xref: /qemu/qemu-img.c (revision f5956d71)
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 #include "qemu/osdep.h"
25 #include "qemu-version.h"
26 #include "qapi/error.h"
27 #include "qapi-visit.h"
28 #include "qapi/qobject-output-visitor.h"
29 #include "qapi/qmp/qerror.h"
30 #include "qapi/qmp/qjson.h"
31 #include "qapi/qmp/qbool.h"
32 #include "qemu/cutils.h"
33 #include "qemu/config-file.h"
34 #include "qemu/option.h"
35 #include "qemu/error-report.h"
36 #include "qemu/log.h"
37 #include "qom/object_interfaces.h"
38 #include "sysemu/sysemu.h"
39 #include "sysemu/block-backend.h"
40 #include "block/block_int.h"
41 #include "block/blockjob.h"
42 #include "block/qapi.h"
43 #include "crypto/init.h"
44 #include "trace/control.h"
45 #include <getopt.h>
46 
47 #define QEMU_IMG_VERSION "qemu-img version " QEMU_VERSION QEMU_PKGVERSION \
48                           "\n" QEMU_COPYRIGHT "\n"
49 
50 typedef struct img_cmd_t {
51     const char *name;
52     int (*handler)(int argc, char **argv);
53 } img_cmd_t;
54 
55 enum {
56     OPTION_OUTPUT = 256,
57     OPTION_BACKING_CHAIN = 257,
58     OPTION_OBJECT = 258,
59     OPTION_IMAGE_OPTS = 259,
60     OPTION_PATTERN = 260,
61     OPTION_FLUSH_INTERVAL = 261,
62     OPTION_NO_DRAIN = 262,
63     OPTION_TARGET_IMAGE_OPTS = 263,
64 };
65 
66 typedef enum OutputFormat {
67     OFORMAT_JSON,
68     OFORMAT_HUMAN,
69 } OutputFormat;
70 
71 /* Default to cache=writeback as data integrity is not important for qemu-img */
72 #define BDRV_DEFAULT_CACHE "writeback"
73 
74 static void format_print(void *opaque, const char *name)
75 {
76     printf(" %s", name);
77 }
78 
79 static void QEMU_NORETURN GCC_FMT_ATTR(1, 2) error_exit(const char *fmt, ...)
80 {
81     va_list ap;
82 
83     error_printf("qemu-img: ");
84 
85     va_start(ap, fmt);
86     error_vprintf(fmt, ap);
87     va_end(ap);
88 
89     error_printf("\nTry 'qemu-img --help' for more information\n");
90     exit(EXIT_FAILURE);
91 }
92 
93 static void QEMU_NORETURN missing_argument(const char *option)
94 {
95     error_exit("missing argument for option '%s'", option);
96 }
97 
98 static void QEMU_NORETURN unrecognized_option(const char *option)
99 {
100     error_exit("unrecognized option '%s'", option);
101 }
102 
103 /* Please keep in synch with qemu-img.texi */
104 static void QEMU_NORETURN help(void)
105 {
106     const char *help_msg =
107            QEMU_IMG_VERSION
108            "usage: qemu-img [standard options] command [command options]\n"
109            "QEMU disk image utility\n"
110            "\n"
111            "    '-h', '--help'       display this help and exit\n"
112            "    '-V', '--version'    output version information and exit\n"
113            "    '-T', '--trace'      [[enable=]<pattern>][,events=<file>][,file=<file>]\n"
114            "                         specify tracing options\n"
115            "\n"
116            "Command syntax:\n"
117 #define DEF(option, callback, arg_string)        \
118            "  " arg_string "\n"
119 #include "qemu-img-cmds.h"
120 #undef DEF
121 #undef GEN_DOCS
122            "\n"
123            "Command parameters:\n"
124            "  'filename' is a disk image filename\n"
125            "  'objectdef' is a QEMU user creatable object definition. See the qemu(1)\n"
126            "    manual page for a description of the object properties. The most common\n"
127            "    object type is a 'secret', which is used to supply passwords and/or\n"
128            "    encryption keys.\n"
129            "  'fmt' is the disk image format. It is guessed automatically in most cases\n"
130            "  'cache' is the cache mode used to write the output disk image, the valid\n"
131            "    options are: 'none', 'writeback' (default, except for convert), 'writethrough',\n"
132            "    'directsync' and 'unsafe' (default for convert)\n"
133            "  'src_cache' is the cache mode used to read input disk images, the valid\n"
134            "    options are the same as for the 'cache' option\n"
135            "  'size' is the disk image size in bytes. Optional suffixes\n"
136            "    'k' or 'K' (kilobyte, 1024), 'M' (megabyte, 1024k), 'G' (gigabyte, 1024M),\n"
137            "    'T' (terabyte, 1024G), 'P' (petabyte, 1024T) and 'E' (exabyte, 1024P)  are\n"
138            "    supported. 'b' is ignored.\n"
139            "  'output_filename' is the destination disk image filename\n"
140            "  'output_fmt' is the destination format\n"
141            "  'options' is a comma separated list of format specific options in a\n"
142            "    name=value format. Use -o ? for an overview of the options supported by the\n"
143            "    used format\n"
144            "  'snapshot_param' is param used for internal snapshot, format\n"
145            "    is 'snapshot.id=[ID],snapshot.name=[NAME]', or\n"
146            "    '[ID_OR_NAME]'\n"
147            "  'snapshot_id_or_name' is deprecated, use 'snapshot_param'\n"
148            "    instead\n"
149            "  '-c' indicates that target image must be compressed (qcow format only)\n"
150            "  '-u' enables unsafe rebasing. It is assumed that old and new backing file\n"
151            "       match exactly. The image doesn't need a working backing file before\n"
152            "       rebasing in this case (useful for renaming the backing file)\n"
153            "  '-h' with or without a command shows this help and lists the supported formats\n"
154            "  '-p' show progress of command (only certain commands)\n"
155            "  '-q' use Quiet mode - do not print any output (except errors)\n"
156            "  '-S' indicates the consecutive number of bytes (defaults to 4k) that must\n"
157            "       contain only zeros for qemu-img to create a sparse image during\n"
158            "       conversion. If the number of bytes is 0, the source will not be scanned for\n"
159            "       unallocated or zero sectors, and the destination image will always be\n"
160            "       fully allocated\n"
161            "  '--output' takes the format in which the output must be done (human or json)\n"
162            "  '-n' skips the target volume creation (useful if the volume is created\n"
163            "       prior to running qemu-img)\n"
164            "\n"
165            "Parameters to check subcommand:\n"
166            "  '-r' tries to repair any inconsistencies that are found during the check.\n"
167            "       '-r leaks' repairs only cluster leaks, whereas '-r all' fixes all\n"
168            "       kinds of errors, with a higher risk of choosing the wrong fix or\n"
169            "       hiding corruption that has already occurred.\n"
170            "\n"
171            "Parameters to convert subcommand:\n"
172            "  '-m' specifies how many coroutines work in parallel during the convert\n"
173            "       process (defaults to 8)\n"
174            "  '-W' allow to write to the target out of order rather than sequential\n"
175            "\n"
176            "Parameters to snapshot subcommand:\n"
177            "  'snapshot' is the name of the snapshot to create, apply or delete\n"
178            "  '-a' applies a snapshot (revert disk to saved state)\n"
179            "  '-c' creates a snapshot\n"
180            "  '-d' deletes a snapshot\n"
181            "  '-l' lists all snapshots in the given image\n"
182            "\n"
183            "Parameters to compare subcommand:\n"
184            "  '-f' first image format\n"
185            "  '-F' second image format\n"
186            "  '-s' run in Strict mode - fail on different image size or sector allocation\n"
187            "\n"
188            "Parameters to dd subcommand:\n"
189            "  'bs=BYTES' read and write up to BYTES bytes at a time "
190            "(default: 512)\n"
191            "  'count=N' copy only N input blocks\n"
192            "  'if=FILE' read from FILE\n"
193            "  'of=FILE' write to FILE\n"
194            "  'skip=N' skip N bs-sized blocks at the start of input\n";
195 
196     printf("%s\nSupported formats:", help_msg);
197     bdrv_iterate_format(format_print, NULL);
198     printf("\n");
199     exit(EXIT_SUCCESS);
200 }
201 
202 static QemuOptsList qemu_object_opts = {
203     .name = "object",
204     .implied_opt_name = "qom-type",
205     .head = QTAILQ_HEAD_INITIALIZER(qemu_object_opts.head),
206     .desc = {
207         { }
208     },
209 };
210 
211 static QemuOptsList qemu_source_opts = {
212     .name = "source",
213     .implied_opt_name = "file",
214     .head = QTAILQ_HEAD_INITIALIZER(qemu_source_opts.head),
215     .desc = {
216         { }
217     },
218 };
219 
220 static int GCC_FMT_ATTR(2, 3) qprintf(bool quiet, const char *fmt, ...)
221 {
222     int ret = 0;
223     if (!quiet) {
224         va_list args;
225         va_start(args, fmt);
226         ret = vprintf(fmt, args);
227         va_end(args);
228     }
229     return ret;
230 }
231 
232 
233 static int print_block_option_help(const char *filename, const char *fmt)
234 {
235     BlockDriver *drv, *proto_drv;
236     QemuOptsList *create_opts = NULL;
237     Error *local_err = NULL;
238 
239     /* Find driver and parse its options */
240     drv = bdrv_find_format(fmt);
241     if (!drv) {
242         error_report("Unknown file format '%s'", fmt);
243         return 1;
244     }
245 
246     create_opts = qemu_opts_append(create_opts, drv->create_opts);
247     if (filename) {
248         proto_drv = bdrv_find_protocol(filename, true, &local_err);
249         if (!proto_drv) {
250             error_report_err(local_err);
251             qemu_opts_free(create_opts);
252             return 1;
253         }
254         create_opts = qemu_opts_append(create_opts, proto_drv->create_opts);
255     }
256 
257     qemu_opts_print_help(create_opts);
258     qemu_opts_free(create_opts);
259     return 0;
260 }
261 
262 
263 static int img_open_password(BlockBackend *blk, const char *filename,
264                              int flags, bool quiet)
265 {
266     BlockDriverState *bs;
267     char password[256];
268 
269     bs = blk_bs(blk);
270     if (bdrv_is_encrypted(bs) && bdrv_key_required(bs) &&
271         !(flags & BDRV_O_NO_IO)) {
272         qprintf(quiet, "Disk image '%s' is encrypted.\n", filename);
273         if (qemu_read_password(password, sizeof(password)) < 0) {
274             error_report("No password given");
275             return -1;
276         }
277         if (bdrv_set_key(bs, password) < 0) {
278             error_report("invalid password");
279             return -1;
280         }
281     }
282     return 0;
283 }
284 
285 
286 static BlockBackend *img_open_opts(const char *optstr,
287                                    QemuOpts *opts, int flags, bool writethrough,
288                                    bool quiet, bool force_share)
289 {
290     QDict *options;
291     Error *local_err = NULL;
292     BlockBackend *blk;
293     options = qemu_opts_to_qdict(opts, NULL);
294     if (force_share) {
295         if (qdict_haskey(options, BDRV_OPT_FORCE_SHARE)
296             && !qdict_get_bool(options, BDRV_OPT_FORCE_SHARE)) {
297             error_report("--force-share/-U conflicts with image options");
298             QDECREF(options);
299             return NULL;
300         }
301         qdict_put_bool(options, BDRV_OPT_FORCE_SHARE, true);
302     }
303     blk = blk_new_open(NULL, NULL, options, flags, &local_err);
304     if (!blk) {
305         error_reportf_err(local_err, "Could not open '%s': ", optstr);
306         return NULL;
307     }
308     blk_set_enable_write_cache(blk, !writethrough);
309 
310     if (img_open_password(blk, optstr, flags, quiet) < 0) {
311         blk_unref(blk);
312         return NULL;
313     }
314     return blk;
315 }
316 
317 static BlockBackend *img_open_file(const char *filename,
318                                    QDict *options,
319                                    const char *fmt, int flags,
320                                    bool writethrough, bool quiet,
321                                    bool force_share)
322 {
323     BlockBackend *blk;
324     Error *local_err = NULL;
325 
326     if (!options) {
327         options = qdict_new();
328     }
329     if (fmt) {
330         qdict_put_str(options, "driver", fmt);
331     }
332 
333     if (force_share) {
334         qdict_put_bool(options, BDRV_OPT_FORCE_SHARE, true);
335     }
336     blk = blk_new_open(filename, NULL, options, flags, &local_err);
337     if (!blk) {
338         error_reportf_err(local_err, "Could not open '%s': ", filename);
339         return NULL;
340     }
341     blk_set_enable_write_cache(blk, !writethrough);
342 
343     if (img_open_password(blk, filename, flags, quiet) < 0) {
344         blk_unref(blk);
345         return NULL;
346     }
347     return blk;
348 }
349 
350 
351 static int img_add_key_secrets(void *opaque,
352                                const char *name, const char *value,
353                                Error **errp)
354 {
355     QDict *options = opaque;
356 
357     if (g_str_has_suffix(name, "key-secret")) {
358         qdict_put(options, name, qstring_from_str(value));
359     }
360 
361     return 0;
362 }
363 
364 static BlockBackend *img_open_new_file(const char *filename,
365                                        QemuOpts *create_opts,
366                                        const char *fmt, int flags,
367                                        bool writethrough, bool quiet,
368                                        bool force_share)
369 {
370     QDict *options = NULL;
371 
372     options = qdict_new();
373     qemu_opt_foreach(create_opts, img_add_key_secrets, options, &error_abort);
374 
375     return img_open_file(filename, options, fmt, flags, writethrough, quiet,
376                          force_share);
377 }
378 
379 
380 static BlockBackend *img_open(bool image_opts,
381                               const char *filename,
382                               const char *fmt, int flags, bool writethrough,
383                               bool quiet, bool force_share)
384 {
385     BlockBackend *blk;
386     if (image_opts) {
387         QemuOpts *opts;
388         if (fmt) {
389             error_report("--image-opts and --format are mutually exclusive");
390             return NULL;
391         }
392         opts = qemu_opts_parse_noisily(qemu_find_opts("source"),
393                                        filename, true);
394         if (!opts) {
395             return NULL;
396         }
397         blk = img_open_opts(filename, opts, flags, writethrough, quiet,
398                             force_share);
399     } else {
400         blk = img_open_file(filename, NULL, fmt, flags, writethrough, quiet,
401                             force_share);
402     }
403     return blk;
404 }
405 
406 
407 static int add_old_style_options(const char *fmt, QemuOpts *opts,
408                                  const char *base_filename,
409                                  const char *base_fmt)
410 {
411     Error *err = NULL;
412 
413     if (base_filename) {
414         qemu_opt_set(opts, BLOCK_OPT_BACKING_FILE, base_filename, &err);
415         if (err) {
416             error_report("Backing file not supported for file format '%s'",
417                          fmt);
418             error_free(err);
419             return -1;
420         }
421     }
422     if (base_fmt) {
423         qemu_opt_set(opts, BLOCK_OPT_BACKING_FMT, base_fmt, &err);
424         if (err) {
425             error_report("Backing file format not supported for file "
426                          "format '%s'", fmt);
427             error_free(err);
428             return -1;
429         }
430     }
431     return 0;
432 }
433 
434 static int64_t cvtnum(const char *s)
435 {
436     int err;
437     uint64_t value;
438 
439     err = qemu_strtosz(s, NULL, &value);
440     if (err < 0) {
441         return err;
442     }
443     if (value > INT64_MAX) {
444         return -ERANGE;
445     }
446     return value;
447 }
448 
449 static int img_create(int argc, char **argv)
450 {
451     int c;
452     uint64_t img_size = -1;
453     const char *fmt = "raw";
454     const char *base_fmt = NULL;
455     const char *filename;
456     const char *base_filename = NULL;
457     char *options = NULL;
458     Error *local_err = NULL;
459     bool quiet = false;
460 
461     for(;;) {
462         static const struct option long_options[] = {
463             {"help", no_argument, 0, 'h'},
464             {"object", required_argument, 0, OPTION_OBJECT},
465             {0, 0, 0, 0}
466         };
467         c = getopt_long(argc, argv, ":F:b:f:ho:q",
468                         long_options, NULL);
469         if (c == -1) {
470             break;
471         }
472         switch(c) {
473         case ':':
474             missing_argument(argv[optind - 1]);
475             break;
476         case '?':
477             unrecognized_option(argv[optind - 1]);
478             break;
479         case 'h':
480             help();
481             break;
482         case 'F':
483             base_fmt = optarg;
484             break;
485         case 'b':
486             base_filename = optarg;
487             break;
488         case 'f':
489             fmt = optarg;
490             break;
491         case 'o':
492             if (!is_valid_option_list(optarg)) {
493                 error_report("Invalid option list: %s", optarg);
494                 goto fail;
495             }
496             if (!options) {
497                 options = g_strdup(optarg);
498             } else {
499                 char *old_options = options;
500                 options = g_strdup_printf("%s,%s", options, optarg);
501                 g_free(old_options);
502             }
503             break;
504         case 'q':
505             quiet = true;
506             break;
507         case OPTION_OBJECT: {
508             QemuOpts *opts;
509             opts = qemu_opts_parse_noisily(&qemu_object_opts,
510                                            optarg, true);
511             if (!opts) {
512                 goto fail;
513             }
514         }   break;
515         }
516     }
517 
518     /* Get the filename */
519     filename = (optind < argc) ? argv[optind] : NULL;
520     if (options && has_help_option(options)) {
521         g_free(options);
522         return print_block_option_help(filename, fmt);
523     }
524 
525     if (optind >= argc) {
526         error_exit("Expecting image file name");
527     }
528     optind++;
529 
530     if (qemu_opts_foreach(&qemu_object_opts,
531                           user_creatable_add_opts_foreach,
532                           NULL, NULL)) {
533         goto fail;
534     }
535 
536     /* Get image size, if specified */
537     if (optind < argc) {
538         int64_t sval;
539 
540         sval = cvtnum(argv[optind++]);
541         if (sval < 0) {
542             if (sval == -ERANGE) {
543                 error_report("Image size must be less than 8 EiB!");
544             } else {
545                 error_report("Invalid image size specified! You may use k, M, "
546                       "G, T, P or E suffixes for ");
547                 error_report("kilobytes, megabytes, gigabytes, terabytes, "
548                              "petabytes and exabytes.");
549             }
550             goto fail;
551         }
552         img_size = (uint64_t)sval;
553     }
554     if (optind != argc) {
555         error_exit("Unexpected argument: %s", argv[optind]);
556     }
557 
558     bdrv_img_create(filename, fmt, base_filename, base_fmt,
559                     options, img_size, 0, quiet, &local_err);
560     if (local_err) {
561         error_reportf_err(local_err, "%s: ", filename);
562         goto fail;
563     }
564 
565     g_free(options);
566     return 0;
567 
568 fail:
569     g_free(options);
570     return 1;
571 }
572 
573 static void dump_json_image_check(ImageCheck *check, bool quiet)
574 {
575     QString *str;
576     QObject *obj;
577     Visitor *v = qobject_output_visitor_new(&obj);
578 
579     visit_type_ImageCheck(v, NULL, &check, &error_abort);
580     visit_complete(v, &obj);
581     str = qobject_to_json_pretty(obj);
582     assert(str != NULL);
583     qprintf(quiet, "%s\n", qstring_get_str(str));
584     qobject_decref(obj);
585     visit_free(v);
586     QDECREF(str);
587 }
588 
589 static void dump_human_image_check(ImageCheck *check, bool quiet)
590 {
591     if (!(check->corruptions || check->leaks || check->check_errors)) {
592         qprintf(quiet, "No errors were found on the image.\n");
593     } else {
594         if (check->corruptions) {
595             qprintf(quiet, "\n%" PRId64 " errors were found on the image.\n"
596                     "Data may be corrupted, or further writes to the image "
597                     "may corrupt it.\n",
598                     check->corruptions);
599         }
600 
601         if (check->leaks) {
602             qprintf(quiet,
603                     "\n%" PRId64 " leaked clusters were found on the image.\n"
604                     "This means waste of disk space, but no harm to data.\n",
605                     check->leaks);
606         }
607 
608         if (check->check_errors) {
609             qprintf(quiet,
610                     "\n%" PRId64
611                     " internal errors have occurred during the check.\n",
612                     check->check_errors);
613         }
614     }
615 
616     if (check->total_clusters != 0 && check->allocated_clusters != 0) {
617         qprintf(quiet, "%" PRId64 "/%" PRId64 " = %0.2f%% allocated, "
618                 "%0.2f%% fragmented, %0.2f%% compressed clusters\n",
619                 check->allocated_clusters, check->total_clusters,
620                 check->allocated_clusters * 100.0 / check->total_clusters,
621                 check->fragmented_clusters * 100.0 / check->allocated_clusters,
622                 check->compressed_clusters * 100.0 /
623                 check->allocated_clusters);
624     }
625 
626     if (check->image_end_offset) {
627         qprintf(quiet,
628                 "Image end offset: %" PRId64 "\n", check->image_end_offset);
629     }
630 }
631 
632 static int collect_image_check(BlockDriverState *bs,
633                    ImageCheck *check,
634                    const char *filename,
635                    const char *fmt,
636                    int fix)
637 {
638     int ret;
639     BdrvCheckResult result;
640 
641     ret = bdrv_check(bs, &result, fix);
642     if (ret < 0) {
643         return ret;
644     }
645 
646     check->filename                 = g_strdup(filename);
647     check->format                   = g_strdup(bdrv_get_format_name(bs));
648     check->check_errors             = result.check_errors;
649     check->corruptions              = result.corruptions;
650     check->has_corruptions          = result.corruptions != 0;
651     check->leaks                    = result.leaks;
652     check->has_leaks                = result.leaks != 0;
653     check->corruptions_fixed        = result.corruptions_fixed;
654     check->has_corruptions_fixed    = result.corruptions != 0;
655     check->leaks_fixed              = result.leaks_fixed;
656     check->has_leaks_fixed          = result.leaks != 0;
657     check->image_end_offset         = result.image_end_offset;
658     check->has_image_end_offset     = result.image_end_offset != 0;
659     check->total_clusters           = result.bfi.total_clusters;
660     check->has_total_clusters       = result.bfi.total_clusters != 0;
661     check->allocated_clusters       = result.bfi.allocated_clusters;
662     check->has_allocated_clusters   = result.bfi.allocated_clusters != 0;
663     check->fragmented_clusters      = result.bfi.fragmented_clusters;
664     check->has_fragmented_clusters  = result.bfi.fragmented_clusters != 0;
665     check->compressed_clusters      = result.bfi.compressed_clusters;
666     check->has_compressed_clusters  = result.bfi.compressed_clusters != 0;
667 
668     return 0;
669 }
670 
671 /*
672  * Checks an image for consistency. Exit codes:
673  *
674  *  0 - Check completed, image is good
675  *  1 - Check not completed because of internal errors
676  *  2 - Check completed, image is corrupted
677  *  3 - Check completed, image has leaked clusters, but is good otherwise
678  * 63 - Checks are not supported by the image format
679  */
680 static int img_check(int argc, char **argv)
681 {
682     int c, ret;
683     OutputFormat output_format = OFORMAT_HUMAN;
684     const char *filename, *fmt, *output, *cache;
685     BlockBackend *blk;
686     BlockDriverState *bs;
687     int fix = 0;
688     int flags = BDRV_O_CHECK;
689     bool writethrough;
690     ImageCheck *check;
691     bool quiet = false;
692     bool image_opts = false;
693     bool force_share = false;
694 
695     fmt = NULL;
696     output = NULL;
697     cache = BDRV_DEFAULT_CACHE;
698 
699     for(;;) {
700         int option_index = 0;
701         static const struct option long_options[] = {
702             {"help", no_argument, 0, 'h'},
703             {"format", required_argument, 0, 'f'},
704             {"repair", required_argument, 0, 'r'},
705             {"output", required_argument, 0, OPTION_OUTPUT},
706             {"object", required_argument, 0, OPTION_OBJECT},
707             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
708             {"force-share", no_argument, 0, 'U'},
709             {0, 0, 0, 0}
710         };
711         c = getopt_long(argc, argv, ":hf:r:T:qU",
712                         long_options, &option_index);
713         if (c == -1) {
714             break;
715         }
716         switch(c) {
717         case ':':
718             missing_argument(argv[optind - 1]);
719             break;
720         case '?':
721             unrecognized_option(argv[optind - 1]);
722             break;
723         case 'h':
724             help();
725             break;
726         case 'f':
727             fmt = optarg;
728             break;
729         case 'r':
730             flags |= BDRV_O_RDWR;
731 
732             if (!strcmp(optarg, "leaks")) {
733                 fix = BDRV_FIX_LEAKS;
734             } else if (!strcmp(optarg, "all")) {
735                 fix = BDRV_FIX_LEAKS | BDRV_FIX_ERRORS;
736             } else {
737                 error_exit("Unknown option value for -r "
738                            "(expecting 'leaks' or 'all'): %s", optarg);
739             }
740             break;
741         case OPTION_OUTPUT:
742             output = optarg;
743             break;
744         case 'T':
745             cache = optarg;
746             break;
747         case 'q':
748             quiet = true;
749             break;
750         case 'U':
751             force_share = true;
752             break;
753         case OPTION_OBJECT: {
754             QemuOpts *opts;
755             opts = qemu_opts_parse_noisily(&qemu_object_opts,
756                                            optarg, true);
757             if (!opts) {
758                 return 1;
759             }
760         }   break;
761         case OPTION_IMAGE_OPTS:
762             image_opts = true;
763             break;
764         }
765     }
766     if (optind != argc - 1) {
767         error_exit("Expecting one image file name");
768     }
769     filename = argv[optind++];
770 
771     if (output && !strcmp(output, "json")) {
772         output_format = OFORMAT_JSON;
773     } else if (output && !strcmp(output, "human")) {
774         output_format = OFORMAT_HUMAN;
775     } else if (output) {
776         error_report("--output must be used with human or json as argument.");
777         return 1;
778     }
779 
780     if (qemu_opts_foreach(&qemu_object_opts,
781                           user_creatable_add_opts_foreach,
782                           NULL, NULL)) {
783         return 1;
784     }
785 
786     ret = bdrv_parse_cache_mode(cache, &flags, &writethrough);
787     if (ret < 0) {
788         error_report("Invalid source cache option: %s", cache);
789         return 1;
790     }
791 
792     blk = img_open(image_opts, filename, fmt, flags, writethrough, quiet,
793                    force_share);
794     if (!blk) {
795         return 1;
796     }
797     bs = blk_bs(blk);
798 
799     check = g_new0(ImageCheck, 1);
800     ret = collect_image_check(bs, check, filename, fmt, fix);
801 
802     if (ret == -ENOTSUP) {
803         error_report("This image format does not support checks");
804         ret = 63;
805         goto fail;
806     }
807 
808     if (check->corruptions_fixed || check->leaks_fixed) {
809         int corruptions_fixed, leaks_fixed;
810 
811         leaks_fixed         = check->leaks_fixed;
812         corruptions_fixed   = check->corruptions_fixed;
813 
814         if (output_format == OFORMAT_HUMAN) {
815             qprintf(quiet,
816                     "The following inconsistencies were found and repaired:\n\n"
817                     "    %" PRId64 " leaked clusters\n"
818                     "    %" PRId64 " corruptions\n\n"
819                     "Double checking the fixed image now...\n",
820                     check->leaks_fixed,
821                     check->corruptions_fixed);
822         }
823 
824         ret = collect_image_check(bs, check, filename, fmt, 0);
825 
826         check->leaks_fixed          = leaks_fixed;
827         check->corruptions_fixed    = corruptions_fixed;
828     }
829 
830     if (!ret) {
831         switch (output_format) {
832         case OFORMAT_HUMAN:
833             dump_human_image_check(check, quiet);
834             break;
835         case OFORMAT_JSON:
836             dump_json_image_check(check, quiet);
837             break;
838         }
839     }
840 
841     if (ret || check->check_errors) {
842         if (ret) {
843             error_report("Check failed: %s", strerror(-ret));
844         } else {
845             error_report("Check failed");
846         }
847         ret = 1;
848         goto fail;
849     }
850 
851     if (check->corruptions) {
852         ret = 2;
853     } else if (check->leaks) {
854         ret = 3;
855     } else {
856         ret = 0;
857     }
858 
859 fail:
860     qapi_free_ImageCheck(check);
861     blk_unref(blk);
862     return ret;
863 }
864 
865 typedef struct CommonBlockJobCBInfo {
866     BlockDriverState *bs;
867     Error **errp;
868 } CommonBlockJobCBInfo;
869 
870 static void common_block_job_cb(void *opaque, int ret)
871 {
872     CommonBlockJobCBInfo *cbi = opaque;
873 
874     if (ret < 0) {
875         error_setg_errno(cbi->errp, -ret, "Block job failed");
876     }
877 }
878 
879 static void run_block_job(BlockJob *job, Error **errp)
880 {
881     AioContext *aio_context = blk_get_aio_context(job->blk);
882     int ret = 0;
883 
884     aio_context_acquire(aio_context);
885     block_job_ref(job);
886     do {
887         aio_poll(aio_context, true);
888         qemu_progress_print(job->len ?
889                             ((float)job->offset / job->len * 100.f) : 0.0f, 0);
890     } while (!job->ready && !job->completed);
891 
892     if (!job->completed) {
893         ret = block_job_complete_sync(job, errp);
894     } else {
895         ret = job->ret;
896     }
897     block_job_unref(job);
898     aio_context_release(aio_context);
899 
900     /* publish completion progress only when success */
901     if (!ret) {
902         qemu_progress_print(100.f, 0);
903     }
904 }
905 
906 static int img_commit(int argc, char **argv)
907 {
908     int c, ret, flags;
909     const char *filename, *fmt, *cache, *base;
910     BlockBackend *blk;
911     BlockDriverState *bs, *base_bs;
912     BlockJob *job;
913     bool progress = false, quiet = false, drop = false;
914     bool writethrough;
915     Error *local_err = NULL;
916     CommonBlockJobCBInfo cbi;
917     bool image_opts = false;
918     AioContext *aio_context;
919 
920     fmt = NULL;
921     cache = BDRV_DEFAULT_CACHE;
922     base = NULL;
923     for(;;) {
924         static const struct option long_options[] = {
925             {"help", no_argument, 0, 'h'},
926             {"object", required_argument, 0, OPTION_OBJECT},
927             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
928             {0, 0, 0, 0}
929         };
930         c = getopt_long(argc, argv, ":f:ht:b:dpq",
931                         long_options, NULL);
932         if (c == -1) {
933             break;
934         }
935         switch(c) {
936         case ':':
937             missing_argument(argv[optind - 1]);
938             break;
939         case '?':
940             unrecognized_option(argv[optind - 1]);
941             break;
942         case 'h':
943             help();
944             break;
945         case 'f':
946             fmt = optarg;
947             break;
948         case 't':
949             cache = optarg;
950             break;
951         case 'b':
952             base = optarg;
953             /* -b implies -d */
954             drop = true;
955             break;
956         case 'd':
957             drop = true;
958             break;
959         case 'p':
960             progress = true;
961             break;
962         case 'q':
963             quiet = true;
964             break;
965         case OPTION_OBJECT: {
966             QemuOpts *opts;
967             opts = qemu_opts_parse_noisily(&qemu_object_opts,
968                                            optarg, true);
969             if (!opts) {
970                 return 1;
971             }
972         }   break;
973         case OPTION_IMAGE_OPTS:
974             image_opts = true;
975             break;
976         }
977     }
978 
979     /* Progress is not shown in Quiet mode */
980     if (quiet) {
981         progress = false;
982     }
983 
984     if (optind != argc - 1) {
985         error_exit("Expecting one image file name");
986     }
987     filename = argv[optind++];
988 
989     if (qemu_opts_foreach(&qemu_object_opts,
990                           user_creatable_add_opts_foreach,
991                           NULL, NULL)) {
992         return 1;
993     }
994 
995     flags = BDRV_O_RDWR | BDRV_O_UNMAP;
996     ret = bdrv_parse_cache_mode(cache, &flags, &writethrough);
997     if (ret < 0) {
998         error_report("Invalid cache option: %s", cache);
999         return 1;
1000     }
1001 
1002     blk = img_open(image_opts, filename, fmt, flags, writethrough, quiet,
1003                    false);
1004     if (!blk) {
1005         return 1;
1006     }
1007     bs = blk_bs(blk);
1008 
1009     qemu_progress_init(progress, 1.f);
1010     qemu_progress_print(0.f, 100);
1011 
1012     if (base) {
1013         base_bs = bdrv_find_backing_image(bs, base);
1014         if (!base_bs) {
1015             error_setg(&local_err,
1016                        "Did not find '%s' in the backing chain of '%s'",
1017                        base, filename);
1018             goto done;
1019         }
1020     } else {
1021         /* This is different from QMP, which by default uses the deepest file in
1022          * the backing chain (i.e., the very base); however, the traditional
1023          * behavior of qemu-img commit is using the immediate backing file. */
1024         base_bs = backing_bs(bs);
1025         if (!base_bs) {
1026             error_setg(&local_err, "Image does not have a backing file");
1027             goto done;
1028         }
1029     }
1030 
1031     cbi = (CommonBlockJobCBInfo){
1032         .errp = &local_err,
1033         .bs   = bs,
1034     };
1035 
1036     aio_context = bdrv_get_aio_context(bs);
1037     aio_context_acquire(aio_context);
1038     commit_active_start("commit", bs, base_bs, BLOCK_JOB_DEFAULT, 0,
1039                         BLOCKDEV_ON_ERROR_REPORT, NULL, common_block_job_cb,
1040                         &cbi, false, &local_err);
1041     aio_context_release(aio_context);
1042     if (local_err) {
1043         goto done;
1044     }
1045 
1046     /* When the block job completes, the BlockBackend reference will point to
1047      * the old backing file. In order to avoid that the top image is already
1048      * deleted, so we can still empty it afterwards, increment the reference
1049      * counter here preemptively. */
1050     if (!drop) {
1051         bdrv_ref(bs);
1052     }
1053 
1054     job = block_job_get("commit");
1055     run_block_job(job, &local_err);
1056     if (local_err) {
1057         goto unref_backing;
1058     }
1059 
1060     if (!drop && bs->drv->bdrv_make_empty) {
1061         ret = bs->drv->bdrv_make_empty(bs);
1062         if (ret) {
1063             error_setg_errno(&local_err, -ret, "Could not empty %s",
1064                              filename);
1065             goto unref_backing;
1066         }
1067     }
1068 
1069 unref_backing:
1070     if (!drop) {
1071         bdrv_unref(bs);
1072     }
1073 
1074 done:
1075     qemu_progress_end();
1076 
1077     blk_unref(blk);
1078 
1079     if (local_err) {
1080         error_report_err(local_err);
1081         return 1;
1082     }
1083 
1084     qprintf(quiet, "Image committed.\n");
1085     return 0;
1086 }
1087 
1088 /*
1089  * Returns true iff the first sector pointed to by 'buf' contains at least
1090  * a non-NUL byte.
1091  *
1092  * 'pnum' is set to the number of sectors (including and immediately following
1093  * the first one) that are known to be in the same allocated/unallocated state.
1094  */
1095 static int is_allocated_sectors(const uint8_t *buf, int n, int *pnum)
1096 {
1097     bool is_zero;
1098     int i;
1099 
1100     if (n <= 0) {
1101         *pnum = 0;
1102         return 0;
1103     }
1104     is_zero = buffer_is_zero(buf, 512);
1105     for(i = 1; i < n; i++) {
1106         buf += 512;
1107         if (is_zero != buffer_is_zero(buf, 512)) {
1108             break;
1109         }
1110     }
1111     *pnum = i;
1112     return !is_zero;
1113 }
1114 
1115 /*
1116  * Like is_allocated_sectors, but if the buffer starts with a used sector,
1117  * up to 'min' consecutive sectors containing zeros are ignored. This avoids
1118  * breaking up write requests for only small sparse areas.
1119  */
1120 static int is_allocated_sectors_min(const uint8_t *buf, int n, int *pnum,
1121     int min)
1122 {
1123     int ret;
1124     int num_checked, num_used;
1125 
1126     if (n < min) {
1127         min = n;
1128     }
1129 
1130     ret = is_allocated_sectors(buf, n, pnum);
1131     if (!ret) {
1132         return ret;
1133     }
1134 
1135     num_used = *pnum;
1136     buf += BDRV_SECTOR_SIZE * *pnum;
1137     n -= *pnum;
1138     num_checked = num_used;
1139 
1140     while (n > 0) {
1141         ret = is_allocated_sectors(buf, n, pnum);
1142 
1143         buf += BDRV_SECTOR_SIZE * *pnum;
1144         n -= *pnum;
1145         num_checked += *pnum;
1146         if (ret) {
1147             num_used = num_checked;
1148         } else if (*pnum >= min) {
1149             break;
1150         }
1151     }
1152 
1153     *pnum = num_used;
1154     return 1;
1155 }
1156 
1157 /*
1158  * Compares two buffers sector by sector. Returns 0 if the first sector of both
1159  * buffers matches, non-zero otherwise.
1160  *
1161  * pnum is set to the number of sectors (including and immediately following
1162  * the first one) that are known to have the same comparison result
1163  */
1164 static int compare_sectors(const uint8_t *buf1, const uint8_t *buf2, int n,
1165     int *pnum)
1166 {
1167     bool res;
1168     int i;
1169 
1170     if (n <= 0) {
1171         *pnum = 0;
1172         return 0;
1173     }
1174 
1175     res = !!memcmp(buf1, buf2, 512);
1176     for(i = 1; i < n; i++) {
1177         buf1 += 512;
1178         buf2 += 512;
1179 
1180         if (!!memcmp(buf1, buf2, 512) != res) {
1181             break;
1182         }
1183     }
1184 
1185     *pnum = i;
1186     return res;
1187 }
1188 
1189 #define IO_BUF_SIZE (2 * 1024 * 1024)
1190 
1191 static int64_t sectors_to_bytes(int64_t sectors)
1192 {
1193     return sectors << BDRV_SECTOR_BITS;
1194 }
1195 
1196 static int64_t sectors_to_process(int64_t total, int64_t from)
1197 {
1198     return MIN(total - from, IO_BUF_SIZE >> BDRV_SECTOR_BITS);
1199 }
1200 
1201 /*
1202  * Check if passed sectors are empty (not allocated or contain only 0 bytes)
1203  *
1204  * Returns 0 in case sectors are filled with 0, 1 if sectors contain non-zero
1205  * data and negative value on error.
1206  *
1207  * @param blk:  BlockBackend for the image
1208  * @param sect_num: Number of first sector to check
1209  * @param sect_count: Number of sectors to check
1210  * @param filename: Name of disk file we are checking (logging purpose)
1211  * @param buffer: Allocated buffer for storing read data
1212  * @param quiet: Flag for quiet mode
1213  */
1214 static int check_empty_sectors(BlockBackend *blk, int64_t sect_num,
1215                                int sect_count, const char *filename,
1216                                uint8_t *buffer, bool quiet)
1217 {
1218     int pnum, ret = 0;
1219     ret = blk_pread(blk, sect_num << BDRV_SECTOR_BITS, buffer,
1220                     sect_count << BDRV_SECTOR_BITS);
1221     if (ret < 0) {
1222         error_report("Error while reading offset %" PRId64 " of %s: %s",
1223                      sectors_to_bytes(sect_num), filename, strerror(-ret));
1224         return ret;
1225     }
1226     ret = is_allocated_sectors(buffer, sect_count, &pnum);
1227     if (ret || pnum != sect_count) {
1228         qprintf(quiet, "Content mismatch at offset %" PRId64 "!\n",
1229                 sectors_to_bytes(ret ? sect_num : sect_num + pnum));
1230         return 1;
1231     }
1232 
1233     return 0;
1234 }
1235 
1236 /*
1237  * Compares two images. Exit codes:
1238  *
1239  * 0 - Images are identical
1240  * 1 - Images differ
1241  * >1 - Error occurred
1242  */
1243 static int img_compare(int argc, char **argv)
1244 {
1245     const char *fmt1 = NULL, *fmt2 = NULL, *cache, *filename1, *filename2;
1246     BlockBackend *blk1, *blk2;
1247     BlockDriverState *bs1, *bs2;
1248     int64_t total_sectors1, total_sectors2;
1249     uint8_t *buf1 = NULL, *buf2 = NULL;
1250     int pnum1, pnum2;
1251     int allocated1, allocated2;
1252     int ret = 0; /* return value - 0 Ident, 1 Different, >1 Error */
1253     bool progress = false, quiet = false, strict = false;
1254     int flags;
1255     bool writethrough;
1256     int64_t total_sectors;
1257     int64_t sector_num = 0;
1258     int64_t nb_sectors;
1259     int c, pnum;
1260     uint64_t progress_base;
1261     bool image_opts = false;
1262     bool force_share = false;
1263 
1264     cache = BDRV_DEFAULT_CACHE;
1265     for (;;) {
1266         static const struct option long_options[] = {
1267             {"help", no_argument, 0, 'h'},
1268             {"object", required_argument, 0, OPTION_OBJECT},
1269             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
1270             {"force-share", no_argument, 0, 'U'},
1271             {0, 0, 0, 0}
1272         };
1273         c = getopt_long(argc, argv, ":hf:F:T:pqsU",
1274                         long_options, NULL);
1275         if (c == -1) {
1276             break;
1277         }
1278         switch (c) {
1279         case ':':
1280             missing_argument(argv[optind - 1]);
1281             break;
1282         case '?':
1283             unrecognized_option(argv[optind - 1]);
1284             break;
1285         case 'h':
1286             help();
1287             break;
1288         case 'f':
1289             fmt1 = optarg;
1290             break;
1291         case 'F':
1292             fmt2 = optarg;
1293             break;
1294         case 'T':
1295             cache = optarg;
1296             break;
1297         case 'p':
1298             progress = true;
1299             break;
1300         case 'q':
1301             quiet = true;
1302             break;
1303         case 's':
1304             strict = true;
1305             break;
1306         case 'U':
1307             force_share = true;
1308             break;
1309         case OPTION_OBJECT: {
1310             QemuOpts *opts;
1311             opts = qemu_opts_parse_noisily(&qemu_object_opts,
1312                                            optarg, true);
1313             if (!opts) {
1314                 ret = 2;
1315                 goto out4;
1316             }
1317         }   break;
1318         case OPTION_IMAGE_OPTS:
1319             image_opts = true;
1320             break;
1321         }
1322     }
1323 
1324     /* Progress is not shown in Quiet mode */
1325     if (quiet) {
1326         progress = false;
1327     }
1328 
1329 
1330     if (optind != argc - 2) {
1331         error_exit("Expecting two image file names");
1332     }
1333     filename1 = argv[optind++];
1334     filename2 = argv[optind++];
1335 
1336     if (qemu_opts_foreach(&qemu_object_opts,
1337                           user_creatable_add_opts_foreach,
1338                           NULL, NULL)) {
1339         ret = 2;
1340         goto out4;
1341     }
1342 
1343     /* Initialize before goto out */
1344     qemu_progress_init(progress, 2.0);
1345 
1346     flags = 0;
1347     ret = bdrv_parse_cache_mode(cache, &flags, &writethrough);
1348     if (ret < 0) {
1349         error_report("Invalid source cache option: %s", cache);
1350         ret = 2;
1351         goto out3;
1352     }
1353 
1354     blk1 = img_open(image_opts, filename1, fmt1, flags, writethrough, quiet,
1355                     force_share);
1356     if (!blk1) {
1357         ret = 2;
1358         goto out3;
1359     }
1360 
1361     blk2 = img_open(image_opts, filename2, fmt2, flags, writethrough, quiet,
1362                     force_share);
1363     if (!blk2) {
1364         ret = 2;
1365         goto out2;
1366     }
1367     bs1 = blk_bs(blk1);
1368     bs2 = blk_bs(blk2);
1369 
1370     buf1 = blk_blockalign(blk1, IO_BUF_SIZE);
1371     buf2 = blk_blockalign(blk2, IO_BUF_SIZE);
1372     total_sectors1 = blk_nb_sectors(blk1);
1373     if (total_sectors1 < 0) {
1374         error_report("Can't get size of %s: %s",
1375                      filename1, strerror(-total_sectors1));
1376         ret = 4;
1377         goto out;
1378     }
1379     total_sectors2 = blk_nb_sectors(blk2);
1380     if (total_sectors2 < 0) {
1381         error_report("Can't get size of %s: %s",
1382                      filename2, strerror(-total_sectors2));
1383         ret = 4;
1384         goto out;
1385     }
1386     total_sectors = MIN(total_sectors1, total_sectors2);
1387     progress_base = MAX(total_sectors1, total_sectors2);
1388 
1389     qemu_progress_print(0, 100);
1390 
1391     if (strict && total_sectors1 != total_sectors2) {
1392         ret = 1;
1393         qprintf(quiet, "Strict mode: Image size mismatch!\n");
1394         goto out;
1395     }
1396 
1397     for (;;) {
1398         int64_t status1, status2;
1399         BlockDriverState *file;
1400 
1401         nb_sectors = sectors_to_process(total_sectors, sector_num);
1402         if (nb_sectors <= 0) {
1403             break;
1404         }
1405         status1 = bdrv_get_block_status_above(bs1, NULL, sector_num,
1406                                               total_sectors1 - sector_num,
1407                                               &pnum1, &file);
1408         if (status1 < 0) {
1409             ret = 3;
1410             error_report("Sector allocation test failed for %s", filename1);
1411             goto out;
1412         }
1413         allocated1 = status1 & BDRV_BLOCK_ALLOCATED;
1414 
1415         status2 = bdrv_get_block_status_above(bs2, NULL, sector_num,
1416                                               total_sectors2 - sector_num,
1417                                               &pnum2, &file);
1418         if (status2 < 0) {
1419             ret = 3;
1420             error_report("Sector allocation test failed for %s", filename2);
1421             goto out;
1422         }
1423         allocated2 = status2 & BDRV_BLOCK_ALLOCATED;
1424         if (pnum1) {
1425             nb_sectors = MIN(nb_sectors, pnum1);
1426         }
1427         if (pnum2) {
1428             nb_sectors = MIN(nb_sectors, pnum2);
1429         }
1430 
1431         if (strict) {
1432             if ((status1 & ~BDRV_BLOCK_OFFSET_MASK) !=
1433                 (status2 & ~BDRV_BLOCK_OFFSET_MASK)) {
1434                 ret = 1;
1435                 qprintf(quiet, "Strict mode: Offset %" PRId64
1436                         " block status mismatch!\n",
1437                         sectors_to_bytes(sector_num));
1438                 goto out;
1439             }
1440         }
1441         if ((status1 & BDRV_BLOCK_ZERO) && (status2 & BDRV_BLOCK_ZERO)) {
1442             nb_sectors = MIN(pnum1, pnum2);
1443         } else if (allocated1 == allocated2) {
1444             if (allocated1) {
1445                 ret = blk_pread(blk1, sector_num << BDRV_SECTOR_BITS, buf1,
1446                                 nb_sectors << BDRV_SECTOR_BITS);
1447                 if (ret < 0) {
1448                     error_report("Error while reading offset %" PRId64 " of %s:"
1449                                  " %s", sectors_to_bytes(sector_num), filename1,
1450                                  strerror(-ret));
1451                     ret = 4;
1452                     goto out;
1453                 }
1454                 ret = blk_pread(blk2, sector_num << BDRV_SECTOR_BITS, buf2,
1455                                 nb_sectors << BDRV_SECTOR_BITS);
1456                 if (ret < 0) {
1457                     error_report("Error while reading offset %" PRId64
1458                                  " of %s: %s", sectors_to_bytes(sector_num),
1459                                  filename2, strerror(-ret));
1460                     ret = 4;
1461                     goto out;
1462                 }
1463                 ret = compare_sectors(buf1, buf2, nb_sectors, &pnum);
1464                 if (ret || pnum != nb_sectors) {
1465                     qprintf(quiet, "Content mismatch at offset %" PRId64 "!\n",
1466                             sectors_to_bytes(
1467                                 ret ? sector_num : sector_num + pnum));
1468                     ret = 1;
1469                     goto out;
1470                 }
1471             }
1472         } else {
1473 
1474             if (allocated1) {
1475                 ret = check_empty_sectors(blk1, sector_num, nb_sectors,
1476                                           filename1, buf1, quiet);
1477             } else {
1478                 ret = check_empty_sectors(blk2, sector_num, nb_sectors,
1479                                           filename2, buf1, quiet);
1480             }
1481             if (ret) {
1482                 if (ret < 0) {
1483                     error_report("Error while reading offset %" PRId64 ": %s",
1484                                  sectors_to_bytes(sector_num), strerror(-ret));
1485                     ret = 4;
1486                 }
1487                 goto out;
1488             }
1489         }
1490         sector_num += nb_sectors;
1491         qemu_progress_print(((float) nb_sectors / progress_base)*100, 100);
1492     }
1493 
1494     if (total_sectors1 != total_sectors2) {
1495         BlockBackend *blk_over;
1496         int64_t total_sectors_over;
1497         const char *filename_over;
1498 
1499         qprintf(quiet, "Warning: Image size mismatch!\n");
1500         if (total_sectors1 > total_sectors2) {
1501             total_sectors_over = total_sectors1;
1502             blk_over = blk1;
1503             filename_over = filename1;
1504         } else {
1505             total_sectors_over = total_sectors2;
1506             blk_over = blk2;
1507             filename_over = filename2;
1508         }
1509 
1510         for (;;) {
1511             int64_t count;
1512 
1513             nb_sectors = sectors_to_process(total_sectors_over, sector_num);
1514             if (nb_sectors <= 0) {
1515                 break;
1516             }
1517             ret = bdrv_is_allocated_above(blk_bs(blk_over), NULL,
1518                                           sector_num * BDRV_SECTOR_SIZE,
1519                                           nb_sectors * BDRV_SECTOR_SIZE,
1520                                           &count);
1521             if (ret < 0) {
1522                 ret = 3;
1523                 error_report("Sector allocation test failed for %s",
1524                              filename_over);
1525                 goto out;
1526 
1527             }
1528             /* TODO relax this once bdrv_is_allocated_above does not enforce
1529              * sector alignment */
1530             assert(QEMU_IS_ALIGNED(count, BDRV_SECTOR_SIZE));
1531             nb_sectors = count >> BDRV_SECTOR_BITS;
1532             if (ret) {
1533                 ret = check_empty_sectors(blk_over, sector_num, nb_sectors,
1534                                           filename_over, buf1, quiet);
1535                 if (ret) {
1536                     if (ret < 0) {
1537                         error_report("Error while reading offset %" PRId64
1538                                      " of %s: %s", sectors_to_bytes(sector_num),
1539                                      filename_over, strerror(-ret));
1540                         ret = 4;
1541                     }
1542                     goto out;
1543                 }
1544             }
1545             sector_num += nb_sectors;
1546             qemu_progress_print(((float) nb_sectors / progress_base)*100, 100);
1547         }
1548     }
1549 
1550     qprintf(quiet, "Images are identical.\n");
1551     ret = 0;
1552 
1553 out:
1554     qemu_vfree(buf1);
1555     qemu_vfree(buf2);
1556     blk_unref(blk2);
1557 out2:
1558     blk_unref(blk1);
1559 out3:
1560     qemu_progress_end();
1561 out4:
1562     return ret;
1563 }
1564 
1565 enum ImgConvertBlockStatus {
1566     BLK_DATA,
1567     BLK_ZERO,
1568     BLK_BACKING_FILE,
1569 };
1570 
1571 #define MAX_COROUTINES 16
1572 
1573 typedef struct ImgConvertState {
1574     BlockBackend **src;
1575     int64_t *src_sectors;
1576     int src_num;
1577     int64_t total_sectors;
1578     int64_t allocated_sectors;
1579     int64_t allocated_done;
1580     int64_t sector_num;
1581     int64_t wr_offs;
1582     enum ImgConvertBlockStatus status;
1583     int64_t sector_next_status;
1584     BlockBackend *target;
1585     bool has_zero_init;
1586     bool compressed;
1587     bool target_has_backing;
1588     bool wr_in_order;
1589     int min_sparse;
1590     size_t cluster_sectors;
1591     size_t buf_sectors;
1592     long num_coroutines;
1593     int running_coroutines;
1594     Coroutine *co[MAX_COROUTINES];
1595     int64_t wait_sector_num[MAX_COROUTINES];
1596     CoMutex lock;
1597     int ret;
1598 } ImgConvertState;
1599 
1600 static void convert_select_part(ImgConvertState *s, int64_t sector_num,
1601                                 int *src_cur, int64_t *src_cur_offset)
1602 {
1603     *src_cur = 0;
1604     *src_cur_offset = 0;
1605     while (sector_num - *src_cur_offset >= s->src_sectors[*src_cur]) {
1606         *src_cur_offset += s->src_sectors[*src_cur];
1607         (*src_cur)++;
1608         assert(*src_cur < s->src_num);
1609     }
1610 }
1611 
1612 static int convert_iteration_sectors(ImgConvertState *s, int64_t sector_num)
1613 {
1614     int64_t ret, src_cur_offset;
1615     int n, src_cur;
1616 
1617     convert_select_part(s, sector_num, &src_cur, &src_cur_offset);
1618 
1619     assert(s->total_sectors > sector_num);
1620     n = MIN(s->total_sectors - sector_num, BDRV_REQUEST_MAX_SECTORS);
1621 
1622     if (s->sector_next_status <= sector_num) {
1623         BlockDriverState *file;
1624         if (s->target_has_backing) {
1625             ret = bdrv_get_block_status(blk_bs(s->src[src_cur]),
1626                                         sector_num - src_cur_offset,
1627                                         n, &n, &file);
1628         } else {
1629             ret = bdrv_get_block_status_above(blk_bs(s->src[src_cur]), NULL,
1630                                               sector_num - src_cur_offset,
1631                                               n, &n, &file);
1632         }
1633         if (ret < 0) {
1634             return ret;
1635         }
1636 
1637         if (ret & BDRV_BLOCK_ZERO) {
1638             s->status = BLK_ZERO;
1639         } else if (ret & BDRV_BLOCK_DATA) {
1640             s->status = BLK_DATA;
1641         } else {
1642             s->status = s->target_has_backing ? BLK_BACKING_FILE : BLK_DATA;
1643         }
1644 
1645         s->sector_next_status = sector_num + n;
1646     }
1647 
1648     n = MIN(n, s->sector_next_status - sector_num);
1649     if (s->status == BLK_DATA) {
1650         n = MIN(n, s->buf_sectors);
1651     }
1652 
1653     /* We need to write complete clusters for compressed images, so if an
1654      * unallocated area is shorter than that, we must consider the whole
1655      * cluster allocated. */
1656     if (s->compressed) {
1657         if (n < s->cluster_sectors) {
1658             n = MIN(s->cluster_sectors, s->total_sectors - sector_num);
1659             s->status = BLK_DATA;
1660         } else {
1661             n = QEMU_ALIGN_DOWN(n, s->cluster_sectors);
1662         }
1663     }
1664 
1665     return n;
1666 }
1667 
1668 static int coroutine_fn convert_co_read(ImgConvertState *s, int64_t sector_num,
1669                                         int nb_sectors, uint8_t *buf)
1670 {
1671     int n, ret;
1672     QEMUIOVector qiov;
1673     struct iovec iov;
1674 
1675     assert(nb_sectors <= s->buf_sectors);
1676     while (nb_sectors > 0) {
1677         BlockBackend *blk;
1678         int src_cur;
1679         int64_t bs_sectors, src_cur_offset;
1680 
1681         /* In the case of compression with multiple source files, we can get a
1682          * nb_sectors that spreads into the next part. So we must be able to
1683          * read across multiple BDSes for one convert_read() call. */
1684         convert_select_part(s, sector_num, &src_cur, &src_cur_offset);
1685         blk = s->src[src_cur];
1686         bs_sectors = s->src_sectors[src_cur];
1687 
1688         n = MIN(nb_sectors, bs_sectors - (sector_num - src_cur_offset));
1689         iov.iov_base = buf;
1690         iov.iov_len = n << BDRV_SECTOR_BITS;
1691         qemu_iovec_init_external(&qiov, &iov, 1);
1692 
1693         ret = blk_co_preadv(
1694                 blk, (sector_num - src_cur_offset) << BDRV_SECTOR_BITS,
1695                 n << BDRV_SECTOR_BITS, &qiov, 0);
1696         if (ret < 0) {
1697             return ret;
1698         }
1699 
1700         sector_num += n;
1701         nb_sectors -= n;
1702         buf += n * BDRV_SECTOR_SIZE;
1703     }
1704 
1705     return 0;
1706 }
1707 
1708 
1709 static int coroutine_fn convert_co_write(ImgConvertState *s, int64_t sector_num,
1710                                          int nb_sectors, uint8_t *buf,
1711                                          enum ImgConvertBlockStatus status)
1712 {
1713     int ret;
1714     QEMUIOVector qiov;
1715     struct iovec iov;
1716 
1717     while (nb_sectors > 0) {
1718         int n = nb_sectors;
1719         BdrvRequestFlags flags = s->compressed ? BDRV_REQ_WRITE_COMPRESSED : 0;
1720 
1721         switch (status) {
1722         case BLK_BACKING_FILE:
1723             /* If we have a backing file, leave clusters unallocated that are
1724              * unallocated in the source image, so that the backing file is
1725              * visible at the respective offset. */
1726             assert(s->target_has_backing);
1727             break;
1728 
1729         case BLK_DATA:
1730             /* If we're told to keep the target fully allocated (-S 0) or there
1731              * is real non-zero data, we must write it. Otherwise we can treat
1732              * it as zero sectors.
1733              * Compressed clusters need to be written as a whole, so in that
1734              * case we can only save the write if the buffer is completely
1735              * zeroed. */
1736             if (!s->min_sparse ||
1737                 (!s->compressed &&
1738                  is_allocated_sectors_min(buf, n, &n, s->min_sparse)) ||
1739                 (s->compressed &&
1740                  !buffer_is_zero(buf, n * BDRV_SECTOR_SIZE)))
1741             {
1742                 iov.iov_base = buf;
1743                 iov.iov_len = n << BDRV_SECTOR_BITS;
1744                 qemu_iovec_init_external(&qiov, &iov, 1);
1745 
1746                 ret = blk_co_pwritev(s->target, sector_num << BDRV_SECTOR_BITS,
1747                                      n << BDRV_SECTOR_BITS, &qiov, flags);
1748                 if (ret < 0) {
1749                     return ret;
1750                 }
1751                 break;
1752             }
1753             /* fall-through */
1754 
1755         case BLK_ZERO:
1756             if (s->has_zero_init) {
1757                 assert(!s->target_has_backing);
1758                 break;
1759             }
1760             ret = blk_co_pwrite_zeroes(s->target,
1761                                        sector_num << BDRV_SECTOR_BITS,
1762                                        n << BDRV_SECTOR_BITS, 0);
1763             if (ret < 0) {
1764                 return ret;
1765             }
1766             break;
1767         }
1768 
1769         sector_num += n;
1770         nb_sectors -= n;
1771         buf += n * BDRV_SECTOR_SIZE;
1772     }
1773 
1774     return 0;
1775 }
1776 
1777 static void coroutine_fn convert_co_do_copy(void *opaque)
1778 {
1779     ImgConvertState *s = opaque;
1780     uint8_t *buf = NULL;
1781     int ret, i;
1782     int index = -1;
1783 
1784     for (i = 0; i < s->num_coroutines; i++) {
1785         if (s->co[i] == qemu_coroutine_self()) {
1786             index = i;
1787             break;
1788         }
1789     }
1790     assert(index >= 0);
1791 
1792     s->running_coroutines++;
1793     buf = blk_blockalign(s->target, s->buf_sectors * BDRV_SECTOR_SIZE);
1794 
1795     while (1) {
1796         int n;
1797         int64_t sector_num;
1798         enum ImgConvertBlockStatus status;
1799 
1800         qemu_co_mutex_lock(&s->lock);
1801         if (s->ret != -EINPROGRESS || s->sector_num >= s->total_sectors) {
1802             qemu_co_mutex_unlock(&s->lock);
1803             break;
1804         }
1805         n = convert_iteration_sectors(s, s->sector_num);
1806         if (n < 0) {
1807             qemu_co_mutex_unlock(&s->lock);
1808             s->ret = n;
1809             break;
1810         }
1811         /* save current sector and allocation status to local variables */
1812         sector_num = s->sector_num;
1813         status = s->status;
1814         if (!s->min_sparse && s->status == BLK_ZERO) {
1815             n = MIN(n, s->buf_sectors);
1816         }
1817         /* increment global sector counter so that other coroutines can
1818          * already continue reading beyond this request */
1819         s->sector_num += n;
1820         qemu_co_mutex_unlock(&s->lock);
1821 
1822         if (status == BLK_DATA || (!s->min_sparse && status == BLK_ZERO)) {
1823             s->allocated_done += n;
1824             qemu_progress_print(100.0 * s->allocated_done /
1825                                         s->allocated_sectors, 0);
1826         }
1827 
1828         if (status == BLK_DATA) {
1829             ret = convert_co_read(s, sector_num, n, buf);
1830             if (ret < 0) {
1831                 error_report("error while reading sector %" PRId64
1832                              ": %s", sector_num, strerror(-ret));
1833                 s->ret = ret;
1834             }
1835         } else if (!s->min_sparse && status == BLK_ZERO) {
1836             status = BLK_DATA;
1837             memset(buf, 0x00, n * BDRV_SECTOR_SIZE);
1838         }
1839 
1840         if (s->wr_in_order) {
1841             /* keep writes in order */
1842             while (s->wr_offs != sector_num && s->ret == -EINPROGRESS) {
1843                 s->wait_sector_num[index] = sector_num;
1844                 qemu_coroutine_yield();
1845             }
1846             s->wait_sector_num[index] = -1;
1847         }
1848 
1849         if (s->ret == -EINPROGRESS) {
1850             ret = convert_co_write(s, sector_num, n, buf, status);
1851             if (ret < 0) {
1852                 error_report("error while writing sector %" PRId64
1853                              ": %s", sector_num, strerror(-ret));
1854                 s->ret = ret;
1855             }
1856         }
1857 
1858         if (s->wr_in_order) {
1859             /* reenter the coroutine that might have waited
1860              * for this write to complete */
1861             s->wr_offs = sector_num + n;
1862             for (i = 0; i < s->num_coroutines; i++) {
1863                 if (s->co[i] && s->wait_sector_num[i] == s->wr_offs) {
1864                     /*
1865                      * A -> B -> A cannot occur because A has
1866                      * s->wait_sector_num[i] == -1 during A -> B.  Therefore
1867                      * B will never enter A during this time window.
1868                      */
1869                     qemu_coroutine_enter(s->co[i]);
1870                     break;
1871                 }
1872             }
1873         }
1874     }
1875 
1876     qemu_vfree(buf);
1877     s->co[index] = NULL;
1878     s->running_coroutines--;
1879     if (!s->running_coroutines && s->ret == -EINPROGRESS) {
1880         /* the convert job finished successfully */
1881         s->ret = 0;
1882     }
1883 }
1884 
1885 static int convert_do_copy(ImgConvertState *s)
1886 {
1887     int ret, i, n;
1888     int64_t sector_num = 0;
1889 
1890     /* Check whether we have zero initialisation or can get it efficiently */
1891     s->has_zero_init = s->min_sparse && !s->target_has_backing
1892                      ? bdrv_has_zero_init(blk_bs(s->target))
1893                      : false;
1894 
1895     if (!s->has_zero_init && !s->target_has_backing &&
1896         bdrv_can_write_zeroes_with_unmap(blk_bs(s->target)))
1897     {
1898         ret = blk_make_zero(s->target, BDRV_REQ_MAY_UNMAP);
1899         if (ret == 0) {
1900             s->has_zero_init = true;
1901         }
1902     }
1903 
1904     /* Allocate buffer for copied data. For compressed images, only one cluster
1905      * can be copied at a time. */
1906     if (s->compressed) {
1907         if (s->cluster_sectors <= 0 || s->cluster_sectors > s->buf_sectors) {
1908             error_report("invalid cluster size");
1909             return -EINVAL;
1910         }
1911         s->buf_sectors = s->cluster_sectors;
1912     }
1913 
1914     while (sector_num < s->total_sectors) {
1915         n = convert_iteration_sectors(s, sector_num);
1916         if (n < 0) {
1917             return n;
1918         }
1919         if (s->status == BLK_DATA || (!s->min_sparse && s->status == BLK_ZERO))
1920         {
1921             s->allocated_sectors += n;
1922         }
1923         sector_num += n;
1924     }
1925 
1926     /* Do the copy */
1927     s->sector_next_status = 0;
1928     s->ret = -EINPROGRESS;
1929 
1930     qemu_co_mutex_init(&s->lock);
1931     for (i = 0; i < s->num_coroutines; i++) {
1932         s->co[i] = qemu_coroutine_create(convert_co_do_copy, s);
1933         s->wait_sector_num[i] = -1;
1934         qemu_coroutine_enter(s->co[i]);
1935     }
1936 
1937     while (s->running_coroutines) {
1938         main_loop_wait(false);
1939     }
1940 
1941     if (s->compressed && !s->ret) {
1942         /* signal EOF to align */
1943         ret = blk_pwrite_compressed(s->target, 0, NULL, 0);
1944         if (ret < 0) {
1945             return ret;
1946         }
1947     }
1948 
1949     return s->ret;
1950 }
1951 
1952 static int img_convert(int argc, char **argv)
1953 {
1954     int c, bs_i, flags, src_flags = 0;
1955     const char *fmt = NULL, *out_fmt = NULL, *cache = "unsafe",
1956                *src_cache = BDRV_DEFAULT_CACHE, *out_baseimg = NULL,
1957                *out_filename, *out_baseimg_param, *snapshot_name = NULL;
1958     BlockDriver *drv = NULL, *proto_drv = NULL;
1959     BlockDriverInfo bdi;
1960     BlockDriverState *out_bs;
1961     QemuOpts *opts = NULL, *sn_opts = NULL;
1962     QemuOptsList *create_opts = NULL;
1963     char *options = NULL;
1964     Error *local_err = NULL;
1965     bool writethrough, src_writethrough, quiet = false, image_opts = false,
1966          skip_create = false, progress = false, tgt_image_opts = false;
1967     int64_t ret = -EINVAL;
1968     bool force_share = false;
1969 
1970     ImgConvertState s = (ImgConvertState) {
1971         /* Need at least 4k of zeros for sparse detection */
1972         .min_sparse         = 8,
1973         .buf_sectors        = IO_BUF_SIZE / BDRV_SECTOR_SIZE,
1974         .wr_in_order        = true,
1975         .num_coroutines     = 8,
1976     };
1977 
1978     for(;;) {
1979         static const struct option long_options[] = {
1980             {"help", no_argument, 0, 'h'},
1981             {"object", required_argument, 0, OPTION_OBJECT},
1982             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
1983             {"force-share", no_argument, 0, 'U'},
1984             {"target-image-opts", no_argument, 0, OPTION_TARGET_IMAGE_OPTS},
1985             {0, 0, 0, 0}
1986         };
1987         c = getopt_long(argc, argv, ":hf:O:B:co:s:l:S:pt:T:qnm:WU",
1988                         long_options, NULL);
1989         if (c == -1) {
1990             break;
1991         }
1992         switch(c) {
1993         case ':':
1994             missing_argument(argv[optind - 1]);
1995             break;
1996         case '?':
1997             unrecognized_option(argv[optind - 1]);
1998             break;
1999         case 'h':
2000             help();
2001             break;
2002         case 'f':
2003             fmt = optarg;
2004             break;
2005         case 'O':
2006             out_fmt = optarg;
2007             break;
2008         case 'B':
2009             out_baseimg = optarg;
2010             break;
2011         case 'c':
2012             s.compressed = true;
2013             break;
2014         case 'o':
2015             if (!is_valid_option_list(optarg)) {
2016                 error_report("Invalid option list: %s", optarg);
2017                 goto fail_getopt;
2018             }
2019             if (!options) {
2020                 options = g_strdup(optarg);
2021             } else {
2022                 char *old_options = options;
2023                 options = g_strdup_printf("%s,%s", options, optarg);
2024                 g_free(old_options);
2025             }
2026             break;
2027         case 's':
2028             snapshot_name = optarg;
2029             break;
2030         case 'l':
2031             if (strstart(optarg, SNAPSHOT_OPT_BASE, NULL)) {
2032                 sn_opts = qemu_opts_parse_noisily(&internal_snapshot_opts,
2033                                                   optarg, false);
2034                 if (!sn_opts) {
2035                     error_report("Failed in parsing snapshot param '%s'",
2036                                  optarg);
2037                     goto fail_getopt;
2038                 }
2039             } else {
2040                 snapshot_name = optarg;
2041             }
2042             break;
2043         case 'S':
2044         {
2045             int64_t sval;
2046 
2047             sval = cvtnum(optarg);
2048             if (sval < 0) {
2049                 error_report("Invalid minimum zero buffer size for sparse output specified");
2050                 goto fail_getopt;
2051             }
2052 
2053             s.min_sparse = sval / BDRV_SECTOR_SIZE;
2054             break;
2055         }
2056         case 'p':
2057             progress = true;
2058             break;
2059         case 't':
2060             cache = optarg;
2061             break;
2062         case 'T':
2063             src_cache = optarg;
2064             break;
2065         case 'q':
2066             quiet = true;
2067             break;
2068         case 'n':
2069             skip_create = true;
2070             break;
2071         case 'm':
2072             if (qemu_strtol(optarg, NULL, 0, &s.num_coroutines) ||
2073                 s.num_coroutines < 1 || s.num_coroutines > MAX_COROUTINES) {
2074                 error_report("Invalid number of coroutines. Allowed number of"
2075                              " coroutines is between 1 and %d", MAX_COROUTINES);
2076                 goto fail_getopt;
2077             }
2078             break;
2079         case 'W':
2080             s.wr_in_order = false;
2081             break;
2082         case 'U':
2083             force_share = true;
2084             break;
2085         case OPTION_OBJECT: {
2086             QemuOpts *object_opts;
2087             object_opts = qemu_opts_parse_noisily(&qemu_object_opts,
2088                                                   optarg, true);
2089             if (!object_opts) {
2090                 goto fail_getopt;
2091             }
2092             break;
2093         }
2094         case OPTION_IMAGE_OPTS:
2095             image_opts = true;
2096             break;
2097         case OPTION_TARGET_IMAGE_OPTS:
2098             tgt_image_opts = true;
2099             break;
2100         }
2101     }
2102 
2103     if (!out_fmt && !tgt_image_opts) {
2104         out_fmt = "raw";
2105     }
2106 
2107     if (qemu_opts_foreach(&qemu_object_opts,
2108                           user_creatable_add_opts_foreach,
2109                           NULL, NULL)) {
2110         goto fail_getopt;
2111     }
2112 
2113     if (!s.wr_in_order && s.compressed) {
2114         error_report("Out of order write and compress are mutually exclusive");
2115         goto fail_getopt;
2116     }
2117 
2118     if (tgt_image_opts && !skip_create) {
2119         error_report("--target-image-opts requires use of -n flag");
2120         goto fail_getopt;
2121     }
2122 
2123     s.src_num = argc - optind - 1;
2124     out_filename = s.src_num >= 1 ? argv[argc - 1] : NULL;
2125 
2126     if (options && has_help_option(options)) {
2127         if (out_fmt) {
2128             ret = print_block_option_help(out_filename, out_fmt);
2129             goto fail_getopt;
2130         } else {
2131             error_report("Option help requires a format be specified");
2132             goto fail_getopt;
2133         }
2134     }
2135 
2136     if (s.src_num < 1) {
2137         error_report("Must specify image file name");
2138         goto fail_getopt;
2139     }
2140 
2141 
2142     /* ret is still -EINVAL until here */
2143     ret = bdrv_parse_cache_mode(src_cache, &src_flags, &src_writethrough);
2144     if (ret < 0) {
2145         error_report("Invalid source cache option: %s", src_cache);
2146         goto fail_getopt;
2147     }
2148 
2149     /* Initialize before goto out */
2150     if (quiet) {
2151         progress = false;
2152     }
2153     qemu_progress_init(progress, 1.0);
2154     qemu_progress_print(0, 100);
2155 
2156     s.src = g_new0(BlockBackend *, s.src_num);
2157     s.src_sectors = g_new(int64_t, s.src_num);
2158 
2159     for (bs_i = 0; bs_i < s.src_num; bs_i++) {
2160         s.src[bs_i] = img_open(image_opts, argv[optind + bs_i],
2161                                fmt, src_flags, src_writethrough, quiet,
2162                                force_share);
2163         if (!s.src[bs_i]) {
2164             ret = -1;
2165             goto out;
2166         }
2167         s.src_sectors[bs_i] = blk_nb_sectors(s.src[bs_i]);
2168         if (s.src_sectors[bs_i] < 0) {
2169             error_report("Could not get size of %s: %s",
2170                          argv[optind + bs_i], strerror(-s.src_sectors[bs_i]));
2171             ret = -1;
2172             goto out;
2173         }
2174         s.total_sectors += s.src_sectors[bs_i];
2175     }
2176 
2177     if (sn_opts) {
2178         bdrv_snapshot_load_tmp(blk_bs(s.src[0]),
2179                                qemu_opt_get(sn_opts, SNAPSHOT_OPT_ID),
2180                                qemu_opt_get(sn_opts, SNAPSHOT_OPT_NAME),
2181                                &local_err);
2182     } else if (snapshot_name != NULL) {
2183         if (s.src_num > 1) {
2184             error_report("No support for concatenating multiple snapshot");
2185             ret = -1;
2186             goto out;
2187         }
2188 
2189         bdrv_snapshot_load_tmp_by_id_or_name(blk_bs(s.src[0]), snapshot_name,
2190                                              &local_err);
2191     }
2192     if (local_err) {
2193         error_reportf_err(local_err, "Failed to load snapshot: ");
2194         ret = -1;
2195         goto out;
2196     }
2197 
2198     if (!skip_create) {
2199         /* Find driver and parse its options */
2200         drv = bdrv_find_format(out_fmt);
2201         if (!drv) {
2202             error_report("Unknown file format '%s'", out_fmt);
2203             ret = -1;
2204             goto out;
2205         }
2206 
2207         proto_drv = bdrv_find_protocol(out_filename, true, &local_err);
2208         if (!proto_drv) {
2209             error_report_err(local_err);
2210             ret = -1;
2211             goto out;
2212         }
2213 
2214         if (!drv->create_opts) {
2215             error_report("Format driver '%s' does not support image creation",
2216                          drv->format_name);
2217             ret = -1;
2218             goto out;
2219         }
2220 
2221         if (!proto_drv->create_opts) {
2222             error_report("Protocol driver '%s' does not support image creation",
2223                          proto_drv->format_name);
2224             ret = -1;
2225             goto out;
2226         }
2227 
2228         create_opts = qemu_opts_append(create_opts, drv->create_opts);
2229         create_opts = qemu_opts_append(create_opts, proto_drv->create_opts);
2230 
2231         opts = qemu_opts_create(create_opts, NULL, 0, &error_abort);
2232         if (options) {
2233             qemu_opts_do_parse(opts, options, NULL, &local_err);
2234             if (local_err) {
2235                 error_report_err(local_err);
2236                 ret = -1;
2237                 goto out;
2238             }
2239         }
2240 
2241         qemu_opt_set_number(opts, BLOCK_OPT_SIZE, s.total_sectors * 512,
2242                             &error_abort);
2243         ret = add_old_style_options(out_fmt, opts, out_baseimg, NULL);
2244         if (ret < 0) {
2245             goto out;
2246         }
2247     }
2248 
2249     /* Get backing file name if -o backing_file was used */
2250     out_baseimg_param = qemu_opt_get(opts, BLOCK_OPT_BACKING_FILE);
2251     if (out_baseimg_param) {
2252         out_baseimg = out_baseimg_param;
2253     }
2254     s.target_has_backing = (bool) out_baseimg;
2255 
2256     if (s.src_num > 1 && out_baseimg) {
2257         error_report("Having a backing file for the target makes no sense when "
2258                      "concatenating multiple input images");
2259         ret = -1;
2260         goto out;
2261     }
2262 
2263     /* Check if compression is supported */
2264     if (s.compressed) {
2265         bool encryption =
2266             qemu_opt_get_bool(opts, BLOCK_OPT_ENCRYPT, false);
2267         const char *preallocation =
2268             qemu_opt_get(opts, BLOCK_OPT_PREALLOC);
2269 
2270         if (drv && !drv->bdrv_co_pwritev_compressed) {
2271             error_report("Compression not supported for this file format");
2272             ret = -1;
2273             goto out;
2274         }
2275 
2276         if (encryption) {
2277             error_report("Compression and encryption not supported at "
2278                          "the same time");
2279             ret = -1;
2280             goto out;
2281         }
2282 
2283         if (preallocation
2284             && strcmp(preallocation, "off"))
2285         {
2286             error_report("Compression and preallocation not supported at "
2287                          "the same time");
2288             ret = -1;
2289             goto out;
2290         }
2291     }
2292 
2293     if (!skip_create) {
2294         /* Create the new image */
2295         ret = bdrv_create(drv, out_filename, opts, &local_err);
2296         if (ret < 0) {
2297             error_reportf_err(local_err, "%s: error while converting %s: ",
2298                               out_filename, out_fmt);
2299             goto out;
2300         }
2301     }
2302 
2303     flags = s.min_sparse ? (BDRV_O_RDWR | BDRV_O_UNMAP) : BDRV_O_RDWR;
2304     ret = bdrv_parse_cache_mode(cache, &flags, &writethrough);
2305     if (ret < 0) {
2306         error_report("Invalid cache option: %s", cache);
2307         goto out;
2308     }
2309 
2310     if (skip_create) {
2311         s.target = img_open(tgt_image_opts, out_filename, out_fmt,
2312                             flags, writethrough, quiet, false);
2313     } else {
2314         /* TODO ultimately we should allow --target-image-opts
2315          * to be used even when -n is not given.
2316          * That has to wait for bdrv_create to be improved
2317          * to allow filenames in option syntax
2318          */
2319         s.target = img_open_new_file(out_filename, opts, out_fmt,
2320                                      flags, writethrough, quiet, false);
2321     }
2322     if (!s.target) {
2323         ret = -1;
2324         goto out;
2325     }
2326     out_bs = blk_bs(s.target);
2327 
2328     if (s.compressed && !out_bs->drv->bdrv_co_pwritev_compressed) {
2329         error_report("Compression not supported for this file format");
2330         ret = -1;
2331         goto out;
2332     }
2333 
2334     /* increase bufsectors from the default 4096 (2M) if opt_transfer
2335      * or discard_alignment of the out_bs is greater. Limit to 32768 (16MB)
2336      * as maximum. */
2337     s.buf_sectors = MIN(32768,
2338                         MAX(s.buf_sectors,
2339                             MAX(out_bs->bl.opt_transfer >> BDRV_SECTOR_BITS,
2340                                 out_bs->bl.pdiscard_alignment >>
2341                                 BDRV_SECTOR_BITS)));
2342 
2343     if (skip_create) {
2344         int64_t output_sectors = blk_nb_sectors(s.target);
2345         if (output_sectors < 0) {
2346             error_report("unable to get output image length: %s",
2347                          strerror(-output_sectors));
2348             ret = -1;
2349             goto out;
2350         } else if (output_sectors < s.total_sectors) {
2351             error_report("output file is smaller than input file");
2352             ret = -1;
2353             goto out;
2354         }
2355     }
2356 
2357     ret = bdrv_get_info(out_bs, &bdi);
2358     if (ret < 0) {
2359         if (s.compressed) {
2360             error_report("could not get block driver info");
2361             goto out;
2362         }
2363     } else {
2364         s.compressed = s.compressed || bdi.needs_compressed_writes;
2365         s.cluster_sectors = bdi.cluster_size / BDRV_SECTOR_SIZE;
2366     }
2367 
2368     ret = convert_do_copy(&s);
2369 out:
2370     if (!ret) {
2371         qemu_progress_print(100, 0);
2372     }
2373     qemu_progress_end();
2374     qemu_opts_del(opts);
2375     qemu_opts_free(create_opts);
2376     qemu_opts_del(sn_opts);
2377     blk_unref(s.target);
2378     if (s.src) {
2379         for (bs_i = 0; bs_i < s.src_num; bs_i++) {
2380             blk_unref(s.src[bs_i]);
2381         }
2382         g_free(s.src);
2383     }
2384     g_free(s.src_sectors);
2385 fail_getopt:
2386     g_free(options);
2387 
2388     return !!ret;
2389 }
2390 
2391 
2392 static void dump_snapshots(BlockDriverState *bs)
2393 {
2394     QEMUSnapshotInfo *sn_tab, *sn;
2395     int nb_sns, i;
2396 
2397     nb_sns = bdrv_snapshot_list(bs, &sn_tab);
2398     if (nb_sns <= 0)
2399         return;
2400     printf("Snapshot list:\n");
2401     bdrv_snapshot_dump(fprintf, stdout, NULL);
2402     printf("\n");
2403     for(i = 0; i < nb_sns; i++) {
2404         sn = &sn_tab[i];
2405         bdrv_snapshot_dump(fprintf, stdout, sn);
2406         printf("\n");
2407     }
2408     g_free(sn_tab);
2409 }
2410 
2411 static void dump_json_image_info_list(ImageInfoList *list)
2412 {
2413     QString *str;
2414     QObject *obj;
2415     Visitor *v = qobject_output_visitor_new(&obj);
2416 
2417     visit_type_ImageInfoList(v, NULL, &list, &error_abort);
2418     visit_complete(v, &obj);
2419     str = qobject_to_json_pretty(obj);
2420     assert(str != NULL);
2421     printf("%s\n", qstring_get_str(str));
2422     qobject_decref(obj);
2423     visit_free(v);
2424     QDECREF(str);
2425 }
2426 
2427 static void dump_json_image_info(ImageInfo *info)
2428 {
2429     QString *str;
2430     QObject *obj;
2431     Visitor *v = qobject_output_visitor_new(&obj);
2432 
2433     visit_type_ImageInfo(v, NULL, &info, &error_abort);
2434     visit_complete(v, &obj);
2435     str = qobject_to_json_pretty(obj);
2436     assert(str != NULL);
2437     printf("%s\n", qstring_get_str(str));
2438     qobject_decref(obj);
2439     visit_free(v);
2440     QDECREF(str);
2441 }
2442 
2443 static void dump_human_image_info_list(ImageInfoList *list)
2444 {
2445     ImageInfoList *elem;
2446     bool delim = false;
2447 
2448     for (elem = list; elem; elem = elem->next) {
2449         if (delim) {
2450             printf("\n");
2451         }
2452         delim = true;
2453 
2454         bdrv_image_info_dump(fprintf, stdout, elem->value);
2455     }
2456 }
2457 
2458 static gboolean str_equal_func(gconstpointer a, gconstpointer b)
2459 {
2460     return strcmp(a, b) == 0;
2461 }
2462 
2463 /**
2464  * Open an image file chain and return an ImageInfoList
2465  *
2466  * @filename: topmost image filename
2467  * @fmt: topmost image format (may be NULL to autodetect)
2468  * @chain: true  - enumerate entire backing file chain
2469  *         false - only topmost image file
2470  *
2471  * Returns a list of ImageInfo objects or NULL if there was an error opening an
2472  * image file.  If there was an error a message will have been printed to
2473  * stderr.
2474  */
2475 static ImageInfoList *collect_image_info_list(bool image_opts,
2476                                               const char *filename,
2477                                               const char *fmt,
2478                                               bool chain, bool force_share)
2479 {
2480     ImageInfoList *head = NULL;
2481     ImageInfoList **last = &head;
2482     GHashTable *filenames;
2483     Error *err = NULL;
2484 
2485     filenames = g_hash_table_new_full(g_str_hash, str_equal_func, NULL, NULL);
2486 
2487     while (filename) {
2488         BlockBackend *blk;
2489         BlockDriverState *bs;
2490         ImageInfo *info;
2491         ImageInfoList *elem;
2492 
2493         if (g_hash_table_lookup_extended(filenames, filename, NULL, NULL)) {
2494             error_report("Backing file '%s' creates an infinite loop.",
2495                          filename);
2496             goto err;
2497         }
2498         g_hash_table_insert(filenames, (gpointer)filename, NULL);
2499 
2500         blk = img_open(image_opts, filename, fmt,
2501                        BDRV_O_NO_BACKING | BDRV_O_NO_IO, false, false,
2502                        force_share);
2503         if (!blk) {
2504             goto err;
2505         }
2506         bs = blk_bs(blk);
2507 
2508         bdrv_query_image_info(bs, &info, &err);
2509         if (err) {
2510             error_report_err(err);
2511             blk_unref(blk);
2512             goto err;
2513         }
2514 
2515         elem = g_new0(ImageInfoList, 1);
2516         elem->value = info;
2517         *last = elem;
2518         last = &elem->next;
2519 
2520         blk_unref(blk);
2521 
2522         filename = fmt = NULL;
2523         if (chain) {
2524             if (info->has_full_backing_filename) {
2525                 filename = info->full_backing_filename;
2526             } else if (info->has_backing_filename) {
2527                 error_report("Could not determine absolute backing filename,"
2528                              " but backing filename '%s' present",
2529                              info->backing_filename);
2530                 goto err;
2531             }
2532             if (info->has_backing_filename_format) {
2533                 fmt = info->backing_filename_format;
2534             }
2535         }
2536     }
2537     g_hash_table_destroy(filenames);
2538     return head;
2539 
2540 err:
2541     qapi_free_ImageInfoList(head);
2542     g_hash_table_destroy(filenames);
2543     return NULL;
2544 }
2545 
2546 static int img_info(int argc, char **argv)
2547 {
2548     int c;
2549     OutputFormat output_format = OFORMAT_HUMAN;
2550     bool chain = false;
2551     const char *filename, *fmt, *output;
2552     ImageInfoList *list;
2553     bool image_opts = false;
2554     bool force_share = false;
2555 
2556     fmt = NULL;
2557     output = NULL;
2558     for(;;) {
2559         int option_index = 0;
2560         static const struct option long_options[] = {
2561             {"help", no_argument, 0, 'h'},
2562             {"format", required_argument, 0, 'f'},
2563             {"output", required_argument, 0, OPTION_OUTPUT},
2564             {"backing-chain", no_argument, 0, OPTION_BACKING_CHAIN},
2565             {"object", required_argument, 0, OPTION_OBJECT},
2566             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
2567             {"force-share", no_argument, 0, 'U'},
2568             {0, 0, 0, 0}
2569         };
2570         c = getopt_long(argc, argv, ":f:hU",
2571                         long_options, &option_index);
2572         if (c == -1) {
2573             break;
2574         }
2575         switch(c) {
2576         case ':':
2577             missing_argument(argv[optind - 1]);
2578             break;
2579         case '?':
2580             unrecognized_option(argv[optind - 1]);
2581             break;
2582         case 'h':
2583             help();
2584             break;
2585         case 'f':
2586             fmt = optarg;
2587             break;
2588         case 'U':
2589             force_share = true;
2590             break;
2591         case OPTION_OUTPUT:
2592             output = optarg;
2593             break;
2594         case OPTION_BACKING_CHAIN:
2595             chain = true;
2596             break;
2597         case OPTION_OBJECT: {
2598             QemuOpts *opts;
2599             opts = qemu_opts_parse_noisily(&qemu_object_opts,
2600                                            optarg, true);
2601             if (!opts) {
2602                 return 1;
2603             }
2604         }   break;
2605         case OPTION_IMAGE_OPTS:
2606             image_opts = true;
2607             break;
2608         }
2609     }
2610     if (optind != argc - 1) {
2611         error_exit("Expecting one image file name");
2612     }
2613     filename = argv[optind++];
2614 
2615     if (output && !strcmp(output, "json")) {
2616         output_format = OFORMAT_JSON;
2617     } else if (output && !strcmp(output, "human")) {
2618         output_format = OFORMAT_HUMAN;
2619     } else if (output) {
2620         error_report("--output must be used with human or json as argument.");
2621         return 1;
2622     }
2623 
2624     if (qemu_opts_foreach(&qemu_object_opts,
2625                           user_creatable_add_opts_foreach,
2626                           NULL, NULL)) {
2627         return 1;
2628     }
2629 
2630     list = collect_image_info_list(image_opts, filename, fmt, chain,
2631                                    force_share);
2632     if (!list) {
2633         return 1;
2634     }
2635 
2636     switch (output_format) {
2637     case OFORMAT_HUMAN:
2638         dump_human_image_info_list(list);
2639         break;
2640     case OFORMAT_JSON:
2641         if (chain) {
2642             dump_json_image_info_list(list);
2643         } else {
2644             dump_json_image_info(list->value);
2645         }
2646         break;
2647     }
2648 
2649     qapi_free_ImageInfoList(list);
2650     return 0;
2651 }
2652 
2653 static void dump_map_entry(OutputFormat output_format, MapEntry *e,
2654                            MapEntry *next)
2655 {
2656     switch (output_format) {
2657     case OFORMAT_HUMAN:
2658         if (e->data && !e->has_offset) {
2659             error_report("File contains external, encrypted or compressed clusters.");
2660             exit(1);
2661         }
2662         if (e->data && !e->zero) {
2663             printf("%#-16"PRIx64"%#-16"PRIx64"%#-16"PRIx64"%s\n",
2664                    e->start, e->length,
2665                    e->has_offset ? e->offset : 0,
2666                    e->has_filename ? e->filename : "");
2667         }
2668         /* This format ignores the distinction between 0, ZERO and ZERO|DATA.
2669          * Modify the flags here to allow more coalescing.
2670          */
2671         if (next && (!next->data || next->zero)) {
2672             next->data = false;
2673             next->zero = true;
2674         }
2675         break;
2676     case OFORMAT_JSON:
2677         printf("%s{ \"start\": %"PRId64", \"length\": %"PRId64","
2678                " \"depth\": %"PRId64", \"zero\": %s, \"data\": %s",
2679                (e->start == 0 ? "[" : ",\n"),
2680                e->start, e->length, e->depth,
2681                e->zero ? "true" : "false",
2682                e->data ? "true" : "false");
2683         if (e->has_offset) {
2684             printf(", \"offset\": %"PRId64"", e->offset);
2685         }
2686         putchar('}');
2687 
2688         if (!next) {
2689             printf("]\n");
2690         }
2691         break;
2692     }
2693 }
2694 
2695 static int get_block_status(BlockDriverState *bs, int64_t sector_num,
2696                             int nb_sectors, MapEntry *e)
2697 {
2698     int64_t ret;
2699     int depth;
2700     BlockDriverState *file;
2701     bool has_offset;
2702 
2703     /* As an optimization, we could cache the current range of unallocated
2704      * clusters in each file of the chain, and avoid querying the same
2705      * range repeatedly.
2706      */
2707 
2708     depth = 0;
2709     for (;;) {
2710         ret = bdrv_get_block_status(bs, sector_num, nb_sectors, &nb_sectors,
2711                                     &file);
2712         if (ret < 0) {
2713             return ret;
2714         }
2715         assert(nb_sectors);
2716         if (ret & (BDRV_BLOCK_ZERO|BDRV_BLOCK_DATA)) {
2717             break;
2718         }
2719         bs = backing_bs(bs);
2720         if (bs == NULL) {
2721             ret = 0;
2722             break;
2723         }
2724 
2725         depth++;
2726     }
2727 
2728     has_offset = !!(ret & BDRV_BLOCK_OFFSET_VALID);
2729 
2730     *e = (MapEntry) {
2731         .start = sector_num * BDRV_SECTOR_SIZE,
2732         .length = nb_sectors * BDRV_SECTOR_SIZE,
2733         .data = !!(ret & BDRV_BLOCK_DATA),
2734         .zero = !!(ret & BDRV_BLOCK_ZERO),
2735         .offset = ret & BDRV_BLOCK_OFFSET_MASK,
2736         .has_offset = has_offset,
2737         .depth = depth,
2738         .has_filename = file && has_offset,
2739         .filename = file && has_offset ? file->filename : NULL,
2740     };
2741 
2742     return 0;
2743 }
2744 
2745 static inline bool entry_mergeable(const MapEntry *curr, const MapEntry *next)
2746 {
2747     if (curr->length == 0) {
2748         return false;
2749     }
2750     if (curr->zero != next->zero ||
2751         curr->data != next->data ||
2752         curr->depth != next->depth ||
2753         curr->has_filename != next->has_filename ||
2754         curr->has_offset != next->has_offset) {
2755         return false;
2756     }
2757     if (curr->has_filename && strcmp(curr->filename, next->filename)) {
2758         return false;
2759     }
2760     if (curr->has_offset && curr->offset + curr->length != next->offset) {
2761         return false;
2762     }
2763     return true;
2764 }
2765 
2766 static int img_map(int argc, char **argv)
2767 {
2768     int c;
2769     OutputFormat output_format = OFORMAT_HUMAN;
2770     BlockBackend *blk;
2771     BlockDriverState *bs;
2772     const char *filename, *fmt, *output;
2773     int64_t length;
2774     MapEntry curr = { .length = 0 }, next;
2775     int ret = 0;
2776     bool image_opts = false;
2777     bool force_share = false;
2778 
2779     fmt = NULL;
2780     output = NULL;
2781     for (;;) {
2782         int option_index = 0;
2783         static const struct option long_options[] = {
2784             {"help", no_argument, 0, 'h'},
2785             {"format", required_argument, 0, 'f'},
2786             {"output", required_argument, 0, OPTION_OUTPUT},
2787             {"object", required_argument, 0, OPTION_OBJECT},
2788             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
2789             {"force-share", no_argument, 0, 'U'},
2790             {0, 0, 0, 0}
2791         };
2792         c = getopt_long(argc, argv, ":f:hU",
2793                         long_options, &option_index);
2794         if (c == -1) {
2795             break;
2796         }
2797         switch (c) {
2798         case ':':
2799             missing_argument(argv[optind - 1]);
2800             break;
2801         case '?':
2802             unrecognized_option(argv[optind - 1]);
2803             break;
2804         case 'h':
2805             help();
2806             break;
2807         case 'f':
2808             fmt = optarg;
2809             break;
2810         case 'U':
2811             force_share = true;
2812             break;
2813         case OPTION_OUTPUT:
2814             output = optarg;
2815             break;
2816         case OPTION_OBJECT: {
2817             QemuOpts *opts;
2818             opts = qemu_opts_parse_noisily(&qemu_object_opts,
2819                                            optarg, true);
2820             if (!opts) {
2821                 return 1;
2822             }
2823         }   break;
2824         case OPTION_IMAGE_OPTS:
2825             image_opts = true;
2826             break;
2827         }
2828     }
2829     if (optind != argc - 1) {
2830         error_exit("Expecting one image file name");
2831     }
2832     filename = argv[optind];
2833 
2834     if (output && !strcmp(output, "json")) {
2835         output_format = OFORMAT_JSON;
2836     } else if (output && !strcmp(output, "human")) {
2837         output_format = OFORMAT_HUMAN;
2838     } else if (output) {
2839         error_report("--output must be used with human or json as argument.");
2840         return 1;
2841     }
2842 
2843     if (qemu_opts_foreach(&qemu_object_opts,
2844                           user_creatable_add_opts_foreach,
2845                           NULL, NULL)) {
2846         return 1;
2847     }
2848 
2849     blk = img_open(image_opts, filename, fmt, 0, false, false, force_share);
2850     if (!blk) {
2851         return 1;
2852     }
2853     bs = blk_bs(blk);
2854 
2855     if (output_format == OFORMAT_HUMAN) {
2856         printf("%-16s%-16s%-16s%s\n", "Offset", "Length", "Mapped to", "File");
2857     }
2858 
2859     length = blk_getlength(blk);
2860     while (curr.start + curr.length < length) {
2861         int64_t nsectors_left;
2862         int64_t sector_num;
2863         int n;
2864 
2865         sector_num = (curr.start + curr.length) >> BDRV_SECTOR_BITS;
2866 
2867         /* Probe up to 1 GiB at a time.  */
2868         nsectors_left = DIV_ROUND_UP(length, BDRV_SECTOR_SIZE) - sector_num;
2869         n = MIN(1 << (30 - BDRV_SECTOR_BITS), nsectors_left);
2870         ret = get_block_status(bs, sector_num, n, &next);
2871 
2872         if (ret < 0) {
2873             error_report("Could not read file metadata: %s", strerror(-ret));
2874             goto out;
2875         }
2876 
2877         if (entry_mergeable(&curr, &next)) {
2878             curr.length += next.length;
2879             continue;
2880         }
2881 
2882         if (curr.length > 0) {
2883             dump_map_entry(output_format, &curr, &next);
2884         }
2885         curr = next;
2886     }
2887 
2888     dump_map_entry(output_format, &curr, NULL);
2889 
2890 out:
2891     blk_unref(blk);
2892     return ret < 0;
2893 }
2894 
2895 #define SNAPSHOT_LIST   1
2896 #define SNAPSHOT_CREATE 2
2897 #define SNAPSHOT_APPLY  3
2898 #define SNAPSHOT_DELETE 4
2899 
2900 static int img_snapshot(int argc, char **argv)
2901 {
2902     BlockBackend *blk;
2903     BlockDriverState *bs;
2904     QEMUSnapshotInfo sn;
2905     char *filename, *snapshot_name = NULL;
2906     int c, ret = 0, bdrv_oflags;
2907     int action = 0;
2908     qemu_timeval tv;
2909     bool quiet = false;
2910     Error *err = NULL;
2911     bool image_opts = false;
2912     bool force_share = false;
2913 
2914     bdrv_oflags = BDRV_O_RDWR;
2915     /* Parse commandline parameters */
2916     for(;;) {
2917         static const struct option long_options[] = {
2918             {"help", no_argument, 0, 'h'},
2919             {"object", required_argument, 0, OPTION_OBJECT},
2920             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
2921             {"force-share", no_argument, 0, 'U'},
2922             {0, 0, 0, 0}
2923         };
2924         c = getopt_long(argc, argv, ":la:c:d:hqU",
2925                         long_options, NULL);
2926         if (c == -1) {
2927             break;
2928         }
2929         switch(c) {
2930         case ':':
2931             missing_argument(argv[optind - 1]);
2932             break;
2933         case '?':
2934             unrecognized_option(argv[optind - 1]);
2935             break;
2936         case 'h':
2937             help();
2938             return 0;
2939         case 'l':
2940             if (action) {
2941                 error_exit("Cannot mix '-l', '-a', '-c', '-d'");
2942                 return 0;
2943             }
2944             action = SNAPSHOT_LIST;
2945             bdrv_oflags &= ~BDRV_O_RDWR; /* no need for RW */
2946             break;
2947         case 'a':
2948             if (action) {
2949                 error_exit("Cannot mix '-l', '-a', '-c', '-d'");
2950                 return 0;
2951             }
2952             action = SNAPSHOT_APPLY;
2953             snapshot_name = optarg;
2954             break;
2955         case 'c':
2956             if (action) {
2957                 error_exit("Cannot mix '-l', '-a', '-c', '-d'");
2958                 return 0;
2959             }
2960             action = SNAPSHOT_CREATE;
2961             snapshot_name = optarg;
2962             break;
2963         case 'd':
2964             if (action) {
2965                 error_exit("Cannot mix '-l', '-a', '-c', '-d'");
2966                 return 0;
2967             }
2968             action = SNAPSHOT_DELETE;
2969             snapshot_name = optarg;
2970             break;
2971         case 'q':
2972             quiet = true;
2973             break;
2974         case 'U':
2975             force_share = true;
2976             break;
2977         case OPTION_OBJECT: {
2978             QemuOpts *opts;
2979             opts = qemu_opts_parse_noisily(&qemu_object_opts,
2980                                            optarg, true);
2981             if (!opts) {
2982                 return 1;
2983             }
2984         }   break;
2985         case OPTION_IMAGE_OPTS:
2986             image_opts = true;
2987             break;
2988         }
2989     }
2990 
2991     if (optind != argc - 1) {
2992         error_exit("Expecting one image file name");
2993     }
2994     filename = argv[optind++];
2995 
2996     if (qemu_opts_foreach(&qemu_object_opts,
2997                           user_creatable_add_opts_foreach,
2998                           NULL, NULL)) {
2999         return 1;
3000     }
3001 
3002     /* Open the image */
3003     blk = img_open(image_opts, filename, NULL, bdrv_oflags, false, quiet,
3004                    force_share);
3005     if (!blk) {
3006         return 1;
3007     }
3008     bs = blk_bs(blk);
3009 
3010     /* Perform the requested action */
3011     switch(action) {
3012     case SNAPSHOT_LIST:
3013         dump_snapshots(bs);
3014         break;
3015 
3016     case SNAPSHOT_CREATE:
3017         memset(&sn, 0, sizeof(sn));
3018         pstrcpy(sn.name, sizeof(sn.name), snapshot_name);
3019 
3020         qemu_gettimeofday(&tv);
3021         sn.date_sec = tv.tv_sec;
3022         sn.date_nsec = tv.tv_usec * 1000;
3023 
3024         ret = bdrv_snapshot_create(bs, &sn);
3025         if (ret) {
3026             error_report("Could not create snapshot '%s': %d (%s)",
3027                 snapshot_name, ret, strerror(-ret));
3028         }
3029         break;
3030 
3031     case SNAPSHOT_APPLY:
3032         ret = bdrv_snapshot_goto(bs, snapshot_name);
3033         if (ret) {
3034             error_report("Could not apply snapshot '%s': %d (%s)",
3035                 snapshot_name, ret, strerror(-ret));
3036         }
3037         break;
3038 
3039     case SNAPSHOT_DELETE:
3040         bdrv_snapshot_delete_by_id_or_name(bs, snapshot_name, &err);
3041         if (err) {
3042             error_reportf_err(err, "Could not delete snapshot '%s': ",
3043                               snapshot_name);
3044             ret = 1;
3045         }
3046         break;
3047     }
3048 
3049     /* Cleanup */
3050     blk_unref(blk);
3051     if (ret) {
3052         return 1;
3053     }
3054     return 0;
3055 }
3056 
3057 static int img_rebase(int argc, char **argv)
3058 {
3059     BlockBackend *blk = NULL, *blk_old_backing = NULL, *blk_new_backing = NULL;
3060     uint8_t *buf_old = NULL;
3061     uint8_t *buf_new = NULL;
3062     BlockDriverState *bs = NULL;
3063     char *filename;
3064     const char *fmt, *cache, *src_cache, *out_basefmt, *out_baseimg;
3065     int c, flags, src_flags, ret;
3066     bool writethrough, src_writethrough;
3067     int unsafe = 0;
3068     bool force_share = false;
3069     int progress = 0;
3070     bool quiet = false;
3071     Error *local_err = NULL;
3072     bool image_opts = false;
3073 
3074     /* Parse commandline parameters */
3075     fmt = NULL;
3076     cache = BDRV_DEFAULT_CACHE;
3077     src_cache = BDRV_DEFAULT_CACHE;
3078     out_baseimg = NULL;
3079     out_basefmt = NULL;
3080     for(;;) {
3081         static const struct option long_options[] = {
3082             {"help", no_argument, 0, 'h'},
3083             {"object", required_argument, 0, OPTION_OBJECT},
3084             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
3085             {"force-share", no_argument, 0, 'U'},
3086             {0, 0, 0, 0}
3087         };
3088         c = getopt_long(argc, argv, ":hf:F:b:upt:T:qU",
3089                         long_options, NULL);
3090         if (c == -1) {
3091             break;
3092         }
3093         switch(c) {
3094         case ':':
3095             missing_argument(argv[optind - 1]);
3096             break;
3097         case '?':
3098             unrecognized_option(argv[optind - 1]);
3099             break;
3100         case 'h':
3101             help();
3102             return 0;
3103         case 'f':
3104             fmt = optarg;
3105             break;
3106         case 'F':
3107             out_basefmt = optarg;
3108             break;
3109         case 'b':
3110             out_baseimg = optarg;
3111             break;
3112         case 'u':
3113             unsafe = 1;
3114             break;
3115         case 'p':
3116             progress = 1;
3117             break;
3118         case 't':
3119             cache = optarg;
3120             break;
3121         case 'T':
3122             src_cache = optarg;
3123             break;
3124         case 'q':
3125             quiet = true;
3126             break;
3127         case OPTION_OBJECT: {
3128             QemuOpts *opts;
3129             opts = qemu_opts_parse_noisily(&qemu_object_opts,
3130                                            optarg, true);
3131             if (!opts) {
3132                 return 1;
3133             }
3134         }   break;
3135         case OPTION_IMAGE_OPTS:
3136             image_opts = true;
3137             break;
3138         case 'U':
3139             force_share = true;
3140             break;
3141         }
3142     }
3143 
3144     if (quiet) {
3145         progress = 0;
3146     }
3147 
3148     if (optind != argc - 1) {
3149         error_exit("Expecting one image file name");
3150     }
3151     if (!unsafe && !out_baseimg) {
3152         error_exit("Must specify backing file (-b) or use unsafe mode (-u)");
3153     }
3154     filename = argv[optind++];
3155 
3156     if (qemu_opts_foreach(&qemu_object_opts,
3157                           user_creatable_add_opts_foreach,
3158                           NULL, NULL)) {
3159         return 1;
3160     }
3161 
3162     qemu_progress_init(progress, 2.0);
3163     qemu_progress_print(0, 100);
3164 
3165     flags = BDRV_O_RDWR | (unsafe ? BDRV_O_NO_BACKING : 0);
3166     ret = bdrv_parse_cache_mode(cache, &flags, &writethrough);
3167     if (ret < 0) {
3168         error_report("Invalid cache option: %s", cache);
3169         goto out;
3170     }
3171 
3172     src_flags = 0;
3173     ret = bdrv_parse_cache_mode(src_cache, &src_flags, &src_writethrough);
3174     if (ret < 0) {
3175         error_report("Invalid source cache option: %s", src_cache);
3176         goto out;
3177     }
3178 
3179     /* The source files are opened read-only, don't care about WCE */
3180     assert((src_flags & BDRV_O_RDWR) == 0);
3181     (void) src_writethrough;
3182 
3183     /*
3184      * Open the images.
3185      *
3186      * Ignore the old backing file for unsafe rebase in case we want to correct
3187      * the reference to a renamed or moved backing file.
3188      */
3189     blk = img_open(image_opts, filename, fmt, flags, writethrough, quiet,
3190                    false);
3191     if (!blk) {
3192         ret = -1;
3193         goto out;
3194     }
3195     bs = blk_bs(blk);
3196 
3197     if (out_basefmt != NULL) {
3198         if (bdrv_find_format(out_basefmt) == NULL) {
3199             error_report("Invalid format name: '%s'", out_basefmt);
3200             ret = -1;
3201             goto out;
3202         }
3203     }
3204 
3205     /* For safe rebasing we need to compare old and new backing file */
3206     if (!unsafe) {
3207         char backing_name[PATH_MAX];
3208         QDict *options = NULL;
3209 
3210         if (bs->backing_format[0] != '\0') {
3211             options = qdict_new();
3212             qdict_put_str(options, "driver", bs->backing_format);
3213         }
3214 
3215         if (force_share) {
3216             if (!options) {
3217                 options = qdict_new();
3218             }
3219             qdict_put_bool(options, BDRV_OPT_FORCE_SHARE, true);
3220         }
3221         bdrv_get_backing_filename(bs, backing_name, sizeof(backing_name));
3222         blk_old_backing = blk_new_open(backing_name, NULL,
3223                                        options, src_flags, &local_err);
3224         if (!blk_old_backing) {
3225             error_reportf_err(local_err,
3226                               "Could not open old backing file '%s': ",
3227                               backing_name);
3228             ret = -1;
3229             goto out;
3230         }
3231 
3232         if (out_baseimg[0]) {
3233             options = qdict_new();
3234             if (out_basefmt) {
3235                 qdict_put_str(options, "driver", out_basefmt);
3236             }
3237             if (force_share) {
3238                 qdict_put_bool(options, BDRV_OPT_FORCE_SHARE, true);
3239             }
3240 
3241             blk_new_backing = blk_new_open(out_baseimg, NULL,
3242                                            options, src_flags, &local_err);
3243             if (!blk_new_backing) {
3244                 error_reportf_err(local_err,
3245                                   "Could not open new backing file '%s': ",
3246                                   out_baseimg);
3247                 ret = -1;
3248                 goto out;
3249             }
3250         }
3251     }
3252 
3253     /*
3254      * Check each unallocated cluster in the COW file. If it is unallocated,
3255      * accesses go to the backing file. We must therefore compare this cluster
3256      * in the old and new backing file, and if they differ we need to copy it
3257      * from the old backing file into the COW file.
3258      *
3259      * If qemu-img crashes during this step, no harm is done. The content of
3260      * the image is the same as the original one at any time.
3261      */
3262     if (!unsafe) {
3263         int64_t num_sectors;
3264         int64_t old_backing_num_sectors;
3265         int64_t new_backing_num_sectors = 0;
3266         uint64_t sector;
3267         int n;
3268         int64_t count;
3269         float local_progress = 0;
3270 
3271         buf_old = blk_blockalign(blk, IO_BUF_SIZE);
3272         buf_new = blk_blockalign(blk, IO_BUF_SIZE);
3273 
3274         num_sectors = blk_nb_sectors(blk);
3275         if (num_sectors < 0) {
3276             error_report("Could not get size of '%s': %s",
3277                          filename, strerror(-num_sectors));
3278             ret = -1;
3279             goto out;
3280         }
3281         old_backing_num_sectors = blk_nb_sectors(blk_old_backing);
3282         if (old_backing_num_sectors < 0) {
3283             char backing_name[PATH_MAX];
3284 
3285             bdrv_get_backing_filename(bs, backing_name, sizeof(backing_name));
3286             error_report("Could not get size of '%s': %s",
3287                          backing_name, strerror(-old_backing_num_sectors));
3288             ret = -1;
3289             goto out;
3290         }
3291         if (blk_new_backing) {
3292             new_backing_num_sectors = blk_nb_sectors(blk_new_backing);
3293             if (new_backing_num_sectors < 0) {
3294                 error_report("Could not get size of '%s': %s",
3295                              out_baseimg, strerror(-new_backing_num_sectors));
3296                 ret = -1;
3297                 goto out;
3298             }
3299         }
3300 
3301         if (num_sectors != 0) {
3302             local_progress = (float)100 /
3303                 (num_sectors / MIN(num_sectors, IO_BUF_SIZE / 512));
3304         }
3305 
3306         for (sector = 0; sector < num_sectors; sector += n) {
3307 
3308             /* How many sectors can we handle with the next read? */
3309             if (sector + (IO_BUF_SIZE / 512) <= num_sectors) {
3310                 n = (IO_BUF_SIZE / 512);
3311             } else {
3312                 n = num_sectors - sector;
3313             }
3314 
3315             /* If the cluster is allocated, we don't need to take action */
3316             ret = bdrv_is_allocated(bs, sector << BDRV_SECTOR_BITS,
3317                                     n << BDRV_SECTOR_BITS, &count);
3318             if (ret < 0) {
3319                 error_report("error while reading image metadata: %s",
3320                              strerror(-ret));
3321                 goto out;
3322             }
3323             /* TODO relax this once bdrv_is_allocated does not enforce
3324              * sector alignment */
3325             assert(QEMU_IS_ALIGNED(count, BDRV_SECTOR_SIZE));
3326             n = count >> BDRV_SECTOR_BITS;
3327             if (ret) {
3328                 continue;
3329             }
3330 
3331             /*
3332              * Read old and new backing file and take into consideration that
3333              * backing files may be smaller than the COW image.
3334              */
3335             if (sector >= old_backing_num_sectors) {
3336                 memset(buf_old, 0, n * BDRV_SECTOR_SIZE);
3337             } else {
3338                 if (sector + n > old_backing_num_sectors) {
3339                     n = old_backing_num_sectors - sector;
3340                 }
3341 
3342                 ret = blk_pread(blk_old_backing, sector << BDRV_SECTOR_BITS,
3343                                 buf_old, n << BDRV_SECTOR_BITS);
3344                 if (ret < 0) {
3345                     error_report("error while reading from old backing file");
3346                     goto out;
3347                 }
3348             }
3349 
3350             if (sector >= new_backing_num_sectors || !blk_new_backing) {
3351                 memset(buf_new, 0, n * BDRV_SECTOR_SIZE);
3352             } else {
3353                 if (sector + n > new_backing_num_sectors) {
3354                     n = new_backing_num_sectors - sector;
3355                 }
3356 
3357                 ret = blk_pread(blk_new_backing, sector << BDRV_SECTOR_BITS,
3358                                 buf_new, n << BDRV_SECTOR_BITS);
3359                 if (ret < 0) {
3360                     error_report("error while reading from new backing file");
3361                     goto out;
3362                 }
3363             }
3364 
3365             /* If they differ, we need to write to the COW file */
3366             uint64_t written = 0;
3367 
3368             while (written < n) {
3369                 int pnum;
3370 
3371                 if (compare_sectors(buf_old + written * 512,
3372                     buf_new + written * 512, n - written, &pnum))
3373                 {
3374                     ret = blk_pwrite(blk,
3375                                      (sector + written) << BDRV_SECTOR_BITS,
3376                                      buf_old + written * 512,
3377                                      pnum << BDRV_SECTOR_BITS, 0);
3378                     if (ret < 0) {
3379                         error_report("Error while writing to COW image: %s",
3380                             strerror(-ret));
3381                         goto out;
3382                     }
3383                 }
3384 
3385                 written += pnum;
3386             }
3387             qemu_progress_print(local_progress, 100);
3388         }
3389     }
3390 
3391     /*
3392      * Change the backing file. All clusters that are different from the old
3393      * backing file are overwritten in the COW file now, so the visible content
3394      * doesn't change when we switch the backing file.
3395      */
3396     if (out_baseimg && *out_baseimg) {
3397         ret = bdrv_change_backing_file(bs, out_baseimg, out_basefmt);
3398     } else {
3399         ret = bdrv_change_backing_file(bs, NULL, NULL);
3400     }
3401 
3402     if (ret == -ENOSPC) {
3403         error_report("Could not change the backing file to '%s': No "
3404                      "space left in the file header", out_baseimg);
3405     } else if (ret < 0) {
3406         error_report("Could not change the backing file to '%s': %s",
3407             out_baseimg, strerror(-ret));
3408     }
3409 
3410     qemu_progress_print(100, 0);
3411     /*
3412      * TODO At this point it is possible to check if any clusters that are
3413      * allocated in the COW file are the same in the backing file. If so, they
3414      * could be dropped from the COW file. Don't do this before switching the
3415      * backing file, in case of a crash this would lead to corruption.
3416      */
3417 out:
3418     qemu_progress_end();
3419     /* Cleanup */
3420     if (!unsafe) {
3421         blk_unref(blk_old_backing);
3422         blk_unref(blk_new_backing);
3423     }
3424     qemu_vfree(buf_old);
3425     qemu_vfree(buf_new);
3426 
3427     blk_unref(blk);
3428     if (ret) {
3429         return 1;
3430     }
3431     return 0;
3432 }
3433 
3434 static int img_resize(int argc, char **argv)
3435 {
3436     Error *err = NULL;
3437     int c, ret, relative;
3438     const char *filename, *fmt, *size;
3439     int64_t n, total_size;
3440     bool quiet = false;
3441     BlockBackend *blk = NULL;
3442     QemuOpts *param;
3443 
3444     static QemuOptsList resize_options = {
3445         .name = "resize_options",
3446         .head = QTAILQ_HEAD_INITIALIZER(resize_options.head),
3447         .desc = {
3448             {
3449                 .name = BLOCK_OPT_SIZE,
3450                 .type = QEMU_OPT_SIZE,
3451                 .help = "Virtual disk size"
3452             }, {
3453                 /* end of list */
3454             }
3455         },
3456     };
3457     bool image_opts = false;
3458 
3459     /* Remove size from argv manually so that negative numbers are not treated
3460      * as options by getopt. */
3461     if (argc < 3) {
3462         error_exit("Not enough arguments");
3463         return 1;
3464     }
3465 
3466     size = argv[--argc];
3467 
3468     /* Parse getopt arguments */
3469     fmt = NULL;
3470     for(;;) {
3471         static const struct option long_options[] = {
3472             {"help", no_argument, 0, 'h'},
3473             {"object", required_argument, 0, OPTION_OBJECT},
3474             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
3475             {0, 0, 0, 0}
3476         };
3477         c = getopt_long(argc, argv, ":f:hq",
3478                         long_options, NULL);
3479         if (c == -1) {
3480             break;
3481         }
3482         switch(c) {
3483         case ':':
3484             missing_argument(argv[optind - 1]);
3485             break;
3486         case '?':
3487             unrecognized_option(argv[optind - 1]);
3488             break;
3489         case 'h':
3490             help();
3491             break;
3492         case 'f':
3493             fmt = optarg;
3494             break;
3495         case 'q':
3496             quiet = true;
3497             break;
3498         case OPTION_OBJECT: {
3499             QemuOpts *opts;
3500             opts = qemu_opts_parse_noisily(&qemu_object_opts,
3501                                            optarg, true);
3502             if (!opts) {
3503                 return 1;
3504             }
3505         }   break;
3506         case OPTION_IMAGE_OPTS:
3507             image_opts = true;
3508             break;
3509         }
3510     }
3511     if (optind != argc - 1) {
3512         error_exit("Expecting one image file name");
3513     }
3514     filename = argv[optind++];
3515 
3516     if (qemu_opts_foreach(&qemu_object_opts,
3517                           user_creatable_add_opts_foreach,
3518                           NULL, NULL)) {
3519         return 1;
3520     }
3521 
3522     /* Choose grow, shrink, or absolute resize mode */
3523     switch (size[0]) {
3524     case '+':
3525         relative = 1;
3526         size++;
3527         break;
3528     case '-':
3529         relative = -1;
3530         size++;
3531         break;
3532     default:
3533         relative = 0;
3534         break;
3535     }
3536 
3537     /* Parse size */
3538     param = qemu_opts_create(&resize_options, NULL, 0, &error_abort);
3539     qemu_opt_set(param, BLOCK_OPT_SIZE, size, &err);
3540     if (err) {
3541         error_report_err(err);
3542         ret = -1;
3543         qemu_opts_del(param);
3544         goto out;
3545     }
3546     n = qemu_opt_get_size(param, BLOCK_OPT_SIZE, 0);
3547     qemu_opts_del(param);
3548 
3549     blk = img_open(image_opts, filename, fmt,
3550                    BDRV_O_RDWR | BDRV_O_RESIZE, false, quiet,
3551                    false);
3552     if (!blk) {
3553         ret = -1;
3554         goto out;
3555     }
3556 
3557     if (relative) {
3558         total_size = blk_getlength(blk) + n * relative;
3559     } else {
3560         total_size = n;
3561     }
3562     if (total_size <= 0) {
3563         error_report("New image size must be positive");
3564         ret = -1;
3565         goto out;
3566     }
3567 
3568     ret = blk_truncate(blk, total_size, &err);
3569     if (!ret) {
3570         qprintf(quiet, "Image resized.\n");
3571     } else {
3572         error_report_err(err);
3573     }
3574 out:
3575     blk_unref(blk);
3576     if (ret) {
3577         return 1;
3578     }
3579     return 0;
3580 }
3581 
3582 static void amend_status_cb(BlockDriverState *bs,
3583                             int64_t offset, int64_t total_work_size,
3584                             void *opaque)
3585 {
3586     qemu_progress_print(100.f * offset / total_work_size, 0);
3587 }
3588 
3589 static int img_amend(int argc, char **argv)
3590 {
3591     Error *err = NULL;
3592     int c, ret = 0;
3593     char *options = NULL;
3594     QemuOptsList *create_opts = NULL;
3595     QemuOpts *opts = NULL;
3596     const char *fmt = NULL, *filename, *cache;
3597     int flags;
3598     bool writethrough;
3599     bool quiet = false, progress = false;
3600     BlockBackend *blk = NULL;
3601     BlockDriverState *bs = NULL;
3602     bool image_opts = false;
3603 
3604     cache = BDRV_DEFAULT_CACHE;
3605     for (;;) {
3606         static const struct option long_options[] = {
3607             {"help", no_argument, 0, 'h'},
3608             {"object", required_argument, 0, OPTION_OBJECT},
3609             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
3610             {0, 0, 0, 0}
3611         };
3612         c = getopt_long(argc, argv, ":ho:f:t:pq",
3613                         long_options, NULL);
3614         if (c == -1) {
3615             break;
3616         }
3617 
3618         switch (c) {
3619         case ':':
3620             missing_argument(argv[optind - 1]);
3621             break;
3622         case '?':
3623             unrecognized_option(argv[optind - 1]);
3624             break;
3625         case 'h':
3626             help();
3627             break;
3628         case 'o':
3629             if (!is_valid_option_list(optarg)) {
3630                 error_report("Invalid option list: %s", optarg);
3631                 ret = -1;
3632                 goto out_no_progress;
3633             }
3634             if (!options) {
3635                 options = g_strdup(optarg);
3636             } else {
3637                 char *old_options = options;
3638                 options = g_strdup_printf("%s,%s", options, optarg);
3639                 g_free(old_options);
3640             }
3641             break;
3642         case 'f':
3643             fmt = optarg;
3644             break;
3645         case 't':
3646             cache = optarg;
3647             break;
3648         case 'p':
3649             progress = true;
3650             break;
3651         case 'q':
3652             quiet = true;
3653             break;
3654         case OPTION_OBJECT:
3655             opts = qemu_opts_parse_noisily(&qemu_object_opts,
3656                                            optarg, true);
3657             if (!opts) {
3658                 ret = -1;
3659                 goto out_no_progress;
3660             }
3661             break;
3662         case OPTION_IMAGE_OPTS:
3663             image_opts = true;
3664             break;
3665         }
3666     }
3667 
3668     if (!options) {
3669         error_exit("Must specify options (-o)");
3670     }
3671 
3672     if (qemu_opts_foreach(&qemu_object_opts,
3673                           user_creatable_add_opts_foreach,
3674                           NULL, NULL)) {
3675         ret = -1;
3676         goto out_no_progress;
3677     }
3678 
3679     if (quiet) {
3680         progress = false;
3681     }
3682     qemu_progress_init(progress, 1.0);
3683 
3684     filename = (optind == argc - 1) ? argv[argc - 1] : NULL;
3685     if (fmt && has_help_option(options)) {
3686         /* If a format is explicitly specified (and possibly no filename is
3687          * given), print option help here */
3688         ret = print_block_option_help(filename, fmt);
3689         goto out;
3690     }
3691 
3692     if (optind != argc - 1) {
3693         error_report("Expecting one image file name");
3694         ret = -1;
3695         goto out;
3696     }
3697 
3698     flags = BDRV_O_RDWR;
3699     ret = bdrv_parse_cache_mode(cache, &flags, &writethrough);
3700     if (ret < 0) {
3701         error_report("Invalid cache option: %s", cache);
3702         goto out;
3703     }
3704 
3705     blk = img_open(image_opts, filename, fmt, flags, writethrough, quiet,
3706                    false);
3707     if (!blk) {
3708         ret = -1;
3709         goto out;
3710     }
3711     bs = blk_bs(blk);
3712 
3713     fmt = bs->drv->format_name;
3714 
3715     if (has_help_option(options)) {
3716         /* If the format was auto-detected, print option help here */
3717         ret = print_block_option_help(filename, fmt);
3718         goto out;
3719     }
3720 
3721     if (!bs->drv->create_opts) {
3722         error_report("Format driver '%s' does not support any options to amend",
3723                      fmt);
3724         ret = -1;
3725         goto out;
3726     }
3727 
3728     create_opts = qemu_opts_append(create_opts, bs->drv->create_opts);
3729     opts = qemu_opts_create(create_opts, NULL, 0, &error_abort);
3730     qemu_opts_do_parse(opts, options, NULL, &err);
3731     if (err) {
3732         error_report_err(err);
3733         ret = -1;
3734         goto out;
3735     }
3736 
3737     /* In case the driver does not call amend_status_cb() */
3738     qemu_progress_print(0.f, 0);
3739     ret = bdrv_amend_options(bs, opts, &amend_status_cb, NULL);
3740     qemu_progress_print(100.f, 0);
3741     if (ret < 0) {
3742         error_report("Error while amending options: %s", strerror(-ret));
3743         goto out;
3744     }
3745 
3746 out:
3747     qemu_progress_end();
3748 
3749 out_no_progress:
3750     blk_unref(blk);
3751     qemu_opts_del(opts);
3752     qemu_opts_free(create_opts);
3753     g_free(options);
3754 
3755     if (ret) {
3756         return 1;
3757     }
3758     return 0;
3759 }
3760 
3761 typedef struct BenchData {
3762     BlockBackend *blk;
3763     uint64_t image_size;
3764     bool write;
3765     int bufsize;
3766     int step;
3767     int nrreq;
3768     int n;
3769     int flush_interval;
3770     bool drain_on_flush;
3771     uint8_t *buf;
3772     QEMUIOVector *qiov;
3773 
3774     int in_flight;
3775     bool in_flush;
3776     uint64_t offset;
3777 } BenchData;
3778 
3779 static void bench_undrained_flush_cb(void *opaque, int ret)
3780 {
3781     if (ret < 0) {
3782         error_report("Failed flush request: %s", strerror(-ret));
3783         exit(EXIT_FAILURE);
3784     }
3785 }
3786 
3787 static void bench_cb(void *opaque, int ret)
3788 {
3789     BenchData *b = opaque;
3790     BlockAIOCB *acb;
3791 
3792     if (ret < 0) {
3793         error_report("Failed request: %s", strerror(-ret));
3794         exit(EXIT_FAILURE);
3795     }
3796 
3797     if (b->in_flush) {
3798         /* Just finished a flush with drained queue: Start next requests */
3799         assert(b->in_flight == 0);
3800         b->in_flush = false;
3801     } else if (b->in_flight > 0) {
3802         int remaining = b->n - b->in_flight;
3803 
3804         b->n--;
3805         b->in_flight--;
3806 
3807         /* Time for flush? Drain queue if requested, then flush */
3808         if (b->flush_interval && remaining % b->flush_interval == 0) {
3809             if (!b->in_flight || !b->drain_on_flush) {
3810                 BlockCompletionFunc *cb;
3811 
3812                 if (b->drain_on_flush) {
3813                     b->in_flush = true;
3814                     cb = bench_cb;
3815                 } else {
3816                     cb = bench_undrained_flush_cb;
3817                 }
3818 
3819                 acb = blk_aio_flush(b->blk, cb, b);
3820                 if (!acb) {
3821                     error_report("Failed to issue flush request");
3822                     exit(EXIT_FAILURE);
3823                 }
3824             }
3825             if (b->drain_on_flush) {
3826                 return;
3827             }
3828         }
3829     }
3830 
3831     while (b->n > b->in_flight && b->in_flight < b->nrreq) {
3832         int64_t offset = b->offset;
3833         /* blk_aio_* might look for completed I/Os and kick bench_cb
3834          * again, so make sure this operation is counted by in_flight
3835          * and b->offset is ready for the next submission.
3836          */
3837         b->in_flight++;
3838         b->offset += b->step;
3839         b->offset %= b->image_size;
3840         if (b->write) {
3841             acb = blk_aio_pwritev(b->blk, offset, b->qiov, 0, bench_cb, b);
3842         } else {
3843             acb = blk_aio_preadv(b->blk, offset, b->qiov, 0, bench_cb, b);
3844         }
3845         if (!acb) {
3846             error_report("Failed to issue request");
3847             exit(EXIT_FAILURE);
3848         }
3849     }
3850 }
3851 
3852 static int img_bench(int argc, char **argv)
3853 {
3854     int c, ret = 0;
3855     const char *fmt = NULL, *filename;
3856     bool quiet = false;
3857     bool image_opts = false;
3858     bool is_write = false;
3859     int count = 75000;
3860     int depth = 64;
3861     int64_t offset = 0;
3862     size_t bufsize = 4096;
3863     int pattern = 0;
3864     size_t step = 0;
3865     int flush_interval = 0;
3866     bool drain_on_flush = true;
3867     int64_t image_size;
3868     BlockBackend *blk = NULL;
3869     BenchData data = {};
3870     int flags = 0;
3871     bool writethrough = false;
3872     struct timeval t1, t2;
3873     int i;
3874     bool force_share = false;
3875 
3876     for (;;) {
3877         static const struct option long_options[] = {
3878             {"help", no_argument, 0, 'h'},
3879             {"flush-interval", required_argument, 0, OPTION_FLUSH_INTERVAL},
3880             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
3881             {"pattern", required_argument, 0, OPTION_PATTERN},
3882             {"no-drain", no_argument, 0, OPTION_NO_DRAIN},
3883             {"force-share", no_argument, 0, 'U'},
3884             {0, 0, 0, 0}
3885         };
3886         c = getopt_long(argc, argv, ":hc:d:f:no:qs:S:t:wU", long_options, NULL);
3887         if (c == -1) {
3888             break;
3889         }
3890 
3891         switch (c) {
3892         case ':':
3893             missing_argument(argv[optind - 1]);
3894             break;
3895         case '?':
3896             unrecognized_option(argv[optind - 1]);
3897             break;
3898         case 'h':
3899             help();
3900             break;
3901         case 'c':
3902         {
3903             unsigned long res;
3904 
3905             if (qemu_strtoul(optarg, NULL, 0, &res) < 0 || res > INT_MAX) {
3906                 error_report("Invalid request count specified");
3907                 return 1;
3908             }
3909             count = res;
3910             break;
3911         }
3912         case 'd':
3913         {
3914             unsigned long res;
3915 
3916             if (qemu_strtoul(optarg, NULL, 0, &res) < 0 || res > INT_MAX) {
3917                 error_report("Invalid queue depth specified");
3918                 return 1;
3919             }
3920             depth = res;
3921             break;
3922         }
3923         case 'f':
3924             fmt = optarg;
3925             break;
3926         case 'n':
3927             flags |= BDRV_O_NATIVE_AIO;
3928             break;
3929         case 'o':
3930         {
3931             offset = cvtnum(optarg);
3932             if (offset < 0) {
3933                 error_report("Invalid offset specified");
3934                 return 1;
3935             }
3936             break;
3937         }
3938             break;
3939         case 'q':
3940             quiet = true;
3941             break;
3942         case 's':
3943         {
3944             int64_t sval;
3945 
3946             sval = cvtnum(optarg);
3947             if (sval < 0 || sval > INT_MAX) {
3948                 error_report("Invalid buffer size specified");
3949                 return 1;
3950             }
3951 
3952             bufsize = sval;
3953             break;
3954         }
3955         case 'S':
3956         {
3957             int64_t sval;
3958 
3959             sval = cvtnum(optarg);
3960             if (sval < 0 || sval > INT_MAX) {
3961                 error_report("Invalid step size specified");
3962                 return 1;
3963             }
3964 
3965             step = sval;
3966             break;
3967         }
3968         case 't':
3969             ret = bdrv_parse_cache_mode(optarg, &flags, &writethrough);
3970             if (ret < 0) {
3971                 error_report("Invalid cache mode");
3972                 ret = -1;
3973                 goto out;
3974             }
3975             break;
3976         case 'w':
3977             flags |= BDRV_O_RDWR;
3978             is_write = true;
3979             break;
3980         case 'U':
3981             force_share = true;
3982             break;
3983         case OPTION_PATTERN:
3984         {
3985             unsigned long res;
3986 
3987             if (qemu_strtoul(optarg, NULL, 0, &res) < 0 || res > 0xff) {
3988                 error_report("Invalid pattern byte specified");
3989                 return 1;
3990             }
3991             pattern = res;
3992             break;
3993         }
3994         case OPTION_FLUSH_INTERVAL:
3995         {
3996             unsigned long res;
3997 
3998             if (qemu_strtoul(optarg, NULL, 0, &res) < 0 || res > INT_MAX) {
3999                 error_report("Invalid flush interval specified");
4000                 return 1;
4001             }
4002             flush_interval = res;
4003             break;
4004         }
4005         case OPTION_NO_DRAIN:
4006             drain_on_flush = false;
4007             break;
4008         case OPTION_IMAGE_OPTS:
4009             image_opts = true;
4010             break;
4011         }
4012     }
4013 
4014     if (optind != argc - 1) {
4015         error_exit("Expecting one image file name");
4016     }
4017     filename = argv[argc - 1];
4018 
4019     if (!is_write && flush_interval) {
4020         error_report("--flush-interval is only available in write tests");
4021         ret = -1;
4022         goto out;
4023     }
4024     if (flush_interval && flush_interval < depth) {
4025         error_report("Flush interval can't be smaller than depth");
4026         ret = -1;
4027         goto out;
4028     }
4029 
4030     blk = img_open(image_opts, filename, fmt, flags, writethrough, quiet,
4031                    force_share);
4032     if (!blk) {
4033         ret = -1;
4034         goto out;
4035     }
4036 
4037     image_size = blk_getlength(blk);
4038     if (image_size < 0) {
4039         ret = image_size;
4040         goto out;
4041     }
4042 
4043     data = (BenchData) {
4044         .blk            = blk,
4045         .image_size     = image_size,
4046         .bufsize        = bufsize,
4047         .step           = step ?: bufsize,
4048         .nrreq          = depth,
4049         .n              = count,
4050         .offset         = offset,
4051         .write          = is_write,
4052         .flush_interval = flush_interval,
4053         .drain_on_flush = drain_on_flush,
4054     };
4055     printf("Sending %d %s requests, %d bytes each, %d in parallel "
4056            "(starting at offset %" PRId64 ", step size %d)\n",
4057            data.n, data.write ? "write" : "read", data.bufsize, data.nrreq,
4058            data.offset, data.step);
4059     if (flush_interval) {
4060         printf("Sending flush every %d requests\n", flush_interval);
4061     }
4062 
4063     data.buf = blk_blockalign(blk, data.nrreq * data.bufsize);
4064     memset(data.buf, pattern, data.nrreq * data.bufsize);
4065 
4066     data.qiov = g_new(QEMUIOVector, data.nrreq);
4067     for (i = 0; i < data.nrreq; i++) {
4068         qemu_iovec_init(&data.qiov[i], 1);
4069         qemu_iovec_add(&data.qiov[i],
4070                        data.buf + i * data.bufsize, data.bufsize);
4071     }
4072 
4073     gettimeofday(&t1, NULL);
4074     bench_cb(&data, 0);
4075 
4076     while (data.n > 0) {
4077         main_loop_wait(false);
4078     }
4079     gettimeofday(&t2, NULL);
4080 
4081     printf("Run completed in %3.3f seconds.\n",
4082            (t2.tv_sec - t1.tv_sec)
4083            + ((double)(t2.tv_usec - t1.tv_usec) / 1000000));
4084 
4085 out:
4086     qemu_vfree(data.buf);
4087     blk_unref(blk);
4088 
4089     if (ret) {
4090         return 1;
4091     }
4092     return 0;
4093 }
4094 
4095 #define C_BS      01
4096 #define C_COUNT   02
4097 #define C_IF      04
4098 #define C_OF      010
4099 #define C_SKIP    020
4100 
4101 struct DdInfo {
4102     unsigned int flags;
4103     int64_t count;
4104 };
4105 
4106 struct DdIo {
4107     int bsz;    /* Block size */
4108     char *filename;
4109     uint8_t *buf;
4110     int64_t offset;
4111 };
4112 
4113 struct DdOpts {
4114     const char *name;
4115     int (*f)(const char *, struct DdIo *, struct DdIo *, struct DdInfo *);
4116     unsigned int flag;
4117 };
4118 
4119 static int img_dd_bs(const char *arg,
4120                      struct DdIo *in, struct DdIo *out,
4121                      struct DdInfo *dd)
4122 {
4123     int64_t res;
4124 
4125     res = cvtnum(arg);
4126 
4127     if (res <= 0 || res > INT_MAX) {
4128         error_report("invalid number: '%s'", arg);
4129         return 1;
4130     }
4131     in->bsz = out->bsz = res;
4132 
4133     return 0;
4134 }
4135 
4136 static int img_dd_count(const char *arg,
4137                         struct DdIo *in, struct DdIo *out,
4138                         struct DdInfo *dd)
4139 {
4140     dd->count = cvtnum(arg);
4141 
4142     if (dd->count < 0) {
4143         error_report("invalid number: '%s'", arg);
4144         return 1;
4145     }
4146 
4147     return 0;
4148 }
4149 
4150 static int img_dd_if(const char *arg,
4151                      struct DdIo *in, struct DdIo *out,
4152                      struct DdInfo *dd)
4153 {
4154     in->filename = g_strdup(arg);
4155 
4156     return 0;
4157 }
4158 
4159 static int img_dd_of(const char *arg,
4160                      struct DdIo *in, struct DdIo *out,
4161                      struct DdInfo *dd)
4162 {
4163     out->filename = g_strdup(arg);
4164 
4165     return 0;
4166 }
4167 
4168 static int img_dd_skip(const char *arg,
4169                        struct DdIo *in, struct DdIo *out,
4170                        struct DdInfo *dd)
4171 {
4172     in->offset = cvtnum(arg);
4173 
4174     if (in->offset < 0) {
4175         error_report("invalid number: '%s'", arg);
4176         return 1;
4177     }
4178 
4179     return 0;
4180 }
4181 
4182 static int img_dd(int argc, char **argv)
4183 {
4184     int ret = 0;
4185     char *arg = NULL;
4186     char *tmp;
4187     BlockDriver *drv = NULL, *proto_drv = NULL;
4188     BlockBackend *blk1 = NULL, *blk2 = NULL;
4189     QemuOpts *opts = NULL;
4190     QemuOptsList *create_opts = NULL;
4191     Error *local_err = NULL;
4192     bool image_opts = false;
4193     int c, i;
4194     const char *out_fmt = "raw";
4195     const char *fmt = NULL;
4196     int64_t size = 0;
4197     int64_t block_count = 0, out_pos, in_pos;
4198     bool force_share = false;
4199     struct DdInfo dd = {
4200         .flags = 0,
4201         .count = 0,
4202     };
4203     struct DdIo in = {
4204         .bsz = 512, /* Block size is by default 512 bytes */
4205         .filename = NULL,
4206         .buf = NULL,
4207         .offset = 0
4208     };
4209     struct DdIo out = {
4210         .bsz = 512,
4211         .filename = NULL,
4212         .buf = NULL,
4213         .offset = 0
4214     };
4215 
4216     const struct DdOpts options[] = {
4217         { "bs", img_dd_bs, C_BS },
4218         { "count", img_dd_count, C_COUNT },
4219         { "if", img_dd_if, C_IF },
4220         { "of", img_dd_of, C_OF },
4221         { "skip", img_dd_skip, C_SKIP },
4222         { NULL, NULL, 0 }
4223     };
4224     const struct option long_options[] = {
4225         { "help", no_argument, 0, 'h'},
4226         { "object", required_argument, 0, OPTION_OBJECT},
4227         { "image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
4228         { "force-share", no_argument, 0, 'U'},
4229         { 0, 0, 0, 0 }
4230     };
4231 
4232     while ((c = getopt_long(argc, argv, ":hf:O:U", long_options, NULL))) {
4233         if (c == EOF) {
4234             break;
4235         }
4236         switch (c) {
4237         case 'O':
4238             out_fmt = optarg;
4239             break;
4240         case 'f':
4241             fmt = optarg;
4242             break;
4243         case ':':
4244             missing_argument(argv[optind - 1]);
4245             break;
4246         case '?':
4247             unrecognized_option(argv[optind - 1]);
4248             break;
4249         case 'h':
4250             help();
4251             break;
4252         case 'U':
4253             force_share = true;
4254             break;
4255         case OPTION_OBJECT:
4256             if (!qemu_opts_parse_noisily(&qemu_object_opts, optarg, true)) {
4257                 ret = -1;
4258                 goto out;
4259             }
4260             break;
4261         case OPTION_IMAGE_OPTS:
4262             image_opts = true;
4263             break;
4264         }
4265     }
4266 
4267     for (i = optind; i < argc; i++) {
4268         int j;
4269         arg = g_strdup(argv[i]);
4270 
4271         tmp = strchr(arg, '=');
4272         if (tmp == NULL) {
4273             error_report("unrecognized operand %s", arg);
4274             ret = -1;
4275             goto out;
4276         }
4277 
4278         *tmp++ = '\0';
4279 
4280         for (j = 0; options[j].name != NULL; j++) {
4281             if (!strcmp(arg, options[j].name)) {
4282                 break;
4283             }
4284         }
4285         if (options[j].name == NULL) {
4286             error_report("unrecognized operand %s", arg);
4287             ret = -1;
4288             goto out;
4289         }
4290 
4291         if (options[j].f(tmp, &in, &out, &dd) != 0) {
4292             ret = -1;
4293             goto out;
4294         }
4295         dd.flags |= options[j].flag;
4296         g_free(arg);
4297         arg = NULL;
4298     }
4299 
4300     if (!(dd.flags & C_IF && dd.flags & C_OF)) {
4301         error_report("Must specify both input and output files");
4302         ret = -1;
4303         goto out;
4304     }
4305 
4306     if (qemu_opts_foreach(&qemu_object_opts,
4307                           user_creatable_add_opts_foreach,
4308                           NULL, NULL)) {
4309         ret = -1;
4310         goto out;
4311     }
4312 
4313     blk1 = img_open(image_opts, in.filename, fmt, 0, false, false,
4314                     force_share);
4315 
4316     if (!blk1) {
4317         ret = -1;
4318         goto out;
4319     }
4320 
4321     drv = bdrv_find_format(out_fmt);
4322     if (!drv) {
4323         error_report("Unknown file format");
4324         ret = -1;
4325         goto out;
4326     }
4327     proto_drv = bdrv_find_protocol(out.filename, true, &local_err);
4328 
4329     if (!proto_drv) {
4330         error_report_err(local_err);
4331         ret = -1;
4332         goto out;
4333     }
4334     if (!drv->create_opts) {
4335         error_report("Format driver '%s' does not support image creation",
4336                      drv->format_name);
4337         ret = -1;
4338         goto out;
4339     }
4340     if (!proto_drv->create_opts) {
4341         error_report("Protocol driver '%s' does not support image creation",
4342                      proto_drv->format_name);
4343         ret = -1;
4344         goto out;
4345     }
4346     create_opts = qemu_opts_append(create_opts, drv->create_opts);
4347     create_opts = qemu_opts_append(create_opts, proto_drv->create_opts);
4348 
4349     opts = qemu_opts_create(create_opts, NULL, 0, &error_abort);
4350 
4351     size = blk_getlength(blk1);
4352     if (size < 0) {
4353         error_report("Failed to get size for '%s'", in.filename);
4354         ret = -1;
4355         goto out;
4356     }
4357 
4358     if (dd.flags & C_COUNT && dd.count <= INT64_MAX / in.bsz &&
4359         dd.count * in.bsz < size) {
4360         size = dd.count * in.bsz;
4361     }
4362 
4363     /* Overflow means the specified offset is beyond input image's size */
4364     if (dd.flags & C_SKIP && (in.offset > INT64_MAX / in.bsz ||
4365                               size < in.bsz * in.offset)) {
4366         qemu_opt_set_number(opts, BLOCK_OPT_SIZE, 0, &error_abort);
4367     } else {
4368         qemu_opt_set_number(opts, BLOCK_OPT_SIZE,
4369                             size - in.bsz * in.offset, &error_abort);
4370     }
4371 
4372     ret = bdrv_create(drv, out.filename, opts, &local_err);
4373     if (ret < 0) {
4374         error_reportf_err(local_err,
4375                           "%s: error while creating output image: ",
4376                           out.filename);
4377         ret = -1;
4378         goto out;
4379     }
4380 
4381     /* TODO, we can't honour --image-opts for the target,
4382      * since it needs to be given in a format compatible
4383      * with the bdrv_create() call above which does not
4384      * support image-opts style.
4385      */
4386     blk2 = img_open_file(out.filename, NULL, out_fmt, BDRV_O_RDWR,
4387                          false, false, false);
4388 
4389     if (!blk2) {
4390         ret = -1;
4391         goto out;
4392     }
4393 
4394     if (dd.flags & C_SKIP && (in.offset > INT64_MAX / in.bsz ||
4395                               size < in.offset * in.bsz)) {
4396         /* We give a warning if the skip option is bigger than the input
4397          * size and create an empty output disk image (i.e. like dd(1)).
4398          */
4399         error_report("%s: cannot skip to specified offset", in.filename);
4400         in_pos = size;
4401     } else {
4402         in_pos = in.offset * in.bsz;
4403     }
4404 
4405     in.buf = g_new(uint8_t, in.bsz);
4406 
4407     for (out_pos = 0; in_pos < size; block_count++) {
4408         int in_ret, out_ret;
4409 
4410         if (in_pos + in.bsz > size) {
4411             in_ret = blk_pread(blk1, in_pos, in.buf, size - in_pos);
4412         } else {
4413             in_ret = blk_pread(blk1, in_pos, in.buf, in.bsz);
4414         }
4415         if (in_ret < 0) {
4416             error_report("error while reading from input image file: %s",
4417                          strerror(-in_ret));
4418             ret = -1;
4419             goto out;
4420         }
4421         in_pos += in_ret;
4422 
4423         out_ret = blk_pwrite(blk2, out_pos, in.buf, in_ret, 0);
4424 
4425         if (out_ret < 0) {
4426             error_report("error while writing to output image file: %s",
4427                          strerror(-out_ret));
4428             ret = -1;
4429             goto out;
4430         }
4431         out_pos += out_ret;
4432     }
4433 
4434 out:
4435     g_free(arg);
4436     qemu_opts_del(opts);
4437     qemu_opts_free(create_opts);
4438     blk_unref(blk1);
4439     blk_unref(blk2);
4440     g_free(in.filename);
4441     g_free(out.filename);
4442     g_free(in.buf);
4443     g_free(out.buf);
4444 
4445     if (ret) {
4446         return 1;
4447     }
4448     return 0;
4449 }
4450 
4451 
4452 static const img_cmd_t img_cmds[] = {
4453 #define DEF(option, callback, arg_string)        \
4454     { option, callback },
4455 #include "qemu-img-cmds.h"
4456 #undef DEF
4457 #undef GEN_DOCS
4458     { NULL, NULL, },
4459 };
4460 
4461 int main(int argc, char **argv)
4462 {
4463     const img_cmd_t *cmd;
4464     const char *cmdname;
4465     Error *local_error = NULL;
4466     char *trace_file = NULL;
4467     int c;
4468     static const struct option long_options[] = {
4469         {"help", no_argument, 0, 'h'},
4470         {"version", no_argument, 0, 'V'},
4471         {"trace", required_argument, NULL, 'T'},
4472         {0, 0, 0, 0}
4473     };
4474 
4475 #ifdef CONFIG_POSIX
4476     signal(SIGPIPE, SIG_IGN);
4477 #endif
4478 
4479     module_call_init(MODULE_INIT_TRACE);
4480     error_set_progname(argv[0]);
4481     qemu_init_exec_dir(argv[0]);
4482 
4483     if (qemu_init_main_loop(&local_error)) {
4484         error_report_err(local_error);
4485         exit(EXIT_FAILURE);
4486     }
4487 
4488     qcrypto_init(&error_fatal);
4489 
4490     module_call_init(MODULE_INIT_QOM);
4491     bdrv_init();
4492     if (argc < 2) {
4493         error_exit("Not enough arguments");
4494     }
4495 
4496     qemu_add_opts(&qemu_object_opts);
4497     qemu_add_opts(&qemu_source_opts);
4498     qemu_add_opts(&qemu_trace_opts);
4499 
4500     while ((c = getopt_long(argc, argv, "+:hVT:", long_options, NULL)) != -1) {
4501         switch (c) {
4502         case ':':
4503             missing_argument(argv[optind - 1]);
4504             return 0;
4505         case '?':
4506             unrecognized_option(argv[optind - 1]);
4507             return 0;
4508         case 'h':
4509             help();
4510             return 0;
4511         case 'V':
4512             printf(QEMU_IMG_VERSION);
4513             return 0;
4514         case 'T':
4515             g_free(trace_file);
4516             trace_file = trace_opt_parse(optarg);
4517             break;
4518         }
4519     }
4520 
4521     cmdname = argv[optind];
4522 
4523     /* reset getopt_long scanning */
4524     argc -= optind;
4525     if (argc < 1) {
4526         return 0;
4527     }
4528     argv += optind;
4529     optind = 0;
4530 
4531     if (!trace_init_backends()) {
4532         exit(1);
4533     }
4534     trace_init_file(trace_file);
4535     qemu_set_log(LOG_TRACE);
4536 
4537     /* find the command */
4538     for (cmd = img_cmds; cmd->name != NULL; cmd++) {
4539         if (!strcmp(cmdname, cmd->name)) {
4540             return cmd->handler(argc, argv);
4541         }
4542     }
4543 
4544     /* not found */
4545     error_exit("Command not found: %s", cmdname);
4546 }
4547