xref: /openbmc/qemu/qemu-img.c (revision 5ee5c14c)
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 "qemu/units.h"
41 #include "qom/object_interfaces.h"
42 #include "sysemu/sysemu.h"
43 #include "sysemu/block-backend.h"
44 #include "block/block_int.h"
45 #include "block/blockjob.h"
46 #include "block/qapi.h"
47 #include "crypto/init.h"
48 #include "trace/control.h"
49 
50 #define QEMU_IMG_VERSION "qemu-img version " QEMU_FULL_VERSION \
51                           "\n" QEMU_COPYRIGHT "\n"
52 
53 typedef struct img_cmd_t {
54     const char *name;
55     int (*handler)(int argc, char **argv);
56 } img_cmd_t;
57 
58 enum {
59     OPTION_OUTPUT = 256,
60     OPTION_BACKING_CHAIN = 257,
61     OPTION_OBJECT = 258,
62     OPTION_IMAGE_OPTS = 259,
63     OPTION_PATTERN = 260,
64     OPTION_FLUSH_INTERVAL = 261,
65     OPTION_NO_DRAIN = 262,
66     OPTION_TARGET_IMAGE_OPTS = 263,
67     OPTION_SIZE = 264,
68     OPTION_PREALLOCATION = 265,
69     OPTION_SHRINK = 266,
70 };
71 
72 typedef enum OutputFormat {
73     OFORMAT_JSON,
74     OFORMAT_HUMAN,
75 } OutputFormat;
76 
77 /* Default to cache=writeback as data integrity is not important for qemu-img */
78 #define BDRV_DEFAULT_CACHE "writeback"
79 
80 static void format_print(void *opaque, const char *name)
81 {
82     printf(" %s", name);
83 }
84 
85 static void QEMU_NORETURN GCC_FMT_ATTR(1, 2) error_exit(const char *fmt, ...)
86 {
87     va_list ap;
88 
89     va_start(ap, fmt);
90     error_vreport(fmt, ap);
91     va_end(ap);
92 
93     error_printf("Try 'qemu-img --help' for more information\n");
94     exit(EXIT_FAILURE);
95 }
96 
97 static void QEMU_NORETURN missing_argument(const char *option)
98 {
99     error_exit("missing argument for option '%s'", option);
100 }
101 
102 static void QEMU_NORETURN unrecognized_option(const char *option)
103 {
104     error_exit("unrecognized option '%s'", option);
105 }
106 
107 /* Please keep in synch with qemu-img.texi */
108 static void QEMU_NORETURN help(void)
109 {
110     const char *help_msg =
111            QEMU_IMG_VERSION
112            "usage: qemu-img [standard options] command [command options]\n"
113            "QEMU disk image utility\n"
114            "\n"
115            "    '-h', '--help'       display this help and exit\n"
116            "    '-V', '--version'    output version information and exit\n"
117            "    '-T', '--trace'      [[enable=]<pattern>][,events=<file>][,file=<file>]\n"
118            "                         specify tracing options\n"
119            "\n"
120            "Command syntax:\n"
121 #define DEF(option, callback, arg_string)        \
122            "  " arg_string "\n"
123 #include "qemu-img-cmds.h"
124 #undef DEF
125            "\n"
126            "Command parameters:\n"
127            "  'filename' is a disk image filename\n"
128            "  'objectdef' is a QEMU user creatable object definition. See the qemu(1)\n"
129            "    manual page for a description of the object properties. The most common\n"
130            "    object type is a 'secret', which is used to supply passwords and/or\n"
131            "    encryption keys.\n"
132            "  'fmt' is the disk image format. It is guessed automatically in most cases\n"
133            "  'cache' is the cache mode used to write the output disk image, the valid\n"
134            "    options are: 'none', 'writeback' (default, except for convert), 'writethrough',\n"
135            "    'directsync' and 'unsafe' (default for convert)\n"
136            "  'src_cache' is the cache mode used to read input disk images, the valid\n"
137            "    options are the same as for the 'cache' option\n"
138            "  'size' is the disk image size in bytes. Optional suffixes\n"
139            "    'k' or 'K' (kilobyte, 1024), 'M' (megabyte, 1024k), 'G' (gigabyte, 1024M),\n"
140            "    'T' (terabyte, 1024G), 'P' (petabyte, 1024T) and 'E' (exabyte, 1024P)  are\n"
141            "    supported. 'b' is ignored.\n"
142            "  'output_filename' is the destination disk image filename\n"
143            "  'output_fmt' is the destination format\n"
144            "  'options' is a comma separated list of format specific options in a\n"
145            "    name=value format. Use -o ? for an overview of the options supported by the\n"
146            "    used format\n"
147            "  'snapshot_param' is param used for internal snapshot, format\n"
148            "    is 'snapshot.id=[ID],snapshot.name=[NAME]', or\n"
149            "    '[ID_OR_NAME]'\n"
150            "  '-c' indicates that target image must be compressed (qcow format only)\n"
151            "  '-u' allows unsafe backing chains. For rebasing, it is assumed that old and\n"
152            "       new backing file match exactly. The image doesn't need a working\n"
153            "       backing file before rebasing in this case (useful for renaming the\n"
154            "       backing file). For image creation, allow creating without attempting\n"
155            "       to open the backing file.\n"
156            "  '-h' with or without a command shows this help and lists the supported formats\n"
157            "  '-p' show progress of command (only certain commands)\n"
158            "  '-q' use Quiet mode - do not print any output (except errors)\n"
159            "  '-S' indicates the consecutive number of bytes (defaults to 4k) that must\n"
160            "       contain only zeros for qemu-img to create a sparse image during\n"
161            "       conversion. If the number of bytes is 0, the source will not be scanned for\n"
162            "       unallocated or zero sectors, and the destination image will always be\n"
163            "       fully allocated\n"
164            "  '--output' takes the format in which the output must be done (human or json)\n"
165            "  '-n' skips the target volume creation (useful if the volume is created\n"
166            "       prior to running qemu-img)\n"
167            "\n"
168            "Parameters to check subcommand:\n"
169            "  '-r' tries to repair any inconsistencies that are found during the check.\n"
170            "       '-r leaks' repairs only cluster leaks, whereas '-r all' fixes all\n"
171            "       kinds of errors, with a higher risk of choosing the wrong fix or\n"
172            "       hiding corruption that has already occurred.\n"
173            "\n"
174            "Parameters to convert subcommand:\n"
175            "  '-m' specifies how many coroutines work in parallel during the convert\n"
176            "       process (defaults to 8)\n"
177            "  '-W' allow to write to the target out of order rather than sequential\n"
178            "\n"
179            "Parameters to snapshot subcommand:\n"
180            "  'snapshot' is the name of the snapshot to create, apply or delete\n"
181            "  '-a' applies a snapshot (revert disk to saved state)\n"
182            "  '-c' creates a snapshot\n"
183            "  '-d' deletes a snapshot\n"
184            "  '-l' lists all snapshots in the given image\n"
185            "\n"
186            "Parameters to compare subcommand:\n"
187            "  '-f' first image format\n"
188            "  '-F' second image format\n"
189            "  '-s' run in Strict mode - fail on different image size or sector allocation\n"
190            "\n"
191            "Parameters to dd subcommand:\n"
192            "  'bs=BYTES' read and write up to BYTES bytes at a time "
193            "(default: 512)\n"
194            "  'count=N' copy only N input blocks\n"
195            "  'if=FILE' read from FILE\n"
196            "  'of=FILE' write to FILE\n"
197            "  'skip=N' skip N bs-sized blocks at the start of input\n";
198 
199     printf("%s\nSupported formats:", help_msg);
200     bdrv_iterate_format(format_print, NULL, false);
201     printf("\n\n" QEMU_HELP_BOTTOM "\n");
202     exit(EXIT_SUCCESS);
203 }
204 
205 static QemuOptsList qemu_object_opts = {
206     .name = "object",
207     .implied_opt_name = "qom-type",
208     .head = QTAILQ_HEAD_INITIALIZER(qemu_object_opts.head),
209     .desc = {
210         { }
211     },
212 };
213 
214 static QemuOptsList qemu_source_opts = {
215     .name = "source",
216     .implied_opt_name = "file",
217     .head = QTAILQ_HEAD_INITIALIZER(qemu_source_opts.head),
218     .desc = {
219         { }
220     },
221 };
222 
223 static int GCC_FMT_ATTR(2, 3) qprintf(bool quiet, const char *fmt, ...)
224 {
225     int ret = 0;
226     if (!quiet) {
227         va_list args;
228         va_start(args, fmt);
229         ret = vprintf(fmt, args);
230         va_end(args);
231     }
232     return ret;
233 }
234 
235 
236 static int print_block_option_help(const char *filename, const char *fmt)
237 {
238     BlockDriver *drv, *proto_drv;
239     QemuOptsList *create_opts = NULL;
240     Error *local_err = NULL;
241 
242     /* Find driver and parse its options */
243     drv = bdrv_find_format(fmt);
244     if (!drv) {
245         error_report("Unknown file format '%s'", fmt);
246         return 1;
247     }
248 
249     if (!drv->create_opts) {
250         error_report("Format driver '%s' does not support image creation", fmt);
251         return 1;
252     }
253 
254     create_opts = qemu_opts_append(create_opts, drv->create_opts);
255     if (filename) {
256         proto_drv = bdrv_find_protocol(filename, true, &local_err);
257         if (!proto_drv) {
258             error_report_err(local_err);
259             qemu_opts_free(create_opts);
260             return 1;
261         }
262         if (!proto_drv->create_opts) {
263             error_report("Protocol driver '%s' does not support image creation",
264                          proto_drv->format_name);
265             qemu_opts_free(create_opts);
266             return 1;
267         }
268         create_opts = qemu_opts_append(create_opts, proto_drv->create_opts);
269     }
270 
271     if (filename) {
272         printf("Supported options:\n");
273     } else {
274         printf("Supported %s options:\n", fmt);
275     }
276     qemu_opts_print_help(create_opts, false);
277     qemu_opts_free(create_opts);
278 
279     if (!filename) {
280         printf("\n"
281                "The protocol level may support further options.\n"
282                "Specify the target filename to include those options.\n");
283     }
284 
285     return 0;
286 }
287 
288 
289 static BlockBackend *img_open_opts(const char *optstr,
290                                    QemuOpts *opts, int flags, bool writethrough,
291                                    bool quiet, bool force_share)
292 {
293     QDict *options;
294     Error *local_err = NULL;
295     BlockBackend *blk;
296     options = qemu_opts_to_qdict(opts, NULL);
297     if (force_share) {
298         if (qdict_haskey(options, BDRV_OPT_FORCE_SHARE)
299             && strcmp(qdict_get_str(options, BDRV_OPT_FORCE_SHARE), "on")) {
300             error_report("--force-share/-U conflicts with image options");
301             qobject_unref(options);
302             return NULL;
303         }
304         qdict_put_str(options, BDRV_OPT_FORCE_SHARE, "on");
305     }
306     blk = blk_new_open(NULL, NULL, options, flags, &local_err);
307     if (!blk) {
308         error_reportf_err(local_err, "Could not open '%s': ", optstr);
309         return NULL;
310     }
311     blk_set_enable_write_cache(blk, !writethrough);
312 
313     return blk;
314 }
315 
316 static BlockBackend *img_open_file(const char *filename,
317                                    QDict *options,
318                                    const char *fmt, int flags,
319                                    bool writethrough, bool quiet,
320                                    bool force_share)
321 {
322     BlockBackend *blk;
323     Error *local_err = NULL;
324 
325     if (!options) {
326         options = qdict_new();
327     }
328     if (fmt) {
329         qdict_put_str(options, "driver", fmt);
330     }
331 
332     if (force_share) {
333         qdict_put_bool(options, BDRV_OPT_FORCE_SHARE, true);
334     }
335     blk = blk_new_open(filename, NULL, options, flags, &local_err);
336     if (!blk) {
337         error_reportf_err(local_err, "Could not open '%s': ", filename);
338         return NULL;
339     }
340     blk_set_enable_write_cache(blk, !writethrough);
341 
342     return blk;
343 }
344 
345 
346 static int img_add_key_secrets(void *opaque,
347                                const char *name, const char *value,
348                                Error **errp)
349 {
350     QDict *options = opaque;
351 
352     if (g_str_has_suffix(name, "key-secret")) {
353         qdict_put_str(options, name, value);
354     }
355 
356     return 0;
357 }
358 
359 
360 static BlockBackend *img_open(bool image_opts,
361                               const char *filename,
362                               const char *fmt, int flags, bool writethrough,
363                               bool quiet, bool force_share)
364 {
365     BlockBackend *blk;
366     if (image_opts) {
367         QemuOpts *opts;
368         if (fmt) {
369             error_report("--image-opts and --format are mutually exclusive");
370             return NULL;
371         }
372         opts = qemu_opts_parse_noisily(qemu_find_opts("source"),
373                                        filename, true);
374         if (!opts) {
375             return NULL;
376         }
377         blk = img_open_opts(filename, opts, flags, writethrough, quiet,
378                             force_share);
379     } else {
380         blk = img_open_file(filename, NULL, fmt, flags, writethrough, quiet,
381                             force_share);
382     }
383     return blk;
384 }
385 
386 
387 static int add_old_style_options(const char *fmt, QemuOpts *opts,
388                                  const char *base_filename,
389                                  const char *base_fmt)
390 {
391     Error *err = NULL;
392 
393     if (base_filename) {
394         qemu_opt_set(opts, BLOCK_OPT_BACKING_FILE, base_filename, &err);
395         if (err) {
396             error_report("Backing file not supported for file format '%s'",
397                          fmt);
398             error_free(err);
399             return -1;
400         }
401     }
402     if (base_fmt) {
403         qemu_opt_set(opts, BLOCK_OPT_BACKING_FMT, base_fmt, &err);
404         if (err) {
405             error_report("Backing file format not supported for file "
406                          "format '%s'", fmt);
407             error_free(err);
408             return -1;
409         }
410     }
411     return 0;
412 }
413 
414 static int64_t cvtnum(const char *s)
415 {
416     int err;
417     uint64_t value;
418 
419     err = qemu_strtosz(s, NULL, &value);
420     if (err < 0) {
421         return err;
422     }
423     if (value > INT64_MAX) {
424         return -ERANGE;
425     }
426     return value;
427 }
428 
429 static int img_create(int argc, char **argv)
430 {
431     int c;
432     uint64_t img_size = -1;
433     const char *fmt = "raw";
434     const char *base_fmt = NULL;
435     const char *filename;
436     const char *base_filename = NULL;
437     char *options = NULL;
438     Error *local_err = NULL;
439     bool quiet = false;
440     int flags = 0;
441 
442     for(;;) {
443         static const struct option long_options[] = {
444             {"help", no_argument, 0, 'h'},
445             {"object", required_argument, 0, OPTION_OBJECT},
446             {0, 0, 0, 0}
447         };
448         c = getopt_long(argc, argv, ":F:b:f:ho:qu",
449                         long_options, NULL);
450         if (c == -1) {
451             break;
452         }
453         switch(c) {
454         case ':':
455             missing_argument(argv[optind - 1]);
456             break;
457         case '?':
458             unrecognized_option(argv[optind - 1]);
459             break;
460         case 'h':
461             help();
462             break;
463         case 'F':
464             base_fmt = optarg;
465             break;
466         case 'b':
467             base_filename = optarg;
468             break;
469         case 'f':
470             fmt = optarg;
471             break;
472         case 'o':
473             if (!is_valid_option_list(optarg)) {
474                 error_report("Invalid option list: %s", optarg);
475                 goto fail;
476             }
477             if (!options) {
478                 options = g_strdup(optarg);
479             } else {
480                 char *old_options = options;
481                 options = g_strdup_printf("%s,%s", options, optarg);
482                 g_free(old_options);
483             }
484             break;
485         case 'q':
486             quiet = true;
487             break;
488         case 'u':
489             flags |= BDRV_O_NO_BACKING;
490             break;
491         case OPTION_OBJECT: {
492             QemuOpts *opts;
493             opts = qemu_opts_parse_noisily(&qemu_object_opts,
494                                            optarg, true);
495             if (!opts) {
496                 goto fail;
497             }
498         }   break;
499         }
500     }
501 
502     /* Get the filename */
503     filename = (optind < argc) ? argv[optind] : NULL;
504     if (options && has_help_option(options)) {
505         g_free(options);
506         return print_block_option_help(filename, fmt);
507     }
508 
509     if (optind >= argc) {
510         error_exit("Expecting image file name");
511     }
512     optind++;
513 
514     if (qemu_opts_foreach(&qemu_object_opts,
515                           user_creatable_add_opts_foreach,
516                           NULL, &error_fatal)) {
517         goto fail;
518     }
519 
520     /* Get image size, if specified */
521     if (optind < argc) {
522         int64_t sval;
523 
524         sval = cvtnum(argv[optind++]);
525         if (sval < 0) {
526             if (sval == -ERANGE) {
527                 error_report("Image size must be less than 8 EiB!");
528             } else {
529                 error_report("Invalid image size specified! You may use k, M, "
530                       "G, T, P or E suffixes for ");
531                 error_report("kilobytes, megabytes, gigabytes, terabytes, "
532                              "petabytes and exabytes.");
533             }
534             goto fail;
535         }
536         img_size = (uint64_t)sval;
537     }
538     if (optind != argc) {
539         error_exit("Unexpected argument: %s", argv[optind]);
540     }
541 
542     bdrv_img_create(filename, fmt, base_filename, base_fmt,
543                     options, img_size, flags, quiet, &local_err);
544     if (local_err) {
545         error_reportf_err(local_err, "%s: ", filename);
546         goto fail;
547     }
548 
549     g_free(options);
550     return 0;
551 
552 fail:
553     g_free(options);
554     return 1;
555 }
556 
557 static void dump_json_image_check(ImageCheck *check, bool quiet)
558 {
559     QString *str;
560     QObject *obj;
561     Visitor *v = qobject_output_visitor_new(&obj);
562 
563     visit_type_ImageCheck(v, NULL, &check, &error_abort);
564     visit_complete(v, &obj);
565     str = qobject_to_json_pretty(obj);
566     assert(str != NULL);
567     qprintf(quiet, "%s\n", qstring_get_str(str));
568     qobject_unref(obj);
569     visit_free(v);
570     qobject_unref(str);
571 }
572 
573 static void dump_human_image_check(ImageCheck *check, bool quiet)
574 {
575     if (!(check->corruptions || check->leaks || check->check_errors)) {
576         qprintf(quiet, "No errors were found on the image.\n");
577     } else {
578         if (check->corruptions) {
579             qprintf(quiet, "\n%" PRId64 " errors were found on the image.\n"
580                     "Data may be corrupted, or further writes to the image "
581                     "may corrupt it.\n",
582                     check->corruptions);
583         }
584 
585         if (check->leaks) {
586             qprintf(quiet,
587                     "\n%" PRId64 " leaked clusters were found on the image.\n"
588                     "This means waste of disk space, but no harm to data.\n",
589                     check->leaks);
590         }
591 
592         if (check->check_errors) {
593             qprintf(quiet,
594                     "\n%" PRId64
595                     " internal errors have occurred during the check.\n",
596                     check->check_errors);
597         }
598     }
599 
600     if (check->total_clusters != 0 && check->allocated_clusters != 0) {
601         qprintf(quiet, "%" PRId64 "/%" PRId64 " = %0.2f%% allocated, "
602                 "%0.2f%% fragmented, %0.2f%% compressed clusters\n",
603                 check->allocated_clusters, check->total_clusters,
604                 check->allocated_clusters * 100.0 / check->total_clusters,
605                 check->fragmented_clusters * 100.0 / check->allocated_clusters,
606                 check->compressed_clusters * 100.0 /
607                 check->allocated_clusters);
608     }
609 
610     if (check->image_end_offset) {
611         qprintf(quiet,
612                 "Image end offset: %" PRId64 "\n", check->image_end_offset);
613     }
614 }
615 
616 static int collect_image_check(BlockDriverState *bs,
617                    ImageCheck *check,
618                    const char *filename,
619                    const char *fmt,
620                    int fix)
621 {
622     int ret;
623     BdrvCheckResult result;
624 
625     ret = bdrv_check(bs, &result, fix);
626     if (ret < 0) {
627         return ret;
628     }
629 
630     check->filename                 = g_strdup(filename);
631     check->format                   = g_strdup(bdrv_get_format_name(bs));
632     check->check_errors             = result.check_errors;
633     check->corruptions              = result.corruptions;
634     check->has_corruptions          = result.corruptions != 0;
635     check->leaks                    = result.leaks;
636     check->has_leaks                = result.leaks != 0;
637     check->corruptions_fixed        = result.corruptions_fixed;
638     check->has_corruptions_fixed    = result.corruptions != 0;
639     check->leaks_fixed              = result.leaks_fixed;
640     check->has_leaks_fixed          = result.leaks != 0;
641     check->image_end_offset         = result.image_end_offset;
642     check->has_image_end_offset     = result.image_end_offset != 0;
643     check->total_clusters           = result.bfi.total_clusters;
644     check->has_total_clusters       = result.bfi.total_clusters != 0;
645     check->allocated_clusters       = result.bfi.allocated_clusters;
646     check->has_allocated_clusters   = result.bfi.allocated_clusters != 0;
647     check->fragmented_clusters      = result.bfi.fragmented_clusters;
648     check->has_fragmented_clusters  = result.bfi.fragmented_clusters != 0;
649     check->compressed_clusters      = result.bfi.compressed_clusters;
650     check->has_compressed_clusters  = result.bfi.compressed_clusters != 0;
651 
652     return 0;
653 }
654 
655 /*
656  * Checks an image for consistency. Exit codes:
657  *
658  *  0 - Check completed, image is good
659  *  1 - Check not completed because of internal errors
660  *  2 - Check completed, image is corrupted
661  *  3 - Check completed, image has leaked clusters, but is good otherwise
662  * 63 - Checks are not supported by the image format
663  */
664 static int img_check(int argc, char **argv)
665 {
666     int c, ret;
667     OutputFormat output_format = OFORMAT_HUMAN;
668     const char *filename, *fmt, *output, *cache;
669     BlockBackend *blk;
670     BlockDriverState *bs;
671     int fix = 0;
672     int flags = BDRV_O_CHECK;
673     bool writethrough;
674     ImageCheck *check;
675     bool quiet = false;
676     bool image_opts = false;
677     bool force_share = false;
678 
679     fmt = NULL;
680     output = NULL;
681     cache = BDRV_DEFAULT_CACHE;
682 
683     for(;;) {
684         int option_index = 0;
685         static const struct option long_options[] = {
686             {"help", no_argument, 0, 'h'},
687             {"format", required_argument, 0, 'f'},
688             {"repair", required_argument, 0, 'r'},
689             {"output", required_argument, 0, OPTION_OUTPUT},
690             {"object", required_argument, 0, OPTION_OBJECT},
691             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
692             {"force-share", no_argument, 0, 'U'},
693             {0, 0, 0, 0}
694         };
695         c = getopt_long(argc, argv, ":hf:r:T:qU",
696                         long_options, &option_index);
697         if (c == -1) {
698             break;
699         }
700         switch(c) {
701         case ':':
702             missing_argument(argv[optind - 1]);
703             break;
704         case '?':
705             unrecognized_option(argv[optind - 1]);
706             break;
707         case 'h':
708             help();
709             break;
710         case 'f':
711             fmt = optarg;
712             break;
713         case 'r':
714             flags |= BDRV_O_RDWR;
715 
716             if (!strcmp(optarg, "leaks")) {
717                 fix = BDRV_FIX_LEAKS;
718             } else if (!strcmp(optarg, "all")) {
719                 fix = BDRV_FIX_LEAKS | BDRV_FIX_ERRORS;
720             } else {
721                 error_exit("Unknown option value for -r "
722                            "(expecting 'leaks' or 'all'): %s", optarg);
723             }
724             break;
725         case OPTION_OUTPUT:
726             output = optarg;
727             break;
728         case 'T':
729             cache = optarg;
730             break;
731         case 'q':
732             quiet = true;
733             break;
734         case 'U':
735             force_share = true;
736             break;
737         case OPTION_OBJECT: {
738             QemuOpts *opts;
739             opts = qemu_opts_parse_noisily(&qemu_object_opts,
740                                            optarg, true);
741             if (!opts) {
742                 return 1;
743             }
744         }   break;
745         case OPTION_IMAGE_OPTS:
746             image_opts = true;
747             break;
748         }
749     }
750     if (optind != argc - 1) {
751         error_exit("Expecting one image file name");
752     }
753     filename = argv[optind++];
754 
755     if (output && !strcmp(output, "json")) {
756         output_format = OFORMAT_JSON;
757     } else if (output && !strcmp(output, "human")) {
758         output_format = OFORMAT_HUMAN;
759     } else if (output) {
760         error_report("--output must be used with human or json as argument.");
761         return 1;
762     }
763 
764     if (qemu_opts_foreach(&qemu_object_opts,
765                           user_creatable_add_opts_foreach,
766                           NULL, &error_fatal)) {
767         return 1;
768     }
769 
770     ret = bdrv_parse_cache_mode(cache, &flags, &writethrough);
771     if (ret < 0) {
772         error_report("Invalid source cache option: %s", cache);
773         return 1;
774     }
775 
776     blk = img_open(image_opts, filename, fmt, flags, writethrough, quiet,
777                    force_share);
778     if (!blk) {
779         return 1;
780     }
781     bs = blk_bs(blk);
782 
783     check = g_new0(ImageCheck, 1);
784     ret = collect_image_check(bs, check, filename, fmt, fix);
785 
786     if (ret == -ENOTSUP) {
787         error_report("This image format does not support checks");
788         ret = 63;
789         goto fail;
790     }
791 
792     if (check->corruptions_fixed || check->leaks_fixed) {
793         int corruptions_fixed, leaks_fixed;
794 
795         leaks_fixed         = check->leaks_fixed;
796         corruptions_fixed   = check->corruptions_fixed;
797 
798         if (output_format == OFORMAT_HUMAN) {
799             qprintf(quiet,
800                     "The following inconsistencies were found and repaired:\n\n"
801                     "    %" PRId64 " leaked clusters\n"
802                     "    %" PRId64 " corruptions\n\n"
803                     "Double checking the fixed image now...\n",
804                     check->leaks_fixed,
805                     check->corruptions_fixed);
806         }
807 
808         ret = collect_image_check(bs, check, filename, fmt, 0);
809 
810         check->leaks_fixed          = leaks_fixed;
811         check->corruptions_fixed    = corruptions_fixed;
812     }
813 
814     if (!ret) {
815         switch (output_format) {
816         case OFORMAT_HUMAN:
817             dump_human_image_check(check, quiet);
818             break;
819         case OFORMAT_JSON:
820             dump_json_image_check(check, quiet);
821             break;
822         }
823     }
824 
825     if (ret || check->check_errors) {
826         if (ret) {
827             error_report("Check failed: %s", strerror(-ret));
828         } else {
829             error_report("Check failed");
830         }
831         ret = 1;
832         goto fail;
833     }
834 
835     if (check->corruptions) {
836         ret = 2;
837     } else if (check->leaks) {
838         ret = 3;
839     } else {
840         ret = 0;
841     }
842 
843 fail:
844     qapi_free_ImageCheck(check);
845     blk_unref(blk);
846     return ret;
847 }
848 
849 typedef struct CommonBlockJobCBInfo {
850     BlockDriverState *bs;
851     Error **errp;
852 } CommonBlockJobCBInfo;
853 
854 static void common_block_job_cb(void *opaque, int ret)
855 {
856     CommonBlockJobCBInfo *cbi = opaque;
857 
858     if (ret < 0) {
859         error_setg_errno(cbi->errp, -ret, "Block job failed");
860     }
861 }
862 
863 static void run_block_job(BlockJob *job, Error **errp)
864 {
865     AioContext *aio_context = blk_get_aio_context(job->blk);
866     int ret = 0;
867 
868     aio_context_acquire(aio_context);
869     job_ref(&job->job);
870     do {
871         float progress = 0.0f;
872         aio_poll(aio_context, true);
873         if (job->job.progress_total) {
874             progress = (float)job->job.progress_current /
875                        job->job.progress_total * 100.f;
876         }
877         qemu_progress_print(progress, 0);
878     } while (!job_is_ready(&job->job) && !job_is_completed(&job->job));
879 
880     if (!job_is_completed(&job->job)) {
881         ret = job_complete_sync(&job->job, errp);
882     } else {
883         ret = job->job.ret;
884     }
885     job_unref(&job->job);
886     aio_context_release(aio_context);
887 
888     /* publish completion progress only when success */
889     if (!ret) {
890         qemu_progress_print(100.f, 0);
891     }
892 }
893 
894 static int img_commit(int argc, char **argv)
895 {
896     int c, ret, flags;
897     const char *filename, *fmt, *cache, *base;
898     BlockBackend *blk;
899     BlockDriverState *bs, *base_bs;
900     BlockJob *job;
901     bool progress = false, quiet = false, drop = false;
902     bool writethrough;
903     Error *local_err = NULL;
904     CommonBlockJobCBInfo cbi;
905     bool image_opts = false;
906     AioContext *aio_context;
907 
908     fmt = NULL;
909     cache = BDRV_DEFAULT_CACHE;
910     base = NULL;
911     for(;;) {
912         static const struct option long_options[] = {
913             {"help", no_argument, 0, 'h'},
914             {"object", required_argument, 0, OPTION_OBJECT},
915             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
916             {0, 0, 0, 0}
917         };
918         c = getopt_long(argc, argv, ":f:ht:b:dpq",
919                         long_options, NULL);
920         if (c == -1) {
921             break;
922         }
923         switch(c) {
924         case ':':
925             missing_argument(argv[optind - 1]);
926             break;
927         case '?':
928             unrecognized_option(argv[optind - 1]);
929             break;
930         case 'h':
931             help();
932             break;
933         case 'f':
934             fmt = optarg;
935             break;
936         case 't':
937             cache = optarg;
938             break;
939         case 'b':
940             base = optarg;
941             /* -b implies -d */
942             drop = true;
943             break;
944         case 'd':
945             drop = true;
946             break;
947         case 'p':
948             progress = true;
949             break;
950         case 'q':
951             quiet = true;
952             break;
953         case OPTION_OBJECT: {
954             QemuOpts *opts;
955             opts = qemu_opts_parse_noisily(&qemu_object_opts,
956                                            optarg, true);
957             if (!opts) {
958                 return 1;
959             }
960         }   break;
961         case OPTION_IMAGE_OPTS:
962             image_opts = true;
963             break;
964         }
965     }
966 
967     /* Progress is not shown in Quiet mode */
968     if (quiet) {
969         progress = false;
970     }
971 
972     if (optind != argc - 1) {
973         error_exit("Expecting one image file name");
974     }
975     filename = argv[optind++];
976 
977     if (qemu_opts_foreach(&qemu_object_opts,
978                           user_creatable_add_opts_foreach,
979                           NULL, &error_fatal)) {
980         return 1;
981     }
982 
983     flags = BDRV_O_RDWR | BDRV_O_UNMAP;
984     ret = bdrv_parse_cache_mode(cache, &flags, &writethrough);
985     if (ret < 0) {
986         error_report("Invalid cache option: %s", cache);
987         return 1;
988     }
989 
990     blk = img_open(image_opts, filename, fmt, flags, writethrough, quiet,
991                    false);
992     if (!blk) {
993         return 1;
994     }
995     bs = blk_bs(blk);
996 
997     qemu_progress_init(progress, 1.f);
998     qemu_progress_print(0.f, 100);
999 
1000     if (base) {
1001         base_bs = bdrv_find_backing_image(bs, base);
1002         if (!base_bs) {
1003             error_setg(&local_err,
1004                        "Did not find '%s' in the backing chain of '%s'",
1005                        base, filename);
1006             goto done;
1007         }
1008     } else {
1009         /* This is different from QMP, which by default uses the deepest file in
1010          * the backing chain (i.e., the very base); however, the traditional
1011          * behavior of qemu-img commit is using the immediate backing file. */
1012         base_bs = backing_bs(bs);
1013         if (!base_bs) {
1014             error_setg(&local_err, "Image does not have a backing file");
1015             goto done;
1016         }
1017     }
1018 
1019     cbi = (CommonBlockJobCBInfo){
1020         .errp = &local_err,
1021         .bs   = bs,
1022     };
1023 
1024     aio_context = bdrv_get_aio_context(bs);
1025     aio_context_acquire(aio_context);
1026     commit_active_start("commit", bs, base_bs, JOB_DEFAULT, 0,
1027                         BLOCKDEV_ON_ERROR_REPORT, NULL, common_block_job_cb,
1028                         &cbi, false, &local_err);
1029     aio_context_release(aio_context);
1030     if (local_err) {
1031         goto done;
1032     }
1033 
1034     /* When the block job completes, the BlockBackend reference will point to
1035      * the old backing file. In order to avoid that the top image is already
1036      * deleted, so we can still empty it afterwards, increment the reference
1037      * counter here preemptively. */
1038     if (!drop) {
1039         bdrv_ref(bs);
1040     }
1041 
1042     job = block_job_get("commit");
1043     assert(job);
1044     run_block_job(job, &local_err);
1045     if (local_err) {
1046         goto unref_backing;
1047     }
1048 
1049     if (!drop && bs->drv->bdrv_make_empty) {
1050         ret = bs->drv->bdrv_make_empty(bs);
1051         if (ret) {
1052             error_setg_errno(&local_err, -ret, "Could not empty %s",
1053                              filename);
1054             goto unref_backing;
1055         }
1056     }
1057 
1058 unref_backing:
1059     if (!drop) {
1060         bdrv_unref(bs);
1061     }
1062 
1063 done:
1064     qemu_progress_end();
1065 
1066     blk_unref(blk);
1067 
1068     if (local_err) {
1069         error_report_err(local_err);
1070         return 1;
1071     }
1072 
1073     qprintf(quiet, "Image committed.\n");
1074     return 0;
1075 }
1076 
1077 /*
1078  * Returns -1 if 'buf' contains only zeroes, otherwise the byte index
1079  * of the first sector boundary within buf where the sector contains a
1080  * non-zero byte.  This function is robust to a buffer that is not
1081  * sector-aligned.
1082  */
1083 static int64_t find_nonzero(const uint8_t *buf, int64_t n)
1084 {
1085     int64_t i;
1086     int64_t end = QEMU_ALIGN_DOWN(n, BDRV_SECTOR_SIZE);
1087 
1088     for (i = 0; i < end; i += BDRV_SECTOR_SIZE) {
1089         if (!buffer_is_zero(buf + i, BDRV_SECTOR_SIZE)) {
1090             return i;
1091         }
1092     }
1093     if (i < n && !buffer_is_zero(buf + i, n - end)) {
1094         return i;
1095     }
1096     return -1;
1097 }
1098 
1099 /*
1100  * Returns true iff the first sector pointed to by 'buf' contains at least
1101  * a non-NUL byte.
1102  *
1103  * 'pnum' is set to the number of sectors (including and immediately following
1104  * the first one) that are known to be in the same allocated/unallocated state.
1105  * The function will try to align the end offset to alignment boundaries so
1106  * that the request will at least end aligned and consequtive requests will
1107  * also start at an aligned offset.
1108  */
1109 static int is_allocated_sectors(const uint8_t *buf, int n, int *pnum,
1110                                 int64_t sector_num, int alignment)
1111 {
1112     bool is_zero;
1113     int i, tail;
1114 
1115     if (n <= 0) {
1116         *pnum = 0;
1117         return 0;
1118     }
1119     is_zero = buffer_is_zero(buf, 512);
1120     for(i = 1; i < n; i++) {
1121         buf += 512;
1122         if (is_zero != buffer_is_zero(buf, 512)) {
1123             break;
1124         }
1125     }
1126 
1127     tail = (sector_num + i) & (alignment - 1);
1128     if (tail) {
1129         if (is_zero && i <= tail) {
1130             /* treat unallocated areas which only consist
1131              * of a small tail as allocated. */
1132             is_zero = false;
1133         }
1134         if (!is_zero) {
1135             /* align up end offset of allocated areas. */
1136             i += alignment - tail;
1137             i = MIN(i, n);
1138         } else {
1139             /* align down end offset of zero areas. */
1140             i -= tail;
1141         }
1142     }
1143     *pnum = i;
1144     return !is_zero;
1145 }
1146 
1147 /*
1148  * Like is_allocated_sectors, but if the buffer starts with a used sector,
1149  * up to 'min' consecutive sectors containing zeros are ignored. This avoids
1150  * breaking up write requests for only small sparse areas.
1151  */
1152 static int is_allocated_sectors_min(const uint8_t *buf, int n, int *pnum,
1153     int min, int64_t sector_num, int alignment)
1154 {
1155     int ret;
1156     int num_checked, num_used;
1157 
1158     if (n < min) {
1159         min = n;
1160     }
1161 
1162     ret = is_allocated_sectors(buf, n, pnum, sector_num, alignment);
1163     if (!ret) {
1164         return ret;
1165     }
1166 
1167     num_used = *pnum;
1168     buf += BDRV_SECTOR_SIZE * *pnum;
1169     n -= *pnum;
1170     sector_num += *pnum;
1171     num_checked = num_used;
1172 
1173     while (n > 0) {
1174         ret = is_allocated_sectors(buf, n, pnum, sector_num, alignment);
1175 
1176         buf += BDRV_SECTOR_SIZE * *pnum;
1177         n -= *pnum;
1178         sector_num += *pnum;
1179         num_checked += *pnum;
1180         if (ret) {
1181             num_used = num_checked;
1182         } else if (*pnum >= min) {
1183             break;
1184         }
1185     }
1186 
1187     *pnum = num_used;
1188     return 1;
1189 }
1190 
1191 /*
1192  * Compares two buffers sector by sector. Returns 0 if the first
1193  * sector of each buffer matches, non-zero otherwise.
1194  *
1195  * pnum is set to the sector-aligned size of the buffer prefix that
1196  * has the same matching status as the first sector.
1197  */
1198 static int compare_buffers(const uint8_t *buf1, const uint8_t *buf2,
1199                            int64_t bytes, int64_t *pnum)
1200 {
1201     bool res;
1202     int64_t i = MIN(bytes, BDRV_SECTOR_SIZE);
1203 
1204     assert(bytes > 0);
1205 
1206     res = !!memcmp(buf1, buf2, i);
1207     while (i < bytes) {
1208         int64_t len = MIN(bytes - i, BDRV_SECTOR_SIZE);
1209 
1210         if (!!memcmp(buf1 + i, buf2 + i, len) != res) {
1211             break;
1212         }
1213         i += len;
1214     }
1215 
1216     *pnum = i;
1217     return res;
1218 }
1219 
1220 #define IO_BUF_SIZE (2 * MiB)
1221 
1222 /*
1223  * Check if passed sectors are empty (not allocated or contain only 0 bytes)
1224  *
1225  * Intended for use by 'qemu-img compare': Returns 0 in case sectors are
1226  * filled with 0, 1 if sectors contain non-zero data (this is a comparison
1227  * failure), and 4 on error (the exit status for read errors), after emitting
1228  * an error message.
1229  *
1230  * @param blk:  BlockBackend for the image
1231  * @param offset: Starting offset to check
1232  * @param bytes: Number of bytes to check
1233  * @param filename: Name of disk file we are checking (logging purpose)
1234  * @param buffer: Allocated buffer for storing read data
1235  * @param quiet: Flag for quiet mode
1236  */
1237 static int check_empty_sectors(BlockBackend *blk, int64_t offset,
1238                                int64_t bytes, const char *filename,
1239                                uint8_t *buffer, bool quiet)
1240 {
1241     int ret = 0;
1242     int64_t idx;
1243 
1244     ret = blk_pread(blk, offset, buffer, bytes);
1245     if (ret < 0) {
1246         error_report("Error while reading offset %" PRId64 " of %s: %s",
1247                      offset, filename, strerror(-ret));
1248         return 4;
1249     }
1250     idx = find_nonzero(buffer, bytes);
1251     if (idx >= 0) {
1252         qprintf(quiet, "Content mismatch at offset %" PRId64 "!\n",
1253                 offset + idx);
1254         return 1;
1255     }
1256 
1257     return 0;
1258 }
1259 
1260 /*
1261  * Compares two images. Exit codes:
1262  *
1263  * 0 - Images are identical
1264  * 1 - Images differ
1265  * >1 - Error occurred
1266  */
1267 static int img_compare(int argc, char **argv)
1268 {
1269     const char *fmt1 = NULL, *fmt2 = NULL, *cache, *filename1, *filename2;
1270     BlockBackend *blk1, *blk2;
1271     BlockDriverState *bs1, *bs2;
1272     int64_t total_size1, total_size2;
1273     uint8_t *buf1 = NULL, *buf2 = NULL;
1274     int64_t pnum1, pnum2;
1275     int allocated1, allocated2;
1276     int ret = 0; /* return value - 0 Ident, 1 Different, >1 Error */
1277     bool progress = false, quiet = false, strict = false;
1278     int flags;
1279     bool writethrough;
1280     int64_t total_size;
1281     int64_t offset = 0;
1282     int64_t chunk;
1283     int c;
1284     uint64_t progress_base;
1285     bool image_opts = false;
1286     bool force_share = false;
1287 
1288     cache = BDRV_DEFAULT_CACHE;
1289     for (;;) {
1290         static const struct option long_options[] = {
1291             {"help", no_argument, 0, 'h'},
1292             {"object", required_argument, 0, OPTION_OBJECT},
1293             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
1294             {"force-share", no_argument, 0, 'U'},
1295             {0, 0, 0, 0}
1296         };
1297         c = getopt_long(argc, argv, ":hf:F:T:pqsU",
1298                         long_options, NULL);
1299         if (c == -1) {
1300             break;
1301         }
1302         switch (c) {
1303         case ':':
1304             missing_argument(argv[optind - 1]);
1305             break;
1306         case '?':
1307             unrecognized_option(argv[optind - 1]);
1308             break;
1309         case 'h':
1310             help();
1311             break;
1312         case 'f':
1313             fmt1 = optarg;
1314             break;
1315         case 'F':
1316             fmt2 = optarg;
1317             break;
1318         case 'T':
1319             cache = optarg;
1320             break;
1321         case 'p':
1322             progress = true;
1323             break;
1324         case 'q':
1325             quiet = true;
1326             break;
1327         case 's':
1328             strict = true;
1329             break;
1330         case 'U':
1331             force_share = true;
1332             break;
1333         case OPTION_OBJECT: {
1334             QemuOpts *opts;
1335             opts = qemu_opts_parse_noisily(&qemu_object_opts,
1336                                            optarg, true);
1337             if (!opts) {
1338                 ret = 2;
1339                 goto out4;
1340             }
1341         }   break;
1342         case OPTION_IMAGE_OPTS:
1343             image_opts = true;
1344             break;
1345         }
1346     }
1347 
1348     /* Progress is not shown in Quiet mode */
1349     if (quiet) {
1350         progress = false;
1351     }
1352 
1353 
1354     if (optind != argc - 2) {
1355         error_exit("Expecting two image file names");
1356     }
1357     filename1 = argv[optind++];
1358     filename2 = argv[optind++];
1359 
1360     if (qemu_opts_foreach(&qemu_object_opts,
1361                           user_creatable_add_opts_foreach,
1362                           NULL, &error_fatal)) {
1363         ret = 2;
1364         goto out4;
1365     }
1366 
1367     /* Initialize before goto out */
1368     qemu_progress_init(progress, 2.0);
1369 
1370     flags = 0;
1371     ret = bdrv_parse_cache_mode(cache, &flags, &writethrough);
1372     if (ret < 0) {
1373         error_report("Invalid source cache option: %s", cache);
1374         ret = 2;
1375         goto out3;
1376     }
1377 
1378     blk1 = img_open(image_opts, filename1, fmt1, flags, writethrough, quiet,
1379                     force_share);
1380     if (!blk1) {
1381         ret = 2;
1382         goto out3;
1383     }
1384 
1385     blk2 = img_open(image_opts, filename2, fmt2, flags, writethrough, quiet,
1386                     force_share);
1387     if (!blk2) {
1388         ret = 2;
1389         goto out2;
1390     }
1391     bs1 = blk_bs(blk1);
1392     bs2 = blk_bs(blk2);
1393 
1394     buf1 = blk_blockalign(blk1, IO_BUF_SIZE);
1395     buf2 = blk_blockalign(blk2, IO_BUF_SIZE);
1396     total_size1 = blk_getlength(blk1);
1397     if (total_size1 < 0) {
1398         error_report("Can't get size of %s: %s",
1399                      filename1, strerror(-total_size1));
1400         ret = 4;
1401         goto out;
1402     }
1403     total_size2 = blk_getlength(blk2);
1404     if (total_size2 < 0) {
1405         error_report("Can't get size of %s: %s",
1406                      filename2, strerror(-total_size2));
1407         ret = 4;
1408         goto out;
1409     }
1410     total_size = MIN(total_size1, total_size2);
1411     progress_base = MAX(total_size1, total_size2);
1412 
1413     qemu_progress_print(0, 100);
1414 
1415     if (strict && total_size1 != total_size2) {
1416         ret = 1;
1417         qprintf(quiet, "Strict mode: Image size mismatch!\n");
1418         goto out;
1419     }
1420 
1421     while (offset < total_size) {
1422         int status1, status2;
1423 
1424         status1 = bdrv_block_status_above(bs1, NULL, offset,
1425                                           total_size1 - offset, &pnum1, NULL,
1426                                           NULL);
1427         if (status1 < 0) {
1428             ret = 3;
1429             error_report("Sector allocation test failed for %s", filename1);
1430             goto out;
1431         }
1432         allocated1 = status1 & BDRV_BLOCK_ALLOCATED;
1433 
1434         status2 = bdrv_block_status_above(bs2, NULL, offset,
1435                                           total_size2 - offset, &pnum2, NULL,
1436                                           NULL);
1437         if (status2 < 0) {
1438             ret = 3;
1439             error_report("Sector allocation test failed for %s", filename2);
1440             goto out;
1441         }
1442         allocated2 = status2 & BDRV_BLOCK_ALLOCATED;
1443 
1444         assert(pnum1 && pnum2);
1445         chunk = MIN(pnum1, pnum2);
1446 
1447         if (strict) {
1448             if (status1 != status2) {
1449                 ret = 1;
1450                 qprintf(quiet, "Strict mode: Offset %" PRId64
1451                         " block status mismatch!\n", offset);
1452                 goto out;
1453             }
1454         }
1455         if ((status1 & BDRV_BLOCK_ZERO) && (status2 & BDRV_BLOCK_ZERO)) {
1456             /* nothing to do */
1457         } else if (allocated1 == allocated2) {
1458             if (allocated1) {
1459                 int64_t pnum;
1460 
1461                 chunk = MIN(chunk, IO_BUF_SIZE);
1462                 ret = blk_pread(blk1, offset, buf1, chunk);
1463                 if (ret < 0) {
1464                     error_report("Error while reading offset %" PRId64
1465                                  " of %s: %s",
1466                                  offset, filename1, strerror(-ret));
1467                     ret = 4;
1468                     goto out;
1469                 }
1470                 ret = blk_pread(blk2, offset, buf2, chunk);
1471                 if (ret < 0) {
1472                     error_report("Error while reading offset %" PRId64
1473                                  " of %s: %s",
1474                                  offset, filename2, strerror(-ret));
1475                     ret = 4;
1476                     goto out;
1477                 }
1478                 ret = compare_buffers(buf1, buf2, chunk, &pnum);
1479                 if (ret || pnum != chunk) {
1480                     qprintf(quiet, "Content mismatch at offset %" PRId64 "!\n",
1481                             offset + (ret ? 0 : pnum));
1482                     ret = 1;
1483                     goto out;
1484                 }
1485             }
1486         } else {
1487             chunk = MIN(chunk, IO_BUF_SIZE);
1488             if (allocated1) {
1489                 ret = check_empty_sectors(blk1, offset, chunk,
1490                                           filename1, buf1, quiet);
1491             } else {
1492                 ret = check_empty_sectors(blk2, offset, chunk,
1493                                           filename2, buf1, quiet);
1494             }
1495             if (ret) {
1496                 goto out;
1497             }
1498         }
1499         offset += chunk;
1500         qemu_progress_print(((float) chunk / progress_base) * 100, 100);
1501     }
1502 
1503     if (total_size1 != total_size2) {
1504         BlockBackend *blk_over;
1505         const char *filename_over;
1506 
1507         qprintf(quiet, "Warning: Image size mismatch!\n");
1508         if (total_size1 > total_size2) {
1509             blk_over = blk1;
1510             filename_over = filename1;
1511         } else {
1512             blk_over = blk2;
1513             filename_over = filename2;
1514         }
1515 
1516         while (offset < progress_base) {
1517             ret = bdrv_block_status_above(blk_bs(blk_over), NULL, offset,
1518                                           progress_base - offset, &chunk,
1519                                           NULL, NULL);
1520             if (ret < 0) {
1521                 ret = 3;
1522                 error_report("Sector allocation test failed for %s",
1523                              filename_over);
1524                 goto out;
1525 
1526             }
1527             if (ret & BDRV_BLOCK_ALLOCATED && !(ret & BDRV_BLOCK_ZERO)) {
1528                 chunk = MIN(chunk, IO_BUF_SIZE);
1529                 ret = check_empty_sectors(blk_over, offset, chunk,
1530                                           filename_over, buf1, quiet);
1531                 if (ret) {
1532                     goto out;
1533                 }
1534             }
1535             offset += chunk;
1536             qemu_progress_print(((float) chunk / progress_base) * 100, 100);
1537         }
1538     }
1539 
1540     qprintf(quiet, "Images are identical.\n");
1541     ret = 0;
1542 
1543 out:
1544     qemu_vfree(buf1);
1545     qemu_vfree(buf2);
1546     blk_unref(blk2);
1547 out2:
1548     blk_unref(blk1);
1549 out3:
1550     qemu_progress_end();
1551 out4:
1552     return ret;
1553 }
1554 
1555 enum ImgConvertBlockStatus {
1556     BLK_DATA,
1557     BLK_ZERO,
1558     BLK_BACKING_FILE,
1559 };
1560 
1561 #define MAX_COROUTINES 16
1562 
1563 typedef struct ImgConvertState {
1564     BlockBackend **src;
1565     int64_t *src_sectors;
1566     int src_num;
1567     int64_t total_sectors;
1568     int64_t allocated_sectors;
1569     int64_t allocated_done;
1570     int64_t sector_num;
1571     int64_t wr_offs;
1572     enum ImgConvertBlockStatus status;
1573     int64_t sector_next_status;
1574     BlockBackend *target;
1575     bool has_zero_init;
1576     bool compressed;
1577     bool unallocated_blocks_are_zero;
1578     bool target_has_backing;
1579     int64_t target_backing_sectors; /* negative if unknown */
1580     bool wr_in_order;
1581     bool copy_range;
1582     int min_sparse;
1583     int alignment;
1584     size_t cluster_sectors;
1585     size_t buf_sectors;
1586     long num_coroutines;
1587     int running_coroutines;
1588     Coroutine *co[MAX_COROUTINES];
1589     int64_t wait_sector_num[MAX_COROUTINES];
1590     CoMutex lock;
1591     int ret;
1592 } ImgConvertState;
1593 
1594 static void convert_select_part(ImgConvertState *s, int64_t sector_num,
1595                                 int *src_cur, int64_t *src_cur_offset)
1596 {
1597     *src_cur = 0;
1598     *src_cur_offset = 0;
1599     while (sector_num - *src_cur_offset >= s->src_sectors[*src_cur]) {
1600         *src_cur_offset += s->src_sectors[*src_cur];
1601         (*src_cur)++;
1602         assert(*src_cur < s->src_num);
1603     }
1604 }
1605 
1606 static int convert_iteration_sectors(ImgConvertState *s, int64_t sector_num)
1607 {
1608     int64_t src_cur_offset;
1609     int ret, n, src_cur;
1610     bool post_backing_zero = false;
1611 
1612     convert_select_part(s, sector_num, &src_cur, &src_cur_offset);
1613 
1614     assert(s->total_sectors > sector_num);
1615     n = MIN(s->total_sectors - sector_num, BDRV_REQUEST_MAX_SECTORS);
1616 
1617     if (s->target_backing_sectors >= 0) {
1618         if (sector_num >= s->target_backing_sectors) {
1619             post_backing_zero = s->unallocated_blocks_are_zero;
1620         } else if (sector_num + n > s->target_backing_sectors) {
1621             /* Split requests around target_backing_sectors (because
1622              * starting from there, zeros are handled differently) */
1623             n = s->target_backing_sectors - sector_num;
1624         }
1625     }
1626 
1627     if (s->sector_next_status <= sector_num) {
1628         int64_t count = n * BDRV_SECTOR_SIZE;
1629 
1630         if (s->target_has_backing) {
1631 
1632             ret = bdrv_block_status(blk_bs(s->src[src_cur]),
1633                                     (sector_num - src_cur_offset) *
1634                                     BDRV_SECTOR_SIZE,
1635                                     count, &count, NULL, NULL);
1636         } else {
1637             ret = bdrv_block_status_above(blk_bs(s->src[src_cur]), NULL,
1638                                           (sector_num - src_cur_offset) *
1639                                           BDRV_SECTOR_SIZE,
1640                                           count, &count, NULL, NULL);
1641         }
1642         if (ret < 0) {
1643             error_report("error while reading block status of sector %" PRId64
1644                          ": %s", sector_num, strerror(-ret));
1645             return ret;
1646         }
1647         n = DIV_ROUND_UP(count, BDRV_SECTOR_SIZE);
1648 
1649         if (ret & BDRV_BLOCK_ZERO) {
1650             s->status = post_backing_zero ? BLK_BACKING_FILE : BLK_ZERO;
1651         } else if (ret & BDRV_BLOCK_DATA) {
1652             s->status = BLK_DATA;
1653         } else {
1654             s->status = s->target_has_backing ? BLK_BACKING_FILE : BLK_DATA;
1655         }
1656 
1657         s->sector_next_status = sector_num + n;
1658     }
1659 
1660     n = MIN(n, s->sector_next_status - sector_num);
1661     if (s->status == BLK_DATA) {
1662         n = MIN(n, s->buf_sectors);
1663     }
1664 
1665     /* We need to write complete clusters for compressed images, so if an
1666      * unallocated area is shorter than that, we must consider the whole
1667      * cluster allocated. */
1668     if (s->compressed) {
1669         if (n < s->cluster_sectors) {
1670             n = MIN(s->cluster_sectors, s->total_sectors - sector_num);
1671             s->status = BLK_DATA;
1672         } else {
1673             n = QEMU_ALIGN_DOWN(n, s->cluster_sectors);
1674         }
1675     }
1676 
1677     return n;
1678 }
1679 
1680 static int coroutine_fn convert_co_read(ImgConvertState *s, int64_t sector_num,
1681                                         int nb_sectors, uint8_t *buf)
1682 {
1683     int n, ret;
1684 
1685     assert(nb_sectors <= s->buf_sectors);
1686     while (nb_sectors > 0) {
1687         BlockBackend *blk;
1688         int src_cur;
1689         int64_t bs_sectors, src_cur_offset;
1690 
1691         /* In the case of compression with multiple source files, we can get a
1692          * nb_sectors that spreads into the next part. So we must be able to
1693          * read across multiple BDSes for one convert_read() call. */
1694         convert_select_part(s, sector_num, &src_cur, &src_cur_offset);
1695         blk = s->src[src_cur];
1696         bs_sectors = s->src_sectors[src_cur];
1697 
1698         n = MIN(nb_sectors, bs_sectors - (sector_num - src_cur_offset));
1699 
1700         ret = blk_co_pread(
1701                 blk, (sector_num - src_cur_offset) << BDRV_SECTOR_BITS,
1702                 n << BDRV_SECTOR_BITS, buf, 0);
1703         if (ret < 0) {
1704             return ret;
1705         }
1706 
1707         sector_num += n;
1708         nb_sectors -= n;
1709         buf += n * BDRV_SECTOR_SIZE;
1710     }
1711 
1712     return 0;
1713 }
1714 
1715 
1716 static int coroutine_fn convert_co_write(ImgConvertState *s, int64_t sector_num,
1717                                          int nb_sectors, uint8_t *buf,
1718                                          enum ImgConvertBlockStatus status)
1719 {
1720     int ret;
1721 
1722     while (nb_sectors > 0) {
1723         int n = nb_sectors;
1724         BdrvRequestFlags flags = s->compressed ? BDRV_REQ_WRITE_COMPRESSED : 0;
1725 
1726         switch (status) {
1727         case BLK_BACKING_FILE:
1728             /* If we have a backing file, leave clusters unallocated that are
1729              * unallocated in the source image, so that the backing file is
1730              * visible at the respective offset. */
1731             assert(s->target_has_backing);
1732             break;
1733 
1734         case BLK_DATA:
1735             /* If we're told to keep the target fully allocated (-S 0) or there
1736              * is real non-zero data, we must write it. Otherwise we can treat
1737              * it as zero sectors.
1738              * Compressed clusters need to be written as a whole, so in that
1739              * case we can only save the write if the buffer is completely
1740              * zeroed. */
1741             if (!s->min_sparse ||
1742                 (!s->compressed &&
1743                  is_allocated_sectors_min(buf, n, &n, s->min_sparse,
1744                                           sector_num, s->alignment)) ||
1745                 (s->compressed &&
1746                  !buffer_is_zero(buf, n * BDRV_SECTOR_SIZE)))
1747             {
1748                 ret = blk_co_pwrite(s->target, sector_num << BDRV_SECTOR_BITS,
1749                                     n << BDRV_SECTOR_BITS, buf, flags);
1750                 if (ret < 0) {
1751                     return ret;
1752                 }
1753                 break;
1754             }
1755             /* fall-through */
1756 
1757         case BLK_ZERO:
1758             if (s->has_zero_init) {
1759                 assert(!s->target_has_backing);
1760                 break;
1761             }
1762             ret = blk_co_pwrite_zeroes(s->target,
1763                                        sector_num << BDRV_SECTOR_BITS,
1764                                        n << BDRV_SECTOR_BITS,
1765                                        BDRV_REQ_MAY_UNMAP);
1766             if (ret < 0) {
1767                 return ret;
1768             }
1769             break;
1770         }
1771 
1772         sector_num += n;
1773         nb_sectors -= n;
1774         buf += n * BDRV_SECTOR_SIZE;
1775     }
1776 
1777     return 0;
1778 }
1779 
1780 static int coroutine_fn convert_co_copy_range(ImgConvertState *s, int64_t sector_num,
1781                                               int nb_sectors)
1782 {
1783     int n, ret;
1784 
1785     while (nb_sectors > 0) {
1786         BlockBackend *blk;
1787         int src_cur;
1788         int64_t bs_sectors, src_cur_offset;
1789         int64_t offset;
1790 
1791         convert_select_part(s, sector_num, &src_cur, &src_cur_offset);
1792         offset = (sector_num - src_cur_offset) << BDRV_SECTOR_BITS;
1793         blk = s->src[src_cur];
1794         bs_sectors = s->src_sectors[src_cur];
1795 
1796         n = MIN(nb_sectors, bs_sectors - (sector_num - src_cur_offset));
1797 
1798         ret = blk_co_copy_range(blk, offset, s->target,
1799                                 sector_num << BDRV_SECTOR_BITS,
1800                                 n << BDRV_SECTOR_BITS, 0, 0);
1801         if (ret < 0) {
1802             return ret;
1803         }
1804 
1805         sector_num += n;
1806         nb_sectors -= n;
1807     }
1808     return 0;
1809 }
1810 
1811 static void coroutine_fn convert_co_do_copy(void *opaque)
1812 {
1813     ImgConvertState *s = opaque;
1814     uint8_t *buf = NULL;
1815     int ret, i;
1816     int index = -1;
1817 
1818     for (i = 0; i < s->num_coroutines; i++) {
1819         if (s->co[i] == qemu_coroutine_self()) {
1820             index = i;
1821             break;
1822         }
1823     }
1824     assert(index >= 0);
1825 
1826     s->running_coroutines++;
1827     buf = blk_blockalign(s->target, s->buf_sectors * BDRV_SECTOR_SIZE);
1828 
1829     while (1) {
1830         int n;
1831         int64_t sector_num;
1832         enum ImgConvertBlockStatus status;
1833         bool copy_range;
1834 
1835         qemu_co_mutex_lock(&s->lock);
1836         if (s->ret != -EINPROGRESS || s->sector_num >= s->total_sectors) {
1837             qemu_co_mutex_unlock(&s->lock);
1838             break;
1839         }
1840         n = convert_iteration_sectors(s, s->sector_num);
1841         if (n < 0) {
1842             qemu_co_mutex_unlock(&s->lock);
1843             s->ret = n;
1844             break;
1845         }
1846         /* save current sector and allocation status to local variables */
1847         sector_num = s->sector_num;
1848         status = s->status;
1849         if (!s->min_sparse && s->status == BLK_ZERO) {
1850             n = MIN(n, s->buf_sectors);
1851         }
1852         /* increment global sector counter so that other coroutines can
1853          * already continue reading beyond this request */
1854         s->sector_num += n;
1855         qemu_co_mutex_unlock(&s->lock);
1856 
1857         if (status == BLK_DATA || (!s->min_sparse && status == BLK_ZERO)) {
1858             s->allocated_done += n;
1859             qemu_progress_print(100.0 * s->allocated_done /
1860                                         s->allocated_sectors, 0);
1861         }
1862 
1863 retry:
1864         copy_range = s->copy_range && s->status == BLK_DATA;
1865         if (status == BLK_DATA && !copy_range) {
1866             ret = convert_co_read(s, sector_num, n, buf);
1867             if (ret < 0) {
1868                 error_report("error while reading sector %" PRId64
1869                              ": %s", sector_num, strerror(-ret));
1870                 s->ret = ret;
1871             }
1872         } else if (!s->min_sparse && status == BLK_ZERO) {
1873             status = BLK_DATA;
1874             memset(buf, 0x00, n * BDRV_SECTOR_SIZE);
1875         }
1876 
1877         if (s->wr_in_order) {
1878             /* keep writes in order */
1879             while (s->wr_offs != sector_num && s->ret == -EINPROGRESS) {
1880                 s->wait_sector_num[index] = sector_num;
1881                 qemu_coroutine_yield();
1882             }
1883             s->wait_sector_num[index] = -1;
1884         }
1885 
1886         if (s->ret == -EINPROGRESS) {
1887             if (copy_range) {
1888                 ret = convert_co_copy_range(s, sector_num, n);
1889                 if (ret) {
1890                     s->copy_range = false;
1891                     goto retry;
1892                 }
1893             } else {
1894                 ret = convert_co_write(s, sector_num, n, buf, status);
1895             }
1896             if (ret < 0) {
1897                 error_report("error while writing sector %" PRId64
1898                              ": %s", sector_num, strerror(-ret));
1899                 s->ret = ret;
1900             }
1901         }
1902 
1903         if (s->wr_in_order) {
1904             /* reenter the coroutine that might have waited
1905              * for this write to complete */
1906             s->wr_offs = sector_num + n;
1907             for (i = 0; i < s->num_coroutines; i++) {
1908                 if (s->co[i] && s->wait_sector_num[i] == s->wr_offs) {
1909                     /*
1910                      * A -> B -> A cannot occur because A has
1911                      * s->wait_sector_num[i] == -1 during A -> B.  Therefore
1912                      * B will never enter A during this time window.
1913                      */
1914                     qemu_coroutine_enter(s->co[i]);
1915                     break;
1916                 }
1917             }
1918         }
1919     }
1920 
1921     qemu_vfree(buf);
1922     s->co[index] = NULL;
1923     s->running_coroutines--;
1924     if (!s->running_coroutines && s->ret == -EINPROGRESS) {
1925         /* the convert job finished successfully */
1926         s->ret = 0;
1927     }
1928 }
1929 
1930 static int convert_do_copy(ImgConvertState *s)
1931 {
1932     int ret, i, n;
1933     int64_t sector_num = 0;
1934 
1935     /* Check whether we have zero initialisation or can get it efficiently */
1936     s->has_zero_init = s->min_sparse && !s->target_has_backing
1937                      ? bdrv_has_zero_init(blk_bs(s->target))
1938                      : false;
1939 
1940     if (!s->has_zero_init && !s->target_has_backing &&
1941         bdrv_can_write_zeroes_with_unmap(blk_bs(s->target)))
1942     {
1943         ret = blk_make_zero(s->target, BDRV_REQ_MAY_UNMAP | BDRV_REQ_NO_FALLBACK);
1944         if (ret == 0) {
1945             s->has_zero_init = true;
1946         }
1947     }
1948 
1949     /* Allocate buffer for copied data. For compressed images, only one cluster
1950      * can be copied at a time. */
1951     if (s->compressed) {
1952         if (s->cluster_sectors <= 0 || s->cluster_sectors > s->buf_sectors) {
1953             error_report("invalid cluster size");
1954             return -EINVAL;
1955         }
1956         s->buf_sectors = s->cluster_sectors;
1957     }
1958 
1959     while (sector_num < s->total_sectors) {
1960         n = convert_iteration_sectors(s, sector_num);
1961         if (n < 0) {
1962             return n;
1963         }
1964         if (s->status == BLK_DATA || (!s->min_sparse && s->status == BLK_ZERO))
1965         {
1966             s->allocated_sectors += n;
1967         }
1968         sector_num += n;
1969     }
1970 
1971     /* Do the copy */
1972     s->sector_next_status = 0;
1973     s->ret = -EINPROGRESS;
1974 
1975     qemu_co_mutex_init(&s->lock);
1976     for (i = 0; i < s->num_coroutines; i++) {
1977         s->co[i] = qemu_coroutine_create(convert_co_do_copy, s);
1978         s->wait_sector_num[i] = -1;
1979         qemu_coroutine_enter(s->co[i]);
1980     }
1981 
1982     while (s->running_coroutines) {
1983         main_loop_wait(false);
1984     }
1985 
1986     if (s->compressed && !s->ret) {
1987         /* signal EOF to align */
1988         ret = blk_pwrite_compressed(s->target, 0, NULL, 0);
1989         if (ret < 0) {
1990             return ret;
1991         }
1992     }
1993 
1994     return s->ret;
1995 }
1996 
1997 #define MAX_BUF_SECTORS 32768
1998 
1999 static int img_convert(int argc, char **argv)
2000 {
2001     int c, bs_i, flags, src_flags = 0;
2002     const char *fmt = NULL, *out_fmt = NULL, *cache = "unsafe",
2003                *src_cache = BDRV_DEFAULT_CACHE, *out_baseimg = NULL,
2004                *out_filename, *out_baseimg_param, *snapshot_name = NULL;
2005     BlockDriver *drv = NULL, *proto_drv = NULL;
2006     BlockDriverInfo bdi;
2007     BlockDriverState *out_bs;
2008     QemuOpts *opts = NULL, *sn_opts = NULL;
2009     QemuOptsList *create_opts = NULL;
2010     QDict *open_opts = NULL;
2011     char *options = NULL;
2012     Error *local_err = NULL;
2013     bool writethrough, src_writethrough, quiet = false, image_opts = false,
2014          skip_create = false, progress = false, tgt_image_opts = false;
2015     int64_t ret = -EINVAL;
2016     bool force_share = false;
2017     bool explict_min_sparse = false;
2018 
2019     ImgConvertState s = (ImgConvertState) {
2020         /* Need at least 4k of zeros for sparse detection */
2021         .min_sparse         = 8,
2022         .copy_range         = false,
2023         .buf_sectors        = IO_BUF_SIZE / BDRV_SECTOR_SIZE,
2024         .wr_in_order        = true,
2025         .num_coroutines     = 8,
2026     };
2027 
2028     for(;;) {
2029         static const struct option long_options[] = {
2030             {"help", no_argument, 0, 'h'},
2031             {"object", required_argument, 0, OPTION_OBJECT},
2032             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
2033             {"force-share", no_argument, 0, 'U'},
2034             {"target-image-opts", no_argument, 0, OPTION_TARGET_IMAGE_OPTS},
2035             {0, 0, 0, 0}
2036         };
2037         c = getopt_long(argc, argv, ":hf:O:B:Cco:l:S:pt:T:qnm:WU",
2038                         long_options, NULL);
2039         if (c == -1) {
2040             break;
2041         }
2042         switch(c) {
2043         case ':':
2044             missing_argument(argv[optind - 1]);
2045             break;
2046         case '?':
2047             unrecognized_option(argv[optind - 1]);
2048             break;
2049         case 'h':
2050             help();
2051             break;
2052         case 'f':
2053             fmt = optarg;
2054             break;
2055         case 'O':
2056             out_fmt = optarg;
2057             break;
2058         case 'B':
2059             out_baseimg = optarg;
2060             break;
2061         case 'C':
2062             s.copy_range = true;
2063             break;
2064         case 'c':
2065             s.compressed = true;
2066             break;
2067         case 'o':
2068             if (!is_valid_option_list(optarg)) {
2069                 error_report("Invalid option list: %s", optarg);
2070                 goto fail_getopt;
2071             }
2072             if (!options) {
2073                 options = g_strdup(optarg);
2074             } else {
2075                 char *old_options = options;
2076                 options = g_strdup_printf("%s,%s", options, optarg);
2077                 g_free(old_options);
2078             }
2079             break;
2080         case 'l':
2081             if (strstart(optarg, SNAPSHOT_OPT_BASE, NULL)) {
2082                 sn_opts = qemu_opts_parse_noisily(&internal_snapshot_opts,
2083                                                   optarg, false);
2084                 if (!sn_opts) {
2085                     error_report("Failed in parsing snapshot param '%s'",
2086                                  optarg);
2087                     goto fail_getopt;
2088                 }
2089             } else {
2090                 snapshot_name = optarg;
2091             }
2092             break;
2093         case 'S':
2094         {
2095             int64_t sval;
2096 
2097             sval = cvtnum(optarg);
2098             if (sval < 0 || sval & (BDRV_SECTOR_SIZE - 1) ||
2099                 sval / BDRV_SECTOR_SIZE > MAX_BUF_SECTORS) {
2100                 error_report("Invalid buffer size for sparse output specified. "
2101                     "Valid sizes are multiples of %llu up to %llu. Select "
2102                     "0 to disable sparse detection (fully allocates output).",
2103                     BDRV_SECTOR_SIZE, MAX_BUF_SECTORS * BDRV_SECTOR_SIZE);
2104                 goto fail_getopt;
2105             }
2106 
2107             s.min_sparse = sval / BDRV_SECTOR_SIZE;
2108             explict_min_sparse = true;
2109             break;
2110         }
2111         case 'p':
2112             progress = true;
2113             break;
2114         case 't':
2115             cache = optarg;
2116             break;
2117         case 'T':
2118             src_cache = optarg;
2119             break;
2120         case 'q':
2121             quiet = true;
2122             break;
2123         case 'n':
2124             skip_create = true;
2125             break;
2126         case 'm':
2127             if (qemu_strtol(optarg, NULL, 0, &s.num_coroutines) ||
2128                 s.num_coroutines < 1 || s.num_coroutines > MAX_COROUTINES) {
2129                 error_report("Invalid number of coroutines. Allowed number of"
2130                              " coroutines is between 1 and %d", MAX_COROUTINES);
2131                 goto fail_getopt;
2132             }
2133             break;
2134         case 'W':
2135             s.wr_in_order = false;
2136             break;
2137         case 'U':
2138             force_share = true;
2139             break;
2140         case OPTION_OBJECT: {
2141             QemuOpts *object_opts;
2142             object_opts = qemu_opts_parse_noisily(&qemu_object_opts,
2143                                                   optarg, true);
2144             if (!object_opts) {
2145                 goto fail_getopt;
2146             }
2147             break;
2148         }
2149         case OPTION_IMAGE_OPTS:
2150             image_opts = true;
2151             break;
2152         case OPTION_TARGET_IMAGE_OPTS:
2153             tgt_image_opts = true;
2154             break;
2155         }
2156     }
2157 
2158     if (!out_fmt && !tgt_image_opts) {
2159         out_fmt = "raw";
2160     }
2161 
2162     if (qemu_opts_foreach(&qemu_object_opts,
2163                           user_creatable_add_opts_foreach,
2164                           NULL, &error_fatal)) {
2165         goto fail_getopt;
2166     }
2167 
2168     if (s.compressed && s.copy_range) {
2169         error_report("Cannot enable copy offloading when -c is used");
2170         goto fail_getopt;
2171     }
2172 
2173     if (explict_min_sparse && s.copy_range) {
2174         error_report("Cannot enable copy offloading when -S is used");
2175         goto fail_getopt;
2176     }
2177 
2178     if (tgt_image_opts && !skip_create) {
2179         error_report("--target-image-opts requires use of -n flag");
2180         goto fail_getopt;
2181     }
2182 
2183     s.src_num = argc - optind - 1;
2184     out_filename = s.src_num >= 1 ? argv[argc - 1] : NULL;
2185 
2186     if (options && has_help_option(options)) {
2187         if (out_fmt) {
2188             ret = print_block_option_help(out_filename, out_fmt);
2189             goto fail_getopt;
2190         } else {
2191             error_report("Option help requires a format be specified");
2192             goto fail_getopt;
2193         }
2194     }
2195 
2196     if (s.src_num < 1) {
2197         error_report("Must specify image file name");
2198         goto fail_getopt;
2199     }
2200 
2201 
2202     /* ret is still -EINVAL until here */
2203     ret = bdrv_parse_cache_mode(src_cache, &src_flags, &src_writethrough);
2204     if (ret < 0) {
2205         error_report("Invalid source cache option: %s", src_cache);
2206         goto fail_getopt;
2207     }
2208 
2209     /* Initialize before goto out */
2210     if (quiet) {
2211         progress = false;
2212     }
2213     qemu_progress_init(progress, 1.0);
2214     qemu_progress_print(0, 100);
2215 
2216     s.src = g_new0(BlockBackend *, s.src_num);
2217     s.src_sectors = g_new(int64_t, s.src_num);
2218 
2219     for (bs_i = 0; bs_i < s.src_num; bs_i++) {
2220         s.src[bs_i] = img_open(image_opts, argv[optind + bs_i],
2221                                fmt, src_flags, src_writethrough, quiet,
2222                                force_share);
2223         if (!s.src[bs_i]) {
2224             ret = -1;
2225             goto out;
2226         }
2227         s.src_sectors[bs_i] = blk_nb_sectors(s.src[bs_i]);
2228         if (s.src_sectors[bs_i] < 0) {
2229             error_report("Could not get size of %s: %s",
2230                          argv[optind + bs_i], strerror(-s.src_sectors[bs_i]));
2231             ret = -1;
2232             goto out;
2233         }
2234         s.total_sectors += s.src_sectors[bs_i];
2235     }
2236 
2237     if (sn_opts) {
2238         bdrv_snapshot_load_tmp(blk_bs(s.src[0]),
2239                                qemu_opt_get(sn_opts, SNAPSHOT_OPT_ID),
2240                                qemu_opt_get(sn_opts, SNAPSHOT_OPT_NAME),
2241                                &local_err);
2242     } else if (snapshot_name != NULL) {
2243         if (s.src_num > 1) {
2244             error_report("No support for concatenating multiple snapshot");
2245             ret = -1;
2246             goto out;
2247         }
2248 
2249         bdrv_snapshot_load_tmp_by_id_or_name(blk_bs(s.src[0]), snapshot_name,
2250                                              &local_err);
2251     }
2252     if (local_err) {
2253         error_reportf_err(local_err, "Failed to load snapshot: ");
2254         ret = -1;
2255         goto out;
2256     }
2257 
2258     if (!skip_create) {
2259         /* Find driver and parse its options */
2260         drv = bdrv_find_format(out_fmt);
2261         if (!drv) {
2262             error_report("Unknown file format '%s'", out_fmt);
2263             ret = -1;
2264             goto out;
2265         }
2266 
2267         proto_drv = bdrv_find_protocol(out_filename, true, &local_err);
2268         if (!proto_drv) {
2269             error_report_err(local_err);
2270             ret = -1;
2271             goto out;
2272         }
2273 
2274         if (!drv->create_opts) {
2275             error_report("Format driver '%s' does not support image creation",
2276                          drv->format_name);
2277             ret = -1;
2278             goto out;
2279         }
2280 
2281         if (!proto_drv->create_opts) {
2282             error_report("Protocol driver '%s' does not support image creation",
2283                          proto_drv->format_name);
2284             ret = -1;
2285             goto out;
2286         }
2287 
2288         create_opts = qemu_opts_append(create_opts, drv->create_opts);
2289         create_opts = qemu_opts_append(create_opts, proto_drv->create_opts);
2290 
2291         opts = qemu_opts_create(create_opts, NULL, 0, &error_abort);
2292         if (options) {
2293             qemu_opts_do_parse(opts, options, NULL, &local_err);
2294             if (local_err) {
2295                 error_report_err(local_err);
2296                 ret = -1;
2297                 goto out;
2298             }
2299         }
2300 
2301         qemu_opt_set_number(opts, BLOCK_OPT_SIZE, s.total_sectors * 512,
2302                             &error_abort);
2303         ret = add_old_style_options(out_fmt, opts, out_baseimg, NULL);
2304         if (ret < 0) {
2305             goto out;
2306         }
2307     }
2308 
2309     /* Get backing file name if -o backing_file was used */
2310     out_baseimg_param = qemu_opt_get(opts, BLOCK_OPT_BACKING_FILE);
2311     if (out_baseimg_param) {
2312         out_baseimg = out_baseimg_param;
2313     }
2314     s.target_has_backing = (bool) out_baseimg;
2315 
2316     if (s.src_num > 1 && out_baseimg) {
2317         error_report("Having a backing file for the target makes no sense when "
2318                      "concatenating multiple input images");
2319         ret = -1;
2320         goto out;
2321     }
2322 
2323     /* Check if compression is supported */
2324     if (s.compressed) {
2325         bool encryption =
2326             qemu_opt_get_bool(opts, BLOCK_OPT_ENCRYPT, false);
2327         const char *encryptfmt =
2328             qemu_opt_get(opts, BLOCK_OPT_ENCRYPT_FORMAT);
2329         const char *preallocation =
2330             qemu_opt_get(opts, BLOCK_OPT_PREALLOC);
2331 
2332         if (drv && !drv->bdrv_co_pwritev_compressed) {
2333             error_report("Compression not supported for this file format");
2334             ret = -1;
2335             goto out;
2336         }
2337 
2338         if (encryption || encryptfmt) {
2339             error_report("Compression and encryption not supported at "
2340                          "the same time");
2341             ret = -1;
2342             goto out;
2343         }
2344 
2345         if (preallocation
2346             && strcmp(preallocation, "off"))
2347         {
2348             error_report("Compression and preallocation not supported at "
2349                          "the same time");
2350             ret = -1;
2351             goto out;
2352         }
2353     }
2354 
2355     /*
2356      * The later open call will need any decryption secrets, and
2357      * bdrv_create() will purge "opts", so extract them now before
2358      * they are lost.
2359      */
2360     if (!skip_create) {
2361         open_opts = qdict_new();
2362         qemu_opt_foreach(opts, img_add_key_secrets, open_opts, &error_abort);
2363     }
2364 
2365     if (!skip_create) {
2366         /* Create the new image */
2367         ret = bdrv_create(drv, out_filename, opts, &local_err);
2368         if (ret < 0) {
2369             error_reportf_err(local_err, "%s: error while converting %s: ",
2370                               out_filename, out_fmt);
2371             goto out;
2372         }
2373     }
2374 
2375     flags = s.min_sparse ? (BDRV_O_RDWR | BDRV_O_UNMAP) : BDRV_O_RDWR;
2376     ret = bdrv_parse_cache_mode(cache, &flags, &writethrough);
2377     if (ret < 0) {
2378         error_report("Invalid cache option: %s", cache);
2379         goto out;
2380     }
2381 
2382     if (skip_create) {
2383         s.target = img_open(tgt_image_opts, out_filename, out_fmt,
2384                             flags, writethrough, quiet, false);
2385     } else {
2386         /* TODO ultimately we should allow --target-image-opts
2387          * to be used even when -n is not given.
2388          * That has to wait for bdrv_create to be improved
2389          * to allow filenames in option syntax
2390          */
2391         s.target = img_open_file(out_filename, open_opts, out_fmt,
2392                                  flags, writethrough, quiet, false);
2393         open_opts = NULL; /* blk_new_open will have freed it */
2394     }
2395     if (!s.target) {
2396         ret = -1;
2397         goto out;
2398     }
2399     out_bs = blk_bs(s.target);
2400 
2401     if (s.compressed && !out_bs->drv->bdrv_co_pwritev_compressed) {
2402         error_report("Compression not supported for this file format");
2403         ret = -1;
2404         goto out;
2405     }
2406 
2407     /* increase bufsectors from the default 4096 (2M) if opt_transfer
2408      * or discard_alignment of the out_bs is greater. Limit to
2409      * MAX_BUF_SECTORS as maximum which is currently 32768 (16MB). */
2410     s.buf_sectors = MIN(MAX_BUF_SECTORS,
2411                         MAX(s.buf_sectors,
2412                             MAX(out_bs->bl.opt_transfer >> BDRV_SECTOR_BITS,
2413                                 out_bs->bl.pdiscard_alignment >>
2414                                 BDRV_SECTOR_BITS)));
2415 
2416     /* try to align the write requests to the destination to avoid unnecessary
2417      * RMW cycles. */
2418     s.alignment = MAX(pow2floor(s.min_sparse),
2419                       DIV_ROUND_UP(out_bs->bl.request_alignment,
2420                                    BDRV_SECTOR_SIZE));
2421     assert(is_power_of_2(s.alignment));
2422 
2423     if (skip_create) {
2424         int64_t output_sectors = blk_nb_sectors(s.target);
2425         if (output_sectors < 0) {
2426             error_report("unable to get output image length: %s",
2427                          strerror(-output_sectors));
2428             ret = -1;
2429             goto out;
2430         } else if (output_sectors < s.total_sectors) {
2431             error_report("output file is smaller than input file");
2432             ret = -1;
2433             goto out;
2434         }
2435     }
2436 
2437     if (s.target_has_backing) {
2438         /* Errors are treated as "backing length unknown" (which means
2439          * s.target_backing_sectors has to be negative, which it will
2440          * be automatically).  The backing file length is used only
2441          * for optimizations, so such a case is not fatal. */
2442         s.target_backing_sectors = bdrv_nb_sectors(out_bs->backing->bs);
2443     } else {
2444         s.target_backing_sectors = -1;
2445     }
2446 
2447     ret = bdrv_get_info(out_bs, &bdi);
2448     if (ret < 0) {
2449         if (s.compressed) {
2450             error_report("could not get block driver info");
2451             goto out;
2452         }
2453     } else {
2454         s.compressed = s.compressed || bdi.needs_compressed_writes;
2455         s.cluster_sectors = bdi.cluster_size / BDRV_SECTOR_SIZE;
2456         s.unallocated_blocks_are_zero = bdi.unallocated_blocks_are_zero;
2457     }
2458 
2459     ret = convert_do_copy(&s);
2460 out:
2461     if (!ret) {
2462         qemu_progress_print(100, 0);
2463     }
2464     qemu_progress_end();
2465     qemu_opts_del(opts);
2466     qemu_opts_free(create_opts);
2467     qemu_opts_del(sn_opts);
2468     qobject_unref(open_opts);
2469     blk_unref(s.target);
2470     if (s.src) {
2471         for (bs_i = 0; bs_i < s.src_num; bs_i++) {
2472             blk_unref(s.src[bs_i]);
2473         }
2474         g_free(s.src);
2475     }
2476     g_free(s.src_sectors);
2477 fail_getopt:
2478     g_free(options);
2479 
2480     return !!ret;
2481 }
2482 
2483 
2484 static void dump_snapshots(BlockDriverState *bs)
2485 {
2486     QEMUSnapshotInfo *sn_tab, *sn;
2487     int nb_sns, i;
2488 
2489     nb_sns = bdrv_snapshot_list(bs, &sn_tab);
2490     if (nb_sns <= 0)
2491         return;
2492     printf("Snapshot list:\n");
2493     bdrv_snapshot_dump(NULL);
2494     printf("\n");
2495     for(i = 0; i < nb_sns; i++) {
2496         sn = &sn_tab[i];
2497         bdrv_snapshot_dump(sn);
2498         printf("\n");
2499     }
2500     g_free(sn_tab);
2501 }
2502 
2503 static void dump_json_image_info_list(ImageInfoList *list)
2504 {
2505     QString *str;
2506     QObject *obj;
2507     Visitor *v = qobject_output_visitor_new(&obj);
2508 
2509     visit_type_ImageInfoList(v, NULL, &list, &error_abort);
2510     visit_complete(v, &obj);
2511     str = qobject_to_json_pretty(obj);
2512     assert(str != NULL);
2513     printf("%s\n", qstring_get_str(str));
2514     qobject_unref(obj);
2515     visit_free(v);
2516     qobject_unref(str);
2517 }
2518 
2519 static void dump_json_image_info(ImageInfo *info)
2520 {
2521     QString *str;
2522     QObject *obj;
2523     Visitor *v = qobject_output_visitor_new(&obj);
2524 
2525     visit_type_ImageInfo(v, NULL, &info, &error_abort);
2526     visit_complete(v, &obj);
2527     str = qobject_to_json_pretty(obj);
2528     assert(str != NULL);
2529     printf("%s\n", qstring_get_str(str));
2530     qobject_unref(obj);
2531     visit_free(v);
2532     qobject_unref(str);
2533 }
2534 
2535 static void dump_human_image_info_list(ImageInfoList *list)
2536 {
2537     ImageInfoList *elem;
2538     bool delim = false;
2539 
2540     for (elem = list; elem; elem = elem->next) {
2541         if (delim) {
2542             printf("\n");
2543         }
2544         delim = true;
2545 
2546         bdrv_image_info_dump(elem->value);
2547     }
2548 }
2549 
2550 static gboolean str_equal_func(gconstpointer a, gconstpointer b)
2551 {
2552     return strcmp(a, b) == 0;
2553 }
2554 
2555 /**
2556  * Open an image file chain and return an ImageInfoList
2557  *
2558  * @filename: topmost image filename
2559  * @fmt: topmost image format (may be NULL to autodetect)
2560  * @chain: true  - enumerate entire backing file chain
2561  *         false - only topmost image file
2562  *
2563  * Returns a list of ImageInfo objects or NULL if there was an error opening an
2564  * image file.  If there was an error a message will have been printed to
2565  * stderr.
2566  */
2567 static ImageInfoList *collect_image_info_list(bool image_opts,
2568                                               const char *filename,
2569                                               const char *fmt,
2570                                               bool chain, bool force_share)
2571 {
2572     ImageInfoList *head = NULL;
2573     ImageInfoList **last = &head;
2574     GHashTable *filenames;
2575     Error *err = NULL;
2576 
2577     filenames = g_hash_table_new_full(g_str_hash, str_equal_func, NULL, NULL);
2578 
2579     while (filename) {
2580         BlockBackend *blk;
2581         BlockDriverState *bs;
2582         ImageInfo *info;
2583         ImageInfoList *elem;
2584 
2585         if (g_hash_table_lookup_extended(filenames, filename, NULL, NULL)) {
2586             error_report("Backing file '%s' creates an infinite loop.",
2587                          filename);
2588             goto err;
2589         }
2590         g_hash_table_insert(filenames, (gpointer)filename, NULL);
2591 
2592         blk = img_open(image_opts, filename, fmt,
2593                        BDRV_O_NO_BACKING | BDRV_O_NO_IO, false, false,
2594                        force_share);
2595         if (!blk) {
2596             goto err;
2597         }
2598         bs = blk_bs(blk);
2599 
2600         bdrv_query_image_info(bs, &info, &err);
2601         if (err) {
2602             error_report_err(err);
2603             blk_unref(blk);
2604             goto err;
2605         }
2606 
2607         elem = g_new0(ImageInfoList, 1);
2608         elem->value = info;
2609         *last = elem;
2610         last = &elem->next;
2611 
2612         blk_unref(blk);
2613 
2614         filename = fmt = NULL;
2615         if (chain) {
2616             if (info->has_full_backing_filename) {
2617                 filename = info->full_backing_filename;
2618             } else if (info->has_backing_filename) {
2619                 error_report("Could not determine absolute backing filename,"
2620                              " but backing filename '%s' present",
2621                              info->backing_filename);
2622                 goto err;
2623             }
2624             if (info->has_backing_filename_format) {
2625                 fmt = info->backing_filename_format;
2626             }
2627         }
2628     }
2629     g_hash_table_destroy(filenames);
2630     return head;
2631 
2632 err:
2633     qapi_free_ImageInfoList(head);
2634     g_hash_table_destroy(filenames);
2635     return NULL;
2636 }
2637 
2638 static int img_info(int argc, char **argv)
2639 {
2640     int c;
2641     OutputFormat output_format = OFORMAT_HUMAN;
2642     bool chain = false;
2643     const char *filename, *fmt, *output;
2644     ImageInfoList *list;
2645     bool image_opts = false;
2646     bool force_share = false;
2647 
2648     fmt = NULL;
2649     output = NULL;
2650     for(;;) {
2651         int option_index = 0;
2652         static const struct option long_options[] = {
2653             {"help", no_argument, 0, 'h'},
2654             {"format", required_argument, 0, 'f'},
2655             {"output", required_argument, 0, OPTION_OUTPUT},
2656             {"backing-chain", no_argument, 0, OPTION_BACKING_CHAIN},
2657             {"object", required_argument, 0, OPTION_OBJECT},
2658             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
2659             {"force-share", no_argument, 0, 'U'},
2660             {0, 0, 0, 0}
2661         };
2662         c = getopt_long(argc, argv, ":f:hU",
2663                         long_options, &option_index);
2664         if (c == -1) {
2665             break;
2666         }
2667         switch(c) {
2668         case ':':
2669             missing_argument(argv[optind - 1]);
2670             break;
2671         case '?':
2672             unrecognized_option(argv[optind - 1]);
2673             break;
2674         case 'h':
2675             help();
2676             break;
2677         case 'f':
2678             fmt = optarg;
2679             break;
2680         case 'U':
2681             force_share = true;
2682             break;
2683         case OPTION_OUTPUT:
2684             output = optarg;
2685             break;
2686         case OPTION_BACKING_CHAIN:
2687             chain = true;
2688             break;
2689         case OPTION_OBJECT: {
2690             QemuOpts *opts;
2691             opts = qemu_opts_parse_noisily(&qemu_object_opts,
2692                                            optarg, true);
2693             if (!opts) {
2694                 return 1;
2695             }
2696         }   break;
2697         case OPTION_IMAGE_OPTS:
2698             image_opts = true;
2699             break;
2700         }
2701     }
2702     if (optind != argc - 1) {
2703         error_exit("Expecting one image file name");
2704     }
2705     filename = argv[optind++];
2706 
2707     if (output && !strcmp(output, "json")) {
2708         output_format = OFORMAT_JSON;
2709     } else if (output && !strcmp(output, "human")) {
2710         output_format = OFORMAT_HUMAN;
2711     } else if (output) {
2712         error_report("--output must be used with human or json as argument.");
2713         return 1;
2714     }
2715 
2716     if (qemu_opts_foreach(&qemu_object_opts,
2717                           user_creatable_add_opts_foreach,
2718                           NULL, &error_fatal)) {
2719         return 1;
2720     }
2721 
2722     list = collect_image_info_list(image_opts, filename, fmt, chain,
2723                                    force_share);
2724     if (!list) {
2725         return 1;
2726     }
2727 
2728     switch (output_format) {
2729     case OFORMAT_HUMAN:
2730         dump_human_image_info_list(list);
2731         break;
2732     case OFORMAT_JSON:
2733         if (chain) {
2734             dump_json_image_info_list(list);
2735         } else {
2736             dump_json_image_info(list->value);
2737         }
2738         break;
2739     }
2740 
2741     qapi_free_ImageInfoList(list);
2742     return 0;
2743 }
2744 
2745 static int dump_map_entry(OutputFormat output_format, MapEntry *e,
2746                           MapEntry *next)
2747 {
2748     switch (output_format) {
2749     case OFORMAT_HUMAN:
2750         if (e->data && !e->has_offset) {
2751             error_report("File contains external, encrypted or compressed clusters.");
2752             return -1;
2753         }
2754         if (e->data && !e->zero) {
2755             printf("%#-16"PRIx64"%#-16"PRIx64"%#-16"PRIx64"%s\n",
2756                    e->start, e->length,
2757                    e->has_offset ? e->offset : 0,
2758                    e->has_filename ? e->filename : "");
2759         }
2760         /* This format ignores the distinction between 0, ZERO and ZERO|DATA.
2761          * Modify the flags here to allow more coalescing.
2762          */
2763         if (next && (!next->data || next->zero)) {
2764             next->data = false;
2765             next->zero = true;
2766         }
2767         break;
2768     case OFORMAT_JSON:
2769         printf("%s{ \"start\": %"PRId64", \"length\": %"PRId64","
2770                " \"depth\": %"PRId64", \"zero\": %s, \"data\": %s",
2771                (e->start == 0 ? "[" : ",\n"),
2772                e->start, e->length, e->depth,
2773                e->zero ? "true" : "false",
2774                e->data ? "true" : "false");
2775         if (e->has_offset) {
2776             printf(", \"offset\": %"PRId64"", e->offset);
2777         }
2778         putchar('}');
2779 
2780         if (!next) {
2781             printf("]\n");
2782         }
2783         break;
2784     }
2785     return 0;
2786 }
2787 
2788 static int get_block_status(BlockDriverState *bs, int64_t offset,
2789                             int64_t bytes, MapEntry *e)
2790 {
2791     int ret;
2792     int depth;
2793     BlockDriverState *file;
2794     bool has_offset;
2795     int64_t map;
2796     char *filename = NULL;
2797 
2798     /* As an optimization, we could cache the current range of unallocated
2799      * clusters in each file of the chain, and avoid querying the same
2800      * range repeatedly.
2801      */
2802 
2803     depth = 0;
2804     for (;;) {
2805         ret = bdrv_block_status(bs, offset, bytes, &bytes, &map, &file);
2806         if (ret < 0) {
2807             return ret;
2808         }
2809         assert(bytes);
2810         if (ret & (BDRV_BLOCK_ZERO|BDRV_BLOCK_DATA)) {
2811             break;
2812         }
2813         bs = backing_bs(bs);
2814         if (bs == NULL) {
2815             ret = 0;
2816             break;
2817         }
2818 
2819         depth++;
2820     }
2821 
2822     has_offset = !!(ret & BDRV_BLOCK_OFFSET_VALID);
2823 
2824     if (file && has_offset) {
2825         bdrv_refresh_filename(file);
2826         filename = file->filename;
2827     }
2828 
2829     *e = (MapEntry) {
2830         .start = offset,
2831         .length = bytes,
2832         .data = !!(ret & BDRV_BLOCK_DATA),
2833         .zero = !!(ret & BDRV_BLOCK_ZERO),
2834         .offset = map,
2835         .has_offset = has_offset,
2836         .depth = depth,
2837         .has_filename = filename,
2838         .filename = filename,
2839     };
2840 
2841     return 0;
2842 }
2843 
2844 static inline bool entry_mergeable(const MapEntry *curr, const MapEntry *next)
2845 {
2846     if (curr->length == 0) {
2847         return false;
2848     }
2849     if (curr->zero != next->zero ||
2850         curr->data != next->data ||
2851         curr->depth != next->depth ||
2852         curr->has_filename != next->has_filename ||
2853         curr->has_offset != next->has_offset) {
2854         return false;
2855     }
2856     if (curr->has_filename && strcmp(curr->filename, next->filename)) {
2857         return false;
2858     }
2859     if (curr->has_offset && curr->offset + curr->length != next->offset) {
2860         return false;
2861     }
2862     return true;
2863 }
2864 
2865 static int img_map(int argc, char **argv)
2866 {
2867     int c;
2868     OutputFormat output_format = OFORMAT_HUMAN;
2869     BlockBackend *blk;
2870     BlockDriverState *bs;
2871     const char *filename, *fmt, *output;
2872     int64_t length;
2873     MapEntry curr = { .length = 0 }, next;
2874     int ret = 0;
2875     bool image_opts = false;
2876     bool force_share = false;
2877 
2878     fmt = NULL;
2879     output = NULL;
2880     for (;;) {
2881         int option_index = 0;
2882         static const struct option long_options[] = {
2883             {"help", no_argument, 0, 'h'},
2884             {"format", required_argument, 0, 'f'},
2885             {"output", required_argument, 0, OPTION_OUTPUT},
2886             {"object", required_argument, 0, OPTION_OBJECT},
2887             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
2888             {"force-share", no_argument, 0, 'U'},
2889             {0, 0, 0, 0}
2890         };
2891         c = getopt_long(argc, argv, ":f:hU",
2892                         long_options, &option_index);
2893         if (c == -1) {
2894             break;
2895         }
2896         switch (c) {
2897         case ':':
2898             missing_argument(argv[optind - 1]);
2899             break;
2900         case '?':
2901             unrecognized_option(argv[optind - 1]);
2902             break;
2903         case 'h':
2904             help();
2905             break;
2906         case 'f':
2907             fmt = optarg;
2908             break;
2909         case 'U':
2910             force_share = true;
2911             break;
2912         case OPTION_OUTPUT:
2913             output = optarg;
2914             break;
2915         case OPTION_OBJECT: {
2916             QemuOpts *opts;
2917             opts = qemu_opts_parse_noisily(&qemu_object_opts,
2918                                            optarg, true);
2919             if (!opts) {
2920                 return 1;
2921             }
2922         }   break;
2923         case OPTION_IMAGE_OPTS:
2924             image_opts = true;
2925             break;
2926         }
2927     }
2928     if (optind != argc - 1) {
2929         error_exit("Expecting one image file name");
2930     }
2931     filename = argv[optind];
2932 
2933     if (output && !strcmp(output, "json")) {
2934         output_format = OFORMAT_JSON;
2935     } else if (output && !strcmp(output, "human")) {
2936         output_format = OFORMAT_HUMAN;
2937     } else if (output) {
2938         error_report("--output must be used with human or json as argument.");
2939         return 1;
2940     }
2941 
2942     if (qemu_opts_foreach(&qemu_object_opts,
2943                           user_creatable_add_opts_foreach,
2944                           NULL, &error_fatal)) {
2945         return 1;
2946     }
2947 
2948     blk = img_open(image_opts, filename, fmt, 0, false, false, force_share);
2949     if (!blk) {
2950         return 1;
2951     }
2952     bs = blk_bs(blk);
2953 
2954     if (output_format == OFORMAT_HUMAN) {
2955         printf("%-16s%-16s%-16s%s\n", "Offset", "Length", "Mapped to", "File");
2956     }
2957 
2958     length = blk_getlength(blk);
2959     while (curr.start + curr.length < length) {
2960         int64_t offset = curr.start + curr.length;
2961         int64_t n;
2962 
2963         /* Probe up to 1 GiB at a time.  */
2964         n = MIN(1 * GiB, length - offset);
2965         ret = get_block_status(bs, offset, n, &next);
2966 
2967         if (ret < 0) {
2968             error_report("Could not read file metadata: %s", strerror(-ret));
2969             goto out;
2970         }
2971 
2972         if (entry_mergeable(&curr, &next)) {
2973             curr.length += next.length;
2974             continue;
2975         }
2976 
2977         if (curr.length > 0) {
2978             ret = dump_map_entry(output_format, &curr, &next);
2979             if (ret < 0) {
2980                 goto out;
2981             }
2982         }
2983         curr = next;
2984     }
2985 
2986     ret = dump_map_entry(output_format, &curr, NULL);
2987 
2988 out:
2989     blk_unref(blk);
2990     return ret < 0;
2991 }
2992 
2993 #define SNAPSHOT_LIST   1
2994 #define SNAPSHOT_CREATE 2
2995 #define SNAPSHOT_APPLY  3
2996 #define SNAPSHOT_DELETE 4
2997 
2998 static int img_snapshot(int argc, char **argv)
2999 {
3000     BlockBackend *blk;
3001     BlockDriverState *bs;
3002     QEMUSnapshotInfo sn;
3003     char *filename, *snapshot_name = NULL;
3004     int c, ret = 0, bdrv_oflags;
3005     int action = 0;
3006     qemu_timeval tv;
3007     bool quiet = false;
3008     Error *err = NULL;
3009     bool image_opts = false;
3010     bool force_share = false;
3011 
3012     bdrv_oflags = BDRV_O_RDWR;
3013     /* Parse commandline parameters */
3014     for(;;) {
3015         static const struct option long_options[] = {
3016             {"help", no_argument, 0, 'h'},
3017             {"object", required_argument, 0, OPTION_OBJECT},
3018             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
3019             {"force-share", no_argument, 0, 'U'},
3020             {0, 0, 0, 0}
3021         };
3022         c = getopt_long(argc, argv, ":la:c:d:hqU",
3023                         long_options, NULL);
3024         if (c == -1) {
3025             break;
3026         }
3027         switch(c) {
3028         case ':':
3029             missing_argument(argv[optind - 1]);
3030             break;
3031         case '?':
3032             unrecognized_option(argv[optind - 1]);
3033             break;
3034         case 'h':
3035             help();
3036             return 0;
3037         case 'l':
3038             if (action) {
3039                 error_exit("Cannot mix '-l', '-a', '-c', '-d'");
3040                 return 0;
3041             }
3042             action = SNAPSHOT_LIST;
3043             bdrv_oflags &= ~BDRV_O_RDWR; /* no need for RW */
3044             break;
3045         case 'a':
3046             if (action) {
3047                 error_exit("Cannot mix '-l', '-a', '-c', '-d'");
3048                 return 0;
3049             }
3050             action = SNAPSHOT_APPLY;
3051             snapshot_name = optarg;
3052             break;
3053         case 'c':
3054             if (action) {
3055                 error_exit("Cannot mix '-l', '-a', '-c', '-d'");
3056                 return 0;
3057             }
3058             action = SNAPSHOT_CREATE;
3059             snapshot_name = optarg;
3060             break;
3061         case 'd':
3062             if (action) {
3063                 error_exit("Cannot mix '-l', '-a', '-c', '-d'");
3064                 return 0;
3065             }
3066             action = SNAPSHOT_DELETE;
3067             snapshot_name = optarg;
3068             break;
3069         case 'q':
3070             quiet = true;
3071             break;
3072         case 'U':
3073             force_share = true;
3074             break;
3075         case OPTION_OBJECT: {
3076             QemuOpts *opts;
3077             opts = qemu_opts_parse_noisily(&qemu_object_opts,
3078                                            optarg, true);
3079             if (!opts) {
3080                 return 1;
3081             }
3082         }   break;
3083         case OPTION_IMAGE_OPTS:
3084             image_opts = true;
3085             break;
3086         }
3087     }
3088 
3089     if (optind != argc - 1) {
3090         error_exit("Expecting one image file name");
3091     }
3092     filename = argv[optind++];
3093 
3094     if (qemu_opts_foreach(&qemu_object_opts,
3095                           user_creatable_add_opts_foreach,
3096                           NULL, &error_fatal)) {
3097         return 1;
3098     }
3099 
3100     /* Open the image */
3101     blk = img_open(image_opts, filename, NULL, bdrv_oflags, false, quiet,
3102                    force_share);
3103     if (!blk) {
3104         return 1;
3105     }
3106     bs = blk_bs(blk);
3107 
3108     /* Perform the requested action */
3109     switch(action) {
3110     case SNAPSHOT_LIST:
3111         dump_snapshots(bs);
3112         break;
3113 
3114     case SNAPSHOT_CREATE:
3115         memset(&sn, 0, sizeof(sn));
3116         pstrcpy(sn.name, sizeof(sn.name), snapshot_name);
3117 
3118         qemu_gettimeofday(&tv);
3119         sn.date_sec = tv.tv_sec;
3120         sn.date_nsec = tv.tv_usec * 1000;
3121 
3122         ret = bdrv_snapshot_create(bs, &sn);
3123         if (ret) {
3124             error_report("Could not create snapshot '%s': %d (%s)",
3125                 snapshot_name, ret, strerror(-ret));
3126         }
3127         break;
3128 
3129     case SNAPSHOT_APPLY:
3130         ret = bdrv_snapshot_goto(bs, snapshot_name, &err);
3131         if (ret) {
3132             error_reportf_err(err, "Could not apply snapshot '%s': ",
3133                               snapshot_name);
3134         }
3135         break;
3136 
3137     case SNAPSHOT_DELETE:
3138         ret = bdrv_snapshot_find(bs, &sn, snapshot_name);
3139         if (ret < 0) {
3140             error_report("Could not delete snapshot '%s': snapshot not "
3141                          "found", snapshot_name);
3142             ret = 1;
3143         } else {
3144             ret = bdrv_snapshot_delete(bs, sn.id_str, sn.name, &err);
3145             if (ret < 0) {
3146                 error_reportf_err(err, "Could not delete snapshot '%s': ",
3147                                   snapshot_name);
3148                 ret = 1;
3149             }
3150         }
3151         break;
3152     }
3153 
3154     /* Cleanup */
3155     blk_unref(blk);
3156     if (ret) {
3157         return 1;
3158     }
3159     return 0;
3160 }
3161 
3162 static int img_rebase(int argc, char **argv)
3163 {
3164     BlockBackend *blk = NULL, *blk_old_backing = NULL, *blk_new_backing = NULL;
3165     uint8_t *buf_old = NULL;
3166     uint8_t *buf_new = NULL;
3167     BlockDriverState *bs = NULL;
3168     char *filename;
3169     const char *fmt, *cache, *src_cache, *out_basefmt, *out_baseimg;
3170     int c, flags, src_flags, ret;
3171     bool writethrough, src_writethrough;
3172     int unsafe = 0;
3173     bool force_share = false;
3174     int progress = 0;
3175     bool quiet = false;
3176     Error *local_err = NULL;
3177     bool image_opts = false;
3178 
3179     /* Parse commandline parameters */
3180     fmt = NULL;
3181     cache = BDRV_DEFAULT_CACHE;
3182     src_cache = BDRV_DEFAULT_CACHE;
3183     out_baseimg = NULL;
3184     out_basefmt = NULL;
3185     for(;;) {
3186         static const struct option long_options[] = {
3187             {"help", no_argument, 0, 'h'},
3188             {"object", required_argument, 0, OPTION_OBJECT},
3189             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
3190             {"force-share", no_argument, 0, 'U'},
3191             {0, 0, 0, 0}
3192         };
3193         c = getopt_long(argc, argv, ":hf:F:b:upt:T:qU",
3194                         long_options, NULL);
3195         if (c == -1) {
3196             break;
3197         }
3198         switch(c) {
3199         case ':':
3200             missing_argument(argv[optind - 1]);
3201             break;
3202         case '?':
3203             unrecognized_option(argv[optind - 1]);
3204             break;
3205         case 'h':
3206             help();
3207             return 0;
3208         case 'f':
3209             fmt = optarg;
3210             break;
3211         case 'F':
3212             out_basefmt = optarg;
3213             break;
3214         case 'b':
3215             out_baseimg = optarg;
3216             break;
3217         case 'u':
3218             unsafe = 1;
3219             break;
3220         case 'p':
3221             progress = 1;
3222             break;
3223         case 't':
3224             cache = optarg;
3225             break;
3226         case 'T':
3227             src_cache = optarg;
3228             break;
3229         case 'q':
3230             quiet = true;
3231             break;
3232         case OPTION_OBJECT: {
3233             QemuOpts *opts;
3234             opts = qemu_opts_parse_noisily(&qemu_object_opts,
3235                                            optarg, true);
3236             if (!opts) {
3237                 return 1;
3238             }
3239         }   break;
3240         case OPTION_IMAGE_OPTS:
3241             image_opts = true;
3242             break;
3243         case 'U':
3244             force_share = true;
3245             break;
3246         }
3247     }
3248 
3249     if (quiet) {
3250         progress = 0;
3251     }
3252 
3253     if (optind != argc - 1) {
3254         error_exit("Expecting one image file name");
3255     }
3256     if (!unsafe && !out_baseimg) {
3257         error_exit("Must specify backing file (-b) or use unsafe mode (-u)");
3258     }
3259     filename = argv[optind++];
3260 
3261     if (qemu_opts_foreach(&qemu_object_opts,
3262                           user_creatable_add_opts_foreach,
3263                           NULL, &error_fatal)) {
3264         return 1;
3265     }
3266 
3267     qemu_progress_init(progress, 2.0);
3268     qemu_progress_print(0, 100);
3269 
3270     flags = BDRV_O_RDWR | (unsafe ? BDRV_O_NO_BACKING : 0);
3271     ret = bdrv_parse_cache_mode(cache, &flags, &writethrough);
3272     if (ret < 0) {
3273         error_report("Invalid cache option: %s", cache);
3274         goto out;
3275     }
3276 
3277     src_flags = 0;
3278     ret = bdrv_parse_cache_mode(src_cache, &src_flags, &src_writethrough);
3279     if (ret < 0) {
3280         error_report("Invalid source cache option: %s", src_cache);
3281         goto out;
3282     }
3283 
3284     /* The source files are opened read-only, don't care about WCE */
3285     assert((src_flags & BDRV_O_RDWR) == 0);
3286     (void) src_writethrough;
3287 
3288     /*
3289      * Open the images.
3290      *
3291      * Ignore the old backing file for unsafe rebase in case we want to correct
3292      * the reference to a renamed or moved backing file.
3293      */
3294     blk = img_open(image_opts, filename, fmt, flags, writethrough, quiet,
3295                    false);
3296     if (!blk) {
3297         ret = -1;
3298         goto out;
3299     }
3300     bs = blk_bs(blk);
3301 
3302     if (out_basefmt != NULL) {
3303         if (bdrv_find_format(out_basefmt) == NULL) {
3304             error_report("Invalid format name: '%s'", out_basefmt);
3305             ret = -1;
3306             goto out;
3307         }
3308     }
3309 
3310     /* For safe rebasing we need to compare old and new backing file */
3311     if (!unsafe) {
3312         char backing_name[PATH_MAX];
3313         QDict *options = NULL;
3314 
3315         if (bs->backing) {
3316             if (bs->backing_format[0] != '\0') {
3317                 options = qdict_new();
3318                 qdict_put_str(options, "driver", bs->backing_format);
3319             }
3320 
3321             if (force_share) {
3322                 if (!options) {
3323                     options = qdict_new();
3324                 }
3325                 qdict_put_bool(options, BDRV_OPT_FORCE_SHARE, true);
3326             }
3327             bdrv_get_backing_filename(bs, backing_name, sizeof(backing_name));
3328             blk_old_backing = blk_new_open(backing_name, NULL,
3329                                            options, src_flags, &local_err);
3330             if (!blk_old_backing) {
3331                 error_reportf_err(local_err,
3332                                   "Could not open old backing file '%s': ",
3333                                   backing_name);
3334                 ret = -1;
3335                 goto out;
3336             }
3337         } else {
3338             blk_old_backing = NULL;
3339         }
3340 
3341         if (out_baseimg[0]) {
3342             const char *overlay_filename;
3343             char *out_real_path;
3344 
3345             options = qdict_new();
3346             if (out_basefmt) {
3347                 qdict_put_str(options, "driver", out_basefmt);
3348             }
3349             if (force_share) {
3350                 qdict_put_bool(options, BDRV_OPT_FORCE_SHARE, true);
3351             }
3352 
3353             bdrv_refresh_filename(bs);
3354             overlay_filename = bs->exact_filename[0] ? bs->exact_filename
3355                                                      : bs->filename;
3356             out_real_path =
3357                 bdrv_get_full_backing_filename_from_filename(overlay_filename,
3358                                                              out_baseimg,
3359                                                              &local_err);
3360             if (local_err) {
3361                 error_reportf_err(local_err,
3362                                   "Could not resolve backing filename: ");
3363                 ret = -1;
3364                 goto out;
3365             }
3366 
3367             blk_new_backing = blk_new_open(out_real_path, NULL,
3368                                            options, src_flags, &local_err);
3369             g_free(out_real_path);
3370             if (!blk_new_backing) {
3371                 error_reportf_err(local_err,
3372                                   "Could not open new backing file '%s': ",
3373                                   out_baseimg);
3374                 ret = -1;
3375                 goto out;
3376             }
3377         }
3378     }
3379 
3380     /*
3381      * Check each unallocated cluster in the COW file. If it is unallocated,
3382      * accesses go to the backing file. We must therefore compare this cluster
3383      * in the old and new backing file, and if they differ we need to copy it
3384      * from the old backing file into the COW file.
3385      *
3386      * If qemu-img crashes during this step, no harm is done. The content of
3387      * the image is the same as the original one at any time.
3388      */
3389     if (!unsafe) {
3390         int64_t size;
3391         int64_t old_backing_size = 0;
3392         int64_t new_backing_size = 0;
3393         uint64_t offset;
3394         int64_t n;
3395         float local_progress = 0;
3396 
3397         buf_old = blk_blockalign(blk, IO_BUF_SIZE);
3398         buf_new = blk_blockalign(blk, IO_BUF_SIZE);
3399 
3400         size = blk_getlength(blk);
3401         if (size < 0) {
3402             error_report("Could not get size of '%s': %s",
3403                          filename, strerror(-size));
3404             ret = -1;
3405             goto out;
3406         }
3407         if (blk_old_backing) {
3408             old_backing_size = blk_getlength(blk_old_backing);
3409             if (old_backing_size < 0) {
3410                 char backing_name[PATH_MAX];
3411 
3412                 bdrv_get_backing_filename(bs, backing_name,
3413                                           sizeof(backing_name));
3414                 error_report("Could not get size of '%s': %s",
3415                              backing_name, strerror(-old_backing_size));
3416                 ret = -1;
3417                 goto out;
3418             }
3419         }
3420         if (blk_new_backing) {
3421             new_backing_size = blk_getlength(blk_new_backing);
3422             if (new_backing_size < 0) {
3423                 error_report("Could not get size of '%s': %s",
3424                              out_baseimg, strerror(-new_backing_size));
3425                 ret = -1;
3426                 goto out;
3427             }
3428         }
3429 
3430         if (size != 0) {
3431             local_progress = (float)100 / (size / MIN(size, IO_BUF_SIZE));
3432         }
3433 
3434         for (offset = 0; offset < size; offset += n) {
3435             bool buf_old_is_zero = false;
3436 
3437             /* How many bytes can we handle with the next read? */
3438             n = MIN(IO_BUF_SIZE, size - offset);
3439 
3440             /* If the cluster is allocated, we don't need to take action */
3441             ret = bdrv_is_allocated(bs, offset, n, &n);
3442             if (ret < 0) {
3443                 error_report("error while reading image metadata: %s",
3444                              strerror(-ret));
3445                 goto out;
3446             }
3447             if (ret) {
3448                 continue;
3449             }
3450 
3451             /*
3452              * Read old and new backing file and take into consideration that
3453              * backing files may be smaller than the COW image.
3454              */
3455             if (offset >= old_backing_size) {
3456                 memset(buf_old, 0, n);
3457                 buf_old_is_zero = true;
3458             } else {
3459                 if (offset + n > old_backing_size) {
3460                     n = old_backing_size - offset;
3461                 }
3462 
3463                 ret = blk_pread(blk_old_backing, offset, buf_old, n);
3464                 if (ret < 0) {
3465                     error_report("error while reading from old backing file");
3466                     goto out;
3467                 }
3468             }
3469 
3470             if (offset >= new_backing_size || !blk_new_backing) {
3471                 memset(buf_new, 0, n);
3472             } else {
3473                 if (offset + n > new_backing_size) {
3474                     n = new_backing_size - offset;
3475                 }
3476 
3477                 ret = blk_pread(blk_new_backing, offset, buf_new, n);
3478                 if (ret < 0) {
3479                     error_report("error while reading from new backing file");
3480                     goto out;
3481                 }
3482             }
3483 
3484             /* If they differ, we need to write to the COW file */
3485             uint64_t written = 0;
3486 
3487             while (written < n) {
3488                 int64_t pnum;
3489 
3490                 if (compare_buffers(buf_old + written, buf_new + written,
3491                                     n - written, &pnum))
3492                 {
3493                     if (buf_old_is_zero) {
3494                         ret = blk_pwrite_zeroes(blk, offset + written, pnum, 0);
3495                     } else {
3496                         ret = blk_pwrite(blk, offset + written,
3497                                          buf_old + written, pnum, 0);
3498                     }
3499                     if (ret < 0) {
3500                         error_report("Error while writing to COW image: %s",
3501                             strerror(-ret));
3502                         goto out;
3503                     }
3504                 }
3505 
3506                 written += pnum;
3507             }
3508             qemu_progress_print(local_progress, 100);
3509         }
3510     }
3511 
3512     /*
3513      * Change the backing file. All clusters that are different from the old
3514      * backing file are overwritten in the COW file now, so the visible content
3515      * doesn't change when we switch the backing file.
3516      */
3517     if (out_baseimg && *out_baseimg) {
3518         ret = bdrv_change_backing_file(bs, out_baseimg, out_basefmt);
3519     } else {
3520         ret = bdrv_change_backing_file(bs, NULL, NULL);
3521     }
3522 
3523     if (ret == -ENOSPC) {
3524         error_report("Could not change the backing file to '%s': No "
3525                      "space left in the file header", out_baseimg);
3526     } else if (ret < 0) {
3527         error_report("Could not change the backing file to '%s': %s",
3528             out_baseimg, strerror(-ret));
3529     }
3530 
3531     qemu_progress_print(100, 0);
3532     /*
3533      * TODO At this point it is possible to check if any clusters that are
3534      * allocated in the COW file are the same in the backing file. If so, they
3535      * could be dropped from the COW file. Don't do this before switching the
3536      * backing file, in case of a crash this would lead to corruption.
3537      */
3538 out:
3539     qemu_progress_end();
3540     /* Cleanup */
3541     if (!unsafe) {
3542         blk_unref(blk_old_backing);
3543         blk_unref(blk_new_backing);
3544     }
3545     qemu_vfree(buf_old);
3546     qemu_vfree(buf_new);
3547 
3548     blk_unref(blk);
3549     if (ret) {
3550         return 1;
3551     }
3552     return 0;
3553 }
3554 
3555 static int img_resize(int argc, char **argv)
3556 {
3557     Error *err = NULL;
3558     int c, ret, relative;
3559     const char *filename, *fmt, *size;
3560     int64_t n, total_size, current_size, new_size;
3561     bool quiet = false;
3562     BlockBackend *blk = NULL;
3563     PreallocMode prealloc = PREALLOC_MODE_OFF;
3564     QemuOpts *param;
3565 
3566     static QemuOptsList resize_options = {
3567         .name = "resize_options",
3568         .head = QTAILQ_HEAD_INITIALIZER(resize_options.head),
3569         .desc = {
3570             {
3571                 .name = BLOCK_OPT_SIZE,
3572                 .type = QEMU_OPT_SIZE,
3573                 .help = "Virtual disk size"
3574             }, {
3575                 /* end of list */
3576             }
3577         },
3578     };
3579     bool image_opts = false;
3580     bool shrink = false;
3581 
3582     /* Remove size from argv manually so that negative numbers are not treated
3583      * as options by getopt. */
3584     if (argc < 3) {
3585         error_exit("Not enough arguments");
3586         return 1;
3587     }
3588 
3589     size = argv[--argc];
3590 
3591     /* Parse getopt arguments */
3592     fmt = NULL;
3593     for(;;) {
3594         static const struct option long_options[] = {
3595             {"help", no_argument, 0, 'h'},
3596             {"object", required_argument, 0, OPTION_OBJECT},
3597             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
3598             {"preallocation", required_argument, 0, OPTION_PREALLOCATION},
3599             {"shrink", no_argument, 0, OPTION_SHRINK},
3600             {0, 0, 0, 0}
3601         };
3602         c = getopt_long(argc, argv, ":f:hq",
3603                         long_options, NULL);
3604         if (c == -1) {
3605             break;
3606         }
3607         switch(c) {
3608         case ':':
3609             missing_argument(argv[optind - 1]);
3610             break;
3611         case '?':
3612             unrecognized_option(argv[optind - 1]);
3613             break;
3614         case 'h':
3615             help();
3616             break;
3617         case 'f':
3618             fmt = optarg;
3619             break;
3620         case 'q':
3621             quiet = true;
3622             break;
3623         case OPTION_OBJECT: {
3624             QemuOpts *opts;
3625             opts = qemu_opts_parse_noisily(&qemu_object_opts,
3626                                            optarg, true);
3627             if (!opts) {
3628                 return 1;
3629             }
3630         }   break;
3631         case OPTION_IMAGE_OPTS:
3632             image_opts = true;
3633             break;
3634         case OPTION_PREALLOCATION:
3635             prealloc = qapi_enum_parse(&PreallocMode_lookup, optarg,
3636                                        PREALLOC_MODE__MAX, NULL);
3637             if (prealloc == PREALLOC_MODE__MAX) {
3638                 error_report("Invalid preallocation mode '%s'", optarg);
3639                 return 1;
3640             }
3641             break;
3642         case OPTION_SHRINK:
3643             shrink = true;
3644             break;
3645         }
3646     }
3647     if (optind != argc - 1) {
3648         error_exit("Expecting image file name and size");
3649     }
3650     filename = argv[optind++];
3651 
3652     if (qemu_opts_foreach(&qemu_object_opts,
3653                           user_creatable_add_opts_foreach,
3654                           NULL, &error_fatal)) {
3655         return 1;
3656     }
3657 
3658     /* Choose grow, shrink, or absolute resize mode */
3659     switch (size[0]) {
3660     case '+':
3661         relative = 1;
3662         size++;
3663         break;
3664     case '-':
3665         relative = -1;
3666         size++;
3667         break;
3668     default:
3669         relative = 0;
3670         break;
3671     }
3672 
3673     /* Parse size */
3674     param = qemu_opts_create(&resize_options, NULL, 0, &error_abort);
3675     qemu_opt_set(param, BLOCK_OPT_SIZE, size, &err);
3676     if (err) {
3677         error_report_err(err);
3678         ret = -1;
3679         qemu_opts_del(param);
3680         goto out;
3681     }
3682     n = qemu_opt_get_size(param, BLOCK_OPT_SIZE, 0);
3683     qemu_opts_del(param);
3684 
3685     blk = img_open(image_opts, filename, fmt,
3686                    BDRV_O_RDWR | BDRV_O_RESIZE, false, quiet,
3687                    false);
3688     if (!blk) {
3689         ret = -1;
3690         goto out;
3691     }
3692 
3693     current_size = blk_getlength(blk);
3694     if (current_size < 0) {
3695         error_report("Failed to inquire current image length: %s",
3696                      strerror(-current_size));
3697         ret = -1;
3698         goto out;
3699     }
3700 
3701     if (relative) {
3702         total_size = current_size + n * relative;
3703     } else {
3704         total_size = n;
3705     }
3706     if (total_size <= 0) {
3707         error_report("New image size must be positive");
3708         ret = -1;
3709         goto out;
3710     }
3711 
3712     if (total_size <= current_size && prealloc != PREALLOC_MODE_OFF) {
3713         error_report("Preallocation can only be used for growing images");
3714         ret = -1;
3715         goto out;
3716     }
3717 
3718     if (total_size < current_size && !shrink) {
3719         warn_report("Shrinking an image will delete all data beyond the "
3720                     "shrunken image's end. Before performing such an "
3721                     "operation, make sure there is no important data there.");
3722 
3723         if (g_strcmp0(bdrv_get_format_name(blk_bs(blk)), "raw") != 0) {
3724             error_report(
3725               "Use the --shrink option to perform a shrink operation.");
3726             ret = -1;
3727             goto out;
3728         } else {
3729             warn_report("Using the --shrink option will suppress this message. "
3730                         "Note that future versions of qemu-img may refuse to "
3731                         "shrink images without this option.");
3732         }
3733     }
3734 
3735     ret = blk_truncate(blk, total_size, prealloc, &err);
3736     if (ret < 0) {
3737         error_report_err(err);
3738         goto out;
3739     }
3740 
3741     new_size = blk_getlength(blk);
3742     if (new_size < 0) {
3743         error_report("Failed to verify truncated image length: %s",
3744                      strerror(-new_size));
3745         ret = -1;
3746         goto out;
3747     }
3748 
3749     /* Some block drivers implement a truncation method, but only so
3750      * the user can cause qemu to refresh the image's size from disk.
3751      * The idea is that the user resizes the image outside of qemu and
3752      * then invokes block_resize to inform qemu about it.
3753      * (This includes iscsi and file-posix for device files.)
3754      * Of course, that is not the behavior someone invoking
3755      * qemu-img resize would find useful, so we catch that behavior
3756      * here and tell the user. */
3757     if (new_size != total_size && new_size == current_size) {
3758         error_report("Image was not resized; resizing may not be supported "
3759                      "for this image");
3760         ret = -1;
3761         goto out;
3762     }
3763 
3764     if (new_size != total_size) {
3765         warn_report("Image should have been resized to %" PRIi64
3766                     " bytes, but was resized to %" PRIi64 " bytes",
3767                     total_size, new_size);
3768     }
3769 
3770     qprintf(quiet, "Image resized.\n");
3771 
3772 out:
3773     blk_unref(blk);
3774     if (ret) {
3775         return 1;
3776     }
3777     return 0;
3778 }
3779 
3780 static void amend_status_cb(BlockDriverState *bs,
3781                             int64_t offset, int64_t total_work_size,
3782                             void *opaque)
3783 {
3784     qemu_progress_print(100.f * offset / total_work_size, 0);
3785 }
3786 
3787 static int print_amend_option_help(const char *format)
3788 {
3789     BlockDriver *drv;
3790 
3791     /* Find driver and parse its options */
3792     drv = bdrv_find_format(format);
3793     if (!drv) {
3794         error_report("Unknown file format '%s'", format);
3795         return 1;
3796     }
3797 
3798     if (!drv->bdrv_amend_options) {
3799         error_report("Format driver '%s' does not support option amendment",
3800                      format);
3801         return 1;
3802     }
3803 
3804     /* Every driver supporting amendment must have create_opts */
3805     assert(drv->create_opts);
3806 
3807     printf("Creation options for '%s':\n", format);
3808     qemu_opts_print_help(drv->create_opts, false);
3809     printf("\nNote that not all of these options may be amendable.\n");
3810     return 0;
3811 }
3812 
3813 static int img_amend(int argc, char **argv)
3814 {
3815     Error *err = NULL;
3816     int c, ret = 0;
3817     char *options = NULL;
3818     QemuOptsList *create_opts = NULL;
3819     QemuOpts *opts = NULL;
3820     const char *fmt = NULL, *filename, *cache;
3821     int flags;
3822     bool writethrough;
3823     bool quiet = false, progress = false;
3824     BlockBackend *blk = NULL;
3825     BlockDriverState *bs = NULL;
3826     bool image_opts = false;
3827 
3828     cache = BDRV_DEFAULT_CACHE;
3829     for (;;) {
3830         static const struct option long_options[] = {
3831             {"help", no_argument, 0, 'h'},
3832             {"object", required_argument, 0, OPTION_OBJECT},
3833             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
3834             {0, 0, 0, 0}
3835         };
3836         c = getopt_long(argc, argv, ":ho:f:t:pq",
3837                         long_options, NULL);
3838         if (c == -1) {
3839             break;
3840         }
3841 
3842         switch (c) {
3843         case ':':
3844             missing_argument(argv[optind - 1]);
3845             break;
3846         case '?':
3847             unrecognized_option(argv[optind - 1]);
3848             break;
3849         case 'h':
3850             help();
3851             break;
3852         case 'o':
3853             if (!is_valid_option_list(optarg)) {
3854                 error_report("Invalid option list: %s", optarg);
3855                 ret = -1;
3856                 goto out_no_progress;
3857             }
3858             if (!options) {
3859                 options = g_strdup(optarg);
3860             } else {
3861                 char *old_options = options;
3862                 options = g_strdup_printf("%s,%s", options, optarg);
3863                 g_free(old_options);
3864             }
3865             break;
3866         case 'f':
3867             fmt = optarg;
3868             break;
3869         case 't':
3870             cache = optarg;
3871             break;
3872         case 'p':
3873             progress = true;
3874             break;
3875         case 'q':
3876             quiet = true;
3877             break;
3878         case OPTION_OBJECT:
3879             opts = qemu_opts_parse_noisily(&qemu_object_opts,
3880                                            optarg, true);
3881             if (!opts) {
3882                 ret = -1;
3883                 goto out_no_progress;
3884             }
3885             break;
3886         case OPTION_IMAGE_OPTS:
3887             image_opts = true;
3888             break;
3889         }
3890     }
3891 
3892     if (!options) {
3893         error_exit("Must specify options (-o)");
3894     }
3895 
3896     if (qemu_opts_foreach(&qemu_object_opts,
3897                           user_creatable_add_opts_foreach,
3898                           NULL, &error_fatal)) {
3899         ret = -1;
3900         goto out_no_progress;
3901     }
3902 
3903     if (quiet) {
3904         progress = false;
3905     }
3906     qemu_progress_init(progress, 1.0);
3907 
3908     filename = (optind == argc - 1) ? argv[argc - 1] : NULL;
3909     if (fmt && has_help_option(options)) {
3910         /* If a format is explicitly specified (and possibly no filename is
3911          * given), print option help here */
3912         ret = print_amend_option_help(fmt);
3913         goto out;
3914     }
3915 
3916     if (optind != argc - 1) {
3917         error_report("Expecting one image file name");
3918         ret = -1;
3919         goto out;
3920     }
3921 
3922     flags = BDRV_O_RDWR;
3923     ret = bdrv_parse_cache_mode(cache, &flags, &writethrough);
3924     if (ret < 0) {
3925         error_report("Invalid cache option: %s", cache);
3926         goto out;
3927     }
3928 
3929     blk = img_open(image_opts, filename, fmt, flags, writethrough, quiet,
3930                    false);
3931     if (!blk) {
3932         ret = -1;
3933         goto out;
3934     }
3935     bs = blk_bs(blk);
3936 
3937     fmt = bs->drv->format_name;
3938 
3939     if (has_help_option(options)) {
3940         /* If the format was auto-detected, print option help here */
3941         ret = print_amend_option_help(fmt);
3942         goto out;
3943     }
3944 
3945     if (!bs->drv->bdrv_amend_options) {
3946         error_report("Format driver '%s' does not support option amendment",
3947                      fmt);
3948         ret = -1;
3949         goto out;
3950     }
3951 
3952     /* Every driver supporting amendment must have create_opts */
3953     assert(bs->drv->create_opts);
3954 
3955     create_opts = qemu_opts_append(create_opts, bs->drv->create_opts);
3956     opts = qemu_opts_create(create_opts, NULL, 0, &error_abort);
3957     qemu_opts_do_parse(opts, options, NULL, &err);
3958     if (err) {
3959         error_report_err(err);
3960         ret = -1;
3961         goto out;
3962     }
3963 
3964     /* In case the driver does not call amend_status_cb() */
3965     qemu_progress_print(0.f, 0);
3966     ret = bdrv_amend_options(bs, opts, &amend_status_cb, NULL, &err);
3967     qemu_progress_print(100.f, 0);
3968     if (ret < 0) {
3969         error_report_err(err);
3970         goto out;
3971     }
3972 
3973 out:
3974     qemu_progress_end();
3975 
3976 out_no_progress:
3977     blk_unref(blk);
3978     qemu_opts_del(opts);
3979     qemu_opts_free(create_opts);
3980     g_free(options);
3981 
3982     if (ret) {
3983         return 1;
3984     }
3985     return 0;
3986 }
3987 
3988 typedef struct BenchData {
3989     BlockBackend *blk;
3990     uint64_t image_size;
3991     bool write;
3992     int bufsize;
3993     int step;
3994     int nrreq;
3995     int n;
3996     int flush_interval;
3997     bool drain_on_flush;
3998     uint8_t *buf;
3999     QEMUIOVector *qiov;
4000 
4001     int in_flight;
4002     bool in_flush;
4003     uint64_t offset;
4004 } BenchData;
4005 
4006 static void bench_undrained_flush_cb(void *opaque, int ret)
4007 {
4008     if (ret < 0) {
4009         error_report("Failed flush request: %s", strerror(-ret));
4010         exit(EXIT_FAILURE);
4011     }
4012 }
4013 
4014 static void bench_cb(void *opaque, int ret)
4015 {
4016     BenchData *b = opaque;
4017     BlockAIOCB *acb;
4018 
4019     if (ret < 0) {
4020         error_report("Failed request: %s", strerror(-ret));
4021         exit(EXIT_FAILURE);
4022     }
4023 
4024     if (b->in_flush) {
4025         /* Just finished a flush with drained queue: Start next requests */
4026         assert(b->in_flight == 0);
4027         b->in_flush = false;
4028     } else if (b->in_flight > 0) {
4029         int remaining = b->n - b->in_flight;
4030 
4031         b->n--;
4032         b->in_flight--;
4033 
4034         /* Time for flush? Drain queue if requested, then flush */
4035         if (b->flush_interval && remaining % b->flush_interval == 0) {
4036             if (!b->in_flight || !b->drain_on_flush) {
4037                 BlockCompletionFunc *cb;
4038 
4039                 if (b->drain_on_flush) {
4040                     b->in_flush = true;
4041                     cb = bench_cb;
4042                 } else {
4043                     cb = bench_undrained_flush_cb;
4044                 }
4045 
4046                 acb = blk_aio_flush(b->blk, cb, b);
4047                 if (!acb) {
4048                     error_report("Failed to issue flush request");
4049                     exit(EXIT_FAILURE);
4050                 }
4051             }
4052             if (b->drain_on_flush) {
4053                 return;
4054             }
4055         }
4056     }
4057 
4058     while (b->n > b->in_flight && b->in_flight < b->nrreq) {
4059         int64_t offset = b->offset;
4060         /* blk_aio_* might look for completed I/Os and kick bench_cb
4061          * again, so make sure this operation is counted by in_flight
4062          * and b->offset is ready for the next submission.
4063          */
4064         b->in_flight++;
4065         b->offset += b->step;
4066         b->offset %= b->image_size;
4067         if (b->write) {
4068             acb = blk_aio_pwritev(b->blk, offset, b->qiov, 0, bench_cb, b);
4069         } else {
4070             acb = blk_aio_preadv(b->blk, offset, b->qiov, 0, bench_cb, b);
4071         }
4072         if (!acb) {
4073             error_report("Failed to issue request");
4074             exit(EXIT_FAILURE);
4075         }
4076     }
4077 }
4078 
4079 static int img_bench(int argc, char **argv)
4080 {
4081     int c, ret = 0;
4082     const char *fmt = NULL, *filename;
4083     bool quiet = false;
4084     bool image_opts = false;
4085     bool is_write = false;
4086     int count = 75000;
4087     int depth = 64;
4088     int64_t offset = 0;
4089     size_t bufsize = 4096;
4090     int pattern = 0;
4091     size_t step = 0;
4092     int flush_interval = 0;
4093     bool drain_on_flush = true;
4094     int64_t image_size;
4095     BlockBackend *blk = NULL;
4096     BenchData data = {};
4097     int flags = 0;
4098     bool writethrough = false;
4099     struct timeval t1, t2;
4100     int i;
4101     bool force_share = false;
4102     size_t buf_size;
4103 
4104     for (;;) {
4105         static const struct option long_options[] = {
4106             {"help", no_argument, 0, 'h'},
4107             {"flush-interval", required_argument, 0, OPTION_FLUSH_INTERVAL},
4108             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
4109             {"pattern", required_argument, 0, OPTION_PATTERN},
4110             {"no-drain", no_argument, 0, OPTION_NO_DRAIN},
4111             {"force-share", no_argument, 0, 'U'},
4112             {0, 0, 0, 0}
4113         };
4114         c = getopt_long(argc, argv, ":hc:d:f:no:qs:S:t:wU", long_options, NULL);
4115         if (c == -1) {
4116             break;
4117         }
4118 
4119         switch (c) {
4120         case ':':
4121             missing_argument(argv[optind - 1]);
4122             break;
4123         case '?':
4124             unrecognized_option(argv[optind - 1]);
4125             break;
4126         case 'h':
4127             help();
4128             break;
4129         case 'c':
4130         {
4131             unsigned long res;
4132 
4133             if (qemu_strtoul(optarg, NULL, 0, &res) < 0 || res > INT_MAX) {
4134                 error_report("Invalid request count specified");
4135                 return 1;
4136             }
4137             count = res;
4138             break;
4139         }
4140         case 'd':
4141         {
4142             unsigned long res;
4143 
4144             if (qemu_strtoul(optarg, NULL, 0, &res) < 0 || res > INT_MAX) {
4145                 error_report("Invalid queue depth specified");
4146                 return 1;
4147             }
4148             depth = res;
4149             break;
4150         }
4151         case 'f':
4152             fmt = optarg;
4153             break;
4154         case 'n':
4155             flags |= BDRV_O_NATIVE_AIO;
4156             break;
4157         case 'o':
4158         {
4159             offset = cvtnum(optarg);
4160             if (offset < 0) {
4161                 error_report("Invalid offset specified");
4162                 return 1;
4163             }
4164             break;
4165         }
4166             break;
4167         case 'q':
4168             quiet = true;
4169             break;
4170         case 's':
4171         {
4172             int64_t sval;
4173 
4174             sval = cvtnum(optarg);
4175             if (sval < 0 || sval > INT_MAX) {
4176                 error_report("Invalid buffer size specified");
4177                 return 1;
4178             }
4179 
4180             bufsize = sval;
4181             break;
4182         }
4183         case 'S':
4184         {
4185             int64_t sval;
4186 
4187             sval = cvtnum(optarg);
4188             if (sval < 0 || sval > INT_MAX) {
4189                 error_report("Invalid step size specified");
4190                 return 1;
4191             }
4192 
4193             step = sval;
4194             break;
4195         }
4196         case 't':
4197             ret = bdrv_parse_cache_mode(optarg, &flags, &writethrough);
4198             if (ret < 0) {
4199                 error_report("Invalid cache mode");
4200                 ret = -1;
4201                 goto out;
4202             }
4203             break;
4204         case 'w':
4205             flags |= BDRV_O_RDWR;
4206             is_write = true;
4207             break;
4208         case 'U':
4209             force_share = true;
4210             break;
4211         case OPTION_PATTERN:
4212         {
4213             unsigned long res;
4214 
4215             if (qemu_strtoul(optarg, NULL, 0, &res) < 0 || res > 0xff) {
4216                 error_report("Invalid pattern byte specified");
4217                 return 1;
4218             }
4219             pattern = res;
4220             break;
4221         }
4222         case OPTION_FLUSH_INTERVAL:
4223         {
4224             unsigned long res;
4225 
4226             if (qemu_strtoul(optarg, NULL, 0, &res) < 0 || res > INT_MAX) {
4227                 error_report("Invalid flush interval specified");
4228                 return 1;
4229             }
4230             flush_interval = res;
4231             break;
4232         }
4233         case OPTION_NO_DRAIN:
4234             drain_on_flush = false;
4235             break;
4236         case OPTION_IMAGE_OPTS:
4237             image_opts = true;
4238             break;
4239         }
4240     }
4241 
4242     if (optind != argc - 1) {
4243         error_exit("Expecting one image file name");
4244     }
4245     filename = argv[argc - 1];
4246 
4247     if (!is_write && flush_interval) {
4248         error_report("--flush-interval is only available in write tests");
4249         ret = -1;
4250         goto out;
4251     }
4252     if (flush_interval && flush_interval < depth) {
4253         error_report("Flush interval can't be smaller than depth");
4254         ret = -1;
4255         goto out;
4256     }
4257 
4258     blk = img_open(image_opts, filename, fmt, flags, writethrough, quiet,
4259                    force_share);
4260     if (!blk) {
4261         ret = -1;
4262         goto out;
4263     }
4264 
4265     image_size = blk_getlength(blk);
4266     if (image_size < 0) {
4267         ret = image_size;
4268         goto out;
4269     }
4270 
4271     data = (BenchData) {
4272         .blk            = blk,
4273         .image_size     = image_size,
4274         .bufsize        = bufsize,
4275         .step           = step ?: bufsize,
4276         .nrreq          = depth,
4277         .n              = count,
4278         .offset         = offset,
4279         .write          = is_write,
4280         .flush_interval = flush_interval,
4281         .drain_on_flush = drain_on_flush,
4282     };
4283     printf("Sending %d %s requests, %d bytes each, %d in parallel "
4284            "(starting at offset %" PRId64 ", step size %d)\n",
4285            data.n, data.write ? "write" : "read", data.bufsize, data.nrreq,
4286            data.offset, data.step);
4287     if (flush_interval) {
4288         printf("Sending flush every %d requests\n", flush_interval);
4289     }
4290 
4291     buf_size = data.nrreq * data.bufsize;
4292     data.buf = blk_blockalign(blk, buf_size);
4293     memset(data.buf, pattern, data.nrreq * data.bufsize);
4294 
4295     blk_register_buf(blk, data.buf, buf_size);
4296 
4297     data.qiov = g_new(QEMUIOVector, data.nrreq);
4298     for (i = 0; i < data.nrreq; i++) {
4299         qemu_iovec_init(&data.qiov[i], 1);
4300         qemu_iovec_add(&data.qiov[i],
4301                        data.buf + i * data.bufsize, data.bufsize);
4302     }
4303 
4304     gettimeofday(&t1, NULL);
4305     bench_cb(&data, 0);
4306 
4307     while (data.n > 0) {
4308         main_loop_wait(false);
4309     }
4310     gettimeofday(&t2, NULL);
4311 
4312     printf("Run completed in %3.3f seconds.\n",
4313            (t2.tv_sec - t1.tv_sec)
4314            + ((double)(t2.tv_usec - t1.tv_usec) / 1000000));
4315 
4316 out:
4317     if (data.buf) {
4318         blk_unregister_buf(blk, data.buf);
4319     }
4320     qemu_vfree(data.buf);
4321     blk_unref(blk);
4322 
4323     if (ret) {
4324         return 1;
4325     }
4326     return 0;
4327 }
4328 
4329 #define C_BS      01
4330 #define C_COUNT   02
4331 #define C_IF      04
4332 #define C_OF      010
4333 #define C_SKIP    020
4334 
4335 struct DdInfo {
4336     unsigned int flags;
4337     int64_t count;
4338 };
4339 
4340 struct DdIo {
4341     int bsz;    /* Block size */
4342     char *filename;
4343     uint8_t *buf;
4344     int64_t offset;
4345 };
4346 
4347 struct DdOpts {
4348     const char *name;
4349     int (*f)(const char *, struct DdIo *, struct DdIo *, struct DdInfo *);
4350     unsigned int flag;
4351 };
4352 
4353 static int img_dd_bs(const char *arg,
4354                      struct DdIo *in, struct DdIo *out,
4355                      struct DdInfo *dd)
4356 {
4357     int64_t res;
4358 
4359     res = cvtnum(arg);
4360 
4361     if (res <= 0 || res > INT_MAX) {
4362         error_report("invalid number: '%s'", arg);
4363         return 1;
4364     }
4365     in->bsz = out->bsz = res;
4366 
4367     return 0;
4368 }
4369 
4370 static int img_dd_count(const char *arg,
4371                         struct DdIo *in, struct DdIo *out,
4372                         struct DdInfo *dd)
4373 {
4374     dd->count = cvtnum(arg);
4375 
4376     if (dd->count < 0) {
4377         error_report("invalid number: '%s'", arg);
4378         return 1;
4379     }
4380 
4381     return 0;
4382 }
4383 
4384 static int img_dd_if(const char *arg,
4385                      struct DdIo *in, struct DdIo *out,
4386                      struct DdInfo *dd)
4387 {
4388     in->filename = g_strdup(arg);
4389 
4390     return 0;
4391 }
4392 
4393 static int img_dd_of(const char *arg,
4394                      struct DdIo *in, struct DdIo *out,
4395                      struct DdInfo *dd)
4396 {
4397     out->filename = g_strdup(arg);
4398 
4399     return 0;
4400 }
4401 
4402 static int img_dd_skip(const char *arg,
4403                        struct DdIo *in, struct DdIo *out,
4404                        struct DdInfo *dd)
4405 {
4406     in->offset = cvtnum(arg);
4407 
4408     if (in->offset < 0) {
4409         error_report("invalid number: '%s'", arg);
4410         return 1;
4411     }
4412 
4413     return 0;
4414 }
4415 
4416 static int img_dd(int argc, char **argv)
4417 {
4418     int ret = 0;
4419     char *arg = NULL;
4420     char *tmp;
4421     BlockDriver *drv = NULL, *proto_drv = NULL;
4422     BlockBackend *blk1 = NULL, *blk2 = NULL;
4423     QemuOpts *opts = NULL;
4424     QemuOptsList *create_opts = NULL;
4425     Error *local_err = NULL;
4426     bool image_opts = false;
4427     int c, i;
4428     const char *out_fmt = "raw";
4429     const char *fmt = NULL;
4430     int64_t size = 0;
4431     int64_t block_count = 0, out_pos, in_pos;
4432     bool force_share = false;
4433     struct DdInfo dd = {
4434         .flags = 0,
4435         .count = 0,
4436     };
4437     struct DdIo in = {
4438         .bsz = 512, /* Block size is by default 512 bytes */
4439         .filename = NULL,
4440         .buf = NULL,
4441         .offset = 0
4442     };
4443     struct DdIo out = {
4444         .bsz = 512,
4445         .filename = NULL,
4446         .buf = NULL,
4447         .offset = 0
4448     };
4449 
4450     const struct DdOpts options[] = {
4451         { "bs", img_dd_bs, C_BS },
4452         { "count", img_dd_count, C_COUNT },
4453         { "if", img_dd_if, C_IF },
4454         { "of", img_dd_of, C_OF },
4455         { "skip", img_dd_skip, C_SKIP },
4456         { NULL, NULL, 0 }
4457     };
4458     const struct option long_options[] = {
4459         { "help", no_argument, 0, 'h'},
4460         { "object", required_argument, 0, OPTION_OBJECT},
4461         { "image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
4462         { "force-share", no_argument, 0, 'U'},
4463         { 0, 0, 0, 0 }
4464     };
4465 
4466     while ((c = getopt_long(argc, argv, ":hf:O:U", long_options, NULL))) {
4467         if (c == EOF) {
4468             break;
4469         }
4470         switch (c) {
4471         case 'O':
4472             out_fmt = optarg;
4473             break;
4474         case 'f':
4475             fmt = optarg;
4476             break;
4477         case ':':
4478             missing_argument(argv[optind - 1]);
4479             break;
4480         case '?':
4481             unrecognized_option(argv[optind - 1]);
4482             break;
4483         case 'h':
4484             help();
4485             break;
4486         case 'U':
4487             force_share = true;
4488             break;
4489         case OPTION_OBJECT:
4490             if (!qemu_opts_parse_noisily(&qemu_object_opts, optarg, true)) {
4491                 ret = -1;
4492                 goto out;
4493             }
4494             break;
4495         case OPTION_IMAGE_OPTS:
4496             image_opts = true;
4497             break;
4498         }
4499     }
4500 
4501     for (i = optind; i < argc; i++) {
4502         int j;
4503         arg = g_strdup(argv[i]);
4504 
4505         tmp = strchr(arg, '=');
4506         if (tmp == NULL) {
4507             error_report("unrecognized operand %s", arg);
4508             ret = -1;
4509             goto out;
4510         }
4511 
4512         *tmp++ = '\0';
4513 
4514         for (j = 0; options[j].name != NULL; j++) {
4515             if (!strcmp(arg, options[j].name)) {
4516                 break;
4517             }
4518         }
4519         if (options[j].name == NULL) {
4520             error_report("unrecognized operand %s", arg);
4521             ret = -1;
4522             goto out;
4523         }
4524 
4525         if (options[j].f(tmp, &in, &out, &dd) != 0) {
4526             ret = -1;
4527             goto out;
4528         }
4529         dd.flags |= options[j].flag;
4530         g_free(arg);
4531         arg = NULL;
4532     }
4533 
4534     if (!(dd.flags & C_IF && dd.flags & C_OF)) {
4535         error_report("Must specify both input and output files");
4536         ret = -1;
4537         goto out;
4538     }
4539 
4540     if (qemu_opts_foreach(&qemu_object_opts,
4541                           user_creatable_add_opts_foreach,
4542                           NULL, &error_fatal)) {
4543         ret = -1;
4544         goto out;
4545     }
4546 
4547     blk1 = img_open(image_opts, in.filename, fmt, 0, false, false,
4548                     force_share);
4549 
4550     if (!blk1) {
4551         ret = -1;
4552         goto out;
4553     }
4554 
4555     drv = bdrv_find_format(out_fmt);
4556     if (!drv) {
4557         error_report("Unknown file format");
4558         ret = -1;
4559         goto out;
4560     }
4561     proto_drv = bdrv_find_protocol(out.filename, true, &local_err);
4562 
4563     if (!proto_drv) {
4564         error_report_err(local_err);
4565         ret = -1;
4566         goto out;
4567     }
4568     if (!drv->create_opts) {
4569         error_report("Format driver '%s' does not support image creation",
4570                      drv->format_name);
4571         ret = -1;
4572         goto out;
4573     }
4574     if (!proto_drv->create_opts) {
4575         error_report("Protocol driver '%s' does not support image creation",
4576                      proto_drv->format_name);
4577         ret = -1;
4578         goto out;
4579     }
4580     create_opts = qemu_opts_append(create_opts, drv->create_opts);
4581     create_opts = qemu_opts_append(create_opts, proto_drv->create_opts);
4582 
4583     opts = qemu_opts_create(create_opts, NULL, 0, &error_abort);
4584 
4585     size = blk_getlength(blk1);
4586     if (size < 0) {
4587         error_report("Failed to get size for '%s'", in.filename);
4588         ret = -1;
4589         goto out;
4590     }
4591 
4592     if (dd.flags & C_COUNT && dd.count <= INT64_MAX / in.bsz &&
4593         dd.count * in.bsz < size) {
4594         size = dd.count * in.bsz;
4595     }
4596 
4597     /* Overflow means the specified offset is beyond input image's size */
4598     if (dd.flags & C_SKIP && (in.offset > INT64_MAX / in.bsz ||
4599                               size < in.bsz * in.offset)) {
4600         qemu_opt_set_number(opts, BLOCK_OPT_SIZE, 0, &error_abort);
4601     } else {
4602         qemu_opt_set_number(opts, BLOCK_OPT_SIZE,
4603                             size - in.bsz * in.offset, &error_abort);
4604     }
4605 
4606     ret = bdrv_create(drv, out.filename, opts, &local_err);
4607     if (ret < 0) {
4608         error_reportf_err(local_err,
4609                           "%s: error while creating output image: ",
4610                           out.filename);
4611         ret = -1;
4612         goto out;
4613     }
4614 
4615     /* TODO, we can't honour --image-opts for the target,
4616      * since it needs to be given in a format compatible
4617      * with the bdrv_create() call above which does not
4618      * support image-opts style.
4619      */
4620     blk2 = img_open_file(out.filename, NULL, out_fmt, BDRV_O_RDWR,
4621                          false, false, false);
4622 
4623     if (!blk2) {
4624         ret = -1;
4625         goto out;
4626     }
4627 
4628     if (dd.flags & C_SKIP && (in.offset > INT64_MAX / in.bsz ||
4629                               size < in.offset * in.bsz)) {
4630         /* We give a warning if the skip option is bigger than the input
4631          * size and create an empty output disk image (i.e. like dd(1)).
4632          */
4633         error_report("%s: cannot skip to specified offset", in.filename);
4634         in_pos = size;
4635     } else {
4636         in_pos = in.offset * in.bsz;
4637     }
4638 
4639     in.buf = g_new(uint8_t, in.bsz);
4640 
4641     for (out_pos = 0; in_pos < size; block_count++) {
4642         int in_ret, out_ret;
4643 
4644         if (in_pos + in.bsz > size) {
4645             in_ret = blk_pread(blk1, in_pos, in.buf, size - in_pos);
4646         } else {
4647             in_ret = blk_pread(blk1, in_pos, in.buf, in.bsz);
4648         }
4649         if (in_ret < 0) {
4650             error_report("error while reading from input image file: %s",
4651                          strerror(-in_ret));
4652             ret = -1;
4653             goto out;
4654         }
4655         in_pos += in_ret;
4656 
4657         out_ret = blk_pwrite(blk2, out_pos, in.buf, in_ret, 0);
4658 
4659         if (out_ret < 0) {
4660             error_report("error while writing to output image file: %s",
4661                          strerror(-out_ret));
4662             ret = -1;
4663             goto out;
4664         }
4665         out_pos += out_ret;
4666     }
4667 
4668 out:
4669     g_free(arg);
4670     qemu_opts_del(opts);
4671     qemu_opts_free(create_opts);
4672     blk_unref(blk1);
4673     blk_unref(blk2);
4674     g_free(in.filename);
4675     g_free(out.filename);
4676     g_free(in.buf);
4677     g_free(out.buf);
4678 
4679     if (ret) {
4680         return 1;
4681     }
4682     return 0;
4683 }
4684 
4685 static void dump_json_block_measure_info(BlockMeasureInfo *info)
4686 {
4687     QString *str;
4688     QObject *obj;
4689     Visitor *v = qobject_output_visitor_new(&obj);
4690 
4691     visit_type_BlockMeasureInfo(v, NULL, &info, &error_abort);
4692     visit_complete(v, &obj);
4693     str = qobject_to_json_pretty(obj);
4694     assert(str != NULL);
4695     printf("%s\n", qstring_get_str(str));
4696     qobject_unref(obj);
4697     visit_free(v);
4698     qobject_unref(str);
4699 }
4700 
4701 static int img_measure(int argc, char **argv)
4702 {
4703     static const struct option long_options[] = {
4704         {"help", no_argument, 0, 'h'},
4705         {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
4706         {"object", required_argument, 0, OPTION_OBJECT},
4707         {"output", required_argument, 0, OPTION_OUTPUT},
4708         {"size", required_argument, 0, OPTION_SIZE},
4709         {"force-share", no_argument, 0, 'U'},
4710         {0, 0, 0, 0}
4711     };
4712     OutputFormat output_format = OFORMAT_HUMAN;
4713     BlockBackend *in_blk = NULL;
4714     BlockDriver *drv;
4715     const char *filename = NULL;
4716     const char *fmt = NULL;
4717     const char *out_fmt = "raw";
4718     char *options = NULL;
4719     char *snapshot_name = NULL;
4720     bool force_share = false;
4721     QemuOpts *opts = NULL;
4722     QemuOpts *object_opts = NULL;
4723     QemuOpts *sn_opts = NULL;
4724     QemuOptsList *create_opts = NULL;
4725     bool image_opts = false;
4726     uint64_t img_size = UINT64_MAX;
4727     BlockMeasureInfo *info = NULL;
4728     Error *local_err = NULL;
4729     int ret = 1;
4730     int c;
4731 
4732     while ((c = getopt_long(argc, argv, "hf:O:o:l:U",
4733                             long_options, NULL)) != -1) {
4734         switch (c) {
4735         case '?':
4736         case 'h':
4737             help();
4738             break;
4739         case 'f':
4740             fmt = optarg;
4741             break;
4742         case 'O':
4743             out_fmt = optarg;
4744             break;
4745         case 'o':
4746             if (!is_valid_option_list(optarg)) {
4747                 error_report("Invalid option list: %s", optarg);
4748                 goto out;
4749             }
4750             if (!options) {
4751                 options = g_strdup(optarg);
4752             } else {
4753                 char *old_options = options;
4754                 options = g_strdup_printf("%s,%s", options, optarg);
4755                 g_free(old_options);
4756             }
4757             break;
4758         case 'l':
4759             if (strstart(optarg, SNAPSHOT_OPT_BASE, NULL)) {
4760                 sn_opts = qemu_opts_parse_noisily(&internal_snapshot_opts,
4761                                                   optarg, false);
4762                 if (!sn_opts) {
4763                     error_report("Failed in parsing snapshot param '%s'",
4764                                  optarg);
4765                     goto out;
4766                 }
4767             } else {
4768                 snapshot_name = optarg;
4769             }
4770             break;
4771         case 'U':
4772             force_share = true;
4773             break;
4774         case OPTION_OBJECT:
4775             object_opts = qemu_opts_parse_noisily(&qemu_object_opts,
4776                                                   optarg, true);
4777             if (!object_opts) {
4778                 goto out;
4779             }
4780             break;
4781         case OPTION_IMAGE_OPTS:
4782             image_opts = true;
4783             break;
4784         case OPTION_OUTPUT:
4785             if (!strcmp(optarg, "json")) {
4786                 output_format = OFORMAT_JSON;
4787             } else if (!strcmp(optarg, "human")) {
4788                 output_format = OFORMAT_HUMAN;
4789             } else {
4790                 error_report("--output must be used with human or json "
4791                              "as argument.");
4792                 goto out;
4793             }
4794             break;
4795         case OPTION_SIZE:
4796         {
4797             int64_t sval;
4798 
4799             sval = cvtnum(optarg);
4800             if (sval < 0) {
4801                 if (sval == -ERANGE) {
4802                     error_report("Image size must be less than 8 EiB!");
4803                 } else {
4804                     error_report("Invalid image size specified! You may use "
4805                                  "k, M, G, T, P or E suffixes for ");
4806                     error_report("kilobytes, megabytes, gigabytes, terabytes, "
4807                                  "petabytes and exabytes.");
4808                 }
4809                 goto out;
4810             }
4811             img_size = (uint64_t)sval;
4812         }
4813         break;
4814         }
4815     }
4816 
4817     if (qemu_opts_foreach(&qemu_object_opts,
4818                           user_creatable_add_opts_foreach,
4819                           NULL, &error_fatal)) {
4820         goto out;
4821     }
4822 
4823     if (argc - optind > 1) {
4824         error_report("At most one filename argument is allowed.");
4825         goto out;
4826     } else if (argc - optind == 1) {
4827         filename = argv[optind];
4828     }
4829 
4830     if (!filename &&
4831         (object_opts || image_opts || fmt || snapshot_name || sn_opts)) {
4832         error_report("--object, --image-opts, -f, and -l "
4833                      "require a filename argument.");
4834         goto out;
4835     }
4836     if (filename && img_size != UINT64_MAX) {
4837         error_report("--size N cannot be used together with a filename.");
4838         goto out;
4839     }
4840     if (!filename && img_size == UINT64_MAX) {
4841         error_report("Either --size N or one filename must be specified.");
4842         goto out;
4843     }
4844 
4845     if (filename) {
4846         in_blk = img_open(image_opts, filename, fmt, 0,
4847                           false, false, force_share);
4848         if (!in_blk) {
4849             goto out;
4850         }
4851 
4852         if (sn_opts) {
4853             bdrv_snapshot_load_tmp(blk_bs(in_blk),
4854                     qemu_opt_get(sn_opts, SNAPSHOT_OPT_ID),
4855                     qemu_opt_get(sn_opts, SNAPSHOT_OPT_NAME),
4856                     &local_err);
4857         } else if (snapshot_name != NULL) {
4858             bdrv_snapshot_load_tmp_by_id_or_name(blk_bs(in_blk),
4859                     snapshot_name, &local_err);
4860         }
4861         if (local_err) {
4862             error_reportf_err(local_err, "Failed to load snapshot: ");
4863             goto out;
4864         }
4865     }
4866 
4867     drv = bdrv_find_format(out_fmt);
4868     if (!drv) {
4869         error_report("Unknown file format '%s'", out_fmt);
4870         goto out;
4871     }
4872     if (!drv->create_opts) {
4873         error_report("Format driver '%s' does not support image creation",
4874                      drv->format_name);
4875         goto out;
4876     }
4877 
4878     create_opts = qemu_opts_append(create_opts, drv->create_opts);
4879     create_opts = qemu_opts_append(create_opts, bdrv_file.create_opts);
4880     opts = qemu_opts_create(create_opts, NULL, 0, &error_abort);
4881     if (options) {
4882         qemu_opts_do_parse(opts, options, NULL, &local_err);
4883         if (local_err) {
4884             error_report_err(local_err);
4885             error_report("Invalid options for file format '%s'", out_fmt);
4886             goto out;
4887         }
4888     }
4889     if (img_size != UINT64_MAX) {
4890         qemu_opt_set_number(opts, BLOCK_OPT_SIZE, img_size, &error_abort);
4891     }
4892 
4893     info = bdrv_measure(drv, opts, in_blk ? blk_bs(in_blk) : NULL, &local_err);
4894     if (local_err) {
4895         error_report_err(local_err);
4896         goto out;
4897     }
4898 
4899     if (output_format == OFORMAT_HUMAN) {
4900         printf("required size: %" PRIu64 "\n", info->required);
4901         printf("fully allocated size: %" PRIu64 "\n", info->fully_allocated);
4902     } else {
4903         dump_json_block_measure_info(info);
4904     }
4905 
4906     ret = 0;
4907 
4908 out:
4909     qapi_free_BlockMeasureInfo(info);
4910     qemu_opts_del(object_opts);
4911     qemu_opts_del(opts);
4912     qemu_opts_del(sn_opts);
4913     qemu_opts_free(create_opts);
4914     g_free(options);
4915     blk_unref(in_blk);
4916     return ret;
4917 }
4918 
4919 static const img_cmd_t img_cmds[] = {
4920 #define DEF(option, callback, arg_string)        \
4921     { option, callback },
4922 #include "qemu-img-cmds.h"
4923 #undef DEF
4924     { NULL, NULL, },
4925 };
4926 
4927 int main(int argc, char **argv)
4928 {
4929     const img_cmd_t *cmd;
4930     const char *cmdname;
4931     Error *local_error = NULL;
4932     char *trace_file = NULL;
4933     int c;
4934     static const struct option long_options[] = {
4935         {"help", no_argument, 0, 'h'},
4936         {"version", no_argument, 0, 'V'},
4937         {"trace", required_argument, NULL, 'T'},
4938         {0, 0, 0, 0}
4939     };
4940 
4941 #ifdef CONFIG_POSIX
4942     signal(SIGPIPE, SIG_IGN);
4943 #endif
4944 
4945     error_init(argv[0]);
4946     module_call_init(MODULE_INIT_TRACE);
4947     qemu_init_exec_dir(argv[0]);
4948 
4949     if (qemu_init_main_loop(&local_error)) {
4950         error_report_err(local_error);
4951         exit(EXIT_FAILURE);
4952     }
4953 
4954     qcrypto_init(&error_fatal);
4955 
4956     module_call_init(MODULE_INIT_QOM);
4957     bdrv_init();
4958     if (argc < 2) {
4959         error_exit("Not enough arguments");
4960     }
4961 
4962     qemu_add_opts(&qemu_object_opts);
4963     qemu_add_opts(&qemu_source_opts);
4964     qemu_add_opts(&qemu_trace_opts);
4965 
4966     while ((c = getopt_long(argc, argv, "+:hVT:", long_options, NULL)) != -1) {
4967         switch (c) {
4968         case ':':
4969             missing_argument(argv[optind - 1]);
4970             return 0;
4971         case '?':
4972             unrecognized_option(argv[optind - 1]);
4973             return 0;
4974         case 'h':
4975             help();
4976             return 0;
4977         case 'V':
4978             printf(QEMU_IMG_VERSION);
4979             return 0;
4980         case 'T':
4981             g_free(trace_file);
4982             trace_file = trace_opt_parse(optarg);
4983             break;
4984         }
4985     }
4986 
4987     cmdname = argv[optind];
4988 
4989     /* reset getopt_long scanning */
4990     argc -= optind;
4991     if (argc < 1) {
4992         return 0;
4993     }
4994     argv += optind;
4995     qemu_reset_optind();
4996 
4997     if (!trace_init_backends()) {
4998         exit(1);
4999     }
5000     trace_init_file(trace_file);
5001     qemu_set_log(LOG_TRACE);
5002 
5003     /* find the command */
5004     for (cmd = img_cmds; cmd->name != NULL; cmd++) {
5005         if (!strcmp(cmdname, cmd->name)) {
5006             return cmd->handler(argc, argv);
5007         }
5008     }
5009 
5010     /* not found */
5011     error_exit("Command not found: %s", cmdname);
5012 }
5013