xref: /openbmc/qemu/qemu-img.c (revision 1a7c00bb)
1 /*
2  * QEMU disk image utility
3  *
4  * Copyright (c) 2003-2008 Fabrice Bellard
5  *
6  * Permission is hereby granted, free of charge, to any person obtaining a copy
7  * of this software and associated documentation files (the "Software"), to deal
8  * in the Software without restriction, including without limitation the rights
9  * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
10  * copies of the Software, and to permit persons to whom the Software is
11  * furnished to do so, subject to the following conditions:
12  *
13  * The above copyright notice and this permission notice shall be included in
14  * all copies or substantial portions of the Software.
15  *
16  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
17  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
18  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
19  * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
20  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
21  * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
22  * THE SOFTWARE.
23  */
24 
25 #include "qemu/osdep.h"
26 #include <getopt.h>
27 
28 #include "qemu-version.h"
29 #include "qapi/error.h"
30 #include "qapi/qapi-visit-block-core.h"
31 #include "qapi/qobject-output-visitor.h"
32 #include "qapi/qmp/qjson.h"
33 #include "qapi/qmp/qdict.h"
34 #include "qapi/qmp/qstring.h"
35 #include "qemu/cutils.h"
36 #include "qemu/config-file.h"
37 #include "qemu/option.h"
38 #include "qemu/error-report.h"
39 #include "qemu/log.h"
40 #include "qom/object_interfaces.h"
41 #include "sysemu/sysemu.h"
42 #include "sysemu/block-backend.h"
43 #include "block/block_int.h"
44 #include "block/blockjob.h"
45 #include "block/qapi.h"
46 #include "crypto/init.h"
47 #include "trace/control.h"
48 
49 #define QEMU_IMG_VERSION "qemu-img version " QEMU_FULL_VERSION \
50                           "\n" QEMU_COPYRIGHT "\n"
51 
52 typedef struct img_cmd_t {
53     const char *name;
54     int (*handler)(int argc, char **argv);
55 } img_cmd_t;
56 
57 enum {
58     OPTION_OUTPUT = 256,
59     OPTION_BACKING_CHAIN = 257,
60     OPTION_OBJECT = 258,
61     OPTION_IMAGE_OPTS = 259,
62     OPTION_PATTERN = 260,
63     OPTION_FLUSH_INTERVAL = 261,
64     OPTION_NO_DRAIN = 262,
65     OPTION_TARGET_IMAGE_OPTS = 263,
66     OPTION_SIZE = 264,
67     OPTION_PREALLOCATION = 265,
68     OPTION_SHRINK = 266,
69 };
70 
71 typedef enum OutputFormat {
72     OFORMAT_JSON,
73     OFORMAT_HUMAN,
74 } OutputFormat;
75 
76 /* Default to cache=writeback as data integrity is not important for qemu-img */
77 #define BDRV_DEFAULT_CACHE "writeback"
78 
79 static void format_print(void *opaque, const char *name)
80 {
81     printf(" %s", name);
82 }
83 
84 static void QEMU_NORETURN GCC_FMT_ATTR(1, 2) error_exit(const char *fmt, ...)
85 {
86     va_list ap;
87 
88     error_printf("qemu-img: ");
89 
90     va_start(ap, fmt);
91     error_vprintf(fmt, ap);
92     va_end(ap);
93 
94     error_printf("\nTry 'qemu-img --help' for more information\n");
95     exit(EXIT_FAILURE);
96 }
97 
98 static void QEMU_NORETURN missing_argument(const char *option)
99 {
100     error_exit("missing argument for option '%s'", option);
101 }
102 
103 static void QEMU_NORETURN unrecognized_option(const char *option)
104 {
105     error_exit("unrecognized option '%s'", option);
106 }
107 
108 /* Please keep in synch with qemu-img.texi */
109 static void QEMU_NORETURN help(void)
110 {
111     const char *help_msg =
112            QEMU_IMG_VERSION
113            "usage: qemu-img [standard options] command [command options]\n"
114            "QEMU disk image utility\n"
115            "\n"
116            "    '-h', '--help'       display this help and exit\n"
117            "    '-V', '--version'    output version information and exit\n"
118            "    '-T', '--trace'      [[enable=]<pattern>][,events=<file>][,file=<file>]\n"
119            "                         specify tracing options\n"
120            "\n"
121            "Command syntax:\n"
122 #define DEF(option, callback, arg_string)        \
123            "  " arg_string "\n"
124 #include "qemu-img-cmds.h"
125 #undef DEF
126            "\n"
127            "Command parameters:\n"
128            "  'filename' is a disk image filename\n"
129            "  'objectdef' is a QEMU user creatable object definition. See the qemu(1)\n"
130            "    manual page for a description of the object properties. The most common\n"
131            "    object type is a 'secret', which is used to supply passwords and/or\n"
132            "    encryption keys.\n"
133            "  'fmt' is the disk image format. It is guessed automatically in most cases\n"
134            "  'cache' is the cache mode used to write the output disk image, the valid\n"
135            "    options are: 'none', 'writeback' (default, except for convert), 'writethrough',\n"
136            "    'directsync' and 'unsafe' (default for convert)\n"
137            "  'src_cache' is the cache mode used to read input disk images, the valid\n"
138            "    options are the same as for the 'cache' option\n"
139            "  'size' is the disk image size in bytes. Optional suffixes\n"
140            "    'k' or 'K' (kilobyte, 1024), 'M' (megabyte, 1024k), 'G' (gigabyte, 1024M),\n"
141            "    'T' (terabyte, 1024G), 'P' (petabyte, 1024T) and 'E' (exabyte, 1024P)  are\n"
142            "    supported. 'b' is ignored.\n"
143            "  'output_filename' is the destination disk image filename\n"
144            "  'output_fmt' is the destination format\n"
145            "  'options' is a comma separated list of format specific options in a\n"
146            "    name=value format. Use -o ? for an overview of the options supported by the\n"
147            "    used format\n"
148            "  'snapshot_param' is param used for internal snapshot, format\n"
149            "    is 'snapshot.id=[ID],snapshot.name=[NAME]', or\n"
150            "    '[ID_OR_NAME]'\n"
151            "  '-c' indicates that target image must be compressed (qcow format only)\n"
152            "  '-u' allows unsafe backing chains. For rebasing, it is assumed that old and\n"
153            "       new backing file match exactly. The image doesn't need a working\n"
154            "       backing file before rebasing in this case (useful for renaming the\n"
155            "       backing file). For image creation, allow creating without attempting\n"
156            "       to open the backing file.\n"
157            "  '-h' with or without a command shows this help and lists the supported formats\n"
158            "  '-p' show progress of command (only certain commands)\n"
159            "  '-q' use Quiet mode - do not print any output (except errors)\n"
160            "  '-S' indicates the consecutive number of bytes (defaults to 4k) that must\n"
161            "       contain only zeros for qemu-img to create a sparse image during\n"
162            "       conversion. If the number of bytes is 0, the source will not be scanned for\n"
163            "       unallocated or zero sectors, and the destination image will always be\n"
164            "       fully allocated\n"
165            "  '--output' takes the format in which the output must be done (human or json)\n"
166            "  '-n' skips the target volume creation (useful if the volume is created\n"
167            "       prior to running qemu-img)\n"
168            "\n"
169            "Parameters to check subcommand:\n"
170            "  '-r' tries to repair any inconsistencies that are found during the check.\n"
171            "       '-r leaks' repairs only cluster leaks, whereas '-r all' fixes all\n"
172            "       kinds of errors, with a higher risk of choosing the wrong fix or\n"
173            "       hiding corruption that has already occurred.\n"
174            "\n"
175            "Parameters to convert subcommand:\n"
176            "  '-m' specifies how many coroutines work in parallel during the convert\n"
177            "       process (defaults to 8)\n"
178            "  '-W' allow to write to the target out of order rather than sequential\n"
179            "\n"
180            "Parameters to snapshot subcommand:\n"
181            "  'snapshot' is the name of the snapshot to create, apply or delete\n"
182            "  '-a' applies a snapshot (revert disk to saved state)\n"
183            "  '-c' creates a snapshot\n"
184            "  '-d' deletes a snapshot\n"
185            "  '-l' lists all snapshots in the given image\n"
186            "\n"
187            "Parameters to compare subcommand:\n"
188            "  '-f' first image format\n"
189            "  '-F' second image format\n"
190            "  '-s' run in Strict mode - fail on different image size or sector allocation\n"
191            "\n"
192            "Parameters to dd subcommand:\n"
193            "  'bs=BYTES' read and write up to BYTES bytes at a time "
194            "(default: 512)\n"
195            "  'count=N' copy only N input blocks\n"
196            "  'if=FILE' read from FILE\n"
197            "  'of=FILE' write to FILE\n"
198            "  'skip=N' skip N bs-sized blocks at the start of input\n";
199 
200     printf("%s\nSupported formats:", help_msg);
201     bdrv_iterate_format(format_print, NULL, false);
202     printf("\n\n" QEMU_HELP_BOTTOM "\n");
203     exit(EXIT_SUCCESS);
204 }
205 
206 static QemuOptsList qemu_object_opts = {
207     .name = "object",
208     .implied_opt_name = "qom-type",
209     .head = QTAILQ_HEAD_INITIALIZER(qemu_object_opts.head),
210     .desc = {
211         { }
212     },
213 };
214 
215 static QemuOptsList qemu_source_opts = {
216     .name = "source",
217     .implied_opt_name = "file",
218     .head = QTAILQ_HEAD_INITIALIZER(qemu_source_opts.head),
219     .desc = {
220         { }
221     },
222 };
223 
224 static int GCC_FMT_ATTR(2, 3) qprintf(bool quiet, const char *fmt, ...)
225 {
226     int ret = 0;
227     if (!quiet) {
228         va_list args;
229         va_start(args, fmt);
230         ret = vprintf(fmt, args);
231         va_end(args);
232     }
233     return ret;
234 }
235 
236 
237 static int print_block_option_help(const char *filename, const char *fmt)
238 {
239     BlockDriver *drv, *proto_drv;
240     QemuOptsList *create_opts = NULL;
241     Error *local_err = NULL;
242 
243     /* Find driver and parse its options */
244     drv = bdrv_find_format(fmt);
245     if (!drv) {
246         error_report("Unknown file format '%s'", fmt);
247         return 1;
248     }
249 
250     if (!drv->create_opts) {
251         error_report("Format driver '%s' does not support image creation", fmt);
252         return 1;
253     }
254 
255     create_opts = qemu_opts_append(create_opts, drv->create_opts);
256     if (filename) {
257         proto_drv = bdrv_find_protocol(filename, true, &local_err);
258         if (!proto_drv) {
259             error_report_err(local_err);
260             qemu_opts_free(create_opts);
261             return 1;
262         }
263         if (!proto_drv->create_opts) {
264             error_report("Protocol driver '%s' does not support image creation",
265                          proto_drv->format_name);
266             qemu_opts_free(create_opts);
267             return 1;
268         }
269         create_opts = qemu_opts_append(create_opts, proto_drv->create_opts);
270     }
271 
272     printf("Supported options:\n");
273     qemu_opts_print_help(create_opts, false);
274     qemu_opts_free(create_opts);
275     return 0;
276 }
277 
278 
279 static BlockBackend *img_open_opts(const char *optstr,
280                                    QemuOpts *opts, int flags, bool writethrough,
281                                    bool quiet, bool force_share)
282 {
283     QDict *options;
284     Error *local_err = NULL;
285     BlockBackend *blk;
286     options = qemu_opts_to_qdict(opts, NULL);
287     if (force_share) {
288         if (qdict_haskey(options, BDRV_OPT_FORCE_SHARE)
289             && strcmp(qdict_get_str(options, BDRV_OPT_FORCE_SHARE), "on")) {
290             error_report("--force-share/-U conflicts with image options");
291             qobject_unref(options);
292             return NULL;
293         }
294         qdict_put_str(options, BDRV_OPT_FORCE_SHARE, "on");
295     }
296     blk = blk_new_open(NULL, NULL, options, flags, &local_err);
297     if (!blk) {
298         error_reportf_err(local_err, "Could not open '%s': ", optstr);
299         return NULL;
300     }
301     blk_set_enable_write_cache(blk, !writethrough);
302 
303     return blk;
304 }
305 
306 static BlockBackend *img_open_file(const char *filename,
307                                    QDict *options,
308                                    const char *fmt, int flags,
309                                    bool writethrough, bool quiet,
310                                    bool force_share)
311 {
312     BlockBackend *blk;
313     Error *local_err = NULL;
314 
315     if (!options) {
316         options = qdict_new();
317     }
318     if (fmt) {
319         qdict_put_str(options, "driver", fmt);
320     }
321 
322     if (force_share) {
323         qdict_put_bool(options, BDRV_OPT_FORCE_SHARE, true);
324     }
325     blk = blk_new_open(filename, NULL, options, flags, &local_err);
326     if (!blk) {
327         error_reportf_err(local_err, "Could not open '%s': ", filename);
328         return NULL;
329     }
330     blk_set_enable_write_cache(blk, !writethrough);
331 
332     return blk;
333 }
334 
335 
336 static int img_add_key_secrets(void *opaque,
337                                const char *name, const char *value,
338                                Error **errp)
339 {
340     QDict *options = opaque;
341 
342     if (g_str_has_suffix(name, "key-secret")) {
343         qdict_put_str(options, name, value);
344     }
345 
346     return 0;
347 }
348 
349 
350 static BlockBackend *img_open(bool image_opts,
351                               const char *filename,
352                               const char *fmt, int flags, bool writethrough,
353                               bool quiet, bool force_share)
354 {
355     BlockBackend *blk;
356     if (image_opts) {
357         QemuOpts *opts;
358         if (fmt) {
359             error_report("--image-opts and --format are mutually exclusive");
360             return NULL;
361         }
362         opts = qemu_opts_parse_noisily(qemu_find_opts("source"),
363                                        filename, true);
364         if (!opts) {
365             return NULL;
366         }
367         blk = img_open_opts(filename, opts, flags, writethrough, quiet,
368                             force_share);
369     } else {
370         blk = img_open_file(filename, NULL, fmt, flags, writethrough, quiet,
371                             force_share);
372     }
373     return blk;
374 }
375 
376 
377 static int add_old_style_options(const char *fmt, QemuOpts *opts,
378                                  const char *base_filename,
379                                  const char *base_fmt)
380 {
381     Error *err = NULL;
382 
383     if (base_filename) {
384         qemu_opt_set(opts, BLOCK_OPT_BACKING_FILE, base_filename, &err);
385         if (err) {
386             error_report("Backing file not supported for file format '%s'",
387                          fmt);
388             error_free(err);
389             return -1;
390         }
391     }
392     if (base_fmt) {
393         qemu_opt_set(opts, BLOCK_OPT_BACKING_FMT, base_fmt, &err);
394         if (err) {
395             error_report("Backing file format not supported for file "
396                          "format '%s'", fmt);
397             error_free(err);
398             return -1;
399         }
400     }
401     return 0;
402 }
403 
404 static int64_t cvtnum(const char *s)
405 {
406     int err;
407     uint64_t value;
408 
409     err = qemu_strtosz(s, NULL, &value);
410     if (err < 0) {
411         return err;
412     }
413     if (value > INT64_MAX) {
414         return -ERANGE;
415     }
416     return value;
417 }
418 
419 static int img_create(int argc, char **argv)
420 {
421     int c;
422     uint64_t img_size = -1;
423     const char *fmt = "raw";
424     const char *base_fmt = NULL;
425     const char *filename;
426     const char *base_filename = NULL;
427     char *options = NULL;
428     Error *local_err = NULL;
429     bool quiet = false;
430     int flags = 0;
431 
432     for(;;) {
433         static const struct option long_options[] = {
434             {"help", no_argument, 0, 'h'},
435             {"object", required_argument, 0, OPTION_OBJECT},
436             {0, 0, 0, 0}
437         };
438         c = getopt_long(argc, argv, ":F:b:f:ho:qu",
439                         long_options, NULL);
440         if (c == -1) {
441             break;
442         }
443         switch(c) {
444         case ':':
445             missing_argument(argv[optind - 1]);
446             break;
447         case '?':
448             unrecognized_option(argv[optind - 1]);
449             break;
450         case 'h':
451             help();
452             break;
453         case 'F':
454             base_fmt = optarg;
455             break;
456         case 'b':
457             base_filename = optarg;
458             break;
459         case 'f':
460             fmt = optarg;
461             break;
462         case 'o':
463             if (!is_valid_option_list(optarg)) {
464                 error_report("Invalid option list: %s", optarg);
465                 goto fail;
466             }
467             if (!options) {
468                 options = g_strdup(optarg);
469             } else {
470                 char *old_options = options;
471                 options = g_strdup_printf("%s,%s", options, optarg);
472                 g_free(old_options);
473             }
474             break;
475         case 'q':
476             quiet = true;
477             break;
478         case 'u':
479             flags |= BDRV_O_NO_BACKING;
480             break;
481         case OPTION_OBJECT: {
482             QemuOpts *opts;
483             opts = qemu_opts_parse_noisily(&qemu_object_opts,
484                                            optarg, true);
485             if (!opts) {
486                 goto fail;
487             }
488         }   break;
489         }
490     }
491 
492     /* Get the filename */
493     filename = (optind < argc) ? argv[optind] : NULL;
494     if (options && has_help_option(options)) {
495         g_free(options);
496         return print_block_option_help(filename, fmt);
497     }
498 
499     if (optind >= argc) {
500         error_exit("Expecting image file name");
501     }
502     optind++;
503 
504     if (qemu_opts_foreach(&qemu_object_opts,
505                           user_creatable_add_opts_foreach,
506                           NULL, &error_fatal)) {
507         goto fail;
508     }
509 
510     /* Get image size, if specified */
511     if (optind < argc) {
512         int64_t sval;
513 
514         sval = cvtnum(argv[optind++]);
515         if (sval < 0) {
516             if (sval == -ERANGE) {
517                 error_report("Image size must be less than 8 EiB!");
518             } else {
519                 error_report("Invalid image size specified! You may use k, M, "
520                       "G, T, P or E suffixes for ");
521                 error_report("kilobytes, megabytes, gigabytes, terabytes, "
522                              "petabytes and exabytes.");
523             }
524             goto fail;
525         }
526         img_size = (uint64_t)sval;
527     }
528     if (optind != argc) {
529         error_exit("Unexpected argument: %s", argv[optind]);
530     }
531 
532     bdrv_img_create(filename, fmt, base_filename, base_fmt,
533                     options, img_size, flags, quiet, &local_err);
534     if (local_err) {
535         error_reportf_err(local_err, "%s: ", filename);
536         goto fail;
537     }
538 
539     g_free(options);
540     return 0;
541 
542 fail:
543     g_free(options);
544     return 1;
545 }
546 
547 static void dump_json_image_check(ImageCheck *check, bool quiet)
548 {
549     QString *str;
550     QObject *obj;
551     Visitor *v = qobject_output_visitor_new(&obj);
552 
553     visit_type_ImageCheck(v, NULL, &check, &error_abort);
554     visit_complete(v, &obj);
555     str = qobject_to_json_pretty(obj);
556     assert(str != NULL);
557     qprintf(quiet, "%s\n", qstring_get_str(str));
558     qobject_unref(obj);
559     visit_free(v);
560     qobject_unref(str);
561 }
562 
563 static void dump_human_image_check(ImageCheck *check, bool quiet)
564 {
565     if (!(check->corruptions || check->leaks || check->check_errors)) {
566         qprintf(quiet, "No errors were found on the image.\n");
567     } else {
568         if (check->corruptions) {
569             qprintf(quiet, "\n%" PRId64 " errors were found on the image.\n"
570                     "Data may be corrupted, or further writes to the image "
571                     "may corrupt it.\n",
572                     check->corruptions);
573         }
574 
575         if (check->leaks) {
576             qprintf(quiet,
577                     "\n%" PRId64 " leaked clusters were found on the image.\n"
578                     "This means waste of disk space, but no harm to data.\n",
579                     check->leaks);
580         }
581 
582         if (check->check_errors) {
583             qprintf(quiet,
584                     "\n%" PRId64
585                     " internal errors have occurred during the check.\n",
586                     check->check_errors);
587         }
588     }
589 
590     if (check->total_clusters != 0 && check->allocated_clusters != 0) {
591         qprintf(quiet, "%" PRId64 "/%" PRId64 " = %0.2f%% allocated, "
592                 "%0.2f%% fragmented, %0.2f%% compressed clusters\n",
593                 check->allocated_clusters, check->total_clusters,
594                 check->allocated_clusters * 100.0 / check->total_clusters,
595                 check->fragmented_clusters * 100.0 / check->allocated_clusters,
596                 check->compressed_clusters * 100.0 /
597                 check->allocated_clusters);
598     }
599 
600     if (check->image_end_offset) {
601         qprintf(quiet,
602                 "Image end offset: %" PRId64 "\n", check->image_end_offset);
603     }
604 }
605 
606 static int collect_image_check(BlockDriverState *bs,
607                    ImageCheck *check,
608                    const char *filename,
609                    const char *fmt,
610                    int fix)
611 {
612     int ret;
613     BdrvCheckResult result;
614 
615     ret = bdrv_check(bs, &result, fix);
616     if (ret < 0) {
617         return ret;
618     }
619 
620     check->filename                 = g_strdup(filename);
621     check->format                   = g_strdup(bdrv_get_format_name(bs));
622     check->check_errors             = result.check_errors;
623     check->corruptions              = result.corruptions;
624     check->has_corruptions          = result.corruptions != 0;
625     check->leaks                    = result.leaks;
626     check->has_leaks                = result.leaks != 0;
627     check->corruptions_fixed        = result.corruptions_fixed;
628     check->has_corruptions_fixed    = result.corruptions != 0;
629     check->leaks_fixed              = result.leaks_fixed;
630     check->has_leaks_fixed          = result.leaks != 0;
631     check->image_end_offset         = result.image_end_offset;
632     check->has_image_end_offset     = result.image_end_offset != 0;
633     check->total_clusters           = result.bfi.total_clusters;
634     check->has_total_clusters       = result.bfi.total_clusters != 0;
635     check->allocated_clusters       = result.bfi.allocated_clusters;
636     check->has_allocated_clusters   = result.bfi.allocated_clusters != 0;
637     check->fragmented_clusters      = result.bfi.fragmented_clusters;
638     check->has_fragmented_clusters  = result.bfi.fragmented_clusters != 0;
639     check->compressed_clusters      = result.bfi.compressed_clusters;
640     check->has_compressed_clusters  = result.bfi.compressed_clusters != 0;
641 
642     return 0;
643 }
644 
645 /*
646  * Checks an image for consistency. Exit codes:
647  *
648  *  0 - Check completed, image is good
649  *  1 - Check not completed because of internal errors
650  *  2 - Check completed, image is corrupted
651  *  3 - Check completed, image has leaked clusters, but is good otherwise
652  * 63 - Checks are not supported by the image format
653  */
654 static int img_check(int argc, char **argv)
655 {
656     int c, ret;
657     OutputFormat output_format = OFORMAT_HUMAN;
658     const char *filename, *fmt, *output, *cache;
659     BlockBackend *blk;
660     BlockDriverState *bs;
661     int fix = 0;
662     int flags = BDRV_O_CHECK;
663     bool writethrough;
664     ImageCheck *check;
665     bool quiet = false;
666     bool image_opts = false;
667     bool force_share = false;
668 
669     fmt = NULL;
670     output = NULL;
671     cache = BDRV_DEFAULT_CACHE;
672 
673     for(;;) {
674         int option_index = 0;
675         static const struct option long_options[] = {
676             {"help", no_argument, 0, 'h'},
677             {"format", required_argument, 0, 'f'},
678             {"repair", required_argument, 0, 'r'},
679             {"output", required_argument, 0, OPTION_OUTPUT},
680             {"object", required_argument, 0, OPTION_OBJECT},
681             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
682             {"force-share", no_argument, 0, 'U'},
683             {0, 0, 0, 0}
684         };
685         c = getopt_long(argc, argv, ":hf:r:T:qU",
686                         long_options, &option_index);
687         if (c == -1) {
688             break;
689         }
690         switch(c) {
691         case ':':
692             missing_argument(argv[optind - 1]);
693             break;
694         case '?':
695             unrecognized_option(argv[optind - 1]);
696             break;
697         case 'h':
698             help();
699             break;
700         case 'f':
701             fmt = optarg;
702             break;
703         case 'r':
704             flags |= BDRV_O_RDWR;
705 
706             if (!strcmp(optarg, "leaks")) {
707                 fix = BDRV_FIX_LEAKS;
708             } else if (!strcmp(optarg, "all")) {
709                 fix = BDRV_FIX_LEAKS | BDRV_FIX_ERRORS;
710             } else {
711                 error_exit("Unknown option value for -r "
712                            "(expecting 'leaks' or 'all'): %s", optarg);
713             }
714             break;
715         case OPTION_OUTPUT:
716             output = optarg;
717             break;
718         case 'T':
719             cache = optarg;
720             break;
721         case 'q':
722             quiet = true;
723             break;
724         case 'U':
725             force_share = true;
726             break;
727         case OPTION_OBJECT: {
728             QemuOpts *opts;
729             opts = qemu_opts_parse_noisily(&qemu_object_opts,
730                                            optarg, true);
731             if (!opts) {
732                 return 1;
733             }
734         }   break;
735         case OPTION_IMAGE_OPTS:
736             image_opts = true;
737             break;
738         }
739     }
740     if (optind != argc - 1) {
741         error_exit("Expecting one image file name");
742     }
743     filename = argv[optind++];
744 
745     if (output && !strcmp(output, "json")) {
746         output_format = OFORMAT_JSON;
747     } else if (output && !strcmp(output, "human")) {
748         output_format = OFORMAT_HUMAN;
749     } else if (output) {
750         error_report("--output must be used with human or json as argument.");
751         return 1;
752     }
753 
754     if (qemu_opts_foreach(&qemu_object_opts,
755                           user_creatable_add_opts_foreach,
756                           NULL, &error_fatal)) {
757         return 1;
758     }
759 
760     ret = bdrv_parse_cache_mode(cache, &flags, &writethrough);
761     if (ret < 0) {
762         error_report("Invalid source cache option: %s", cache);
763         return 1;
764     }
765 
766     blk = img_open(image_opts, filename, fmt, flags, writethrough, quiet,
767                    force_share);
768     if (!blk) {
769         return 1;
770     }
771     bs = blk_bs(blk);
772 
773     check = g_new0(ImageCheck, 1);
774     ret = collect_image_check(bs, check, filename, fmt, fix);
775 
776     if (ret == -ENOTSUP) {
777         error_report("This image format does not support checks");
778         ret = 63;
779         goto fail;
780     }
781 
782     if (check->corruptions_fixed || check->leaks_fixed) {
783         int corruptions_fixed, leaks_fixed;
784 
785         leaks_fixed         = check->leaks_fixed;
786         corruptions_fixed   = check->corruptions_fixed;
787 
788         if (output_format == OFORMAT_HUMAN) {
789             qprintf(quiet,
790                     "The following inconsistencies were found and repaired:\n\n"
791                     "    %" PRId64 " leaked clusters\n"
792                     "    %" PRId64 " corruptions\n\n"
793                     "Double checking the fixed image now...\n",
794                     check->leaks_fixed,
795                     check->corruptions_fixed);
796         }
797 
798         ret = collect_image_check(bs, check, filename, fmt, 0);
799 
800         check->leaks_fixed          = leaks_fixed;
801         check->corruptions_fixed    = corruptions_fixed;
802     }
803 
804     if (!ret) {
805         switch (output_format) {
806         case OFORMAT_HUMAN:
807             dump_human_image_check(check, quiet);
808             break;
809         case OFORMAT_JSON:
810             dump_json_image_check(check, quiet);
811             break;
812         }
813     }
814 
815     if (ret || check->check_errors) {
816         if (ret) {
817             error_report("Check failed: %s", strerror(-ret));
818         } else {
819             error_report("Check failed");
820         }
821         ret = 1;
822         goto fail;
823     }
824 
825     if (check->corruptions) {
826         ret = 2;
827     } else if (check->leaks) {
828         ret = 3;
829     } else {
830         ret = 0;
831     }
832 
833 fail:
834     qapi_free_ImageCheck(check);
835     blk_unref(blk);
836     return ret;
837 }
838 
839 typedef struct CommonBlockJobCBInfo {
840     BlockDriverState *bs;
841     Error **errp;
842 } CommonBlockJobCBInfo;
843 
844 static void common_block_job_cb(void *opaque, int ret)
845 {
846     CommonBlockJobCBInfo *cbi = opaque;
847 
848     if (ret < 0) {
849         error_setg_errno(cbi->errp, -ret, "Block job failed");
850     }
851 }
852 
853 static void run_block_job(BlockJob *job, Error **errp)
854 {
855     AioContext *aio_context = blk_get_aio_context(job->blk);
856     int ret = 0;
857 
858     aio_context_acquire(aio_context);
859     job_ref(&job->job);
860     do {
861         float progress = 0.0f;
862         aio_poll(aio_context, true);
863         if (job->job.progress_total) {
864             progress = (float)job->job.progress_current /
865                        job->job.progress_total * 100.f;
866         }
867         qemu_progress_print(progress, 0);
868     } while (!job_is_ready(&job->job) && !job_is_completed(&job->job));
869 
870     if (!job_is_completed(&job->job)) {
871         ret = job_complete_sync(&job->job, errp);
872     } else {
873         ret = job->job.ret;
874     }
875     job_unref(&job->job);
876     aio_context_release(aio_context);
877 
878     /* publish completion progress only when success */
879     if (!ret) {
880         qemu_progress_print(100.f, 0);
881     }
882 }
883 
884 static int img_commit(int argc, char **argv)
885 {
886     int c, ret, flags;
887     const char *filename, *fmt, *cache, *base;
888     BlockBackend *blk;
889     BlockDriverState *bs, *base_bs;
890     BlockJob *job;
891     bool progress = false, quiet = false, drop = false;
892     bool writethrough;
893     Error *local_err = NULL;
894     CommonBlockJobCBInfo cbi;
895     bool image_opts = false;
896     AioContext *aio_context;
897 
898     fmt = NULL;
899     cache = BDRV_DEFAULT_CACHE;
900     base = NULL;
901     for(;;) {
902         static const struct option long_options[] = {
903             {"help", no_argument, 0, 'h'},
904             {"object", required_argument, 0, OPTION_OBJECT},
905             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
906             {0, 0, 0, 0}
907         };
908         c = getopt_long(argc, argv, ":f:ht:b:dpq",
909                         long_options, NULL);
910         if (c == -1) {
911             break;
912         }
913         switch(c) {
914         case ':':
915             missing_argument(argv[optind - 1]);
916             break;
917         case '?':
918             unrecognized_option(argv[optind - 1]);
919             break;
920         case 'h':
921             help();
922             break;
923         case 'f':
924             fmt = optarg;
925             break;
926         case 't':
927             cache = optarg;
928             break;
929         case 'b':
930             base = optarg;
931             /* -b implies -d */
932             drop = true;
933             break;
934         case 'd':
935             drop = true;
936             break;
937         case 'p':
938             progress = true;
939             break;
940         case 'q':
941             quiet = true;
942             break;
943         case OPTION_OBJECT: {
944             QemuOpts *opts;
945             opts = qemu_opts_parse_noisily(&qemu_object_opts,
946                                            optarg, true);
947             if (!opts) {
948                 return 1;
949             }
950         }   break;
951         case OPTION_IMAGE_OPTS:
952             image_opts = true;
953             break;
954         }
955     }
956 
957     /* Progress is not shown in Quiet mode */
958     if (quiet) {
959         progress = false;
960     }
961 
962     if (optind != argc - 1) {
963         error_exit("Expecting one image file name");
964     }
965     filename = argv[optind++];
966 
967     if (qemu_opts_foreach(&qemu_object_opts,
968                           user_creatable_add_opts_foreach,
969                           NULL, &error_fatal)) {
970         return 1;
971     }
972 
973     flags = BDRV_O_RDWR | BDRV_O_UNMAP;
974     ret = bdrv_parse_cache_mode(cache, &flags, &writethrough);
975     if (ret < 0) {
976         error_report("Invalid cache option: %s", cache);
977         return 1;
978     }
979 
980     blk = img_open(image_opts, filename, fmt, flags, writethrough, quiet,
981                    false);
982     if (!blk) {
983         return 1;
984     }
985     bs = blk_bs(blk);
986 
987     qemu_progress_init(progress, 1.f);
988     qemu_progress_print(0.f, 100);
989 
990     if (base) {
991         base_bs = bdrv_find_backing_image(bs, base);
992         if (!base_bs) {
993             error_setg(&local_err,
994                        "Did not find '%s' in the backing chain of '%s'",
995                        base, filename);
996             goto done;
997         }
998     } else {
999         /* This is different from QMP, which by default uses the deepest file in
1000          * the backing chain (i.e., the very base); however, the traditional
1001          * behavior of qemu-img commit is using the immediate backing file. */
1002         base_bs = backing_bs(bs);
1003         if (!base_bs) {
1004             error_setg(&local_err, "Image does not have a backing file");
1005             goto done;
1006         }
1007     }
1008 
1009     cbi = (CommonBlockJobCBInfo){
1010         .errp = &local_err,
1011         .bs   = bs,
1012     };
1013 
1014     aio_context = bdrv_get_aio_context(bs);
1015     aio_context_acquire(aio_context);
1016     commit_active_start("commit", bs, base_bs, JOB_DEFAULT, 0,
1017                         BLOCKDEV_ON_ERROR_REPORT, NULL, common_block_job_cb,
1018                         &cbi, false, &local_err);
1019     aio_context_release(aio_context);
1020     if (local_err) {
1021         goto done;
1022     }
1023 
1024     /* When the block job completes, the BlockBackend reference will point to
1025      * the old backing file. In order to avoid that the top image is already
1026      * deleted, so we can still empty it afterwards, increment the reference
1027      * counter here preemptively. */
1028     if (!drop) {
1029         bdrv_ref(bs);
1030     }
1031 
1032     job = block_job_get("commit");
1033     assert(job);
1034     run_block_job(job, &local_err);
1035     if (local_err) {
1036         goto unref_backing;
1037     }
1038 
1039     if (!drop && bs->drv->bdrv_make_empty) {
1040         ret = bs->drv->bdrv_make_empty(bs);
1041         if (ret) {
1042             error_setg_errno(&local_err, -ret, "Could not empty %s",
1043                              filename);
1044             goto unref_backing;
1045         }
1046     }
1047 
1048 unref_backing:
1049     if (!drop) {
1050         bdrv_unref(bs);
1051     }
1052 
1053 done:
1054     qemu_progress_end();
1055 
1056     blk_unref(blk);
1057 
1058     if (local_err) {
1059         error_report_err(local_err);
1060         return 1;
1061     }
1062 
1063     qprintf(quiet, "Image committed.\n");
1064     return 0;
1065 }
1066 
1067 /*
1068  * Returns -1 if 'buf' contains only zeroes, otherwise the byte index
1069  * of the first sector boundary within buf where the sector contains a
1070  * non-zero byte.  This function is robust to a buffer that is not
1071  * sector-aligned.
1072  */
1073 static int64_t find_nonzero(const uint8_t *buf, int64_t n)
1074 {
1075     int64_t i;
1076     int64_t end = QEMU_ALIGN_DOWN(n, BDRV_SECTOR_SIZE);
1077 
1078     for (i = 0; i < end; i += BDRV_SECTOR_SIZE) {
1079         if (!buffer_is_zero(buf + i, BDRV_SECTOR_SIZE)) {
1080             return i;
1081         }
1082     }
1083     if (i < n && !buffer_is_zero(buf + i, n - end)) {
1084         return i;
1085     }
1086     return -1;
1087 }
1088 
1089 /*
1090  * Returns true iff the first sector pointed to by 'buf' contains at least
1091  * a non-NUL byte.
1092  *
1093  * 'pnum' is set to the number of sectors (including and immediately following
1094  * the first one) that are known to be in the same allocated/unallocated state.
1095  * The function will try to align the end offset to alignment boundaries so
1096  * that the request will at least end aligned and consequtive requests will
1097  * also start at an aligned offset.
1098  */
1099 static int is_allocated_sectors(const uint8_t *buf, int n, int *pnum,
1100                                 int64_t sector_num, int alignment)
1101 {
1102     bool is_zero;
1103     int i, tail;
1104 
1105     if (n <= 0) {
1106         *pnum = 0;
1107         return 0;
1108     }
1109     is_zero = buffer_is_zero(buf, 512);
1110     for(i = 1; i < n; i++) {
1111         buf += 512;
1112         if (is_zero != buffer_is_zero(buf, 512)) {
1113             break;
1114         }
1115     }
1116 
1117     tail = (sector_num + i) & (alignment - 1);
1118     if (tail) {
1119         if (is_zero && i <= tail) {
1120             /* treat unallocated areas which only consist
1121              * of a small tail as allocated. */
1122             is_zero = false;
1123         }
1124         if (!is_zero) {
1125             /* align up end offset of allocated areas. */
1126             i += alignment - tail;
1127             i = MIN(i, n);
1128         } else {
1129             /* align down end offset of zero areas. */
1130             i -= tail;
1131         }
1132     }
1133     *pnum = i;
1134     return !is_zero;
1135 }
1136 
1137 /*
1138  * Like is_allocated_sectors, but if the buffer starts with a used sector,
1139  * up to 'min' consecutive sectors containing zeros are ignored. This avoids
1140  * breaking up write requests for only small sparse areas.
1141  */
1142 static int is_allocated_sectors_min(const uint8_t *buf, int n, int *pnum,
1143     int min, int64_t sector_num, int alignment)
1144 {
1145     int ret;
1146     int num_checked, num_used;
1147 
1148     if (n < min) {
1149         min = n;
1150     }
1151 
1152     ret = is_allocated_sectors(buf, n, pnum, sector_num, alignment);
1153     if (!ret) {
1154         return ret;
1155     }
1156 
1157     num_used = *pnum;
1158     buf += BDRV_SECTOR_SIZE * *pnum;
1159     n -= *pnum;
1160     sector_num += *pnum;
1161     num_checked = num_used;
1162 
1163     while (n > 0) {
1164         ret = is_allocated_sectors(buf, n, pnum, sector_num, alignment);
1165 
1166         buf += BDRV_SECTOR_SIZE * *pnum;
1167         n -= *pnum;
1168         sector_num += *pnum;
1169         num_checked += *pnum;
1170         if (ret) {
1171             num_used = num_checked;
1172         } else if (*pnum >= min) {
1173             break;
1174         }
1175     }
1176 
1177     *pnum = num_used;
1178     return 1;
1179 }
1180 
1181 /*
1182  * Compares two buffers sector by sector. Returns 0 if the first
1183  * sector of each buffer matches, non-zero otherwise.
1184  *
1185  * pnum is set to the sector-aligned size of the buffer prefix that
1186  * has the same matching status as the first sector.
1187  */
1188 static int compare_buffers(const uint8_t *buf1, const uint8_t *buf2,
1189                            int64_t bytes, int64_t *pnum)
1190 {
1191     bool res;
1192     int64_t i = MIN(bytes, BDRV_SECTOR_SIZE);
1193 
1194     assert(bytes > 0);
1195 
1196     res = !!memcmp(buf1, buf2, i);
1197     while (i < bytes) {
1198         int64_t len = MIN(bytes - i, BDRV_SECTOR_SIZE);
1199 
1200         if (!!memcmp(buf1 + i, buf2 + i, len) != res) {
1201             break;
1202         }
1203         i += len;
1204     }
1205 
1206     *pnum = i;
1207     return res;
1208 }
1209 
1210 #define IO_BUF_SIZE (2 * 1024 * 1024)
1211 
1212 /*
1213  * Check if passed sectors are empty (not allocated or contain only 0 bytes)
1214  *
1215  * Intended for use by 'qemu-img compare': Returns 0 in case sectors are
1216  * filled with 0, 1 if sectors contain non-zero data (this is a comparison
1217  * failure), and 4 on error (the exit status for read errors), after emitting
1218  * an error message.
1219  *
1220  * @param blk:  BlockBackend for the image
1221  * @param offset: Starting offset to check
1222  * @param bytes: Number of bytes to check
1223  * @param filename: Name of disk file we are checking (logging purpose)
1224  * @param buffer: Allocated buffer for storing read data
1225  * @param quiet: Flag for quiet mode
1226  */
1227 static int check_empty_sectors(BlockBackend *blk, int64_t offset,
1228                                int64_t bytes, const char *filename,
1229                                uint8_t *buffer, bool quiet)
1230 {
1231     int ret = 0;
1232     int64_t idx;
1233 
1234     ret = blk_pread(blk, offset, buffer, bytes);
1235     if (ret < 0) {
1236         error_report("Error while reading offset %" PRId64 " of %s: %s",
1237                      offset, filename, strerror(-ret));
1238         return 4;
1239     }
1240     idx = find_nonzero(buffer, bytes);
1241     if (idx >= 0) {
1242         qprintf(quiet, "Content mismatch at offset %" PRId64 "!\n",
1243                 offset + idx);
1244         return 1;
1245     }
1246 
1247     return 0;
1248 }
1249 
1250 /*
1251  * Compares two images. Exit codes:
1252  *
1253  * 0 - Images are identical
1254  * 1 - Images differ
1255  * >1 - Error occurred
1256  */
1257 static int img_compare(int argc, char **argv)
1258 {
1259     const char *fmt1 = NULL, *fmt2 = NULL, *cache, *filename1, *filename2;
1260     BlockBackend *blk1, *blk2;
1261     BlockDriverState *bs1, *bs2;
1262     int64_t total_size1, total_size2;
1263     uint8_t *buf1 = NULL, *buf2 = NULL;
1264     int64_t pnum1, pnum2;
1265     int allocated1, allocated2;
1266     int ret = 0; /* return value - 0 Ident, 1 Different, >1 Error */
1267     bool progress = false, quiet = false, strict = false;
1268     int flags;
1269     bool writethrough;
1270     int64_t total_size;
1271     int64_t offset = 0;
1272     int64_t chunk;
1273     int c;
1274     uint64_t progress_base;
1275     bool image_opts = false;
1276     bool force_share = false;
1277 
1278     cache = BDRV_DEFAULT_CACHE;
1279     for (;;) {
1280         static const struct option long_options[] = {
1281             {"help", no_argument, 0, 'h'},
1282             {"object", required_argument, 0, OPTION_OBJECT},
1283             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
1284             {"force-share", no_argument, 0, 'U'},
1285             {0, 0, 0, 0}
1286         };
1287         c = getopt_long(argc, argv, ":hf:F:T:pqsU",
1288                         long_options, NULL);
1289         if (c == -1) {
1290             break;
1291         }
1292         switch (c) {
1293         case ':':
1294             missing_argument(argv[optind - 1]);
1295             break;
1296         case '?':
1297             unrecognized_option(argv[optind - 1]);
1298             break;
1299         case 'h':
1300             help();
1301             break;
1302         case 'f':
1303             fmt1 = optarg;
1304             break;
1305         case 'F':
1306             fmt2 = optarg;
1307             break;
1308         case 'T':
1309             cache = optarg;
1310             break;
1311         case 'p':
1312             progress = true;
1313             break;
1314         case 'q':
1315             quiet = true;
1316             break;
1317         case 's':
1318             strict = true;
1319             break;
1320         case 'U':
1321             force_share = true;
1322             break;
1323         case OPTION_OBJECT: {
1324             QemuOpts *opts;
1325             opts = qemu_opts_parse_noisily(&qemu_object_opts,
1326                                            optarg, true);
1327             if (!opts) {
1328                 ret = 2;
1329                 goto out4;
1330             }
1331         }   break;
1332         case OPTION_IMAGE_OPTS:
1333             image_opts = true;
1334             break;
1335         }
1336     }
1337 
1338     /* Progress is not shown in Quiet mode */
1339     if (quiet) {
1340         progress = false;
1341     }
1342 
1343 
1344     if (optind != argc - 2) {
1345         error_exit("Expecting two image file names");
1346     }
1347     filename1 = argv[optind++];
1348     filename2 = argv[optind++];
1349 
1350     if (qemu_opts_foreach(&qemu_object_opts,
1351                           user_creatable_add_opts_foreach,
1352                           NULL, &error_fatal)) {
1353         ret = 2;
1354         goto out4;
1355     }
1356 
1357     /* Initialize before goto out */
1358     qemu_progress_init(progress, 2.0);
1359 
1360     flags = 0;
1361     ret = bdrv_parse_cache_mode(cache, &flags, &writethrough);
1362     if (ret < 0) {
1363         error_report("Invalid source cache option: %s", cache);
1364         ret = 2;
1365         goto out3;
1366     }
1367 
1368     blk1 = img_open(image_opts, filename1, fmt1, flags, writethrough, quiet,
1369                     force_share);
1370     if (!blk1) {
1371         ret = 2;
1372         goto out3;
1373     }
1374 
1375     blk2 = img_open(image_opts, filename2, fmt2, flags, writethrough, quiet,
1376                     force_share);
1377     if (!blk2) {
1378         ret = 2;
1379         goto out2;
1380     }
1381     bs1 = blk_bs(blk1);
1382     bs2 = blk_bs(blk2);
1383 
1384     buf1 = blk_blockalign(blk1, IO_BUF_SIZE);
1385     buf2 = blk_blockalign(blk2, IO_BUF_SIZE);
1386     total_size1 = blk_getlength(blk1);
1387     if (total_size1 < 0) {
1388         error_report("Can't get size of %s: %s",
1389                      filename1, strerror(-total_size1));
1390         ret = 4;
1391         goto out;
1392     }
1393     total_size2 = blk_getlength(blk2);
1394     if (total_size2 < 0) {
1395         error_report("Can't get size of %s: %s",
1396                      filename2, strerror(-total_size2));
1397         ret = 4;
1398         goto out;
1399     }
1400     total_size = MIN(total_size1, total_size2);
1401     progress_base = MAX(total_size1, total_size2);
1402 
1403     qemu_progress_print(0, 100);
1404 
1405     if (strict && total_size1 != total_size2) {
1406         ret = 1;
1407         qprintf(quiet, "Strict mode: Image size mismatch!\n");
1408         goto out;
1409     }
1410 
1411     while (offset < total_size) {
1412         int status1, status2;
1413 
1414         status1 = bdrv_block_status_above(bs1, NULL, offset,
1415                                           total_size1 - offset, &pnum1, NULL,
1416                                           NULL);
1417         if (status1 < 0) {
1418             ret = 3;
1419             error_report("Sector allocation test failed for %s", filename1);
1420             goto out;
1421         }
1422         allocated1 = status1 & BDRV_BLOCK_ALLOCATED;
1423 
1424         status2 = bdrv_block_status_above(bs2, NULL, offset,
1425                                           total_size2 - offset, &pnum2, NULL,
1426                                           NULL);
1427         if (status2 < 0) {
1428             ret = 3;
1429             error_report("Sector allocation test failed for %s", filename2);
1430             goto out;
1431         }
1432         allocated2 = status2 & BDRV_BLOCK_ALLOCATED;
1433 
1434         assert(pnum1 && pnum2);
1435         chunk = MIN(pnum1, pnum2);
1436 
1437         if (strict) {
1438             if (status1 != status2) {
1439                 ret = 1;
1440                 qprintf(quiet, "Strict mode: Offset %" PRId64
1441                         " block status mismatch!\n", offset);
1442                 goto out;
1443             }
1444         }
1445         if ((status1 & BDRV_BLOCK_ZERO) && (status2 & BDRV_BLOCK_ZERO)) {
1446             /* nothing to do */
1447         } else if (allocated1 == allocated2) {
1448             if (allocated1) {
1449                 int64_t pnum;
1450 
1451                 chunk = MIN(chunk, IO_BUF_SIZE);
1452                 ret = blk_pread(blk1, offset, buf1, chunk);
1453                 if (ret < 0) {
1454                     error_report("Error while reading offset %" PRId64
1455                                  " of %s: %s",
1456                                  offset, filename1, strerror(-ret));
1457                     ret = 4;
1458                     goto out;
1459                 }
1460                 ret = blk_pread(blk2, offset, buf2, chunk);
1461                 if (ret < 0) {
1462                     error_report("Error while reading offset %" PRId64
1463                                  " of %s: %s",
1464                                  offset, filename2, strerror(-ret));
1465                     ret = 4;
1466                     goto out;
1467                 }
1468                 ret = compare_buffers(buf1, buf2, chunk, &pnum);
1469                 if (ret || pnum != chunk) {
1470                     qprintf(quiet, "Content mismatch at offset %" PRId64 "!\n",
1471                             offset + (ret ? 0 : pnum));
1472                     ret = 1;
1473                     goto out;
1474                 }
1475             }
1476         } else {
1477             chunk = MIN(chunk, IO_BUF_SIZE);
1478             if (allocated1) {
1479                 ret = check_empty_sectors(blk1, offset, chunk,
1480                                           filename1, buf1, quiet);
1481             } else {
1482                 ret = check_empty_sectors(blk2, offset, chunk,
1483                                           filename2, buf1, quiet);
1484             }
1485             if (ret) {
1486                 goto out;
1487             }
1488         }
1489         offset += chunk;
1490         qemu_progress_print(((float) chunk / progress_base) * 100, 100);
1491     }
1492 
1493     if (total_size1 != total_size2) {
1494         BlockBackend *blk_over;
1495         const char *filename_over;
1496 
1497         qprintf(quiet, "Warning: Image size mismatch!\n");
1498         if (total_size1 > total_size2) {
1499             blk_over = blk1;
1500             filename_over = filename1;
1501         } else {
1502             blk_over = blk2;
1503             filename_over = filename2;
1504         }
1505 
1506         while (offset < progress_base) {
1507             ret = bdrv_block_status_above(blk_bs(blk_over), NULL, offset,
1508                                           progress_base - offset, &chunk,
1509                                           NULL, NULL);
1510             if (ret < 0) {
1511                 ret = 3;
1512                 error_report("Sector allocation test failed for %s",
1513                              filename_over);
1514                 goto out;
1515 
1516             }
1517             if (ret & BDRV_BLOCK_ALLOCATED && !(ret & BDRV_BLOCK_ZERO)) {
1518                 chunk = MIN(chunk, IO_BUF_SIZE);
1519                 ret = check_empty_sectors(blk_over, offset, chunk,
1520                                           filename_over, buf1, quiet);
1521                 if (ret) {
1522                     goto out;
1523                 }
1524             }
1525             offset += chunk;
1526             qemu_progress_print(((float) chunk / progress_base) * 100, 100);
1527         }
1528     }
1529 
1530     qprintf(quiet, "Images are identical.\n");
1531     ret = 0;
1532 
1533 out:
1534     qemu_vfree(buf1);
1535     qemu_vfree(buf2);
1536     blk_unref(blk2);
1537 out2:
1538     blk_unref(blk1);
1539 out3:
1540     qemu_progress_end();
1541 out4:
1542     return ret;
1543 }
1544 
1545 enum ImgConvertBlockStatus {
1546     BLK_DATA,
1547     BLK_ZERO,
1548     BLK_BACKING_FILE,
1549 };
1550 
1551 #define MAX_COROUTINES 16
1552 
1553 typedef struct ImgConvertState {
1554     BlockBackend **src;
1555     int64_t *src_sectors;
1556     int src_num;
1557     int64_t total_sectors;
1558     int64_t allocated_sectors;
1559     int64_t allocated_done;
1560     int64_t sector_num;
1561     int64_t wr_offs;
1562     enum ImgConvertBlockStatus status;
1563     int64_t sector_next_status;
1564     BlockBackend *target;
1565     bool has_zero_init;
1566     bool compressed;
1567     bool unallocated_blocks_are_zero;
1568     bool target_has_backing;
1569     int64_t target_backing_sectors; /* negative if unknown */
1570     bool wr_in_order;
1571     bool copy_range;
1572     int min_sparse;
1573     int alignment;
1574     size_t cluster_sectors;
1575     size_t buf_sectors;
1576     long num_coroutines;
1577     int running_coroutines;
1578     Coroutine *co[MAX_COROUTINES];
1579     int64_t wait_sector_num[MAX_COROUTINES];
1580     CoMutex lock;
1581     int ret;
1582 } ImgConvertState;
1583 
1584 static void convert_select_part(ImgConvertState *s, int64_t sector_num,
1585                                 int *src_cur, int64_t *src_cur_offset)
1586 {
1587     *src_cur = 0;
1588     *src_cur_offset = 0;
1589     while (sector_num - *src_cur_offset >= s->src_sectors[*src_cur]) {
1590         *src_cur_offset += s->src_sectors[*src_cur];
1591         (*src_cur)++;
1592         assert(*src_cur < s->src_num);
1593     }
1594 }
1595 
1596 static int convert_iteration_sectors(ImgConvertState *s, int64_t sector_num)
1597 {
1598     int64_t src_cur_offset;
1599     int ret, n, src_cur;
1600     bool post_backing_zero = false;
1601 
1602     convert_select_part(s, sector_num, &src_cur, &src_cur_offset);
1603 
1604     assert(s->total_sectors > sector_num);
1605     n = MIN(s->total_sectors - sector_num, BDRV_REQUEST_MAX_SECTORS);
1606 
1607     if (s->target_backing_sectors >= 0) {
1608         if (sector_num >= s->target_backing_sectors) {
1609             post_backing_zero = s->unallocated_blocks_are_zero;
1610         } else if (sector_num + n > s->target_backing_sectors) {
1611             /* Split requests around target_backing_sectors (because
1612              * starting from there, zeros are handled differently) */
1613             n = s->target_backing_sectors - sector_num;
1614         }
1615     }
1616 
1617     if (s->sector_next_status <= sector_num) {
1618         int64_t count = n * BDRV_SECTOR_SIZE;
1619 
1620         if (s->target_has_backing) {
1621 
1622             ret = bdrv_block_status(blk_bs(s->src[src_cur]),
1623                                     (sector_num - src_cur_offset) *
1624                                     BDRV_SECTOR_SIZE,
1625                                     count, &count, NULL, NULL);
1626         } else {
1627             ret = bdrv_block_status_above(blk_bs(s->src[src_cur]), NULL,
1628                                           (sector_num - src_cur_offset) *
1629                                           BDRV_SECTOR_SIZE,
1630                                           count, &count, NULL, NULL);
1631         }
1632         if (ret < 0) {
1633             return ret;
1634         }
1635         n = DIV_ROUND_UP(count, BDRV_SECTOR_SIZE);
1636 
1637         if (ret & BDRV_BLOCK_ZERO) {
1638             s->status = post_backing_zero ? BLK_BACKING_FILE : BLK_ZERO;
1639         } else if (ret & BDRV_BLOCK_DATA) {
1640             s->status = BLK_DATA;
1641         } else {
1642             s->status = s->target_has_backing ? BLK_BACKING_FILE : BLK_DATA;
1643         }
1644 
1645         s->sector_next_status = sector_num + n;
1646     }
1647 
1648     n = MIN(n, s->sector_next_status - sector_num);
1649     if (s->status == BLK_DATA) {
1650         n = MIN(n, s->buf_sectors);
1651     }
1652 
1653     /* We need to write complete clusters for compressed images, so if an
1654      * unallocated area is shorter than that, we must consider the whole
1655      * cluster allocated. */
1656     if (s->compressed) {
1657         if (n < s->cluster_sectors) {
1658             n = MIN(s->cluster_sectors, s->total_sectors - sector_num);
1659             s->status = BLK_DATA;
1660         } else {
1661             n = QEMU_ALIGN_DOWN(n, s->cluster_sectors);
1662         }
1663     }
1664 
1665     return n;
1666 }
1667 
1668 static int coroutine_fn convert_co_read(ImgConvertState *s, int64_t sector_num,
1669                                         int nb_sectors, uint8_t *buf)
1670 {
1671     int n, ret;
1672     QEMUIOVector qiov;
1673 
1674     assert(nb_sectors <= s->buf_sectors);
1675     while (nb_sectors > 0) {
1676         BlockBackend *blk;
1677         int src_cur;
1678         int64_t bs_sectors, src_cur_offset;
1679 
1680         /* In the case of compression with multiple source files, we can get a
1681          * nb_sectors that spreads into the next part. So we must be able to
1682          * read across multiple BDSes for one convert_read() call. */
1683         convert_select_part(s, sector_num, &src_cur, &src_cur_offset);
1684         blk = s->src[src_cur];
1685         bs_sectors = s->src_sectors[src_cur];
1686 
1687         n = MIN(nb_sectors, bs_sectors - (sector_num - src_cur_offset));
1688         qemu_iovec_init_buf(&qiov, buf, n << BDRV_SECTOR_BITS);
1689 
1690         ret = blk_co_preadv(
1691                 blk, (sector_num - src_cur_offset) << BDRV_SECTOR_BITS,
1692                 n << BDRV_SECTOR_BITS, &qiov, 0);
1693         if (ret < 0) {
1694             return ret;
1695         }
1696 
1697         sector_num += n;
1698         nb_sectors -= n;
1699         buf += n * BDRV_SECTOR_SIZE;
1700     }
1701 
1702     return 0;
1703 }
1704 
1705 
1706 static int coroutine_fn convert_co_write(ImgConvertState *s, int64_t sector_num,
1707                                          int nb_sectors, uint8_t *buf,
1708                                          enum ImgConvertBlockStatus status)
1709 {
1710     int ret;
1711     QEMUIOVector qiov;
1712 
1713     while (nb_sectors > 0) {
1714         int n = nb_sectors;
1715         BdrvRequestFlags flags = s->compressed ? BDRV_REQ_WRITE_COMPRESSED : 0;
1716 
1717         switch (status) {
1718         case BLK_BACKING_FILE:
1719             /* If we have a backing file, leave clusters unallocated that are
1720              * unallocated in the source image, so that the backing file is
1721              * visible at the respective offset. */
1722             assert(s->target_has_backing);
1723             break;
1724 
1725         case BLK_DATA:
1726             /* If we're told to keep the target fully allocated (-S 0) or there
1727              * is real non-zero data, we must write it. Otherwise we can treat
1728              * it as zero sectors.
1729              * Compressed clusters need to be written as a whole, so in that
1730              * case we can only save the write if the buffer is completely
1731              * zeroed. */
1732             if (!s->min_sparse ||
1733                 (!s->compressed &&
1734                  is_allocated_sectors_min(buf, n, &n, s->min_sparse,
1735                                           sector_num, s->alignment)) ||
1736                 (s->compressed &&
1737                  !buffer_is_zero(buf, n * BDRV_SECTOR_SIZE)))
1738             {
1739                 qemu_iovec_init_buf(&qiov, buf, n << BDRV_SECTOR_BITS);
1740 
1741                 ret = blk_co_pwritev(s->target, sector_num << BDRV_SECTOR_BITS,
1742                                      n << BDRV_SECTOR_BITS, &qiov, flags);
1743                 if (ret < 0) {
1744                     return ret;
1745                 }
1746                 break;
1747             }
1748             /* fall-through */
1749 
1750         case BLK_ZERO:
1751             if (s->has_zero_init) {
1752                 assert(!s->target_has_backing);
1753                 break;
1754             }
1755             ret = blk_co_pwrite_zeroes(s->target,
1756                                        sector_num << BDRV_SECTOR_BITS,
1757                                        n << BDRV_SECTOR_BITS, 0);
1758             if (ret < 0) {
1759                 return ret;
1760             }
1761             break;
1762         }
1763 
1764         sector_num += n;
1765         nb_sectors -= n;
1766         buf += n * BDRV_SECTOR_SIZE;
1767     }
1768 
1769     return 0;
1770 }
1771 
1772 static int coroutine_fn convert_co_copy_range(ImgConvertState *s, int64_t sector_num,
1773                                               int nb_sectors)
1774 {
1775     int n, ret;
1776 
1777     while (nb_sectors > 0) {
1778         BlockBackend *blk;
1779         int src_cur;
1780         int64_t bs_sectors, src_cur_offset;
1781         int64_t offset;
1782 
1783         convert_select_part(s, sector_num, &src_cur, &src_cur_offset);
1784         offset = (sector_num - src_cur_offset) << BDRV_SECTOR_BITS;
1785         blk = s->src[src_cur];
1786         bs_sectors = s->src_sectors[src_cur];
1787 
1788         n = MIN(nb_sectors, bs_sectors - (sector_num - src_cur_offset));
1789 
1790         ret = blk_co_copy_range(blk, offset, s->target,
1791                                 sector_num << BDRV_SECTOR_BITS,
1792                                 n << BDRV_SECTOR_BITS, 0, 0);
1793         if (ret < 0) {
1794             return ret;
1795         }
1796 
1797         sector_num += n;
1798         nb_sectors -= n;
1799     }
1800     return 0;
1801 }
1802 
1803 static void coroutine_fn convert_co_do_copy(void *opaque)
1804 {
1805     ImgConvertState *s = opaque;
1806     uint8_t *buf = NULL;
1807     int ret, i;
1808     int index = -1;
1809 
1810     for (i = 0; i < s->num_coroutines; i++) {
1811         if (s->co[i] == qemu_coroutine_self()) {
1812             index = i;
1813             break;
1814         }
1815     }
1816     assert(index >= 0);
1817 
1818     s->running_coroutines++;
1819     buf = blk_blockalign(s->target, s->buf_sectors * BDRV_SECTOR_SIZE);
1820 
1821     while (1) {
1822         int n;
1823         int64_t sector_num;
1824         enum ImgConvertBlockStatus status;
1825         bool copy_range;
1826 
1827         qemu_co_mutex_lock(&s->lock);
1828         if (s->ret != -EINPROGRESS || s->sector_num >= s->total_sectors) {
1829             qemu_co_mutex_unlock(&s->lock);
1830             break;
1831         }
1832         n = convert_iteration_sectors(s, s->sector_num);
1833         if (n < 0) {
1834             qemu_co_mutex_unlock(&s->lock);
1835             s->ret = n;
1836             break;
1837         }
1838         /* save current sector and allocation status to local variables */
1839         sector_num = s->sector_num;
1840         status = s->status;
1841         if (!s->min_sparse && s->status == BLK_ZERO) {
1842             n = MIN(n, s->buf_sectors);
1843         }
1844         /* increment global sector counter so that other coroutines can
1845          * already continue reading beyond this request */
1846         s->sector_num += n;
1847         qemu_co_mutex_unlock(&s->lock);
1848 
1849         if (status == BLK_DATA || (!s->min_sparse && status == BLK_ZERO)) {
1850             s->allocated_done += n;
1851             qemu_progress_print(100.0 * s->allocated_done /
1852                                         s->allocated_sectors, 0);
1853         }
1854 
1855 retry:
1856         copy_range = s->copy_range && s->status == BLK_DATA;
1857         if (status == BLK_DATA && !copy_range) {
1858             ret = convert_co_read(s, sector_num, n, buf);
1859             if (ret < 0) {
1860                 error_report("error while reading sector %" PRId64
1861                              ": %s", sector_num, strerror(-ret));
1862                 s->ret = ret;
1863             }
1864         } else if (!s->min_sparse && status == BLK_ZERO) {
1865             status = BLK_DATA;
1866             memset(buf, 0x00, n * BDRV_SECTOR_SIZE);
1867         }
1868 
1869         if (s->wr_in_order) {
1870             /* keep writes in order */
1871             while (s->wr_offs != sector_num && s->ret == -EINPROGRESS) {
1872                 s->wait_sector_num[index] = sector_num;
1873                 qemu_coroutine_yield();
1874             }
1875             s->wait_sector_num[index] = -1;
1876         }
1877 
1878         if (s->ret == -EINPROGRESS) {
1879             if (copy_range) {
1880                 ret = convert_co_copy_range(s, sector_num, n);
1881                 if (ret) {
1882                     s->copy_range = false;
1883                     goto retry;
1884                 }
1885             } else {
1886                 ret = convert_co_write(s, sector_num, n, buf, status);
1887             }
1888             if (ret < 0) {
1889                 error_report("error while writing sector %" PRId64
1890                              ": %s", sector_num, strerror(-ret));
1891                 s->ret = ret;
1892             }
1893         }
1894 
1895         if (s->wr_in_order) {
1896             /* reenter the coroutine that might have waited
1897              * for this write to complete */
1898             s->wr_offs = sector_num + n;
1899             for (i = 0; i < s->num_coroutines; i++) {
1900                 if (s->co[i] && s->wait_sector_num[i] == s->wr_offs) {
1901                     /*
1902                      * A -> B -> A cannot occur because A has
1903                      * s->wait_sector_num[i] == -1 during A -> B.  Therefore
1904                      * B will never enter A during this time window.
1905                      */
1906                     qemu_coroutine_enter(s->co[i]);
1907                     break;
1908                 }
1909             }
1910         }
1911     }
1912 
1913     qemu_vfree(buf);
1914     s->co[index] = NULL;
1915     s->running_coroutines--;
1916     if (!s->running_coroutines && s->ret == -EINPROGRESS) {
1917         /* the convert job finished successfully */
1918         s->ret = 0;
1919     }
1920 }
1921 
1922 static int convert_do_copy(ImgConvertState *s)
1923 {
1924     int ret, i, n;
1925     int64_t sector_num = 0;
1926 
1927     /* Check whether we have zero initialisation or can get it efficiently */
1928     s->has_zero_init = s->min_sparse && !s->target_has_backing
1929                      ? bdrv_has_zero_init(blk_bs(s->target))
1930                      : false;
1931 
1932     if (!s->has_zero_init && !s->target_has_backing &&
1933         bdrv_can_write_zeroes_with_unmap(blk_bs(s->target)))
1934     {
1935         ret = blk_make_zero(s->target, BDRV_REQ_MAY_UNMAP);
1936         if (ret == 0) {
1937             s->has_zero_init = true;
1938         }
1939     }
1940 
1941     /* Allocate buffer for copied data. For compressed images, only one cluster
1942      * can be copied at a time. */
1943     if (s->compressed) {
1944         if (s->cluster_sectors <= 0 || s->cluster_sectors > s->buf_sectors) {
1945             error_report("invalid cluster size");
1946             return -EINVAL;
1947         }
1948         s->buf_sectors = s->cluster_sectors;
1949     }
1950 
1951     while (sector_num < s->total_sectors) {
1952         n = convert_iteration_sectors(s, sector_num);
1953         if (n < 0) {
1954             return n;
1955         }
1956         if (s->status == BLK_DATA || (!s->min_sparse && s->status == BLK_ZERO))
1957         {
1958             s->allocated_sectors += n;
1959         }
1960         sector_num += n;
1961     }
1962 
1963     /* Do the copy */
1964     s->sector_next_status = 0;
1965     s->ret = -EINPROGRESS;
1966 
1967     qemu_co_mutex_init(&s->lock);
1968     for (i = 0; i < s->num_coroutines; i++) {
1969         s->co[i] = qemu_coroutine_create(convert_co_do_copy, s);
1970         s->wait_sector_num[i] = -1;
1971         qemu_coroutine_enter(s->co[i]);
1972     }
1973 
1974     while (s->running_coroutines) {
1975         main_loop_wait(false);
1976     }
1977 
1978     if (s->compressed && !s->ret) {
1979         /* signal EOF to align */
1980         ret = blk_pwrite_compressed(s->target, 0, NULL, 0);
1981         if (ret < 0) {
1982             return ret;
1983         }
1984     }
1985 
1986     return s->ret;
1987 }
1988 
1989 #define MAX_BUF_SECTORS 32768
1990 
1991 static int img_convert(int argc, char **argv)
1992 {
1993     int c, bs_i, flags, src_flags = 0;
1994     const char *fmt = NULL, *out_fmt = NULL, *cache = "unsafe",
1995                *src_cache = BDRV_DEFAULT_CACHE, *out_baseimg = NULL,
1996                *out_filename, *out_baseimg_param, *snapshot_name = NULL;
1997     BlockDriver *drv = NULL, *proto_drv = NULL;
1998     BlockDriverInfo bdi;
1999     BlockDriverState *out_bs;
2000     QemuOpts *opts = NULL, *sn_opts = NULL;
2001     QemuOptsList *create_opts = NULL;
2002     QDict *open_opts = NULL;
2003     char *options = NULL;
2004     Error *local_err = NULL;
2005     bool writethrough, src_writethrough, quiet = false, image_opts = false,
2006          skip_create = false, progress = false, tgt_image_opts = false;
2007     int64_t ret = -EINVAL;
2008     bool force_share = false;
2009     bool explict_min_sparse = false;
2010 
2011     ImgConvertState s = (ImgConvertState) {
2012         /* Need at least 4k of zeros for sparse detection */
2013         .min_sparse         = 8,
2014         .copy_range         = false,
2015         .buf_sectors        = IO_BUF_SIZE / BDRV_SECTOR_SIZE,
2016         .wr_in_order        = true,
2017         .num_coroutines     = 8,
2018     };
2019 
2020     for(;;) {
2021         static const struct option long_options[] = {
2022             {"help", no_argument, 0, 'h'},
2023             {"object", required_argument, 0, OPTION_OBJECT},
2024             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
2025             {"force-share", no_argument, 0, 'U'},
2026             {"target-image-opts", no_argument, 0, OPTION_TARGET_IMAGE_OPTS},
2027             {0, 0, 0, 0}
2028         };
2029         c = getopt_long(argc, argv, ":hf:O:B:Cco:l:S:pt:T:qnm:WU",
2030                         long_options, NULL);
2031         if (c == -1) {
2032             break;
2033         }
2034         switch(c) {
2035         case ':':
2036             missing_argument(argv[optind - 1]);
2037             break;
2038         case '?':
2039             unrecognized_option(argv[optind - 1]);
2040             break;
2041         case 'h':
2042             help();
2043             break;
2044         case 'f':
2045             fmt = optarg;
2046             break;
2047         case 'O':
2048             out_fmt = optarg;
2049             break;
2050         case 'B':
2051             out_baseimg = optarg;
2052             break;
2053         case 'C':
2054             s.copy_range = true;
2055             break;
2056         case 'c':
2057             s.compressed = true;
2058             break;
2059         case 'o':
2060             if (!is_valid_option_list(optarg)) {
2061                 error_report("Invalid option list: %s", optarg);
2062                 goto fail_getopt;
2063             }
2064             if (!options) {
2065                 options = g_strdup(optarg);
2066             } else {
2067                 char *old_options = options;
2068                 options = g_strdup_printf("%s,%s", options, optarg);
2069                 g_free(old_options);
2070             }
2071             break;
2072         case 'l':
2073             if (strstart(optarg, SNAPSHOT_OPT_BASE, NULL)) {
2074                 sn_opts = qemu_opts_parse_noisily(&internal_snapshot_opts,
2075                                                   optarg, false);
2076                 if (!sn_opts) {
2077                     error_report("Failed in parsing snapshot param '%s'",
2078                                  optarg);
2079                     goto fail_getopt;
2080                 }
2081             } else {
2082                 snapshot_name = optarg;
2083             }
2084             break;
2085         case 'S':
2086         {
2087             int64_t sval;
2088 
2089             sval = cvtnum(optarg);
2090             if (sval < 0 || sval & (BDRV_SECTOR_SIZE - 1) ||
2091                 sval / BDRV_SECTOR_SIZE > MAX_BUF_SECTORS) {
2092                 error_report("Invalid buffer size for sparse output specified. "
2093                     "Valid sizes are multiples of %llu up to %llu. Select "
2094                     "0 to disable sparse detection (fully allocates output).",
2095                     BDRV_SECTOR_SIZE, MAX_BUF_SECTORS * BDRV_SECTOR_SIZE);
2096                 goto fail_getopt;
2097             }
2098 
2099             s.min_sparse = sval / BDRV_SECTOR_SIZE;
2100             explict_min_sparse = true;
2101             break;
2102         }
2103         case 'p':
2104             progress = true;
2105             break;
2106         case 't':
2107             cache = optarg;
2108             break;
2109         case 'T':
2110             src_cache = optarg;
2111             break;
2112         case 'q':
2113             quiet = true;
2114             break;
2115         case 'n':
2116             skip_create = true;
2117             break;
2118         case 'm':
2119             if (qemu_strtol(optarg, NULL, 0, &s.num_coroutines) ||
2120                 s.num_coroutines < 1 || s.num_coroutines > MAX_COROUTINES) {
2121                 error_report("Invalid number of coroutines. Allowed number of"
2122                              " coroutines is between 1 and %d", MAX_COROUTINES);
2123                 goto fail_getopt;
2124             }
2125             break;
2126         case 'W':
2127             s.wr_in_order = false;
2128             break;
2129         case 'U':
2130             force_share = true;
2131             break;
2132         case OPTION_OBJECT: {
2133             QemuOpts *object_opts;
2134             object_opts = qemu_opts_parse_noisily(&qemu_object_opts,
2135                                                   optarg, true);
2136             if (!object_opts) {
2137                 goto fail_getopt;
2138             }
2139             break;
2140         }
2141         case OPTION_IMAGE_OPTS:
2142             image_opts = true;
2143             break;
2144         case OPTION_TARGET_IMAGE_OPTS:
2145             tgt_image_opts = true;
2146             break;
2147         }
2148     }
2149 
2150     if (!out_fmt && !tgt_image_opts) {
2151         out_fmt = "raw";
2152     }
2153 
2154     if (qemu_opts_foreach(&qemu_object_opts,
2155                           user_creatable_add_opts_foreach,
2156                           NULL, &error_fatal)) {
2157         goto fail_getopt;
2158     }
2159 
2160     if (s.compressed && s.copy_range) {
2161         error_report("Cannot enable copy offloading when -c is used");
2162         goto fail_getopt;
2163     }
2164 
2165     if (explict_min_sparse && s.copy_range) {
2166         error_report("Cannot enable copy offloading when -S is used");
2167         goto fail_getopt;
2168     }
2169 
2170     if (tgt_image_opts && !skip_create) {
2171         error_report("--target-image-opts requires use of -n flag");
2172         goto fail_getopt;
2173     }
2174 
2175     s.src_num = argc - optind - 1;
2176     out_filename = s.src_num >= 1 ? argv[argc - 1] : NULL;
2177 
2178     if (options && has_help_option(options)) {
2179         if (out_fmt) {
2180             ret = print_block_option_help(out_filename, out_fmt);
2181             goto fail_getopt;
2182         } else {
2183             error_report("Option help requires a format be specified");
2184             goto fail_getopt;
2185         }
2186     }
2187 
2188     if (s.src_num < 1) {
2189         error_report("Must specify image file name");
2190         goto fail_getopt;
2191     }
2192 
2193 
2194     /* ret is still -EINVAL until here */
2195     ret = bdrv_parse_cache_mode(src_cache, &src_flags, &src_writethrough);
2196     if (ret < 0) {
2197         error_report("Invalid source cache option: %s", src_cache);
2198         goto fail_getopt;
2199     }
2200 
2201     /* Initialize before goto out */
2202     if (quiet) {
2203         progress = false;
2204     }
2205     qemu_progress_init(progress, 1.0);
2206     qemu_progress_print(0, 100);
2207 
2208     s.src = g_new0(BlockBackend *, s.src_num);
2209     s.src_sectors = g_new(int64_t, s.src_num);
2210 
2211     for (bs_i = 0; bs_i < s.src_num; bs_i++) {
2212         s.src[bs_i] = img_open(image_opts, argv[optind + bs_i],
2213                                fmt, src_flags, src_writethrough, quiet,
2214                                force_share);
2215         if (!s.src[bs_i]) {
2216             ret = -1;
2217             goto out;
2218         }
2219         s.src_sectors[bs_i] = blk_nb_sectors(s.src[bs_i]);
2220         if (s.src_sectors[bs_i] < 0) {
2221             error_report("Could not get size of %s: %s",
2222                          argv[optind + bs_i], strerror(-s.src_sectors[bs_i]));
2223             ret = -1;
2224             goto out;
2225         }
2226         s.total_sectors += s.src_sectors[bs_i];
2227     }
2228 
2229     if (sn_opts) {
2230         bdrv_snapshot_load_tmp(blk_bs(s.src[0]),
2231                                qemu_opt_get(sn_opts, SNAPSHOT_OPT_ID),
2232                                qemu_opt_get(sn_opts, SNAPSHOT_OPT_NAME),
2233                                &local_err);
2234     } else if (snapshot_name != NULL) {
2235         if (s.src_num > 1) {
2236             error_report("No support for concatenating multiple snapshot");
2237             ret = -1;
2238             goto out;
2239         }
2240 
2241         bdrv_snapshot_load_tmp_by_id_or_name(blk_bs(s.src[0]), snapshot_name,
2242                                              &local_err);
2243     }
2244     if (local_err) {
2245         error_reportf_err(local_err, "Failed to load snapshot: ");
2246         ret = -1;
2247         goto out;
2248     }
2249 
2250     if (!skip_create) {
2251         /* Find driver and parse its options */
2252         drv = bdrv_find_format(out_fmt);
2253         if (!drv) {
2254             error_report("Unknown file format '%s'", out_fmt);
2255             ret = -1;
2256             goto out;
2257         }
2258 
2259         proto_drv = bdrv_find_protocol(out_filename, true, &local_err);
2260         if (!proto_drv) {
2261             error_report_err(local_err);
2262             ret = -1;
2263             goto out;
2264         }
2265 
2266         if (!drv->create_opts) {
2267             error_report("Format driver '%s' does not support image creation",
2268                          drv->format_name);
2269             ret = -1;
2270             goto out;
2271         }
2272 
2273         if (!proto_drv->create_opts) {
2274             error_report("Protocol driver '%s' does not support image creation",
2275                          proto_drv->format_name);
2276             ret = -1;
2277             goto out;
2278         }
2279 
2280         create_opts = qemu_opts_append(create_opts, drv->create_opts);
2281         create_opts = qemu_opts_append(create_opts, proto_drv->create_opts);
2282 
2283         opts = qemu_opts_create(create_opts, NULL, 0, &error_abort);
2284         if (options) {
2285             qemu_opts_do_parse(opts, options, NULL, &local_err);
2286             if (local_err) {
2287                 error_report_err(local_err);
2288                 ret = -1;
2289                 goto out;
2290             }
2291         }
2292 
2293         qemu_opt_set_number(opts, BLOCK_OPT_SIZE, s.total_sectors * 512,
2294                             &error_abort);
2295         ret = add_old_style_options(out_fmt, opts, out_baseimg, NULL);
2296         if (ret < 0) {
2297             goto out;
2298         }
2299     }
2300 
2301     /* Get backing file name if -o backing_file was used */
2302     out_baseimg_param = qemu_opt_get(opts, BLOCK_OPT_BACKING_FILE);
2303     if (out_baseimg_param) {
2304         out_baseimg = out_baseimg_param;
2305     }
2306     s.target_has_backing = (bool) out_baseimg;
2307 
2308     if (s.src_num > 1 && out_baseimg) {
2309         error_report("Having a backing file for the target makes no sense when "
2310                      "concatenating multiple input images");
2311         ret = -1;
2312         goto out;
2313     }
2314 
2315     /* Check if compression is supported */
2316     if (s.compressed) {
2317         bool encryption =
2318             qemu_opt_get_bool(opts, BLOCK_OPT_ENCRYPT, false);
2319         const char *encryptfmt =
2320             qemu_opt_get(opts, BLOCK_OPT_ENCRYPT_FORMAT);
2321         const char *preallocation =
2322             qemu_opt_get(opts, BLOCK_OPT_PREALLOC);
2323 
2324         if (drv && !drv->bdrv_co_pwritev_compressed) {
2325             error_report("Compression not supported for this file format");
2326             ret = -1;
2327             goto out;
2328         }
2329 
2330         if (encryption || encryptfmt) {
2331             error_report("Compression and encryption not supported at "
2332                          "the same time");
2333             ret = -1;
2334             goto out;
2335         }
2336 
2337         if (preallocation
2338             && strcmp(preallocation, "off"))
2339         {
2340             error_report("Compression and preallocation not supported at "
2341                          "the same time");
2342             ret = -1;
2343             goto out;
2344         }
2345     }
2346 
2347     /*
2348      * The later open call will need any decryption secrets, and
2349      * bdrv_create() will purge "opts", so extract them now before
2350      * they are lost.
2351      */
2352     if (!skip_create) {
2353         open_opts = qdict_new();
2354         qemu_opt_foreach(opts, img_add_key_secrets, open_opts, &error_abort);
2355     }
2356 
2357     if (!skip_create) {
2358         /* Create the new image */
2359         ret = bdrv_create(drv, out_filename, opts, &local_err);
2360         if (ret < 0) {
2361             error_reportf_err(local_err, "%s: error while converting %s: ",
2362                               out_filename, out_fmt);
2363             goto out;
2364         }
2365     }
2366 
2367     flags = s.min_sparse ? (BDRV_O_RDWR | BDRV_O_UNMAP) : BDRV_O_RDWR;
2368     ret = bdrv_parse_cache_mode(cache, &flags, &writethrough);
2369     if (ret < 0) {
2370         error_report("Invalid cache option: %s", cache);
2371         goto out;
2372     }
2373 
2374     if (skip_create) {
2375         s.target = img_open(tgt_image_opts, out_filename, out_fmt,
2376                             flags, writethrough, quiet, false);
2377     } else {
2378         /* TODO ultimately we should allow --target-image-opts
2379          * to be used even when -n is not given.
2380          * That has to wait for bdrv_create to be improved
2381          * to allow filenames in option syntax
2382          */
2383         s.target = img_open_file(out_filename, open_opts, out_fmt,
2384                                  flags, writethrough, quiet, false);
2385         open_opts = NULL; /* blk_new_open will have freed it */
2386     }
2387     if (!s.target) {
2388         ret = -1;
2389         goto out;
2390     }
2391     out_bs = blk_bs(s.target);
2392 
2393     if (s.compressed && !out_bs->drv->bdrv_co_pwritev_compressed) {
2394         error_report("Compression not supported for this file format");
2395         ret = -1;
2396         goto out;
2397     }
2398 
2399     /* increase bufsectors from the default 4096 (2M) if opt_transfer
2400      * or discard_alignment of the out_bs is greater. Limit to
2401      * MAX_BUF_SECTORS as maximum which is currently 32768 (16MB). */
2402     s.buf_sectors = MIN(MAX_BUF_SECTORS,
2403                         MAX(s.buf_sectors,
2404                             MAX(out_bs->bl.opt_transfer >> BDRV_SECTOR_BITS,
2405                                 out_bs->bl.pdiscard_alignment >>
2406                                 BDRV_SECTOR_BITS)));
2407 
2408     /* try to align the write requests to the destination to avoid unnecessary
2409      * RMW cycles. */
2410     s.alignment = MAX(pow2floor(s.min_sparse),
2411                       DIV_ROUND_UP(out_bs->bl.request_alignment,
2412                                    BDRV_SECTOR_SIZE));
2413     assert(is_power_of_2(s.alignment));
2414 
2415     if (skip_create) {
2416         int64_t output_sectors = blk_nb_sectors(s.target);
2417         if (output_sectors < 0) {
2418             error_report("unable to get output image length: %s",
2419                          strerror(-output_sectors));
2420             ret = -1;
2421             goto out;
2422         } else if (output_sectors < s.total_sectors) {
2423             error_report("output file is smaller than input file");
2424             ret = -1;
2425             goto out;
2426         }
2427     }
2428 
2429     if (s.target_has_backing) {
2430         /* Errors are treated as "backing length unknown" (which means
2431          * s.target_backing_sectors has to be negative, which it will
2432          * be automatically).  The backing file length is used only
2433          * for optimizations, so such a case is not fatal. */
2434         s.target_backing_sectors = bdrv_nb_sectors(out_bs->backing->bs);
2435     } else {
2436         s.target_backing_sectors = -1;
2437     }
2438 
2439     ret = bdrv_get_info(out_bs, &bdi);
2440     if (ret < 0) {
2441         if (s.compressed) {
2442             error_report("could not get block driver info");
2443             goto out;
2444         }
2445     } else {
2446         s.compressed = s.compressed || bdi.needs_compressed_writes;
2447         s.cluster_sectors = bdi.cluster_size / BDRV_SECTOR_SIZE;
2448         s.unallocated_blocks_are_zero = bdi.unallocated_blocks_are_zero;
2449     }
2450 
2451     ret = convert_do_copy(&s);
2452 out:
2453     if (!ret) {
2454         qemu_progress_print(100, 0);
2455     }
2456     qemu_progress_end();
2457     qemu_opts_del(opts);
2458     qemu_opts_free(create_opts);
2459     qemu_opts_del(sn_opts);
2460     qobject_unref(open_opts);
2461     blk_unref(s.target);
2462     if (s.src) {
2463         for (bs_i = 0; bs_i < s.src_num; bs_i++) {
2464             blk_unref(s.src[bs_i]);
2465         }
2466         g_free(s.src);
2467     }
2468     g_free(s.src_sectors);
2469 fail_getopt:
2470     g_free(options);
2471 
2472     return !!ret;
2473 }
2474 
2475 
2476 static void dump_snapshots(BlockDriverState *bs)
2477 {
2478     QEMUSnapshotInfo *sn_tab, *sn;
2479     int nb_sns, i;
2480 
2481     nb_sns = bdrv_snapshot_list(bs, &sn_tab);
2482     if (nb_sns <= 0)
2483         return;
2484     printf("Snapshot list:\n");
2485     bdrv_snapshot_dump(fprintf, stdout, NULL);
2486     printf("\n");
2487     for(i = 0; i < nb_sns; i++) {
2488         sn = &sn_tab[i];
2489         bdrv_snapshot_dump(fprintf, stdout, sn);
2490         printf("\n");
2491     }
2492     g_free(sn_tab);
2493 }
2494 
2495 static void dump_json_image_info_list(ImageInfoList *list)
2496 {
2497     QString *str;
2498     QObject *obj;
2499     Visitor *v = qobject_output_visitor_new(&obj);
2500 
2501     visit_type_ImageInfoList(v, NULL, &list, &error_abort);
2502     visit_complete(v, &obj);
2503     str = qobject_to_json_pretty(obj);
2504     assert(str != NULL);
2505     printf("%s\n", qstring_get_str(str));
2506     qobject_unref(obj);
2507     visit_free(v);
2508     qobject_unref(str);
2509 }
2510 
2511 static void dump_json_image_info(ImageInfo *info)
2512 {
2513     QString *str;
2514     QObject *obj;
2515     Visitor *v = qobject_output_visitor_new(&obj);
2516 
2517     visit_type_ImageInfo(v, NULL, &info, &error_abort);
2518     visit_complete(v, &obj);
2519     str = qobject_to_json_pretty(obj);
2520     assert(str != NULL);
2521     printf("%s\n", qstring_get_str(str));
2522     qobject_unref(obj);
2523     visit_free(v);
2524     qobject_unref(str);
2525 }
2526 
2527 static void dump_human_image_info_list(ImageInfoList *list)
2528 {
2529     ImageInfoList *elem;
2530     bool delim = false;
2531 
2532     for (elem = list; elem; elem = elem->next) {
2533         if (delim) {
2534             printf("\n");
2535         }
2536         delim = true;
2537 
2538         bdrv_image_info_dump(fprintf, stdout, elem->value);
2539     }
2540 }
2541 
2542 static gboolean str_equal_func(gconstpointer a, gconstpointer b)
2543 {
2544     return strcmp(a, b) == 0;
2545 }
2546 
2547 /**
2548  * Open an image file chain and return an ImageInfoList
2549  *
2550  * @filename: topmost image filename
2551  * @fmt: topmost image format (may be NULL to autodetect)
2552  * @chain: true  - enumerate entire backing file chain
2553  *         false - only topmost image file
2554  *
2555  * Returns a list of ImageInfo objects or NULL if there was an error opening an
2556  * image file.  If there was an error a message will have been printed to
2557  * stderr.
2558  */
2559 static ImageInfoList *collect_image_info_list(bool image_opts,
2560                                               const char *filename,
2561                                               const char *fmt,
2562                                               bool chain, bool force_share)
2563 {
2564     ImageInfoList *head = NULL;
2565     ImageInfoList **last = &head;
2566     GHashTable *filenames;
2567     Error *err = NULL;
2568 
2569     filenames = g_hash_table_new_full(g_str_hash, str_equal_func, NULL, NULL);
2570 
2571     while (filename) {
2572         BlockBackend *blk;
2573         BlockDriverState *bs;
2574         ImageInfo *info;
2575         ImageInfoList *elem;
2576 
2577         if (g_hash_table_lookup_extended(filenames, filename, NULL, NULL)) {
2578             error_report("Backing file '%s' creates an infinite loop.",
2579                          filename);
2580             goto err;
2581         }
2582         g_hash_table_insert(filenames, (gpointer)filename, NULL);
2583 
2584         blk = img_open(image_opts, filename, fmt,
2585                        BDRV_O_NO_BACKING | BDRV_O_NO_IO, false, false,
2586                        force_share);
2587         if (!blk) {
2588             goto err;
2589         }
2590         bs = blk_bs(blk);
2591 
2592         bdrv_query_image_info(bs, &info, &err);
2593         if (err) {
2594             error_report_err(err);
2595             blk_unref(blk);
2596             goto err;
2597         }
2598 
2599         elem = g_new0(ImageInfoList, 1);
2600         elem->value = info;
2601         *last = elem;
2602         last = &elem->next;
2603 
2604         blk_unref(blk);
2605 
2606         filename = fmt = NULL;
2607         if (chain) {
2608             if (info->has_full_backing_filename) {
2609                 filename = info->full_backing_filename;
2610             } else if (info->has_backing_filename) {
2611                 error_report("Could not determine absolute backing filename,"
2612                              " but backing filename '%s' present",
2613                              info->backing_filename);
2614                 goto err;
2615             }
2616             if (info->has_backing_filename_format) {
2617                 fmt = info->backing_filename_format;
2618             }
2619         }
2620     }
2621     g_hash_table_destroy(filenames);
2622     return head;
2623 
2624 err:
2625     qapi_free_ImageInfoList(head);
2626     g_hash_table_destroy(filenames);
2627     return NULL;
2628 }
2629 
2630 static int img_info(int argc, char **argv)
2631 {
2632     int c;
2633     OutputFormat output_format = OFORMAT_HUMAN;
2634     bool chain = false;
2635     const char *filename, *fmt, *output;
2636     ImageInfoList *list;
2637     bool image_opts = false;
2638     bool force_share = false;
2639 
2640     fmt = NULL;
2641     output = NULL;
2642     for(;;) {
2643         int option_index = 0;
2644         static const struct option long_options[] = {
2645             {"help", no_argument, 0, 'h'},
2646             {"format", required_argument, 0, 'f'},
2647             {"output", required_argument, 0, OPTION_OUTPUT},
2648             {"backing-chain", no_argument, 0, OPTION_BACKING_CHAIN},
2649             {"object", required_argument, 0, OPTION_OBJECT},
2650             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
2651             {"force-share", no_argument, 0, 'U'},
2652             {0, 0, 0, 0}
2653         };
2654         c = getopt_long(argc, argv, ":f:hU",
2655                         long_options, &option_index);
2656         if (c == -1) {
2657             break;
2658         }
2659         switch(c) {
2660         case ':':
2661             missing_argument(argv[optind - 1]);
2662             break;
2663         case '?':
2664             unrecognized_option(argv[optind - 1]);
2665             break;
2666         case 'h':
2667             help();
2668             break;
2669         case 'f':
2670             fmt = optarg;
2671             break;
2672         case 'U':
2673             force_share = true;
2674             break;
2675         case OPTION_OUTPUT:
2676             output = optarg;
2677             break;
2678         case OPTION_BACKING_CHAIN:
2679             chain = true;
2680             break;
2681         case OPTION_OBJECT: {
2682             QemuOpts *opts;
2683             opts = qemu_opts_parse_noisily(&qemu_object_opts,
2684                                            optarg, true);
2685             if (!opts) {
2686                 return 1;
2687             }
2688         }   break;
2689         case OPTION_IMAGE_OPTS:
2690             image_opts = true;
2691             break;
2692         }
2693     }
2694     if (optind != argc - 1) {
2695         error_exit("Expecting one image file name");
2696     }
2697     filename = argv[optind++];
2698 
2699     if (output && !strcmp(output, "json")) {
2700         output_format = OFORMAT_JSON;
2701     } else if (output && !strcmp(output, "human")) {
2702         output_format = OFORMAT_HUMAN;
2703     } else if (output) {
2704         error_report("--output must be used with human or json as argument.");
2705         return 1;
2706     }
2707 
2708     if (qemu_opts_foreach(&qemu_object_opts,
2709                           user_creatable_add_opts_foreach,
2710                           NULL, &error_fatal)) {
2711         return 1;
2712     }
2713 
2714     list = collect_image_info_list(image_opts, filename, fmt, chain,
2715                                    force_share);
2716     if (!list) {
2717         return 1;
2718     }
2719 
2720     switch (output_format) {
2721     case OFORMAT_HUMAN:
2722         dump_human_image_info_list(list);
2723         break;
2724     case OFORMAT_JSON:
2725         if (chain) {
2726             dump_json_image_info_list(list);
2727         } else {
2728             dump_json_image_info(list->value);
2729         }
2730         break;
2731     }
2732 
2733     qapi_free_ImageInfoList(list);
2734     return 0;
2735 }
2736 
2737 static void dump_map_entry(OutputFormat output_format, MapEntry *e,
2738                            MapEntry *next)
2739 {
2740     switch (output_format) {
2741     case OFORMAT_HUMAN:
2742         if (e->data && !e->has_offset) {
2743             error_report("File contains external, encrypted or compressed clusters.");
2744             exit(1);
2745         }
2746         if (e->data && !e->zero) {
2747             printf("%#-16"PRIx64"%#-16"PRIx64"%#-16"PRIx64"%s\n",
2748                    e->start, e->length,
2749                    e->has_offset ? e->offset : 0,
2750                    e->has_filename ? e->filename : "");
2751         }
2752         /* This format ignores the distinction between 0, ZERO and ZERO|DATA.
2753          * Modify the flags here to allow more coalescing.
2754          */
2755         if (next && (!next->data || next->zero)) {
2756             next->data = false;
2757             next->zero = true;
2758         }
2759         break;
2760     case OFORMAT_JSON:
2761         printf("%s{ \"start\": %"PRId64", \"length\": %"PRId64","
2762                " \"depth\": %"PRId64", \"zero\": %s, \"data\": %s",
2763                (e->start == 0 ? "[" : ",\n"),
2764                e->start, e->length, e->depth,
2765                e->zero ? "true" : "false",
2766                e->data ? "true" : "false");
2767         if (e->has_offset) {
2768             printf(", \"offset\": %"PRId64"", e->offset);
2769         }
2770         putchar('}');
2771 
2772         if (!next) {
2773             printf("]\n");
2774         }
2775         break;
2776     }
2777 }
2778 
2779 static int get_block_status(BlockDriverState *bs, int64_t offset,
2780                             int64_t bytes, MapEntry *e)
2781 {
2782     int ret;
2783     int depth;
2784     BlockDriverState *file;
2785     bool has_offset;
2786     int64_t map;
2787     char *filename = NULL;
2788 
2789     /* As an optimization, we could cache the current range of unallocated
2790      * clusters in each file of the chain, and avoid querying the same
2791      * range repeatedly.
2792      */
2793 
2794     depth = 0;
2795     for (;;) {
2796         ret = bdrv_block_status(bs, offset, bytes, &bytes, &map, &file);
2797         if (ret < 0) {
2798             return ret;
2799         }
2800         assert(bytes);
2801         if (ret & (BDRV_BLOCK_ZERO|BDRV_BLOCK_DATA)) {
2802             break;
2803         }
2804         bs = backing_bs(bs);
2805         if (bs == NULL) {
2806             ret = 0;
2807             break;
2808         }
2809 
2810         depth++;
2811     }
2812 
2813     has_offset = !!(ret & BDRV_BLOCK_OFFSET_VALID);
2814 
2815     if (file && has_offset) {
2816         bdrv_refresh_filename(file);
2817         filename = file->filename;
2818     }
2819 
2820     *e = (MapEntry) {
2821         .start = offset,
2822         .length = bytes,
2823         .data = !!(ret & BDRV_BLOCK_DATA),
2824         .zero = !!(ret & BDRV_BLOCK_ZERO),
2825         .offset = map,
2826         .has_offset = has_offset,
2827         .depth = depth,
2828         .has_filename = filename,
2829         .filename = filename,
2830     };
2831 
2832     return 0;
2833 }
2834 
2835 static inline bool entry_mergeable(const MapEntry *curr, const MapEntry *next)
2836 {
2837     if (curr->length == 0) {
2838         return false;
2839     }
2840     if (curr->zero != next->zero ||
2841         curr->data != next->data ||
2842         curr->depth != next->depth ||
2843         curr->has_filename != next->has_filename ||
2844         curr->has_offset != next->has_offset) {
2845         return false;
2846     }
2847     if (curr->has_filename && strcmp(curr->filename, next->filename)) {
2848         return false;
2849     }
2850     if (curr->has_offset && curr->offset + curr->length != next->offset) {
2851         return false;
2852     }
2853     return true;
2854 }
2855 
2856 static int img_map(int argc, char **argv)
2857 {
2858     int c;
2859     OutputFormat output_format = OFORMAT_HUMAN;
2860     BlockBackend *blk;
2861     BlockDriverState *bs;
2862     const char *filename, *fmt, *output;
2863     int64_t length;
2864     MapEntry curr = { .length = 0 }, next;
2865     int ret = 0;
2866     bool image_opts = false;
2867     bool force_share = false;
2868 
2869     fmt = NULL;
2870     output = NULL;
2871     for (;;) {
2872         int option_index = 0;
2873         static const struct option long_options[] = {
2874             {"help", no_argument, 0, 'h'},
2875             {"format", required_argument, 0, 'f'},
2876             {"output", required_argument, 0, OPTION_OUTPUT},
2877             {"object", required_argument, 0, OPTION_OBJECT},
2878             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
2879             {"force-share", no_argument, 0, 'U'},
2880             {0, 0, 0, 0}
2881         };
2882         c = getopt_long(argc, argv, ":f:hU",
2883                         long_options, &option_index);
2884         if (c == -1) {
2885             break;
2886         }
2887         switch (c) {
2888         case ':':
2889             missing_argument(argv[optind - 1]);
2890             break;
2891         case '?':
2892             unrecognized_option(argv[optind - 1]);
2893             break;
2894         case 'h':
2895             help();
2896             break;
2897         case 'f':
2898             fmt = optarg;
2899             break;
2900         case 'U':
2901             force_share = true;
2902             break;
2903         case OPTION_OUTPUT:
2904             output = optarg;
2905             break;
2906         case OPTION_OBJECT: {
2907             QemuOpts *opts;
2908             opts = qemu_opts_parse_noisily(&qemu_object_opts,
2909                                            optarg, true);
2910             if (!opts) {
2911                 return 1;
2912             }
2913         }   break;
2914         case OPTION_IMAGE_OPTS:
2915             image_opts = true;
2916             break;
2917         }
2918     }
2919     if (optind != argc - 1) {
2920         error_exit("Expecting one image file name");
2921     }
2922     filename = argv[optind];
2923 
2924     if (output && !strcmp(output, "json")) {
2925         output_format = OFORMAT_JSON;
2926     } else if (output && !strcmp(output, "human")) {
2927         output_format = OFORMAT_HUMAN;
2928     } else if (output) {
2929         error_report("--output must be used with human or json as argument.");
2930         return 1;
2931     }
2932 
2933     if (qemu_opts_foreach(&qemu_object_opts,
2934                           user_creatable_add_opts_foreach,
2935                           NULL, &error_fatal)) {
2936         return 1;
2937     }
2938 
2939     blk = img_open(image_opts, filename, fmt, 0, false, false, force_share);
2940     if (!blk) {
2941         return 1;
2942     }
2943     bs = blk_bs(blk);
2944 
2945     if (output_format == OFORMAT_HUMAN) {
2946         printf("%-16s%-16s%-16s%s\n", "Offset", "Length", "Mapped to", "File");
2947     }
2948 
2949     length = blk_getlength(blk);
2950     while (curr.start + curr.length < length) {
2951         int64_t offset = curr.start + curr.length;
2952         int64_t n;
2953 
2954         /* Probe up to 1 GiB at a time.  */
2955         n = MIN(1 << 30, length - offset);
2956         ret = get_block_status(bs, offset, n, &next);
2957 
2958         if (ret < 0) {
2959             error_report("Could not read file metadata: %s", strerror(-ret));
2960             goto out;
2961         }
2962 
2963         if (entry_mergeable(&curr, &next)) {
2964             curr.length += next.length;
2965             continue;
2966         }
2967 
2968         if (curr.length > 0) {
2969             dump_map_entry(output_format, &curr, &next);
2970         }
2971         curr = next;
2972     }
2973 
2974     dump_map_entry(output_format, &curr, NULL);
2975 
2976 out:
2977     blk_unref(blk);
2978     return ret < 0;
2979 }
2980 
2981 #define SNAPSHOT_LIST   1
2982 #define SNAPSHOT_CREATE 2
2983 #define SNAPSHOT_APPLY  3
2984 #define SNAPSHOT_DELETE 4
2985 
2986 static int img_snapshot(int argc, char **argv)
2987 {
2988     BlockBackend *blk;
2989     BlockDriverState *bs;
2990     QEMUSnapshotInfo sn;
2991     char *filename, *snapshot_name = NULL;
2992     int c, ret = 0, bdrv_oflags;
2993     int action = 0;
2994     qemu_timeval tv;
2995     bool quiet = false;
2996     Error *err = NULL;
2997     bool image_opts = false;
2998     bool force_share = false;
2999 
3000     bdrv_oflags = BDRV_O_RDWR;
3001     /* Parse commandline parameters */
3002     for(;;) {
3003         static const struct option long_options[] = {
3004             {"help", no_argument, 0, 'h'},
3005             {"object", required_argument, 0, OPTION_OBJECT},
3006             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
3007             {"force-share", no_argument, 0, 'U'},
3008             {0, 0, 0, 0}
3009         };
3010         c = getopt_long(argc, argv, ":la:c:d:hqU",
3011                         long_options, NULL);
3012         if (c == -1) {
3013             break;
3014         }
3015         switch(c) {
3016         case ':':
3017             missing_argument(argv[optind - 1]);
3018             break;
3019         case '?':
3020             unrecognized_option(argv[optind - 1]);
3021             break;
3022         case 'h':
3023             help();
3024             return 0;
3025         case 'l':
3026             if (action) {
3027                 error_exit("Cannot mix '-l', '-a', '-c', '-d'");
3028                 return 0;
3029             }
3030             action = SNAPSHOT_LIST;
3031             bdrv_oflags &= ~BDRV_O_RDWR; /* no need for RW */
3032             break;
3033         case 'a':
3034             if (action) {
3035                 error_exit("Cannot mix '-l', '-a', '-c', '-d'");
3036                 return 0;
3037             }
3038             action = SNAPSHOT_APPLY;
3039             snapshot_name = optarg;
3040             break;
3041         case 'c':
3042             if (action) {
3043                 error_exit("Cannot mix '-l', '-a', '-c', '-d'");
3044                 return 0;
3045             }
3046             action = SNAPSHOT_CREATE;
3047             snapshot_name = optarg;
3048             break;
3049         case 'd':
3050             if (action) {
3051                 error_exit("Cannot mix '-l', '-a', '-c', '-d'");
3052                 return 0;
3053             }
3054             action = SNAPSHOT_DELETE;
3055             snapshot_name = optarg;
3056             break;
3057         case 'q':
3058             quiet = true;
3059             break;
3060         case 'U':
3061             force_share = true;
3062             break;
3063         case OPTION_OBJECT: {
3064             QemuOpts *opts;
3065             opts = qemu_opts_parse_noisily(&qemu_object_opts,
3066                                            optarg, true);
3067             if (!opts) {
3068                 return 1;
3069             }
3070         }   break;
3071         case OPTION_IMAGE_OPTS:
3072             image_opts = true;
3073             break;
3074         }
3075     }
3076 
3077     if (optind != argc - 1) {
3078         error_exit("Expecting one image file name");
3079     }
3080     filename = argv[optind++];
3081 
3082     if (qemu_opts_foreach(&qemu_object_opts,
3083                           user_creatable_add_opts_foreach,
3084                           NULL, &error_fatal)) {
3085         return 1;
3086     }
3087 
3088     /* Open the image */
3089     blk = img_open(image_opts, filename, NULL, bdrv_oflags, false, quiet,
3090                    force_share);
3091     if (!blk) {
3092         return 1;
3093     }
3094     bs = blk_bs(blk);
3095 
3096     /* Perform the requested action */
3097     switch(action) {
3098     case SNAPSHOT_LIST:
3099         dump_snapshots(bs);
3100         break;
3101 
3102     case SNAPSHOT_CREATE:
3103         memset(&sn, 0, sizeof(sn));
3104         pstrcpy(sn.name, sizeof(sn.name), snapshot_name);
3105 
3106         qemu_gettimeofday(&tv);
3107         sn.date_sec = tv.tv_sec;
3108         sn.date_nsec = tv.tv_usec * 1000;
3109 
3110         ret = bdrv_snapshot_create(bs, &sn);
3111         if (ret) {
3112             error_report("Could not create snapshot '%s': %d (%s)",
3113                 snapshot_name, ret, strerror(-ret));
3114         }
3115         break;
3116 
3117     case SNAPSHOT_APPLY:
3118         ret = bdrv_snapshot_goto(bs, snapshot_name, &err);
3119         if (ret) {
3120             error_reportf_err(err, "Could not apply snapshot '%s': ",
3121                               snapshot_name);
3122         }
3123         break;
3124 
3125     case SNAPSHOT_DELETE:
3126         ret = bdrv_snapshot_find(bs, &sn, snapshot_name);
3127         if (ret < 0) {
3128             error_report("Could not delete snapshot '%s': snapshot not "
3129                          "found", snapshot_name);
3130             ret = 1;
3131         } else {
3132             ret = bdrv_snapshot_delete(bs, sn.id_str, sn.name, &err);
3133             if (ret < 0) {
3134                 error_reportf_err(err, "Could not delete snapshot '%s': ",
3135                                   snapshot_name);
3136                 ret = 1;
3137             }
3138         }
3139         break;
3140     }
3141 
3142     /* Cleanup */
3143     blk_unref(blk);
3144     if (ret) {
3145         return 1;
3146     }
3147     return 0;
3148 }
3149 
3150 static int img_rebase(int argc, char **argv)
3151 {
3152     BlockBackend *blk = NULL, *blk_old_backing = NULL, *blk_new_backing = NULL;
3153     uint8_t *buf_old = NULL;
3154     uint8_t *buf_new = NULL;
3155     BlockDriverState *bs = NULL;
3156     char *filename;
3157     const char *fmt, *cache, *src_cache, *out_basefmt, *out_baseimg;
3158     int c, flags, src_flags, ret;
3159     bool writethrough, src_writethrough;
3160     int unsafe = 0;
3161     bool force_share = false;
3162     int progress = 0;
3163     bool quiet = false;
3164     Error *local_err = NULL;
3165     bool image_opts = false;
3166 
3167     /* Parse commandline parameters */
3168     fmt = NULL;
3169     cache = BDRV_DEFAULT_CACHE;
3170     src_cache = BDRV_DEFAULT_CACHE;
3171     out_baseimg = NULL;
3172     out_basefmt = NULL;
3173     for(;;) {
3174         static const struct option long_options[] = {
3175             {"help", no_argument, 0, 'h'},
3176             {"object", required_argument, 0, OPTION_OBJECT},
3177             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
3178             {"force-share", no_argument, 0, 'U'},
3179             {0, 0, 0, 0}
3180         };
3181         c = getopt_long(argc, argv, ":hf:F:b:upt:T:qU",
3182                         long_options, NULL);
3183         if (c == -1) {
3184             break;
3185         }
3186         switch(c) {
3187         case ':':
3188             missing_argument(argv[optind - 1]);
3189             break;
3190         case '?':
3191             unrecognized_option(argv[optind - 1]);
3192             break;
3193         case 'h':
3194             help();
3195             return 0;
3196         case 'f':
3197             fmt = optarg;
3198             break;
3199         case 'F':
3200             out_basefmt = optarg;
3201             break;
3202         case 'b':
3203             out_baseimg = optarg;
3204             break;
3205         case 'u':
3206             unsafe = 1;
3207             break;
3208         case 'p':
3209             progress = 1;
3210             break;
3211         case 't':
3212             cache = optarg;
3213             break;
3214         case 'T':
3215             src_cache = optarg;
3216             break;
3217         case 'q':
3218             quiet = true;
3219             break;
3220         case OPTION_OBJECT: {
3221             QemuOpts *opts;
3222             opts = qemu_opts_parse_noisily(&qemu_object_opts,
3223                                            optarg, true);
3224             if (!opts) {
3225                 return 1;
3226             }
3227         }   break;
3228         case OPTION_IMAGE_OPTS:
3229             image_opts = true;
3230             break;
3231         case 'U':
3232             force_share = true;
3233             break;
3234         }
3235     }
3236 
3237     if (quiet) {
3238         progress = 0;
3239     }
3240 
3241     if (optind != argc - 1) {
3242         error_exit("Expecting one image file name");
3243     }
3244     if (!unsafe && !out_baseimg) {
3245         error_exit("Must specify backing file (-b) or use unsafe mode (-u)");
3246     }
3247     filename = argv[optind++];
3248 
3249     if (qemu_opts_foreach(&qemu_object_opts,
3250                           user_creatable_add_opts_foreach,
3251                           NULL, &error_fatal)) {
3252         return 1;
3253     }
3254 
3255     qemu_progress_init(progress, 2.0);
3256     qemu_progress_print(0, 100);
3257 
3258     flags = BDRV_O_RDWR | (unsafe ? BDRV_O_NO_BACKING : 0);
3259     ret = bdrv_parse_cache_mode(cache, &flags, &writethrough);
3260     if (ret < 0) {
3261         error_report("Invalid cache option: %s", cache);
3262         goto out;
3263     }
3264 
3265     src_flags = 0;
3266     ret = bdrv_parse_cache_mode(src_cache, &src_flags, &src_writethrough);
3267     if (ret < 0) {
3268         error_report("Invalid source cache option: %s", src_cache);
3269         goto out;
3270     }
3271 
3272     /* The source files are opened read-only, don't care about WCE */
3273     assert((src_flags & BDRV_O_RDWR) == 0);
3274     (void) src_writethrough;
3275 
3276     /*
3277      * Open the images.
3278      *
3279      * Ignore the old backing file for unsafe rebase in case we want to correct
3280      * the reference to a renamed or moved backing file.
3281      */
3282     blk = img_open(image_opts, filename, fmt, flags, writethrough, quiet,
3283                    false);
3284     if (!blk) {
3285         ret = -1;
3286         goto out;
3287     }
3288     bs = blk_bs(blk);
3289 
3290     if (out_basefmt != NULL) {
3291         if (bdrv_find_format(out_basefmt) == NULL) {
3292             error_report("Invalid format name: '%s'", out_basefmt);
3293             ret = -1;
3294             goto out;
3295         }
3296     }
3297 
3298     /* For safe rebasing we need to compare old and new backing file */
3299     if (!unsafe) {
3300         char backing_name[PATH_MAX];
3301         QDict *options = NULL;
3302 
3303         if (bs->backing_format[0] != '\0') {
3304             options = qdict_new();
3305             qdict_put_str(options, "driver", bs->backing_format);
3306         }
3307 
3308         if (force_share) {
3309             if (!options) {
3310                 options = qdict_new();
3311             }
3312             qdict_put_bool(options, BDRV_OPT_FORCE_SHARE, true);
3313         }
3314         bdrv_get_backing_filename(bs, backing_name, sizeof(backing_name));
3315         blk_old_backing = blk_new_open(backing_name, NULL,
3316                                        options, src_flags, &local_err);
3317         if (!blk_old_backing) {
3318             error_reportf_err(local_err,
3319                               "Could not open old backing file '%s': ",
3320                               backing_name);
3321             ret = -1;
3322             goto out;
3323         }
3324 
3325         if (out_baseimg[0]) {
3326             const char *overlay_filename;
3327             char *out_real_path;
3328 
3329             options = qdict_new();
3330             if (out_basefmt) {
3331                 qdict_put_str(options, "driver", out_basefmt);
3332             }
3333             if (force_share) {
3334                 qdict_put_bool(options, BDRV_OPT_FORCE_SHARE, true);
3335             }
3336 
3337             bdrv_refresh_filename(bs);
3338             overlay_filename = bs->exact_filename[0] ? bs->exact_filename
3339                                                      : bs->filename;
3340             out_real_path =
3341                 bdrv_get_full_backing_filename_from_filename(overlay_filename,
3342                                                              out_baseimg,
3343                                                              &local_err);
3344             if (local_err) {
3345                 error_reportf_err(local_err,
3346                                   "Could not resolve backing filename: ");
3347                 ret = -1;
3348                 goto out;
3349             }
3350 
3351             blk_new_backing = blk_new_open(out_real_path, NULL,
3352                                            options, src_flags, &local_err);
3353             g_free(out_real_path);
3354             if (!blk_new_backing) {
3355                 error_reportf_err(local_err,
3356                                   "Could not open new backing file '%s': ",
3357                                   out_baseimg);
3358                 ret = -1;
3359                 goto out;
3360             }
3361         }
3362     }
3363 
3364     /*
3365      * Check each unallocated cluster in the COW file. If it is unallocated,
3366      * accesses go to the backing file. We must therefore compare this cluster
3367      * in the old and new backing file, and if they differ we need to copy it
3368      * from the old backing file into the COW file.
3369      *
3370      * If qemu-img crashes during this step, no harm is done. The content of
3371      * the image is the same as the original one at any time.
3372      */
3373     if (!unsafe) {
3374         int64_t size;
3375         int64_t old_backing_size;
3376         int64_t new_backing_size = 0;
3377         uint64_t offset;
3378         int64_t n;
3379         float local_progress = 0;
3380 
3381         buf_old = blk_blockalign(blk, IO_BUF_SIZE);
3382         buf_new = blk_blockalign(blk, IO_BUF_SIZE);
3383 
3384         size = blk_getlength(blk);
3385         if (size < 0) {
3386             error_report("Could not get size of '%s': %s",
3387                          filename, strerror(-size));
3388             ret = -1;
3389             goto out;
3390         }
3391         old_backing_size = blk_getlength(blk_old_backing);
3392         if (old_backing_size < 0) {
3393             char backing_name[PATH_MAX];
3394 
3395             bdrv_get_backing_filename(bs, backing_name, sizeof(backing_name));
3396             error_report("Could not get size of '%s': %s",
3397                          backing_name, strerror(-old_backing_size));
3398             ret = -1;
3399             goto out;
3400         }
3401         if (blk_new_backing) {
3402             new_backing_size = blk_getlength(blk_new_backing);
3403             if (new_backing_size < 0) {
3404                 error_report("Could not get size of '%s': %s",
3405                              out_baseimg, strerror(-new_backing_size));
3406                 ret = -1;
3407                 goto out;
3408             }
3409         }
3410 
3411         if (size != 0) {
3412             local_progress = (float)100 / (size / MIN(size, IO_BUF_SIZE));
3413         }
3414 
3415         for (offset = 0; offset < size; offset += n) {
3416             /* How many bytes can we handle with the next read? */
3417             n = MIN(IO_BUF_SIZE, size - offset);
3418 
3419             /* If the cluster is allocated, we don't need to take action */
3420             ret = bdrv_is_allocated(bs, offset, n, &n);
3421             if (ret < 0) {
3422                 error_report("error while reading image metadata: %s",
3423                              strerror(-ret));
3424                 goto out;
3425             }
3426             if (ret) {
3427                 continue;
3428             }
3429 
3430             /*
3431              * Read old and new backing file and take into consideration that
3432              * backing files may be smaller than the COW image.
3433              */
3434             if (offset >= old_backing_size) {
3435                 memset(buf_old, 0, n);
3436             } else {
3437                 if (offset + n > old_backing_size) {
3438                     n = old_backing_size - offset;
3439                 }
3440 
3441                 ret = blk_pread(blk_old_backing, offset, buf_old, n);
3442                 if (ret < 0) {
3443                     error_report("error while reading from old backing file");
3444                     goto out;
3445                 }
3446             }
3447 
3448             if (offset >= new_backing_size || !blk_new_backing) {
3449                 memset(buf_new, 0, n);
3450             } else {
3451                 if (offset + n > new_backing_size) {
3452                     n = new_backing_size - offset;
3453                 }
3454 
3455                 ret = blk_pread(blk_new_backing, offset, buf_new, n);
3456                 if (ret < 0) {
3457                     error_report("error while reading from new backing file");
3458                     goto out;
3459                 }
3460             }
3461 
3462             /* If they differ, we need to write to the COW file */
3463             uint64_t written = 0;
3464 
3465             while (written < n) {
3466                 int64_t pnum;
3467 
3468                 if (compare_buffers(buf_old + written, buf_new + written,
3469                                     n - written, &pnum))
3470                 {
3471                     ret = blk_pwrite(blk, offset + written,
3472                                      buf_old + written, pnum, 0);
3473                     if (ret < 0) {
3474                         error_report("Error while writing to COW image: %s",
3475                             strerror(-ret));
3476                         goto out;
3477                     }
3478                 }
3479 
3480                 written += pnum;
3481             }
3482             qemu_progress_print(local_progress, 100);
3483         }
3484     }
3485 
3486     /*
3487      * Change the backing file. All clusters that are different from the old
3488      * backing file are overwritten in the COW file now, so the visible content
3489      * doesn't change when we switch the backing file.
3490      */
3491     if (out_baseimg && *out_baseimg) {
3492         ret = bdrv_change_backing_file(bs, out_baseimg, out_basefmt);
3493     } else {
3494         ret = bdrv_change_backing_file(bs, NULL, NULL);
3495     }
3496 
3497     if (ret == -ENOSPC) {
3498         error_report("Could not change the backing file to '%s': No "
3499                      "space left in the file header", out_baseimg);
3500     } else if (ret < 0) {
3501         error_report("Could not change the backing file to '%s': %s",
3502             out_baseimg, strerror(-ret));
3503     }
3504 
3505     qemu_progress_print(100, 0);
3506     /*
3507      * TODO At this point it is possible to check if any clusters that are
3508      * allocated in the COW file are the same in the backing file. If so, they
3509      * could be dropped from the COW file. Don't do this before switching the
3510      * backing file, in case of a crash this would lead to corruption.
3511      */
3512 out:
3513     qemu_progress_end();
3514     /* Cleanup */
3515     if (!unsafe) {
3516         blk_unref(blk_old_backing);
3517         blk_unref(blk_new_backing);
3518     }
3519     qemu_vfree(buf_old);
3520     qemu_vfree(buf_new);
3521 
3522     blk_unref(blk);
3523     if (ret) {
3524         return 1;
3525     }
3526     return 0;
3527 }
3528 
3529 static int img_resize(int argc, char **argv)
3530 {
3531     Error *err = NULL;
3532     int c, ret, relative;
3533     const char *filename, *fmt, *size;
3534     int64_t n, total_size, current_size, new_size;
3535     bool quiet = false;
3536     BlockBackend *blk = NULL;
3537     PreallocMode prealloc = PREALLOC_MODE_OFF;
3538     QemuOpts *param;
3539 
3540     static QemuOptsList resize_options = {
3541         .name = "resize_options",
3542         .head = QTAILQ_HEAD_INITIALIZER(resize_options.head),
3543         .desc = {
3544             {
3545                 .name = BLOCK_OPT_SIZE,
3546                 .type = QEMU_OPT_SIZE,
3547                 .help = "Virtual disk size"
3548             }, {
3549                 /* end of list */
3550             }
3551         },
3552     };
3553     bool image_opts = false;
3554     bool shrink = false;
3555 
3556     /* Remove size from argv manually so that negative numbers are not treated
3557      * as options by getopt. */
3558     if (argc < 3) {
3559         error_exit("Not enough arguments");
3560         return 1;
3561     }
3562 
3563     size = argv[--argc];
3564 
3565     /* Parse getopt arguments */
3566     fmt = NULL;
3567     for(;;) {
3568         static const struct option long_options[] = {
3569             {"help", no_argument, 0, 'h'},
3570             {"object", required_argument, 0, OPTION_OBJECT},
3571             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
3572             {"preallocation", required_argument, 0, OPTION_PREALLOCATION},
3573             {"shrink", no_argument, 0, OPTION_SHRINK},
3574             {0, 0, 0, 0}
3575         };
3576         c = getopt_long(argc, argv, ":f:hq",
3577                         long_options, NULL);
3578         if (c == -1) {
3579             break;
3580         }
3581         switch(c) {
3582         case ':':
3583             missing_argument(argv[optind - 1]);
3584             break;
3585         case '?':
3586             unrecognized_option(argv[optind - 1]);
3587             break;
3588         case 'h':
3589             help();
3590             break;
3591         case 'f':
3592             fmt = optarg;
3593             break;
3594         case 'q':
3595             quiet = true;
3596             break;
3597         case OPTION_OBJECT: {
3598             QemuOpts *opts;
3599             opts = qemu_opts_parse_noisily(&qemu_object_opts,
3600                                            optarg, true);
3601             if (!opts) {
3602                 return 1;
3603             }
3604         }   break;
3605         case OPTION_IMAGE_OPTS:
3606             image_opts = true;
3607             break;
3608         case OPTION_PREALLOCATION:
3609             prealloc = qapi_enum_parse(&PreallocMode_lookup, optarg,
3610                                        PREALLOC_MODE__MAX, NULL);
3611             if (prealloc == PREALLOC_MODE__MAX) {
3612                 error_report("Invalid preallocation mode '%s'", optarg);
3613                 return 1;
3614             }
3615             break;
3616         case OPTION_SHRINK:
3617             shrink = true;
3618             break;
3619         }
3620     }
3621     if (optind != argc - 1) {
3622         error_exit("Expecting image file name and size");
3623     }
3624     filename = argv[optind++];
3625 
3626     if (qemu_opts_foreach(&qemu_object_opts,
3627                           user_creatable_add_opts_foreach,
3628                           NULL, &error_fatal)) {
3629         return 1;
3630     }
3631 
3632     /* Choose grow, shrink, or absolute resize mode */
3633     switch (size[0]) {
3634     case '+':
3635         relative = 1;
3636         size++;
3637         break;
3638     case '-':
3639         relative = -1;
3640         size++;
3641         break;
3642     default:
3643         relative = 0;
3644         break;
3645     }
3646 
3647     /* Parse size */
3648     param = qemu_opts_create(&resize_options, NULL, 0, &error_abort);
3649     qemu_opt_set(param, BLOCK_OPT_SIZE, size, &err);
3650     if (err) {
3651         error_report_err(err);
3652         ret = -1;
3653         qemu_opts_del(param);
3654         goto out;
3655     }
3656     n = qemu_opt_get_size(param, BLOCK_OPT_SIZE, 0);
3657     qemu_opts_del(param);
3658 
3659     blk = img_open(image_opts, filename, fmt,
3660                    BDRV_O_RDWR | BDRV_O_RESIZE, false, quiet,
3661                    false);
3662     if (!blk) {
3663         ret = -1;
3664         goto out;
3665     }
3666 
3667     current_size = blk_getlength(blk);
3668     if (current_size < 0) {
3669         error_report("Failed to inquire current image length: %s",
3670                      strerror(-current_size));
3671         ret = -1;
3672         goto out;
3673     }
3674 
3675     if (relative) {
3676         total_size = current_size + n * relative;
3677     } else {
3678         total_size = n;
3679     }
3680     if (total_size <= 0) {
3681         error_report("New image size must be positive");
3682         ret = -1;
3683         goto out;
3684     }
3685 
3686     if (total_size <= current_size && prealloc != PREALLOC_MODE_OFF) {
3687         error_report("Preallocation can only be used for growing images");
3688         ret = -1;
3689         goto out;
3690     }
3691 
3692     if (total_size < current_size && !shrink) {
3693         warn_report("Shrinking an image will delete all data beyond the "
3694                     "shrunken image's end. Before performing such an "
3695                     "operation, make sure there is no important data there.");
3696 
3697         if (g_strcmp0(bdrv_get_format_name(blk_bs(blk)), "raw") != 0) {
3698             error_report(
3699               "Use the --shrink option to perform a shrink operation.");
3700             ret = -1;
3701             goto out;
3702         } else {
3703             warn_report("Using the --shrink option will suppress this message. "
3704                         "Note that future versions of qemu-img may refuse to "
3705                         "shrink images without this option.");
3706         }
3707     }
3708 
3709     ret = blk_truncate(blk, total_size, prealloc, &err);
3710     if (ret < 0) {
3711         error_report_err(err);
3712         goto out;
3713     }
3714 
3715     new_size = blk_getlength(blk);
3716     if (new_size < 0) {
3717         error_report("Failed to verify truncated image length: %s",
3718                      strerror(-new_size));
3719         ret = -1;
3720         goto out;
3721     }
3722 
3723     /* Some block drivers implement a truncation method, but only so
3724      * the user can cause qemu to refresh the image's size from disk.
3725      * The idea is that the user resizes the image outside of qemu and
3726      * then invokes block_resize to inform qemu about it.
3727      * (This includes iscsi and file-posix for device files.)
3728      * Of course, that is not the behavior someone invoking
3729      * qemu-img resize would find useful, so we catch that behavior
3730      * here and tell the user. */
3731     if (new_size != total_size && new_size == current_size) {
3732         error_report("Image was not resized; resizing may not be supported "
3733                      "for this image");
3734         ret = -1;
3735         goto out;
3736     }
3737 
3738     if (new_size != total_size) {
3739         warn_report("Image should have been resized to %" PRIi64
3740                     " bytes, but was resized to %" PRIi64 " bytes",
3741                     total_size, new_size);
3742     }
3743 
3744     qprintf(quiet, "Image resized.\n");
3745 
3746 out:
3747     blk_unref(blk);
3748     if (ret) {
3749         return 1;
3750     }
3751     return 0;
3752 }
3753 
3754 static void amend_status_cb(BlockDriverState *bs,
3755                             int64_t offset, int64_t total_work_size,
3756                             void *opaque)
3757 {
3758     qemu_progress_print(100.f * offset / total_work_size, 0);
3759 }
3760 
3761 static int print_amend_option_help(const char *format)
3762 {
3763     BlockDriver *drv;
3764 
3765     /* Find driver and parse its options */
3766     drv = bdrv_find_format(format);
3767     if (!drv) {
3768         error_report("Unknown file format '%s'", format);
3769         return 1;
3770     }
3771 
3772     if (!drv->bdrv_amend_options) {
3773         error_report("Format driver '%s' does not support option amendment",
3774                      format);
3775         return 1;
3776     }
3777 
3778     /* Every driver supporting amendment must have create_opts */
3779     assert(drv->create_opts);
3780 
3781     printf("Creation options for '%s':\n", format);
3782     qemu_opts_print_help(drv->create_opts, false);
3783     printf("\nNote that not all of these options may be amendable.\n");
3784     return 0;
3785 }
3786 
3787 static int img_amend(int argc, char **argv)
3788 {
3789     Error *err = NULL;
3790     int c, ret = 0;
3791     char *options = NULL;
3792     QemuOptsList *create_opts = NULL;
3793     QemuOpts *opts = NULL;
3794     const char *fmt = NULL, *filename, *cache;
3795     int flags;
3796     bool writethrough;
3797     bool quiet = false, progress = false;
3798     BlockBackend *blk = NULL;
3799     BlockDriverState *bs = NULL;
3800     bool image_opts = false;
3801 
3802     cache = BDRV_DEFAULT_CACHE;
3803     for (;;) {
3804         static const struct option long_options[] = {
3805             {"help", no_argument, 0, 'h'},
3806             {"object", required_argument, 0, OPTION_OBJECT},
3807             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
3808             {0, 0, 0, 0}
3809         };
3810         c = getopt_long(argc, argv, ":ho:f:t:pq",
3811                         long_options, NULL);
3812         if (c == -1) {
3813             break;
3814         }
3815 
3816         switch (c) {
3817         case ':':
3818             missing_argument(argv[optind - 1]);
3819             break;
3820         case '?':
3821             unrecognized_option(argv[optind - 1]);
3822             break;
3823         case 'h':
3824             help();
3825             break;
3826         case 'o':
3827             if (!is_valid_option_list(optarg)) {
3828                 error_report("Invalid option list: %s", optarg);
3829                 ret = -1;
3830                 goto out_no_progress;
3831             }
3832             if (!options) {
3833                 options = g_strdup(optarg);
3834             } else {
3835                 char *old_options = options;
3836                 options = g_strdup_printf("%s,%s", options, optarg);
3837                 g_free(old_options);
3838             }
3839             break;
3840         case 'f':
3841             fmt = optarg;
3842             break;
3843         case 't':
3844             cache = optarg;
3845             break;
3846         case 'p':
3847             progress = true;
3848             break;
3849         case 'q':
3850             quiet = true;
3851             break;
3852         case OPTION_OBJECT:
3853             opts = qemu_opts_parse_noisily(&qemu_object_opts,
3854                                            optarg, true);
3855             if (!opts) {
3856                 ret = -1;
3857                 goto out_no_progress;
3858             }
3859             break;
3860         case OPTION_IMAGE_OPTS:
3861             image_opts = true;
3862             break;
3863         }
3864     }
3865 
3866     if (!options) {
3867         error_exit("Must specify options (-o)");
3868     }
3869 
3870     if (qemu_opts_foreach(&qemu_object_opts,
3871                           user_creatable_add_opts_foreach,
3872                           NULL, &error_fatal)) {
3873         ret = -1;
3874         goto out_no_progress;
3875     }
3876 
3877     if (quiet) {
3878         progress = false;
3879     }
3880     qemu_progress_init(progress, 1.0);
3881 
3882     filename = (optind == argc - 1) ? argv[argc - 1] : NULL;
3883     if (fmt && has_help_option(options)) {
3884         /* If a format is explicitly specified (and possibly no filename is
3885          * given), print option help here */
3886         ret = print_amend_option_help(fmt);
3887         goto out;
3888     }
3889 
3890     if (optind != argc - 1) {
3891         error_report("Expecting one image file name");
3892         ret = -1;
3893         goto out;
3894     }
3895 
3896     flags = BDRV_O_RDWR;
3897     ret = bdrv_parse_cache_mode(cache, &flags, &writethrough);
3898     if (ret < 0) {
3899         error_report("Invalid cache option: %s", cache);
3900         goto out;
3901     }
3902 
3903     blk = img_open(image_opts, filename, fmt, flags, writethrough, quiet,
3904                    false);
3905     if (!blk) {
3906         ret = -1;
3907         goto out;
3908     }
3909     bs = blk_bs(blk);
3910 
3911     fmt = bs->drv->format_name;
3912 
3913     if (has_help_option(options)) {
3914         /* If the format was auto-detected, print option help here */
3915         ret = print_amend_option_help(fmt);
3916         goto out;
3917     }
3918 
3919     if (!bs->drv->bdrv_amend_options) {
3920         error_report("Format driver '%s' does not support option amendment",
3921                      fmt);
3922         ret = -1;
3923         goto out;
3924     }
3925 
3926     /* Every driver supporting amendment must have create_opts */
3927     assert(bs->drv->create_opts);
3928 
3929     create_opts = qemu_opts_append(create_opts, bs->drv->create_opts);
3930     opts = qemu_opts_create(create_opts, NULL, 0, &error_abort);
3931     qemu_opts_do_parse(opts, options, NULL, &err);
3932     if (err) {
3933         error_report_err(err);
3934         ret = -1;
3935         goto out;
3936     }
3937 
3938     /* In case the driver does not call amend_status_cb() */
3939     qemu_progress_print(0.f, 0);
3940     ret = bdrv_amend_options(bs, opts, &amend_status_cb, NULL, &err);
3941     qemu_progress_print(100.f, 0);
3942     if (ret < 0) {
3943         error_report_err(err);
3944         goto out;
3945     }
3946 
3947 out:
3948     qemu_progress_end();
3949 
3950 out_no_progress:
3951     blk_unref(blk);
3952     qemu_opts_del(opts);
3953     qemu_opts_free(create_opts);
3954     g_free(options);
3955 
3956     if (ret) {
3957         return 1;
3958     }
3959     return 0;
3960 }
3961 
3962 typedef struct BenchData {
3963     BlockBackend *blk;
3964     uint64_t image_size;
3965     bool write;
3966     int bufsize;
3967     int step;
3968     int nrreq;
3969     int n;
3970     int flush_interval;
3971     bool drain_on_flush;
3972     uint8_t *buf;
3973     QEMUIOVector *qiov;
3974 
3975     int in_flight;
3976     bool in_flush;
3977     uint64_t offset;
3978 } BenchData;
3979 
3980 static void bench_undrained_flush_cb(void *opaque, int ret)
3981 {
3982     if (ret < 0) {
3983         error_report("Failed flush request: %s", strerror(-ret));
3984         exit(EXIT_FAILURE);
3985     }
3986 }
3987 
3988 static void bench_cb(void *opaque, int ret)
3989 {
3990     BenchData *b = opaque;
3991     BlockAIOCB *acb;
3992 
3993     if (ret < 0) {
3994         error_report("Failed request: %s", strerror(-ret));
3995         exit(EXIT_FAILURE);
3996     }
3997 
3998     if (b->in_flush) {
3999         /* Just finished a flush with drained queue: Start next requests */
4000         assert(b->in_flight == 0);
4001         b->in_flush = false;
4002     } else if (b->in_flight > 0) {
4003         int remaining = b->n - b->in_flight;
4004 
4005         b->n--;
4006         b->in_flight--;
4007 
4008         /* Time for flush? Drain queue if requested, then flush */
4009         if (b->flush_interval && remaining % b->flush_interval == 0) {
4010             if (!b->in_flight || !b->drain_on_flush) {
4011                 BlockCompletionFunc *cb;
4012 
4013                 if (b->drain_on_flush) {
4014                     b->in_flush = true;
4015                     cb = bench_cb;
4016                 } else {
4017                     cb = bench_undrained_flush_cb;
4018                 }
4019 
4020                 acb = blk_aio_flush(b->blk, cb, b);
4021                 if (!acb) {
4022                     error_report("Failed to issue flush request");
4023                     exit(EXIT_FAILURE);
4024                 }
4025             }
4026             if (b->drain_on_flush) {
4027                 return;
4028             }
4029         }
4030     }
4031 
4032     while (b->n > b->in_flight && b->in_flight < b->nrreq) {
4033         int64_t offset = b->offset;
4034         /* blk_aio_* might look for completed I/Os and kick bench_cb
4035          * again, so make sure this operation is counted by in_flight
4036          * and b->offset is ready for the next submission.
4037          */
4038         b->in_flight++;
4039         b->offset += b->step;
4040         b->offset %= b->image_size;
4041         if (b->write) {
4042             acb = blk_aio_pwritev(b->blk, offset, b->qiov, 0, bench_cb, b);
4043         } else {
4044             acb = blk_aio_preadv(b->blk, offset, b->qiov, 0, bench_cb, b);
4045         }
4046         if (!acb) {
4047             error_report("Failed to issue request");
4048             exit(EXIT_FAILURE);
4049         }
4050     }
4051 }
4052 
4053 static int img_bench(int argc, char **argv)
4054 {
4055     int c, ret = 0;
4056     const char *fmt = NULL, *filename;
4057     bool quiet = false;
4058     bool image_opts = false;
4059     bool is_write = false;
4060     int count = 75000;
4061     int depth = 64;
4062     int64_t offset = 0;
4063     size_t bufsize = 4096;
4064     int pattern = 0;
4065     size_t step = 0;
4066     int flush_interval = 0;
4067     bool drain_on_flush = true;
4068     int64_t image_size;
4069     BlockBackend *blk = NULL;
4070     BenchData data = {};
4071     int flags = 0;
4072     bool writethrough = false;
4073     struct timeval t1, t2;
4074     int i;
4075     bool force_share = false;
4076     size_t buf_size;
4077 
4078     for (;;) {
4079         static const struct option long_options[] = {
4080             {"help", no_argument, 0, 'h'},
4081             {"flush-interval", required_argument, 0, OPTION_FLUSH_INTERVAL},
4082             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
4083             {"pattern", required_argument, 0, OPTION_PATTERN},
4084             {"no-drain", no_argument, 0, OPTION_NO_DRAIN},
4085             {"force-share", no_argument, 0, 'U'},
4086             {0, 0, 0, 0}
4087         };
4088         c = getopt_long(argc, argv, ":hc:d:f:no:qs:S:t:wU", long_options, NULL);
4089         if (c == -1) {
4090             break;
4091         }
4092 
4093         switch (c) {
4094         case ':':
4095             missing_argument(argv[optind - 1]);
4096             break;
4097         case '?':
4098             unrecognized_option(argv[optind - 1]);
4099             break;
4100         case 'h':
4101             help();
4102             break;
4103         case 'c':
4104         {
4105             unsigned long res;
4106 
4107             if (qemu_strtoul(optarg, NULL, 0, &res) < 0 || res > INT_MAX) {
4108                 error_report("Invalid request count specified");
4109                 return 1;
4110             }
4111             count = res;
4112             break;
4113         }
4114         case 'd':
4115         {
4116             unsigned long res;
4117 
4118             if (qemu_strtoul(optarg, NULL, 0, &res) < 0 || res > INT_MAX) {
4119                 error_report("Invalid queue depth specified");
4120                 return 1;
4121             }
4122             depth = res;
4123             break;
4124         }
4125         case 'f':
4126             fmt = optarg;
4127             break;
4128         case 'n':
4129             flags |= BDRV_O_NATIVE_AIO;
4130             break;
4131         case 'o':
4132         {
4133             offset = cvtnum(optarg);
4134             if (offset < 0) {
4135                 error_report("Invalid offset specified");
4136                 return 1;
4137             }
4138             break;
4139         }
4140             break;
4141         case 'q':
4142             quiet = true;
4143             break;
4144         case 's':
4145         {
4146             int64_t sval;
4147 
4148             sval = cvtnum(optarg);
4149             if (sval < 0 || sval > INT_MAX) {
4150                 error_report("Invalid buffer size specified");
4151                 return 1;
4152             }
4153 
4154             bufsize = sval;
4155             break;
4156         }
4157         case 'S':
4158         {
4159             int64_t sval;
4160 
4161             sval = cvtnum(optarg);
4162             if (sval < 0 || sval > INT_MAX) {
4163                 error_report("Invalid step size specified");
4164                 return 1;
4165             }
4166 
4167             step = sval;
4168             break;
4169         }
4170         case 't':
4171             ret = bdrv_parse_cache_mode(optarg, &flags, &writethrough);
4172             if (ret < 0) {
4173                 error_report("Invalid cache mode");
4174                 ret = -1;
4175                 goto out;
4176             }
4177             break;
4178         case 'w':
4179             flags |= BDRV_O_RDWR;
4180             is_write = true;
4181             break;
4182         case 'U':
4183             force_share = true;
4184             break;
4185         case OPTION_PATTERN:
4186         {
4187             unsigned long res;
4188 
4189             if (qemu_strtoul(optarg, NULL, 0, &res) < 0 || res > 0xff) {
4190                 error_report("Invalid pattern byte specified");
4191                 return 1;
4192             }
4193             pattern = res;
4194             break;
4195         }
4196         case OPTION_FLUSH_INTERVAL:
4197         {
4198             unsigned long res;
4199 
4200             if (qemu_strtoul(optarg, NULL, 0, &res) < 0 || res > INT_MAX) {
4201                 error_report("Invalid flush interval specified");
4202                 return 1;
4203             }
4204             flush_interval = res;
4205             break;
4206         }
4207         case OPTION_NO_DRAIN:
4208             drain_on_flush = false;
4209             break;
4210         case OPTION_IMAGE_OPTS:
4211             image_opts = true;
4212             break;
4213         }
4214     }
4215 
4216     if (optind != argc - 1) {
4217         error_exit("Expecting one image file name");
4218     }
4219     filename = argv[argc - 1];
4220 
4221     if (!is_write && flush_interval) {
4222         error_report("--flush-interval is only available in write tests");
4223         ret = -1;
4224         goto out;
4225     }
4226     if (flush_interval && flush_interval < depth) {
4227         error_report("Flush interval can't be smaller than depth");
4228         ret = -1;
4229         goto out;
4230     }
4231 
4232     blk = img_open(image_opts, filename, fmt, flags, writethrough, quiet,
4233                    force_share);
4234     if (!blk) {
4235         ret = -1;
4236         goto out;
4237     }
4238 
4239     image_size = blk_getlength(blk);
4240     if (image_size < 0) {
4241         ret = image_size;
4242         goto out;
4243     }
4244 
4245     data = (BenchData) {
4246         .blk            = blk,
4247         .image_size     = image_size,
4248         .bufsize        = bufsize,
4249         .step           = step ?: bufsize,
4250         .nrreq          = depth,
4251         .n              = count,
4252         .offset         = offset,
4253         .write          = is_write,
4254         .flush_interval = flush_interval,
4255         .drain_on_flush = drain_on_flush,
4256     };
4257     printf("Sending %d %s requests, %d bytes each, %d in parallel "
4258            "(starting at offset %" PRId64 ", step size %d)\n",
4259            data.n, data.write ? "write" : "read", data.bufsize, data.nrreq,
4260            data.offset, data.step);
4261     if (flush_interval) {
4262         printf("Sending flush every %d requests\n", flush_interval);
4263     }
4264 
4265     buf_size = data.nrreq * data.bufsize;
4266     data.buf = blk_blockalign(blk, buf_size);
4267     memset(data.buf, pattern, data.nrreq * data.bufsize);
4268 
4269     blk_register_buf(blk, data.buf, buf_size);
4270 
4271     data.qiov = g_new(QEMUIOVector, data.nrreq);
4272     for (i = 0; i < data.nrreq; i++) {
4273         qemu_iovec_init(&data.qiov[i], 1);
4274         qemu_iovec_add(&data.qiov[i],
4275                        data.buf + i * data.bufsize, data.bufsize);
4276     }
4277 
4278     gettimeofday(&t1, NULL);
4279     bench_cb(&data, 0);
4280 
4281     while (data.n > 0) {
4282         main_loop_wait(false);
4283     }
4284     gettimeofday(&t2, NULL);
4285 
4286     printf("Run completed in %3.3f seconds.\n",
4287            (t2.tv_sec - t1.tv_sec)
4288            + ((double)(t2.tv_usec - t1.tv_usec) / 1000000));
4289 
4290 out:
4291     if (data.buf) {
4292         blk_unregister_buf(blk, data.buf);
4293     }
4294     qemu_vfree(data.buf);
4295     blk_unref(blk);
4296 
4297     if (ret) {
4298         return 1;
4299     }
4300     return 0;
4301 }
4302 
4303 #define C_BS      01
4304 #define C_COUNT   02
4305 #define C_IF      04
4306 #define C_OF      010
4307 #define C_SKIP    020
4308 
4309 struct DdInfo {
4310     unsigned int flags;
4311     int64_t count;
4312 };
4313 
4314 struct DdIo {
4315     int bsz;    /* Block size */
4316     char *filename;
4317     uint8_t *buf;
4318     int64_t offset;
4319 };
4320 
4321 struct DdOpts {
4322     const char *name;
4323     int (*f)(const char *, struct DdIo *, struct DdIo *, struct DdInfo *);
4324     unsigned int flag;
4325 };
4326 
4327 static int img_dd_bs(const char *arg,
4328                      struct DdIo *in, struct DdIo *out,
4329                      struct DdInfo *dd)
4330 {
4331     int64_t res;
4332 
4333     res = cvtnum(arg);
4334 
4335     if (res <= 0 || res > INT_MAX) {
4336         error_report("invalid number: '%s'", arg);
4337         return 1;
4338     }
4339     in->bsz = out->bsz = res;
4340 
4341     return 0;
4342 }
4343 
4344 static int img_dd_count(const char *arg,
4345                         struct DdIo *in, struct DdIo *out,
4346                         struct DdInfo *dd)
4347 {
4348     dd->count = cvtnum(arg);
4349 
4350     if (dd->count < 0) {
4351         error_report("invalid number: '%s'", arg);
4352         return 1;
4353     }
4354 
4355     return 0;
4356 }
4357 
4358 static int img_dd_if(const char *arg,
4359                      struct DdIo *in, struct DdIo *out,
4360                      struct DdInfo *dd)
4361 {
4362     in->filename = g_strdup(arg);
4363 
4364     return 0;
4365 }
4366 
4367 static int img_dd_of(const char *arg,
4368                      struct DdIo *in, struct DdIo *out,
4369                      struct DdInfo *dd)
4370 {
4371     out->filename = g_strdup(arg);
4372 
4373     return 0;
4374 }
4375 
4376 static int img_dd_skip(const char *arg,
4377                        struct DdIo *in, struct DdIo *out,
4378                        struct DdInfo *dd)
4379 {
4380     in->offset = cvtnum(arg);
4381 
4382     if (in->offset < 0) {
4383         error_report("invalid number: '%s'", arg);
4384         return 1;
4385     }
4386 
4387     return 0;
4388 }
4389 
4390 static int img_dd(int argc, char **argv)
4391 {
4392     int ret = 0;
4393     char *arg = NULL;
4394     char *tmp;
4395     BlockDriver *drv = NULL, *proto_drv = NULL;
4396     BlockBackend *blk1 = NULL, *blk2 = NULL;
4397     QemuOpts *opts = NULL;
4398     QemuOptsList *create_opts = NULL;
4399     Error *local_err = NULL;
4400     bool image_opts = false;
4401     int c, i;
4402     const char *out_fmt = "raw";
4403     const char *fmt = NULL;
4404     int64_t size = 0;
4405     int64_t block_count = 0, out_pos, in_pos;
4406     bool force_share = false;
4407     struct DdInfo dd = {
4408         .flags = 0,
4409         .count = 0,
4410     };
4411     struct DdIo in = {
4412         .bsz = 512, /* Block size is by default 512 bytes */
4413         .filename = NULL,
4414         .buf = NULL,
4415         .offset = 0
4416     };
4417     struct DdIo out = {
4418         .bsz = 512,
4419         .filename = NULL,
4420         .buf = NULL,
4421         .offset = 0
4422     };
4423 
4424     const struct DdOpts options[] = {
4425         { "bs", img_dd_bs, C_BS },
4426         { "count", img_dd_count, C_COUNT },
4427         { "if", img_dd_if, C_IF },
4428         { "of", img_dd_of, C_OF },
4429         { "skip", img_dd_skip, C_SKIP },
4430         { NULL, NULL, 0 }
4431     };
4432     const struct option long_options[] = {
4433         { "help", no_argument, 0, 'h'},
4434         { "object", required_argument, 0, OPTION_OBJECT},
4435         { "image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
4436         { "force-share", no_argument, 0, 'U'},
4437         { 0, 0, 0, 0 }
4438     };
4439 
4440     while ((c = getopt_long(argc, argv, ":hf:O:U", long_options, NULL))) {
4441         if (c == EOF) {
4442             break;
4443         }
4444         switch (c) {
4445         case 'O':
4446             out_fmt = optarg;
4447             break;
4448         case 'f':
4449             fmt = optarg;
4450             break;
4451         case ':':
4452             missing_argument(argv[optind - 1]);
4453             break;
4454         case '?':
4455             unrecognized_option(argv[optind - 1]);
4456             break;
4457         case 'h':
4458             help();
4459             break;
4460         case 'U':
4461             force_share = true;
4462             break;
4463         case OPTION_OBJECT:
4464             if (!qemu_opts_parse_noisily(&qemu_object_opts, optarg, true)) {
4465                 ret = -1;
4466                 goto out;
4467             }
4468             break;
4469         case OPTION_IMAGE_OPTS:
4470             image_opts = true;
4471             break;
4472         }
4473     }
4474 
4475     for (i = optind; i < argc; i++) {
4476         int j;
4477         arg = g_strdup(argv[i]);
4478 
4479         tmp = strchr(arg, '=');
4480         if (tmp == NULL) {
4481             error_report("unrecognized operand %s", arg);
4482             ret = -1;
4483             goto out;
4484         }
4485 
4486         *tmp++ = '\0';
4487 
4488         for (j = 0; options[j].name != NULL; j++) {
4489             if (!strcmp(arg, options[j].name)) {
4490                 break;
4491             }
4492         }
4493         if (options[j].name == NULL) {
4494             error_report("unrecognized operand %s", arg);
4495             ret = -1;
4496             goto out;
4497         }
4498 
4499         if (options[j].f(tmp, &in, &out, &dd) != 0) {
4500             ret = -1;
4501             goto out;
4502         }
4503         dd.flags |= options[j].flag;
4504         g_free(arg);
4505         arg = NULL;
4506     }
4507 
4508     if (!(dd.flags & C_IF && dd.flags & C_OF)) {
4509         error_report("Must specify both input and output files");
4510         ret = -1;
4511         goto out;
4512     }
4513 
4514     if (qemu_opts_foreach(&qemu_object_opts,
4515                           user_creatable_add_opts_foreach,
4516                           NULL, &error_fatal)) {
4517         ret = -1;
4518         goto out;
4519     }
4520 
4521     blk1 = img_open(image_opts, in.filename, fmt, 0, false, false,
4522                     force_share);
4523 
4524     if (!blk1) {
4525         ret = -1;
4526         goto out;
4527     }
4528 
4529     drv = bdrv_find_format(out_fmt);
4530     if (!drv) {
4531         error_report("Unknown file format");
4532         ret = -1;
4533         goto out;
4534     }
4535     proto_drv = bdrv_find_protocol(out.filename, true, &local_err);
4536 
4537     if (!proto_drv) {
4538         error_report_err(local_err);
4539         ret = -1;
4540         goto out;
4541     }
4542     if (!drv->create_opts) {
4543         error_report("Format driver '%s' does not support image creation",
4544                      drv->format_name);
4545         ret = -1;
4546         goto out;
4547     }
4548     if (!proto_drv->create_opts) {
4549         error_report("Protocol driver '%s' does not support image creation",
4550                      proto_drv->format_name);
4551         ret = -1;
4552         goto out;
4553     }
4554     create_opts = qemu_opts_append(create_opts, drv->create_opts);
4555     create_opts = qemu_opts_append(create_opts, proto_drv->create_opts);
4556 
4557     opts = qemu_opts_create(create_opts, NULL, 0, &error_abort);
4558 
4559     size = blk_getlength(blk1);
4560     if (size < 0) {
4561         error_report("Failed to get size for '%s'", in.filename);
4562         ret = -1;
4563         goto out;
4564     }
4565 
4566     if (dd.flags & C_COUNT && dd.count <= INT64_MAX / in.bsz &&
4567         dd.count * in.bsz < size) {
4568         size = dd.count * in.bsz;
4569     }
4570 
4571     /* Overflow means the specified offset is beyond input image's size */
4572     if (dd.flags & C_SKIP && (in.offset > INT64_MAX / in.bsz ||
4573                               size < in.bsz * in.offset)) {
4574         qemu_opt_set_number(opts, BLOCK_OPT_SIZE, 0, &error_abort);
4575     } else {
4576         qemu_opt_set_number(opts, BLOCK_OPT_SIZE,
4577                             size - in.bsz * in.offset, &error_abort);
4578     }
4579 
4580     ret = bdrv_create(drv, out.filename, opts, &local_err);
4581     if (ret < 0) {
4582         error_reportf_err(local_err,
4583                           "%s: error while creating output image: ",
4584                           out.filename);
4585         ret = -1;
4586         goto out;
4587     }
4588 
4589     /* TODO, we can't honour --image-opts for the target,
4590      * since it needs to be given in a format compatible
4591      * with the bdrv_create() call above which does not
4592      * support image-opts style.
4593      */
4594     blk2 = img_open_file(out.filename, NULL, out_fmt, BDRV_O_RDWR,
4595                          false, false, false);
4596 
4597     if (!blk2) {
4598         ret = -1;
4599         goto out;
4600     }
4601 
4602     if (dd.flags & C_SKIP && (in.offset > INT64_MAX / in.bsz ||
4603                               size < in.offset * in.bsz)) {
4604         /* We give a warning if the skip option is bigger than the input
4605          * size and create an empty output disk image (i.e. like dd(1)).
4606          */
4607         error_report("%s: cannot skip to specified offset", in.filename);
4608         in_pos = size;
4609     } else {
4610         in_pos = in.offset * in.bsz;
4611     }
4612 
4613     in.buf = g_new(uint8_t, in.bsz);
4614 
4615     for (out_pos = 0; in_pos < size; block_count++) {
4616         int in_ret, out_ret;
4617 
4618         if (in_pos + in.bsz > size) {
4619             in_ret = blk_pread(blk1, in_pos, in.buf, size - in_pos);
4620         } else {
4621             in_ret = blk_pread(blk1, in_pos, in.buf, in.bsz);
4622         }
4623         if (in_ret < 0) {
4624             error_report("error while reading from input image file: %s",
4625                          strerror(-in_ret));
4626             ret = -1;
4627             goto out;
4628         }
4629         in_pos += in_ret;
4630 
4631         out_ret = blk_pwrite(blk2, out_pos, in.buf, in_ret, 0);
4632 
4633         if (out_ret < 0) {
4634             error_report("error while writing to output image file: %s",
4635                          strerror(-out_ret));
4636             ret = -1;
4637             goto out;
4638         }
4639         out_pos += out_ret;
4640     }
4641 
4642 out:
4643     g_free(arg);
4644     qemu_opts_del(opts);
4645     qemu_opts_free(create_opts);
4646     blk_unref(blk1);
4647     blk_unref(blk2);
4648     g_free(in.filename);
4649     g_free(out.filename);
4650     g_free(in.buf);
4651     g_free(out.buf);
4652 
4653     if (ret) {
4654         return 1;
4655     }
4656     return 0;
4657 }
4658 
4659 static void dump_json_block_measure_info(BlockMeasureInfo *info)
4660 {
4661     QString *str;
4662     QObject *obj;
4663     Visitor *v = qobject_output_visitor_new(&obj);
4664 
4665     visit_type_BlockMeasureInfo(v, NULL, &info, &error_abort);
4666     visit_complete(v, &obj);
4667     str = qobject_to_json_pretty(obj);
4668     assert(str != NULL);
4669     printf("%s\n", qstring_get_str(str));
4670     qobject_unref(obj);
4671     visit_free(v);
4672     qobject_unref(str);
4673 }
4674 
4675 static int img_measure(int argc, char **argv)
4676 {
4677     static const struct option long_options[] = {
4678         {"help", no_argument, 0, 'h'},
4679         {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
4680         {"object", required_argument, 0, OPTION_OBJECT},
4681         {"output", required_argument, 0, OPTION_OUTPUT},
4682         {"size", required_argument, 0, OPTION_SIZE},
4683         {"force-share", no_argument, 0, 'U'},
4684         {0, 0, 0, 0}
4685     };
4686     OutputFormat output_format = OFORMAT_HUMAN;
4687     BlockBackend *in_blk = NULL;
4688     BlockDriver *drv;
4689     const char *filename = NULL;
4690     const char *fmt = NULL;
4691     const char *out_fmt = "raw";
4692     char *options = NULL;
4693     char *snapshot_name = NULL;
4694     bool force_share = false;
4695     QemuOpts *opts = NULL;
4696     QemuOpts *object_opts = NULL;
4697     QemuOpts *sn_opts = NULL;
4698     QemuOptsList *create_opts = NULL;
4699     bool image_opts = false;
4700     uint64_t img_size = UINT64_MAX;
4701     BlockMeasureInfo *info = NULL;
4702     Error *local_err = NULL;
4703     int ret = 1;
4704     int c;
4705 
4706     while ((c = getopt_long(argc, argv, "hf:O:o:l:U",
4707                             long_options, NULL)) != -1) {
4708         switch (c) {
4709         case '?':
4710         case 'h':
4711             help();
4712             break;
4713         case 'f':
4714             fmt = optarg;
4715             break;
4716         case 'O':
4717             out_fmt = optarg;
4718             break;
4719         case 'o':
4720             if (!is_valid_option_list(optarg)) {
4721                 error_report("Invalid option list: %s", optarg);
4722                 goto out;
4723             }
4724             if (!options) {
4725                 options = g_strdup(optarg);
4726             } else {
4727                 char *old_options = options;
4728                 options = g_strdup_printf("%s,%s", options, optarg);
4729                 g_free(old_options);
4730             }
4731             break;
4732         case 'l':
4733             if (strstart(optarg, SNAPSHOT_OPT_BASE, NULL)) {
4734                 sn_opts = qemu_opts_parse_noisily(&internal_snapshot_opts,
4735                                                   optarg, false);
4736                 if (!sn_opts) {
4737                     error_report("Failed in parsing snapshot param '%s'",
4738                                  optarg);
4739                     goto out;
4740                 }
4741             } else {
4742                 snapshot_name = optarg;
4743             }
4744             break;
4745         case 'U':
4746             force_share = true;
4747             break;
4748         case OPTION_OBJECT:
4749             object_opts = qemu_opts_parse_noisily(&qemu_object_opts,
4750                                                   optarg, true);
4751             if (!object_opts) {
4752                 goto out;
4753             }
4754             break;
4755         case OPTION_IMAGE_OPTS:
4756             image_opts = true;
4757             break;
4758         case OPTION_OUTPUT:
4759             if (!strcmp(optarg, "json")) {
4760                 output_format = OFORMAT_JSON;
4761             } else if (!strcmp(optarg, "human")) {
4762                 output_format = OFORMAT_HUMAN;
4763             } else {
4764                 error_report("--output must be used with human or json "
4765                              "as argument.");
4766                 goto out;
4767             }
4768             break;
4769         case OPTION_SIZE:
4770         {
4771             int64_t sval;
4772 
4773             sval = cvtnum(optarg);
4774             if (sval < 0) {
4775                 if (sval == -ERANGE) {
4776                     error_report("Image size must be less than 8 EiB!");
4777                 } else {
4778                     error_report("Invalid image size specified! You may use "
4779                                  "k, M, G, T, P or E suffixes for ");
4780                     error_report("kilobytes, megabytes, gigabytes, terabytes, "
4781                                  "petabytes and exabytes.");
4782                 }
4783                 goto out;
4784             }
4785             img_size = (uint64_t)sval;
4786         }
4787         break;
4788         }
4789     }
4790 
4791     if (qemu_opts_foreach(&qemu_object_opts,
4792                           user_creatable_add_opts_foreach,
4793                           NULL, &error_fatal)) {
4794         goto out;
4795     }
4796 
4797     if (argc - optind > 1) {
4798         error_report("At most one filename argument is allowed.");
4799         goto out;
4800     } else if (argc - optind == 1) {
4801         filename = argv[optind];
4802     }
4803 
4804     if (!filename &&
4805         (object_opts || image_opts || fmt || snapshot_name || sn_opts)) {
4806         error_report("--object, --image-opts, -f, and -l "
4807                      "require a filename argument.");
4808         goto out;
4809     }
4810     if (filename && img_size != UINT64_MAX) {
4811         error_report("--size N cannot be used together with a filename.");
4812         goto out;
4813     }
4814     if (!filename && img_size == UINT64_MAX) {
4815         error_report("Either --size N or one filename must be specified.");
4816         goto out;
4817     }
4818 
4819     if (filename) {
4820         in_blk = img_open(image_opts, filename, fmt, 0,
4821                           false, false, force_share);
4822         if (!in_blk) {
4823             goto out;
4824         }
4825 
4826         if (sn_opts) {
4827             bdrv_snapshot_load_tmp(blk_bs(in_blk),
4828                     qemu_opt_get(sn_opts, SNAPSHOT_OPT_ID),
4829                     qemu_opt_get(sn_opts, SNAPSHOT_OPT_NAME),
4830                     &local_err);
4831         } else if (snapshot_name != NULL) {
4832             bdrv_snapshot_load_tmp_by_id_or_name(blk_bs(in_blk),
4833                     snapshot_name, &local_err);
4834         }
4835         if (local_err) {
4836             error_reportf_err(local_err, "Failed to load snapshot: ");
4837             goto out;
4838         }
4839     }
4840 
4841     drv = bdrv_find_format(out_fmt);
4842     if (!drv) {
4843         error_report("Unknown file format '%s'", out_fmt);
4844         goto out;
4845     }
4846     if (!drv->create_opts) {
4847         error_report("Format driver '%s' does not support image creation",
4848                      drv->format_name);
4849         goto out;
4850     }
4851 
4852     create_opts = qemu_opts_append(create_opts, drv->create_opts);
4853     create_opts = qemu_opts_append(create_opts, bdrv_file.create_opts);
4854     opts = qemu_opts_create(create_opts, NULL, 0, &error_abort);
4855     if (options) {
4856         qemu_opts_do_parse(opts, options, NULL, &local_err);
4857         if (local_err) {
4858             error_report_err(local_err);
4859             error_report("Invalid options for file format '%s'", out_fmt);
4860             goto out;
4861         }
4862     }
4863     if (img_size != UINT64_MAX) {
4864         qemu_opt_set_number(opts, BLOCK_OPT_SIZE, img_size, &error_abort);
4865     }
4866 
4867     info = bdrv_measure(drv, opts, in_blk ? blk_bs(in_blk) : NULL, &local_err);
4868     if (local_err) {
4869         error_report_err(local_err);
4870         goto out;
4871     }
4872 
4873     if (output_format == OFORMAT_HUMAN) {
4874         printf("required size: %" PRIu64 "\n", info->required);
4875         printf("fully allocated size: %" PRIu64 "\n", info->fully_allocated);
4876     } else {
4877         dump_json_block_measure_info(info);
4878     }
4879 
4880     ret = 0;
4881 
4882 out:
4883     qapi_free_BlockMeasureInfo(info);
4884     qemu_opts_del(object_opts);
4885     qemu_opts_del(opts);
4886     qemu_opts_del(sn_opts);
4887     qemu_opts_free(create_opts);
4888     g_free(options);
4889     blk_unref(in_blk);
4890     return ret;
4891 }
4892 
4893 static const img_cmd_t img_cmds[] = {
4894 #define DEF(option, callback, arg_string)        \
4895     { option, callback },
4896 #include "qemu-img-cmds.h"
4897 #undef DEF
4898     { NULL, NULL, },
4899 };
4900 
4901 int main(int argc, char **argv)
4902 {
4903     const img_cmd_t *cmd;
4904     const char *cmdname;
4905     Error *local_error = NULL;
4906     char *trace_file = NULL;
4907     int c;
4908     static const struct option long_options[] = {
4909         {"help", no_argument, 0, 'h'},
4910         {"version", no_argument, 0, 'V'},
4911         {"trace", required_argument, NULL, 'T'},
4912         {0, 0, 0, 0}
4913     };
4914 
4915 #ifdef CONFIG_POSIX
4916     signal(SIGPIPE, SIG_IGN);
4917 #endif
4918 
4919     module_call_init(MODULE_INIT_TRACE);
4920     error_set_progname(argv[0]);
4921     qemu_init_exec_dir(argv[0]);
4922 
4923     if (qemu_init_main_loop(&local_error)) {
4924         error_report_err(local_error);
4925         exit(EXIT_FAILURE);
4926     }
4927 
4928     qcrypto_init(&error_fatal);
4929 
4930     module_call_init(MODULE_INIT_QOM);
4931     bdrv_init();
4932     if (argc < 2) {
4933         error_exit("Not enough arguments");
4934     }
4935 
4936     qemu_add_opts(&qemu_object_opts);
4937     qemu_add_opts(&qemu_source_opts);
4938     qemu_add_opts(&qemu_trace_opts);
4939 
4940     while ((c = getopt_long(argc, argv, "+:hVT:", long_options, NULL)) != -1) {
4941         switch (c) {
4942         case ':':
4943             missing_argument(argv[optind - 1]);
4944             return 0;
4945         case '?':
4946             unrecognized_option(argv[optind - 1]);
4947             return 0;
4948         case 'h':
4949             help();
4950             return 0;
4951         case 'V':
4952             printf(QEMU_IMG_VERSION);
4953             return 0;
4954         case 'T':
4955             g_free(trace_file);
4956             trace_file = trace_opt_parse(optarg);
4957             break;
4958         }
4959     }
4960 
4961     cmdname = argv[optind];
4962 
4963     /* reset getopt_long scanning */
4964     argc -= optind;
4965     if (argc < 1) {
4966         return 0;
4967     }
4968     argv += optind;
4969     qemu_reset_optind();
4970 
4971     if (!trace_init_backends()) {
4972         exit(1);
4973     }
4974     trace_init_file(trace_file);
4975     qemu_set_log(LOG_TRACE);
4976 
4977     /* find the command */
4978     for (cmd = img_cmds; cmd->name != NULL; cmd++) {
4979         if (!strcmp(cmdname, cmd->name)) {
4980             return cmd->handler(argc, argv);
4981         }
4982     }
4983 
4984     /* not found */
4985     error_exit("Command not found: %s", cmdname);
4986 }
4987