xref: /openbmc/qemu/qemu-img.c (revision bf7b1eab)
1 /*
2  * QEMU disk image utility
3  *
4  * Copyright (c) 2003-2008 Fabrice Bellard
5  *
6  * Permission is hereby granted, free of charge, to any person obtaining a copy
7  * of this software and associated documentation files (the "Software"), to deal
8  * in the Software without restriction, including without limitation the rights
9  * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
10  * copies of the Software, and to permit persons to whom the Software is
11  * furnished to do so, subject to the following conditions:
12  *
13  * The above copyright notice and this permission notice shall be included in
14  * all copies or substantial portions of the Software.
15  *
16  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
17  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
18  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
19  * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
20  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
21  * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
22  * THE SOFTWARE.
23  */
24 
25 #include "qemu/osdep.h"
26 #include <getopt.h>
27 
28 #include "qemu-common.h"
29 #include "qemu-version.h"
30 #include "qapi/error.h"
31 #include "qapi/qapi-commands-block-core.h"
32 #include "qapi/qapi-visit-block-core.h"
33 #include "qapi/qobject-output-visitor.h"
34 #include "qapi/qmp/qjson.h"
35 #include "qapi/qmp/qdict.h"
36 #include "qemu/cutils.h"
37 #include "qemu/config-file.h"
38 #include "qemu/option.h"
39 #include "qemu/error-report.h"
40 #include "qemu/log.h"
41 #include "qemu/main-loop.h"
42 #include "qemu/module.h"
43 #include "qemu/sockets.h"
44 #include "qemu/units.h"
45 #include "qom/object_interfaces.h"
46 #include "sysemu/block-backend.h"
47 #include "block/block_int.h"
48 #include "block/blockjob.h"
49 #include "block/qapi.h"
50 #include "crypto/init.h"
51 #include "trace/control.h"
52 #include "qemu/throttle.h"
53 #include "block/throttle-groups.h"
54 
55 #define QEMU_IMG_VERSION "qemu-img version " QEMU_FULL_VERSION \
56                           "\n" QEMU_COPYRIGHT "\n"
57 
58 typedef struct img_cmd_t {
59     const char *name;
60     int (*handler)(int argc, char **argv);
61 } img_cmd_t;
62 
63 enum {
64     OPTION_OUTPUT = 256,
65     OPTION_BACKING_CHAIN = 257,
66     OPTION_OBJECT = 258,
67     OPTION_IMAGE_OPTS = 259,
68     OPTION_PATTERN = 260,
69     OPTION_FLUSH_INTERVAL = 261,
70     OPTION_NO_DRAIN = 262,
71     OPTION_TARGET_IMAGE_OPTS = 263,
72     OPTION_SIZE = 264,
73     OPTION_PREALLOCATION = 265,
74     OPTION_SHRINK = 266,
75     OPTION_SALVAGE = 267,
76     OPTION_TARGET_IS_ZERO = 268,
77     OPTION_ADD = 269,
78     OPTION_REMOVE = 270,
79     OPTION_CLEAR = 271,
80     OPTION_ENABLE = 272,
81     OPTION_DISABLE = 273,
82     OPTION_MERGE = 274,
83     OPTION_BITMAPS = 275,
84     OPTION_FORCE = 276,
85     OPTION_SKIP_BROKEN = 277,
86 };
87 
88 typedef enum OutputFormat {
89     OFORMAT_JSON,
90     OFORMAT_HUMAN,
91 } OutputFormat;
92 
93 /* Default to cache=writeback as data integrity is not important for qemu-img */
94 #define BDRV_DEFAULT_CACHE "writeback"
95 
96 static void format_print(void *opaque, const char *name)
97 {
98     printf(" %s", name);
99 }
100 
101 static void QEMU_NORETURN GCC_FMT_ATTR(1, 2) error_exit(const char *fmt, ...)
102 {
103     va_list ap;
104 
105     va_start(ap, fmt);
106     error_vreport(fmt, ap);
107     va_end(ap);
108 
109     error_printf("Try 'qemu-img --help' for more information\n");
110     exit(EXIT_FAILURE);
111 }
112 
113 static void QEMU_NORETURN missing_argument(const char *option)
114 {
115     error_exit("missing argument for option '%s'", option);
116 }
117 
118 static void QEMU_NORETURN unrecognized_option(const char *option)
119 {
120     error_exit("unrecognized option '%s'", option);
121 }
122 
123 /* Please keep in synch with docs/tools/qemu-img.rst */
124 static void QEMU_NORETURN help(void)
125 {
126     const char *help_msg =
127            QEMU_IMG_VERSION
128            "usage: qemu-img [standard options] command [command options]\n"
129            "QEMU disk image utility\n"
130            "\n"
131            "    '-h', '--help'       display this help and exit\n"
132            "    '-V', '--version'    output version information and exit\n"
133            "    '-T', '--trace'      [[enable=]<pattern>][,events=<file>][,file=<file>]\n"
134            "                         specify tracing options\n"
135            "\n"
136            "Command syntax:\n"
137 #define DEF(option, callback, arg_string)        \
138            "  " arg_string "\n"
139 #include "qemu-img-cmds.h"
140 #undef DEF
141            "\n"
142            "Command parameters:\n"
143            "  'filename' is a disk image filename\n"
144            "  'objectdef' is a QEMU user creatable object definition. See the qemu(1)\n"
145            "    manual page for a description of the object properties. The most common\n"
146            "    object type is a 'secret', which is used to supply passwords and/or\n"
147            "    encryption keys.\n"
148            "  'fmt' is the disk image format. It is guessed automatically in most cases\n"
149            "  'cache' is the cache mode used to write the output disk image, the valid\n"
150            "    options are: 'none', 'writeback' (default, except for convert), 'writethrough',\n"
151            "    'directsync' and 'unsafe' (default for convert)\n"
152            "  'src_cache' is the cache mode used to read input disk images, the valid\n"
153            "    options are the same as for the 'cache' option\n"
154            "  'size' is the disk image size in bytes. Optional suffixes\n"
155            "    'k' or 'K' (kilobyte, 1024), 'M' (megabyte, 1024k), 'G' (gigabyte, 1024M),\n"
156            "    'T' (terabyte, 1024G), 'P' (petabyte, 1024T) and 'E' (exabyte, 1024P)  are\n"
157            "    supported. 'b' is ignored.\n"
158            "  'output_filename' is the destination disk image filename\n"
159            "  'output_fmt' is the destination format\n"
160            "  'options' is a comma separated list of format specific options in a\n"
161            "    name=value format. Use -o ? for an overview of the options supported by the\n"
162            "    used format\n"
163            "  'snapshot_param' is param used for internal snapshot, format\n"
164            "    is 'snapshot.id=[ID],snapshot.name=[NAME]', or\n"
165            "    '[ID_OR_NAME]'\n"
166            "  '-c' indicates that target image must be compressed (qcow format only)\n"
167            "  '-u' allows unsafe backing chains. For rebasing, it is assumed that old and\n"
168            "       new backing file match exactly. The image doesn't need a working\n"
169            "       backing file before rebasing in this case (useful for renaming the\n"
170            "       backing file). For image creation, allow creating without attempting\n"
171            "       to open the backing file.\n"
172            "  '-h' with or without a command shows this help and lists the supported formats\n"
173            "  '-p' show progress of command (only certain commands)\n"
174            "  '-q' use Quiet mode - do not print any output (except errors)\n"
175            "  '-S' indicates the consecutive number of bytes (defaults to 4k) that must\n"
176            "       contain only zeros for qemu-img to create a sparse image during\n"
177            "       conversion. If the number of bytes is 0, the source will not be scanned for\n"
178            "       unallocated or zero sectors, and the destination image will always be\n"
179            "       fully allocated\n"
180            "  '--output' takes the format in which the output must be done (human or json)\n"
181            "  '-n' skips the target volume creation (useful if the volume is created\n"
182            "       prior to running qemu-img)\n"
183            "\n"
184            "Parameters to bitmap subcommand:\n"
185            "  'bitmap' is the name of the bitmap to manipulate, through one or more\n"
186            "       actions from '--add', '--remove', '--clear', '--enable', '--disable',\n"
187            "       or '--merge source'\n"
188            "  '-g granularity' sets the granularity for '--add' actions\n"
189            "  '-b source' and '-F src_fmt' tell '--merge' actions to find the source\n"
190            "       bitmaps from an alternative file\n"
191            "\n"
192            "Parameters to check subcommand:\n"
193            "  '-r' tries to repair any inconsistencies that are found during the check.\n"
194            "       '-r leaks' repairs only cluster leaks, whereas '-r all' fixes all\n"
195            "       kinds of errors, with a higher risk of choosing the wrong fix or\n"
196            "       hiding corruption that has already occurred.\n"
197            "\n"
198            "Parameters to convert subcommand:\n"
199            "  '--bitmaps' copies all top-level persistent bitmaps to destination\n"
200            "  '-m' specifies how many coroutines work in parallel during the convert\n"
201            "       process (defaults to 8)\n"
202            "  '-W' allow to write to the target out of order rather than sequential\n"
203            "\n"
204            "Parameters to snapshot subcommand:\n"
205            "  'snapshot' is the name of the snapshot to create, apply or delete\n"
206            "  '-a' applies a snapshot (revert disk to saved state)\n"
207            "  '-c' creates a snapshot\n"
208            "  '-d' deletes a snapshot\n"
209            "  '-l' lists all snapshots in the given image\n"
210            "\n"
211            "Parameters to compare subcommand:\n"
212            "  '-f' first image format\n"
213            "  '-F' second image format\n"
214            "  '-s' run in Strict mode - fail on different image size or sector allocation\n"
215            "\n"
216            "Parameters to dd subcommand:\n"
217            "  'bs=BYTES' read and write up to BYTES bytes at a time "
218            "(default: 512)\n"
219            "  'count=N' copy only N input blocks\n"
220            "  'if=FILE' read from FILE\n"
221            "  'of=FILE' write to FILE\n"
222            "  'skip=N' skip N bs-sized blocks at the start of input\n";
223 
224     printf("%s\nSupported formats:", help_msg);
225     bdrv_iterate_format(format_print, NULL, false);
226     printf("\n\n" QEMU_HELP_BOTTOM "\n");
227     exit(EXIT_SUCCESS);
228 }
229 
230 /*
231  * Is @optarg safe for accumulate_options()?
232  * It is when multiple of them can be joined together separated by ','.
233  * To make that work, @optarg must not start with ',' (or else a
234  * separating ',' preceding it gets escaped), and it must not end with
235  * an odd number of ',' (or else a separating ',' following it gets
236  * escaped), or be empty (or else a separating ',' preceding it can
237  * escape a separating ',' following it).
238  *
239  */
240 static bool is_valid_option_list(const char *optarg)
241 {
242     size_t len = strlen(optarg);
243     size_t i;
244 
245     if (!optarg[0] || optarg[0] == ',') {
246         return false;
247     }
248 
249     for (i = len; i > 0 && optarg[i - 1] == ','; i--) {
250     }
251     if ((len - i) % 2) {
252         return false;
253     }
254 
255     return true;
256 }
257 
258 static int accumulate_options(char **options, char *optarg)
259 {
260     char *new_options;
261 
262     if (!is_valid_option_list(optarg)) {
263         error_report("Invalid option list: %s", optarg);
264         return -1;
265     }
266 
267     if (!*options) {
268         *options = g_strdup(optarg);
269     } else {
270         new_options = g_strdup_printf("%s,%s", *options, optarg);
271         g_free(*options);
272         *options = new_options;
273     }
274     return 0;
275 }
276 
277 static QemuOptsList qemu_source_opts = {
278     .name = "source",
279     .implied_opt_name = "file",
280     .head = QTAILQ_HEAD_INITIALIZER(qemu_source_opts.head),
281     .desc = {
282         { }
283     },
284 };
285 
286 static int GCC_FMT_ATTR(2, 3) qprintf(bool quiet, const char *fmt, ...)
287 {
288     int ret = 0;
289     if (!quiet) {
290         va_list args;
291         va_start(args, fmt);
292         ret = vprintf(fmt, args);
293         va_end(args);
294     }
295     return ret;
296 }
297 
298 
299 static int print_block_option_help(const char *filename, const char *fmt)
300 {
301     BlockDriver *drv, *proto_drv;
302     QemuOptsList *create_opts = NULL;
303     Error *local_err = NULL;
304 
305     /* Find driver and parse its options */
306     drv = bdrv_find_format(fmt);
307     if (!drv) {
308         error_report("Unknown file format '%s'", fmt);
309         return 1;
310     }
311 
312     if (!drv->create_opts) {
313         error_report("Format driver '%s' does not support image creation", fmt);
314         return 1;
315     }
316 
317     create_opts = qemu_opts_append(create_opts, drv->create_opts);
318     if (filename) {
319         proto_drv = bdrv_find_protocol(filename, true, &local_err);
320         if (!proto_drv) {
321             error_report_err(local_err);
322             qemu_opts_free(create_opts);
323             return 1;
324         }
325         if (!proto_drv->create_opts) {
326             error_report("Protocol driver '%s' does not support image creation",
327                          proto_drv->format_name);
328             qemu_opts_free(create_opts);
329             return 1;
330         }
331         create_opts = qemu_opts_append(create_opts, proto_drv->create_opts);
332     }
333 
334     if (filename) {
335         printf("Supported options:\n");
336     } else {
337         printf("Supported %s options:\n", fmt);
338     }
339     qemu_opts_print_help(create_opts, false);
340     qemu_opts_free(create_opts);
341 
342     if (!filename) {
343         printf("\n"
344                "The protocol level may support further options.\n"
345                "Specify the target filename to include those options.\n");
346     }
347 
348     return 0;
349 }
350 
351 
352 static BlockBackend *img_open_opts(const char *optstr,
353                                    QemuOpts *opts, int flags, bool writethrough,
354                                    bool quiet, bool force_share)
355 {
356     QDict *options;
357     Error *local_err = NULL;
358     BlockBackend *blk;
359     options = qemu_opts_to_qdict(opts, NULL);
360     if (force_share) {
361         if (qdict_haskey(options, BDRV_OPT_FORCE_SHARE)
362             && strcmp(qdict_get_str(options, BDRV_OPT_FORCE_SHARE), "on")) {
363             error_report("--force-share/-U conflicts with image options");
364             qobject_unref(options);
365             return NULL;
366         }
367         qdict_put_str(options, BDRV_OPT_FORCE_SHARE, "on");
368     }
369     blk = blk_new_open(NULL, NULL, options, flags, &local_err);
370     if (!blk) {
371         error_reportf_err(local_err, "Could not open '%s': ", optstr);
372         return NULL;
373     }
374     blk_set_enable_write_cache(blk, !writethrough);
375 
376     return blk;
377 }
378 
379 static BlockBackend *img_open_file(const char *filename,
380                                    QDict *options,
381                                    const char *fmt, int flags,
382                                    bool writethrough, bool quiet,
383                                    bool force_share)
384 {
385     BlockBackend *blk;
386     Error *local_err = NULL;
387 
388     if (!options) {
389         options = qdict_new();
390     }
391     if (fmt) {
392         qdict_put_str(options, "driver", fmt);
393     }
394 
395     if (force_share) {
396         qdict_put_bool(options, BDRV_OPT_FORCE_SHARE, true);
397     }
398     blk = blk_new_open(filename, NULL, options, flags, &local_err);
399     if (!blk) {
400         error_reportf_err(local_err, "Could not open '%s': ", filename);
401         return NULL;
402     }
403     blk_set_enable_write_cache(blk, !writethrough);
404 
405     return blk;
406 }
407 
408 
409 static int img_add_key_secrets(void *opaque,
410                                const char *name, const char *value,
411                                Error **errp)
412 {
413     QDict *options = opaque;
414 
415     if (g_str_has_suffix(name, "key-secret")) {
416         qdict_put_str(options, name, value);
417     }
418 
419     return 0;
420 }
421 
422 
423 static BlockBackend *img_open(bool image_opts,
424                               const char *filename,
425                               const char *fmt, int flags, bool writethrough,
426                               bool quiet, bool force_share)
427 {
428     BlockBackend *blk;
429     if (image_opts) {
430         QemuOpts *opts;
431         if (fmt) {
432             error_report("--image-opts and --format are mutually exclusive");
433             return NULL;
434         }
435         opts = qemu_opts_parse_noisily(qemu_find_opts("source"),
436                                        filename, true);
437         if (!opts) {
438             return NULL;
439         }
440         blk = img_open_opts(filename, opts, flags, writethrough, quiet,
441                             force_share);
442     } else {
443         blk = img_open_file(filename, NULL, fmt, flags, writethrough, quiet,
444                             force_share);
445     }
446     return blk;
447 }
448 
449 
450 static int add_old_style_options(const char *fmt, QemuOpts *opts,
451                                  const char *base_filename,
452                                  const char *base_fmt)
453 {
454     if (base_filename) {
455         if (!qemu_opt_set(opts, BLOCK_OPT_BACKING_FILE, base_filename,
456                           NULL)) {
457             error_report("Backing file not supported for file format '%s'",
458                          fmt);
459             return -1;
460         }
461     }
462     if (base_fmt) {
463         if (!qemu_opt_set(opts, BLOCK_OPT_BACKING_FMT, base_fmt, NULL)) {
464             error_report("Backing file format not supported for file "
465                          "format '%s'", fmt);
466             return -1;
467         }
468     }
469     return 0;
470 }
471 
472 static int64_t cvtnum_full(const char *name, const char *value, int64_t min,
473                            int64_t max)
474 {
475     int err;
476     uint64_t res;
477 
478     err = qemu_strtosz(value, NULL, &res);
479     if (err < 0 && err != -ERANGE) {
480         error_report("Invalid %s specified. You may use "
481                      "k, M, G, T, P or E suffixes for", name);
482         error_report("kilobytes, megabytes, gigabytes, terabytes, "
483                      "petabytes and exabytes.");
484         return err;
485     }
486     if (err == -ERANGE || res > max || res < min) {
487         error_report("Invalid %s specified. Must be between %" PRId64
488                      " and %" PRId64 ".", name, min, max);
489         return -ERANGE;
490     }
491     return res;
492 }
493 
494 static int64_t cvtnum(const char *name, const char *value)
495 {
496     return cvtnum_full(name, value, 0, INT64_MAX);
497 }
498 
499 static int img_create(int argc, char **argv)
500 {
501     int c;
502     uint64_t img_size = -1;
503     const char *fmt = "raw";
504     const char *base_fmt = NULL;
505     const char *filename;
506     const char *base_filename = NULL;
507     char *options = NULL;
508     Error *local_err = NULL;
509     bool quiet = false;
510     int flags = 0;
511 
512     for(;;) {
513         static const struct option long_options[] = {
514             {"help", no_argument, 0, 'h'},
515             {"object", required_argument, 0, OPTION_OBJECT},
516             {0, 0, 0, 0}
517         };
518         c = getopt_long(argc, argv, ":F:b:f:ho:qu",
519                         long_options, NULL);
520         if (c == -1) {
521             break;
522         }
523         switch(c) {
524         case ':':
525             missing_argument(argv[optind - 1]);
526             break;
527         case '?':
528             unrecognized_option(argv[optind - 1]);
529             break;
530         case 'h':
531             help();
532             break;
533         case 'F':
534             base_fmt = optarg;
535             break;
536         case 'b':
537             base_filename = optarg;
538             break;
539         case 'f':
540             fmt = optarg;
541             break;
542         case 'o':
543             if (accumulate_options(&options, optarg) < 0) {
544                 goto fail;
545             }
546             break;
547         case 'q':
548             quiet = true;
549             break;
550         case 'u':
551             flags |= BDRV_O_NO_BACKING;
552             break;
553         case OPTION_OBJECT:
554             user_creatable_process_cmdline(optarg);
555             break;
556         }
557     }
558 
559     /* Get the filename */
560     filename = (optind < argc) ? argv[optind] : NULL;
561     if (options && has_help_option(options)) {
562         g_free(options);
563         return print_block_option_help(filename, fmt);
564     }
565 
566     if (optind >= argc) {
567         error_exit("Expecting image file name");
568     }
569     optind++;
570 
571     /* Get image size, if specified */
572     if (optind < argc) {
573         int64_t sval;
574 
575         sval = cvtnum("image size", argv[optind++]);
576         if (sval < 0) {
577             goto fail;
578         }
579         img_size = (uint64_t)sval;
580     }
581     if (optind != argc) {
582         error_exit("Unexpected argument: %s", argv[optind]);
583     }
584 
585     bdrv_img_create(filename, fmt, base_filename, base_fmt,
586                     options, img_size, flags, quiet, &local_err);
587     if (local_err) {
588         error_reportf_err(local_err, "%s: ", filename);
589         goto fail;
590     }
591 
592     g_free(options);
593     return 0;
594 
595 fail:
596     g_free(options);
597     return 1;
598 }
599 
600 static void dump_json_image_check(ImageCheck *check, bool quiet)
601 {
602     GString *str;
603     QObject *obj;
604     Visitor *v = qobject_output_visitor_new(&obj);
605 
606     visit_type_ImageCheck(v, NULL, &check, &error_abort);
607     visit_complete(v, &obj);
608     str = qobject_to_json_pretty(obj, true);
609     assert(str != NULL);
610     qprintf(quiet, "%s\n", str->str);
611     qobject_unref(obj);
612     visit_free(v);
613     g_string_free(str, true);
614 }
615 
616 static void dump_human_image_check(ImageCheck *check, bool quiet)
617 {
618     if (!(check->corruptions || check->leaks || check->check_errors)) {
619         qprintf(quiet, "No errors were found on the image.\n");
620     } else {
621         if (check->corruptions) {
622             qprintf(quiet, "\n%" PRId64 " errors were found on the image.\n"
623                     "Data may be corrupted, or further writes to the image "
624                     "may corrupt it.\n",
625                     check->corruptions);
626         }
627 
628         if (check->leaks) {
629             qprintf(quiet,
630                     "\n%" PRId64 " leaked clusters were found on the image.\n"
631                     "This means waste of disk space, but no harm to data.\n",
632                     check->leaks);
633         }
634 
635         if (check->check_errors) {
636             qprintf(quiet,
637                     "\n%" PRId64
638                     " internal errors have occurred during the check.\n",
639                     check->check_errors);
640         }
641     }
642 
643     if (check->total_clusters != 0 && check->allocated_clusters != 0) {
644         qprintf(quiet, "%" PRId64 "/%" PRId64 " = %0.2f%% allocated, "
645                 "%0.2f%% fragmented, %0.2f%% compressed clusters\n",
646                 check->allocated_clusters, check->total_clusters,
647                 check->allocated_clusters * 100.0 / check->total_clusters,
648                 check->fragmented_clusters * 100.0 / check->allocated_clusters,
649                 check->compressed_clusters * 100.0 /
650                 check->allocated_clusters);
651     }
652 
653     if (check->image_end_offset) {
654         qprintf(quiet,
655                 "Image end offset: %" PRId64 "\n", check->image_end_offset);
656     }
657 }
658 
659 static int collect_image_check(BlockDriverState *bs,
660                    ImageCheck *check,
661                    const char *filename,
662                    const char *fmt,
663                    int fix)
664 {
665     int ret;
666     BdrvCheckResult result;
667 
668     ret = bdrv_check(bs, &result, fix);
669     if (ret < 0) {
670         return ret;
671     }
672 
673     check->filename                 = g_strdup(filename);
674     check->format                   = g_strdup(bdrv_get_format_name(bs));
675     check->check_errors             = result.check_errors;
676     check->corruptions              = result.corruptions;
677     check->has_corruptions          = result.corruptions != 0;
678     check->leaks                    = result.leaks;
679     check->has_leaks                = result.leaks != 0;
680     check->corruptions_fixed        = result.corruptions_fixed;
681     check->has_corruptions_fixed    = result.corruptions_fixed != 0;
682     check->leaks_fixed              = result.leaks_fixed;
683     check->has_leaks_fixed          = result.leaks_fixed != 0;
684     check->image_end_offset         = result.image_end_offset;
685     check->has_image_end_offset     = result.image_end_offset != 0;
686     check->total_clusters           = result.bfi.total_clusters;
687     check->has_total_clusters       = result.bfi.total_clusters != 0;
688     check->allocated_clusters       = result.bfi.allocated_clusters;
689     check->has_allocated_clusters   = result.bfi.allocated_clusters != 0;
690     check->fragmented_clusters      = result.bfi.fragmented_clusters;
691     check->has_fragmented_clusters  = result.bfi.fragmented_clusters != 0;
692     check->compressed_clusters      = result.bfi.compressed_clusters;
693     check->has_compressed_clusters  = result.bfi.compressed_clusters != 0;
694 
695     return 0;
696 }
697 
698 /*
699  * Checks an image for consistency. Exit codes:
700  *
701  *  0 - Check completed, image is good
702  *  1 - Check not completed because of internal errors
703  *  2 - Check completed, image is corrupted
704  *  3 - Check completed, image has leaked clusters, but is good otherwise
705  * 63 - Checks are not supported by the image format
706  */
707 static int img_check(int argc, char **argv)
708 {
709     int c, ret;
710     OutputFormat output_format = OFORMAT_HUMAN;
711     const char *filename, *fmt, *output, *cache;
712     BlockBackend *blk;
713     BlockDriverState *bs;
714     int fix = 0;
715     int flags = BDRV_O_CHECK;
716     bool writethrough;
717     ImageCheck *check;
718     bool quiet = false;
719     bool image_opts = false;
720     bool force_share = false;
721 
722     fmt = NULL;
723     output = NULL;
724     cache = BDRV_DEFAULT_CACHE;
725 
726     for(;;) {
727         int option_index = 0;
728         static const struct option long_options[] = {
729             {"help", no_argument, 0, 'h'},
730             {"format", required_argument, 0, 'f'},
731             {"repair", required_argument, 0, 'r'},
732             {"output", required_argument, 0, OPTION_OUTPUT},
733             {"object", required_argument, 0, OPTION_OBJECT},
734             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
735             {"force-share", no_argument, 0, 'U'},
736             {0, 0, 0, 0}
737         };
738         c = getopt_long(argc, argv, ":hf:r:T:qU",
739                         long_options, &option_index);
740         if (c == -1) {
741             break;
742         }
743         switch(c) {
744         case ':':
745             missing_argument(argv[optind - 1]);
746             break;
747         case '?':
748             unrecognized_option(argv[optind - 1]);
749             break;
750         case 'h':
751             help();
752             break;
753         case 'f':
754             fmt = optarg;
755             break;
756         case 'r':
757             flags |= BDRV_O_RDWR;
758 
759             if (!strcmp(optarg, "leaks")) {
760                 fix = BDRV_FIX_LEAKS;
761             } else if (!strcmp(optarg, "all")) {
762                 fix = BDRV_FIX_LEAKS | BDRV_FIX_ERRORS;
763             } else {
764                 error_exit("Unknown option value for -r "
765                            "(expecting 'leaks' or 'all'): %s", optarg);
766             }
767             break;
768         case OPTION_OUTPUT:
769             output = optarg;
770             break;
771         case 'T':
772             cache = optarg;
773             break;
774         case 'q':
775             quiet = true;
776             break;
777         case 'U':
778             force_share = true;
779             break;
780         case OPTION_OBJECT:
781             user_creatable_process_cmdline(optarg);
782             break;
783         case OPTION_IMAGE_OPTS:
784             image_opts = true;
785             break;
786         }
787     }
788     if (optind != argc - 1) {
789         error_exit("Expecting one image file name");
790     }
791     filename = argv[optind++];
792 
793     if (output && !strcmp(output, "json")) {
794         output_format = OFORMAT_JSON;
795     } else if (output && !strcmp(output, "human")) {
796         output_format = OFORMAT_HUMAN;
797     } else if (output) {
798         error_report("--output must be used with human or json as argument.");
799         return 1;
800     }
801 
802     ret = bdrv_parse_cache_mode(cache, &flags, &writethrough);
803     if (ret < 0) {
804         error_report("Invalid source cache option: %s", cache);
805         return 1;
806     }
807 
808     blk = img_open(image_opts, filename, fmt, flags, writethrough, quiet,
809                    force_share);
810     if (!blk) {
811         return 1;
812     }
813     bs = blk_bs(blk);
814 
815     check = g_new0(ImageCheck, 1);
816     ret = collect_image_check(bs, check, filename, fmt, fix);
817 
818     if (ret == -ENOTSUP) {
819         error_report("This image format does not support checks");
820         ret = 63;
821         goto fail;
822     }
823 
824     if (check->corruptions_fixed || check->leaks_fixed) {
825         int corruptions_fixed, leaks_fixed;
826         bool has_leaks_fixed, has_corruptions_fixed;
827 
828         leaks_fixed         = check->leaks_fixed;
829         has_leaks_fixed     = check->has_leaks_fixed;
830         corruptions_fixed   = check->corruptions_fixed;
831         has_corruptions_fixed = check->has_corruptions_fixed;
832 
833         if (output_format == OFORMAT_HUMAN) {
834             qprintf(quiet,
835                     "The following inconsistencies were found and repaired:\n\n"
836                     "    %" PRId64 " leaked clusters\n"
837                     "    %" PRId64 " corruptions\n\n"
838                     "Double checking the fixed image now...\n",
839                     check->leaks_fixed,
840                     check->corruptions_fixed);
841         }
842 
843         qapi_free_ImageCheck(check);
844         check = g_new0(ImageCheck, 1);
845         ret = collect_image_check(bs, check, filename, fmt, 0);
846 
847         check->leaks_fixed          = leaks_fixed;
848         check->has_leaks_fixed      = has_leaks_fixed;
849         check->corruptions_fixed    = corruptions_fixed;
850         check->has_corruptions_fixed = has_corruptions_fixed;
851     }
852 
853     if (!ret) {
854         switch (output_format) {
855         case OFORMAT_HUMAN:
856             dump_human_image_check(check, quiet);
857             break;
858         case OFORMAT_JSON:
859             dump_json_image_check(check, quiet);
860             break;
861         }
862     }
863 
864     if (ret || check->check_errors) {
865         if (ret) {
866             error_report("Check failed: %s", strerror(-ret));
867         } else {
868             error_report("Check failed");
869         }
870         ret = 1;
871         goto fail;
872     }
873 
874     if (check->corruptions) {
875         ret = 2;
876     } else if (check->leaks) {
877         ret = 3;
878     } else {
879         ret = 0;
880     }
881 
882 fail:
883     qapi_free_ImageCheck(check);
884     blk_unref(blk);
885     return ret;
886 }
887 
888 typedef struct CommonBlockJobCBInfo {
889     BlockDriverState *bs;
890     Error **errp;
891 } CommonBlockJobCBInfo;
892 
893 static void common_block_job_cb(void *opaque, int ret)
894 {
895     CommonBlockJobCBInfo *cbi = opaque;
896 
897     if (ret < 0) {
898         error_setg_errno(cbi->errp, -ret, "Block job failed");
899     }
900 }
901 
902 static void run_block_job(BlockJob *job, Error **errp)
903 {
904     uint64_t progress_current, progress_total;
905     AioContext *aio_context = blk_get_aio_context(job->blk);
906     int ret = 0;
907 
908     aio_context_acquire(aio_context);
909     job_ref(&job->job);
910     do {
911         float progress = 0.0f;
912         aio_poll(aio_context, true);
913 
914         progress_get_snapshot(&job->job.progress, &progress_current,
915                               &progress_total);
916         if (progress_total) {
917             progress = (float)progress_current / progress_total * 100.f;
918         }
919         qemu_progress_print(progress, 0);
920     } while (!job_is_ready(&job->job) && !job_is_completed(&job->job));
921 
922     if (!job_is_completed(&job->job)) {
923         ret = job_complete_sync(&job->job, errp);
924     } else {
925         ret = job->job.ret;
926     }
927     job_unref(&job->job);
928     aio_context_release(aio_context);
929 
930     /* publish completion progress only when success */
931     if (!ret) {
932         qemu_progress_print(100.f, 0);
933     }
934 }
935 
936 static int img_commit(int argc, char **argv)
937 {
938     int c, ret, flags;
939     const char *filename, *fmt, *cache, *base;
940     BlockBackend *blk;
941     BlockDriverState *bs, *base_bs;
942     BlockJob *job;
943     bool progress = false, quiet = false, drop = false;
944     bool writethrough;
945     Error *local_err = NULL;
946     CommonBlockJobCBInfo cbi;
947     bool image_opts = false;
948     AioContext *aio_context;
949     int64_t rate_limit = 0;
950 
951     fmt = NULL;
952     cache = BDRV_DEFAULT_CACHE;
953     base = NULL;
954     for(;;) {
955         static const struct option long_options[] = {
956             {"help", no_argument, 0, 'h'},
957             {"object", required_argument, 0, OPTION_OBJECT},
958             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
959             {0, 0, 0, 0}
960         };
961         c = getopt_long(argc, argv, ":f:ht:b:dpqr:",
962                         long_options, NULL);
963         if (c == -1) {
964             break;
965         }
966         switch(c) {
967         case ':':
968             missing_argument(argv[optind - 1]);
969             break;
970         case '?':
971             unrecognized_option(argv[optind - 1]);
972             break;
973         case 'h':
974             help();
975             break;
976         case 'f':
977             fmt = optarg;
978             break;
979         case 't':
980             cache = optarg;
981             break;
982         case 'b':
983             base = optarg;
984             /* -b implies -d */
985             drop = true;
986             break;
987         case 'd':
988             drop = true;
989             break;
990         case 'p':
991             progress = true;
992             break;
993         case 'q':
994             quiet = true;
995             break;
996         case 'r':
997             rate_limit = cvtnum("rate limit", optarg);
998             if (rate_limit < 0) {
999                 return 1;
1000             }
1001             break;
1002         case OPTION_OBJECT:
1003             user_creatable_process_cmdline(optarg);
1004             break;
1005         case OPTION_IMAGE_OPTS:
1006             image_opts = true;
1007             break;
1008         }
1009     }
1010 
1011     /* Progress is not shown in Quiet mode */
1012     if (quiet) {
1013         progress = false;
1014     }
1015 
1016     if (optind != argc - 1) {
1017         error_exit("Expecting one image file name");
1018     }
1019     filename = argv[optind++];
1020 
1021     flags = BDRV_O_RDWR | BDRV_O_UNMAP;
1022     ret = bdrv_parse_cache_mode(cache, &flags, &writethrough);
1023     if (ret < 0) {
1024         error_report("Invalid cache option: %s", cache);
1025         return 1;
1026     }
1027 
1028     blk = img_open(image_opts, filename, fmt, flags, writethrough, quiet,
1029                    false);
1030     if (!blk) {
1031         return 1;
1032     }
1033     bs = blk_bs(blk);
1034 
1035     qemu_progress_init(progress, 1.f);
1036     qemu_progress_print(0.f, 100);
1037 
1038     if (base) {
1039         base_bs = bdrv_find_backing_image(bs, base);
1040         if (!base_bs) {
1041             error_setg(&local_err,
1042                        "Did not find '%s' in the backing chain of '%s'",
1043                        base, filename);
1044             goto done;
1045         }
1046     } else {
1047         /* This is different from QMP, which by default uses the deepest file in
1048          * the backing chain (i.e., the very base); however, the traditional
1049          * behavior of qemu-img commit is using the immediate backing file. */
1050         base_bs = bdrv_backing_chain_next(bs);
1051         if (!base_bs) {
1052             error_setg(&local_err, "Image does not have a backing file");
1053             goto done;
1054         }
1055     }
1056 
1057     cbi = (CommonBlockJobCBInfo){
1058         .errp = &local_err,
1059         .bs   = bs,
1060     };
1061 
1062     aio_context = bdrv_get_aio_context(bs);
1063     aio_context_acquire(aio_context);
1064     commit_active_start("commit", bs, base_bs, JOB_DEFAULT, rate_limit,
1065                         BLOCKDEV_ON_ERROR_REPORT, NULL, common_block_job_cb,
1066                         &cbi, false, &local_err);
1067     aio_context_release(aio_context);
1068     if (local_err) {
1069         goto done;
1070     }
1071 
1072     /* When the block job completes, the BlockBackend reference will point to
1073      * the old backing file. In order to avoid that the top image is already
1074      * deleted, so we can still empty it afterwards, increment the reference
1075      * counter here preemptively. */
1076     if (!drop) {
1077         bdrv_ref(bs);
1078     }
1079 
1080     job = block_job_get("commit");
1081     assert(job);
1082     run_block_job(job, &local_err);
1083     if (local_err) {
1084         goto unref_backing;
1085     }
1086 
1087     if (!drop) {
1088         BlockBackend *old_backing_blk;
1089 
1090         old_backing_blk = blk_new_with_bs(bs, BLK_PERM_WRITE, BLK_PERM_ALL,
1091                                           &local_err);
1092         if (!old_backing_blk) {
1093             goto unref_backing;
1094         }
1095         ret = blk_make_empty(old_backing_blk, &local_err);
1096         blk_unref(old_backing_blk);
1097         if (ret == -ENOTSUP) {
1098             error_free(local_err);
1099             local_err = NULL;
1100         } else if (ret < 0) {
1101             goto unref_backing;
1102         }
1103     }
1104 
1105 unref_backing:
1106     if (!drop) {
1107         bdrv_unref(bs);
1108     }
1109 
1110 done:
1111     qemu_progress_end();
1112 
1113     blk_unref(blk);
1114 
1115     if (local_err) {
1116         error_report_err(local_err);
1117         return 1;
1118     }
1119 
1120     qprintf(quiet, "Image committed.\n");
1121     return 0;
1122 }
1123 
1124 /*
1125  * Returns -1 if 'buf' contains only zeroes, otherwise the byte index
1126  * of the first sector boundary within buf where the sector contains a
1127  * non-zero byte.  This function is robust to a buffer that is not
1128  * sector-aligned.
1129  */
1130 static int64_t find_nonzero(const uint8_t *buf, int64_t n)
1131 {
1132     int64_t i;
1133     int64_t end = QEMU_ALIGN_DOWN(n, BDRV_SECTOR_SIZE);
1134 
1135     for (i = 0; i < end; i += BDRV_SECTOR_SIZE) {
1136         if (!buffer_is_zero(buf + i, BDRV_SECTOR_SIZE)) {
1137             return i;
1138         }
1139     }
1140     if (i < n && !buffer_is_zero(buf + i, n - end)) {
1141         return i;
1142     }
1143     return -1;
1144 }
1145 
1146 /*
1147  * Returns true iff the first sector pointed to by 'buf' contains at least
1148  * a non-NUL byte.
1149  *
1150  * 'pnum' is set to the number of sectors (including and immediately following
1151  * the first one) that are known to be in the same allocated/unallocated state.
1152  * The function will try to align the end offset to alignment boundaries so
1153  * that the request will at least end aligned and consecutive requests will
1154  * also start at an aligned offset.
1155  */
1156 static int is_allocated_sectors(const uint8_t *buf, int n, int *pnum,
1157                                 int64_t sector_num, int alignment)
1158 {
1159     bool is_zero;
1160     int i, tail;
1161 
1162     if (n <= 0) {
1163         *pnum = 0;
1164         return 0;
1165     }
1166     is_zero = buffer_is_zero(buf, BDRV_SECTOR_SIZE);
1167     for(i = 1; i < n; i++) {
1168         buf += BDRV_SECTOR_SIZE;
1169         if (is_zero != buffer_is_zero(buf, BDRV_SECTOR_SIZE)) {
1170             break;
1171         }
1172     }
1173 
1174     tail = (sector_num + i) & (alignment - 1);
1175     if (tail) {
1176         if (is_zero && i <= tail) {
1177             /* treat unallocated areas which only consist
1178              * of a small tail as allocated. */
1179             is_zero = false;
1180         }
1181         if (!is_zero) {
1182             /* align up end offset of allocated areas. */
1183             i += alignment - tail;
1184             i = MIN(i, n);
1185         } else {
1186             /* align down end offset of zero areas. */
1187             i -= tail;
1188         }
1189     }
1190     *pnum = i;
1191     return !is_zero;
1192 }
1193 
1194 /*
1195  * Like is_allocated_sectors, but if the buffer starts with a used sector,
1196  * up to 'min' consecutive sectors containing zeros are ignored. This avoids
1197  * breaking up write requests for only small sparse areas.
1198  */
1199 static int is_allocated_sectors_min(const uint8_t *buf, int n, int *pnum,
1200     int min, int64_t sector_num, int alignment)
1201 {
1202     int ret;
1203     int num_checked, num_used;
1204 
1205     if (n < min) {
1206         min = n;
1207     }
1208 
1209     ret = is_allocated_sectors(buf, n, pnum, sector_num, alignment);
1210     if (!ret) {
1211         return ret;
1212     }
1213 
1214     num_used = *pnum;
1215     buf += BDRV_SECTOR_SIZE * *pnum;
1216     n -= *pnum;
1217     sector_num += *pnum;
1218     num_checked = num_used;
1219 
1220     while (n > 0) {
1221         ret = is_allocated_sectors(buf, n, pnum, sector_num, alignment);
1222 
1223         buf += BDRV_SECTOR_SIZE * *pnum;
1224         n -= *pnum;
1225         sector_num += *pnum;
1226         num_checked += *pnum;
1227         if (ret) {
1228             num_used = num_checked;
1229         } else if (*pnum >= min) {
1230             break;
1231         }
1232     }
1233 
1234     *pnum = num_used;
1235     return 1;
1236 }
1237 
1238 /*
1239  * Compares two buffers sector by sector. Returns 0 if the first
1240  * sector of each buffer matches, non-zero otherwise.
1241  *
1242  * pnum is set to the sector-aligned size of the buffer prefix that
1243  * has the same matching status as the first sector.
1244  */
1245 static int compare_buffers(const uint8_t *buf1, const uint8_t *buf2,
1246                            int64_t bytes, int64_t *pnum)
1247 {
1248     bool res;
1249     int64_t i = MIN(bytes, BDRV_SECTOR_SIZE);
1250 
1251     assert(bytes > 0);
1252 
1253     res = !!memcmp(buf1, buf2, i);
1254     while (i < bytes) {
1255         int64_t len = MIN(bytes - i, BDRV_SECTOR_SIZE);
1256 
1257         if (!!memcmp(buf1 + i, buf2 + i, len) != res) {
1258             break;
1259         }
1260         i += len;
1261     }
1262 
1263     *pnum = i;
1264     return res;
1265 }
1266 
1267 #define IO_BUF_SIZE (2 * MiB)
1268 
1269 /*
1270  * Check if passed sectors are empty (not allocated or contain only 0 bytes)
1271  *
1272  * Intended for use by 'qemu-img compare': Returns 0 in case sectors are
1273  * filled with 0, 1 if sectors contain non-zero data (this is a comparison
1274  * failure), and 4 on error (the exit status for read errors), after emitting
1275  * an error message.
1276  *
1277  * @param blk:  BlockBackend for the image
1278  * @param offset: Starting offset to check
1279  * @param bytes: Number of bytes to check
1280  * @param filename: Name of disk file we are checking (logging purpose)
1281  * @param buffer: Allocated buffer for storing read data
1282  * @param quiet: Flag for quiet mode
1283  */
1284 static int check_empty_sectors(BlockBackend *blk, int64_t offset,
1285                                int64_t bytes, const char *filename,
1286                                uint8_t *buffer, bool quiet)
1287 {
1288     int ret = 0;
1289     int64_t idx;
1290 
1291     ret = blk_pread(blk, offset, buffer, bytes);
1292     if (ret < 0) {
1293         error_report("Error while reading offset %" PRId64 " of %s: %s",
1294                      offset, filename, strerror(-ret));
1295         return 4;
1296     }
1297     idx = find_nonzero(buffer, bytes);
1298     if (idx >= 0) {
1299         qprintf(quiet, "Content mismatch at offset %" PRId64 "!\n",
1300                 offset + idx);
1301         return 1;
1302     }
1303 
1304     return 0;
1305 }
1306 
1307 /*
1308  * Compares two images. Exit codes:
1309  *
1310  * 0 - Images are identical or the requested help was printed
1311  * 1 - Images differ
1312  * >1 - Error occurred
1313  */
1314 static int img_compare(int argc, char **argv)
1315 {
1316     const char *fmt1 = NULL, *fmt2 = NULL, *cache, *filename1, *filename2;
1317     BlockBackend *blk1, *blk2;
1318     BlockDriverState *bs1, *bs2;
1319     int64_t total_size1, total_size2;
1320     uint8_t *buf1 = NULL, *buf2 = NULL;
1321     int64_t pnum1, pnum2;
1322     int allocated1, allocated2;
1323     int ret = 0; /* return value - 0 Ident, 1 Different, >1 Error */
1324     bool progress = false, quiet = false, strict = false;
1325     int flags;
1326     bool writethrough;
1327     int64_t total_size;
1328     int64_t offset = 0;
1329     int64_t chunk;
1330     int c;
1331     uint64_t progress_base;
1332     bool image_opts = false;
1333     bool force_share = false;
1334 
1335     cache = BDRV_DEFAULT_CACHE;
1336     for (;;) {
1337         static const struct option long_options[] = {
1338             {"help", no_argument, 0, 'h'},
1339             {"object", required_argument, 0, OPTION_OBJECT},
1340             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
1341             {"force-share", no_argument, 0, 'U'},
1342             {0, 0, 0, 0}
1343         };
1344         c = getopt_long(argc, argv, ":hf:F:T:pqsU",
1345                         long_options, NULL);
1346         if (c == -1) {
1347             break;
1348         }
1349         switch (c) {
1350         case ':':
1351             missing_argument(argv[optind - 1]);
1352             break;
1353         case '?':
1354             unrecognized_option(argv[optind - 1]);
1355             break;
1356         case 'h':
1357             help();
1358             break;
1359         case 'f':
1360             fmt1 = optarg;
1361             break;
1362         case 'F':
1363             fmt2 = optarg;
1364             break;
1365         case 'T':
1366             cache = optarg;
1367             break;
1368         case 'p':
1369             progress = true;
1370             break;
1371         case 'q':
1372             quiet = true;
1373             break;
1374         case 's':
1375             strict = true;
1376             break;
1377         case 'U':
1378             force_share = true;
1379             break;
1380         case OPTION_OBJECT:
1381             {
1382                 Error *local_err = NULL;
1383 
1384                 if (!user_creatable_add_from_str(optarg, &local_err)) {
1385                     if (local_err) {
1386                         error_report_err(local_err);
1387                         exit(2);
1388                     } else {
1389                         /* Help was printed */
1390                         exit(EXIT_SUCCESS);
1391                     }
1392                 }
1393                 break;
1394             }
1395         case OPTION_IMAGE_OPTS:
1396             image_opts = true;
1397             break;
1398         }
1399     }
1400 
1401     /* Progress is not shown in Quiet mode */
1402     if (quiet) {
1403         progress = false;
1404     }
1405 
1406 
1407     if (optind != argc - 2) {
1408         error_exit("Expecting two image file names");
1409     }
1410     filename1 = argv[optind++];
1411     filename2 = argv[optind++];
1412 
1413     /* Initialize before goto out */
1414     qemu_progress_init(progress, 2.0);
1415 
1416     flags = 0;
1417     ret = bdrv_parse_cache_mode(cache, &flags, &writethrough);
1418     if (ret < 0) {
1419         error_report("Invalid source cache option: %s", cache);
1420         ret = 2;
1421         goto out3;
1422     }
1423 
1424     blk1 = img_open(image_opts, filename1, fmt1, flags, writethrough, quiet,
1425                     force_share);
1426     if (!blk1) {
1427         ret = 2;
1428         goto out3;
1429     }
1430 
1431     blk2 = img_open(image_opts, filename2, fmt2, flags, writethrough, quiet,
1432                     force_share);
1433     if (!blk2) {
1434         ret = 2;
1435         goto out2;
1436     }
1437     bs1 = blk_bs(blk1);
1438     bs2 = blk_bs(blk2);
1439 
1440     buf1 = blk_blockalign(blk1, IO_BUF_SIZE);
1441     buf2 = blk_blockalign(blk2, IO_BUF_SIZE);
1442     total_size1 = blk_getlength(blk1);
1443     if (total_size1 < 0) {
1444         error_report("Can't get size of %s: %s",
1445                      filename1, strerror(-total_size1));
1446         ret = 4;
1447         goto out;
1448     }
1449     total_size2 = blk_getlength(blk2);
1450     if (total_size2 < 0) {
1451         error_report("Can't get size of %s: %s",
1452                      filename2, strerror(-total_size2));
1453         ret = 4;
1454         goto out;
1455     }
1456     total_size = MIN(total_size1, total_size2);
1457     progress_base = MAX(total_size1, total_size2);
1458 
1459     qemu_progress_print(0, 100);
1460 
1461     if (strict && total_size1 != total_size2) {
1462         ret = 1;
1463         qprintf(quiet, "Strict mode: Image size mismatch!\n");
1464         goto out;
1465     }
1466 
1467     while (offset < total_size) {
1468         int status1, status2;
1469 
1470         status1 = bdrv_block_status_above(bs1, NULL, offset,
1471                                           total_size1 - offset, &pnum1, NULL,
1472                                           NULL);
1473         if (status1 < 0) {
1474             ret = 3;
1475             error_report("Sector allocation test failed for %s", filename1);
1476             goto out;
1477         }
1478         allocated1 = status1 & BDRV_BLOCK_ALLOCATED;
1479 
1480         status2 = bdrv_block_status_above(bs2, NULL, offset,
1481                                           total_size2 - offset, &pnum2, NULL,
1482                                           NULL);
1483         if (status2 < 0) {
1484             ret = 3;
1485             error_report("Sector allocation test failed for %s", filename2);
1486             goto out;
1487         }
1488         allocated2 = status2 & BDRV_BLOCK_ALLOCATED;
1489 
1490         assert(pnum1 && pnum2);
1491         chunk = MIN(pnum1, pnum2);
1492 
1493         if (strict) {
1494             if (status1 != status2) {
1495                 ret = 1;
1496                 qprintf(quiet, "Strict mode: Offset %" PRId64
1497                         " block status mismatch!\n", offset);
1498                 goto out;
1499             }
1500         }
1501         if ((status1 & BDRV_BLOCK_ZERO) && (status2 & BDRV_BLOCK_ZERO)) {
1502             /* nothing to do */
1503         } else if (allocated1 == allocated2) {
1504             if (allocated1) {
1505                 int64_t pnum;
1506 
1507                 chunk = MIN(chunk, IO_BUF_SIZE);
1508                 ret = blk_pread(blk1, offset, buf1, chunk);
1509                 if (ret < 0) {
1510                     error_report("Error while reading offset %" PRId64
1511                                  " of %s: %s",
1512                                  offset, filename1, strerror(-ret));
1513                     ret = 4;
1514                     goto out;
1515                 }
1516                 ret = blk_pread(blk2, offset, buf2, chunk);
1517                 if (ret < 0) {
1518                     error_report("Error while reading offset %" PRId64
1519                                  " of %s: %s",
1520                                  offset, filename2, strerror(-ret));
1521                     ret = 4;
1522                     goto out;
1523                 }
1524                 ret = compare_buffers(buf1, buf2, chunk, &pnum);
1525                 if (ret || pnum != chunk) {
1526                     qprintf(quiet, "Content mismatch at offset %" PRId64 "!\n",
1527                             offset + (ret ? 0 : pnum));
1528                     ret = 1;
1529                     goto out;
1530                 }
1531             }
1532         } else {
1533             chunk = MIN(chunk, IO_BUF_SIZE);
1534             if (allocated1) {
1535                 ret = check_empty_sectors(blk1, offset, chunk,
1536                                           filename1, buf1, quiet);
1537             } else {
1538                 ret = check_empty_sectors(blk2, offset, chunk,
1539                                           filename2, buf1, quiet);
1540             }
1541             if (ret) {
1542                 goto out;
1543             }
1544         }
1545         offset += chunk;
1546         qemu_progress_print(((float) chunk / progress_base) * 100, 100);
1547     }
1548 
1549     if (total_size1 != total_size2) {
1550         BlockBackend *blk_over;
1551         const char *filename_over;
1552 
1553         qprintf(quiet, "Warning: Image size mismatch!\n");
1554         if (total_size1 > total_size2) {
1555             blk_over = blk1;
1556             filename_over = filename1;
1557         } else {
1558             blk_over = blk2;
1559             filename_over = filename2;
1560         }
1561 
1562         while (offset < progress_base) {
1563             ret = bdrv_block_status_above(blk_bs(blk_over), NULL, offset,
1564                                           progress_base - offset, &chunk,
1565                                           NULL, NULL);
1566             if (ret < 0) {
1567                 ret = 3;
1568                 error_report("Sector allocation test failed for %s",
1569                              filename_over);
1570                 goto out;
1571 
1572             }
1573             if (ret & BDRV_BLOCK_ALLOCATED && !(ret & BDRV_BLOCK_ZERO)) {
1574                 chunk = MIN(chunk, IO_BUF_SIZE);
1575                 ret = check_empty_sectors(blk_over, offset, chunk,
1576                                           filename_over, buf1, quiet);
1577                 if (ret) {
1578                     goto out;
1579                 }
1580             }
1581             offset += chunk;
1582             qemu_progress_print(((float) chunk / progress_base) * 100, 100);
1583         }
1584     }
1585 
1586     qprintf(quiet, "Images are identical.\n");
1587     ret = 0;
1588 
1589 out:
1590     qemu_vfree(buf1);
1591     qemu_vfree(buf2);
1592     blk_unref(blk2);
1593 out2:
1594     blk_unref(blk1);
1595 out3:
1596     qemu_progress_end();
1597     return ret;
1598 }
1599 
1600 /* Convenience wrapper around qmp_block_dirty_bitmap_merge */
1601 static void do_dirty_bitmap_merge(const char *dst_node, const char *dst_name,
1602                                   const char *src_node, const char *src_name,
1603                                   Error **errp)
1604 {
1605     BlockDirtyBitmapMergeSource *merge_src;
1606     BlockDirtyBitmapMergeSourceList *list = NULL;
1607 
1608     merge_src = g_new0(BlockDirtyBitmapMergeSource, 1);
1609     merge_src->type = QTYPE_QDICT;
1610     merge_src->u.external.node = g_strdup(src_node);
1611     merge_src->u.external.name = g_strdup(src_name);
1612     QAPI_LIST_PREPEND(list, merge_src);
1613     qmp_block_dirty_bitmap_merge(dst_node, dst_name, list, errp);
1614     qapi_free_BlockDirtyBitmapMergeSourceList(list);
1615 }
1616 
1617 enum ImgConvertBlockStatus {
1618     BLK_DATA,
1619     BLK_ZERO,
1620     BLK_BACKING_FILE,
1621 };
1622 
1623 #define MAX_COROUTINES 16
1624 #define CONVERT_THROTTLE_GROUP "img_convert"
1625 
1626 typedef struct ImgConvertState {
1627     BlockBackend **src;
1628     int64_t *src_sectors;
1629     int *src_alignment;
1630     int src_num;
1631     int64_t total_sectors;
1632     int64_t allocated_sectors;
1633     int64_t allocated_done;
1634     int64_t sector_num;
1635     int64_t wr_offs;
1636     enum ImgConvertBlockStatus status;
1637     int64_t sector_next_status;
1638     BlockBackend *target;
1639     bool has_zero_init;
1640     bool compressed;
1641     bool target_is_new;
1642     bool target_has_backing;
1643     int64_t target_backing_sectors; /* negative if unknown */
1644     bool wr_in_order;
1645     bool copy_range;
1646     bool salvage;
1647     bool quiet;
1648     int min_sparse;
1649     int alignment;
1650     size_t cluster_sectors;
1651     size_t buf_sectors;
1652     long num_coroutines;
1653     int running_coroutines;
1654     Coroutine *co[MAX_COROUTINES];
1655     int64_t wait_sector_num[MAX_COROUTINES];
1656     CoMutex lock;
1657     int ret;
1658 } ImgConvertState;
1659 
1660 static void convert_select_part(ImgConvertState *s, int64_t sector_num,
1661                                 int *src_cur, int64_t *src_cur_offset)
1662 {
1663     *src_cur = 0;
1664     *src_cur_offset = 0;
1665     while (sector_num - *src_cur_offset >= s->src_sectors[*src_cur]) {
1666         *src_cur_offset += s->src_sectors[*src_cur];
1667         (*src_cur)++;
1668         assert(*src_cur < s->src_num);
1669     }
1670 }
1671 
1672 static int convert_iteration_sectors(ImgConvertState *s, int64_t sector_num)
1673 {
1674     int64_t src_cur_offset;
1675     int ret, n, src_cur;
1676     bool post_backing_zero = false;
1677 
1678     convert_select_part(s, sector_num, &src_cur, &src_cur_offset);
1679 
1680     assert(s->total_sectors > sector_num);
1681     n = MIN(s->total_sectors - sector_num, BDRV_REQUEST_MAX_SECTORS);
1682 
1683     if (s->target_backing_sectors >= 0) {
1684         if (sector_num >= s->target_backing_sectors) {
1685             post_backing_zero = true;
1686         } else if (sector_num + n > s->target_backing_sectors) {
1687             /* Split requests around target_backing_sectors (because
1688              * starting from there, zeros are handled differently) */
1689             n = s->target_backing_sectors - sector_num;
1690         }
1691     }
1692 
1693     if (s->sector_next_status <= sector_num) {
1694         uint64_t offset = (sector_num - src_cur_offset) * BDRV_SECTOR_SIZE;
1695         int64_t count;
1696         int tail;
1697         BlockDriverState *src_bs = blk_bs(s->src[src_cur]);
1698         BlockDriverState *base;
1699 
1700         if (s->target_has_backing) {
1701             base = bdrv_cow_bs(bdrv_skip_filters(src_bs));
1702         } else {
1703             base = NULL;
1704         }
1705 
1706         do {
1707             count = n * BDRV_SECTOR_SIZE;
1708 
1709             ret = bdrv_block_status_above(src_bs, base, offset, count, &count,
1710                                           NULL, NULL);
1711 
1712             if (ret < 0) {
1713                 if (s->salvage) {
1714                     if (n == 1) {
1715                         if (!s->quiet) {
1716                             warn_report("error while reading block status at "
1717                                         "offset %" PRIu64 ": %s", offset,
1718                                         strerror(-ret));
1719                         }
1720                         /* Just try to read the data, then */
1721                         ret = BDRV_BLOCK_DATA;
1722                         count = BDRV_SECTOR_SIZE;
1723                     } else {
1724                         /* Retry on a shorter range */
1725                         n = DIV_ROUND_UP(n, 4);
1726                     }
1727                 } else {
1728                     error_report("error while reading block status at offset "
1729                                  "%" PRIu64 ": %s", offset, strerror(-ret));
1730                     return ret;
1731                 }
1732             }
1733         } while (ret < 0);
1734 
1735         n = DIV_ROUND_UP(count, BDRV_SECTOR_SIZE);
1736 
1737         /*
1738          * Avoid that s->sector_next_status becomes unaligned to the source
1739          * request alignment and/or cluster size to avoid unnecessary read
1740          * cycles.
1741          */
1742         tail = (sector_num - src_cur_offset + n) % s->src_alignment[src_cur];
1743         if (n > tail) {
1744             n -= tail;
1745         }
1746 
1747         if (ret & BDRV_BLOCK_ZERO) {
1748             s->status = post_backing_zero ? BLK_BACKING_FILE : BLK_ZERO;
1749         } else if (ret & BDRV_BLOCK_DATA) {
1750             s->status = BLK_DATA;
1751         } else {
1752             s->status = s->target_has_backing ? BLK_BACKING_FILE : BLK_DATA;
1753         }
1754 
1755         s->sector_next_status = sector_num + n;
1756     }
1757 
1758     n = MIN(n, s->sector_next_status - sector_num);
1759     if (s->status == BLK_DATA) {
1760         n = MIN(n, s->buf_sectors);
1761     }
1762 
1763     /* We need to write complete clusters for compressed images, so if an
1764      * unallocated area is shorter than that, we must consider the whole
1765      * cluster allocated. */
1766     if (s->compressed) {
1767         if (n < s->cluster_sectors) {
1768             n = MIN(s->cluster_sectors, s->total_sectors - sector_num);
1769             s->status = BLK_DATA;
1770         } else {
1771             n = QEMU_ALIGN_DOWN(n, s->cluster_sectors);
1772         }
1773     }
1774 
1775     return n;
1776 }
1777 
1778 static int coroutine_fn convert_co_read(ImgConvertState *s, int64_t sector_num,
1779                                         int nb_sectors, uint8_t *buf)
1780 {
1781     uint64_t single_read_until = 0;
1782     int n, ret;
1783 
1784     assert(nb_sectors <= s->buf_sectors);
1785     while (nb_sectors > 0) {
1786         BlockBackend *blk;
1787         int src_cur;
1788         int64_t bs_sectors, src_cur_offset;
1789         uint64_t offset;
1790 
1791         /* In the case of compression with multiple source files, we can get a
1792          * nb_sectors that spreads into the next part. So we must be able to
1793          * read across multiple BDSes for one convert_read() call. */
1794         convert_select_part(s, sector_num, &src_cur, &src_cur_offset);
1795         blk = s->src[src_cur];
1796         bs_sectors = s->src_sectors[src_cur];
1797 
1798         offset = (sector_num - src_cur_offset) << BDRV_SECTOR_BITS;
1799 
1800         n = MIN(nb_sectors, bs_sectors - (sector_num - src_cur_offset));
1801         if (single_read_until > offset) {
1802             n = 1;
1803         }
1804 
1805         ret = blk_co_pread(blk, offset, n << BDRV_SECTOR_BITS, buf, 0);
1806         if (ret < 0) {
1807             if (s->salvage) {
1808                 if (n > 1) {
1809                     single_read_until = offset + (n << BDRV_SECTOR_BITS);
1810                     continue;
1811                 } else {
1812                     if (!s->quiet) {
1813                         warn_report("error while reading offset %" PRIu64
1814                                     ": %s", offset, strerror(-ret));
1815                     }
1816                     memset(buf, 0, BDRV_SECTOR_SIZE);
1817                 }
1818             } else {
1819                 return ret;
1820             }
1821         }
1822 
1823         sector_num += n;
1824         nb_sectors -= n;
1825         buf += n * BDRV_SECTOR_SIZE;
1826     }
1827 
1828     return 0;
1829 }
1830 
1831 
1832 static int coroutine_fn convert_co_write(ImgConvertState *s, int64_t sector_num,
1833                                          int nb_sectors, uint8_t *buf,
1834                                          enum ImgConvertBlockStatus status)
1835 {
1836     int ret;
1837 
1838     while (nb_sectors > 0) {
1839         int n = nb_sectors;
1840         BdrvRequestFlags flags = s->compressed ? BDRV_REQ_WRITE_COMPRESSED : 0;
1841 
1842         switch (status) {
1843         case BLK_BACKING_FILE:
1844             /* If we have a backing file, leave clusters unallocated that are
1845              * unallocated in the source image, so that the backing file is
1846              * visible at the respective offset. */
1847             assert(s->target_has_backing);
1848             break;
1849 
1850         case BLK_DATA:
1851             /* If we're told to keep the target fully allocated (-S 0) or there
1852              * is real non-zero data, we must write it. Otherwise we can treat
1853              * it as zero sectors.
1854              * Compressed clusters need to be written as a whole, so in that
1855              * case we can only save the write if the buffer is completely
1856              * zeroed. */
1857             if (!s->min_sparse ||
1858                 (!s->compressed &&
1859                  is_allocated_sectors_min(buf, n, &n, s->min_sparse,
1860                                           sector_num, s->alignment)) ||
1861                 (s->compressed &&
1862                  !buffer_is_zero(buf, n * BDRV_SECTOR_SIZE)))
1863             {
1864                 ret = blk_co_pwrite(s->target, sector_num << BDRV_SECTOR_BITS,
1865                                     n << BDRV_SECTOR_BITS, buf, flags);
1866                 if (ret < 0) {
1867                     return ret;
1868                 }
1869                 break;
1870             }
1871             /* fall-through */
1872 
1873         case BLK_ZERO:
1874             if (s->has_zero_init) {
1875                 assert(!s->target_has_backing);
1876                 break;
1877             }
1878             ret = blk_co_pwrite_zeroes(s->target,
1879                                        sector_num << BDRV_SECTOR_BITS,
1880                                        n << BDRV_SECTOR_BITS,
1881                                        BDRV_REQ_MAY_UNMAP);
1882             if (ret < 0) {
1883                 return ret;
1884             }
1885             break;
1886         }
1887 
1888         sector_num += n;
1889         nb_sectors -= n;
1890         buf += n * BDRV_SECTOR_SIZE;
1891     }
1892 
1893     return 0;
1894 }
1895 
1896 static int coroutine_fn convert_co_copy_range(ImgConvertState *s, int64_t sector_num,
1897                                               int nb_sectors)
1898 {
1899     int n, ret;
1900 
1901     while (nb_sectors > 0) {
1902         BlockBackend *blk;
1903         int src_cur;
1904         int64_t bs_sectors, src_cur_offset;
1905         int64_t offset;
1906 
1907         convert_select_part(s, sector_num, &src_cur, &src_cur_offset);
1908         offset = (sector_num - src_cur_offset) << BDRV_SECTOR_BITS;
1909         blk = s->src[src_cur];
1910         bs_sectors = s->src_sectors[src_cur];
1911 
1912         n = MIN(nb_sectors, bs_sectors - (sector_num - src_cur_offset));
1913 
1914         ret = blk_co_copy_range(blk, offset, s->target,
1915                                 sector_num << BDRV_SECTOR_BITS,
1916                                 n << BDRV_SECTOR_BITS, 0, 0);
1917         if (ret < 0) {
1918             return ret;
1919         }
1920 
1921         sector_num += n;
1922         nb_sectors -= n;
1923     }
1924     return 0;
1925 }
1926 
1927 static void coroutine_fn convert_co_do_copy(void *opaque)
1928 {
1929     ImgConvertState *s = opaque;
1930     uint8_t *buf = NULL;
1931     int ret, i;
1932     int index = -1;
1933 
1934     for (i = 0; i < s->num_coroutines; i++) {
1935         if (s->co[i] == qemu_coroutine_self()) {
1936             index = i;
1937             break;
1938         }
1939     }
1940     assert(index >= 0);
1941 
1942     s->running_coroutines++;
1943     buf = blk_blockalign(s->target, s->buf_sectors * BDRV_SECTOR_SIZE);
1944 
1945     while (1) {
1946         int n;
1947         int64_t sector_num;
1948         enum ImgConvertBlockStatus status;
1949         bool copy_range;
1950 
1951         qemu_co_mutex_lock(&s->lock);
1952         if (s->ret != -EINPROGRESS || s->sector_num >= s->total_sectors) {
1953             qemu_co_mutex_unlock(&s->lock);
1954             break;
1955         }
1956         n = convert_iteration_sectors(s, s->sector_num);
1957         if (n < 0) {
1958             qemu_co_mutex_unlock(&s->lock);
1959             s->ret = n;
1960             break;
1961         }
1962         /* save current sector and allocation status to local variables */
1963         sector_num = s->sector_num;
1964         status = s->status;
1965         if (!s->min_sparse && s->status == BLK_ZERO) {
1966             n = MIN(n, s->buf_sectors);
1967         }
1968         /* increment global sector counter so that other coroutines can
1969          * already continue reading beyond this request */
1970         s->sector_num += n;
1971         qemu_co_mutex_unlock(&s->lock);
1972 
1973         if (status == BLK_DATA || (!s->min_sparse && status == BLK_ZERO)) {
1974             s->allocated_done += n;
1975             qemu_progress_print(100.0 * s->allocated_done /
1976                                         s->allocated_sectors, 0);
1977         }
1978 
1979 retry:
1980         copy_range = s->copy_range && s->status == BLK_DATA;
1981         if (status == BLK_DATA && !copy_range) {
1982             ret = convert_co_read(s, sector_num, n, buf);
1983             if (ret < 0) {
1984                 error_report("error while reading at byte %lld: %s",
1985                              sector_num * BDRV_SECTOR_SIZE, strerror(-ret));
1986                 s->ret = ret;
1987             }
1988         } else if (!s->min_sparse && status == BLK_ZERO) {
1989             status = BLK_DATA;
1990             memset(buf, 0x00, n * BDRV_SECTOR_SIZE);
1991         }
1992 
1993         if (s->wr_in_order) {
1994             /* keep writes in order */
1995             while (s->wr_offs != sector_num && s->ret == -EINPROGRESS) {
1996                 s->wait_sector_num[index] = sector_num;
1997                 qemu_coroutine_yield();
1998             }
1999             s->wait_sector_num[index] = -1;
2000         }
2001 
2002         if (s->ret == -EINPROGRESS) {
2003             if (copy_range) {
2004                 ret = convert_co_copy_range(s, sector_num, n);
2005                 if (ret) {
2006                     s->copy_range = false;
2007                     goto retry;
2008                 }
2009             } else {
2010                 ret = convert_co_write(s, sector_num, n, buf, status);
2011             }
2012             if (ret < 0) {
2013                 error_report("error while writing at byte %lld: %s",
2014                              sector_num * BDRV_SECTOR_SIZE, strerror(-ret));
2015                 s->ret = ret;
2016             }
2017         }
2018 
2019         if (s->wr_in_order) {
2020             /* reenter the coroutine that might have waited
2021              * for this write to complete */
2022             s->wr_offs = sector_num + n;
2023             for (i = 0; i < s->num_coroutines; i++) {
2024                 if (s->co[i] && s->wait_sector_num[i] == s->wr_offs) {
2025                     /*
2026                      * A -> B -> A cannot occur because A has
2027                      * s->wait_sector_num[i] == -1 during A -> B.  Therefore
2028                      * B will never enter A during this time window.
2029                      */
2030                     qemu_coroutine_enter(s->co[i]);
2031                     break;
2032                 }
2033             }
2034         }
2035     }
2036 
2037     qemu_vfree(buf);
2038     s->co[index] = NULL;
2039     s->running_coroutines--;
2040     if (!s->running_coroutines && s->ret == -EINPROGRESS) {
2041         /* the convert job finished successfully */
2042         s->ret = 0;
2043     }
2044 }
2045 
2046 static int convert_do_copy(ImgConvertState *s)
2047 {
2048     int ret, i, n;
2049     int64_t sector_num = 0;
2050 
2051     /* Check whether we have zero initialisation or can get it efficiently */
2052     if (!s->has_zero_init && s->target_is_new && s->min_sparse &&
2053         !s->target_has_backing) {
2054         s->has_zero_init = bdrv_has_zero_init(blk_bs(s->target));
2055     }
2056 
2057     /* Allocate buffer for copied data. For compressed images, only one cluster
2058      * can be copied at a time. */
2059     if (s->compressed) {
2060         if (s->cluster_sectors <= 0 || s->cluster_sectors > s->buf_sectors) {
2061             error_report("invalid cluster size");
2062             return -EINVAL;
2063         }
2064         s->buf_sectors = s->cluster_sectors;
2065     }
2066 
2067     while (sector_num < s->total_sectors) {
2068         n = convert_iteration_sectors(s, sector_num);
2069         if (n < 0) {
2070             return n;
2071         }
2072         if (s->status == BLK_DATA || (!s->min_sparse && s->status == BLK_ZERO))
2073         {
2074             s->allocated_sectors += n;
2075         }
2076         sector_num += n;
2077     }
2078 
2079     /* Do the copy */
2080     s->sector_next_status = 0;
2081     s->ret = -EINPROGRESS;
2082 
2083     qemu_co_mutex_init(&s->lock);
2084     for (i = 0; i < s->num_coroutines; i++) {
2085         s->co[i] = qemu_coroutine_create(convert_co_do_copy, s);
2086         s->wait_sector_num[i] = -1;
2087         qemu_coroutine_enter(s->co[i]);
2088     }
2089 
2090     while (s->running_coroutines) {
2091         main_loop_wait(false);
2092     }
2093 
2094     if (s->compressed && !s->ret) {
2095         /* signal EOF to align */
2096         ret = blk_pwrite_compressed(s->target, 0, NULL, 0);
2097         if (ret < 0) {
2098             return ret;
2099         }
2100     }
2101 
2102     return s->ret;
2103 }
2104 
2105 /* Check that bitmaps can be copied, or output an error */
2106 static int convert_check_bitmaps(BlockDriverState *src, bool skip_broken)
2107 {
2108     BdrvDirtyBitmap *bm;
2109 
2110     if (!bdrv_supports_persistent_dirty_bitmap(src)) {
2111         error_report("Source lacks bitmap support");
2112         return -1;
2113     }
2114     FOR_EACH_DIRTY_BITMAP(src, bm) {
2115         if (!bdrv_dirty_bitmap_get_persistence(bm)) {
2116             continue;
2117         }
2118         if (!skip_broken && bdrv_dirty_bitmap_inconsistent(bm)) {
2119             error_report("Cannot copy inconsistent bitmap '%s'",
2120                          bdrv_dirty_bitmap_name(bm));
2121             error_printf("Try --skip-broken-bitmaps, or "
2122                          "use 'qemu-img bitmap --remove' to delete it\n");
2123             return -1;
2124         }
2125     }
2126     return 0;
2127 }
2128 
2129 static int convert_copy_bitmaps(BlockDriverState *src, BlockDriverState *dst,
2130                                 bool skip_broken)
2131 {
2132     BdrvDirtyBitmap *bm;
2133     Error *err = NULL;
2134 
2135     FOR_EACH_DIRTY_BITMAP(src, bm) {
2136         const char *name;
2137 
2138         if (!bdrv_dirty_bitmap_get_persistence(bm)) {
2139             continue;
2140         }
2141         name = bdrv_dirty_bitmap_name(bm);
2142         if (skip_broken && bdrv_dirty_bitmap_inconsistent(bm)) {
2143             warn_report("Skipping inconsistent bitmap '%s'", name);
2144             continue;
2145         }
2146         qmp_block_dirty_bitmap_add(dst->node_name, name,
2147                                    true, bdrv_dirty_bitmap_granularity(bm),
2148                                    true, true,
2149                                    true, !bdrv_dirty_bitmap_enabled(bm),
2150                                    &err);
2151         if (err) {
2152             error_reportf_err(err, "Failed to create bitmap %s: ", name);
2153             return -1;
2154         }
2155 
2156         do_dirty_bitmap_merge(dst->node_name, name, src->node_name, name,
2157                               &err);
2158         if (err) {
2159             error_reportf_err(err, "Failed to populate bitmap %s: ", name);
2160             qmp_block_dirty_bitmap_remove(dst->node_name, name, NULL);
2161             return -1;
2162         }
2163     }
2164 
2165     return 0;
2166 }
2167 
2168 #define MAX_BUF_SECTORS 32768
2169 
2170 static void set_rate_limit(BlockBackend *blk, int64_t rate_limit)
2171 {
2172     ThrottleConfig cfg;
2173 
2174     throttle_config_init(&cfg);
2175     cfg.buckets[THROTTLE_BPS_WRITE].avg = rate_limit;
2176 
2177     blk_io_limits_enable(blk, CONVERT_THROTTLE_GROUP);
2178     blk_set_io_limits(blk, &cfg);
2179 }
2180 
2181 static int img_convert(int argc, char **argv)
2182 {
2183     int c, bs_i, flags, src_flags = BDRV_O_NO_SHARE;
2184     const char *fmt = NULL, *out_fmt = NULL, *cache = "unsafe",
2185                *src_cache = BDRV_DEFAULT_CACHE, *out_baseimg = NULL,
2186                *out_filename, *out_baseimg_param, *snapshot_name = NULL;
2187     BlockDriver *drv = NULL, *proto_drv = NULL;
2188     BlockDriverInfo bdi;
2189     BlockDriverState *out_bs;
2190     QemuOpts *opts = NULL, *sn_opts = NULL;
2191     QemuOptsList *create_opts = NULL;
2192     QDict *open_opts = NULL;
2193     char *options = NULL;
2194     Error *local_err = NULL;
2195     bool writethrough, src_writethrough, image_opts = false,
2196          skip_create = false, progress = false, tgt_image_opts = false;
2197     int64_t ret = -EINVAL;
2198     bool force_share = false;
2199     bool explict_min_sparse = false;
2200     bool bitmaps = false;
2201     bool skip_broken = false;
2202     int64_t rate_limit = 0;
2203 
2204     ImgConvertState s = (ImgConvertState) {
2205         /* Need at least 4k of zeros for sparse detection */
2206         .min_sparse         = 8,
2207         .copy_range         = false,
2208         .buf_sectors        = IO_BUF_SIZE / BDRV_SECTOR_SIZE,
2209         .wr_in_order        = true,
2210         .num_coroutines     = 8,
2211     };
2212 
2213     for(;;) {
2214         static const struct option long_options[] = {
2215             {"help", no_argument, 0, 'h'},
2216             {"object", required_argument, 0, OPTION_OBJECT},
2217             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
2218             {"force-share", no_argument, 0, 'U'},
2219             {"target-image-opts", no_argument, 0, OPTION_TARGET_IMAGE_OPTS},
2220             {"salvage", no_argument, 0, OPTION_SALVAGE},
2221             {"target-is-zero", no_argument, 0, OPTION_TARGET_IS_ZERO},
2222             {"bitmaps", no_argument, 0, OPTION_BITMAPS},
2223             {"skip-broken-bitmaps", no_argument, 0, OPTION_SKIP_BROKEN},
2224             {0, 0, 0, 0}
2225         };
2226         c = getopt_long(argc, argv, ":hf:O:B:Cco:l:S:pt:T:qnm:WUr:",
2227                         long_options, NULL);
2228         if (c == -1) {
2229             break;
2230         }
2231         switch(c) {
2232         case ':':
2233             missing_argument(argv[optind - 1]);
2234             break;
2235         case '?':
2236             unrecognized_option(argv[optind - 1]);
2237             break;
2238         case 'h':
2239             help();
2240             break;
2241         case 'f':
2242             fmt = optarg;
2243             break;
2244         case 'O':
2245             out_fmt = optarg;
2246             break;
2247         case 'B':
2248             out_baseimg = optarg;
2249             break;
2250         case 'C':
2251             s.copy_range = true;
2252             break;
2253         case 'c':
2254             s.compressed = true;
2255             break;
2256         case 'o':
2257             if (accumulate_options(&options, optarg) < 0) {
2258                 goto fail_getopt;
2259             }
2260             break;
2261         case 'l':
2262             if (strstart(optarg, SNAPSHOT_OPT_BASE, NULL)) {
2263                 sn_opts = qemu_opts_parse_noisily(&internal_snapshot_opts,
2264                                                   optarg, false);
2265                 if (!sn_opts) {
2266                     error_report("Failed in parsing snapshot param '%s'",
2267                                  optarg);
2268                     goto fail_getopt;
2269                 }
2270             } else {
2271                 snapshot_name = optarg;
2272             }
2273             break;
2274         case 'S':
2275         {
2276             int64_t sval;
2277 
2278             sval = cvtnum("buffer size for sparse output", optarg);
2279             if (sval < 0) {
2280                 goto fail_getopt;
2281             } else if (!QEMU_IS_ALIGNED(sval, BDRV_SECTOR_SIZE) ||
2282                 sval / BDRV_SECTOR_SIZE > MAX_BUF_SECTORS) {
2283                 error_report("Invalid buffer size for sparse output specified. "
2284                     "Valid sizes are multiples of %llu up to %llu. Select "
2285                     "0 to disable sparse detection (fully allocates output).",
2286                     BDRV_SECTOR_SIZE, MAX_BUF_SECTORS * BDRV_SECTOR_SIZE);
2287                 goto fail_getopt;
2288             }
2289 
2290             s.min_sparse = sval / BDRV_SECTOR_SIZE;
2291             explict_min_sparse = true;
2292             break;
2293         }
2294         case 'p':
2295             progress = true;
2296             break;
2297         case 't':
2298             cache = optarg;
2299             break;
2300         case 'T':
2301             src_cache = optarg;
2302             break;
2303         case 'q':
2304             s.quiet = true;
2305             break;
2306         case 'n':
2307             skip_create = true;
2308             break;
2309         case 'm':
2310             if (qemu_strtol(optarg, NULL, 0, &s.num_coroutines) ||
2311                 s.num_coroutines < 1 || s.num_coroutines > MAX_COROUTINES) {
2312                 error_report("Invalid number of coroutines. Allowed number of"
2313                              " coroutines is between 1 and %d", MAX_COROUTINES);
2314                 goto fail_getopt;
2315             }
2316             break;
2317         case 'W':
2318             s.wr_in_order = false;
2319             break;
2320         case 'U':
2321             force_share = true;
2322             break;
2323         case 'r':
2324             rate_limit = cvtnum("rate limit", optarg);
2325             if (rate_limit < 0) {
2326                 goto fail_getopt;
2327             }
2328             break;
2329         case OPTION_OBJECT:
2330             user_creatable_process_cmdline(optarg);
2331             break;
2332         case OPTION_IMAGE_OPTS:
2333             image_opts = true;
2334             break;
2335         case OPTION_SALVAGE:
2336             s.salvage = true;
2337             break;
2338         case OPTION_TARGET_IMAGE_OPTS:
2339             tgt_image_opts = true;
2340             break;
2341         case OPTION_TARGET_IS_ZERO:
2342             /*
2343              * The user asserting that the target is blank has the
2344              * same effect as the target driver supporting zero
2345              * initialisation.
2346              */
2347             s.has_zero_init = true;
2348             break;
2349         case OPTION_BITMAPS:
2350             bitmaps = true;
2351             break;
2352         case OPTION_SKIP_BROKEN:
2353             skip_broken = true;
2354             break;
2355         }
2356     }
2357 
2358     if (!out_fmt && !tgt_image_opts) {
2359         out_fmt = "raw";
2360     }
2361 
2362     if (skip_broken && !bitmaps) {
2363         error_report("Use of --skip-broken-bitmaps requires --bitmaps");
2364         goto fail_getopt;
2365     }
2366 
2367     if (s.compressed && s.copy_range) {
2368         error_report("Cannot enable copy offloading when -c is used");
2369         goto fail_getopt;
2370     }
2371 
2372     if (explict_min_sparse && s.copy_range) {
2373         error_report("Cannot enable copy offloading when -S is used");
2374         goto fail_getopt;
2375     }
2376 
2377     if (s.copy_range && s.salvage) {
2378         error_report("Cannot use copy offloading in salvaging mode");
2379         goto fail_getopt;
2380     }
2381 
2382     if (tgt_image_opts && !skip_create) {
2383         error_report("--target-image-opts requires use of -n flag");
2384         goto fail_getopt;
2385     }
2386 
2387     if (skip_create && options) {
2388         error_report("-o has no effect when skipping image creation");
2389         goto fail_getopt;
2390     }
2391 
2392     if (s.has_zero_init && !skip_create) {
2393         error_report("--target-is-zero requires use of -n flag");
2394         goto fail_getopt;
2395     }
2396 
2397     s.src_num = argc - optind - 1;
2398     out_filename = s.src_num >= 1 ? argv[argc - 1] : NULL;
2399 
2400     if (options && has_help_option(options)) {
2401         if (out_fmt) {
2402             ret = print_block_option_help(out_filename, out_fmt);
2403             goto fail_getopt;
2404         } else {
2405             error_report("Option help requires a format be specified");
2406             goto fail_getopt;
2407         }
2408     }
2409 
2410     if (s.src_num < 1) {
2411         error_report("Must specify image file name");
2412         goto fail_getopt;
2413     }
2414 
2415     /* ret is still -EINVAL until here */
2416     ret = bdrv_parse_cache_mode(src_cache, &src_flags, &src_writethrough);
2417     if (ret < 0) {
2418         error_report("Invalid source cache option: %s", src_cache);
2419         goto fail_getopt;
2420     }
2421 
2422     /* Initialize before goto out */
2423     if (s.quiet) {
2424         progress = false;
2425     }
2426     qemu_progress_init(progress, 1.0);
2427     qemu_progress_print(0, 100);
2428 
2429     s.src = g_new0(BlockBackend *, s.src_num);
2430     s.src_sectors = g_new(int64_t, s.src_num);
2431     s.src_alignment = g_new(int, s.src_num);
2432 
2433     for (bs_i = 0; bs_i < s.src_num; bs_i++) {
2434         BlockDriverState *src_bs;
2435         s.src[bs_i] = img_open(image_opts, argv[optind + bs_i],
2436                                fmt, src_flags, src_writethrough, s.quiet,
2437                                force_share);
2438         if (!s.src[bs_i]) {
2439             ret = -1;
2440             goto out;
2441         }
2442         s.src_sectors[bs_i] = blk_nb_sectors(s.src[bs_i]);
2443         if (s.src_sectors[bs_i] < 0) {
2444             error_report("Could not get size of %s: %s",
2445                          argv[optind + bs_i], strerror(-s.src_sectors[bs_i]));
2446             ret = -1;
2447             goto out;
2448         }
2449         src_bs = blk_bs(s.src[bs_i]);
2450         s.src_alignment[bs_i] = DIV_ROUND_UP(src_bs->bl.request_alignment,
2451                                              BDRV_SECTOR_SIZE);
2452         if (!bdrv_get_info(src_bs, &bdi)) {
2453             s.src_alignment[bs_i] = MAX(s.src_alignment[bs_i],
2454                                         bdi.cluster_size / BDRV_SECTOR_SIZE);
2455         }
2456         s.total_sectors += s.src_sectors[bs_i];
2457     }
2458 
2459     if (sn_opts) {
2460         bdrv_snapshot_load_tmp(blk_bs(s.src[0]),
2461                                qemu_opt_get(sn_opts, SNAPSHOT_OPT_ID),
2462                                qemu_opt_get(sn_opts, SNAPSHOT_OPT_NAME),
2463                                &local_err);
2464     } else if (snapshot_name != NULL) {
2465         if (s.src_num > 1) {
2466             error_report("No support for concatenating multiple snapshot");
2467             ret = -1;
2468             goto out;
2469         }
2470 
2471         bdrv_snapshot_load_tmp_by_id_or_name(blk_bs(s.src[0]), snapshot_name,
2472                                              &local_err);
2473     }
2474     if (local_err) {
2475         error_reportf_err(local_err, "Failed to load snapshot: ");
2476         ret = -1;
2477         goto out;
2478     }
2479 
2480     if (!skip_create) {
2481         /* Find driver and parse its options */
2482         drv = bdrv_find_format(out_fmt);
2483         if (!drv) {
2484             error_report("Unknown file format '%s'", out_fmt);
2485             ret = -1;
2486             goto out;
2487         }
2488 
2489         proto_drv = bdrv_find_protocol(out_filename, true, &local_err);
2490         if (!proto_drv) {
2491             error_report_err(local_err);
2492             ret = -1;
2493             goto out;
2494         }
2495 
2496         if (!drv->create_opts) {
2497             error_report("Format driver '%s' does not support image creation",
2498                          drv->format_name);
2499             ret = -1;
2500             goto out;
2501         }
2502 
2503         if (!proto_drv->create_opts) {
2504             error_report("Protocol driver '%s' does not support image creation",
2505                          proto_drv->format_name);
2506             ret = -1;
2507             goto out;
2508         }
2509 
2510         create_opts = qemu_opts_append(create_opts, drv->create_opts);
2511         create_opts = qemu_opts_append(create_opts, proto_drv->create_opts);
2512 
2513         opts = qemu_opts_create(create_opts, NULL, 0, &error_abort);
2514         if (options) {
2515             if (!qemu_opts_do_parse(opts, options, NULL, &local_err)) {
2516                 error_report_err(local_err);
2517                 ret = -1;
2518                 goto out;
2519             }
2520         }
2521 
2522         qemu_opt_set_number(opts, BLOCK_OPT_SIZE,
2523                             s.total_sectors * BDRV_SECTOR_SIZE, &error_abort);
2524         ret = add_old_style_options(out_fmt, opts, out_baseimg, NULL);
2525         if (ret < 0) {
2526             goto out;
2527         }
2528     }
2529 
2530     /* Get backing file name if -o backing_file was used */
2531     out_baseimg_param = qemu_opt_get(opts, BLOCK_OPT_BACKING_FILE);
2532     if (out_baseimg_param) {
2533         out_baseimg = out_baseimg_param;
2534     }
2535     s.target_has_backing = (bool) out_baseimg;
2536 
2537     if (s.has_zero_init && s.target_has_backing) {
2538         error_report("Cannot use --target-is-zero when the destination "
2539                      "image has a backing file");
2540         goto out;
2541     }
2542 
2543     if (s.src_num > 1 && out_baseimg) {
2544         error_report("Having a backing file for the target makes no sense when "
2545                      "concatenating multiple input images");
2546         ret = -1;
2547         goto out;
2548     }
2549 
2550     if (out_baseimg_param) {
2551         if (!qemu_opt_get(opts, BLOCK_OPT_BACKING_FMT)) {
2552             error_report("Use of backing file requires explicit "
2553                          "backing format");
2554             ret = -1;
2555             goto out;
2556         }
2557     }
2558 
2559     /* Check if compression is supported */
2560     if (s.compressed) {
2561         bool encryption =
2562             qemu_opt_get_bool(opts, BLOCK_OPT_ENCRYPT, false);
2563         const char *encryptfmt =
2564             qemu_opt_get(opts, BLOCK_OPT_ENCRYPT_FORMAT);
2565         const char *preallocation =
2566             qemu_opt_get(opts, BLOCK_OPT_PREALLOC);
2567 
2568         if (drv && !block_driver_can_compress(drv)) {
2569             error_report("Compression not supported for this file format");
2570             ret = -1;
2571             goto out;
2572         }
2573 
2574         if (encryption || encryptfmt) {
2575             error_report("Compression and encryption not supported at "
2576                          "the same time");
2577             ret = -1;
2578             goto out;
2579         }
2580 
2581         if (preallocation
2582             && strcmp(preallocation, "off"))
2583         {
2584             error_report("Compression and preallocation not supported at "
2585                          "the same time");
2586             ret = -1;
2587             goto out;
2588         }
2589     }
2590 
2591     /* Determine if bitmaps need copying */
2592     if (bitmaps) {
2593         if (s.src_num > 1) {
2594             error_report("Copying bitmaps only possible with single source");
2595             ret = -1;
2596             goto out;
2597         }
2598         ret = convert_check_bitmaps(blk_bs(s.src[0]), skip_broken);
2599         if (ret < 0) {
2600             goto out;
2601         }
2602     }
2603 
2604     /*
2605      * The later open call will need any decryption secrets, and
2606      * bdrv_create() will purge "opts", so extract them now before
2607      * they are lost.
2608      */
2609     if (!skip_create) {
2610         open_opts = qdict_new();
2611         qemu_opt_foreach(opts, img_add_key_secrets, open_opts, &error_abort);
2612 
2613         /* Create the new image */
2614         ret = bdrv_create(drv, out_filename, opts, &local_err);
2615         if (ret < 0) {
2616             error_reportf_err(local_err, "%s: error while converting %s: ",
2617                               out_filename, out_fmt);
2618             goto out;
2619         }
2620     }
2621 
2622     s.target_is_new = !skip_create;
2623 
2624     flags = s.min_sparse ? (BDRV_O_RDWR | BDRV_O_UNMAP) : BDRV_O_RDWR;
2625     ret = bdrv_parse_cache_mode(cache, &flags, &writethrough);
2626     if (ret < 0) {
2627         error_report("Invalid cache option: %s", cache);
2628         goto out;
2629     }
2630 
2631     if (skip_create) {
2632         s.target = img_open(tgt_image_opts, out_filename, out_fmt,
2633                             flags, writethrough, s.quiet, false);
2634     } else {
2635         /* TODO ultimately we should allow --target-image-opts
2636          * to be used even when -n is not given.
2637          * That has to wait for bdrv_create to be improved
2638          * to allow filenames in option syntax
2639          */
2640         s.target = img_open_file(out_filename, open_opts, out_fmt,
2641                                  flags, writethrough, s.quiet, false);
2642         open_opts = NULL; /* blk_new_open will have freed it */
2643     }
2644     if (!s.target) {
2645         ret = -1;
2646         goto out;
2647     }
2648     out_bs = blk_bs(s.target);
2649 
2650     if (bitmaps && !bdrv_supports_persistent_dirty_bitmap(out_bs)) {
2651         error_report("Format driver '%s' does not support bitmaps",
2652                      out_bs->drv->format_name);
2653         ret = -1;
2654         goto out;
2655     }
2656 
2657     if (s.compressed && !block_driver_can_compress(out_bs->drv)) {
2658         error_report("Compression not supported for this file format");
2659         ret = -1;
2660         goto out;
2661     }
2662 
2663     /* increase bufsectors from the default 4096 (2M) if opt_transfer
2664      * or discard_alignment of the out_bs is greater. Limit to
2665      * MAX_BUF_SECTORS as maximum which is currently 32768 (16MB). */
2666     s.buf_sectors = MIN(MAX_BUF_SECTORS,
2667                         MAX(s.buf_sectors,
2668                             MAX(out_bs->bl.opt_transfer >> BDRV_SECTOR_BITS,
2669                                 out_bs->bl.pdiscard_alignment >>
2670                                 BDRV_SECTOR_BITS)));
2671 
2672     /* try to align the write requests to the destination to avoid unnecessary
2673      * RMW cycles. */
2674     s.alignment = MAX(pow2floor(s.min_sparse),
2675                       DIV_ROUND_UP(out_bs->bl.request_alignment,
2676                                    BDRV_SECTOR_SIZE));
2677     assert(is_power_of_2(s.alignment));
2678 
2679     if (skip_create) {
2680         int64_t output_sectors = blk_nb_sectors(s.target);
2681         if (output_sectors < 0) {
2682             error_report("unable to get output image length: %s",
2683                          strerror(-output_sectors));
2684             ret = -1;
2685             goto out;
2686         } else if (output_sectors < s.total_sectors) {
2687             error_report("output file is smaller than input file");
2688             ret = -1;
2689             goto out;
2690         }
2691     }
2692 
2693     if (s.target_has_backing && s.target_is_new) {
2694         /* Errors are treated as "backing length unknown" (which means
2695          * s.target_backing_sectors has to be negative, which it will
2696          * be automatically).  The backing file length is used only
2697          * for optimizations, so such a case is not fatal. */
2698         s.target_backing_sectors =
2699             bdrv_nb_sectors(bdrv_backing_chain_next(out_bs));
2700     } else {
2701         s.target_backing_sectors = -1;
2702     }
2703 
2704     ret = bdrv_get_info(out_bs, &bdi);
2705     if (ret < 0) {
2706         if (s.compressed) {
2707             error_report("could not get block driver info");
2708             goto out;
2709         }
2710     } else {
2711         s.compressed = s.compressed || bdi.needs_compressed_writes;
2712         s.cluster_sectors = bdi.cluster_size / BDRV_SECTOR_SIZE;
2713     }
2714 
2715     if (rate_limit) {
2716         set_rate_limit(s.target, rate_limit);
2717     }
2718 
2719     ret = convert_do_copy(&s);
2720 
2721     /* Now copy the bitmaps */
2722     if (bitmaps && ret == 0) {
2723         ret = convert_copy_bitmaps(blk_bs(s.src[0]), out_bs, skip_broken);
2724     }
2725 
2726 out:
2727     if (!ret) {
2728         qemu_progress_print(100, 0);
2729     }
2730     qemu_progress_end();
2731     qemu_opts_del(opts);
2732     qemu_opts_free(create_opts);
2733     qobject_unref(open_opts);
2734     blk_unref(s.target);
2735     if (s.src) {
2736         for (bs_i = 0; bs_i < s.src_num; bs_i++) {
2737             blk_unref(s.src[bs_i]);
2738         }
2739         g_free(s.src);
2740     }
2741     g_free(s.src_sectors);
2742     g_free(s.src_alignment);
2743 fail_getopt:
2744     qemu_opts_del(sn_opts);
2745     g_free(options);
2746 
2747     return !!ret;
2748 }
2749 
2750 
2751 static void dump_snapshots(BlockDriverState *bs)
2752 {
2753     QEMUSnapshotInfo *sn_tab, *sn;
2754     int nb_sns, i;
2755 
2756     nb_sns = bdrv_snapshot_list(bs, &sn_tab);
2757     if (nb_sns <= 0)
2758         return;
2759     printf("Snapshot list:\n");
2760     bdrv_snapshot_dump(NULL);
2761     printf("\n");
2762     for(i = 0; i < nb_sns; i++) {
2763         sn = &sn_tab[i];
2764         bdrv_snapshot_dump(sn);
2765         printf("\n");
2766     }
2767     g_free(sn_tab);
2768 }
2769 
2770 static void dump_json_image_info_list(ImageInfoList *list)
2771 {
2772     GString *str;
2773     QObject *obj;
2774     Visitor *v = qobject_output_visitor_new(&obj);
2775 
2776     visit_type_ImageInfoList(v, NULL, &list, &error_abort);
2777     visit_complete(v, &obj);
2778     str = qobject_to_json_pretty(obj, true);
2779     assert(str != NULL);
2780     printf("%s\n", str->str);
2781     qobject_unref(obj);
2782     visit_free(v);
2783     g_string_free(str, true);
2784 }
2785 
2786 static void dump_json_image_info(ImageInfo *info)
2787 {
2788     GString *str;
2789     QObject *obj;
2790     Visitor *v = qobject_output_visitor_new(&obj);
2791 
2792     visit_type_ImageInfo(v, NULL, &info, &error_abort);
2793     visit_complete(v, &obj);
2794     str = qobject_to_json_pretty(obj, true);
2795     assert(str != NULL);
2796     printf("%s\n", str->str);
2797     qobject_unref(obj);
2798     visit_free(v);
2799     g_string_free(str, true);
2800 }
2801 
2802 static void dump_human_image_info_list(ImageInfoList *list)
2803 {
2804     ImageInfoList *elem;
2805     bool delim = false;
2806 
2807     for (elem = list; elem; elem = elem->next) {
2808         if (delim) {
2809             printf("\n");
2810         }
2811         delim = true;
2812 
2813         bdrv_image_info_dump(elem->value);
2814     }
2815 }
2816 
2817 static gboolean str_equal_func(gconstpointer a, gconstpointer b)
2818 {
2819     return strcmp(a, b) == 0;
2820 }
2821 
2822 /**
2823  * Open an image file chain and return an ImageInfoList
2824  *
2825  * @filename: topmost image filename
2826  * @fmt: topmost image format (may be NULL to autodetect)
2827  * @chain: true  - enumerate entire backing file chain
2828  *         false - only topmost image file
2829  *
2830  * Returns a list of ImageInfo objects or NULL if there was an error opening an
2831  * image file.  If there was an error a message will have been printed to
2832  * stderr.
2833  */
2834 static ImageInfoList *collect_image_info_list(bool image_opts,
2835                                               const char *filename,
2836                                               const char *fmt,
2837                                               bool chain, bool force_share)
2838 {
2839     ImageInfoList *head = NULL;
2840     ImageInfoList **tail = &head;
2841     GHashTable *filenames;
2842     Error *err = NULL;
2843 
2844     filenames = g_hash_table_new_full(g_str_hash, str_equal_func, NULL, NULL);
2845 
2846     while (filename) {
2847         BlockBackend *blk;
2848         BlockDriverState *bs;
2849         ImageInfo *info;
2850 
2851         if (g_hash_table_lookup_extended(filenames, filename, NULL, NULL)) {
2852             error_report("Backing file '%s' creates an infinite loop.",
2853                          filename);
2854             goto err;
2855         }
2856         g_hash_table_insert(filenames, (gpointer)filename, NULL);
2857 
2858         blk = img_open(image_opts, filename, fmt,
2859                        BDRV_O_NO_BACKING | BDRV_O_NO_IO, false, false,
2860                        force_share);
2861         if (!blk) {
2862             goto err;
2863         }
2864         bs = blk_bs(blk);
2865 
2866         bdrv_query_image_info(bs, &info, &err);
2867         if (err) {
2868             error_report_err(err);
2869             blk_unref(blk);
2870             goto err;
2871         }
2872 
2873         QAPI_LIST_APPEND(tail, info);
2874 
2875         blk_unref(blk);
2876 
2877         /* Clear parameters that only apply to the topmost image */
2878         filename = fmt = NULL;
2879         image_opts = false;
2880 
2881         if (chain) {
2882             if (info->has_full_backing_filename) {
2883                 filename = info->full_backing_filename;
2884             } else if (info->has_backing_filename) {
2885                 error_report("Could not determine absolute backing filename,"
2886                              " but backing filename '%s' present",
2887                              info->backing_filename);
2888                 goto err;
2889             }
2890             if (info->has_backing_filename_format) {
2891                 fmt = info->backing_filename_format;
2892             }
2893         }
2894     }
2895     g_hash_table_destroy(filenames);
2896     return head;
2897 
2898 err:
2899     qapi_free_ImageInfoList(head);
2900     g_hash_table_destroy(filenames);
2901     return NULL;
2902 }
2903 
2904 static int img_info(int argc, char **argv)
2905 {
2906     int c;
2907     OutputFormat output_format = OFORMAT_HUMAN;
2908     bool chain = false;
2909     const char *filename, *fmt, *output;
2910     ImageInfoList *list;
2911     bool image_opts = false;
2912     bool force_share = false;
2913 
2914     fmt = NULL;
2915     output = NULL;
2916     for(;;) {
2917         int option_index = 0;
2918         static const struct option long_options[] = {
2919             {"help", no_argument, 0, 'h'},
2920             {"format", required_argument, 0, 'f'},
2921             {"output", required_argument, 0, OPTION_OUTPUT},
2922             {"backing-chain", no_argument, 0, OPTION_BACKING_CHAIN},
2923             {"object", required_argument, 0, OPTION_OBJECT},
2924             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
2925             {"force-share", no_argument, 0, 'U'},
2926             {0, 0, 0, 0}
2927         };
2928         c = getopt_long(argc, argv, ":f:hU",
2929                         long_options, &option_index);
2930         if (c == -1) {
2931             break;
2932         }
2933         switch(c) {
2934         case ':':
2935             missing_argument(argv[optind - 1]);
2936             break;
2937         case '?':
2938             unrecognized_option(argv[optind - 1]);
2939             break;
2940         case 'h':
2941             help();
2942             break;
2943         case 'f':
2944             fmt = optarg;
2945             break;
2946         case 'U':
2947             force_share = true;
2948             break;
2949         case OPTION_OUTPUT:
2950             output = optarg;
2951             break;
2952         case OPTION_BACKING_CHAIN:
2953             chain = true;
2954             break;
2955         case OPTION_OBJECT:
2956             user_creatable_process_cmdline(optarg);
2957             break;
2958         case OPTION_IMAGE_OPTS:
2959             image_opts = true;
2960             break;
2961         }
2962     }
2963     if (optind != argc - 1) {
2964         error_exit("Expecting one image file name");
2965     }
2966     filename = argv[optind++];
2967 
2968     if (output && !strcmp(output, "json")) {
2969         output_format = OFORMAT_JSON;
2970     } else if (output && !strcmp(output, "human")) {
2971         output_format = OFORMAT_HUMAN;
2972     } else if (output) {
2973         error_report("--output must be used with human or json as argument.");
2974         return 1;
2975     }
2976 
2977     list = collect_image_info_list(image_opts, filename, fmt, chain,
2978                                    force_share);
2979     if (!list) {
2980         return 1;
2981     }
2982 
2983     switch (output_format) {
2984     case OFORMAT_HUMAN:
2985         dump_human_image_info_list(list);
2986         break;
2987     case OFORMAT_JSON:
2988         if (chain) {
2989             dump_json_image_info_list(list);
2990         } else {
2991             dump_json_image_info(list->value);
2992         }
2993         break;
2994     }
2995 
2996     qapi_free_ImageInfoList(list);
2997     return 0;
2998 }
2999 
3000 static int dump_map_entry(OutputFormat output_format, MapEntry *e,
3001                           MapEntry *next)
3002 {
3003     switch (output_format) {
3004     case OFORMAT_HUMAN:
3005         if (e->data && !e->has_offset) {
3006             error_report("File contains external, encrypted or compressed clusters.");
3007             return -1;
3008         }
3009         if (e->data && !e->zero) {
3010             printf("%#-16"PRIx64"%#-16"PRIx64"%#-16"PRIx64"%s\n",
3011                    e->start, e->length,
3012                    e->has_offset ? e->offset : 0,
3013                    e->has_filename ? e->filename : "");
3014         }
3015         /* This format ignores the distinction between 0, ZERO and ZERO|DATA.
3016          * Modify the flags here to allow more coalescing.
3017          */
3018         if (next && (!next->data || next->zero)) {
3019             next->data = false;
3020             next->zero = true;
3021         }
3022         break;
3023     case OFORMAT_JSON:
3024         printf("{ \"start\": %"PRId64", \"length\": %"PRId64","
3025                " \"depth\": %"PRId64", \"present\": %s, \"zero\": %s,"
3026                " \"data\": %s", e->start, e->length, e->depth,
3027                e->present ? "true" : "false",
3028                e->zero ? "true" : "false",
3029                e->data ? "true" : "false");
3030         if (e->has_offset) {
3031             printf(", \"offset\": %"PRId64"", e->offset);
3032         }
3033         putchar('}');
3034 
3035         if (next) {
3036             puts(",");
3037         }
3038         break;
3039     }
3040     return 0;
3041 }
3042 
3043 static int get_block_status(BlockDriverState *bs, int64_t offset,
3044                             int64_t bytes, MapEntry *e)
3045 {
3046     int ret;
3047     int depth;
3048     BlockDriverState *file;
3049     bool has_offset;
3050     int64_t map;
3051     char *filename = NULL;
3052 
3053     /* As an optimization, we could cache the current range of unallocated
3054      * clusters in each file of the chain, and avoid querying the same
3055      * range repeatedly.
3056      */
3057 
3058     depth = 0;
3059     for (;;) {
3060         bs = bdrv_skip_filters(bs);
3061         ret = bdrv_block_status(bs, offset, bytes, &bytes, &map, &file);
3062         if (ret < 0) {
3063             return ret;
3064         }
3065         assert(bytes);
3066         if (ret & (BDRV_BLOCK_ZERO|BDRV_BLOCK_DATA)) {
3067             break;
3068         }
3069         bs = bdrv_cow_bs(bs);
3070         if (bs == NULL) {
3071             ret = 0;
3072             break;
3073         }
3074 
3075         depth++;
3076     }
3077 
3078     has_offset = !!(ret & BDRV_BLOCK_OFFSET_VALID);
3079 
3080     if (file && has_offset) {
3081         bdrv_refresh_filename(file);
3082         filename = file->filename;
3083     }
3084 
3085     *e = (MapEntry) {
3086         .start = offset,
3087         .length = bytes,
3088         .data = !!(ret & BDRV_BLOCK_DATA),
3089         .zero = !!(ret & BDRV_BLOCK_ZERO),
3090         .offset = map,
3091         .has_offset = has_offset,
3092         .depth = depth,
3093         .present = !!(ret & BDRV_BLOCK_ALLOCATED),
3094         .has_filename = filename,
3095         .filename = filename,
3096     };
3097 
3098     return 0;
3099 }
3100 
3101 static inline bool entry_mergeable(const MapEntry *curr, const MapEntry *next)
3102 {
3103     if (curr->length == 0) {
3104         return false;
3105     }
3106     if (curr->zero != next->zero ||
3107         curr->data != next->data ||
3108         curr->depth != next->depth ||
3109         curr->present != next->present ||
3110         curr->has_filename != next->has_filename ||
3111         curr->has_offset != next->has_offset) {
3112         return false;
3113     }
3114     if (curr->has_filename && strcmp(curr->filename, next->filename)) {
3115         return false;
3116     }
3117     if (curr->has_offset && curr->offset + curr->length != next->offset) {
3118         return false;
3119     }
3120     return true;
3121 }
3122 
3123 static int img_map(int argc, char **argv)
3124 {
3125     int c;
3126     OutputFormat output_format = OFORMAT_HUMAN;
3127     BlockBackend *blk;
3128     BlockDriverState *bs;
3129     const char *filename, *fmt, *output;
3130     int64_t length;
3131     MapEntry curr = { .length = 0 }, next;
3132     int ret = 0;
3133     bool image_opts = false;
3134     bool force_share = false;
3135     int64_t start_offset = 0;
3136     int64_t max_length = -1;
3137 
3138     fmt = NULL;
3139     output = NULL;
3140     for (;;) {
3141         int option_index = 0;
3142         static const struct option long_options[] = {
3143             {"help", no_argument, 0, 'h'},
3144             {"format", required_argument, 0, 'f'},
3145             {"output", required_argument, 0, OPTION_OUTPUT},
3146             {"object", required_argument, 0, OPTION_OBJECT},
3147             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
3148             {"force-share", no_argument, 0, 'U'},
3149             {"start-offset", required_argument, 0, 's'},
3150             {"max-length", required_argument, 0, 'l'},
3151             {0, 0, 0, 0}
3152         };
3153         c = getopt_long(argc, argv, ":f:s:l:hU",
3154                         long_options, &option_index);
3155         if (c == -1) {
3156             break;
3157         }
3158         switch (c) {
3159         case ':':
3160             missing_argument(argv[optind - 1]);
3161             break;
3162         case '?':
3163             unrecognized_option(argv[optind - 1]);
3164             break;
3165         case 'h':
3166             help();
3167             break;
3168         case 'f':
3169             fmt = optarg;
3170             break;
3171         case 'U':
3172             force_share = true;
3173             break;
3174         case OPTION_OUTPUT:
3175             output = optarg;
3176             break;
3177         case 's':
3178             start_offset = cvtnum("start offset", optarg);
3179             if (start_offset < 0) {
3180                 return 1;
3181             }
3182             break;
3183         case 'l':
3184             max_length = cvtnum("max length", optarg);
3185             if (max_length < 0) {
3186                 return 1;
3187             }
3188             break;
3189         case OPTION_OBJECT:
3190             user_creatable_process_cmdline(optarg);
3191             break;
3192         case OPTION_IMAGE_OPTS:
3193             image_opts = true;
3194             break;
3195         }
3196     }
3197     if (optind != argc - 1) {
3198         error_exit("Expecting one image file name");
3199     }
3200     filename = argv[optind];
3201 
3202     if (output && !strcmp(output, "json")) {
3203         output_format = OFORMAT_JSON;
3204     } else if (output && !strcmp(output, "human")) {
3205         output_format = OFORMAT_HUMAN;
3206     } else if (output) {
3207         error_report("--output must be used with human or json as argument.");
3208         return 1;
3209     }
3210 
3211     blk = img_open(image_opts, filename, fmt, 0, false, false, force_share);
3212     if (!blk) {
3213         return 1;
3214     }
3215     bs = blk_bs(blk);
3216 
3217     if (output_format == OFORMAT_HUMAN) {
3218         printf("%-16s%-16s%-16s%s\n", "Offset", "Length", "Mapped to", "File");
3219     } else if (output_format == OFORMAT_JSON) {
3220         putchar('[');
3221     }
3222 
3223     length = blk_getlength(blk);
3224     if (length < 0) {
3225         error_report("Failed to get size for '%s'", filename);
3226         return 1;
3227     }
3228     if (max_length != -1) {
3229         length = MIN(start_offset + max_length, length);
3230     }
3231 
3232     curr.start = start_offset;
3233     while (curr.start + curr.length < length) {
3234         int64_t offset = curr.start + curr.length;
3235         int64_t n = length - offset;
3236 
3237         ret = get_block_status(bs, offset, n, &next);
3238         if (ret < 0) {
3239             error_report("Could not read file metadata: %s", strerror(-ret));
3240             goto out;
3241         }
3242 
3243         if (entry_mergeable(&curr, &next)) {
3244             curr.length += next.length;
3245             continue;
3246         }
3247 
3248         if (curr.length > 0) {
3249             ret = dump_map_entry(output_format, &curr, &next);
3250             if (ret < 0) {
3251                 goto out;
3252             }
3253         }
3254         curr = next;
3255     }
3256 
3257     ret = dump_map_entry(output_format, &curr, NULL);
3258     if (output_format == OFORMAT_JSON) {
3259         puts("]");
3260     }
3261 
3262 out:
3263     blk_unref(blk);
3264     return ret < 0;
3265 }
3266 
3267 #define SNAPSHOT_LIST   1
3268 #define SNAPSHOT_CREATE 2
3269 #define SNAPSHOT_APPLY  3
3270 #define SNAPSHOT_DELETE 4
3271 
3272 static int img_snapshot(int argc, char **argv)
3273 {
3274     BlockBackend *blk;
3275     BlockDriverState *bs;
3276     QEMUSnapshotInfo sn;
3277     char *filename, *snapshot_name = NULL;
3278     int c, ret = 0, bdrv_oflags;
3279     int action = 0;
3280     qemu_timeval tv;
3281     bool quiet = false;
3282     Error *err = NULL;
3283     bool image_opts = false;
3284     bool force_share = false;
3285 
3286     bdrv_oflags = BDRV_O_RDWR;
3287     /* Parse commandline parameters */
3288     for(;;) {
3289         static const struct option long_options[] = {
3290             {"help", no_argument, 0, 'h'},
3291             {"object", required_argument, 0, OPTION_OBJECT},
3292             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
3293             {"force-share", no_argument, 0, 'U'},
3294             {0, 0, 0, 0}
3295         };
3296         c = getopt_long(argc, argv, ":la:c:d:hqU",
3297                         long_options, NULL);
3298         if (c == -1) {
3299             break;
3300         }
3301         switch(c) {
3302         case ':':
3303             missing_argument(argv[optind - 1]);
3304             break;
3305         case '?':
3306             unrecognized_option(argv[optind - 1]);
3307             break;
3308         case 'h':
3309             help();
3310             return 0;
3311         case 'l':
3312             if (action) {
3313                 error_exit("Cannot mix '-l', '-a', '-c', '-d'");
3314                 return 0;
3315             }
3316             action = SNAPSHOT_LIST;
3317             bdrv_oflags &= ~BDRV_O_RDWR; /* no need for RW */
3318             break;
3319         case 'a':
3320             if (action) {
3321                 error_exit("Cannot mix '-l', '-a', '-c', '-d'");
3322                 return 0;
3323             }
3324             action = SNAPSHOT_APPLY;
3325             snapshot_name = optarg;
3326             break;
3327         case 'c':
3328             if (action) {
3329                 error_exit("Cannot mix '-l', '-a', '-c', '-d'");
3330                 return 0;
3331             }
3332             action = SNAPSHOT_CREATE;
3333             snapshot_name = optarg;
3334             break;
3335         case 'd':
3336             if (action) {
3337                 error_exit("Cannot mix '-l', '-a', '-c', '-d'");
3338                 return 0;
3339             }
3340             action = SNAPSHOT_DELETE;
3341             snapshot_name = optarg;
3342             break;
3343         case 'q':
3344             quiet = true;
3345             break;
3346         case 'U':
3347             force_share = true;
3348             break;
3349         case OPTION_OBJECT:
3350             user_creatable_process_cmdline(optarg);
3351             break;
3352         case OPTION_IMAGE_OPTS:
3353             image_opts = true;
3354             break;
3355         }
3356     }
3357 
3358     if (optind != argc - 1) {
3359         error_exit("Expecting one image file name");
3360     }
3361     filename = argv[optind++];
3362 
3363     /* Open the image */
3364     blk = img_open(image_opts, filename, NULL, bdrv_oflags, false, quiet,
3365                    force_share);
3366     if (!blk) {
3367         return 1;
3368     }
3369     bs = blk_bs(blk);
3370 
3371     /* Perform the requested action */
3372     switch(action) {
3373     case SNAPSHOT_LIST:
3374         dump_snapshots(bs);
3375         break;
3376 
3377     case SNAPSHOT_CREATE:
3378         memset(&sn, 0, sizeof(sn));
3379         pstrcpy(sn.name, sizeof(sn.name), snapshot_name);
3380 
3381         qemu_gettimeofday(&tv);
3382         sn.date_sec = tv.tv_sec;
3383         sn.date_nsec = tv.tv_usec * 1000;
3384 
3385         ret = bdrv_snapshot_create(bs, &sn);
3386         if (ret) {
3387             error_report("Could not create snapshot '%s': %d (%s)",
3388                 snapshot_name, ret, strerror(-ret));
3389         }
3390         break;
3391 
3392     case SNAPSHOT_APPLY:
3393         ret = bdrv_snapshot_goto(bs, snapshot_name, &err);
3394         if (ret) {
3395             error_reportf_err(err, "Could not apply snapshot '%s': ",
3396                               snapshot_name);
3397         }
3398         break;
3399 
3400     case SNAPSHOT_DELETE:
3401         ret = bdrv_snapshot_find(bs, &sn, snapshot_name);
3402         if (ret < 0) {
3403             error_report("Could not delete snapshot '%s': snapshot not "
3404                          "found", snapshot_name);
3405             ret = 1;
3406         } else {
3407             ret = bdrv_snapshot_delete(bs, sn.id_str, sn.name, &err);
3408             if (ret < 0) {
3409                 error_reportf_err(err, "Could not delete snapshot '%s': ",
3410                                   snapshot_name);
3411                 ret = 1;
3412             }
3413         }
3414         break;
3415     }
3416 
3417     /* Cleanup */
3418     blk_unref(blk);
3419     if (ret) {
3420         return 1;
3421     }
3422     return 0;
3423 }
3424 
3425 static int img_rebase(int argc, char **argv)
3426 {
3427     BlockBackend *blk = NULL, *blk_old_backing = NULL, *blk_new_backing = NULL;
3428     uint8_t *buf_old = NULL;
3429     uint8_t *buf_new = NULL;
3430     BlockDriverState *bs = NULL, *prefix_chain_bs = NULL;
3431     BlockDriverState *unfiltered_bs;
3432     char *filename;
3433     const char *fmt, *cache, *src_cache, *out_basefmt, *out_baseimg;
3434     int c, flags, src_flags, ret;
3435     bool writethrough, src_writethrough;
3436     int unsafe = 0;
3437     bool force_share = false;
3438     int progress = 0;
3439     bool quiet = false;
3440     Error *local_err = NULL;
3441     bool image_opts = false;
3442 
3443     /* Parse commandline parameters */
3444     fmt = NULL;
3445     cache = BDRV_DEFAULT_CACHE;
3446     src_cache = BDRV_DEFAULT_CACHE;
3447     out_baseimg = NULL;
3448     out_basefmt = NULL;
3449     for(;;) {
3450         static const struct option long_options[] = {
3451             {"help", no_argument, 0, 'h'},
3452             {"object", required_argument, 0, OPTION_OBJECT},
3453             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
3454             {"force-share", no_argument, 0, 'U'},
3455             {0, 0, 0, 0}
3456         };
3457         c = getopt_long(argc, argv, ":hf:F:b:upt:T:qU",
3458                         long_options, NULL);
3459         if (c == -1) {
3460             break;
3461         }
3462         switch(c) {
3463         case ':':
3464             missing_argument(argv[optind - 1]);
3465             break;
3466         case '?':
3467             unrecognized_option(argv[optind - 1]);
3468             break;
3469         case 'h':
3470             help();
3471             return 0;
3472         case 'f':
3473             fmt = optarg;
3474             break;
3475         case 'F':
3476             out_basefmt = optarg;
3477             break;
3478         case 'b':
3479             out_baseimg = optarg;
3480             break;
3481         case 'u':
3482             unsafe = 1;
3483             break;
3484         case 'p':
3485             progress = 1;
3486             break;
3487         case 't':
3488             cache = optarg;
3489             break;
3490         case 'T':
3491             src_cache = optarg;
3492             break;
3493         case 'q':
3494             quiet = true;
3495             break;
3496         case OPTION_OBJECT:
3497             user_creatable_process_cmdline(optarg);
3498             break;
3499         case OPTION_IMAGE_OPTS:
3500             image_opts = true;
3501             break;
3502         case 'U':
3503             force_share = true;
3504             break;
3505         }
3506     }
3507 
3508     if (quiet) {
3509         progress = 0;
3510     }
3511 
3512     if (optind != argc - 1) {
3513         error_exit("Expecting one image file name");
3514     }
3515     if (!unsafe && !out_baseimg) {
3516         error_exit("Must specify backing file (-b) or use unsafe mode (-u)");
3517     }
3518     filename = argv[optind++];
3519 
3520     qemu_progress_init(progress, 2.0);
3521     qemu_progress_print(0, 100);
3522 
3523     flags = BDRV_O_RDWR | (unsafe ? BDRV_O_NO_BACKING : 0);
3524     ret = bdrv_parse_cache_mode(cache, &flags, &writethrough);
3525     if (ret < 0) {
3526         error_report("Invalid cache option: %s", cache);
3527         goto out;
3528     }
3529 
3530     src_flags = 0;
3531     ret = bdrv_parse_cache_mode(src_cache, &src_flags, &src_writethrough);
3532     if (ret < 0) {
3533         error_report("Invalid source cache option: %s", src_cache);
3534         goto out;
3535     }
3536 
3537     /* The source files are opened read-only, don't care about WCE */
3538     assert((src_flags & BDRV_O_RDWR) == 0);
3539     (void) src_writethrough;
3540 
3541     /*
3542      * Open the images.
3543      *
3544      * Ignore the old backing file for unsafe rebase in case we want to correct
3545      * the reference to a renamed or moved backing file.
3546      */
3547     blk = img_open(image_opts, filename, fmt, flags, writethrough, quiet,
3548                    false);
3549     if (!blk) {
3550         ret = -1;
3551         goto out;
3552     }
3553     bs = blk_bs(blk);
3554 
3555     unfiltered_bs = bdrv_skip_filters(bs);
3556 
3557     if (out_basefmt != NULL) {
3558         if (bdrv_find_format(out_basefmt) == NULL) {
3559             error_report("Invalid format name: '%s'", out_basefmt);
3560             ret = -1;
3561             goto out;
3562         }
3563     }
3564 
3565     /* For safe rebasing we need to compare old and new backing file */
3566     if (!unsafe) {
3567         QDict *options = NULL;
3568         BlockDriverState *base_bs = bdrv_cow_bs(unfiltered_bs);
3569 
3570         if (base_bs) {
3571             blk_old_backing = blk_new(qemu_get_aio_context(),
3572                                       BLK_PERM_CONSISTENT_READ,
3573                                       BLK_PERM_ALL);
3574             ret = blk_insert_bs(blk_old_backing, base_bs,
3575                                 &local_err);
3576             if (ret < 0) {
3577                 error_reportf_err(local_err,
3578                                   "Could not reuse old backing file '%s': ",
3579                                   base_bs->filename);
3580                 goto out;
3581             }
3582         } else {
3583             blk_old_backing = NULL;
3584         }
3585 
3586         if (out_baseimg[0]) {
3587             const char *overlay_filename;
3588             char *out_real_path;
3589 
3590             options = qdict_new();
3591             if (out_basefmt) {
3592                 qdict_put_str(options, "driver", out_basefmt);
3593             }
3594             if (force_share) {
3595                 qdict_put_bool(options, BDRV_OPT_FORCE_SHARE, true);
3596             }
3597 
3598             bdrv_refresh_filename(bs);
3599             overlay_filename = bs->exact_filename[0] ? bs->exact_filename
3600                                                      : bs->filename;
3601             out_real_path =
3602                 bdrv_get_full_backing_filename_from_filename(overlay_filename,
3603                                                              out_baseimg,
3604                                                              &local_err);
3605             if (local_err) {
3606                 qobject_unref(options);
3607                 error_reportf_err(local_err,
3608                                   "Could not resolve backing filename: ");
3609                 ret = -1;
3610                 goto out;
3611             }
3612 
3613             /*
3614              * Find out whether we rebase an image on top of a previous image
3615              * in its chain.
3616              */
3617             prefix_chain_bs = bdrv_find_backing_image(bs, out_real_path);
3618             if (prefix_chain_bs) {
3619                 qobject_unref(options);
3620                 g_free(out_real_path);
3621 
3622                 blk_new_backing = blk_new(qemu_get_aio_context(),
3623                                           BLK_PERM_CONSISTENT_READ,
3624                                           BLK_PERM_ALL);
3625                 ret = blk_insert_bs(blk_new_backing, prefix_chain_bs,
3626                                     &local_err);
3627                 if (ret < 0) {
3628                     error_reportf_err(local_err,
3629                                       "Could not reuse backing file '%s': ",
3630                                       out_baseimg);
3631                     goto out;
3632                 }
3633             } else {
3634                 blk_new_backing = blk_new_open(out_real_path, NULL,
3635                                                options, src_flags, &local_err);
3636                 g_free(out_real_path);
3637                 if (!blk_new_backing) {
3638                     error_reportf_err(local_err,
3639                                       "Could not open new backing file '%s': ",
3640                                       out_baseimg);
3641                     ret = -1;
3642                     goto out;
3643                 }
3644             }
3645         }
3646     }
3647 
3648     /*
3649      * Check each unallocated cluster in the COW file. If it is unallocated,
3650      * accesses go to the backing file. We must therefore compare this cluster
3651      * in the old and new backing file, and if they differ we need to copy it
3652      * from the old backing file into the COW file.
3653      *
3654      * If qemu-img crashes during this step, no harm is done. The content of
3655      * the image is the same as the original one at any time.
3656      */
3657     if (!unsafe) {
3658         int64_t size;
3659         int64_t old_backing_size = 0;
3660         int64_t new_backing_size = 0;
3661         uint64_t offset;
3662         int64_t n;
3663         float local_progress = 0;
3664 
3665         buf_old = blk_blockalign(blk, IO_BUF_SIZE);
3666         buf_new = blk_blockalign(blk, IO_BUF_SIZE);
3667 
3668         size = blk_getlength(blk);
3669         if (size < 0) {
3670             error_report("Could not get size of '%s': %s",
3671                          filename, strerror(-size));
3672             ret = -1;
3673             goto out;
3674         }
3675         if (blk_old_backing) {
3676             old_backing_size = blk_getlength(blk_old_backing);
3677             if (old_backing_size < 0) {
3678                 char backing_name[PATH_MAX];
3679 
3680                 bdrv_get_backing_filename(bs, backing_name,
3681                                           sizeof(backing_name));
3682                 error_report("Could not get size of '%s': %s",
3683                              backing_name, strerror(-old_backing_size));
3684                 ret = -1;
3685                 goto out;
3686             }
3687         }
3688         if (blk_new_backing) {
3689             new_backing_size = blk_getlength(blk_new_backing);
3690             if (new_backing_size < 0) {
3691                 error_report("Could not get size of '%s': %s",
3692                              out_baseimg, strerror(-new_backing_size));
3693                 ret = -1;
3694                 goto out;
3695             }
3696         }
3697 
3698         if (size != 0) {
3699             local_progress = (float)100 / (size / MIN(size, IO_BUF_SIZE));
3700         }
3701 
3702         for (offset = 0; offset < size; offset += n) {
3703             bool buf_old_is_zero = false;
3704 
3705             /* How many bytes can we handle with the next read? */
3706             n = MIN(IO_BUF_SIZE, size - offset);
3707 
3708             /* If the cluster is allocated, we don't need to take action */
3709             ret = bdrv_is_allocated(unfiltered_bs, offset, n, &n);
3710             if (ret < 0) {
3711                 error_report("error while reading image metadata: %s",
3712                              strerror(-ret));
3713                 goto out;
3714             }
3715             if (ret) {
3716                 continue;
3717             }
3718 
3719             if (prefix_chain_bs) {
3720                 /*
3721                  * If cluster wasn't changed since prefix_chain, we don't need
3722                  * to take action
3723                  */
3724                 ret = bdrv_is_allocated_above(bdrv_cow_bs(unfiltered_bs),
3725                                               prefix_chain_bs, false,
3726                                               offset, n, &n);
3727                 if (ret < 0) {
3728                     error_report("error while reading image metadata: %s",
3729                                  strerror(-ret));
3730                     goto out;
3731                 }
3732                 if (!ret) {
3733                     continue;
3734                 }
3735             }
3736 
3737             /*
3738              * Read old and new backing file and take into consideration that
3739              * backing files may be smaller than the COW image.
3740              */
3741             if (offset >= old_backing_size) {
3742                 memset(buf_old, 0, n);
3743                 buf_old_is_zero = true;
3744             } else {
3745                 if (offset + n > old_backing_size) {
3746                     n = old_backing_size - offset;
3747                 }
3748 
3749                 ret = blk_pread(blk_old_backing, offset, buf_old, n);
3750                 if (ret < 0) {
3751                     error_report("error while reading from old backing file");
3752                     goto out;
3753                 }
3754             }
3755 
3756             if (offset >= new_backing_size || !blk_new_backing) {
3757                 memset(buf_new, 0, n);
3758             } else {
3759                 if (offset + n > new_backing_size) {
3760                     n = new_backing_size - offset;
3761                 }
3762 
3763                 ret = blk_pread(blk_new_backing, offset, buf_new, n);
3764                 if (ret < 0) {
3765                     error_report("error while reading from new backing file");
3766                     goto out;
3767                 }
3768             }
3769 
3770             /* If they differ, we need to write to the COW file */
3771             uint64_t written = 0;
3772 
3773             while (written < n) {
3774                 int64_t pnum;
3775 
3776                 if (compare_buffers(buf_old + written, buf_new + written,
3777                                     n - written, &pnum))
3778                 {
3779                     if (buf_old_is_zero) {
3780                         ret = blk_pwrite_zeroes(blk, offset + written, pnum, 0);
3781                     } else {
3782                         ret = blk_pwrite(blk, offset + written,
3783                                          buf_old + written, pnum, 0);
3784                     }
3785                     if (ret < 0) {
3786                         error_report("Error while writing to COW image: %s",
3787                             strerror(-ret));
3788                         goto out;
3789                     }
3790                 }
3791 
3792                 written += pnum;
3793             }
3794             qemu_progress_print(local_progress, 100);
3795         }
3796     }
3797 
3798     /*
3799      * Change the backing file. All clusters that are different from the old
3800      * backing file are overwritten in the COW file now, so the visible content
3801      * doesn't change when we switch the backing file.
3802      */
3803     if (out_baseimg && *out_baseimg) {
3804         ret = bdrv_change_backing_file(unfiltered_bs, out_baseimg, out_basefmt,
3805                                        true);
3806     } else {
3807         ret = bdrv_change_backing_file(unfiltered_bs, NULL, NULL, false);
3808     }
3809 
3810     if (ret == -ENOSPC) {
3811         error_report("Could not change the backing file to '%s': No "
3812                      "space left in the file header", out_baseimg);
3813     } else if (ret == -EINVAL && out_baseimg && !out_basefmt) {
3814         error_report("Could not change the backing file to '%s': backing "
3815                      "format must be specified", out_baseimg);
3816     } else if (ret < 0) {
3817         error_report("Could not change the backing file to '%s': %s",
3818             out_baseimg, strerror(-ret));
3819     }
3820 
3821     qemu_progress_print(100, 0);
3822     /*
3823      * TODO At this point it is possible to check if any clusters that are
3824      * allocated in the COW file are the same in the backing file. If so, they
3825      * could be dropped from the COW file. Don't do this before switching the
3826      * backing file, in case of a crash this would lead to corruption.
3827      */
3828 out:
3829     qemu_progress_end();
3830     /* Cleanup */
3831     if (!unsafe) {
3832         blk_unref(blk_old_backing);
3833         blk_unref(blk_new_backing);
3834     }
3835     qemu_vfree(buf_old);
3836     qemu_vfree(buf_new);
3837 
3838     blk_unref(blk);
3839     if (ret) {
3840         return 1;
3841     }
3842     return 0;
3843 }
3844 
3845 static int img_resize(int argc, char **argv)
3846 {
3847     Error *err = NULL;
3848     int c, ret, relative;
3849     const char *filename, *fmt, *size;
3850     int64_t n, total_size, current_size;
3851     bool quiet = false;
3852     BlockBackend *blk = NULL;
3853     PreallocMode prealloc = PREALLOC_MODE_OFF;
3854     QemuOpts *param;
3855 
3856     static QemuOptsList resize_options = {
3857         .name = "resize_options",
3858         .head = QTAILQ_HEAD_INITIALIZER(resize_options.head),
3859         .desc = {
3860             {
3861                 .name = BLOCK_OPT_SIZE,
3862                 .type = QEMU_OPT_SIZE,
3863                 .help = "Virtual disk size"
3864             }, {
3865                 /* end of list */
3866             }
3867         },
3868     };
3869     bool image_opts = false;
3870     bool shrink = false;
3871 
3872     /* Remove size from argv manually so that negative numbers are not treated
3873      * as options by getopt. */
3874     if (argc < 3) {
3875         error_exit("Not enough arguments");
3876         return 1;
3877     }
3878 
3879     size = argv[--argc];
3880 
3881     /* Parse getopt arguments */
3882     fmt = NULL;
3883     for(;;) {
3884         static const struct option long_options[] = {
3885             {"help", no_argument, 0, 'h'},
3886             {"object", required_argument, 0, OPTION_OBJECT},
3887             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
3888             {"preallocation", required_argument, 0, OPTION_PREALLOCATION},
3889             {"shrink", no_argument, 0, OPTION_SHRINK},
3890             {0, 0, 0, 0}
3891         };
3892         c = getopt_long(argc, argv, ":f:hq",
3893                         long_options, NULL);
3894         if (c == -1) {
3895             break;
3896         }
3897         switch(c) {
3898         case ':':
3899             missing_argument(argv[optind - 1]);
3900             break;
3901         case '?':
3902             unrecognized_option(argv[optind - 1]);
3903             break;
3904         case 'h':
3905             help();
3906             break;
3907         case 'f':
3908             fmt = optarg;
3909             break;
3910         case 'q':
3911             quiet = true;
3912             break;
3913         case OPTION_OBJECT:
3914             user_creatable_process_cmdline(optarg);
3915             break;
3916         case OPTION_IMAGE_OPTS:
3917             image_opts = true;
3918             break;
3919         case OPTION_PREALLOCATION:
3920             prealloc = qapi_enum_parse(&PreallocMode_lookup, optarg,
3921                                        PREALLOC_MODE__MAX, NULL);
3922             if (prealloc == PREALLOC_MODE__MAX) {
3923                 error_report("Invalid preallocation mode '%s'", optarg);
3924                 return 1;
3925             }
3926             break;
3927         case OPTION_SHRINK:
3928             shrink = true;
3929             break;
3930         }
3931     }
3932     if (optind != argc - 1) {
3933         error_exit("Expecting image file name and size");
3934     }
3935     filename = argv[optind++];
3936 
3937     /* Choose grow, shrink, or absolute resize mode */
3938     switch (size[0]) {
3939     case '+':
3940         relative = 1;
3941         size++;
3942         break;
3943     case '-':
3944         relative = -1;
3945         size++;
3946         break;
3947     default:
3948         relative = 0;
3949         break;
3950     }
3951 
3952     /* Parse size */
3953     param = qemu_opts_create(&resize_options, NULL, 0, &error_abort);
3954     if (!qemu_opt_set(param, BLOCK_OPT_SIZE, size, &err)) {
3955         error_report_err(err);
3956         ret = -1;
3957         qemu_opts_del(param);
3958         goto out;
3959     }
3960     n = qemu_opt_get_size(param, BLOCK_OPT_SIZE, 0);
3961     qemu_opts_del(param);
3962 
3963     blk = img_open(image_opts, filename, fmt,
3964                    BDRV_O_RDWR | BDRV_O_RESIZE, false, quiet,
3965                    false);
3966     if (!blk) {
3967         ret = -1;
3968         goto out;
3969     }
3970 
3971     current_size = blk_getlength(blk);
3972     if (current_size < 0) {
3973         error_report("Failed to inquire current image length: %s",
3974                      strerror(-current_size));
3975         ret = -1;
3976         goto out;
3977     }
3978 
3979     if (relative) {
3980         total_size = current_size + n * relative;
3981     } else {
3982         total_size = n;
3983     }
3984     if (total_size <= 0) {
3985         error_report("New image size must be positive");
3986         ret = -1;
3987         goto out;
3988     }
3989 
3990     if (total_size <= current_size && prealloc != PREALLOC_MODE_OFF) {
3991         error_report("Preallocation can only be used for growing images");
3992         ret = -1;
3993         goto out;
3994     }
3995 
3996     if (total_size < current_size && !shrink) {
3997         error_report("Use the --shrink option to perform a shrink operation.");
3998         warn_report("Shrinking an image will delete all data beyond the "
3999                     "shrunken image's end. Before performing such an "
4000                     "operation, make sure there is no important data there.");
4001         ret = -1;
4002         goto out;
4003     }
4004 
4005     /*
4006      * The user expects the image to have the desired size after
4007      * resizing, so pass @exact=true.  It is of no use to report
4008      * success when the image has not actually been resized.
4009      */
4010     ret = blk_truncate(blk, total_size, true, prealloc, 0, &err);
4011     if (!ret) {
4012         qprintf(quiet, "Image resized.\n");
4013     } else {
4014         error_report_err(err);
4015     }
4016 out:
4017     blk_unref(blk);
4018     if (ret) {
4019         return 1;
4020     }
4021     return 0;
4022 }
4023 
4024 static void amend_status_cb(BlockDriverState *bs,
4025                             int64_t offset, int64_t total_work_size,
4026                             void *opaque)
4027 {
4028     qemu_progress_print(100.f * offset / total_work_size, 0);
4029 }
4030 
4031 static int print_amend_option_help(const char *format)
4032 {
4033     BlockDriver *drv;
4034 
4035     /* Find driver and parse its options */
4036     drv = bdrv_find_format(format);
4037     if (!drv) {
4038         error_report("Unknown file format '%s'", format);
4039         return 1;
4040     }
4041 
4042     if (!drv->bdrv_amend_options) {
4043         error_report("Format driver '%s' does not support option amendment",
4044                      format);
4045         return 1;
4046     }
4047 
4048     /* Every driver supporting amendment must have amend_opts */
4049     assert(drv->amend_opts);
4050 
4051     printf("Amend options for '%s':\n", format);
4052     qemu_opts_print_help(drv->amend_opts, false);
4053     return 0;
4054 }
4055 
4056 static int img_amend(int argc, char **argv)
4057 {
4058     Error *err = NULL;
4059     int c, ret = 0;
4060     char *options = NULL;
4061     QemuOptsList *amend_opts = NULL;
4062     QemuOpts *opts = NULL;
4063     const char *fmt = NULL, *filename, *cache;
4064     int flags;
4065     bool writethrough;
4066     bool quiet = false, progress = false;
4067     BlockBackend *blk = NULL;
4068     BlockDriverState *bs = NULL;
4069     bool image_opts = false;
4070     bool force = false;
4071 
4072     cache = BDRV_DEFAULT_CACHE;
4073     for (;;) {
4074         static const struct option long_options[] = {
4075             {"help", no_argument, 0, 'h'},
4076             {"object", required_argument, 0, OPTION_OBJECT},
4077             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
4078             {"force", no_argument, 0, OPTION_FORCE},
4079             {0, 0, 0, 0}
4080         };
4081         c = getopt_long(argc, argv, ":ho:f:t:pq",
4082                         long_options, NULL);
4083         if (c == -1) {
4084             break;
4085         }
4086 
4087         switch (c) {
4088         case ':':
4089             missing_argument(argv[optind - 1]);
4090             break;
4091         case '?':
4092             unrecognized_option(argv[optind - 1]);
4093             break;
4094         case 'h':
4095             help();
4096             break;
4097         case 'o':
4098             if (accumulate_options(&options, optarg) < 0) {
4099                 ret = -1;
4100                 goto out_no_progress;
4101             }
4102             break;
4103         case 'f':
4104             fmt = optarg;
4105             break;
4106         case 't':
4107             cache = optarg;
4108             break;
4109         case 'p':
4110             progress = true;
4111             break;
4112         case 'q':
4113             quiet = true;
4114             break;
4115         case OPTION_OBJECT:
4116             user_creatable_process_cmdline(optarg);
4117             break;
4118         case OPTION_IMAGE_OPTS:
4119             image_opts = true;
4120             break;
4121         case OPTION_FORCE:
4122             force = true;
4123             break;
4124         }
4125     }
4126 
4127     if (!options) {
4128         error_exit("Must specify options (-o)");
4129     }
4130 
4131     if (quiet) {
4132         progress = false;
4133     }
4134     qemu_progress_init(progress, 1.0);
4135 
4136     filename = (optind == argc - 1) ? argv[argc - 1] : NULL;
4137     if (fmt && has_help_option(options)) {
4138         /* If a format is explicitly specified (and possibly no filename is
4139          * given), print option help here */
4140         ret = print_amend_option_help(fmt);
4141         goto out;
4142     }
4143 
4144     if (optind != argc - 1) {
4145         error_report("Expecting one image file name");
4146         ret = -1;
4147         goto out;
4148     }
4149 
4150     flags = BDRV_O_RDWR;
4151     ret = bdrv_parse_cache_mode(cache, &flags, &writethrough);
4152     if (ret < 0) {
4153         error_report("Invalid cache option: %s", cache);
4154         goto out;
4155     }
4156 
4157     blk = img_open(image_opts, filename, fmt, flags, writethrough, quiet,
4158                    false);
4159     if (!blk) {
4160         ret = -1;
4161         goto out;
4162     }
4163     bs = blk_bs(blk);
4164 
4165     fmt = bs->drv->format_name;
4166 
4167     if (has_help_option(options)) {
4168         /* If the format was auto-detected, print option help here */
4169         ret = print_amend_option_help(fmt);
4170         goto out;
4171     }
4172 
4173     if (!bs->drv->bdrv_amend_options) {
4174         error_report("Format driver '%s' does not support option amendment",
4175                      fmt);
4176         ret = -1;
4177         goto out;
4178     }
4179 
4180     /* Every driver supporting amendment must have amend_opts */
4181     assert(bs->drv->amend_opts);
4182 
4183     amend_opts = qemu_opts_append(amend_opts, bs->drv->amend_opts);
4184     opts = qemu_opts_create(amend_opts, NULL, 0, &error_abort);
4185     if (!qemu_opts_do_parse(opts, options, NULL, &err)) {
4186         /* Try to parse options using the create options */
4187         amend_opts = qemu_opts_append(amend_opts, bs->drv->create_opts);
4188         qemu_opts_del(opts);
4189         opts = qemu_opts_create(amend_opts, NULL, 0, &error_abort);
4190         if (qemu_opts_do_parse(opts, options, NULL, NULL)) {
4191             error_append_hint(&err,
4192                               "This option is only supported for image creation\n");
4193         }
4194 
4195         error_report_err(err);
4196         ret = -1;
4197         goto out;
4198     }
4199 
4200     /* In case the driver does not call amend_status_cb() */
4201     qemu_progress_print(0.f, 0);
4202     ret = bdrv_amend_options(bs, opts, &amend_status_cb, NULL, force, &err);
4203     qemu_progress_print(100.f, 0);
4204     if (ret < 0) {
4205         error_report_err(err);
4206         goto out;
4207     }
4208 
4209 out:
4210     qemu_progress_end();
4211 
4212 out_no_progress:
4213     blk_unref(blk);
4214     qemu_opts_del(opts);
4215     qemu_opts_free(amend_opts);
4216     g_free(options);
4217 
4218     if (ret) {
4219         return 1;
4220     }
4221     return 0;
4222 }
4223 
4224 typedef struct BenchData {
4225     BlockBackend *blk;
4226     uint64_t image_size;
4227     bool write;
4228     int bufsize;
4229     int step;
4230     int nrreq;
4231     int n;
4232     int flush_interval;
4233     bool drain_on_flush;
4234     uint8_t *buf;
4235     QEMUIOVector *qiov;
4236 
4237     int in_flight;
4238     bool in_flush;
4239     uint64_t offset;
4240 } BenchData;
4241 
4242 static void bench_undrained_flush_cb(void *opaque, int ret)
4243 {
4244     if (ret < 0) {
4245         error_report("Failed flush request: %s", strerror(-ret));
4246         exit(EXIT_FAILURE);
4247     }
4248 }
4249 
4250 static void bench_cb(void *opaque, int ret)
4251 {
4252     BenchData *b = opaque;
4253     BlockAIOCB *acb;
4254 
4255     if (ret < 0) {
4256         error_report("Failed request: %s", strerror(-ret));
4257         exit(EXIT_FAILURE);
4258     }
4259 
4260     if (b->in_flush) {
4261         /* Just finished a flush with drained queue: Start next requests */
4262         assert(b->in_flight == 0);
4263         b->in_flush = false;
4264     } else if (b->in_flight > 0) {
4265         int remaining = b->n - b->in_flight;
4266 
4267         b->n--;
4268         b->in_flight--;
4269 
4270         /* Time for flush? Drain queue if requested, then flush */
4271         if (b->flush_interval && remaining % b->flush_interval == 0) {
4272             if (!b->in_flight || !b->drain_on_flush) {
4273                 BlockCompletionFunc *cb;
4274 
4275                 if (b->drain_on_flush) {
4276                     b->in_flush = true;
4277                     cb = bench_cb;
4278                 } else {
4279                     cb = bench_undrained_flush_cb;
4280                 }
4281 
4282                 acb = blk_aio_flush(b->blk, cb, b);
4283                 if (!acb) {
4284                     error_report("Failed to issue flush request");
4285                     exit(EXIT_FAILURE);
4286                 }
4287             }
4288             if (b->drain_on_flush) {
4289                 return;
4290             }
4291         }
4292     }
4293 
4294     while (b->n > b->in_flight && b->in_flight < b->nrreq) {
4295         int64_t offset = b->offset;
4296         /* blk_aio_* might look for completed I/Os and kick bench_cb
4297          * again, so make sure this operation is counted by in_flight
4298          * and b->offset is ready for the next submission.
4299          */
4300         b->in_flight++;
4301         b->offset += b->step;
4302         b->offset %= b->image_size;
4303         if (b->write) {
4304             acb = blk_aio_pwritev(b->blk, offset, b->qiov, 0, bench_cb, b);
4305         } else {
4306             acb = blk_aio_preadv(b->blk, offset, b->qiov, 0, bench_cb, b);
4307         }
4308         if (!acb) {
4309             error_report("Failed to issue request");
4310             exit(EXIT_FAILURE);
4311         }
4312     }
4313 }
4314 
4315 static int img_bench(int argc, char **argv)
4316 {
4317     int c, ret = 0;
4318     const char *fmt = NULL, *filename;
4319     bool quiet = false;
4320     bool image_opts = false;
4321     bool is_write = false;
4322     int count = 75000;
4323     int depth = 64;
4324     int64_t offset = 0;
4325     size_t bufsize = 4096;
4326     int pattern = 0;
4327     size_t step = 0;
4328     int flush_interval = 0;
4329     bool drain_on_flush = true;
4330     int64_t image_size;
4331     BlockBackend *blk = NULL;
4332     BenchData data = {};
4333     int flags = 0;
4334     bool writethrough = false;
4335     struct timeval t1, t2;
4336     int i;
4337     bool force_share = false;
4338     size_t buf_size;
4339 
4340     for (;;) {
4341         static const struct option long_options[] = {
4342             {"help", no_argument, 0, 'h'},
4343             {"flush-interval", required_argument, 0, OPTION_FLUSH_INTERVAL},
4344             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
4345             {"pattern", required_argument, 0, OPTION_PATTERN},
4346             {"no-drain", no_argument, 0, OPTION_NO_DRAIN},
4347             {"force-share", no_argument, 0, 'U'},
4348             {0, 0, 0, 0}
4349         };
4350         c = getopt_long(argc, argv, ":hc:d:f:ni:o:qs:S:t:wU", long_options,
4351                         NULL);
4352         if (c == -1) {
4353             break;
4354         }
4355 
4356         switch (c) {
4357         case ':':
4358             missing_argument(argv[optind - 1]);
4359             break;
4360         case '?':
4361             unrecognized_option(argv[optind - 1]);
4362             break;
4363         case 'h':
4364             help();
4365             break;
4366         case 'c':
4367         {
4368             unsigned long res;
4369 
4370             if (qemu_strtoul(optarg, NULL, 0, &res) < 0 || res > INT_MAX) {
4371                 error_report("Invalid request count specified");
4372                 return 1;
4373             }
4374             count = res;
4375             break;
4376         }
4377         case 'd':
4378         {
4379             unsigned long res;
4380 
4381             if (qemu_strtoul(optarg, NULL, 0, &res) < 0 || res > INT_MAX) {
4382                 error_report("Invalid queue depth specified");
4383                 return 1;
4384             }
4385             depth = res;
4386             break;
4387         }
4388         case 'f':
4389             fmt = optarg;
4390             break;
4391         case 'n':
4392             flags |= BDRV_O_NATIVE_AIO;
4393             break;
4394         case 'i':
4395             ret = bdrv_parse_aio(optarg, &flags);
4396             if (ret < 0) {
4397                 error_report("Invalid aio option: %s", optarg);
4398                 ret = -1;
4399                 goto out;
4400             }
4401             break;
4402         case 'o':
4403         {
4404             offset = cvtnum("offset", optarg);
4405             if (offset < 0) {
4406                 return 1;
4407             }
4408             break;
4409         }
4410             break;
4411         case 'q':
4412             quiet = true;
4413             break;
4414         case 's':
4415         {
4416             int64_t sval;
4417 
4418             sval = cvtnum_full("buffer size", optarg, 0, INT_MAX);
4419             if (sval < 0) {
4420                 return 1;
4421             }
4422 
4423             bufsize = sval;
4424             break;
4425         }
4426         case 'S':
4427         {
4428             int64_t sval;
4429 
4430             sval = cvtnum_full("step_size", optarg, 0, INT_MAX);
4431             if (sval < 0) {
4432                 return 1;
4433             }
4434 
4435             step = sval;
4436             break;
4437         }
4438         case 't':
4439             ret = bdrv_parse_cache_mode(optarg, &flags, &writethrough);
4440             if (ret < 0) {
4441                 error_report("Invalid cache mode");
4442                 ret = -1;
4443                 goto out;
4444             }
4445             break;
4446         case 'w':
4447             flags |= BDRV_O_RDWR;
4448             is_write = true;
4449             break;
4450         case 'U':
4451             force_share = true;
4452             break;
4453         case OPTION_PATTERN:
4454         {
4455             unsigned long res;
4456 
4457             if (qemu_strtoul(optarg, NULL, 0, &res) < 0 || res > 0xff) {
4458                 error_report("Invalid pattern byte specified");
4459                 return 1;
4460             }
4461             pattern = res;
4462             break;
4463         }
4464         case OPTION_FLUSH_INTERVAL:
4465         {
4466             unsigned long res;
4467 
4468             if (qemu_strtoul(optarg, NULL, 0, &res) < 0 || res > INT_MAX) {
4469                 error_report("Invalid flush interval specified");
4470                 return 1;
4471             }
4472             flush_interval = res;
4473             break;
4474         }
4475         case OPTION_NO_DRAIN:
4476             drain_on_flush = false;
4477             break;
4478         case OPTION_IMAGE_OPTS:
4479             image_opts = true;
4480             break;
4481         }
4482     }
4483 
4484     if (optind != argc - 1) {
4485         error_exit("Expecting one image file name");
4486     }
4487     filename = argv[argc - 1];
4488 
4489     if (!is_write && flush_interval) {
4490         error_report("--flush-interval is only available in write tests");
4491         ret = -1;
4492         goto out;
4493     }
4494     if (flush_interval && flush_interval < depth) {
4495         error_report("Flush interval can't be smaller than depth");
4496         ret = -1;
4497         goto out;
4498     }
4499 
4500     blk = img_open(image_opts, filename, fmt, flags, writethrough, quiet,
4501                    force_share);
4502     if (!blk) {
4503         ret = -1;
4504         goto out;
4505     }
4506 
4507     image_size = blk_getlength(blk);
4508     if (image_size < 0) {
4509         ret = image_size;
4510         goto out;
4511     }
4512 
4513     data = (BenchData) {
4514         .blk            = blk,
4515         .image_size     = image_size,
4516         .bufsize        = bufsize,
4517         .step           = step ?: bufsize,
4518         .nrreq          = depth,
4519         .n              = count,
4520         .offset         = offset,
4521         .write          = is_write,
4522         .flush_interval = flush_interval,
4523         .drain_on_flush = drain_on_flush,
4524     };
4525     printf("Sending %d %s requests, %d bytes each, %d in parallel "
4526            "(starting at offset %" PRId64 ", step size %d)\n",
4527            data.n, data.write ? "write" : "read", data.bufsize, data.nrreq,
4528            data.offset, data.step);
4529     if (flush_interval) {
4530         printf("Sending flush every %d requests\n", flush_interval);
4531     }
4532 
4533     buf_size = data.nrreq * data.bufsize;
4534     data.buf = blk_blockalign(blk, buf_size);
4535     memset(data.buf, pattern, data.nrreq * data.bufsize);
4536 
4537     blk_register_buf(blk, data.buf, buf_size);
4538 
4539     data.qiov = g_new(QEMUIOVector, data.nrreq);
4540     for (i = 0; i < data.nrreq; i++) {
4541         qemu_iovec_init(&data.qiov[i], 1);
4542         qemu_iovec_add(&data.qiov[i],
4543                        data.buf + i * data.bufsize, data.bufsize);
4544     }
4545 
4546     gettimeofday(&t1, NULL);
4547     bench_cb(&data, 0);
4548 
4549     while (data.n > 0) {
4550         main_loop_wait(false);
4551     }
4552     gettimeofday(&t2, NULL);
4553 
4554     printf("Run completed in %3.3f seconds.\n",
4555            (t2.tv_sec - t1.tv_sec)
4556            + ((double)(t2.tv_usec - t1.tv_usec) / 1000000));
4557 
4558 out:
4559     if (data.buf) {
4560         blk_unregister_buf(blk, data.buf);
4561     }
4562     qemu_vfree(data.buf);
4563     blk_unref(blk);
4564 
4565     if (ret) {
4566         return 1;
4567     }
4568     return 0;
4569 }
4570 
4571 enum ImgBitmapAct {
4572     BITMAP_ADD,
4573     BITMAP_REMOVE,
4574     BITMAP_CLEAR,
4575     BITMAP_ENABLE,
4576     BITMAP_DISABLE,
4577     BITMAP_MERGE,
4578 };
4579 typedef struct ImgBitmapAction {
4580     enum ImgBitmapAct act;
4581     const char *src; /* only used for merge */
4582     QSIMPLEQ_ENTRY(ImgBitmapAction) next;
4583 } ImgBitmapAction;
4584 
4585 static int img_bitmap(int argc, char **argv)
4586 {
4587     Error *err = NULL;
4588     int c, ret = 1;
4589     QemuOpts *opts = NULL;
4590     const char *fmt = NULL, *src_fmt = NULL, *src_filename = NULL;
4591     const char *filename, *bitmap;
4592     BlockBackend *blk = NULL, *src = NULL;
4593     BlockDriverState *bs = NULL, *src_bs = NULL;
4594     bool image_opts = false;
4595     int64_t granularity = 0;
4596     bool add = false, merge = false;
4597     QSIMPLEQ_HEAD(, ImgBitmapAction) actions;
4598     ImgBitmapAction *act, *act_next;
4599     const char *op;
4600 
4601     QSIMPLEQ_INIT(&actions);
4602 
4603     for (;;) {
4604         static const struct option long_options[] = {
4605             {"help", no_argument, 0, 'h'},
4606             {"object", required_argument, 0, OPTION_OBJECT},
4607             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
4608             {"add", no_argument, 0, OPTION_ADD},
4609             {"remove", no_argument, 0, OPTION_REMOVE},
4610             {"clear", no_argument, 0, OPTION_CLEAR},
4611             {"enable", no_argument, 0, OPTION_ENABLE},
4612             {"disable", no_argument, 0, OPTION_DISABLE},
4613             {"merge", required_argument, 0, OPTION_MERGE},
4614             {"granularity", required_argument, 0, 'g'},
4615             {"source-file", required_argument, 0, 'b'},
4616             {"source-format", required_argument, 0, 'F'},
4617             {0, 0, 0, 0}
4618         };
4619         c = getopt_long(argc, argv, ":b:f:F:g:h", long_options, NULL);
4620         if (c == -1) {
4621             break;
4622         }
4623 
4624         switch (c) {
4625         case ':':
4626             missing_argument(argv[optind - 1]);
4627             break;
4628         case '?':
4629             unrecognized_option(argv[optind - 1]);
4630             break;
4631         case 'h':
4632             help();
4633             break;
4634         case 'b':
4635             src_filename = optarg;
4636             break;
4637         case 'f':
4638             fmt = optarg;
4639             break;
4640         case 'F':
4641             src_fmt = optarg;
4642             break;
4643         case 'g':
4644             granularity = cvtnum("granularity", optarg);
4645             if (granularity < 0) {
4646                 return 1;
4647             }
4648             break;
4649         case OPTION_ADD:
4650             act = g_new0(ImgBitmapAction, 1);
4651             act->act = BITMAP_ADD;
4652             QSIMPLEQ_INSERT_TAIL(&actions, act, next);
4653             add = true;
4654             break;
4655         case OPTION_REMOVE:
4656             act = g_new0(ImgBitmapAction, 1);
4657             act->act = BITMAP_REMOVE;
4658             QSIMPLEQ_INSERT_TAIL(&actions, act, next);
4659             break;
4660         case OPTION_CLEAR:
4661             act = g_new0(ImgBitmapAction, 1);
4662             act->act = BITMAP_CLEAR;
4663             QSIMPLEQ_INSERT_TAIL(&actions, act, next);
4664             break;
4665         case OPTION_ENABLE:
4666             act = g_new0(ImgBitmapAction, 1);
4667             act->act = BITMAP_ENABLE;
4668             QSIMPLEQ_INSERT_TAIL(&actions, act, next);
4669             break;
4670         case OPTION_DISABLE:
4671             act = g_new0(ImgBitmapAction, 1);
4672             act->act = BITMAP_DISABLE;
4673             QSIMPLEQ_INSERT_TAIL(&actions, act, next);
4674             break;
4675         case OPTION_MERGE:
4676             act = g_new0(ImgBitmapAction, 1);
4677             act->act = BITMAP_MERGE;
4678             act->src = optarg;
4679             QSIMPLEQ_INSERT_TAIL(&actions, act, next);
4680             merge = true;
4681             break;
4682         case OPTION_OBJECT:
4683             user_creatable_process_cmdline(optarg);
4684             break;
4685         case OPTION_IMAGE_OPTS:
4686             image_opts = true;
4687             break;
4688         }
4689     }
4690 
4691     if (QSIMPLEQ_EMPTY(&actions)) {
4692         error_report("Need at least one of --add, --remove, --clear, "
4693                      "--enable, --disable, or --merge");
4694         goto out;
4695     }
4696 
4697     if (granularity && !add) {
4698         error_report("granularity only supported with --add");
4699         goto out;
4700     }
4701     if (src_fmt && !src_filename) {
4702         error_report("-F only supported with -b");
4703         goto out;
4704     }
4705     if (src_filename && !merge) {
4706         error_report("Merge bitmap source file only supported with "
4707                      "--merge");
4708         goto out;
4709     }
4710 
4711     if (optind != argc - 2) {
4712         error_report("Expecting filename and bitmap name");
4713         goto out;
4714     }
4715 
4716     filename = argv[optind];
4717     bitmap = argv[optind + 1];
4718 
4719     /*
4720      * No need to open backing chains; we will be manipulating bitmaps
4721      * directly in this image without reference to image contents.
4722      */
4723     blk = img_open(image_opts, filename, fmt, BDRV_O_RDWR | BDRV_O_NO_BACKING,
4724                    false, false, false);
4725     if (!blk) {
4726         goto out;
4727     }
4728     bs = blk_bs(blk);
4729     if (src_filename) {
4730         src = img_open(false, src_filename, src_fmt, BDRV_O_NO_BACKING,
4731                        false, false, false);
4732         if (!src) {
4733             goto out;
4734         }
4735         src_bs = blk_bs(src);
4736     } else {
4737         src_bs = bs;
4738     }
4739 
4740     QSIMPLEQ_FOREACH_SAFE(act, &actions, next, act_next) {
4741         switch (act->act) {
4742         case BITMAP_ADD:
4743             qmp_block_dirty_bitmap_add(bs->node_name, bitmap,
4744                                        !!granularity, granularity, true, true,
4745                                        false, false, &err);
4746             op = "add";
4747             break;
4748         case BITMAP_REMOVE:
4749             qmp_block_dirty_bitmap_remove(bs->node_name, bitmap, &err);
4750             op = "remove";
4751             break;
4752         case BITMAP_CLEAR:
4753             qmp_block_dirty_bitmap_clear(bs->node_name, bitmap, &err);
4754             op = "clear";
4755             break;
4756         case BITMAP_ENABLE:
4757             qmp_block_dirty_bitmap_enable(bs->node_name, bitmap, &err);
4758             op = "enable";
4759             break;
4760         case BITMAP_DISABLE:
4761             qmp_block_dirty_bitmap_disable(bs->node_name, bitmap, &err);
4762             op = "disable";
4763             break;
4764         case BITMAP_MERGE:
4765             do_dirty_bitmap_merge(bs->node_name, bitmap, src_bs->node_name,
4766                                   act->src, &err);
4767             op = "merge";
4768             break;
4769         default:
4770             g_assert_not_reached();
4771         }
4772 
4773         if (err) {
4774             error_reportf_err(err, "Operation %s on bitmap %s failed: ",
4775                               op, bitmap);
4776             goto out;
4777         }
4778         g_free(act);
4779     }
4780 
4781     ret = 0;
4782 
4783  out:
4784     blk_unref(src);
4785     blk_unref(blk);
4786     qemu_opts_del(opts);
4787     return ret;
4788 }
4789 
4790 #define C_BS      01
4791 #define C_COUNT   02
4792 #define C_IF      04
4793 #define C_OF      010
4794 #define C_SKIP    020
4795 
4796 struct DdInfo {
4797     unsigned int flags;
4798     int64_t count;
4799 };
4800 
4801 struct DdIo {
4802     int bsz;    /* Block size */
4803     char *filename;
4804     uint8_t *buf;
4805     int64_t offset;
4806 };
4807 
4808 struct DdOpts {
4809     const char *name;
4810     int (*f)(const char *, struct DdIo *, struct DdIo *, struct DdInfo *);
4811     unsigned int flag;
4812 };
4813 
4814 static int img_dd_bs(const char *arg,
4815                      struct DdIo *in, struct DdIo *out,
4816                      struct DdInfo *dd)
4817 {
4818     int64_t res;
4819 
4820     res = cvtnum_full("bs", arg, 1, INT_MAX);
4821 
4822     if (res < 0) {
4823         return 1;
4824     }
4825     in->bsz = out->bsz = res;
4826 
4827     return 0;
4828 }
4829 
4830 static int img_dd_count(const char *arg,
4831                         struct DdIo *in, struct DdIo *out,
4832                         struct DdInfo *dd)
4833 {
4834     dd->count = cvtnum("count", arg);
4835 
4836     if (dd->count < 0) {
4837         return 1;
4838     }
4839 
4840     return 0;
4841 }
4842 
4843 static int img_dd_if(const char *arg,
4844                      struct DdIo *in, struct DdIo *out,
4845                      struct DdInfo *dd)
4846 {
4847     in->filename = g_strdup(arg);
4848 
4849     return 0;
4850 }
4851 
4852 static int img_dd_of(const char *arg,
4853                      struct DdIo *in, struct DdIo *out,
4854                      struct DdInfo *dd)
4855 {
4856     out->filename = g_strdup(arg);
4857 
4858     return 0;
4859 }
4860 
4861 static int img_dd_skip(const char *arg,
4862                        struct DdIo *in, struct DdIo *out,
4863                        struct DdInfo *dd)
4864 {
4865     in->offset = cvtnum("skip", arg);
4866 
4867     if (in->offset < 0) {
4868         return 1;
4869     }
4870 
4871     return 0;
4872 }
4873 
4874 static int img_dd(int argc, char **argv)
4875 {
4876     int ret = 0;
4877     char *arg = NULL;
4878     char *tmp;
4879     BlockDriver *drv = NULL, *proto_drv = NULL;
4880     BlockBackend *blk1 = NULL, *blk2 = NULL;
4881     QemuOpts *opts = NULL;
4882     QemuOptsList *create_opts = NULL;
4883     Error *local_err = NULL;
4884     bool image_opts = false;
4885     int c, i;
4886     const char *out_fmt = "raw";
4887     const char *fmt = NULL;
4888     int64_t size = 0;
4889     int64_t block_count = 0, out_pos, in_pos;
4890     bool force_share = false;
4891     struct DdInfo dd = {
4892         .flags = 0,
4893         .count = 0,
4894     };
4895     struct DdIo in = {
4896         .bsz = 512, /* Block size is by default 512 bytes */
4897         .filename = NULL,
4898         .buf = NULL,
4899         .offset = 0
4900     };
4901     struct DdIo out = {
4902         .bsz = 512,
4903         .filename = NULL,
4904         .buf = NULL,
4905         .offset = 0
4906     };
4907 
4908     const struct DdOpts options[] = {
4909         { "bs", img_dd_bs, C_BS },
4910         { "count", img_dd_count, C_COUNT },
4911         { "if", img_dd_if, C_IF },
4912         { "of", img_dd_of, C_OF },
4913         { "skip", img_dd_skip, C_SKIP },
4914         { NULL, NULL, 0 }
4915     };
4916     const struct option long_options[] = {
4917         { "help", no_argument, 0, 'h'},
4918         { "object", required_argument, 0, OPTION_OBJECT},
4919         { "image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
4920         { "force-share", no_argument, 0, 'U'},
4921         { 0, 0, 0, 0 }
4922     };
4923 
4924     while ((c = getopt_long(argc, argv, ":hf:O:U", long_options, NULL))) {
4925         if (c == EOF) {
4926             break;
4927         }
4928         switch (c) {
4929         case 'O':
4930             out_fmt = optarg;
4931             break;
4932         case 'f':
4933             fmt = optarg;
4934             break;
4935         case ':':
4936             missing_argument(argv[optind - 1]);
4937             break;
4938         case '?':
4939             unrecognized_option(argv[optind - 1]);
4940             break;
4941         case 'h':
4942             help();
4943             break;
4944         case 'U':
4945             force_share = true;
4946             break;
4947         case OPTION_OBJECT:
4948             user_creatable_process_cmdline(optarg);
4949             break;
4950         case OPTION_IMAGE_OPTS:
4951             image_opts = true;
4952             break;
4953         }
4954     }
4955 
4956     for (i = optind; i < argc; i++) {
4957         int j;
4958         arg = g_strdup(argv[i]);
4959 
4960         tmp = strchr(arg, '=');
4961         if (tmp == NULL) {
4962             error_report("unrecognized operand %s", arg);
4963             ret = -1;
4964             goto out;
4965         }
4966 
4967         *tmp++ = '\0';
4968 
4969         for (j = 0; options[j].name != NULL; j++) {
4970             if (!strcmp(arg, options[j].name)) {
4971                 break;
4972             }
4973         }
4974         if (options[j].name == NULL) {
4975             error_report("unrecognized operand %s", arg);
4976             ret = -1;
4977             goto out;
4978         }
4979 
4980         if (options[j].f(tmp, &in, &out, &dd) != 0) {
4981             ret = -1;
4982             goto out;
4983         }
4984         dd.flags |= options[j].flag;
4985         g_free(arg);
4986         arg = NULL;
4987     }
4988 
4989     if (!(dd.flags & C_IF && dd.flags & C_OF)) {
4990         error_report("Must specify both input and output files");
4991         ret = -1;
4992         goto out;
4993     }
4994 
4995     blk1 = img_open(image_opts, in.filename, fmt, 0, false, false,
4996                     force_share);
4997 
4998     if (!blk1) {
4999         ret = -1;
5000         goto out;
5001     }
5002 
5003     drv = bdrv_find_format(out_fmt);
5004     if (!drv) {
5005         error_report("Unknown file format");
5006         ret = -1;
5007         goto out;
5008     }
5009     proto_drv = bdrv_find_protocol(out.filename, true, &local_err);
5010 
5011     if (!proto_drv) {
5012         error_report_err(local_err);
5013         ret = -1;
5014         goto out;
5015     }
5016     if (!drv->create_opts) {
5017         error_report("Format driver '%s' does not support image creation",
5018                      drv->format_name);
5019         ret = -1;
5020         goto out;
5021     }
5022     if (!proto_drv->create_opts) {
5023         error_report("Protocol driver '%s' does not support image creation",
5024                      proto_drv->format_name);
5025         ret = -1;
5026         goto out;
5027     }
5028     create_opts = qemu_opts_append(create_opts, drv->create_opts);
5029     create_opts = qemu_opts_append(create_opts, proto_drv->create_opts);
5030 
5031     opts = qemu_opts_create(create_opts, NULL, 0, &error_abort);
5032 
5033     size = blk_getlength(blk1);
5034     if (size < 0) {
5035         error_report("Failed to get size for '%s'", in.filename);
5036         ret = -1;
5037         goto out;
5038     }
5039 
5040     if (dd.flags & C_COUNT && dd.count <= INT64_MAX / in.bsz &&
5041         dd.count * in.bsz < size) {
5042         size = dd.count * in.bsz;
5043     }
5044 
5045     /* Overflow means the specified offset is beyond input image's size */
5046     if (dd.flags & C_SKIP && (in.offset > INT64_MAX / in.bsz ||
5047                               size < in.bsz * in.offset)) {
5048         qemu_opt_set_number(opts, BLOCK_OPT_SIZE, 0, &error_abort);
5049     } else {
5050         qemu_opt_set_number(opts, BLOCK_OPT_SIZE,
5051                             size - in.bsz * in.offset, &error_abort);
5052     }
5053 
5054     ret = bdrv_create(drv, out.filename, opts, &local_err);
5055     if (ret < 0) {
5056         error_reportf_err(local_err,
5057                           "%s: error while creating output image: ",
5058                           out.filename);
5059         ret = -1;
5060         goto out;
5061     }
5062 
5063     /* TODO, we can't honour --image-opts for the target,
5064      * since it needs to be given in a format compatible
5065      * with the bdrv_create() call above which does not
5066      * support image-opts style.
5067      */
5068     blk2 = img_open_file(out.filename, NULL, out_fmt, BDRV_O_RDWR,
5069                          false, false, false);
5070 
5071     if (!blk2) {
5072         ret = -1;
5073         goto out;
5074     }
5075 
5076     if (dd.flags & C_SKIP && (in.offset > INT64_MAX / in.bsz ||
5077                               size < in.offset * in.bsz)) {
5078         /* We give a warning if the skip option is bigger than the input
5079          * size and create an empty output disk image (i.e. like dd(1)).
5080          */
5081         error_report("%s: cannot skip to specified offset", in.filename);
5082         in_pos = size;
5083     } else {
5084         in_pos = in.offset * in.bsz;
5085     }
5086 
5087     in.buf = g_new(uint8_t, in.bsz);
5088 
5089     for (out_pos = 0; in_pos < size; block_count++) {
5090         int in_ret, out_ret;
5091 
5092         if (in_pos + in.bsz > size) {
5093             in_ret = blk_pread(blk1, in_pos, in.buf, size - in_pos);
5094         } else {
5095             in_ret = blk_pread(blk1, in_pos, in.buf, in.bsz);
5096         }
5097         if (in_ret < 0) {
5098             error_report("error while reading from input image file: %s",
5099                          strerror(-in_ret));
5100             ret = -1;
5101             goto out;
5102         }
5103         in_pos += in_ret;
5104 
5105         out_ret = blk_pwrite(blk2, out_pos, in.buf, in_ret, 0);
5106 
5107         if (out_ret < 0) {
5108             error_report("error while writing to output image file: %s",
5109                          strerror(-out_ret));
5110             ret = -1;
5111             goto out;
5112         }
5113         out_pos += out_ret;
5114     }
5115 
5116 out:
5117     g_free(arg);
5118     qemu_opts_del(opts);
5119     qemu_opts_free(create_opts);
5120     blk_unref(blk1);
5121     blk_unref(blk2);
5122     g_free(in.filename);
5123     g_free(out.filename);
5124     g_free(in.buf);
5125     g_free(out.buf);
5126 
5127     if (ret) {
5128         return 1;
5129     }
5130     return 0;
5131 }
5132 
5133 static void dump_json_block_measure_info(BlockMeasureInfo *info)
5134 {
5135     GString *str;
5136     QObject *obj;
5137     Visitor *v = qobject_output_visitor_new(&obj);
5138 
5139     visit_type_BlockMeasureInfo(v, NULL, &info, &error_abort);
5140     visit_complete(v, &obj);
5141     str = qobject_to_json_pretty(obj, true);
5142     assert(str != NULL);
5143     printf("%s\n", str->str);
5144     qobject_unref(obj);
5145     visit_free(v);
5146     g_string_free(str, true);
5147 }
5148 
5149 static int img_measure(int argc, char **argv)
5150 {
5151     static const struct option long_options[] = {
5152         {"help", no_argument, 0, 'h'},
5153         {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
5154         {"object", required_argument, 0, OPTION_OBJECT},
5155         {"output", required_argument, 0, OPTION_OUTPUT},
5156         {"size", required_argument, 0, OPTION_SIZE},
5157         {"force-share", no_argument, 0, 'U'},
5158         {0, 0, 0, 0}
5159     };
5160     OutputFormat output_format = OFORMAT_HUMAN;
5161     BlockBackend *in_blk = NULL;
5162     BlockDriver *drv;
5163     const char *filename = NULL;
5164     const char *fmt = NULL;
5165     const char *out_fmt = "raw";
5166     char *options = NULL;
5167     char *snapshot_name = NULL;
5168     bool force_share = false;
5169     QemuOpts *opts = NULL;
5170     QemuOpts *object_opts = NULL;
5171     QemuOpts *sn_opts = NULL;
5172     QemuOptsList *create_opts = NULL;
5173     bool image_opts = false;
5174     uint64_t img_size = UINT64_MAX;
5175     BlockMeasureInfo *info = NULL;
5176     Error *local_err = NULL;
5177     int ret = 1;
5178     int c;
5179 
5180     while ((c = getopt_long(argc, argv, "hf:O:o:l:U",
5181                             long_options, NULL)) != -1) {
5182         switch (c) {
5183         case '?':
5184         case 'h':
5185             help();
5186             break;
5187         case 'f':
5188             fmt = optarg;
5189             break;
5190         case 'O':
5191             out_fmt = optarg;
5192             break;
5193         case 'o':
5194             if (accumulate_options(&options, optarg) < 0) {
5195                 goto out;
5196             }
5197             break;
5198         case 'l':
5199             if (strstart(optarg, SNAPSHOT_OPT_BASE, NULL)) {
5200                 sn_opts = qemu_opts_parse_noisily(&internal_snapshot_opts,
5201                                                   optarg, false);
5202                 if (!sn_opts) {
5203                     error_report("Failed in parsing snapshot param '%s'",
5204                                  optarg);
5205                     goto out;
5206                 }
5207             } else {
5208                 snapshot_name = optarg;
5209             }
5210             break;
5211         case 'U':
5212             force_share = true;
5213             break;
5214         case OPTION_OBJECT:
5215             user_creatable_process_cmdline(optarg);
5216             break;
5217         case OPTION_IMAGE_OPTS:
5218             image_opts = true;
5219             break;
5220         case OPTION_OUTPUT:
5221             if (!strcmp(optarg, "json")) {
5222                 output_format = OFORMAT_JSON;
5223             } else if (!strcmp(optarg, "human")) {
5224                 output_format = OFORMAT_HUMAN;
5225             } else {
5226                 error_report("--output must be used with human or json "
5227                              "as argument.");
5228                 goto out;
5229             }
5230             break;
5231         case OPTION_SIZE:
5232         {
5233             int64_t sval;
5234 
5235             sval = cvtnum("image size", optarg);
5236             if (sval < 0) {
5237                 goto out;
5238             }
5239             img_size = (uint64_t)sval;
5240         }
5241         break;
5242         }
5243     }
5244 
5245     if (argc - optind > 1) {
5246         error_report("At most one filename argument is allowed.");
5247         goto out;
5248     } else if (argc - optind == 1) {
5249         filename = argv[optind];
5250     }
5251 
5252     if (!filename && (image_opts || fmt || snapshot_name || sn_opts)) {
5253         error_report("--image-opts, -f, and -l require a filename argument.");
5254         goto out;
5255     }
5256     if (filename && img_size != UINT64_MAX) {
5257         error_report("--size N cannot be used together with a filename.");
5258         goto out;
5259     }
5260     if (!filename && img_size == UINT64_MAX) {
5261         error_report("Either --size N or one filename must be specified.");
5262         goto out;
5263     }
5264 
5265     if (filename) {
5266         in_blk = img_open(image_opts, filename, fmt, 0,
5267                           false, false, force_share);
5268         if (!in_blk) {
5269             goto out;
5270         }
5271 
5272         if (sn_opts) {
5273             bdrv_snapshot_load_tmp(blk_bs(in_blk),
5274                     qemu_opt_get(sn_opts, SNAPSHOT_OPT_ID),
5275                     qemu_opt_get(sn_opts, SNAPSHOT_OPT_NAME),
5276                     &local_err);
5277         } else if (snapshot_name != NULL) {
5278             bdrv_snapshot_load_tmp_by_id_or_name(blk_bs(in_blk),
5279                     snapshot_name, &local_err);
5280         }
5281         if (local_err) {
5282             error_reportf_err(local_err, "Failed to load snapshot: ");
5283             goto out;
5284         }
5285     }
5286 
5287     drv = bdrv_find_format(out_fmt);
5288     if (!drv) {
5289         error_report("Unknown file format '%s'", out_fmt);
5290         goto out;
5291     }
5292     if (!drv->create_opts) {
5293         error_report("Format driver '%s' does not support image creation",
5294                      drv->format_name);
5295         goto out;
5296     }
5297 
5298     create_opts = qemu_opts_append(create_opts, drv->create_opts);
5299     create_opts = qemu_opts_append(create_opts, bdrv_file.create_opts);
5300     opts = qemu_opts_create(create_opts, NULL, 0, &error_abort);
5301     if (options) {
5302         if (!qemu_opts_do_parse(opts, options, NULL, &local_err)) {
5303             error_report_err(local_err);
5304             error_report("Invalid options for file format '%s'", out_fmt);
5305             goto out;
5306         }
5307     }
5308     if (img_size != UINT64_MAX) {
5309         qemu_opt_set_number(opts, BLOCK_OPT_SIZE, img_size, &error_abort);
5310     }
5311 
5312     info = bdrv_measure(drv, opts, in_blk ? blk_bs(in_blk) : NULL, &local_err);
5313     if (local_err) {
5314         error_report_err(local_err);
5315         goto out;
5316     }
5317 
5318     if (output_format == OFORMAT_HUMAN) {
5319         printf("required size: %" PRIu64 "\n", info->required);
5320         printf("fully allocated size: %" PRIu64 "\n", info->fully_allocated);
5321         if (info->has_bitmaps) {
5322             printf("bitmaps size: %" PRIu64 "\n", info->bitmaps);
5323         }
5324     } else {
5325         dump_json_block_measure_info(info);
5326     }
5327 
5328     ret = 0;
5329 
5330 out:
5331     qapi_free_BlockMeasureInfo(info);
5332     qemu_opts_del(object_opts);
5333     qemu_opts_del(opts);
5334     qemu_opts_del(sn_opts);
5335     qemu_opts_free(create_opts);
5336     g_free(options);
5337     blk_unref(in_blk);
5338     return ret;
5339 }
5340 
5341 static const img_cmd_t img_cmds[] = {
5342 #define DEF(option, callback, arg_string)        \
5343     { option, callback },
5344 #include "qemu-img-cmds.h"
5345 #undef DEF
5346     { NULL, NULL, },
5347 };
5348 
5349 int main(int argc, char **argv)
5350 {
5351     const img_cmd_t *cmd;
5352     const char *cmdname;
5353     Error *local_error = NULL;
5354     int c;
5355     static const struct option long_options[] = {
5356         {"help", no_argument, 0, 'h'},
5357         {"version", no_argument, 0, 'V'},
5358         {"trace", required_argument, NULL, 'T'},
5359         {0, 0, 0, 0}
5360     };
5361 
5362 #ifdef CONFIG_POSIX
5363     signal(SIGPIPE, SIG_IGN);
5364 #endif
5365 
5366     socket_init();
5367     error_init(argv[0]);
5368     module_call_init(MODULE_INIT_TRACE);
5369     qemu_init_exec_dir(argv[0]);
5370 
5371     if (qemu_init_main_loop(&local_error)) {
5372         error_report_err(local_error);
5373         exit(EXIT_FAILURE);
5374     }
5375 
5376     qcrypto_init(&error_fatal);
5377 
5378     module_call_init(MODULE_INIT_QOM);
5379     bdrv_init();
5380     if (argc < 2) {
5381         error_exit("Not enough arguments");
5382     }
5383 
5384     qemu_add_opts(&qemu_source_opts);
5385     qemu_add_opts(&qemu_trace_opts);
5386 
5387     while ((c = getopt_long(argc, argv, "+:hVT:", long_options, NULL)) != -1) {
5388         switch (c) {
5389         case ':':
5390             missing_argument(argv[optind - 1]);
5391             return 0;
5392         case '?':
5393             unrecognized_option(argv[optind - 1]);
5394             return 0;
5395         case 'h':
5396             help();
5397             return 0;
5398         case 'V':
5399             printf(QEMU_IMG_VERSION);
5400             return 0;
5401         case 'T':
5402             trace_opt_parse(optarg);
5403             break;
5404         }
5405     }
5406 
5407     cmdname = argv[optind];
5408 
5409     /* reset getopt_long scanning */
5410     argc -= optind;
5411     if (argc < 1) {
5412         return 0;
5413     }
5414     argv += optind;
5415     qemu_reset_optind();
5416 
5417     if (!trace_init_backends()) {
5418         exit(1);
5419     }
5420     trace_init_file();
5421     qemu_set_log(LOG_TRACE);
5422 
5423     /* find the command */
5424     for (cmd = img_cmds; cmd->name != NULL; cmd++) {
5425         if (!strcmp(cmdname, cmd->name)) {
5426             return cmd->handler(argc, argv);
5427         }
5428     }
5429 
5430     /* not found */
5431     error_exit("Command not found: %s", cmdname);
5432 }
5433