xref: /openbmc/qemu/qemu-img.c (revision 438c78da)
1 /*
2  * QEMU disk image utility
3  *
4  * Copyright (c) 2003-2008 Fabrice Bellard
5  *
6  * Permission is hereby granted, free of charge, to any person obtaining a copy
7  * of this software and associated documentation files (the "Software"), to deal
8  * in the Software without restriction, including without limitation the rights
9  * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
10  * copies of the Software, and to permit persons to whom the Software is
11  * furnished to do so, subject to the following conditions:
12  *
13  * The above copyright notice and this permission notice shall be included in
14  * all copies or substantial portions of the Software.
15  *
16  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
17  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
18  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
19  * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
20  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
21  * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
22  * THE SOFTWARE.
23  */
24 
25 #include "qemu/osdep.h"
26 #include <getopt.h>
27 
28 #include "qemu-version.h"
29 #include "qapi/error.h"
30 #include "qapi/qapi-visit-block-core.h"
31 #include "qapi/qobject-output-visitor.h"
32 #include "qapi/qmp/qjson.h"
33 #include "qapi/qmp/qdict.h"
34 #include "qapi/qmp/qstring.h"
35 #include "qemu/cutils.h"
36 #include "qemu/config-file.h"
37 #include "qemu/option.h"
38 #include "qemu/error-report.h"
39 #include "qemu/log.h"
40 #include "qom/object_interfaces.h"
41 #include "sysemu/sysemu.h"
42 #include "sysemu/block-backend.h"
43 #include "block/block_int.h"
44 #include "block/blockjob.h"
45 #include "block/qapi.h"
46 #include "crypto/init.h"
47 #include "trace/control.h"
48 
49 #define QEMU_IMG_VERSION "qemu-img version " QEMU_FULL_VERSION \
50                           "\n" QEMU_COPYRIGHT "\n"
51 
52 typedef struct img_cmd_t {
53     const char *name;
54     int (*handler)(int argc, char **argv);
55 } img_cmd_t;
56 
57 enum {
58     OPTION_OUTPUT = 256,
59     OPTION_BACKING_CHAIN = 257,
60     OPTION_OBJECT = 258,
61     OPTION_IMAGE_OPTS = 259,
62     OPTION_PATTERN = 260,
63     OPTION_FLUSH_INTERVAL = 261,
64     OPTION_NO_DRAIN = 262,
65     OPTION_TARGET_IMAGE_OPTS = 263,
66     OPTION_SIZE = 264,
67     OPTION_PREALLOCATION = 265,
68     OPTION_SHRINK = 266,
69 };
70 
71 typedef enum OutputFormat {
72     OFORMAT_JSON,
73     OFORMAT_HUMAN,
74 } OutputFormat;
75 
76 /* Default to cache=writeback as data integrity is not important for qemu-img */
77 #define BDRV_DEFAULT_CACHE "writeback"
78 
79 static void format_print(void *opaque, const char *name)
80 {
81     printf(" %s", name);
82 }
83 
84 static void QEMU_NORETURN GCC_FMT_ATTR(1, 2) error_exit(const char *fmt, ...)
85 {
86     va_list ap;
87 
88     error_printf("qemu-img: ");
89 
90     va_start(ap, fmt);
91     error_vprintf(fmt, ap);
92     va_end(ap);
93 
94     error_printf("\nTry 'qemu-img --help' for more information\n");
95     exit(EXIT_FAILURE);
96 }
97 
98 static void QEMU_NORETURN missing_argument(const char *option)
99 {
100     error_exit("missing argument for option '%s'", option);
101 }
102 
103 static void QEMU_NORETURN unrecognized_option(const char *option)
104 {
105     error_exit("unrecognized option '%s'", option);
106 }
107 
108 /* Please keep in synch with qemu-img.texi */
109 static void QEMU_NORETURN help(void)
110 {
111     const char *help_msg =
112            QEMU_IMG_VERSION
113            "usage: qemu-img [standard options] command [command options]\n"
114            "QEMU disk image utility\n"
115            "\n"
116            "    '-h', '--help'       display this help and exit\n"
117            "    '-V', '--version'    output version information and exit\n"
118            "    '-T', '--trace'      [[enable=]<pattern>][,events=<file>][,file=<file>]\n"
119            "                         specify tracing options\n"
120            "\n"
121            "Command syntax:\n"
122 #define DEF(option, callback, arg_string)        \
123            "  " arg_string "\n"
124 #include "qemu-img-cmds.h"
125 #undef DEF
126            "\n"
127            "Command parameters:\n"
128            "  'filename' is a disk image filename\n"
129            "  'objectdef' is a QEMU user creatable object definition. See the qemu(1)\n"
130            "    manual page for a description of the object properties. The most common\n"
131            "    object type is a 'secret', which is used to supply passwords and/or\n"
132            "    encryption keys.\n"
133            "  'fmt' is the disk image format. It is guessed automatically in most cases\n"
134            "  'cache' is the cache mode used to write the output disk image, the valid\n"
135            "    options are: 'none', 'writeback' (default, except for convert), 'writethrough',\n"
136            "    'directsync' and 'unsafe' (default for convert)\n"
137            "  'src_cache' is the cache mode used to read input disk images, the valid\n"
138            "    options are the same as for the 'cache' option\n"
139            "  'size' is the disk image size in bytes. Optional suffixes\n"
140            "    'k' or 'K' (kilobyte, 1024), 'M' (megabyte, 1024k), 'G' (gigabyte, 1024M),\n"
141            "    'T' (terabyte, 1024G), 'P' (petabyte, 1024T) and 'E' (exabyte, 1024P)  are\n"
142            "    supported. 'b' is ignored.\n"
143            "  'output_filename' is the destination disk image filename\n"
144            "  'output_fmt' is the destination format\n"
145            "  'options' is a comma separated list of format specific options in a\n"
146            "    name=value format. Use -o ? for an overview of the options supported by the\n"
147            "    used format\n"
148            "  'snapshot_param' is param used for internal snapshot, format\n"
149            "    is 'snapshot.id=[ID],snapshot.name=[NAME]', or\n"
150            "    '[ID_OR_NAME]'\n"
151            "  '-c' indicates that target image must be compressed (qcow format only)\n"
152            "  '-u' allows unsafe backing chains. For rebasing, it is assumed that old and\n"
153            "       new backing file match exactly. The image doesn't need a working\n"
154            "       backing file before rebasing in this case (useful for renaming the\n"
155            "       backing file). For image creation, allow creating without attempting\n"
156            "       to open the backing file.\n"
157            "  '-h' with or without a command shows this help and lists the supported formats\n"
158            "  '-p' show progress of command (only certain commands)\n"
159            "  '-q' use Quiet mode - do not print any output (except errors)\n"
160            "  '-S' indicates the consecutive number of bytes (defaults to 4k) that must\n"
161            "       contain only zeros for qemu-img to create a sparse image during\n"
162            "       conversion. If the number of bytes is 0, the source will not be scanned for\n"
163            "       unallocated or zero sectors, and the destination image will always be\n"
164            "       fully allocated\n"
165            "  '--output' takes the format in which the output must be done (human or json)\n"
166            "  '-n' skips the target volume creation (useful if the volume is created\n"
167            "       prior to running qemu-img)\n"
168            "\n"
169            "Parameters to check subcommand:\n"
170            "  '-r' tries to repair any inconsistencies that are found during the check.\n"
171            "       '-r leaks' repairs only cluster leaks, whereas '-r all' fixes all\n"
172            "       kinds of errors, with a higher risk of choosing the wrong fix or\n"
173            "       hiding corruption that has already occurred.\n"
174            "\n"
175            "Parameters to convert subcommand:\n"
176            "  '-m' specifies how many coroutines work in parallel during the convert\n"
177            "       process (defaults to 8)\n"
178            "  '-W' allow to write to the target out of order rather than sequential\n"
179            "\n"
180            "Parameters to snapshot subcommand:\n"
181            "  'snapshot' is the name of the snapshot to create, apply or delete\n"
182            "  '-a' applies a snapshot (revert disk to saved state)\n"
183            "  '-c' creates a snapshot\n"
184            "  '-d' deletes a snapshot\n"
185            "  '-l' lists all snapshots in the given image\n"
186            "\n"
187            "Parameters to compare subcommand:\n"
188            "  '-f' first image format\n"
189            "  '-F' second image format\n"
190            "  '-s' run in Strict mode - fail on different image size or sector allocation\n"
191            "\n"
192            "Parameters to dd subcommand:\n"
193            "  'bs=BYTES' read and write up to BYTES bytes at a time "
194            "(default: 512)\n"
195            "  'count=N' copy only N input blocks\n"
196            "  'if=FILE' read from FILE\n"
197            "  'of=FILE' write to FILE\n"
198            "  'skip=N' skip N bs-sized blocks at the start of input\n";
199 
200     printf("%s\nSupported formats:", help_msg);
201     bdrv_iterate_format(format_print, NULL);
202     printf("\n\n" QEMU_HELP_BOTTOM "\n");
203     exit(EXIT_SUCCESS);
204 }
205 
206 static QemuOptsList qemu_object_opts = {
207     .name = "object",
208     .implied_opt_name = "qom-type",
209     .head = QTAILQ_HEAD_INITIALIZER(qemu_object_opts.head),
210     .desc = {
211         { }
212     },
213 };
214 
215 static QemuOptsList qemu_source_opts = {
216     .name = "source",
217     .implied_opt_name = "file",
218     .head = QTAILQ_HEAD_INITIALIZER(qemu_source_opts.head),
219     .desc = {
220         { }
221     },
222 };
223 
224 static int GCC_FMT_ATTR(2, 3) qprintf(bool quiet, const char *fmt, ...)
225 {
226     int ret = 0;
227     if (!quiet) {
228         va_list args;
229         va_start(args, fmt);
230         ret = vprintf(fmt, args);
231         va_end(args);
232     }
233     return ret;
234 }
235 
236 
237 static int print_block_option_help(const char *filename, const char *fmt)
238 {
239     BlockDriver *drv, *proto_drv;
240     QemuOptsList *create_opts = NULL;
241     Error *local_err = NULL;
242 
243     /* Find driver and parse its options */
244     drv = bdrv_find_format(fmt);
245     if (!drv) {
246         error_report("Unknown file format '%s'", fmt);
247         return 1;
248     }
249 
250     if (!drv->create_opts) {
251         error_report("Format driver '%s' does not support image creation", fmt);
252         return 1;
253     }
254 
255     create_opts = qemu_opts_append(create_opts, drv->create_opts);
256     if (filename) {
257         proto_drv = bdrv_find_protocol(filename, true, &local_err);
258         if (!proto_drv) {
259             error_report_err(local_err);
260             qemu_opts_free(create_opts);
261             return 1;
262         }
263         if (!proto_drv->create_opts) {
264             error_report("Protocal driver '%s' does not support image creation",
265                          proto_drv->format_name);
266             return 1;
267         }
268         create_opts = qemu_opts_append(create_opts, proto_drv->create_opts);
269     }
270 
271     printf("Supported options:\n");
272     qemu_opts_print_help(create_opts);
273     qemu_opts_free(create_opts);
274     return 0;
275 }
276 
277 
278 static BlockBackend *img_open_opts(const char *optstr,
279                                    QemuOpts *opts, int flags, bool writethrough,
280                                    bool quiet, bool force_share)
281 {
282     QDict *options;
283     Error *local_err = NULL;
284     BlockBackend *blk;
285     options = qemu_opts_to_qdict(opts, NULL);
286     if (force_share) {
287         if (qdict_haskey(options, BDRV_OPT_FORCE_SHARE)
288             && strcmp(qdict_get_str(options, BDRV_OPT_FORCE_SHARE), "on")) {
289             error_report("--force-share/-U conflicts with image options");
290             qobject_unref(options);
291             return NULL;
292         }
293         qdict_put_str(options, BDRV_OPT_FORCE_SHARE, "on");
294     }
295     blk = blk_new_open(NULL, NULL, options, flags, &local_err);
296     if (!blk) {
297         error_reportf_err(local_err, "Could not open '%s': ", optstr);
298         return NULL;
299     }
300     blk_set_enable_write_cache(blk, !writethrough);
301 
302     return blk;
303 }
304 
305 static BlockBackend *img_open_file(const char *filename,
306                                    QDict *options,
307                                    const char *fmt, int flags,
308                                    bool writethrough, bool quiet,
309                                    bool force_share)
310 {
311     BlockBackend *blk;
312     Error *local_err = NULL;
313 
314     if (!options) {
315         options = qdict_new();
316     }
317     if (fmt) {
318         qdict_put_str(options, "driver", fmt);
319     }
320 
321     if (force_share) {
322         qdict_put_bool(options, BDRV_OPT_FORCE_SHARE, true);
323     }
324     blk = blk_new_open(filename, NULL, options, flags, &local_err);
325     if (!blk) {
326         error_reportf_err(local_err, "Could not open '%s': ", filename);
327         return NULL;
328     }
329     blk_set_enable_write_cache(blk, !writethrough);
330 
331     return blk;
332 }
333 
334 
335 static int img_add_key_secrets(void *opaque,
336                                const char *name, const char *value,
337                                Error **errp)
338 {
339     QDict *options = opaque;
340 
341     if (g_str_has_suffix(name, "key-secret")) {
342         qdict_put_str(options, name, value);
343     }
344 
345     return 0;
346 }
347 
348 
349 static BlockBackend *img_open(bool image_opts,
350                               const char *filename,
351                               const char *fmt, int flags, bool writethrough,
352                               bool quiet, bool force_share)
353 {
354     BlockBackend *blk;
355     if (image_opts) {
356         QemuOpts *opts;
357         if (fmt) {
358             error_report("--image-opts and --format are mutually exclusive");
359             return NULL;
360         }
361         opts = qemu_opts_parse_noisily(qemu_find_opts("source"),
362                                        filename, true);
363         if (!opts) {
364             return NULL;
365         }
366         blk = img_open_opts(filename, opts, flags, writethrough, quiet,
367                             force_share);
368     } else {
369         blk = img_open_file(filename, NULL, fmt, flags, writethrough, quiet,
370                             force_share);
371     }
372     return blk;
373 }
374 
375 
376 static int add_old_style_options(const char *fmt, QemuOpts *opts,
377                                  const char *base_filename,
378                                  const char *base_fmt)
379 {
380     Error *err = NULL;
381 
382     if (base_filename) {
383         qemu_opt_set(opts, BLOCK_OPT_BACKING_FILE, base_filename, &err);
384         if (err) {
385             error_report("Backing file not supported for file format '%s'",
386                          fmt);
387             error_free(err);
388             return -1;
389         }
390     }
391     if (base_fmt) {
392         qemu_opt_set(opts, BLOCK_OPT_BACKING_FMT, base_fmt, &err);
393         if (err) {
394             error_report("Backing file format not supported for file "
395                          "format '%s'", fmt);
396             error_free(err);
397             return -1;
398         }
399     }
400     return 0;
401 }
402 
403 static int64_t cvtnum(const char *s)
404 {
405     int err;
406     uint64_t value;
407 
408     err = qemu_strtosz(s, NULL, &value);
409     if (err < 0) {
410         return err;
411     }
412     if (value > INT64_MAX) {
413         return -ERANGE;
414     }
415     return value;
416 }
417 
418 static int img_create(int argc, char **argv)
419 {
420     int c;
421     uint64_t img_size = -1;
422     const char *fmt = "raw";
423     const char *base_fmt = NULL;
424     const char *filename;
425     const char *base_filename = NULL;
426     char *options = NULL;
427     Error *local_err = NULL;
428     bool quiet = false;
429     int flags = 0;
430 
431     for(;;) {
432         static const struct option long_options[] = {
433             {"help", no_argument, 0, 'h'},
434             {"object", required_argument, 0, OPTION_OBJECT},
435             {0, 0, 0, 0}
436         };
437         c = getopt_long(argc, argv, ":F:b:f:ho:qu",
438                         long_options, NULL);
439         if (c == -1) {
440             break;
441         }
442         switch(c) {
443         case ':':
444             missing_argument(argv[optind - 1]);
445             break;
446         case '?':
447             unrecognized_option(argv[optind - 1]);
448             break;
449         case 'h':
450             help();
451             break;
452         case 'F':
453             base_fmt = optarg;
454             break;
455         case 'b':
456             base_filename = optarg;
457             break;
458         case 'f':
459             fmt = optarg;
460             break;
461         case 'o':
462             if (!is_valid_option_list(optarg)) {
463                 error_report("Invalid option list: %s", optarg);
464                 goto fail;
465             }
466             if (!options) {
467                 options = g_strdup(optarg);
468             } else {
469                 char *old_options = options;
470                 options = g_strdup_printf("%s,%s", options, optarg);
471                 g_free(old_options);
472             }
473             break;
474         case 'q':
475             quiet = true;
476             break;
477         case 'u':
478             flags |= BDRV_O_NO_BACKING;
479             break;
480         case OPTION_OBJECT: {
481             QemuOpts *opts;
482             opts = qemu_opts_parse_noisily(&qemu_object_opts,
483                                            optarg, true);
484             if (!opts) {
485                 goto fail;
486             }
487         }   break;
488         }
489     }
490 
491     /* Get the filename */
492     filename = (optind < argc) ? argv[optind] : NULL;
493     if (options && has_help_option(options)) {
494         g_free(options);
495         return print_block_option_help(filename, fmt);
496     }
497 
498     if (optind >= argc) {
499         error_exit("Expecting image file name");
500     }
501     optind++;
502 
503     if (qemu_opts_foreach(&qemu_object_opts,
504                           user_creatable_add_opts_foreach,
505                           NULL, NULL)) {
506         goto fail;
507     }
508 
509     /* Get image size, if specified */
510     if (optind < argc) {
511         int64_t sval;
512 
513         sval = cvtnum(argv[optind++]);
514         if (sval < 0) {
515             if (sval == -ERANGE) {
516                 error_report("Image size must be less than 8 EiB!");
517             } else {
518                 error_report("Invalid image size specified! You may use k, M, "
519                       "G, T, P or E suffixes for ");
520                 error_report("kilobytes, megabytes, gigabytes, terabytes, "
521                              "petabytes and exabytes.");
522             }
523             goto fail;
524         }
525         img_size = (uint64_t)sval;
526     }
527     if (optind != argc) {
528         error_exit("Unexpected argument: %s", argv[optind]);
529     }
530 
531     bdrv_img_create(filename, fmt, base_filename, base_fmt,
532                     options, img_size, flags, quiet, &local_err);
533     if (local_err) {
534         error_reportf_err(local_err, "%s: ", filename);
535         goto fail;
536     }
537 
538     g_free(options);
539     return 0;
540 
541 fail:
542     g_free(options);
543     return 1;
544 }
545 
546 static void dump_json_image_check(ImageCheck *check, bool quiet)
547 {
548     QString *str;
549     QObject *obj;
550     Visitor *v = qobject_output_visitor_new(&obj);
551 
552     visit_type_ImageCheck(v, NULL, &check, &error_abort);
553     visit_complete(v, &obj);
554     str = qobject_to_json_pretty(obj);
555     assert(str != NULL);
556     qprintf(quiet, "%s\n", qstring_get_str(str));
557     qobject_unref(obj);
558     visit_free(v);
559     qobject_unref(str);
560 }
561 
562 static void dump_human_image_check(ImageCheck *check, bool quiet)
563 {
564     if (!(check->corruptions || check->leaks || check->check_errors)) {
565         qprintf(quiet, "No errors were found on the image.\n");
566     } else {
567         if (check->corruptions) {
568             qprintf(quiet, "\n%" PRId64 " errors were found on the image.\n"
569                     "Data may be corrupted, or further writes to the image "
570                     "may corrupt it.\n",
571                     check->corruptions);
572         }
573 
574         if (check->leaks) {
575             qprintf(quiet,
576                     "\n%" PRId64 " leaked clusters were found on the image.\n"
577                     "This means waste of disk space, but no harm to data.\n",
578                     check->leaks);
579         }
580 
581         if (check->check_errors) {
582             qprintf(quiet,
583                     "\n%" PRId64
584                     " internal errors have occurred during the check.\n",
585                     check->check_errors);
586         }
587     }
588 
589     if (check->total_clusters != 0 && check->allocated_clusters != 0) {
590         qprintf(quiet, "%" PRId64 "/%" PRId64 " = %0.2f%% allocated, "
591                 "%0.2f%% fragmented, %0.2f%% compressed clusters\n",
592                 check->allocated_clusters, check->total_clusters,
593                 check->allocated_clusters * 100.0 / check->total_clusters,
594                 check->fragmented_clusters * 100.0 / check->allocated_clusters,
595                 check->compressed_clusters * 100.0 /
596                 check->allocated_clusters);
597     }
598 
599     if (check->image_end_offset) {
600         qprintf(quiet,
601                 "Image end offset: %" PRId64 "\n", check->image_end_offset);
602     }
603 }
604 
605 static int collect_image_check(BlockDriverState *bs,
606                    ImageCheck *check,
607                    const char *filename,
608                    const char *fmt,
609                    int fix)
610 {
611     int ret;
612     BdrvCheckResult result;
613 
614     ret = bdrv_check(bs, &result, fix);
615     if (ret < 0) {
616         return ret;
617     }
618 
619     check->filename                 = g_strdup(filename);
620     check->format                   = g_strdup(bdrv_get_format_name(bs));
621     check->check_errors             = result.check_errors;
622     check->corruptions              = result.corruptions;
623     check->has_corruptions          = result.corruptions != 0;
624     check->leaks                    = result.leaks;
625     check->has_leaks                = result.leaks != 0;
626     check->corruptions_fixed        = result.corruptions_fixed;
627     check->has_corruptions_fixed    = result.corruptions != 0;
628     check->leaks_fixed              = result.leaks_fixed;
629     check->has_leaks_fixed          = result.leaks != 0;
630     check->image_end_offset         = result.image_end_offset;
631     check->has_image_end_offset     = result.image_end_offset != 0;
632     check->total_clusters           = result.bfi.total_clusters;
633     check->has_total_clusters       = result.bfi.total_clusters != 0;
634     check->allocated_clusters       = result.bfi.allocated_clusters;
635     check->has_allocated_clusters   = result.bfi.allocated_clusters != 0;
636     check->fragmented_clusters      = result.bfi.fragmented_clusters;
637     check->has_fragmented_clusters  = result.bfi.fragmented_clusters != 0;
638     check->compressed_clusters      = result.bfi.compressed_clusters;
639     check->has_compressed_clusters  = result.bfi.compressed_clusters != 0;
640 
641     return 0;
642 }
643 
644 /*
645  * Checks an image for consistency. Exit codes:
646  *
647  *  0 - Check completed, image is good
648  *  1 - Check not completed because of internal errors
649  *  2 - Check completed, image is corrupted
650  *  3 - Check completed, image has leaked clusters, but is good otherwise
651  * 63 - Checks are not supported by the image format
652  */
653 static int img_check(int argc, char **argv)
654 {
655     int c, ret;
656     OutputFormat output_format = OFORMAT_HUMAN;
657     const char *filename, *fmt, *output, *cache;
658     BlockBackend *blk;
659     BlockDriverState *bs;
660     int fix = 0;
661     int flags = BDRV_O_CHECK;
662     bool writethrough;
663     ImageCheck *check;
664     bool quiet = false;
665     bool image_opts = false;
666     bool force_share = false;
667 
668     fmt = NULL;
669     output = NULL;
670     cache = BDRV_DEFAULT_CACHE;
671 
672     for(;;) {
673         int option_index = 0;
674         static const struct option long_options[] = {
675             {"help", no_argument, 0, 'h'},
676             {"format", required_argument, 0, 'f'},
677             {"repair", required_argument, 0, 'r'},
678             {"output", required_argument, 0, OPTION_OUTPUT},
679             {"object", required_argument, 0, OPTION_OBJECT},
680             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
681             {"force-share", no_argument, 0, 'U'},
682             {0, 0, 0, 0}
683         };
684         c = getopt_long(argc, argv, ":hf:r:T:qU",
685                         long_options, &option_index);
686         if (c == -1) {
687             break;
688         }
689         switch(c) {
690         case ':':
691             missing_argument(argv[optind - 1]);
692             break;
693         case '?':
694             unrecognized_option(argv[optind - 1]);
695             break;
696         case 'h':
697             help();
698             break;
699         case 'f':
700             fmt = optarg;
701             break;
702         case 'r':
703             flags |= BDRV_O_RDWR;
704 
705             if (!strcmp(optarg, "leaks")) {
706                 fix = BDRV_FIX_LEAKS;
707             } else if (!strcmp(optarg, "all")) {
708                 fix = BDRV_FIX_LEAKS | BDRV_FIX_ERRORS;
709             } else {
710                 error_exit("Unknown option value for -r "
711                            "(expecting 'leaks' or 'all'): %s", optarg);
712             }
713             break;
714         case OPTION_OUTPUT:
715             output = optarg;
716             break;
717         case 'T':
718             cache = optarg;
719             break;
720         case 'q':
721             quiet = true;
722             break;
723         case 'U':
724             force_share = true;
725             break;
726         case OPTION_OBJECT: {
727             QemuOpts *opts;
728             opts = qemu_opts_parse_noisily(&qemu_object_opts,
729                                            optarg, true);
730             if (!opts) {
731                 return 1;
732             }
733         }   break;
734         case OPTION_IMAGE_OPTS:
735             image_opts = true;
736             break;
737         }
738     }
739     if (optind != argc - 1) {
740         error_exit("Expecting one image file name");
741     }
742     filename = argv[optind++];
743 
744     if (output && !strcmp(output, "json")) {
745         output_format = OFORMAT_JSON;
746     } else if (output && !strcmp(output, "human")) {
747         output_format = OFORMAT_HUMAN;
748     } else if (output) {
749         error_report("--output must be used with human or json as argument.");
750         return 1;
751     }
752 
753     if (qemu_opts_foreach(&qemu_object_opts,
754                           user_creatable_add_opts_foreach,
755                           NULL, NULL)) {
756         return 1;
757     }
758 
759     ret = bdrv_parse_cache_mode(cache, &flags, &writethrough);
760     if (ret < 0) {
761         error_report("Invalid source cache option: %s", cache);
762         return 1;
763     }
764 
765     blk = img_open(image_opts, filename, fmt, flags, writethrough, quiet,
766                    force_share);
767     if (!blk) {
768         return 1;
769     }
770     bs = blk_bs(blk);
771 
772     check = g_new0(ImageCheck, 1);
773     ret = collect_image_check(bs, check, filename, fmt, fix);
774 
775     if (ret == -ENOTSUP) {
776         error_report("This image format does not support checks");
777         ret = 63;
778         goto fail;
779     }
780 
781     if (check->corruptions_fixed || check->leaks_fixed) {
782         int corruptions_fixed, leaks_fixed;
783 
784         leaks_fixed         = check->leaks_fixed;
785         corruptions_fixed   = check->corruptions_fixed;
786 
787         if (output_format == OFORMAT_HUMAN) {
788             qprintf(quiet,
789                     "The following inconsistencies were found and repaired:\n\n"
790                     "    %" PRId64 " leaked clusters\n"
791                     "    %" PRId64 " corruptions\n\n"
792                     "Double checking the fixed image now...\n",
793                     check->leaks_fixed,
794                     check->corruptions_fixed);
795         }
796 
797         ret = collect_image_check(bs, check, filename, fmt, 0);
798 
799         check->leaks_fixed          = leaks_fixed;
800         check->corruptions_fixed    = corruptions_fixed;
801     }
802 
803     if (!ret) {
804         switch (output_format) {
805         case OFORMAT_HUMAN:
806             dump_human_image_check(check, quiet);
807             break;
808         case OFORMAT_JSON:
809             dump_json_image_check(check, quiet);
810             break;
811         }
812     }
813 
814     if (ret || check->check_errors) {
815         if (ret) {
816             error_report("Check failed: %s", strerror(-ret));
817         } else {
818             error_report("Check failed");
819         }
820         ret = 1;
821         goto fail;
822     }
823 
824     if (check->corruptions) {
825         ret = 2;
826     } else if (check->leaks) {
827         ret = 3;
828     } else {
829         ret = 0;
830     }
831 
832 fail:
833     qapi_free_ImageCheck(check);
834     blk_unref(blk);
835     return ret;
836 }
837 
838 typedef struct CommonBlockJobCBInfo {
839     BlockDriverState *bs;
840     Error **errp;
841 } CommonBlockJobCBInfo;
842 
843 static void common_block_job_cb(void *opaque, int ret)
844 {
845     CommonBlockJobCBInfo *cbi = opaque;
846 
847     if (ret < 0) {
848         error_setg_errno(cbi->errp, -ret, "Block job failed");
849     }
850 }
851 
852 static void run_block_job(BlockJob *job, Error **errp)
853 {
854     AioContext *aio_context = blk_get_aio_context(job->blk);
855     int ret = 0;
856 
857     aio_context_acquire(aio_context);
858     job_ref(&job->job);
859     do {
860         float progress = 0.0f;
861         aio_poll(aio_context, true);
862         if (job->job.progress_total) {
863             progress = (float)job->job.progress_current /
864                        job->job.progress_total * 100.f;
865         }
866         qemu_progress_print(progress, 0);
867     } while (!job_is_ready(&job->job) && !job_is_completed(&job->job));
868 
869     if (!job_is_completed(&job->job)) {
870         ret = job_complete_sync(&job->job, errp);
871     } else {
872         ret = job->job.ret;
873     }
874     job_unref(&job->job);
875     aio_context_release(aio_context);
876 
877     /* publish completion progress only when success */
878     if (!ret) {
879         qemu_progress_print(100.f, 0);
880     }
881 }
882 
883 static int img_commit(int argc, char **argv)
884 {
885     int c, ret, flags;
886     const char *filename, *fmt, *cache, *base;
887     BlockBackend *blk;
888     BlockDriverState *bs, *base_bs;
889     BlockJob *job;
890     bool progress = false, quiet = false, drop = false;
891     bool writethrough;
892     Error *local_err = NULL;
893     CommonBlockJobCBInfo cbi;
894     bool image_opts = false;
895     AioContext *aio_context;
896 
897     fmt = NULL;
898     cache = BDRV_DEFAULT_CACHE;
899     base = NULL;
900     for(;;) {
901         static const struct option long_options[] = {
902             {"help", no_argument, 0, 'h'},
903             {"object", required_argument, 0, OPTION_OBJECT},
904             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
905             {0, 0, 0, 0}
906         };
907         c = getopt_long(argc, argv, ":f:ht:b:dpq",
908                         long_options, NULL);
909         if (c == -1) {
910             break;
911         }
912         switch(c) {
913         case ':':
914             missing_argument(argv[optind - 1]);
915             break;
916         case '?':
917             unrecognized_option(argv[optind - 1]);
918             break;
919         case 'h':
920             help();
921             break;
922         case 'f':
923             fmt = optarg;
924             break;
925         case 't':
926             cache = optarg;
927             break;
928         case 'b':
929             base = optarg;
930             /* -b implies -d */
931             drop = true;
932             break;
933         case 'd':
934             drop = true;
935             break;
936         case 'p':
937             progress = true;
938             break;
939         case 'q':
940             quiet = true;
941             break;
942         case OPTION_OBJECT: {
943             QemuOpts *opts;
944             opts = qemu_opts_parse_noisily(&qemu_object_opts,
945                                            optarg, true);
946             if (!opts) {
947                 return 1;
948             }
949         }   break;
950         case OPTION_IMAGE_OPTS:
951             image_opts = true;
952             break;
953         }
954     }
955 
956     /* Progress is not shown in Quiet mode */
957     if (quiet) {
958         progress = false;
959     }
960 
961     if (optind != argc - 1) {
962         error_exit("Expecting one image file name");
963     }
964     filename = argv[optind++];
965 
966     if (qemu_opts_foreach(&qemu_object_opts,
967                           user_creatable_add_opts_foreach,
968                           NULL, NULL)) {
969         return 1;
970     }
971 
972     flags = BDRV_O_RDWR | BDRV_O_UNMAP;
973     ret = bdrv_parse_cache_mode(cache, &flags, &writethrough);
974     if (ret < 0) {
975         error_report("Invalid cache option: %s", cache);
976         return 1;
977     }
978 
979     blk = img_open(image_opts, filename, fmt, flags, writethrough, quiet,
980                    false);
981     if (!blk) {
982         return 1;
983     }
984     bs = blk_bs(blk);
985 
986     qemu_progress_init(progress, 1.f);
987     qemu_progress_print(0.f, 100);
988 
989     if (base) {
990         base_bs = bdrv_find_backing_image(bs, base);
991         if (!base_bs) {
992             error_setg(&local_err,
993                        "Did not find '%s' in the backing chain of '%s'",
994                        base, filename);
995             goto done;
996         }
997     } else {
998         /* This is different from QMP, which by default uses the deepest file in
999          * the backing chain (i.e., the very base); however, the traditional
1000          * behavior of qemu-img commit is using the immediate backing file. */
1001         base_bs = backing_bs(bs);
1002         if (!base_bs) {
1003             error_setg(&local_err, "Image does not have a backing file");
1004             goto done;
1005         }
1006     }
1007 
1008     cbi = (CommonBlockJobCBInfo){
1009         .errp = &local_err,
1010         .bs   = bs,
1011     };
1012 
1013     aio_context = bdrv_get_aio_context(bs);
1014     aio_context_acquire(aio_context);
1015     commit_active_start("commit", bs, base_bs, JOB_DEFAULT, 0,
1016                         BLOCKDEV_ON_ERROR_REPORT, NULL, common_block_job_cb,
1017                         &cbi, false, &local_err);
1018     aio_context_release(aio_context);
1019     if (local_err) {
1020         goto done;
1021     }
1022 
1023     /* When the block job completes, the BlockBackend reference will point to
1024      * the old backing file. In order to avoid that the top image is already
1025      * deleted, so we can still empty it afterwards, increment the reference
1026      * counter here preemptively. */
1027     if (!drop) {
1028         bdrv_ref(bs);
1029     }
1030 
1031     job = block_job_get("commit");
1032     run_block_job(job, &local_err);
1033     if (local_err) {
1034         goto unref_backing;
1035     }
1036 
1037     if (!drop && bs->drv->bdrv_make_empty) {
1038         ret = bs->drv->bdrv_make_empty(bs);
1039         if (ret) {
1040             error_setg_errno(&local_err, -ret, "Could not empty %s",
1041                              filename);
1042             goto unref_backing;
1043         }
1044     }
1045 
1046 unref_backing:
1047     if (!drop) {
1048         bdrv_unref(bs);
1049     }
1050 
1051 done:
1052     qemu_progress_end();
1053 
1054     blk_unref(blk);
1055 
1056     if (local_err) {
1057         error_report_err(local_err);
1058         return 1;
1059     }
1060 
1061     qprintf(quiet, "Image committed.\n");
1062     return 0;
1063 }
1064 
1065 /*
1066  * Returns -1 if 'buf' contains only zeroes, otherwise the byte index
1067  * of the first sector boundary within buf where the sector contains a
1068  * non-zero byte.  This function is robust to a buffer that is not
1069  * sector-aligned.
1070  */
1071 static int64_t find_nonzero(const uint8_t *buf, int64_t n)
1072 {
1073     int64_t i;
1074     int64_t end = QEMU_ALIGN_DOWN(n, BDRV_SECTOR_SIZE);
1075 
1076     for (i = 0; i < end; i += BDRV_SECTOR_SIZE) {
1077         if (!buffer_is_zero(buf + i, BDRV_SECTOR_SIZE)) {
1078             return i;
1079         }
1080     }
1081     if (i < n && !buffer_is_zero(buf + i, n - end)) {
1082         return i;
1083     }
1084     return -1;
1085 }
1086 
1087 /*
1088  * Returns true iff the first sector pointed to by 'buf' contains at least
1089  * a non-NUL byte.
1090  *
1091  * 'pnum' is set to the number of sectors (including and immediately following
1092  * the first one) that are known to be in the same allocated/unallocated state.
1093  * The function will try to align the end offset to alignment boundaries so
1094  * that the request will at least end aligned and consequtive requests will
1095  * also start at an aligned offset.
1096  */
1097 static int is_allocated_sectors(const uint8_t *buf, int n, int *pnum,
1098                                 int64_t sector_num, int alignment)
1099 {
1100     bool is_zero;
1101     int i, tail;
1102 
1103     if (n <= 0) {
1104         *pnum = 0;
1105         return 0;
1106     }
1107     is_zero = buffer_is_zero(buf, 512);
1108     for(i = 1; i < n; i++) {
1109         buf += 512;
1110         if (is_zero != buffer_is_zero(buf, 512)) {
1111             break;
1112         }
1113     }
1114 
1115     tail = (sector_num + i) & (alignment - 1);
1116     if (tail) {
1117         if (is_zero && i <= tail) {
1118             /* treat unallocated areas which only consist
1119              * of a small tail as allocated. */
1120             is_zero = false;
1121         }
1122         if (!is_zero) {
1123             /* align up end offset of allocated areas. */
1124             i += alignment - tail;
1125             i = MIN(i, n);
1126         } else {
1127             /* align down end offset of zero areas. */
1128             i -= tail;
1129         }
1130     }
1131     *pnum = i;
1132     return !is_zero;
1133 }
1134 
1135 /*
1136  * Like is_allocated_sectors, but if the buffer starts with a used sector,
1137  * up to 'min' consecutive sectors containing zeros are ignored. This avoids
1138  * breaking up write requests for only small sparse areas.
1139  */
1140 static int is_allocated_sectors_min(const uint8_t *buf, int n, int *pnum,
1141     int min, int64_t sector_num, int alignment)
1142 {
1143     int ret;
1144     int num_checked, num_used;
1145 
1146     if (n < min) {
1147         min = n;
1148     }
1149 
1150     ret = is_allocated_sectors(buf, n, pnum, sector_num, alignment);
1151     if (!ret) {
1152         return ret;
1153     }
1154 
1155     num_used = *pnum;
1156     buf += BDRV_SECTOR_SIZE * *pnum;
1157     n -= *pnum;
1158     sector_num += *pnum;
1159     num_checked = num_used;
1160 
1161     while (n > 0) {
1162         ret = is_allocated_sectors(buf, n, pnum, sector_num, alignment);
1163 
1164         buf += BDRV_SECTOR_SIZE * *pnum;
1165         n -= *pnum;
1166         sector_num += *pnum;
1167         num_checked += *pnum;
1168         if (ret) {
1169             num_used = num_checked;
1170         } else if (*pnum >= min) {
1171             break;
1172         }
1173     }
1174 
1175     *pnum = num_used;
1176     return 1;
1177 }
1178 
1179 /*
1180  * Compares two buffers sector by sector. Returns 0 if the first
1181  * sector of each buffer matches, non-zero otherwise.
1182  *
1183  * pnum is set to the sector-aligned size of the buffer prefix that
1184  * has the same matching status as the first sector.
1185  */
1186 static int compare_buffers(const uint8_t *buf1, const uint8_t *buf2,
1187                            int64_t bytes, int64_t *pnum)
1188 {
1189     bool res;
1190     int64_t i = MIN(bytes, BDRV_SECTOR_SIZE);
1191 
1192     assert(bytes > 0);
1193 
1194     res = !!memcmp(buf1, buf2, i);
1195     while (i < bytes) {
1196         int64_t len = MIN(bytes - i, BDRV_SECTOR_SIZE);
1197 
1198         if (!!memcmp(buf1 + i, buf2 + i, len) != res) {
1199             break;
1200         }
1201         i += len;
1202     }
1203 
1204     *pnum = i;
1205     return res;
1206 }
1207 
1208 #define IO_BUF_SIZE (2 * 1024 * 1024)
1209 
1210 /*
1211  * Check if passed sectors are empty (not allocated or contain only 0 bytes)
1212  *
1213  * Intended for use by 'qemu-img compare': Returns 0 in case sectors are
1214  * filled with 0, 1 if sectors contain non-zero data (this is a comparison
1215  * failure), and 4 on error (the exit status for read errors), after emitting
1216  * an error message.
1217  *
1218  * @param blk:  BlockBackend for the image
1219  * @param offset: Starting offset to check
1220  * @param bytes: Number of bytes to check
1221  * @param filename: Name of disk file we are checking (logging purpose)
1222  * @param buffer: Allocated buffer for storing read data
1223  * @param quiet: Flag for quiet mode
1224  */
1225 static int check_empty_sectors(BlockBackend *blk, int64_t offset,
1226                                int64_t bytes, const char *filename,
1227                                uint8_t *buffer, bool quiet)
1228 {
1229     int ret = 0;
1230     int64_t idx;
1231 
1232     ret = blk_pread(blk, offset, buffer, bytes);
1233     if (ret < 0) {
1234         error_report("Error while reading offset %" PRId64 " of %s: %s",
1235                      offset, filename, strerror(-ret));
1236         return 4;
1237     }
1238     idx = find_nonzero(buffer, bytes);
1239     if (idx >= 0) {
1240         qprintf(quiet, "Content mismatch at offset %" PRId64 "!\n",
1241                 offset + idx);
1242         return 1;
1243     }
1244 
1245     return 0;
1246 }
1247 
1248 /*
1249  * Compares two images. Exit codes:
1250  *
1251  * 0 - Images are identical
1252  * 1 - Images differ
1253  * >1 - Error occurred
1254  */
1255 static int img_compare(int argc, char **argv)
1256 {
1257     const char *fmt1 = NULL, *fmt2 = NULL, *cache, *filename1, *filename2;
1258     BlockBackend *blk1, *blk2;
1259     BlockDriverState *bs1, *bs2;
1260     int64_t total_size1, total_size2;
1261     uint8_t *buf1 = NULL, *buf2 = NULL;
1262     int64_t pnum1, pnum2;
1263     int allocated1, allocated2;
1264     int ret = 0; /* return value - 0 Ident, 1 Different, >1 Error */
1265     bool progress = false, quiet = false, strict = false;
1266     int flags;
1267     bool writethrough;
1268     int64_t total_size;
1269     int64_t offset = 0;
1270     int64_t chunk;
1271     int c;
1272     uint64_t progress_base;
1273     bool image_opts = false;
1274     bool force_share = false;
1275 
1276     cache = BDRV_DEFAULT_CACHE;
1277     for (;;) {
1278         static const struct option long_options[] = {
1279             {"help", no_argument, 0, 'h'},
1280             {"object", required_argument, 0, OPTION_OBJECT},
1281             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
1282             {"force-share", no_argument, 0, 'U'},
1283             {0, 0, 0, 0}
1284         };
1285         c = getopt_long(argc, argv, ":hf:F:T:pqsU",
1286                         long_options, NULL);
1287         if (c == -1) {
1288             break;
1289         }
1290         switch (c) {
1291         case ':':
1292             missing_argument(argv[optind - 1]);
1293             break;
1294         case '?':
1295             unrecognized_option(argv[optind - 1]);
1296             break;
1297         case 'h':
1298             help();
1299             break;
1300         case 'f':
1301             fmt1 = optarg;
1302             break;
1303         case 'F':
1304             fmt2 = optarg;
1305             break;
1306         case 'T':
1307             cache = optarg;
1308             break;
1309         case 'p':
1310             progress = true;
1311             break;
1312         case 'q':
1313             quiet = true;
1314             break;
1315         case 's':
1316             strict = true;
1317             break;
1318         case 'U':
1319             force_share = true;
1320             break;
1321         case OPTION_OBJECT: {
1322             QemuOpts *opts;
1323             opts = qemu_opts_parse_noisily(&qemu_object_opts,
1324                                            optarg, true);
1325             if (!opts) {
1326                 ret = 2;
1327                 goto out4;
1328             }
1329         }   break;
1330         case OPTION_IMAGE_OPTS:
1331             image_opts = true;
1332             break;
1333         }
1334     }
1335 
1336     /* Progress is not shown in Quiet mode */
1337     if (quiet) {
1338         progress = false;
1339     }
1340 
1341 
1342     if (optind != argc - 2) {
1343         error_exit("Expecting two image file names");
1344     }
1345     filename1 = argv[optind++];
1346     filename2 = argv[optind++];
1347 
1348     if (qemu_opts_foreach(&qemu_object_opts,
1349                           user_creatable_add_opts_foreach,
1350                           NULL, NULL)) {
1351         ret = 2;
1352         goto out4;
1353     }
1354 
1355     /* Initialize before goto out */
1356     qemu_progress_init(progress, 2.0);
1357 
1358     flags = 0;
1359     ret = bdrv_parse_cache_mode(cache, &flags, &writethrough);
1360     if (ret < 0) {
1361         error_report("Invalid source cache option: %s", cache);
1362         ret = 2;
1363         goto out3;
1364     }
1365 
1366     blk1 = img_open(image_opts, filename1, fmt1, flags, writethrough, quiet,
1367                     force_share);
1368     if (!blk1) {
1369         ret = 2;
1370         goto out3;
1371     }
1372 
1373     blk2 = img_open(image_opts, filename2, fmt2, flags, writethrough, quiet,
1374                     force_share);
1375     if (!blk2) {
1376         ret = 2;
1377         goto out2;
1378     }
1379     bs1 = blk_bs(blk1);
1380     bs2 = blk_bs(blk2);
1381 
1382     buf1 = blk_blockalign(blk1, IO_BUF_SIZE);
1383     buf2 = blk_blockalign(blk2, IO_BUF_SIZE);
1384     total_size1 = blk_getlength(blk1);
1385     if (total_size1 < 0) {
1386         error_report("Can't get size of %s: %s",
1387                      filename1, strerror(-total_size1));
1388         ret = 4;
1389         goto out;
1390     }
1391     total_size2 = blk_getlength(blk2);
1392     if (total_size2 < 0) {
1393         error_report("Can't get size of %s: %s",
1394                      filename2, strerror(-total_size2));
1395         ret = 4;
1396         goto out;
1397     }
1398     total_size = MIN(total_size1, total_size2);
1399     progress_base = MAX(total_size1, total_size2);
1400 
1401     qemu_progress_print(0, 100);
1402 
1403     if (strict && total_size1 != total_size2) {
1404         ret = 1;
1405         qprintf(quiet, "Strict mode: Image size mismatch!\n");
1406         goto out;
1407     }
1408 
1409     while (offset < total_size) {
1410         int status1, status2;
1411 
1412         status1 = bdrv_block_status_above(bs1, NULL, offset,
1413                                           total_size1 - offset, &pnum1, NULL,
1414                                           NULL);
1415         if (status1 < 0) {
1416             ret = 3;
1417             error_report("Sector allocation test failed for %s", filename1);
1418             goto out;
1419         }
1420         allocated1 = status1 & BDRV_BLOCK_ALLOCATED;
1421 
1422         status2 = bdrv_block_status_above(bs2, NULL, offset,
1423                                           total_size2 - offset, &pnum2, NULL,
1424                                           NULL);
1425         if (status2 < 0) {
1426             ret = 3;
1427             error_report("Sector allocation test failed for %s", filename2);
1428             goto out;
1429         }
1430         allocated2 = status2 & BDRV_BLOCK_ALLOCATED;
1431 
1432         assert(pnum1 && pnum2);
1433         chunk = MIN(pnum1, pnum2);
1434 
1435         if (strict) {
1436             if (status1 != status2) {
1437                 ret = 1;
1438                 qprintf(quiet, "Strict mode: Offset %" PRId64
1439                         " block status mismatch!\n", offset);
1440                 goto out;
1441             }
1442         }
1443         if ((status1 & BDRV_BLOCK_ZERO) && (status2 & BDRV_BLOCK_ZERO)) {
1444             /* nothing to do */
1445         } else if (allocated1 == allocated2) {
1446             if (allocated1) {
1447                 int64_t pnum;
1448 
1449                 chunk = MIN(chunk, IO_BUF_SIZE);
1450                 ret = blk_pread(blk1, offset, buf1, chunk);
1451                 if (ret < 0) {
1452                     error_report("Error while reading offset %" PRId64
1453                                  " of %s: %s",
1454                                  offset, filename1, strerror(-ret));
1455                     ret = 4;
1456                     goto out;
1457                 }
1458                 ret = blk_pread(blk2, offset, buf2, chunk);
1459                 if (ret < 0) {
1460                     error_report("Error while reading offset %" PRId64
1461                                  " of %s: %s",
1462                                  offset, filename2, strerror(-ret));
1463                     ret = 4;
1464                     goto out;
1465                 }
1466                 ret = compare_buffers(buf1, buf2, chunk, &pnum);
1467                 if (ret || pnum != chunk) {
1468                     qprintf(quiet, "Content mismatch at offset %" PRId64 "!\n",
1469                             offset + (ret ? 0 : pnum));
1470                     ret = 1;
1471                     goto out;
1472                 }
1473             }
1474         } else {
1475             chunk = MIN(chunk, IO_BUF_SIZE);
1476             if (allocated1) {
1477                 ret = check_empty_sectors(blk1, offset, chunk,
1478                                           filename1, buf1, quiet);
1479             } else {
1480                 ret = check_empty_sectors(blk2, offset, chunk,
1481                                           filename2, buf1, quiet);
1482             }
1483             if (ret) {
1484                 goto out;
1485             }
1486         }
1487         offset += chunk;
1488         qemu_progress_print(((float) chunk / progress_base) * 100, 100);
1489     }
1490 
1491     if (total_size1 != total_size2) {
1492         BlockBackend *blk_over;
1493         const char *filename_over;
1494 
1495         qprintf(quiet, "Warning: Image size mismatch!\n");
1496         if (total_size1 > total_size2) {
1497             blk_over = blk1;
1498             filename_over = filename1;
1499         } else {
1500             blk_over = blk2;
1501             filename_over = filename2;
1502         }
1503 
1504         while (offset < progress_base) {
1505             ret = bdrv_block_status_above(blk_bs(blk_over), NULL, offset,
1506                                           progress_base - offset, &chunk,
1507                                           NULL, NULL);
1508             if (ret < 0) {
1509                 ret = 3;
1510                 error_report("Sector allocation test failed for %s",
1511                              filename_over);
1512                 goto out;
1513 
1514             }
1515             if (ret & BDRV_BLOCK_ALLOCATED && !(ret & BDRV_BLOCK_ZERO)) {
1516                 chunk = MIN(chunk, IO_BUF_SIZE);
1517                 ret = check_empty_sectors(blk_over, offset, chunk,
1518                                           filename_over, buf1, quiet);
1519                 if (ret) {
1520                     goto out;
1521                 }
1522             }
1523             offset += chunk;
1524             qemu_progress_print(((float) chunk / progress_base) * 100, 100);
1525         }
1526     }
1527 
1528     qprintf(quiet, "Images are identical.\n");
1529     ret = 0;
1530 
1531 out:
1532     qemu_vfree(buf1);
1533     qemu_vfree(buf2);
1534     blk_unref(blk2);
1535 out2:
1536     blk_unref(blk1);
1537 out3:
1538     qemu_progress_end();
1539 out4:
1540     return ret;
1541 }
1542 
1543 enum ImgConvertBlockStatus {
1544     BLK_DATA,
1545     BLK_ZERO,
1546     BLK_BACKING_FILE,
1547 };
1548 
1549 #define MAX_COROUTINES 16
1550 
1551 typedef struct ImgConvertState {
1552     BlockBackend **src;
1553     int64_t *src_sectors;
1554     int src_num;
1555     int64_t total_sectors;
1556     int64_t allocated_sectors;
1557     int64_t allocated_done;
1558     int64_t sector_num;
1559     int64_t wr_offs;
1560     enum ImgConvertBlockStatus status;
1561     int64_t sector_next_status;
1562     BlockBackend *target;
1563     bool has_zero_init;
1564     bool compressed;
1565     bool unallocated_blocks_are_zero;
1566     bool target_has_backing;
1567     int64_t target_backing_sectors; /* negative if unknown */
1568     bool wr_in_order;
1569     bool copy_range;
1570     int min_sparse;
1571     int alignment;
1572     size_t cluster_sectors;
1573     size_t buf_sectors;
1574     long num_coroutines;
1575     int running_coroutines;
1576     Coroutine *co[MAX_COROUTINES];
1577     int64_t wait_sector_num[MAX_COROUTINES];
1578     CoMutex lock;
1579     int ret;
1580 } ImgConvertState;
1581 
1582 static void convert_select_part(ImgConvertState *s, int64_t sector_num,
1583                                 int *src_cur, int64_t *src_cur_offset)
1584 {
1585     *src_cur = 0;
1586     *src_cur_offset = 0;
1587     while (sector_num - *src_cur_offset >= s->src_sectors[*src_cur]) {
1588         *src_cur_offset += s->src_sectors[*src_cur];
1589         (*src_cur)++;
1590         assert(*src_cur < s->src_num);
1591     }
1592 }
1593 
1594 static int convert_iteration_sectors(ImgConvertState *s, int64_t sector_num)
1595 {
1596     int64_t src_cur_offset;
1597     int ret, n, src_cur;
1598     bool post_backing_zero = false;
1599 
1600     convert_select_part(s, sector_num, &src_cur, &src_cur_offset);
1601 
1602     assert(s->total_sectors > sector_num);
1603     n = MIN(s->total_sectors - sector_num, BDRV_REQUEST_MAX_SECTORS);
1604 
1605     if (s->target_backing_sectors >= 0) {
1606         if (sector_num >= s->target_backing_sectors) {
1607             post_backing_zero = s->unallocated_blocks_are_zero;
1608         } else if (sector_num + n > s->target_backing_sectors) {
1609             /* Split requests around target_backing_sectors (because
1610              * starting from there, zeros are handled differently) */
1611             n = s->target_backing_sectors - sector_num;
1612         }
1613     }
1614 
1615     if (s->sector_next_status <= sector_num) {
1616         int64_t count = n * BDRV_SECTOR_SIZE;
1617 
1618         if (s->target_has_backing) {
1619 
1620             ret = bdrv_block_status(blk_bs(s->src[src_cur]),
1621                                     (sector_num - src_cur_offset) *
1622                                     BDRV_SECTOR_SIZE,
1623                                     count, &count, NULL, NULL);
1624         } else {
1625             ret = bdrv_block_status_above(blk_bs(s->src[src_cur]), NULL,
1626                                           (sector_num - src_cur_offset) *
1627                                           BDRV_SECTOR_SIZE,
1628                                           count, &count, NULL, NULL);
1629         }
1630         if (ret < 0) {
1631             return ret;
1632         }
1633         n = DIV_ROUND_UP(count, BDRV_SECTOR_SIZE);
1634 
1635         if (ret & BDRV_BLOCK_ZERO) {
1636             s->status = post_backing_zero ? BLK_BACKING_FILE : BLK_ZERO;
1637         } else if (ret & BDRV_BLOCK_DATA) {
1638             s->status = BLK_DATA;
1639         } else {
1640             s->status = s->target_has_backing ? BLK_BACKING_FILE : BLK_DATA;
1641         }
1642 
1643         s->sector_next_status = sector_num + n;
1644     }
1645 
1646     n = MIN(n, s->sector_next_status - sector_num);
1647     if (s->status == BLK_DATA) {
1648         n = MIN(n, s->buf_sectors);
1649     }
1650 
1651     /* We need to write complete clusters for compressed images, so if an
1652      * unallocated area is shorter than that, we must consider the whole
1653      * cluster allocated. */
1654     if (s->compressed) {
1655         if (n < s->cluster_sectors) {
1656             n = MIN(s->cluster_sectors, s->total_sectors - sector_num);
1657             s->status = BLK_DATA;
1658         } else {
1659             n = QEMU_ALIGN_DOWN(n, s->cluster_sectors);
1660         }
1661     }
1662 
1663     return n;
1664 }
1665 
1666 static int coroutine_fn convert_co_read(ImgConvertState *s, int64_t sector_num,
1667                                         int nb_sectors, uint8_t *buf)
1668 {
1669     int n, ret;
1670     QEMUIOVector qiov;
1671     struct iovec iov;
1672 
1673     assert(nb_sectors <= s->buf_sectors);
1674     while (nb_sectors > 0) {
1675         BlockBackend *blk;
1676         int src_cur;
1677         int64_t bs_sectors, src_cur_offset;
1678 
1679         /* In the case of compression with multiple source files, we can get a
1680          * nb_sectors that spreads into the next part. So we must be able to
1681          * read across multiple BDSes for one convert_read() call. */
1682         convert_select_part(s, sector_num, &src_cur, &src_cur_offset);
1683         blk = s->src[src_cur];
1684         bs_sectors = s->src_sectors[src_cur];
1685 
1686         n = MIN(nb_sectors, bs_sectors - (sector_num - src_cur_offset));
1687         iov.iov_base = buf;
1688         iov.iov_len = n << BDRV_SECTOR_BITS;
1689         qemu_iovec_init_external(&qiov, &iov, 1);
1690 
1691         ret = blk_co_preadv(
1692                 blk, (sector_num - src_cur_offset) << BDRV_SECTOR_BITS,
1693                 n << BDRV_SECTOR_BITS, &qiov, 0);
1694         if (ret < 0) {
1695             return ret;
1696         }
1697 
1698         sector_num += n;
1699         nb_sectors -= n;
1700         buf += n * BDRV_SECTOR_SIZE;
1701     }
1702 
1703     return 0;
1704 }
1705 
1706 
1707 static int coroutine_fn convert_co_write(ImgConvertState *s, int64_t sector_num,
1708                                          int nb_sectors, uint8_t *buf,
1709                                          enum ImgConvertBlockStatus status)
1710 {
1711     int ret;
1712     QEMUIOVector qiov;
1713     struct iovec iov;
1714 
1715     while (nb_sectors > 0) {
1716         int n = nb_sectors;
1717         BdrvRequestFlags flags = s->compressed ? BDRV_REQ_WRITE_COMPRESSED : 0;
1718 
1719         switch (status) {
1720         case BLK_BACKING_FILE:
1721             /* If we have a backing file, leave clusters unallocated that are
1722              * unallocated in the source image, so that the backing file is
1723              * visible at the respective offset. */
1724             assert(s->target_has_backing);
1725             break;
1726 
1727         case BLK_DATA:
1728             /* If we're told to keep the target fully allocated (-S 0) or there
1729              * is real non-zero data, we must write it. Otherwise we can treat
1730              * it as zero sectors.
1731              * Compressed clusters need to be written as a whole, so in that
1732              * case we can only save the write if the buffer is completely
1733              * zeroed. */
1734             if (!s->min_sparse ||
1735                 (!s->compressed &&
1736                  is_allocated_sectors_min(buf, n, &n, s->min_sparse,
1737                                           sector_num, s->alignment)) ||
1738                 (s->compressed &&
1739                  !buffer_is_zero(buf, n * BDRV_SECTOR_SIZE)))
1740             {
1741                 iov.iov_base = buf;
1742                 iov.iov_len = n << BDRV_SECTOR_BITS;
1743                 qemu_iovec_init_external(&qiov, &iov, 1);
1744 
1745                 ret = blk_co_pwritev(s->target, sector_num << BDRV_SECTOR_BITS,
1746                                      n << BDRV_SECTOR_BITS, &qiov, flags);
1747                 if (ret < 0) {
1748                     return ret;
1749                 }
1750                 break;
1751             }
1752             /* fall-through */
1753 
1754         case BLK_ZERO:
1755             if (s->has_zero_init) {
1756                 assert(!s->target_has_backing);
1757                 break;
1758             }
1759             ret = blk_co_pwrite_zeroes(s->target,
1760                                        sector_num << BDRV_SECTOR_BITS,
1761                                        n << BDRV_SECTOR_BITS, 0);
1762             if (ret < 0) {
1763                 return ret;
1764             }
1765             break;
1766         }
1767 
1768         sector_num += n;
1769         nb_sectors -= n;
1770         buf += n * BDRV_SECTOR_SIZE;
1771     }
1772 
1773     return 0;
1774 }
1775 
1776 static int coroutine_fn convert_co_copy_range(ImgConvertState *s, int64_t sector_num,
1777                                               int nb_sectors)
1778 {
1779     int n, ret;
1780 
1781     while (nb_sectors > 0) {
1782         BlockBackend *blk;
1783         int src_cur;
1784         int64_t bs_sectors, src_cur_offset;
1785         int64_t offset;
1786 
1787         convert_select_part(s, sector_num, &src_cur, &src_cur_offset);
1788         offset = (sector_num - src_cur_offset) << BDRV_SECTOR_BITS;
1789         blk = s->src[src_cur];
1790         bs_sectors = s->src_sectors[src_cur];
1791 
1792         n = MIN(nb_sectors, bs_sectors - (sector_num - src_cur_offset));
1793 
1794         ret = blk_co_copy_range(blk, offset, s->target,
1795                                 sector_num << BDRV_SECTOR_BITS,
1796                                 n << BDRV_SECTOR_BITS, 0, 0);
1797         if (ret < 0) {
1798             return ret;
1799         }
1800 
1801         sector_num += n;
1802         nb_sectors -= n;
1803     }
1804     return 0;
1805 }
1806 
1807 static void coroutine_fn convert_co_do_copy(void *opaque)
1808 {
1809     ImgConvertState *s = opaque;
1810     uint8_t *buf = NULL;
1811     int ret, i;
1812     int index = -1;
1813 
1814     for (i = 0; i < s->num_coroutines; i++) {
1815         if (s->co[i] == qemu_coroutine_self()) {
1816             index = i;
1817             break;
1818         }
1819     }
1820     assert(index >= 0);
1821 
1822     s->running_coroutines++;
1823     buf = blk_blockalign(s->target, s->buf_sectors * BDRV_SECTOR_SIZE);
1824 
1825     while (1) {
1826         int n;
1827         int64_t sector_num;
1828         enum ImgConvertBlockStatus status;
1829         bool copy_range;
1830 
1831         qemu_co_mutex_lock(&s->lock);
1832         if (s->ret != -EINPROGRESS || s->sector_num >= s->total_sectors) {
1833             qemu_co_mutex_unlock(&s->lock);
1834             break;
1835         }
1836         n = convert_iteration_sectors(s, s->sector_num);
1837         if (n < 0) {
1838             qemu_co_mutex_unlock(&s->lock);
1839             s->ret = n;
1840             break;
1841         }
1842         /* save current sector and allocation status to local variables */
1843         sector_num = s->sector_num;
1844         status = s->status;
1845         if (!s->min_sparse && s->status == BLK_ZERO) {
1846             n = MIN(n, s->buf_sectors);
1847         }
1848         /* increment global sector counter so that other coroutines can
1849          * already continue reading beyond this request */
1850         s->sector_num += n;
1851         qemu_co_mutex_unlock(&s->lock);
1852 
1853         if (status == BLK_DATA || (!s->min_sparse && status == BLK_ZERO)) {
1854             s->allocated_done += n;
1855             qemu_progress_print(100.0 * s->allocated_done /
1856                                         s->allocated_sectors, 0);
1857         }
1858 
1859 retry:
1860         copy_range = s->copy_range && s->status == BLK_DATA;
1861         if (status == BLK_DATA && !copy_range) {
1862             ret = convert_co_read(s, sector_num, n, buf);
1863             if (ret < 0) {
1864                 error_report("error while reading sector %" PRId64
1865                              ": %s", sector_num, strerror(-ret));
1866                 s->ret = ret;
1867             }
1868         } else if (!s->min_sparse && status == BLK_ZERO) {
1869             status = BLK_DATA;
1870             memset(buf, 0x00, n * BDRV_SECTOR_SIZE);
1871         }
1872 
1873         if (s->wr_in_order) {
1874             /* keep writes in order */
1875             while (s->wr_offs != sector_num && s->ret == -EINPROGRESS) {
1876                 s->wait_sector_num[index] = sector_num;
1877                 qemu_coroutine_yield();
1878             }
1879             s->wait_sector_num[index] = -1;
1880         }
1881 
1882         if (s->ret == -EINPROGRESS) {
1883             if (copy_range) {
1884                 ret = convert_co_copy_range(s, sector_num, n);
1885                 if (ret) {
1886                     s->copy_range = false;
1887                     goto retry;
1888                 }
1889             } else {
1890                 ret = convert_co_write(s, sector_num, n, buf, status);
1891             }
1892             if (ret < 0) {
1893                 error_report("error while writing sector %" PRId64
1894                              ": %s", sector_num, strerror(-ret));
1895                 s->ret = ret;
1896             }
1897         }
1898 
1899         if (s->wr_in_order) {
1900             /* reenter the coroutine that might have waited
1901              * for this write to complete */
1902             s->wr_offs = sector_num + n;
1903             for (i = 0; i < s->num_coroutines; i++) {
1904                 if (s->co[i] && s->wait_sector_num[i] == s->wr_offs) {
1905                     /*
1906                      * A -> B -> A cannot occur because A has
1907                      * s->wait_sector_num[i] == -1 during A -> B.  Therefore
1908                      * B will never enter A during this time window.
1909                      */
1910                     qemu_coroutine_enter(s->co[i]);
1911                     break;
1912                 }
1913             }
1914         }
1915     }
1916 
1917     qemu_vfree(buf);
1918     s->co[index] = NULL;
1919     s->running_coroutines--;
1920     if (!s->running_coroutines && s->ret == -EINPROGRESS) {
1921         /* the convert job finished successfully */
1922         s->ret = 0;
1923     }
1924 }
1925 
1926 static int convert_do_copy(ImgConvertState *s)
1927 {
1928     int ret, i, n;
1929     int64_t sector_num = 0;
1930 
1931     /* Check whether we have zero initialisation or can get it efficiently */
1932     s->has_zero_init = s->min_sparse && !s->target_has_backing
1933                      ? bdrv_has_zero_init(blk_bs(s->target))
1934                      : false;
1935 
1936     if (!s->has_zero_init && !s->target_has_backing &&
1937         bdrv_can_write_zeroes_with_unmap(blk_bs(s->target)))
1938     {
1939         ret = blk_make_zero(s->target, BDRV_REQ_MAY_UNMAP);
1940         if (ret == 0) {
1941             s->has_zero_init = true;
1942         }
1943     }
1944 
1945     /* Allocate buffer for copied data. For compressed images, only one cluster
1946      * can be copied at a time. */
1947     if (s->compressed) {
1948         if (s->cluster_sectors <= 0 || s->cluster_sectors > s->buf_sectors) {
1949             error_report("invalid cluster size");
1950             return -EINVAL;
1951         }
1952         s->buf_sectors = s->cluster_sectors;
1953     }
1954 
1955     while (sector_num < s->total_sectors) {
1956         n = convert_iteration_sectors(s, sector_num);
1957         if (n < 0) {
1958             return n;
1959         }
1960         if (s->status == BLK_DATA || (!s->min_sparse && s->status == BLK_ZERO))
1961         {
1962             s->allocated_sectors += n;
1963         }
1964         sector_num += n;
1965     }
1966 
1967     /* Do the copy */
1968     s->sector_next_status = 0;
1969     s->ret = -EINPROGRESS;
1970 
1971     qemu_co_mutex_init(&s->lock);
1972     for (i = 0; i < s->num_coroutines; i++) {
1973         s->co[i] = qemu_coroutine_create(convert_co_do_copy, s);
1974         s->wait_sector_num[i] = -1;
1975         qemu_coroutine_enter(s->co[i]);
1976     }
1977 
1978     while (s->running_coroutines) {
1979         main_loop_wait(false);
1980     }
1981 
1982     if (s->compressed && !s->ret) {
1983         /* signal EOF to align */
1984         ret = blk_pwrite_compressed(s->target, 0, NULL, 0);
1985         if (ret < 0) {
1986             return ret;
1987         }
1988     }
1989 
1990     return s->ret;
1991 }
1992 
1993 #define MAX_BUF_SECTORS 32768
1994 
1995 static int img_convert(int argc, char **argv)
1996 {
1997     int c, bs_i, flags, src_flags = 0;
1998     const char *fmt = NULL, *out_fmt = NULL, *cache = "unsafe",
1999                *src_cache = BDRV_DEFAULT_CACHE, *out_baseimg = NULL,
2000                *out_filename, *out_baseimg_param, *snapshot_name = NULL;
2001     BlockDriver *drv = NULL, *proto_drv = NULL;
2002     BlockDriverInfo bdi;
2003     BlockDriverState *out_bs;
2004     QemuOpts *opts = NULL, *sn_opts = NULL;
2005     QemuOptsList *create_opts = NULL;
2006     QDict *open_opts = NULL;
2007     char *options = NULL;
2008     Error *local_err = NULL;
2009     bool writethrough, src_writethrough, quiet = false, image_opts = false,
2010          skip_create = false, progress = false, tgt_image_opts = false;
2011     int64_t ret = -EINVAL;
2012     bool force_share = false;
2013     bool explict_min_sparse = false;
2014 
2015     ImgConvertState s = (ImgConvertState) {
2016         /* Need at least 4k of zeros for sparse detection */
2017         .min_sparse         = 8,
2018         .copy_range         = false,
2019         .buf_sectors        = IO_BUF_SIZE / BDRV_SECTOR_SIZE,
2020         .wr_in_order        = true,
2021         .num_coroutines     = 8,
2022     };
2023 
2024     for(;;) {
2025         static const struct option long_options[] = {
2026             {"help", no_argument, 0, 'h'},
2027             {"object", required_argument, 0, OPTION_OBJECT},
2028             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
2029             {"force-share", no_argument, 0, 'U'},
2030             {"target-image-opts", no_argument, 0, OPTION_TARGET_IMAGE_OPTS},
2031             {0, 0, 0, 0}
2032         };
2033         c = getopt_long(argc, argv, ":hf:O:B:Cco:l:S:pt:T:qnm:WU",
2034                         long_options, NULL);
2035         if (c == -1) {
2036             break;
2037         }
2038         switch(c) {
2039         case ':':
2040             missing_argument(argv[optind - 1]);
2041             break;
2042         case '?':
2043             unrecognized_option(argv[optind - 1]);
2044             break;
2045         case 'h':
2046             help();
2047             break;
2048         case 'f':
2049             fmt = optarg;
2050             break;
2051         case 'O':
2052             out_fmt = optarg;
2053             break;
2054         case 'B':
2055             out_baseimg = optarg;
2056             break;
2057         case 'C':
2058             s.copy_range = true;
2059             break;
2060         case 'c':
2061             s.compressed = true;
2062             break;
2063         case 'o':
2064             if (!is_valid_option_list(optarg)) {
2065                 error_report("Invalid option list: %s", optarg);
2066                 goto fail_getopt;
2067             }
2068             if (!options) {
2069                 options = g_strdup(optarg);
2070             } else {
2071                 char *old_options = options;
2072                 options = g_strdup_printf("%s,%s", options, optarg);
2073                 g_free(old_options);
2074             }
2075             break;
2076         case 'l':
2077             if (strstart(optarg, SNAPSHOT_OPT_BASE, NULL)) {
2078                 sn_opts = qemu_opts_parse_noisily(&internal_snapshot_opts,
2079                                                   optarg, false);
2080                 if (!sn_opts) {
2081                     error_report("Failed in parsing snapshot param '%s'",
2082                                  optarg);
2083                     goto fail_getopt;
2084                 }
2085             } else {
2086                 snapshot_name = optarg;
2087             }
2088             break;
2089         case 'S':
2090         {
2091             int64_t sval;
2092 
2093             sval = cvtnum(optarg);
2094             if (sval < 0 || sval & (BDRV_SECTOR_SIZE - 1) ||
2095                 sval / BDRV_SECTOR_SIZE > MAX_BUF_SECTORS) {
2096                 error_report("Invalid buffer size for sparse output specified. "
2097                     "Valid sizes are multiples of %llu up to %llu. Select "
2098                     "0 to disable sparse detection (fully allocates output).",
2099                     BDRV_SECTOR_SIZE, MAX_BUF_SECTORS * BDRV_SECTOR_SIZE);
2100                 goto fail_getopt;
2101             }
2102 
2103             s.min_sparse = sval / BDRV_SECTOR_SIZE;
2104             explict_min_sparse = true;
2105             break;
2106         }
2107         case 'p':
2108             progress = true;
2109             break;
2110         case 't':
2111             cache = optarg;
2112             break;
2113         case 'T':
2114             src_cache = optarg;
2115             break;
2116         case 'q':
2117             quiet = true;
2118             break;
2119         case 'n':
2120             skip_create = true;
2121             break;
2122         case 'm':
2123             if (qemu_strtol(optarg, NULL, 0, &s.num_coroutines) ||
2124                 s.num_coroutines < 1 || s.num_coroutines > MAX_COROUTINES) {
2125                 error_report("Invalid number of coroutines. Allowed number of"
2126                              " coroutines is between 1 and %d", MAX_COROUTINES);
2127                 goto fail_getopt;
2128             }
2129             break;
2130         case 'W':
2131             s.wr_in_order = false;
2132             break;
2133         case 'U':
2134             force_share = true;
2135             break;
2136         case OPTION_OBJECT: {
2137             QemuOpts *object_opts;
2138             object_opts = qemu_opts_parse_noisily(&qemu_object_opts,
2139                                                   optarg, true);
2140             if (!object_opts) {
2141                 goto fail_getopt;
2142             }
2143             break;
2144         }
2145         case OPTION_IMAGE_OPTS:
2146             image_opts = true;
2147             break;
2148         case OPTION_TARGET_IMAGE_OPTS:
2149             tgt_image_opts = true;
2150             break;
2151         }
2152     }
2153 
2154     if (!out_fmt && !tgt_image_opts) {
2155         out_fmt = "raw";
2156     }
2157 
2158     if (qemu_opts_foreach(&qemu_object_opts,
2159                           user_creatable_add_opts_foreach,
2160                           NULL, NULL)) {
2161         goto fail_getopt;
2162     }
2163 
2164     if (s.compressed && s.copy_range) {
2165         error_report("Cannot enable copy offloading when -c is used");
2166         goto fail_getopt;
2167     }
2168 
2169     if (explict_min_sparse && s.copy_range) {
2170         error_report("Cannot enable copy offloading when -S is used");
2171         goto fail_getopt;
2172     }
2173 
2174     if (tgt_image_opts && !skip_create) {
2175         error_report("--target-image-opts requires use of -n flag");
2176         goto fail_getopt;
2177     }
2178 
2179     s.src_num = argc - optind - 1;
2180     out_filename = s.src_num >= 1 ? argv[argc - 1] : NULL;
2181 
2182     if (options && has_help_option(options)) {
2183         if (out_fmt) {
2184             ret = print_block_option_help(out_filename, out_fmt);
2185             goto fail_getopt;
2186         } else {
2187             error_report("Option help requires a format be specified");
2188             goto fail_getopt;
2189         }
2190     }
2191 
2192     if (s.src_num < 1) {
2193         error_report("Must specify image file name");
2194         goto fail_getopt;
2195     }
2196 
2197 
2198     /* ret is still -EINVAL until here */
2199     ret = bdrv_parse_cache_mode(src_cache, &src_flags, &src_writethrough);
2200     if (ret < 0) {
2201         error_report("Invalid source cache option: %s", src_cache);
2202         goto fail_getopt;
2203     }
2204 
2205     /* Initialize before goto out */
2206     if (quiet) {
2207         progress = false;
2208     }
2209     qemu_progress_init(progress, 1.0);
2210     qemu_progress_print(0, 100);
2211 
2212     s.src = g_new0(BlockBackend *, s.src_num);
2213     s.src_sectors = g_new(int64_t, s.src_num);
2214 
2215     for (bs_i = 0; bs_i < s.src_num; bs_i++) {
2216         s.src[bs_i] = img_open(image_opts, argv[optind + bs_i],
2217                                fmt, src_flags, src_writethrough, quiet,
2218                                force_share);
2219         if (!s.src[bs_i]) {
2220             ret = -1;
2221             goto out;
2222         }
2223         s.src_sectors[bs_i] = blk_nb_sectors(s.src[bs_i]);
2224         if (s.src_sectors[bs_i] < 0) {
2225             error_report("Could not get size of %s: %s",
2226                          argv[optind + bs_i], strerror(-s.src_sectors[bs_i]));
2227             ret = -1;
2228             goto out;
2229         }
2230         s.total_sectors += s.src_sectors[bs_i];
2231     }
2232 
2233     if (sn_opts) {
2234         bdrv_snapshot_load_tmp(blk_bs(s.src[0]),
2235                                qemu_opt_get(sn_opts, SNAPSHOT_OPT_ID),
2236                                qemu_opt_get(sn_opts, SNAPSHOT_OPT_NAME),
2237                                &local_err);
2238     } else if (snapshot_name != NULL) {
2239         if (s.src_num > 1) {
2240             error_report("No support for concatenating multiple snapshot");
2241             ret = -1;
2242             goto out;
2243         }
2244 
2245         bdrv_snapshot_load_tmp_by_id_or_name(blk_bs(s.src[0]), snapshot_name,
2246                                              &local_err);
2247     }
2248     if (local_err) {
2249         error_reportf_err(local_err, "Failed to load snapshot: ");
2250         ret = -1;
2251         goto out;
2252     }
2253 
2254     if (!skip_create) {
2255         /* Find driver and parse its options */
2256         drv = bdrv_find_format(out_fmt);
2257         if (!drv) {
2258             error_report("Unknown file format '%s'", out_fmt);
2259             ret = -1;
2260             goto out;
2261         }
2262 
2263         proto_drv = bdrv_find_protocol(out_filename, true, &local_err);
2264         if (!proto_drv) {
2265             error_report_err(local_err);
2266             ret = -1;
2267             goto out;
2268         }
2269 
2270         if (!drv->create_opts) {
2271             error_report("Format driver '%s' does not support image creation",
2272                          drv->format_name);
2273             ret = -1;
2274             goto out;
2275         }
2276 
2277         if (!proto_drv->create_opts) {
2278             error_report("Protocol driver '%s' does not support image creation",
2279                          proto_drv->format_name);
2280             ret = -1;
2281             goto out;
2282         }
2283 
2284         create_opts = qemu_opts_append(create_opts, drv->create_opts);
2285         create_opts = qemu_opts_append(create_opts, proto_drv->create_opts);
2286 
2287         opts = qemu_opts_create(create_opts, NULL, 0, &error_abort);
2288         if (options) {
2289             qemu_opts_do_parse(opts, options, NULL, &local_err);
2290             if (local_err) {
2291                 error_report_err(local_err);
2292                 ret = -1;
2293                 goto out;
2294             }
2295         }
2296 
2297         qemu_opt_set_number(opts, BLOCK_OPT_SIZE, s.total_sectors * 512,
2298                             &error_abort);
2299         ret = add_old_style_options(out_fmt, opts, out_baseimg, NULL);
2300         if (ret < 0) {
2301             goto out;
2302         }
2303     }
2304 
2305     /* Get backing file name if -o backing_file was used */
2306     out_baseimg_param = qemu_opt_get(opts, BLOCK_OPT_BACKING_FILE);
2307     if (out_baseimg_param) {
2308         out_baseimg = out_baseimg_param;
2309     }
2310     s.target_has_backing = (bool) out_baseimg;
2311 
2312     if (s.src_num > 1 && out_baseimg) {
2313         error_report("Having a backing file for the target makes no sense when "
2314                      "concatenating multiple input images");
2315         ret = -1;
2316         goto out;
2317     }
2318 
2319     /* Check if compression is supported */
2320     if (s.compressed) {
2321         bool encryption =
2322             qemu_opt_get_bool(opts, BLOCK_OPT_ENCRYPT, false);
2323         const char *encryptfmt =
2324             qemu_opt_get(opts, BLOCK_OPT_ENCRYPT_FORMAT);
2325         const char *preallocation =
2326             qemu_opt_get(opts, BLOCK_OPT_PREALLOC);
2327 
2328         if (drv && !drv->bdrv_co_pwritev_compressed) {
2329             error_report("Compression not supported for this file format");
2330             ret = -1;
2331             goto out;
2332         }
2333 
2334         if (encryption || encryptfmt) {
2335             error_report("Compression and encryption not supported at "
2336                          "the same time");
2337             ret = -1;
2338             goto out;
2339         }
2340 
2341         if (preallocation
2342             && strcmp(preallocation, "off"))
2343         {
2344             error_report("Compression and preallocation not supported at "
2345                          "the same time");
2346             ret = -1;
2347             goto out;
2348         }
2349     }
2350 
2351     /*
2352      * The later open call will need any decryption secrets, and
2353      * bdrv_create() will purge "opts", so extract them now before
2354      * they are lost.
2355      */
2356     if (!skip_create) {
2357         open_opts = qdict_new();
2358         qemu_opt_foreach(opts, img_add_key_secrets, open_opts, &error_abort);
2359     }
2360 
2361     if (!skip_create) {
2362         /* Create the new image */
2363         ret = bdrv_create(drv, out_filename, opts, &local_err);
2364         if (ret < 0) {
2365             error_reportf_err(local_err, "%s: error while converting %s: ",
2366                               out_filename, out_fmt);
2367             goto out;
2368         }
2369     }
2370 
2371     flags = s.min_sparse ? (BDRV_O_RDWR | BDRV_O_UNMAP) : BDRV_O_RDWR;
2372     ret = bdrv_parse_cache_mode(cache, &flags, &writethrough);
2373     if (ret < 0) {
2374         error_report("Invalid cache option: %s", cache);
2375         goto out;
2376     }
2377 
2378     if (skip_create) {
2379         s.target = img_open(tgt_image_opts, out_filename, out_fmt,
2380                             flags, writethrough, quiet, false);
2381     } else {
2382         /* TODO ultimately we should allow --target-image-opts
2383          * to be used even when -n is not given.
2384          * That has to wait for bdrv_create to be improved
2385          * to allow filenames in option syntax
2386          */
2387         s.target = img_open_file(out_filename, open_opts, out_fmt,
2388                                  flags, writethrough, quiet, false);
2389         open_opts = NULL; /* blk_new_open will have freed it */
2390     }
2391     if (!s.target) {
2392         ret = -1;
2393         goto out;
2394     }
2395     out_bs = blk_bs(s.target);
2396 
2397     if (s.compressed && !out_bs->drv->bdrv_co_pwritev_compressed) {
2398         error_report("Compression not supported for this file format");
2399         ret = -1;
2400         goto out;
2401     }
2402 
2403     /* increase bufsectors from the default 4096 (2M) if opt_transfer
2404      * or discard_alignment of the out_bs is greater. Limit to
2405      * MAX_BUF_SECTORS as maximum which is currently 32768 (16MB). */
2406     s.buf_sectors = MIN(MAX_BUF_SECTORS,
2407                         MAX(s.buf_sectors,
2408                             MAX(out_bs->bl.opt_transfer >> BDRV_SECTOR_BITS,
2409                                 out_bs->bl.pdiscard_alignment >>
2410                                 BDRV_SECTOR_BITS)));
2411 
2412     /* try to align the write requests to the destination to avoid unnecessary
2413      * RMW cycles. */
2414     s.alignment = MAX(pow2floor(s.min_sparse),
2415                       DIV_ROUND_UP(out_bs->bl.request_alignment,
2416                                    BDRV_SECTOR_SIZE));
2417     assert(is_power_of_2(s.alignment));
2418 
2419     if (skip_create) {
2420         int64_t output_sectors = blk_nb_sectors(s.target);
2421         if (output_sectors < 0) {
2422             error_report("unable to get output image length: %s",
2423                          strerror(-output_sectors));
2424             ret = -1;
2425             goto out;
2426         } else if (output_sectors < s.total_sectors) {
2427             error_report("output file is smaller than input file");
2428             ret = -1;
2429             goto out;
2430         }
2431     }
2432 
2433     if (s.target_has_backing) {
2434         /* Errors are treated as "backing length unknown" (which means
2435          * s.target_backing_sectors has to be negative, which it will
2436          * be automatically).  The backing file length is used only
2437          * for optimizations, so such a case is not fatal. */
2438         s.target_backing_sectors = bdrv_nb_sectors(out_bs->backing->bs);
2439     } else {
2440         s.target_backing_sectors = -1;
2441     }
2442 
2443     ret = bdrv_get_info(out_bs, &bdi);
2444     if (ret < 0) {
2445         if (s.compressed) {
2446             error_report("could not get block driver info");
2447             goto out;
2448         }
2449     } else {
2450         s.compressed = s.compressed || bdi.needs_compressed_writes;
2451         s.cluster_sectors = bdi.cluster_size / BDRV_SECTOR_SIZE;
2452         s.unallocated_blocks_are_zero = bdi.unallocated_blocks_are_zero;
2453     }
2454 
2455     ret = convert_do_copy(&s);
2456 out:
2457     if (!ret) {
2458         qemu_progress_print(100, 0);
2459     }
2460     qemu_progress_end();
2461     qemu_opts_del(opts);
2462     qemu_opts_free(create_opts);
2463     qemu_opts_del(sn_opts);
2464     qobject_unref(open_opts);
2465     blk_unref(s.target);
2466     if (s.src) {
2467         for (bs_i = 0; bs_i < s.src_num; bs_i++) {
2468             blk_unref(s.src[bs_i]);
2469         }
2470         g_free(s.src);
2471     }
2472     g_free(s.src_sectors);
2473 fail_getopt:
2474     g_free(options);
2475 
2476     return !!ret;
2477 }
2478 
2479 
2480 static void dump_snapshots(BlockDriverState *bs)
2481 {
2482     QEMUSnapshotInfo *sn_tab, *sn;
2483     int nb_sns, i;
2484 
2485     nb_sns = bdrv_snapshot_list(bs, &sn_tab);
2486     if (nb_sns <= 0)
2487         return;
2488     printf("Snapshot list:\n");
2489     bdrv_snapshot_dump(fprintf, stdout, NULL);
2490     printf("\n");
2491     for(i = 0; i < nb_sns; i++) {
2492         sn = &sn_tab[i];
2493         bdrv_snapshot_dump(fprintf, stdout, sn);
2494         printf("\n");
2495     }
2496     g_free(sn_tab);
2497 }
2498 
2499 static void dump_json_image_info_list(ImageInfoList *list)
2500 {
2501     QString *str;
2502     QObject *obj;
2503     Visitor *v = qobject_output_visitor_new(&obj);
2504 
2505     visit_type_ImageInfoList(v, NULL, &list, &error_abort);
2506     visit_complete(v, &obj);
2507     str = qobject_to_json_pretty(obj);
2508     assert(str != NULL);
2509     printf("%s\n", qstring_get_str(str));
2510     qobject_unref(obj);
2511     visit_free(v);
2512     qobject_unref(str);
2513 }
2514 
2515 static void dump_json_image_info(ImageInfo *info)
2516 {
2517     QString *str;
2518     QObject *obj;
2519     Visitor *v = qobject_output_visitor_new(&obj);
2520 
2521     visit_type_ImageInfo(v, NULL, &info, &error_abort);
2522     visit_complete(v, &obj);
2523     str = qobject_to_json_pretty(obj);
2524     assert(str != NULL);
2525     printf("%s\n", qstring_get_str(str));
2526     qobject_unref(obj);
2527     visit_free(v);
2528     qobject_unref(str);
2529 }
2530 
2531 static void dump_human_image_info_list(ImageInfoList *list)
2532 {
2533     ImageInfoList *elem;
2534     bool delim = false;
2535 
2536     for (elem = list; elem; elem = elem->next) {
2537         if (delim) {
2538             printf("\n");
2539         }
2540         delim = true;
2541 
2542         bdrv_image_info_dump(fprintf, stdout, elem->value);
2543     }
2544 }
2545 
2546 static gboolean str_equal_func(gconstpointer a, gconstpointer b)
2547 {
2548     return strcmp(a, b) == 0;
2549 }
2550 
2551 /**
2552  * Open an image file chain and return an ImageInfoList
2553  *
2554  * @filename: topmost image filename
2555  * @fmt: topmost image format (may be NULL to autodetect)
2556  * @chain: true  - enumerate entire backing file chain
2557  *         false - only topmost image file
2558  *
2559  * Returns a list of ImageInfo objects or NULL if there was an error opening an
2560  * image file.  If there was an error a message will have been printed to
2561  * stderr.
2562  */
2563 static ImageInfoList *collect_image_info_list(bool image_opts,
2564                                               const char *filename,
2565                                               const char *fmt,
2566                                               bool chain, bool force_share)
2567 {
2568     ImageInfoList *head = NULL;
2569     ImageInfoList **last = &head;
2570     GHashTable *filenames;
2571     Error *err = NULL;
2572 
2573     filenames = g_hash_table_new_full(g_str_hash, str_equal_func, NULL, NULL);
2574 
2575     while (filename) {
2576         BlockBackend *blk;
2577         BlockDriverState *bs;
2578         ImageInfo *info;
2579         ImageInfoList *elem;
2580 
2581         if (g_hash_table_lookup_extended(filenames, filename, NULL, NULL)) {
2582             error_report("Backing file '%s' creates an infinite loop.",
2583                          filename);
2584             goto err;
2585         }
2586         g_hash_table_insert(filenames, (gpointer)filename, NULL);
2587 
2588         blk = img_open(image_opts, filename, fmt,
2589                        BDRV_O_NO_BACKING | BDRV_O_NO_IO, false, false,
2590                        force_share);
2591         if (!blk) {
2592             goto err;
2593         }
2594         bs = blk_bs(blk);
2595 
2596         bdrv_query_image_info(bs, &info, &err);
2597         if (err) {
2598             error_report_err(err);
2599             blk_unref(blk);
2600             goto err;
2601         }
2602 
2603         elem = g_new0(ImageInfoList, 1);
2604         elem->value = info;
2605         *last = elem;
2606         last = &elem->next;
2607 
2608         blk_unref(blk);
2609 
2610         filename = fmt = NULL;
2611         if (chain) {
2612             if (info->has_full_backing_filename) {
2613                 filename = info->full_backing_filename;
2614             } else if (info->has_backing_filename) {
2615                 error_report("Could not determine absolute backing filename,"
2616                              " but backing filename '%s' present",
2617                              info->backing_filename);
2618                 goto err;
2619             }
2620             if (info->has_backing_filename_format) {
2621                 fmt = info->backing_filename_format;
2622             }
2623         }
2624     }
2625     g_hash_table_destroy(filenames);
2626     return head;
2627 
2628 err:
2629     qapi_free_ImageInfoList(head);
2630     g_hash_table_destroy(filenames);
2631     return NULL;
2632 }
2633 
2634 static int img_info(int argc, char **argv)
2635 {
2636     int c;
2637     OutputFormat output_format = OFORMAT_HUMAN;
2638     bool chain = false;
2639     const char *filename, *fmt, *output;
2640     ImageInfoList *list;
2641     bool image_opts = false;
2642     bool force_share = false;
2643 
2644     fmt = NULL;
2645     output = NULL;
2646     for(;;) {
2647         int option_index = 0;
2648         static const struct option long_options[] = {
2649             {"help", no_argument, 0, 'h'},
2650             {"format", required_argument, 0, 'f'},
2651             {"output", required_argument, 0, OPTION_OUTPUT},
2652             {"backing-chain", no_argument, 0, OPTION_BACKING_CHAIN},
2653             {"object", required_argument, 0, OPTION_OBJECT},
2654             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
2655             {"force-share", no_argument, 0, 'U'},
2656             {0, 0, 0, 0}
2657         };
2658         c = getopt_long(argc, argv, ":f:hU",
2659                         long_options, &option_index);
2660         if (c == -1) {
2661             break;
2662         }
2663         switch(c) {
2664         case ':':
2665             missing_argument(argv[optind - 1]);
2666             break;
2667         case '?':
2668             unrecognized_option(argv[optind - 1]);
2669             break;
2670         case 'h':
2671             help();
2672             break;
2673         case 'f':
2674             fmt = optarg;
2675             break;
2676         case 'U':
2677             force_share = true;
2678             break;
2679         case OPTION_OUTPUT:
2680             output = optarg;
2681             break;
2682         case OPTION_BACKING_CHAIN:
2683             chain = true;
2684             break;
2685         case OPTION_OBJECT: {
2686             QemuOpts *opts;
2687             opts = qemu_opts_parse_noisily(&qemu_object_opts,
2688                                            optarg, true);
2689             if (!opts) {
2690                 return 1;
2691             }
2692         }   break;
2693         case OPTION_IMAGE_OPTS:
2694             image_opts = true;
2695             break;
2696         }
2697     }
2698     if (optind != argc - 1) {
2699         error_exit("Expecting one image file name");
2700     }
2701     filename = argv[optind++];
2702 
2703     if (output && !strcmp(output, "json")) {
2704         output_format = OFORMAT_JSON;
2705     } else if (output && !strcmp(output, "human")) {
2706         output_format = OFORMAT_HUMAN;
2707     } else if (output) {
2708         error_report("--output must be used with human or json as argument.");
2709         return 1;
2710     }
2711 
2712     if (qemu_opts_foreach(&qemu_object_opts,
2713                           user_creatable_add_opts_foreach,
2714                           NULL, NULL)) {
2715         return 1;
2716     }
2717 
2718     list = collect_image_info_list(image_opts, filename, fmt, chain,
2719                                    force_share);
2720     if (!list) {
2721         return 1;
2722     }
2723 
2724     switch (output_format) {
2725     case OFORMAT_HUMAN:
2726         dump_human_image_info_list(list);
2727         break;
2728     case OFORMAT_JSON:
2729         if (chain) {
2730             dump_json_image_info_list(list);
2731         } else {
2732             dump_json_image_info(list->value);
2733         }
2734         break;
2735     }
2736 
2737     qapi_free_ImageInfoList(list);
2738     return 0;
2739 }
2740 
2741 static void dump_map_entry(OutputFormat output_format, MapEntry *e,
2742                            MapEntry *next)
2743 {
2744     switch (output_format) {
2745     case OFORMAT_HUMAN:
2746         if (e->data && !e->has_offset) {
2747             error_report("File contains external, encrypted or compressed clusters.");
2748             exit(1);
2749         }
2750         if (e->data && !e->zero) {
2751             printf("%#-16"PRIx64"%#-16"PRIx64"%#-16"PRIx64"%s\n",
2752                    e->start, e->length,
2753                    e->has_offset ? e->offset : 0,
2754                    e->has_filename ? e->filename : "");
2755         }
2756         /* This format ignores the distinction between 0, ZERO and ZERO|DATA.
2757          * Modify the flags here to allow more coalescing.
2758          */
2759         if (next && (!next->data || next->zero)) {
2760             next->data = false;
2761             next->zero = true;
2762         }
2763         break;
2764     case OFORMAT_JSON:
2765         printf("%s{ \"start\": %"PRId64", \"length\": %"PRId64","
2766                " \"depth\": %"PRId64", \"zero\": %s, \"data\": %s",
2767                (e->start == 0 ? "[" : ",\n"),
2768                e->start, e->length, e->depth,
2769                e->zero ? "true" : "false",
2770                e->data ? "true" : "false");
2771         if (e->has_offset) {
2772             printf(", \"offset\": %"PRId64"", e->offset);
2773         }
2774         putchar('}');
2775 
2776         if (!next) {
2777             printf("]\n");
2778         }
2779         break;
2780     }
2781 }
2782 
2783 static int get_block_status(BlockDriverState *bs, int64_t offset,
2784                             int64_t bytes, MapEntry *e)
2785 {
2786     int ret;
2787     int depth;
2788     BlockDriverState *file;
2789     bool has_offset;
2790     int64_t map;
2791 
2792     /* As an optimization, we could cache the current range of unallocated
2793      * clusters in each file of the chain, and avoid querying the same
2794      * range repeatedly.
2795      */
2796 
2797     depth = 0;
2798     for (;;) {
2799         ret = bdrv_block_status(bs, offset, bytes, &bytes, &map, &file);
2800         if (ret < 0) {
2801             return ret;
2802         }
2803         assert(bytes);
2804         if (ret & (BDRV_BLOCK_ZERO|BDRV_BLOCK_DATA)) {
2805             break;
2806         }
2807         bs = backing_bs(bs);
2808         if (bs == NULL) {
2809             ret = 0;
2810             break;
2811         }
2812 
2813         depth++;
2814     }
2815 
2816     has_offset = !!(ret & BDRV_BLOCK_OFFSET_VALID);
2817 
2818     *e = (MapEntry) {
2819         .start = offset,
2820         .length = bytes,
2821         .data = !!(ret & BDRV_BLOCK_DATA),
2822         .zero = !!(ret & BDRV_BLOCK_ZERO),
2823         .offset = map,
2824         .has_offset = has_offset,
2825         .depth = depth,
2826         .has_filename = file && has_offset,
2827         .filename = file && has_offset ? file->filename : NULL,
2828     };
2829 
2830     return 0;
2831 }
2832 
2833 static inline bool entry_mergeable(const MapEntry *curr, const MapEntry *next)
2834 {
2835     if (curr->length == 0) {
2836         return false;
2837     }
2838     if (curr->zero != next->zero ||
2839         curr->data != next->data ||
2840         curr->depth != next->depth ||
2841         curr->has_filename != next->has_filename ||
2842         curr->has_offset != next->has_offset) {
2843         return false;
2844     }
2845     if (curr->has_filename && strcmp(curr->filename, next->filename)) {
2846         return false;
2847     }
2848     if (curr->has_offset && curr->offset + curr->length != next->offset) {
2849         return false;
2850     }
2851     return true;
2852 }
2853 
2854 static int img_map(int argc, char **argv)
2855 {
2856     int c;
2857     OutputFormat output_format = OFORMAT_HUMAN;
2858     BlockBackend *blk;
2859     BlockDriverState *bs;
2860     const char *filename, *fmt, *output;
2861     int64_t length;
2862     MapEntry curr = { .length = 0 }, next;
2863     int ret = 0;
2864     bool image_opts = false;
2865     bool force_share = false;
2866 
2867     fmt = NULL;
2868     output = NULL;
2869     for (;;) {
2870         int option_index = 0;
2871         static const struct option long_options[] = {
2872             {"help", no_argument, 0, 'h'},
2873             {"format", required_argument, 0, 'f'},
2874             {"output", required_argument, 0, OPTION_OUTPUT},
2875             {"object", required_argument, 0, OPTION_OBJECT},
2876             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
2877             {"force-share", no_argument, 0, 'U'},
2878             {0, 0, 0, 0}
2879         };
2880         c = getopt_long(argc, argv, ":f:hU",
2881                         long_options, &option_index);
2882         if (c == -1) {
2883             break;
2884         }
2885         switch (c) {
2886         case ':':
2887             missing_argument(argv[optind - 1]);
2888             break;
2889         case '?':
2890             unrecognized_option(argv[optind - 1]);
2891             break;
2892         case 'h':
2893             help();
2894             break;
2895         case 'f':
2896             fmt = optarg;
2897             break;
2898         case 'U':
2899             force_share = true;
2900             break;
2901         case OPTION_OUTPUT:
2902             output = optarg;
2903             break;
2904         case OPTION_OBJECT: {
2905             QemuOpts *opts;
2906             opts = qemu_opts_parse_noisily(&qemu_object_opts,
2907                                            optarg, true);
2908             if (!opts) {
2909                 return 1;
2910             }
2911         }   break;
2912         case OPTION_IMAGE_OPTS:
2913             image_opts = true;
2914             break;
2915         }
2916     }
2917     if (optind != argc - 1) {
2918         error_exit("Expecting one image file name");
2919     }
2920     filename = argv[optind];
2921 
2922     if (output && !strcmp(output, "json")) {
2923         output_format = OFORMAT_JSON;
2924     } else if (output && !strcmp(output, "human")) {
2925         output_format = OFORMAT_HUMAN;
2926     } else if (output) {
2927         error_report("--output must be used with human or json as argument.");
2928         return 1;
2929     }
2930 
2931     if (qemu_opts_foreach(&qemu_object_opts,
2932                           user_creatable_add_opts_foreach,
2933                           NULL, NULL)) {
2934         return 1;
2935     }
2936 
2937     blk = img_open(image_opts, filename, fmt, 0, false, false, force_share);
2938     if (!blk) {
2939         return 1;
2940     }
2941     bs = blk_bs(blk);
2942 
2943     if (output_format == OFORMAT_HUMAN) {
2944         printf("%-16s%-16s%-16s%s\n", "Offset", "Length", "Mapped to", "File");
2945     }
2946 
2947     length = blk_getlength(blk);
2948     while (curr.start + curr.length < length) {
2949         int64_t offset = curr.start + curr.length;
2950         int64_t n;
2951 
2952         /* Probe up to 1 GiB at a time.  */
2953         n = MIN(1 << 30, length - offset);
2954         ret = get_block_status(bs, offset, n, &next);
2955 
2956         if (ret < 0) {
2957             error_report("Could not read file metadata: %s", strerror(-ret));
2958             goto out;
2959         }
2960 
2961         if (entry_mergeable(&curr, &next)) {
2962             curr.length += next.length;
2963             continue;
2964         }
2965 
2966         if (curr.length > 0) {
2967             dump_map_entry(output_format, &curr, &next);
2968         }
2969         curr = next;
2970     }
2971 
2972     dump_map_entry(output_format, &curr, NULL);
2973 
2974 out:
2975     blk_unref(blk);
2976     return ret < 0;
2977 }
2978 
2979 #define SNAPSHOT_LIST   1
2980 #define SNAPSHOT_CREATE 2
2981 #define SNAPSHOT_APPLY  3
2982 #define SNAPSHOT_DELETE 4
2983 
2984 static int img_snapshot(int argc, char **argv)
2985 {
2986     BlockBackend *blk;
2987     BlockDriverState *bs;
2988     QEMUSnapshotInfo sn;
2989     char *filename, *snapshot_name = NULL;
2990     int c, ret = 0, bdrv_oflags;
2991     int action = 0;
2992     qemu_timeval tv;
2993     bool quiet = false;
2994     Error *err = NULL;
2995     bool image_opts = false;
2996     bool force_share = false;
2997 
2998     bdrv_oflags = BDRV_O_RDWR;
2999     /* Parse commandline parameters */
3000     for(;;) {
3001         static const struct option long_options[] = {
3002             {"help", no_argument, 0, 'h'},
3003             {"object", required_argument, 0, OPTION_OBJECT},
3004             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
3005             {"force-share", no_argument, 0, 'U'},
3006             {0, 0, 0, 0}
3007         };
3008         c = getopt_long(argc, argv, ":la:c:d:hqU",
3009                         long_options, NULL);
3010         if (c == -1) {
3011             break;
3012         }
3013         switch(c) {
3014         case ':':
3015             missing_argument(argv[optind - 1]);
3016             break;
3017         case '?':
3018             unrecognized_option(argv[optind - 1]);
3019             break;
3020         case 'h':
3021             help();
3022             return 0;
3023         case 'l':
3024             if (action) {
3025                 error_exit("Cannot mix '-l', '-a', '-c', '-d'");
3026                 return 0;
3027             }
3028             action = SNAPSHOT_LIST;
3029             bdrv_oflags &= ~BDRV_O_RDWR; /* no need for RW */
3030             break;
3031         case 'a':
3032             if (action) {
3033                 error_exit("Cannot mix '-l', '-a', '-c', '-d'");
3034                 return 0;
3035             }
3036             action = SNAPSHOT_APPLY;
3037             snapshot_name = optarg;
3038             break;
3039         case 'c':
3040             if (action) {
3041                 error_exit("Cannot mix '-l', '-a', '-c', '-d'");
3042                 return 0;
3043             }
3044             action = SNAPSHOT_CREATE;
3045             snapshot_name = optarg;
3046             break;
3047         case 'd':
3048             if (action) {
3049                 error_exit("Cannot mix '-l', '-a', '-c', '-d'");
3050                 return 0;
3051             }
3052             action = SNAPSHOT_DELETE;
3053             snapshot_name = optarg;
3054             break;
3055         case 'q':
3056             quiet = true;
3057             break;
3058         case 'U':
3059             force_share = true;
3060             break;
3061         case OPTION_OBJECT: {
3062             QemuOpts *opts;
3063             opts = qemu_opts_parse_noisily(&qemu_object_opts,
3064                                            optarg, true);
3065             if (!opts) {
3066                 return 1;
3067             }
3068         }   break;
3069         case OPTION_IMAGE_OPTS:
3070             image_opts = true;
3071             break;
3072         }
3073     }
3074 
3075     if (optind != argc - 1) {
3076         error_exit("Expecting one image file name");
3077     }
3078     filename = argv[optind++];
3079 
3080     if (qemu_opts_foreach(&qemu_object_opts,
3081                           user_creatable_add_opts_foreach,
3082                           NULL, NULL)) {
3083         return 1;
3084     }
3085 
3086     /* Open the image */
3087     blk = img_open(image_opts, filename, NULL, bdrv_oflags, false, quiet,
3088                    force_share);
3089     if (!blk) {
3090         return 1;
3091     }
3092     bs = blk_bs(blk);
3093 
3094     /* Perform the requested action */
3095     switch(action) {
3096     case SNAPSHOT_LIST:
3097         dump_snapshots(bs);
3098         break;
3099 
3100     case SNAPSHOT_CREATE:
3101         memset(&sn, 0, sizeof(sn));
3102         pstrcpy(sn.name, sizeof(sn.name), snapshot_name);
3103 
3104         qemu_gettimeofday(&tv);
3105         sn.date_sec = tv.tv_sec;
3106         sn.date_nsec = tv.tv_usec * 1000;
3107 
3108         ret = bdrv_snapshot_create(bs, &sn);
3109         if (ret) {
3110             error_report("Could not create snapshot '%s': %d (%s)",
3111                 snapshot_name, ret, strerror(-ret));
3112         }
3113         break;
3114 
3115     case SNAPSHOT_APPLY:
3116         ret = bdrv_snapshot_goto(bs, snapshot_name, &err);
3117         if (ret) {
3118             error_reportf_err(err, "Could not apply snapshot '%s': ",
3119                               snapshot_name);
3120         }
3121         break;
3122 
3123     case SNAPSHOT_DELETE:
3124         bdrv_snapshot_delete_by_id_or_name(bs, snapshot_name, &err);
3125         if (err) {
3126             error_reportf_err(err, "Could not delete snapshot '%s': ",
3127                               snapshot_name);
3128             ret = 1;
3129         }
3130         break;
3131     }
3132 
3133     /* Cleanup */
3134     blk_unref(blk);
3135     if (ret) {
3136         return 1;
3137     }
3138     return 0;
3139 }
3140 
3141 static int img_rebase(int argc, char **argv)
3142 {
3143     BlockBackend *blk = NULL, *blk_old_backing = NULL, *blk_new_backing = NULL;
3144     uint8_t *buf_old = NULL;
3145     uint8_t *buf_new = NULL;
3146     BlockDriverState *bs = NULL;
3147     char *filename;
3148     const char *fmt, *cache, *src_cache, *out_basefmt, *out_baseimg;
3149     int c, flags, src_flags, ret;
3150     bool writethrough, src_writethrough;
3151     int unsafe = 0;
3152     bool force_share = false;
3153     int progress = 0;
3154     bool quiet = false;
3155     Error *local_err = NULL;
3156     bool image_opts = false;
3157 
3158     /* Parse commandline parameters */
3159     fmt = NULL;
3160     cache = BDRV_DEFAULT_CACHE;
3161     src_cache = BDRV_DEFAULT_CACHE;
3162     out_baseimg = NULL;
3163     out_basefmt = NULL;
3164     for(;;) {
3165         static const struct option long_options[] = {
3166             {"help", no_argument, 0, 'h'},
3167             {"object", required_argument, 0, OPTION_OBJECT},
3168             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
3169             {"force-share", no_argument, 0, 'U'},
3170             {0, 0, 0, 0}
3171         };
3172         c = getopt_long(argc, argv, ":hf:F:b:upt:T:qU",
3173                         long_options, NULL);
3174         if (c == -1) {
3175             break;
3176         }
3177         switch(c) {
3178         case ':':
3179             missing_argument(argv[optind - 1]);
3180             break;
3181         case '?':
3182             unrecognized_option(argv[optind - 1]);
3183             break;
3184         case 'h':
3185             help();
3186             return 0;
3187         case 'f':
3188             fmt = optarg;
3189             break;
3190         case 'F':
3191             out_basefmt = optarg;
3192             break;
3193         case 'b':
3194             out_baseimg = optarg;
3195             break;
3196         case 'u':
3197             unsafe = 1;
3198             break;
3199         case 'p':
3200             progress = 1;
3201             break;
3202         case 't':
3203             cache = optarg;
3204             break;
3205         case 'T':
3206             src_cache = optarg;
3207             break;
3208         case 'q':
3209             quiet = true;
3210             break;
3211         case OPTION_OBJECT: {
3212             QemuOpts *opts;
3213             opts = qemu_opts_parse_noisily(&qemu_object_opts,
3214                                            optarg, true);
3215             if (!opts) {
3216                 return 1;
3217             }
3218         }   break;
3219         case OPTION_IMAGE_OPTS:
3220             image_opts = true;
3221             break;
3222         case 'U':
3223             force_share = true;
3224             break;
3225         }
3226     }
3227 
3228     if (quiet) {
3229         progress = 0;
3230     }
3231 
3232     if (optind != argc - 1) {
3233         error_exit("Expecting one image file name");
3234     }
3235     if (!unsafe && !out_baseimg) {
3236         error_exit("Must specify backing file (-b) or use unsafe mode (-u)");
3237     }
3238     filename = argv[optind++];
3239 
3240     if (qemu_opts_foreach(&qemu_object_opts,
3241                           user_creatable_add_opts_foreach,
3242                           NULL, NULL)) {
3243         return 1;
3244     }
3245 
3246     qemu_progress_init(progress, 2.0);
3247     qemu_progress_print(0, 100);
3248 
3249     flags = BDRV_O_RDWR | (unsafe ? BDRV_O_NO_BACKING : 0);
3250     ret = bdrv_parse_cache_mode(cache, &flags, &writethrough);
3251     if (ret < 0) {
3252         error_report("Invalid cache option: %s", cache);
3253         goto out;
3254     }
3255 
3256     src_flags = 0;
3257     ret = bdrv_parse_cache_mode(src_cache, &src_flags, &src_writethrough);
3258     if (ret < 0) {
3259         error_report("Invalid source cache option: %s", src_cache);
3260         goto out;
3261     }
3262 
3263     /* The source files are opened read-only, don't care about WCE */
3264     assert((src_flags & BDRV_O_RDWR) == 0);
3265     (void) src_writethrough;
3266 
3267     /*
3268      * Open the images.
3269      *
3270      * Ignore the old backing file for unsafe rebase in case we want to correct
3271      * the reference to a renamed or moved backing file.
3272      */
3273     blk = img_open(image_opts, filename, fmt, flags, writethrough, quiet,
3274                    false);
3275     if (!blk) {
3276         ret = -1;
3277         goto out;
3278     }
3279     bs = blk_bs(blk);
3280 
3281     if (out_basefmt != NULL) {
3282         if (bdrv_find_format(out_basefmt) == NULL) {
3283             error_report("Invalid format name: '%s'", out_basefmt);
3284             ret = -1;
3285             goto out;
3286         }
3287     }
3288 
3289     /* For safe rebasing we need to compare old and new backing file */
3290     if (!unsafe) {
3291         char backing_name[PATH_MAX];
3292         QDict *options = NULL;
3293 
3294         if (bs->backing_format[0] != '\0') {
3295             options = qdict_new();
3296             qdict_put_str(options, "driver", bs->backing_format);
3297         }
3298 
3299         if (force_share) {
3300             if (!options) {
3301                 options = qdict_new();
3302             }
3303             qdict_put_bool(options, BDRV_OPT_FORCE_SHARE, true);
3304         }
3305         bdrv_get_backing_filename(bs, backing_name, sizeof(backing_name));
3306         blk_old_backing = blk_new_open(backing_name, NULL,
3307                                        options, src_flags, &local_err);
3308         if (!blk_old_backing) {
3309             error_reportf_err(local_err,
3310                               "Could not open old backing file '%s': ",
3311                               backing_name);
3312             ret = -1;
3313             goto out;
3314         }
3315 
3316         if (out_baseimg[0]) {
3317             const char *overlay_filename;
3318             char *out_real_path;
3319 
3320             options = qdict_new();
3321             if (out_basefmt) {
3322                 qdict_put_str(options, "driver", out_basefmt);
3323             }
3324             if (force_share) {
3325                 qdict_put_bool(options, BDRV_OPT_FORCE_SHARE, true);
3326             }
3327 
3328             overlay_filename = bs->exact_filename[0] ? bs->exact_filename
3329                                                      : bs->filename;
3330             out_real_path = g_malloc(PATH_MAX);
3331 
3332             bdrv_get_full_backing_filename_from_filename(overlay_filename,
3333                                                          out_baseimg,
3334                                                          out_real_path,
3335                                                          PATH_MAX,
3336                                                          &local_err);
3337             if (local_err) {
3338                 error_reportf_err(local_err,
3339                                   "Could not resolve backing filename: ");
3340                 ret = -1;
3341                 g_free(out_real_path);
3342                 goto out;
3343             }
3344 
3345             blk_new_backing = blk_new_open(out_real_path, NULL,
3346                                            options, src_flags, &local_err);
3347             g_free(out_real_path);
3348             if (!blk_new_backing) {
3349                 error_reportf_err(local_err,
3350                                   "Could not open new backing file '%s': ",
3351                                   out_baseimg);
3352                 ret = -1;
3353                 goto out;
3354             }
3355         }
3356     }
3357 
3358     /*
3359      * Check each unallocated cluster in the COW file. If it is unallocated,
3360      * accesses go to the backing file. We must therefore compare this cluster
3361      * in the old and new backing file, and if they differ we need to copy it
3362      * from the old backing file into the COW file.
3363      *
3364      * If qemu-img crashes during this step, no harm is done. The content of
3365      * the image is the same as the original one at any time.
3366      */
3367     if (!unsafe) {
3368         int64_t size;
3369         int64_t old_backing_size;
3370         int64_t new_backing_size = 0;
3371         uint64_t offset;
3372         int64_t n;
3373         float local_progress = 0;
3374 
3375         buf_old = blk_blockalign(blk, IO_BUF_SIZE);
3376         buf_new = blk_blockalign(blk, IO_BUF_SIZE);
3377 
3378         size = blk_getlength(blk);
3379         if (size < 0) {
3380             error_report("Could not get size of '%s': %s",
3381                          filename, strerror(-size));
3382             ret = -1;
3383             goto out;
3384         }
3385         old_backing_size = blk_getlength(blk_old_backing);
3386         if (old_backing_size < 0) {
3387             char backing_name[PATH_MAX];
3388 
3389             bdrv_get_backing_filename(bs, backing_name, sizeof(backing_name));
3390             error_report("Could not get size of '%s': %s",
3391                          backing_name, strerror(-old_backing_size));
3392             ret = -1;
3393             goto out;
3394         }
3395         if (blk_new_backing) {
3396             new_backing_size = blk_getlength(blk_new_backing);
3397             if (new_backing_size < 0) {
3398                 error_report("Could not get size of '%s': %s",
3399                              out_baseimg, strerror(-new_backing_size));
3400                 ret = -1;
3401                 goto out;
3402             }
3403         }
3404 
3405         if (size != 0) {
3406             local_progress = (float)100 / (size / MIN(size, IO_BUF_SIZE));
3407         }
3408 
3409         for (offset = 0; offset < size; offset += n) {
3410             /* How many bytes can we handle with the next read? */
3411             n = MIN(IO_BUF_SIZE, size - offset);
3412 
3413             /* If the cluster is allocated, we don't need to take action */
3414             ret = bdrv_is_allocated(bs, offset, n, &n);
3415             if (ret < 0) {
3416                 error_report("error while reading image metadata: %s",
3417                              strerror(-ret));
3418                 goto out;
3419             }
3420             if (ret) {
3421                 continue;
3422             }
3423 
3424             /*
3425              * Read old and new backing file and take into consideration that
3426              * backing files may be smaller than the COW image.
3427              */
3428             if (offset >= old_backing_size) {
3429                 memset(buf_old, 0, n);
3430             } else {
3431                 if (offset + n > old_backing_size) {
3432                     n = old_backing_size - offset;
3433                 }
3434 
3435                 ret = blk_pread(blk_old_backing, offset, buf_old, n);
3436                 if (ret < 0) {
3437                     error_report("error while reading from old backing file");
3438                     goto out;
3439                 }
3440             }
3441 
3442             if (offset >= new_backing_size || !blk_new_backing) {
3443                 memset(buf_new, 0, n);
3444             } else {
3445                 if (offset + n > new_backing_size) {
3446                     n = new_backing_size - offset;
3447                 }
3448 
3449                 ret = blk_pread(blk_new_backing, offset, buf_new, n);
3450                 if (ret < 0) {
3451                     error_report("error while reading from new backing file");
3452                     goto out;
3453                 }
3454             }
3455 
3456             /* If they differ, we need to write to the COW file */
3457             uint64_t written = 0;
3458 
3459             while (written < n) {
3460                 int64_t pnum;
3461 
3462                 if (compare_buffers(buf_old + written, buf_new + written,
3463                                     n - written, &pnum))
3464                 {
3465                     ret = blk_pwrite(blk, offset + written,
3466                                      buf_old + written, pnum, 0);
3467                     if (ret < 0) {
3468                         error_report("Error while writing to COW image: %s",
3469                             strerror(-ret));
3470                         goto out;
3471                     }
3472                 }
3473 
3474                 written += pnum;
3475             }
3476             qemu_progress_print(local_progress, 100);
3477         }
3478     }
3479 
3480     /*
3481      * Change the backing file. All clusters that are different from the old
3482      * backing file are overwritten in the COW file now, so the visible content
3483      * doesn't change when we switch the backing file.
3484      */
3485     if (out_baseimg && *out_baseimg) {
3486         ret = bdrv_change_backing_file(bs, out_baseimg, out_basefmt);
3487     } else {
3488         ret = bdrv_change_backing_file(bs, NULL, NULL);
3489     }
3490 
3491     if (ret == -ENOSPC) {
3492         error_report("Could not change the backing file to '%s': No "
3493                      "space left in the file header", out_baseimg);
3494     } else if (ret < 0) {
3495         error_report("Could not change the backing file to '%s': %s",
3496             out_baseimg, strerror(-ret));
3497     }
3498 
3499     qemu_progress_print(100, 0);
3500     /*
3501      * TODO At this point it is possible to check if any clusters that are
3502      * allocated in the COW file are the same in the backing file. If so, they
3503      * could be dropped from the COW file. Don't do this before switching the
3504      * backing file, in case of a crash this would lead to corruption.
3505      */
3506 out:
3507     qemu_progress_end();
3508     /* Cleanup */
3509     if (!unsafe) {
3510         blk_unref(blk_old_backing);
3511         blk_unref(blk_new_backing);
3512     }
3513     qemu_vfree(buf_old);
3514     qemu_vfree(buf_new);
3515 
3516     blk_unref(blk);
3517     if (ret) {
3518         return 1;
3519     }
3520     return 0;
3521 }
3522 
3523 static int img_resize(int argc, char **argv)
3524 {
3525     Error *err = NULL;
3526     int c, ret, relative;
3527     const char *filename, *fmt, *size;
3528     int64_t n, total_size, current_size, new_size;
3529     bool quiet = false;
3530     BlockBackend *blk = NULL;
3531     PreallocMode prealloc = PREALLOC_MODE_OFF;
3532     QemuOpts *param;
3533 
3534     static QemuOptsList resize_options = {
3535         .name = "resize_options",
3536         .head = QTAILQ_HEAD_INITIALIZER(resize_options.head),
3537         .desc = {
3538             {
3539                 .name = BLOCK_OPT_SIZE,
3540                 .type = QEMU_OPT_SIZE,
3541                 .help = "Virtual disk size"
3542             }, {
3543                 /* end of list */
3544             }
3545         },
3546     };
3547     bool image_opts = false;
3548     bool shrink = false;
3549 
3550     /* Remove size from argv manually so that negative numbers are not treated
3551      * as options by getopt. */
3552     if (argc < 3) {
3553         error_exit("Not enough arguments");
3554         return 1;
3555     }
3556 
3557     size = argv[--argc];
3558 
3559     /* Parse getopt arguments */
3560     fmt = NULL;
3561     for(;;) {
3562         static const struct option long_options[] = {
3563             {"help", no_argument, 0, 'h'},
3564             {"object", required_argument, 0, OPTION_OBJECT},
3565             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
3566             {"preallocation", required_argument, 0, OPTION_PREALLOCATION},
3567             {"shrink", no_argument, 0, OPTION_SHRINK},
3568             {0, 0, 0, 0}
3569         };
3570         c = getopt_long(argc, argv, ":f:hq",
3571                         long_options, NULL);
3572         if (c == -1) {
3573             break;
3574         }
3575         switch(c) {
3576         case ':':
3577             missing_argument(argv[optind - 1]);
3578             break;
3579         case '?':
3580             unrecognized_option(argv[optind - 1]);
3581             break;
3582         case 'h':
3583             help();
3584             break;
3585         case 'f':
3586             fmt = optarg;
3587             break;
3588         case 'q':
3589             quiet = true;
3590             break;
3591         case OPTION_OBJECT: {
3592             QemuOpts *opts;
3593             opts = qemu_opts_parse_noisily(&qemu_object_opts,
3594                                            optarg, true);
3595             if (!opts) {
3596                 return 1;
3597             }
3598         }   break;
3599         case OPTION_IMAGE_OPTS:
3600             image_opts = true;
3601             break;
3602         case OPTION_PREALLOCATION:
3603             prealloc = qapi_enum_parse(&PreallocMode_lookup, optarg,
3604                                        PREALLOC_MODE__MAX, NULL);
3605             if (prealloc == PREALLOC_MODE__MAX) {
3606                 error_report("Invalid preallocation mode '%s'", optarg);
3607                 return 1;
3608             }
3609             break;
3610         case OPTION_SHRINK:
3611             shrink = true;
3612             break;
3613         }
3614     }
3615     if (optind != argc - 1) {
3616         error_exit("Expecting image file name and size");
3617     }
3618     filename = argv[optind++];
3619 
3620     if (qemu_opts_foreach(&qemu_object_opts,
3621                           user_creatable_add_opts_foreach,
3622                           NULL, NULL)) {
3623         return 1;
3624     }
3625 
3626     /* Choose grow, shrink, or absolute resize mode */
3627     switch (size[0]) {
3628     case '+':
3629         relative = 1;
3630         size++;
3631         break;
3632     case '-':
3633         relative = -1;
3634         size++;
3635         break;
3636     default:
3637         relative = 0;
3638         break;
3639     }
3640 
3641     /* Parse size */
3642     param = qemu_opts_create(&resize_options, NULL, 0, &error_abort);
3643     qemu_opt_set(param, BLOCK_OPT_SIZE, size, &err);
3644     if (err) {
3645         error_report_err(err);
3646         ret = -1;
3647         qemu_opts_del(param);
3648         goto out;
3649     }
3650     n = qemu_opt_get_size(param, BLOCK_OPT_SIZE, 0);
3651     qemu_opts_del(param);
3652 
3653     blk = img_open(image_opts, filename, fmt,
3654                    BDRV_O_RDWR | BDRV_O_RESIZE, false, quiet,
3655                    false);
3656     if (!blk) {
3657         ret = -1;
3658         goto out;
3659     }
3660 
3661     current_size = blk_getlength(blk);
3662     if (current_size < 0) {
3663         error_report("Failed to inquire current image length: %s",
3664                      strerror(-current_size));
3665         ret = -1;
3666         goto out;
3667     }
3668 
3669     if (relative) {
3670         total_size = current_size + n * relative;
3671     } else {
3672         total_size = n;
3673     }
3674     if (total_size <= 0) {
3675         error_report("New image size must be positive");
3676         ret = -1;
3677         goto out;
3678     }
3679 
3680     if (total_size <= current_size && prealloc != PREALLOC_MODE_OFF) {
3681         error_report("Preallocation can only be used for growing images");
3682         ret = -1;
3683         goto out;
3684     }
3685 
3686     if (total_size < current_size && !shrink) {
3687         warn_report("Shrinking an image will delete all data beyond the "
3688                     "shrunken image's end. Before performing such an "
3689                     "operation, make sure there is no important data there.");
3690 
3691         if (g_strcmp0(bdrv_get_format_name(blk_bs(blk)), "raw") != 0) {
3692             error_report(
3693               "Use the --shrink option to perform a shrink operation.");
3694             ret = -1;
3695             goto out;
3696         } else {
3697             warn_report("Using the --shrink option will suppress this message. "
3698                         "Note that future versions of qemu-img may refuse to "
3699                         "shrink images without this option.");
3700         }
3701     }
3702 
3703     ret = blk_truncate(blk, total_size, prealloc, &err);
3704     if (ret < 0) {
3705         error_report_err(err);
3706         goto out;
3707     }
3708 
3709     new_size = blk_getlength(blk);
3710     if (new_size < 0) {
3711         error_report("Failed to verify truncated image length: %s",
3712                      strerror(-new_size));
3713         ret = -1;
3714         goto out;
3715     }
3716 
3717     /* Some block drivers implement a truncation method, but only so
3718      * the user can cause qemu to refresh the image's size from disk.
3719      * The idea is that the user resizes the image outside of qemu and
3720      * then invokes block_resize to inform qemu about it.
3721      * (This includes iscsi and file-posix for device files.)
3722      * Of course, that is not the behavior someone invoking
3723      * qemu-img resize would find useful, so we catch that behavior
3724      * here and tell the user. */
3725     if (new_size != total_size && new_size == current_size) {
3726         error_report("Image was not resized; resizing may not be supported "
3727                      "for this image");
3728         ret = -1;
3729         goto out;
3730     }
3731 
3732     if (new_size != total_size) {
3733         warn_report("Image should have been resized to %" PRIi64
3734                     " bytes, but was resized to %" PRIi64 " bytes",
3735                     total_size, new_size);
3736     }
3737 
3738     qprintf(quiet, "Image resized.\n");
3739 
3740 out:
3741     blk_unref(blk);
3742     if (ret) {
3743         return 1;
3744     }
3745     return 0;
3746 }
3747 
3748 static void amend_status_cb(BlockDriverState *bs,
3749                             int64_t offset, int64_t total_work_size,
3750                             void *opaque)
3751 {
3752     qemu_progress_print(100.f * offset / total_work_size, 0);
3753 }
3754 
3755 static int print_amend_option_help(const char *format)
3756 {
3757     BlockDriver *drv;
3758 
3759     /* Find driver and parse its options */
3760     drv = bdrv_find_format(format);
3761     if (!drv) {
3762         error_report("Unknown file format '%s'", format);
3763         return 1;
3764     }
3765 
3766     if (!drv->bdrv_amend_options) {
3767         error_report("Format driver '%s' does not support option amendment",
3768                      format);
3769         return 1;
3770     }
3771 
3772     /* Every driver supporting amendment must have create_opts */
3773     assert(drv->create_opts);
3774 
3775     printf("Creation options for '%s':\n", format);
3776     qemu_opts_print_help(drv->create_opts);
3777     printf("\nNote that not all of these options may be amendable.\n");
3778     return 0;
3779 }
3780 
3781 static int img_amend(int argc, char **argv)
3782 {
3783     Error *err = NULL;
3784     int c, ret = 0;
3785     char *options = NULL;
3786     QemuOptsList *create_opts = NULL;
3787     QemuOpts *opts = NULL;
3788     const char *fmt = NULL, *filename, *cache;
3789     int flags;
3790     bool writethrough;
3791     bool quiet = false, progress = false;
3792     BlockBackend *blk = NULL;
3793     BlockDriverState *bs = NULL;
3794     bool image_opts = false;
3795 
3796     cache = BDRV_DEFAULT_CACHE;
3797     for (;;) {
3798         static const struct option long_options[] = {
3799             {"help", no_argument, 0, 'h'},
3800             {"object", required_argument, 0, OPTION_OBJECT},
3801             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
3802             {0, 0, 0, 0}
3803         };
3804         c = getopt_long(argc, argv, ":ho:f:t:pq",
3805                         long_options, NULL);
3806         if (c == -1) {
3807             break;
3808         }
3809 
3810         switch (c) {
3811         case ':':
3812             missing_argument(argv[optind - 1]);
3813             break;
3814         case '?':
3815             unrecognized_option(argv[optind - 1]);
3816             break;
3817         case 'h':
3818             help();
3819             break;
3820         case 'o':
3821             if (!is_valid_option_list(optarg)) {
3822                 error_report("Invalid option list: %s", optarg);
3823                 ret = -1;
3824                 goto out_no_progress;
3825             }
3826             if (!options) {
3827                 options = g_strdup(optarg);
3828             } else {
3829                 char *old_options = options;
3830                 options = g_strdup_printf("%s,%s", options, optarg);
3831                 g_free(old_options);
3832             }
3833             break;
3834         case 'f':
3835             fmt = optarg;
3836             break;
3837         case 't':
3838             cache = optarg;
3839             break;
3840         case 'p':
3841             progress = true;
3842             break;
3843         case 'q':
3844             quiet = true;
3845             break;
3846         case OPTION_OBJECT:
3847             opts = qemu_opts_parse_noisily(&qemu_object_opts,
3848                                            optarg, true);
3849             if (!opts) {
3850                 ret = -1;
3851                 goto out_no_progress;
3852             }
3853             break;
3854         case OPTION_IMAGE_OPTS:
3855             image_opts = true;
3856             break;
3857         }
3858     }
3859 
3860     if (!options) {
3861         error_exit("Must specify options (-o)");
3862     }
3863 
3864     if (qemu_opts_foreach(&qemu_object_opts,
3865                           user_creatable_add_opts_foreach,
3866                           NULL, NULL)) {
3867         ret = -1;
3868         goto out_no_progress;
3869     }
3870 
3871     if (quiet) {
3872         progress = false;
3873     }
3874     qemu_progress_init(progress, 1.0);
3875 
3876     filename = (optind == argc - 1) ? argv[argc - 1] : NULL;
3877     if (fmt && has_help_option(options)) {
3878         /* If a format is explicitly specified (and possibly no filename is
3879          * given), print option help here */
3880         ret = print_amend_option_help(fmt);
3881         goto out;
3882     }
3883 
3884     if (optind != argc - 1) {
3885         error_report("Expecting one image file name");
3886         ret = -1;
3887         goto out;
3888     }
3889 
3890     flags = BDRV_O_RDWR;
3891     ret = bdrv_parse_cache_mode(cache, &flags, &writethrough);
3892     if (ret < 0) {
3893         error_report("Invalid cache option: %s", cache);
3894         goto out;
3895     }
3896 
3897     blk = img_open(image_opts, filename, fmt, flags, writethrough, quiet,
3898                    false);
3899     if (!blk) {
3900         ret = -1;
3901         goto out;
3902     }
3903     bs = blk_bs(blk);
3904 
3905     fmt = bs->drv->format_name;
3906 
3907     if (has_help_option(options)) {
3908         /* If the format was auto-detected, print option help here */
3909         ret = print_amend_option_help(fmt);
3910         goto out;
3911     }
3912 
3913     if (!bs->drv->bdrv_amend_options) {
3914         error_report("Format driver '%s' does not support option amendment",
3915                      fmt);
3916         ret = -1;
3917         goto out;
3918     }
3919 
3920     /* Every driver supporting amendment must have create_opts */
3921     assert(bs->drv->create_opts);
3922 
3923     create_opts = qemu_opts_append(create_opts, bs->drv->create_opts);
3924     opts = qemu_opts_create(create_opts, NULL, 0, &error_abort);
3925     qemu_opts_do_parse(opts, options, NULL, &err);
3926     if (err) {
3927         error_report_err(err);
3928         ret = -1;
3929         goto out;
3930     }
3931 
3932     /* In case the driver does not call amend_status_cb() */
3933     qemu_progress_print(0.f, 0);
3934     ret = bdrv_amend_options(bs, opts, &amend_status_cb, NULL, &err);
3935     qemu_progress_print(100.f, 0);
3936     if (ret < 0) {
3937         error_report_err(err);
3938         goto out;
3939     }
3940 
3941 out:
3942     qemu_progress_end();
3943 
3944 out_no_progress:
3945     blk_unref(blk);
3946     qemu_opts_del(opts);
3947     qemu_opts_free(create_opts);
3948     g_free(options);
3949 
3950     if (ret) {
3951         return 1;
3952     }
3953     return 0;
3954 }
3955 
3956 typedef struct BenchData {
3957     BlockBackend *blk;
3958     uint64_t image_size;
3959     bool write;
3960     int bufsize;
3961     int step;
3962     int nrreq;
3963     int n;
3964     int flush_interval;
3965     bool drain_on_flush;
3966     uint8_t *buf;
3967     QEMUIOVector *qiov;
3968 
3969     int in_flight;
3970     bool in_flush;
3971     uint64_t offset;
3972 } BenchData;
3973 
3974 static void bench_undrained_flush_cb(void *opaque, int ret)
3975 {
3976     if (ret < 0) {
3977         error_report("Failed flush request: %s", strerror(-ret));
3978         exit(EXIT_FAILURE);
3979     }
3980 }
3981 
3982 static void bench_cb(void *opaque, int ret)
3983 {
3984     BenchData *b = opaque;
3985     BlockAIOCB *acb;
3986 
3987     if (ret < 0) {
3988         error_report("Failed request: %s", strerror(-ret));
3989         exit(EXIT_FAILURE);
3990     }
3991 
3992     if (b->in_flush) {
3993         /* Just finished a flush with drained queue: Start next requests */
3994         assert(b->in_flight == 0);
3995         b->in_flush = false;
3996     } else if (b->in_flight > 0) {
3997         int remaining = b->n - b->in_flight;
3998 
3999         b->n--;
4000         b->in_flight--;
4001 
4002         /* Time for flush? Drain queue if requested, then flush */
4003         if (b->flush_interval && remaining % b->flush_interval == 0) {
4004             if (!b->in_flight || !b->drain_on_flush) {
4005                 BlockCompletionFunc *cb;
4006 
4007                 if (b->drain_on_flush) {
4008                     b->in_flush = true;
4009                     cb = bench_cb;
4010                 } else {
4011                     cb = bench_undrained_flush_cb;
4012                 }
4013 
4014                 acb = blk_aio_flush(b->blk, cb, b);
4015                 if (!acb) {
4016                     error_report("Failed to issue flush request");
4017                     exit(EXIT_FAILURE);
4018                 }
4019             }
4020             if (b->drain_on_flush) {
4021                 return;
4022             }
4023         }
4024     }
4025 
4026     while (b->n > b->in_flight && b->in_flight < b->nrreq) {
4027         int64_t offset = b->offset;
4028         /* blk_aio_* might look for completed I/Os and kick bench_cb
4029          * again, so make sure this operation is counted by in_flight
4030          * and b->offset is ready for the next submission.
4031          */
4032         b->in_flight++;
4033         b->offset += b->step;
4034         b->offset %= b->image_size;
4035         if (b->write) {
4036             acb = blk_aio_pwritev(b->blk, offset, b->qiov, 0, bench_cb, b);
4037         } else {
4038             acb = blk_aio_preadv(b->blk, offset, b->qiov, 0, bench_cb, b);
4039         }
4040         if (!acb) {
4041             error_report("Failed to issue request");
4042             exit(EXIT_FAILURE);
4043         }
4044     }
4045 }
4046 
4047 static int img_bench(int argc, char **argv)
4048 {
4049     int c, ret = 0;
4050     const char *fmt = NULL, *filename;
4051     bool quiet = false;
4052     bool image_opts = false;
4053     bool is_write = false;
4054     int count = 75000;
4055     int depth = 64;
4056     int64_t offset = 0;
4057     size_t bufsize = 4096;
4058     int pattern = 0;
4059     size_t step = 0;
4060     int flush_interval = 0;
4061     bool drain_on_flush = true;
4062     int64_t image_size;
4063     BlockBackend *blk = NULL;
4064     BenchData data = {};
4065     int flags = 0;
4066     bool writethrough = false;
4067     struct timeval t1, t2;
4068     int i;
4069     bool force_share = false;
4070     size_t buf_size;
4071 
4072     for (;;) {
4073         static const struct option long_options[] = {
4074             {"help", no_argument, 0, 'h'},
4075             {"flush-interval", required_argument, 0, OPTION_FLUSH_INTERVAL},
4076             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
4077             {"pattern", required_argument, 0, OPTION_PATTERN},
4078             {"no-drain", no_argument, 0, OPTION_NO_DRAIN},
4079             {"force-share", no_argument, 0, 'U'},
4080             {0, 0, 0, 0}
4081         };
4082         c = getopt_long(argc, argv, ":hc:d:f:no:qs:S:t:wU", long_options, NULL);
4083         if (c == -1) {
4084             break;
4085         }
4086 
4087         switch (c) {
4088         case ':':
4089             missing_argument(argv[optind - 1]);
4090             break;
4091         case '?':
4092             unrecognized_option(argv[optind - 1]);
4093             break;
4094         case 'h':
4095             help();
4096             break;
4097         case 'c':
4098         {
4099             unsigned long res;
4100 
4101             if (qemu_strtoul(optarg, NULL, 0, &res) < 0 || res > INT_MAX) {
4102                 error_report("Invalid request count specified");
4103                 return 1;
4104             }
4105             count = res;
4106             break;
4107         }
4108         case 'd':
4109         {
4110             unsigned long res;
4111 
4112             if (qemu_strtoul(optarg, NULL, 0, &res) < 0 || res > INT_MAX) {
4113                 error_report("Invalid queue depth specified");
4114                 return 1;
4115             }
4116             depth = res;
4117             break;
4118         }
4119         case 'f':
4120             fmt = optarg;
4121             break;
4122         case 'n':
4123             flags |= BDRV_O_NATIVE_AIO;
4124             break;
4125         case 'o':
4126         {
4127             offset = cvtnum(optarg);
4128             if (offset < 0) {
4129                 error_report("Invalid offset specified");
4130                 return 1;
4131             }
4132             break;
4133         }
4134             break;
4135         case 'q':
4136             quiet = true;
4137             break;
4138         case 's':
4139         {
4140             int64_t sval;
4141 
4142             sval = cvtnum(optarg);
4143             if (sval < 0 || sval > INT_MAX) {
4144                 error_report("Invalid buffer size specified");
4145                 return 1;
4146             }
4147 
4148             bufsize = sval;
4149             break;
4150         }
4151         case 'S':
4152         {
4153             int64_t sval;
4154 
4155             sval = cvtnum(optarg);
4156             if (sval < 0 || sval > INT_MAX) {
4157                 error_report("Invalid step size specified");
4158                 return 1;
4159             }
4160 
4161             step = sval;
4162             break;
4163         }
4164         case 't':
4165             ret = bdrv_parse_cache_mode(optarg, &flags, &writethrough);
4166             if (ret < 0) {
4167                 error_report("Invalid cache mode");
4168                 ret = -1;
4169                 goto out;
4170             }
4171             break;
4172         case 'w':
4173             flags |= BDRV_O_RDWR;
4174             is_write = true;
4175             break;
4176         case 'U':
4177             force_share = true;
4178             break;
4179         case OPTION_PATTERN:
4180         {
4181             unsigned long res;
4182 
4183             if (qemu_strtoul(optarg, NULL, 0, &res) < 0 || res > 0xff) {
4184                 error_report("Invalid pattern byte specified");
4185                 return 1;
4186             }
4187             pattern = res;
4188             break;
4189         }
4190         case OPTION_FLUSH_INTERVAL:
4191         {
4192             unsigned long res;
4193 
4194             if (qemu_strtoul(optarg, NULL, 0, &res) < 0 || res > INT_MAX) {
4195                 error_report("Invalid flush interval specified");
4196                 return 1;
4197             }
4198             flush_interval = res;
4199             break;
4200         }
4201         case OPTION_NO_DRAIN:
4202             drain_on_flush = false;
4203             break;
4204         case OPTION_IMAGE_OPTS:
4205             image_opts = true;
4206             break;
4207         }
4208     }
4209 
4210     if (optind != argc - 1) {
4211         error_exit("Expecting one image file name");
4212     }
4213     filename = argv[argc - 1];
4214 
4215     if (!is_write && flush_interval) {
4216         error_report("--flush-interval is only available in write tests");
4217         ret = -1;
4218         goto out;
4219     }
4220     if (flush_interval && flush_interval < depth) {
4221         error_report("Flush interval can't be smaller than depth");
4222         ret = -1;
4223         goto out;
4224     }
4225 
4226     blk = img_open(image_opts, filename, fmt, flags, writethrough, quiet,
4227                    force_share);
4228     if (!blk) {
4229         ret = -1;
4230         goto out;
4231     }
4232 
4233     image_size = blk_getlength(blk);
4234     if (image_size < 0) {
4235         ret = image_size;
4236         goto out;
4237     }
4238 
4239     data = (BenchData) {
4240         .blk            = blk,
4241         .image_size     = image_size,
4242         .bufsize        = bufsize,
4243         .step           = step ?: bufsize,
4244         .nrreq          = depth,
4245         .n              = count,
4246         .offset         = offset,
4247         .write          = is_write,
4248         .flush_interval = flush_interval,
4249         .drain_on_flush = drain_on_flush,
4250     };
4251     printf("Sending %d %s requests, %d bytes each, %d in parallel "
4252            "(starting at offset %" PRId64 ", step size %d)\n",
4253            data.n, data.write ? "write" : "read", data.bufsize, data.nrreq,
4254            data.offset, data.step);
4255     if (flush_interval) {
4256         printf("Sending flush every %d requests\n", flush_interval);
4257     }
4258 
4259     buf_size = data.nrreq * data.bufsize;
4260     data.buf = blk_blockalign(blk, buf_size);
4261     memset(data.buf, pattern, data.nrreq * data.bufsize);
4262 
4263     blk_register_buf(blk, data.buf, buf_size);
4264 
4265     data.qiov = g_new(QEMUIOVector, data.nrreq);
4266     for (i = 0; i < data.nrreq; i++) {
4267         qemu_iovec_init(&data.qiov[i], 1);
4268         qemu_iovec_add(&data.qiov[i],
4269                        data.buf + i * data.bufsize, data.bufsize);
4270     }
4271 
4272     gettimeofday(&t1, NULL);
4273     bench_cb(&data, 0);
4274 
4275     while (data.n > 0) {
4276         main_loop_wait(false);
4277     }
4278     gettimeofday(&t2, NULL);
4279 
4280     printf("Run completed in %3.3f seconds.\n",
4281            (t2.tv_sec - t1.tv_sec)
4282            + ((double)(t2.tv_usec - t1.tv_usec) / 1000000));
4283 
4284 out:
4285     if (data.buf) {
4286         blk_unregister_buf(blk, data.buf);
4287     }
4288     qemu_vfree(data.buf);
4289     blk_unref(blk);
4290 
4291     if (ret) {
4292         return 1;
4293     }
4294     return 0;
4295 }
4296 
4297 #define C_BS      01
4298 #define C_COUNT   02
4299 #define C_IF      04
4300 #define C_OF      010
4301 #define C_SKIP    020
4302 
4303 struct DdInfo {
4304     unsigned int flags;
4305     int64_t count;
4306 };
4307 
4308 struct DdIo {
4309     int bsz;    /* Block size */
4310     char *filename;
4311     uint8_t *buf;
4312     int64_t offset;
4313 };
4314 
4315 struct DdOpts {
4316     const char *name;
4317     int (*f)(const char *, struct DdIo *, struct DdIo *, struct DdInfo *);
4318     unsigned int flag;
4319 };
4320 
4321 static int img_dd_bs(const char *arg,
4322                      struct DdIo *in, struct DdIo *out,
4323                      struct DdInfo *dd)
4324 {
4325     int64_t res;
4326 
4327     res = cvtnum(arg);
4328 
4329     if (res <= 0 || res > INT_MAX) {
4330         error_report("invalid number: '%s'", arg);
4331         return 1;
4332     }
4333     in->bsz = out->bsz = res;
4334 
4335     return 0;
4336 }
4337 
4338 static int img_dd_count(const char *arg,
4339                         struct DdIo *in, struct DdIo *out,
4340                         struct DdInfo *dd)
4341 {
4342     dd->count = cvtnum(arg);
4343 
4344     if (dd->count < 0) {
4345         error_report("invalid number: '%s'", arg);
4346         return 1;
4347     }
4348 
4349     return 0;
4350 }
4351 
4352 static int img_dd_if(const char *arg,
4353                      struct DdIo *in, struct DdIo *out,
4354                      struct DdInfo *dd)
4355 {
4356     in->filename = g_strdup(arg);
4357 
4358     return 0;
4359 }
4360 
4361 static int img_dd_of(const char *arg,
4362                      struct DdIo *in, struct DdIo *out,
4363                      struct DdInfo *dd)
4364 {
4365     out->filename = g_strdup(arg);
4366 
4367     return 0;
4368 }
4369 
4370 static int img_dd_skip(const char *arg,
4371                        struct DdIo *in, struct DdIo *out,
4372                        struct DdInfo *dd)
4373 {
4374     in->offset = cvtnum(arg);
4375 
4376     if (in->offset < 0) {
4377         error_report("invalid number: '%s'", arg);
4378         return 1;
4379     }
4380 
4381     return 0;
4382 }
4383 
4384 static int img_dd(int argc, char **argv)
4385 {
4386     int ret = 0;
4387     char *arg = NULL;
4388     char *tmp;
4389     BlockDriver *drv = NULL, *proto_drv = NULL;
4390     BlockBackend *blk1 = NULL, *blk2 = NULL;
4391     QemuOpts *opts = NULL;
4392     QemuOptsList *create_opts = NULL;
4393     Error *local_err = NULL;
4394     bool image_opts = false;
4395     int c, i;
4396     const char *out_fmt = "raw";
4397     const char *fmt = NULL;
4398     int64_t size = 0;
4399     int64_t block_count = 0, out_pos, in_pos;
4400     bool force_share = false;
4401     struct DdInfo dd = {
4402         .flags = 0,
4403         .count = 0,
4404     };
4405     struct DdIo in = {
4406         .bsz = 512, /* Block size is by default 512 bytes */
4407         .filename = NULL,
4408         .buf = NULL,
4409         .offset = 0
4410     };
4411     struct DdIo out = {
4412         .bsz = 512,
4413         .filename = NULL,
4414         .buf = NULL,
4415         .offset = 0
4416     };
4417 
4418     const struct DdOpts options[] = {
4419         { "bs", img_dd_bs, C_BS },
4420         { "count", img_dd_count, C_COUNT },
4421         { "if", img_dd_if, C_IF },
4422         { "of", img_dd_of, C_OF },
4423         { "skip", img_dd_skip, C_SKIP },
4424         { NULL, NULL, 0 }
4425     };
4426     const struct option long_options[] = {
4427         { "help", no_argument, 0, 'h'},
4428         { "object", required_argument, 0, OPTION_OBJECT},
4429         { "image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
4430         { "force-share", no_argument, 0, 'U'},
4431         { 0, 0, 0, 0 }
4432     };
4433 
4434     while ((c = getopt_long(argc, argv, ":hf:O:U", long_options, NULL))) {
4435         if (c == EOF) {
4436             break;
4437         }
4438         switch (c) {
4439         case 'O':
4440             out_fmt = optarg;
4441             break;
4442         case 'f':
4443             fmt = optarg;
4444             break;
4445         case ':':
4446             missing_argument(argv[optind - 1]);
4447             break;
4448         case '?':
4449             unrecognized_option(argv[optind - 1]);
4450             break;
4451         case 'h':
4452             help();
4453             break;
4454         case 'U':
4455             force_share = true;
4456             break;
4457         case OPTION_OBJECT:
4458             if (!qemu_opts_parse_noisily(&qemu_object_opts, optarg, true)) {
4459                 ret = -1;
4460                 goto out;
4461             }
4462             break;
4463         case OPTION_IMAGE_OPTS:
4464             image_opts = true;
4465             break;
4466         }
4467     }
4468 
4469     for (i = optind; i < argc; i++) {
4470         int j;
4471         arg = g_strdup(argv[i]);
4472 
4473         tmp = strchr(arg, '=');
4474         if (tmp == NULL) {
4475             error_report("unrecognized operand %s", arg);
4476             ret = -1;
4477             goto out;
4478         }
4479 
4480         *tmp++ = '\0';
4481 
4482         for (j = 0; options[j].name != NULL; j++) {
4483             if (!strcmp(arg, options[j].name)) {
4484                 break;
4485             }
4486         }
4487         if (options[j].name == NULL) {
4488             error_report("unrecognized operand %s", arg);
4489             ret = -1;
4490             goto out;
4491         }
4492 
4493         if (options[j].f(tmp, &in, &out, &dd) != 0) {
4494             ret = -1;
4495             goto out;
4496         }
4497         dd.flags |= options[j].flag;
4498         g_free(arg);
4499         arg = NULL;
4500     }
4501 
4502     if (!(dd.flags & C_IF && dd.flags & C_OF)) {
4503         error_report("Must specify both input and output files");
4504         ret = -1;
4505         goto out;
4506     }
4507 
4508     if (qemu_opts_foreach(&qemu_object_opts,
4509                           user_creatable_add_opts_foreach,
4510                           NULL, NULL)) {
4511         ret = -1;
4512         goto out;
4513     }
4514 
4515     blk1 = img_open(image_opts, in.filename, fmt, 0, false, false,
4516                     force_share);
4517 
4518     if (!blk1) {
4519         ret = -1;
4520         goto out;
4521     }
4522 
4523     drv = bdrv_find_format(out_fmt);
4524     if (!drv) {
4525         error_report("Unknown file format");
4526         ret = -1;
4527         goto out;
4528     }
4529     proto_drv = bdrv_find_protocol(out.filename, true, &local_err);
4530 
4531     if (!proto_drv) {
4532         error_report_err(local_err);
4533         ret = -1;
4534         goto out;
4535     }
4536     if (!drv->create_opts) {
4537         error_report("Format driver '%s' does not support image creation",
4538                      drv->format_name);
4539         ret = -1;
4540         goto out;
4541     }
4542     if (!proto_drv->create_opts) {
4543         error_report("Protocol driver '%s' does not support image creation",
4544                      proto_drv->format_name);
4545         ret = -1;
4546         goto out;
4547     }
4548     create_opts = qemu_opts_append(create_opts, drv->create_opts);
4549     create_opts = qemu_opts_append(create_opts, proto_drv->create_opts);
4550 
4551     opts = qemu_opts_create(create_opts, NULL, 0, &error_abort);
4552 
4553     size = blk_getlength(blk1);
4554     if (size < 0) {
4555         error_report("Failed to get size for '%s'", in.filename);
4556         ret = -1;
4557         goto out;
4558     }
4559 
4560     if (dd.flags & C_COUNT && dd.count <= INT64_MAX / in.bsz &&
4561         dd.count * in.bsz < size) {
4562         size = dd.count * in.bsz;
4563     }
4564 
4565     /* Overflow means the specified offset is beyond input image's size */
4566     if (dd.flags & C_SKIP && (in.offset > INT64_MAX / in.bsz ||
4567                               size < in.bsz * in.offset)) {
4568         qemu_opt_set_number(opts, BLOCK_OPT_SIZE, 0, &error_abort);
4569     } else {
4570         qemu_opt_set_number(opts, BLOCK_OPT_SIZE,
4571                             size - in.bsz * in.offset, &error_abort);
4572     }
4573 
4574     ret = bdrv_create(drv, out.filename, opts, &local_err);
4575     if (ret < 0) {
4576         error_reportf_err(local_err,
4577                           "%s: error while creating output image: ",
4578                           out.filename);
4579         ret = -1;
4580         goto out;
4581     }
4582 
4583     /* TODO, we can't honour --image-opts for the target,
4584      * since it needs to be given in a format compatible
4585      * with the bdrv_create() call above which does not
4586      * support image-opts style.
4587      */
4588     blk2 = img_open_file(out.filename, NULL, out_fmt, BDRV_O_RDWR,
4589                          false, false, false);
4590 
4591     if (!blk2) {
4592         ret = -1;
4593         goto out;
4594     }
4595 
4596     if (dd.flags & C_SKIP && (in.offset > INT64_MAX / in.bsz ||
4597                               size < in.offset * in.bsz)) {
4598         /* We give a warning if the skip option is bigger than the input
4599          * size and create an empty output disk image (i.e. like dd(1)).
4600          */
4601         error_report("%s: cannot skip to specified offset", in.filename);
4602         in_pos = size;
4603     } else {
4604         in_pos = in.offset * in.bsz;
4605     }
4606 
4607     in.buf = g_new(uint8_t, in.bsz);
4608 
4609     for (out_pos = 0; in_pos < size; block_count++) {
4610         int in_ret, out_ret;
4611 
4612         if (in_pos + in.bsz > size) {
4613             in_ret = blk_pread(blk1, in_pos, in.buf, size - in_pos);
4614         } else {
4615             in_ret = blk_pread(blk1, in_pos, in.buf, in.bsz);
4616         }
4617         if (in_ret < 0) {
4618             error_report("error while reading from input image file: %s",
4619                          strerror(-in_ret));
4620             ret = -1;
4621             goto out;
4622         }
4623         in_pos += in_ret;
4624 
4625         out_ret = blk_pwrite(blk2, out_pos, in.buf, in_ret, 0);
4626 
4627         if (out_ret < 0) {
4628             error_report("error while writing to output image file: %s",
4629                          strerror(-out_ret));
4630             ret = -1;
4631             goto out;
4632         }
4633         out_pos += out_ret;
4634     }
4635 
4636 out:
4637     g_free(arg);
4638     qemu_opts_del(opts);
4639     qemu_opts_free(create_opts);
4640     blk_unref(blk1);
4641     blk_unref(blk2);
4642     g_free(in.filename);
4643     g_free(out.filename);
4644     g_free(in.buf);
4645     g_free(out.buf);
4646 
4647     if (ret) {
4648         return 1;
4649     }
4650     return 0;
4651 }
4652 
4653 static void dump_json_block_measure_info(BlockMeasureInfo *info)
4654 {
4655     QString *str;
4656     QObject *obj;
4657     Visitor *v = qobject_output_visitor_new(&obj);
4658 
4659     visit_type_BlockMeasureInfo(v, NULL, &info, &error_abort);
4660     visit_complete(v, &obj);
4661     str = qobject_to_json_pretty(obj);
4662     assert(str != NULL);
4663     printf("%s\n", qstring_get_str(str));
4664     qobject_unref(obj);
4665     visit_free(v);
4666     qobject_unref(str);
4667 }
4668 
4669 static int img_measure(int argc, char **argv)
4670 {
4671     static const struct option long_options[] = {
4672         {"help", no_argument, 0, 'h'},
4673         {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
4674         {"object", required_argument, 0, OPTION_OBJECT},
4675         {"output", required_argument, 0, OPTION_OUTPUT},
4676         {"size", required_argument, 0, OPTION_SIZE},
4677         {"force-share", no_argument, 0, 'U'},
4678         {0, 0, 0, 0}
4679     };
4680     OutputFormat output_format = OFORMAT_HUMAN;
4681     BlockBackend *in_blk = NULL;
4682     BlockDriver *drv;
4683     const char *filename = NULL;
4684     const char *fmt = NULL;
4685     const char *out_fmt = "raw";
4686     char *options = NULL;
4687     char *snapshot_name = NULL;
4688     bool force_share = false;
4689     QemuOpts *opts = NULL;
4690     QemuOpts *object_opts = NULL;
4691     QemuOpts *sn_opts = NULL;
4692     QemuOptsList *create_opts = NULL;
4693     bool image_opts = false;
4694     uint64_t img_size = UINT64_MAX;
4695     BlockMeasureInfo *info = NULL;
4696     Error *local_err = NULL;
4697     int ret = 1;
4698     int c;
4699 
4700     while ((c = getopt_long(argc, argv, "hf:O:o:l:U",
4701                             long_options, NULL)) != -1) {
4702         switch (c) {
4703         case '?':
4704         case 'h':
4705             help();
4706             break;
4707         case 'f':
4708             fmt = optarg;
4709             break;
4710         case 'O':
4711             out_fmt = optarg;
4712             break;
4713         case 'o':
4714             if (!is_valid_option_list(optarg)) {
4715                 error_report("Invalid option list: %s", optarg);
4716                 goto out;
4717             }
4718             if (!options) {
4719                 options = g_strdup(optarg);
4720             } else {
4721                 char *old_options = options;
4722                 options = g_strdup_printf("%s,%s", options, optarg);
4723                 g_free(old_options);
4724             }
4725             break;
4726         case 'l':
4727             if (strstart(optarg, SNAPSHOT_OPT_BASE, NULL)) {
4728                 sn_opts = qemu_opts_parse_noisily(&internal_snapshot_opts,
4729                                                   optarg, false);
4730                 if (!sn_opts) {
4731                     error_report("Failed in parsing snapshot param '%s'",
4732                                  optarg);
4733                     goto out;
4734                 }
4735             } else {
4736                 snapshot_name = optarg;
4737             }
4738             break;
4739         case 'U':
4740             force_share = true;
4741             break;
4742         case OPTION_OBJECT:
4743             object_opts = qemu_opts_parse_noisily(&qemu_object_opts,
4744                                                   optarg, true);
4745             if (!object_opts) {
4746                 goto out;
4747             }
4748             break;
4749         case OPTION_IMAGE_OPTS:
4750             image_opts = true;
4751             break;
4752         case OPTION_OUTPUT:
4753             if (!strcmp(optarg, "json")) {
4754                 output_format = OFORMAT_JSON;
4755             } else if (!strcmp(optarg, "human")) {
4756                 output_format = OFORMAT_HUMAN;
4757             } else {
4758                 error_report("--output must be used with human or json "
4759                              "as argument.");
4760                 goto out;
4761             }
4762             break;
4763         case OPTION_SIZE:
4764         {
4765             int64_t sval;
4766 
4767             sval = cvtnum(optarg);
4768             if (sval < 0) {
4769                 if (sval == -ERANGE) {
4770                     error_report("Image size must be less than 8 EiB!");
4771                 } else {
4772                     error_report("Invalid image size specified! You may use "
4773                                  "k, M, G, T, P or E suffixes for ");
4774                     error_report("kilobytes, megabytes, gigabytes, terabytes, "
4775                                  "petabytes and exabytes.");
4776                 }
4777                 goto out;
4778             }
4779             img_size = (uint64_t)sval;
4780         }
4781         break;
4782         }
4783     }
4784 
4785     if (qemu_opts_foreach(&qemu_object_opts,
4786                           user_creatable_add_opts_foreach,
4787                           NULL, NULL)) {
4788         goto out;
4789     }
4790 
4791     if (argc - optind > 1) {
4792         error_report("At most one filename argument is allowed.");
4793         goto out;
4794     } else if (argc - optind == 1) {
4795         filename = argv[optind];
4796     }
4797 
4798     if (!filename &&
4799         (object_opts || image_opts || fmt || snapshot_name || sn_opts)) {
4800         error_report("--object, --image-opts, -f, and -l "
4801                      "require a filename argument.");
4802         goto out;
4803     }
4804     if (filename && img_size != UINT64_MAX) {
4805         error_report("--size N cannot be used together with a filename.");
4806         goto out;
4807     }
4808     if (!filename && img_size == UINT64_MAX) {
4809         error_report("Either --size N or one filename must be specified.");
4810         goto out;
4811     }
4812 
4813     if (filename) {
4814         in_blk = img_open(image_opts, filename, fmt, 0,
4815                           false, false, force_share);
4816         if (!in_blk) {
4817             goto out;
4818         }
4819 
4820         if (sn_opts) {
4821             bdrv_snapshot_load_tmp(blk_bs(in_blk),
4822                     qemu_opt_get(sn_opts, SNAPSHOT_OPT_ID),
4823                     qemu_opt_get(sn_opts, SNAPSHOT_OPT_NAME),
4824                     &local_err);
4825         } else if (snapshot_name != NULL) {
4826             bdrv_snapshot_load_tmp_by_id_or_name(blk_bs(in_blk),
4827                     snapshot_name, &local_err);
4828         }
4829         if (local_err) {
4830             error_reportf_err(local_err, "Failed to load snapshot: ");
4831             goto out;
4832         }
4833     }
4834 
4835     drv = bdrv_find_format(out_fmt);
4836     if (!drv) {
4837         error_report("Unknown file format '%s'", out_fmt);
4838         goto out;
4839     }
4840     if (!drv->create_opts) {
4841         error_report("Format driver '%s' does not support image creation",
4842                      drv->format_name);
4843         goto out;
4844     }
4845 
4846     create_opts = qemu_opts_append(create_opts, drv->create_opts);
4847     create_opts = qemu_opts_append(create_opts, bdrv_file.create_opts);
4848     opts = qemu_opts_create(create_opts, NULL, 0, &error_abort);
4849     if (options) {
4850         qemu_opts_do_parse(opts, options, NULL, &local_err);
4851         if (local_err) {
4852             error_report_err(local_err);
4853             error_report("Invalid options for file format '%s'", out_fmt);
4854             goto out;
4855         }
4856     }
4857     if (img_size != UINT64_MAX) {
4858         qemu_opt_set_number(opts, BLOCK_OPT_SIZE, img_size, &error_abort);
4859     }
4860 
4861     info = bdrv_measure(drv, opts, in_blk ? blk_bs(in_blk) : NULL, &local_err);
4862     if (local_err) {
4863         error_report_err(local_err);
4864         goto out;
4865     }
4866 
4867     if (output_format == OFORMAT_HUMAN) {
4868         printf("required size: %" PRIu64 "\n", info->required);
4869         printf("fully allocated size: %" PRIu64 "\n", info->fully_allocated);
4870     } else {
4871         dump_json_block_measure_info(info);
4872     }
4873 
4874     ret = 0;
4875 
4876 out:
4877     qapi_free_BlockMeasureInfo(info);
4878     qemu_opts_del(object_opts);
4879     qemu_opts_del(opts);
4880     qemu_opts_del(sn_opts);
4881     qemu_opts_free(create_opts);
4882     g_free(options);
4883     blk_unref(in_blk);
4884     return ret;
4885 }
4886 
4887 static const img_cmd_t img_cmds[] = {
4888 #define DEF(option, callback, arg_string)        \
4889     { option, callback },
4890 #include "qemu-img-cmds.h"
4891 #undef DEF
4892     { NULL, NULL, },
4893 };
4894 
4895 int main(int argc, char **argv)
4896 {
4897     const img_cmd_t *cmd;
4898     const char *cmdname;
4899     Error *local_error = NULL;
4900     char *trace_file = NULL;
4901     int c;
4902     static const struct option long_options[] = {
4903         {"help", no_argument, 0, 'h'},
4904         {"version", no_argument, 0, 'V'},
4905         {"trace", required_argument, NULL, 'T'},
4906         {0, 0, 0, 0}
4907     };
4908 
4909 #ifdef CONFIG_POSIX
4910     signal(SIGPIPE, SIG_IGN);
4911 #endif
4912 
4913     module_call_init(MODULE_INIT_TRACE);
4914     error_set_progname(argv[0]);
4915     qemu_init_exec_dir(argv[0]);
4916 
4917     if (qemu_init_main_loop(&local_error)) {
4918         error_report_err(local_error);
4919         exit(EXIT_FAILURE);
4920     }
4921 
4922     qcrypto_init(&error_fatal);
4923 
4924     module_call_init(MODULE_INIT_QOM);
4925     bdrv_init();
4926     if (argc < 2) {
4927         error_exit("Not enough arguments");
4928     }
4929 
4930     qemu_add_opts(&qemu_object_opts);
4931     qemu_add_opts(&qemu_source_opts);
4932     qemu_add_opts(&qemu_trace_opts);
4933 
4934     while ((c = getopt_long(argc, argv, "+:hVT:", long_options, NULL)) != -1) {
4935         switch (c) {
4936         case ':':
4937             missing_argument(argv[optind - 1]);
4938             return 0;
4939         case '?':
4940             unrecognized_option(argv[optind - 1]);
4941             return 0;
4942         case 'h':
4943             help();
4944             return 0;
4945         case 'V':
4946             printf(QEMU_IMG_VERSION);
4947             return 0;
4948         case 'T':
4949             g_free(trace_file);
4950             trace_file = trace_opt_parse(optarg);
4951             break;
4952         }
4953     }
4954 
4955     cmdname = argv[optind];
4956 
4957     /* reset getopt_long scanning */
4958     argc -= optind;
4959     if (argc < 1) {
4960         return 0;
4961     }
4962     argv += optind;
4963     optind = 0;
4964 
4965     if (!trace_init_backends()) {
4966         exit(1);
4967     }
4968     trace_init_file(trace_file);
4969     qemu_set_log(LOG_TRACE);
4970 
4971     /* find the command */
4972     for (cmd = img_cmds; cmd->name != NULL; cmd++) {
4973         if (!strcmp(cmdname, cmd->name)) {
4974             return cmd->handler(argc, argv);
4975         }
4976     }
4977 
4978     /* not found */
4979     error_exit("Command not found: %s", cmdname);
4980 }
4981