xref: /openbmc/qemu/qemu-img.c (revision 10358b6a)
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 "qapi-visit.h"
25 #include "qapi/qmp-output-visitor.h"
26 #include "qapi/qmp/qjson.h"
27 #include "qemu-common.h"
28 #include "qemu/option.h"
29 #include "qemu/error-report.h"
30 #include "qemu/osdep.h"
31 #include "sysemu/sysemu.h"
32 #include "block/block_int.h"
33 #include "block/qapi.h"
34 #include <getopt.h>
35 #include <glib.h>
36 
37 #define QEMU_IMG_VERSION "qemu-img version " QEMU_VERSION \
38                           ", Copyright (c) 2004-2008 Fabrice Bellard\n"
39 
40 typedef struct img_cmd_t {
41     const char *name;
42     int (*handler)(int argc, char **argv);
43 } img_cmd_t;
44 
45 enum {
46     OPTION_OUTPUT = 256,
47     OPTION_BACKING_CHAIN = 257,
48 };
49 
50 typedef enum OutputFormat {
51     OFORMAT_JSON,
52     OFORMAT_HUMAN,
53 } OutputFormat;
54 
55 /* Default to cache=writeback as data integrity is not important for qemu-tcg. */
56 #define BDRV_O_FLAGS BDRV_O_CACHE_WB
57 #define BDRV_DEFAULT_CACHE "writeback"
58 
59 static gint compare_data(gconstpointer a, gconstpointer b, gpointer user)
60 {
61     return g_strcmp0(a, b);
62 }
63 
64 static void print_format(gpointer data, gpointer user)
65 {
66     printf(" %s", (char *)data);
67 }
68 
69 static void add_format_to_seq(void *opaque, const char *fmt_name)
70 {
71     GSequence *seq = opaque;
72 
73     g_sequence_insert_sorted(seq, (gpointer)fmt_name,
74                              compare_data, NULL);
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 command [command options]\n"
97            "QEMU disk image utility\n"
98            "\n"
99            "Command syntax:\n"
100 #define DEF(option, callback, arg_string)        \
101            "  " arg_string "\n"
102 #include "qemu-img-cmds.h"
103 #undef DEF
104 #undef GEN_DOCS
105            "\n"
106            "Command parameters:\n"
107            "  'filename' is a disk image filename\n"
108            "  'fmt' is the disk image format. It is guessed automatically in most cases\n"
109            "  'cache' is the cache mode used to write the output disk image, the valid\n"
110            "    options are: 'none', 'writeback' (default, except for convert), 'writethrough',\n"
111            "    'directsync' and 'unsafe' (default for convert)\n"
112            "  'size' is the disk image size in bytes. Optional suffixes\n"
113            "    'k' or 'K' (kilobyte, 1024), 'M' (megabyte, 1024k), 'G' (gigabyte, 1024M),\n"
114            "    'T' (terabyte, 1024G), 'P' (petabyte, 1024T) and 'E' (exabyte, 1024P)  are\n"
115            "    supported. 'b' is ignored.\n"
116            "  'output_filename' is the destination disk image filename\n"
117            "  'output_fmt' is the destination format\n"
118            "  'options' is a comma separated list of format specific options in a\n"
119            "    name=value format. Use -o ? for an overview of the options supported by the\n"
120            "    used format\n"
121            "  'snapshot_param' is param used for internal snapshot, format\n"
122            "    is 'snapshot.id=[ID],snapshot.name=[NAME]', or\n"
123            "    '[ID_OR_NAME]'\n"
124            "  'snapshot_id_or_name' is deprecated, use 'snapshot_param'\n"
125            "    instead\n"
126            "  '-c' indicates that target image must be compressed (qcow format only)\n"
127            "  '-u' enables unsafe rebasing. It is assumed that old and new backing file\n"
128            "       match exactly. The image doesn't need a working backing file before\n"
129            "       rebasing in this case (useful for renaming the backing file)\n"
130            "  '-h' with or without a command shows this help and lists the supported formats\n"
131            "  '-p' show progress of command (only certain commands)\n"
132            "  '-q' use Quiet mode - do not print any output (except errors)\n"
133            "  '-S' indicates the consecutive number of bytes (defaults to 4k) that must\n"
134            "       contain only zeros for qemu-img to create a sparse image during\n"
135            "       conversion. If the number of bytes is 0, the source will not be scanned for\n"
136            "       unallocated or zero sectors, and the destination image will always be\n"
137            "       fully allocated\n"
138            "  '--output' takes the format in which the output must be done (human or json)\n"
139            "  '-n' skips the target volume creation (useful if the volume is created\n"
140            "       prior to running qemu-img)\n"
141            "\n"
142            "Parameters to check subcommand:\n"
143            "  '-r' tries to repair any inconsistencies that are found during the check.\n"
144            "       '-r leaks' repairs only cluster leaks, whereas '-r all' fixes all\n"
145            "       kinds of errors, with a higher risk of choosing the wrong fix or\n"
146            "       hiding corruption that has already occurred.\n"
147            "\n"
148            "Parameters to snapshot subcommand:\n"
149            "  'snapshot' is the name of the snapshot to create, apply or delete\n"
150            "  '-a' applies a snapshot (revert disk to saved state)\n"
151            "  '-c' creates a snapshot\n"
152            "  '-d' deletes a snapshot\n"
153            "  '-l' lists all snapshots in the given image\n"
154            "\n"
155            "Parameters to compare subcommand:\n"
156            "  '-f' first image format\n"
157            "  '-F' second image format\n"
158            "  '-s' run in Strict mode - fail on different image size or sector allocation\n";
159     GSequence *seq;
160 
161     printf("%s\nSupported formats:", help_msg);
162     seq = g_sequence_new(NULL);
163     bdrv_iterate_format(add_format_to_seq, seq);
164     g_sequence_foreach(seq, print_format, NULL);
165     printf("\n");
166     g_sequence_free(seq);
167 
168     exit(EXIT_SUCCESS);
169 }
170 
171 static int GCC_FMT_ATTR(2, 3) qprintf(bool quiet, const char *fmt, ...)
172 {
173     int ret = 0;
174     if (!quiet) {
175         va_list args;
176         va_start(args, fmt);
177         ret = vprintf(fmt, args);
178         va_end(args);
179     }
180     return ret;
181 }
182 
183 #if defined(WIN32)
184 /* XXX: put correct support for win32 */
185 static int read_password(char *buf, int buf_size)
186 {
187     int c, i;
188     printf("Password: ");
189     fflush(stdout);
190     i = 0;
191     for(;;) {
192         c = getchar();
193         if (c == '\n')
194             break;
195         if (i < (buf_size - 1))
196             buf[i++] = c;
197     }
198     buf[i] = '\0';
199     return 0;
200 }
201 
202 #else
203 
204 #include <termios.h>
205 
206 static struct termios oldtty;
207 
208 static void term_exit(void)
209 {
210     tcsetattr (0, TCSANOW, &oldtty);
211 }
212 
213 static void term_init(void)
214 {
215     struct termios tty;
216 
217     tcgetattr (0, &tty);
218     oldtty = tty;
219 
220     tty.c_iflag &= ~(IGNBRK|BRKINT|PARMRK|ISTRIP
221                           |INLCR|IGNCR|ICRNL|IXON);
222     tty.c_oflag |= OPOST;
223     tty.c_lflag &= ~(ECHO|ECHONL|ICANON|IEXTEN);
224     tty.c_cflag &= ~(CSIZE|PARENB);
225     tty.c_cflag |= CS8;
226     tty.c_cc[VMIN] = 1;
227     tty.c_cc[VTIME] = 0;
228 
229     tcsetattr (0, TCSANOW, &tty);
230 
231     atexit(term_exit);
232 }
233 
234 static int read_password(char *buf, int buf_size)
235 {
236     uint8_t ch;
237     int i, ret;
238 
239     printf("password: ");
240     fflush(stdout);
241     term_init();
242     i = 0;
243     for(;;) {
244         ret = read(0, &ch, 1);
245         if (ret == -1) {
246             if (errno == EAGAIN || errno == EINTR) {
247                 continue;
248             } else {
249                 ret = -1;
250                 break;
251             }
252         } else if (ret == 0) {
253             ret = -1;
254             break;
255         } else {
256             if (ch == '\r') {
257                 ret = 0;
258                 break;
259             }
260             if (i < (buf_size - 1))
261                 buf[i++] = ch;
262         }
263     }
264     term_exit();
265     buf[i] = '\0';
266     printf("\n");
267     return ret;
268 }
269 #endif
270 
271 static int print_block_option_help(const char *filename, const char *fmt)
272 {
273     BlockDriver *drv, *proto_drv;
274     QemuOptsList *create_opts = NULL;
275 
276     /* Find driver and parse its options */
277     drv = bdrv_find_format(fmt);
278     if (!drv) {
279         error_report("Unknown file format '%s'", fmt);
280         return 1;
281     }
282 
283     create_opts = qemu_opts_append(create_opts, drv->create_opts);
284     if (filename) {
285         proto_drv = bdrv_find_protocol(filename, true);
286         if (!proto_drv) {
287             error_report("Unknown protocol '%s'", filename);
288             qemu_opts_free(create_opts);
289             return 1;
290         }
291         create_opts = qemu_opts_append(create_opts, proto_drv->create_opts);
292     }
293 
294     qemu_opts_print_help(create_opts);
295     qemu_opts_free(create_opts);
296     return 0;
297 }
298 
299 static BlockDriverState *bdrv_new_open(const char *id,
300                                        const char *filename,
301                                        const char *fmt,
302                                        int flags,
303                                        bool require_io,
304                                        bool quiet)
305 {
306     BlockDriverState *bs;
307     BlockDriver *drv;
308     char password[256];
309     Error *local_err = NULL;
310     int ret;
311 
312     bs = bdrv_new(id, &error_abort);
313 
314     if (fmt) {
315         drv = bdrv_find_format(fmt);
316         if (!drv) {
317             error_report("Unknown file format '%s'", fmt);
318             goto fail;
319         }
320     } else {
321         drv = NULL;
322     }
323 
324     ret = bdrv_open(&bs, filename, NULL, NULL, flags, drv, &local_err);
325     if (ret < 0) {
326         error_report("Could not open '%s': %s", filename,
327                      error_get_pretty(local_err));
328         error_free(local_err);
329         goto fail;
330     }
331 
332     if (bdrv_is_encrypted(bs) && require_io) {
333         qprintf(quiet, "Disk image '%s' is encrypted.\n", filename);
334         if (read_password(password, sizeof(password)) < 0) {
335             error_report("No password given");
336             goto fail;
337         }
338         if (bdrv_set_key(bs, password) < 0) {
339             error_report("invalid password");
340             goto fail;
341         }
342     }
343     return bs;
344 fail:
345     bdrv_unref(bs);
346     return NULL;
347 }
348 
349 static int add_old_style_options(const char *fmt, QemuOpts *opts,
350                                  const char *base_filename,
351                                  const char *base_fmt)
352 {
353     if (base_filename) {
354         if (qemu_opt_set(opts, BLOCK_OPT_BACKING_FILE, base_filename)) {
355             error_report("Backing file not supported for file format '%s'",
356                          fmt);
357             return -1;
358         }
359     }
360     if (base_fmt) {
361         if (qemu_opt_set(opts, BLOCK_OPT_BACKING_FMT, base_fmt)) {
362             error_report("Backing file format not supported for file "
363                          "format '%s'", fmt);
364             return -1;
365         }
366     }
367     return 0;
368 }
369 
370 static int img_create(int argc, char **argv)
371 {
372     int c;
373     uint64_t img_size = -1;
374     const char *fmt = "raw";
375     const char *base_fmt = NULL;
376     const char *filename;
377     const char *base_filename = NULL;
378     char *options = NULL;
379     Error *local_err = NULL;
380     bool quiet = false;
381 
382     for(;;) {
383         c = getopt(argc, argv, "F:b:f:he6o:q");
384         if (c == -1) {
385             break;
386         }
387         switch(c) {
388         case '?':
389         case 'h':
390             help();
391             break;
392         case 'F':
393             base_fmt = optarg;
394             break;
395         case 'b':
396             base_filename = optarg;
397             break;
398         case 'f':
399             fmt = optarg;
400             break;
401         case 'e':
402             error_report("option -e is deprecated, please use \'-o "
403                   "encryption\' instead!");
404             goto fail;
405         case '6':
406             error_report("option -6 is deprecated, please use \'-o "
407                   "compat6\' instead!");
408             goto fail;
409         case 'o':
410             if (!is_valid_option_list(optarg)) {
411                 error_report("Invalid option list: %s", optarg);
412                 goto fail;
413             }
414             if (!options) {
415                 options = g_strdup(optarg);
416             } else {
417                 char *old_options = options;
418                 options = g_strdup_printf("%s,%s", options, optarg);
419                 g_free(old_options);
420             }
421             break;
422         case 'q':
423             quiet = true;
424             break;
425         }
426     }
427 
428     /* Get the filename */
429     filename = (optind < argc) ? argv[optind] : NULL;
430     if (options && has_help_option(options)) {
431         g_free(options);
432         return print_block_option_help(filename, fmt);
433     }
434 
435     if (optind >= argc) {
436         error_exit("Expecting image file name");
437     }
438     optind++;
439 
440     /* Get image size, if specified */
441     if (optind < argc) {
442         int64_t sval;
443         char *end;
444         sval = strtosz_suffix(argv[optind++], &end, STRTOSZ_DEFSUFFIX_B);
445         if (sval < 0 || *end) {
446             if (sval == -ERANGE) {
447                 error_report("Image size must be less than 8 EiB!");
448             } else {
449                 error_report("Invalid image size specified! You may use k, M, "
450                       "G, T, P or E suffixes for ");
451                 error_report("kilobytes, megabytes, gigabytes, terabytes, "
452                              "petabytes and exabytes.");
453             }
454             goto fail;
455         }
456         img_size = (uint64_t)sval;
457     }
458     if (optind != argc) {
459         error_exit("Unexpected argument: %s", argv[optind]);
460     }
461 
462     bdrv_img_create(filename, fmt, base_filename, base_fmt,
463                     options, img_size, BDRV_O_FLAGS, &local_err, quiet);
464     if (local_err) {
465         error_report("%s: %s", filename, error_get_pretty(local_err));
466         error_free(local_err);
467         goto fail;
468     }
469 
470     g_free(options);
471     return 0;
472 
473 fail:
474     g_free(options);
475     return 1;
476 }
477 
478 static void dump_json_image_check(ImageCheck *check, bool quiet)
479 {
480     Error *local_err = NULL;
481     QString *str;
482     QmpOutputVisitor *ov = qmp_output_visitor_new();
483     QObject *obj;
484     visit_type_ImageCheck(qmp_output_get_visitor(ov),
485                           &check, NULL, &local_err);
486     obj = qmp_output_get_qobject(ov);
487     str = qobject_to_json_pretty(obj);
488     assert(str != NULL);
489     qprintf(quiet, "%s\n", qstring_get_str(str));
490     qobject_decref(obj);
491     qmp_output_visitor_cleanup(ov);
492     QDECREF(str);
493 }
494 
495 static void dump_human_image_check(ImageCheck *check, bool quiet)
496 {
497     if (!(check->corruptions || check->leaks || check->check_errors)) {
498         qprintf(quiet, "No errors were found on the image.\n");
499     } else {
500         if (check->corruptions) {
501             qprintf(quiet, "\n%" PRId64 " errors were found on the image.\n"
502                     "Data may be corrupted, or further writes to the image "
503                     "may corrupt it.\n",
504                     check->corruptions);
505         }
506 
507         if (check->leaks) {
508             qprintf(quiet,
509                     "\n%" PRId64 " leaked clusters were found on the image.\n"
510                     "This means waste of disk space, but no harm to data.\n",
511                     check->leaks);
512         }
513 
514         if (check->check_errors) {
515             qprintf(quiet,
516                     "\n%" PRId64
517                     " internal errors have occurred during the check.\n",
518                     check->check_errors);
519         }
520     }
521 
522     if (check->total_clusters != 0 && check->allocated_clusters != 0) {
523         qprintf(quiet, "%" PRId64 "/%" PRId64 " = %0.2f%% allocated, "
524                 "%0.2f%% fragmented, %0.2f%% compressed clusters\n",
525                 check->allocated_clusters, check->total_clusters,
526                 check->allocated_clusters * 100.0 / check->total_clusters,
527                 check->fragmented_clusters * 100.0 / check->allocated_clusters,
528                 check->compressed_clusters * 100.0 /
529                 check->allocated_clusters);
530     }
531 
532     if (check->image_end_offset) {
533         qprintf(quiet,
534                 "Image end offset: %" PRId64 "\n", check->image_end_offset);
535     }
536 }
537 
538 static int collect_image_check(BlockDriverState *bs,
539                    ImageCheck *check,
540                    const char *filename,
541                    const char *fmt,
542                    int fix)
543 {
544     int ret;
545     BdrvCheckResult result;
546 
547     ret = bdrv_check(bs, &result, fix);
548     if (ret < 0) {
549         return ret;
550     }
551 
552     check->filename                 = g_strdup(filename);
553     check->format                   = g_strdup(bdrv_get_format_name(bs));
554     check->check_errors             = result.check_errors;
555     check->corruptions              = result.corruptions;
556     check->has_corruptions          = result.corruptions != 0;
557     check->leaks                    = result.leaks;
558     check->has_leaks                = result.leaks != 0;
559     check->corruptions_fixed        = result.corruptions_fixed;
560     check->has_corruptions_fixed    = result.corruptions != 0;
561     check->leaks_fixed              = result.leaks_fixed;
562     check->has_leaks_fixed          = result.leaks != 0;
563     check->image_end_offset         = result.image_end_offset;
564     check->has_image_end_offset     = result.image_end_offset != 0;
565     check->total_clusters           = result.bfi.total_clusters;
566     check->has_total_clusters       = result.bfi.total_clusters != 0;
567     check->allocated_clusters       = result.bfi.allocated_clusters;
568     check->has_allocated_clusters   = result.bfi.allocated_clusters != 0;
569     check->fragmented_clusters      = result.bfi.fragmented_clusters;
570     check->has_fragmented_clusters  = result.bfi.fragmented_clusters != 0;
571     check->compressed_clusters      = result.bfi.compressed_clusters;
572     check->has_compressed_clusters  = result.bfi.compressed_clusters != 0;
573 
574     return 0;
575 }
576 
577 /*
578  * Checks an image for consistency. Exit codes:
579  *
580  *  0 - Check completed, image is good
581  *  1 - Check not completed because of internal errors
582  *  2 - Check completed, image is corrupted
583  *  3 - Check completed, image has leaked clusters, but is good otherwise
584  * 63 - Checks are not supported by the image format
585  */
586 static int img_check(int argc, char **argv)
587 {
588     int c, ret;
589     OutputFormat output_format = OFORMAT_HUMAN;
590     const char *filename, *fmt, *output;
591     BlockDriverState *bs;
592     int fix = 0;
593     int flags = BDRV_O_FLAGS | BDRV_O_CHECK;
594     ImageCheck *check;
595     bool quiet = false;
596 
597     fmt = NULL;
598     output = NULL;
599     for(;;) {
600         int option_index = 0;
601         static const struct option long_options[] = {
602             {"help", no_argument, 0, 'h'},
603             {"format", required_argument, 0, 'f'},
604             {"repair", required_argument, 0, 'r'},
605             {"output", required_argument, 0, OPTION_OUTPUT},
606             {0, 0, 0, 0}
607         };
608         c = getopt_long(argc, argv, "f:hr:q",
609                         long_options, &option_index);
610         if (c == -1) {
611             break;
612         }
613         switch(c) {
614         case '?':
615         case 'h':
616             help();
617             break;
618         case 'f':
619             fmt = optarg;
620             break;
621         case 'r':
622             flags |= BDRV_O_RDWR;
623 
624             if (!strcmp(optarg, "leaks")) {
625                 fix = BDRV_FIX_LEAKS;
626             } else if (!strcmp(optarg, "all")) {
627                 fix = BDRV_FIX_LEAKS | BDRV_FIX_ERRORS;
628             } else {
629                 error_exit("Unknown option value for -r "
630                            "(expecting 'leaks' or 'all'): %s", optarg);
631             }
632             break;
633         case OPTION_OUTPUT:
634             output = optarg;
635             break;
636         case 'q':
637             quiet = true;
638             break;
639         }
640     }
641     if (optind != argc - 1) {
642         error_exit("Expecting one image file name");
643     }
644     filename = argv[optind++];
645 
646     if (output && !strcmp(output, "json")) {
647         output_format = OFORMAT_JSON;
648     } else if (output && !strcmp(output, "human")) {
649         output_format = OFORMAT_HUMAN;
650     } else if (output) {
651         error_report("--output must be used with human or json as argument.");
652         return 1;
653     }
654 
655     bs = bdrv_new_open("image", filename, fmt, flags, true, quiet);
656     if (!bs) {
657         return 1;
658     }
659 
660     check = g_new0(ImageCheck, 1);
661     ret = collect_image_check(bs, check, filename, fmt, fix);
662 
663     if (ret == -ENOTSUP) {
664         error_report("This image format does not support checks");
665         ret = 63;
666         goto fail;
667     }
668 
669     if (check->corruptions_fixed || check->leaks_fixed) {
670         int corruptions_fixed, leaks_fixed;
671 
672         leaks_fixed         = check->leaks_fixed;
673         corruptions_fixed   = check->corruptions_fixed;
674 
675         if (output_format == OFORMAT_HUMAN) {
676             qprintf(quiet,
677                     "The following inconsistencies were found and repaired:\n\n"
678                     "    %" PRId64 " leaked clusters\n"
679                     "    %" PRId64 " corruptions\n\n"
680                     "Double checking the fixed image now...\n",
681                     check->leaks_fixed,
682                     check->corruptions_fixed);
683         }
684 
685         ret = collect_image_check(bs, check, filename, fmt, 0);
686 
687         check->leaks_fixed          = leaks_fixed;
688         check->corruptions_fixed    = corruptions_fixed;
689     }
690 
691     switch (output_format) {
692     case OFORMAT_HUMAN:
693         dump_human_image_check(check, quiet);
694         break;
695     case OFORMAT_JSON:
696         dump_json_image_check(check, quiet);
697         break;
698     }
699 
700     if (ret || check->check_errors) {
701         ret = 1;
702         goto fail;
703     }
704 
705     if (check->corruptions) {
706         ret = 2;
707     } else if (check->leaks) {
708         ret = 3;
709     } else {
710         ret = 0;
711     }
712 
713 fail:
714     qapi_free_ImageCheck(check);
715     bdrv_unref(bs);
716 
717     return ret;
718 }
719 
720 static int img_commit(int argc, char **argv)
721 {
722     int c, ret, flags;
723     const char *filename, *fmt, *cache;
724     BlockDriverState *bs;
725     bool quiet = false;
726 
727     fmt = NULL;
728     cache = BDRV_DEFAULT_CACHE;
729     for(;;) {
730         c = getopt(argc, argv, "f:ht:q");
731         if (c == -1) {
732             break;
733         }
734         switch(c) {
735         case '?':
736         case 'h':
737             help();
738             break;
739         case 'f':
740             fmt = optarg;
741             break;
742         case 't':
743             cache = optarg;
744             break;
745         case 'q':
746             quiet = true;
747             break;
748         }
749     }
750     if (optind != argc - 1) {
751         error_exit("Expecting one image file name");
752     }
753     filename = argv[optind++];
754 
755     flags = BDRV_O_RDWR;
756     ret = bdrv_parse_cache_flags(cache, &flags);
757     if (ret < 0) {
758         error_report("Invalid cache option: %s", cache);
759         return -1;
760     }
761 
762     bs = bdrv_new_open("image", filename, fmt, flags, true, quiet);
763     if (!bs) {
764         return 1;
765     }
766     ret = bdrv_commit(bs);
767     switch(ret) {
768     case 0:
769         qprintf(quiet, "Image committed.\n");
770         break;
771     case -ENOENT:
772         error_report("No disk inserted");
773         break;
774     case -EACCES:
775         error_report("Image is read-only");
776         break;
777     case -ENOTSUP:
778         error_report("Image is already committed");
779         break;
780     default:
781         error_report("Error while committing image");
782         break;
783     }
784 
785     bdrv_unref(bs);
786     if (ret) {
787         return 1;
788     }
789     return 0;
790 }
791 
792 /*
793  * Returns true iff the first sector pointed to by 'buf' contains at least
794  * a non-NUL byte.
795  *
796  * 'pnum' is set to the number of sectors (including and immediately following
797  * the first one) that are known to be in the same allocated/unallocated state.
798  */
799 static int is_allocated_sectors(const uint8_t *buf, int n, int *pnum)
800 {
801     bool is_zero;
802     int i;
803 
804     if (n <= 0) {
805         *pnum = 0;
806         return 0;
807     }
808     is_zero = buffer_is_zero(buf, 512);
809     for(i = 1; i < n; i++) {
810         buf += 512;
811         if (is_zero != buffer_is_zero(buf, 512)) {
812             break;
813         }
814     }
815     *pnum = i;
816     return !is_zero;
817 }
818 
819 /*
820  * Like is_allocated_sectors, but if the buffer starts with a used sector,
821  * up to 'min' consecutive sectors containing zeros are ignored. This avoids
822  * breaking up write requests for only small sparse areas.
823  */
824 static int is_allocated_sectors_min(const uint8_t *buf, int n, int *pnum,
825     int min)
826 {
827     int ret;
828     int num_checked, num_used;
829 
830     if (n < min) {
831         min = n;
832     }
833 
834     ret = is_allocated_sectors(buf, n, pnum);
835     if (!ret) {
836         return ret;
837     }
838 
839     num_used = *pnum;
840     buf += BDRV_SECTOR_SIZE * *pnum;
841     n -= *pnum;
842     num_checked = num_used;
843 
844     while (n > 0) {
845         ret = is_allocated_sectors(buf, n, pnum);
846 
847         buf += BDRV_SECTOR_SIZE * *pnum;
848         n -= *pnum;
849         num_checked += *pnum;
850         if (ret) {
851             num_used = num_checked;
852         } else if (*pnum >= min) {
853             break;
854         }
855     }
856 
857     *pnum = num_used;
858     return 1;
859 }
860 
861 /*
862  * Compares two buffers sector by sector. Returns 0 if the first sector of both
863  * buffers matches, non-zero otherwise.
864  *
865  * pnum is set to the number of sectors (including and immediately following
866  * the first one) that are known to have the same comparison result
867  */
868 static int compare_sectors(const uint8_t *buf1, const uint8_t *buf2, int n,
869     int *pnum)
870 {
871     int res, i;
872 
873     if (n <= 0) {
874         *pnum = 0;
875         return 0;
876     }
877 
878     res = !!memcmp(buf1, buf2, 512);
879     for(i = 1; i < n; i++) {
880         buf1 += 512;
881         buf2 += 512;
882 
883         if (!!memcmp(buf1, buf2, 512) != res) {
884             break;
885         }
886     }
887 
888     *pnum = i;
889     return res;
890 }
891 
892 #define IO_BUF_SIZE (2 * 1024 * 1024)
893 
894 static int64_t sectors_to_bytes(int64_t sectors)
895 {
896     return sectors << BDRV_SECTOR_BITS;
897 }
898 
899 static int64_t sectors_to_process(int64_t total, int64_t from)
900 {
901     return MIN(total - from, IO_BUF_SIZE >> BDRV_SECTOR_BITS);
902 }
903 
904 /*
905  * Check if passed sectors are empty (not allocated or contain only 0 bytes)
906  *
907  * Returns 0 in case sectors are filled with 0, 1 if sectors contain non-zero
908  * data and negative value on error.
909  *
910  * @param bs:  Driver used for accessing file
911  * @param sect_num: Number of first sector to check
912  * @param sect_count: Number of sectors to check
913  * @param filename: Name of disk file we are checking (logging purpose)
914  * @param buffer: Allocated buffer for storing read data
915  * @param quiet: Flag for quiet mode
916  */
917 static int check_empty_sectors(BlockDriverState *bs, int64_t sect_num,
918                                int sect_count, const char *filename,
919                                uint8_t *buffer, bool quiet)
920 {
921     int pnum, ret = 0;
922     ret = bdrv_read(bs, sect_num, buffer, sect_count);
923     if (ret < 0) {
924         error_report("Error while reading offset %" PRId64 " of %s: %s",
925                      sectors_to_bytes(sect_num), filename, strerror(-ret));
926         return ret;
927     }
928     ret = is_allocated_sectors(buffer, sect_count, &pnum);
929     if (ret || pnum != sect_count) {
930         qprintf(quiet, "Content mismatch at offset %" PRId64 "!\n",
931                 sectors_to_bytes(ret ? sect_num : sect_num + pnum));
932         return 1;
933     }
934 
935     return 0;
936 }
937 
938 /*
939  * Compares two images. Exit codes:
940  *
941  * 0 - Images are identical
942  * 1 - Images differ
943  * >1 - Error occurred
944  */
945 static int img_compare(int argc, char **argv)
946 {
947     const char *fmt1 = NULL, *fmt2 = NULL, *filename1, *filename2;
948     BlockDriverState *bs1, *bs2;
949     int64_t total_sectors1, total_sectors2;
950     uint8_t *buf1 = NULL, *buf2 = NULL;
951     int pnum1, pnum2;
952     int allocated1, allocated2;
953     int ret = 0; /* return value - 0 Ident, 1 Different, >1 Error */
954     bool progress = false, quiet = false, strict = false;
955     int64_t total_sectors;
956     int64_t sector_num = 0;
957     int64_t nb_sectors;
958     int c, pnum;
959     uint64_t bs_sectors;
960     uint64_t progress_base;
961 
962     for (;;) {
963         c = getopt(argc, argv, "hpf:F:sq");
964         if (c == -1) {
965             break;
966         }
967         switch (c) {
968         case '?':
969         case 'h':
970             help();
971             break;
972         case 'f':
973             fmt1 = optarg;
974             break;
975         case 'F':
976             fmt2 = optarg;
977             break;
978         case 'p':
979             progress = true;
980             break;
981         case 'q':
982             quiet = true;
983             break;
984         case 's':
985             strict = true;
986             break;
987         }
988     }
989 
990     /* Progress is not shown in Quiet mode */
991     if (quiet) {
992         progress = false;
993     }
994 
995 
996     if (optind != argc - 2) {
997         error_exit("Expecting two image file names");
998     }
999     filename1 = argv[optind++];
1000     filename2 = argv[optind++];
1001 
1002     /* Initialize before goto out */
1003     qemu_progress_init(progress, 2.0);
1004 
1005     bs1 = bdrv_new_open("image 1", filename1, fmt1, BDRV_O_FLAGS, true, quiet);
1006     if (!bs1) {
1007         error_report("Can't open file %s", filename1);
1008         ret = 2;
1009         goto out3;
1010     }
1011 
1012     bs2 = bdrv_new_open("image 2", filename2, fmt2, BDRV_O_FLAGS, true, quiet);
1013     if (!bs2) {
1014         error_report("Can't open file %s", filename2);
1015         ret = 2;
1016         goto out2;
1017     }
1018 
1019     buf1 = qemu_blockalign(bs1, IO_BUF_SIZE);
1020     buf2 = qemu_blockalign(bs2, IO_BUF_SIZE);
1021     bdrv_get_geometry(bs1, &bs_sectors);
1022     total_sectors1 = bs_sectors;
1023     bdrv_get_geometry(bs2, &bs_sectors);
1024     total_sectors2 = bs_sectors;
1025     total_sectors = MIN(total_sectors1, total_sectors2);
1026     progress_base = MAX(total_sectors1, total_sectors2);
1027 
1028     qemu_progress_print(0, 100);
1029 
1030     if (strict && total_sectors1 != total_sectors2) {
1031         ret = 1;
1032         qprintf(quiet, "Strict mode: Image size mismatch!\n");
1033         goto out;
1034     }
1035 
1036     for (;;) {
1037         nb_sectors = sectors_to_process(total_sectors, sector_num);
1038         if (nb_sectors <= 0) {
1039             break;
1040         }
1041         allocated1 = bdrv_is_allocated_above(bs1, NULL, sector_num, nb_sectors,
1042                                              &pnum1);
1043         if (allocated1 < 0) {
1044             ret = 3;
1045             error_report("Sector allocation test failed for %s", filename1);
1046             goto out;
1047         }
1048 
1049         allocated2 = bdrv_is_allocated_above(bs2, NULL, sector_num, nb_sectors,
1050                                              &pnum2);
1051         if (allocated2 < 0) {
1052             ret = 3;
1053             error_report("Sector allocation test failed for %s", filename2);
1054             goto out;
1055         }
1056         nb_sectors = MIN(pnum1, pnum2);
1057 
1058         if (allocated1 == allocated2) {
1059             if (allocated1) {
1060                 ret = bdrv_read(bs1, sector_num, buf1, nb_sectors);
1061                 if (ret < 0) {
1062                     error_report("Error while reading offset %" PRId64 " of %s:"
1063                                  " %s", sectors_to_bytes(sector_num), filename1,
1064                                  strerror(-ret));
1065                     ret = 4;
1066                     goto out;
1067                 }
1068                 ret = bdrv_read(bs2, sector_num, buf2, nb_sectors);
1069                 if (ret < 0) {
1070                     error_report("Error while reading offset %" PRId64
1071                                  " of %s: %s", sectors_to_bytes(sector_num),
1072                                  filename2, strerror(-ret));
1073                     ret = 4;
1074                     goto out;
1075                 }
1076                 ret = compare_sectors(buf1, buf2, nb_sectors, &pnum);
1077                 if (ret || pnum != nb_sectors) {
1078                     qprintf(quiet, "Content mismatch at offset %" PRId64 "!\n",
1079                             sectors_to_bytes(
1080                                 ret ? sector_num : sector_num + pnum));
1081                     ret = 1;
1082                     goto out;
1083                 }
1084             }
1085         } else {
1086             if (strict) {
1087                 ret = 1;
1088                 qprintf(quiet, "Strict mode: Offset %" PRId64
1089                         " allocation mismatch!\n",
1090                         sectors_to_bytes(sector_num));
1091                 goto out;
1092             }
1093 
1094             if (allocated1) {
1095                 ret = check_empty_sectors(bs1, sector_num, nb_sectors,
1096                                           filename1, buf1, quiet);
1097             } else {
1098                 ret = check_empty_sectors(bs2, sector_num, nb_sectors,
1099                                           filename2, buf1, quiet);
1100             }
1101             if (ret) {
1102                 if (ret < 0) {
1103                     error_report("Error while reading offset %" PRId64 ": %s",
1104                                  sectors_to_bytes(sector_num), strerror(-ret));
1105                     ret = 4;
1106                 }
1107                 goto out;
1108             }
1109         }
1110         sector_num += nb_sectors;
1111         qemu_progress_print(((float) nb_sectors / progress_base)*100, 100);
1112     }
1113 
1114     if (total_sectors1 != total_sectors2) {
1115         BlockDriverState *bs_over;
1116         int64_t total_sectors_over;
1117         const char *filename_over;
1118 
1119         qprintf(quiet, "Warning: Image size mismatch!\n");
1120         if (total_sectors1 > total_sectors2) {
1121             total_sectors_over = total_sectors1;
1122             bs_over = bs1;
1123             filename_over = filename1;
1124         } else {
1125             total_sectors_over = total_sectors2;
1126             bs_over = bs2;
1127             filename_over = filename2;
1128         }
1129 
1130         for (;;) {
1131             nb_sectors = sectors_to_process(total_sectors_over, sector_num);
1132             if (nb_sectors <= 0) {
1133                 break;
1134             }
1135             ret = bdrv_is_allocated_above(bs_over, NULL, sector_num,
1136                                           nb_sectors, &pnum);
1137             if (ret < 0) {
1138                 ret = 3;
1139                 error_report("Sector allocation test failed for %s",
1140                              filename_over);
1141                 goto out;
1142 
1143             }
1144             nb_sectors = pnum;
1145             if (ret) {
1146                 ret = check_empty_sectors(bs_over, sector_num, nb_sectors,
1147                                           filename_over, buf1, quiet);
1148                 if (ret) {
1149                     if (ret < 0) {
1150                         error_report("Error while reading offset %" PRId64
1151                                      " of %s: %s", sectors_to_bytes(sector_num),
1152                                      filename_over, strerror(-ret));
1153                         ret = 4;
1154                     }
1155                     goto out;
1156                 }
1157             }
1158             sector_num += nb_sectors;
1159             qemu_progress_print(((float) nb_sectors / progress_base)*100, 100);
1160         }
1161     }
1162 
1163     qprintf(quiet, "Images are identical.\n");
1164     ret = 0;
1165 
1166 out:
1167     bdrv_unref(bs2);
1168     qemu_vfree(buf1);
1169     qemu_vfree(buf2);
1170 out2:
1171     bdrv_unref(bs1);
1172 out3:
1173     qemu_progress_end();
1174     return ret;
1175 }
1176 
1177 static int img_convert(int argc, char **argv)
1178 {
1179     int c, n, n1, bs_n, bs_i, compress, cluster_sectors, skip_create;
1180     int64_t ret = 0;
1181     int progress = 0, flags;
1182     const char *fmt, *out_fmt, *cache, *out_baseimg, *out_filename;
1183     BlockDriver *drv, *proto_drv;
1184     BlockDriverState **bs = NULL, *out_bs = NULL;
1185     int64_t total_sectors, nb_sectors, sector_num, bs_offset;
1186     uint64_t bs_sectors;
1187     uint8_t * buf = NULL;
1188     size_t bufsectors = IO_BUF_SIZE / BDRV_SECTOR_SIZE;
1189     const uint8_t *buf1;
1190     BlockDriverInfo bdi;
1191     QemuOpts *opts = NULL;
1192     QemuOptsList *create_opts = NULL;
1193     const char *out_baseimg_param;
1194     char *options = NULL;
1195     const char *snapshot_name = NULL;
1196     int min_sparse = 8; /* Need at least 4k of zeros for sparse detection */
1197     bool quiet = false;
1198     Error *local_err = NULL;
1199     QemuOpts *sn_opts = NULL;
1200 
1201     fmt = NULL;
1202     out_fmt = "raw";
1203     cache = "unsafe";
1204     out_baseimg = NULL;
1205     compress = 0;
1206     skip_create = 0;
1207     for(;;) {
1208         c = getopt(argc, argv, "f:O:B:s:hce6o:pS:t:qnl:");
1209         if (c == -1) {
1210             break;
1211         }
1212         switch(c) {
1213         case '?':
1214         case 'h':
1215             help();
1216             break;
1217         case 'f':
1218             fmt = optarg;
1219             break;
1220         case 'O':
1221             out_fmt = optarg;
1222             break;
1223         case 'B':
1224             out_baseimg = optarg;
1225             break;
1226         case 'c':
1227             compress = 1;
1228             break;
1229         case 'e':
1230             error_report("option -e is deprecated, please use \'-o "
1231                   "encryption\' instead!");
1232             ret = -1;
1233             goto fail_getopt;
1234         case '6':
1235             error_report("option -6 is deprecated, please use \'-o "
1236                   "compat6\' instead!");
1237             ret = -1;
1238             goto fail_getopt;
1239         case 'o':
1240             if (!is_valid_option_list(optarg)) {
1241                 error_report("Invalid option list: %s", optarg);
1242                 ret = -1;
1243                 goto fail_getopt;
1244             }
1245             if (!options) {
1246                 options = g_strdup(optarg);
1247             } else {
1248                 char *old_options = options;
1249                 options = g_strdup_printf("%s,%s", options, optarg);
1250                 g_free(old_options);
1251             }
1252             break;
1253         case 's':
1254             snapshot_name = optarg;
1255             break;
1256         case 'l':
1257             if (strstart(optarg, SNAPSHOT_OPT_BASE, NULL)) {
1258                 sn_opts = qemu_opts_parse(&internal_snapshot_opts, optarg, 0);
1259                 if (!sn_opts) {
1260                     error_report("Failed in parsing snapshot param '%s'",
1261                                  optarg);
1262                     ret = -1;
1263                     goto fail_getopt;
1264                 }
1265             } else {
1266                 snapshot_name = optarg;
1267             }
1268             break;
1269         case 'S':
1270         {
1271             int64_t sval;
1272             char *end;
1273             sval = strtosz_suffix(optarg, &end, STRTOSZ_DEFSUFFIX_B);
1274             if (sval < 0 || *end) {
1275                 error_report("Invalid minimum zero buffer size for sparse output specified");
1276                 ret = -1;
1277                 goto fail_getopt;
1278             }
1279 
1280             min_sparse = sval / BDRV_SECTOR_SIZE;
1281             break;
1282         }
1283         case 'p':
1284             progress = 1;
1285             break;
1286         case 't':
1287             cache = optarg;
1288             break;
1289         case 'q':
1290             quiet = true;
1291             break;
1292         case 'n':
1293             skip_create = 1;
1294             break;
1295         }
1296     }
1297 
1298     /* Initialize before goto out */
1299     if (quiet) {
1300         progress = 0;
1301     }
1302     qemu_progress_init(progress, 1.0);
1303 
1304 
1305     bs_n = argc - optind - 1;
1306     out_filename = bs_n >= 1 ? argv[argc - 1] : NULL;
1307 
1308     if (options && has_help_option(options)) {
1309         ret = print_block_option_help(out_filename, out_fmt);
1310         goto out;
1311     }
1312 
1313     if (bs_n < 1) {
1314         error_exit("Must specify image file name");
1315     }
1316 
1317 
1318     if (bs_n > 1 && out_baseimg) {
1319         error_report("-B makes no sense when concatenating multiple input "
1320                      "images");
1321         ret = -1;
1322         goto out;
1323     }
1324 
1325     qemu_progress_print(0, 100);
1326 
1327     bs = g_malloc0(bs_n * sizeof(BlockDriverState *));
1328 
1329     total_sectors = 0;
1330     for (bs_i = 0; bs_i < bs_n; bs_i++) {
1331         char *id = bs_n > 1 ? g_strdup_printf("source %d", bs_i)
1332                             : g_strdup("source");
1333         bs[bs_i] = bdrv_new_open(id, argv[optind + bs_i], fmt, BDRV_O_FLAGS,
1334                                  true, quiet);
1335         g_free(id);
1336         if (!bs[bs_i]) {
1337             error_report("Could not open '%s'", argv[optind + bs_i]);
1338             ret = -1;
1339             goto out;
1340         }
1341         bdrv_get_geometry(bs[bs_i], &bs_sectors);
1342         total_sectors += bs_sectors;
1343     }
1344 
1345     if (sn_opts) {
1346         ret = bdrv_snapshot_load_tmp(bs[0],
1347                                      qemu_opt_get(sn_opts, SNAPSHOT_OPT_ID),
1348                                      qemu_opt_get(sn_opts, SNAPSHOT_OPT_NAME),
1349                                      &local_err);
1350     } else if (snapshot_name != NULL) {
1351         if (bs_n > 1) {
1352             error_report("No support for concatenating multiple snapshot");
1353             ret = -1;
1354             goto out;
1355         }
1356 
1357         bdrv_snapshot_load_tmp_by_id_or_name(bs[0], snapshot_name, &local_err);
1358     }
1359     if (local_err) {
1360         error_report("Failed to load snapshot: %s",
1361                      error_get_pretty(local_err));
1362         error_free(local_err);
1363         ret = -1;
1364         goto out;
1365     }
1366 
1367     /* Find driver and parse its options */
1368     drv = bdrv_find_format(out_fmt);
1369     if (!drv) {
1370         error_report("Unknown file format '%s'", out_fmt);
1371         ret = -1;
1372         goto out;
1373     }
1374 
1375     proto_drv = bdrv_find_protocol(out_filename, true);
1376     if (!proto_drv) {
1377         error_report("Unknown protocol '%s'", out_filename);
1378         ret = -1;
1379         goto out;
1380     }
1381 
1382     create_opts = qemu_opts_append(create_opts, drv->create_opts);
1383     create_opts = qemu_opts_append(create_opts, proto_drv->create_opts);
1384 
1385     opts = qemu_opts_create(create_opts, NULL, 0, &error_abort);
1386     if (options && qemu_opts_do_parse(opts, options, NULL)) {
1387         error_report("Invalid options for file format '%s'", out_fmt);
1388         ret = -1;
1389         goto out;
1390     }
1391 
1392     qemu_opt_set_number(opts, BLOCK_OPT_SIZE, total_sectors * 512);
1393     ret = add_old_style_options(out_fmt, opts, out_baseimg, NULL);
1394     if (ret < 0) {
1395         goto out;
1396     }
1397 
1398     /* Get backing file name if -o backing_file was used */
1399     out_baseimg_param = qemu_opt_get(opts, BLOCK_OPT_BACKING_FILE);
1400     if (out_baseimg_param) {
1401         out_baseimg = out_baseimg_param;
1402     }
1403 
1404     /* Check if compression is supported */
1405     if (compress) {
1406         bool encryption =
1407             qemu_opt_get_bool(opts, BLOCK_OPT_ENCRYPT, false);
1408         const char *preallocation =
1409             qemu_opt_get(opts, BLOCK_OPT_PREALLOC);
1410 
1411         if (!drv->bdrv_write_compressed) {
1412             error_report("Compression not supported for this file format");
1413             ret = -1;
1414             goto out;
1415         }
1416 
1417         if (encryption) {
1418             error_report("Compression and encryption not supported at "
1419                          "the same time");
1420             ret = -1;
1421             goto out;
1422         }
1423 
1424         if (preallocation
1425             && strcmp(preallocation, "off"))
1426         {
1427             error_report("Compression and preallocation not supported at "
1428                          "the same time");
1429             ret = -1;
1430             goto out;
1431         }
1432     }
1433 
1434     if (!skip_create) {
1435         /* Create the new image */
1436         ret = bdrv_create(drv, out_filename, opts, &local_err);
1437         if (ret < 0) {
1438             error_report("%s: error while converting %s: %s",
1439                          out_filename, out_fmt, error_get_pretty(local_err));
1440             error_free(local_err);
1441             goto out;
1442         }
1443     }
1444 
1445     flags = min_sparse ? (BDRV_O_RDWR | BDRV_O_UNMAP) : BDRV_O_RDWR;
1446     ret = bdrv_parse_cache_flags(cache, &flags);
1447     if (ret < 0) {
1448         error_report("Invalid cache option: %s", cache);
1449         goto out;
1450     }
1451 
1452     out_bs = bdrv_new_open("target", out_filename, out_fmt, flags, true, quiet);
1453     if (!out_bs) {
1454         ret = -1;
1455         goto out;
1456     }
1457 
1458     bs_i = 0;
1459     bs_offset = 0;
1460     bdrv_get_geometry(bs[0], &bs_sectors);
1461 
1462     /* increase bufsectors from the default 4096 (2M) if opt_transfer_length
1463      * or discard_alignment of the out_bs is greater. Limit to 32768 (16MB)
1464      * as maximum. */
1465     bufsectors = MIN(32768,
1466                      MAX(bufsectors, MAX(out_bs->bl.opt_transfer_length,
1467                                          out_bs->bl.discard_alignment))
1468                     );
1469 
1470     buf = qemu_blockalign(out_bs, bufsectors * BDRV_SECTOR_SIZE);
1471 
1472     if (skip_create) {
1473         int64_t output_length = bdrv_getlength(out_bs);
1474         if (output_length < 0) {
1475             error_report("unable to get output image length: %s\n",
1476                          strerror(-output_length));
1477             ret = -1;
1478             goto out;
1479         } else if (output_length < total_sectors << BDRV_SECTOR_BITS) {
1480             error_report("output file is smaller than input file");
1481             ret = -1;
1482             goto out;
1483         }
1484     }
1485 
1486     cluster_sectors = 0;
1487     ret = bdrv_get_info(out_bs, &bdi);
1488     if (ret < 0) {
1489         if (compress) {
1490             error_report("could not get block driver info");
1491             goto out;
1492         }
1493     } else {
1494         compress = compress || bdi.needs_compressed_writes;
1495         cluster_sectors = bdi.cluster_size / BDRV_SECTOR_SIZE;
1496     }
1497 
1498     if (compress) {
1499         if (cluster_sectors <= 0 || cluster_sectors > bufsectors) {
1500             error_report("invalid cluster size");
1501             ret = -1;
1502             goto out;
1503         }
1504         sector_num = 0;
1505 
1506         nb_sectors = total_sectors;
1507 
1508         for(;;) {
1509             int64_t bs_num;
1510             int remainder;
1511             uint8_t *buf2;
1512 
1513             nb_sectors = total_sectors - sector_num;
1514             if (nb_sectors <= 0)
1515                 break;
1516             if (nb_sectors >= cluster_sectors)
1517                 n = cluster_sectors;
1518             else
1519                 n = nb_sectors;
1520 
1521             bs_num = sector_num - bs_offset;
1522             assert (bs_num >= 0);
1523             remainder = n;
1524             buf2 = buf;
1525             while (remainder > 0) {
1526                 int nlow;
1527                 while (bs_num == bs_sectors) {
1528                     bs_i++;
1529                     assert (bs_i < bs_n);
1530                     bs_offset += bs_sectors;
1531                     bdrv_get_geometry(bs[bs_i], &bs_sectors);
1532                     bs_num = 0;
1533                     /* printf("changing part: sector_num=%" PRId64 ", "
1534                        "bs_i=%d, bs_offset=%" PRId64 ", bs_sectors=%" PRId64
1535                        "\n", sector_num, bs_i, bs_offset, bs_sectors); */
1536                 }
1537                 assert (bs_num < bs_sectors);
1538 
1539                 nlow = (remainder > bs_sectors - bs_num) ? bs_sectors - bs_num : remainder;
1540 
1541                 ret = bdrv_read(bs[bs_i], bs_num, buf2, nlow);
1542                 if (ret < 0) {
1543                     error_report("error while reading sector %" PRId64 ": %s",
1544                                  bs_num, strerror(-ret));
1545                     goto out;
1546                 }
1547 
1548                 buf2 += nlow * 512;
1549                 bs_num += nlow;
1550 
1551                 remainder -= nlow;
1552             }
1553             assert (remainder == 0);
1554 
1555             if (!buffer_is_zero(buf, n * BDRV_SECTOR_SIZE)) {
1556                 ret = bdrv_write_compressed(out_bs, sector_num, buf, n);
1557                 if (ret != 0) {
1558                     error_report("error while compressing sector %" PRId64
1559                                  ": %s", sector_num, strerror(-ret));
1560                     goto out;
1561                 }
1562             }
1563             sector_num += n;
1564             qemu_progress_print(100.0 * sector_num / total_sectors, 0);
1565         }
1566         /* signal EOF to align */
1567         bdrv_write_compressed(out_bs, 0, NULL, 0);
1568     } else {
1569         int64_t sectors_to_read, sectors_read, sector_num_next_status;
1570         bool count_allocated_sectors;
1571         int has_zero_init = min_sparse ? bdrv_has_zero_init(out_bs) : 0;
1572 
1573         if (!has_zero_init && bdrv_can_write_zeroes_with_unmap(out_bs)) {
1574             ret = bdrv_make_zero(out_bs, BDRV_REQ_MAY_UNMAP);
1575             if (ret < 0) {
1576                 goto out;
1577             }
1578             has_zero_init = 1;
1579         }
1580 
1581         sectors_to_read = total_sectors;
1582         count_allocated_sectors = progress && (out_baseimg || has_zero_init);
1583 restart:
1584         sector_num = 0; // total number of sectors converted so far
1585         sectors_read = 0;
1586         sector_num_next_status = 0;
1587 
1588         for(;;) {
1589             nb_sectors = total_sectors - sector_num;
1590             if (nb_sectors <= 0) {
1591                 if (count_allocated_sectors) {
1592                     sectors_to_read = sectors_read;
1593                     count_allocated_sectors = false;
1594                     goto restart;
1595                 }
1596                 ret = 0;
1597                 break;
1598             }
1599 
1600             while (sector_num - bs_offset >= bs_sectors) {
1601                 bs_i ++;
1602                 assert (bs_i < bs_n);
1603                 bs_offset += bs_sectors;
1604                 bdrv_get_geometry(bs[bs_i], &bs_sectors);
1605                 /* printf("changing part: sector_num=%" PRId64 ", bs_i=%d, "
1606                   "bs_offset=%" PRId64 ", bs_sectors=%" PRId64 "\n",
1607                    sector_num, bs_i, bs_offset, bs_sectors); */
1608             }
1609 
1610             if ((out_baseimg || has_zero_init) &&
1611                 sector_num >= sector_num_next_status) {
1612                 n = nb_sectors > INT_MAX ? INT_MAX : nb_sectors;
1613                 ret = bdrv_get_block_status(bs[bs_i], sector_num - bs_offset,
1614                                             n, &n1);
1615                 if (ret < 0) {
1616                     error_report("error while reading block status of sector %"
1617                                  PRId64 ": %s", sector_num - bs_offset,
1618                                  strerror(-ret));
1619                     goto out;
1620                 }
1621                 /* If the output image is zero initialized, we are not working
1622                  * on a shared base and the input is zero we can skip the next
1623                  * n1 sectors */
1624                 if (has_zero_init && !out_baseimg && (ret & BDRV_BLOCK_ZERO)) {
1625                     sector_num += n1;
1626                     continue;
1627                 }
1628                 /* If the output image is being created as a copy on write
1629                  * image, assume that sectors which are unallocated in the
1630                  * input image are present in both the output's and input's
1631                  * base images (no need to copy them). */
1632                 if (out_baseimg) {
1633                     if (!(ret & BDRV_BLOCK_DATA)) {
1634                         sector_num += n1;
1635                         continue;
1636                     }
1637                     /* The next 'n1' sectors are allocated in the input image.
1638                      * Copy only those as they may be followed by unallocated
1639                      * sectors. */
1640                     nb_sectors = n1;
1641                 }
1642                 /* avoid redundant callouts to get_block_status */
1643                 sector_num_next_status = sector_num + n1;
1644             }
1645 
1646             n = MIN(nb_sectors, bufsectors);
1647 
1648             /* round down request length to an aligned sector, but
1649              * do not bother doing this on short requests. They happen
1650              * when we found an all-zero area, and the next sector to
1651              * write will not be sector_num + n. */
1652             if (cluster_sectors > 0 && n >= cluster_sectors) {
1653                 int64_t next_aligned_sector = (sector_num + n);
1654                 next_aligned_sector -= next_aligned_sector % cluster_sectors;
1655                 if (sector_num + n > next_aligned_sector) {
1656                     n = next_aligned_sector - sector_num;
1657                 }
1658             }
1659 
1660             n = MIN(n, bs_sectors - (sector_num - bs_offset));
1661 
1662             sectors_read += n;
1663             if (count_allocated_sectors) {
1664                 sector_num += n;
1665                 continue;
1666             }
1667 
1668             n1 = n;
1669             ret = bdrv_read(bs[bs_i], sector_num - bs_offset, buf, n);
1670             if (ret < 0) {
1671                 error_report("error while reading sector %" PRId64 ": %s",
1672                              sector_num - bs_offset, strerror(-ret));
1673                 goto out;
1674             }
1675             /* NOTE: at the same time we convert, we do not write zero
1676                sectors to have a chance to compress the image. Ideally, we
1677                should add a specific call to have the info to go faster */
1678             buf1 = buf;
1679             while (n > 0) {
1680                 if (!has_zero_init ||
1681                     is_allocated_sectors_min(buf1, n, &n1, min_sparse)) {
1682                     ret = bdrv_write(out_bs, sector_num, buf1, n1);
1683                     if (ret < 0) {
1684                         error_report("error while writing sector %" PRId64
1685                                      ": %s", sector_num, strerror(-ret));
1686                         goto out;
1687                     }
1688                 }
1689                 sector_num += n1;
1690                 n -= n1;
1691                 buf1 += n1 * 512;
1692             }
1693             qemu_progress_print(100.0 * sectors_read / sectors_to_read, 0);
1694         }
1695     }
1696 out:
1697     if (!ret) {
1698         qemu_progress_print(100, 0);
1699     }
1700     qemu_progress_end();
1701     qemu_opts_del(opts);
1702     qemu_opts_free(create_opts);
1703     qemu_vfree(buf);
1704     if (sn_opts) {
1705         qemu_opts_del(sn_opts);
1706     }
1707     if (out_bs) {
1708         bdrv_unref(out_bs);
1709     }
1710     if (bs) {
1711         for (bs_i = 0; bs_i < bs_n; bs_i++) {
1712             if (bs[bs_i]) {
1713                 bdrv_unref(bs[bs_i]);
1714             }
1715         }
1716         g_free(bs);
1717     }
1718 fail_getopt:
1719     g_free(options);
1720 
1721     if (ret) {
1722         return 1;
1723     }
1724     return 0;
1725 }
1726 
1727 
1728 static void dump_snapshots(BlockDriverState *bs)
1729 {
1730     QEMUSnapshotInfo *sn_tab, *sn;
1731     int nb_sns, i;
1732 
1733     nb_sns = bdrv_snapshot_list(bs, &sn_tab);
1734     if (nb_sns <= 0)
1735         return;
1736     printf("Snapshot list:\n");
1737     bdrv_snapshot_dump(fprintf, stdout, NULL);
1738     printf("\n");
1739     for(i = 0; i < nb_sns; i++) {
1740         sn = &sn_tab[i];
1741         bdrv_snapshot_dump(fprintf, stdout, sn);
1742         printf("\n");
1743     }
1744     g_free(sn_tab);
1745 }
1746 
1747 static void dump_json_image_info_list(ImageInfoList *list)
1748 {
1749     Error *local_err = NULL;
1750     QString *str;
1751     QmpOutputVisitor *ov = qmp_output_visitor_new();
1752     QObject *obj;
1753     visit_type_ImageInfoList(qmp_output_get_visitor(ov),
1754                              &list, NULL, &local_err);
1755     obj = qmp_output_get_qobject(ov);
1756     str = qobject_to_json_pretty(obj);
1757     assert(str != NULL);
1758     printf("%s\n", qstring_get_str(str));
1759     qobject_decref(obj);
1760     qmp_output_visitor_cleanup(ov);
1761     QDECREF(str);
1762 }
1763 
1764 static void dump_json_image_info(ImageInfo *info)
1765 {
1766     Error *local_err = NULL;
1767     QString *str;
1768     QmpOutputVisitor *ov = qmp_output_visitor_new();
1769     QObject *obj;
1770     visit_type_ImageInfo(qmp_output_get_visitor(ov),
1771                          &info, NULL, &local_err);
1772     obj = qmp_output_get_qobject(ov);
1773     str = qobject_to_json_pretty(obj);
1774     assert(str != NULL);
1775     printf("%s\n", qstring_get_str(str));
1776     qobject_decref(obj);
1777     qmp_output_visitor_cleanup(ov);
1778     QDECREF(str);
1779 }
1780 
1781 static void dump_human_image_info_list(ImageInfoList *list)
1782 {
1783     ImageInfoList *elem;
1784     bool delim = false;
1785 
1786     for (elem = list; elem; elem = elem->next) {
1787         if (delim) {
1788             printf("\n");
1789         }
1790         delim = true;
1791 
1792         bdrv_image_info_dump(fprintf, stdout, elem->value);
1793     }
1794 }
1795 
1796 static gboolean str_equal_func(gconstpointer a, gconstpointer b)
1797 {
1798     return strcmp(a, b) == 0;
1799 }
1800 
1801 /**
1802  * Open an image file chain and return an ImageInfoList
1803  *
1804  * @filename: topmost image filename
1805  * @fmt: topmost image format (may be NULL to autodetect)
1806  * @chain: true  - enumerate entire backing file chain
1807  *         false - only topmost image file
1808  *
1809  * Returns a list of ImageInfo objects or NULL if there was an error opening an
1810  * image file.  If there was an error a message will have been printed to
1811  * stderr.
1812  */
1813 static ImageInfoList *collect_image_info_list(const char *filename,
1814                                               const char *fmt,
1815                                               bool chain)
1816 {
1817     ImageInfoList *head = NULL;
1818     ImageInfoList **last = &head;
1819     GHashTable *filenames;
1820     Error *err = NULL;
1821 
1822     filenames = g_hash_table_new_full(g_str_hash, str_equal_func, NULL, NULL);
1823 
1824     while (filename) {
1825         BlockDriverState *bs;
1826         ImageInfo *info;
1827         ImageInfoList *elem;
1828 
1829         if (g_hash_table_lookup_extended(filenames, filename, NULL, NULL)) {
1830             error_report("Backing file '%s' creates an infinite loop.",
1831                          filename);
1832             goto err;
1833         }
1834         g_hash_table_insert(filenames, (gpointer)filename, NULL);
1835 
1836         bs = bdrv_new_open("image", filename, fmt,
1837                            BDRV_O_FLAGS | BDRV_O_NO_BACKING, false, false);
1838         if (!bs) {
1839             goto err;
1840         }
1841 
1842         bdrv_query_image_info(bs, &info, &err);
1843         if (err) {
1844             error_report("%s", error_get_pretty(err));
1845             error_free(err);
1846             bdrv_unref(bs);
1847             goto err;
1848         }
1849 
1850         elem = g_new0(ImageInfoList, 1);
1851         elem->value = info;
1852         *last = elem;
1853         last = &elem->next;
1854 
1855         bdrv_unref(bs);
1856 
1857         filename = fmt = NULL;
1858         if (chain) {
1859             if (info->has_full_backing_filename) {
1860                 filename = info->full_backing_filename;
1861             } else if (info->has_backing_filename) {
1862                 filename = info->backing_filename;
1863             }
1864             if (info->has_backing_filename_format) {
1865                 fmt = info->backing_filename_format;
1866             }
1867         }
1868     }
1869     g_hash_table_destroy(filenames);
1870     return head;
1871 
1872 err:
1873     qapi_free_ImageInfoList(head);
1874     g_hash_table_destroy(filenames);
1875     return NULL;
1876 }
1877 
1878 static int img_info(int argc, char **argv)
1879 {
1880     int c;
1881     OutputFormat output_format = OFORMAT_HUMAN;
1882     bool chain = false;
1883     const char *filename, *fmt, *output;
1884     ImageInfoList *list;
1885 
1886     fmt = NULL;
1887     output = NULL;
1888     for(;;) {
1889         int option_index = 0;
1890         static const struct option long_options[] = {
1891             {"help", no_argument, 0, 'h'},
1892             {"format", required_argument, 0, 'f'},
1893             {"output", required_argument, 0, OPTION_OUTPUT},
1894             {"backing-chain", no_argument, 0, OPTION_BACKING_CHAIN},
1895             {0, 0, 0, 0}
1896         };
1897         c = getopt_long(argc, argv, "f:h",
1898                         long_options, &option_index);
1899         if (c == -1) {
1900             break;
1901         }
1902         switch(c) {
1903         case '?':
1904         case 'h':
1905             help();
1906             break;
1907         case 'f':
1908             fmt = optarg;
1909             break;
1910         case OPTION_OUTPUT:
1911             output = optarg;
1912             break;
1913         case OPTION_BACKING_CHAIN:
1914             chain = true;
1915             break;
1916         }
1917     }
1918     if (optind != argc - 1) {
1919         error_exit("Expecting one image file name");
1920     }
1921     filename = argv[optind++];
1922 
1923     if (output && !strcmp(output, "json")) {
1924         output_format = OFORMAT_JSON;
1925     } else if (output && !strcmp(output, "human")) {
1926         output_format = OFORMAT_HUMAN;
1927     } else if (output) {
1928         error_report("--output must be used with human or json as argument.");
1929         return 1;
1930     }
1931 
1932     list = collect_image_info_list(filename, fmt, chain);
1933     if (!list) {
1934         return 1;
1935     }
1936 
1937     switch (output_format) {
1938     case OFORMAT_HUMAN:
1939         dump_human_image_info_list(list);
1940         break;
1941     case OFORMAT_JSON:
1942         if (chain) {
1943             dump_json_image_info_list(list);
1944         } else {
1945             dump_json_image_info(list->value);
1946         }
1947         break;
1948     }
1949 
1950     qapi_free_ImageInfoList(list);
1951     return 0;
1952 }
1953 
1954 
1955 typedef struct MapEntry {
1956     int flags;
1957     int depth;
1958     int64_t start;
1959     int64_t length;
1960     int64_t offset;
1961     BlockDriverState *bs;
1962 } MapEntry;
1963 
1964 static void dump_map_entry(OutputFormat output_format, MapEntry *e,
1965                            MapEntry *next)
1966 {
1967     switch (output_format) {
1968     case OFORMAT_HUMAN:
1969         if ((e->flags & BDRV_BLOCK_DATA) &&
1970             !(e->flags & BDRV_BLOCK_OFFSET_VALID)) {
1971             error_report("File contains external, encrypted or compressed clusters.");
1972             exit(1);
1973         }
1974         if ((e->flags & (BDRV_BLOCK_DATA|BDRV_BLOCK_ZERO)) == BDRV_BLOCK_DATA) {
1975             printf("%#-16"PRIx64"%#-16"PRIx64"%#-16"PRIx64"%s\n",
1976                    e->start, e->length, e->offset, e->bs->filename);
1977         }
1978         /* This format ignores the distinction between 0, ZERO and ZERO|DATA.
1979          * Modify the flags here to allow more coalescing.
1980          */
1981         if (next &&
1982             (next->flags & (BDRV_BLOCK_DATA|BDRV_BLOCK_ZERO)) != BDRV_BLOCK_DATA) {
1983             next->flags &= ~BDRV_BLOCK_DATA;
1984             next->flags |= BDRV_BLOCK_ZERO;
1985         }
1986         break;
1987     case OFORMAT_JSON:
1988         printf("%s{ \"start\": %"PRId64", \"length\": %"PRId64", \"depth\": %d,"
1989                " \"zero\": %s, \"data\": %s",
1990                (e->start == 0 ? "[" : ",\n"),
1991                e->start, e->length, e->depth,
1992                (e->flags & BDRV_BLOCK_ZERO) ? "true" : "false",
1993                (e->flags & BDRV_BLOCK_DATA) ? "true" : "false");
1994         if (e->flags & BDRV_BLOCK_OFFSET_VALID) {
1995             printf(", \"offset\": %"PRId64"", e->offset);
1996         }
1997         putchar('}');
1998 
1999         if (!next) {
2000             printf("]\n");
2001         }
2002         break;
2003     }
2004 }
2005 
2006 static int get_block_status(BlockDriverState *bs, int64_t sector_num,
2007                             int nb_sectors, MapEntry *e)
2008 {
2009     int64_t ret;
2010     int depth;
2011 
2012     /* As an optimization, we could cache the current range of unallocated
2013      * clusters in each file of the chain, and avoid querying the same
2014      * range repeatedly.
2015      */
2016 
2017     depth = 0;
2018     for (;;) {
2019         ret = bdrv_get_block_status(bs, sector_num, nb_sectors, &nb_sectors);
2020         if (ret < 0) {
2021             return ret;
2022         }
2023         assert(nb_sectors);
2024         if (ret & (BDRV_BLOCK_ZERO|BDRV_BLOCK_DATA)) {
2025             break;
2026         }
2027         bs = bs->backing_hd;
2028         if (bs == NULL) {
2029             ret = 0;
2030             break;
2031         }
2032 
2033         depth++;
2034     }
2035 
2036     e->start = sector_num * BDRV_SECTOR_SIZE;
2037     e->length = nb_sectors * BDRV_SECTOR_SIZE;
2038     e->flags = ret & ~BDRV_BLOCK_OFFSET_MASK;
2039     e->offset = ret & BDRV_BLOCK_OFFSET_MASK;
2040     e->depth = depth;
2041     e->bs = bs;
2042     return 0;
2043 }
2044 
2045 static int img_map(int argc, char **argv)
2046 {
2047     int c;
2048     OutputFormat output_format = OFORMAT_HUMAN;
2049     BlockDriverState *bs;
2050     const char *filename, *fmt, *output;
2051     int64_t length;
2052     MapEntry curr = { .length = 0 }, next;
2053     int ret = 0;
2054 
2055     fmt = NULL;
2056     output = NULL;
2057     for (;;) {
2058         int option_index = 0;
2059         static const struct option long_options[] = {
2060             {"help", no_argument, 0, 'h'},
2061             {"format", required_argument, 0, 'f'},
2062             {"output", required_argument, 0, OPTION_OUTPUT},
2063             {0, 0, 0, 0}
2064         };
2065         c = getopt_long(argc, argv, "f:h",
2066                         long_options, &option_index);
2067         if (c == -1) {
2068             break;
2069         }
2070         switch (c) {
2071         case '?':
2072         case 'h':
2073             help();
2074             break;
2075         case 'f':
2076             fmt = optarg;
2077             break;
2078         case OPTION_OUTPUT:
2079             output = optarg;
2080             break;
2081         }
2082     }
2083     if (optind != argc - 1) {
2084         error_exit("Expecting one image file name");
2085     }
2086     filename = argv[optind];
2087 
2088     if (output && !strcmp(output, "json")) {
2089         output_format = OFORMAT_JSON;
2090     } else if (output && !strcmp(output, "human")) {
2091         output_format = OFORMAT_HUMAN;
2092     } else if (output) {
2093         error_report("--output must be used with human or json as argument.");
2094         return 1;
2095     }
2096 
2097     bs = bdrv_new_open("image", filename, fmt, BDRV_O_FLAGS, true, false);
2098     if (!bs) {
2099         return 1;
2100     }
2101 
2102     if (output_format == OFORMAT_HUMAN) {
2103         printf("%-16s%-16s%-16s%s\n", "Offset", "Length", "Mapped to", "File");
2104     }
2105 
2106     length = bdrv_getlength(bs);
2107     while (curr.start + curr.length < length) {
2108         int64_t nsectors_left;
2109         int64_t sector_num;
2110         int n;
2111 
2112         sector_num = (curr.start + curr.length) >> BDRV_SECTOR_BITS;
2113 
2114         /* Probe up to 1 GiB at a time.  */
2115         nsectors_left = DIV_ROUND_UP(length, BDRV_SECTOR_SIZE) - sector_num;
2116         n = MIN(1 << (30 - BDRV_SECTOR_BITS), nsectors_left);
2117         ret = get_block_status(bs, sector_num, n, &next);
2118 
2119         if (ret < 0) {
2120             error_report("Could not read file metadata: %s", strerror(-ret));
2121             goto out;
2122         }
2123 
2124         if (curr.length != 0 && curr.flags == next.flags &&
2125             curr.depth == next.depth &&
2126             ((curr.flags & BDRV_BLOCK_OFFSET_VALID) == 0 ||
2127              curr.offset + curr.length == next.offset)) {
2128             curr.length += next.length;
2129             continue;
2130         }
2131 
2132         if (curr.length > 0) {
2133             dump_map_entry(output_format, &curr, &next);
2134         }
2135         curr = next;
2136     }
2137 
2138     dump_map_entry(output_format, &curr, NULL);
2139 
2140 out:
2141     bdrv_unref(bs);
2142     return ret < 0;
2143 }
2144 
2145 #define SNAPSHOT_LIST   1
2146 #define SNAPSHOT_CREATE 2
2147 #define SNAPSHOT_APPLY  3
2148 #define SNAPSHOT_DELETE 4
2149 
2150 static int img_snapshot(int argc, char **argv)
2151 {
2152     BlockDriverState *bs;
2153     QEMUSnapshotInfo sn;
2154     char *filename, *snapshot_name = NULL;
2155     int c, ret = 0, bdrv_oflags;
2156     int action = 0;
2157     qemu_timeval tv;
2158     bool quiet = false;
2159     Error *err = NULL;
2160 
2161     bdrv_oflags = BDRV_O_FLAGS | BDRV_O_RDWR;
2162     /* Parse commandline parameters */
2163     for(;;) {
2164         c = getopt(argc, argv, "la:c:d:hq");
2165         if (c == -1) {
2166             break;
2167         }
2168         switch(c) {
2169         case '?':
2170         case 'h':
2171             help();
2172             return 0;
2173         case 'l':
2174             if (action) {
2175                 error_exit("Cannot mix '-l', '-a', '-c', '-d'");
2176                 return 0;
2177             }
2178             action = SNAPSHOT_LIST;
2179             bdrv_oflags &= ~BDRV_O_RDWR; /* no need for RW */
2180             break;
2181         case 'a':
2182             if (action) {
2183                 error_exit("Cannot mix '-l', '-a', '-c', '-d'");
2184                 return 0;
2185             }
2186             action = SNAPSHOT_APPLY;
2187             snapshot_name = optarg;
2188             break;
2189         case 'c':
2190             if (action) {
2191                 error_exit("Cannot mix '-l', '-a', '-c', '-d'");
2192                 return 0;
2193             }
2194             action = SNAPSHOT_CREATE;
2195             snapshot_name = optarg;
2196             break;
2197         case 'd':
2198             if (action) {
2199                 error_exit("Cannot mix '-l', '-a', '-c', '-d'");
2200                 return 0;
2201             }
2202             action = SNAPSHOT_DELETE;
2203             snapshot_name = optarg;
2204             break;
2205         case 'q':
2206             quiet = true;
2207             break;
2208         }
2209     }
2210 
2211     if (optind != argc - 1) {
2212         error_exit("Expecting one image file name");
2213     }
2214     filename = argv[optind++];
2215 
2216     /* Open the image */
2217     bs = bdrv_new_open("image", filename, NULL, bdrv_oflags, true, quiet);
2218     if (!bs) {
2219         return 1;
2220     }
2221 
2222     /* Perform the requested action */
2223     switch(action) {
2224     case SNAPSHOT_LIST:
2225         dump_snapshots(bs);
2226         break;
2227 
2228     case SNAPSHOT_CREATE:
2229         memset(&sn, 0, sizeof(sn));
2230         pstrcpy(sn.name, sizeof(sn.name), snapshot_name);
2231 
2232         qemu_gettimeofday(&tv);
2233         sn.date_sec = tv.tv_sec;
2234         sn.date_nsec = tv.tv_usec * 1000;
2235 
2236         ret = bdrv_snapshot_create(bs, &sn);
2237         if (ret) {
2238             error_report("Could not create snapshot '%s': %d (%s)",
2239                 snapshot_name, ret, strerror(-ret));
2240         }
2241         break;
2242 
2243     case SNAPSHOT_APPLY:
2244         ret = bdrv_snapshot_goto(bs, snapshot_name);
2245         if (ret) {
2246             error_report("Could not apply snapshot '%s': %d (%s)",
2247                 snapshot_name, ret, strerror(-ret));
2248         }
2249         break;
2250 
2251     case SNAPSHOT_DELETE:
2252         bdrv_snapshot_delete_by_id_or_name(bs, snapshot_name, &err);
2253         if (err) {
2254             error_report("Could not delete snapshot '%s': (%s)",
2255                          snapshot_name, error_get_pretty(err));
2256             error_free(err);
2257             ret = 1;
2258         }
2259         break;
2260     }
2261 
2262     /* Cleanup */
2263     bdrv_unref(bs);
2264     if (ret) {
2265         return 1;
2266     }
2267     return 0;
2268 }
2269 
2270 static int img_rebase(int argc, char **argv)
2271 {
2272     BlockDriverState *bs, *bs_old_backing = NULL, *bs_new_backing = NULL;
2273     BlockDriver *old_backing_drv, *new_backing_drv;
2274     char *filename;
2275     const char *fmt, *cache, *out_basefmt, *out_baseimg;
2276     int c, flags, ret;
2277     int unsafe = 0;
2278     int progress = 0;
2279     bool quiet = false;
2280     Error *local_err = NULL;
2281 
2282     /* Parse commandline parameters */
2283     fmt = NULL;
2284     cache = BDRV_DEFAULT_CACHE;
2285     out_baseimg = NULL;
2286     out_basefmt = NULL;
2287     for(;;) {
2288         c = getopt(argc, argv, "uhf:F:b:pt:q");
2289         if (c == -1) {
2290             break;
2291         }
2292         switch(c) {
2293         case '?':
2294         case 'h':
2295             help();
2296             return 0;
2297         case 'f':
2298             fmt = optarg;
2299             break;
2300         case 'F':
2301             out_basefmt = optarg;
2302             break;
2303         case 'b':
2304             out_baseimg = optarg;
2305             break;
2306         case 'u':
2307             unsafe = 1;
2308             break;
2309         case 'p':
2310             progress = 1;
2311             break;
2312         case 't':
2313             cache = optarg;
2314             break;
2315         case 'q':
2316             quiet = true;
2317             break;
2318         }
2319     }
2320 
2321     if (quiet) {
2322         progress = 0;
2323     }
2324 
2325     if (optind != argc - 1) {
2326         error_exit("Expecting one image file name");
2327     }
2328     if (!unsafe && !out_baseimg) {
2329         error_exit("Must specify backing file (-b) or use unsafe mode (-u)");
2330     }
2331     filename = argv[optind++];
2332 
2333     qemu_progress_init(progress, 2.0);
2334     qemu_progress_print(0, 100);
2335 
2336     flags = BDRV_O_RDWR | (unsafe ? BDRV_O_NO_BACKING : 0);
2337     ret = bdrv_parse_cache_flags(cache, &flags);
2338     if (ret < 0) {
2339         error_report("Invalid cache option: %s", cache);
2340         return -1;
2341     }
2342 
2343     /*
2344      * Open the images.
2345      *
2346      * Ignore the old backing file for unsafe rebase in case we want to correct
2347      * the reference to a renamed or moved backing file.
2348      */
2349     bs = bdrv_new_open("image", filename, fmt, flags, true, quiet);
2350     if (!bs) {
2351         return 1;
2352     }
2353 
2354     /* Find the right drivers for the backing files */
2355     old_backing_drv = NULL;
2356     new_backing_drv = NULL;
2357 
2358     if (!unsafe && bs->backing_format[0] != '\0') {
2359         old_backing_drv = bdrv_find_format(bs->backing_format);
2360         if (old_backing_drv == NULL) {
2361             error_report("Invalid format name: '%s'", bs->backing_format);
2362             ret = -1;
2363             goto out;
2364         }
2365     }
2366 
2367     if (out_basefmt != NULL) {
2368         new_backing_drv = bdrv_find_format(out_basefmt);
2369         if (new_backing_drv == NULL) {
2370             error_report("Invalid format name: '%s'", out_basefmt);
2371             ret = -1;
2372             goto out;
2373         }
2374     }
2375 
2376     /* For safe rebasing we need to compare old and new backing file */
2377     if (unsafe) {
2378         /* Make the compiler happy */
2379         bs_old_backing = NULL;
2380         bs_new_backing = NULL;
2381     } else {
2382         char backing_name[1024];
2383 
2384         bs_old_backing = bdrv_new("old_backing", &error_abort);
2385         bdrv_get_backing_filename(bs, backing_name, sizeof(backing_name));
2386         ret = bdrv_open(&bs_old_backing, backing_name, NULL, NULL, BDRV_O_FLAGS,
2387                         old_backing_drv, &local_err);
2388         if (ret) {
2389             error_report("Could not open old backing file '%s': %s",
2390                          backing_name, error_get_pretty(local_err));
2391             error_free(local_err);
2392             goto out;
2393         }
2394         if (out_baseimg[0]) {
2395             bs_new_backing = bdrv_new("new_backing", &error_abort);
2396             ret = bdrv_open(&bs_new_backing, out_baseimg, NULL, NULL,
2397                             BDRV_O_FLAGS, new_backing_drv, &local_err);
2398             if (ret) {
2399                 error_report("Could not open new backing file '%s': %s",
2400                              out_baseimg, error_get_pretty(local_err));
2401                 error_free(local_err);
2402                 goto out;
2403             }
2404         }
2405     }
2406 
2407     /*
2408      * Check each unallocated cluster in the COW file. If it is unallocated,
2409      * accesses go to the backing file. We must therefore compare this cluster
2410      * in the old and new backing file, and if they differ we need to copy it
2411      * from the old backing file into the COW file.
2412      *
2413      * If qemu-img crashes during this step, no harm is done. The content of
2414      * the image is the same as the original one at any time.
2415      */
2416     if (!unsafe) {
2417         uint64_t num_sectors;
2418         uint64_t old_backing_num_sectors;
2419         uint64_t new_backing_num_sectors = 0;
2420         uint64_t sector;
2421         int n;
2422         uint8_t * buf_old;
2423         uint8_t * buf_new;
2424         float local_progress = 0;
2425 
2426         buf_old = qemu_blockalign(bs, IO_BUF_SIZE);
2427         buf_new = qemu_blockalign(bs, IO_BUF_SIZE);
2428 
2429         bdrv_get_geometry(bs, &num_sectors);
2430         bdrv_get_geometry(bs_old_backing, &old_backing_num_sectors);
2431         if (bs_new_backing) {
2432             bdrv_get_geometry(bs_new_backing, &new_backing_num_sectors);
2433         }
2434 
2435         if (num_sectors != 0) {
2436             local_progress = (float)100 /
2437                 (num_sectors / MIN(num_sectors, IO_BUF_SIZE / 512));
2438         }
2439 
2440         for (sector = 0; sector < num_sectors; sector += n) {
2441 
2442             /* How many sectors can we handle with the next read? */
2443             if (sector + (IO_BUF_SIZE / 512) <= num_sectors) {
2444                 n = (IO_BUF_SIZE / 512);
2445             } else {
2446                 n = num_sectors - sector;
2447             }
2448 
2449             /* If the cluster is allocated, we don't need to take action */
2450             ret = bdrv_is_allocated(bs, sector, n, &n);
2451             if (ret < 0) {
2452                 error_report("error while reading image metadata: %s",
2453                              strerror(-ret));
2454                 goto out;
2455             }
2456             if (ret) {
2457                 continue;
2458             }
2459 
2460             /*
2461              * Read old and new backing file and take into consideration that
2462              * backing files may be smaller than the COW image.
2463              */
2464             if (sector >= old_backing_num_sectors) {
2465                 memset(buf_old, 0, n * BDRV_SECTOR_SIZE);
2466             } else {
2467                 if (sector + n > old_backing_num_sectors) {
2468                     n = old_backing_num_sectors - sector;
2469                 }
2470 
2471                 ret = bdrv_read(bs_old_backing, sector, buf_old, n);
2472                 if (ret < 0) {
2473                     error_report("error while reading from old backing file");
2474                     goto out;
2475                 }
2476             }
2477 
2478             if (sector >= new_backing_num_sectors || !bs_new_backing) {
2479                 memset(buf_new, 0, n * BDRV_SECTOR_SIZE);
2480             } else {
2481                 if (sector + n > new_backing_num_sectors) {
2482                     n = new_backing_num_sectors - sector;
2483                 }
2484 
2485                 ret = bdrv_read(bs_new_backing, sector, buf_new, n);
2486                 if (ret < 0) {
2487                     error_report("error while reading from new backing file");
2488                     goto out;
2489                 }
2490             }
2491 
2492             /* If they differ, we need to write to the COW file */
2493             uint64_t written = 0;
2494 
2495             while (written < n) {
2496                 int pnum;
2497 
2498                 if (compare_sectors(buf_old + written * 512,
2499                     buf_new + written * 512, n - written, &pnum))
2500                 {
2501                     ret = bdrv_write(bs, sector + written,
2502                         buf_old + written * 512, pnum);
2503                     if (ret < 0) {
2504                         error_report("Error while writing to COW image: %s",
2505                             strerror(-ret));
2506                         goto out;
2507                     }
2508                 }
2509 
2510                 written += pnum;
2511             }
2512             qemu_progress_print(local_progress, 100);
2513         }
2514 
2515         qemu_vfree(buf_old);
2516         qemu_vfree(buf_new);
2517     }
2518 
2519     /*
2520      * Change the backing file. All clusters that are different from the old
2521      * backing file are overwritten in the COW file now, so the visible content
2522      * doesn't change when we switch the backing file.
2523      */
2524     if (out_baseimg && *out_baseimg) {
2525         ret = bdrv_change_backing_file(bs, out_baseimg, out_basefmt);
2526     } else {
2527         ret = bdrv_change_backing_file(bs, NULL, NULL);
2528     }
2529 
2530     if (ret == -ENOSPC) {
2531         error_report("Could not change the backing file to '%s': No "
2532                      "space left in the file header", out_baseimg);
2533     } else if (ret < 0) {
2534         error_report("Could not change the backing file to '%s': %s",
2535             out_baseimg, strerror(-ret));
2536     }
2537 
2538     qemu_progress_print(100, 0);
2539     /*
2540      * TODO At this point it is possible to check if any clusters that are
2541      * allocated in the COW file are the same in the backing file. If so, they
2542      * could be dropped from the COW file. Don't do this before switching the
2543      * backing file, in case of a crash this would lead to corruption.
2544      */
2545 out:
2546     qemu_progress_end();
2547     /* Cleanup */
2548     if (!unsafe) {
2549         if (bs_old_backing != NULL) {
2550             bdrv_unref(bs_old_backing);
2551         }
2552         if (bs_new_backing != NULL) {
2553             bdrv_unref(bs_new_backing);
2554         }
2555     }
2556 
2557     bdrv_unref(bs);
2558     if (ret) {
2559         return 1;
2560     }
2561     return 0;
2562 }
2563 
2564 static int img_resize(int argc, char **argv)
2565 {
2566     int c, ret, relative;
2567     const char *filename, *fmt, *size;
2568     int64_t n, total_size;
2569     bool quiet = false;
2570     BlockDriverState *bs = NULL;
2571     QemuOpts *param;
2572     static QemuOptsList resize_options = {
2573         .name = "resize_options",
2574         .head = QTAILQ_HEAD_INITIALIZER(resize_options.head),
2575         .desc = {
2576             {
2577                 .name = BLOCK_OPT_SIZE,
2578                 .type = QEMU_OPT_SIZE,
2579                 .help = "Virtual disk size"
2580             }, {
2581                 /* end of list */
2582             }
2583         },
2584     };
2585 
2586     /* Remove size from argv manually so that negative numbers are not treated
2587      * as options by getopt. */
2588     if (argc < 3) {
2589         error_exit("Not enough arguments");
2590         return 1;
2591     }
2592 
2593     size = argv[--argc];
2594 
2595     /* Parse getopt arguments */
2596     fmt = NULL;
2597     for(;;) {
2598         c = getopt(argc, argv, "f:hq");
2599         if (c == -1) {
2600             break;
2601         }
2602         switch(c) {
2603         case '?':
2604         case 'h':
2605             help();
2606             break;
2607         case 'f':
2608             fmt = optarg;
2609             break;
2610         case 'q':
2611             quiet = true;
2612             break;
2613         }
2614     }
2615     if (optind != argc - 1) {
2616         error_exit("Expecting one image file name");
2617     }
2618     filename = argv[optind++];
2619 
2620     /* Choose grow, shrink, or absolute resize mode */
2621     switch (size[0]) {
2622     case '+':
2623         relative = 1;
2624         size++;
2625         break;
2626     case '-':
2627         relative = -1;
2628         size++;
2629         break;
2630     default:
2631         relative = 0;
2632         break;
2633     }
2634 
2635     /* Parse size */
2636     param = qemu_opts_create(&resize_options, NULL, 0, &error_abort);
2637     if (qemu_opt_set(param, BLOCK_OPT_SIZE, size)) {
2638         /* Error message already printed when size parsing fails */
2639         ret = -1;
2640         qemu_opts_del(param);
2641         goto out;
2642     }
2643     n = qemu_opt_get_size(param, BLOCK_OPT_SIZE, 0);
2644     qemu_opts_del(param);
2645 
2646     bs = bdrv_new_open("image", filename, fmt, BDRV_O_FLAGS | BDRV_O_RDWR,
2647                        true, quiet);
2648     if (!bs) {
2649         ret = -1;
2650         goto out;
2651     }
2652 
2653     if (relative) {
2654         total_size = bdrv_getlength(bs) + n * relative;
2655     } else {
2656         total_size = n;
2657     }
2658     if (total_size <= 0) {
2659         error_report("New image size must be positive");
2660         ret = -1;
2661         goto out;
2662     }
2663 
2664     ret = bdrv_truncate(bs, total_size);
2665     switch (ret) {
2666     case 0:
2667         qprintf(quiet, "Image resized.\n");
2668         break;
2669     case -ENOTSUP:
2670         error_report("This image does not support resize");
2671         break;
2672     case -EACCES:
2673         error_report("Image is read-only");
2674         break;
2675     default:
2676         error_report("Error resizing image (%d)", -ret);
2677         break;
2678     }
2679 out:
2680     if (bs) {
2681         bdrv_unref(bs);
2682     }
2683     if (ret) {
2684         return 1;
2685     }
2686     return 0;
2687 }
2688 
2689 static int img_amend(int argc, char **argv)
2690 {
2691     int c, ret = 0;
2692     char *options = NULL;
2693     QemuOptsList *create_opts = NULL;
2694     QemuOpts *opts = NULL;
2695     const char *fmt = NULL, *filename;
2696     bool quiet = false;
2697     BlockDriverState *bs = NULL;
2698 
2699     for (;;) {
2700         c = getopt(argc, argv, "hqf:o:");
2701         if (c == -1) {
2702             break;
2703         }
2704 
2705         switch (c) {
2706             case 'h':
2707             case '?':
2708                 help();
2709                 break;
2710             case 'o':
2711                 if (!is_valid_option_list(optarg)) {
2712                     error_report("Invalid option list: %s", optarg);
2713                     ret = -1;
2714                     goto out;
2715                 }
2716                 if (!options) {
2717                     options = g_strdup(optarg);
2718                 } else {
2719                     char *old_options = options;
2720                     options = g_strdup_printf("%s,%s", options, optarg);
2721                     g_free(old_options);
2722                 }
2723                 break;
2724             case 'f':
2725                 fmt = optarg;
2726                 break;
2727             case 'q':
2728                 quiet = true;
2729                 break;
2730         }
2731     }
2732 
2733     if (!options) {
2734         error_exit("Must specify options (-o)");
2735     }
2736 
2737     filename = (optind == argc - 1) ? argv[argc - 1] : NULL;
2738     if (fmt && has_help_option(options)) {
2739         /* If a format is explicitly specified (and possibly no filename is
2740          * given), print option help here */
2741         ret = print_block_option_help(filename, fmt);
2742         goto out;
2743     }
2744 
2745     if (optind != argc - 1) {
2746         error_exit("Expecting one image file name");
2747     }
2748 
2749     bs = bdrv_new_open("image", filename, fmt,
2750                        BDRV_O_FLAGS | BDRV_O_RDWR, true, quiet);
2751     if (!bs) {
2752         error_report("Could not open image '%s'", filename);
2753         ret = -1;
2754         goto out;
2755     }
2756 
2757     fmt = bs->drv->format_name;
2758 
2759     if (has_help_option(options)) {
2760         /* If the format was auto-detected, print option help here */
2761         ret = print_block_option_help(filename, fmt);
2762         goto out;
2763     }
2764 
2765     create_opts = qemu_opts_append(create_opts, bs->drv->create_opts);
2766     opts = qemu_opts_create(create_opts, NULL, 0, &error_abort);
2767     if (options && qemu_opts_do_parse(opts, options, NULL)) {
2768         error_report("Invalid options for file format '%s'", fmt);
2769         ret = -1;
2770         goto out;
2771     }
2772 
2773     ret = bdrv_amend_options(bs, opts);
2774     if (ret < 0) {
2775         error_report("Error while amending options: %s", strerror(-ret));
2776         goto out;
2777     }
2778 
2779 out:
2780     if (bs) {
2781         bdrv_unref(bs);
2782     }
2783     qemu_opts_del(opts);
2784     qemu_opts_free(create_opts);
2785     g_free(options);
2786 
2787     if (ret) {
2788         return 1;
2789     }
2790     return 0;
2791 }
2792 
2793 static const img_cmd_t img_cmds[] = {
2794 #define DEF(option, callback, arg_string)        \
2795     { option, callback },
2796 #include "qemu-img-cmds.h"
2797 #undef DEF
2798 #undef GEN_DOCS
2799     { NULL, NULL, },
2800 };
2801 
2802 int main(int argc, char **argv)
2803 {
2804     const img_cmd_t *cmd;
2805     const char *cmdname;
2806     int c;
2807     static const struct option long_options[] = {
2808         {"help", no_argument, 0, 'h'},
2809         {"version", no_argument, 0, 'v'},
2810         {0, 0, 0, 0}
2811     };
2812 
2813 #ifdef CONFIG_POSIX
2814     signal(SIGPIPE, SIG_IGN);
2815 #endif
2816 
2817     error_set_progname(argv[0]);
2818     qemu_init_exec_dir(argv[0]);
2819 
2820     qemu_init_main_loop();
2821     bdrv_init();
2822     if (argc < 2) {
2823         error_exit("Not enough arguments");
2824     }
2825     cmdname = argv[1];
2826 
2827     /* find the command */
2828     for (cmd = img_cmds; cmd->name != NULL; cmd++) {
2829         if (!strcmp(cmdname, cmd->name)) {
2830             return cmd->handler(argc - 1, argv + 1);
2831         }
2832     }
2833 
2834     c = getopt_long(argc, argv, "h", long_options, NULL);
2835 
2836     if (c == 'h') {
2837         help();
2838     }
2839     if (c == 'v') {
2840         printf(QEMU_IMG_VERSION);
2841         return 0;
2842     }
2843 
2844     /* not found */
2845     error_exit("Command not found: %s", cmdname);
2846 }
2847