xref: /qemu/qemu-img.c (revision f03868bd5653265e97b253102d77d83ea85efdea)
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            "  'snapshot_id_or_name' is deprecated, use 'snapshot_param'\n"
152            "    instead\n"
153            "  '-c' indicates that target image must be compressed (qcow format only)\n"
154            "  '-u' allows unsafe backing chains. For rebasing, it is assumed that old and\n"
155            "       new backing file match exactly. The image doesn't need a working\n"
156            "       backing file before rebasing in this case (useful for renaming the\n"
157            "       backing file). For image creation, allow creating without attempting\n"
158            "       to open the backing file.\n"
159            "  '-h' with or without a command shows this help and lists the supported formats\n"
160            "  '-p' show progress of command (only certain commands)\n"
161            "  '-q' use Quiet mode - do not print any output (except errors)\n"
162            "  '-S' indicates the consecutive number of bytes (defaults to 4k) that must\n"
163            "       contain only zeros for qemu-img to create a sparse image during\n"
164            "       conversion. If the number of bytes is 0, the source will not be scanned for\n"
165            "       unallocated or zero sectors, and the destination image will always be\n"
166            "       fully allocated\n"
167            "  '--output' takes the format in which the output must be done (human or json)\n"
168            "  '-n' skips the target volume creation (useful if the volume is created\n"
169            "       prior to running qemu-img)\n"
170            "\n"
171            "Parameters to check subcommand:\n"
172            "  '-r' tries to repair any inconsistencies that are found during the check.\n"
173            "       '-r leaks' repairs only cluster leaks, whereas '-r all' fixes all\n"
174            "       kinds of errors, with a higher risk of choosing the wrong fix or\n"
175            "       hiding corruption that has already occurred.\n"
176            "\n"
177            "Parameters to convert subcommand:\n"
178            "  '-m' specifies how many coroutines work in parallel during the convert\n"
179            "       process (defaults to 8)\n"
180            "  '-W' allow to write to the target out of order rather than sequential\n"
181            "\n"
182            "Parameters to snapshot subcommand:\n"
183            "  'snapshot' is the name of the snapshot to create, apply or delete\n"
184            "  '-a' applies a snapshot (revert disk to saved state)\n"
185            "  '-c' creates a snapshot\n"
186            "  '-d' deletes a snapshot\n"
187            "  '-l' lists all snapshots in the given image\n"
188            "\n"
189            "Parameters to compare subcommand:\n"
190            "  '-f' first image format\n"
191            "  '-F' second image format\n"
192            "  '-s' run in Strict mode - fail on different image size or sector allocation\n"
193            "\n"
194            "Parameters to dd subcommand:\n"
195            "  'bs=BYTES' read and write up to BYTES bytes at a time "
196            "(default: 512)\n"
197            "  'count=N' copy only N input blocks\n"
198            "  'if=FILE' read from FILE\n"
199            "  'of=FILE' write to FILE\n"
200            "  'skip=N' skip N bs-sized blocks at the start of input\n";
201 
202     printf("%s\nSupported formats:", help_msg);
203     bdrv_iterate_format(format_print, NULL);
204     printf("\n\n" QEMU_HELP_BOTTOM "\n");
205     exit(EXIT_SUCCESS);
206 }
207 
208 static QemuOptsList qemu_object_opts = {
209     .name = "object",
210     .implied_opt_name = "qom-type",
211     .head = QTAILQ_HEAD_INITIALIZER(qemu_object_opts.head),
212     .desc = {
213         { }
214     },
215 };
216 
217 static QemuOptsList qemu_source_opts = {
218     .name = "source",
219     .implied_opt_name = "file",
220     .head = QTAILQ_HEAD_INITIALIZER(qemu_source_opts.head),
221     .desc = {
222         { }
223     },
224 };
225 
226 static int GCC_FMT_ATTR(2, 3) qprintf(bool quiet, const char *fmt, ...)
227 {
228     int ret = 0;
229     if (!quiet) {
230         va_list args;
231         va_start(args, fmt);
232         ret = vprintf(fmt, args);
233         va_end(args);
234     }
235     return ret;
236 }
237 
238 
239 static int print_block_option_help(const char *filename, const char *fmt)
240 {
241     BlockDriver *drv, *proto_drv;
242     QemuOptsList *create_opts = NULL;
243     Error *local_err = NULL;
244 
245     /* Find driver and parse its options */
246     drv = bdrv_find_format(fmt);
247     if (!drv) {
248         error_report("Unknown file format '%s'", fmt);
249         return 1;
250     }
251 
252     create_opts = qemu_opts_append(create_opts, drv->create_opts);
253     if (filename) {
254         proto_drv = bdrv_find_protocol(filename, true, &local_err);
255         if (!proto_drv) {
256             error_report_err(local_err);
257             qemu_opts_free(create_opts);
258             return 1;
259         }
260         create_opts = qemu_opts_append(create_opts, proto_drv->create_opts);
261     }
262 
263     qemu_opts_print_help(create_opts);
264     qemu_opts_free(create_opts);
265     return 0;
266 }
267 
268 
269 static BlockBackend *img_open_opts(const char *optstr,
270                                    QemuOpts *opts, int flags, bool writethrough,
271                                    bool quiet, bool force_share)
272 {
273     QDict *options;
274     Error *local_err = NULL;
275     BlockBackend *blk;
276     options = qemu_opts_to_qdict(opts, NULL);
277     if (force_share) {
278         if (qdict_haskey(options, BDRV_OPT_FORCE_SHARE)
279             && strcmp(qdict_get_str(options, BDRV_OPT_FORCE_SHARE), "on")) {
280             error_report("--force-share/-U conflicts with image options");
281             qobject_unref(options);
282             return NULL;
283         }
284         qdict_put_str(options, BDRV_OPT_FORCE_SHARE, "on");
285     }
286     blk = blk_new_open(NULL, NULL, options, flags, &local_err);
287     if (!blk) {
288         error_reportf_err(local_err, "Could not open '%s': ", optstr);
289         return NULL;
290     }
291     blk_set_enable_write_cache(blk, !writethrough);
292 
293     return blk;
294 }
295 
296 static BlockBackend *img_open_file(const char *filename,
297                                    QDict *options,
298                                    const char *fmt, int flags,
299                                    bool writethrough, bool quiet,
300                                    bool force_share)
301 {
302     BlockBackend *blk;
303     Error *local_err = NULL;
304 
305     if (!options) {
306         options = qdict_new();
307     }
308     if (fmt) {
309         qdict_put_str(options, "driver", fmt);
310     }
311 
312     if (force_share) {
313         qdict_put_bool(options, BDRV_OPT_FORCE_SHARE, true);
314     }
315     blk = blk_new_open(filename, NULL, options, flags, &local_err);
316     if (!blk) {
317         error_reportf_err(local_err, "Could not open '%s': ", filename);
318         return NULL;
319     }
320     blk_set_enable_write_cache(blk, !writethrough);
321 
322     return blk;
323 }
324 
325 
326 static int img_add_key_secrets(void *opaque,
327                                const char *name, const char *value,
328                                Error **errp)
329 {
330     QDict *options = opaque;
331 
332     if (g_str_has_suffix(name, "key-secret")) {
333         qdict_put_str(options, name, value);
334     }
335 
336     return 0;
337 }
338 
339 static BlockBackend *img_open_new_file(const char *filename,
340                                        QemuOpts *create_opts,
341                                        const char *fmt, int flags,
342                                        bool writethrough, bool quiet,
343                                        bool force_share)
344 {
345     QDict *options = NULL;
346 
347     options = qdict_new();
348     qemu_opt_foreach(create_opts, img_add_key_secrets, options, &error_abort);
349 
350     return img_open_file(filename, options, fmt, flags, writethrough, quiet,
351                          force_share);
352 }
353 
354 
355 static BlockBackend *img_open(bool image_opts,
356                               const char *filename,
357                               const char *fmt, int flags, bool writethrough,
358                               bool quiet, bool force_share)
359 {
360     BlockBackend *blk;
361     if (image_opts) {
362         QemuOpts *opts;
363         if (fmt) {
364             error_report("--image-opts and --format are mutually exclusive");
365             return NULL;
366         }
367         opts = qemu_opts_parse_noisily(qemu_find_opts("source"),
368                                        filename, true);
369         if (!opts) {
370             return NULL;
371         }
372         blk = img_open_opts(filename, opts, flags, writethrough, quiet,
373                             force_share);
374     } else {
375         blk = img_open_file(filename, NULL, fmt, flags, writethrough, quiet,
376                             force_share);
377     }
378     return blk;
379 }
380 
381 
382 static int add_old_style_options(const char *fmt, QemuOpts *opts,
383                                  const char *base_filename,
384                                  const char *base_fmt)
385 {
386     Error *err = NULL;
387 
388     if (base_filename) {
389         qemu_opt_set(opts, BLOCK_OPT_BACKING_FILE, base_filename, &err);
390         if (err) {
391             error_report("Backing file not supported for file format '%s'",
392                          fmt);
393             error_free(err);
394             return -1;
395         }
396     }
397     if (base_fmt) {
398         qemu_opt_set(opts, BLOCK_OPT_BACKING_FMT, base_fmt, &err);
399         if (err) {
400             error_report("Backing file format not supported for file "
401                          "format '%s'", fmt);
402             error_free(err);
403             return -1;
404         }
405     }
406     return 0;
407 }
408 
409 static int64_t cvtnum(const char *s)
410 {
411     int err;
412     uint64_t value;
413 
414     err = qemu_strtosz(s, NULL, &value);
415     if (err < 0) {
416         return err;
417     }
418     if (value > INT64_MAX) {
419         return -ERANGE;
420     }
421     return value;
422 }
423 
424 static int img_create(int argc, char **argv)
425 {
426     int c;
427     uint64_t img_size = -1;
428     const char *fmt = "raw";
429     const char *base_fmt = NULL;
430     const char *filename;
431     const char *base_filename = NULL;
432     char *options = NULL;
433     Error *local_err = NULL;
434     bool quiet = false;
435     int flags = 0;
436 
437     for(;;) {
438         static const struct option long_options[] = {
439             {"help", no_argument, 0, 'h'},
440             {"object", required_argument, 0, OPTION_OBJECT},
441             {0, 0, 0, 0}
442         };
443         c = getopt_long(argc, argv, ":F:b:f:ho:qu",
444                         long_options, NULL);
445         if (c == -1) {
446             break;
447         }
448         switch(c) {
449         case ':':
450             missing_argument(argv[optind - 1]);
451             break;
452         case '?':
453             unrecognized_option(argv[optind - 1]);
454             break;
455         case 'h':
456             help();
457             break;
458         case 'F':
459             base_fmt = optarg;
460             break;
461         case 'b':
462             base_filename = optarg;
463             break;
464         case 'f':
465             fmt = optarg;
466             break;
467         case 'o':
468             if (!is_valid_option_list(optarg)) {
469                 error_report("Invalid option list: %s", optarg);
470                 goto fail;
471             }
472             if (!options) {
473                 options = g_strdup(optarg);
474             } else {
475                 char *old_options = options;
476                 options = g_strdup_printf("%s,%s", options, optarg);
477                 g_free(old_options);
478             }
479             break;
480         case 'q':
481             quiet = true;
482             break;
483         case 'u':
484             flags |= BDRV_O_NO_BACKING;
485             break;
486         case OPTION_OBJECT: {
487             QemuOpts *opts;
488             opts = qemu_opts_parse_noisily(&qemu_object_opts,
489                                            optarg, true);
490             if (!opts) {
491                 goto fail;
492             }
493         }   break;
494         }
495     }
496 
497     /* Get the filename */
498     filename = (optind < argc) ? argv[optind] : NULL;
499     if (options && has_help_option(options)) {
500         g_free(options);
501         return print_block_option_help(filename, fmt);
502     }
503 
504     if (optind >= argc) {
505         error_exit("Expecting image file name");
506     }
507     optind++;
508 
509     if (qemu_opts_foreach(&qemu_object_opts,
510                           user_creatable_add_opts_foreach,
511                           NULL, NULL)) {
512         goto fail;
513     }
514 
515     /* Get image size, if specified */
516     if (optind < argc) {
517         int64_t sval;
518 
519         sval = cvtnum(argv[optind++]);
520         if (sval < 0) {
521             if (sval == -ERANGE) {
522                 error_report("Image size must be less than 8 EiB!");
523             } else {
524                 error_report("Invalid image size specified! You may use k, M, "
525                       "G, T, P or E suffixes for ");
526                 error_report("kilobytes, megabytes, gigabytes, terabytes, "
527                              "petabytes and exabytes.");
528             }
529             goto fail;
530         }
531         img_size = (uint64_t)sval;
532     }
533     if (optind != argc) {
534         error_exit("Unexpected argument: %s", argv[optind]);
535     }
536 
537     bdrv_img_create(filename, fmt, base_filename, base_fmt,
538                     options, img_size, flags, quiet, &local_err);
539     if (local_err) {
540         error_reportf_err(local_err, "%s: ", filename);
541         goto fail;
542     }
543 
544     g_free(options);
545     return 0;
546 
547 fail:
548     g_free(options);
549     return 1;
550 }
551 
552 static void dump_json_image_check(ImageCheck *check, bool quiet)
553 {
554     QString *str;
555     QObject *obj;
556     Visitor *v = qobject_output_visitor_new(&obj);
557 
558     visit_type_ImageCheck(v, NULL, &check, &error_abort);
559     visit_complete(v, &obj);
560     str = qobject_to_json_pretty(obj);
561     assert(str != NULL);
562     qprintf(quiet, "%s\n", qstring_get_str(str));
563     qobject_unref(obj);
564     visit_free(v);
565     qobject_unref(str);
566 }
567 
568 static void dump_human_image_check(ImageCheck *check, bool quiet)
569 {
570     if (!(check->corruptions || check->leaks || check->check_errors)) {
571         qprintf(quiet, "No errors were found on the image.\n");
572     } else {
573         if (check->corruptions) {
574             qprintf(quiet, "\n%" PRId64 " errors were found on the image.\n"
575                     "Data may be corrupted, or further writes to the image "
576                     "may corrupt it.\n",
577                     check->corruptions);
578         }
579 
580         if (check->leaks) {
581             qprintf(quiet,
582                     "\n%" PRId64 " leaked clusters were found on the image.\n"
583                     "This means waste of disk space, but no harm to data.\n",
584                     check->leaks);
585         }
586 
587         if (check->check_errors) {
588             qprintf(quiet,
589                     "\n%" PRId64
590                     " internal errors have occurred during the check.\n",
591                     check->check_errors);
592         }
593     }
594 
595     if (check->total_clusters != 0 && check->allocated_clusters != 0) {
596         qprintf(quiet, "%" PRId64 "/%" PRId64 " = %0.2f%% allocated, "
597                 "%0.2f%% fragmented, %0.2f%% compressed clusters\n",
598                 check->allocated_clusters, check->total_clusters,
599                 check->allocated_clusters * 100.0 / check->total_clusters,
600                 check->fragmented_clusters * 100.0 / check->allocated_clusters,
601                 check->compressed_clusters * 100.0 /
602                 check->allocated_clusters);
603     }
604 
605     if (check->image_end_offset) {
606         qprintf(quiet,
607                 "Image end offset: %" PRId64 "\n", check->image_end_offset);
608     }
609 }
610 
611 static int collect_image_check(BlockDriverState *bs,
612                    ImageCheck *check,
613                    const char *filename,
614                    const char *fmt,
615                    int fix)
616 {
617     int ret;
618     BdrvCheckResult result;
619 
620     ret = bdrv_check(bs, &result, fix);
621     if (ret < 0) {
622         return ret;
623     }
624 
625     check->filename                 = g_strdup(filename);
626     check->format                   = g_strdup(bdrv_get_format_name(bs));
627     check->check_errors             = result.check_errors;
628     check->corruptions              = result.corruptions;
629     check->has_corruptions          = result.corruptions != 0;
630     check->leaks                    = result.leaks;
631     check->has_leaks                = result.leaks != 0;
632     check->corruptions_fixed        = result.corruptions_fixed;
633     check->has_corruptions_fixed    = result.corruptions != 0;
634     check->leaks_fixed              = result.leaks_fixed;
635     check->has_leaks_fixed          = result.leaks != 0;
636     check->image_end_offset         = result.image_end_offset;
637     check->has_image_end_offset     = result.image_end_offset != 0;
638     check->total_clusters           = result.bfi.total_clusters;
639     check->has_total_clusters       = result.bfi.total_clusters != 0;
640     check->allocated_clusters       = result.bfi.allocated_clusters;
641     check->has_allocated_clusters   = result.bfi.allocated_clusters != 0;
642     check->fragmented_clusters      = result.bfi.fragmented_clusters;
643     check->has_fragmented_clusters  = result.bfi.fragmented_clusters != 0;
644     check->compressed_clusters      = result.bfi.compressed_clusters;
645     check->has_compressed_clusters  = result.bfi.compressed_clusters != 0;
646 
647     return 0;
648 }
649 
650 /*
651  * Checks an image for consistency. Exit codes:
652  *
653  *  0 - Check completed, image is good
654  *  1 - Check not completed because of internal errors
655  *  2 - Check completed, image is corrupted
656  *  3 - Check completed, image has leaked clusters, but is good otherwise
657  * 63 - Checks are not supported by the image format
658  */
659 static int img_check(int argc, char **argv)
660 {
661     int c, ret;
662     OutputFormat output_format = OFORMAT_HUMAN;
663     const char *filename, *fmt, *output, *cache;
664     BlockBackend *blk;
665     BlockDriverState *bs;
666     int fix = 0;
667     int flags = BDRV_O_CHECK;
668     bool writethrough;
669     ImageCheck *check;
670     bool quiet = false;
671     bool image_opts = false;
672     bool force_share = false;
673 
674     fmt = NULL;
675     output = NULL;
676     cache = BDRV_DEFAULT_CACHE;
677 
678     for(;;) {
679         int option_index = 0;
680         static const struct option long_options[] = {
681             {"help", no_argument, 0, 'h'},
682             {"format", required_argument, 0, 'f'},
683             {"repair", required_argument, 0, 'r'},
684             {"output", required_argument, 0, OPTION_OUTPUT},
685             {"object", required_argument, 0, OPTION_OBJECT},
686             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
687             {"force-share", no_argument, 0, 'U'},
688             {0, 0, 0, 0}
689         };
690         c = getopt_long(argc, argv, ":hf:r:T:qU",
691                         long_options, &option_index);
692         if (c == -1) {
693             break;
694         }
695         switch(c) {
696         case ':':
697             missing_argument(argv[optind - 1]);
698             break;
699         case '?':
700             unrecognized_option(argv[optind - 1]);
701             break;
702         case 'h':
703             help();
704             break;
705         case 'f':
706             fmt = optarg;
707             break;
708         case 'r':
709             flags |= BDRV_O_RDWR;
710 
711             if (!strcmp(optarg, "leaks")) {
712                 fix = BDRV_FIX_LEAKS;
713             } else if (!strcmp(optarg, "all")) {
714                 fix = BDRV_FIX_LEAKS | BDRV_FIX_ERRORS;
715             } else {
716                 error_exit("Unknown option value for -r "
717                            "(expecting 'leaks' or 'all'): %s", optarg);
718             }
719             break;
720         case OPTION_OUTPUT:
721             output = optarg;
722             break;
723         case 'T':
724             cache = optarg;
725             break;
726         case 'q':
727             quiet = true;
728             break;
729         case 'U':
730             force_share = true;
731             break;
732         case OPTION_OBJECT: {
733             QemuOpts *opts;
734             opts = qemu_opts_parse_noisily(&qemu_object_opts,
735                                            optarg, true);
736             if (!opts) {
737                 return 1;
738             }
739         }   break;
740         case OPTION_IMAGE_OPTS:
741             image_opts = true;
742             break;
743         }
744     }
745     if (optind != argc - 1) {
746         error_exit("Expecting one image file name");
747     }
748     filename = argv[optind++];
749 
750     if (output && !strcmp(output, "json")) {
751         output_format = OFORMAT_JSON;
752     } else if (output && !strcmp(output, "human")) {
753         output_format = OFORMAT_HUMAN;
754     } else if (output) {
755         error_report("--output must be used with human or json as argument.");
756         return 1;
757     }
758 
759     if (qemu_opts_foreach(&qemu_object_opts,
760                           user_creatable_add_opts_foreach,
761                           NULL, NULL)) {
762         return 1;
763     }
764 
765     ret = bdrv_parse_cache_mode(cache, &flags, &writethrough);
766     if (ret < 0) {
767         error_report("Invalid source cache option: %s", cache);
768         return 1;
769     }
770 
771     blk = img_open(image_opts, filename, fmt, flags, writethrough, quiet,
772                    force_share);
773     if (!blk) {
774         return 1;
775     }
776     bs = blk_bs(blk);
777 
778     check = g_new0(ImageCheck, 1);
779     ret = collect_image_check(bs, check, filename, fmt, fix);
780 
781     if (ret == -ENOTSUP) {
782         error_report("This image format does not support checks");
783         ret = 63;
784         goto fail;
785     }
786 
787     if (check->corruptions_fixed || check->leaks_fixed) {
788         int corruptions_fixed, leaks_fixed;
789 
790         leaks_fixed         = check->leaks_fixed;
791         corruptions_fixed   = check->corruptions_fixed;
792 
793         if (output_format == OFORMAT_HUMAN) {
794             qprintf(quiet,
795                     "The following inconsistencies were found and repaired:\n\n"
796                     "    %" PRId64 " leaked clusters\n"
797                     "    %" PRId64 " corruptions\n\n"
798                     "Double checking the fixed image now...\n",
799                     check->leaks_fixed,
800                     check->corruptions_fixed);
801         }
802 
803         ret = collect_image_check(bs, check, filename, fmt, 0);
804 
805         check->leaks_fixed          = leaks_fixed;
806         check->corruptions_fixed    = corruptions_fixed;
807     }
808 
809     if (!ret) {
810         switch (output_format) {
811         case OFORMAT_HUMAN:
812             dump_human_image_check(check, quiet);
813             break;
814         case OFORMAT_JSON:
815             dump_json_image_check(check, quiet);
816             break;
817         }
818     }
819 
820     if (ret || check->check_errors) {
821         if (ret) {
822             error_report("Check failed: %s", strerror(-ret));
823         } else {
824             error_report("Check failed");
825         }
826         ret = 1;
827         goto fail;
828     }
829 
830     if (check->corruptions) {
831         ret = 2;
832     } else if (check->leaks) {
833         ret = 3;
834     } else {
835         ret = 0;
836     }
837 
838 fail:
839     qapi_free_ImageCheck(check);
840     blk_unref(blk);
841     return ret;
842 }
843 
844 typedef struct CommonBlockJobCBInfo {
845     BlockDriverState *bs;
846     Error **errp;
847 } CommonBlockJobCBInfo;
848 
849 static void common_block_job_cb(void *opaque, int ret)
850 {
851     CommonBlockJobCBInfo *cbi = opaque;
852 
853     if (ret < 0) {
854         error_setg_errno(cbi->errp, -ret, "Block job failed");
855     }
856 }
857 
858 static void run_block_job(BlockJob *job, Error **errp)
859 {
860     AioContext *aio_context = blk_get_aio_context(job->blk);
861     int ret = 0;
862 
863     aio_context_acquire(aio_context);
864     job_ref(&job->job);
865     do {
866         float progress = 0.0f;
867         aio_poll(aio_context, true);
868         if (job->job.progress_total) {
869             progress = (float)job->job.progress_current /
870                        job->job.progress_total * 100.f;
871         }
872         qemu_progress_print(progress, 0);
873     } while (!job_is_ready(&job->job) && !job_is_completed(&job->job));
874 
875     if (!job_is_completed(&job->job)) {
876         ret = job_complete_sync(&job->job, errp);
877     } else {
878         ret = job->job.ret;
879     }
880     job_unref(&job->job);
881     aio_context_release(aio_context);
882 
883     /* publish completion progress only when success */
884     if (!ret) {
885         qemu_progress_print(100.f, 0);
886     }
887 }
888 
889 static int img_commit(int argc, char **argv)
890 {
891     int c, ret, flags;
892     const char *filename, *fmt, *cache, *base;
893     BlockBackend *blk;
894     BlockDriverState *bs, *base_bs;
895     BlockJob *job;
896     bool progress = false, quiet = false, drop = false;
897     bool writethrough;
898     Error *local_err = NULL;
899     CommonBlockJobCBInfo cbi;
900     bool image_opts = false;
901     AioContext *aio_context;
902 
903     fmt = NULL;
904     cache = BDRV_DEFAULT_CACHE;
905     base = NULL;
906     for(;;) {
907         static const struct option long_options[] = {
908             {"help", no_argument, 0, 'h'},
909             {"object", required_argument, 0, OPTION_OBJECT},
910             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
911             {0, 0, 0, 0}
912         };
913         c = getopt_long(argc, argv, ":f:ht:b:dpq",
914                         long_options, NULL);
915         if (c == -1) {
916             break;
917         }
918         switch(c) {
919         case ':':
920             missing_argument(argv[optind - 1]);
921             break;
922         case '?':
923             unrecognized_option(argv[optind - 1]);
924             break;
925         case 'h':
926             help();
927             break;
928         case 'f':
929             fmt = optarg;
930             break;
931         case 't':
932             cache = optarg;
933             break;
934         case 'b':
935             base = optarg;
936             /* -b implies -d */
937             drop = true;
938             break;
939         case 'd':
940             drop = true;
941             break;
942         case 'p':
943             progress = true;
944             break;
945         case 'q':
946             quiet = true;
947             break;
948         case OPTION_OBJECT: {
949             QemuOpts *opts;
950             opts = qemu_opts_parse_noisily(&qemu_object_opts,
951                                            optarg, true);
952             if (!opts) {
953                 return 1;
954             }
955         }   break;
956         case OPTION_IMAGE_OPTS:
957             image_opts = true;
958             break;
959         }
960     }
961 
962     /* Progress is not shown in Quiet mode */
963     if (quiet) {
964         progress = false;
965     }
966 
967     if (optind != argc - 1) {
968         error_exit("Expecting one image file name");
969     }
970     filename = argv[optind++];
971 
972     if (qemu_opts_foreach(&qemu_object_opts,
973                           user_creatable_add_opts_foreach,
974                           NULL, NULL)) {
975         return 1;
976     }
977 
978     flags = BDRV_O_RDWR | BDRV_O_UNMAP;
979     ret = bdrv_parse_cache_mode(cache, &flags, &writethrough);
980     if (ret < 0) {
981         error_report("Invalid cache option: %s", cache);
982         return 1;
983     }
984 
985     blk = img_open(image_opts, filename, fmt, flags, writethrough, quiet,
986                    false);
987     if (!blk) {
988         return 1;
989     }
990     bs = blk_bs(blk);
991 
992     qemu_progress_init(progress, 1.f);
993     qemu_progress_print(0.f, 100);
994 
995     if (base) {
996         base_bs = bdrv_find_backing_image(bs, base);
997         if (!base_bs) {
998             error_setg(&local_err,
999                        "Did not find '%s' in the backing chain of '%s'",
1000                        base, filename);
1001             goto done;
1002         }
1003     } else {
1004         /* This is different from QMP, which by default uses the deepest file in
1005          * the backing chain (i.e., the very base); however, the traditional
1006          * behavior of qemu-img commit is using the immediate backing file. */
1007         base_bs = backing_bs(bs);
1008         if (!base_bs) {
1009             error_setg(&local_err, "Image does not have a backing file");
1010             goto done;
1011         }
1012     }
1013 
1014     cbi = (CommonBlockJobCBInfo){
1015         .errp = &local_err,
1016         .bs   = bs,
1017     };
1018 
1019     aio_context = bdrv_get_aio_context(bs);
1020     aio_context_acquire(aio_context);
1021     commit_active_start("commit", bs, base_bs, JOB_DEFAULT, 0,
1022                         BLOCKDEV_ON_ERROR_REPORT, NULL, common_block_job_cb,
1023                         &cbi, false, &local_err);
1024     aio_context_release(aio_context);
1025     if (local_err) {
1026         goto done;
1027     }
1028 
1029     /* When the block job completes, the BlockBackend reference will point to
1030      * the old backing file. In order to avoid that the top image is already
1031      * deleted, so we can still empty it afterwards, increment the reference
1032      * counter here preemptively. */
1033     if (!drop) {
1034         bdrv_ref(bs);
1035     }
1036 
1037     job = block_job_get("commit");
1038     run_block_job(job, &local_err);
1039     if (local_err) {
1040         goto unref_backing;
1041     }
1042 
1043     if (!drop && bs->drv->bdrv_make_empty) {
1044         ret = bs->drv->bdrv_make_empty(bs);
1045         if (ret) {
1046             error_setg_errno(&local_err, -ret, "Could not empty %s",
1047                              filename);
1048             goto unref_backing;
1049         }
1050     }
1051 
1052 unref_backing:
1053     if (!drop) {
1054         bdrv_unref(bs);
1055     }
1056 
1057 done:
1058     qemu_progress_end();
1059 
1060     blk_unref(blk);
1061 
1062     if (local_err) {
1063         error_report_err(local_err);
1064         return 1;
1065     }
1066 
1067     qprintf(quiet, "Image committed.\n");
1068     return 0;
1069 }
1070 
1071 /*
1072  * Returns -1 if 'buf' contains only zeroes, otherwise the byte index
1073  * of the first sector boundary within buf where the sector contains a
1074  * non-zero byte.  This function is robust to a buffer that is not
1075  * sector-aligned.
1076  */
1077 static int64_t find_nonzero(const uint8_t *buf, int64_t n)
1078 {
1079     int64_t i;
1080     int64_t end = QEMU_ALIGN_DOWN(n, BDRV_SECTOR_SIZE);
1081 
1082     for (i = 0; i < end; i += BDRV_SECTOR_SIZE) {
1083         if (!buffer_is_zero(buf + i, BDRV_SECTOR_SIZE)) {
1084             return i;
1085         }
1086     }
1087     if (i < n && !buffer_is_zero(buf + i, n - end)) {
1088         return i;
1089     }
1090     return -1;
1091 }
1092 
1093 /*
1094  * Returns true iff the first sector pointed to by 'buf' contains at least
1095  * a non-NUL byte.
1096  *
1097  * 'pnum' is set to the number of sectors (including and immediately following
1098  * the first one) that are known to be in the same allocated/unallocated state.
1099  */
1100 static int is_allocated_sectors(const uint8_t *buf, int n, int *pnum)
1101 {
1102     bool is_zero;
1103     int i;
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     *pnum = i;
1117     return !is_zero;
1118 }
1119 
1120 /*
1121  * Like is_allocated_sectors, but if the buffer starts with a used sector,
1122  * up to 'min' consecutive sectors containing zeros are ignored. This avoids
1123  * breaking up write requests for only small sparse areas.
1124  */
1125 static int is_allocated_sectors_min(const uint8_t *buf, int n, int *pnum,
1126     int min)
1127 {
1128     int ret;
1129     int num_checked, num_used;
1130 
1131     if (n < min) {
1132         min = n;
1133     }
1134 
1135     ret = is_allocated_sectors(buf, n, pnum);
1136     if (!ret) {
1137         return ret;
1138     }
1139 
1140     num_used = *pnum;
1141     buf += BDRV_SECTOR_SIZE * *pnum;
1142     n -= *pnum;
1143     num_checked = num_used;
1144 
1145     while (n > 0) {
1146         ret = is_allocated_sectors(buf, n, pnum);
1147 
1148         buf += BDRV_SECTOR_SIZE * *pnum;
1149         n -= *pnum;
1150         num_checked += *pnum;
1151         if (ret) {
1152             num_used = num_checked;
1153         } else if (*pnum >= min) {
1154             break;
1155         }
1156     }
1157 
1158     *pnum = num_used;
1159     return 1;
1160 }
1161 
1162 /*
1163  * Compares two buffers sector by sector. Returns 0 if the first
1164  * sector of each buffer matches, non-zero otherwise.
1165  *
1166  * pnum is set to the sector-aligned size of the buffer prefix that
1167  * has the same matching status as the first sector.
1168  */
1169 static int compare_buffers(const uint8_t *buf1, const uint8_t *buf2,
1170                            int64_t bytes, int64_t *pnum)
1171 {
1172     bool res;
1173     int64_t i = MIN(bytes, BDRV_SECTOR_SIZE);
1174 
1175     assert(bytes > 0);
1176 
1177     res = !!memcmp(buf1, buf2, i);
1178     while (i < bytes) {
1179         int64_t len = MIN(bytes - i, BDRV_SECTOR_SIZE);
1180 
1181         if (!!memcmp(buf1 + i, buf2 + i, len) != res) {
1182             break;
1183         }
1184         i += len;
1185     }
1186 
1187     *pnum = i;
1188     return res;
1189 }
1190 
1191 #define IO_BUF_SIZE (2 * 1024 * 1024)
1192 
1193 /*
1194  * Check if passed sectors are empty (not allocated or contain only 0 bytes)
1195  *
1196  * Intended for use by 'qemu-img compare': Returns 0 in case sectors are
1197  * filled with 0, 1 if sectors contain non-zero data (this is a comparison
1198  * failure), and 4 on error (the exit status for read errors), after emitting
1199  * an error message.
1200  *
1201  * @param blk:  BlockBackend for the image
1202  * @param offset: Starting offset to check
1203  * @param bytes: Number of bytes to check
1204  * @param filename: Name of disk file we are checking (logging purpose)
1205  * @param buffer: Allocated buffer for storing read data
1206  * @param quiet: Flag for quiet mode
1207  */
1208 static int check_empty_sectors(BlockBackend *blk, int64_t offset,
1209                                int64_t bytes, const char *filename,
1210                                uint8_t *buffer, bool quiet)
1211 {
1212     int ret = 0;
1213     int64_t idx;
1214 
1215     ret = blk_pread(blk, offset, buffer, bytes);
1216     if (ret < 0) {
1217         error_report("Error while reading offset %" PRId64 " of %s: %s",
1218                      offset, filename, strerror(-ret));
1219         return 4;
1220     }
1221     idx = find_nonzero(buffer, bytes);
1222     if (idx >= 0) {
1223         qprintf(quiet, "Content mismatch at offset %" PRId64 "!\n",
1224                 offset + idx);
1225         return 1;
1226     }
1227 
1228     return 0;
1229 }
1230 
1231 /*
1232  * Compares two images. Exit codes:
1233  *
1234  * 0 - Images are identical
1235  * 1 - Images differ
1236  * >1 - Error occurred
1237  */
1238 static int img_compare(int argc, char **argv)
1239 {
1240     const char *fmt1 = NULL, *fmt2 = NULL, *cache, *filename1, *filename2;
1241     BlockBackend *blk1, *blk2;
1242     BlockDriverState *bs1, *bs2;
1243     int64_t total_size1, total_size2;
1244     uint8_t *buf1 = NULL, *buf2 = NULL;
1245     int64_t pnum1, pnum2;
1246     int allocated1, allocated2;
1247     int ret = 0; /* return value - 0 Ident, 1 Different, >1 Error */
1248     bool progress = false, quiet = false, strict = false;
1249     int flags;
1250     bool writethrough;
1251     int64_t total_size;
1252     int64_t offset = 0;
1253     int64_t chunk;
1254     int c;
1255     uint64_t progress_base;
1256     bool image_opts = false;
1257     bool force_share = false;
1258 
1259     cache = BDRV_DEFAULT_CACHE;
1260     for (;;) {
1261         static const struct option long_options[] = {
1262             {"help", no_argument, 0, 'h'},
1263             {"object", required_argument, 0, OPTION_OBJECT},
1264             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
1265             {"force-share", no_argument, 0, 'U'},
1266             {0, 0, 0, 0}
1267         };
1268         c = getopt_long(argc, argv, ":hf:F:T:pqsU",
1269                         long_options, NULL);
1270         if (c == -1) {
1271             break;
1272         }
1273         switch (c) {
1274         case ':':
1275             missing_argument(argv[optind - 1]);
1276             break;
1277         case '?':
1278             unrecognized_option(argv[optind - 1]);
1279             break;
1280         case 'h':
1281             help();
1282             break;
1283         case 'f':
1284             fmt1 = optarg;
1285             break;
1286         case 'F':
1287             fmt2 = optarg;
1288             break;
1289         case 'T':
1290             cache = optarg;
1291             break;
1292         case 'p':
1293             progress = true;
1294             break;
1295         case 'q':
1296             quiet = true;
1297             break;
1298         case 's':
1299             strict = true;
1300             break;
1301         case 'U':
1302             force_share = true;
1303             break;
1304         case OPTION_OBJECT: {
1305             QemuOpts *opts;
1306             opts = qemu_opts_parse_noisily(&qemu_object_opts,
1307                                            optarg, true);
1308             if (!opts) {
1309                 ret = 2;
1310                 goto out4;
1311             }
1312         }   break;
1313         case OPTION_IMAGE_OPTS:
1314             image_opts = true;
1315             break;
1316         }
1317     }
1318 
1319     /* Progress is not shown in Quiet mode */
1320     if (quiet) {
1321         progress = false;
1322     }
1323 
1324 
1325     if (optind != argc - 2) {
1326         error_exit("Expecting two image file names");
1327     }
1328     filename1 = argv[optind++];
1329     filename2 = argv[optind++];
1330 
1331     if (qemu_opts_foreach(&qemu_object_opts,
1332                           user_creatable_add_opts_foreach,
1333                           NULL, NULL)) {
1334         ret = 2;
1335         goto out4;
1336     }
1337 
1338     /* Initialize before goto out */
1339     qemu_progress_init(progress, 2.0);
1340 
1341     flags = 0;
1342     ret = bdrv_parse_cache_mode(cache, &flags, &writethrough);
1343     if (ret < 0) {
1344         error_report("Invalid source cache option: %s", cache);
1345         ret = 2;
1346         goto out3;
1347     }
1348 
1349     blk1 = img_open(image_opts, filename1, fmt1, flags, writethrough, quiet,
1350                     force_share);
1351     if (!blk1) {
1352         ret = 2;
1353         goto out3;
1354     }
1355 
1356     blk2 = img_open(image_opts, filename2, fmt2, flags, writethrough, quiet,
1357                     force_share);
1358     if (!blk2) {
1359         ret = 2;
1360         goto out2;
1361     }
1362     bs1 = blk_bs(blk1);
1363     bs2 = blk_bs(blk2);
1364 
1365     buf1 = blk_blockalign(blk1, IO_BUF_SIZE);
1366     buf2 = blk_blockalign(blk2, IO_BUF_SIZE);
1367     total_size1 = blk_getlength(blk1);
1368     if (total_size1 < 0) {
1369         error_report("Can't get size of %s: %s",
1370                      filename1, strerror(-total_size1));
1371         ret = 4;
1372         goto out;
1373     }
1374     total_size2 = blk_getlength(blk2);
1375     if (total_size2 < 0) {
1376         error_report("Can't get size of %s: %s",
1377                      filename2, strerror(-total_size2));
1378         ret = 4;
1379         goto out;
1380     }
1381     total_size = MIN(total_size1, total_size2);
1382     progress_base = MAX(total_size1, total_size2);
1383 
1384     qemu_progress_print(0, 100);
1385 
1386     if (strict && total_size1 != total_size2) {
1387         ret = 1;
1388         qprintf(quiet, "Strict mode: Image size mismatch!\n");
1389         goto out;
1390     }
1391 
1392     while (offset < total_size) {
1393         int status1, status2;
1394 
1395         status1 = bdrv_block_status_above(bs1, NULL, offset,
1396                                           total_size1 - offset, &pnum1, NULL,
1397                                           NULL);
1398         if (status1 < 0) {
1399             ret = 3;
1400             error_report("Sector allocation test failed for %s", filename1);
1401             goto out;
1402         }
1403         allocated1 = status1 & BDRV_BLOCK_ALLOCATED;
1404 
1405         status2 = bdrv_block_status_above(bs2, NULL, offset,
1406                                           total_size2 - offset, &pnum2, NULL,
1407                                           NULL);
1408         if (status2 < 0) {
1409             ret = 3;
1410             error_report("Sector allocation test failed for %s", filename2);
1411             goto out;
1412         }
1413         allocated2 = status2 & BDRV_BLOCK_ALLOCATED;
1414 
1415         assert(pnum1 && pnum2);
1416         chunk = MIN(pnum1, pnum2);
1417 
1418         if (strict) {
1419             if (status1 != status2) {
1420                 ret = 1;
1421                 qprintf(quiet, "Strict mode: Offset %" PRId64
1422                         " block status mismatch!\n", offset);
1423                 goto out;
1424             }
1425         }
1426         if ((status1 & BDRV_BLOCK_ZERO) && (status2 & BDRV_BLOCK_ZERO)) {
1427             /* nothing to do */
1428         } else if (allocated1 == allocated2) {
1429             if (allocated1) {
1430                 int64_t pnum;
1431 
1432                 chunk = MIN(chunk, IO_BUF_SIZE);
1433                 ret = blk_pread(blk1, offset, buf1, chunk);
1434                 if (ret < 0) {
1435                     error_report("Error while reading offset %" PRId64
1436                                  " of %s: %s",
1437                                  offset, filename1, strerror(-ret));
1438                     ret = 4;
1439                     goto out;
1440                 }
1441                 ret = blk_pread(blk2, offset, buf2, chunk);
1442                 if (ret < 0) {
1443                     error_report("Error while reading offset %" PRId64
1444                                  " of %s: %s",
1445                                  offset, filename2, strerror(-ret));
1446                     ret = 4;
1447                     goto out;
1448                 }
1449                 ret = compare_buffers(buf1, buf2, chunk, &pnum);
1450                 if (ret || pnum != chunk) {
1451                     qprintf(quiet, "Content mismatch at offset %" PRId64 "!\n",
1452                             offset + (ret ? 0 : pnum));
1453                     ret = 1;
1454                     goto out;
1455                 }
1456             }
1457         } else {
1458             chunk = MIN(chunk, IO_BUF_SIZE);
1459             if (allocated1) {
1460                 ret = check_empty_sectors(blk1, offset, chunk,
1461                                           filename1, buf1, quiet);
1462             } else {
1463                 ret = check_empty_sectors(blk2, offset, chunk,
1464                                           filename2, buf1, quiet);
1465             }
1466             if (ret) {
1467                 goto out;
1468             }
1469         }
1470         offset += chunk;
1471         qemu_progress_print(((float) chunk / progress_base) * 100, 100);
1472     }
1473 
1474     if (total_size1 != total_size2) {
1475         BlockBackend *blk_over;
1476         const char *filename_over;
1477 
1478         qprintf(quiet, "Warning: Image size mismatch!\n");
1479         if (total_size1 > total_size2) {
1480             blk_over = blk1;
1481             filename_over = filename1;
1482         } else {
1483             blk_over = blk2;
1484             filename_over = filename2;
1485         }
1486 
1487         while (offset < progress_base) {
1488             ret = bdrv_block_status_above(blk_bs(blk_over), NULL, offset,
1489                                           progress_base - offset, &chunk,
1490                                           NULL, NULL);
1491             if (ret < 0) {
1492                 ret = 3;
1493                 error_report("Sector allocation test failed for %s",
1494                              filename_over);
1495                 goto out;
1496 
1497             }
1498             if (ret & BDRV_BLOCK_ALLOCATED && !(ret & BDRV_BLOCK_ZERO)) {
1499                 chunk = MIN(chunk, IO_BUF_SIZE);
1500                 ret = check_empty_sectors(blk_over, offset, chunk,
1501                                           filename_over, buf1, quiet);
1502                 if (ret) {
1503                     goto out;
1504                 }
1505             }
1506             offset += chunk;
1507             qemu_progress_print(((float) chunk / progress_base) * 100, 100);
1508         }
1509     }
1510 
1511     qprintf(quiet, "Images are identical.\n");
1512     ret = 0;
1513 
1514 out:
1515     qemu_vfree(buf1);
1516     qemu_vfree(buf2);
1517     blk_unref(blk2);
1518 out2:
1519     blk_unref(blk1);
1520 out3:
1521     qemu_progress_end();
1522 out4:
1523     return ret;
1524 }
1525 
1526 enum ImgConvertBlockStatus {
1527     BLK_DATA,
1528     BLK_ZERO,
1529     BLK_BACKING_FILE,
1530 };
1531 
1532 #define MAX_COROUTINES 16
1533 
1534 typedef struct ImgConvertState {
1535     BlockBackend **src;
1536     int64_t *src_sectors;
1537     int src_num;
1538     int64_t total_sectors;
1539     int64_t allocated_sectors;
1540     int64_t allocated_done;
1541     int64_t sector_num;
1542     int64_t wr_offs;
1543     enum ImgConvertBlockStatus status;
1544     int64_t sector_next_status;
1545     BlockBackend *target;
1546     bool has_zero_init;
1547     bool compressed;
1548     bool target_has_backing;
1549     bool wr_in_order;
1550     bool copy_range;
1551     int min_sparse;
1552     size_t cluster_sectors;
1553     size_t buf_sectors;
1554     long num_coroutines;
1555     int running_coroutines;
1556     Coroutine *co[MAX_COROUTINES];
1557     int64_t wait_sector_num[MAX_COROUTINES];
1558     CoMutex lock;
1559     int ret;
1560 } ImgConvertState;
1561 
1562 static void convert_select_part(ImgConvertState *s, int64_t sector_num,
1563                                 int *src_cur, int64_t *src_cur_offset)
1564 {
1565     *src_cur = 0;
1566     *src_cur_offset = 0;
1567     while (sector_num - *src_cur_offset >= s->src_sectors[*src_cur]) {
1568         *src_cur_offset += s->src_sectors[*src_cur];
1569         (*src_cur)++;
1570         assert(*src_cur < s->src_num);
1571     }
1572 }
1573 
1574 static int convert_iteration_sectors(ImgConvertState *s, int64_t sector_num)
1575 {
1576     int64_t src_cur_offset;
1577     int ret, n, src_cur;
1578 
1579     convert_select_part(s, sector_num, &src_cur, &src_cur_offset);
1580 
1581     assert(s->total_sectors > sector_num);
1582     n = MIN(s->total_sectors - sector_num, BDRV_REQUEST_MAX_SECTORS);
1583 
1584     if (s->sector_next_status <= sector_num) {
1585         int64_t count = n * BDRV_SECTOR_SIZE;
1586 
1587         if (s->target_has_backing) {
1588 
1589             ret = bdrv_block_status(blk_bs(s->src[src_cur]),
1590                                     (sector_num - src_cur_offset) *
1591                                     BDRV_SECTOR_SIZE,
1592                                     count, &count, NULL, NULL);
1593         } else {
1594             ret = bdrv_block_status_above(blk_bs(s->src[src_cur]), NULL,
1595                                           (sector_num - src_cur_offset) *
1596                                           BDRV_SECTOR_SIZE,
1597                                           count, &count, NULL, NULL);
1598         }
1599         if (ret < 0) {
1600             return ret;
1601         }
1602         n = DIV_ROUND_UP(count, BDRV_SECTOR_SIZE);
1603 
1604         if (ret & BDRV_BLOCK_ZERO) {
1605             s->status = BLK_ZERO;
1606         } else if (ret & BDRV_BLOCK_DATA) {
1607             s->status = BLK_DATA;
1608         } else {
1609             s->status = s->target_has_backing ? BLK_BACKING_FILE : BLK_DATA;
1610         }
1611 
1612         s->sector_next_status = sector_num + n;
1613     }
1614 
1615     n = MIN(n, s->sector_next_status - sector_num);
1616     if (s->status == BLK_DATA) {
1617         n = MIN(n, s->buf_sectors);
1618     }
1619 
1620     /* We need to write complete clusters for compressed images, so if an
1621      * unallocated area is shorter than that, we must consider the whole
1622      * cluster allocated. */
1623     if (s->compressed) {
1624         if (n < s->cluster_sectors) {
1625             n = MIN(s->cluster_sectors, s->total_sectors - sector_num);
1626             s->status = BLK_DATA;
1627         } else {
1628             n = QEMU_ALIGN_DOWN(n, s->cluster_sectors);
1629         }
1630     }
1631 
1632     return n;
1633 }
1634 
1635 static int coroutine_fn convert_co_read(ImgConvertState *s, int64_t sector_num,
1636                                         int nb_sectors, uint8_t *buf)
1637 {
1638     int n, ret;
1639     QEMUIOVector qiov;
1640     struct iovec iov;
1641 
1642     assert(nb_sectors <= s->buf_sectors);
1643     while (nb_sectors > 0) {
1644         BlockBackend *blk;
1645         int src_cur;
1646         int64_t bs_sectors, src_cur_offset;
1647 
1648         /* In the case of compression with multiple source files, we can get a
1649          * nb_sectors that spreads into the next part. So we must be able to
1650          * read across multiple BDSes for one convert_read() call. */
1651         convert_select_part(s, sector_num, &src_cur, &src_cur_offset);
1652         blk = s->src[src_cur];
1653         bs_sectors = s->src_sectors[src_cur];
1654 
1655         n = MIN(nb_sectors, bs_sectors - (sector_num - src_cur_offset));
1656         iov.iov_base = buf;
1657         iov.iov_len = n << BDRV_SECTOR_BITS;
1658         qemu_iovec_init_external(&qiov, &iov, 1);
1659 
1660         ret = blk_co_preadv(
1661                 blk, (sector_num - src_cur_offset) << BDRV_SECTOR_BITS,
1662                 n << BDRV_SECTOR_BITS, &qiov, 0);
1663         if (ret < 0) {
1664             return ret;
1665         }
1666 
1667         sector_num += n;
1668         nb_sectors -= n;
1669         buf += n * BDRV_SECTOR_SIZE;
1670     }
1671 
1672     return 0;
1673 }
1674 
1675 
1676 static int coroutine_fn convert_co_write(ImgConvertState *s, int64_t sector_num,
1677                                          int nb_sectors, uint8_t *buf,
1678                                          enum ImgConvertBlockStatus status)
1679 {
1680     int ret;
1681     QEMUIOVector qiov;
1682     struct iovec iov;
1683 
1684     while (nb_sectors > 0) {
1685         int n = nb_sectors;
1686         BdrvRequestFlags flags = s->compressed ? BDRV_REQ_WRITE_COMPRESSED : 0;
1687 
1688         switch (status) {
1689         case BLK_BACKING_FILE:
1690             /* If we have a backing file, leave clusters unallocated that are
1691              * unallocated in the source image, so that the backing file is
1692              * visible at the respective offset. */
1693             assert(s->target_has_backing);
1694             break;
1695 
1696         case BLK_DATA:
1697             /* If we're told to keep the target fully allocated (-S 0) or there
1698              * is real non-zero data, we must write it. Otherwise we can treat
1699              * it as zero sectors.
1700              * Compressed clusters need to be written as a whole, so in that
1701              * case we can only save the write if the buffer is completely
1702              * zeroed. */
1703             if (!s->min_sparse ||
1704                 (!s->compressed &&
1705                  is_allocated_sectors_min(buf, n, &n, s->min_sparse)) ||
1706                 (s->compressed &&
1707                  !buffer_is_zero(buf, n * BDRV_SECTOR_SIZE)))
1708             {
1709                 iov.iov_base = buf;
1710                 iov.iov_len = n << BDRV_SECTOR_BITS;
1711                 qemu_iovec_init_external(&qiov, &iov, 1);
1712 
1713                 ret = blk_co_pwritev(s->target, sector_num << BDRV_SECTOR_BITS,
1714                                      n << BDRV_SECTOR_BITS, &qiov, flags);
1715                 if (ret < 0) {
1716                     return ret;
1717                 }
1718                 break;
1719             }
1720             /* fall-through */
1721 
1722         case BLK_ZERO:
1723             if (s->has_zero_init) {
1724                 assert(!s->target_has_backing);
1725                 break;
1726             }
1727             ret = blk_co_pwrite_zeroes(s->target,
1728                                        sector_num << BDRV_SECTOR_BITS,
1729                                        n << BDRV_SECTOR_BITS, 0);
1730             if (ret < 0) {
1731                 return ret;
1732             }
1733             break;
1734         }
1735 
1736         sector_num += n;
1737         nb_sectors -= n;
1738         buf += n * BDRV_SECTOR_SIZE;
1739     }
1740 
1741     return 0;
1742 }
1743 
1744 static int coroutine_fn convert_co_copy_range(ImgConvertState *s, int64_t sector_num,
1745                                               int nb_sectors)
1746 {
1747     int n, ret;
1748 
1749     while (nb_sectors > 0) {
1750         BlockBackend *blk;
1751         int src_cur;
1752         int64_t bs_sectors, src_cur_offset;
1753         int64_t offset;
1754 
1755         convert_select_part(s, sector_num, &src_cur, &src_cur_offset);
1756         offset = (sector_num - src_cur_offset) << BDRV_SECTOR_BITS;
1757         blk = s->src[src_cur];
1758         bs_sectors = s->src_sectors[src_cur];
1759 
1760         n = MIN(nb_sectors, bs_sectors - (sector_num - src_cur_offset));
1761 
1762         ret = blk_co_copy_range(blk, offset, s->target,
1763                                 sector_num << BDRV_SECTOR_BITS,
1764                                 n << BDRV_SECTOR_BITS, 0);
1765         if (ret < 0) {
1766             return ret;
1767         }
1768 
1769         sector_num += n;
1770         nb_sectors -= n;
1771     }
1772     return 0;
1773 }
1774 
1775 static void coroutine_fn convert_co_do_copy(void *opaque)
1776 {
1777     ImgConvertState *s = opaque;
1778     uint8_t *buf = NULL;
1779     int ret, i;
1780     int index = -1;
1781 
1782     for (i = 0; i < s->num_coroutines; i++) {
1783         if (s->co[i] == qemu_coroutine_self()) {
1784             index = i;
1785             break;
1786         }
1787     }
1788     assert(index >= 0);
1789 
1790     s->running_coroutines++;
1791     buf = blk_blockalign(s->target, s->buf_sectors * BDRV_SECTOR_SIZE);
1792 
1793     while (1) {
1794         int n;
1795         int64_t sector_num;
1796         enum ImgConvertBlockStatus status;
1797         bool copy_range;
1798 
1799         qemu_co_mutex_lock(&s->lock);
1800         if (s->ret != -EINPROGRESS || s->sector_num >= s->total_sectors) {
1801             qemu_co_mutex_unlock(&s->lock);
1802             break;
1803         }
1804         n = convert_iteration_sectors(s, s->sector_num);
1805         if (n < 0) {
1806             qemu_co_mutex_unlock(&s->lock);
1807             s->ret = n;
1808             break;
1809         }
1810         /* save current sector and allocation status to local variables */
1811         sector_num = s->sector_num;
1812         status = s->status;
1813         if (!s->min_sparse && s->status == BLK_ZERO) {
1814             n = MIN(n, s->buf_sectors);
1815         }
1816         /* increment global sector counter so that other coroutines can
1817          * already continue reading beyond this request */
1818         s->sector_num += n;
1819         qemu_co_mutex_unlock(&s->lock);
1820 
1821         if (status == BLK_DATA || (!s->min_sparse && status == BLK_ZERO)) {
1822             s->allocated_done += n;
1823             qemu_progress_print(100.0 * s->allocated_done /
1824                                         s->allocated_sectors, 0);
1825         }
1826 
1827 retry:
1828         copy_range = s->copy_range && s->status == BLK_DATA;
1829         if (status == BLK_DATA && !copy_range) {
1830             ret = convert_co_read(s, sector_num, n, buf);
1831             if (ret < 0) {
1832                 error_report("error while reading sector %" PRId64
1833                              ": %s", sector_num, strerror(-ret));
1834                 s->ret = ret;
1835             }
1836         } else if (!s->min_sparse && status == BLK_ZERO) {
1837             status = BLK_DATA;
1838             memset(buf, 0x00, n * BDRV_SECTOR_SIZE);
1839         }
1840 
1841         if (s->wr_in_order) {
1842             /* keep writes in order */
1843             while (s->wr_offs != sector_num && s->ret == -EINPROGRESS) {
1844                 s->wait_sector_num[index] = sector_num;
1845                 qemu_coroutine_yield();
1846             }
1847             s->wait_sector_num[index] = -1;
1848         }
1849 
1850         if (s->ret == -EINPROGRESS) {
1851             if (copy_range) {
1852                 ret = convert_co_copy_range(s, sector_num, n);
1853                 if (ret) {
1854                     s->copy_range = false;
1855                     goto retry;
1856                 }
1857             } else {
1858                 ret = convert_co_write(s, sector_num, n, buf, status);
1859             }
1860             if (ret < 0) {
1861                 error_report("error while writing sector %" PRId64
1862                              ": %s", sector_num, strerror(-ret));
1863                 s->ret = ret;
1864             }
1865         }
1866 
1867         if (s->wr_in_order) {
1868             /* reenter the coroutine that might have waited
1869              * for this write to complete */
1870             s->wr_offs = sector_num + n;
1871             for (i = 0; i < s->num_coroutines; i++) {
1872                 if (s->co[i] && s->wait_sector_num[i] == s->wr_offs) {
1873                     /*
1874                      * A -> B -> A cannot occur because A has
1875                      * s->wait_sector_num[i] == -1 during A -> B.  Therefore
1876                      * B will never enter A during this time window.
1877                      */
1878                     qemu_coroutine_enter(s->co[i]);
1879                     break;
1880                 }
1881             }
1882         }
1883     }
1884 
1885     qemu_vfree(buf);
1886     s->co[index] = NULL;
1887     s->running_coroutines--;
1888     if (!s->running_coroutines && s->ret == -EINPROGRESS) {
1889         /* the convert job finished successfully */
1890         s->ret = 0;
1891     }
1892 }
1893 
1894 static int convert_do_copy(ImgConvertState *s)
1895 {
1896     int ret, i, n;
1897     int64_t sector_num = 0;
1898 
1899     /* Check whether we have zero initialisation or can get it efficiently */
1900     s->has_zero_init = s->min_sparse && !s->target_has_backing
1901                      ? bdrv_has_zero_init(blk_bs(s->target))
1902                      : false;
1903 
1904     if (!s->has_zero_init && !s->target_has_backing &&
1905         bdrv_can_write_zeroes_with_unmap(blk_bs(s->target)))
1906     {
1907         ret = blk_make_zero(s->target, BDRV_REQ_MAY_UNMAP);
1908         if (ret == 0) {
1909             s->has_zero_init = true;
1910         }
1911     }
1912 
1913     /* Allocate buffer for copied data. For compressed images, only one cluster
1914      * can be copied at a time. */
1915     if (s->compressed) {
1916         if (s->cluster_sectors <= 0 || s->cluster_sectors > s->buf_sectors) {
1917             error_report("invalid cluster size");
1918             return -EINVAL;
1919         }
1920         s->buf_sectors = s->cluster_sectors;
1921     }
1922 
1923     while (sector_num < s->total_sectors) {
1924         n = convert_iteration_sectors(s, sector_num);
1925         if (n < 0) {
1926             return n;
1927         }
1928         if (s->status == BLK_DATA || (!s->min_sparse && s->status == BLK_ZERO))
1929         {
1930             s->allocated_sectors += n;
1931         }
1932         sector_num += n;
1933     }
1934 
1935     /* Do the copy */
1936     s->sector_next_status = 0;
1937     s->ret = -EINPROGRESS;
1938 
1939     qemu_co_mutex_init(&s->lock);
1940     for (i = 0; i < s->num_coroutines; i++) {
1941         s->co[i] = qemu_coroutine_create(convert_co_do_copy, s);
1942         s->wait_sector_num[i] = -1;
1943         qemu_coroutine_enter(s->co[i]);
1944     }
1945 
1946     while (s->running_coroutines) {
1947         main_loop_wait(false);
1948     }
1949 
1950     if (s->compressed && !s->ret) {
1951         /* signal EOF to align */
1952         ret = blk_pwrite_compressed(s->target, 0, NULL, 0);
1953         if (ret < 0) {
1954             return ret;
1955         }
1956     }
1957 
1958     return s->ret;
1959 }
1960 
1961 static int img_convert(int argc, char **argv)
1962 {
1963     int c, bs_i, flags, src_flags = 0;
1964     const char *fmt = NULL, *out_fmt = NULL, *cache = "unsafe",
1965                *src_cache = BDRV_DEFAULT_CACHE, *out_baseimg = NULL,
1966                *out_filename, *out_baseimg_param, *snapshot_name = NULL;
1967     BlockDriver *drv = NULL, *proto_drv = NULL;
1968     BlockDriverInfo bdi;
1969     BlockDriverState *out_bs;
1970     QemuOpts *opts = NULL, *sn_opts = NULL;
1971     QemuOptsList *create_opts = NULL;
1972     char *options = NULL;
1973     Error *local_err = NULL;
1974     bool writethrough, src_writethrough, quiet = false, image_opts = false,
1975          skip_create = false, progress = false, tgt_image_opts = false;
1976     int64_t ret = -EINVAL;
1977     bool force_share = false;
1978 
1979     ImgConvertState s = (ImgConvertState) {
1980         /* Need at least 4k of zeros for sparse detection */
1981         .min_sparse         = 8,
1982         .copy_range         = true,
1983         .buf_sectors        = IO_BUF_SIZE / BDRV_SECTOR_SIZE,
1984         .wr_in_order        = true,
1985         .num_coroutines     = 8,
1986     };
1987 
1988     for(;;) {
1989         static const struct option long_options[] = {
1990             {"help", no_argument, 0, 'h'},
1991             {"object", required_argument, 0, OPTION_OBJECT},
1992             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
1993             {"force-share", no_argument, 0, 'U'},
1994             {"target-image-opts", no_argument, 0, OPTION_TARGET_IMAGE_OPTS},
1995             {0, 0, 0, 0}
1996         };
1997         c = getopt_long(argc, argv, ":hf:O:B:co:s:l:S:pt:T:qnm:WU",
1998                         long_options, NULL);
1999         if (c == -1) {
2000             break;
2001         }
2002         switch(c) {
2003         case ':':
2004             missing_argument(argv[optind - 1]);
2005             break;
2006         case '?':
2007             unrecognized_option(argv[optind - 1]);
2008             break;
2009         case 'h':
2010             help();
2011             break;
2012         case 'f':
2013             fmt = optarg;
2014             break;
2015         case 'O':
2016             out_fmt = optarg;
2017             break;
2018         case 'B':
2019             out_baseimg = optarg;
2020             break;
2021         case 'c':
2022             s.compressed = true;
2023             s.copy_range = false;
2024             break;
2025         case 'o':
2026             if (!is_valid_option_list(optarg)) {
2027                 error_report("Invalid option list: %s", optarg);
2028                 goto fail_getopt;
2029             }
2030             if (!options) {
2031                 options = g_strdup(optarg);
2032             } else {
2033                 char *old_options = options;
2034                 options = g_strdup_printf("%s,%s", options, optarg);
2035                 g_free(old_options);
2036             }
2037             break;
2038         case 's':
2039             snapshot_name = optarg;
2040             break;
2041         case 'l':
2042             if (strstart(optarg, SNAPSHOT_OPT_BASE, NULL)) {
2043                 sn_opts = qemu_opts_parse_noisily(&internal_snapshot_opts,
2044                                                   optarg, false);
2045                 if (!sn_opts) {
2046                     error_report("Failed in parsing snapshot param '%s'",
2047                                  optarg);
2048                     goto fail_getopt;
2049                 }
2050             } else {
2051                 snapshot_name = optarg;
2052             }
2053             break;
2054         case 'S':
2055         {
2056             int64_t sval;
2057 
2058             sval = cvtnum(optarg);
2059             if (sval < 0) {
2060                 error_report("Invalid minimum zero buffer size for sparse output specified");
2061                 goto fail_getopt;
2062             }
2063 
2064             s.min_sparse = sval / BDRV_SECTOR_SIZE;
2065             s.copy_range = false;
2066             break;
2067         }
2068         case 'p':
2069             progress = true;
2070             break;
2071         case 't':
2072             cache = optarg;
2073             break;
2074         case 'T':
2075             src_cache = optarg;
2076             break;
2077         case 'q':
2078             quiet = true;
2079             break;
2080         case 'n':
2081             skip_create = true;
2082             break;
2083         case 'm':
2084             if (qemu_strtol(optarg, NULL, 0, &s.num_coroutines) ||
2085                 s.num_coroutines < 1 || s.num_coroutines > MAX_COROUTINES) {
2086                 error_report("Invalid number of coroutines. Allowed number of"
2087                              " coroutines is between 1 and %d", MAX_COROUTINES);
2088                 goto fail_getopt;
2089             }
2090             break;
2091         case 'W':
2092             s.wr_in_order = false;
2093             break;
2094         case 'U':
2095             force_share = true;
2096             break;
2097         case OPTION_OBJECT: {
2098             QemuOpts *object_opts;
2099             object_opts = qemu_opts_parse_noisily(&qemu_object_opts,
2100                                                   optarg, true);
2101             if (!object_opts) {
2102                 goto fail_getopt;
2103             }
2104             break;
2105         }
2106         case OPTION_IMAGE_OPTS:
2107             image_opts = true;
2108             break;
2109         case OPTION_TARGET_IMAGE_OPTS:
2110             tgt_image_opts = true;
2111             break;
2112         }
2113     }
2114 
2115     if (!out_fmt && !tgt_image_opts) {
2116         out_fmt = "raw";
2117     }
2118 
2119     if (qemu_opts_foreach(&qemu_object_opts,
2120                           user_creatable_add_opts_foreach,
2121                           NULL, NULL)) {
2122         goto fail_getopt;
2123     }
2124 
2125     if (!s.wr_in_order && s.compressed) {
2126         error_report("Out of order write and compress are mutually exclusive");
2127         goto fail_getopt;
2128     }
2129 
2130     if (tgt_image_opts && !skip_create) {
2131         error_report("--target-image-opts requires use of -n flag");
2132         goto fail_getopt;
2133     }
2134 
2135     s.src_num = argc - optind - 1;
2136     out_filename = s.src_num >= 1 ? argv[argc - 1] : NULL;
2137 
2138     if (options && has_help_option(options)) {
2139         if (out_fmt) {
2140             ret = print_block_option_help(out_filename, out_fmt);
2141             goto fail_getopt;
2142         } else {
2143             error_report("Option help requires a format be specified");
2144             goto fail_getopt;
2145         }
2146     }
2147 
2148     if (s.src_num < 1) {
2149         error_report("Must specify image file name");
2150         goto fail_getopt;
2151     }
2152 
2153 
2154     /* ret is still -EINVAL until here */
2155     ret = bdrv_parse_cache_mode(src_cache, &src_flags, &src_writethrough);
2156     if (ret < 0) {
2157         error_report("Invalid source cache option: %s", src_cache);
2158         goto fail_getopt;
2159     }
2160 
2161     /* Initialize before goto out */
2162     if (quiet) {
2163         progress = false;
2164     }
2165     qemu_progress_init(progress, 1.0);
2166     qemu_progress_print(0, 100);
2167 
2168     s.src = g_new0(BlockBackend *, s.src_num);
2169     s.src_sectors = g_new(int64_t, s.src_num);
2170 
2171     for (bs_i = 0; bs_i < s.src_num; bs_i++) {
2172         s.src[bs_i] = img_open(image_opts, argv[optind + bs_i],
2173                                fmt, src_flags, src_writethrough, quiet,
2174                                force_share);
2175         if (!s.src[bs_i]) {
2176             ret = -1;
2177             goto out;
2178         }
2179         s.src_sectors[bs_i] = blk_nb_sectors(s.src[bs_i]);
2180         if (s.src_sectors[bs_i] < 0) {
2181             error_report("Could not get size of %s: %s",
2182                          argv[optind + bs_i], strerror(-s.src_sectors[bs_i]));
2183             ret = -1;
2184             goto out;
2185         }
2186         s.total_sectors += s.src_sectors[bs_i];
2187     }
2188 
2189     if (sn_opts) {
2190         bdrv_snapshot_load_tmp(blk_bs(s.src[0]),
2191                                qemu_opt_get(sn_opts, SNAPSHOT_OPT_ID),
2192                                qemu_opt_get(sn_opts, SNAPSHOT_OPT_NAME),
2193                                &local_err);
2194     } else if (snapshot_name != NULL) {
2195         if (s.src_num > 1) {
2196             error_report("No support for concatenating multiple snapshot");
2197             ret = -1;
2198             goto out;
2199         }
2200 
2201         bdrv_snapshot_load_tmp_by_id_or_name(blk_bs(s.src[0]), snapshot_name,
2202                                              &local_err);
2203     }
2204     if (local_err) {
2205         error_reportf_err(local_err, "Failed to load snapshot: ");
2206         ret = -1;
2207         goto out;
2208     }
2209 
2210     if (!skip_create) {
2211         /* Find driver and parse its options */
2212         drv = bdrv_find_format(out_fmt);
2213         if (!drv) {
2214             error_report("Unknown file format '%s'", out_fmt);
2215             ret = -1;
2216             goto out;
2217         }
2218 
2219         proto_drv = bdrv_find_protocol(out_filename, true, &local_err);
2220         if (!proto_drv) {
2221             error_report_err(local_err);
2222             ret = -1;
2223             goto out;
2224         }
2225 
2226         if (!drv->create_opts) {
2227             error_report("Format driver '%s' does not support image creation",
2228                          drv->format_name);
2229             ret = -1;
2230             goto out;
2231         }
2232 
2233         if (!proto_drv->create_opts) {
2234             error_report("Protocol driver '%s' does not support image creation",
2235                          proto_drv->format_name);
2236             ret = -1;
2237             goto out;
2238         }
2239 
2240         create_opts = qemu_opts_append(create_opts, drv->create_opts);
2241         create_opts = qemu_opts_append(create_opts, proto_drv->create_opts);
2242 
2243         opts = qemu_opts_create(create_opts, NULL, 0, &error_abort);
2244         if (options) {
2245             qemu_opts_do_parse(opts, options, NULL, &local_err);
2246             if (local_err) {
2247                 error_report_err(local_err);
2248                 ret = -1;
2249                 goto out;
2250             }
2251         }
2252 
2253         qemu_opt_set_number(opts, BLOCK_OPT_SIZE, s.total_sectors * 512,
2254                             &error_abort);
2255         ret = add_old_style_options(out_fmt, opts, out_baseimg, NULL);
2256         if (ret < 0) {
2257             goto out;
2258         }
2259     }
2260 
2261     /* Get backing file name if -o backing_file was used */
2262     out_baseimg_param = qemu_opt_get(opts, BLOCK_OPT_BACKING_FILE);
2263     if (out_baseimg_param) {
2264         out_baseimg = out_baseimg_param;
2265     }
2266     s.target_has_backing = (bool) out_baseimg;
2267 
2268     if (s.src_num > 1 && out_baseimg) {
2269         error_report("Having a backing file for the target makes no sense when "
2270                      "concatenating multiple input images");
2271         ret = -1;
2272         goto out;
2273     }
2274 
2275     /* Check if compression is supported */
2276     if (s.compressed) {
2277         bool encryption =
2278             qemu_opt_get_bool(opts, BLOCK_OPT_ENCRYPT, false);
2279         const char *encryptfmt =
2280             qemu_opt_get(opts, BLOCK_OPT_ENCRYPT_FORMAT);
2281         const char *preallocation =
2282             qemu_opt_get(opts, BLOCK_OPT_PREALLOC);
2283 
2284         if (drv && !drv->bdrv_co_pwritev_compressed) {
2285             error_report("Compression not supported for this file format");
2286             ret = -1;
2287             goto out;
2288         }
2289 
2290         if (encryption || encryptfmt) {
2291             error_report("Compression and encryption not supported at "
2292                          "the same time");
2293             ret = -1;
2294             goto out;
2295         }
2296 
2297         if (preallocation
2298             && strcmp(preallocation, "off"))
2299         {
2300             error_report("Compression and preallocation not supported at "
2301                          "the same time");
2302             ret = -1;
2303             goto out;
2304         }
2305     }
2306 
2307     if (!skip_create) {
2308         /* Create the new image */
2309         ret = bdrv_create(drv, out_filename, opts, &local_err);
2310         if (ret < 0) {
2311             error_reportf_err(local_err, "%s: error while converting %s: ",
2312                               out_filename, out_fmt);
2313             goto out;
2314         }
2315     }
2316 
2317     flags = s.min_sparse ? (BDRV_O_RDWR | BDRV_O_UNMAP) : BDRV_O_RDWR;
2318     ret = bdrv_parse_cache_mode(cache, &flags, &writethrough);
2319     if (ret < 0) {
2320         error_report("Invalid cache option: %s", cache);
2321         goto out;
2322     }
2323 
2324     if (skip_create) {
2325         s.target = img_open(tgt_image_opts, out_filename, out_fmt,
2326                             flags, writethrough, quiet, false);
2327     } else {
2328         /* TODO ultimately we should allow --target-image-opts
2329          * to be used even when -n is not given.
2330          * That has to wait for bdrv_create to be improved
2331          * to allow filenames in option syntax
2332          */
2333         s.target = img_open_new_file(out_filename, opts, out_fmt,
2334                                      flags, writethrough, quiet, false);
2335     }
2336     if (!s.target) {
2337         ret = -1;
2338         goto out;
2339     }
2340     out_bs = blk_bs(s.target);
2341 
2342     if (s.compressed && !out_bs->drv->bdrv_co_pwritev_compressed) {
2343         error_report("Compression not supported for this file format");
2344         ret = -1;
2345         goto out;
2346     }
2347 
2348     /* increase bufsectors from the default 4096 (2M) if opt_transfer
2349      * or discard_alignment of the out_bs is greater. Limit to 32768 (16MB)
2350      * as maximum. */
2351     s.buf_sectors = MIN(32768,
2352                         MAX(s.buf_sectors,
2353                             MAX(out_bs->bl.opt_transfer >> BDRV_SECTOR_BITS,
2354                                 out_bs->bl.pdiscard_alignment >>
2355                                 BDRV_SECTOR_BITS)));
2356 
2357     if (skip_create) {
2358         int64_t output_sectors = blk_nb_sectors(s.target);
2359         if (output_sectors < 0) {
2360             error_report("unable to get output image length: %s",
2361                          strerror(-output_sectors));
2362             ret = -1;
2363             goto out;
2364         } else if (output_sectors < s.total_sectors) {
2365             error_report("output file is smaller than input file");
2366             ret = -1;
2367             goto out;
2368         }
2369     }
2370 
2371     ret = bdrv_get_info(out_bs, &bdi);
2372     if (ret < 0) {
2373         if (s.compressed) {
2374             error_report("could not get block driver info");
2375             goto out;
2376         }
2377     } else {
2378         s.compressed = s.compressed || bdi.needs_compressed_writes;
2379         s.cluster_sectors = bdi.cluster_size / BDRV_SECTOR_SIZE;
2380     }
2381 
2382     ret = convert_do_copy(&s);
2383 out:
2384     if (!ret) {
2385         qemu_progress_print(100, 0);
2386     }
2387     qemu_progress_end();
2388     qemu_opts_del(opts);
2389     qemu_opts_free(create_opts);
2390     qemu_opts_del(sn_opts);
2391     blk_unref(s.target);
2392     if (s.src) {
2393         for (bs_i = 0; bs_i < s.src_num; bs_i++) {
2394             blk_unref(s.src[bs_i]);
2395         }
2396         g_free(s.src);
2397     }
2398     g_free(s.src_sectors);
2399 fail_getopt:
2400     g_free(options);
2401 
2402     return !!ret;
2403 }
2404 
2405 
2406 static void dump_snapshots(BlockDriverState *bs)
2407 {
2408     QEMUSnapshotInfo *sn_tab, *sn;
2409     int nb_sns, i;
2410 
2411     nb_sns = bdrv_snapshot_list(bs, &sn_tab);
2412     if (nb_sns <= 0)
2413         return;
2414     printf("Snapshot list:\n");
2415     bdrv_snapshot_dump(fprintf, stdout, NULL);
2416     printf("\n");
2417     for(i = 0; i < nb_sns; i++) {
2418         sn = &sn_tab[i];
2419         bdrv_snapshot_dump(fprintf, stdout, sn);
2420         printf("\n");
2421     }
2422     g_free(sn_tab);
2423 }
2424 
2425 static void dump_json_image_info_list(ImageInfoList *list)
2426 {
2427     QString *str;
2428     QObject *obj;
2429     Visitor *v = qobject_output_visitor_new(&obj);
2430 
2431     visit_type_ImageInfoList(v, NULL, &list, &error_abort);
2432     visit_complete(v, &obj);
2433     str = qobject_to_json_pretty(obj);
2434     assert(str != NULL);
2435     printf("%s\n", qstring_get_str(str));
2436     qobject_unref(obj);
2437     visit_free(v);
2438     qobject_unref(str);
2439 }
2440 
2441 static void dump_json_image_info(ImageInfo *info)
2442 {
2443     QString *str;
2444     QObject *obj;
2445     Visitor *v = qobject_output_visitor_new(&obj);
2446 
2447     visit_type_ImageInfo(v, NULL, &info, &error_abort);
2448     visit_complete(v, &obj);
2449     str = qobject_to_json_pretty(obj);
2450     assert(str != NULL);
2451     printf("%s\n", qstring_get_str(str));
2452     qobject_unref(obj);
2453     visit_free(v);
2454     qobject_unref(str);
2455 }
2456 
2457 static void dump_human_image_info_list(ImageInfoList *list)
2458 {
2459     ImageInfoList *elem;
2460     bool delim = false;
2461 
2462     for (elem = list; elem; elem = elem->next) {
2463         if (delim) {
2464             printf("\n");
2465         }
2466         delim = true;
2467 
2468         bdrv_image_info_dump(fprintf, stdout, elem->value);
2469     }
2470 }
2471 
2472 static gboolean str_equal_func(gconstpointer a, gconstpointer b)
2473 {
2474     return strcmp(a, b) == 0;
2475 }
2476 
2477 /**
2478  * Open an image file chain and return an ImageInfoList
2479  *
2480  * @filename: topmost image filename
2481  * @fmt: topmost image format (may be NULL to autodetect)
2482  * @chain: true  - enumerate entire backing file chain
2483  *         false - only topmost image file
2484  *
2485  * Returns a list of ImageInfo objects or NULL if there was an error opening an
2486  * image file.  If there was an error a message will have been printed to
2487  * stderr.
2488  */
2489 static ImageInfoList *collect_image_info_list(bool image_opts,
2490                                               const char *filename,
2491                                               const char *fmt,
2492                                               bool chain, bool force_share)
2493 {
2494     ImageInfoList *head = NULL;
2495     ImageInfoList **last = &head;
2496     GHashTable *filenames;
2497     Error *err = NULL;
2498 
2499     filenames = g_hash_table_new_full(g_str_hash, str_equal_func, NULL, NULL);
2500 
2501     while (filename) {
2502         BlockBackend *blk;
2503         BlockDriverState *bs;
2504         ImageInfo *info;
2505         ImageInfoList *elem;
2506 
2507         if (g_hash_table_lookup_extended(filenames, filename, NULL, NULL)) {
2508             error_report("Backing file '%s' creates an infinite loop.",
2509                          filename);
2510             goto err;
2511         }
2512         g_hash_table_insert(filenames, (gpointer)filename, NULL);
2513 
2514         blk = img_open(image_opts, filename, fmt,
2515                        BDRV_O_NO_BACKING | BDRV_O_NO_IO, false, false,
2516                        force_share);
2517         if (!blk) {
2518             goto err;
2519         }
2520         bs = blk_bs(blk);
2521 
2522         bdrv_query_image_info(bs, &info, &err);
2523         if (err) {
2524             error_report_err(err);
2525             blk_unref(blk);
2526             goto err;
2527         }
2528 
2529         elem = g_new0(ImageInfoList, 1);
2530         elem->value = info;
2531         *last = elem;
2532         last = &elem->next;
2533 
2534         blk_unref(blk);
2535 
2536         filename = fmt = NULL;
2537         if (chain) {
2538             if (info->has_full_backing_filename) {
2539                 filename = info->full_backing_filename;
2540             } else if (info->has_backing_filename) {
2541                 error_report("Could not determine absolute backing filename,"
2542                              " but backing filename '%s' present",
2543                              info->backing_filename);
2544                 goto err;
2545             }
2546             if (info->has_backing_filename_format) {
2547                 fmt = info->backing_filename_format;
2548             }
2549         }
2550     }
2551     g_hash_table_destroy(filenames);
2552     return head;
2553 
2554 err:
2555     qapi_free_ImageInfoList(head);
2556     g_hash_table_destroy(filenames);
2557     return NULL;
2558 }
2559 
2560 static int img_info(int argc, char **argv)
2561 {
2562     int c;
2563     OutputFormat output_format = OFORMAT_HUMAN;
2564     bool chain = false;
2565     const char *filename, *fmt, *output;
2566     ImageInfoList *list;
2567     bool image_opts = false;
2568     bool force_share = false;
2569 
2570     fmt = NULL;
2571     output = NULL;
2572     for(;;) {
2573         int option_index = 0;
2574         static const struct option long_options[] = {
2575             {"help", no_argument, 0, 'h'},
2576             {"format", required_argument, 0, 'f'},
2577             {"output", required_argument, 0, OPTION_OUTPUT},
2578             {"backing-chain", no_argument, 0, OPTION_BACKING_CHAIN},
2579             {"object", required_argument, 0, OPTION_OBJECT},
2580             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
2581             {"force-share", no_argument, 0, 'U'},
2582             {0, 0, 0, 0}
2583         };
2584         c = getopt_long(argc, argv, ":f:hU",
2585                         long_options, &option_index);
2586         if (c == -1) {
2587             break;
2588         }
2589         switch(c) {
2590         case ':':
2591             missing_argument(argv[optind - 1]);
2592             break;
2593         case '?':
2594             unrecognized_option(argv[optind - 1]);
2595             break;
2596         case 'h':
2597             help();
2598             break;
2599         case 'f':
2600             fmt = optarg;
2601             break;
2602         case 'U':
2603             force_share = true;
2604             break;
2605         case OPTION_OUTPUT:
2606             output = optarg;
2607             break;
2608         case OPTION_BACKING_CHAIN:
2609             chain = true;
2610             break;
2611         case OPTION_OBJECT: {
2612             QemuOpts *opts;
2613             opts = qemu_opts_parse_noisily(&qemu_object_opts,
2614                                            optarg, true);
2615             if (!opts) {
2616                 return 1;
2617             }
2618         }   break;
2619         case OPTION_IMAGE_OPTS:
2620             image_opts = true;
2621             break;
2622         }
2623     }
2624     if (optind != argc - 1) {
2625         error_exit("Expecting one image file name");
2626     }
2627     filename = argv[optind++];
2628 
2629     if (output && !strcmp(output, "json")) {
2630         output_format = OFORMAT_JSON;
2631     } else if (output && !strcmp(output, "human")) {
2632         output_format = OFORMAT_HUMAN;
2633     } else if (output) {
2634         error_report("--output must be used with human or json as argument.");
2635         return 1;
2636     }
2637 
2638     if (qemu_opts_foreach(&qemu_object_opts,
2639                           user_creatable_add_opts_foreach,
2640                           NULL, NULL)) {
2641         return 1;
2642     }
2643 
2644     list = collect_image_info_list(image_opts, filename, fmt, chain,
2645                                    force_share);
2646     if (!list) {
2647         return 1;
2648     }
2649 
2650     switch (output_format) {
2651     case OFORMAT_HUMAN:
2652         dump_human_image_info_list(list);
2653         break;
2654     case OFORMAT_JSON:
2655         if (chain) {
2656             dump_json_image_info_list(list);
2657         } else {
2658             dump_json_image_info(list->value);
2659         }
2660         break;
2661     }
2662 
2663     qapi_free_ImageInfoList(list);
2664     return 0;
2665 }
2666 
2667 static void dump_map_entry(OutputFormat output_format, MapEntry *e,
2668                            MapEntry *next)
2669 {
2670     switch (output_format) {
2671     case OFORMAT_HUMAN:
2672         if (e->data && !e->has_offset) {
2673             error_report("File contains external, encrypted or compressed clusters.");
2674             exit(1);
2675         }
2676         if (e->data && !e->zero) {
2677             printf("%#-16"PRIx64"%#-16"PRIx64"%#-16"PRIx64"%s\n",
2678                    e->start, e->length,
2679                    e->has_offset ? e->offset : 0,
2680                    e->has_filename ? e->filename : "");
2681         }
2682         /* This format ignores the distinction between 0, ZERO and ZERO|DATA.
2683          * Modify the flags here to allow more coalescing.
2684          */
2685         if (next && (!next->data || next->zero)) {
2686             next->data = false;
2687             next->zero = true;
2688         }
2689         break;
2690     case OFORMAT_JSON:
2691         printf("%s{ \"start\": %"PRId64", \"length\": %"PRId64","
2692                " \"depth\": %"PRId64", \"zero\": %s, \"data\": %s",
2693                (e->start == 0 ? "[" : ",\n"),
2694                e->start, e->length, e->depth,
2695                e->zero ? "true" : "false",
2696                e->data ? "true" : "false");
2697         if (e->has_offset) {
2698             printf(", \"offset\": %"PRId64"", e->offset);
2699         }
2700         putchar('}');
2701 
2702         if (!next) {
2703             printf("]\n");
2704         }
2705         break;
2706     }
2707 }
2708 
2709 static int get_block_status(BlockDriverState *bs, int64_t offset,
2710                             int64_t bytes, MapEntry *e)
2711 {
2712     int ret;
2713     int depth;
2714     BlockDriverState *file;
2715     bool has_offset;
2716     int64_t map;
2717 
2718     /* As an optimization, we could cache the current range of unallocated
2719      * clusters in each file of the chain, and avoid querying the same
2720      * range repeatedly.
2721      */
2722 
2723     depth = 0;
2724     for (;;) {
2725         ret = bdrv_block_status(bs, offset, bytes, &bytes, &map, &file);
2726         if (ret < 0) {
2727             return ret;
2728         }
2729         assert(bytes);
2730         if (ret & (BDRV_BLOCK_ZERO|BDRV_BLOCK_DATA)) {
2731             break;
2732         }
2733         bs = backing_bs(bs);
2734         if (bs == NULL) {
2735             ret = 0;
2736             break;
2737         }
2738 
2739         depth++;
2740     }
2741 
2742     has_offset = !!(ret & BDRV_BLOCK_OFFSET_VALID);
2743 
2744     *e = (MapEntry) {
2745         .start = offset,
2746         .length = bytes,
2747         .data = !!(ret & BDRV_BLOCK_DATA),
2748         .zero = !!(ret & BDRV_BLOCK_ZERO),
2749         .offset = map,
2750         .has_offset = has_offset,
2751         .depth = depth,
2752         .has_filename = file && has_offset,
2753         .filename = file && has_offset ? file->filename : NULL,
2754     };
2755 
2756     return 0;
2757 }
2758 
2759 static inline bool entry_mergeable(const MapEntry *curr, const MapEntry *next)
2760 {
2761     if (curr->length == 0) {
2762         return false;
2763     }
2764     if (curr->zero != next->zero ||
2765         curr->data != next->data ||
2766         curr->depth != next->depth ||
2767         curr->has_filename != next->has_filename ||
2768         curr->has_offset != next->has_offset) {
2769         return false;
2770     }
2771     if (curr->has_filename && strcmp(curr->filename, next->filename)) {
2772         return false;
2773     }
2774     if (curr->has_offset && curr->offset + curr->length != next->offset) {
2775         return false;
2776     }
2777     return true;
2778 }
2779 
2780 static int img_map(int argc, char **argv)
2781 {
2782     int c;
2783     OutputFormat output_format = OFORMAT_HUMAN;
2784     BlockBackend *blk;
2785     BlockDriverState *bs;
2786     const char *filename, *fmt, *output;
2787     int64_t length;
2788     MapEntry curr = { .length = 0 }, next;
2789     int ret = 0;
2790     bool image_opts = false;
2791     bool force_share = false;
2792 
2793     fmt = NULL;
2794     output = NULL;
2795     for (;;) {
2796         int option_index = 0;
2797         static const struct option long_options[] = {
2798             {"help", no_argument, 0, 'h'},
2799             {"format", required_argument, 0, 'f'},
2800             {"output", required_argument, 0, OPTION_OUTPUT},
2801             {"object", required_argument, 0, OPTION_OBJECT},
2802             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
2803             {"force-share", no_argument, 0, 'U'},
2804             {0, 0, 0, 0}
2805         };
2806         c = getopt_long(argc, argv, ":f:hU",
2807                         long_options, &option_index);
2808         if (c == -1) {
2809             break;
2810         }
2811         switch (c) {
2812         case ':':
2813             missing_argument(argv[optind - 1]);
2814             break;
2815         case '?':
2816             unrecognized_option(argv[optind - 1]);
2817             break;
2818         case 'h':
2819             help();
2820             break;
2821         case 'f':
2822             fmt = optarg;
2823             break;
2824         case 'U':
2825             force_share = true;
2826             break;
2827         case OPTION_OUTPUT:
2828             output = optarg;
2829             break;
2830         case OPTION_OBJECT: {
2831             QemuOpts *opts;
2832             opts = qemu_opts_parse_noisily(&qemu_object_opts,
2833                                            optarg, true);
2834             if (!opts) {
2835                 return 1;
2836             }
2837         }   break;
2838         case OPTION_IMAGE_OPTS:
2839             image_opts = true;
2840             break;
2841         }
2842     }
2843     if (optind != argc - 1) {
2844         error_exit("Expecting one image file name");
2845     }
2846     filename = argv[optind];
2847 
2848     if (output && !strcmp(output, "json")) {
2849         output_format = OFORMAT_JSON;
2850     } else if (output && !strcmp(output, "human")) {
2851         output_format = OFORMAT_HUMAN;
2852     } else if (output) {
2853         error_report("--output must be used with human or json as argument.");
2854         return 1;
2855     }
2856 
2857     if (qemu_opts_foreach(&qemu_object_opts,
2858                           user_creatable_add_opts_foreach,
2859                           NULL, NULL)) {
2860         return 1;
2861     }
2862 
2863     blk = img_open(image_opts, filename, fmt, 0, false, false, force_share);
2864     if (!blk) {
2865         return 1;
2866     }
2867     bs = blk_bs(blk);
2868 
2869     if (output_format == OFORMAT_HUMAN) {
2870         printf("%-16s%-16s%-16s%s\n", "Offset", "Length", "Mapped to", "File");
2871     }
2872 
2873     length = blk_getlength(blk);
2874     while (curr.start + curr.length < length) {
2875         int64_t offset = curr.start + curr.length;
2876         int64_t n;
2877 
2878         /* Probe up to 1 GiB at a time.  */
2879         n = QEMU_ALIGN_DOWN(MIN(1 << 30, length - offset), BDRV_SECTOR_SIZE);
2880         ret = get_block_status(bs, offset, n, &next);
2881 
2882         if (ret < 0) {
2883             error_report("Could not read file metadata: %s", strerror(-ret));
2884             goto out;
2885         }
2886 
2887         if (entry_mergeable(&curr, &next)) {
2888             curr.length += next.length;
2889             continue;
2890         }
2891 
2892         if (curr.length > 0) {
2893             dump_map_entry(output_format, &curr, &next);
2894         }
2895         curr = next;
2896     }
2897 
2898     dump_map_entry(output_format, &curr, NULL);
2899 
2900 out:
2901     blk_unref(blk);
2902     return ret < 0;
2903 }
2904 
2905 #define SNAPSHOT_LIST   1
2906 #define SNAPSHOT_CREATE 2
2907 #define SNAPSHOT_APPLY  3
2908 #define SNAPSHOT_DELETE 4
2909 
2910 static int img_snapshot(int argc, char **argv)
2911 {
2912     BlockBackend *blk;
2913     BlockDriverState *bs;
2914     QEMUSnapshotInfo sn;
2915     char *filename, *snapshot_name = NULL;
2916     int c, ret = 0, bdrv_oflags;
2917     int action = 0;
2918     qemu_timeval tv;
2919     bool quiet = false;
2920     Error *err = NULL;
2921     bool image_opts = false;
2922     bool force_share = false;
2923 
2924     bdrv_oflags = BDRV_O_RDWR;
2925     /* Parse commandline parameters */
2926     for(;;) {
2927         static const struct option long_options[] = {
2928             {"help", no_argument, 0, 'h'},
2929             {"object", required_argument, 0, OPTION_OBJECT},
2930             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
2931             {"force-share", no_argument, 0, 'U'},
2932             {0, 0, 0, 0}
2933         };
2934         c = getopt_long(argc, argv, ":la:c:d:hqU",
2935                         long_options, NULL);
2936         if (c == -1) {
2937             break;
2938         }
2939         switch(c) {
2940         case ':':
2941             missing_argument(argv[optind - 1]);
2942             break;
2943         case '?':
2944             unrecognized_option(argv[optind - 1]);
2945             break;
2946         case 'h':
2947             help();
2948             return 0;
2949         case 'l':
2950             if (action) {
2951                 error_exit("Cannot mix '-l', '-a', '-c', '-d'");
2952                 return 0;
2953             }
2954             action = SNAPSHOT_LIST;
2955             bdrv_oflags &= ~BDRV_O_RDWR; /* no need for RW */
2956             break;
2957         case 'a':
2958             if (action) {
2959                 error_exit("Cannot mix '-l', '-a', '-c', '-d'");
2960                 return 0;
2961             }
2962             action = SNAPSHOT_APPLY;
2963             snapshot_name = optarg;
2964             break;
2965         case 'c':
2966             if (action) {
2967                 error_exit("Cannot mix '-l', '-a', '-c', '-d'");
2968                 return 0;
2969             }
2970             action = SNAPSHOT_CREATE;
2971             snapshot_name = optarg;
2972             break;
2973         case 'd':
2974             if (action) {
2975                 error_exit("Cannot mix '-l', '-a', '-c', '-d'");
2976                 return 0;
2977             }
2978             action = SNAPSHOT_DELETE;
2979             snapshot_name = optarg;
2980             break;
2981         case 'q':
2982             quiet = true;
2983             break;
2984         case 'U':
2985             force_share = true;
2986             break;
2987         case OPTION_OBJECT: {
2988             QemuOpts *opts;
2989             opts = qemu_opts_parse_noisily(&qemu_object_opts,
2990                                            optarg, true);
2991             if (!opts) {
2992                 return 1;
2993             }
2994         }   break;
2995         case OPTION_IMAGE_OPTS:
2996             image_opts = true;
2997             break;
2998         }
2999     }
3000 
3001     if (optind != argc - 1) {
3002         error_exit("Expecting one image file name");
3003     }
3004     filename = argv[optind++];
3005 
3006     if (qemu_opts_foreach(&qemu_object_opts,
3007                           user_creatable_add_opts_foreach,
3008                           NULL, NULL)) {
3009         return 1;
3010     }
3011 
3012     /* Open the image */
3013     blk = img_open(image_opts, filename, NULL, bdrv_oflags, false, quiet,
3014                    force_share);
3015     if (!blk) {
3016         return 1;
3017     }
3018     bs = blk_bs(blk);
3019 
3020     /* Perform the requested action */
3021     switch(action) {
3022     case SNAPSHOT_LIST:
3023         dump_snapshots(bs);
3024         break;
3025 
3026     case SNAPSHOT_CREATE:
3027         memset(&sn, 0, sizeof(sn));
3028         pstrcpy(sn.name, sizeof(sn.name), snapshot_name);
3029 
3030         qemu_gettimeofday(&tv);
3031         sn.date_sec = tv.tv_sec;
3032         sn.date_nsec = tv.tv_usec * 1000;
3033 
3034         ret = bdrv_snapshot_create(bs, &sn);
3035         if (ret) {
3036             error_report("Could not create snapshot '%s': %d (%s)",
3037                 snapshot_name, ret, strerror(-ret));
3038         }
3039         break;
3040 
3041     case SNAPSHOT_APPLY:
3042         ret = bdrv_snapshot_goto(bs, snapshot_name, &err);
3043         if (ret) {
3044             error_reportf_err(err, "Could not apply snapshot '%s': ",
3045                               snapshot_name);
3046         }
3047         break;
3048 
3049     case SNAPSHOT_DELETE:
3050         bdrv_snapshot_delete_by_id_or_name(bs, snapshot_name, &err);
3051         if (err) {
3052             error_reportf_err(err, "Could not delete snapshot '%s': ",
3053                               snapshot_name);
3054             ret = 1;
3055         }
3056         break;
3057     }
3058 
3059     /* Cleanup */
3060     blk_unref(blk);
3061     if (ret) {
3062         return 1;
3063     }
3064     return 0;
3065 }
3066 
3067 static int img_rebase(int argc, char **argv)
3068 {
3069     BlockBackend *blk = NULL, *blk_old_backing = NULL, *blk_new_backing = NULL;
3070     uint8_t *buf_old = NULL;
3071     uint8_t *buf_new = NULL;
3072     BlockDriverState *bs = NULL;
3073     char *filename;
3074     const char *fmt, *cache, *src_cache, *out_basefmt, *out_baseimg;
3075     int c, flags, src_flags, ret;
3076     bool writethrough, src_writethrough;
3077     int unsafe = 0;
3078     bool force_share = false;
3079     int progress = 0;
3080     bool quiet = false;
3081     Error *local_err = NULL;
3082     bool image_opts = false;
3083 
3084     /* Parse commandline parameters */
3085     fmt = NULL;
3086     cache = BDRV_DEFAULT_CACHE;
3087     src_cache = BDRV_DEFAULT_CACHE;
3088     out_baseimg = NULL;
3089     out_basefmt = NULL;
3090     for(;;) {
3091         static const struct option long_options[] = {
3092             {"help", no_argument, 0, 'h'},
3093             {"object", required_argument, 0, OPTION_OBJECT},
3094             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
3095             {"force-share", no_argument, 0, 'U'},
3096             {0, 0, 0, 0}
3097         };
3098         c = getopt_long(argc, argv, ":hf:F:b:upt:T:qU",
3099                         long_options, NULL);
3100         if (c == -1) {
3101             break;
3102         }
3103         switch(c) {
3104         case ':':
3105             missing_argument(argv[optind - 1]);
3106             break;
3107         case '?':
3108             unrecognized_option(argv[optind - 1]);
3109             break;
3110         case 'h':
3111             help();
3112             return 0;
3113         case 'f':
3114             fmt = optarg;
3115             break;
3116         case 'F':
3117             out_basefmt = optarg;
3118             break;
3119         case 'b':
3120             out_baseimg = optarg;
3121             break;
3122         case 'u':
3123             unsafe = 1;
3124             break;
3125         case 'p':
3126             progress = 1;
3127             break;
3128         case 't':
3129             cache = optarg;
3130             break;
3131         case 'T':
3132             src_cache = optarg;
3133             break;
3134         case 'q':
3135             quiet = true;
3136             break;
3137         case OPTION_OBJECT: {
3138             QemuOpts *opts;
3139             opts = qemu_opts_parse_noisily(&qemu_object_opts,
3140                                            optarg, true);
3141             if (!opts) {
3142                 return 1;
3143             }
3144         }   break;
3145         case OPTION_IMAGE_OPTS:
3146             image_opts = true;
3147             break;
3148         case 'U':
3149             force_share = true;
3150             break;
3151         }
3152     }
3153 
3154     if (quiet) {
3155         progress = 0;
3156     }
3157 
3158     if (optind != argc - 1) {
3159         error_exit("Expecting one image file name");
3160     }
3161     if (!unsafe && !out_baseimg) {
3162         error_exit("Must specify backing file (-b) or use unsafe mode (-u)");
3163     }
3164     filename = argv[optind++];
3165 
3166     if (qemu_opts_foreach(&qemu_object_opts,
3167                           user_creatable_add_opts_foreach,
3168                           NULL, NULL)) {
3169         return 1;
3170     }
3171 
3172     qemu_progress_init(progress, 2.0);
3173     qemu_progress_print(0, 100);
3174 
3175     flags = BDRV_O_RDWR | (unsafe ? BDRV_O_NO_BACKING : 0);
3176     ret = bdrv_parse_cache_mode(cache, &flags, &writethrough);
3177     if (ret < 0) {
3178         error_report("Invalid cache option: %s", cache);
3179         goto out;
3180     }
3181 
3182     src_flags = 0;
3183     ret = bdrv_parse_cache_mode(src_cache, &src_flags, &src_writethrough);
3184     if (ret < 0) {
3185         error_report("Invalid source cache option: %s", src_cache);
3186         goto out;
3187     }
3188 
3189     /* The source files are opened read-only, don't care about WCE */
3190     assert((src_flags & BDRV_O_RDWR) == 0);
3191     (void) src_writethrough;
3192 
3193     /*
3194      * Open the images.
3195      *
3196      * Ignore the old backing file for unsafe rebase in case we want to correct
3197      * the reference to a renamed or moved backing file.
3198      */
3199     blk = img_open(image_opts, filename, fmt, flags, writethrough, quiet,
3200                    false);
3201     if (!blk) {
3202         ret = -1;
3203         goto out;
3204     }
3205     bs = blk_bs(blk);
3206 
3207     if (out_basefmt != NULL) {
3208         if (bdrv_find_format(out_basefmt) == NULL) {
3209             error_report("Invalid format name: '%s'", out_basefmt);
3210             ret = -1;
3211             goto out;
3212         }
3213     }
3214 
3215     /* For safe rebasing we need to compare old and new backing file */
3216     if (!unsafe) {
3217         char backing_name[PATH_MAX];
3218         QDict *options = NULL;
3219 
3220         if (bs->backing_format[0] != '\0') {
3221             options = qdict_new();
3222             qdict_put_str(options, "driver", bs->backing_format);
3223         }
3224 
3225         if (force_share) {
3226             if (!options) {
3227                 options = qdict_new();
3228             }
3229             qdict_put_bool(options, BDRV_OPT_FORCE_SHARE, true);
3230         }
3231         bdrv_get_backing_filename(bs, backing_name, sizeof(backing_name));
3232         blk_old_backing = blk_new_open(backing_name, NULL,
3233                                        options, src_flags, &local_err);
3234         if (!blk_old_backing) {
3235             error_reportf_err(local_err,
3236                               "Could not open old backing file '%s': ",
3237                               backing_name);
3238             ret = -1;
3239             goto out;
3240         }
3241 
3242         if (out_baseimg[0]) {
3243             options = qdict_new();
3244             if (out_basefmt) {
3245                 qdict_put_str(options, "driver", out_basefmt);
3246             }
3247             if (force_share) {
3248                 qdict_put_bool(options, BDRV_OPT_FORCE_SHARE, true);
3249             }
3250 
3251             blk_new_backing = blk_new_open(out_baseimg, NULL,
3252                                            options, src_flags, &local_err);
3253             if (!blk_new_backing) {
3254                 error_reportf_err(local_err,
3255                                   "Could not open new backing file '%s': ",
3256                                   out_baseimg);
3257                 ret = -1;
3258                 goto out;
3259             }
3260         }
3261     }
3262 
3263     /*
3264      * Check each unallocated cluster in the COW file. If it is unallocated,
3265      * accesses go to the backing file. We must therefore compare this cluster
3266      * in the old and new backing file, and if they differ we need to copy it
3267      * from the old backing file into the COW file.
3268      *
3269      * If qemu-img crashes during this step, no harm is done. The content of
3270      * the image is the same as the original one at any time.
3271      */
3272     if (!unsafe) {
3273         int64_t size;
3274         int64_t old_backing_size;
3275         int64_t new_backing_size = 0;
3276         uint64_t offset;
3277         int64_t n;
3278         float local_progress = 0;
3279 
3280         buf_old = blk_blockalign(blk, IO_BUF_SIZE);
3281         buf_new = blk_blockalign(blk, IO_BUF_SIZE);
3282 
3283         size = blk_getlength(blk);
3284         if (size < 0) {
3285             error_report("Could not get size of '%s': %s",
3286                          filename, strerror(-size));
3287             ret = -1;
3288             goto out;
3289         }
3290         old_backing_size = blk_getlength(blk_old_backing);
3291         if (old_backing_size < 0) {
3292             char backing_name[PATH_MAX];
3293 
3294             bdrv_get_backing_filename(bs, backing_name, sizeof(backing_name));
3295             error_report("Could not get size of '%s': %s",
3296                          backing_name, strerror(-old_backing_size));
3297             ret = -1;
3298             goto out;
3299         }
3300         if (blk_new_backing) {
3301             new_backing_size = blk_getlength(blk_new_backing);
3302             if (new_backing_size < 0) {
3303                 error_report("Could not get size of '%s': %s",
3304                              out_baseimg, strerror(-new_backing_size));
3305                 ret = -1;
3306                 goto out;
3307             }
3308         }
3309 
3310         if (size != 0) {
3311             local_progress = (float)100 / (size / MIN(size, IO_BUF_SIZE));
3312         }
3313 
3314         for (offset = 0; offset < size; offset += n) {
3315             /* How many bytes can we handle with the next read? */
3316             n = MIN(IO_BUF_SIZE, size - offset);
3317 
3318             /* If the cluster is allocated, we don't need to take action */
3319             ret = bdrv_is_allocated(bs, offset, n, &n);
3320             if (ret < 0) {
3321                 error_report("error while reading image metadata: %s",
3322                              strerror(-ret));
3323                 goto out;
3324             }
3325             if (ret) {
3326                 continue;
3327             }
3328 
3329             /*
3330              * Read old and new backing file and take into consideration that
3331              * backing files may be smaller than the COW image.
3332              */
3333             if (offset >= old_backing_size) {
3334                 memset(buf_old, 0, n);
3335             } else {
3336                 if (offset + n > old_backing_size) {
3337                     n = old_backing_size - offset;
3338                 }
3339 
3340                 ret = blk_pread(blk_old_backing, offset, buf_old, n);
3341                 if (ret < 0) {
3342                     error_report("error while reading from old backing file");
3343                     goto out;
3344                 }
3345             }
3346 
3347             if (offset >= new_backing_size || !blk_new_backing) {
3348                 memset(buf_new, 0, n);
3349             } else {
3350                 if (offset + n > new_backing_size) {
3351                     n = new_backing_size - offset;
3352                 }
3353 
3354                 ret = blk_pread(blk_new_backing, offset, buf_new, n);
3355                 if (ret < 0) {
3356                     error_report("error while reading from new backing file");
3357                     goto out;
3358                 }
3359             }
3360 
3361             /* If they differ, we need to write to the COW file */
3362             uint64_t written = 0;
3363 
3364             while (written < n) {
3365                 int64_t pnum;
3366 
3367                 if (compare_buffers(buf_old + written, buf_new + written,
3368                                     n - written, &pnum))
3369                 {
3370                     ret = blk_pwrite(blk, offset + written,
3371                                      buf_old + written, pnum, 0);
3372                     if (ret < 0) {
3373                         error_report("Error while writing to COW image: %s",
3374                             strerror(-ret));
3375                         goto out;
3376                     }
3377                 }
3378 
3379                 written += pnum;
3380             }
3381             qemu_progress_print(local_progress, 100);
3382         }
3383     }
3384 
3385     /*
3386      * Change the backing file. All clusters that are different from the old
3387      * backing file are overwritten in the COW file now, so the visible content
3388      * doesn't change when we switch the backing file.
3389      */
3390     if (out_baseimg && *out_baseimg) {
3391         ret = bdrv_change_backing_file(bs, out_baseimg, out_basefmt);
3392     } else {
3393         ret = bdrv_change_backing_file(bs, NULL, NULL);
3394     }
3395 
3396     if (ret == -ENOSPC) {
3397         error_report("Could not change the backing file to '%s': No "
3398                      "space left in the file header", out_baseimg);
3399     } else if (ret < 0) {
3400         error_report("Could not change the backing file to '%s': %s",
3401             out_baseimg, strerror(-ret));
3402     }
3403 
3404     qemu_progress_print(100, 0);
3405     /*
3406      * TODO At this point it is possible to check if any clusters that are
3407      * allocated in the COW file are the same in the backing file. If so, they
3408      * could be dropped from the COW file. Don't do this before switching the
3409      * backing file, in case of a crash this would lead to corruption.
3410      */
3411 out:
3412     qemu_progress_end();
3413     /* Cleanup */
3414     if (!unsafe) {
3415         blk_unref(blk_old_backing);
3416         blk_unref(blk_new_backing);
3417     }
3418     qemu_vfree(buf_old);
3419     qemu_vfree(buf_new);
3420 
3421     blk_unref(blk);
3422     if (ret) {
3423         return 1;
3424     }
3425     return 0;
3426 }
3427 
3428 static int img_resize(int argc, char **argv)
3429 {
3430     Error *err = NULL;
3431     int c, ret, relative;
3432     const char *filename, *fmt, *size;
3433     int64_t n, total_size, current_size, new_size;
3434     bool quiet = false;
3435     BlockBackend *blk = NULL;
3436     PreallocMode prealloc = PREALLOC_MODE_OFF;
3437     QemuOpts *param;
3438 
3439     static QemuOptsList resize_options = {
3440         .name = "resize_options",
3441         .head = QTAILQ_HEAD_INITIALIZER(resize_options.head),
3442         .desc = {
3443             {
3444                 .name = BLOCK_OPT_SIZE,
3445                 .type = QEMU_OPT_SIZE,
3446                 .help = "Virtual disk size"
3447             }, {
3448                 /* end of list */
3449             }
3450         },
3451     };
3452     bool image_opts = false;
3453     bool shrink = false;
3454 
3455     /* Remove size from argv manually so that negative numbers are not treated
3456      * as options by getopt. */
3457     if (argc < 3) {
3458         error_exit("Not enough arguments");
3459         return 1;
3460     }
3461 
3462     size = argv[--argc];
3463 
3464     /* Parse getopt arguments */
3465     fmt = NULL;
3466     for(;;) {
3467         static const struct option long_options[] = {
3468             {"help", no_argument, 0, 'h'},
3469             {"object", required_argument, 0, OPTION_OBJECT},
3470             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
3471             {"preallocation", required_argument, 0, OPTION_PREALLOCATION},
3472             {"shrink", no_argument, 0, OPTION_SHRINK},
3473             {0, 0, 0, 0}
3474         };
3475         c = getopt_long(argc, argv, ":f:hq",
3476                         long_options, NULL);
3477         if (c == -1) {
3478             break;
3479         }
3480         switch(c) {
3481         case ':':
3482             missing_argument(argv[optind - 1]);
3483             break;
3484         case '?':
3485             unrecognized_option(argv[optind - 1]);
3486             break;
3487         case 'h':
3488             help();
3489             break;
3490         case 'f':
3491             fmt = optarg;
3492             break;
3493         case 'q':
3494             quiet = true;
3495             break;
3496         case OPTION_OBJECT: {
3497             QemuOpts *opts;
3498             opts = qemu_opts_parse_noisily(&qemu_object_opts,
3499                                            optarg, true);
3500             if (!opts) {
3501                 return 1;
3502             }
3503         }   break;
3504         case OPTION_IMAGE_OPTS:
3505             image_opts = true;
3506             break;
3507         case OPTION_PREALLOCATION:
3508             prealloc = qapi_enum_parse(&PreallocMode_lookup, optarg,
3509                                        PREALLOC_MODE__MAX, NULL);
3510             if (prealloc == PREALLOC_MODE__MAX) {
3511                 error_report("Invalid preallocation mode '%s'", optarg);
3512                 return 1;
3513             }
3514             break;
3515         case OPTION_SHRINK:
3516             shrink = true;
3517             break;
3518         }
3519     }
3520     if (optind != argc - 1) {
3521         error_exit("Expecting image file name and size");
3522     }
3523     filename = argv[optind++];
3524 
3525     if (qemu_opts_foreach(&qemu_object_opts,
3526                           user_creatable_add_opts_foreach,
3527                           NULL, NULL)) {
3528         return 1;
3529     }
3530 
3531     /* Choose grow, shrink, or absolute resize mode */
3532     switch (size[0]) {
3533     case '+':
3534         relative = 1;
3535         size++;
3536         break;
3537     case '-':
3538         relative = -1;
3539         size++;
3540         break;
3541     default:
3542         relative = 0;
3543         break;
3544     }
3545 
3546     /* Parse size */
3547     param = qemu_opts_create(&resize_options, NULL, 0, &error_abort);
3548     qemu_opt_set(param, BLOCK_OPT_SIZE, size, &err);
3549     if (err) {
3550         error_report_err(err);
3551         ret = -1;
3552         qemu_opts_del(param);
3553         goto out;
3554     }
3555     n = qemu_opt_get_size(param, BLOCK_OPT_SIZE, 0);
3556     qemu_opts_del(param);
3557 
3558     blk = img_open(image_opts, filename, fmt,
3559                    BDRV_O_RDWR | BDRV_O_RESIZE, false, quiet,
3560                    false);
3561     if (!blk) {
3562         ret = -1;
3563         goto out;
3564     }
3565 
3566     current_size = blk_getlength(blk);
3567     if (current_size < 0) {
3568         error_report("Failed to inquire current image length: %s",
3569                      strerror(-current_size));
3570         ret = -1;
3571         goto out;
3572     }
3573 
3574     if (relative) {
3575         total_size = current_size + n * relative;
3576     } else {
3577         total_size = n;
3578     }
3579     if (total_size <= 0) {
3580         error_report("New image size must be positive");
3581         ret = -1;
3582         goto out;
3583     }
3584 
3585     if (total_size <= current_size && prealloc != PREALLOC_MODE_OFF) {
3586         error_report("Preallocation can only be used for growing images");
3587         ret = -1;
3588         goto out;
3589     }
3590 
3591     if (total_size < current_size && !shrink) {
3592         warn_report("Shrinking an image will delete all data beyond the "
3593                     "shrunken image's end. Before performing such an "
3594                     "operation, make sure there is no important data there.");
3595 
3596         if (g_strcmp0(bdrv_get_format_name(blk_bs(blk)), "raw") != 0) {
3597             error_report(
3598               "Use the --shrink option to perform a shrink operation.");
3599             ret = -1;
3600             goto out;
3601         } else {
3602             warn_report("Using the --shrink option will suppress this message. "
3603                         "Note that future versions of qemu-img may refuse to "
3604                         "shrink images without this option.");
3605         }
3606     }
3607 
3608     ret = blk_truncate(blk, total_size, prealloc, &err);
3609     if (ret < 0) {
3610         error_report_err(err);
3611         goto out;
3612     }
3613 
3614     new_size = blk_getlength(blk);
3615     if (new_size < 0) {
3616         error_report("Failed to verify truncated image length: %s",
3617                      strerror(-new_size));
3618         ret = -1;
3619         goto out;
3620     }
3621 
3622     /* Some block drivers implement a truncation method, but only so
3623      * the user can cause qemu to refresh the image's size from disk.
3624      * The idea is that the user resizes the image outside of qemu and
3625      * then invokes block_resize to inform qemu about it.
3626      * (This includes iscsi and file-posix for device files.)
3627      * Of course, that is not the behavior someone invoking
3628      * qemu-img resize would find useful, so we catch that behavior
3629      * here and tell the user. */
3630     if (new_size != total_size && new_size == current_size) {
3631         error_report("Image was not resized; resizing may not be supported "
3632                      "for this image");
3633         ret = -1;
3634         goto out;
3635     }
3636 
3637     if (new_size != total_size) {
3638         warn_report("Image should have been resized to %" PRIi64
3639                     " bytes, but was resized to %" PRIi64 " bytes",
3640                     total_size, new_size);
3641     }
3642 
3643     qprintf(quiet, "Image resized.\n");
3644 
3645 out:
3646     blk_unref(blk);
3647     if (ret) {
3648         return 1;
3649     }
3650     return 0;
3651 }
3652 
3653 static void amend_status_cb(BlockDriverState *bs,
3654                             int64_t offset, int64_t total_work_size,
3655                             void *opaque)
3656 {
3657     qemu_progress_print(100.f * offset / total_work_size, 0);
3658 }
3659 
3660 static int img_amend(int argc, char **argv)
3661 {
3662     Error *err = NULL;
3663     int c, ret = 0;
3664     char *options = NULL;
3665     QemuOptsList *create_opts = NULL;
3666     QemuOpts *opts = NULL;
3667     const char *fmt = NULL, *filename, *cache;
3668     int flags;
3669     bool writethrough;
3670     bool quiet = false, progress = false;
3671     BlockBackend *blk = NULL;
3672     BlockDriverState *bs = NULL;
3673     bool image_opts = false;
3674 
3675     cache = BDRV_DEFAULT_CACHE;
3676     for (;;) {
3677         static const struct option long_options[] = {
3678             {"help", no_argument, 0, 'h'},
3679             {"object", required_argument, 0, OPTION_OBJECT},
3680             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
3681             {0, 0, 0, 0}
3682         };
3683         c = getopt_long(argc, argv, ":ho:f:t:pq",
3684                         long_options, NULL);
3685         if (c == -1) {
3686             break;
3687         }
3688 
3689         switch (c) {
3690         case ':':
3691             missing_argument(argv[optind - 1]);
3692             break;
3693         case '?':
3694             unrecognized_option(argv[optind - 1]);
3695             break;
3696         case 'h':
3697             help();
3698             break;
3699         case 'o':
3700             if (!is_valid_option_list(optarg)) {
3701                 error_report("Invalid option list: %s", optarg);
3702                 ret = -1;
3703                 goto out_no_progress;
3704             }
3705             if (!options) {
3706                 options = g_strdup(optarg);
3707             } else {
3708                 char *old_options = options;
3709                 options = g_strdup_printf("%s,%s", options, optarg);
3710                 g_free(old_options);
3711             }
3712             break;
3713         case 'f':
3714             fmt = optarg;
3715             break;
3716         case 't':
3717             cache = optarg;
3718             break;
3719         case 'p':
3720             progress = true;
3721             break;
3722         case 'q':
3723             quiet = true;
3724             break;
3725         case OPTION_OBJECT:
3726             opts = qemu_opts_parse_noisily(&qemu_object_opts,
3727                                            optarg, true);
3728             if (!opts) {
3729                 ret = -1;
3730                 goto out_no_progress;
3731             }
3732             break;
3733         case OPTION_IMAGE_OPTS:
3734             image_opts = true;
3735             break;
3736         }
3737     }
3738 
3739     if (!options) {
3740         error_exit("Must specify options (-o)");
3741     }
3742 
3743     if (qemu_opts_foreach(&qemu_object_opts,
3744                           user_creatable_add_opts_foreach,
3745                           NULL, NULL)) {
3746         ret = -1;
3747         goto out_no_progress;
3748     }
3749 
3750     if (quiet) {
3751         progress = false;
3752     }
3753     qemu_progress_init(progress, 1.0);
3754 
3755     filename = (optind == argc - 1) ? argv[argc - 1] : NULL;
3756     if (fmt && has_help_option(options)) {
3757         /* If a format is explicitly specified (and possibly no filename is
3758          * given), print option help here */
3759         ret = print_block_option_help(filename, fmt);
3760         goto out;
3761     }
3762 
3763     if (optind != argc - 1) {
3764         error_report("Expecting one image file name");
3765         ret = -1;
3766         goto out;
3767     }
3768 
3769     flags = BDRV_O_RDWR;
3770     ret = bdrv_parse_cache_mode(cache, &flags, &writethrough);
3771     if (ret < 0) {
3772         error_report("Invalid cache option: %s", cache);
3773         goto out;
3774     }
3775 
3776     blk = img_open(image_opts, filename, fmt, flags, writethrough, quiet,
3777                    false);
3778     if (!blk) {
3779         ret = -1;
3780         goto out;
3781     }
3782     bs = blk_bs(blk);
3783 
3784     fmt = bs->drv->format_name;
3785 
3786     if (has_help_option(options)) {
3787         /* If the format was auto-detected, print option help here */
3788         ret = print_block_option_help(filename, fmt);
3789         goto out;
3790     }
3791 
3792     if (!bs->drv->create_opts) {
3793         error_report("Format driver '%s' does not support any options to amend",
3794                      fmt);
3795         ret = -1;
3796         goto out;
3797     }
3798 
3799     create_opts = qemu_opts_append(create_opts, bs->drv->create_opts);
3800     opts = qemu_opts_create(create_opts, NULL, 0, &error_abort);
3801     qemu_opts_do_parse(opts, options, NULL, &err);
3802     if (err) {
3803         error_report_err(err);
3804         ret = -1;
3805         goto out;
3806     }
3807 
3808     /* In case the driver does not call amend_status_cb() */
3809     qemu_progress_print(0.f, 0);
3810     ret = bdrv_amend_options(bs, opts, &amend_status_cb, NULL);
3811     qemu_progress_print(100.f, 0);
3812     if (ret < 0) {
3813         error_report("Error while amending options: %s", strerror(-ret));
3814         goto out;
3815     }
3816 
3817 out:
3818     qemu_progress_end();
3819 
3820 out_no_progress:
3821     blk_unref(blk);
3822     qemu_opts_del(opts);
3823     qemu_opts_free(create_opts);
3824     g_free(options);
3825 
3826     if (ret) {
3827         return 1;
3828     }
3829     return 0;
3830 }
3831 
3832 typedef struct BenchData {
3833     BlockBackend *blk;
3834     uint64_t image_size;
3835     bool write;
3836     int bufsize;
3837     int step;
3838     int nrreq;
3839     int n;
3840     int flush_interval;
3841     bool drain_on_flush;
3842     uint8_t *buf;
3843     QEMUIOVector *qiov;
3844 
3845     int in_flight;
3846     bool in_flush;
3847     uint64_t offset;
3848 } BenchData;
3849 
3850 static void bench_undrained_flush_cb(void *opaque, int ret)
3851 {
3852     if (ret < 0) {
3853         error_report("Failed flush request: %s", strerror(-ret));
3854         exit(EXIT_FAILURE);
3855     }
3856 }
3857 
3858 static void bench_cb(void *opaque, int ret)
3859 {
3860     BenchData *b = opaque;
3861     BlockAIOCB *acb;
3862 
3863     if (ret < 0) {
3864         error_report("Failed request: %s", strerror(-ret));
3865         exit(EXIT_FAILURE);
3866     }
3867 
3868     if (b->in_flush) {
3869         /* Just finished a flush with drained queue: Start next requests */
3870         assert(b->in_flight == 0);
3871         b->in_flush = false;
3872     } else if (b->in_flight > 0) {
3873         int remaining = b->n - b->in_flight;
3874 
3875         b->n--;
3876         b->in_flight--;
3877 
3878         /* Time for flush? Drain queue if requested, then flush */
3879         if (b->flush_interval && remaining % b->flush_interval == 0) {
3880             if (!b->in_flight || !b->drain_on_flush) {
3881                 BlockCompletionFunc *cb;
3882 
3883                 if (b->drain_on_flush) {
3884                     b->in_flush = true;
3885                     cb = bench_cb;
3886                 } else {
3887                     cb = bench_undrained_flush_cb;
3888                 }
3889 
3890                 acb = blk_aio_flush(b->blk, cb, b);
3891                 if (!acb) {
3892                     error_report("Failed to issue flush request");
3893                     exit(EXIT_FAILURE);
3894                 }
3895             }
3896             if (b->drain_on_flush) {
3897                 return;
3898             }
3899         }
3900     }
3901 
3902     while (b->n > b->in_flight && b->in_flight < b->nrreq) {
3903         int64_t offset = b->offset;
3904         /* blk_aio_* might look for completed I/Os and kick bench_cb
3905          * again, so make sure this operation is counted by in_flight
3906          * and b->offset is ready for the next submission.
3907          */
3908         b->in_flight++;
3909         b->offset += b->step;
3910         b->offset %= b->image_size;
3911         if (b->write) {
3912             acb = blk_aio_pwritev(b->blk, offset, b->qiov, 0, bench_cb, b);
3913         } else {
3914             acb = blk_aio_preadv(b->blk, offset, b->qiov, 0, bench_cb, b);
3915         }
3916         if (!acb) {
3917             error_report("Failed to issue request");
3918             exit(EXIT_FAILURE);
3919         }
3920     }
3921 }
3922 
3923 static int img_bench(int argc, char **argv)
3924 {
3925     int c, ret = 0;
3926     const char *fmt = NULL, *filename;
3927     bool quiet = false;
3928     bool image_opts = false;
3929     bool is_write = false;
3930     int count = 75000;
3931     int depth = 64;
3932     int64_t offset = 0;
3933     size_t bufsize = 4096;
3934     int pattern = 0;
3935     size_t step = 0;
3936     int flush_interval = 0;
3937     bool drain_on_flush = true;
3938     int64_t image_size;
3939     BlockBackend *blk = NULL;
3940     BenchData data = {};
3941     int flags = 0;
3942     bool writethrough = false;
3943     struct timeval t1, t2;
3944     int i;
3945     bool force_share = false;
3946     size_t buf_size;
3947 
3948     for (;;) {
3949         static const struct option long_options[] = {
3950             {"help", no_argument, 0, 'h'},
3951             {"flush-interval", required_argument, 0, OPTION_FLUSH_INTERVAL},
3952             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
3953             {"pattern", required_argument, 0, OPTION_PATTERN},
3954             {"no-drain", no_argument, 0, OPTION_NO_DRAIN},
3955             {"force-share", no_argument, 0, 'U'},
3956             {0, 0, 0, 0}
3957         };
3958         c = getopt_long(argc, argv, ":hc:d:f:no:qs:S:t:wU", long_options, NULL);
3959         if (c == -1) {
3960             break;
3961         }
3962 
3963         switch (c) {
3964         case ':':
3965             missing_argument(argv[optind - 1]);
3966             break;
3967         case '?':
3968             unrecognized_option(argv[optind - 1]);
3969             break;
3970         case 'h':
3971             help();
3972             break;
3973         case 'c':
3974         {
3975             unsigned long res;
3976 
3977             if (qemu_strtoul(optarg, NULL, 0, &res) < 0 || res > INT_MAX) {
3978                 error_report("Invalid request count specified");
3979                 return 1;
3980             }
3981             count = res;
3982             break;
3983         }
3984         case 'd':
3985         {
3986             unsigned long res;
3987 
3988             if (qemu_strtoul(optarg, NULL, 0, &res) < 0 || res > INT_MAX) {
3989                 error_report("Invalid queue depth specified");
3990                 return 1;
3991             }
3992             depth = res;
3993             break;
3994         }
3995         case 'f':
3996             fmt = optarg;
3997             break;
3998         case 'n':
3999             flags |= BDRV_O_NATIVE_AIO;
4000             break;
4001         case 'o':
4002         {
4003             offset = cvtnum(optarg);
4004             if (offset < 0) {
4005                 error_report("Invalid offset specified");
4006                 return 1;
4007             }
4008             break;
4009         }
4010             break;
4011         case 'q':
4012             quiet = true;
4013             break;
4014         case 's':
4015         {
4016             int64_t sval;
4017 
4018             sval = cvtnum(optarg);
4019             if (sval < 0 || sval > INT_MAX) {
4020                 error_report("Invalid buffer size specified");
4021                 return 1;
4022             }
4023 
4024             bufsize = sval;
4025             break;
4026         }
4027         case 'S':
4028         {
4029             int64_t sval;
4030 
4031             sval = cvtnum(optarg);
4032             if (sval < 0 || sval > INT_MAX) {
4033                 error_report("Invalid step size specified");
4034                 return 1;
4035             }
4036 
4037             step = sval;
4038             break;
4039         }
4040         case 't':
4041             ret = bdrv_parse_cache_mode(optarg, &flags, &writethrough);
4042             if (ret < 0) {
4043                 error_report("Invalid cache mode");
4044                 ret = -1;
4045                 goto out;
4046             }
4047             break;
4048         case 'w':
4049             flags |= BDRV_O_RDWR;
4050             is_write = true;
4051             break;
4052         case 'U':
4053             force_share = true;
4054             break;
4055         case OPTION_PATTERN:
4056         {
4057             unsigned long res;
4058 
4059             if (qemu_strtoul(optarg, NULL, 0, &res) < 0 || res > 0xff) {
4060                 error_report("Invalid pattern byte specified");
4061                 return 1;
4062             }
4063             pattern = res;
4064             break;
4065         }
4066         case OPTION_FLUSH_INTERVAL:
4067         {
4068             unsigned long res;
4069 
4070             if (qemu_strtoul(optarg, NULL, 0, &res) < 0 || res > INT_MAX) {
4071                 error_report("Invalid flush interval specified");
4072                 return 1;
4073             }
4074             flush_interval = res;
4075             break;
4076         }
4077         case OPTION_NO_DRAIN:
4078             drain_on_flush = false;
4079             break;
4080         case OPTION_IMAGE_OPTS:
4081             image_opts = true;
4082             break;
4083         }
4084     }
4085 
4086     if (optind != argc - 1) {
4087         error_exit("Expecting one image file name");
4088     }
4089     filename = argv[argc - 1];
4090 
4091     if (!is_write && flush_interval) {
4092         error_report("--flush-interval is only available in write tests");
4093         ret = -1;
4094         goto out;
4095     }
4096     if (flush_interval && flush_interval < depth) {
4097         error_report("Flush interval can't be smaller than depth");
4098         ret = -1;
4099         goto out;
4100     }
4101 
4102     blk = img_open(image_opts, filename, fmt, flags, writethrough, quiet,
4103                    force_share);
4104     if (!blk) {
4105         ret = -1;
4106         goto out;
4107     }
4108 
4109     image_size = blk_getlength(blk);
4110     if (image_size < 0) {
4111         ret = image_size;
4112         goto out;
4113     }
4114 
4115     data = (BenchData) {
4116         .blk            = blk,
4117         .image_size     = image_size,
4118         .bufsize        = bufsize,
4119         .step           = step ?: bufsize,
4120         .nrreq          = depth,
4121         .n              = count,
4122         .offset         = offset,
4123         .write          = is_write,
4124         .flush_interval = flush_interval,
4125         .drain_on_flush = drain_on_flush,
4126     };
4127     printf("Sending %d %s requests, %d bytes each, %d in parallel "
4128            "(starting at offset %" PRId64 ", step size %d)\n",
4129            data.n, data.write ? "write" : "read", data.bufsize, data.nrreq,
4130            data.offset, data.step);
4131     if (flush_interval) {
4132         printf("Sending flush every %d requests\n", flush_interval);
4133     }
4134 
4135     buf_size = data.nrreq * data.bufsize;
4136     data.buf = blk_blockalign(blk, buf_size);
4137     memset(data.buf, pattern, data.nrreq * data.bufsize);
4138 
4139     blk_register_buf(blk, data.buf, buf_size);
4140 
4141     data.qiov = g_new(QEMUIOVector, data.nrreq);
4142     for (i = 0; i < data.nrreq; i++) {
4143         qemu_iovec_init(&data.qiov[i], 1);
4144         qemu_iovec_add(&data.qiov[i],
4145                        data.buf + i * data.bufsize, data.bufsize);
4146     }
4147 
4148     gettimeofday(&t1, NULL);
4149     bench_cb(&data, 0);
4150 
4151     while (data.n > 0) {
4152         main_loop_wait(false);
4153     }
4154     gettimeofday(&t2, NULL);
4155 
4156     printf("Run completed in %3.3f seconds.\n",
4157            (t2.tv_sec - t1.tv_sec)
4158            + ((double)(t2.tv_usec - t1.tv_usec) / 1000000));
4159 
4160 out:
4161     if (data.buf) {
4162         blk_unregister_buf(blk, data.buf);
4163     }
4164     qemu_vfree(data.buf);
4165     blk_unref(blk);
4166 
4167     if (ret) {
4168         return 1;
4169     }
4170     return 0;
4171 }
4172 
4173 #define C_BS      01
4174 #define C_COUNT   02
4175 #define C_IF      04
4176 #define C_OF      010
4177 #define C_SKIP    020
4178 
4179 struct DdInfo {
4180     unsigned int flags;
4181     int64_t count;
4182 };
4183 
4184 struct DdIo {
4185     int bsz;    /* Block size */
4186     char *filename;
4187     uint8_t *buf;
4188     int64_t offset;
4189 };
4190 
4191 struct DdOpts {
4192     const char *name;
4193     int (*f)(const char *, struct DdIo *, struct DdIo *, struct DdInfo *);
4194     unsigned int flag;
4195 };
4196 
4197 static int img_dd_bs(const char *arg,
4198                      struct DdIo *in, struct DdIo *out,
4199                      struct DdInfo *dd)
4200 {
4201     int64_t res;
4202 
4203     res = cvtnum(arg);
4204 
4205     if (res <= 0 || res > INT_MAX) {
4206         error_report("invalid number: '%s'", arg);
4207         return 1;
4208     }
4209     in->bsz = out->bsz = res;
4210 
4211     return 0;
4212 }
4213 
4214 static int img_dd_count(const char *arg,
4215                         struct DdIo *in, struct DdIo *out,
4216                         struct DdInfo *dd)
4217 {
4218     dd->count = cvtnum(arg);
4219 
4220     if (dd->count < 0) {
4221         error_report("invalid number: '%s'", arg);
4222         return 1;
4223     }
4224 
4225     return 0;
4226 }
4227 
4228 static int img_dd_if(const char *arg,
4229                      struct DdIo *in, struct DdIo *out,
4230                      struct DdInfo *dd)
4231 {
4232     in->filename = g_strdup(arg);
4233 
4234     return 0;
4235 }
4236 
4237 static int img_dd_of(const char *arg,
4238                      struct DdIo *in, struct DdIo *out,
4239                      struct DdInfo *dd)
4240 {
4241     out->filename = g_strdup(arg);
4242 
4243     return 0;
4244 }
4245 
4246 static int img_dd_skip(const char *arg,
4247                        struct DdIo *in, struct DdIo *out,
4248                        struct DdInfo *dd)
4249 {
4250     in->offset = cvtnum(arg);
4251 
4252     if (in->offset < 0) {
4253         error_report("invalid number: '%s'", arg);
4254         return 1;
4255     }
4256 
4257     return 0;
4258 }
4259 
4260 static int img_dd(int argc, char **argv)
4261 {
4262     int ret = 0;
4263     char *arg = NULL;
4264     char *tmp;
4265     BlockDriver *drv = NULL, *proto_drv = NULL;
4266     BlockBackend *blk1 = NULL, *blk2 = NULL;
4267     QemuOpts *opts = NULL;
4268     QemuOptsList *create_opts = NULL;
4269     Error *local_err = NULL;
4270     bool image_opts = false;
4271     int c, i;
4272     const char *out_fmt = "raw";
4273     const char *fmt = NULL;
4274     int64_t size = 0;
4275     int64_t block_count = 0, out_pos, in_pos;
4276     bool force_share = false;
4277     struct DdInfo dd = {
4278         .flags = 0,
4279         .count = 0,
4280     };
4281     struct DdIo in = {
4282         .bsz = 512, /* Block size is by default 512 bytes */
4283         .filename = NULL,
4284         .buf = NULL,
4285         .offset = 0
4286     };
4287     struct DdIo out = {
4288         .bsz = 512,
4289         .filename = NULL,
4290         .buf = NULL,
4291         .offset = 0
4292     };
4293 
4294     const struct DdOpts options[] = {
4295         { "bs", img_dd_bs, C_BS },
4296         { "count", img_dd_count, C_COUNT },
4297         { "if", img_dd_if, C_IF },
4298         { "of", img_dd_of, C_OF },
4299         { "skip", img_dd_skip, C_SKIP },
4300         { NULL, NULL, 0 }
4301     };
4302     const struct option long_options[] = {
4303         { "help", no_argument, 0, 'h'},
4304         { "object", required_argument, 0, OPTION_OBJECT},
4305         { "image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
4306         { "force-share", no_argument, 0, 'U'},
4307         { 0, 0, 0, 0 }
4308     };
4309 
4310     while ((c = getopt_long(argc, argv, ":hf:O:U", long_options, NULL))) {
4311         if (c == EOF) {
4312             break;
4313         }
4314         switch (c) {
4315         case 'O':
4316             out_fmt = optarg;
4317             break;
4318         case 'f':
4319             fmt = optarg;
4320             break;
4321         case ':':
4322             missing_argument(argv[optind - 1]);
4323             break;
4324         case '?':
4325             unrecognized_option(argv[optind - 1]);
4326             break;
4327         case 'h':
4328             help();
4329             break;
4330         case 'U':
4331             force_share = true;
4332             break;
4333         case OPTION_OBJECT:
4334             if (!qemu_opts_parse_noisily(&qemu_object_opts, optarg, true)) {
4335                 ret = -1;
4336                 goto out;
4337             }
4338             break;
4339         case OPTION_IMAGE_OPTS:
4340             image_opts = true;
4341             break;
4342         }
4343     }
4344 
4345     for (i = optind; i < argc; i++) {
4346         int j;
4347         arg = g_strdup(argv[i]);
4348 
4349         tmp = strchr(arg, '=');
4350         if (tmp == NULL) {
4351             error_report("unrecognized operand %s", arg);
4352             ret = -1;
4353             goto out;
4354         }
4355 
4356         *tmp++ = '\0';
4357 
4358         for (j = 0; options[j].name != NULL; j++) {
4359             if (!strcmp(arg, options[j].name)) {
4360                 break;
4361             }
4362         }
4363         if (options[j].name == NULL) {
4364             error_report("unrecognized operand %s", arg);
4365             ret = -1;
4366             goto out;
4367         }
4368 
4369         if (options[j].f(tmp, &in, &out, &dd) != 0) {
4370             ret = -1;
4371             goto out;
4372         }
4373         dd.flags |= options[j].flag;
4374         g_free(arg);
4375         arg = NULL;
4376     }
4377 
4378     if (!(dd.flags & C_IF && dd.flags & C_OF)) {
4379         error_report("Must specify both input and output files");
4380         ret = -1;
4381         goto out;
4382     }
4383 
4384     if (qemu_opts_foreach(&qemu_object_opts,
4385                           user_creatable_add_opts_foreach,
4386                           NULL, NULL)) {
4387         ret = -1;
4388         goto out;
4389     }
4390 
4391     blk1 = img_open(image_opts, in.filename, fmt, 0, false, false,
4392                     force_share);
4393 
4394     if (!blk1) {
4395         ret = -1;
4396         goto out;
4397     }
4398 
4399     drv = bdrv_find_format(out_fmt);
4400     if (!drv) {
4401         error_report("Unknown file format");
4402         ret = -1;
4403         goto out;
4404     }
4405     proto_drv = bdrv_find_protocol(out.filename, true, &local_err);
4406 
4407     if (!proto_drv) {
4408         error_report_err(local_err);
4409         ret = -1;
4410         goto out;
4411     }
4412     if (!drv->create_opts) {
4413         error_report("Format driver '%s' does not support image creation",
4414                      drv->format_name);
4415         ret = -1;
4416         goto out;
4417     }
4418     if (!proto_drv->create_opts) {
4419         error_report("Protocol driver '%s' does not support image creation",
4420                      proto_drv->format_name);
4421         ret = -1;
4422         goto out;
4423     }
4424     create_opts = qemu_opts_append(create_opts, drv->create_opts);
4425     create_opts = qemu_opts_append(create_opts, proto_drv->create_opts);
4426 
4427     opts = qemu_opts_create(create_opts, NULL, 0, &error_abort);
4428 
4429     size = blk_getlength(blk1);
4430     if (size < 0) {
4431         error_report("Failed to get size for '%s'", in.filename);
4432         ret = -1;
4433         goto out;
4434     }
4435 
4436     if (dd.flags & C_COUNT && dd.count <= INT64_MAX / in.bsz &&
4437         dd.count * in.bsz < size) {
4438         size = dd.count * in.bsz;
4439     }
4440 
4441     /* Overflow means the specified offset is beyond input image's size */
4442     if (dd.flags & C_SKIP && (in.offset > INT64_MAX / in.bsz ||
4443                               size < in.bsz * in.offset)) {
4444         qemu_opt_set_number(opts, BLOCK_OPT_SIZE, 0, &error_abort);
4445     } else {
4446         qemu_opt_set_number(opts, BLOCK_OPT_SIZE,
4447                             size - in.bsz * in.offset, &error_abort);
4448     }
4449 
4450     ret = bdrv_create(drv, out.filename, opts, &local_err);
4451     if (ret < 0) {
4452         error_reportf_err(local_err,
4453                           "%s: error while creating output image: ",
4454                           out.filename);
4455         ret = -1;
4456         goto out;
4457     }
4458 
4459     /* TODO, we can't honour --image-opts for the target,
4460      * since it needs to be given in a format compatible
4461      * with the bdrv_create() call above which does not
4462      * support image-opts style.
4463      */
4464     blk2 = img_open_file(out.filename, NULL, out_fmt, BDRV_O_RDWR,
4465                          false, false, false);
4466 
4467     if (!blk2) {
4468         ret = -1;
4469         goto out;
4470     }
4471 
4472     if (dd.flags & C_SKIP && (in.offset > INT64_MAX / in.bsz ||
4473                               size < in.offset * in.bsz)) {
4474         /* We give a warning if the skip option is bigger than the input
4475          * size and create an empty output disk image (i.e. like dd(1)).
4476          */
4477         error_report("%s: cannot skip to specified offset", in.filename);
4478         in_pos = size;
4479     } else {
4480         in_pos = in.offset * in.bsz;
4481     }
4482 
4483     in.buf = g_new(uint8_t, in.bsz);
4484 
4485     for (out_pos = 0; in_pos < size; block_count++) {
4486         int in_ret, out_ret;
4487 
4488         if (in_pos + in.bsz > size) {
4489             in_ret = blk_pread(blk1, in_pos, in.buf, size - in_pos);
4490         } else {
4491             in_ret = blk_pread(blk1, in_pos, in.buf, in.bsz);
4492         }
4493         if (in_ret < 0) {
4494             error_report("error while reading from input image file: %s",
4495                          strerror(-in_ret));
4496             ret = -1;
4497             goto out;
4498         }
4499         in_pos += in_ret;
4500 
4501         out_ret = blk_pwrite(blk2, out_pos, in.buf, in_ret, 0);
4502 
4503         if (out_ret < 0) {
4504             error_report("error while writing to output image file: %s",
4505                          strerror(-out_ret));
4506             ret = -1;
4507             goto out;
4508         }
4509         out_pos += out_ret;
4510     }
4511 
4512 out:
4513     g_free(arg);
4514     qemu_opts_del(opts);
4515     qemu_opts_free(create_opts);
4516     blk_unref(blk1);
4517     blk_unref(blk2);
4518     g_free(in.filename);
4519     g_free(out.filename);
4520     g_free(in.buf);
4521     g_free(out.buf);
4522 
4523     if (ret) {
4524         return 1;
4525     }
4526     return 0;
4527 }
4528 
4529 static void dump_json_block_measure_info(BlockMeasureInfo *info)
4530 {
4531     QString *str;
4532     QObject *obj;
4533     Visitor *v = qobject_output_visitor_new(&obj);
4534 
4535     visit_type_BlockMeasureInfo(v, NULL, &info, &error_abort);
4536     visit_complete(v, &obj);
4537     str = qobject_to_json_pretty(obj);
4538     assert(str != NULL);
4539     printf("%s\n", qstring_get_str(str));
4540     qobject_unref(obj);
4541     visit_free(v);
4542     qobject_unref(str);
4543 }
4544 
4545 static int img_measure(int argc, char **argv)
4546 {
4547     static const struct option long_options[] = {
4548         {"help", no_argument, 0, 'h'},
4549         {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
4550         {"object", required_argument, 0, OPTION_OBJECT},
4551         {"output", required_argument, 0, OPTION_OUTPUT},
4552         {"size", required_argument, 0, OPTION_SIZE},
4553         {"force-share", no_argument, 0, 'U'},
4554         {0, 0, 0, 0}
4555     };
4556     OutputFormat output_format = OFORMAT_HUMAN;
4557     BlockBackend *in_blk = NULL;
4558     BlockDriver *drv;
4559     const char *filename = NULL;
4560     const char *fmt = NULL;
4561     const char *out_fmt = "raw";
4562     char *options = NULL;
4563     char *snapshot_name = NULL;
4564     bool force_share = false;
4565     QemuOpts *opts = NULL;
4566     QemuOpts *object_opts = NULL;
4567     QemuOpts *sn_opts = NULL;
4568     QemuOptsList *create_opts = NULL;
4569     bool image_opts = false;
4570     uint64_t img_size = UINT64_MAX;
4571     BlockMeasureInfo *info = NULL;
4572     Error *local_err = NULL;
4573     int ret = 1;
4574     int c;
4575 
4576     while ((c = getopt_long(argc, argv, "hf:O:o:l:U",
4577                             long_options, NULL)) != -1) {
4578         switch (c) {
4579         case '?':
4580         case 'h':
4581             help();
4582             break;
4583         case 'f':
4584             fmt = optarg;
4585             break;
4586         case 'O':
4587             out_fmt = optarg;
4588             break;
4589         case 'o':
4590             if (!is_valid_option_list(optarg)) {
4591                 error_report("Invalid option list: %s", optarg);
4592                 goto out;
4593             }
4594             if (!options) {
4595                 options = g_strdup(optarg);
4596             } else {
4597                 char *old_options = options;
4598                 options = g_strdup_printf("%s,%s", options, optarg);
4599                 g_free(old_options);
4600             }
4601             break;
4602         case 'l':
4603             if (strstart(optarg, SNAPSHOT_OPT_BASE, NULL)) {
4604                 sn_opts = qemu_opts_parse_noisily(&internal_snapshot_opts,
4605                                                   optarg, false);
4606                 if (!sn_opts) {
4607                     error_report("Failed in parsing snapshot param '%s'",
4608                                  optarg);
4609                     goto out;
4610                 }
4611             } else {
4612                 snapshot_name = optarg;
4613             }
4614             break;
4615         case 'U':
4616             force_share = true;
4617             break;
4618         case OPTION_OBJECT:
4619             object_opts = qemu_opts_parse_noisily(&qemu_object_opts,
4620                                                   optarg, true);
4621             if (!object_opts) {
4622                 goto out;
4623             }
4624             break;
4625         case OPTION_IMAGE_OPTS:
4626             image_opts = true;
4627             break;
4628         case OPTION_OUTPUT:
4629             if (!strcmp(optarg, "json")) {
4630                 output_format = OFORMAT_JSON;
4631             } else if (!strcmp(optarg, "human")) {
4632                 output_format = OFORMAT_HUMAN;
4633             } else {
4634                 error_report("--output must be used with human or json "
4635                              "as argument.");
4636                 goto out;
4637             }
4638             break;
4639         case OPTION_SIZE:
4640         {
4641             int64_t sval;
4642 
4643             sval = cvtnum(optarg);
4644             if (sval < 0) {
4645                 if (sval == -ERANGE) {
4646                     error_report("Image size must be less than 8 EiB!");
4647                 } else {
4648                     error_report("Invalid image size specified! You may use "
4649                                  "k, M, G, T, P or E suffixes for ");
4650                     error_report("kilobytes, megabytes, gigabytes, terabytes, "
4651                                  "petabytes and exabytes.");
4652                 }
4653                 goto out;
4654             }
4655             img_size = (uint64_t)sval;
4656         }
4657         break;
4658         }
4659     }
4660 
4661     if (qemu_opts_foreach(&qemu_object_opts,
4662                           user_creatable_add_opts_foreach,
4663                           NULL, NULL)) {
4664         goto out;
4665     }
4666 
4667     if (argc - optind > 1) {
4668         error_report("At most one filename argument is allowed.");
4669         goto out;
4670     } else if (argc - optind == 1) {
4671         filename = argv[optind];
4672     }
4673 
4674     if (!filename &&
4675         (object_opts || image_opts || fmt || snapshot_name || sn_opts)) {
4676         error_report("--object, --image-opts, -f, and -l "
4677                      "require a filename argument.");
4678         goto out;
4679     }
4680     if (filename && img_size != UINT64_MAX) {
4681         error_report("--size N cannot be used together with a filename.");
4682         goto out;
4683     }
4684     if (!filename && img_size == UINT64_MAX) {
4685         error_report("Either --size N or one filename must be specified.");
4686         goto out;
4687     }
4688 
4689     if (filename) {
4690         in_blk = img_open(image_opts, filename, fmt, 0,
4691                           false, false, force_share);
4692         if (!in_blk) {
4693             goto out;
4694         }
4695 
4696         if (sn_opts) {
4697             bdrv_snapshot_load_tmp(blk_bs(in_blk),
4698                     qemu_opt_get(sn_opts, SNAPSHOT_OPT_ID),
4699                     qemu_opt_get(sn_opts, SNAPSHOT_OPT_NAME),
4700                     &local_err);
4701         } else if (snapshot_name != NULL) {
4702             bdrv_snapshot_load_tmp_by_id_or_name(blk_bs(in_blk),
4703                     snapshot_name, &local_err);
4704         }
4705         if (local_err) {
4706             error_reportf_err(local_err, "Failed to load snapshot: ");
4707             goto out;
4708         }
4709     }
4710 
4711     drv = bdrv_find_format(out_fmt);
4712     if (!drv) {
4713         error_report("Unknown file format '%s'", out_fmt);
4714         goto out;
4715     }
4716     if (!drv->create_opts) {
4717         error_report("Format driver '%s' does not support image creation",
4718                      drv->format_name);
4719         goto out;
4720     }
4721 
4722     create_opts = qemu_opts_append(create_opts, drv->create_opts);
4723     create_opts = qemu_opts_append(create_opts, bdrv_file.create_opts);
4724     opts = qemu_opts_create(create_opts, NULL, 0, &error_abort);
4725     if (options) {
4726         qemu_opts_do_parse(opts, options, NULL, &local_err);
4727         if (local_err) {
4728             error_report_err(local_err);
4729             error_report("Invalid options for file format '%s'", out_fmt);
4730             goto out;
4731         }
4732     }
4733     if (img_size != UINT64_MAX) {
4734         qemu_opt_set_number(opts, BLOCK_OPT_SIZE, img_size, &error_abort);
4735     }
4736 
4737     info = bdrv_measure(drv, opts, in_blk ? blk_bs(in_blk) : NULL, &local_err);
4738     if (local_err) {
4739         error_report_err(local_err);
4740         goto out;
4741     }
4742 
4743     if (output_format == OFORMAT_HUMAN) {
4744         printf("required size: %" PRIu64 "\n", info->required);
4745         printf("fully allocated size: %" PRIu64 "\n", info->fully_allocated);
4746     } else {
4747         dump_json_block_measure_info(info);
4748     }
4749 
4750     ret = 0;
4751 
4752 out:
4753     qapi_free_BlockMeasureInfo(info);
4754     qemu_opts_del(object_opts);
4755     qemu_opts_del(opts);
4756     qemu_opts_del(sn_opts);
4757     qemu_opts_free(create_opts);
4758     g_free(options);
4759     blk_unref(in_blk);
4760     return ret;
4761 }
4762 
4763 static const img_cmd_t img_cmds[] = {
4764 #define DEF(option, callback, arg_string)        \
4765     { option, callback },
4766 #include "qemu-img-cmds.h"
4767 #undef DEF
4768     { NULL, NULL, },
4769 };
4770 
4771 int main(int argc, char **argv)
4772 {
4773     const img_cmd_t *cmd;
4774     const char *cmdname;
4775     Error *local_error = NULL;
4776     char *trace_file = NULL;
4777     int c;
4778     static const struct option long_options[] = {
4779         {"help", no_argument, 0, 'h'},
4780         {"version", no_argument, 0, 'V'},
4781         {"trace", required_argument, NULL, 'T'},
4782         {0, 0, 0, 0}
4783     };
4784 
4785 #ifdef CONFIG_POSIX
4786     signal(SIGPIPE, SIG_IGN);
4787 #endif
4788 
4789     module_call_init(MODULE_INIT_TRACE);
4790     error_set_progname(argv[0]);
4791     qemu_init_exec_dir(argv[0]);
4792 
4793     if (qemu_init_main_loop(&local_error)) {
4794         error_report_err(local_error);
4795         exit(EXIT_FAILURE);
4796     }
4797 
4798     qcrypto_init(&error_fatal);
4799 
4800     module_call_init(MODULE_INIT_QOM);
4801     bdrv_init();
4802     if (argc < 2) {
4803         error_exit("Not enough arguments");
4804     }
4805 
4806     qemu_add_opts(&qemu_object_opts);
4807     qemu_add_opts(&qemu_source_opts);
4808     qemu_add_opts(&qemu_trace_opts);
4809 
4810     while ((c = getopt_long(argc, argv, "+:hVT:", long_options, NULL)) != -1) {
4811         switch (c) {
4812         case ':':
4813             missing_argument(argv[optind - 1]);
4814             return 0;
4815         case '?':
4816             unrecognized_option(argv[optind - 1]);
4817             return 0;
4818         case 'h':
4819             help();
4820             return 0;
4821         case 'V':
4822             printf(QEMU_IMG_VERSION);
4823             return 0;
4824         case 'T':
4825             g_free(trace_file);
4826             trace_file = trace_opt_parse(optarg);
4827             break;
4828         }
4829     }
4830 
4831     cmdname = argv[optind];
4832 
4833     /* reset getopt_long scanning */
4834     argc -= optind;
4835     if (argc < 1) {
4836         return 0;
4837     }
4838     argv += optind;
4839     optind = 0;
4840 
4841     if (!trace_init_backends()) {
4842         exit(1);
4843     }
4844     trace_init_file(trace_file);
4845     qemu_set_log(LOG_TRACE);
4846 
4847     /* find the command */
4848     for (cmd = img_cmds; cmd->name != NULL; cmd++) {
4849         if (!strcmp(cmdname, cmd->name)) {
4850             return cmd->handler(argc, argv);
4851         }
4852     }
4853 
4854     /* not found */
4855     error_exit("Command not found: %s", cmdname);
4856 }
4857