xref: /openbmc/qemu/qemu-img.c (revision 977a7204)
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-version.h"
29 #include "qapi/error.h"
30 #include "qapi/qapi-visit-block-core.h"
31 #include "qapi/qobject-output-visitor.h"
32 #include "qapi/qmp/qjson.h"
33 #include "qapi/qmp/qdict.h"
34 #include "qapi/qmp/qstring.h"
35 #include "qemu/cutils.h"
36 #include "qemu/config-file.h"
37 #include "qemu/option.h"
38 #include "qemu/error-report.h"
39 #include "qemu/log.h"
40 #include "qom/object_interfaces.h"
41 #include "sysemu/sysemu.h"
42 #include "sysemu/block-backend.h"
43 #include "block/block_int.h"
44 #include "block/blockjob.h"
45 #include "block/qapi.h"
46 #include "crypto/init.h"
47 #include "trace/control.h"
48 
49 #define QEMU_IMG_VERSION "qemu-img version " QEMU_FULL_VERSION \
50                           "\n" QEMU_COPYRIGHT "\n"
51 
52 typedef struct img_cmd_t {
53     const char *name;
54     int (*handler)(int argc, char **argv);
55 } img_cmd_t;
56 
57 enum {
58     OPTION_OUTPUT = 256,
59     OPTION_BACKING_CHAIN = 257,
60     OPTION_OBJECT = 258,
61     OPTION_IMAGE_OPTS = 259,
62     OPTION_PATTERN = 260,
63     OPTION_FLUSH_INTERVAL = 261,
64     OPTION_NO_DRAIN = 262,
65     OPTION_TARGET_IMAGE_OPTS = 263,
66     OPTION_SIZE = 264,
67     OPTION_PREALLOCATION = 265,
68     OPTION_SHRINK = 266,
69 };
70 
71 typedef enum OutputFormat {
72     OFORMAT_JSON,
73     OFORMAT_HUMAN,
74 } OutputFormat;
75 
76 /* Default to cache=writeback as data integrity is not important for qemu-img */
77 #define BDRV_DEFAULT_CACHE "writeback"
78 
79 static void format_print(void *opaque, const char *name)
80 {
81     printf(" %s", name);
82 }
83 
84 static void QEMU_NORETURN GCC_FMT_ATTR(1, 2) error_exit(const char *fmt, ...)
85 {
86     va_list ap;
87 
88     error_printf("qemu-img: ");
89 
90     va_start(ap, fmt);
91     error_vprintf(fmt, ap);
92     va_end(ap);
93 
94     error_printf("\nTry 'qemu-img --help' for more information\n");
95     exit(EXIT_FAILURE);
96 }
97 
98 static void QEMU_NORETURN missing_argument(const char *option)
99 {
100     error_exit("missing argument for option '%s'", option);
101 }
102 
103 static void QEMU_NORETURN unrecognized_option(const char *option)
104 {
105     error_exit("unrecognized option '%s'", option);
106 }
107 
108 /* Please keep in synch with qemu-img.texi */
109 static void QEMU_NORETURN help(void)
110 {
111     const char *help_msg =
112            QEMU_IMG_VERSION
113            "usage: qemu-img [standard options] command [command options]\n"
114            "QEMU disk image utility\n"
115            "\n"
116            "    '-h', '--help'       display this help and exit\n"
117            "    '-V', '--version'    output version information and exit\n"
118            "    '-T', '--trace'      [[enable=]<pattern>][,events=<file>][,file=<file>]\n"
119            "                         specify tracing options\n"
120            "\n"
121            "Command syntax:\n"
122 #define DEF(option, callback, arg_string)        \
123            "  " arg_string "\n"
124 #include "qemu-img-cmds.h"
125 #undef DEF
126            "\n"
127            "Command parameters:\n"
128            "  'filename' is a disk image filename\n"
129            "  'objectdef' is a QEMU user creatable object definition. See the qemu(1)\n"
130            "    manual page for a description of the object properties. The most common\n"
131            "    object type is a 'secret', which is used to supply passwords and/or\n"
132            "    encryption keys.\n"
133            "  'fmt' is the disk image format. It is guessed automatically in most cases\n"
134            "  'cache' is the cache mode used to write the output disk image, the valid\n"
135            "    options are: 'none', 'writeback' (default, except for convert), 'writethrough',\n"
136            "    'directsync' and 'unsafe' (default for convert)\n"
137            "  'src_cache' is the cache mode used to read input disk images, the valid\n"
138            "    options are the same as for the 'cache' option\n"
139            "  'size' is the disk image size in bytes. Optional suffixes\n"
140            "    'k' or 'K' (kilobyte, 1024), 'M' (megabyte, 1024k), 'G' (gigabyte, 1024M),\n"
141            "    'T' (terabyte, 1024G), 'P' (petabyte, 1024T) and 'E' (exabyte, 1024P)  are\n"
142            "    supported. 'b' is ignored.\n"
143            "  'output_filename' is the destination disk image filename\n"
144            "  'output_fmt' is the destination format\n"
145            "  'options' is a comma separated list of format specific options in a\n"
146            "    name=value format. Use -o ? for an overview of the options supported by the\n"
147            "    used format\n"
148            "  'snapshot_param' is param used for internal snapshot, format\n"
149            "    is 'snapshot.id=[ID],snapshot.name=[NAME]', or\n"
150            "    '[ID_OR_NAME]'\n"
151            "  '-c' indicates that target image must be compressed (qcow format only)\n"
152            "  '-u' allows unsafe backing chains. For rebasing, it is assumed that old and\n"
153            "       new backing file match exactly. The image doesn't need a working\n"
154            "       backing file before rebasing in this case (useful for renaming the\n"
155            "       backing file). For image creation, allow creating without attempting\n"
156            "       to open the backing file.\n"
157            "  '-h' with or without a command shows this help and lists the supported formats\n"
158            "  '-p' show progress of command (only certain commands)\n"
159            "  '-q' use Quiet mode - do not print any output (except errors)\n"
160            "  '-S' indicates the consecutive number of bytes (defaults to 4k) that must\n"
161            "       contain only zeros for qemu-img to create a sparse image during\n"
162            "       conversion. If the number of bytes is 0, the source will not be scanned for\n"
163            "       unallocated or zero sectors, and the destination image will always be\n"
164            "       fully allocated\n"
165            "  '--output' takes the format in which the output must be done (human or json)\n"
166            "  '-n' skips the target volume creation (useful if the volume is created\n"
167            "       prior to running qemu-img)\n"
168            "\n"
169            "Parameters to check subcommand:\n"
170            "  '-r' tries to repair any inconsistencies that are found during the check.\n"
171            "       '-r leaks' repairs only cluster leaks, whereas '-r all' fixes all\n"
172            "       kinds of errors, with a higher risk of choosing the wrong fix or\n"
173            "       hiding corruption that has already occurred.\n"
174            "\n"
175            "Parameters to convert subcommand:\n"
176            "  '-m' specifies how many coroutines work in parallel during the convert\n"
177            "       process (defaults to 8)\n"
178            "  '-W' allow to write to the target out of order rather than sequential\n"
179            "\n"
180            "Parameters to snapshot subcommand:\n"
181            "  'snapshot' is the name of the snapshot to create, apply or delete\n"
182            "  '-a' applies a snapshot (revert disk to saved state)\n"
183            "  '-c' creates a snapshot\n"
184            "  '-d' deletes a snapshot\n"
185            "  '-l' lists all snapshots in the given image\n"
186            "\n"
187            "Parameters to compare subcommand:\n"
188            "  '-f' first image format\n"
189            "  '-F' second image format\n"
190            "  '-s' run in Strict mode - fail on different image size or sector allocation\n"
191            "\n"
192            "Parameters to dd subcommand:\n"
193            "  'bs=BYTES' read and write up to BYTES bytes at a time "
194            "(default: 512)\n"
195            "  'count=N' copy only N input blocks\n"
196            "  'if=FILE' read from FILE\n"
197            "  'of=FILE' write to FILE\n"
198            "  'skip=N' skip N bs-sized blocks at the start of input\n";
199 
200     printf("%s\nSupported formats:", help_msg);
201     bdrv_iterate_format(format_print, NULL);
202     printf("\n\n" QEMU_HELP_BOTTOM "\n");
203     exit(EXIT_SUCCESS);
204 }
205 
206 static QemuOptsList qemu_object_opts = {
207     .name = "object",
208     .implied_opt_name = "qom-type",
209     .head = QTAILQ_HEAD_INITIALIZER(qemu_object_opts.head),
210     .desc = {
211         { }
212     },
213 };
214 
215 static QemuOptsList qemu_source_opts = {
216     .name = "source",
217     .implied_opt_name = "file",
218     .head = QTAILQ_HEAD_INITIALIZER(qemu_source_opts.head),
219     .desc = {
220         { }
221     },
222 };
223 
224 static int GCC_FMT_ATTR(2, 3) qprintf(bool quiet, const char *fmt, ...)
225 {
226     int ret = 0;
227     if (!quiet) {
228         va_list args;
229         va_start(args, fmt);
230         ret = vprintf(fmt, args);
231         va_end(args);
232     }
233     return ret;
234 }
235 
236 
237 static int print_block_option_help(const char *filename, const char *fmt)
238 {
239     BlockDriver *drv, *proto_drv;
240     QemuOptsList *create_opts = NULL;
241     Error *local_err = NULL;
242 
243     /* Find driver and parse its options */
244     drv = bdrv_find_format(fmt);
245     if (!drv) {
246         error_report("Unknown file format '%s'", fmt);
247         return 1;
248     }
249 
250     if (!drv->create_opts) {
251         error_report("Format driver '%s' does not support image creation", fmt);
252         return 1;
253     }
254 
255     create_opts = qemu_opts_append(create_opts, drv->create_opts);
256     if (filename) {
257         proto_drv = bdrv_find_protocol(filename, true, &local_err);
258         if (!proto_drv) {
259             error_report_err(local_err);
260             qemu_opts_free(create_opts);
261             return 1;
262         }
263         if (!proto_drv->create_opts) {
264             error_report("Protocal driver '%s' does not support image creation",
265                          proto_drv->format_name);
266             return 1;
267         }
268         create_opts = qemu_opts_append(create_opts, proto_drv->create_opts);
269     }
270 
271     printf("Supported options:\n");
272     qemu_opts_print_help(create_opts);
273     qemu_opts_free(create_opts);
274     return 0;
275 }
276 
277 
278 static BlockBackend *img_open_opts(const char *optstr,
279                                    QemuOpts *opts, int flags, bool writethrough,
280                                    bool quiet, bool force_share)
281 {
282     QDict *options;
283     Error *local_err = NULL;
284     BlockBackend *blk;
285     options = qemu_opts_to_qdict(opts, NULL);
286     if (force_share) {
287         if (qdict_haskey(options, BDRV_OPT_FORCE_SHARE)
288             && strcmp(qdict_get_str(options, BDRV_OPT_FORCE_SHARE), "on")) {
289             error_report("--force-share/-U conflicts with image options");
290             qobject_unref(options);
291             return NULL;
292         }
293         qdict_put_str(options, BDRV_OPT_FORCE_SHARE, "on");
294     }
295     blk = blk_new_open(NULL, NULL, options, flags, &local_err);
296     if (!blk) {
297         error_reportf_err(local_err, "Could not open '%s': ", optstr);
298         return NULL;
299     }
300     blk_set_enable_write_cache(blk, !writethrough);
301 
302     return blk;
303 }
304 
305 static BlockBackend *img_open_file(const char *filename,
306                                    QDict *options,
307                                    const char *fmt, int flags,
308                                    bool writethrough, bool quiet,
309                                    bool force_share)
310 {
311     BlockBackend *blk;
312     Error *local_err = NULL;
313 
314     if (!options) {
315         options = qdict_new();
316     }
317     if (fmt) {
318         qdict_put_str(options, "driver", fmt);
319     }
320 
321     if (force_share) {
322         qdict_put_bool(options, BDRV_OPT_FORCE_SHARE, true);
323     }
324     blk = blk_new_open(filename, NULL, options, flags, &local_err);
325     if (!blk) {
326         error_reportf_err(local_err, "Could not open '%s': ", filename);
327         return NULL;
328     }
329     blk_set_enable_write_cache(blk, !writethrough);
330 
331     return blk;
332 }
333 
334 
335 static int img_add_key_secrets(void *opaque,
336                                const char *name, const char *value,
337                                Error **errp)
338 {
339     QDict *options = opaque;
340 
341     if (g_str_has_suffix(name, "key-secret")) {
342         qdict_put_str(options, name, value);
343     }
344 
345     return 0;
346 }
347 
348 static BlockBackend *img_open_new_file(const char *filename,
349                                        QemuOpts *create_opts,
350                                        const char *fmt, int flags,
351                                        bool writethrough, bool quiet,
352                                        bool force_share)
353 {
354     QDict *options = NULL;
355 
356     options = qdict_new();
357     qemu_opt_foreach(create_opts, img_add_key_secrets, options, &error_abort);
358 
359     return img_open_file(filename, options, fmt, flags, writethrough, quiet,
360                          force_share);
361 }
362 
363 
364 static BlockBackend *img_open(bool image_opts,
365                               const char *filename,
366                               const char *fmt, int flags, bool writethrough,
367                               bool quiet, bool force_share)
368 {
369     BlockBackend *blk;
370     if (image_opts) {
371         QemuOpts *opts;
372         if (fmt) {
373             error_report("--image-opts and --format are mutually exclusive");
374             return NULL;
375         }
376         opts = qemu_opts_parse_noisily(qemu_find_opts("source"),
377                                        filename, true);
378         if (!opts) {
379             return NULL;
380         }
381         blk = img_open_opts(filename, opts, flags, writethrough, quiet,
382                             force_share);
383     } else {
384         blk = img_open_file(filename, NULL, fmt, flags, writethrough, quiet,
385                             force_share);
386     }
387     return blk;
388 }
389 
390 
391 static int add_old_style_options(const char *fmt, QemuOpts *opts,
392                                  const char *base_filename,
393                                  const char *base_fmt)
394 {
395     Error *err = NULL;
396 
397     if (base_filename) {
398         qemu_opt_set(opts, BLOCK_OPT_BACKING_FILE, base_filename, &err);
399         if (err) {
400             error_report("Backing file not supported for file format '%s'",
401                          fmt);
402             error_free(err);
403             return -1;
404         }
405     }
406     if (base_fmt) {
407         qemu_opt_set(opts, BLOCK_OPT_BACKING_FMT, base_fmt, &err);
408         if (err) {
409             error_report("Backing file format not supported for file "
410                          "format '%s'", fmt);
411             error_free(err);
412             return -1;
413         }
414     }
415     return 0;
416 }
417 
418 static int64_t cvtnum(const char *s)
419 {
420     int err;
421     uint64_t value;
422 
423     err = qemu_strtosz(s, NULL, &value);
424     if (err < 0) {
425         return err;
426     }
427     if (value > INT64_MAX) {
428         return -ERANGE;
429     }
430     return value;
431 }
432 
433 static int img_create(int argc, char **argv)
434 {
435     int c;
436     uint64_t img_size = -1;
437     const char *fmt = "raw";
438     const char *base_fmt = NULL;
439     const char *filename;
440     const char *base_filename = NULL;
441     char *options = NULL;
442     Error *local_err = NULL;
443     bool quiet = false;
444     int flags = 0;
445 
446     for(;;) {
447         static const struct option long_options[] = {
448             {"help", no_argument, 0, 'h'},
449             {"object", required_argument, 0, OPTION_OBJECT},
450             {0, 0, 0, 0}
451         };
452         c = getopt_long(argc, argv, ":F:b:f:ho:qu",
453                         long_options, NULL);
454         if (c == -1) {
455             break;
456         }
457         switch(c) {
458         case ':':
459             missing_argument(argv[optind - 1]);
460             break;
461         case '?':
462             unrecognized_option(argv[optind - 1]);
463             break;
464         case 'h':
465             help();
466             break;
467         case 'F':
468             base_fmt = optarg;
469             break;
470         case 'b':
471             base_filename = optarg;
472             break;
473         case 'f':
474             fmt = optarg;
475             break;
476         case 'o':
477             if (!is_valid_option_list(optarg)) {
478                 error_report("Invalid option list: %s", optarg);
479                 goto fail;
480             }
481             if (!options) {
482                 options = g_strdup(optarg);
483             } else {
484                 char *old_options = options;
485                 options = g_strdup_printf("%s,%s", options, optarg);
486                 g_free(old_options);
487             }
488             break;
489         case 'q':
490             quiet = true;
491             break;
492         case 'u':
493             flags |= BDRV_O_NO_BACKING;
494             break;
495         case OPTION_OBJECT: {
496             QemuOpts *opts;
497             opts = qemu_opts_parse_noisily(&qemu_object_opts,
498                                            optarg, true);
499             if (!opts) {
500                 goto fail;
501             }
502         }   break;
503         }
504     }
505 
506     /* Get the filename */
507     filename = (optind < argc) ? argv[optind] : NULL;
508     if (options && has_help_option(options)) {
509         g_free(options);
510         return print_block_option_help(filename, fmt);
511     }
512 
513     if (optind >= argc) {
514         error_exit("Expecting image file name");
515     }
516     optind++;
517 
518     if (qemu_opts_foreach(&qemu_object_opts,
519                           user_creatable_add_opts_foreach,
520                           NULL, NULL)) {
521         goto fail;
522     }
523 
524     /* Get image size, if specified */
525     if (optind < argc) {
526         int64_t sval;
527 
528         sval = cvtnum(argv[optind++]);
529         if (sval < 0) {
530             if (sval == -ERANGE) {
531                 error_report("Image size must be less than 8 EiB!");
532             } else {
533                 error_report("Invalid image size specified! You may use k, M, "
534                       "G, T, P or E suffixes for ");
535                 error_report("kilobytes, megabytes, gigabytes, terabytes, "
536                              "petabytes and exabytes.");
537             }
538             goto fail;
539         }
540         img_size = (uint64_t)sval;
541     }
542     if (optind != argc) {
543         error_exit("Unexpected argument: %s", argv[optind]);
544     }
545 
546     bdrv_img_create(filename, fmt, base_filename, base_fmt,
547                     options, img_size, flags, quiet, &local_err);
548     if (local_err) {
549         error_reportf_err(local_err, "%s: ", filename);
550         goto fail;
551     }
552 
553     g_free(options);
554     return 0;
555 
556 fail:
557     g_free(options);
558     return 1;
559 }
560 
561 static void dump_json_image_check(ImageCheck *check, bool quiet)
562 {
563     QString *str;
564     QObject *obj;
565     Visitor *v = qobject_output_visitor_new(&obj);
566 
567     visit_type_ImageCheck(v, NULL, &check, &error_abort);
568     visit_complete(v, &obj);
569     str = qobject_to_json_pretty(obj);
570     assert(str != NULL);
571     qprintf(quiet, "%s\n", qstring_get_str(str));
572     qobject_unref(obj);
573     visit_free(v);
574     qobject_unref(str);
575 }
576 
577 static void dump_human_image_check(ImageCheck *check, bool quiet)
578 {
579     if (!(check->corruptions || check->leaks || check->check_errors)) {
580         qprintf(quiet, "No errors were found on the image.\n");
581     } else {
582         if (check->corruptions) {
583             qprintf(quiet, "\n%" PRId64 " errors were found on the image.\n"
584                     "Data may be corrupted, or further writes to the image "
585                     "may corrupt it.\n",
586                     check->corruptions);
587         }
588 
589         if (check->leaks) {
590             qprintf(quiet,
591                     "\n%" PRId64 " leaked clusters were found on the image.\n"
592                     "This means waste of disk space, but no harm to data.\n",
593                     check->leaks);
594         }
595 
596         if (check->check_errors) {
597             qprintf(quiet,
598                     "\n%" PRId64
599                     " internal errors have occurred during the check.\n",
600                     check->check_errors);
601         }
602     }
603 
604     if (check->total_clusters != 0 && check->allocated_clusters != 0) {
605         qprintf(quiet, "%" PRId64 "/%" PRId64 " = %0.2f%% allocated, "
606                 "%0.2f%% fragmented, %0.2f%% compressed clusters\n",
607                 check->allocated_clusters, check->total_clusters,
608                 check->allocated_clusters * 100.0 / check->total_clusters,
609                 check->fragmented_clusters * 100.0 / check->allocated_clusters,
610                 check->compressed_clusters * 100.0 /
611                 check->allocated_clusters);
612     }
613 
614     if (check->image_end_offset) {
615         qprintf(quiet,
616                 "Image end offset: %" PRId64 "\n", check->image_end_offset);
617     }
618 }
619 
620 static int collect_image_check(BlockDriverState *bs,
621                    ImageCheck *check,
622                    const char *filename,
623                    const char *fmt,
624                    int fix)
625 {
626     int ret;
627     BdrvCheckResult result;
628 
629     ret = bdrv_check(bs, &result, fix);
630     if (ret < 0) {
631         return ret;
632     }
633 
634     check->filename                 = g_strdup(filename);
635     check->format                   = g_strdup(bdrv_get_format_name(bs));
636     check->check_errors             = result.check_errors;
637     check->corruptions              = result.corruptions;
638     check->has_corruptions          = result.corruptions != 0;
639     check->leaks                    = result.leaks;
640     check->has_leaks                = result.leaks != 0;
641     check->corruptions_fixed        = result.corruptions_fixed;
642     check->has_corruptions_fixed    = result.corruptions != 0;
643     check->leaks_fixed              = result.leaks_fixed;
644     check->has_leaks_fixed          = result.leaks != 0;
645     check->image_end_offset         = result.image_end_offset;
646     check->has_image_end_offset     = result.image_end_offset != 0;
647     check->total_clusters           = result.bfi.total_clusters;
648     check->has_total_clusters       = result.bfi.total_clusters != 0;
649     check->allocated_clusters       = result.bfi.allocated_clusters;
650     check->has_allocated_clusters   = result.bfi.allocated_clusters != 0;
651     check->fragmented_clusters      = result.bfi.fragmented_clusters;
652     check->has_fragmented_clusters  = result.bfi.fragmented_clusters != 0;
653     check->compressed_clusters      = result.bfi.compressed_clusters;
654     check->has_compressed_clusters  = result.bfi.compressed_clusters != 0;
655 
656     return 0;
657 }
658 
659 /*
660  * Checks an image for consistency. Exit codes:
661  *
662  *  0 - Check completed, image is good
663  *  1 - Check not completed because of internal errors
664  *  2 - Check completed, image is corrupted
665  *  3 - Check completed, image has leaked clusters, but is good otherwise
666  * 63 - Checks are not supported by the image format
667  */
668 static int img_check(int argc, char **argv)
669 {
670     int c, ret;
671     OutputFormat output_format = OFORMAT_HUMAN;
672     const char *filename, *fmt, *output, *cache;
673     BlockBackend *blk;
674     BlockDriverState *bs;
675     int fix = 0;
676     int flags = BDRV_O_CHECK;
677     bool writethrough;
678     ImageCheck *check;
679     bool quiet = false;
680     bool image_opts = false;
681     bool force_share = false;
682 
683     fmt = NULL;
684     output = NULL;
685     cache = BDRV_DEFAULT_CACHE;
686 
687     for(;;) {
688         int option_index = 0;
689         static const struct option long_options[] = {
690             {"help", no_argument, 0, 'h'},
691             {"format", required_argument, 0, 'f'},
692             {"repair", required_argument, 0, 'r'},
693             {"output", required_argument, 0, OPTION_OUTPUT},
694             {"object", required_argument, 0, OPTION_OBJECT},
695             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
696             {"force-share", no_argument, 0, 'U'},
697             {0, 0, 0, 0}
698         };
699         c = getopt_long(argc, argv, ":hf:r:T:qU",
700                         long_options, &option_index);
701         if (c == -1) {
702             break;
703         }
704         switch(c) {
705         case ':':
706             missing_argument(argv[optind - 1]);
707             break;
708         case '?':
709             unrecognized_option(argv[optind - 1]);
710             break;
711         case 'h':
712             help();
713             break;
714         case 'f':
715             fmt = optarg;
716             break;
717         case 'r':
718             flags |= BDRV_O_RDWR;
719 
720             if (!strcmp(optarg, "leaks")) {
721                 fix = BDRV_FIX_LEAKS;
722             } else if (!strcmp(optarg, "all")) {
723                 fix = BDRV_FIX_LEAKS | BDRV_FIX_ERRORS;
724             } else {
725                 error_exit("Unknown option value for -r "
726                            "(expecting 'leaks' or 'all'): %s", optarg);
727             }
728             break;
729         case OPTION_OUTPUT:
730             output = optarg;
731             break;
732         case 'T':
733             cache = optarg;
734             break;
735         case 'q':
736             quiet = true;
737             break;
738         case 'U':
739             force_share = true;
740             break;
741         case OPTION_OBJECT: {
742             QemuOpts *opts;
743             opts = qemu_opts_parse_noisily(&qemu_object_opts,
744                                            optarg, true);
745             if (!opts) {
746                 return 1;
747             }
748         }   break;
749         case OPTION_IMAGE_OPTS:
750             image_opts = true;
751             break;
752         }
753     }
754     if (optind != argc - 1) {
755         error_exit("Expecting one image file name");
756     }
757     filename = argv[optind++];
758 
759     if (output && !strcmp(output, "json")) {
760         output_format = OFORMAT_JSON;
761     } else if (output && !strcmp(output, "human")) {
762         output_format = OFORMAT_HUMAN;
763     } else if (output) {
764         error_report("--output must be used with human or json as argument.");
765         return 1;
766     }
767 
768     if (qemu_opts_foreach(&qemu_object_opts,
769                           user_creatable_add_opts_foreach,
770                           NULL, NULL)) {
771         return 1;
772     }
773 
774     ret = bdrv_parse_cache_mode(cache, &flags, &writethrough);
775     if (ret < 0) {
776         error_report("Invalid source cache option: %s", cache);
777         return 1;
778     }
779 
780     blk = img_open(image_opts, filename, fmt, flags, writethrough, quiet,
781                    force_share);
782     if (!blk) {
783         return 1;
784     }
785     bs = blk_bs(blk);
786 
787     check = g_new0(ImageCheck, 1);
788     ret = collect_image_check(bs, check, filename, fmt, fix);
789 
790     if (ret == -ENOTSUP) {
791         error_report("This image format does not support checks");
792         ret = 63;
793         goto fail;
794     }
795 
796     if (check->corruptions_fixed || check->leaks_fixed) {
797         int corruptions_fixed, leaks_fixed;
798 
799         leaks_fixed         = check->leaks_fixed;
800         corruptions_fixed   = check->corruptions_fixed;
801 
802         if (output_format == OFORMAT_HUMAN) {
803             qprintf(quiet,
804                     "The following inconsistencies were found and repaired:\n\n"
805                     "    %" PRId64 " leaked clusters\n"
806                     "    %" PRId64 " corruptions\n\n"
807                     "Double checking the fixed image now...\n",
808                     check->leaks_fixed,
809                     check->corruptions_fixed);
810         }
811 
812         ret = collect_image_check(bs, check, filename, fmt, 0);
813 
814         check->leaks_fixed          = leaks_fixed;
815         check->corruptions_fixed    = corruptions_fixed;
816     }
817 
818     if (!ret) {
819         switch (output_format) {
820         case OFORMAT_HUMAN:
821             dump_human_image_check(check, quiet);
822             break;
823         case OFORMAT_JSON:
824             dump_json_image_check(check, quiet);
825             break;
826         }
827     }
828 
829     if (ret || check->check_errors) {
830         if (ret) {
831             error_report("Check failed: %s", strerror(-ret));
832         } else {
833             error_report("Check failed");
834         }
835         ret = 1;
836         goto fail;
837     }
838 
839     if (check->corruptions) {
840         ret = 2;
841     } else if (check->leaks) {
842         ret = 3;
843     } else {
844         ret = 0;
845     }
846 
847 fail:
848     qapi_free_ImageCheck(check);
849     blk_unref(blk);
850     return ret;
851 }
852 
853 typedef struct CommonBlockJobCBInfo {
854     BlockDriverState *bs;
855     Error **errp;
856 } CommonBlockJobCBInfo;
857 
858 static void common_block_job_cb(void *opaque, int ret)
859 {
860     CommonBlockJobCBInfo *cbi = opaque;
861 
862     if (ret < 0) {
863         error_setg_errno(cbi->errp, -ret, "Block job failed");
864     }
865 }
866 
867 static void run_block_job(BlockJob *job, Error **errp)
868 {
869     AioContext *aio_context = blk_get_aio_context(job->blk);
870     int ret = 0;
871 
872     aio_context_acquire(aio_context);
873     job_ref(&job->job);
874     do {
875         float progress = 0.0f;
876         aio_poll(aio_context, true);
877         if (job->job.progress_total) {
878             progress = (float)job->job.progress_current /
879                        job->job.progress_total * 100.f;
880         }
881         qemu_progress_print(progress, 0);
882     } while (!job_is_ready(&job->job) && !job_is_completed(&job->job));
883 
884     if (!job_is_completed(&job->job)) {
885         ret = job_complete_sync(&job->job, errp);
886     } else {
887         ret = job->job.ret;
888     }
889     job_unref(&job->job);
890     aio_context_release(aio_context);
891 
892     /* publish completion progress only when success */
893     if (!ret) {
894         qemu_progress_print(100.f, 0);
895     }
896 }
897 
898 static int img_commit(int argc, char **argv)
899 {
900     int c, ret, flags;
901     const char *filename, *fmt, *cache, *base;
902     BlockBackend *blk;
903     BlockDriverState *bs, *base_bs;
904     BlockJob *job;
905     bool progress = false, quiet = false, drop = false;
906     bool writethrough;
907     Error *local_err = NULL;
908     CommonBlockJobCBInfo cbi;
909     bool image_opts = false;
910     AioContext *aio_context;
911 
912     fmt = NULL;
913     cache = BDRV_DEFAULT_CACHE;
914     base = NULL;
915     for(;;) {
916         static const struct option long_options[] = {
917             {"help", no_argument, 0, 'h'},
918             {"object", required_argument, 0, OPTION_OBJECT},
919             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
920             {0, 0, 0, 0}
921         };
922         c = getopt_long(argc, argv, ":f:ht:b:dpq",
923                         long_options, NULL);
924         if (c == -1) {
925             break;
926         }
927         switch(c) {
928         case ':':
929             missing_argument(argv[optind - 1]);
930             break;
931         case '?':
932             unrecognized_option(argv[optind - 1]);
933             break;
934         case 'h':
935             help();
936             break;
937         case 'f':
938             fmt = optarg;
939             break;
940         case 't':
941             cache = optarg;
942             break;
943         case 'b':
944             base = optarg;
945             /* -b implies -d */
946             drop = true;
947             break;
948         case 'd':
949             drop = true;
950             break;
951         case 'p':
952             progress = true;
953             break;
954         case 'q':
955             quiet = true;
956             break;
957         case OPTION_OBJECT: {
958             QemuOpts *opts;
959             opts = qemu_opts_parse_noisily(&qemu_object_opts,
960                                            optarg, true);
961             if (!opts) {
962                 return 1;
963             }
964         }   break;
965         case OPTION_IMAGE_OPTS:
966             image_opts = true;
967             break;
968         }
969     }
970 
971     /* Progress is not shown in Quiet mode */
972     if (quiet) {
973         progress = false;
974     }
975 
976     if (optind != argc - 1) {
977         error_exit("Expecting one image file name");
978     }
979     filename = argv[optind++];
980 
981     if (qemu_opts_foreach(&qemu_object_opts,
982                           user_creatable_add_opts_foreach,
983                           NULL, NULL)) {
984         return 1;
985     }
986 
987     flags = BDRV_O_RDWR | BDRV_O_UNMAP;
988     ret = bdrv_parse_cache_mode(cache, &flags, &writethrough);
989     if (ret < 0) {
990         error_report("Invalid cache option: %s", cache);
991         return 1;
992     }
993 
994     blk = img_open(image_opts, filename, fmt, flags, writethrough, quiet,
995                    false);
996     if (!blk) {
997         return 1;
998     }
999     bs = blk_bs(blk);
1000 
1001     qemu_progress_init(progress, 1.f);
1002     qemu_progress_print(0.f, 100);
1003 
1004     if (base) {
1005         base_bs = bdrv_find_backing_image(bs, base);
1006         if (!base_bs) {
1007             error_setg(&local_err,
1008                        "Did not find '%s' in the backing chain of '%s'",
1009                        base, filename);
1010             goto done;
1011         }
1012     } else {
1013         /* This is different from QMP, which by default uses the deepest file in
1014          * the backing chain (i.e., the very base); however, the traditional
1015          * behavior of qemu-img commit is using the immediate backing file. */
1016         base_bs = backing_bs(bs);
1017         if (!base_bs) {
1018             error_setg(&local_err, "Image does not have a backing file");
1019             goto done;
1020         }
1021     }
1022 
1023     cbi = (CommonBlockJobCBInfo){
1024         .errp = &local_err,
1025         .bs   = bs,
1026     };
1027 
1028     aio_context = bdrv_get_aio_context(bs);
1029     aio_context_acquire(aio_context);
1030     commit_active_start("commit", bs, base_bs, JOB_DEFAULT, 0,
1031                         BLOCKDEV_ON_ERROR_REPORT, NULL, common_block_job_cb,
1032                         &cbi, false, &local_err);
1033     aio_context_release(aio_context);
1034     if (local_err) {
1035         goto done;
1036     }
1037 
1038     /* When the block job completes, the BlockBackend reference will point to
1039      * the old backing file. In order to avoid that the top image is already
1040      * deleted, so we can still empty it afterwards, increment the reference
1041      * counter here preemptively. */
1042     if (!drop) {
1043         bdrv_ref(bs);
1044     }
1045 
1046     job = block_job_get("commit");
1047     run_block_job(job, &local_err);
1048     if (local_err) {
1049         goto unref_backing;
1050     }
1051 
1052     if (!drop && bs->drv->bdrv_make_empty) {
1053         ret = bs->drv->bdrv_make_empty(bs);
1054         if (ret) {
1055             error_setg_errno(&local_err, -ret, "Could not empty %s",
1056                              filename);
1057             goto unref_backing;
1058         }
1059     }
1060 
1061 unref_backing:
1062     if (!drop) {
1063         bdrv_unref(bs);
1064     }
1065 
1066 done:
1067     qemu_progress_end();
1068 
1069     blk_unref(blk);
1070 
1071     if (local_err) {
1072         error_report_err(local_err);
1073         return 1;
1074     }
1075 
1076     qprintf(quiet, "Image committed.\n");
1077     return 0;
1078 }
1079 
1080 /*
1081  * Returns -1 if 'buf' contains only zeroes, otherwise the byte index
1082  * of the first sector boundary within buf where the sector contains a
1083  * non-zero byte.  This function is robust to a buffer that is not
1084  * sector-aligned.
1085  */
1086 static int64_t find_nonzero(const uint8_t *buf, int64_t n)
1087 {
1088     int64_t i;
1089     int64_t end = QEMU_ALIGN_DOWN(n, BDRV_SECTOR_SIZE);
1090 
1091     for (i = 0; i < end; i += BDRV_SECTOR_SIZE) {
1092         if (!buffer_is_zero(buf + i, BDRV_SECTOR_SIZE)) {
1093             return i;
1094         }
1095     }
1096     if (i < n && !buffer_is_zero(buf + i, n - end)) {
1097         return i;
1098     }
1099     return -1;
1100 }
1101 
1102 /*
1103  * Returns true iff the first sector pointed to by 'buf' contains at least
1104  * a non-NUL byte.
1105  *
1106  * 'pnum' is set to the number of sectors (including and immediately following
1107  * the first one) that are known to be in the same allocated/unallocated state.
1108  * The function will try to align the end offset to alignment boundaries so
1109  * that the request will at least end aligned and consequtive requests will
1110  * also start at an aligned offset.
1111  */
1112 static int is_allocated_sectors(const uint8_t *buf, int n, int *pnum,
1113                                 int64_t sector_num, int alignment)
1114 {
1115     bool is_zero;
1116     int i, tail;
1117 
1118     if (n <= 0) {
1119         *pnum = 0;
1120         return 0;
1121     }
1122     is_zero = buffer_is_zero(buf, 512);
1123     for(i = 1; i < n; i++) {
1124         buf += 512;
1125         if (is_zero != buffer_is_zero(buf, 512)) {
1126             break;
1127         }
1128     }
1129 
1130     tail = (sector_num + i) & (alignment - 1);
1131     if (tail) {
1132         if (is_zero && i <= tail) {
1133             /* treat unallocated areas which only consist
1134              * of a small tail as allocated. */
1135             is_zero = false;
1136         }
1137         if (!is_zero) {
1138             /* align up end offset of allocated areas. */
1139             i += alignment - tail;
1140             i = MIN(i, n);
1141         } else {
1142             /* align down end offset of zero areas. */
1143             i -= tail;
1144         }
1145     }
1146     *pnum = i;
1147     return !is_zero;
1148 }
1149 
1150 /*
1151  * Like is_allocated_sectors, but if the buffer starts with a used sector,
1152  * up to 'min' consecutive sectors containing zeros are ignored. This avoids
1153  * breaking up write requests for only small sparse areas.
1154  */
1155 static int is_allocated_sectors_min(const uint8_t *buf, int n, int *pnum,
1156     int min, int64_t sector_num, int alignment)
1157 {
1158     int ret;
1159     int num_checked, num_used;
1160 
1161     if (n < min) {
1162         min = n;
1163     }
1164 
1165     ret = is_allocated_sectors(buf, n, pnum, sector_num, alignment);
1166     if (!ret) {
1167         return ret;
1168     }
1169 
1170     num_used = *pnum;
1171     buf += BDRV_SECTOR_SIZE * *pnum;
1172     n -= *pnum;
1173     sector_num += *pnum;
1174     num_checked = num_used;
1175 
1176     while (n > 0) {
1177         ret = is_allocated_sectors(buf, n, pnum, sector_num, alignment);
1178 
1179         buf += BDRV_SECTOR_SIZE * *pnum;
1180         n -= *pnum;
1181         sector_num += *pnum;
1182         num_checked += *pnum;
1183         if (ret) {
1184             num_used = num_checked;
1185         } else if (*pnum >= min) {
1186             break;
1187         }
1188     }
1189 
1190     *pnum = num_used;
1191     return 1;
1192 }
1193 
1194 /*
1195  * Compares two buffers sector by sector. Returns 0 if the first
1196  * sector of each buffer matches, non-zero otherwise.
1197  *
1198  * pnum is set to the sector-aligned size of the buffer prefix that
1199  * has the same matching status as the first sector.
1200  */
1201 static int compare_buffers(const uint8_t *buf1, const uint8_t *buf2,
1202                            int64_t bytes, int64_t *pnum)
1203 {
1204     bool res;
1205     int64_t i = MIN(bytes, BDRV_SECTOR_SIZE);
1206 
1207     assert(bytes > 0);
1208 
1209     res = !!memcmp(buf1, buf2, i);
1210     while (i < bytes) {
1211         int64_t len = MIN(bytes - i, BDRV_SECTOR_SIZE);
1212 
1213         if (!!memcmp(buf1 + i, buf2 + i, len) != res) {
1214             break;
1215         }
1216         i += len;
1217     }
1218 
1219     *pnum = i;
1220     return res;
1221 }
1222 
1223 #define IO_BUF_SIZE (2 * 1024 * 1024)
1224 
1225 /*
1226  * Check if passed sectors are empty (not allocated or contain only 0 bytes)
1227  *
1228  * Intended for use by 'qemu-img compare': Returns 0 in case sectors are
1229  * filled with 0, 1 if sectors contain non-zero data (this is a comparison
1230  * failure), and 4 on error (the exit status for read errors), after emitting
1231  * an error message.
1232  *
1233  * @param blk:  BlockBackend for the image
1234  * @param offset: Starting offset to check
1235  * @param bytes: Number of bytes to check
1236  * @param filename: Name of disk file we are checking (logging purpose)
1237  * @param buffer: Allocated buffer for storing read data
1238  * @param quiet: Flag for quiet mode
1239  */
1240 static int check_empty_sectors(BlockBackend *blk, int64_t offset,
1241                                int64_t bytes, const char *filename,
1242                                uint8_t *buffer, bool quiet)
1243 {
1244     int ret = 0;
1245     int64_t idx;
1246 
1247     ret = blk_pread(blk, offset, buffer, bytes);
1248     if (ret < 0) {
1249         error_report("Error while reading offset %" PRId64 " of %s: %s",
1250                      offset, filename, strerror(-ret));
1251         return 4;
1252     }
1253     idx = find_nonzero(buffer, bytes);
1254     if (idx >= 0) {
1255         qprintf(quiet, "Content mismatch at offset %" PRId64 "!\n",
1256                 offset + idx);
1257         return 1;
1258     }
1259 
1260     return 0;
1261 }
1262 
1263 /*
1264  * Compares two images. Exit codes:
1265  *
1266  * 0 - Images are identical
1267  * 1 - Images differ
1268  * >1 - Error occurred
1269  */
1270 static int img_compare(int argc, char **argv)
1271 {
1272     const char *fmt1 = NULL, *fmt2 = NULL, *cache, *filename1, *filename2;
1273     BlockBackend *blk1, *blk2;
1274     BlockDriverState *bs1, *bs2;
1275     int64_t total_size1, total_size2;
1276     uint8_t *buf1 = NULL, *buf2 = NULL;
1277     int64_t pnum1, pnum2;
1278     int allocated1, allocated2;
1279     int ret = 0; /* return value - 0 Ident, 1 Different, >1 Error */
1280     bool progress = false, quiet = false, strict = false;
1281     int flags;
1282     bool writethrough;
1283     int64_t total_size;
1284     int64_t offset = 0;
1285     int64_t chunk;
1286     int c;
1287     uint64_t progress_base;
1288     bool image_opts = false;
1289     bool force_share = false;
1290 
1291     cache = BDRV_DEFAULT_CACHE;
1292     for (;;) {
1293         static const struct option long_options[] = {
1294             {"help", no_argument, 0, 'h'},
1295             {"object", required_argument, 0, OPTION_OBJECT},
1296             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
1297             {"force-share", no_argument, 0, 'U'},
1298             {0, 0, 0, 0}
1299         };
1300         c = getopt_long(argc, argv, ":hf:F:T:pqsU",
1301                         long_options, NULL);
1302         if (c == -1) {
1303             break;
1304         }
1305         switch (c) {
1306         case ':':
1307             missing_argument(argv[optind - 1]);
1308             break;
1309         case '?':
1310             unrecognized_option(argv[optind - 1]);
1311             break;
1312         case 'h':
1313             help();
1314             break;
1315         case 'f':
1316             fmt1 = optarg;
1317             break;
1318         case 'F':
1319             fmt2 = optarg;
1320             break;
1321         case 'T':
1322             cache = optarg;
1323             break;
1324         case 'p':
1325             progress = true;
1326             break;
1327         case 'q':
1328             quiet = true;
1329             break;
1330         case 's':
1331             strict = true;
1332             break;
1333         case 'U':
1334             force_share = true;
1335             break;
1336         case OPTION_OBJECT: {
1337             QemuOpts *opts;
1338             opts = qemu_opts_parse_noisily(&qemu_object_opts,
1339                                            optarg, true);
1340             if (!opts) {
1341                 ret = 2;
1342                 goto out4;
1343             }
1344         }   break;
1345         case OPTION_IMAGE_OPTS:
1346             image_opts = true;
1347             break;
1348         }
1349     }
1350 
1351     /* Progress is not shown in Quiet mode */
1352     if (quiet) {
1353         progress = false;
1354     }
1355 
1356 
1357     if (optind != argc - 2) {
1358         error_exit("Expecting two image file names");
1359     }
1360     filename1 = argv[optind++];
1361     filename2 = argv[optind++];
1362 
1363     if (qemu_opts_foreach(&qemu_object_opts,
1364                           user_creatable_add_opts_foreach,
1365                           NULL, NULL)) {
1366         ret = 2;
1367         goto out4;
1368     }
1369 
1370     /* Initialize before goto out */
1371     qemu_progress_init(progress, 2.0);
1372 
1373     flags = 0;
1374     ret = bdrv_parse_cache_mode(cache, &flags, &writethrough);
1375     if (ret < 0) {
1376         error_report("Invalid source cache option: %s", cache);
1377         ret = 2;
1378         goto out3;
1379     }
1380 
1381     blk1 = img_open(image_opts, filename1, fmt1, flags, writethrough, quiet,
1382                     force_share);
1383     if (!blk1) {
1384         ret = 2;
1385         goto out3;
1386     }
1387 
1388     blk2 = img_open(image_opts, filename2, fmt2, flags, writethrough, quiet,
1389                     force_share);
1390     if (!blk2) {
1391         ret = 2;
1392         goto out2;
1393     }
1394     bs1 = blk_bs(blk1);
1395     bs2 = blk_bs(blk2);
1396 
1397     buf1 = blk_blockalign(blk1, IO_BUF_SIZE);
1398     buf2 = blk_blockalign(blk2, IO_BUF_SIZE);
1399     total_size1 = blk_getlength(blk1);
1400     if (total_size1 < 0) {
1401         error_report("Can't get size of %s: %s",
1402                      filename1, strerror(-total_size1));
1403         ret = 4;
1404         goto out;
1405     }
1406     total_size2 = blk_getlength(blk2);
1407     if (total_size2 < 0) {
1408         error_report("Can't get size of %s: %s",
1409                      filename2, strerror(-total_size2));
1410         ret = 4;
1411         goto out;
1412     }
1413     total_size = MIN(total_size1, total_size2);
1414     progress_base = MAX(total_size1, total_size2);
1415 
1416     qemu_progress_print(0, 100);
1417 
1418     if (strict && total_size1 != total_size2) {
1419         ret = 1;
1420         qprintf(quiet, "Strict mode: Image size mismatch!\n");
1421         goto out;
1422     }
1423 
1424     while (offset < total_size) {
1425         int status1, status2;
1426 
1427         status1 = bdrv_block_status_above(bs1, NULL, offset,
1428                                           total_size1 - offset, &pnum1, NULL,
1429                                           NULL);
1430         if (status1 < 0) {
1431             ret = 3;
1432             error_report("Sector allocation test failed for %s", filename1);
1433             goto out;
1434         }
1435         allocated1 = status1 & BDRV_BLOCK_ALLOCATED;
1436 
1437         status2 = bdrv_block_status_above(bs2, NULL, offset,
1438                                           total_size2 - offset, &pnum2, NULL,
1439                                           NULL);
1440         if (status2 < 0) {
1441             ret = 3;
1442             error_report("Sector allocation test failed for %s", filename2);
1443             goto out;
1444         }
1445         allocated2 = status2 & BDRV_BLOCK_ALLOCATED;
1446 
1447         assert(pnum1 && pnum2);
1448         chunk = MIN(pnum1, pnum2);
1449 
1450         if (strict) {
1451             if (status1 != status2) {
1452                 ret = 1;
1453                 qprintf(quiet, "Strict mode: Offset %" PRId64
1454                         " block status mismatch!\n", offset);
1455                 goto out;
1456             }
1457         }
1458         if ((status1 & BDRV_BLOCK_ZERO) && (status2 & BDRV_BLOCK_ZERO)) {
1459             /* nothing to do */
1460         } else if (allocated1 == allocated2) {
1461             if (allocated1) {
1462                 int64_t pnum;
1463 
1464                 chunk = MIN(chunk, IO_BUF_SIZE);
1465                 ret = blk_pread(blk1, offset, buf1, chunk);
1466                 if (ret < 0) {
1467                     error_report("Error while reading offset %" PRId64
1468                                  " of %s: %s",
1469                                  offset, filename1, strerror(-ret));
1470                     ret = 4;
1471                     goto out;
1472                 }
1473                 ret = blk_pread(blk2, offset, buf2, chunk);
1474                 if (ret < 0) {
1475                     error_report("Error while reading offset %" PRId64
1476                                  " of %s: %s",
1477                                  offset, filename2, strerror(-ret));
1478                     ret = 4;
1479                     goto out;
1480                 }
1481                 ret = compare_buffers(buf1, buf2, chunk, &pnum);
1482                 if (ret || pnum != chunk) {
1483                     qprintf(quiet, "Content mismatch at offset %" PRId64 "!\n",
1484                             offset + (ret ? 0 : pnum));
1485                     ret = 1;
1486                     goto out;
1487                 }
1488             }
1489         } else {
1490             chunk = MIN(chunk, IO_BUF_SIZE);
1491             if (allocated1) {
1492                 ret = check_empty_sectors(blk1, offset, chunk,
1493                                           filename1, buf1, quiet);
1494             } else {
1495                 ret = check_empty_sectors(blk2, offset, chunk,
1496                                           filename2, buf1, quiet);
1497             }
1498             if (ret) {
1499                 goto out;
1500             }
1501         }
1502         offset += chunk;
1503         qemu_progress_print(((float) chunk / progress_base) * 100, 100);
1504     }
1505 
1506     if (total_size1 != total_size2) {
1507         BlockBackend *blk_over;
1508         const char *filename_over;
1509 
1510         qprintf(quiet, "Warning: Image size mismatch!\n");
1511         if (total_size1 > total_size2) {
1512             blk_over = blk1;
1513             filename_over = filename1;
1514         } else {
1515             blk_over = blk2;
1516             filename_over = filename2;
1517         }
1518 
1519         while (offset < progress_base) {
1520             ret = bdrv_block_status_above(blk_bs(blk_over), NULL, offset,
1521                                           progress_base - offset, &chunk,
1522                                           NULL, NULL);
1523             if (ret < 0) {
1524                 ret = 3;
1525                 error_report("Sector allocation test failed for %s",
1526                              filename_over);
1527                 goto out;
1528 
1529             }
1530             if (ret & BDRV_BLOCK_ALLOCATED && !(ret & BDRV_BLOCK_ZERO)) {
1531                 chunk = MIN(chunk, IO_BUF_SIZE);
1532                 ret = check_empty_sectors(blk_over, offset, chunk,
1533                                           filename_over, buf1, quiet);
1534                 if (ret) {
1535                     goto out;
1536                 }
1537             }
1538             offset += chunk;
1539             qemu_progress_print(((float) chunk / progress_base) * 100, 100);
1540         }
1541     }
1542 
1543     qprintf(quiet, "Images are identical.\n");
1544     ret = 0;
1545 
1546 out:
1547     qemu_vfree(buf1);
1548     qemu_vfree(buf2);
1549     blk_unref(blk2);
1550 out2:
1551     blk_unref(blk1);
1552 out3:
1553     qemu_progress_end();
1554 out4:
1555     return ret;
1556 }
1557 
1558 enum ImgConvertBlockStatus {
1559     BLK_DATA,
1560     BLK_ZERO,
1561     BLK_BACKING_FILE,
1562 };
1563 
1564 #define MAX_COROUTINES 16
1565 
1566 typedef struct ImgConvertState {
1567     BlockBackend **src;
1568     int64_t *src_sectors;
1569     int src_num;
1570     int64_t total_sectors;
1571     int64_t allocated_sectors;
1572     int64_t allocated_done;
1573     int64_t sector_num;
1574     int64_t wr_offs;
1575     enum ImgConvertBlockStatus status;
1576     int64_t sector_next_status;
1577     BlockBackend *target;
1578     bool has_zero_init;
1579     bool compressed;
1580     bool unallocated_blocks_are_zero;
1581     bool target_has_backing;
1582     int64_t target_backing_sectors; /* negative if unknown */
1583     bool wr_in_order;
1584     bool copy_range;
1585     int min_sparse;
1586     int alignment;
1587     size_t cluster_sectors;
1588     size_t buf_sectors;
1589     long num_coroutines;
1590     int running_coroutines;
1591     Coroutine *co[MAX_COROUTINES];
1592     int64_t wait_sector_num[MAX_COROUTINES];
1593     CoMutex lock;
1594     int ret;
1595 } ImgConvertState;
1596 
1597 static void convert_select_part(ImgConvertState *s, int64_t sector_num,
1598                                 int *src_cur, int64_t *src_cur_offset)
1599 {
1600     *src_cur = 0;
1601     *src_cur_offset = 0;
1602     while (sector_num - *src_cur_offset >= s->src_sectors[*src_cur]) {
1603         *src_cur_offset += s->src_sectors[*src_cur];
1604         (*src_cur)++;
1605         assert(*src_cur < s->src_num);
1606     }
1607 }
1608 
1609 static int convert_iteration_sectors(ImgConvertState *s, int64_t sector_num)
1610 {
1611     int64_t src_cur_offset;
1612     int ret, n, src_cur;
1613     bool post_backing_zero = false;
1614 
1615     convert_select_part(s, sector_num, &src_cur, &src_cur_offset);
1616 
1617     assert(s->total_sectors > sector_num);
1618     n = MIN(s->total_sectors - sector_num, BDRV_REQUEST_MAX_SECTORS);
1619 
1620     if (s->target_backing_sectors >= 0) {
1621         if (sector_num >= s->target_backing_sectors) {
1622             post_backing_zero = s->unallocated_blocks_are_zero;
1623         } else if (sector_num + n > s->target_backing_sectors) {
1624             /* Split requests around target_backing_sectors (because
1625              * starting from there, zeros are handled differently) */
1626             n = s->target_backing_sectors - sector_num;
1627         }
1628     }
1629 
1630     if (s->sector_next_status <= sector_num) {
1631         int64_t count = n * BDRV_SECTOR_SIZE;
1632 
1633         if (s->target_has_backing) {
1634 
1635             ret = bdrv_block_status(blk_bs(s->src[src_cur]),
1636                                     (sector_num - src_cur_offset) *
1637                                     BDRV_SECTOR_SIZE,
1638                                     count, &count, NULL, NULL);
1639         } else {
1640             ret = bdrv_block_status_above(blk_bs(s->src[src_cur]), NULL,
1641                                           (sector_num - src_cur_offset) *
1642                                           BDRV_SECTOR_SIZE,
1643                                           count, &count, NULL, NULL);
1644         }
1645         if (ret < 0) {
1646             return ret;
1647         }
1648         n = DIV_ROUND_UP(count, BDRV_SECTOR_SIZE);
1649 
1650         if (ret & BDRV_BLOCK_ZERO) {
1651             s->status = post_backing_zero ? BLK_BACKING_FILE : BLK_ZERO;
1652         } else if (ret & BDRV_BLOCK_DATA) {
1653             s->status = BLK_DATA;
1654         } else {
1655             s->status = s->target_has_backing ? BLK_BACKING_FILE : BLK_DATA;
1656         }
1657 
1658         s->sector_next_status = sector_num + n;
1659     }
1660 
1661     n = MIN(n, s->sector_next_status - sector_num);
1662     if (s->status == BLK_DATA) {
1663         n = MIN(n, s->buf_sectors);
1664     }
1665 
1666     /* We need to write complete clusters for compressed images, so if an
1667      * unallocated area is shorter than that, we must consider the whole
1668      * cluster allocated. */
1669     if (s->compressed) {
1670         if (n < s->cluster_sectors) {
1671             n = MIN(s->cluster_sectors, s->total_sectors - sector_num);
1672             s->status = BLK_DATA;
1673         } else {
1674             n = QEMU_ALIGN_DOWN(n, s->cluster_sectors);
1675         }
1676     }
1677 
1678     return n;
1679 }
1680 
1681 static int coroutine_fn convert_co_read(ImgConvertState *s, int64_t sector_num,
1682                                         int nb_sectors, uint8_t *buf)
1683 {
1684     int n, ret;
1685     QEMUIOVector qiov;
1686     struct iovec iov;
1687 
1688     assert(nb_sectors <= s->buf_sectors);
1689     while (nb_sectors > 0) {
1690         BlockBackend *blk;
1691         int src_cur;
1692         int64_t bs_sectors, src_cur_offset;
1693 
1694         /* In the case of compression with multiple source files, we can get a
1695          * nb_sectors that spreads into the next part. So we must be able to
1696          * read across multiple BDSes for one convert_read() call. */
1697         convert_select_part(s, sector_num, &src_cur, &src_cur_offset);
1698         blk = s->src[src_cur];
1699         bs_sectors = s->src_sectors[src_cur];
1700 
1701         n = MIN(nb_sectors, bs_sectors - (sector_num - src_cur_offset));
1702         iov.iov_base = buf;
1703         iov.iov_len = n << BDRV_SECTOR_BITS;
1704         qemu_iovec_init_external(&qiov, &iov, 1);
1705 
1706         ret = blk_co_preadv(
1707                 blk, (sector_num - src_cur_offset) << BDRV_SECTOR_BITS,
1708                 n << BDRV_SECTOR_BITS, &qiov, 0);
1709         if (ret < 0) {
1710             return ret;
1711         }
1712 
1713         sector_num += n;
1714         nb_sectors -= n;
1715         buf += n * BDRV_SECTOR_SIZE;
1716     }
1717 
1718     return 0;
1719 }
1720 
1721 
1722 static int coroutine_fn convert_co_write(ImgConvertState *s, int64_t sector_num,
1723                                          int nb_sectors, uint8_t *buf,
1724                                          enum ImgConvertBlockStatus status)
1725 {
1726     int ret;
1727     QEMUIOVector qiov;
1728     struct iovec iov;
1729 
1730     while (nb_sectors > 0) {
1731         int n = nb_sectors;
1732         BdrvRequestFlags flags = s->compressed ? BDRV_REQ_WRITE_COMPRESSED : 0;
1733 
1734         switch (status) {
1735         case BLK_BACKING_FILE:
1736             /* If we have a backing file, leave clusters unallocated that are
1737              * unallocated in the source image, so that the backing file is
1738              * visible at the respective offset. */
1739             assert(s->target_has_backing);
1740             break;
1741 
1742         case BLK_DATA:
1743             /* If we're told to keep the target fully allocated (-S 0) or there
1744              * is real non-zero data, we must write it. Otherwise we can treat
1745              * it as zero sectors.
1746              * Compressed clusters need to be written as a whole, so in that
1747              * case we can only save the write if the buffer is completely
1748              * zeroed. */
1749             if (!s->min_sparse ||
1750                 (!s->compressed &&
1751                  is_allocated_sectors_min(buf, n, &n, s->min_sparse,
1752                                           sector_num, s->alignment)) ||
1753                 (s->compressed &&
1754                  !buffer_is_zero(buf, n * BDRV_SECTOR_SIZE)))
1755             {
1756                 iov.iov_base = buf;
1757                 iov.iov_len = n << BDRV_SECTOR_BITS;
1758                 qemu_iovec_init_external(&qiov, &iov, 1);
1759 
1760                 ret = blk_co_pwritev(s->target, sector_num << BDRV_SECTOR_BITS,
1761                                      n << BDRV_SECTOR_BITS, &qiov, flags);
1762                 if (ret < 0) {
1763                     return ret;
1764                 }
1765                 break;
1766             }
1767             /* fall-through */
1768 
1769         case BLK_ZERO:
1770             if (s->has_zero_init) {
1771                 assert(!s->target_has_backing);
1772                 break;
1773             }
1774             ret = blk_co_pwrite_zeroes(s->target,
1775                                        sector_num << BDRV_SECTOR_BITS,
1776                                        n << BDRV_SECTOR_BITS, 0);
1777             if (ret < 0) {
1778                 return ret;
1779             }
1780             break;
1781         }
1782 
1783         sector_num += n;
1784         nb_sectors -= n;
1785         buf += n * BDRV_SECTOR_SIZE;
1786     }
1787 
1788     return 0;
1789 }
1790 
1791 static int coroutine_fn convert_co_copy_range(ImgConvertState *s, int64_t sector_num,
1792                                               int nb_sectors)
1793 {
1794     int n, ret;
1795 
1796     while (nb_sectors > 0) {
1797         BlockBackend *blk;
1798         int src_cur;
1799         int64_t bs_sectors, src_cur_offset;
1800         int64_t offset;
1801 
1802         convert_select_part(s, sector_num, &src_cur, &src_cur_offset);
1803         offset = (sector_num - src_cur_offset) << BDRV_SECTOR_BITS;
1804         blk = s->src[src_cur];
1805         bs_sectors = s->src_sectors[src_cur];
1806 
1807         n = MIN(nb_sectors, bs_sectors - (sector_num - src_cur_offset));
1808 
1809         ret = blk_co_copy_range(blk, offset, s->target,
1810                                 sector_num << BDRV_SECTOR_BITS,
1811                                 n << BDRV_SECTOR_BITS, 0, 0);
1812         if (ret < 0) {
1813             return ret;
1814         }
1815 
1816         sector_num += n;
1817         nb_sectors -= n;
1818     }
1819     return 0;
1820 }
1821 
1822 static void coroutine_fn convert_co_do_copy(void *opaque)
1823 {
1824     ImgConvertState *s = opaque;
1825     uint8_t *buf = NULL;
1826     int ret, i;
1827     int index = -1;
1828 
1829     for (i = 0; i < s->num_coroutines; i++) {
1830         if (s->co[i] == qemu_coroutine_self()) {
1831             index = i;
1832             break;
1833         }
1834     }
1835     assert(index >= 0);
1836 
1837     s->running_coroutines++;
1838     buf = blk_blockalign(s->target, s->buf_sectors * BDRV_SECTOR_SIZE);
1839 
1840     while (1) {
1841         int n;
1842         int64_t sector_num;
1843         enum ImgConvertBlockStatus status;
1844         bool copy_range;
1845 
1846         qemu_co_mutex_lock(&s->lock);
1847         if (s->ret != -EINPROGRESS || s->sector_num >= s->total_sectors) {
1848             qemu_co_mutex_unlock(&s->lock);
1849             break;
1850         }
1851         n = convert_iteration_sectors(s, s->sector_num);
1852         if (n < 0) {
1853             qemu_co_mutex_unlock(&s->lock);
1854             s->ret = n;
1855             break;
1856         }
1857         /* save current sector and allocation status to local variables */
1858         sector_num = s->sector_num;
1859         status = s->status;
1860         if (!s->min_sparse && s->status == BLK_ZERO) {
1861             n = MIN(n, s->buf_sectors);
1862         }
1863         /* increment global sector counter so that other coroutines can
1864          * already continue reading beyond this request */
1865         s->sector_num += n;
1866         qemu_co_mutex_unlock(&s->lock);
1867 
1868         if (status == BLK_DATA || (!s->min_sparse && status == BLK_ZERO)) {
1869             s->allocated_done += n;
1870             qemu_progress_print(100.0 * s->allocated_done /
1871                                         s->allocated_sectors, 0);
1872         }
1873 
1874 retry:
1875         copy_range = s->copy_range && s->status == BLK_DATA;
1876         if (status == BLK_DATA && !copy_range) {
1877             ret = convert_co_read(s, sector_num, n, buf);
1878             if (ret < 0) {
1879                 error_report("error while reading sector %" PRId64
1880                              ": %s", sector_num, strerror(-ret));
1881                 s->ret = ret;
1882             }
1883         } else if (!s->min_sparse && status == BLK_ZERO) {
1884             status = BLK_DATA;
1885             memset(buf, 0x00, n * BDRV_SECTOR_SIZE);
1886         }
1887 
1888         if (s->wr_in_order) {
1889             /* keep writes in order */
1890             while (s->wr_offs != sector_num && s->ret == -EINPROGRESS) {
1891                 s->wait_sector_num[index] = sector_num;
1892                 qemu_coroutine_yield();
1893             }
1894             s->wait_sector_num[index] = -1;
1895         }
1896 
1897         if (s->ret == -EINPROGRESS) {
1898             if (copy_range) {
1899                 ret = convert_co_copy_range(s, sector_num, n);
1900                 if (ret) {
1901                     s->copy_range = false;
1902                     goto retry;
1903                 }
1904             } else {
1905                 ret = convert_co_write(s, sector_num, n, buf, status);
1906             }
1907             if (ret < 0) {
1908                 error_report("error while writing sector %" PRId64
1909                              ": %s", sector_num, strerror(-ret));
1910                 s->ret = ret;
1911             }
1912         }
1913 
1914         if (s->wr_in_order) {
1915             /* reenter the coroutine that might have waited
1916              * for this write to complete */
1917             s->wr_offs = sector_num + n;
1918             for (i = 0; i < s->num_coroutines; i++) {
1919                 if (s->co[i] && s->wait_sector_num[i] == s->wr_offs) {
1920                     /*
1921                      * A -> B -> A cannot occur because A has
1922                      * s->wait_sector_num[i] == -1 during A -> B.  Therefore
1923                      * B will never enter A during this time window.
1924                      */
1925                     qemu_coroutine_enter(s->co[i]);
1926                     break;
1927                 }
1928             }
1929         }
1930     }
1931 
1932     qemu_vfree(buf);
1933     s->co[index] = NULL;
1934     s->running_coroutines--;
1935     if (!s->running_coroutines && s->ret == -EINPROGRESS) {
1936         /* the convert job finished successfully */
1937         s->ret = 0;
1938     }
1939 }
1940 
1941 static int convert_do_copy(ImgConvertState *s)
1942 {
1943     int ret, i, n;
1944     int64_t sector_num = 0;
1945 
1946     /* Check whether we have zero initialisation or can get it efficiently */
1947     s->has_zero_init = s->min_sparse && !s->target_has_backing
1948                      ? bdrv_has_zero_init(blk_bs(s->target))
1949                      : false;
1950 
1951     if (!s->has_zero_init && !s->target_has_backing &&
1952         bdrv_can_write_zeroes_with_unmap(blk_bs(s->target)))
1953     {
1954         ret = blk_make_zero(s->target, BDRV_REQ_MAY_UNMAP);
1955         if (ret == 0) {
1956             s->has_zero_init = true;
1957         }
1958     }
1959 
1960     /* Allocate buffer for copied data. For compressed images, only one cluster
1961      * can be copied at a time. */
1962     if (s->compressed) {
1963         if (s->cluster_sectors <= 0 || s->cluster_sectors > s->buf_sectors) {
1964             error_report("invalid cluster size");
1965             return -EINVAL;
1966         }
1967         s->buf_sectors = s->cluster_sectors;
1968     }
1969 
1970     while (sector_num < s->total_sectors) {
1971         n = convert_iteration_sectors(s, sector_num);
1972         if (n < 0) {
1973             return n;
1974         }
1975         if (s->status == BLK_DATA || (!s->min_sparse && s->status == BLK_ZERO))
1976         {
1977             s->allocated_sectors += n;
1978         }
1979         sector_num += n;
1980     }
1981 
1982     /* Do the copy */
1983     s->sector_next_status = 0;
1984     s->ret = -EINPROGRESS;
1985 
1986     qemu_co_mutex_init(&s->lock);
1987     for (i = 0; i < s->num_coroutines; i++) {
1988         s->co[i] = qemu_coroutine_create(convert_co_do_copy, s);
1989         s->wait_sector_num[i] = -1;
1990         qemu_coroutine_enter(s->co[i]);
1991     }
1992 
1993     while (s->running_coroutines) {
1994         main_loop_wait(false);
1995     }
1996 
1997     if (s->compressed && !s->ret) {
1998         /* signal EOF to align */
1999         ret = blk_pwrite_compressed(s->target, 0, NULL, 0);
2000         if (ret < 0) {
2001             return ret;
2002         }
2003     }
2004 
2005     return s->ret;
2006 }
2007 
2008 #define MAX_BUF_SECTORS 32768
2009 
2010 static int img_convert(int argc, char **argv)
2011 {
2012     int c, bs_i, flags, src_flags = 0;
2013     const char *fmt = NULL, *out_fmt = NULL, *cache = "unsafe",
2014                *src_cache = BDRV_DEFAULT_CACHE, *out_baseimg = NULL,
2015                *out_filename, *out_baseimg_param, *snapshot_name = NULL;
2016     BlockDriver *drv = NULL, *proto_drv = NULL;
2017     BlockDriverInfo bdi;
2018     BlockDriverState *out_bs;
2019     QemuOpts *opts = NULL, *sn_opts = NULL;
2020     QemuOptsList *create_opts = NULL;
2021     char *options = NULL;
2022     Error *local_err = NULL;
2023     bool writethrough, src_writethrough, quiet = false, image_opts = false,
2024          skip_create = false, progress = false, tgt_image_opts = false;
2025     int64_t ret = -EINVAL;
2026     bool force_share = false;
2027 
2028     ImgConvertState s = (ImgConvertState) {
2029         /* Need at least 4k of zeros for sparse detection */
2030         .min_sparse         = 8,
2031         .copy_range         = true,
2032         .buf_sectors        = IO_BUF_SIZE / BDRV_SECTOR_SIZE,
2033         .wr_in_order        = true,
2034         .num_coroutines     = 8,
2035     };
2036 
2037     for(;;) {
2038         static const struct option long_options[] = {
2039             {"help", no_argument, 0, 'h'},
2040             {"object", required_argument, 0, OPTION_OBJECT},
2041             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
2042             {"force-share", no_argument, 0, 'U'},
2043             {"target-image-opts", no_argument, 0, OPTION_TARGET_IMAGE_OPTS},
2044             {0, 0, 0, 0}
2045         };
2046         c = getopt_long(argc, argv, ":hf:O:B:co:l:S:pt:T:qnm:WU",
2047                         long_options, NULL);
2048         if (c == -1) {
2049             break;
2050         }
2051         switch(c) {
2052         case ':':
2053             missing_argument(argv[optind - 1]);
2054             break;
2055         case '?':
2056             unrecognized_option(argv[optind - 1]);
2057             break;
2058         case 'h':
2059             help();
2060             break;
2061         case 'f':
2062             fmt = optarg;
2063             break;
2064         case 'O':
2065             out_fmt = optarg;
2066             break;
2067         case 'B':
2068             out_baseimg = optarg;
2069             break;
2070         case 'c':
2071             s.compressed = true;
2072             s.copy_range = false;
2073             break;
2074         case 'o':
2075             if (!is_valid_option_list(optarg)) {
2076                 error_report("Invalid option list: %s", optarg);
2077                 goto fail_getopt;
2078             }
2079             if (!options) {
2080                 options = g_strdup(optarg);
2081             } else {
2082                 char *old_options = options;
2083                 options = g_strdup_printf("%s,%s", options, optarg);
2084                 g_free(old_options);
2085             }
2086             break;
2087         case 'l':
2088             if (strstart(optarg, SNAPSHOT_OPT_BASE, NULL)) {
2089                 sn_opts = qemu_opts_parse_noisily(&internal_snapshot_opts,
2090                                                   optarg, false);
2091                 if (!sn_opts) {
2092                     error_report("Failed in parsing snapshot param '%s'",
2093                                  optarg);
2094                     goto fail_getopt;
2095                 }
2096             } else {
2097                 snapshot_name = optarg;
2098             }
2099             break;
2100         case 'S':
2101         {
2102             int64_t sval;
2103 
2104             sval = cvtnum(optarg);
2105             if (sval < 0 || sval & (BDRV_SECTOR_SIZE - 1) ||
2106                 sval / BDRV_SECTOR_SIZE > MAX_BUF_SECTORS) {
2107                 error_report("Invalid buffer size for sparse output specified. "
2108                     "Valid sizes are multiples of %llu up to %llu. Select "
2109                     "0 to disable sparse detection (fully allocates output).",
2110                     BDRV_SECTOR_SIZE, MAX_BUF_SECTORS * BDRV_SECTOR_SIZE);
2111                 goto fail_getopt;
2112             }
2113 
2114             s.min_sparse = sval / BDRV_SECTOR_SIZE;
2115             s.copy_range = false;
2116             break;
2117         }
2118         case 'p':
2119             progress = true;
2120             break;
2121         case 't':
2122             cache = optarg;
2123             break;
2124         case 'T':
2125             src_cache = optarg;
2126             break;
2127         case 'q':
2128             quiet = true;
2129             break;
2130         case 'n':
2131             skip_create = true;
2132             break;
2133         case 'm':
2134             if (qemu_strtol(optarg, NULL, 0, &s.num_coroutines) ||
2135                 s.num_coroutines < 1 || s.num_coroutines > MAX_COROUTINES) {
2136                 error_report("Invalid number of coroutines. Allowed number of"
2137                              " coroutines is between 1 and %d", MAX_COROUTINES);
2138                 goto fail_getopt;
2139             }
2140             break;
2141         case 'W':
2142             s.wr_in_order = false;
2143             break;
2144         case 'U':
2145             force_share = true;
2146             break;
2147         case OPTION_OBJECT: {
2148             QemuOpts *object_opts;
2149             object_opts = qemu_opts_parse_noisily(&qemu_object_opts,
2150                                                   optarg, true);
2151             if (!object_opts) {
2152                 goto fail_getopt;
2153             }
2154             break;
2155         }
2156         case OPTION_IMAGE_OPTS:
2157             image_opts = true;
2158             break;
2159         case OPTION_TARGET_IMAGE_OPTS:
2160             tgt_image_opts = true;
2161             break;
2162         }
2163     }
2164 
2165     if (!out_fmt && !tgt_image_opts) {
2166         out_fmt = "raw";
2167     }
2168 
2169     if (qemu_opts_foreach(&qemu_object_opts,
2170                           user_creatable_add_opts_foreach,
2171                           NULL, NULL)) {
2172         goto fail_getopt;
2173     }
2174 
2175     if (tgt_image_opts && !skip_create) {
2176         error_report("--target-image-opts requires use of -n flag");
2177         goto fail_getopt;
2178     }
2179 
2180     s.src_num = argc - optind - 1;
2181     out_filename = s.src_num >= 1 ? argv[argc - 1] : NULL;
2182 
2183     if (options && has_help_option(options)) {
2184         if (out_fmt) {
2185             ret = print_block_option_help(out_filename, out_fmt);
2186             goto fail_getopt;
2187         } else {
2188             error_report("Option help requires a format be specified");
2189             goto fail_getopt;
2190         }
2191     }
2192 
2193     if (s.src_num < 1) {
2194         error_report("Must specify image file name");
2195         goto fail_getopt;
2196     }
2197 
2198 
2199     /* ret is still -EINVAL until here */
2200     ret = bdrv_parse_cache_mode(src_cache, &src_flags, &src_writethrough);
2201     if (ret < 0) {
2202         error_report("Invalid source cache option: %s", src_cache);
2203         goto fail_getopt;
2204     }
2205 
2206     /* Initialize before goto out */
2207     if (quiet) {
2208         progress = false;
2209     }
2210     qemu_progress_init(progress, 1.0);
2211     qemu_progress_print(0, 100);
2212 
2213     s.src = g_new0(BlockBackend *, s.src_num);
2214     s.src_sectors = g_new(int64_t, s.src_num);
2215 
2216     for (bs_i = 0; bs_i < s.src_num; bs_i++) {
2217         s.src[bs_i] = img_open(image_opts, argv[optind + bs_i],
2218                                fmt, src_flags, src_writethrough, quiet,
2219                                force_share);
2220         if (!s.src[bs_i]) {
2221             ret = -1;
2222             goto out;
2223         }
2224         s.src_sectors[bs_i] = blk_nb_sectors(s.src[bs_i]);
2225         if (s.src_sectors[bs_i] < 0) {
2226             error_report("Could not get size of %s: %s",
2227                          argv[optind + bs_i], strerror(-s.src_sectors[bs_i]));
2228             ret = -1;
2229             goto out;
2230         }
2231         s.total_sectors += s.src_sectors[bs_i];
2232     }
2233 
2234     if (sn_opts) {
2235         bdrv_snapshot_load_tmp(blk_bs(s.src[0]),
2236                                qemu_opt_get(sn_opts, SNAPSHOT_OPT_ID),
2237                                qemu_opt_get(sn_opts, SNAPSHOT_OPT_NAME),
2238                                &local_err);
2239     } else if (snapshot_name != NULL) {
2240         if (s.src_num > 1) {
2241             error_report("No support for concatenating multiple snapshot");
2242             ret = -1;
2243             goto out;
2244         }
2245 
2246         bdrv_snapshot_load_tmp_by_id_or_name(blk_bs(s.src[0]), snapshot_name,
2247                                              &local_err);
2248     }
2249     if (local_err) {
2250         error_reportf_err(local_err, "Failed to load snapshot: ");
2251         ret = -1;
2252         goto out;
2253     }
2254 
2255     if (!skip_create) {
2256         /* Find driver and parse its options */
2257         drv = bdrv_find_format(out_fmt);
2258         if (!drv) {
2259             error_report("Unknown file format '%s'", out_fmt);
2260             ret = -1;
2261             goto out;
2262         }
2263 
2264         proto_drv = bdrv_find_protocol(out_filename, true, &local_err);
2265         if (!proto_drv) {
2266             error_report_err(local_err);
2267             ret = -1;
2268             goto out;
2269         }
2270 
2271         if (!drv->create_opts) {
2272             error_report("Format driver '%s' does not support image creation",
2273                          drv->format_name);
2274             ret = -1;
2275             goto out;
2276         }
2277 
2278         if (!proto_drv->create_opts) {
2279             error_report("Protocol driver '%s' does not support image creation",
2280                          proto_drv->format_name);
2281             ret = -1;
2282             goto out;
2283         }
2284 
2285         create_opts = qemu_opts_append(create_opts, drv->create_opts);
2286         create_opts = qemu_opts_append(create_opts, proto_drv->create_opts);
2287 
2288         opts = qemu_opts_create(create_opts, NULL, 0, &error_abort);
2289         if (options) {
2290             qemu_opts_do_parse(opts, options, NULL, &local_err);
2291             if (local_err) {
2292                 error_report_err(local_err);
2293                 ret = -1;
2294                 goto out;
2295             }
2296         }
2297 
2298         qemu_opt_set_number(opts, BLOCK_OPT_SIZE, s.total_sectors * 512,
2299                             &error_abort);
2300         ret = add_old_style_options(out_fmt, opts, out_baseimg, NULL);
2301         if (ret < 0) {
2302             goto out;
2303         }
2304     }
2305 
2306     /* Get backing file name if -o backing_file was used */
2307     out_baseimg_param = qemu_opt_get(opts, BLOCK_OPT_BACKING_FILE);
2308     if (out_baseimg_param) {
2309         out_baseimg = out_baseimg_param;
2310     }
2311     s.target_has_backing = (bool) out_baseimg;
2312 
2313     if (s.src_num > 1 && out_baseimg) {
2314         error_report("Having a backing file for the target makes no sense when "
2315                      "concatenating multiple input images");
2316         ret = -1;
2317         goto out;
2318     }
2319 
2320     /* Check if compression is supported */
2321     if (s.compressed) {
2322         bool encryption =
2323             qemu_opt_get_bool(opts, BLOCK_OPT_ENCRYPT, false);
2324         const char *encryptfmt =
2325             qemu_opt_get(opts, BLOCK_OPT_ENCRYPT_FORMAT);
2326         const char *preallocation =
2327             qemu_opt_get(opts, BLOCK_OPT_PREALLOC);
2328 
2329         if (drv && !drv->bdrv_co_pwritev_compressed) {
2330             error_report("Compression not supported for this file format");
2331             ret = -1;
2332             goto out;
2333         }
2334 
2335         if (encryption || encryptfmt) {
2336             error_report("Compression and encryption not supported at "
2337                          "the same time");
2338             ret = -1;
2339             goto out;
2340         }
2341 
2342         if (preallocation
2343             && strcmp(preallocation, "off"))
2344         {
2345             error_report("Compression and preallocation not supported at "
2346                          "the same time");
2347             ret = -1;
2348             goto out;
2349         }
2350     }
2351 
2352     if (!skip_create) {
2353         /* Create the new image */
2354         ret = bdrv_create(drv, out_filename, opts, &local_err);
2355         if (ret < 0) {
2356             error_reportf_err(local_err, "%s: error while converting %s: ",
2357                               out_filename, out_fmt);
2358             goto out;
2359         }
2360     }
2361 
2362     flags = s.min_sparse ? (BDRV_O_RDWR | BDRV_O_UNMAP) : BDRV_O_RDWR;
2363     ret = bdrv_parse_cache_mode(cache, &flags, &writethrough);
2364     if (ret < 0) {
2365         error_report("Invalid cache option: %s", cache);
2366         goto out;
2367     }
2368 
2369     if (skip_create) {
2370         s.target = img_open(tgt_image_opts, out_filename, out_fmt,
2371                             flags, writethrough, quiet, false);
2372     } else {
2373         /* TODO ultimately we should allow --target-image-opts
2374          * to be used even when -n is not given.
2375          * That has to wait for bdrv_create to be improved
2376          * to allow filenames in option syntax
2377          */
2378         s.target = img_open_new_file(out_filename, opts, out_fmt,
2379                                      flags, writethrough, quiet, false);
2380     }
2381     if (!s.target) {
2382         ret = -1;
2383         goto out;
2384     }
2385     out_bs = blk_bs(s.target);
2386 
2387     if (s.compressed && !out_bs->drv->bdrv_co_pwritev_compressed) {
2388         error_report("Compression not supported for this file format");
2389         ret = -1;
2390         goto out;
2391     }
2392 
2393     /* increase bufsectors from the default 4096 (2M) if opt_transfer
2394      * or discard_alignment of the out_bs is greater. Limit to
2395      * MAX_BUF_SECTORS as maximum which is currently 32768 (16MB). */
2396     s.buf_sectors = MIN(MAX_BUF_SECTORS,
2397                         MAX(s.buf_sectors,
2398                             MAX(out_bs->bl.opt_transfer >> BDRV_SECTOR_BITS,
2399                                 out_bs->bl.pdiscard_alignment >>
2400                                 BDRV_SECTOR_BITS)));
2401 
2402     /* try to align the write requests to the destination to avoid unnecessary
2403      * RMW cycles. */
2404     s.alignment = MAX(pow2floor(s.min_sparse),
2405                       DIV_ROUND_UP(out_bs->bl.request_alignment,
2406                                    BDRV_SECTOR_SIZE));
2407     assert(is_power_of_2(s.alignment));
2408 
2409     if (skip_create) {
2410         int64_t output_sectors = blk_nb_sectors(s.target);
2411         if (output_sectors < 0) {
2412             error_report("unable to get output image length: %s",
2413                          strerror(-output_sectors));
2414             ret = -1;
2415             goto out;
2416         } else if (output_sectors < s.total_sectors) {
2417             error_report("output file is smaller than input file");
2418             ret = -1;
2419             goto out;
2420         }
2421     }
2422 
2423     if (s.target_has_backing) {
2424         /* Errors are treated as "backing length unknown" (which means
2425          * s.target_backing_sectors has to be negative, which it will
2426          * be automatically).  The backing file length is used only
2427          * for optimizations, so such a case is not fatal. */
2428         s.target_backing_sectors = bdrv_nb_sectors(out_bs->backing->bs);
2429     } else {
2430         s.target_backing_sectors = -1;
2431     }
2432 
2433     ret = bdrv_get_info(out_bs, &bdi);
2434     if (ret < 0) {
2435         if (s.compressed) {
2436             error_report("could not get block driver info");
2437             goto out;
2438         }
2439     } else {
2440         s.compressed = s.compressed || bdi.needs_compressed_writes;
2441         s.cluster_sectors = bdi.cluster_size / BDRV_SECTOR_SIZE;
2442         s.unallocated_blocks_are_zero = bdi.unallocated_blocks_are_zero;
2443     }
2444 
2445     ret = convert_do_copy(&s);
2446 out:
2447     if (!ret) {
2448         qemu_progress_print(100, 0);
2449     }
2450     qemu_progress_end();
2451     qemu_opts_del(opts);
2452     qemu_opts_free(create_opts);
2453     qemu_opts_del(sn_opts);
2454     blk_unref(s.target);
2455     if (s.src) {
2456         for (bs_i = 0; bs_i < s.src_num; bs_i++) {
2457             blk_unref(s.src[bs_i]);
2458         }
2459         g_free(s.src);
2460     }
2461     g_free(s.src_sectors);
2462 fail_getopt:
2463     g_free(options);
2464 
2465     return !!ret;
2466 }
2467 
2468 
2469 static void dump_snapshots(BlockDriverState *bs)
2470 {
2471     QEMUSnapshotInfo *sn_tab, *sn;
2472     int nb_sns, i;
2473 
2474     nb_sns = bdrv_snapshot_list(bs, &sn_tab);
2475     if (nb_sns <= 0)
2476         return;
2477     printf("Snapshot list:\n");
2478     bdrv_snapshot_dump(fprintf, stdout, NULL);
2479     printf("\n");
2480     for(i = 0; i < nb_sns; i++) {
2481         sn = &sn_tab[i];
2482         bdrv_snapshot_dump(fprintf, stdout, sn);
2483         printf("\n");
2484     }
2485     g_free(sn_tab);
2486 }
2487 
2488 static void dump_json_image_info_list(ImageInfoList *list)
2489 {
2490     QString *str;
2491     QObject *obj;
2492     Visitor *v = qobject_output_visitor_new(&obj);
2493 
2494     visit_type_ImageInfoList(v, NULL, &list, &error_abort);
2495     visit_complete(v, &obj);
2496     str = qobject_to_json_pretty(obj);
2497     assert(str != NULL);
2498     printf("%s\n", qstring_get_str(str));
2499     qobject_unref(obj);
2500     visit_free(v);
2501     qobject_unref(str);
2502 }
2503 
2504 static void dump_json_image_info(ImageInfo *info)
2505 {
2506     QString *str;
2507     QObject *obj;
2508     Visitor *v = qobject_output_visitor_new(&obj);
2509 
2510     visit_type_ImageInfo(v, NULL, &info, &error_abort);
2511     visit_complete(v, &obj);
2512     str = qobject_to_json_pretty(obj);
2513     assert(str != NULL);
2514     printf("%s\n", qstring_get_str(str));
2515     qobject_unref(obj);
2516     visit_free(v);
2517     qobject_unref(str);
2518 }
2519 
2520 static void dump_human_image_info_list(ImageInfoList *list)
2521 {
2522     ImageInfoList *elem;
2523     bool delim = false;
2524 
2525     for (elem = list; elem; elem = elem->next) {
2526         if (delim) {
2527             printf("\n");
2528         }
2529         delim = true;
2530 
2531         bdrv_image_info_dump(fprintf, stdout, elem->value);
2532     }
2533 }
2534 
2535 static gboolean str_equal_func(gconstpointer a, gconstpointer b)
2536 {
2537     return strcmp(a, b) == 0;
2538 }
2539 
2540 /**
2541  * Open an image file chain and return an ImageInfoList
2542  *
2543  * @filename: topmost image filename
2544  * @fmt: topmost image format (may be NULL to autodetect)
2545  * @chain: true  - enumerate entire backing file chain
2546  *         false - only topmost image file
2547  *
2548  * Returns a list of ImageInfo objects or NULL if there was an error opening an
2549  * image file.  If there was an error a message will have been printed to
2550  * stderr.
2551  */
2552 static ImageInfoList *collect_image_info_list(bool image_opts,
2553                                               const char *filename,
2554                                               const char *fmt,
2555                                               bool chain, bool force_share)
2556 {
2557     ImageInfoList *head = NULL;
2558     ImageInfoList **last = &head;
2559     GHashTable *filenames;
2560     Error *err = NULL;
2561 
2562     filenames = g_hash_table_new_full(g_str_hash, str_equal_func, NULL, NULL);
2563 
2564     while (filename) {
2565         BlockBackend *blk;
2566         BlockDriverState *bs;
2567         ImageInfo *info;
2568         ImageInfoList *elem;
2569 
2570         if (g_hash_table_lookup_extended(filenames, filename, NULL, NULL)) {
2571             error_report("Backing file '%s' creates an infinite loop.",
2572                          filename);
2573             goto err;
2574         }
2575         g_hash_table_insert(filenames, (gpointer)filename, NULL);
2576 
2577         blk = img_open(image_opts, filename, fmt,
2578                        BDRV_O_NO_BACKING | BDRV_O_NO_IO, false, false,
2579                        force_share);
2580         if (!blk) {
2581             goto err;
2582         }
2583         bs = blk_bs(blk);
2584 
2585         bdrv_query_image_info(bs, &info, &err);
2586         if (err) {
2587             error_report_err(err);
2588             blk_unref(blk);
2589             goto err;
2590         }
2591 
2592         elem = g_new0(ImageInfoList, 1);
2593         elem->value = info;
2594         *last = elem;
2595         last = &elem->next;
2596 
2597         blk_unref(blk);
2598 
2599         filename = fmt = NULL;
2600         if (chain) {
2601             if (info->has_full_backing_filename) {
2602                 filename = info->full_backing_filename;
2603             } else if (info->has_backing_filename) {
2604                 error_report("Could not determine absolute backing filename,"
2605                              " but backing filename '%s' present",
2606                              info->backing_filename);
2607                 goto err;
2608             }
2609             if (info->has_backing_filename_format) {
2610                 fmt = info->backing_filename_format;
2611             }
2612         }
2613     }
2614     g_hash_table_destroy(filenames);
2615     return head;
2616 
2617 err:
2618     qapi_free_ImageInfoList(head);
2619     g_hash_table_destroy(filenames);
2620     return NULL;
2621 }
2622 
2623 static int img_info(int argc, char **argv)
2624 {
2625     int c;
2626     OutputFormat output_format = OFORMAT_HUMAN;
2627     bool chain = false;
2628     const char *filename, *fmt, *output;
2629     ImageInfoList *list;
2630     bool image_opts = false;
2631     bool force_share = false;
2632 
2633     fmt = NULL;
2634     output = NULL;
2635     for(;;) {
2636         int option_index = 0;
2637         static const struct option long_options[] = {
2638             {"help", no_argument, 0, 'h'},
2639             {"format", required_argument, 0, 'f'},
2640             {"output", required_argument, 0, OPTION_OUTPUT},
2641             {"backing-chain", no_argument, 0, OPTION_BACKING_CHAIN},
2642             {"object", required_argument, 0, OPTION_OBJECT},
2643             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
2644             {"force-share", no_argument, 0, 'U'},
2645             {0, 0, 0, 0}
2646         };
2647         c = getopt_long(argc, argv, ":f:hU",
2648                         long_options, &option_index);
2649         if (c == -1) {
2650             break;
2651         }
2652         switch(c) {
2653         case ':':
2654             missing_argument(argv[optind - 1]);
2655             break;
2656         case '?':
2657             unrecognized_option(argv[optind - 1]);
2658             break;
2659         case 'h':
2660             help();
2661             break;
2662         case 'f':
2663             fmt = optarg;
2664             break;
2665         case 'U':
2666             force_share = true;
2667             break;
2668         case OPTION_OUTPUT:
2669             output = optarg;
2670             break;
2671         case OPTION_BACKING_CHAIN:
2672             chain = true;
2673             break;
2674         case OPTION_OBJECT: {
2675             QemuOpts *opts;
2676             opts = qemu_opts_parse_noisily(&qemu_object_opts,
2677                                            optarg, true);
2678             if (!opts) {
2679                 return 1;
2680             }
2681         }   break;
2682         case OPTION_IMAGE_OPTS:
2683             image_opts = true;
2684             break;
2685         }
2686     }
2687     if (optind != argc - 1) {
2688         error_exit("Expecting one image file name");
2689     }
2690     filename = argv[optind++];
2691 
2692     if (output && !strcmp(output, "json")) {
2693         output_format = OFORMAT_JSON;
2694     } else if (output && !strcmp(output, "human")) {
2695         output_format = OFORMAT_HUMAN;
2696     } else if (output) {
2697         error_report("--output must be used with human or json as argument.");
2698         return 1;
2699     }
2700 
2701     if (qemu_opts_foreach(&qemu_object_opts,
2702                           user_creatable_add_opts_foreach,
2703                           NULL, NULL)) {
2704         return 1;
2705     }
2706 
2707     list = collect_image_info_list(image_opts, filename, fmt, chain,
2708                                    force_share);
2709     if (!list) {
2710         return 1;
2711     }
2712 
2713     switch (output_format) {
2714     case OFORMAT_HUMAN:
2715         dump_human_image_info_list(list);
2716         break;
2717     case OFORMAT_JSON:
2718         if (chain) {
2719             dump_json_image_info_list(list);
2720         } else {
2721             dump_json_image_info(list->value);
2722         }
2723         break;
2724     }
2725 
2726     qapi_free_ImageInfoList(list);
2727     return 0;
2728 }
2729 
2730 static void dump_map_entry(OutputFormat output_format, MapEntry *e,
2731                            MapEntry *next)
2732 {
2733     switch (output_format) {
2734     case OFORMAT_HUMAN:
2735         if (e->data && !e->has_offset) {
2736             error_report("File contains external, encrypted or compressed clusters.");
2737             exit(1);
2738         }
2739         if (e->data && !e->zero) {
2740             printf("%#-16"PRIx64"%#-16"PRIx64"%#-16"PRIx64"%s\n",
2741                    e->start, e->length,
2742                    e->has_offset ? e->offset : 0,
2743                    e->has_filename ? e->filename : "");
2744         }
2745         /* This format ignores the distinction between 0, ZERO and ZERO|DATA.
2746          * Modify the flags here to allow more coalescing.
2747          */
2748         if (next && (!next->data || next->zero)) {
2749             next->data = false;
2750             next->zero = true;
2751         }
2752         break;
2753     case OFORMAT_JSON:
2754         printf("%s{ \"start\": %"PRId64", \"length\": %"PRId64","
2755                " \"depth\": %"PRId64", \"zero\": %s, \"data\": %s",
2756                (e->start == 0 ? "[" : ",\n"),
2757                e->start, e->length, e->depth,
2758                e->zero ? "true" : "false",
2759                e->data ? "true" : "false");
2760         if (e->has_offset) {
2761             printf(", \"offset\": %"PRId64"", e->offset);
2762         }
2763         putchar('}');
2764 
2765         if (!next) {
2766             printf("]\n");
2767         }
2768         break;
2769     }
2770 }
2771 
2772 static int get_block_status(BlockDriverState *bs, int64_t offset,
2773                             int64_t bytes, MapEntry *e)
2774 {
2775     int ret;
2776     int depth;
2777     BlockDriverState *file;
2778     bool has_offset;
2779     int64_t map;
2780 
2781     /* As an optimization, we could cache the current range of unallocated
2782      * clusters in each file of the chain, and avoid querying the same
2783      * range repeatedly.
2784      */
2785 
2786     depth = 0;
2787     for (;;) {
2788         ret = bdrv_block_status(bs, offset, bytes, &bytes, &map, &file);
2789         if (ret < 0) {
2790             return ret;
2791         }
2792         assert(bytes);
2793         if (ret & (BDRV_BLOCK_ZERO|BDRV_BLOCK_DATA)) {
2794             break;
2795         }
2796         bs = backing_bs(bs);
2797         if (bs == NULL) {
2798             ret = 0;
2799             break;
2800         }
2801 
2802         depth++;
2803     }
2804 
2805     has_offset = !!(ret & BDRV_BLOCK_OFFSET_VALID);
2806 
2807     *e = (MapEntry) {
2808         .start = offset,
2809         .length = bytes,
2810         .data = !!(ret & BDRV_BLOCK_DATA),
2811         .zero = !!(ret & BDRV_BLOCK_ZERO),
2812         .offset = map,
2813         .has_offset = has_offset,
2814         .depth = depth,
2815         .has_filename = file && has_offset,
2816         .filename = file && has_offset ? file->filename : NULL,
2817     };
2818 
2819     return 0;
2820 }
2821 
2822 static inline bool entry_mergeable(const MapEntry *curr, const MapEntry *next)
2823 {
2824     if (curr->length == 0) {
2825         return false;
2826     }
2827     if (curr->zero != next->zero ||
2828         curr->data != next->data ||
2829         curr->depth != next->depth ||
2830         curr->has_filename != next->has_filename ||
2831         curr->has_offset != next->has_offset) {
2832         return false;
2833     }
2834     if (curr->has_filename && strcmp(curr->filename, next->filename)) {
2835         return false;
2836     }
2837     if (curr->has_offset && curr->offset + curr->length != next->offset) {
2838         return false;
2839     }
2840     return true;
2841 }
2842 
2843 static int img_map(int argc, char **argv)
2844 {
2845     int c;
2846     OutputFormat output_format = OFORMAT_HUMAN;
2847     BlockBackend *blk;
2848     BlockDriverState *bs;
2849     const char *filename, *fmt, *output;
2850     int64_t length;
2851     MapEntry curr = { .length = 0 }, next;
2852     int ret = 0;
2853     bool image_opts = false;
2854     bool force_share = false;
2855 
2856     fmt = NULL;
2857     output = NULL;
2858     for (;;) {
2859         int option_index = 0;
2860         static const struct option long_options[] = {
2861             {"help", no_argument, 0, 'h'},
2862             {"format", required_argument, 0, 'f'},
2863             {"output", required_argument, 0, OPTION_OUTPUT},
2864             {"object", required_argument, 0, OPTION_OBJECT},
2865             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
2866             {"force-share", no_argument, 0, 'U'},
2867             {0, 0, 0, 0}
2868         };
2869         c = getopt_long(argc, argv, ":f:hU",
2870                         long_options, &option_index);
2871         if (c == -1) {
2872             break;
2873         }
2874         switch (c) {
2875         case ':':
2876             missing_argument(argv[optind - 1]);
2877             break;
2878         case '?':
2879             unrecognized_option(argv[optind - 1]);
2880             break;
2881         case 'h':
2882             help();
2883             break;
2884         case 'f':
2885             fmt = optarg;
2886             break;
2887         case 'U':
2888             force_share = true;
2889             break;
2890         case OPTION_OUTPUT:
2891             output = optarg;
2892             break;
2893         case OPTION_OBJECT: {
2894             QemuOpts *opts;
2895             opts = qemu_opts_parse_noisily(&qemu_object_opts,
2896                                            optarg, true);
2897             if (!opts) {
2898                 return 1;
2899             }
2900         }   break;
2901         case OPTION_IMAGE_OPTS:
2902             image_opts = true;
2903             break;
2904         }
2905     }
2906     if (optind != argc - 1) {
2907         error_exit("Expecting one image file name");
2908     }
2909     filename = argv[optind];
2910 
2911     if (output && !strcmp(output, "json")) {
2912         output_format = OFORMAT_JSON;
2913     } else if (output && !strcmp(output, "human")) {
2914         output_format = OFORMAT_HUMAN;
2915     } else if (output) {
2916         error_report("--output must be used with human or json as argument.");
2917         return 1;
2918     }
2919 
2920     if (qemu_opts_foreach(&qemu_object_opts,
2921                           user_creatable_add_opts_foreach,
2922                           NULL, NULL)) {
2923         return 1;
2924     }
2925 
2926     blk = img_open(image_opts, filename, fmt, 0, false, false, force_share);
2927     if (!blk) {
2928         return 1;
2929     }
2930     bs = blk_bs(blk);
2931 
2932     if (output_format == OFORMAT_HUMAN) {
2933         printf("%-16s%-16s%-16s%s\n", "Offset", "Length", "Mapped to", "File");
2934     }
2935 
2936     length = blk_getlength(blk);
2937     while (curr.start + curr.length < length) {
2938         int64_t offset = curr.start + curr.length;
2939         int64_t n;
2940 
2941         /* Probe up to 1 GiB at a time.  */
2942         n = MIN(1 << 30, length - offset);
2943         ret = get_block_status(bs, offset, n, &next);
2944 
2945         if (ret < 0) {
2946             error_report("Could not read file metadata: %s", strerror(-ret));
2947             goto out;
2948         }
2949 
2950         if (entry_mergeable(&curr, &next)) {
2951             curr.length += next.length;
2952             continue;
2953         }
2954 
2955         if (curr.length > 0) {
2956             dump_map_entry(output_format, &curr, &next);
2957         }
2958         curr = next;
2959     }
2960 
2961     dump_map_entry(output_format, &curr, NULL);
2962 
2963 out:
2964     blk_unref(blk);
2965     return ret < 0;
2966 }
2967 
2968 #define SNAPSHOT_LIST   1
2969 #define SNAPSHOT_CREATE 2
2970 #define SNAPSHOT_APPLY  3
2971 #define SNAPSHOT_DELETE 4
2972 
2973 static int img_snapshot(int argc, char **argv)
2974 {
2975     BlockBackend *blk;
2976     BlockDriverState *bs;
2977     QEMUSnapshotInfo sn;
2978     char *filename, *snapshot_name = NULL;
2979     int c, ret = 0, bdrv_oflags;
2980     int action = 0;
2981     qemu_timeval tv;
2982     bool quiet = false;
2983     Error *err = NULL;
2984     bool image_opts = false;
2985     bool force_share = false;
2986 
2987     bdrv_oflags = BDRV_O_RDWR;
2988     /* Parse commandline parameters */
2989     for(;;) {
2990         static const struct option long_options[] = {
2991             {"help", no_argument, 0, 'h'},
2992             {"object", required_argument, 0, OPTION_OBJECT},
2993             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
2994             {"force-share", no_argument, 0, 'U'},
2995             {0, 0, 0, 0}
2996         };
2997         c = getopt_long(argc, argv, ":la:c:d:hqU",
2998                         long_options, NULL);
2999         if (c == -1) {
3000             break;
3001         }
3002         switch(c) {
3003         case ':':
3004             missing_argument(argv[optind - 1]);
3005             break;
3006         case '?':
3007             unrecognized_option(argv[optind - 1]);
3008             break;
3009         case 'h':
3010             help();
3011             return 0;
3012         case 'l':
3013             if (action) {
3014                 error_exit("Cannot mix '-l', '-a', '-c', '-d'");
3015                 return 0;
3016             }
3017             action = SNAPSHOT_LIST;
3018             bdrv_oflags &= ~BDRV_O_RDWR; /* no need for RW */
3019             break;
3020         case 'a':
3021             if (action) {
3022                 error_exit("Cannot mix '-l', '-a', '-c', '-d'");
3023                 return 0;
3024             }
3025             action = SNAPSHOT_APPLY;
3026             snapshot_name = optarg;
3027             break;
3028         case 'c':
3029             if (action) {
3030                 error_exit("Cannot mix '-l', '-a', '-c', '-d'");
3031                 return 0;
3032             }
3033             action = SNAPSHOT_CREATE;
3034             snapshot_name = optarg;
3035             break;
3036         case 'd':
3037             if (action) {
3038                 error_exit("Cannot mix '-l', '-a', '-c', '-d'");
3039                 return 0;
3040             }
3041             action = SNAPSHOT_DELETE;
3042             snapshot_name = optarg;
3043             break;
3044         case 'q':
3045             quiet = true;
3046             break;
3047         case 'U':
3048             force_share = true;
3049             break;
3050         case OPTION_OBJECT: {
3051             QemuOpts *opts;
3052             opts = qemu_opts_parse_noisily(&qemu_object_opts,
3053                                            optarg, true);
3054             if (!opts) {
3055                 return 1;
3056             }
3057         }   break;
3058         case OPTION_IMAGE_OPTS:
3059             image_opts = true;
3060             break;
3061         }
3062     }
3063 
3064     if (optind != argc - 1) {
3065         error_exit("Expecting one image file name");
3066     }
3067     filename = argv[optind++];
3068 
3069     if (qemu_opts_foreach(&qemu_object_opts,
3070                           user_creatable_add_opts_foreach,
3071                           NULL, NULL)) {
3072         return 1;
3073     }
3074 
3075     /* Open the image */
3076     blk = img_open(image_opts, filename, NULL, bdrv_oflags, false, quiet,
3077                    force_share);
3078     if (!blk) {
3079         return 1;
3080     }
3081     bs = blk_bs(blk);
3082 
3083     /* Perform the requested action */
3084     switch(action) {
3085     case SNAPSHOT_LIST:
3086         dump_snapshots(bs);
3087         break;
3088 
3089     case SNAPSHOT_CREATE:
3090         memset(&sn, 0, sizeof(sn));
3091         pstrcpy(sn.name, sizeof(sn.name), snapshot_name);
3092 
3093         qemu_gettimeofday(&tv);
3094         sn.date_sec = tv.tv_sec;
3095         sn.date_nsec = tv.tv_usec * 1000;
3096 
3097         ret = bdrv_snapshot_create(bs, &sn);
3098         if (ret) {
3099             error_report("Could not create snapshot '%s': %d (%s)",
3100                 snapshot_name, ret, strerror(-ret));
3101         }
3102         break;
3103 
3104     case SNAPSHOT_APPLY:
3105         ret = bdrv_snapshot_goto(bs, snapshot_name, &err);
3106         if (ret) {
3107             error_reportf_err(err, "Could not apply snapshot '%s': ",
3108                               snapshot_name);
3109         }
3110         break;
3111 
3112     case SNAPSHOT_DELETE:
3113         bdrv_snapshot_delete_by_id_or_name(bs, snapshot_name, &err);
3114         if (err) {
3115             error_reportf_err(err, "Could not delete snapshot '%s': ",
3116                               snapshot_name);
3117             ret = 1;
3118         }
3119         break;
3120     }
3121 
3122     /* Cleanup */
3123     blk_unref(blk);
3124     if (ret) {
3125         return 1;
3126     }
3127     return 0;
3128 }
3129 
3130 static int img_rebase(int argc, char **argv)
3131 {
3132     BlockBackend *blk = NULL, *blk_old_backing = NULL, *blk_new_backing = NULL;
3133     uint8_t *buf_old = NULL;
3134     uint8_t *buf_new = NULL;
3135     BlockDriverState *bs = NULL;
3136     char *filename;
3137     const char *fmt, *cache, *src_cache, *out_basefmt, *out_baseimg;
3138     int c, flags, src_flags, ret;
3139     bool writethrough, src_writethrough;
3140     int unsafe = 0;
3141     bool force_share = false;
3142     int progress = 0;
3143     bool quiet = false;
3144     Error *local_err = NULL;
3145     bool image_opts = false;
3146 
3147     /* Parse commandline parameters */
3148     fmt = NULL;
3149     cache = BDRV_DEFAULT_CACHE;
3150     src_cache = BDRV_DEFAULT_CACHE;
3151     out_baseimg = NULL;
3152     out_basefmt = NULL;
3153     for(;;) {
3154         static const struct option long_options[] = {
3155             {"help", no_argument, 0, 'h'},
3156             {"object", required_argument, 0, OPTION_OBJECT},
3157             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
3158             {"force-share", no_argument, 0, 'U'},
3159             {0, 0, 0, 0}
3160         };
3161         c = getopt_long(argc, argv, ":hf:F:b:upt:T:qU",
3162                         long_options, NULL);
3163         if (c == -1) {
3164             break;
3165         }
3166         switch(c) {
3167         case ':':
3168             missing_argument(argv[optind - 1]);
3169             break;
3170         case '?':
3171             unrecognized_option(argv[optind - 1]);
3172             break;
3173         case 'h':
3174             help();
3175             return 0;
3176         case 'f':
3177             fmt = optarg;
3178             break;
3179         case 'F':
3180             out_basefmt = optarg;
3181             break;
3182         case 'b':
3183             out_baseimg = optarg;
3184             break;
3185         case 'u':
3186             unsafe = 1;
3187             break;
3188         case 'p':
3189             progress = 1;
3190             break;
3191         case 't':
3192             cache = optarg;
3193             break;
3194         case 'T':
3195             src_cache = optarg;
3196             break;
3197         case 'q':
3198             quiet = true;
3199             break;
3200         case OPTION_OBJECT: {
3201             QemuOpts *opts;
3202             opts = qemu_opts_parse_noisily(&qemu_object_opts,
3203                                            optarg, true);
3204             if (!opts) {
3205                 return 1;
3206             }
3207         }   break;
3208         case OPTION_IMAGE_OPTS:
3209             image_opts = true;
3210             break;
3211         case 'U':
3212             force_share = true;
3213             break;
3214         }
3215     }
3216 
3217     if (quiet) {
3218         progress = 0;
3219     }
3220 
3221     if (optind != argc - 1) {
3222         error_exit("Expecting one image file name");
3223     }
3224     if (!unsafe && !out_baseimg) {
3225         error_exit("Must specify backing file (-b) or use unsafe mode (-u)");
3226     }
3227     filename = argv[optind++];
3228 
3229     if (qemu_opts_foreach(&qemu_object_opts,
3230                           user_creatable_add_opts_foreach,
3231                           NULL, NULL)) {
3232         return 1;
3233     }
3234 
3235     qemu_progress_init(progress, 2.0);
3236     qemu_progress_print(0, 100);
3237 
3238     flags = BDRV_O_RDWR | (unsafe ? BDRV_O_NO_BACKING : 0);
3239     ret = bdrv_parse_cache_mode(cache, &flags, &writethrough);
3240     if (ret < 0) {
3241         error_report("Invalid cache option: %s", cache);
3242         goto out;
3243     }
3244 
3245     src_flags = 0;
3246     ret = bdrv_parse_cache_mode(src_cache, &src_flags, &src_writethrough);
3247     if (ret < 0) {
3248         error_report("Invalid source cache option: %s", src_cache);
3249         goto out;
3250     }
3251 
3252     /* The source files are opened read-only, don't care about WCE */
3253     assert((src_flags & BDRV_O_RDWR) == 0);
3254     (void) src_writethrough;
3255 
3256     /*
3257      * Open the images.
3258      *
3259      * Ignore the old backing file for unsafe rebase in case we want to correct
3260      * the reference to a renamed or moved backing file.
3261      */
3262     blk = img_open(image_opts, filename, fmt, flags, writethrough, quiet,
3263                    false);
3264     if (!blk) {
3265         ret = -1;
3266         goto out;
3267     }
3268     bs = blk_bs(blk);
3269 
3270     if (out_basefmt != NULL) {
3271         if (bdrv_find_format(out_basefmt) == NULL) {
3272             error_report("Invalid format name: '%s'", out_basefmt);
3273             ret = -1;
3274             goto out;
3275         }
3276     }
3277 
3278     /* For safe rebasing we need to compare old and new backing file */
3279     if (!unsafe) {
3280         char backing_name[PATH_MAX];
3281         QDict *options = NULL;
3282 
3283         if (bs->backing_format[0] != '\0') {
3284             options = qdict_new();
3285             qdict_put_str(options, "driver", bs->backing_format);
3286         }
3287 
3288         if (force_share) {
3289             if (!options) {
3290                 options = qdict_new();
3291             }
3292             qdict_put_bool(options, BDRV_OPT_FORCE_SHARE, true);
3293         }
3294         bdrv_get_backing_filename(bs, backing_name, sizeof(backing_name));
3295         blk_old_backing = blk_new_open(backing_name, NULL,
3296                                        options, src_flags, &local_err);
3297         if (!blk_old_backing) {
3298             error_reportf_err(local_err,
3299                               "Could not open old backing file '%s': ",
3300                               backing_name);
3301             ret = -1;
3302             goto out;
3303         }
3304 
3305         if (out_baseimg[0]) {
3306             const char *overlay_filename;
3307             char *out_real_path;
3308 
3309             options = qdict_new();
3310             if (out_basefmt) {
3311                 qdict_put_str(options, "driver", out_basefmt);
3312             }
3313             if (force_share) {
3314                 qdict_put_bool(options, BDRV_OPT_FORCE_SHARE, true);
3315             }
3316 
3317             overlay_filename = bs->exact_filename[0] ? bs->exact_filename
3318                                                      : bs->filename;
3319             out_real_path = g_malloc(PATH_MAX);
3320 
3321             bdrv_get_full_backing_filename_from_filename(overlay_filename,
3322                                                          out_baseimg,
3323                                                          out_real_path,
3324                                                          PATH_MAX,
3325                                                          &local_err);
3326             if (local_err) {
3327                 error_reportf_err(local_err,
3328                                   "Could not resolve backing filename: ");
3329                 ret = -1;
3330                 g_free(out_real_path);
3331                 goto out;
3332             }
3333 
3334             blk_new_backing = blk_new_open(out_real_path, NULL,
3335                                            options, src_flags, &local_err);
3336             g_free(out_real_path);
3337             if (!blk_new_backing) {
3338                 error_reportf_err(local_err,
3339                                   "Could not open new backing file '%s': ",
3340                                   out_baseimg);
3341                 ret = -1;
3342                 goto out;
3343             }
3344         }
3345     }
3346 
3347     /*
3348      * Check each unallocated cluster in the COW file. If it is unallocated,
3349      * accesses go to the backing file. We must therefore compare this cluster
3350      * in the old and new backing file, and if they differ we need to copy it
3351      * from the old backing file into the COW file.
3352      *
3353      * If qemu-img crashes during this step, no harm is done. The content of
3354      * the image is the same as the original one at any time.
3355      */
3356     if (!unsafe) {
3357         int64_t size;
3358         int64_t old_backing_size;
3359         int64_t new_backing_size = 0;
3360         uint64_t offset;
3361         int64_t n;
3362         float local_progress = 0;
3363 
3364         buf_old = blk_blockalign(blk, IO_BUF_SIZE);
3365         buf_new = blk_blockalign(blk, IO_BUF_SIZE);
3366 
3367         size = blk_getlength(blk);
3368         if (size < 0) {
3369             error_report("Could not get size of '%s': %s",
3370                          filename, strerror(-size));
3371             ret = -1;
3372             goto out;
3373         }
3374         old_backing_size = blk_getlength(blk_old_backing);
3375         if (old_backing_size < 0) {
3376             char backing_name[PATH_MAX];
3377 
3378             bdrv_get_backing_filename(bs, backing_name, sizeof(backing_name));
3379             error_report("Could not get size of '%s': %s",
3380                          backing_name, strerror(-old_backing_size));
3381             ret = -1;
3382             goto out;
3383         }
3384         if (blk_new_backing) {
3385             new_backing_size = blk_getlength(blk_new_backing);
3386             if (new_backing_size < 0) {
3387                 error_report("Could not get size of '%s': %s",
3388                              out_baseimg, strerror(-new_backing_size));
3389                 ret = -1;
3390                 goto out;
3391             }
3392         }
3393 
3394         if (size != 0) {
3395             local_progress = (float)100 / (size / MIN(size, IO_BUF_SIZE));
3396         }
3397 
3398         for (offset = 0; offset < size; offset += n) {
3399             /* How many bytes can we handle with the next read? */
3400             n = MIN(IO_BUF_SIZE, size - offset);
3401 
3402             /* If the cluster is allocated, we don't need to take action */
3403             ret = bdrv_is_allocated(bs, offset, n, &n);
3404             if (ret < 0) {
3405                 error_report("error while reading image metadata: %s",
3406                              strerror(-ret));
3407                 goto out;
3408             }
3409             if (ret) {
3410                 continue;
3411             }
3412 
3413             /*
3414              * Read old and new backing file and take into consideration that
3415              * backing files may be smaller than the COW image.
3416              */
3417             if (offset >= old_backing_size) {
3418                 memset(buf_old, 0, n);
3419             } else {
3420                 if (offset + n > old_backing_size) {
3421                     n = old_backing_size - offset;
3422                 }
3423 
3424                 ret = blk_pread(blk_old_backing, offset, buf_old, n);
3425                 if (ret < 0) {
3426                     error_report("error while reading from old backing file");
3427                     goto out;
3428                 }
3429             }
3430 
3431             if (offset >= new_backing_size || !blk_new_backing) {
3432                 memset(buf_new, 0, n);
3433             } else {
3434                 if (offset + n > new_backing_size) {
3435                     n = new_backing_size - offset;
3436                 }
3437 
3438                 ret = blk_pread(blk_new_backing, offset, buf_new, n);
3439                 if (ret < 0) {
3440                     error_report("error while reading from new backing file");
3441                     goto out;
3442                 }
3443             }
3444 
3445             /* If they differ, we need to write to the COW file */
3446             uint64_t written = 0;
3447 
3448             while (written < n) {
3449                 int64_t pnum;
3450 
3451                 if (compare_buffers(buf_old + written, buf_new + written,
3452                                     n - written, &pnum))
3453                 {
3454                     ret = blk_pwrite(blk, offset + written,
3455                                      buf_old + written, pnum, 0);
3456                     if (ret < 0) {
3457                         error_report("Error while writing to COW image: %s",
3458                             strerror(-ret));
3459                         goto out;
3460                     }
3461                 }
3462 
3463                 written += pnum;
3464             }
3465             qemu_progress_print(local_progress, 100);
3466         }
3467     }
3468 
3469     /*
3470      * Change the backing file. All clusters that are different from the old
3471      * backing file are overwritten in the COW file now, so the visible content
3472      * doesn't change when we switch the backing file.
3473      */
3474     if (out_baseimg && *out_baseimg) {
3475         ret = bdrv_change_backing_file(bs, out_baseimg, out_basefmt);
3476     } else {
3477         ret = bdrv_change_backing_file(bs, NULL, NULL);
3478     }
3479 
3480     if (ret == -ENOSPC) {
3481         error_report("Could not change the backing file to '%s': No "
3482                      "space left in the file header", out_baseimg);
3483     } else if (ret < 0) {
3484         error_report("Could not change the backing file to '%s': %s",
3485             out_baseimg, strerror(-ret));
3486     }
3487 
3488     qemu_progress_print(100, 0);
3489     /*
3490      * TODO At this point it is possible to check if any clusters that are
3491      * allocated in the COW file are the same in the backing file. If so, they
3492      * could be dropped from the COW file. Don't do this before switching the
3493      * backing file, in case of a crash this would lead to corruption.
3494      */
3495 out:
3496     qemu_progress_end();
3497     /* Cleanup */
3498     if (!unsafe) {
3499         blk_unref(blk_old_backing);
3500         blk_unref(blk_new_backing);
3501     }
3502     qemu_vfree(buf_old);
3503     qemu_vfree(buf_new);
3504 
3505     blk_unref(blk);
3506     if (ret) {
3507         return 1;
3508     }
3509     return 0;
3510 }
3511 
3512 static int img_resize(int argc, char **argv)
3513 {
3514     Error *err = NULL;
3515     int c, ret, relative;
3516     const char *filename, *fmt, *size;
3517     int64_t n, total_size, current_size, new_size;
3518     bool quiet = false;
3519     BlockBackend *blk = NULL;
3520     PreallocMode prealloc = PREALLOC_MODE_OFF;
3521     QemuOpts *param;
3522 
3523     static QemuOptsList resize_options = {
3524         .name = "resize_options",
3525         .head = QTAILQ_HEAD_INITIALIZER(resize_options.head),
3526         .desc = {
3527             {
3528                 .name = BLOCK_OPT_SIZE,
3529                 .type = QEMU_OPT_SIZE,
3530                 .help = "Virtual disk size"
3531             }, {
3532                 /* end of list */
3533             }
3534         },
3535     };
3536     bool image_opts = false;
3537     bool shrink = false;
3538 
3539     /* Remove size from argv manually so that negative numbers are not treated
3540      * as options by getopt. */
3541     if (argc < 3) {
3542         error_exit("Not enough arguments");
3543         return 1;
3544     }
3545 
3546     size = argv[--argc];
3547 
3548     /* Parse getopt arguments */
3549     fmt = NULL;
3550     for(;;) {
3551         static const struct option long_options[] = {
3552             {"help", no_argument, 0, 'h'},
3553             {"object", required_argument, 0, OPTION_OBJECT},
3554             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
3555             {"preallocation", required_argument, 0, OPTION_PREALLOCATION},
3556             {"shrink", no_argument, 0, OPTION_SHRINK},
3557             {0, 0, 0, 0}
3558         };
3559         c = getopt_long(argc, argv, ":f:hq",
3560                         long_options, NULL);
3561         if (c == -1) {
3562             break;
3563         }
3564         switch(c) {
3565         case ':':
3566             missing_argument(argv[optind - 1]);
3567             break;
3568         case '?':
3569             unrecognized_option(argv[optind - 1]);
3570             break;
3571         case 'h':
3572             help();
3573             break;
3574         case 'f':
3575             fmt = optarg;
3576             break;
3577         case 'q':
3578             quiet = true;
3579             break;
3580         case OPTION_OBJECT: {
3581             QemuOpts *opts;
3582             opts = qemu_opts_parse_noisily(&qemu_object_opts,
3583                                            optarg, true);
3584             if (!opts) {
3585                 return 1;
3586             }
3587         }   break;
3588         case OPTION_IMAGE_OPTS:
3589             image_opts = true;
3590             break;
3591         case OPTION_PREALLOCATION:
3592             prealloc = qapi_enum_parse(&PreallocMode_lookup, optarg,
3593                                        PREALLOC_MODE__MAX, NULL);
3594             if (prealloc == PREALLOC_MODE__MAX) {
3595                 error_report("Invalid preallocation mode '%s'", optarg);
3596                 return 1;
3597             }
3598             break;
3599         case OPTION_SHRINK:
3600             shrink = true;
3601             break;
3602         }
3603     }
3604     if (optind != argc - 1) {
3605         error_exit("Expecting image file name and size");
3606     }
3607     filename = argv[optind++];
3608 
3609     if (qemu_opts_foreach(&qemu_object_opts,
3610                           user_creatable_add_opts_foreach,
3611                           NULL, NULL)) {
3612         return 1;
3613     }
3614 
3615     /* Choose grow, shrink, or absolute resize mode */
3616     switch (size[0]) {
3617     case '+':
3618         relative = 1;
3619         size++;
3620         break;
3621     case '-':
3622         relative = -1;
3623         size++;
3624         break;
3625     default:
3626         relative = 0;
3627         break;
3628     }
3629 
3630     /* Parse size */
3631     param = qemu_opts_create(&resize_options, NULL, 0, &error_abort);
3632     qemu_opt_set(param, BLOCK_OPT_SIZE, size, &err);
3633     if (err) {
3634         error_report_err(err);
3635         ret = -1;
3636         qemu_opts_del(param);
3637         goto out;
3638     }
3639     n = qemu_opt_get_size(param, BLOCK_OPT_SIZE, 0);
3640     qemu_opts_del(param);
3641 
3642     blk = img_open(image_opts, filename, fmt,
3643                    BDRV_O_RDWR | BDRV_O_RESIZE, false, quiet,
3644                    false);
3645     if (!blk) {
3646         ret = -1;
3647         goto out;
3648     }
3649 
3650     current_size = blk_getlength(blk);
3651     if (current_size < 0) {
3652         error_report("Failed to inquire current image length: %s",
3653                      strerror(-current_size));
3654         ret = -1;
3655         goto out;
3656     }
3657 
3658     if (relative) {
3659         total_size = current_size + n * relative;
3660     } else {
3661         total_size = n;
3662     }
3663     if (total_size <= 0) {
3664         error_report("New image size must be positive");
3665         ret = -1;
3666         goto out;
3667     }
3668 
3669     if (total_size <= current_size && prealloc != PREALLOC_MODE_OFF) {
3670         error_report("Preallocation can only be used for growing images");
3671         ret = -1;
3672         goto out;
3673     }
3674 
3675     if (total_size < current_size && !shrink) {
3676         warn_report("Shrinking an image will delete all data beyond the "
3677                     "shrunken image's end. Before performing such an "
3678                     "operation, make sure there is no important data there.");
3679 
3680         if (g_strcmp0(bdrv_get_format_name(blk_bs(blk)), "raw") != 0) {
3681             error_report(
3682               "Use the --shrink option to perform a shrink operation.");
3683             ret = -1;
3684             goto out;
3685         } else {
3686             warn_report("Using the --shrink option will suppress this message. "
3687                         "Note that future versions of qemu-img may refuse to "
3688                         "shrink images without this option.");
3689         }
3690     }
3691 
3692     ret = blk_truncate(blk, total_size, prealloc, &err);
3693     if (ret < 0) {
3694         error_report_err(err);
3695         goto out;
3696     }
3697 
3698     new_size = blk_getlength(blk);
3699     if (new_size < 0) {
3700         error_report("Failed to verify truncated image length: %s",
3701                      strerror(-new_size));
3702         ret = -1;
3703         goto out;
3704     }
3705 
3706     /* Some block drivers implement a truncation method, but only so
3707      * the user can cause qemu to refresh the image's size from disk.
3708      * The idea is that the user resizes the image outside of qemu and
3709      * then invokes block_resize to inform qemu about it.
3710      * (This includes iscsi and file-posix for device files.)
3711      * Of course, that is not the behavior someone invoking
3712      * qemu-img resize would find useful, so we catch that behavior
3713      * here and tell the user. */
3714     if (new_size != total_size && new_size == current_size) {
3715         error_report("Image was not resized; resizing may not be supported "
3716                      "for this image");
3717         ret = -1;
3718         goto out;
3719     }
3720 
3721     if (new_size != total_size) {
3722         warn_report("Image should have been resized to %" PRIi64
3723                     " bytes, but was resized to %" PRIi64 " bytes",
3724                     total_size, new_size);
3725     }
3726 
3727     qprintf(quiet, "Image resized.\n");
3728 
3729 out:
3730     blk_unref(blk);
3731     if (ret) {
3732         return 1;
3733     }
3734     return 0;
3735 }
3736 
3737 static void amend_status_cb(BlockDriverState *bs,
3738                             int64_t offset, int64_t total_work_size,
3739                             void *opaque)
3740 {
3741     qemu_progress_print(100.f * offset / total_work_size, 0);
3742 }
3743 
3744 static int print_amend_option_help(const char *format)
3745 {
3746     BlockDriver *drv;
3747 
3748     /* Find driver and parse its options */
3749     drv = bdrv_find_format(format);
3750     if (!drv) {
3751         error_report("Unknown file format '%s'", format);
3752         return 1;
3753     }
3754 
3755     if (!drv->bdrv_amend_options) {
3756         error_report("Format driver '%s' does not support option amendment",
3757                      format);
3758         return 1;
3759     }
3760 
3761     /* Every driver supporting amendment must have create_opts */
3762     assert(drv->create_opts);
3763 
3764     printf("Creation options for '%s':\n", format);
3765     qemu_opts_print_help(drv->create_opts);
3766     printf("\nNote that not all of these options may be amendable.\n");
3767     return 0;
3768 }
3769 
3770 static int img_amend(int argc, char **argv)
3771 {
3772     Error *err = NULL;
3773     int c, ret = 0;
3774     char *options = NULL;
3775     QemuOptsList *create_opts = NULL;
3776     QemuOpts *opts = NULL;
3777     const char *fmt = NULL, *filename, *cache;
3778     int flags;
3779     bool writethrough;
3780     bool quiet = false, progress = false;
3781     BlockBackend *blk = NULL;
3782     BlockDriverState *bs = NULL;
3783     bool image_opts = false;
3784 
3785     cache = BDRV_DEFAULT_CACHE;
3786     for (;;) {
3787         static const struct option long_options[] = {
3788             {"help", no_argument, 0, 'h'},
3789             {"object", required_argument, 0, OPTION_OBJECT},
3790             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
3791             {0, 0, 0, 0}
3792         };
3793         c = getopt_long(argc, argv, ":ho:f:t:pq",
3794                         long_options, NULL);
3795         if (c == -1) {
3796             break;
3797         }
3798 
3799         switch (c) {
3800         case ':':
3801             missing_argument(argv[optind - 1]);
3802             break;
3803         case '?':
3804             unrecognized_option(argv[optind - 1]);
3805             break;
3806         case 'h':
3807             help();
3808             break;
3809         case 'o':
3810             if (!is_valid_option_list(optarg)) {
3811                 error_report("Invalid option list: %s", optarg);
3812                 ret = -1;
3813                 goto out_no_progress;
3814             }
3815             if (!options) {
3816                 options = g_strdup(optarg);
3817             } else {
3818                 char *old_options = options;
3819                 options = g_strdup_printf("%s,%s", options, optarg);
3820                 g_free(old_options);
3821             }
3822             break;
3823         case 'f':
3824             fmt = optarg;
3825             break;
3826         case 't':
3827             cache = optarg;
3828             break;
3829         case 'p':
3830             progress = true;
3831             break;
3832         case 'q':
3833             quiet = true;
3834             break;
3835         case OPTION_OBJECT:
3836             opts = qemu_opts_parse_noisily(&qemu_object_opts,
3837                                            optarg, true);
3838             if (!opts) {
3839                 ret = -1;
3840                 goto out_no_progress;
3841             }
3842             break;
3843         case OPTION_IMAGE_OPTS:
3844             image_opts = true;
3845             break;
3846         }
3847     }
3848 
3849     if (!options) {
3850         error_exit("Must specify options (-o)");
3851     }
3852 
3853     if (qemu_opts_foreach(&qemu_object_opts,
3854                           user_creatable_add_opts_foreach,
3855                           NULL, NULL)) {
3856         ret = -1;
3857         goto out_no_progress;
3858     }
3859 
3860     if (quiet) {
3861         progress = false;
3862     }
3863     qemu_progress_init(progress, 1.0);
3864 
3865     filename = (optind == argc - 1) ? argv[argc - 1] : NULL;
3866     if (fmt && has_help_option(options)) {
3867         /* If a format is explicitly specified (and possibly no filename is
3868          * given), print option help here */
3869         ret = print_amend_option_help(fmt);
3870         goto out;
3871     }
3872 
3873     if (optind != argc - 1) {
3874         error_report("Expecting one image file name");
3875         ret = -1;
3876         goto out;
3877     }
3878 
3879     flags = BDRV_O_RDWR;
3880     ret = bdrv_parse_cache_mode(cache, &flags, &writethrough);
3881     if (ret < 0) {
3882         error_report("Invalid cache option: %s", cache);
3883         goto out;
3884     }
3885 
3886     blk = img_open(image_opts, filename, fmt, flags, writethrough, quiet,
3887                    false);
3888     if (!blk) {
3889         ret = -1;
3890         goto out;
3891     }
3892     bs = blk_bs(blk);
3893 
3894     fmt = bs->drv->format_name;
3895 
3896     if (has_help_option(options)) {
3897         /* If the format was auto-detected, print option help here */
3898         ret = print_amend_option_help(fmt);
3899         goto out;
3900     }
3901 
3902     if (!bs->drv->bdrv_amend_options) {
3903         error_report("Format driver '%s' does not support option amendment",
3904                      fmt);
3905         ret = -1;
3906         goto out;
3907     }
3908 
3909     /* Every driver supporting amendment must have create_opts */
3910     assert(bs->drv->create_opts);
3911 
3912     create_opts = qemu_opts_append(create_opts, bs->drv->create_opts);
3913     opts = qemu_opts_create(create_opts, NULL, 0, &error_abort);
3914     qemu_opts_do_parse(opts, options, NULL, &err);
3915     if (err) {
3916         error_report_err(err);
3917         ret = -1;
3918         goto out;
3919     }
3920 
3921     /* In case the driver does not call amend_status_cb() */
3922     qemu_progress_print(0.f, 0);
3923     ret = bdrv_amend_options(bs, opts, &amend_status_cb, NULL, &err);
3924     qemu_progress_print(100.f, 0);
3925     if (ret < 0) {
3926         error_report_err(err);
3927         goto out;
3928     }
3929 
3930 out:
3931     qemu_progress_end();
3932 
3933 out_no_progress:
3934     blk_unref(blk);
3935     qemu_opts_del(opts);
3936     qemu_opts_free(create_opts);
3937     g_free(options);
3938 
3939     if (ret) {
3940         return 1;
3941     }
3942     return 0;
3943 }
3944 
3945 typedef struct BenchData {
3946     BlockBackend *blk;
3947     uint64_t image_size;
3948     bool write;
3949     int bufsize;
3950     int step;
3951     int nrreq;
3952     int n;
3953     int flush_interval;
3954     bool drain_on_flush;
3955     uint8_t *buf;
3956     QEMUIOVector *qiov;
3957 
3958     int in_flight;
3959     bool in_flush;
3960     uint64_t offset;
3961 } BenchData;
3962 
3963 static void bench_undrained_flush_cb(void *opaque, int ret)
3964 {
3965     if (ret < 0) {
3966         error_report("Failed flush request: %s", strerror(-ret));
3967         exit(EXIT_FAILURE);
3968     }
3969 }
3970 
3971 static void bench_cb(void *opaque, int ret)
3972 {
3973     BenchData *b = opaque;
3974     BlockAIOCB *acb;
3975 
3976     if (ret < 0) {
3977         error_report("Failed request: %s", strerror(-ret));
3978         exit(EXIT_FAILURE);
3979     }
3980 
3981     if (b->in_flush) {
3982         /* Just finished a flush with drained queue: Start next requests */
3983         assert(b->in_flight == 0);
3984         b->in_flush = false;
3985     } else if (b->in_flight > 0) {
3986         int remaining = b->n - b->in_flight;
3987 
3988         b->n--;
3989         b->in_flight--;
3990 
3991         /* Time for flush? Drain queue if requested, then flush */
3992         if (b->flush_interval && remaining % b->flush_interval == 0) {
3993             if (!b->in_flight || !b->drain_on_flush) {
3994                 BlockCompletionFunc *cb;
3995 
3996                 if (b->drain_on_flush) {
3997                     b->in_flush = true;
3998                     cb = bench_cb;
3999                 } else {
4000                     cb = bench_undrained_flush_cb;
4001                 }
4002 
4003                 acb = blk_aio_flush(b->blk, cb, b);
4004                 if (!acb) {
4005                     error_report("Failed to issue flush request");
4006                     exit(EXIT_FAILURE);
4007                 }
4008             }
4009             if (b->drain_on_flush) {
4010                 return;
4011             }
4012         }
4013     }
4014 
4015     while (b->n > b->in_flight && b->in_flight < b->nrreq) {
4016         int64_t offset = b->offset;
4017         /* blk_aio_* might look for completed I/Os and kick bench_cb
4018          * again, so make sure this operation is counted by in_flight
4019          * and b->offset is ready for the next submission.
4020          */
4021         b->in_flight++;
4022         b->offset += b->step;
4023         b->offset %= b->image_size;
4024         if (b->write) {
4025             acb = blk_aio_pwritev(b->blk, offset, b->qiov, 0, bench_cb, b);
4026         } else {
4027             acb = blk_aio_preadv(b->blk, offset, b->qiov, 0, bench_cb, b);
4028         }
4029         if (!acb) {
4030             error_report("Failed to issue request");
4031             exit(EXIT_FAILURE);
4032         }
4033     }
4034 }
4035 
4036 static int img_bench(int argc, char **argv)
4037 {
4038     int c, ret = 0;
4039     const char *fmt = NULL, *filename;
4040     bool quiet = false;
4041     bool image_opts = false;
4042     bool is_write = false;
4043     int count = 75000;
4044     int depth = 64;
4045     int64_t offset = 0;
4046     size_t bufsize = 4096;
4047     int pattern = 0;
4048     size_t step = 0;
4049     int flush_interval = 0;
4050     bool drain_on_flush = true;
4051     int64_t image_size;
4052     BlockBackend *blk = NULL;
4053     BenchData data = {};
4054     int flags = 0;
4055     bool writethrough = false;
4056     struct timeval t1, t2;
4057     int i;
4058     bool force_share = false;
4059     size_t buf_size;
4060 
4061     for (;;) {
4062         static const struct option long_options[] = {
4063             {"help", no_argument, 0, 'h'},
4064             {"flush-interval", required_argument, 0, OPTION_FLUSH_INTERVAL},
4065             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
4066             {"pattern", required_argument, 0, OPTION_PATTERN},
4067             {"no-drain", no_argument, 0, OPTION_NO_DRAIN},
4068             {"force-share", no_argument, 0, 'U'},
4069             {0, 0, 0, 0}
4070         };
4071         c = getopt_long(argc, argv, ":hc:d:f:no:qs:S:t:wU", long_options, NULL);
4072         if (c == -1) {
4073             break;
4074         }
4075 
4076         switch (c) {
4077         case ':':
4078             missing_argument(argv[optind - 1]);
4079             break;
4080         case '?':
4081             unrecognized_option(argv[optind - 1]);
4082             break;
4083         case 'h':
4084             help();
4085             break;
4086         case 'c':
4087         {
4088             unsigned long res;
4089 
4090             if (qemu_strtoul(optarg, NULL, 0, &res) < 0 || res > INT_MAX) {
4091                 error_report("Invalid request count specified");
4092                 return 1;
4093             }
4094             count = res;
4095             break;
4096         }
4097         case 'd':
4098         {
4099             unsigned long res;
4100 
4101             if (qemu_strtoul(optarg, NULL, 0, &res) < 0 || res > INT_MAX) {
4102                 error_report("Invalid queue depth specified");
4103                 return 1;
4104             }
4105             depth = res;
4106             break;
4107         }
4108         case 'f':
4109             fmt = optarg;
4110             break;
4111         case 'n':
4112             flags |= BDRV_O_NATIVE_AIO;
4113             break;
4114         case 'o':
4115         {
4116             offset = cvtnum(optarg);
4117             if (offset < 0) {
4118                 error_report("Invalid offset specified");
4119                 return 1;
4120             }
4121             break;
4122         }
4123             break;
4124         case 'q':
4125             quiet = true;
4126             break;
4127         case 's':
4128         {
4129             int64_t sval;
4130 
4131             sval = cvtnum(optarg);
4132             if (sval < 0 || sval > INT_MAX) {
4133                 error_report("Invalid buffer size specified");
4134                 return 1;
4135             }
4136 
4137             bufsize = sval;
4138             break;
4139         }
4140         case 'S':
4141         {
4142             int64_t sval;
4143 
4144             sval = cvtnum(optarg);
4145             if (sval < 0 || sval > INT_MAX) {
4146                 error_report("Invalid step size specified");
4147                 return 1;
4148             }
4149 
4150             step = sval;
4151             break;
4152         }
4153         case 't':
4154             ret = bdrv_parse_cache_mode(optarg, &flags, &writethrough);
4155             if (ret < 0) {
4156                 error_report("Invalid cache mode");
4157                 ret = -1;
4158                 goto out;
4159             }
4160             break;
4161         case 'w':
4162             flags |= BDRV_O_RDWR;
4163             is_write = true;
4164             break;
4165         case 'U':
4166             force_share = true;
4167             break;
4168         case OPTION_PATTERN:
4169         {
4170             unsigned long res;
4171 
4172             if (qemu_strtoul(optarg, NULL, 0, &res) < 0 || res > 0xff) {
4173                 error_report("Invalid pattern byte specified");
4174                 return 1;
4175             }
4176             pattern = res;
4177             break;
4178         }
4179         case OPTION_FLUSH_INTERVAL:
4180         {
4181             unsigned long res;
4182 
4183             if (qemu_strtoul(optarg, NULL, 0, &res) < 0 || res > INT_MAX) {
4184                 error_report("Invalid flush interval specified");
4185                 return 1;
4186             }
4187             flush_interval = res;
4188             break;
4189         }
4190         case OPTION_NO_DRAIN:
4191             drain_on_flush = false;
4192             break;
4193         case OPTION_IMAGE_OPTS:
4194             image_opts = true;
4195             break;
4196         }
4197     }
4198 
4199     if (optind != argc - 1) {
4200         error_exit("Expecting one image file name");
4201     }
4202     filename = argv[argc - 1];
4203 
4204     if (!is_write && flush_interval) {
4205         error_report("--flush-interval is only available in write tests");
4206         ret = -1;
4207         goto out;
4208     }
4209     if (flush_interval && flush_interval < depth) {
4210         error_report("Flush interval can't be smaller than depth");
4211         ret = -1;
4212         goto out;
4213     }
4214 
4215     blk = img_open(image_opts, filename, fmt, flags, writethrough, quiet,
4216                    force_share);
4217     if (!blk) {
4218         ret = -1;
4219         goto out;
4220     }
4221 
4222     image_size = blk_getlength(blk);
4223     if (image_size < 0) {
4224         ret = image_size;
4225         goto out;
4226     }
4227 
4228     data = (BenchData) {
4229         .blk            = blk,
4230         .image_size     = image_size,
4231         .bufsize        = bufsize,
4232         .step           = step ?: bufsize,
4233         .nrreq          = depth,
4234         .n              = count,
4235         .offset         = offset,
4236         .write          = is_write,
4237         .flush_interval = flush_interval,
4238         .drain_on_flush = drain_on_flush,
4239     };
4240     printf("Sending %d %s requests, %d bytes each, %d in parallel "
4241            "(starting at offset %" PRId64 ", step size %d)\n",
4242            data.n, data.write ? "write" : "read", data.bufsize, data.nrreq,
4243            data.offset, data.step);
4244     if (flush_interval) {
4245         printf("Sending flush every %d requests\n", flush_interval);
4246     }
4247 
4248     buf_size = data.nrreq * data.bufsize;
4249     data.buf = blk_blockalign(blk, buf_size);
4250     memset(data.buf, pattern, data.nrreq * data.bufsize);
4251 
4252     blk_register_buf(blk, data.buf, buf_size);
4253 
4254     data.qiov = g_new(QEMUIOVector, data.nrreq);
4255     for (i = 0; i < data.nrreq; i++) {
4256         qemu_iovec_init(&data.qiov[i], 1);
4257         qemu_iovec_add(&data.qiov[i],
4258                        data.buf + i * data.bufsize, data.bufsize);
4259     }
4260 
4261     gettimeofday(&t1, NULL);
4262     bench_cb(&data, 0);
4263 
4264     while (data.n > 0) {
4265         main_loop_wait(false);
4266     }
4267     gettimeofday(&t2, NULL);
4268 
4269     printf("Run completed in %3.3f seconds.\n",
4270            (t2.tv_sec - t1.tv_sec)
4271            + ((double)(t2.tv_usec - t1.tv_usec) / 1000000));
4272 
4273 out:
4274     if (data.buf) {
4275         blk_unregister_buf(blk, data.buf);
4276     }
4277     qemu_vfree(data.buf);
4278     blk_unref(blk);
4279 
4280     if (ret) {
4281         return 1;
4282     }
4283     return 0;
4284 }
4285 
4286 #define C_BS      01
4287 #define C_COUNT   02
4288 #define C_IF      04
4289 #define C_OF      010
4290 #define C_SKIP    020
4291 
4292 struct DdInfo {
4293     unsigned int flags;
4294     int64_t count;
4295 };
4296 
4297 struct DdIo {
4298     int bsz;    /* Block size */
4299     char *filename;
4300     uint8_t *buf;
4301     int64_t offset;
4302 };
4303 
4304 struct DdOpts {
4305     const char *name;
4306     int (*f)(const char *, struct DdIo *, struct DdIo *, struct DdInfo *);
4307     unsigned int flag;
4308 };
4309 
4310 static int img_dd_bs(const char *arg,
4311                      struct DdIo *in, struct DdIo *out,
4312                      struct DdInfo *dd)
4313 {
4314     int64_t res;
4315 
4316     res = cvtnum(arg);
4317 
4318     if (res <= 0 || res > INT_MAX) {
4319         error_report("invalid number: '%s'", arg);
4320         return 1;
4321     }
4322     in->bsz = out->bsz = res;
4323 
4324     return 0;
4325 }
4326 
4327 static int img_dd_count(const char *arg,
4328                         struct DdIo *in, struct DdIo *out,
4329                         struct DdInfo *dd)
4330 {
4331     dd->count = cvtnum(arg);
4332 
4333     if (dd->count < 0) {
4334         error_report("invalid number: '%s'", arg);
4335         return 1;
4336     }
4337 
4338     return 0;
4339 }
4340 
4341 static int img_dd_if(const char *arg,
4342                      struct DdIo *in, struct DdIo *out,
4343                      struct DdInfo *dd)
4344 {
4345     in->filename = g_strdup(arg);
4346 
4347     return 0;
4348 }
4349 
4350 static int img_dd_of(const char *arg,
4351                      struct DdIo *in, struct DdIo *out,
4352                      struct DdInfo *dd)
4353 {
4354     out->filename = g_strdup(arg);
4355 
4356     return 0;
4357 }
4358 
4359 static int img_dd_skip(const char *arg,
4360                        struct DdIo *in, struct DdIo *out,
4361                        struct DdInfo *dd)
4362 {
4363     in->offset = cvtnum(arg);
4364 
4365     if (in->offset < 0) {
4366         error_report("invalid number: '%s'", arg);
4367         return 1;
4368     }
4369 
4370     return 0;
4371 }
4372 
4373 static int img_dd(int argc, char **argv)
4374 {
4375     int ret = 0;
4376     char *arg = NULL;
4377     char *tmp;
4378     BlockDriver *drv = NULL, *proto_drv = NULL;
4379     BlockBackend *blk1 = NULL, *blk2 = NULL;
4380     QemuOpts *opts = NULL;
4381     QemuOptsList *create_opts = NULL;
4382     Error *local_err = NULL;
4383     bool image_opts = false;
4384     int c, i;
4385     const char *out_fmt = "raw";
4386     const char *fmt = NULL;
4387     int64_t size = 0;
4388     int64_t block_count = 0, out_pos, in_pos;
4389     bool force_share = false;
4390     struct DdInfo dd = {
4391         .flags = 0,
4392         .count = 0,
4393     };
4394     struct DdIo in = {
4395         .bsz = 512, /* Block size is by default 512 bytes */
4396         .filename = NULL,
4397         .buf = NULL,
4398         .offset = 0
4399     };
4400     struct DdIo out = {
4401         .bsz = 512,
4402         .filename = NULL,
4403         .buf = NULL,
4404         .offset = 0
4405     };
4406 
4407     const struct DdOpts options[] = {
4408         { "bs", img_dd_bs, C_BS },
4409         { "count", img_dd_count, C_COUNT },
4410         { "if", img_dd_if, C_IF },
4411         { "of", img_dd_of, C_OF },
4412         { "skip", img_dd_skip, C_SKIP },
4413         { NULL, NULL, 0 }
4414     };
4415     const struct option long_options[] = {
4416         { "help", no_argument, 0, 'h'},
4417         { "object", required_argument, 0, OPTION_OBJECT},
4418         { "image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
4419         { "force-share", no_argument, 0, 'U'},
4420         { 0, 0, 0, 0 }
4421     };
4422 
4423     while ((c = getopt_long(argc, argv, ":hf:O:U", long_options, NULL))) {
4424         if (c == EOF) {
4425             break;
4426         }
4427         switch (c) {
4428         case 'O':
4429             out_fmt = optarg;
4430             break;
4431         case 'f':
4432             fmt = optarg;
4433             break;
4434         case ':':
4435             missing_argument(argv[optind - 1]);
4436             break;
4437         case '?':
4438             unrecognized_option(argv[optind - 1]);
4439             break;
4440         case 'h':
4441             help();
4442             break;
4443         case 'U':
4444             force_share = true;
4445             break;
4446         case OPTION_OBJECT:
4447             if (!qemu_opts_parse_noisily(&qemu_object_opts, optarg, true)) {
4448                 ret = -1;
4449                 goto out;
4450             }
4451             break;
4452         case OPTION_IMAGE_OPTS:
4453             image_opts = true;
4454             break;
4455         }
4456     }
4457 
4458     for (i = optind; i < argc; i++) {
4459         int j;
4460         arg = g_strdup(argv[i]);
4461 
4462         tmp = strchr(arg, '=');
4463         if (tmp == NULL) {
4464             error_report("unrecognized operand %s", arg);
4465             ret = -1;
4466             goto out;
4467         }
4468 
4469         *tmp++ = '\0';
4470 
4471         for (j = 0; options[j].name != NULL; j++) {
4472             if (!strcmp(arg, options[j].name)) {
4473                 break;
4474             }
4475         }
4476         if (options[j].name == NULL) {
4477             error_report("unrecognized operand %s", arg);
4478             ret = -1;
4479             goto out;
4480         }
4481 
4482         if (options[j].f(tmp, &in, &out, &dd) != 0) {
4483             ret = -1;
4484             goto out;
4485         }
4486         dd.flags |= options[j].flag;
4487         g_free(arg);
4488         arg = NULL;
4489     }
4490 
4491     if (!(dd.flags & C_IF && dd.flags & C_OF)) {
4492         error_report("Must specify both input and output files");
4493         ret = -1;
4494         goto out;
4495     }
4496 
4497     if (qemu_opts_foreach(&qemu_object_opts,
4498                           user_creatable_add_opts_foreach,
4499                           NULL, NULL)) {
4500         ret = -1;
4501         goto out;
4502     }
4503 
4504     blk1 = img_open(image_opts, in.filename, fmt, 0, false, false,
4505                     force_share);
4506 
4507     if (!blk1) {
4508         ret = -1;
4509         goto out;
4510     }
4511 
4512     drv = bdrv_find_format(out_fmt);
4513     if (!drv) {
4514         error_report("Unknown file format");
4515         ret = -1;
4516         goto out;
4517     }
4518     proto_drv = bdrv_find_protocol(out.filename, true, &local_err);
4519 
4520     if (!proto_drv) {
4521         error_report_err(local_err);
4522         ret = -1;
4523         goto out;
4524     }
4525     if (!drv->create_opts) {
4526         error_report("Format driver '%s' does not support image creation",
4527                      drv->format_name);
4528         ret = -1;
4529         goto out;
4530     }
4531     if (!proto_drv->create_opts) {
4532         error_report("Protocol driver '%s' does not support image creation",
4533                      proto_drv->format_name);
4534         ret = -1;
4535         goto out;
4536     }
4537     create_opts = qemu_opts_append(create_opts, drv->create_opts);
4538     create_opts = qemu_opts_append(create_opts, proto_drv->create_opts);
4539 
4540     opts = qemu_opts_create(create_opts, NULL, 0, &error_abort);
4541 
4542     size = blk_getlength(blk1);
4543     if (size < 0) {
4544         error_report("Failed to get size for '%s'", in.filename);
4545         ret = -1;
4546         goto out;
4547     }
4548 
4549     if (dd.flags & C_COUNT && dd.count <= INT64_MAX / in.bsz &&
4550         dd.count * in.bsz < size) {
4551         size = dd.count * in.bsz;
4552     }
4553 
4554     /* Overflow means the specified offset is beyond input image's size */
4555     if (dd.flags & C_SKIP && (in.offset > INT64_MAX / in.bsz ||
4556                               size < in.bsz * in.offset)) {
4557         qemu_opt_set_number(opts, BLOCK_OPT_SIZE, 0, &error_abort);
4558     } else {
4559         qemu_opt_set_number(opts, BLOCK_OPT_SIZE,
4560                             size - in.bsz * in.offset, &error_abort);
4561     }
4562 
4563     ret = bdrv_create(drv, out.filename, opts, &local_err);
4564     if (ret < 0) {
4565         error_reportf_err(local_err,
4566                           "%s: error while creating output image: ",
4567                           out.filename);
4568         ret = -1;
4569         goto out;
4570     }
4571 
4572     /* TODO, we can't honour --image-opts for the target,
4573      * since it needs to be given in a format compatible
4574      * with the bdrv_create() call above which does not
4575      * support image-opts style.
4576      */
4577     blk2 = img_open_file(out.filename, NULL, out_fmt, BDRV_O_RDWR,
4578                          false, false, false);
4579 
4580     if (!blk2) {
4581         ret = -1;
4582         goto out;
4583     }
4584 
4585     if (dd.flags & C_SKIP && (in.offset > INT64_MAX / in.bsz ||
4586                               size < in.offset * in.bsz)) {
4587         /* We give a warning if the skip option is bigger than the input
4588          * size and create an empty output disk image (i.e. like dd(1)).
4589          */
4590         error_report("%s: cannot skip to specified offset", in.filename);
4591         in_pos = size;
4592     } else {
4593         in_pos = in.offset * in.bsz;
4594     }
4595 
4596     in.buf = g_new(uint8_t, in.bsz);
4597 
4598     for (out_pos = 0; in_pos < size; block_count++) {
4599         int in_ret, out_ret;
4600 
4601         if (in_pos + in.bsz > size) {
4602             in_ret = blk_pread(blk1, in_pos, in.buf, size - in_pos);
4603         } else {
4604             in_ret = blk_pread(blk1, in_pos, in.buf, in.bsz);
4605         }
4606         if (in_ret < 0) {
4607             error_report("error while reading from input image file: %s",
4608                          strerror(-in_ret));
4609             ret = -1;
4610             goto out;
4611         }
4612         in_pos += in_ret;
4613 
4614         out_ret = blk_pwrite(blk2, out_pos, in.buf, in_ret, 0);
4615 
4616         if (out_ret < 0) {
4617             error_report("error while writing to output image file: %s",
4618                          strerror(-out_ret));
4619             ret = -1;
4620             goto out;
4621         }
4622         out_pos += out_ret;
4623     }
4624 
4625 out:
4626     g_free(arg);
4627     qemu_opts_del(opts);
4628     qemu_opts_free(create_opts);
4629     blk_unref(blk1);
4630     blk_unref(blk2);
4631     g_free(in.filename);
4632     g_free(out.filename);
4633     g_free(in.buf);
4634     g_free(out.buf);
4635 
4636     if (ret) {
4637         return 1;
4638     }
4639     return 0;
4640 }
4641 
4642 static void dump_json_block_measure_info(BlockMeasureInfo *info)
4643 {
4644     QString *str;
4645     QObject *obj;
4646     Visitor *v = qobject_output_visitor_new(&obj);
4647 
4648     visit_type_BlockMeasureInfo(v, NULL, &info, &error_abort);
4649     visit_complete(v, &obj);
4650     str = qobject_to_json_pretty(obj);
4651     assert(str != NULL);
4652     printf("%s\n", qstring_get_str(str));
4653     qobject_unref(obj);
4654     visit_free(v);
4655     qobject_unref(str);
4656 }
4657 
4658 static int img_measure(int argc, char **argv)
4659 {
4660     static const struct option long_options[] = {
4661         {"help", no_argument, 0, 'h'},
4662         {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
4663         {"object", required_argument, 0, OPTION_OBJECT},
4664         {"output", required_argument, 0, OPTION_OUTPUT},
4665         {"size", required_argument, 0, OPTION_SIZE},
4666         {"force-share", no_argument, 0, 'U'},
4667         {0, 0, 0, 0}
4668     };
4669     OutputFormat output_format = OFORMAT_HUMAN;
4670     BlockBackend *in_blk = NULL;
4671     BlockDriver *drv;
4672     const char *filename = NULL;
4673     const char *fmt = NULL;
4674     const char *out_fmt = "raw";
4675     char *options = NULL;
4676     char *snapshot_name = NULL;
4677     bool force_share = false;
4678     QemuOpts *opts = NULL;
4679     QemuOpts *object_opts = NULL;
4680     QemuOpts *sn_opts = NULL;
4681     QemuOptsList *create_opts = NULL;
4682     bool image_opts = false;
4683     uint64_t img_size = UINT64_MAX;
4684     BlockMeasureInfo *info = NULL;
4685     Error *local_err = NULL;
4686     int ret = 1;
4687     int c;
4688 
4689     while ((c = getopt_long(argc, argv, "hf:O:o:l:U",
4690                             long_options, NULL)) != -1) {
4691         switch (c) {
4692         case '?':
4693         case 'h':
4694             help();
4695             break;
4696         case 'f':
4697             fmt = optarg;
4698             break;
4699         case 'O':
4700             out_fmt = optarg;
4701             break;
4702         case 'o':
4703             if (!is_valid_option_list(optarg)) {
4704                 error_report("Invalid option list: %s", optarg);
4705                 goto out;
4706             }
4707             if (!options) {
4708                 options = g_strdup(optarg);
4709             } else {
4710                 char *old_options = options;
4711                 options = g_strdup_printf("%s,%s", options, optarg);
4712                 g_free(old_options);
4713             }
4714             break;
4715         case 'l':
4716             if (strstart(optarg, SNAPSHOT_OPT_BASE, NULL)) {
4717                 sn_opts = qemu_opts_parse_noisily(&internal_snapshot_opts,
4718                                                   optarg, false);
4719                 if (!sn_opts) {
4720                     error_report("Failed in parsing snapshot param '%s'",
4721                                  optarg);
4722                     goto out;
4723                 }
4724             } else {
4725                 snapshot_name = optarg;
4726             }
4727             break;
4728         case 'U':
4729             force_share = true;
4730             break;
4731         case OPTION_OBJECT:
4732             object_opts = qemu_opts_parse_noisily(&qemu_object_opts,
4733                                                   optarg, true);
4734             if (!object_opts) {
4735                 goto out;
4736             }
4737             break;
4738         case OPTION_IMAGE_OPTS:
4739             image_opts = true;
4740             break;
4741         case OPTION_OUTPUT:
4742             if (!strcmp(optarg, "json")) {
4743                 output_format = OFORMAT_JSON;
4744             } else if (!strcmp(optarg, "human")) {
4745                 output_format = OFORMAT_HUMAN;
4746             } else {
4747                 error_report("--output must be used with human or json "
4748                              "as argument.");
4749                 goto out;
4750             }
4751             break;
4752         case OPTION_SIZE:
4753         {
4754             int64_t sval;
4755 
4756             sval = cvtnum(optarg);
4757             if (sval < 0) {
4758                 if (sval == -ERANGE) {
4759                     error_report("Image size must be less than 8 EiB!");
4760                 } else {
4761                     error_report("Invalid image size specified! You may use "
4762                                  "k, M, G, T, P or E suffixes for ");
4763                     error_report("kilobytes, megabytes, gigabytes, terabytes, "
4764                                  "petabytes and exabytes.");
4765                 }
4766                 goto out;
4767             }
4768             img_size = (uint64_t)sval;
4769         }
4770         break;
4771         }
4772     }
4773 
4774     if (qemu_opts_foreach(&qemu_object_opts,
4775                           user_creatable_add_opts_foreach,
4776                           NULL, NULL)) {
4777         goto out;
4778     }
4779 
4780     if (argc - optind > 1) {
4781         error_report("At most one filename argument is allowed.");
4782         goto out;
4783     } else if (argc - optind == 1) {
4784         filename = argv[optind];
4785     }
4786 
4787     if (!filename &&
4788         (object_opts || image_opts || fmt || snapshot_name || sn_opts)) {
4789         error_report("--object, --image-opts, -f, and -l "
4790                      "require a filename argument.");
4791         goto out;
4792     }
4793     if (filename && img_size != UINT64_MAX) {
4794         error_report("--size N cannot be used together with a filename.");
4795         goto out;
4796     }
4797     if (!filename && img_size == UINT64_MAX) {
4798         error_report("Either --size N or one filename must be specified.");
4799         goto out;
4800     }
4801 
4802     if (filename) {
4803         in_blk = img_open(image_opts, filename, fmt, 0,
4804                           false, false, force_share);
4805         if (!in_blk) {
4806             goto out;
4807         }
4808 
4809         if (sn_opts) {
4810             bdrv_snapshot_load_tmp(blk_bs(in_blk),
4811                     qemu_opt_get(sn_opts, SNAPSHOT_OPT_ID),
4812                     qemu_opt_get(sn_opts, SNAPSHOT_OPT_NAME),
4813                     &local_err);
4814         } else if (snapshot_name != NULL) {
4815             bdrv_snapshot_load_tmp_by_id_or_name(blk_bs(in_blk),
4816                     snapshot_name, &local_err);
4817         }
4818         if (local_err) {
4819             error_reportf_err(local_err, "Failed to load snapshot: ");
4820             goto out;
4821         }
4822     }
4823 
4824     drv = bdrv_find_format(out_fmt);
4825     if (!drv) {
4826         error_report("Unknown file format '%s'", out_fmt);
4827         goto out;
4828     }
4829     if (!drv->create_opts) {
4830         error_report("Format driver '%s' does not support image creation",
4831                      drv->format_name);
4832         goto out;
4833     }
4834 
4835     create_opts = qemu_opts_append(create_opts, drv->create_opts);
4836     create_opts = qemu_opts_append(create_opts, bdrv_file.create_opts);
4837     opts = qemu_opts_create(create_opts, NULL, 0, &error_abort);
4838     if (options) {
4839         qemu_opts_do_parse(opts, options, NULL, &local_err);
4840         if (local_err) {
4841             error_report_err(local_err);
4842             error_report("Invalid options for file format '%s'", out_fmt);
4843             goto out;
4844         }
4845     }
4846     if (img_size != UINT64_MAX) {
4847         qemu_opt_set_number(opts, BLOCK_OPT_SIZE, img_size, &error_abort);
4848     }
4849 
4850     info = bdrv_measure(drv, opts, in_blk ? blk_bs(in_blk) : NULL, &local_err);
4851     if (local_err) {
4852         error_report_err(local_err);
4853         goto out;
4854     }
4855 
4856     if (output_format == OFORMAT_HUMAN) {
4857         printf("required size: %" PRIu64 "\n", info->required);
4858         printf("fully allocated size: %" PRIu64 "\n", info->fully_allocated);
4859     } else {
4860         dump_json_block_measure_info(info);
4861     }
4862 
4863     ret = 0;
4864 
4865 out:
4866     qapi_free_BlockMeasureInfo(info);
4867     qemu_opts_del(object_opts);
4868     qemu_opts_del(opts);
4869     qemu_opts_del(sn_opts);
4870     qemu_opts_free(create_opts);
4871     g_free(options);
4872     blk_unref(in_blk);
4873     return ret;
4874 }
4875 
4876 static const img_cmd_t img_cmds[] = {
4877 #define DEF(option, callback, arg_string)        \
4878     { option, callback },
4879 #include "qemu-img-cmds.h"
4880 #undef DEF
4881     { NULL, NULL, },
4882 };
4883 
4884 int main(int argc, char **argv)
4885 {
4886     const img_cmd_t *cmd;
4887     const char *cmdname;
4888     Error *local_error = NULL;
4889     char *trace_file = NULL;
4890     int c;
4891     static const struct option long_options[] = {
4892         {"help", no_argument, 0, 'h'},
4893         {"version", no_argument, 0, 'V'},
4894         {"trace", required_argument, NULL, 'T'},
4895         {0, 0, 0, 0}
4896     };
4897 
4898 #ifdef CONFIG_POSIX
4899     signal(SIGPIPE, SIG_IGN);
4900 #endif
4901 
4902     module_call_init(MODULE_INIT_TRACE);
4903     error_set_progname(argv[0]);
4904     qemu_init_exec_dir(argv[0]);
4905 
4906     if (qemu_init_main_loop(&local_error)) {
4907         error_report_err(local_error);
4908         exit(EXIT_FAILURE);
4909     }
4910 
4911     qcrypto_init(&error_fatal);
4912 
4913     module_call_init(MODULE_INIT_QOM);
4914     bdrv_init();
4915     if (argc < 2) {
4916         error_exit("Not enough arguments");
4917     }
4918 
4919     qemu_add_opts(&qemu_object_opts);
4920     qemu_add_opts(&qemu_source_opts);
4921     qemu_add_opts(&qemu_trace_opts);
4922 
4923     while ((c = getopt_long(argc, argv, "+:hVT:", long_options, NULL)) != -1) {
4924         switch (c) {
4925         case ':':
4926             missing_argument(argv[optind - 1]);
4927             return 0;
4928         case '?':
4929             unrecognized_option(argv[optind - 1]);
4930             return 0;
4931         case 'h':
4932             help();
4933             return 0;
4934         case 'V':
4935             printf(QEMU_IMG_VERSION);
4936             return 0;
4937         case 'T':
4938             g_free(trace_file);
4939             trace_file = trace_opt_parse(optarg);
4940             break;
4941         }
4942     }
4943 
4944     cmdname = argv[optind];
4945 
4946     /* reset getopt_long scanning */
4947     argc -= optind;
4948     if (argc < 1) {
4949         return 0;
4950     }
4951     argv += optind;
4952     optind = 0;
4953 
4954     if (!trace_init_backends()) {
4955         exit(1);
4956     }
4957     trace_init_file(trace_file);
4958     qemu_set_log(LOG_TRACE);
4959 
4960     /* find the command */
4961     for (cmd = img_cmds; cmd->name != NULL; cmd++) {
4962         if (!strcmp(cmdname, cmd->name)) {
4963             return cmd->handler(argc, argv);
4964         }
4965     }
4966 
4967     /* not found */
4968     error_exit("Command not found: %s", cmdname);
4969 }
4970