
3442             unsafe = 1;
3443             break;
3444         case 'p':
3445             progress = 1;
3446             break;
3447         case 't':
3448             cache = optarg;
3449             break;
3450         case 'T':
3451             src_cache = optarg;
3452             break;
3453         case 'q':
3454             quiet = true;
3455             break;
3456         case OPTION_OBJECT:
3457             user_creatable_process_cmdline(optarg);
3458             break;
3459         case OPTION_IMAGE_OPTS:
3460             image_opts = true;
3461             break;
3462         case 'U':
3463             force_share = true;
3464             break;
3465         }
3466     }
3467 
3468     if (quiet) {
3469         progress = 0;
3470     }
3471 
3472     if (optind != argc - 1) {
3473         error_exit("Expecting one image file name");
3474     }
3475     if (!unsafe && !out_baseimg) {
3476         error_exit("Must specify backing file (-b) or use unsafe mode (-u)");
3477     }
3478     filename = argv[optind++];
3479 
3480     qemu_progress_init(progress, 2.0);
3481     qemu_progress_print(0, 100);
3482 
3483     flags = BDRV_O_RDWR | (unsafe ? BDRV_O_NO_BACKING : 0);
3484     ret = bdrv_parse_cache_mode(cache, &flags, &writethrough);
3485     if (ret < 0) {
3486         error_report("Invalid cache option: %s", cache);
3487         goto out;
3488     }
3489 
3490     src_flags = 0;
3491     ret = bdrv_parse_cache_mode(src_cache, &src_flags, &src_writethrough);
3492     if (ret < 0) {
3493         error_report("Invalid source cache option: %s", src_cache);
3494         goto out;
3495     }
3496 
3497     /* The source files are opened read-only, don't care about WCE */
3498     assert((src_flags & BDRV_O_RDWR) == 0);
3499     (void) src_writethrough;
3500 
3501     /*
3502      * Open the images.
3503      *
3504      * Ignore the old backing file for unsafe rebase in case we want to correct
3505      * the reference to a renamed or moved backing file.
3506      */
3507     blk = img_open(image_opts, filename, fmt, flags, writethrough, quiet,
3508                    false);
3509     if (!blk) {
3510         ret = -1;
3511         goto out;
3512     }
3513     bs = blk_bs(blk);
3514 
3515     unfiltered_bs = bdrv_skip_filters(bs);
3516 
3517     if (out_basefmt != NULL) {
3518         if (bdrv_find_format(out_basefmt) == NULL) {
3519             error_report("Invalid format name: '%s'", out_basefmt);
3520             ret = -1;
3521             goto out;
3522         }
3523     }
3524 
3525     /* For safe rebasing we need to compare old and new backing file */
3526     if (!unsafe) {
3527         QDict *options = NULL;
3528         BlockDriverState *base_bs = bdrv_cow_bs(unfiltered_bs);
3529 
3530         if (base_bs) {
3531             blk_old_backing = blk_new(qemu_get_aio_context(),
3532                                       BLK_PERM_CONSISTENT_READ,
3533                                       BLK_PERM_ALL);
3534             ret = blk_insert_bs(blk_old_backing, base_bs,
3535                                 &local_err);
3536             if (ret < 0) {
3537                 error_reportf_err(local_err,
3538                                   "Could not reuse old backing file '%s': ",
3539                                   base_bs->filename);
3540                 goto out;
3541             }
3542         } else {
3543             blk_old_backing = NULL;
3544         }
3545 
3546         if (out_baseimg[0]) {
3547             const char *overlay_filename;
3548             char *out_real_path;
3549 
3550             options = qdict_new();
3551             if (out_basefmt) {
3552                 qdict_put_str(options, "driver", out_basefmt);
3553             }
3554             if (force_share) {
3555                 qdict_put_bool(options, BDRV_OPT_FORCE_SHARE, true);
3556             }
3557 
3558             bdrv_refresh_filename(bs);
3559             overlay_filename = bs->exact_filename[0] ? bs->exact_filename
3560                                                      : bs->filename;
3561             out_real_path =
3562                 bdrv_get_full_backing_filename_from_filename(overlay_filename,
3563                                                              out_baseimg,
3564                                                              &local_err);
3565             if (local_err) {
3566                 qobject_unref(options);
3567                 error_reportf_err(local_err,
3568                                   "Could not resolve backing filename: ");
3569                 ret = -1;
3570                 goto out;
3571             }
3572 
3573             /*
3574              * Find out whether we rebase an image on top of a previous image
3575              * in its chain.
3576              */
3577             prefix_chain_bs = bdrv_find_backing_image(bs, out_real_path);
3578             if (prefix_chain_bs) {
3579                 qobject_unref(options);
3580                 g_free(out_real_path);
3581 
3582                 blk_new_backing = blk_new(qemu_get_aio_context(),
3583                                           BLK_PERM_CONSISTENT_READ,
3584                                           BLK_PERM_ALL);
3585                 ret = blk_insert_bs(blk_new_backing, prefix_chain_bs,
3586                                     &local_err);
3587                 if (ret < 0) {
3588                     error_reportf_err(local_err,
3589                                       "Could not reuse backing file '%s': ",
3590                                       out_baseimg);
3591                     goto out;
3592                 }
3593             } else {
3594                 blk_new_backing = blk_new_open(out_real_path, NULL,
3595                                                options, src_flags, &local_err);
3596                 g_free(out_real_path);
3597                 if (!blk_new_backing) {
3598                     error_reportf_err(local_err,
3599                                       "Could not open new backing file '%s': ",
3600                                       out_baseimg);
3601                     ret = -1;
3602                     goto out;
3603                 }
3604             }
3605         }
3606     }
3607 
3608     /*
3609      * Check each unallocated cluster in the COW file. If it is unallocated,
3610      * accesses go to the backing file. We must therefore compare this cluster
3611      * in the old and new backing file, and if they differ we need to copy it
3612      * from the old backing file into the COW file.
3613      *
3614      * If qemu-img crashes during this step, no harm is done. The content of
3615      * the image is the same as the original one at any time.
3616      */
3617     if (!unsafe) {
3618         int64_t size;
3619         int64_t old_backing_size = 0;
3620         int64_t new_backing_size = 0;
3621         uint64_t offset;
3622         int64_t n;
3623         float local_progress = 0;
3624 
3625         buf_old = blk_blockalign(blk, IO_BUF_SIZE);
3626         buf_new = blk_blockalign(blk, IO_BUF_SIZE);
3627 
3628         size = blk_getlength(blk);
3629         if (size < 0) {
3630             error_report("Could not get size of '%s': %s",
3631                          filename, strerror(-size));
3632             ret = -1;
3633             goto out;
3634         }
3635         if (blk_old_backing) {
3636             old_backing_size = blk_getlength(blk_old_backing);
3637             if (old_backing_size < 0) {
3638                 char backing_name[PATH_MAX];
3639 
3640                 bdrv_get_backing_filename(bs, backing_name,
3641                                           sizeof(backing_name));
3642                 error_report("Could not get size of '%s': %s",
3643                              backing_name, strerror(-old_backing_size));
3644                 ret = -1;
3645                 goto out;
3646             }
3647         }
3648         if (blk_new_backing) {
3649             new_backing_size = blk_getlength(blk_new_backing);
3650             if (new_backing_size < 0) {
3651                 error_report("Could not get size of '%s': %s",
3652                              out_baseimg, strerror(-new_backing_size));
3653                 ret = -1;
3654                 goto out;
3655             }
3656         }
3657 
3658         if (size != 0) {
3659             local_progress = (float)100 / (size / MIN(size, IO_BUF_SIZE));
3660         }
3661 
3662         for (offset = 0; offset < size; offset += n) {
3663             bool buf_old_is_zero = false;
3664 
3665             /* How many bytes can we handle with the next read? */
3666             n = MIN(IO_BUF_SIZE, size - offset);
3667 
3668             /* If the cluster is allocated, we don't need to take action */
3669             ret = bdrv_is_allocated(unfiltered_bs, offset, n, &n);
3670             if (ret < 0) {
3671                 error_report("error while reading image metadata: %s",
3672                              strerror(-ret));
3673                 goto out;
3674             }
3675             if (ret) {
3676                 continue;
3677             }
3678 
3679             if (prefix_chain_bs) {
3680                 /*
3681                  * If cluster wasn't changed since prefix_chain, we don't need
3682                  * to take action
3683                  */
3684                 ret = bdrv_is_allocated_above(bdrv_cow_bs(unfiltered_bs),
3685                                               prefix_chain_bs, false,
3686                                               offset, n, &n);
3687                 if (ret < 0) {
3688                     error_report("error while reading image metadata: %s",
3689                                  strerror(-ret));
3690                     goto out;
3691                 }
3692                 if (!ret) {
3693                     continue;
3694                 }
3695             }
3696 
3697             /*
3698              * Read old and new backing file and take into consideration that
3699              * backing files may be smaller than the COW image.
3700              */
3701             if (offset >= old_backing_size) {
3702                 memset(buf_old, 0, n);
3703                 buf_old_is_zero = true;
3704             } else {
3705                 if (offset + n > old_backing_size) {
3706                     n = old_backing_size - offset;
3707                 }
3708 
3709                 ret = blk_pread(blk_old_backing, offset, buf_old, n);
3710                 if (ret < 0) {
3711                     error_report("error while reading from old backing file");
3712                     goto out;
3713                 }
3714             }
3715 
3716             if (offset >= new_backing_size || !blk_new_backing) {
3717                 memset(buf_new, 0, n);
3718             } else {
3719                 if (offset + n > new_backing_size) {
3720                     n = new_backing_size - offset;
3721                 }
3722 
3723                 ret = blk_pread(blk_new_backing, offset, buf_new, n);
3724                 if (ret < 0) {
3725                     error_report("error while reading from new backing file");
3726                     goto out;
3727                 }
3728             }
3729 
3730             /* If they differ, we need to write to the COW file */
3731             uint64_t written = 0;
3732 
3733             while (written < n) {
3734                 int64_t pnum;
3735 
3736                 if (compare_buffers(buf_old + written, buf_new + written,
3737                                     n - written, &pnum))
3738                 {
3739                     if (buf_old_is_zero) {
3740                         ret = blk_pwrite_zeroes(blk, offset + written, pnum, 0);
3741                     } else {
3742                         ret = blk_pwrite(blk, offset + written,
3743                                          buf_old + written, pnum, 0);
3744                     }
3745                     if (ret < 0) {
3746                         error_report("Error while writing to COW image: %s",
3747                             strerror(-ret));
3748                         goto out;
3749                     }
3750                 }
3751 
3752                 written += pnum;
3753             }
3754             qemu_progress_print(local_progress, 100);
3755         }
3756     }
3757 
3758     /*
3759      * Change the backing file. All clusters that are different from the old
3760      * backing file are overwritten in the COW file now, so the visible content
3761      * doesn't change when we switch the backing file.
3762      */
3763     if (out_baseimg && *out_baseimg) {
3764         ret = bdrv_change_backing_file(unfiltered_bs, out_baseimg, out_basefmt,
3765                                        true);
3766     } else {
3767         ret = bdrv_change_backing_file(unfiltered_bs, NULL, NULL, false);
3768     }
3769 
3770     if (ret == -ENOSPC) {
3771         error_report("Could not change the backing file to '%s': No "
3772                      "space left in the file header", out_baseimg);
3773     } else if (ret == -EINVAL && out_baseimg && !out_basefmt) {
3774         error_report("Could not change the backing file to '%s': backing "
3775                      "format must be specified", out_baseimg);
3776     } else if (ret < 0) {
3777         error_report("Could not change the backing file to '%s': %s",
3778             out_baseimg, strerror(-ret));
3779     }
3780 
3781     qemu_progress_print(100, 0);
3782     /*
3783      * TODO At this point it is possible to check if any clusters that are
3784      * allocated in the COW file are the same in the backing file. If so, they
3785      * could be dropped from the COW file. Don't do this before switching the
3786      * backing file, in case of a crash this would lead to corruption.
3787      */
3788 out:
3789     qemu_progress_end();
3790     /* Cleanup */
3791     if (!unsafe) {
3792         blk_unref(blk_old_backing);
3793         blk_unref(blk_new_backing);
3794     }
3795     qemu_vfree(buf_old);
3796     qemu_vfree(buf_new);
3797 
3798     blk_unref(blk);
3799     if (ret) {
3800         return 1;
3801     }
3802     return 0;
3803 }
3804 
3805 static int img_resize(int argc, char **argv)
3806 {
3807     Error *err = NULL;
3808     int c, ret, relative;
3809     const char *filename, *fmt, *size;
3810     int64_t n, total_size, current_size;
3811     bool quiet = false;
3812     BlockBackend *blk = NULL;
3813     PreallocMode prealloc = PREALLOC_MODE_OFF;
3814     QemuOpts *param;
3815 
3816     static QemuOptsList resize_options = {
3817         .name = "resize_options",
3818         .head = QTAILQ_HEAD_INITIALIZER(resize_options.head),
3819         .desc = {
3820             {
3821                 .name = BLOCK_OPT_SIZE,
3822                 .type = QEMU_OPT_SIZE,
3823                 .help = "Virtual disk size"
3824             }, {
3825                 /* end of list */
3826             }
3827         },
3828     };
3829     bool image_opts = false;
3830     bool shrink = false;
3831 
3832     /* Remove size from argv manually so that negative numbers are not treated
3833      * as options by getopt. */
3834     if (argc < 3) {
3835         error_exit("Not enough arguments");
3836         return 1;
3837     }
3838 
3839     size = argv[--argc];
3840 
3841     /* Parse getopt arguments */
3842     fmt = NULL;
3843     for(;;) {
3844         static const struct option long_options[] = {
3845             {"help", no_argument, 0, 'h'},
3846             {"object", required_argument, 0, OPTION_OBJECT},
3847             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
3848             {"preallocation", required_argument, 0, OPTION_PREALLOCATION},
3849             {"shrink", no_argument, 0, OPTION_SHRINK},
3850             {0, 0, 0, 0}
3851         };
3852         c = getopt_long(argc, argv, ":f:hq",
3853                         long_options, NULL);
3854         if (c == -1) {
3855             break;
3856         }
3857         switch(c) {
3858         case ':':
3859             missing_argument(argv[optind - 1]);
3860             break;
3861         case '?':
3862             unrecognized_option(argv[optind - 1]);
3863             break;
3864         case 'h':
3865             help();
3866             break;
3867         case 'f':
3868             fmt = optarg;
3869             break;
3870         case 'q':
3871             quiet = true;
3872             break;
3873         case OPTION_OBJECT:
3874             user_creatable_process_cmdline(optarg);
3875             break;
3876         case OPTION_IMAGE_OPTS:
3877             image_opts = true;
3878             break;
3879         case OPTION_PREALLOCATION:
3880             prealloc = qapi_enum_parse(&PreallocMode_lookup, optarg,
3881                                        PREALLOC_MODE__MAX, NULL);
3882             if (prealloc == PREALLOC_MODE__MAX) {
3883                 error_report("Invalid preallocation mode '%s'", optarg);
3884                 return 1;
3885             }
3886             break;
3887         case OPTION_SHRINK:
3888             shrink = true;
3889             break;
3890         }
3891     }
3892     if (optind != argc - 1) {
3893         error_exit("Expecting image file name and size");
3894     }
3895     filename = argv[optind++];
3896 
3897     /* Choose grow, shrink, or absolute resize mode */
3898     switch (size[0]) {
3899     case '+':
3900         relative = 1;
3901         size++;
3902         break;
3903     case '-':
3904         relative = -1;
3905         size++;
3906         break;
3907     default:
3908         relative = 0;
3909         break;
3910     }
3911 
3912     /* Parse size */
3913     param = qemu_opts_create(&resize_options, NULL, 0, &error_abort);
3914     if (!qemu_opt_set(param, BLOCK_OPT_SIZE, size, &err)) {
3915         error_report_err(err);
3916         ret = -1;
3917         qemu_opts_del(param);
3918         goto out;
3919     }
3920     n = qemu_opt_get_size(param, BLOCK_OPT_SIZE, 0);
3921     qemu_opts_del(param);
3922 
3923     blk = img_open(image_opts, filename, fmt,
3924                    BDRV_O_RDWR | BDRV_O_RESIZE, false, quiet,
3925                    false);
3926     if (!blk) {
3927         ret = -1;
3928         goto out;
3929     }
3930 
3931     current_size = blk_getlength(blk);
3932     if (current_size < 0) {
3933         error_report("Failed to inquire current image length: %s",
3934                      strerror(-current_size));
3935         ret = -1;
3936         goto out;
3937     }
3938 
3939     if (relative) {
3940         total_size = current_size + n * relative;
3941     } else {
3942         total_size = n;
3943     }
3944     if (total_size <= 0) {
3945         error_report("New image size must be positive");
3946         ret = -1;
3947         goto out;
3948     }
3949 
3950     if (total_size <= current_size && prealloc != PREALLOC_MODE_OFF) {
3951         error_report("Preallocation can only be used for growing images");
3952         ret = -1;
3953         goto out;
3954     }
3955 
3956     if (total_size < current_size && !shrink) {
3957         error_report("Use the --shrink option to perform a shrink operation.");
3958         warn_report("Shrinking an image will delete all data beyond the "
3959                     "shrunken image's end. Before performing such an "
3960                     "operation, make sure there is no important data there.");
3961         ret = -1;
3962         goto out;
3963     }
3964 
3965     /*
3966      * The user expects the image to have the desired size after
3967      * resizing, so pass @exact=true.  It is of no use to report
3968      * success when the image has not actually been resized.
3969      */
3970     ret = blk_truncate(blk, total_size, true, prealloc, 0, &err);
3971     if (!ret) {
3972         qprintf(quiet, "Image resized.\n");
3973     } else {
3974         error_report_err(err);
3975     }
3976 out:
3977     blk_unref(blk);
3978     if (ret) {
3979         return 1;
3980     }
3981     return 0;
3982 }
3983 
3984 static void amend_status_cb(BlockDriverState *bs,
3985                             int64_t offset, int64_t total_work_size,
3986                             void *opaque)
3987 {
3988     qemu_progress_print(100.f * offset / total_work_size, 0);
3989 }
3990 
3991 static int print_amend_option_help(const char *format)
3992 {
3993     BlockDriver *drv;
3994 
3995     /* Find driver and parse its options */
3996     drv = bdrv_find_format(format);
3997     if (!drv) {
3998         error_report("Unknown file format '%s'", format);
3999         return 1;
4000     }
4001 
4002     if (!drv->bdrv_amend_options) {
4003         error_report("Format driver '%s' does not support option amendment",
4004                      format);
4005         return 1;
4006     }
4007 
4008     /* Every driver supporting amendment must have amend_opts */
4009     assert(drv->amend_opts);
4010 
4011     printf("Amend options for '%s':\n", format);
4012     qemu_opts_print_help(drv->amend_opts, false);
4013     return 0;
4014 }
4015 
4016 static int img_amend(int argc, char **argv)
4017 {
4018     Error *err = NULL;
4019     int c, ret = 0;
4020     char *options = NULL;
4021     QemuOptsList *amend_opts = NULL;
4022     QemuOpts *opts = NULL;
4023     const char *fmt = NULL, *filename, *cache;
4024     int flags;
4025     bool writethrough;
4026     bool quiet = false, progress = false;
4027     BlockBackend *blk = NULL;
4028     BlockDriverState *bs = NULL;
4029     bool image_opts = false;
4030     bool force = false;
4031 
4032     cache = BDRV_DEFAULT_CACHE;
4033     for (;;) {
4034         static const struct option long_options[] = {
4035             {"help", no_argument, 0, 'h'},
4036             {"object", required_argument, 0, OPTION_OBJECT},
4037             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
4038             {"force", no_argument, 0, OPTION_FORCE},
4039             {0, 0, 0, 0}
4040         };
4041         c = getopt_long(argc, argv, ":ho:f:t:pq",
4042                         long_options, NULL);
4043         if (c == -1) {
4044             break;
4045         }
4046 
4047         switch (c) {
4048         case ':':
4049             missing_argument(argv[optind - 1]);
4050             break;
4051         case '?':
4052             unrecognized_option(argv[optind - 1]);
4053             break;
4054         case 'h':
4055             help();
4056             break;
4057         case 'o':
4058             if (accumulate_options(&options, optarg) < 0) {
4059                 ret = -1;
4060                 goto out_no_progress;
4061             }
4062             break;
4063         case 'f':
4064             fmt = optarg;
4065             break;
4066         case 't':
4067             cache = optarg;
4068             break;
4069         case 'p':
4070             progress = true;
4071             break;
4072         case 'q':
4073             quiet = true;
4074             break;
4075         case OPTION_OBJECT:
4076             user_creatable_process_cmdline(optarg);
4077             break;
4078         case OPTION_IMAGE_OPTS:
4079             image_opts = true;
4080             break;
4081         case OPTION_FORCE:
4082             force = true;
4083             break;
4084         }
4085     }
4086 
4087     if (!options) {
4088         error_exit("Must specify options (-o)");
4089     }
4090 
4091     if (quiet) {
4092         progress = false;
4093     }
4094     qemu_progress_init(progress, 1.0);
4095 
4096     filename = (optind == argc - 1) ? argv[argc - 1] : NULL;
4097     if (fmt && has_help_option(options)) {
4098         /* If a format is explicitly specified (and possibly no filename is
4099          * given), print option help here */
4100         ret = print_amend_option_help(fmt);
4101         goto out;
4102     }
4103 
4104     if (optind != argc - 1) {
4105         error_report("Expecting one image file name");
4106         ret = -1;
4107         goto out;
4108     }
4109 
4110     flags = BDRV_O_RDWR;
4111     ret = bdrv_parse_cache_mode(cache, &flags, &writethrough);
4112     if (ret < 0) {
4113         error_report("Invalid cache option: %s", cache);
4114         goto out;
4115     }
4116 
4117     blk = img_open(image_opts, filename, fmt, flags, writethrough, quiet,
4118                    false);
4119     if (!blk) {
4120         ret = -1;
4121         goto out;
4122     }
4123     bs = blk_bs(blk);
4124 
4125     fmt = bs->drv->format_name;
4126 
4127     if (has_help_option(options)) {
4128         /* If the format was auto-detected, print option help here */
4129         ret = print_amend_option_help(fmt);
4130         goto out;
4131     }
4132 
4133     if (!bs->drv->bdrv_amend_options) {
4134         error_report("Format driver '%s' does not support option amendment",
4135                      fmt);
4136         ret = -1;
4137         goto out;
4138     }
4139 
4140     /* Every driver supporting amendment must have amend_opts */
4141     assert(bs->drv->amend_opts);
4142 
4143     amend_opts = qemu_opts_append(amend_opts, bs->drv->amend_opts);
4144     opts = qemu_opts_create(amend_opts, NULL, 0, &error_abort);
4145     if (!qemu_opts_do_parse(opts, options, NULL, &err)) {
4146         /* Try to parse options using the create options */
4147         amend_opts = qemu_opts_append(amend_opts, bs->drv->create_opts);
4148         qemu_opts_del(opts);
4149         opts = qemu_opts_create(amend_opts, NULL, 0, &error_abort);
4150         if (qemu_opts_do_parse(opts, options, NULL, NULL)) {
4151             error_append_hint(&err,
4152                               "This option is only supported for image creation\n");
4153         }
4154 
4155         error_report_err(err);
4156         ret = -1;
4157         goto out;
4158     }
4159 
4160     /* In case the driver does not call amend_status_cb() */
4161     qemu_progress_print(0.f, 0);
4162     ret = bdrv_amend_options(bs, opts, &amend_status_cb, NULL, force, &err);
4163     qemu_progress_print(100.f, 0);
4164     if (ret < 0) {
4165         error_report_err(err);
4166         goto out;
4167     }
4168 
4169 out:
4170     qemu_progress_end();
4171 
4172 out_no_progress:
4173     blk_unref(blk);
4174     qemu_opts_del(opts);
4175     qemu_opts_free(amend_opts);
4176     g_free(options);
4177 
4178     if (ret) {
4179         return 1;
4180     }
4181     return 0;
4182 }
4183 
4184 typedef struct BenchData {
4185     BlockBackend *blk;
4186     uint64_t image_size;
4187     bool write;
4188     int bufsize;
4189     int step;
4190     int nrreq;
4191     int n;
4192     int flush_interval;
4193     bool drain_on_flush;
4194     uint8_t *buf;
4195     QEMUIOVector *qiov;
4196 
4197     int in_flight;
4198     bool in_flush;
4199     uint64_t offset;
4200 } BenchData;
4201 
4202 static void bench_undrained_flush_cb(void *opaque, int ret)
4203 {
4204     if (ret < 0) {
4205         error_report("Failed flush request: %s", strerror(-ret));
4206         exit(EXIT_FAILURE);
4207     }
4208 }
4209 
4210 static void bench_cb(void *opaque, int ret)
4211 {
4212     BenchData *b = opaque;
4213     BlockAIOCB *acb;
4214 
4215     if (ret < 0) {
4216         error_report("Failed request: %s", strerror(-ret));
4217         exit(EXIT_FAILURE);
4218     }
4219 
4220     if (b->in_flush) {
4221         /* Just finished a flush with drained queue: Start next requests */
4222         assert(b->in_flight == 0);
4223         b->in_flush = false;
4224     } else if (b->in_flight > 0) {
4225         int remaining = b->n - b->in_flight;
4226 
4227         b->n--;
4228         b->in_flight--;
4229 
4230         /* Time for flush? Drain queue if requested, then flush */
4231         if (b->flush_interval && remaining % b->flush_interval == 0) {
4232             if (!b->in_flight || !b->drain_on_flush) {
4233                 BlockCompletionFunc *cb;
4234 
4235                 if (b->drain_on_flush) {
4236                     b->in_flush = true;
4237                     cb = bench_cb;
4238                 } else {
4239                     cb = bench_undrained_flush_cb;
4240                 }
4241 
4242                 acb = blk_aio_flush(b->blk, cb, b);
4243                 if (!acb) {
4244                     error_report("Failed to issue flush request");
4245                     exit(EXIT_FAILURE);
4246                 }
4247             }
4248             if (b->drain_on_flush) {
4249                 return;
4250             }
4251         }
4252     }
4253 
4254     while (b->n > b->in_flight && b->in_flight < b->nrreq) {
4255         int64_t offset = b->offset;
4256         /* blk_aio_* might look for completed I/Os and kick bench_cb
4257          * again, so make sure this operation is counted by in_flight
4258          * and b->offset is ready for the next submission.
4259          */
4260         b->in_flight++;
4261         b->offset += b->step;
4262         b->offset %= b->image_size;
4263         if (b->write) {
4264             acb = blk_aio_pwritev(b->blk, offset, b->qiov, 0, bench_cb, b);
4265         } else {
4266             acb = blk_aio_preadv(b->blk, offset, b->qiov, 0, bench_cb, b);
4267         }
4268         if (!acb) {
4269             error_report("Failed to issue request");
4270             exit(EXIT_FAILURE);
4271         }
4272     }
4273 }
4274 
4275 static int img_bench(int argc, char **argv)
4276 {
4277     int c, ret = 0;
4278     const char *fmt = NULL, *filename;
4279     bool quiet = false;
4280     bool image_opts = false;
4281     bool is_write = false;
4282     int count = 75000;
4283     int depth = 64;
4284     int64_t offset = 0;
4285     size_t bufsize = 4096;
4286     int pattern = 0;
4287     size_t step = 0;
4288     int flush_interval = 0;
4289     bool drain_on_flush = true;
4290     int64_t image_size;
4291     BlockBackend *blk = NULL;
4292     BenchData data = {};
4293     int flags = 0;
4294     bool writethrough = false;
4295     struct timeval t1, t2;
4296     int i;
4297     bool force_share = false;
4298     size_t buf_size;
4299 
4300     for (;;) {
4301         static const struct option long_options[] = {
4302             {"help", no_argument, 0, 'h'},
4303             {"flush-interval", required_argument, 0, OPTION_FLUSH_INTERVAL},
4304             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
4305             {"pattern", required_argument, 0, OPTION_PATTERN},
4306             {"no-drain", no_argument, 0, OPTION_NO_DRAIN},
4307             {"force-share", no_argument, 0, 'U'},
4308             {0, 0, 0, 0}
4309         };
4310         c = getopt_long(argc, argv, ":hc:d:f:ni:o:qs:S:t:wU", long_options,
4311                         NULL);
4312         if (c == -1) {
4313             break;
4314         }
4315 
4316         switch (c) {
4317         case ':':
4318             missing_argument(argv[optind - 1]);
4319             break;
4320         case '?':
4321             unrecognized_option(argv[optind - 1]);
4322             break;
4323         case 'h':
4324             help();
4325             break;
4326         case 'c':
4327         {
4328             unsigned long res;
4329 
4330             if (qemu_strtoul(optarg, NULL, 0, &res) < 0 || res > INT_MAX) {
4331                 error_report("Invalid request count specified");
4332                 return 1;
4333             }
4334             count = res;
4335             break;
4336         }
4337         case 'd':
4338         {
4339             unsigned long res;
4340 
4341             if (qemu_strtoul(optarg, NULL, 0, &res) < 0 || res > INT_MAX) {
4342                 error_report("Invalid queue depth specified");
4343                 return 1;
4344             }
4345             depth = res;
4346             break;
4347         }
4348         case 'f':
4349             fmt = optarg;
4350             break;
4351         case 'n':
4352             flags |= BDRV_O_NATIVE_AIO;
4353             break;
4354         case 'i':
4355             ret = bdrv_parse_aio(optarg, &flags);
4356             if (ret < 0) {
4357                 error_report("Invalid aio option: %s", optarg);
4358                 ret = -1;
4359                 goto out;
4360             }
4361             break;
4362         case 'o':
4363         {
4364             offset = cvtnum("offset", optarg);
4365             if (offset < 0) {
4366                 return 1;
4367             }
4368             break;
4369         }
4370             break;
4371         case 'q':
4372             quiet = true;
4373             break;
4374         case 's':
4375         {
4376             int64_t sval;
4377 
4378             sval = cvtnum_full("buffer size", optarg, 0, INT_MAX);
4379             if (sval < 0) {
4380                 return 1;
4381             }
4382 
4383             bufsize = sval;
4384             break;
4385         }
4386         case 'S':
4387         {
4388             int64_t sval;
4389 
4390             sval = cvtnum_full("step_size", optarg, 0, INT_MAX);
4391             if (sval < 0) {
4392                 return 1;
4393             }
4394 
4395             step = sval;
4396             break;
4397         }
4398         case 't':
4399             ret = bdrv_parse_cache_mode(optarg, &flags, &writethrough);
4400             if (ret < 0) {
4401                 error_report("Invalid cache mode");
4402                 ret = -1;
4403                 goto out;
4404             }
4405             break;
4406         case 'w':
4407             flags |= BDRV_O_RDWR;
4408             is_write = true;
4409             break;
4410         case 'U':
4411             force_share = true;
4412             break;
4413         case OPTION_PATTERN:
4414         {
4415             unsigned long res;
4416 
4417             if (qemu_strtoul(optarg, NULL, 0, &res) < 0 || res > 0xff) {
4418                 error_report("Invalid pattern byte specified");
4419                 return 1;
4420             }
4421             pattern = res;
4422             break;
4423         }
4424         case OPTION_FLUSH_INTERVAL:
4425         {
4426             unsigned long res;
4427 
4428             if (qemu_strtoul(optarg, NULL, 0, &res) < 0 || res > INT_MAX) {
4429                 error_report("Invalid flush interval specified");
4430                 return 1;
4431             }
4432             flush_interval = res;
4433             break;
4434         }
4435         case OPTION_NO_DRAIN:
4436             drain_on_flush = false;
4437             break;
4438         case OPTION_IMAGE_OPTS:
4439             image_opts = true;
4440             break;
4441         }
4442     }
4443 
4444     if (optind != argc - 1) {
4445         error_exit("Expecting one image file name");
4446     }
4447     filename = argv[argc - 1];
4448 
4449     if (!is_write && flush_interval) {
4450         error_report("--flush-interval is only available in write tests");
4451         ret = -1;
4452         goto out;
4453     }
4454     if (flush_interval && flush_interval < depth) {
4455         error_report("Flush interval can't be smaller than depth");
4456         ret = -1;
4457         goto out;
4458     }
4459 
4460     blk = img_open(image_opts, filename, fmt, flags, writethrough, quiet,
4461                    force_share);
4462     if (!blk) {
4463         ret = -1;
4464         goto out;
4465     }
4466 
4467     image_size = blk_getlength(blk);
4468     if (image_size < 0) {
4469         ret = image_size;
4470         goto out;
4471     }
4472 
4473     data = (BenchData) {
4474         .blk            = blk,
4475         .image_size     = image_size,
4476         .bufsize        = bufsize,
4477         .step           = step ?: bufsize,
4478         .nrreq          = depth,
4479         .n              = count,
4480         .offset         = offset,
4481         .write          = is_write,
4482         .flush_interval = flush_interval,
4483         .drain_on_flush = drain_on_flush,
4484     };
4485     printf("Sending %d %s requests, %d bytes each, %d in parallel "
4486            "(starting at offset %" PRId64 ", step size %d)\n",
4487            data.n, data.write ? "write" : "read", data.bufsize, data.nrreq,
4488            data.offset, data.step);
4489     if (flush_interval) {
4490         printf("Sending flush every %d requests\n", flush_interval);
4491     }
4492 
4493     buf_size = data.nrreq * data.bufsize;
4494     data.buf = blk_blockalign(blk, buf_size);
4495     memset(data.buf, pattern, data.nrreq * data.bufsize);
4496 
4497     blk_register_buf(blk, data.buf, buf_size);
4498 
4499     data.qiov = g_new(QEMUIOVector, data.nrreq);
4500     for (i = 0; i < data.nrreq; i++) {
4501         qemu_iovec_init(&data.qiov[i], 1);
4502         qemu_iovec_add(&data.qiov[i],
4503                        data.buf + i * data.bufsize, data.bufsize);
4504     }
4505 
4506     gettimeofday(&t1, NULL);
4507     bench_cb(&data, 0);
4508 
4509     while (data.n > 0) {
4510         main_loop_wait(false);
4511     }
4512     gettimeofday(&t2, NULL);
4513 
4514     printf("Run completed in %3.3f seconds.\n",
4515            (t2.tv_sec - t1.tv_sec)
4516            + ((double)(t2.tv_usec - t1.tv_usec) / 1000000));
4517 
4518 out:
4519     if (data.buf) {
4520         blk_unregister_buf(blk, data.buf);
4521     }
4522     qemu_vfree(data.buf);
4523     blk_unref(blk);
4524 
4525     if (ret) {
4526         return 1;
4527     }
4528     return 0;
4529 }
4530 
4531 enum ImgBitmapAct {
4532     BITMAP_ADD,
4533     BITMAP_REMOVE,
4534     BITMAP_CLEAR,
4535     BITMAP_ENABLE,
4536     BITMAP_DISABLE,
4537     BITMAP_MERGE,
4538 };
4539 typedef struct ImgBitmapAction {
4540     enum ImgBitmapAct act;
4541     const char *src; /* only used for merge */
4542     QSIMPLEQ_ENTRY(ImgBitmapAction) next;
4543 } ImgBitmapAction;
4544 
4545 static int img_bitmap(int argc, char **argv)
4546 {
4547     Error *err = NULL;
4548     int c, ret = 1;
4549     QemuOpts *opts = NULL;
4550     const char *fmt = NULL, *src_fmt = NULL, *src_filename = NULL;
4551     const char *filename, *bitmap;
4552     BlockBackend *blk = NULL, *src = NULL;
4553     BlockDriverState *bs = NULL, *src_bs = NULL;
4554     bool image_opts = false;
4555     int64_t granularity = 0;
4556     bool add = false, merge = false;
4557     QSIMPLEQ_HEAD(, ImgBitmapAction) actions;
4558     ImgBitmapAction *act, *act_next;
4559     const char *op;
4560 
4561     QSIMPLEQ_INIT(&actions);
4562 
4563     for (;;) {
4564         static const struct option long_options[] = {
4565             {"help", no_argument, 0, 'h'},
4566             {"object", required_argument, 0, OPTION_OBJECT},
4567             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
4568             {"add", no_argument, 0, OPTION_ADD},
4569             {"remove", no_argument, 0, OPTION_REMOVE},
4570             {"clear", no_argument, 0, OPTION_CLEAR},
4571             {"enable", no_argument, 0, OPTION_ENABLE},
4572             {"disable", no_argument, 0, OPTION_DISABLE},
4573             {"merge", required_argument, 0, OPTION_MERGE},
4574             {"granularity", required_argument, 0, 'g'},
4575             {"source-file", required_argument, 0, 'b'},
4576             {"source-format", required_argument, 0, 'F'},
4577             {0, 0, 0, 0}
4578         };
4579         c = getopt_long(argc, argv, ":b:f:F:g:h", long_options, NULL);
4580         if (c == -1) {
4581             break;
4582         }
4583 
4584         switch (c) {
4585         case ':':
4586             missing_argument(argv[optind - 1]);
4587             break;
4588         case '?':
4589             unrecognized_option(argv[optind - 1]);
4590             break;
4591         case 'h':
4592             help();
4593             break;
4594         case 'b':
4595             src_filename = optarg;
4596             break;
4597         case 'f':
4598             fmt = optarg;
4599             break;
4600         case 'F':
4601             src_fmt = optarg;
4602             break;
4603         case 'g':
4604             granularity = cvtnum("granularity", optarg);
4605             if (granularity < 0) {
4606                 return 1;
4607             }
4608             break;
4609         case OPTION_ADD:
4610             act = g_new0(ImgBitmapAction, 1);
4611             act->act = BITMAP_ADD;
4612             QSIMPLEQ_INSERT_TAIL(&actions, act, next);
4613             add = true;
4614             break;
4615         case OPTION_REMOVE:
4616             act = g_new0(ImgBitmapAction, 1);
4617             act->act = BITMAP_REMOVE;
4618             QSIMPLEQ_INSERT_TAIL(&actions, act, next);
4619             break;
4620         case OPTION_CLEAR:
4621             act = g_new0(ImgBitmapAction, 1);
4622             act->act = BITMAP_CLEAR;
4623             QSIMPLEQ_INSERT_TAIL(&actions, act, next);
4624             break;
4625         case OPTION_ENABLE:
4626             act = g_new0(ImgBitmapAction, 1);
4627             act->act = BITMAP_ENABLE;
4628             QSIMPLEQ_INSERT_TAIL(&actions, act, next);
4629             break;
4630         case OPTION_DISABLE:
4631             act = g_new0(ImgBitmapAction, 1);
4632             act->act = BITMAP_DISABLE;
4633             QSIMPLEQ_INSERT_TAIL(&actions, act, next);
4634             break;
4635         case OPTION_MERGE:
4636             act = g_new0(ImgBitmapAction, 1);
4637             act->act = BITMAP_MERGE;
4638             act->src = optarg;
4639             QSIMPLEQ_INSERT_TAIL(&actions, act, next);
4640             merge = true;
4641             break;
4642         case OPTION_OBJECT:
4643             user_creatable_process_cmdline(optarg);
4644             break;
4645         case OPTION_IMAGE_OPTS:
4646             image_opts = true;
4647             break;
4648         }
4649     }
4650 
4651     if (QSIMPLEQ_EMPTY(&actions)) {
4652         error_report("Need at least one of --add, --remove, --clear, "
4653                      "--enable, --disable, or --merge");
4654         goto out;
4655     }
4656 
4657     if (granularity && !add) {
4658         error_report("granularity only supported with --add");
4659         goto out;
4660     }
4661     if (src_fmt && !src_filename) {
4662         error_report("-F only supported with -b");
4663         goto out;
4664     }
4665     if (src_filename && !merge) {
4666         error_report("Merge bitmap source file only supported with "
4667                      "--merge");
4668         goto out;
4669     }
4670 
4671     if (optind != argc - 2) {
4672         error_report("Expecting filename and bitmap name");
4673         goto out;
4674     }
4675 
4676     filename = argv[optind];
4677     bitmap = argv[optind + 1];
4678 
4679     /*
4680      * No need to open backing chains; we will be manipulating bitmaps
4681      * directly in this image without reference to image contents.
4682      */
4683     blk = img_open(image_opts, filename, fmt, BDRV_O_RDWR | BDRV_O_NO_BACKING,
4684                    false, false, false);
4685     if (!blk) {
4686         goto out;
4687     }
4688     bs = blk_bs(blk);
4689     if (src_filename) {
4690         src = img_open(false, src_filename, src_fmt, BDRV_O_NO_BACKING,
4691                        false, false, false);
4692         if (!src) {
4693             goto out;
4694         }
4695         src_bs = blk_bs(src);
4696     } else {
4697         src_bs = bs;
4698     }
4699 
4700     QSIMPLEQ_FOREACH_SAFE(act, &actions, next, act_next) {
4701         switch (act->act) {
4702         case BITMAP_ADD:
4703             qmp_block_dirty_bitmap_add(bs->node_name, bitmap,
4704                                        !!granularity, granularity, true, true,
4705                                        false, false, &err);
4706             op = "add";
4707             break;
4708         case BITMAP_REMOVE:
4709             qmp_block_dirty_bitmap_remove(bs->node_name, bitmap, &err);
4710             op = "remove";
4711             break;
4712         case BITMAP_CLEAR:
4713             qmp_block_dirty_bitmap_clear(bs->node_name, bitmap, &err);
4714             op = "clear";
4715             break;
4716         case BITMAP_ENABLE:
4717             qmp_block_dirty_bitmap_enable(bs->node_name, bitmap, &err);
4718             op = "enable";
4719             break;
4720         case BITMAP_DISABLE:
4721             qmp_block_dirty_bitmap_disable(bs->node_name, bitmap, &err);
4722             op = "disable";
4723             break;
4724         case BITMAP_MERGE:
4725             do_dirty_bitmap_merge(bs->node_name, bitmap, src_bs->node_name,
4726                                   act->src, &err);
4727             op = "merge";
4728             break;
4729         default:
4730             g_assert_not_reached();
4731         }
4732 
4733         if (err) {
4734             error_reportf_err(err, "Operation %s on bitmap %s failed: ",
4735                               op, bitmap);
4736             goto out;
4737         }
4738         g_free(act);
4739     }
4740 
4741     ret = 0;
4742 
4743  out:
4744     blk_unref(src);
4745     blk_unref(blk);
4746     qemu_opts_del(opts);
4747     return ret;
4748 }
4749 
4750 #define C_BS      01
4751 #define C_COUNT   02
4752 #define C_IF      04
4753 #define C_OF      010
4754 #define C_SKIP    020
4755 
4756 struct DdInfo {
4757     unsigned int flags;
4758     int64_t count;
4759 };
4760 
4761 struct DdIo {
4762     int bsz;    /* Block size */
4763     char *filename;
4764     uint8_t *buf;
4765     int64_t offset;
4766 };
4767 
4768 struct DdOpts {
4769     const char *name;
4770     int (*f)(const char *, struct DdIo *, struct DdIo *, struct DdInfo *);
4771     unsigned int flag;
4772 };
4773 
4774 static int img_dd_bs(const char *arg,
4775                      struct DdIo *in, struct DdIo *out,
4776                      struct DdInfo *dd)
4777 {
4778     int64_t res;
4779 
4780     res = cvtnum_full("bs", arg, 1, INT_MAX);
4781 
4782     if (res < 0) {
4783         return 1;
4784     }
4785     in->bsz = out->bsz = res;
4786 
4787     return 0;
4788 }
4789 
4790 static int img_dd_count(const char *arg,
4791                         struct DdIo *in, struct DdIo *out,
4792                         struct DdInfo *dd)
4793 {
4794     dd->count = cvtnum("count", arg);
4795 
4796     if (dd->count < 0) {
4797         return 1;
4798     }
4799 
4800     return 0;
4801 }
4802 
4803 static int img_dd_if(const char *arg,
4804                      struct DdIo *in, struct DdIo *out,
4805                      struct DdInfo *dd)
4806 {
4807     in->filename = g_strdup(arg);
4808 
4809     return 0;
4810 }
4811 
4812 static int img_dd_of(const char *arg,
4813                      struct DdIo *in, struct DdIo *out,
4814                      struct DdInfo *dd)
4815 {
4816     out->filename = g_strdup(arg);
4817 
4818     return 0;
4819 }
4820 
4821 static int img_dd_skip(const char *arg,
4822                        struct DdIo *in, struct DdIo *out,
4823                        struct DdInfo *dd)
4824 {
4825     in->offset = cvtnum("skip", arg);
4826 
4827     if (in->offset < 0) {
4828         return 1;
4829     }
4830 
4831     return 0;
4832 }
4833 
4834 static int img_dd(int argc, char **argv)
4835 {
4836     int ret = 0;
4837     char *arg = NULL;
4838     char *tmp;
4839     BlockDriver *drv = NULL, *proto_drv = NULL;
4840     BlockBackend *blk1 = NULL, *blk2 = NULL;
4841     QemuOpts *opts = NULL;
4842     QemuOptsList *create_opts = NULL;
4843     Error *local_err = NULL;
4844     bool image_opts = false;
4845     int c, i;
4846     const char *out_fmt = "raw";
4847     const char *fmt = NULL;
4848     int64_t size = 0;
4849     int64_t block_count = 0, out_pos, in_pos;
4850     bool force_share = false;
4851     struct DdInfo dd = {
4852         .flags = 0,
4853         .count = 0,
4854     };
4855     struct DdIo in = {
4856         .bsz = 512, /* Block size is by default 512 bytes */
4857         .filename = NULL,
4858         .buf = NULL,
4859         .offset = 0
4860     };
4861     struct DdIo out = {
4862         .bsz = 512,
4863         .filename = NULL,
4864         .buf = NULL,
4865         .offset = 0
4866     };
4867 
4868     const struct DdOpts options[] = {
4869         { "bs", img_dd_bs, C_BS },
4870         { "count", img_dd_count, C_COUNT },
4871         { "if", img_dd_if, C_IF },
4872         { "of", img_dd_of, C_OF },
4873         { "skip", img_dd_skip, C_SKIP },
4874         { NULL, NULL, 0 }
4875     };
4876     const struct option long_options[] = {
4877         { "help", no_argument, 0, 'h'},
4878         { "object", required_argument, 0, OPTION_OBJECT},
4879         { "image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
4880         { "force-share", no_argument, 0, 'U'},
4881         { 0, 0, 0, 0 }
4882     };
4883 
4884     while ((c = getopt_long(argc, argv, ":hf:O:U", long_options, NULL))) {
4885         if (c == EOF) {
4886             break;
4887         }
4888         switch (c) {
4889         case 'O':
4890             out_fmt = optarg;
4891             break;
4892         case 'f':
4893             fmt = optarg;
4894             break;
4895         case ':':
4896             missing_argument(argv[optind - 1]);
4897             break;
4898         case '?':
4899             unrecognized_option(argv[optind - 1]);
4900             break;
4901         case 'h':
4902             help();
4903             break;
4904         case 'U':
4905             force_share = true;
4906             break;
4907         case OPTION_OBJECT:
4908             user_creatable_process_cmdline(optarg);
4909             break;
4910         case OPTION_IMAGE_OPTS:
4911             image_opts = true;
4912             break;
4913         }
4914     }
4915 
4916     for (i = optind; i < argc; i++) {
4917         int j;
4918         arg = g_strdup(argv[i]);
4919 
4920         tmp = strchr(arg, '=');
4921         if (tmp == NULL) {
4922             error_report("unrecognized operand %s", arg);
4923             ret = -1;
4924             goto out;
4925         }
4926 
4927         *tmp++ = '\0';
4928 
4929         for (j = 0; options[j].name != NULL; j++) {
4930             if (!strcmp(arg, options[j].name)) {
4931                 break;
4932             }
4933         }
4934         if (options[j].name == NULL) {
4935             error_report("unrecognized operand %s", arg);
4936             ret = -1;
4937             goto out;
4938         }
4939 
4940         if (options[j].f(tmp, &in, &out, &dd) != 0) {
4941             ret = -1;
4942             goto out;
4943         }
4944         dd.flags |= options[j].flag;
4945         g_free(arg);
4946         arg = NULL;
4947     }
4948 
4949     if (!(dd.flags & C_IF && dd.flags & C_OF)) {
4950         error_report("Must specify both input and output files");
4951         ret = -1;
4952         goto out;
4953     }
4954 
4955     blk1 = img_open(image_opts, in.filename, fmt, 0, false, false,
4956                     force_share);
4957 
4958     if (!blk1) {
4959         ret = -1;
4960         goto out;
4961     }
4962 
4963     drv = bdrv_find_format(out_fmt);
4964     if (!drv) {
4965         error_report("Unknown file format");
4966         ret = -1;
4967         goto out;
4968     }
4969     proto_drv = bdrv_find_protocol(out.filename, true, &local_err);
4970 
4971     if (!proto_drv) {
4972         error_report_err(local_err);
4973         ret = -1;
4974         goto out;
4975     }
4976     if (!drv->create_opts) {
4977         error_report("Format driver '%s' does not support image creation",
4978                      drv->format_name);
4979         ret = -1;
4980         goto out;
4981     }
4982     if (!proto_drv->create_opts) {
4983         error_report("Protocol driver '%s' does not support image creation",
4984                      proto_drv->format_name);
4985         ret = -1;
4986         goto out;
4987     }
4988     create_opts = qemu_opts_append(create_opts, drv->create_opts);
4989     create_opts = qemu_opts_append(create_opts, proto_drv->create_opts);
4990 
4991     opts = qemu_opts_create(create_opts, NULL, 0, &error_abort);
4992 
4993     size = blk_getlength(blk1);
4994     if (size < 0) {
4995         error_report("Failed to get size for '%s'", in.filename);
4996         ret = -1;
4997         goto out;
4998     }
4999 
5000     if (dd.flags & C_COUNT && dd.count <= INT64_MAX / in.bsz &&
5001         dd.count * in.bsz < size) {
5002         size = dd.count * in.bsz;
5003     }
5004 
5005     /* Overflow means the specified offset is beyond input image's size */
5006     if (dd.flags & C_SKIP && (in.offset > INT64_MAX / in.bsz ||
5007                               size < in.bsz * in.offset)) {
5008         qemu_opt_set_number(opts, BLOCK_OPT_SIZE, 0, &error_abort);
5009     } else {
5010         qemu_opt_set_number(opts, BLOCK_OPT_SIZE,
5011                             size - in.bsz * in.offset, &error_abort);
5012     }
5013 
5014     ret = bdrv_create(drv, out.filename, opts, &local_err);
5015     if (ret < 0) {
5016         error_reportf_err(local_err,
5017                           "%s: error while creating output image: ",
5018                           out.filename);
5019         ret = -1;
5020         goto out;
5021     }
5022 
5023     /* TODO, we can't honour --image-opts for the target,
5024      * since it needs to be given in a format compatible
5025      * with the bdrv_create() call above which does not
5026      * support image-opts style.
5027      */
5028     blk2 = img_open_file(out.filename, NULL, out_fmt, BDRV_O_RDWR,
5029                          false, false, false);
5030 
5031     if (!blk2) {
5032         ret = -1;
5033         goto out;
5034     }
5035 
5036     if (dd.flags & C_SKIP && (in.offset > INT64_MAX / in.bsz ||
5037                               size < in.offset * in.bsz)) {
5038         /* We give a warning if the skip option is bigger than the input
5039          * size and create an empty output disk image (i.e. like dd(1)).
5040          */
5041         error_report("%s: cannot skip to specified offset", in.filename);
5042         in_pos = size;
5043     } else {
5044         in_pos = in.offset * in.bsz;
5045     }
5046 
5047     in.buf = g_new(uint8_t, in.bsz);
5048 
5049     for (out_pos = 0; in_pos < size; block_count++) {
5050         int in_ret, out_ret;
5051 
5052         if (in_pos + in.bsz > size) {
5053             in_ret = blk_pread(blk1, in_pos, in.buf, size - in_pos);
5054         } else {
5055             in_ret = blk_pread(blk1, in_pos, in.buf, in.bsz);
5056         }
5057         if (in_ret < 0) {
5058             error_report("error while reading from input image file: %s",
5059                          strerror(-in_ret));
5060             ret = -1;
5061             goto out;
5062         }
5063         in_pos += in_ret;
5064 
5065         out_ret = blk_pwrite(blk2, out_pos, in.buf, in_ret, 0);
5066 
5067         if (out_ret < 0) {
5068             error_report("error while writing to output image file: %s",
5069                          strerror(-out_ret));
5070             ret = -1;
5071             goto out;
5072         }
5073         out_pos += out_ret;
5074     }
5075 
5076 out:
5077     g_free(arg);
5078     qemu_opts_del(opts);
5079     qemu_opts_free(create_opts);
5080     blk_unref(blk1);
5081     blk_unref(blk2);
5082     g_free(in.filename);
5083     g_free(out.filename);
5084     g_free(in.buf);
5085     g_free(out.buf);
5086 
5087     if (ret) {
5088         return 1;
5089     }
5090     return 0;
5091 }
5092 
5093 static void dump_json_block_measure_info(BlockMeasureInfo *info)
5094 {
5095     GString *str;
5096     QObject *obj;
5097     Visitor *v = qobject_output_visitor_new(&obj);
5098 
5099     visit_type_BlockMeasureInfo(v, NULL, &info, &error_abort);
5100     visit_complete(v, &obj);
5101     str = qobject_to_json_pretty(obj, true);
5102     assert(str != NULL);
5103     printf("%s\n", str->str);
5104     qobject_unref(obj);
5105     visit_free(v);
5106     g_string_free(str, true);
5107 }
5108 
5109 static int img_measure(int argc, char **argv)
5110 {
5111     static const struct option long_options[] = {
5112         {"help", no_argument, 0, 'h'},
5113         {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
5114         {"object", required_argument, 0, OPTION_OBJECT},
5115         {"output", required_argument, 0, OPTION_OUTPUT},
5116         {"size", required_argument, 0, OPTION_SIZE},
5117         {"force-share", no_argument, 0, 'U'},
5118         {0, 0, 0, 0}
5119     };
5120     OutputFormat output_format = OFORMAT_HUMAN;
5121     BlockBackend *in_blk = NULL;
5122     BlockDriver *drv;
5123     const char *filename = NULL;
5124     const char *fmt = NULL;
5125     const char *out_fmt = "raw";
5126     char *options = NULL;
5127     char *snapshot_name = NULL;
5128     bool force_share = false;
5129     QemuOpts *opts = NULL;
5130     QemuOpts *object_opts = NULL;
5131     QemuOpts *sn_opts = NULL;
5132     QemuOptsList *create_opts = NULL;
5133     bool image_opts = false;
5134     uint64_t img_size = UINT64_MAX;
5135     BlockMeasureInfo *info = NULL;
5136     Error *local_err = NULL;
5137     int ret = 1;
5138     int c;
5139 
5140     while ((c = getopt_long(argc, argv, "hf:O:o:l:U",
5141                             long_options, NULL)) != -1) {
5142         switch (c) {
5143         case '?':
5144         case 'h':
5145             help();
5146             break;
5147         case 'f':
5148             fmt = optarg;
5149             break;
5150         case 'O':
5151             out_fmt = optarg;
5152             break;
5153         case 'o':
5154             if (accumulate_options(&options, optarg) < 0) {
5155                 goto out;
5156             }
5157             break;
5158         case 'l':
5159             if (strstart(optarg, SNAPSHOT_OPT_BASE, NULL)) {
5160                 sn_opts = qemu_opts_parse_noisily(&internal_snapshot_opts,
5161                                                   optarg, false);
5162                 if (!sn_opts) {
5163                     error_report("Failed in parsing snapshot param '%s'",
5164                                  optarg);
5165                     goto out;
5166                 }
5167             } else {
5168                 snapshot_name = optarg;
5169             }
5170             break;
5171         case 'U':
5172             force_share = true;
5173             break;
5174         case OPTION_OBJECT:
5175             user_creatable_process_cmdline(optarg);
5176             break;
5177         case OPTION_IMAGE_OPTS:
5178             image_opts = true;
5179             break;
5180         case OPTION_OUTPUT:
5181             if (!strcmp(optarg, "json")) {
5182                 output_format = OFORMAT_JSON;
5183             } else if (!strcmp(optarg, "human")) {
5184                 output_format = OFORMAT_HUMAN;
5185             } else {
5186                 error_report("--output must be used with human or json "
5187                              "as argument.");
5188                 goto out;
5189             }
5190             break;
5191         case OPTION_SIZE:
5192         {
5193             int64_t sval;
5194 
5195             sval = cvtnum("image size", optarg);
5196             if (sval < 0) {
5197                 goto out;
5198             }
5199             img_size = (uint64_t)sval;
5200         }
5201         break;
5202         }
5203     }
5204 
5205     if (argc - optind > 1) {
5206         error_report("At most one filename argument is allowed.");
5207         goto out;
5208     } else if (argc - optind == 1) {
5209         filename = argv[optind];
5210     }
5211 
5212     if (!filename && (image_opts || fmt || snapshot_name || sn_opts)) {
5213         error_report("--image-opts, -f, and -l require a filename argument.");
5214         goto out;
5215     }
5216     if (filename && img_size != UINT64_MAX) {
5217         error_report("--size N cannot be used together with a filename.");
5218         goto out;
5219     }
5220     if (!filename && img_size == UINT64_MAX) {
5221         error_report("Either --size N or one filename must be specified.");
5222         goto out;
5223     }
5224 
5225     if (filename) {
5226         in_blk = img_open(image_opts, filename, fmt, 0,
5227                           false, false, force_share);
5228         if (!in_blk) {
5229             goto out;
5230         }
5231 
5232         if (sn_opts) {
5233             bdrv_snapshot_load_tmp(blk_bs(in_blk),
5234                     qemu_opt_get(sn_opts, SNAPSHOT_OPT_ID),
5235                     qemu_opt_get(sn_opts, SNAPSHOT_OPT_NAME),
5236                     &local_err);
5237         } else if (snapshot_name != NULL) {
5238             bdrv_snapshot_load_tmp_by_id_or_name(blk_bs(in_blk),
5239                     snapshot_name, &local_err);
5240         }
5241         if (local_err) {
5242             error_reportf_err(local_err, "Failed to load snapshot: ");
5243             goto out;
5244         }
5245     }
5246 
5247     drv = bdrv_find_format(out_fmt);
5248     if (!drv) {
5249         error_report("Unknown file format '%s'", out_fmt);
5250         goto out;
5251     }
5252     if (!drv->create_opts) {
5253         error_report("Format driver '%s' does not support image creation",
5254                      drv->format_name);
5255         goto out;
5256     }
5257 
5258     create_opts = qemu_opts_append(create_opts, drv->create_opts);
5259     create_opts = qemu_opts_append(create_opts, bdrv_file.create_opts);
5260     opts = qemu_opts_create(create_opts, NULL, 0, &error_abort);
5261     if (options) {
5262         if (!qemu_opts_do_parse(opts, options, NULL, &local_err)) {
5263             error_report_err(local_err);
5264             error_report("Invalid options for file format '%s'", out_fmt);
5265             goto out;
5266         }
5267     }
5268     if (img_size != UINT64_MAX) {
5269         qemu_opt_set_number(opts, BLOCK_OPT_SIZE, img_size, &error_abort);
5270     }
5271 
5272     info = bdrv_measure(drv, opts, in_blk ? blk_bs(in_blk) : NULL, &local_err);
5273     if (local_err) {
5274         error_report_err(local_err);
5275         goto out;
5276     }
5277 
5278     if (output_format == OFORMAT_HUMAN) {
5279         printf("required size: %" PRIu64 "\n", info->required);
5280         printf("fully allocated size: %" PRIu64 "\n", info->fully_allocated);
5281         if (info->has_bitmaps) {
5282             printf("bitmaps size: %" PRIu64 "\n", info->bitmaps);
5283         }
5284     } else {
5285         dump_json_block_measure_info(info);
5286     }
5287 
5288     ret = 0;
5289 
5290 out:
5291     qapi_free_BlockMeasureInfo(info);
5292     qemu_opts_del(object_opts);
5293     qemu_opts_del(opts);
5294     qemu_opts_del(sn_opts);
5295     qemu_opts_free(create_opts);
5296     g_free(options);
5297     blk_unref(in_blk);
5298     return ret;
5299 }
5300 
5301 static const img_cmd_t img_cmds[] = {
5302 #define DEF(option, callback, arg_string)        \
5303     { option, callback },
5304 #include "qemu-img-cmds.h"
5305 #undef DEF
5306     { NULL, NULL, },
5307 };
5308 
5309 int main(int argc, char **argv)
5310 {
5311     const img_cmd_t *cmd;
5312     const char *cmdname;
5313     Error *local_error = NULL;
5314     int c;
5315     static const struct option long_options[] = {
5316         {"help", no_argument, 0, 'h'},
5317         {"version", no_argument, 0, 'V'},
5318         {"trace", required_argument, NULL, 'T'},
5319         {0, 0, 0, 0}
5320     };
5321 
5322 #ifdef CONFIG_POSIX
5323     signal(SIGPIPE, SIG_IGN);
5324 #endif
5325 
5326     socket_init();
5327     error_init(argv[0]);
5328     module_call_init(MODULE_INIT_TRACE);
5329     qemu_init_exec_dir(argv[0]);
5330 
5331     if (qemu_init_main_loop(&local_error)) {
5332         error_report_err(local_error);
5333         exit(EXIT_FAILURE);
5334     }
5335 
5336     qcrypto_init(&error_fatal);
5337 
5338     module_call_init(MODULE_INIT_QOM);
5339     bdrv_init();
5340     if (argc < 2) {
5341         error_exit("Not enough arguments");
5342     }
5343 
5344     qemu_add_opts(&qemu_source_opts);
5345     qemu_add_opts(&qemu_trace_opts);
5346 
5347     while ((c = getopt_long(argc, argv, "+:hVT:", long_options, NULL)) != -1) {
5348         switch (c) {
5349         case ':':
5350             missing_argument(argv[optind - 1]);
5351             return 0;
5352         case '?':
5353             unrecognized_option(argv[optind - 1]);
5354             return 0;
5355         case 'h':
5356             help();
5357             return 0;
5358         case 'V':
5359             printf(QEMU_IMG_VERSION);
5360             return 0;
5361         case 'T':
5362             trace_opt_parse(optarg);
5363             break;
5364         }
5365     }
5366 
5367     cmdname = argv[optind];
5368 
5369     /* reset getopt_long scanning */
5370     argc -= optind;
5371     if (argc < 1) {
5372         return 0;
5373     }
5374     argv += optind;
5375     qemu_reset_optind();
5376 
5377     if (!trace_init_backends()) {
5378         exit(1);
5379     }
5380     trace_init_file();
5381     qemu_set_log(LOG_TRACE);
5382 
5383     /* find the command */
5384     for (cmd = img_cmds; cmd->name != NULL; cmd++) {
5385         if (!strcmp(cmdname, cmd->name)) {
5386             return cmd->handler(argc, argv);
5387         }
5388     }
5389 
5390     /* not found */
5391     error_exit("Command not found: %s", cmdname);
5392 }
5393