xref: /openbmc/qemu/qemu-img.c (revision a4d50b1d)
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/qmp-output-visitor.h"
29 #include "qapi/qmp/qerror.h"
30 #include "qapi/qmp/qjson.h"
31 #include "qemu/cutils.h"
32 #include "qemu/config-file.h"
33 #include "qemu/option.h"
34 #include "qemu/error-report.h"
35 #include "qemu/log.h"
36 #include "qom/object_interfaces.h"
37 #include "sysemu/sysemu.h"
38 #include "sysemu/block-backend.h"
39 #include "block/block_int.h"
40 #include "block/blockjob.h"
41 #include "block/qapi.h"
42 #include "crypto/init.h"
43 #include "trace/control.h"
44 #include <getopt.h>
45 
46 #define QEMU_IMG_VERSION "qemu-img version " QEMU_VERSION QEMU_PKGVERSION \
47                           ", " QEMU_COPYRIGHT "\n"
48 
49 typedef struct img_cmd_t {
50     const char *name;
51     int (*handler)(int argc, char **argv);
52 } img_cmd_t;
53 
54 enum {
55     OPTION_OUTPUT = 256,
56     OPTION_BACKING_CHAIN = 257,
57     OPTION_OBJECT = 258,
58     OPTION_IMAGE_OPTS = 259,
59     OPTION_PATTERN = 260,
60     OPTION_FLUSH_INTERVAL = 261,
61     OPTION_NO_DRAIN = 262,
62 };
63 
64 typedef enum OutputFormat {
65     OFORMAT_JSON,
66     OFORMAT_HUMAN,
67 } OutputFormat;
68 
69 /* Default to cache=writeback as data integrity is not important for qemu-img */
70 #define BDRV_DEFAULT_CACHE "writeback"
71 
72 static void format_print(void *opaque, const char *name)
73 {
74     printf(" %s", name);
75 }
76 
77 static void QEMU_NORETURN GCC_FMT_ATTR(1, 2) error_exit(const char *fmt, ...)
78 {
79     va_list ap;
80 
81     error_printf("qemu-img: ");
82 
83     va_start(ap, fmt);
84     error_vprintf(fmt, ap);
85     va_end(ap);
86 
87     error_printf("\nTry 'qemu-img --help' for more information\n");
88     exit(EXIT_FAILURE);
89 }
90 
91 /* Please keep in synch with qemu-img.texi */
92 static void QEMU_NORETURN help(void)
93 {
94     const char *help_msg =
95            QEMU_IMG_VERSION
96            "usage: qemu-img [standard options] command [command options]\n"
97            "QEMU disk image utility\n"
98            "\n"
99            "    '-h', '--help'       display this help and exit\n"
100            "    '-V', '--version'    output version information and exit\n"
101            "    '-T', '--trace'      [[enable=]<pattern>][,events=<file>][,file=<file>]\n"
102            "                         specify tracing options\n"
103            "\n"
104            "Command syntax:\n"
105 #define DEF(option, callback, arg_string)        \
106            "  " arg_string "\n"
107 #include "qemu-img-cmds.h"
108 #undef DEF
109 #undef GEN_DOCS
110            "\n"
111            "Command parameters:\n"
112            "  'filename' is a disk image filename\n"
113            "  'objectdef' is a QEMU user creatable object definition. See the qemu(1)\n"
114            "    manual page for a description of the object properties. The most common\n"
115            "    object type is a 'secret', which is used to supply passwords and/or\n"
116            "    encryption keys.\n"
117            "  'fmt' is the disk image format. It is guessed automatically in most cases\n"
118            "  'cache' is the cache mode used to write the output disk image, the valid\n"
119            "    options are: 'none', 'writeback' (default, except for convert), 'writethrough',\n"
120            "    'directsync' and 'unsafe' (default for convert)\n"
121            "  'src_cache' is the cache mode used to read input disk images, the valid\n"
122            "    options are the same as for the 'cache' option\n"
123            "  'size' is the disk image size in bytes. Optional suffixes\n"
124            "    'k' or 'K' (kilobyte, 1024), 'M' (megabyte, 1024k), 'G' (gigabyte, 1024M),\n"
125            "    'T' (terabyte, 1024G), 'P' (petabyte, 1024T) and 'E' (exabyte, 1024P)  are\n"
126            "    supported. 'b' is ignored.\n"
127            "  'output_filename' is the destination disk image filename\n"
128            "  'output_fmt' is the destination format\n"
129            "  'options' is a comma separated list of format specific options in a\n"
130            "    name=value format. Use -o ? for an overview of the options supported by the\n"
131            "    used format\n"
132            "  'snapshot_param' is param used for internal snapshot, format\n"
133            "    is 'snapshot.id=[ID],snapshot.name=[NAME]', or\n"
134            "    '[ID_OR_NAME]'\n"
135            "  'snapshot_id_or_name' is deprecated, use 'snapshot_param'\n"
136            "    instead\n"
137            "  '-c' indicates that target image must be compressed (qcow format only)\n"
138            "  '-u' enables unsafe rebasing. It is assumed that old and new backing file\n"
139            "       match exactly. The image doesn't need a working backing file before\n"
140            "       rebasing in this case (useful for renaming the backing file)\n"
141            "  '-h' with or without a command shows this help and lists the supported formats\n"
142            "  '-p' show progress of command (only certain commands)\n"
143            "  '-q' use Quiet mode - do not print any output (except errors)\n"
144            "  '-S' indicates the consecutive number of bytes (defaults to 4k) that must\n"
145            "       contain only zeros for qemu-img to create a sparse image during\n"
146            "       conversion. If the number of bytes is 0, the source will not be scanned for\n"
147            "       unallocated or zero sectors, and the destination image will always be\n"
148            "       fully allocated\n"
149            "  '--output' takes the format in which the output must be done (human or json)\n"
150            "  '-n' skips the target volume creation (useful if the volume is created\n"
151            "       prior to running qemu-img)\n"
152            "\n"
153            "Parameters to check subcommand:\n"
154            "  '-r' tries to repair any inconsistencies that are found during the check.\n"
155            "       '-r leaks' repairs only cluster leaks, whereas '-r all' fixes all\n"
156            "       kinds of errors, with a higher risk of choosing the wrong fix or\n"
157            "       hiding corruption that has already occurred.\n"
158            "\n"
159            "Parameters to snapshot subcommand:\n"
160            "  'snapshot' is the name of the snapshot to create, apply or delete\n"
161            "  '-a' applies a snapshot (revert disk to saved state)\n"
162            "  '-c' creates a snapshot\n"
163            "  '-d' deletes a snapshot\n"
164            "  '-l' lists all snapshots in the given image\n"
165            "\n"
166            "Parameters to compare subcommand:\n"
167            "  '-f' first image format\n"
168            "  '-F' second image format\n"
169            "  '-s' run in Strict mode - fail on different image size or sector allocation\n";
170 
171     printf("%s\nSupported formats:", help_msg);
172     bdrv_iterate_format(format_print, NULL);
173     printf("\n");
174     exit(EXIT_SUCCESS);
175 }
176 
177 static QemuOptsList qemu_object_opts = {
178     .name = "object",
179     .implied_opt_name = "qom-type",
180     .head = QTAILQ_HEAD_INITIALIZER(qemu_object_opts.head),
181     .desc = {
182         { }
183     },
184 };
185 
186 static QemuOptsList qemu_source_opts = {
187     .name = "source",
188     .implied_opt_name = "file",
189     .head = QTAILQ_HEAD_INITIALIZER(qemu_source_opts.head),
190     .desc = {
191         { }
192     },
193 };
194 
195 static int GCC_FMT_ATTR(2, 3) qprintf(bool quiet, const char *fmt, ...)
196 {
197     int ret = 0;
198     if (!quiet) {
199         va_list args;
200         va_start(args, fmt);
201         ret = vprintf(fmt, args);
202         va_end(args);
203     }
204     return ret;
205 }
206 
207 
208 static int print_block_option_help(const char *filename, const char *fmt)
209 {
210     BlockDriver *drv, *proto_drv;
211     QemuOptsList *create_opts = NULL;
212     Error *local_err = NULL;
213 
214     /* Find driver and parse its options */
215     drv = bdrv_find_format(fmt);
216     if (!drv) {
217         error_report("Unknown file format '%s'", fmt);
218         return 1;
219     }
220 
221     create_opts = qemu_opts_append(create_opts, drv->create_opts);
222     if (filename) {
223         proto_drv = bdrv_find_protocol(filename, true, &local_err);
224         if (!proto_drv) {
225             error_report_err(local_err);
226             qemu_opts_free(create_opts);
227             return 1;
228         }
229         create_opts = qemu_opts_append(create_opts, proto_drv->create_opts);
230     }
231 
232     qemu_opts_print_help(create_opts);
233     qemu_opts_free(create_opts);
234     return 0;
235 }
236 
237 
238 static int img_open_password(BlockBackend *blk, const char *filename,
239                              int flags, bool quiet)
240 {
241     BlockDriverState *bs;
242     char password[256];
243 
244     bs = blk_bs(blk);
245     if (bdrv_is_encrypted(bs) && bdrv_key_required(bs) &&
246         !(flags & BDRV_O_NO_IO)) {
247         qprintf(quiet, "Disk image '%s' is encrypted.\n", filename);
248         if (qemu_read_password(password, sizeof(password)) < 0) {
249             error_report("No password given");
250             return -1;
251         }
252         if (bdrv_set_key(bs, password) < 0) {
253             error_report("invalid password");
254             return -1;
255         }
256     }
257     return 0;
258 }
259 
260 
261 static BlockBackend *img_open_opts(const char *optstr,
262                                    QemuOpts *opts, int flags, bool writethrough,
263                                    bool quiet)
264 {
265     QDict *options;
266     Error *local_err = NULL;
267     BlockBackend *blk;
268     options = qemu_opts_to_qdict(opts, NULL);
269     blk = blk_new_open(NULL, NULL, options, flags, &local_err);
270     if (!blk) {
271         error_reportf_err(local_err, "Could not open '%s': ", optstr);
272         return NULL;
273     }
274     blk_set_enable_write_cache(blk, !writethrough);
275 
276     if (img_open_password(blk, optstr, flags, quiet) < 0) {
277         blk_unref(blk);
278         return NULL;
279     }
280     return blk;
281 }
282 
283 static BlockBackend *img_open_file(const char *filename,
284                                    const char *fmt, int flags,
285                                    bool writethrough, bool quiet)
286 {
287     BlockBackend *blk;
288     Error *local_err = NULL;
289     QDict *options = NULL;
290 
291     if (fmt) {
292         options = qdict_new();
293         qdict_put(options, "driver", qstring_from_str(fmt));
294     }
295 
296     blk = blk_new_open(filename, NULL, options, flags, &local_err);
297     if (!blk) {
298         error_reportf_err(local_err, "Could not open '%s': ", filename);
299         return NULL;
300     }
301     blk_set_enable_write_cache(blk, !writethrough);
302 
303     if (img_open_password(blk, filename, flags, quiet) < 0) {
304         blk_unref(blk);
305         return NULL;
306     }
307     return blk;
308 }
309 
310 
311 static BlockBackend *img_open(bool image_opts,
312                               const char *filename,
313                               const char *fmt, int flags, bool writethrough,
314                               bool quiet)
315 {
316     BlockBackend *blk;
317     if (image_opts) {
318         QemuOpts *opts;
319         if (fmt) {
320             error_report("--image-opts and --format are mutually exclusive");
321             return NULL;
322         }
323         opts = qemu_opts_parse_noisily(qemu_find_opts("source"),
324                                        filename, true);
325         if (!opts) {
326             return NULL;
327         }
328         blk = img_open_opts(filename, opts, flags, writethrough, quiet);
329     } else {
330         blk = img_open_file(filename, fmt, flags, writethrough, quiet);
331     }
332     return blk;
333 }
334 
335 
336 static int add_old_style_options(const char *fmt, QemuOpts *opts,
337                                  const char *base_filename,
338                                  const char *base_fmt)
339 {
340     Error *err = NULL;
341 
342     if (base_filename) {
343         qemu_opt_set(opts, BLOCK_OPT_BACKING_FILE, base_filename, &err);
344         if (err) {
345             error_report("Backing file not supported for file format '%s'",
346                          fmt);
347             error_free(err);
348             return -1;
349         }
350     }
351     if (base_fmt) {
352         qemu_opt_set(opts, BLOCK_OPT_BACKING_FMT, base_fmt, &err);
353         if (err) {
354             error_report("Backing file format not supported for file "
355                          "format '%s'", fmt);
356             error_free(err);
357             return -1;
358         }
359     }
360     return 0;
361 }
362 
363 static int img_create(int argc, char **argv)
364 {
365     int c;
366     uint64_t img_size = -1;
367     const char *fmt = "raw";
368     const char *base_fmt = NULL;
369     const char *filename;
370     const char *base_filename = NULL;
371     char *options = NULL;
372     Error *local_err = NULL;
373     bool quiet = false;
374 
375     for(;;) {
376         static const struct option long_options[] = {
377             {"help", no_argument, 0, 'h'},
378             {"object", required_argument, 0, OPTION_OBJECT},
379             {0, 0, 0, 0}
380         };
381         c = getopt_long(argc, argv, "F:b:f:he6o:q",
382                         long_options, NULL);
383         if (c == -1) {
384             break;
385         }
386         switch(c) {
387         case '?':
388         case 'h':
389             help();
390             break;
391         case 'F':
392             base_fmt = optarg;
393             break;
394         case 'b':
395             base_filename = optarg;
396             break;
397         case 'f':
398             fmt = optarg;
399             break;
400         case 'e':
401             error_report("option -e is deprecated, please use \'-o "
402                   "encryption\' instead!");
403             goto fail;
404         case '6':
405             error_report("option -6 is deprecated, please use \'-o "
406                   "compat6\' instead!");
407             goto fail;
408         case 'o':
409             if (!is_valid_option_list(optarg)) {
410                 error_report("Invalid option list: %s", optarg);
411                 goto fail;
412             }
413             if (!options) {
414                 options = g_strdup(optarg);
415             } else {
416                 char *old_options = options;
417                 options = g_strdup_printf("%s,%s", options, optarg);
418                 g_free(old_options);
419             }
420             break;
421         case 'q':
422             quiet = true;
423             break;
424         case OPTION_OBJECT: {
425             QemuOpts *opts;
426             opts = qemu_opts_parse_noisily(&qemu_object_opts,
427                                            optarg, true);
428             if (!opts) {
429                 goto fail;
430             }
431         }   break;
432         }
433     }
434 
435     /* Get the filename */
436     filename = (optind < argc) ? argv[optind] : NULL;
437     if (options && has_help_option(options)) {
438         g_free(options);
439         return print_block_option_help(filename, fmt);
440     }
441 
442     if (optind >= argc) {
443         error_exit("Expecting image file name");
444     }
445     optind++;
446 
447     if (qemu_opts_foreach(&qemu_object_opts,
448                           user_creatable_add_opts_foreach,
449                           NULL, NULL)) {
450         goto fail;
451     }
452 
453     /* Get image size, if specified */
454     if (optind < argc) {
455         int64_t sval;
456         char *end;
457         sval = qemu_strtosz_suffix(argv[optind++], &end,
458                                    QEMU_STRTOSZ_DEFSUFFIX_B);
459         if (sval < 0 || *end) {
460             if (sval == -ERANGE) {
461                 error_report("Image size must be less than 8 EiB!");
462             } else {
463                 error_report("Invalid image size specified! You may use k, M, "
464                       "G, T, P or E suffixes for ");
465                 error_report("kilobytes, megabytes, gigabytes, terabytes, "
466                              "petabytes and exabytes.");
467             }
468             goto fail;
469         }
470         img_size = (uint64_t)sval;
471     }
472     if (optind != argc) {
473         error_exit("Unexpected argument: %s", argv[optind]);
474     }
475 
476     bdrv_img_create(filename, fmt, base_filename, base_fmt,
477                     options, img_size, 0, &local_err, quiet);
478     if (local_err) {
479         error_reportf_err(local_err, "%s: ", filename);
480         goto fail;
481     }
482 
483     g_free(options);
484     return 0;
485 
486 fail:
487     g_free(options);
488     return 1;
489 }
490 
491 static void dump_json_image_check(ImageCheck *check, bool quiet)
492 {
493     QString *str;
494     QObject *obj;
495     Visitor *v = qmp_output_visitor_new(&obj);
496 
497     visit_type_ImageCheck(v, NULL, &check, &error_abort);
498     visit_complete(v, &obj);
499     str = qobject_to_json_pretty(obj);
500     assert(str != NULL);
501     qprintf(quiet, "%s\n", qstring_get_str(str));
502     qobject_decref(obj);
503     visit_free(v);
504     QDECREF(str);
505 }
506 
507 static void dump_human_image_check(ImageCheck *check, bool quiet)
508 {
509     if (!(check->corruptions || check->leaks || check->check_errors)) {
510         qprintf(quiet, "No errors were found on the image.\n");
511     } else {
512         if (check->corruptions) {
513             qprintf(quiet, "\n%" PRId64 " errors were found on the image.\n"
514                     "Data may be corrupted, or further writes to the image "
515                     "may corrupt it.\n",
516                     check->corruptions);
517         }
518 
519         if (check->leaks) {
520             qprintf(quiet,
521                     "\n%" PRId64 " leaked clusters were found on the image.\n"
522                     "This means waste of disk space, but no harm to data.\n",
523                     check->leaks);
524         }
525 
526         if (check->check_errors) {
527             qprintf(quiet,
528                     "\n%" PRId64
529                     " internal errors have occurred during the check.\n",
530                     check->check_errors);
531         }
532     }
533 
534     if (check->total_clusters != 0 && check->allocated_clusters != 0) {
535         qprintf(quiet, "%" PRId64 "/%" PRId64 " = %0.2f%% allocated, "
536                 "%0.2f%% fragmented, %0.2f%% compressed clusters\n",
537                 check->allocated_clusters, check->total_clusters,
538                 check->allocated_clusters * 100.0 / check->total_clusters,
539                 check->fragmented_clusters * 100.0 / check->allocated_clusters,
540                 check->compressed_clusters * 100.0 /
541                 check->allocated_clusters);
542     }
543 
544     if (check->image_end_offset) {
545         qprintf(quiet,
546                 "Image end offset: %" PRId64 "\n", check->image_end_offset);
547     }
548 }
549 
550 static int collect_image_check(BlockDriverState *bs,
551                    ImageCheck *check,
552                    const char *filename,
553                    const char *fmt,
554                    int fix)
555 {
556     int ret;
557     BdrvCheckResult result;
558 
559     ret = bdrv_check(bs, &result, fix);
560     if (ret < 0) {
561         return ret;
562     }
563 
564     check->filename                 = g_strdup(filename);
565     check->format                   = g_strdup(bdrv_get_format_name(bs));
566     check->check_errors             = result.check_errors;
567     check->corruptions              = result.corruptions;
568     check->has_corruptions          = result.corruptions != 0;
569     check->leaks                    = result.leaks;
570     check->has_leaks                = result.leaks != 0;
571     check->corruptions_fixed        = result.corruptions_fixed;
572     check->has_corruptions_fixed    = result.corruptions != 0;
573     check->leaks_fixed              = result.leaks_fixed;
574     check->has_leaks_fixed          = result.leaks != 0;
575     check->image_end_offset         = result.image_end_offset;
576     check->has_image_end_offset     = result.image_end_offset != 0;
577     check->total_clusters           = result.bfi.total_clusters;
578     check->has_total_clusters       = result.bfi.total_clusters != 0;
579     check->allocated_clusters       = result.bfi.allocated_clusters;
580     check->has_allocated_clusters   = result.bfi.allocated_clusters != 0;
581     check->fragmented_clusters      = result.bfi.fragmented_clusters;
582     check->has_fragmented_clusters  = result.bfi.fragmented_clusters != 0;
583     check->compressed_clusters      = result.bfi.compressed_clusters;
584     check->has_compressed_clusters  = result.bfi.compressed_clusters != 0;
585 
586     return 0;
587 }
588 
589 /*
590  * Checks an image for consistency. Exit codes:
591  *
592  *  0 - Check completed, image is good
593  *  1 - Check not completed because of internal errors
594  *  2 - Check completed, image is corrupted
595  *  3 - Check completed, image has leaked clusters, but is good otherwise
596  * 63 - Checks are not supported by the image format
597  */
598 static int img_check(int argc, char **argv)
599 {
600     int c, ret;
601     OutputFormat output_format = OFORMAT_HUMAN;
602     const char *filename, *fmt, *output, *cache;
603     BlockBackend *blk;
604     BlockDriverState *bs;
605     int fix = 0;
606     int flags = BDRV_O_CHECK;
607     bool writethrough;
608     ImageCheck *check;
609     bool quiet = false;
610     bool image_opts = false;
611 
612     fmt = NULL;
613     output = NULL;
614     cache = BDRV_DEFAULT_CACHE;
615 
616     for(;;) {
617         int option_index = 0;
618         static const struct option long_options[] = {
619             {"help", no_argument, 0, 'h'},
620             {"format", required_argument, 0, 'f'},
621             {"repair", required_argument, 0, 'r'},
622             {"output", required_argument, 0, OPTION_OUTPUT},
623             {"object", required_argument, 0, OPTION_OBJECT},
624             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
625             {0, 0, 0, 0}
626         };
627         c = getopt_long(argc, argv, "hf:r:T:q",
628                         long_options, &option_index);
629         if (c == -1) {
630             break;
631         }
632         switch(c) {
633         case '?':
634         case 'h':
635             help();
636             break;
637         case 'f':
638             fmt = optarg;
639             break;
640         case 'r':
641             flags |= BDRV_O_RDWR;
642 
643             if (!strcmp(optarg, "leaks")) {
644                 fix = BDRV_FIX_LEAKS;
645             } else if (!strcmp(optarg, "all")) {
646                 fix = BDRV_FIX_LEAKS | BDRV_FIX_ERRORS;
647             } else {
648                 error_exit("Unknown option value for -r "
649                            "(expecting 'leaks' or 'all'): %s", optarg);
650             }
651             break;
652         case OPTION_OUTPUT:
653             output = optarg;
654             break;
655         case 'T':
656             cache = optarg;
657             break;
658         case 'q':
659             quiet = true;
660             break;
661         case OPTION_OBJECT: {
662             QemuOpts *opts;
663             opts = qemu_opts_parse_noisily(&qemu_object_opts,
664                                            optarg, true);
665             if (!opts) {
666                 return 1;
667             }
668         }   break;
669         case OPTION_IMAGE_OPTS:
670             image_opts = true;
671             break;
672         }
673     }
674     if (optind != argc - 1) {
675         error_exit("Expecting one image file name");
676     }
677     filename = argv[optind++];
678 
679     if (output && !strcmp(output, "json")) {
680         output_format = OFORMAT_JSON;
681     } else if (output && !strcmp(output, "human")) {
682         output_format = OFORMAT_HUMAN;
683     } else if (output) {
684         error_report("--output must be used with human or json as argument.");
685         return 1;
686     }
687 
688     if (qemu_opts_foreach(&qemu_object_opts,
689                           user_creatable_add_opts_foreach,
690                           NULL, NULL)) {
691         return 1;
692     }
693 
694     ret = bdrv_parse_cache_mode(cache, &flags, &writethrough);
695     if (ret < 0) {
696         error_report("Invalid source cache option: %s", cache);
697         return 1;
698     }
699 
700     blk = img_open(image_opts, filename, fmt, flags, writethrough, quiet);
701     if (!blk) {
702         return 1;
703     }
704     bs = blk_bs(blk);
705 
706     check = g_new0(ImageCheck, 1);
707     ret = collect_image_check(bs, check, filename, fmt, fix);
708 
709     if (ret == -ENOTSUP) {
710         error_report("This image format does not support checks");
711         ret = 63;
712         goto fail;
713     }
714 
715     if (check->corruptions_fixed || check->leaks_fixed) {
716         int corruptions_fixed, leaks_fixed;
717 
718         leaks_fixed         = check->leaks_fixed;
719         corruptions_fixed   = check->corruptions_fixed;
720 
721         if (output_format == OFORMAT_HUMAN) {
722             qprintf(quiet,
723                     "The following inconsistencies were found and repaired:\n\n"
724                     "    %" PRId64 " leaked clusters\n"
725                     "    %" PRId64 " corruptions\n\n"
726                     "Double checking the fixed image now...\n",
727                     check->leaks_fixed,
728                     check->corruptions_fixed);
729         }
730 
731         ret = collect_image_check(bs, check, filename, fmt, 0);
732 
733         check->leaks_fixed          = leaks_fixed;
734         check->corruptions_fixed    = corruptions_fixed;
735     }
736 
737     if (!ret) {
738         switch (output_format) {
739         case OFORMAT_HUMAN:
740             dump_human_image_check(check, quiet);
741             break;
742         case OFORMAT_JSON:
743             dump_json_image_check(check, quiet);
744             break;
745         }
746     }
747 
748     if (ret || check->check_errors) {
749         if (ret) {
750             error_report("Check failed: %s", strerror(-ret));
751         } else {
752             error_report("Check failed");
753         }
754         ret = 1;
755         goto fail;
756     }
757 
758     if (check->corruptions) {
759         ret = 2;
760     } else if (check->leaks) {
761         ret = 3;
762     } else {
763         ret = 0;
764     }
765 
766 fail:
767     qapi_free_ImageCheck(check);
768     blk_unref(blk);
769     return ret;
770 }
771 
772 typedef struct CommonBlockJobCBInfo {
773     BlockDriverState *bs;
774     Error **errp;
775 } CommonBlockJobCBInfo;
776 
777 static void common_block_job_cb(void *opaque, int ret)
778 {
779     CommonBlockJobCBInfo *cbi = opaque;
780 
781     if (ret < 0) {
782         error_setg_errno(cbi->errp, -ret, "Block job failed");
783     }
784 }
785 
786 static void run_block_job(BlockJob *job, Error **errp)
787 {
788     AioContext *aio_context = blk_get_aio_context(job->blk);
789 
790     do {
791         aio_poll(aio_context, true);
792         qemu_progress_print(job->len ?
793                             ((float)job->offset / job->len * 100.f) : 0.0f, 0);
794     } while (!job->ready);
795 
796     block_job_complete_sync(job, errp);
797 
798     /* A block job may finish instantaneously without publishing any progress,
799      * so just signal completion here */
800     qemu_progress_print(100.f, 0);
801 }
802 
803 static int img_commit(int argc, char **argv)
804 {
805     int c, ret, flags;
806     const char *filename, *fmt, *cache, *base;
807     BlockBackend *blk;
808     BlockDriverState *bs, *base_bs;
809     bool progress = false, quiet = false, drop = false;
810     bool writethrough;
811     Error *local_err = NULL;
812     CommonBlockJobCBInfo cbi;
813     bool image_opts = false;
814 
815     fmt = NULL;
816     cache = BDRV_DEFAULT_CACHE;
817     base = NULL;
818     for(;;) {
819         static const struct option long_options[] = {
820             {"help", no_argument, 0, 'h'},
821             {"object", required_argument, 0, OPTION_OBJECT},
822             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
823             {0, 0, 0, 0}
824         };
825         c = getopt_long(argc, argv, "f:ht:b:dpq",
826                         long_options, NULL);
827         if (c == -1) {
828             break;
829         }
830         switch(c) {
831         case '?':
832         case 'h':
833             help();
834             break;
835         case 'f':
836             fmt = optarg;
837             break;
838         case 't':
839             cache = optarg;
840             break;
841         case 'b':
842             base = optarg;
843             /* -b implies -d */
844             drop = true;
845             break;
846         case 'd':
847             drop = true;
848             break;
849         case 'p':
850             progress = true;
851             break;
852         case 'q':
853             quiet = true;
854             break;
855         case OPTION_OBJECT: {
856             QemuOpts *opts;
857             opts = qemu_opts_parse_noisily(&qemu_object_opts,
858                                            optarg, true);
859             if (!opts) {
860                 return 1;
861             }
862         }   break;
863         case OPTION_IMAGE_OPTS:
864             image_opts = true;
865             break;
866         }
867     }
868 
869     /* Progress is not shown in Quiet mode */
870     if (quiet) {
871         progress = false;
872     }
873 
874     if (optind != argc - 1) {
875         error_exit("Expecting one image file name");
876     }
877     filename = argv[optind++];
878 
879     if (qemu_opts_foreach(&qemu_object_opts,
880                           user_creatable_add_opts_foreach,
881                           NULL, NULL)) {
882         return 1;
883     }
884 
885     flags = BDRV_O_RDWR | BDRV_O_UNMAP;
886     ret = bdrv_parse_cache_mode(cache, &flags, &writethrough);
887     if (ret < 0) {
888         error_report("Invalid cache option: %s", cache);
889         return 1;
890     }
891 
892     blk = img_open(image_opts, filename, fmt, flags, writethrough, quiet);
893     if (!blk) {
894         return 1;
895     }
896     bs = blk_bs(blk);
897 
898     qemu_progress_init(progress, 1.f);
899     qemu_progress_print(0.f, 100);
900 
901     if (base) {
902         base_bs = bdrv_find_backing_image(bs, base);
903         if (!base_bs) {
904             error_setg(&local_err, QERR_BASE_NOT_FOUND, base);
905             goto done;
906         }
907     } else {
908         /* This is different from QMP, which by default uses the deepest file in
909          * the backing chain (i.e., the very base); however, the traditional
910          * behavior of qemu-img commit is using the immediate backing file. */
911         base_bs = backing_bs(bs);
912         if (!base_bs) {
913             error_setg(&local_err, "Image does not have a backing file");
914             goto done;
915         }
916     }
917 
918     cbi = (CommonBlockJobCBInfo){
919         .errp = &local_err,
920         .bs   = bs,
921     };
922 
923     commit_active_start("commit", bs, base_bs, 0, BLOCKDEV_ON_ERROR_REPORT,
924                         common_block_job_cb, &cbi, &local_err);
925     if (local_err) {
926         goto done;
927     }
928 
929     /* When the block job completes, the BlockBackend reference will point to
930      * the old backing file. In order to avoid that the top image is already
931      * deleted, so we can still empty it afterwards, increment the reference
932      * counter here preemptively. */
933     if (!drop) {
934         bdrv_ref(bs);
935     }
936 
937     run_block_job(bs->job, &local_err);
938     if (local_err) {
939         goto unref_backing;
940     }
941 
942     if (!drop && bs->drv->bdrv_make_empty) {
943         ret = bs->drv->bdrv_make_empty(bs);
944         if (ret) {
945             error_setg_errno(&local_err, -ret, "Could not empty %s",
946                              filename);
947             goto unref_backing;
948         }
949     }
950 
951 unref_backing:
952     if (!drop) {
953         bdrv_unref(bs);
954     }
955 
956 done:
957     qemu_progress_end();
958 
959     blk_unref(blk);
960 
961     if (local_err) {
962         error_report_err(local_err);
963         return 1;
964     }
965 
966     qprintf(quiet, "Image committed.\n");
967     return 0;
968 }
969 
970 /*
971  * Returns true iff the first sector pointed to by 'buf' contains at least
972  * a non-NUL byte.
973  *
974  * 'pnum' is set to the number of sectors (including and immediately following
975  * the first one) that are known to be in the same allocated/unallocated state.
976  */
977 static int is_allocated_sectors(const uint8_t *buf, int n, int *pnum)
978 {
979     bool is_zero;
980     int i;
981 
982     if (n <= 0) {
983         *pnum = 0;
984         return 0;
985     }
986     is_zero = buffer_is_zero(buf, 512);
987     for(i = 1; i < n; i++) {
988         buf += 512;
989         if (is_zero != buffer_is_zero(buf, 512)) {
990             break;
991         }
992     }
993     *pnum = i;
994     return !is_zero;
995 }
996 
997 /*
998  * Like is_allocated_sectors, but if the buffer starts with a used sector,
999  * up to 'min' consecutive sectors containing zeros are ignored. This avoids
1000  * breaking up write requests for only small sparse areas.
1001  */
1002 static int is_allocated_sectors_min(const uint8_t *buf, int n, int *pnum,
1003     int min)
1004 {
1005     int ret;
1006     int num_checked, num_used;
1007 
1008     if (n < min) {
1009         min = n;
1010     }
1011 
1012     ret = is_allocated_sectors(buf, n, pnum);
1013     if (!ret) {
1014         return ret;
1015     }
1016 
1017     num_used = *pnum;
1018     buf += BDRV_SECTOR_SIZE * *pnum;
1019     n -= *pnum;
1020     num_checked = num_used;
1021 
1022     while (n > 0) {
1023         ret = is_allocated_sectors(buf, n, pnum);
1024 
1025         buf += BDRV_SECTOR_SIZE * *pnum;
1026         n -= *pnum;
1027         num_checked += *pnum;
1028         if (ret) {
1029             num_used = num_checked;
1030         } else if (*pnum >= min) {
1031             break;
1032         }
1033     }
1034 
1035     *pnum = num_used;
1036     return 1;
1037 }
1038 
1039 /*
1040  * Compares two buffers sector by sector. Returns 0 if the first sector of both
1041  * buffers matches, non-zero otherwise.
1042  *
1043  * pnum is set to the number of sectors (including and immediately following
1044  * the first one) that are known to have the same comparison result
1045  */
1046 static int compare_sectors(const uint8_t *buf1, const uint8_t *buf2, int n,
1047     int *pnum)
1048 {
1049     bool res;
1050     int i;
1051 
1052     if (n <= 0) {
1053         *pnum = 0;
1054         return 0;
1055     }
1056 
1057     res = !!memcmp(buf1, buf2, 512);
1058     for(i = 1; i < n; i++) {
1059         buf1 += 512;
1060         buf2 += 512;
1061 
1062         if (!!memcmp(buf1, buf2, 512) != res) {
1063             break;
1064         }
1065     }
1066 
1067     *pnum = i;
1068     return res;
1069 }
1070 
1071 #define IO_BUF_SIZE (2 * 1024 * 1024)
1072 
1073 static int64_t sectors_to_bytes(int64_t sectors)
1074 {
1075     return sectors << BDRV_SECTOR_BITS;
1076 }
1077 
1078 static int64_t sectors_to_process(int64_t total, int64_t from)
1079 {
1080     return MIN(total - from, IO_BUF_SIZE >> BDRV_SECTOR_BITS);
1081 }
1082 
1083 /*
1084  * Check if passed sectors are empty (not allocated or contain only 0 bytes)
1085  *
1086  * Returns 0 in case sectors are filled with 0, 1 if sectors contain non-zero
1087  * data and negative value on error.
1088  *
1089  * @param blk:  BlockBackend for the image
1090  * @param sect_num: Number of first sector to check
1091  * @param sect_count: Number of sectors to check
1092  * @param filename: Name of disk file we are checking (logging purpose)
1093  * @param buffer: Allocated buffer for storing read data
1094  * @param quiet: Flag for quiet mode
1095  */
1096 static int check_empty_sectors(BlockBackend *blk, int64_t sect_num,
1097                                int sect_count, const char *filename,
1098                                uint8_t *buffer, bool quiet)
1099 {
1100     int pnum, ret = 0;
1101     ret = blk_pread(blk, sect_num << BDRV_SECTOR_BITS, buffer,
1102                     sect_count << BDRV_SECTOR_BITS);
1103     if (ret < 0) {
1104         error_report("Error while reading offset %" PRId64 " of %s: %s",
1105                      sectors_to_bytes(sect_num), filename, strerror(-ret));
1106         return ret;
1107     }
1108     ret = is_allocated_sectors(buffer, sect_count, &pnum);
1109     if (ret || pnum != sect_count) {
1110         qprintf(quiet, "Content mismatch at offset %" PRId64 "!\n",
1111                 sectors_to_bytes(ret ? sect_num : sect_num + pnum));
1112         return 1;
1113     }
1114 
1115     return 0;
1116 }
1117 
1118 /*
1119  * Compares two images. Exit codes:
1120  *
1121  * 0 - Images are identical
1122  * 1 - Images differ
1123  * >1 - Error occurred
1124  */
1125 static int img_compare(int argc, char **argv)
1126 {
1127     const char *fmt1 = NULL, *fmt2 = NULL, *cache, *filename1, *filename2;
1128     BlockBackend *blk1, *blk2;
1129     BlockDriverState *bs1, *bs2;
1130     int64_t total_sectors1, total_sectors2;
1131     uint8_t *buf1 = NULL, *buf2 = NULL;
1132     int pnum1, pnum2;
1133     int allocated1, allocated2;
1134     int ret = 0; /* return value - 0 Ident, 1 Different, >1 Error */
1135     bool progress = false, quiet = false, strict = false;
1136     int flags;
1137     bool writethrough;
1138     int64_t total_sectors;
1139     int64_t sector_num = 0;
1140     int64_t nb_sectors;
1141     int c, pnum;
1142     uint64_t progress_base;
1143     bool image_opts = false;
1144 
1145     cache = BDRV_DEFAULT_CACHE;
1146     for (;;) {
1147         static const struct option long_options[] = {
1148             {"help", no_argument, 0, 'h'},
1149             {"object", required_argument, 0, OPTION_OBJECT},
1150             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
1151             {0, 0, 0, 0}
1152         };
1153         c = getopt_long(argc, argv, "hf:F:T:pqs",
1154                         long_options, NULL);
1155         if (c == -1) {
1156             break;
1157         }
1158         switch (c) {
1159         case '?':
1160         case 'h':
1161             help();
1162             break;
1163         case 'f':
1164             fmt1 = optarg;
1165             break;
1166         case 'F':
1167             fmt2 = optarg;
1168             break;
1169         case 'T':
1170             cache = optarg;
1171             break;
1172         case 'p':
1173             progress = true;
1174             break;
1175         case 'q':
1176             quiet = true;
1177             break;
1178         case 's':
1179             strict = true;
1180             break;
1181         case OPTION_OBJECT: {
1182             QemuOpts *opts;
1183             opts = qemu_opts_parse_noisily(&qemu_object_opts,
1184                                            optarg, true);
1185             if (!opts) {
1186                 ret = 2;
1187                 goto out4;
1188             }
1189         }   break;
1190         case OPTION_IMAGE_OPTS:
1191             image_opts = true;
1192             break;
1193         }
1194     }
1195 
1196     /* Progress is not shown in Quiet mode */
1197     if (quiet) {
1198         progress = false;
1199     }
1200 
1201 
1202     if (optind != argc - 2) {
1203         error_exit("Expecting two image file names");
1204     }
1205     filename1 = argv[optind++];
1206     filename2 = argv[optind++];
1207 
1208     if (qemu_opts_foreach(&qemu_object_opts,
1209                           user_creatable_add_opts_foreach,
1210                           NULL, NULL)) {
1211         ret = 2;
1212         goto out4;
1213     }
1214 
1215     /* Initialize before goto out */
1216     qemu_progress_init(progress, 2.0);
1217 
1218     flags = 0;
1219     ret = bdrv_parse_cache_mode(cache, &flags, &writethrough);
1220     if (ret < 0) {
1221         error_report("Invalid source cache option: %s", cache);
1222         ret = 2;
1223         goto out3;
1224     }
1225 
1226     blk1 = img_open(image_opts, filename1, fmt1, flags, writethrough, quiet);
1227     if (!blk1) {
1228         ret = 2;
1229         goto out3;
1230     }
1231 
1232     blk2 = img_open(image_opts, filename2, fmt2, flags, writethrough, quiet);
1233     if (!blk2) {
1234         ret = 2;
1235         goto out2;
1236     }
1237     bs1 = blk_bs(blk1);
1238     bs2 = blk_bs(blk2);
1239 
1240     buf1 = blk_blockalign(blk1, IO_BUF_SIZE);
1241     buf2 = blk_blockalign(blk2, IO_BUF_SIZE);
1242     total_sectors1 = blk_nb_sectors(blk1);
1243     if (total_sectors1 < 0) {
1244         error_report("Can't get size of %s: %s",
1245                      filename1, strerror(-total_sectors1));
1246         ret = 4;
1247         goto out;
1248     }
1249     total_sectors2 = blk_nb_sectors(blk2);
1250     if (total_sectors2 < 0) {
1251         error_report("Can't get size of %s: %s",
1252                      filename2, strerror(-total_sectors2));
1253         ret = 4;
1254         goto out;
1255     }
1256     total_sectors = MIN(total_sectors1, total_sectors2);
1257     progress_base = MAX(total_sectors1, total_sectors2);
1258 
1259     qemu_progress_print(0, 100);
1260 
1261     if (strict && total_sectors1 != total_sectors2) {
1262         ret = 1;
1263         qprintf(quiet, "Strict mode: Image size mismatch!\n");
1264         goto out;
1265     }
1266 
1267     for (;;) {
1268         int64_t status1, status2;
1269         BlockDriverState *file;
1270 
1271         nb_sectors = sectors_to_process(total_sectors, sector_num);
1272         if (nb_sectors <= 0) {
1273             break;
1274         }
1275         status1 = bdrv_get_block_status_above(bs1, NULL, sector_num,
1276                                               total_sectors1 - sector_num,
1277                                               &pnum1, &file);
1278         if (status1 < 0) {
1279             ret = 3;
1280             error_report("Sector allocation test failed for %s", filename1);
1281             goto out;
1282         }
1283         allocated1 = status1 & BDRV_BLOCK_ALLOCATED;
1284 
1285         status2 = bdrv_get_block_status_above(bs2, NULL, sector_num,
1286                                               total_sectors2 - sector_num,
1287                                               &pnum2, &file);
1288         if (status2 < 0) {
1289             ret = 3;
1290             error_report("Sector allocation test failed for %s", filename2);
1291             goto out;
1292         }
1293         allocated2 = status2 & BDRV_BLOCK_ALLOCATED;
1294         if (pnum1) {
1295             nb_sectors = MIN(nb_sectors, pnum1);
1296         }
1297         if (pnum2) {
1298             nb_sectors = MIN(nb_sectors, pnum2);
1299         }
1300 
1301         if (strict) {
1302             if ((status1 & ~BDRV_BLOCK_OFFSET_MASK) !=
1303                 (status2 & ~BDRV_BLOCK_OFFSET_MASK)) {
1304                 ret = 1;
1305                 qprintf(quiet, "Strict mode: Offset %" PRId64
1306                         " block status mismatch!\n",
1307                         sectors_to_bytes(sector_num));
1308                 goto out;
1309             }
1310         }
1311         if ((status1 & BDRV_BLOCK_ZERO) && (status2 & BDRV_BLOCK_ZERO)) {
1312             nb_sectors = MIN(pnum1, pnum2);
1313         } else if (allocated1 == allocated2) {
1314             if (allocated1) {
1315                 ret = blk_pread(blk1, sector_num << BDRV_SECTOR_BITS, buf1,
1316                                 nb_sectors << BDRV_SECTOR_BITS);
1317                 if (ret < 0) {
1318                     error_report("Error while reading offset %" PRId64 " of %s:"
1319                                  " %s", sectors_to_bytes(sector_num), filename1,
1320                                  strerror(-ret));
1321                     ret = 4;
1322                     goto out;
1323                 }
1324                 ret = blk_pread(blk2, sector_num << BDRV_SECTOR_BITS, buf2,
1325                                 nb_sectors << BDRV_SECTOR_BITS);
1326                 if (ret < 0) {
1327                     error_report("Error while reading offset %" PRId64
1328                                  " of %s: %s", sectors_to_bytes(sector_num),
1329                                  filename2, strerror(-ret));
1330                     ret = 4;
1331                     goto out;
1332                 }
1333                 ret = compare_sectors(buf1, buf2, nb_sectors, &pnum);
1334                 if (ret || pnum != nb_sectors) {
1335                     qprintf(quiet, "Content mismatch at offset %" PRId64 "!\n",
1336                             sectors_to_bytes(
1337                                 ret ? sector_num : sector_num + pnum));
1338                     ret = 1;
1339                     goto out;
1340                 }
1341             }
1342         } else {
1343 
1344             if (allocated1) {
1345                 ret = check_empty_sectors(blk1, sector_num, nb_sectors,
1346                                           filename1, buf1, quiet);
1347             } else {
1348                 ret = check_empty_sectors(blk2, sector_num, nb_sectors,
1349                                           filename2, buf1, quiet);
1350             }
1351             if (ret) {
1352                 if (ret < 0) {
1353                     error_report("Error while reading offset %" PRId64 ": %s",
1354                                  sectors_to_bytes(sector_num), strerror(-ret));
1355                     ret = 4;
1356                 }
1357                 goto out;
1358             }
1359         }
1360         sector_num += nb_sectors;
1361         qemu_progress_print(((float) nb_sectors / progress_base)*100, 100);
1362     }
1363 
1364     if (total_sectors1 != total_sectors2) {
1365         BlockBackend *blk_over;
1366         int64_t total_sectors_over;
1367         const char *filename_over;
1368 
1369         qprintf(quiet, "Warning: Image size mismatch!\n");
1370         if (total_sectors1 > total_sectors2) {
1371             total_sectors_over = total_sectors1;
1372             blk_over = blk1;
1373             filename_over = filename1;
1374         } else {
1375             total_sectors_over = total_sectors2;
1376             blk_over = blk2;
1377             filename_over = filename2;
1378         }
1379 
1380         for (;;) {
1381             nb_sectors = sectors_to_process(total_sectors_over, sector_num);
1382             if (nb_sectors <= 0) {
1383                 break;
1384             }
1385             ret = bdrv_is_allocated_above(blk_bs(blk_over), NULL, sector_num,
1386                                           nb_sectors, &pnum);
1387             if (ret < 0) {
1388                 ret = 3;
1389                 error_report("Sector allocation test failed for %s",
1390                              filename_over);
1391                 goto out;
1392 
1393             }
1394             nb_sectors = pnum;
1395             if (ret) {
1396                 ret = check_empty_sectors(blk_over, sector_num, nb_sectors,
1397                                           filename_over, buf1, quiet);
1398                 if (ret) {
1399                     if (ret < 0) {
1400                         error_report("Error while reading offset %" PRId64
1401                                      " of %s: %s", sectors_to_bytes(sector_num),
1402                                      filename_over, strerror(-ret));
1403                         ret = 4;
1404                     }
1405                     goto out;
1406                 }
1407             }
1408             sector_num += nb_sectors;
1409             qemu_progress_print(((float) nb_sectors / progress_base)*100, 100);
1410         }
1411     }
1412 
1413     qprintf(quiet, "Images are identical.\n");
1414     ret = 0;
1415 
1416 out:
1417     qemu_vfree(buf1);
1418     qemu_vfree(buf2);
1419     blk_unref(blk2);
1420 out2:
1421     blk_unref(blk1);
1422 out3:
1423     qemu_progress_end();
1424 out4:
1425     return ret;
1426 }
1427 
1428 enum ImgConvertBlockStatus {
1429     BLK_DATA,
1430     BLK_ZERO,
1431     BLK_BACKING_FILE,
1432 };
1433 
1434 typedef struct ImgConvertState {
1435     BlockBackend **src;
1436     int64_t *src_sectors;
1437     int src_cur, src_num;
1438     int64_t src_cur_offset;
1439     int64_t total_sectors;
1440     int64_t allocated_sectors;
1441     enum ImgConvertBlockStatus status;
1442     int64_t sector_next_status;
1443     BlockBackend *target;
1444     bool has_zero_init;
1445     bool compressed;
1446     bool target_has_backing;
1447     int min_sparse;
1448     size_t cluster_sectors;
1449     size_t buf_sectors;
1450 } ImgConvertState;
1451 
1452 static void convert_select_part(ImgConvertState *s, int64_t sector_num)
1453 {
1454     assert(sector_num >= s->src_cur_offset);
1455     while (sector_num - s->src_cur_offset >= s->src_sectors[s->src_cur]) {
1456         s->src_cur_offset += s->src_sectors[s->src_cur];
1457         s->src_cur++;
1458         assert(s->src_cur < s->src_num);
1459     }
1460 }
1461 
1462 static int convert_iteration_sectors(ImgConvertState *s, int64_t sector_num)
1463 {
1464     int64_t ret;
1465     int n;
1466 
1467     convert_select_part(s, sector_num);
1468 
1469     assert(s->total_sectors > sector_num);
1470     n = MIN(s->total_sectors - sector_num, BDRV_REQUEST_MAX_SECTORS);
1471 
1472     if (s->sector_next_status <= sector_num) {
1473         BlockDriverState *file;
1474         ret = bdrv_get_block_status(blk_bs(s->src[s->src_cur]),
1475                                     sector_num - s->src_cur_offset,
1476                                     n, &n, &file);
1477         if (ret < 0) {
1478             return ret;
1479         }
1480 
1481         if (ret & BDRV_BLOCK_ZERO) {
1482             s->status = BLK_ZERO;
1483         } else if (ret & BDRV_BLOCK_DATA) {
1484             s->status = BLK_DATA;
1485         } else if (!s->target_has_backing) {
1486             /* Without a target backing file we must copy over the contents of
1487              * the backing file as well. */
1488             /* Check block status of the backing file chain to avoid
1489              * needlessly reading zeroes and limiting the iteration to the
1490              * buffer size */
1491             ret = bdrv_get_block_status_above(blk_bs(s->src[s->src_cur]), NULL,
1492                                               sector_num - s->src_cur_offset,
1493                                               n, &n, &file);
1494             if (ret < 0) {
1495                 return ret;
1496             }
1497 
1498             if (ret & BDRV_BLOCK_ZERO) {
1499                 s->status = BLK_ZERO;
1500             } else {
1501                 s->status = BLK_DATA;
1502             }
1503         } else {
1504             s->status = BLK_BACKING_FILE;
1505         }
1506 
1507         s->sector_next_status = sector_num + n;
1508     }
1509 
1510     n = MIN(n, s->sector_next_status - sector_num);
1511     if (s->status == BLK_DATA) {
1512         n = MIN(n, s->buf_sectors);
1513     }
1514 
1515     /* We need to write complete clusters for compressed images, so if an
1516      * unallocated area is shorter than that, we must consider the whole
1517      * cluster allocated. */
1518     if (s->compressed) {
1519         if (n < s->cluster_sectors) {
1520             n = MIN(s->cluster_sectors, s->total_sectors - sector_num);
1521             s->status = BLK_DATA;
1522         } else {
1523             n = QEMU_ALIGN_DOWN(n, s->cluster_sectors);
1524         }
1525     }
1526 
1527     return n;
1528 }
1529 
1530 static int convert_read(ImgConvertState *s, int64_t sector_num, int nb_sectors,
1531                         uint8_t *buf)
1532 {
1533     int n;
1534     int ret;
1535 
1536     assert(nb_sectors <= s->buf_sectors);
1537     while (nb_sectors > 0) {
1538         BlockBackend *blk;
1539         int64_t bs_sectors;
1540 
1541         /* In the case of compression with multiple source files, we can get a
1542          * nb_sectors that spreads into the next part. So we must be able to
1543          * read across multiple BDSes for one convert_read() call. */
1544         convert_select_part(s, sector_num);
1545         blk = s->src[s->src_cur];
1546         bs_sectors = s->src_sectors[s->src_cur];
1547 
1548         n = MIN(nb_sectors, bs_sectors - (sector_num - s->src_cur_offset));
1549         ret = blk_pread(blk,
1550                         (sector_num - s->src_cur_offset) << BDRV_SECTOR_BITS,
1551                         buf, n << BDRV_SECTOR_BITS);
1552         if (ret < 0) {
1553             return ret;
1554         }
1555 
1556         sector_num += n;
1557         nb_sectors -= n;
1558         buf += n * BDRV_SECTOR_SIZE;
1559     }
1560 
1561     return 0;
1562 }
1563 
1564 static int convert_write(ImgConvertState *s, int64_t sector_num, int nb_sectors,
1565                          const uint8_t *buf)
1566 {
1567     int ret;
1568 
1569     while (nb_sectors > 0) {
1570         int n = nb_sectors;
1571 
1572         switch (s->status) {
1573         case BLK_BACKING_FILE:
1574             /* If we have a backing file, leave clusters unallocated that are
1575              * unallocated in the source image, so that the backing file is
1576              * visible at the respective offset. */
1577             assert(s->target_has_backing);
1578             break;
1579 
1580         case BLK_DATA:
1581             /* We must always write compressed clusters as a whole, so don't
1582              * try to find zeroed parts in the buffer. We can only save the
1583              * write if the buffer is completely zeroed and we're allowed to
1584              * keep the target sparse. */
1585             if (s->compressed) {
1586                 if (s->has_zero_init && s->min_sparse &&
1587                     buffer_is_zero(buf, n * BDRV_SECTOR_SIZE))
1588                 {
1589                     assert(!s->target_has_backing);
1590                     break;
1591                 }
1592 
1593                 ret = blk_write_compressed(s->target, sector_num, buf, n);
1594                 if (ret < 0) {
1595                     return ret;
1596                 }
1597                 break;
1598             }
1599 
1600             /* If there is real non-zero data or we're told to keep the target
1601              * fully allocated (-S 0), we must write it. Otherwise we can treat
1602              * it as zero sectors. */
1603             if (!s->min_sparse ||
1604                 is_allocated_sectors_min(buf, n, &n, s->min_sparse))
1605             {
1606                 ret = blk_pwrite(s->target, sector_num << BDRV_SECTOR_BITS,
1607                                  buf, n << BDRV_SECTOR_BITS, 0);
1608                 if (ret < 0) {
1609                     return ret;
1610                 }
1611                 break;
1612             }
1613             /* fall-through */
1614 
1615         case BLK_ZERO:
1616             if (s->has_zero_init) {
1617                 break;
1618             }
1619             ret = blk_pwrite_zeroes(s->target, sector_num << BDRV_SECTOR_BITS,
1620                                     n << BDRV_SECTOR_BITS, 0);
1621             if (ret < 0) {
1622                 return ret;
1623             }
1624             break;
1625         }
1626 
1627         sector_num += n;
1628         nb_sectors -= n;
1629         buf += n * BDRV_SECTOR_SIZE;
1630     }
1631 
1632     return 0;
1633 }
1634 
1635 static int convert_do_copy(ImgConvertState *s)
1636 {
1637     uint8_t *buf = NULL;
1638     int64_t sector_num, allocated_done;
1639     int ret;
1640     int n;
1641 
1642     /* Check whether we have zero initialisation or can get it efficiently */
1643     s->has_zero_init = s->min_sparse && !s->target_has_backing
1644                      ? bdrv_has_zero_init(blk_bs(s->target))
1645                      : false;
1646 
1647     if (!s->has_zero_init && !s->target_has_backing &&
1648         bdrv_can_write_zeroes_with_unmap(blk_bs(s->target)))
1649     {
1650         ret = blk_make_zero(s->target, BDRV_REQ_MAY_UNMAP);
1651         if (ret == 0) {
1652             s->has_zero_init = true;
1653         }
1654     }
1655 
1656     /* Allocate buffer for copied data. For compressed images, only one cluster
1657      * can be copied at a time. */
1658     if (s->compressed) {
1659         if (s->cluster_sectors <= 0 || s->cluster_sectors > s->buf_sectors) {
1660             error_report("invalid cluster size");
1661             ret = -EINVAL;
1662             goto fail;
1663         }
1664         s->buf_sectors = s->cluster_sectors;
1665     }
1666     buf = blk_blockalign(s->target, s->buf_sectors * BDRV_SECTOR_SIZE);
1667 
1668     /* Calculate allocated sectors for progress */
1669     s->allocated_sectors = 0;
1670     sector_num = 0;
1671     while (sector_num < s->total_sectors) {
1672         n = convert_iteration_sectors(s, sector_num);
1673         if (n < 0) {
1674             ret = n;
1675             goto fail;
1676         }
1677         if (s->status == BLK_DATA || (!s->min_sparse && s->status == BLK_ZERO))
1678         {
1679             s->allocated_sectors += n;
1680         }
1681         sector_num += n;
1682     }
1683 
1684     /* Do the copy */
1685     s->src_cur = 0;
1686     s->src_cur_offset = 0;
1687     s->sector_next_status = 0;
1688 
1689     sector_num = 0;
1690     allocated_done = 0;
1691 
1692     while (sector_num < s->total_sectors) {
1693         n = convert_iteration_sectors(s, sector_num);
1694         if (n < 0) {
1695             ret = n;
1696             goto fail;
1697         }
1698         if (s->status == BLK_DATA || (!s->min_sparse && s->status == BLK_ZERO))
1699         {
1700             allocated_done += n;
1701             qemu_progress_print(100.0 * allocated_done / s->allocated_sectors,
1702                                 0);
1703         }
1704 
1705         if (s->status == BLK_DATA) {
1706             ret = convert_read(s, sector_num, n, buf);
1707             if (ret < 0) {
1708                 error_report("error while reading sector %" PRId64
1709                              ": %s", sector_num, strerror(-ret));
1710                 goto fail;
1711             }
1712         } else if (!s->min_sparse && s->status == BLK_ZERO) {
1713             n = MIN(n, s->buf_sectors);
1714             memset(buf, 0, n * BDRV_SECTOR_SIZE);
1715             s->status = BLK_DATA;
1716         }
1717 
1718         ret = convert_write(s, sector_num, n, buf);
1719         if (ret < 0) {
1720             error_report("error while writing sector %" PRId64
1721                          ": %s", sector_num, strerror(-ret));
1722             goto fail;
1723         }
1724 
1725         sector_num += n;
1726     }
1727 
1728     if (s->compressed) {
1729         /* signal EOF to align */
1730         ret = blk_write_compressed(s->target, 0, NULL, 0);
1731         if (ret < 0) {
1732             goto fail;
1733         }
1734     }
1735 
1736     ret = 0;
1737 fail:
1738     qemu_vfree(buf);
1739     return ret;
1740 }
1741 
1742 static int img_convert(int argc, char **argv)
1743 {
1744     int c, bs_n, bs_i, compress, cluster_sectors, skip_create;
1745     int64_t ret = 0;
1746     int progress = 0, flags, src_flags;
1747     bool writethrough, src_writethrough;
1748     const char *fmt, *out_fmt, *cache, *src_cache, *out_baseimg, *out_filename;
1749     BlockDriver *drv, *proto_drv;
1750     BlockBackend **blk = NULL, *out_blk = NULL;
1751     BlockDriverState **bs = NULL, *out_bs = NULL;
1752     int64_t total_sectors;
1753     int64_t *bs_sectors = NULL;
1754     size_t bufsectors = IO_BUF_SIZE / BDRV_SECTOR_SIZE;
1755     BlockDriverInfo bdi;
1756     QemuOpts *opts = NULL;
1757     QemuOptsList *create_opts = NULL;
1758     const char *out_baseimg_param;
1759     char *options = NULL;
1760     const char *snapshot_name = NULL;
1761     int min_sparse = 8; /* Need at least 4k of zeros for sparse detection */
1762     bool quiet = false;
1763     Error *local_err = NULL;
1764     QemuOpts *sn_opts = NULL;
1765     ImgConvertState state;
1766     bool image_opts = false;
1767 
1768     fmt = NULL;
1769     out_fmt = "raw";
1770     cache = "unsafe";
1771     src_cache = BDRV_DEFAULT_CACHE;
1772     out_baseimg = NULL;
1773     compress = 0;
1774     skip_create = 0;
1775     for(;;) {
1776         static const struct option long_options[] = {
1777             {"help", no_argument, 0, 'h'},
1778             {"object", required_argument, 0, OPTION_OBJECT},
1779             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
1780             {0, 0, 0, 0}
1781         };
1782         c = getopt_long(argc, argv, "hf:O:B:ce6o:s:l:S:pt:T:qn",
1783                         long_options, NULL);
1784         if (c == -1) {
1785             break;
1786         }
1787         switch(c) {
1788         case '?':
1789         case 'h':
1790             help();
1791             break;
1792         case 'f':
1793             fmt = optarg;
1794             break;
1795         case 'O':
1796             out_fmt = optarg;
1797             break;
1798         case 'B':
1799             out_baseimg = optarg;
1800             break;
1801         case 'c':
1802             compress = 1;
1803             break;
1804         case 'e':
1805             error_report("option -e is deprecated, please use \'-o "
1806                   "encryption\' instead!");
1807             ret = -1;
1808             goto fail_getopt;
1809         case '6':
1810             error_report("option -6 is deprecated, please use \'-o "
1811                   "compat6\' instead!");
1812             ret = -1;
1813             goto fail_getopt;
1814         case 'o':
1815             if (!is_valid_option_list(optarg)) {
1816                 error_report("Invalid option list: %s", optarg);
1817                 ret = -1;
1818                 goto fail_getopt;
1819             }
1820             if (!options) {
1821                 options = g_strdup(optarg);
1822             } else {
1823                 char *old_options = options;
1824                 options = g_strdup_printf("%s,%s", options, optarg);
1825                 g_free(old_options);
1826             }
1827             break;
1828         case 's':
1829             snapshot_name = optarg;
1830             break;
1831         case 'l':
1832             if (strstart(optarg, SNAPSHOT_OPT_BASE, NULL)) {
1833                 sn_opts = qemu_opts_parse_noisily(&internal_snapshot_opts,
1834                                                   optarg, false);
1835                 if (!sn_opts) {
1836                     error_report("Failed in parsing snapshot param '%s'",
1837                                  optarg);
1838                     ret = -1;
1839                     goto fail_getopt;
1840                 }
1841             } else {
1842                 snapshot_name = optarg;
1843             }
1844             break;
1845         case 'S':
1846         {
1847             int64_t sval;
1848             char *end;
1849             sval = qemu_strtosz_suffix(optarg, &end, QEMU_STRTOSZ_DEFSUFFIX_B);
1850             if (sval < 0 || *end) {
1851                 error_report("Invalid minimum zero buffer size for sparse output specified");
1852                 ret = -1;
1853                 goto fail_getopt;
1854             }
1855 
1856             min_sparse = sval / BDRV_SECTOR_SIZE;
1857             break;
1858         }
1859         case 'p':
1860             progress = 1;
1861             break;
1862         case 't':
1863             cache = optarg;
1864             break;
1865         case 'T':
1866             src_cache = optarg;
1867             break;
1868         case 'q':
1869             quiet = true;
1870             break;
1871         case 'n':
1872             skip_create = 1;
1873             break;
1874         case OPTION_OBJECT:
1875             opts = qemu_opts_parse_noisily(&qemu_object_opts,
1876                                            optarg, true);
1877             if (!opts) {
1878                 goto fail_getopt;
1879             }
1880             break;
1881         case OPTION_IMAGE_OPTS:
1882             image_opts = true;
1883             break;
1884         }
1885     }
1886 
1887     if (qemu_opts_foreach(&qemu_object_opts,
1888                           user_creatable_add_opts_foreach,
1889                           NULL, NULL)) {
1890         goto fail_getopt;
1891     }
1892 
1893     /* Initialize before goto out */
1894     if (quiet) {
1895         progress = 0;
1896     }
1897     qemu_progress_init(progress, 1.0);
1898 
1899     bs_n = argc - optind - 1;
1900     out_filename = bs_n >= 1 ? argv[argc - 1] : NULL;
1901 
1902     if (options && has_help_option(options)) {
1903         ret = print_block_option_help(out_filename, out_fmt);
1904         goto out;
1905     }
1906 
1907     if (bs_n < 1) {
1908         error_exit("Must specify image file name");
1909     }
1910 
1911 
1912     if (bs_n > 1 && out_baseimg) {
1913         error_report("-B makes no sense when concatenating multiple input "
1914                      "images");
1915         ret = -1;
1916         goto out;
1917     }
1918 
1919     src_flags = 0;
1920     ret = bdrv_parse_cache_mode(src_cache, &src_flags, &src_writethrough);
1921     if (ret < 0) {
1922         error_report("Invalid source cache option: %s", src_cache);
1923         goto out;
1924     }
1925 
1926     qemu_progress_print(0, 100);
1927 
1928     blk = g_new0(BlockBackend *, bs_n);
1929     bs = g_new0(BlockDriverState *, bs_n);
1930     bs_sectors = g_new(int64_t, bs_n);
1931 
1932     total_sectors = 0;
1933     for (bs_i = 0; bs_i < bs_n; bs_i++) {
1934         blk[bs_i] = img_open(image_opts, argv[optind + bs_i],
1935                              fmt, src_flags, src_writethrough, quiet);
1936         if (!blk[bs_i]) {
1937             ret = -1;
1938             goto out;
1939         }
1940         bs[bs_i] = blk_bs(blk[bs_i]);
1941         bs_sectors[bs_i] = blk_nb_sectors(blk[bs_i]);
1942         if (bs_sectors[bs_i] < 0) {
1943             error_report("Could not get size of %s: %s",
1944                          argv[optind + bs_i], strerror(-bs_sectors[bs_i]));
1945             ret = -1;
1946             goto out;
1947         }
1948         total_sectors += bs_sectors[bs_i];
1949     }
1950 
1951     if (sn_opts) {
1952         ret = bdrv_snapshot_load_tmp(bs[0],
1953                                      qemu_opt_get(sn_opts, SNAPSHOT_OPT_ID),
1954                                      qemu_opt_get(sn_opts, SNAPSHOT_OPT_NAME),
1955                                      &local_err);
1956     } else if (snapshot_name != NULL) {
1957         if (bs_n > 1) {
1958             error_report("No support for concatenating multiple snapshot");
1959             ret = -1;
1960             goto out;
1961         }
1962 
1963         bdrv_snapshot_load_tmp_by_id_or_name(bs[0], snapshot_name, &local_err);
1964     }
1965     if (local_err) {
1966         error_reportf_err(local_err, "Failed to load snapshot: ");
1967         ret = -1;
1968         goto out;
1969     }
1970 
1971     /* Find driver and parse its options */
1972     drv = bdrv_find_format(out_fmt);
1973     if (!drv) {
1974         error_report("Unknown file format '%s'", out_fmt);
1975         ret = -1;
1976         goto out;
1977     }
1978 
1979     proto_drv = bdrv_find_protocol(out_filename, true, &local_err);
1980     if (!proto_drv) {
1981         error_report_err(local_err);
1982         ret = -1;
1983         goto out;
1984     }
1985 
1986     if (!skip_create) {
1987         if (!drv->create_opts) {
1988             error_report("Format driver '%s' does not support image creation",
1989                          drv->format_name);
1990             ret = -1;
1991             goto out;
1992         }
1993 
1994         if (!proto_drv->create_opts) {
1995             error_report("Protocol driver '%s' does not support image creation",
1996                          proto_drv->format_name);
1997             ret = -1;
1998             goto out;
1999         }
2000 
2001         create_opts = qemu_opts_append(create_opts, drv->create_opts);
2002         create_opts = qemu_opts_append(create_opts, proto_drv->create_opts);
2003 
2004         opts = qemu_opts_create(create_opts, NULL, 0, &error_abort);
2005         if (options) {
2006             qemu_opts_do_parse(opts, options, NULL, &local_err);
2007             if (local_err) {
2008                 error_report_err(local_err);
2009                 ret = -1;
2010                 goto out;
2011             }
2012         }
2013 
2014         qemu_opt_set_number(opts, BLOCK_OPT_SIZE, total_sectors * 512,
2015                             &error_abort);
2016         ret = add_old_style_options(out_fmt, opts, out_baseimg, NULL);
2017         if (ret < 0) {
2018             goto out;
2019         }
2020     }
2021 
2022     /* Get backing file name if -o backing_file was used */
2023     out_baseimg_param = qemu_opt_get(opts, BLOCK_OPT_BACKING_FILE);
2024     if (out_baseimg_param) {
2025         out_baseimg = out_baseimg_param;
2026     }
2027 
2028     /* Check if compression is supported */
2029     if (compress) {
2030         bool encryption =
2031             qemu_opt_get_bool(opts, BLOCK_OPT_ENCRYPT, false);
2032         const char *preallocation =
2033             qemu_opt_get(opts, BLOCK_OPT_PREALLOC);
2034 
2035         if (!drv->bdrv_write_compressed) {
2036             error_report("Compression not supported for this file format");
2037             ret = -1;
2038             goto out;
2039         }
2040 
2041         if (encryption) {
2042             error_report("Compression and encryption not supported at "
2043                          "the same time");
2044             ret = -1;
2045             goto out;
2046         }
2047 
2048         if (preallocation
2049             && strcmp(preallocation, "off"))
2050         {
2051             error_report("Compression and preallocation not supported at "
2052                          "the same time");
2053             ret = -1;
2054             goto out;
2055         }
2056     }
2057 
2058     if (!skip_create) {
2059         /* Create the new image */
2060         ret = bdrv_create(drv, out_filename, opts, &local_err);
2061         if (ret < 0) {
2062             error_reportf_err(local_err, "%s: error while converting %s: ",
2063                               out_filename, out_fmt);
2064             goto out;
2065         }
2066     }
2067 
2068     flags = min_sparse ? (BDRV_O_RDWR | BDRV_O_UNMAP) : BDRV_O_RDWR;
2069     ret = bdrv_parse_cache_mode(cache, &flags, &writethrough);
2070     if (ret < 0) {
2071         error_report("Invalid cache option: %s", cache);
2072         goto out;
2073     }
2074 
2075     /* XXX we should allow --image-opts to trigger use of
2076      * img_open() here, but then we have trouble with
2077      * the bdrv_create() call which takes different params.
2078      * Not critical right now, so fix can wait...
2079      */
2080     out_blk = img_open_file(out_filename, out_fmt, flags, writethrough, quiet);
2081     if (!out_blk) {
2082         ret = -1;
2083         goto out;
2084     }
2085     out_bs = blk_bs(out_blk);
2086 
2087     /* increase bufsectors from the default 4096 (2M) if opt_transfer
2088      * or discard_alignment of the out_bs is greater. Limit to 32768 (16MB)
2089      * as maximum. */
2090     bufsectors = MIN(32768,
2091                      MAX(bufsectors,
2092                          MAX(out_bs->bl.opt_transfer >> BDRV_SECTOR_BITS,
2093                              out_bs->bl.pdiscard_alignment >>
2094                              BDRV_SECTOR_BITS)));
2095 
2096     if (skip_create) {
2097         int64_t output_sectors = blk_nb_sectors(out_blk);
2098         if (output_sectors < 0) {
2099             error_report("unable to get output image length: %s",
2100                          strerror(-output_sectors));
2101             ret = -1;
2102             goto out;
2103         } else if (output_sectors < total_sectors) {
2104             error_report("output file is smaller than input file");
2105             ret = -1;
2106             goto out;
2107         }
2108     }
2109 
2110     cluster_sectors = 0;
2111     ret = bdrv_get_info(out_bs, &bdi);
2112     if (ret < 0) {
2113         if (compress) {
2114             error_report("could not get block driver info");
2115             goto out;
2116         }
2117     } else {
2118         compress = compress || bdi.needs_compressed_writes;
2119         cluster_sectors = bdi.cluster_size / BDRV_SECTOR_SIZE;
2120     }
2121 
2122     state = (ImgConvertState) {
2123         .src                = blk,
2124         .src_sectors        = bs_sectors,
2125         .src_num            = bs_n,
2126         .total_sectors      = total_sectors,
2127         .target             = out_blk,
2128         .compressed         = compress,
2129         .target_has_backing = (bool) out_baseimg,
2130         .min_sparse         = min_sparse,
2131         .cluster_sectors    = cluster_sectors,
2132         .buf_sectors        = bufsectors,
2133     };
2134     ret = convert_do_copy(&state);
2135 
2136 out:
2137     if (!ret) {
2138         qemu_progress_print(100, 0);
2139     }
2140     qemu_progress_end();
2141     qemu_opts_del(opts);
2142     qemu_opts_free(create_opts);
2143     qemu_opts_del(sn_opts);
2144     blk_unref(out_blk);
2145     g_free(bs);
2146     if (blk) {
2147         for (bs_i = 0; bs_i < bs_n; bs_i++) {
2148             blk_unref(blk[bs_i]);
2149         }
2150         g_free(blk);
2151     }
2152     g_free(bs_sectors);
2153 fail_getopt:
2154     g_free(options);
2155 
2156     if (ret) {
2157         return 1;
2158     }
2159     return 0;
2160 }
2161 
2162 
2163 static void dump_snapshots(BlockDriverState *bs)
2164 {
2165     QEMUSnapshotInfo *sn_tab, *sn;
2166     int nb_sns, i;
2167 
2168     nb_sns = bdrv_snapshot_list(bs, &sn_tab);
2169     if (nb_sns <= 0)
2170         return;
2171     printf("Snapshot list:\n");
2172     bdrv_snapshot_dump(fprintf, stdout, NULL);
2173     printf("\n");
2174     for(i = 0; i < nb_sns; i++) {
2175         sn = &sn_tab[i];
2176         bdrv_snapshot_dump(fprintf, stdout, sn);
2177         printf("\n");
2178     }
2179     g_free(sn_tab);
2180 }
2181 
2182 static void dump_json_image_info_list(ImageInfoList *list)
2183 {
2184     QString *str;
2185     QObject *obj;
2186     Visitor *v = qmp_output_visitor_new(&obj);
2187 
2188     visit_type_ImageInfoList(v, NULL, &list, &error_abort);
2189     visit_complete(v, &obj);
2190     str = qobject_to_json_pretty(obj);
2191     assert(str != NULL);
2192     printf("%s\n", qstring_get_str(str));
2193     qobject_decref(obj);
2194     visit_free(v);
2195     QDECREF(str);
2196 }
2197 
2198 static void dump_json_image_info(ImageInfo *info)
2199 {
2200     QString *str;
2201     QObject *obj;
2202     Visitor *v = qmp_output_visitor_new(&obj);
2203 
2204     visit_type_ImageInfo(v, NULL, &info, &error_abort);
2205     visit_complete(v, &obj);
2206     str = qobject_to_json_pretty(obj);
2207     assert(str != NULL);
2208     printf("%s\n", qstring_get_str(str));
2209     qobject_decref(obj);
2210     visit_free(v);
2211     QDECREF(str);
2212 }
2213 
2214 static void dump_human_image_info_list(ImageInfoList *list)
2215 {
2216     ImageInfoList *elem;
2217     bool delim = false;
2218 
2219     for (elem = list; elem; elem = elem->next) {
2220         if (delim) {
2221             printf("\n");
2222         }
2223         delim = true;
2224 
2225         bdrv_image_info_dump(fprintf, stdout, elem->value);
2226     }
2227 }
2228 
2229 static gboolean str_equal_func(gconstpointer a, gconstpointer b)
2230 {
2231     return strcmp(a, b) == 0;
2232 }
2233 
2234 /**
2235  * Open an image file chain and return an ImageInfoList
2236  *
2237  * @filename: topmost image filename
2238  * @fmt: topmost image format (may be NULL to autodetect)
2239  * @chain: true  - enumerate entire backing file chain
2240  *         false - only topmost image file
2241  *
2242  * Returns a list of ImageInfo objects or NULL if there was an error opening an
2243  * image file.  If there was an error a message will have been printed to
2244  * stderr.
2245  */
2246 static ImageInfoList *collect_image_info_list(bool image_opts,
2247                                               const char *filename,
2248                                               const char *fmt,
2249                                               bool chain)
2250 {
2251     ImageInfoList *head = NULL;
2252     ImageInfoList **last = &head;
2253     GHashTable *filenames;
2254     Error *err = NULL;
2255 
2256     filenames = g_hash_table_new_full(g_str_hash, str_equal_func, NULL, NULL);
2257 
2258     while (filename) {
2259         BlockBackend *blk;
2260         BlockDriverState *bs;
2261         ImageInfo *info;
2262         ImageInfoList *elem;
2263 
2264         if (g_hash_table_lookup_extended(filenames, filename, NULL, NULL)) {
2265             error_report("Backing file '%s' creates an infinite loop.",
2266                          filename);
2267             goto err;
2268         }
2269         g_hash_table_insert(filenames, (gpointer)filename, NULL);
2270 
2271         blk = img_open(image_opts, filename, fmt,
2272                        BDRV_O_NO_BACKING | BDRV_O_NO_IO, false, false);
2273         if (!blk) {
2274             goto err;
2275         }
2276         bs = blk_bs(blk);
2277 
2278         bdrv_query_image_info(bs, &info, &err);
2279         if (err) {
2280             error_report_err(err);
2281             blk_unref(blk);
2282             goto err;
2283         }
2284 
2285         elem = g_new0(ImageInfoList, 1);
2286         elem->value = info;
2287         *last = elem;
2288         last = &elem->next;
2289 
2290         blk_unref(blk);
2291 
2292         filename = fmt = NULL;
2293         if (chain) {
2294             if (info->has_full_backing_filename) {
2295                 filename = info->full_backing_filename;
2296             } else if (info->has_backing_filename) {
2297                 error_report("Could not determine absolute backing filename,"
2298                              " but backing filename '%s' present",
2299                              info->backing_filename);
2300                 goto err;
2301             }
2302             if (info->has_backing_filename_format) {
2303                 fmt = info->backing_filename_format;
2304             }
2305         }
2306     }
2307     g_hash_table_destroy(filenames);
2308     return head;
2309 
2310 err:
2311     qapi_free_ImageInfoList(head);
2312     g_hash_table_destroy(filenames);
2313     return NULL;
2314 }
2315 
2316 static int img_info(int argc, char **argv)
2317 {
2318     int c;
2319     OutputFormat output_format = OFORMAT_HUMAN;
2320     bool chain = false;
2321     const char *filename, *fmt, *output;
2322     ImageInfoList *list;
2323     bool image_opts = false;
2324 
2325     fmt = NULL;
2326     output = NULL;
2327     for(;;) {
2328         int option_index = 0;
2329         static const struct option long_options[] = {
2330             {"help", no_argument, 0, 'h'},
2331             {"format", required_argument, 0, 'f'},
2332             {"output", required_argument, 0, OPTION_OUTPUT},
2333             {"backing-chain", no_argument, 0, OPTION_BACKING_CHAIN},
2334             {"object", required_argument, 0, OPTION_OBJECT},
2335             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
2336             {0, 0, 0, 0}
2337         };
2338         c = getopt_long(argc, argv, "f:h",
2339                         long_options, &option_index);
2340         if (c == -1) {
2341             break;
2342         }
2343         switch(c) {
2344         case '?':
2345         case 'h':
2346             help();
2347             break;
2348         case 'f':
2349             fmt = optarg;
2350             break;
2351         case OPTION_OUTPUT:
2352             output = optarg;
2353             break;
2354         case OPTION_BACKING_CHAIN:
2355             chain = true;
2356             break;
2357         case OPTION_OBJECT: {
2358             QemuOpts *opts;
2359             opts = qemu_opts_parse_noisily(&qemu_object_opts,
2360                                            optarg, true);
2361             if (!opts) {
2362                 return 1;
2363             }
2364         }   break;
2365         case OPTION_IMAGE_OPTS:
2366             image_opts = true;
2367             break;
2368         }
2369     }
2370     if (optind != argc - 1) {
2371         error_exit("Expecting one image file name");
2372     }
2373     filename = argv[optind++];
2374 
2375     if (output && !strcmp(output, "json")) {
2376         output_format = OFORMAT_JSON;
2377     } else if (output && !strcmp(output, "human")) {
2378         output_format = OFORMAT_HUMAN;
2379     } else if (output) {
2380         error_report("--output must be used with human or json as argument.");
2381         return 1;
2382     }
2383 
2384     if (qemu_opts_foreach(&qemu_object_opts,
2385                           user_creatable_add_opts_foreach,
2386                           NULL, NULL)) {
2387         return 1;
2388     }
2389 
2390     list = collect_image_info_list(image_opts, filename, fmt, chain);
2391     if (!list) {
2392         return 1;
2393     }
2394 
2395     switch (output_format) {
2396     case OFORMAT_HUMAN:
2397         dump_human_image_info_list(list);
2398         break;
2399     case OFORMAT_JSON:
2400         if (chain) {
2401             dump_json_image_info_list(list);
2402         } else {
2403             dump_json_image_info(list->value);
2404         }
2405         break;
2406     }
2407 
2408     qapi_free_ImageInfoList(list);
2409     return 0;
2410 }
2411 
2412 static void dump_map_entry(OutputFormat output_format, MapEntry *e,
2413                            MapEntry *next)
2414 {
2415     switch (output_format) {
2416     case OFORMAT_HUMAN:
2417         if (e->data && !e->has_offset) {
2418             error_report("File contains external, encrypted or compressed clusters.");
2419             exit(1);
2420         }
2421         if (e->data && !e->zero) {
2422             printf("%#-16"PRIx64"%#-16"PRIx64"%#-16"PRIx64"%s\n",
2423                    e->start, e->length,
2424                    e->has_offset ? e->offset : 0,
2425                    e->has_filename ? e->filename : "");
2426         }
2427         /* This format ignores the distinction between 0, ZERO and ZERO|DATA.
2428          * Modify the flags here to allow more coalescing.
2429          */
2430         if (next && (!next->data || next->zero)) {
2431             next->data = false;
2432             next->zero = true;
2433         }
2434         break;
2435     case OFORMAT_JSON:
2436         printf("%s{ \"start\": %"PRId64", \"length\": %"PRId64","
2437                " \"depth\": %"PRId64", \"zero\": %s, \"data\": %s",
2438                (e->start == 0 ? "[" : ",\n"),
2439                e->start, e->length, e->depth,
2440                e->zero ? "true" : "false",
2441                e->data ? "true" : "false");
2442         if (e->has_offset) {
2443             printf(", \"offset\": %"PRId64"", e->offset);
2444         }
2445         putchar('}');
2446 
2447         if (!next) {
2448             printf("]\n");
2449         }
2450         break;
2451     }
2452 }
2453 
2454 static int get_block_status(BlockDriverState *bs, int64_t sector_num,
2455                             int nb_sectors, MapEntry *e)
2456 {
2457     int64_t ret;
2458     int depth;
2459     BlockDriverState *file;
2460     bool has_offset;
2461 
2462     /* As an optimization, we could cache the current range of unallocated
2463      * clusters in each file of the chain, and avoid querying the same
2464      * range repeatedly.
2465      */
2466 
2467     depth = 0;
2468     for (;;) {
2469         ret = bdrv_get_block_status(bs, sector_num, nb_sectors, &nb_sectors,
2470                                     &file);
2471         if (ret < 0) {
2472             return ret;
2473         }
2474         assert(nb_sectors);
2475         if (ret & (BDRV_BLOCK_ZERO|BDRV_BLOCK_DATA)) {
2476             break;
2477         }
2478         bs = backing_bs(bs);
2479         if (bs == NULL) {
2480             ret = 0;
2481             break;
2482         }
2483 
2484         depth++;
2485     }
2486 
2487     has_offset = !!(ret & BDRV_BLOCK_OFFSET_VALID);
2488 
2489     *e = (MapEntry) {
2490         .start = sector_num * BDRV_SECTOR_SIZE,
2491         .length = nb_sectors * BDRV_SECTOR_SIZE,
2492         .data = !!(ret & BDRV_BLOCK_DATA),
2493         .zero = !!(ret & BDRV_BLOCK_ZERO),
2494         .offset = ret & BDRV_BLOCK_OFFSET_MASK,
2495         .has_offset = has_offset,
2496         .depth = depth,
2497         .has_filename = file && has_offset,
2498         .filename = file && has_offset ? file->filename : NULL,
2499     };
2500 
2501     return 0;
2502 }
2503 
2504 static inline bool entry_mergeable(const MapEntry *curr, const MapEntry *next)
2505 {
2506     if (curr->length == 0) {
2507         return false;
2508     }
2509     if (curr->zero != next->zero ||
2510         curr->data != next->data ||
2511         curr->depth != next->depth ||
2512         curr->has_filename != next->has_filename ||
2513         curr->has_offset != next->has_offset) {
2514         return false;
2515     }
2516     if (curr->has_filename && strcmp(curr->filename, next->filename)) {
2517         return false;
2518     }
2519     if (curr->has_offset && curr->offset + curr->length != next->offset) {
2520         return false;
2521     }
2522     return true;
2523 }
2524 
2525 static int img_map(int argc, char **argv)
2526 {
2527     int c;
2528     OutputFormat output_format = OFORMAT_HUMAN;
2529     BlockBackend *blk;
2530     BlockDriverState *bs;
2531     const char *filename, *fmt, *output;
2532     int64_t length;
2533     MapEntry curr = { .length = 0 }, next;
2534     int ret = 0;
2535     bool image_opts = false;
2536 
2537     fmt = NULL;
2538     output = NULL;
2539     for (;;) {
2540         int option_index = 0;
2541         static const struct option long_options[] = {
2542             {"help", no_argument, 0, 'h'},
2543             {"format", required_argument, 0, 'f'},
2544             {"output", required_argument, 0, OPTION_OUTPUT},
2545             {"object", required_argument, 0, OPTION_OBJECT},
2546             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
2547             {0, 0, 0, 0}
2548         };
2549         c = getopt_long(argc, argv, "f:h",
2550                         long_options, &option_index);
2551         if (c == -1) {
2552             break;
2553         }
2554         switch (c) {
2555         case '?':
2556         case 'h':
2557             help();
2558             break;
2559         case 'f':
2560             fmt = optarg;
2561             break;
2562         case OPTION_OUTPUT:
2563             output = optarg;
2564             break;
2565         case OPTION_OBJECT: {
2566             QemuOpts *opts;
2567             opts = qemu_opts_parse_noisily(&qemu_object_opts,
2568                                            optarg, true);
2569             if (!opts) {
2570                 return 1;
2571             }
2572         }   break;
2573         case OPTION_IMAGE_OPTS:
2574             image_opts = true;
2575             break;
2576         }
2577     }
2578     if (optind != argc - 1) {
2579         error_exit("Expecting one image file name");
2580     }
2581     filename = argv[optind];
2582 
2583     if (output && !strcmp(output, "json")) {
2584         output_format = OFORMAT_JSON;
2585     } else if (output && !strcmp(output, "human")) {
2586         output_format = OFORMAT_HUMAN;
2587     } else if (output) {
2588         error_report("--output must be used with human or json as argument.");
2589         return 1;
2590     }
2591 
2592     if (qemu_opts_foreach(&qemu_object_opts,
2593                           user_creatable_add_opts_foreach,
2594                           NULL, NULL)) {
2595         return 1;
2596     }
2597 
2598     blk = img_open(image_opts, filename, fmt, 0, false, false);
2599     if (!blk) {
2600         return 1;
2601     }
2602     bs = blk_bs(blk);
2603 
2604     if (output_format == OFORMAT_HUMAN) {
2605         printf("%-16s%-16s%-16s%s\n", "Offset", "Length", "Mapped to", "File");
2606     }
2607 
2608     length = blk_getlength(blk);
2609     while (curr.start + curr.length < length) {
2610         int64_t nsectors_left;
2611         int64_t sector_num;
2612         int n;
2613 
2614         sector_num = (curr.start + curr.length) >> BDRV_SECTOR_BITS;
2615 
2616         /* Probe up to 1 GiB at a time.  */
2617         nsectors_left = DIV_ROUND_UP(length, BDRV_SECTOR_SIZE) - sector_num;
2618         n = MIN(1 << (30 - BDRV_SECTOR_BITS), nsectors_left);
2619         ret = get_block_status(bs, sector_num, n, &next);
2620 
2621         if (ret < 0) {
2622             error_report("Could not read file metadata: %s", strerror(-ret));
2623             goto out;
2624         }
2625 
2626         if (entry_mergeable(&curr, &next)) {
2627             curr.length += next.length;
2628             continue;
2629         }
2630 
2631         if (curr.length > 0) {
2632             dump_map_entry(output_format, &curr, &next);
2633         }
2634         curr = next;
2635     }
2636 
2637     dump_map_entry(output_format, &curr, NULL);
2638 
2639 out:
2640     blk_unref(blk);
2641     return ret < 0;
2642 }
2643 
2644 #define SNAPSHOT_LIST   1
2645 #define SNAPSHOT_CREATE 2
2646 #define SNAPSHOT_APPLY  3
2647 #define SNAPSHOT_DELETE 4
2648 
2649 static int img_snapshot(int argc, char **argv)
2650 {
2651     BlockBackend *blk;
2652     BlockDriverState *bs;
2653     QEMUSnapshotInfo sn;
2654     char *filename, *snapshot_name = NULL;
2655     int c, ret = 0, bdrv_oflags;
2656     int action = 0;
2657     qemu_timeval tv;
2658     bool quiet = false;
2659     Error *err = NULL;
2660     bool image_opts = false;
2661 
2662     bdrv_oflags = BDRV_O_RDWR;
2663     /* Parse commandline parameters */
2664     for(;;) {
2665         static const struct option long_options[] = {
2666             {"help", no_argument, 0, 'h'},
2667             {"object", required_argument, 0, OPTION_OBJECT},
2668             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
2669             {0, 0, 0, 0}
2670         };
2671         c = getopt_long(argc, argv, "la:c:d:hq",
2672                         long_options, NULL);
2673         if (c == -1) {
2674             break;
2675         }
2676         switch(c) {
2677         case '?':
2678         case 'h':
2679             help();
2680             return 0;
2681         case 'l':
2682             if (action) {
2683                 error_exit("Cannot mix '-l', '-a', '-c', '-d'");
2684                 return 0;
2685             }
2686             action = SNAPSHOT_LIST;
2687             bdrv_oflags &= ~BDRV_O_RDWR; /* no need for RW */
2688             break;
2689         case 'a':
2690             if (action) {
2691                 error_exit("Cannot mix '-l', '-a', '-c', '-d'");
2692                 return 0;
2693             }
2694             action = SNAPSHOT_APPLY;
2695             snapshot_name = optarg;
2696             break;
2697         case 'c':
2698             if (action) {
2699                 error_exit("Cannot mix '-l', '-a', '-c', '-d'");
2700                 return 0;
2701             }
2702             action = SNAPSHOT_CREATE;
2703             snapshot_name = optarg;
2704             break;
2705         case 'd':
2706             if (action) {
2707                 error_exit("Cannot mix '-l', '-a', '-c', '-d'");
2708                 return 0;
2709             }
2710             action = SNAPSHOT_DELETE;
2711             snapshot_name = optarg;
2712             break;
2713         case 'q':
2714             quiet = true;
2715             break;
2716         case OPTION_OBJECT: {
2717             QemuOpts *opts;
2718             opts = qemu_opts_parse_noisily(&qemu_object_opts,
2719                                            optarg, true);
2720             if (!opts) {
2721                 return 1;
2722             }
2723         }   break;
2724         case OPTION_IMAGE_OPTS:
2725             image_opts = true;
2726             break;
2727         }
2728     }
2729 
2730     if (optind != argc - 1) {
2731         error_exit("Expecting one image file name");
2732     }
2733     filename = argv[optind++];
2734 
2735     if (qemu_opts_foreach(&qemu_object_opts,
2736                           user_creatable_add_opts_foreach,
2737                           NULL, NULL)) {
2738         return 1;
2739     }
2740 
2741     /* Open the image */
2742     blk = img_open(image_opts, filename, NULL, bdrv_oflags, false, quiet);
2743     if (!blk) {
2744         return 1;
2745     }
2746     bs = blk_bs(blk);
2747 
2748     /* Perform the requested action */
2749     switch(action) {
2750     case SNAPSHOT_LIST:
2751         dump_snapshots(bs);
2752         break;
2753 
2754     case SNAPSHOT_CREATE:
2755         memset(&sn, 0, sizeof(sn));
2756         pstrcpy(sn.name, sizeof(sn.name), snapshot_name);
2757 
2758         qemu_gettimeofday(&tv);
2759         sn.date_sec = tv.tv_sec;
2760         sn.date_nsec = tv.tv_usec * 1000;
2761 
2762         ret = bdrv_snapshot_create(bs, &sn);
2763         if (ret) {
2764             error_report("Could not create snapshot '%s': %d (%s)",
2765                 snapshot_name, ret, strerror(-ret));
2766         }
2767         break;
2768 
2769     case SNAPSHOT_APPLY:
2770         ret = bdrv_snapshot_goto(bs, snapshot_name);
2771         if (ret) {
2772             error_report("Could not apply snapshot '%s': %d (%s)",
2773                 snapshot_name, ret, strerror(-ret));
2774         }
2775         break;
2776 
2777     case SNAPSHOT_DELETE:
2778         bdrv_snapshot_delete_by_id_or_name(bs, snapshot_name, &err);
2779         if (err) {
2780             error_reportf_err(err, "Could not delete snapshot '%s': ",
2781                               snapshot_name);
2782             ret = 1;
2783         }
2784         break;
2785     }
2786 
2787     /* Cleanup */
2788     blk_unref(blk);
2789     if (ret) {
2790         return 1;
2791     }
2792     return 0;
2793 }
2794 
2795 static int img_rebase(int argc, char **argv)
2796 {
2797     BlockBackend *blk = NULL, *blk_old_backing = NULL, *blk_new_backing = NULL;
2798     uint8_t *buf_old = NULL;
2799     uint8_t *buf_new = NULL;
2800     BlockDriverState *bs = NULL;
2801     char *filename;
2802     const char *fmt, *cache, *src_cache, *out_basefmt, *out_baseimg;
2803     int c, flags, src_flags, ret;
2804     bool writethrough, src_writethrough;
2805     int unsafe = 0;
2806     int progress = 0;
2807     bool quiet = false;
2808     Error *local_err = NULL;
2809     bool image_opts = false;
2810 
2811     /* Parse commandline parameters */
2812     fmt = NULL;
2813     cache = BDRV_DEFAULT_CACHE;
2814     src_cache = BDRV_DEFAULT_CACHE;
2815     out_baseimg = NULL;
2816     out_basefmt = NULL;
2817     for(;;) {
2818         static const struct option long_options[] = {
2819             {"help", no_argument, 0, 'h'},
2820             {"object", required_argument, 0, OPTION_OBJECT},
2821             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
2822             {0, 0, 0, 0}
2823         };
2824         c = getopt_long(argc, argv, "hf:F:b:upt:T:q",
2825                         long_options, NULL);
2826         if (c == -1) {
2827             break;
2828         }
2829         switch(c) {
2830         case '?':
2831         case 'h':
2832             help();
2833             return 0;
2834         case 'f':
2835             fmt = optarg;
2836             break;
2837         case 'F':
2838             out_basefmt = optarg;
2839             break;
2840         case 'b':
2841             out_baseimg = optarg;
2842             break;
2843         case 'u':
2844             unsafe = 1;
2845             break;
2846         case 'p':
2847             progress = 1;
2848             break;
2849         case 't':
2850             cache = optarg;
2851             break;
2852         case 'T':
2853             src_cache = optarg;
2854             break;
2855         case 'q':
2856             quiet = true;
2857             break;
2858         case OPTION_OBJECT: {
2859             QemuOpts *opts;
2860             opts = qemu_opts_parse_noisily(&qemu_object_opts,
2861                                            optarg, true);
2862             if (!opts) {
2863                 return 1;
2864             }
2865         }   break;
2866         case OPTION_IMAGE_OPTS:
2867             image_opts = true;
2868             break;
2869         }
2870     }
2871 
2872     if (quiet) {
2873         progress = 0;
2874     }
2875 
2876     if (optind != argc - 1) {
2877         error_exit("Expecting one image file name");
2878     }
2879     if (!unsafe && !out_baseimg) {
2880         error_exit("Must specify backing file (-b) or use unsafe mode (-u)");
2881     }
2882     filename = argv[optind++];
2883 
2884     if (qemu_opts_foreach(&qemu_object_opts,
2885                           user_creatable_add_opts_foreach,
2886                           NULL, NULL)) {
2887         return 1;
2888     }
2889 
2890     qemu_progress_init(progress, 2.0);
2891     qemu_progress_print(0, 100);
2892 
2893     flags = BDRV_O_RDWR | (unsafe ? BDRV_O_NO_BACKING : 0);
2894     ret = bdrv_parse_cache_mode(cache, &flags, &writethrough);
2895     if (ret < 0) {
2896         error_report("Invalid cache option: %s", cache);
2897         goto out;
2898     }
2899 
2900     src_flags = 0;
2901     ret = bdrv_parse_cache_mode(src_cache, &src_flags, &src_writethrough);
2902     if (ret < 0) {
2903         error_report("Invalid source cache option: %s", src_cache);
2904         goto out;
2905     }
2906 
2907     /* The source files are opened read-only, don't care about WCE */
2908     assert((src_flags & BDRV_O_RDWR) == 0);
2909     (void) src_writethrough;
2910 
2911     /*
2912      * Open the images.
2913      *
2914      * Ignore the old backing file for unsafe rebase in case we want to correct
2915      * the reference to a renamed or moved backing file.
2916      */
2917     blk = img_open(image_opts, filename, fmt, flags, writethrough, quiet);
2918     if (!blk) {
2919         ret = -1;
2920         goto out;
2921     }
2922     bs = blk_bs(blk);
2923 
2924     if (out_basefmt != NULL) {
2925         if (bdrv_find_format(out_basefmt) == NULL) {
2926             error_report("Invalid format name: '%s'", out_basefmt);
2927             ret = -1;
2928             goto out;
2929         }
2930     }
2931 
2932     /* For safe rebasing we need to compare old and new backing file */
2933     if (!unsafe) {
2934         char backing_name[PATH_MAX];
2935         QDict *options = NULL;
2936 
2937         if (bs->backing_format[0] != '\0') {
2938             options = qdict_new();
2939             qdict_put(options, "driver", qstring_from_str(bs->backing_format));
2940         }
2941 
2942         bdrv_get_backing_filename(bs, backing_name, sizeof(backing_name));
2943         blk_old_backing = blk_new_open(backing_name, NULL,
2944                                        options, src_flags, &local_err);
2945         if (!blk_old_backing) {
2946             error_reportf_err(local_err,
2947                               "Could not open old backing file '%s': ",
2948                               backing_name);
2949             goto out;
2950         }
2951 
2952         if (out_baseimg[0]) {
2953             if (out_basefmt) {
2954                 options = qdict_new();
2955                 qdict_put(options, "driver", qstring_from_str(out_basefmt));
2956             } else {
2957                 options = NULL;
2958             }
2959 
2960             blk_new_backing = blk_new_open(out_baseimg, NULL,
2961                                            options, src_flags, &local_err);
2962             if (!blk_new_backing) {
2963                 error_reportf_err(local_err,
2964                                   "Could not open new backing file '%s': ",
2965                                   out_baseimg);
2966                 goto out;
2967             }
2968         }
2969     }
2970 
2971     /*
2972      * Check each unallocated cluster in the COW file. If it is unallocated,
2973      * accesses go to the backing file. We must therefore compare this cluster
2974      * in the old and new backing file, and if they differ we need to copy it
2975      * from the old backing file into the COW file.
2976      *
2977      * If qemu-img crashes during this step, no harm is done. The content of
2978      * the image is the same as the original one at any time.
2979      */
2980     if (!unsafe) {
2981         int64_t num_sectors;
2982         int64_t old_backing_num_sectors;
2983         int64_t new_backing_num_sectors = 0;
2984         uint64_t sector;
2985         int n;
2986         float local_progress = 0;
2987 
2988         buf_old = blk_blockalign(blk, IO_BUF_SIZE);
2989         buf_new = blk_blockalign(blk, IO_BUF_SIZE);
2990 
2991         num_sectors = blk_nb_sectors(blk);
2992         if (num_sectors < 0) {
2993             error_report("Could not get size of '%s': %s",
2994                          filename, strerror(-num_sectors));
2995             ret = -1;
2996             goto out;
2997         }
2998         old_backing_num_sectors = blk_nb_sectors(blk_old_backing);
2999         if (old_backing_num_sectors < 0) {
3000             char backing_name[PATH_MAX];
3001 
3002             bdrv_get_backing_filename(bs, backing_name, sizeof(backing_name));
3003             error_report("Could not get size of '%s': %s",
3004                          backing_name, strerror(-old_backing_num_sectors));
3005             ret = -1;
3006             goto out;
3007         }
3008         if (blk_new_backing) {
3009             new_backing_num_sectors = blk_nb_sectors(blk_new_backing);
3010             if (new_backing_num_sectors < 0) {
3011                 error_report("Could not get size of '%s': %s",
3012                              out_baseimg, strerror(-new_backing_num_sectors));
3013                 ret = -1;
3014                 goto out;
3015             }
3016         }
3017 
3018         if (num_sectors != 0) {
3019             local_progress = (float)100 /
3020                 (num_sectors / MIN(num_sectors, IO_BUF_SIZE / 512));
3021         }
3022 
3023         for (sector = 0; sector < num_sectors; sector += n) {
3024 
3025             /* How many sectors can we handle with the next read? */
3026             if (sector + (IO_BUF_SIZE / 512) <= num_sectors) {
3027                 n = (IO_BUF_SIZE / 512);
3028             } else {
3029                 n = num_sectors - sector;
3030             }
3031 
3032             /* If the cluster is allocated, we don't need to take action */
3033             ret = bdrv_is_allocated(bs, sector, n, &n);
3034             if (ret < 0) {
3035                 error_report("error while reading image metadata: %s",
3036                              strerror(-ret));
3037                 goto out;
3038             }
3039             if (ret) {
3040                 continue;
3041             }
3042 
3043             /*
3044              * Read old and new backing file and take into consideration that
3045              * backing files may be smaller than the COW image.
3046              */
3047             if (sector >= old_backing_num_sectors) {
3048                 memset(buf_old, 0, n * BDRV_SECTOR_SIZE);
3049             } else {
3050                 if (sector + n > old_backing_num_sectors) {
3051                     n = old_backing_num_sectors - sector;
3052                 }
3053 
3054                 ret = blk_pread(blk_old_backing, sector << BDRV_SECTOR_BITS,
3055                                 buf_old, n << BDRV_SECTOR_BITS);
3056                 if (ret < 0) {
3057                     error_report("error while reading from old backing file");
3058                     goto out;
3059                 }
3060             }
3061 
3062             if (sector >= new_backing_num_sectors || !blk_new_backing) {
3063                 memset(buf_new, 0, n * BDRV_SECTOR_SIZE);
3064             } else {
3065                 if (sector + n > new_backing_num_sectors) {
3066                     n = new_backing_num_sectors - sector;
3067                 }
3068 
3069                 ret = blk_pread(blk_new_backing, sector << BDRV_SECTOR_BITS,
3070                                 buf_new, n << BDRV_SECTOR_BITS);
3071                 if (ret < 0) {
3072                     error_report("error while reading from new backing file");
3073                     goto out;
3074                 }
3075             }
3076 
3077             /* If they differ, we need to write to the COW file */
3078             uint64_t written = 0;
3079 
3080             while (written < n) {
3081                 int pnum;
3082 
3083                 if (compare_sectors(buf_old + written * 512,
3084                     buf_new + written * 512, n - written, &pnum))
3085                 {
3086                     ret = blk_pwrite(blk,
3087                                      (sector + written) << BDRV_SECTOR_BITS,
3088                                      buf_old + written * 512,
3089                                      pnum << BDRV_SECTOR_BITS, 0);
3090                     if (ret < 0) {
3091                         error_report("Error while writing to COW image: %s",
3092                             strerror(-ret));
3093                         goto out;
3094                     }
3095                 }
3096 
3097                 written += pnum;
3098             }
3099             qemu_progress_print(local_progress, 100);
3100         }
3101     }
3102 
3103     /*
3104      * Change the backing file. All clusters that are different from the old
3105      * backing file are overwritten in the COW file now, so the visible content
3106      * doesn't change when we switch the backing file.
3107      */
3108     if (out_baseimg && *out_baseimg) {
3109         ret = bdrv_change_backing_file(bs, out_baseimg, out_basefmt);
3110     } else {
3111         ret = bdrv_change_backing_file(bs, NULL, NULL);
3112     }
3113 
3114     if (ret == -ENOSPC) {
3115         error_report("Could not change the backing file to '%s': No "
3116                      "space left in the file header", out_baseimg);
3117     } else if (ret < 0) {
3118         error_report("Could not change the backing file to '%s': %s",
3119             out_baseimg, strerror(-ret));
3120     }
3121 
3122     qemu_progress_print(100, 0);
3123     /*
3124      * TODO At this point it is possible to check if any clusters that are
3125      * allocated in the COW file are the same in the backing file. If so, they
3126      * could be dropped from the COW file. Don't do this before switching the
3127      * backing file, in case of a crash this would lead to corruption.
3128      */
3129 out:
3130     qemu_progress_end();
3131     /* Cleanup */
3132     if (!unsafe) {
3133         blk_unref(blk_old_backing);
3134         blk_unref(blk_new_backing);
3135     }
3136     qemu_vfree(buf_old);
3137     qemu_vfree(buf_new);
3138 
3139     blk_unref(blk);
3140     if (ret) {
3141         return 1;
3142     }
3143     return 0;
3144 }
3145 
3146 static int img_resize(int argc, char **argv)
3147 {
3148     Error *err = NULL;
3149     int c, ret, relative;
3150     const char *filename, *fmt, *size;
3151     int64_t n, total_size;
3152     bool quiet = false;
3153     BlockBackend *blk = NULL;
3154     QemuOpts *param;
3155 
3156     static QemuOptsList resize_options = {
3157         .name = "resize_options",
3158         .head = QTAILQ_HEAD_INITIALIZER(resize_options.head),
3159         .desc = {
3160             {
3161                 .name = BLOCK_OPT_SIZE,
3162                 .type = QEMU_OPT_SIZE,
3163                 .help = "Virtual disk size"
3164             }, {
3165                 /* end of list */
3166             }
3167         },
3168     };
3169     bool image_opts = false;
3170 
3171     /* Remove size from argv manually so that negative numbers are not treated
3172      * as options by getopt. */
3173     if (argc < 3) {
3174         error_exit("Not enough arguments");
3175         return 1;
3176     }
3177 
3178     size = argv[--argc];
3179 
3180     /* Parse getopt arguments */
3181     fmt = NULL;
3182     for(;;) {
3183         static const struct option long_options[] = {
3184             {"help", no_argument, 0, 'h'},
3185             {"object", required_argument, 0, OPTION_OBJECT},
3186             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
3187             {0, 0, 0, 0}
3188         };
3189         c = getopt_long(argc, argv, "f:hq",
3190                         long_options, NULL);
3191         if (c == -1) {
3192             break;
3193         }
3194         switch(c) {
3195         case '?':
3196         case 'h':
3197             help();
3198             break;
3199         case 'f':
3200             fmt = optarg;
3201             break;
3202         case 'q':
3203             quiet = true;
3204             break;
3205         case OPTION_OBJECT: {
3206             QemuOpts *opts;
3207             opts = qemu_opts_parse_noisily(&qemu_object_opts,
3208                                            optarg, true);
3209             if (!opts) {
3210                 return 1;
3211             }
3212         }   break;
3213         case OPTION_IMAGE_OPTS:
3214             image_opts = true;
3215             break;
3216         }
3217     }
3218     if (optind != argc - 1) {
3219         error_exit("Expecting one image file name");
3220     }
3221     filename = argv[optind++];
3222 
3223     if (qemu_opts_foreach(&qemu_object_opts,
3224                           user_creatable_add_opts_foreach,
3225                           NULL, NULL)) {
3226         return 1;
3227     }
3228 
3229     /* Choose grow, shrink, or absolute resize mode */
3230     switch (size[0]) {
3231     case '+':
3232         relative = 1;
3233         size++;
3234         break;
3235     case '-':
3236         relative = -1;
3237         size++;
3238         break;
3239     default:
3240         relative = 0;
3241         break;
3242     }
3243 
3244     /* Parse size */
3245     param = qemu_opts_create(&resize_options, NULL, 0, &error_abort);
3246     qemu_opt_set(param, BLOCK_OPT_SIZE, size, &err);
3247     if (err) {
3248         error_report_err(err);
3249         ret = -1;
3250         qemu_opts_del(param);
3251         goto out;
3252     }
3253     n = qemu_opt_get_size(param, BLOCK_OPT_SIZE, 0);
3254     qemu_opts_del(param);
3255 
3256     blk = img_open(image_opts, filename, fmt,
3257                    BDRV_O_RDWR, false, quiet);
3258     if (!blk) {
3259         ret = -1;
3260         goto out;
3261     }
3262 
3263     if (relative) {
3264         total_size = blk_getlength(blk) + n * relative;
3265     } else {
3266         total_size = n;
3267     }
3268     if (total_size <= 0) {
3269         error_report("New image size must be positive");
3270         ret = -1;
3271         goto out;
3272     }
3273 
3274     ret = blk_truncate(blk, total_size);
3275     switch (ret) {
3276     case 0:
3277         qprintf(quiet, "Image resized.\n");
3278         break;
3279     case -ENOTSUP:
3280         error_report("This image does not support resize");
3281         break;
3282     case -EACCES:
3283         error_report("Image is read-only");
3284         break;
3285     default:
3286         error_report("Error resizing image: %s", strerror(-ret));
3287         break;
3288     }
3289 out:
3290     blk_unref(blk);
3291     if (ret) {
3292         return 1;
3293     }
3294     return 0;
3295 }
3296 
3297 static void amend_status_cb(BlockDriverState *bs,
3298                             int64_t offset, int64_t total_work_size,
3299                             void *opaque)
3300 {
3301     qemu_progress_print(100.f * offset / total_work_size, 0);
3302 }
3303 
3304 static int img_amend(int argc, char **argv)
3305 {
3306     Error *err = NULL;
3307     int c, ret = 0;
3308     char *options = NULL;
3309     QemuOptsList *create_opts = NULL;
3310     QemuOpts *opts = NULL;
3311     const char *fmt = NULL, *filename, *cache;
3312     int flags;
3313     bool writethrough;
3314     bool quiet = false, progress = false;
3315     BlockBackend *blk = NULL;
3316     BlockDriverState *bs = NULL;
3317     bool image_opts = false;
3318 
3319     cache = BDRV_DEFAULT_CACHE;
3320     for (;;) {
3321         static const struct option long_options[] = {
3322             {"help", no_argument, 0, 'h'},
3323             {"object", required_argument, 0, OPTION_OBJECT},
3324             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
3325             {0, 0, 0, 0}
3326         };
3327         c = getopt_long(argc, argv, "ho:f:t:pq",
3328                         long_options, NULL);
3329         if (c == -1) {
3330             break;
3331         }
3332 
3333         switch (c) {
3334             case 'h':
3335             case '?':
3336                 help();
3337                 break;
3338             case 'o':
3339                 if (!is_valid_option_list(optarg)) {
3340                     error_report("Invalid option list: %s", optarg);
3341                     ret = -1;
3342                     goto out_no_progress;
3343                 }
3344                 if (!options) {
3345                     options = g_strdup(optarg);
3346                 } else {
3347                     char *old_options = options;
3348                     options = g_strdup_printf("%s,%s", options, optarg);
3349                     g_free(old_options);
3350                 }
3351                 break;
3352             case 'f':
3353                 fmt = optarg;
3354                 break;
3355             case 't':
3356                 cache = optarg;
3357                 break;
3358             case 'p':
3359                 progress = true;
3360                 break;
3361             case 'q':
3362                 quiet = true;
3363                 break;
3364             case OPTION_OBJECT:
3365                 opts = qemu_opts_parse_noisily(&qemu_object_opts,
3366                                                optarg, true);
3367                 if (!opts) {
3368                     ret = -1;
3369                     goto out_no_progress;
3370                 }
3371                 break;
3372             case OPTION_IMAGE_OPTS:
3373                 image_opts = true;
3374                 break;
3375         }
3376     }
3377 
3378     if (!options) {
3379         error_exit("Must specify options (-o)");
3380     }
3381 
3382     if (qemu_opts_foreach(&qemu_object_opts,
3383                           user_creatable_add_opts_foreach,
3384                           NULL, NULL)) {
3385         ret = -1;
3386         goto out_no_progress;
3387     }
3388 
3389     if (quiet) {
3390         progress = false;
3391     }
3392     qemu_progress_init(progress, 1.0);
3393 
3394     filename = (optind == argc - 1) ? argv[argc - 1] : NULL;
3395     if (fmt && has_help_option(options)) {
3396         /* If a format is explicitly specified (and possibly no filename is
3397          * given), print option help here */
3398         ret = print_block_option_help(filename, fmt);
3399         goto out;
3400     }
3401 
3402     if (optind != argc - 1) {
3403         error_report("Expecting one image file name");
3404         ret = -1;
3405         goto out;
3406     }
3407 
3408     flags = BDRV_O_RDWR;
3409     ret = bdrv_parse_cache_mode(cache, &flags, &writethrough);
3410     if (ret < 0) {
3411         error_report("Invalid cache option: %s", cache);
3412         goto out;
3413     }
3414 
3415     blk = img_open(image_opts, filename, fmt, flags, writethrough, quiet);
3416     if (!blk) {
3417         ret = -1;
3418         goto out;
3419     }
3420     bs = blk_bs(blk);
3421 
3422     fmt = bs->drv->format_name;
3423 
3424     if (has_help_option(options)) {
3425         /* If the format was auto-detected, print option help here */
3426         ret = print_block_option_help(filename, fmt);
3427         goto out;
3428     }
3429 
3430     if (!bs->drv->create_opts) {
3431         error_report("Format driver '%s' does not support any options to amend",
3432                      fmt);
3433         ret = -1;
3434         goto out;
3435     }
3436 
3437     create_opts = qemu_opts_append(create_opts, bs->drv->create_opts);
3438     opts = qemu_opts_create(create_opts, NULL, 0, &error_abort);
3439     if (options) {
3440         qemu_opts_do_parse(opts, options, NULL, &err);
3441         if (err) {
3442             error_report_err(err);
3443             ret = -1;
3444             goto out;
3445         }
3446     }
3447 
3448     /* In case the driver does not call amend_status_cb() */
3449     qemu_progress_print(0.f, 0);
3450     ret = bdrv_amend_options(bs, opts, &amend_status_cb, NULL);
3451     qemu_progress_print(100.f, 0);
3452     if (ret < 0) {
3453         error_report("Error while amending options: %s", strerror(-ret));
3454         goto out;
3455     }
3456 
3457 out:
3458     qemu_progress_end();
3459 
3460 out_no_progress:
3461     blk_unref(blk);
3462     qemu_opts_del(opts);
3463     qemu_opts_free(create_opts);
3464     g_free(options);
3465 
3466     if (ret) {
3467         return 1;
3468     }
3469     return 0;
3470 }
3471 
3472 typedef struct BenchData {
3473     BlockBackend *blk;
3474     uint64_t image_size;
3475     bool write;
3476     int bufsize;
3477     int step;
3478     int nrreq;
3479     int n;
3480     int flush_interval;
3481     bool drain_on_flush;
3482     uint8_t *buf;
3483     QEMUIOVector *qiov;
3484 
3485     int in_flight;
3486     bool in_flush;
3487     uint64_t offset;
3488 } BenchData;
3489 
3490 static void bench_undrained_flush_cb(void *opaque, int ret)
3491 {
3492     if (ret < 0) {
3493         error_report("Failed flush request: %s", strerror(-ret));
3494         exit(EXIT_FAILURE);
3495     }
3496 }
3497 
3498 static void bench_cb(void *opaque, int ret)
3499 {
3500     BenchData *b = opaque;
3501     BlockAIOCB *acb;
3502 
3503     if (ret < 0) {
3504         error_report("Failed request: %s", strerror(-ret));
3505         exit(EXIT_FAILURE);
3506     }
3507 
3508     if (b->in_flush) {
3509         /* Just finished a flush with drained queue: Start next requests */
3510         assert(b->in_flight == 0);
3511         b->in_flush = false;
3512     } else if (b->in_flight > 0) {
3513         int remaining = b->n - b->in_flight;
3514 
3515         b->n--;
3516         b->in_flight--;
3517 
3518         /* Time for flush? Drain queue if requested, then flush */
3519         if (b->flush_interval && remaining % b->flush_interval == 0) {
3520             if (!b->in_flight || !b->drain_on_flush) {
3521                 BlockCompletionFunc *cb;
3522 
3523                 if (b->drain_on_flush) {
3524                     b->in_flush = true;
3525                     cb = bench_cb;
3526                 } else {
3527                     cb = bench_undrained_flush_cb;
3528                 }
3529 
3530                 acb = blk_aio_flush(b->blk, cb, b);
3531                 if (!acb) {
3532                     error_report("Failed to issue flush request");
3533                     exit(EXIT_FAILURE);
3534                 }
3535             }
3536             if (b->drain_on_flush) {
3537                 return;
3538             }
3539         }
3540     }
3541 
3542     while (b->n > b->in_flight && b->in_flight < b->nrreq) {
3543         if (b->write) {
3544             acb = blk_aio_pwritev(b->blk, b->offset, b->qiov, 0,
3545                                   bench_cb, b);
3546         } else {
3547             acb = blk_aio_preadv(b->blk, b->offset, b->qiov, 0,
3548                                  bench_cb, b);
3549         }
3550         if (!acb) {
3551             error_report("Failed to issue request");
3552             exit(EXIT_FAILURE);
3553         }
3554         b->in_flight++;
3555         b->offset += b->step;
3556         b->offset %= b->image_size;
3557     }
3558 }
3559 
3560 static int img_bench(int argc, char **argv)
3561 {
3562     int c, ret = 0;
3563     const char *fmt = NULL, *filename;
3564     bool quiet = false;
3565     bool image_opts = false;
3566     bool is_write = false;
3567     int count = 75000;
3568     int depth = 64;
3569     int64_t offset = 0;
3570     size_t bufsize = 4096;
3571     int pattern = 0;
3572     size_t step = 0;
3573     int flush_interval = 0;
3574     bool drain_on_flush = true;
3575     int64_t image_size;
3576     BlockBackend *blk = NULL;
3577     BenchData data = {};
3578     int flags = 0;
3579     bool writethrough = false;
3580     struct timeval t1, t2;
3581     int i;
3582 
3583     for (;;) {
3584         static const struct option long_options[] = {
3585             {"help", no_argument, 0, 'h'},
3586             {"flush-interval", required_argument, 0, OPTION_FLUSH_INTERVAL},
3587             {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
3588             {"pattern", required_argument, 0, OPTION_PATTERN},
3589             {"no-drain", no_argument, 0, OPTION_NO_DRAIN},
3590             {0, 0, 0, 0}
3591         };
3592         c = getopt_long(argc, argv, "hc:d:f:no:qs:S:t:w", long_options, NULL);
3593         if (c == -1) {
3594             break;
3595         }
3596 
3597         switch (c) {
3598         case 'h':
3599         case '?':
3600             help();
3601             break;
3602         case 'c':
3603         {
3604             char *end;
3605             errno = 0;
3606             count = strtoul(optarg, &end, 0);
3607             if (errno || *end || count > INT_MAX) {
3608                 error_report("Invalid request count specified");
3609                 return 1;
3610             }
3611             break;
3612         }
3613         case 'd':
3614         {
3615             char *end;
3616             errno = 0;
3617             depth = strtoul(optarg, &end, 0);
3618             if (errno || *end || depth > INT_MAX) {
3619                 error_report("Invalid queue depth specified");
3620                 return 1;
3621             }
3622             break;
3623         }
3624         case 'f':
3625             fmt = optarg;
3626             break;
3627         case 'n':
3628             flags |= BDRV_O_NATIVE_AIO;
3629             break;
3630         case 'o':
3631         {
3632             char *end;
3633             errno = 0;
3634             offset = qemu_strtosz_suffix(optarg, &end,
3635                                          QEMU_STRTOSZ_DEFSUFFIX_B);
3636             if (offset < 0|| *end) {
3637                 error_report("Invalid offset specified");
3638                 return 1;
3639             }
3640             break;
3641         }
3642             break;
3643         case 'q':
3644             quiet = true;
3645             break;
3646         case 's':
3647         {
3648             int64_t sval;
3649             char *end;
3650 
3651             sval = qemu_strtosz_suffix(optarg, &end, QEMU_STRTOSZ_DEFSUFFIX_B);
3652             if (sval < 0 || sval > INT_MAX || *end) {
3653                 error_report("Invalid buffer size specified");
3654                 return 1;
3655             }
3656 
3657             bufsize = sval;
3658             break;
3659         }
3660         case 'S':
3661         {
3662             int64_t sval;
3663             char *end;
3664 
3665             sval = qemu_strtosz_suffix(optarg, &end, QEMU_STRTOSZ_DEFSUFFIX_B);
3666             if (sval < 0 || sval > INT_MAX || *end) {
3667                 error_report("Invalid step size specified");
3668                 return 1;
3669             }
3670 
3671             step = sval;
3672             break;
3673         }
3674         case 't':
3675             ret = bdrv_parse_cache_mode(optarg, &flags, &writethrough);
3676             if (ret < 0) {
3677                 error_report("Invalid cache mode");
3678                 ret = -1;
3679                 goto out;
3680             }
3681             break;
3682         case 'w':
3683             flags |= BDRV_O_RDWR;
3684             is_write = true;
3685             break;
3686         case OPTION_PATTERN:
3687         {
3688             char *end;
3689             errno = 0;
3690             pattern = strtoul(optarg, &end, 0);
3691             if (errno || *end || pattern > 0xff) {
3692                 error_report("Invalid pattern byte specified");
3693                 return 1;
3694             }
3695             break;
3696         }
3697         case OPTION_FLUSH_INTERVAL:
3698         {
3699             char *end;
3700             errno = 0;
3701             flush_interval = strtoul(optarg, &end, 0);
3702             if (errno || *end || flush_interval > INT_MAX) {
3703                 error_report("Invalid flush interval specified");
3704                 return 1;
3705             }
3706             break;
3707         }
3708         case OPTION_NO_DRAIN:
3709             drain_on_flush = false;
3710             break;
3711         case OPTION_IMAGE_OPTS:
3712             image_opts = true;
3713             break;
3714         }
3715     }
3716 
3717     if (optind != argc - 1) {
3718         error_exit("Expecting one image file name");
3719     }
3720     filename = argv[argc - 1];
3721 
3722     if (!is_write && flush_interval) {
3723         error_report("--flush-interval is only available in write tests");
3724         ret = -1;
3725         goto out;
3726     }
3727     if (flush_interval && flush_interval < depth) {
3728         error_report("Flush interval can't be smaller than depth");
3729         ret = -1;
3730         goto out;
3731     }
3732 
3733     blk = img_open(image_opts, filename, fmt, flags, writethrough, quiet);
3734     if (!blk) {
3735         ret = -1;
3736         goto out;
3737     }
3738 
3739     image_size = blk_getlength(blk);
3740     if (image_size < 0) {
3741         ret = image_size;
3742         goto out;
3743     }
3744 
3745     data = (BenchData) {
3746         .blk            = blk,
3747         .image_size     = image_size,
3748         .bufsize        = bufsize,
3749         .step           = step ?: bufsize,
3750         .nrreq          = depth,
3751         .n              = count,
3752         .offset         = offset,
3753         .write          = is_write,
3754         .flush_interval = flush_interval,
3755         .drain_on_flush = drain_on_flush,
3756     };
3757     printf("Sending %d %s requests, %d bytes each, %d in parallel "
3758            "(starting at offset %" PRId64 ", step size %d)\n",
3759            data.n, data.write ? "write" : "read", data.bufsize, data.nrreq,
3760            data.offset, data.step);
3761     if (flush_interval) {
3762         printf("Sending flush every %d requests\n", flush_interval);
3763     }
3764 
3765     data.buf = blk_blockalign(blk, data.nrreq * data.bufsize);
3766     memset(data.buf, pattern, data.nrreq * data.bufsize);
3767 
3768     data.qiov = g_new(QEMUIOVector, data.nrreq);
3769     for (i = 0; i < data.nrreq; i++) {
3770         qemu_iovec_init(&data.qiov[i], 1);
3771         qemu_iovec_add(&data.qiov[i],
3772                        data.buf + i * data.bufsize, data.bufsize);
3773     }
3774 
3775     gettimeofday(&t1, NULL);
3776     bench_cb(&data, 0);
3777 
3778     while (data.n > 0) {
3779         main_loop_wait(false);
3780     }
3781     gettimeofday(&t2, NULL);
3782 
3783     printf("Run completed in %3.3f seconds.\n",
3784            (t2.tv_sec - t1.tv_sec)
3785            + ((double)(t2.tv_usec - t1.tv_usec) / 1000000));
3786 
3787 out:
3788     qemu_vfree(data.buf);
3789     blk_unref(blk);
3790 
3791     if (ret) {
3792         return 1;
3793     }
3794     return 0;
3795 }
3796 
3797 
3798 static const img_cmd_t img_cmds[] = {
3799 #define DEF(option, callback, arg_string)        \
3800     { option, callback },
3801 #include "qemu-img-cmds.h"
3802 #undef DEF
3803 #undef GEN_DOCS
3804     { NULL, NULL, },
3805 };
3806 
3807 int main(int argc, char **argv)
3808 {
3809     const img_cmd_t *cmd;
3810     const char *cmdname;
3811     Error *local_error = NULL;
3812     char *trace_file = NULL;
3813     int c;
3814     static const struct option long_options[] = {
3815         {"help", no_argument, 0, 'h'},
3816         {"version", no_argument, 0, 'V'},
3817         {"trace", required_argument, NULL, 'T'},
3818         {0, 0, 0, 0}
3819     };
3820 
3821 #ifdef CONFIG_POSIX
3822     signal(SIGPIPE, SIG_IGN);
3823 #endif
3824 
3825     error_set_progname(argv[0]);
3826     qemu_init_exec_dir(argv[0]);
3827 
3828     if (qemu_init_main_loop(&local_error)) {
3829         error_report_err(local_error);
3830         exit(EXIT_FAILURE);
3831     }
3832 
3833     qcrypto_init(&error_fatal);
3834 
3835     module_call_init(MODULE_INIT_QOM);
3836     bdrv_init();
3837     if (argc < 2) {
3838         error_exit("Not enough arguments");
3839     }
3840 
3841     qemu_add_opts(&qemu_object_opts);
3842     qemu_add_opts(&qemu_source_opts);
3843     qemu_add_opts(&qemu_trace_opts);
3844 
3845     while ((c = getopt_long(argc, argv, "+hVT:", long_options, NULL)) != -1) {
3846         switch (c) {
3847         case 'h':
3848             help();
3849             return 0;
3850         case 'V':
3851             printf(QEMU_IMG_VERSION);
3852             return 0;
3853         case 'T':
3854             g_free(trace_file);
3855             trace_file = trace_opt_parse(optarg);
3856             break;
3857         }
3858     }
3859 
3860     cmdname = argv[optind];
3861 
3862     /* reset getopt_long scanning */
3863     argc -= optind;
3864     if (argc < 1) {
3865         return 0;
3866     }
3867     argv += optind;
3868     optind = 0;
3869 
3870     if (!trace_init_backends()) {
3871         exit(1);
3872     }
3873     trace_init_file(trace_file);
3874     qemu_set_log(LOG_TRACE);
3875 
3876     /* find the command */
3877     for (cmd = img_cmds; cmd->name != NULL; cmd++) {
3878         if (!strcmp(cmdname, cmd->name)) {
3879             return cmd->handler(argc, argv);
3880         }
3881     }
3882 
3883     /* not found */
3884     error_exit("Command not found: %s", cmdname);
3885 }
3886