]> git.proxmox.com Git - mirror_qemu.git/blob - qemu-img.c
qemu-img: use blk_co_pwrite_zeroes for zero sectors when compressed
[mirror_qemu.git] / qemu-img.c
1 /*
2 * QEMU disk image utility
3 *
4 * Copyright (c) 2003-2008 Fabrice Bellard
5 *
6 * Permission is hereby granted, free of charge, to any person obtaining a copy
7 * of this software and associated documentation files (the "Software"), to deal
8 * in the Software without restriction, including without limitation the rights
9 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
10 * copies of the Software, and to permit persons to whom the Software is
11 * furnished to do so, subject to the following conditions:
12 *
13 * The above copyright notice and this permission notice shall be included in
14 * all copies or substantial portions of the Software.
15 *
16 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
17 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
18 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
19 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
20 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
21 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
22 * THE SOFTWARE.
23 */
24 #include "qemu/osdep.h"
25 #include "qemu-version.h"
26 #include "qapi/error.h"
27 #include "qapi-visit.h"
28 #include "qapi/qobject-output-visitor.h"
29 #include "qapi/qmp/qerror.h"
30 #include "qapi/qmp/qjson.h"
31 #include "qemu/cutils.h"
32 #include "qemu/config-file.h"
33 #include "qemu/option.h"
34 #include "qemu/error-report.h"
35 #include "qemu/log.h"
36 #include "qom/object_interfaces.h"
37 #include "sysemu/sysemu.h"
38 #include "sysemu/block-backend.h"
39 #include "block/block_int.h"
40 #include "block/blockjob.h"
41 #include "block/qapi.h"
42 #include "crypto/init.h"
43 #include "trace/control.h"
44 #include <getopt.h>
45
46 #define QEMU_IMG_VERSION "qemu-img version " QEMU_VERSION QEMU_PKGVERSION \
47 "\n" QEMU_COPYRIGHT "\n"
48
49 typedef struct img_cmd_t {
50 const char *name;
51 int (*handler)(int argc, char **argv);
52 } img_cmd_t;
53
54 enum {
55 OPTION_OUTPUT = 256,
56 OPTION_BACKING_CHAIN = 257,
57 OPTION_OBJECT = 258,
58 OPTION_IMAGE_OPTS = 259,
59 OPTION_PATTERN = 260,
60 OPTION_FLUSH_INTERVAL = 261,
61 OPTION_NO_DRAIN = 262,
62 };
63
64 typedef enum OutputFormat {
65 OFORMAT_JSON,
66 OFORMAT_HUMAN,
67 } OutputFormat;
68
69 /* Default to cache=writeback as data integrity is not important for qemu-img */
70 #define BDRV_DEFAULT_CACHE "writeback"
71
72 static void format_print(void *opaque, const char *name)
73 {
74 printf(" %s", name);
75 }
76
77 static void QEMU_NORETURN GCC_FMT_ATTR(1, 2) error_exit(const char *fmt, ...)
78 {
79 va_list ap;
80
81 error_printf("qemu-img: ");
82
83 va_start(ap, fmt);
84 error_vprintf(fmt, ap);
85 va_end(ap);
86
87 error_printf("\nTry 'qemu-img --help' for more information\n");
88 exit(EXIT_FAILURE);
89 }
90
91 static void QEMU_NORETURN missing_argument(const char *option)
92 {
93 error_exit("missing argument for option '%s'", option);
94 }
95
96 static void QEMU_NORETURN unrecognized_option(const char *option)
97 {
98 error_exit("unrecognized option '%s'", option);
99 }
100
101 /* Please keep in synch with qemu-img.texi */
102 static void QEMU_NORETURN help(void)
103 {
104 const char *help_msg =
105 QEMU_IMG_VERSION
106 "usage: qemu-img [standard options] command [command options]\n"
107 "QEMU disk image utility\n"
108 "\n"
109 " '-h', '--help' display this help and exit\n"
110 " '-V', '--version' output version information and exit\n"
111 " '-T', '--trace' [[enable=]<pattern>][,events=<file>][,file=<file>]\n"
112 " specify tracing options\n"
113 "\n"
114 "Command syntax:\n"
115 #define DEF(option, callback, arg_string) \
116 " " arg_string "\n"
117 #include "qemu-img-cmds.h"
118 #undef DEF
119 #undef GEN_DOCS
120 "\n"
121 "Command parameters:\n"
122 " 'filename' is a disk image filename\n"
123 " 'objectdef' is a QEMU user creatable object definition. See the qemu(1)\n"
124 " manual page for a description of the object properties. The most common\n"
125 " object type is a 'secret', which is used to supply passwords and/or\n"
126 " encryption keys.\n"
127 " 'fmt' is the disk image format. It is guessed automatically in most cases\n"
128 " 'cache' is the cache mode used to write the output disk image, the valid\n"
129 " options are: 'none', 'writeback' (default, except for convert), 'writethrough',\n"
130 " 'directsync' and 'unsafe' (default for convert)\n"
131 " 'src_cache' is the cache mode used to read input disk images, the valid\n"
132 " options are the same as for the 'cache' option\n"
133 " 'size' is the disk image size in bytes. Optional suffixes\n"
134 " 'k' or 'K' (kilobyte, 1024), 'M' (megabyte, 1024k), 'G' (gigabyte, 1024M),\n"
135 " 'T' (terabyte, 1024G), 'P' (petabyte, 1024T) and 'E' (exabyte, 1024P) are\n"
136 " supported. 'b' is ignored.\n"
137 " 'output_filename' is the destination disk image filename\n"
138 " 'output_fmt' is the destination format\n"
139 " 'options' is a comma separated list of format specific options in a\n"
140 " name=value format. Use -o ? for an overview of the options supported by the\n"
141 " used format\n"
142 " 'snapshot_param' is param used for internal snapshot, format\n"
143 " is 'snapshot.id=[ID],snapshot.name=[NAME]', or\n"
144 " '[ID_OR_NAME]'\n"
145 " 'snapshot_id_or_name' is deprecated, use 'snapshot_param'\n"
146 " instead\n"
147 " '-c' indicates that target image must be compressed (qcow format only)\n"
148 " '-u' enables unsafe rebasing. It is assumed that old and new backing file\n"
149 " match exactly. The image doesn't need a working backing file before\n"
150 " rebasing in this case (useful for renaming the backing file)\n"
151 " '-h' with or without a command shows this help and lists the supported formats\n"
152 " '-p' show progress of command (only certain commands)\n"
153 " '-q' use Quiet mode - do not print any output (except errors)\n"
154 " '-S' indicates the consecutive number of bytes (defaults to 4k) that must\n"
155 " contain only zeros for qemu-img to create a sparse image during\n"
156 " conversion. If the number of bytes is 0, the source will not be scanned for\n"
157 " unallocated or zero sectors, and the destination image will always be\n"
158 " fully allocated\n"
159 " '--output' takes the format in which the output must be done (human or json)\n"
160 " '-n' skips the target volume creation (useful if the volume is created\n"
161 " prior to running qemu-img)\n"
162 "\n"
163 "Parameters to check subcommand:\n"
164 " '-r' tries to repair any inconsistencies that are found during the check.\n"
165 " '-r leaks' repairs only cluster leaks, whereas '-r all' fixes all\n"
166 " kinds of errors, with a higher risk of choosing the wrong fix or\n"
167 " hiding corruption that has already occurred.\n"
168 "\n"
169 "Parameters to convert subcommand:\n"
170 " '-m' specifies how many coroutines work in parallel during the convert\n"
171 " process (defaults to 8)\n"
172 " '-W' allow to write to the target out of order rather than sequential\n"
173 "\n"
174 "Parameters to snapshot subcommand:\n"
175 " 'snapshot' is the name of the snapshot to create, apply or delete\n"
176 " '-a' applies a snapshot (revert disk to saved state)\n"
177 " '-c' creates a snapshot\n"
178 " '-d' deletes a snapshot\n"
179 " '-l' lists all snapshots in the given image\n"
180 "\n"
181 "Parameters to compare subcommand:\n"
182 " '-f' first image format\n"
183 " '-F' second image format\n"
184 " '-s' run in Strict mode - fail on different image size or sector allocation\n"
185 "\n"
186 "Parameters to dd subcommand:\n"
187 " 'bs=BYTES' read and write up to BYTES bytes at a time "
188 "(default: 512)\n"
189 " 'count=N' copy only N input blocks\n"
190 " 'if=FILE' read from FILE\n"
191 " 'of=FILE' write to FILE\n"
192 " 'skip=N' skip N bs-sized blocks at the start of input\n";
193
194 printf("%s\nSupported formats:", help_msg);
195 bdrv_iterate_format(format_print, NULL);
196 printf("\n");
197 exit(EXIT_SUCCESS);
198 }
199
200 static QemuOptsList qemu_object_opts = {
201 .name = "object",
202 .implied_opt_name = "qom-type",
203 .head = QTAILQ_HEAD_INITIALIZER(qemu_object_opts.head),
204 .desc = {
205 { }
206 },
207 };
208
209 static QemuOptsList qemu_source_opts = {
210 .name = "source",
211 .implied_opt_name = "file",
212 .head = QTAILQ_HEAD_INITIALIZER(qemu_source_opts.head),
213 .desc = {
214 { }
215 },
216 };
217
218 static int GCC_FMT_ATTR(2, 3) qprintf(bool quiet, const char *fmt, ...)
219 {
220 int ret = 0;
221 if (!quiet) {
222 va_list args;
223 va_start(args, fmt);
224 ret = vprintf(fmt, args);
225 va_end(args);
226 }
227 return ret;
228 }
229
230
231 static int print_block_option_help(const char *filename, const char *fmt)
232 {
233 BlockDriver *drv, *proto_drv;
234 QemuOptsList *create_opts = NULL;
235 Error *local_err = NULL;
236
237 /* Find driver and parse its options */
238 drv = bdrv_find_format(fmt);
239 if (!drv) {
240 error_report("Unknown file format '%s'", fmt);
241 return 1;
242 }
243
244 create_opts = qemu_opts_append(create_opts, drv->create_opts);
245 if (filename) {
246 proto_drv = bdrv_find_protocol(filename, true, &local_err);
247 if (!proto_drv) {
248 error_report_err(local_err);
249 qemu_opts_free(create_opts);
250 return 1;
251 }
252 create_opts = qemu_opts_append(create_opts, proto_drv->create_opts);
253 }
254
255 qemu_opts_print_help(create_opts);
256 qemu_opts_free(create_opts);
257 return 0;
258 }
259
260
261 static int img_open_password(BlockBackend *blk, const char *filename,
262 int flags, bool quiet)
263 {
264 BlockDriverState *bs;
265 char password[256];
266
267 bs = blk_bs(blk);
268 if (bdrv_is_encrypted(bs) && bdrv_key_required(bs) &&
269 !(flags & BDRV_O_NO_IO)) {
270 qprintf(quiet, "Disk image '%s' is encrypted.\n", filename);
271 if (qemu_read_password(password, sizeof(password)) < 0) {
272 error_report("No password given");
273 return -1;
274 }
275 if (bdrv_set_key(bs, password) < 0) {
276 error_report("invalid password");
277 return -1;
278 }
279 }
280 return 0;
281 }
282
283
284 static BlockBackend *img_open_opts(const char *optstr,
285 QemuOpts *opts, int flags, bool writethrough,
286 bool quiet)
287 {
288 QDict *options;
289 Error *local_err = NULL;
290 BlockBackend *blk;
291 options = qemu_opts_to_qdict(opts, NULL);
292 blk = blk_new_open(NULL, NULL, options, flags, &local_err);
293 if (!blk) {
294 error_reportf_err(local_err, "Could not open '%s': ", optstr);
295 return NULL;
296 }
297 blk_set_enable_write_cache(blk, !writethrough);
298
299 if (img_open_password(blk, optstr, flags, quiet) < 0) {
300 blk_unref(blk);
301 return NULL;
302 }
303 return blk;
304 }
305
306 static BlockBackend *img_open_file(const char *filename,
307 const char *fmt, int flags,
308 bool writethrough, bool quiet)
309 {
310 BlockBackend *blk;
311 Error *local_err = NULL;
312 QDict *options = NULL;
313
314 if (fmt) {
315 options = qdict_new();
316 qdict_put(options, "driver", qstring_from_str(fmt));
317 }
318
319 blk = blk_new_open(filename, NULL, options, flags, &local_err);
320 if (!blk) {
321 error_reportf_err(local_err, "Could not open '%s': ", filename);
322 return NULL;
323 }
324 blk_set_enable_write_cache(blk, !writethrough);
325
326 if (img_open_password(blk, filename, flags, quiet) < 0) {
327 blk_unref(blk);
328 return NULL;
329 }
330 return blk;
331 }
332
333
334 static BlockBackend *img_open(bool image_opts,
335 const char *filename,
336 const char *fmt, int flags, bool writethrough,
337 bool quiet)
338 {
339 BlockBackend *blk;
340 if (image_opts) {
341 QemuOpts *opts;
342 if (fmt) {
343 error_report("--image-opts and --format are mutually exclusive");
344 return NULL;
345 }
346 opts = qemu_opts_parse_noisily(qemu_find_opts("source"),
347 filename, true);
348 if (!opts) {
349 return NULL;
350 }
351 blk = img_open_opts(filename, opts, flags, writethrough, quiet);
352 } else {
353 blk = img_open_file(filename, fmt, flags, writethrough, quiet);
354 }
355 return blk;
356 }
357
358
359 static int add_old_style_options(const char *fmt, QemuOpts *opts,
360 const char *base_filename,
361 const char *base_fmt)
362 {
363 Error *err = NULL;
364
365 if (base_filename) {
366 qemu_opt_set(opts, BLOCK_OPT_BACKING_FILE, base_filename, &err);
367 if (err) {
368 error_report("Backing file not supported for file format '%s'",
369 fmt);
370 error_free(err);
371 return -1;
372 }
373 }
374 if (base_fmt) {
375 qemu_opt_set(opts, BLOCK_OPT_BACKING_FMT, base_fmt, &err);
376 if (err) {
377 error_report("Backing file format not supported for file "
378 "format '%s'", fmt);
379 error_free(err);
380 return -1;
381 }
382 }
383 return 0;
384 }
385
386 static int64_t cvtnum(const char *s)
387 {
388 int err;
389 uint64_t value;
390
391 err = qemu_strtosz(s, NULL, &value);
392 if (err < 0) {
393 return err;
394 }
395 if (value > INT64_MAX) {
396 return -ERANGE;
397 }
398 return value;
399 }
400
401 static int img_create(int argc, char **argv)
402 {
403 int c;
404 uint64_t img_size = -1;
405 const char *fmt = "raw";
406 const char *base_fmt = NULL;
407 const char *filename;
408 const char *base_filename = NULL;
409 char *options = NULL;
410 Error *local_err = NULL;
411 bool quiet = false;
412
413 for(;;) {
414 static const struct option long_options[] = {
415 {"help", no_argument, 0, 'h'},
416 {"object", required_argument, 0, OPTION_OBJECT},
417 {0, 0, 0, 0}
418 };
419 c = getopt_long(argc, argv, ":F:b:f:he6o:q",
420 long_options, NULL);
421 if (c == -1) {
422 break;
423 }
424 switch(c) {
425 case ':':
426 missing_argument(argv[optind - 1]);
427 break;
428 case '?':
429 unrecognized_option(argv[optind - 1]);
430 break;
431 case 'h':
432 help();
433 break;
434 case 'F':
435 base_fmt = optarg;
436 break;
437 case 'b':
438 base_filename = optarg;
439 break;
440 case 'f':
441 fmt = optarg;
442 break;
443 case 'e':
444 error_report("option -e is deprecated, please use \'-o "
445 "encryption\' instead!");
446 goto fail;
447 case '6':
448 error_report("option -6 is deprecated, please use \'-o "
449 "compat6\' instead!");
450 goto fail;
451 case 'o':
452 if (!is_valid_option_list(optarg)) {
453 error_report("Invalid option list: %s", optarg);
454 goto fail;
455 }
456 if (!options) {
457 options = g_strdup(optarg);
458 } else {
459 char *old_options = options;
460 options = g_strdup_printf("%s,%s", options, optarg);
461 g_free(old_options);
462 }
463 break;
464 case 'q':
465 quiet = true;
466 break;
467 case OPTION_OBJECT: {
468 QemuOpts *opts;
469 opts = qemu_opts_parse_noisily(&qemu_object_opts,
470 optarg, true);
471 if (!opts) {
472 goto fail;
473 }
474 } break;
475 }
476 }
477
478 /* Get the filename */
479 filename = (optind < argc) ? argv[optind] : NULL;
480 if (options && has_help_option(options)) {
481 g_free(options);
482 return print_block_option_help(filename, fmt);
483 }
484
485 if (optind >= argc) {
486 error_exit("Expecting image file name");
487 }
488 optind++;
489
490 if (qemu_opts_foreach(&qemu_object_opts,
491 user_creatable_add_opts_foreach,
492 NULL, NULL)) {
493 goto fail;
494 }
495
496 /* Get image size, if specified */
497 if (optind < argc) {
498 int64_t sval;
499
500 sval = cvtnum(argv[optind++]);
501 if (sval < 0) {
502 if (sval == -ERANGE) {
503 error_report("Image size must be less than 8 EiB!");
504 } else {
505 error_report("Invalid image size specified! You may use k, M, "
506 "G, T, P or E suffixes for ");
507 error_report("kilobytes, megabytes, gigabytes, terabytes, "
508 "petabytes and exabytes.");
509 }
510 goto fail;
511 }
512 img_size = (uint64_t)sval;
513 }
514 if (optind != argc) {
515 error_exit("Unexpected argument: %s", argv[optind]);
516 }
517
518 bdrv_img_create(filename, fmt, base_filename, base_fmt,
519 options, img_size, 0, quiet, &local_err);
520 if (local_err) {
521 error_reportf_err(local_err, "%s: ", filename);
522 goto fail;
523 }
524
525 g_free(options);
526 return 0;
527
528 fail:
529 g_free(options);
530 return 1;
531 }
532
533 static void dump_json_image_check(ImageCheck *check, bool quiet)
534 {
535 QString *str;
536 QObject *obj;
537 Visitor *v = qobject_output_visitor_new(&obj);
538
539 visit_type_ImageCheck(v, NULL, &check, &error_abort);
540 visit_complete(v, &obj);
541 str = qobject_to_json_pretty(obj);
542 assert(str != NULL);
543 qprintf(quiet, "%s\n", qstring_get_str(str));
544 qobject_decref(obj);
545 visit_free(v);
546 QDECREF(str);
547 }
548
549 static void dump_human_image_check(ImageCheck *check, bool quiet)
550 {
551 if (!(check->corruptions || check->leaks || check->check_errors)) {
552 qprintf(quiet, "No errors were found on the image.\n");
553 } else {
554 if (check->corruptions) {
555 qprintf(quiet, "\n%" PRId64 " errors were found on the image.\n"
556 "Data may be corrupted, or further writes to the image "
557 "may corrupt it.\n",
558 check->corruptions);
559 }
560
561 if (check->leaks) {
562 qprintf(quiet,
563 "\n%" PRId64 " leaked clusters were found on the image.\n"
564 "This means waste of disk space, but no harm to data.\n",
565 check->leaks);
566 }
567
568 if (check->check_errors) {
569 qprintf(quiet,
570 "\n%" PRId64
571 " internal errors have occurred during the check.\n",
572 check->check_errors);
573 }
574 }
575
576 if (check->total_clusters != 0 && check->allocated_clusters != 0) {
577 qprintf(quiet, "%" PRId64 "/%" PRId64 " = %0.2f%% allocated, "
578 "%0.2f%% fragmented, %0.2f%% compressed clusters\n",
579 check->allocated_clusters, check->total_clusters,
580 check->allocated_clusters * 100.0 / check->total_clusters,
581 check->fragmented_clusters * 100.0 / check->allocated_clusters,
582 check->compressed_clusters * 100.0 /
583 check->allocated_clusters);
584 }
585
586 if (check->image_end_offset) {
587 qprintf(quiet,
588 "Image end offset: %" PRId64 "\n", check->image_end_offset);
589 }
590 }
591
592 static int collect_image_check(BlockDriverState *bs,
593 ImageCheck *check,
594 const char *filename,
595 const char *fmt,
596 int fix)
597 {
598 int ret;
599 BdrvCheckResult result;
600
601 ret = bdrv_check(bs, &result, fix);
602 if (ret < 0) {
603 return ret;
604 }
605
606 check->filename = g_strdup(filename);
607 check->format = g_strdup(bdrv_get_format_name(bs));
608 check->check_errors = result.check_errors;
609 check->corruptions = result.corruptions;
610 check->has_corruptions = result.corruptions != 0;
611 check->leaks = result.leaks;
612 check->has_leaks = result.leaks != 0;
613 check->corruptions_fixed = result.corruptions_fixed;
614 check->has_corruptions_fixed = result.corruptions != 0;
615 check->leaks_fixed = result.leaks_fixed;
616 check->has_leaks_fixed = result.leaks != 0;
617 check->image_end_offset = result.image_end_offset;
618 check->has_image_end_offset = result.image_end_offset != 0;
619 check->total_clusters = result.bfi.total_clusters;
620 check->has_total_clusters = result.bfi.total_clusters != 0;
621 check->allocated_clusters = result.bfi.allocated_clusters;
622 check->has_allocated_clusters = result.bfi.allocated_clusters != 0;
623 check->fragmented_clusters = result.bfi.fragmented_clusters;
624 check->has_fragmented_clusters = result.bfi.fragmented_clusters != 0;
625 check->compressed_clusters = result.bfi.compressed_clusters;
626 check->has_compressed_clusters = result.bfi.compressed_clusters != 0;
627
628 return 0;
629 }
630
631 /*
632 * Checks an image for consistency. Exit codes:
633 *
634 * 0 - Check completed, image is good
635 * 1 - Check not completed because of internal errors
636 * 2 - Check completed, image is corrupted
637 * 3 - Check completed, image has leaked clusters, but is good otherwise
638 * 63 - Checks are not supported by the image format
639 */
640 static int img_check(int argc, char **argv)
641 {
642 int c, ret;
643 OutputFormat output_format = OFORMAT_HUMAN;
644 const char *filename, *fmt, *output, *cache;
645 BlockBackend *blk;
646 BlockDriverState *bs;
647 int fix = 0;
648 int flags = BDRV_O_CHECK;
649 bool writethrough;
650 ImageCheck *check;
651 bool quiet = false;
652 bool image_opts = false;
653
654 fmt = NULL;
655 output = NULL;
656 cache = BDRV_DEFAULT_CACHE;
657
658 for(;;) {
659 int option_index = 0;
660 static const struct option long_options[] = {
661 {"help", no_argument, 0, 'h'},
662 {"format", required_argument, 0, 'f'},
663 {"repair", required_argument, 0, 'r'},
664 {"output", required_argument, 0, OPTION_OUTPUT},
665 {"object", required_argument, 0, OPTION_OBJECT},
666 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
667 {0, 0, 0, 0}
668 };
669 c = getopt_long(argc, argv, ":hf:r:T:q",
670 long_options, &option_index);
671 if (c == -1) {
672 break;
673 }
674 switch(c) {
675 case ':':
676 missing_argument(argv[optind - 1]);
677 break;
678 case '?':
679 unrecognized_option(argv[optind - 1]);
680 break;
681 case 'h':
682 help();
683 break;
684 case 'f':
685 fmt = optarg;
686 break;
687 case 'r':
688 flags |= BDRV_O_RDWR;
689
690 if (!strcmp(optarg, "leaks")) {
691 fix = BDRV_FIX_LEAKS;
692 } else if (!strcmp(optarg, "all")) {
693 fix = BDRV_FIX_LEAKS | BDRV_FIX_ERRORS;
694 } else {
695 error_exit("Unknown option value for -r "
696 "(expecting 'leaks' or 'all'): %s", optarg);
697 }
698 break;
699 case OPTION_OUTPUT:
700 output = optarg;
701 break;
702 case 'T':
703 cache = optarg;
704 break;
705 case 'q':
706 quiet = true;
707 break;
708 case OPTION_OBJECT: {
709 QemuOpts *opts;
710 opts = qemu_opts_parse_noisily(&qemu_object_opts,
711 optarg, true);
712 if (!opts) {
713 return 1;
714 }
715 } break;
716 case OPTION_IMAGE_OPTS:
717 image_opts = true;
718 break;
719 }
720 }
721 if (optind != argc - 1) {
722 error_exit("Expecting one image file name");
723 }
724 filename = argv[optind++];
725
726 if (output && !strcmp(output, "json")) {
727 output_format = OFORMAT_JSON;
728 } else if (output && !strcmp(output, "human")) {
729 output_format = OFORMAT_HUMAN;
730 } else if (output) {
731 error_report("--output must be used with human or json as argument.");
732 return 1;
733 }
734
735 if (qemu_opts_foreach(&qemu_object_opts,
736 user_creatable_add_opts_foreach,
737 NULL, NULL)) {
738 return 1;
739 }
740
741 ret = bdrv_parse_cache_mode(cache, &flags, &writethrough);
742 if (ret < 0) {
743 error_report("Invalid source cache option: %s", cache);
744 return 1;
745 }
746
747 blk = img_open(image_opts, filename, fmt, flags, writethrough, quiet);
748 if (!blk) {
749 return 1;
750 }
751 bs = blk_bs(blk);
752
753 check = g_new0(ImageCheck, 1);
754 ret = collect_image_check(bs, check, filename, fmt, fix);
755
756 if (ret == -ENOTSUP) {
757 error_report("This image format does not support checks");
758 ret = 63;
759 goto fail;
760 }
761
762 if (check->corruptions_fixed || check->leaks_fixed) {
763 int corruptions_fixed, leaks_fixed;
764
765 leaks_fixed = check->leaks_fixed;
766 corruptions_fixed = check->corruptions_fixed;
767
768 if (output_format == OFORMAT_HUMAN) {
769 qprintf(quiet,
770 "The following inconsistencies were found and repaired:\n\n"
771 " %" PRId64 " leaked clusters\n"
772 " %" PRId64 " corruptions\n\n"
773 "Double checking the fixed image now...\n",
774 check->leaks_fixed,
775 check->corruptions_fixed);
776 }
777
778 ret = collect_image_check(bs, check, filename, fmt, 0);
779
780 check->leaks_fixed = leaks_fixed;
781 check->corruptions_fixed = corruptions_fixed;
782 }
783
784 if (!ret) {
785 switch (output_format) {
786 case OFORMAT_HUMAN:
787 dump_human_image_check(check, quiet);
788 break;
789 case OFORMAT_JSON:
790 dump_json_image_check(check, quiet);
791 break;
792 }
793 }
794
795 if (ret || check->check_errors) {
796 if (ret) {
797 error_report("Check failed: %s", strerror(-ret));
798 } else {
799 error_report("Check failed");
800 }
801 ret = 1;
802 goto fail;
803 }
804
805 if (check->corruptions) {
806 ret = 2;
807 } else if (check->leaks) {
808 ret = 3;
809 } else {
810 ret = 0;
811 }
812
813 fail:
814 qapi_free_ImageCheck(check);
815 blk_unref(blk);
816 return ret;
817 }
818
819 typedef struct CommonBlockJobCBInfo {
820 BlockDriverState *bs;
821 Error **errp;
822 } CommonBlockJobCBInfo;
823
824 static void common_block_job_cb(void *opaque, int ret)
825 {
826 CommonBlockJobCBInfo *cbi = opaque;
827
828 if (ret < 0) {
829 error_setg_errno(cbi->errp, -ret, "Block job failed");
830 }
831 }
832
833 static void run_block_job(BlockJob *job, Error **errp)
834 {
835 AioContext *aio_context = blk_get_aio_context(job->blk);
836
837 /* FIXME In error cases, the job simply goes away and we access a dangling
838 * pointer below. */
839 aio_context_acquire(aio_context);
840 do {
841 aio_poll(aio_context, true);
842 qemu_progress_print(job->len ?
843 ((float)job->offset / job->len * 100.f) : 0.0f, 0);
844 } while (!job->ready);
845
846 block_job_complete_sync(job, errp);
847 aio_context_release(aio_context);
848
849 /* A block job may finish instantaneously without publishing any progress,
850 * so just signal completion here */
851 qemu_progress_print(100.f, 0);
852 }
853
854 static int img_commit(int argc, char **argv)
855 {
856 int c, ret, flags;
857 const char *filename, *fmt, *cache, *base;
858 BlockBackend *blk;
859 BlockDriverState *bs, *base_bs;
860 BlockJob *job;
861 bool progress = false, quiet = false, drop = false;
862 bool writethrough;
863 Error *local_err = NULL;
864 CommonBlockJobCBInfo cbi;
865 bool image_opts = false;
866 AioContext *aio_context;
867
868 fmt = NULL;
869 cache = BDRV_DEFAULT_CACHE;
870 base = NULL;
871 for(;;) {
872 static const struct option long_options[] = {
873 {"help", no_argument, 0, 'h'},
874 {"object", required_argument, 0, OPTION_OBJECT},
875 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
876 {0, 0, 0, 0}
877 };
878 c = getopt_long(argc, argv, ":f:ht:b:dpq",
879 long_options, NULL);
880 if (c == -1) {
881 break;
882 }
883 switch(c) {
884 case ':':
885 missing_argument(argv[optind - 1]);
886 break;
887 case '?':
888 unrecognized_option(argv[optind - 1]);
889 break;
890 case 'h':
891 help();
892 break;
893 case 'f':
894 fmt = optarg;
895 break;
896 case 't':
897 cache = optarg;
898 break;
899 case 'b':
900 base = optarg;
901 /* -b implies -d */
902 drop = true;
903 break;
904 case 'd':
905 drop = true;
906 break;
907 case 'p':
908 progress = true;
909 break;
910 case 'q':
911 quiet = true;
912 break;
913 case OPTION_OBJECT: {
914 QemuOpts *opts;
915 opts = qemu_opts_parse_noisily(&qemu_object_opts,
916 optarg, true);
917 if (!opts) {
918 return 1;
919 }
920 } break;
921 case OPTION_IMAGE_OPTS:
922 image_opts = true;
923 break;
924 }
925 }
926
927 /* Progress is not shown in Quiet mode */
928 if (quiet) {
929 progress = false;
930 }
931
932 if (optind != argc - 1) {
933 error_exit("Expecting one image file name");
934 }
935 filename = argv[optind++];
936
937 if (qemu_opts_foreach(&qemu_object_opts,
938 user_creatable_add_opts_foreach,
939 NULL, NULL)) {
940 return 1;
941 }
942
943 flags = BDRV_O_RDWR | BDRV_O_UNMAP;
944 ret = bdrv_parse_cache_mode(cache, &flags, &writethrough);
945 if (ret < 0) {
946 error_report("Invalid cache option: %s", cache);
947 return 1;
948 }
949
950 blk = img_open(image_opts, filename, fmt, flags, writethrough, quiet);
951 if (!blk) {
952 return 1;
953 }
954 bs = blk_bs(blk);
955
956 qemu_progress_init(progress, 1.f);
957 qemu_progress_print(0.f, 100);
958
959 if (base) {
960 base_bs = bdrv_find_backing_image(bs, base);
961 if (!base_bs) {
962 error_setg(&local_err,
963 "Did not find '%s' in the backing chain of '%s'",
964 base, filename);
965 goto done;
966 }
967 } else {
968 /* This is different from QMP, which by default uses the deepest file in
969 * the backing chain (i.e., the very base); however, the traditional
970 * behavior of qemu-img commit is using the immediate backing file. */
971 base_bs = backing_bs(bs);
972 if (!base_bs) {
973 error_setg(&local_err, "Image does not have a backing file");
974 goto done;
975 }
976 }
977
978 cbi = (CommonBlockJobCBInfo){
979 .errp = &local_err,
980 .bs = bs,
981 };
982
983 aio_context = bdrv_get_aio_context(bs);
984 aio_context_acquire(aio_context);
985 commit_active_start("commit", bs, base_bs, BLOCK_JOB_DEFAULT, 0,
986 BLOCKDEV_ON_ERROR_REPORT, NULL, common_block_job_cb,
987 &cbi, false, &local_err);
988 aio_context_release(aio_context);
989 if (local_err) {
990 goto done;
991 }
992
993 /* When the block job completes, the BlockBackend reference will point to
994 * the old backing file. In order to avoid that the top image is already
995 * deleted, so we can still empty it afterwards, increment the reference
996 * counter here preemptively. */
997 if (!drop) {
998 bdrv_ref(bs);
999 }
1000
1001 job = block_job_get("commit");
1002 run_block_job(job, &local_err);
1003 if (local_err) {
1004 goto unref_backing;
1005 }
1006
1007 if (!drop && bs->drv->bdrv_make_empty) {
1008 ret = bs->drv->bdrv_make_empty(bs);
1009 if (ret) {
1010 error_setg_errno(&local_err, -ret, "Could not empty %s",
1011 filename);
1012 goto unref_backing;
1013 }
1014 }
1015
1016 unref_backing:
1017 if (!drop) {
1018 bdrv_unref(bs);
1019 }
1020
1021 done:
1022 qemu_progress_end();
1023
1024 blk_unref(blk);
1025
1026 if (local_err) {
1027 error_report_err(local_err);
1028 return 1;
1029 }
1030
1031 qprintf(quiet, "Image committed.\n");
1032 return 0;
1033 }
1034
1035 /*
1036 * Returns true iff the first sector pointed to by 'buf' contains at least
1037 * a non-NUL byte.
1038 *
1039 * 'pnum' is set to the number of sectors (including and immediately following
1040 * the first one) that are known to be in the same allocated/unallocated state.
1041 */
1042 static int is_allocated_sectors(const uint8_t *buf, int n, int *pnum)
1043 {
1044 bool is_zero;
1045 int i;
1046
1047 if (n <= 0) {
1048 *pnum = 0;
1049 return 0;
1050 }
1051 is_zero = buffer_is_zero(buf, 512);
1052 for(i = 1; i < n; i++) {
1053 buf += 512;
1054 if (is_zero != buffer_is_zero(buf, 512)) {
1055 break;
1056 }
1057 }
1058 *pnum = i;
1059 return !is_zero;
1060 }
1061
1062 /*
1063 * Like is_allocated_sectors, but if the buffer starts with a used sector,
1064 * up to 'min' consecutive sectors containing zeros are ignored. This avoids
1065 * breaking up write requests for only small sparse areas.
1066 */
1067 static int is_allocated_sectors_min(const uint8_t *buf, int n, int *pnum,
1068 int min)
1069 {
1070 int ret;
1071 int num_checked, num_used;
1072
1073 if (n < min) {
1074 min = n;
1075 }
1076
1077 ret = is_allocated_sectors(buf, n, pnum);
1078 if (!ret) {
1079 return ret;
1080 }
1081
1082 num_used = *pnum;
1083 buf += BDRV_SECTOR_SIZE * *pnum;
1084 n -= *pnum;
1085 num_checked = num_used;
1086
1087 while (n > 0) {
1088 ret = is_allocated_sectors(buf, n, pnum);
1089
1090 buf += BDRV_SECTOR_SIZE * *pnum;
1091 n -= *pnum;
1092 num_checked += *pnum;
1093 if (ret) {
1094 num_used = num_checked;
1095 } else if (*pnum >= min) {
1096 break;
1097 }
1098 }
1099
1100 *pnum = num_used;
1101 return 1;
1102 }
1103
1104 /*
1105 * Compares two buffers sector by sector. Returns 0 if the first sector of both
1106 * buffers matches, non-zero otherwise.
1107 *
1108 * pnum is set to the number of sectors (including and immediately following
1109 * the first one) that are known to have the same comparison result
1110 */
1111 static int compare_sectors(const uint8_t *buf1, const uint8_t *buf2, int n,
1112 int *pnum)
1113 {
1114 bool res;
1115 int i;
1116
1117 if (n <= 0) {
1118 *pnum = 0;
1119 return 0;
1120 }
1121
1122 res = !!memcmp(buf1, buf2, 512);
1123 for(i = 1; i < n; i++) {
1124 buf1 += 512;
1125 buf2 += 512;
1126
1127 if (!!memcmp(buf1, buf2, 512) != res) {
1128 break;
1129 }
1130 }
1131
1132 *pnum = i;
1133 return res;
1134 }
1135
1136 #define IO_BUF_SIZE (2 * 1024 * 1024)
1137
1138 static int64_t sectors_to_bytes(int64_t sectors)
1139 {
1140 return sectors << BDRV_SECTOR_BITS;
1141 }
1142
1143 static int64_t sectors_to_process(int64_t total, int64_t from)
1144 {
1145 return MIN(total - from, IO_BUF_SIZE >> BDRV_SECTOR_BITS);
1146 }
1147
1148 /*
1149 * Check if passed sectors are empty (not allocated or contain only 0 bytes)
1150 *
1151 * Returns 0 in case sectors are filled with 0, 1 if sectors contain non-zero
1152 * data and negative value on error.
1153 *
1154 * @param blk: BlockBackend for the image
1155 * @param sect_num: Number of first sector to check
1156 * @param sect_count: Number of sectors to check
1157 * @param filename: Name of disk file we are checking (logging purpose)
1158 * @param buffer: Allocated buffer for storing read data
1159 * @param quiet: Flag for quiet mode
1160 */
1161 static int check_empty_sectors(BlockBackend *blk, int64_t sect_num,
1162 int sect_count, const char *filename,
1163 uint8_t *buffer, bool quiet)
1164 {
1165 int pnum, ret = 0;
1166 ret = blk_pread(blk, sect_num << BDRV_SECTOR_BITS, buffer,
1167 sect_count << BDRV_SECTOR_BITS);
1168 if (ret < 0) {
1169 error_report("Error while reading offset %" PRId64 " of %s: %s",
1170 sectors_to_bytes(sect_num), filename, strerror(-ret));
1171 return ret;
1172 }
1173 ret = is_allocated_sectors(buffer, sect_count, &pnum);
1174 if (ret || pnum != sect_count) {
1175 qprintf(quiet, "Content mismatch at offset %" PRId64 "!\n",
1176 sectors_to_bytes(ret ? sect_num : sect_num + pnum));
1177 return 1;
1178 }
1179
1180 return 0;
1181 }
1182
1183 /*
1184 * Compares two images. Exit codes:
1185 *
1186 * 0 - Images are identical
1187 * 1 - Images differ
1188 * >1 - Error occurred
1189 */
1190 static int img_compare(int argc, char **argv)
1191 {
1192 const char *fmt1 = NULL, *fmt2 = NULL, *cache, *filename1, *filename2;
1193 BlockBackend *blk1, *blk2;
1194 BlockDriverState *bs1, *bs2;
1195 int64_t total_sectors1, total_sectors2;
1196 uint8_t *buf1 = NULL, *buf2 = NULL;
1197 int pnum1, pnum2;
1198 int allocated1, allocated2;
1199 int ret = 0; /* return value - 0 Ident, 1 Different, >1 Error */
1200 bool progress = false, quiet = false, strict = false;
1201 int flags;
1202 bool writethrough;
1203 int64_t total_sectors;
1204 int64_t sector_num = 0;
1205 int64_t nb_sectors;
1206 int c, pnum;
1207 uint64_t progress_base;
1208 bool image_opts = false;
1209
1210 cache = BDRV_DEFAULT_CACHE;
1211 for (;;) {
1212 static const struct option long_options[] = {
1213 {"help", no_argument, 0, 'h'},
1214 {"object", required_argument, 0, OPTION_OBJECT},
1215 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
1216 {0, 0, 0, 0}
1217 };
1218 c = getopt_long(argc, argv, ":hf:F:T:pqs",
1219 long_options, NULL);
1220 if (c == -1) {
1221 break;
1222 }
1223 switch (c) {
1224 case ':':
1225 missing_argument(argv[optind - 1]);
1226 break;
1227 case '?':
1228 unrecognized_option(argv[optind - 1]);
1229 break;
1230 case 'h':
1231 help();
1232 break;
1233 case 'f':
1234 fmt1 = optarg;
1235 break;
1236 case 'F':
1237 fmt2 = optarg;
1238 break;
1239 case 'T':
1240 cache = optarg;
1241 break;
1242 case 'p':
1243 progress = true;
1244 break;
1245 case 'q':
1246 quiet = true;
1247 break;
1248 case 's':
1249 strict = true;
1250 break;
1251 case OPTION_OBJECT: {
1252 QemuOpts *opts;
1253 opts = qemu_opts_parse_noisily(&qemu_object_opts,
1254 optarg, true);
1255 if (!opts) {
1256 ret = 2;
1257 goto out4;
1258 }
1259 } break;
1260 case OPTION_IMAGE_OPTS:
1261 image_opts = true;
1262 break;
1263 }
1264 }
1265
1266 /* Progress is not shown in Quiet mode */
1267 if (quiet) {
1268 progress = false;
1269 }
1270
1271
1272 if (optind != argc - 2) {
1273 error_exit("Expecting two image file names");
1274 }
1275 filename1 = argv[optind++];
1276 filename2 = argv[optind++];
1277
1278 if (qemu_opts_foreach(&qemu_object_opts,
1279 user_creatable_add_opts_foreach,
1280 NULL, NULL)) {
1281 ret = 2;
1282 goto out4;
1283 }
1284
1285 /* Initialize before goto out */
1286 qemu_progress_init(progress, 2.0);
1287
1288 flags = 0;
1289 ret = bdrv_parse_cache_mode(cache, &flags, &writethrough);
1290 if (ret < 0) {
1291 error_report("Invalid source cache option: %s", cache);
1292 ret = 2;
1293 goto out3;
1294 }
1295
1296 blk1 = img_open(image_opts, filename1, fmt1, flags, writethrough, quiet);
1297 if (!blk1) {
1298 ret = 2;
1299 goto out3;
1300 }
1301
1302 blk2 = img_open(image_opts, filename2, fmt2, flags, writethrough, quiet);
1303 if (!blk2) {
1304 ret = 2;
1305 goto out2;
1306 }
1307 bs1 = blk_bs(blk1);
1308 bs2 = blk_bs(blk2);
1309
1310 buf1 = blk_blockalign(blk1, IO_BUF_SIZE);
1311 buf2 = blk_blockalign(blk2, IO_BUF_SIZE);
1312 total_sectors1 = blk_nb_sectors(blk1);
1313 if (total_sectors1 < 0) {
1314 error_report("Can't get size of %s: %s",
1315 filename1, strerror(-total_sectors1));
1316 ret = 4;
1317 goto out;
1318 }
1319 total_sectors2 = blk_nb_sectors(blk2);
1320 if (total_sectors2 < 0) {
1321 error_report("Can't get size of %s: %s",
1322 filename2, strerror(-total_sectors2));
1323 ret = 4;
1324 goto out;
1325 }
1326 total_sectors = MIN(total_sectors1, total_sectors2);
1327 progress_base = MAX(total_sectors1, total_sectors2);
1328
1329 qemu_progress_print(0, 100);
1330
1331 if (strict && total_sectors1 != total_sectors2) {
1332 ret = 1;
1333 qprintf(quiet, "Strict mode: Image size mismatch!\n");
1334 goto out;
1335 }
1336
1337 for (;;) {
1338 int64_t status1, status2;
1339 BlockDriverState *file;
1340
1341 nb_sectors = sectors_to_process(total_sectors, sector_num);
1342 if (nb_sectors <= 0) {
1343 break;
1344 }
1345 status1 = bdrv_get_block_status_above(bs1, NULL, sector_num,
1346 total_sectors1 - sector_num,
1347 &pnum1, &file);
1348 if (status1 < 0) {
1349 ret = 3;
1350 error_report("Sector allocation test failed for %s", filename1);
1351 goto out;
1352 }
1353 allocated1 = status1 & BDRV_BLOCK_ALLOCATED;
1354
1355 status2 = bdrv_get_block_status_above(bs2, NULL, sector_num,
1356 total_sectors2 - sector_num,
1357 &pnum2, &file);
1358 if (status2 < 0) {
1359 ret = 3;
1360 error_report("Sector allocation test failed for %s", filename2);
1361 goto out;
1362 }
1363 allocated2 = status2 & BDRV_BLOCK_ALLOCATED;
1364 if (pnum1) {
1365 nb_sectors = MIN(nb_sectors, pnum1);
1366 }
1367 if (pnum2) {
1368 nb_sectors = MIN(nb_sectors, pnum2);
1369 }
1370
1371 if (strict) {
1372 if ((status1 & ~BDRV_BLOCK_OFFSET_MASK) !=
1373 (status2 & ~BDRV_BLOCK_OFFSET_MASK)) {
1374 ret = 1;
1375 qprintf(quiet, "Strict mode: Offset %" PRId64
1376 " block status mismatch!\n",
1377 sectors_to_bytes(sector_num));
1378 goto out;
1379 }
1380 }
1381 if ((status1 & BDRV_BLOCK_ZERO) && (status2 & BDRV_BLOCK_ZERO)) {
1382 nb_sectors = MIN(pnum1, pnum2);
1383 } else if (allocated1 == allocated2) {
1384 if (allocated1) {
1385 ret = blk_pread(blk1, sector_num << BDRV_SECTOR_BITS, buf1,
1386 nb_sectors << BDRV_SECTOR_BITS);
1387 if (ret < 0) {
1388 error_report("Error while reading offset %" PRId64 " of %s:"
1389 " %s", sectors_to_bytes(sector_num), filename1,
1390 strerror(-ret));
1391 ret = 4;
1392 goto out;
1393 }
1394 ret = blk_pread(blk2, sector_num << BDRV_SECTOR_BITS, buf2,
1395 nb_sectors << BDRV_SECTOR_BITS);
1396 if (ret < 0) {
1397 error_report("Error while reading offset %" PRId64
1398 " of %s: %s", sectors_to_bytes(sector_num),
1399 filename2, strerror(-ret));
1400 ret = 4;
1401 goto out;
1402 }
1403 ret = compare_sectors(buf1, buf2, nb_sectors, &pnum);
1404 if (ret || pnum != nb_sectors) {
1405 qprintf(quiet, "Content mismatch at offset %" PRId64 "!\n",
1406 sectors_to_bytes(
1407 ret ? sector_num : sector_num + pnum));
1408 ret = 1;
1409 goto out;
1410 }
1411 }
1412 } else {
1413
1414 if (allocated1) {
1415 ret = check_empty_sectors(blk1, sector_num, nb_sectors,
1416 filename1, buf1, quiet);
1417 } else {
1418 ret = check_empty_sectors(blk2, sector_num, nb_sectors,
1419 filename2, buf1, quiet);
1420 }
1421 if (ret) {
1422 if (ret < 0) {
1423 error_report("Error while reading offset %" PRId64 ": %s",
1424 sectors_to_bytes(sector_num), strerror(-ret));
1425 ret = 4;
1426 }
1427 goto out;
1428 }
1429 }
1430 sector_num += nb_sectors;
1431 qemu_progress_print(((float) nb_sectors / progress_base)*100, 100);
1432 }
1433
1434 if (total_sectors1 != total_sectors2) {
1435 BlockBackend *blk_over;
1436 int64_t total_sectors_over;
1437 const char *filename_over;
1438
1439 qprintf(quiet, "Warning: Image size mismatch!\n");
1440 if (total_sectors1 > total_sectors2) {
1441 total_sectors_over = total_sectors1;
1442 blk_over = blk1;
1443 filename_over = filename1;
1444 } else {
1445 total_sectors_over = total_sectors2;
1446 blk_over = blk2;
1447 filename_over = filename2;
1448 }
1449
1450 for (;;) {
1451 nb_sectors = sectors_to_process(total_sectors_over, sector_num);
1452 if (nb_sectors <= 0) {
1453 break;
1454 }
1455 ret = bdrv_is_allocated_above(blk_bs(blk_over), NULL, sector_num,
1456 nb_sectors, &pnum);
1457 if (ret < 0) {
1458 ret = 3;
1459 error_report("Sector allocation test failed for %s",
1460 filename_over);
1461 goto out;
1462
1463 }
1464 nb_sectors = pnum;
1465 if (ret) {
1466 ret = check_empty_sectors(blk_over, sector_num, nb_sectors,
1467 filename_over, buf1, quiet);
1468 if (ret) {
1469 if (ret < 0) {
1470 error_report("Error while reading offset %" PRId64
1471 " of %s: %s", sectors_to_bytes(sector_num),
1472 filename_over, strerror(-ret));
1473 ret = 4;
1474 }
1475 goto out;
1476 }
1477 }
1478 sector_num += nb_sectors;
1479 qemu_progress_print(((float) nb_sectors / progress_base)*100, 100);
1480 }
1481 }
1482
1483 qprintf(quiet, "Images are identical.\n");
1484 ret = 0;
1485
1486 out:
1487 qemu_vfree(buf1);
1488 qemu_vfree(buf2);
1489 blk_unref(blk2);
1490 out2:
1491 blk_unref(blk1);
1492 out3:
1493 qemu_progress_end();
1494 out4:
1495 return ret;
1496 }
1497
1498 enum ImgConvertBlockStatus {
1499 BLK_DATA,
1500 BLK_ZERO,
1501 BLK_BACKING_FILE,
1502 };
1503
1504 #define MAX_COROUTINES 16
1505
1506 typedef struct ImgConvertState {
1507 BlockBackend **src;
1508 int64_t *src_sectors;
1509 int src_num;
1510 int64_t total_sectors;
1511 int64_t allocated_sectors;
1512 int64_t allocated_done;
1513 int64_t sector_num;
1514 int64_t wr_offs;
1515 enum ImgConvertBlockStatus status;
1516 int64_t sector_next_status;
1517 BlockBackend *target;
1518 bool has_zero_init;
1519 bool compressed;
1520 bool target_has_backing;
1521 bool wr_in_order;
1522 int min_sparse;
1523 size_t cluster_sectors;
1524 size_t buf_sectors;
1525 long num_coroutines;
1526 int running_coroutines;
1527 Coroutine *co[MAX_COROUTINES];
1528 int64_t wait_sector_num[MAX_COROUTINES];
1529 CoMutex lock;
1530 int ret;
1531 } ImgConvertState;
1532
1533 static void convert_select_part(ImgConvertState *s, int64_t sector_num,
1534 int *src_cur, int64_t *src_cur_offset)
1535 {
1536 *src_cur = 0;
1537 *src_cur_offset = 0;
1538 while (sector_num - *src_cur_offset >= s->src_sectors[*src_cur]) {
1539 *src_cur_offset += s->src_sectors[*src_cur];
1540 (*src_cur)++;
1541 assert(*src_cur < s->src_num);
1542 }
1543 }
1544
1545 static int convert_iteration_sectors(ImgConvertState *s, int64_t sector_num)
1546 {
1547 int64_t ret, src_cur_offset;
1548 int n, src_cur;
1549
1550 convert_select_part(s, sector_num, &src_cur, &src_cur_offset);
1551
1552 assert(s->total_sectors > sector_num);
1553 n = MIN(s->total_sectors - sector_num, BDRV_REQUEST_MAX_SECTORS);
1554
1555 if (s->sector_next_status <= sector_num) {
1556 BlockDriverState *file;
1557 if (s->target_has_backing) {
1558 ret = bdrv_get_block_status(blk_bs(s->src[src_cur]),
1559 sector_num - src_cur_offset,
1560 n, &n, &file);
1561 } else {
1562 ret = bdrv_get_block_status_above(blk_bs(s->src[src_cur]), NULL,
1563 sector_num - src_cur_offset,
1564 n, &n, &file);
1565 }
1566 if (ret < 0) {
1567 return ret;
1568 }
1569
1570 if (ret & BDRV_BLOCK_ZERO) {
1571 s->status = BLK_ZERO;
1572 } else if (ret & BDRV_BLOCK_DATA) {
1573 s->status = BLK_DATA;
1574 } else {
1575 s->status = s->target_has_backing ? BLK_BACKING_FILE : BLK_DATA;
1576 }
1577
1578 s->sector_next_status = sector_num + n;
1579 }
1580
1581 n = MIN(n, s->sector_next_status - sector_num);
1582 if (s->status == BLK_DATA) {
1583 n = MIN(n, s->buf_sectors);
1584 }
1585
1586 /* We need to write complete clusters for compressed images, so if an
1587 * unallocated area is shorter than that, we must consider the whole
1588 * cluster allocated. */
1589 if (s->compressed) {
1590 if (n < s->cluster_sectors) {
1591 n = MIN(s->cluster_sectors, s->total_sectors - sector_num);
1592 s->status = BLK_DATA;
1593 } else {
1594 n = QEMU_ALIGN_DOWN(n, s->cluster_sectors);
1595 }
1596 }
1597
1598 return n;
1599 }
1600
1601 static int coroutine_fn convert_co_read(ImgConvertState *s, int64_t sector_num,
1602 int nb_sectors, uint8_t *buf)
1603 {
1604 int n, ret;
1605 QEMUIOVector qiov;
1606 struct iovec iov;
1607
1608 assert(nb_sectors <= s->buf_sectors);
1609 while (nb_sectors > 0) {
1610 BlockBackend *blk;
1611 int src_cur;
1612 int64_t bs_sectors, src_cur_offset;
1613
1614 /* In the case of compression with multiple source files, we can get a
1615 * nb_sectors that spreads into the next part. So we must be able to
1616 * read across multiple BDSes for one convert_read() call. */
1617 convert_select_part(s, sector_num, &src_cur, &src_cur_offset);
1618 blk = s->src[src_cur];
1619 bs_sectors = s->src_sectors[src_cur];
1620
1621 n = MIN(nb_sectors, bs_sectors - (sector_num - src_cur_offset));
1622 iov.iov_base = buf;
1623 iov.iov_len = n << BDRV_SECTOR_BITS;
1624 qemu_iovec_init_external(&qiov, &iov, 1);
1625
1626 ret = blk_co_preadv(
1627 blk, (sector_num - src_cur_offset) << BDRV_SECTOR_BITS,
1628 n << BDRV_SECTOR_BITS, &qiov, 0);
1629 if (ret < 0) {
1630 return ret;
1631 }
1632
1633 sector_num += n;
1634 nb_sectors -= n;
1635 buf += n * BDRV_SECTOR_SIZE;
1636 }
1637
1638 return 0;
1639 }
1640
1641
1642 static int coroutine_fn convert_co_write(ImgConvertState *s, int64_t sector_num,
1643 int nb_sectors, uint8_t *buf,
1644 enum ImgConvertBlockStatus status)
1645 {
1646 int ret;
1647 QEMUIOVector qiov;
1648 struct iovec iov;
1649
1650 while (nb_sectors > 0) {
1651 int n = nb_sectors;
1652 BdrvRequestFlags flags = s->compressed ? BDRV_REQ_WRITE_COMPRESSED : 0;
1653
1654 switch (status) {
1655 case BLK_BACKING_FILE:
1656 /* If we have a backing file, leave clusters unallocated that are
1657 * unallocated in the source image, so that the backing file is
1658 * visible at the respective offset. */
1659 assert(s->target_has_backing);
1660 break;
1661
1662 case BLK_DATA:
1663 /* If we're told to keep the target fully allocated (-S 0) or there
1664 * is real non-zero data, we must write it. Otherwise we can treat
1665 * it as zero sectors.
1666 * Compressed clusters need to be written as a whole, so in that
1667 * case we can only save the write if the buffer is completely
1668 * zeroed. */
1669 if (!s->min_sparse ||
1670 (!s->compressed &&
1671 is_allocated_sectors_min(buf, n, &n, s->min_sparse)) ||
1672 (s->compressed &&
1673 !buffer_is_zero(buf, n * BDRV_SECTOR_SIZE)))
1674 {
1675 iov.iov_base = buf;
1676 iov.iov_len = n << BDRV_SECTOR_BITS;
1677 qemu_iovec_init_external(&qiov, &iov, 1);
1678
1679 ret = blk_co_pwritev(s->target, sector_num << BDRV_SECTOR_BITS,
1680 n << BDRV_SECTOR_BITS, &qiov, flags);
1681 if (ret < 0) {
1682 return ret;
1683 }
1684 break;
1685 }
1686 /* fall-through */
1687
1688 case BLK_ZERO:
1689 if (s->has_zero_init) {
1690 assert(!s->target_has_backing);
1691 break;
1692 }
1693 ret = blk_co_pwrite_zeroes(s->target,
1694 sector_num << BDRV_SECTOR_BITS,
1695 n << BDRV_SECTOR_BITS, 0);
1696 if (ret < 0) {
1697 return ret;
1698 }
1699 break;
1700 }
1701
1702 sector_num += n;
1703 nb_sectors -= n;
1704 buf += n * BDRV_SECTOR_SIZE;
1705 }
1706
1707 return 0;
1708 }
1709
1710 static void coroutine_fn convert_co_do_copy(void *opaque)
1711 {
1712 ImgConvertState *s = opaque;
1713 uint8_t *buf = NULL;
1714 int ret, i;
1715 int index = -1;
1716
1717 for (i = 0; i < s->num_coroutines; i++) {
1718 if (s->co[i] == qemu_coroutine_self()) {
1719 index = i;
1720 break;
1721 }
1722 }
1723 assert(index >= 0);
1724
1725 s->running_coroutines++;
1726 buf = blk_blockalign(s->target, s->buf_sectors * BDRV_SECTOR_SIZE);
1727
1728 while (1) {
1729 int n;
1730 int64_t sector_num;
1731 enum ImgConvertBlockStatus status;
1732
1733 qemu_co_mutex_lock(&s->lock);
1734 if (s->ret != -EINPROGRESS || s->sector_num >= s->total_sectors) {
1735 qemu_co_mutex_unlock(&s->lock);
1736 goto out;
1737 }
1738 n = convert_iteration_sectors(s, s->sector_num);
1739 if (n < 0) {
1740 qemu_co_mutex_unlock(&s->lock);
1741 s->ret = n;
1742 goto out;
1743 }
1744 /* save current sector and allocation status to local variables */
1745 sector_num = s->sector_num;
1746 status = s->status;
1747 if (!s->min_sparse && s->status == BLK_ZERO) {
1748 n = MIN(n, s->buf_sectors);
1749 }
1750 /* increment global sector counter so that other coroutines can
1751 * already continue reading beyond this request */
1752 s->sector_num += n;
1753 qemu_co_mutex_unlock(&s->lock);
1754
1755 if (status == BLK_DATA || (!s->min_sparse && status == BLK_ZERO)) {
1756 s->allocated_done += n;
1757 qemu_progress_print(100.0 * s->allocated_done /
1758 s->allocated_sectors, 0);
1759 }
1760
1761 if (status == BLK_DATA) {
1762 ret = convert_co_read(s, sector_num, n, buf);
1763 if (ret < 0) {
1764 error_report("error while reading sector %" PRId64
1765 ": %s", sector_num, strerror(-ret));
1766 s->ret = ret;
1767 goto out;
1768 }
1769 } else if (!s->min_sparse && status == BLK_ZERO) {
1770 status = BLK_DATA;
1771 memset(buf, 0x00, n * BDRV_SECTOR_SIZE);
1772 }
1773
1774 if (s->wr_in_order) {
1775 /* keep writes in order */
1776 while (s->wr_offs != sector_num) {
1777 if (s->ret != -EINPROGRESS) {
1778 goto out;
1779 }
1780 s->wait_sector_num[index] = sector_num;
1781 qemu_coroutine_yield();
1782 }
1783 s->wait_sector_num[index] = -1;
1784 }
1785
1786 ret = convert_co_write(s, sector_num, n, buf, status);
1787 if (ret < 0) {
1788 error_report("error while writing sector %" PRId64
1789 ": %s", sector_num, strerror(-ret));
1790 s->ret = ret;
1791 goto out;
1792 }
1793
1794 if (s->wr_in_order) {
1795 /* reenter the coroutine that might have waited
1796 * for this write to complete */
1797 s->wr_offs = sector_num + n;
1798 for (i = 0; i < s->num_coroutines; i++) {
1799 if (s->co[i] && s->wait_sector_num[i] == s->wr_offs) {
1800 /*
1801 * A -> B -> A cannot occur because A has
1802 * s->wait_sector_num[i] == -1 during A -> B. Therefore
1803 * B will never enter A during this time window.
1804 */
1805 qemu_coroutine_enter(s->co[i]);
1806 break;
1807 }
1808 }
1809 }
1810 }
1811
1812 out:
1813 qemu_vfree(buf);
1814 s->co[index] = NULL;
1815 s->running_coroutines--;
1816 if (!s->running_coroutines && s->ret == -EINPROGRESS) {
1817 /* the convert job finished successfully */
1818 s->ret = 0;
1819 }
1820 }
1821
1822 static int convert_do_copy(ImgConvertState *s)
1823 {
1824 int ret, i, n;
1825 int64_t sector_num = 0;
1826
1827 /* Check whether we have zero initialisation or can get it efficiently */
1828 s->has_zero_init = s->min_sparse && !s->target_has_backing
1829 ? bdrv_has_zero_init(blk_bs(s->target))
1830 : false;
1831
1832 if (!s->has_zero_init && !s->target_has_backing &&
1833 bdrv_can_write_zeroes_with_unmap(blk_bs(s->target)))
1834 {
1835 ret = blk_make_zero(s->target, BDRV_REQ_MAY_UNMAP);
1836 if (ret == 0) {
1837 s->has_zero_init = true;
1838 }
1839 }
1840
1841 /* Allocate buffer for copied data. For compressed images, only one cluster
1842 * can be copied at a time. */
1843 if (s->compressed) {
1844 if (s->cluster_sectors <= 0 || s->cluster_sectors > s->buf_sectors) {
1845 error_report("invalid cluster size");
1846 return -EINVAL;
1847 }
1848 s->buf_sectors = s->cluster_sectors;
1849 }
1850
1851 while (sector_num < s->total_sectors) {
1852 n = convert_iteration_sectors(s, sector_num);
1853 if (n < 0) {
1854 return n;
1855 }
1856 if (s->status == BLK_DATA || (!s->min_sparse && s->status == BLK_ZERO))
1857 {
1858 s->allocated_sectors += n;
1859 }
1860 sector_num += n;
1861 }
1862
1863 /* Do the copy */
1864 s->sector_next_status = 0;
1865 s->ret = -EINPROGRESS;
1866
1867 qemu_co_mutex_init(&s->lock);
1868 for (i = 0; i < s->num_coroutines; i++) {
1869 s->co[i] = qemu_coroutine_create(convert_co_do_copy, s);
1870 s->wait_sector_num[i] = -1;
1871 qemu_coroutine_enter(s->co[i]);
1872 }
1873
1874 while (s->ret == -EINPROGRESS) {
1875 main_loop_wait(false);
1876 }
1877
1878 if (s->compressed && !s->ret) {
1879 /* signal EOF to align */
1880 ret = blk_pwrite_compressed(s->target, 0, NULL, 0);
1881 if (ret < 0) {
1882 return ret;
1883 }
1884 }
1885
1886 return s->ret;
1887 }
1888
1889 static int img_convert(int argc, char **argv)
1890 {
1891 int c, bs_i, flags, src_flags = 0;
1892 const char *fmt = NULL, *out_fmt = "raw", *cache = "unsafe",
1893 *src_cache = BDRV_DEFAULT_CACHE, *out_baseimg = NULL,
1894 *out_filename, *out_baseimg_param, *snapshot_name = NULL;
1895 BlockDriver *drv, *proto_drv;
1896 BlockDriverInfo bdi;
1897 BlockDriverState *out_bs;
1898 QemuOpts *opts = NULL, *sn_opts = NULL;
1899 QemuOptsList *create_opts = NULL;
1900 char *options = NULL;
1901 Error *local_err = NULL;
1902 bool writethrough, src_writethrough, quiet = false, image_opts = false,
1903 skip_create = false, progress = false;
1904 int64_t ret = -EINVAL;
1905
1906 ImgConvertState s = (ImgConvertState) {
1907 /* Need at least 4k of zeros for sparse detection */
1908 .min_sparse = 8,
1909 .buf_sectors = IO_BUF_SIZE / BDRV_SECTOR_SIZE,
1910 .wr_in_order = true,
1911 .num_coroutines = 8,
1912 };
1913
1914 for(;;) {
1915 static const struct option long_options[] = {
1916 {"help", no_argument, 0, 'h'},
1917 {"object", required_argument, 0, OPTION_OBJECT},
1918 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
1919 {0, 0, 0, 0}
1920 };
1921 c = getopt_long(argc, argv, ":hf:O:B:ce6o:s:l:S:pt:T:qnm:W",
1922 long_options, NULL);
1923 if (c == -1) {
1924 break;
1925 }
1926 switch(c) {
1927 case ':':
1928 missing_argument(argv[optind - 1]);
1929 break;
1930 case '?':
1931 unrecognized_option(argv[optind - 1]);
1932 break;
1933 case 'h':
1934 help();
1935 break;
1936 case 'f':
1937 fmt = optarg;
1938 break;
1939 case 'O':
1940 out_fmt = optarg;
1941 break;
1942 case 'B':
1943 out_baseimg = optarg;
1944 break;
1945 case 'c':
1946 s.compressed = true;
1947 break;
1948 case 'e':
1949 error_report("option -e is deprecated, please use \'-o "
1950 "encryption\' instead!");
1951 goto fail_getopt;
1952 case '6':
1953 error_report("option -6 is deprecated, please use \'-o "
1954 "compat6\' instead!");
1955 goto fail_getopt;
1956 case 'o':
1957 if (!is_valid_option_list(optarg)) {
1958 error_report("Invalid option list: %s", optarg);
1959 goto fail_getopt;
1960 }
1961 if (!options) {
1962 options = g_strdup(optarg);
1963 } else {
1964 char *old_options = options;
1965 options = g_strdup_printf("%s,%s", options, optarg);
1966 g_free(old_options);
1967 }
1968 break;
1969 case 's':
1970 snapshot_name = optarg;
1971 break;
1972 case 'l':
1973 if (strstart(optarg, SNAPSHOT_OPT_BASE, NULL)) {
1974 sn_opts = qemu_opts_parse_noisily(&internal_snapshot_opts,
1975 optarg, false);
1976 if (!sn_opts) {
1977 error_report("Failed in parsing snapshot param '%s'",
1978 optarg);
1979 goto fail_getopt;
1980 }
1981 } else {
1982 snapshot_name = optarg;
1983 }
1984 break;
1985 case 'S':
1986 {
1987 int64_t sval;
1988
1989 sval = cvtnum(optarg);
1990 if (sval < 0) {
1991 error_report("Invalid minimum zero buffer size for sparse output specified");
1992 goto fail_getopt;
1993 }
1994
1995 s.min_sparse = sval / BDRV_SECTOR_SIZE;
1996 break;
1997 }
1998 case 'p':
1999 progress = true;
2000 break;
2001 case 't':
2002 cache = optarg;
2003 break;
2004 case 'T':
2005 src_cache = optarg;
2006 break;
2007 case 'q':
2008 quiet = true;
2009 break;
2010 case 'n':
2011 skip_create = true;
2012 break;
2013 case 'm':
2014 if (qemu_strtol(optarg, NULL, 0, &s.num_coroutines) ||
2015 s.num_coroutines < 1 || s.num_coroutines > MAX_COROUTINES) {
2016 error_report("Invalid number of coroutines. Allowed number of"
2017 " coroutines is between 1 and %d", MAX_COROUTINES);
2018 goto fail_getopt;
2019 }
2020 break;
2021 case 'W':
2022 s.wr_in_order = false;
2023 break;
2024 case OPTION_OBJECT: {
2025 QemuOpts *object_opts;
2026 object_opts = qemu_opts_parse_noisily(&qemu_object_opts,
2027 optarg, true);
2028 if (!object_opts) {
2029 goto fail_getopt;
2030 }
2031 break;
2032 }
2033 case OPTION_IMAGE_OPTS:
2034 image_opts = true;
2035 break;
2036 }
2037 }
2038
2039 if (qemu_opts_foreach(&qemu_object_opts,
2040 user_creatable_add_opts_foreach,
2041 NULL, NULL)) {
2042 goto fail_getopt;
2043 }
2044
2045 if (!s.wr_in_order && s.compressed) {
2046 error_report("Out of order write and compress are mutually exclusive");
2047 goto fail_getopt;
2048 }
2049
2050 s.src_num = argc - optind - 1;
2051 out_filename = s.src_num >= 1 ? argv[argc - 1] : NULL;
2052
2053 if (options && has_help_option(options)) {
2054 ret = print_block_option_help(out_filename, out_fmt);
2055 goto fail_getopt;
2056 }
2057
2058 if (s.src_num < 1) {
2059 error_report("Must specify image file name");
2060 goto fail_getopt;
2061 }
2062
2063
2064 /* ret is still -EINVAL until here */
2065 ret = bdrv_parse_cache_mode(src_cache, &src_flags, &src_writethrough);
2066 if (ret < 0) {
2067 error_report("Invalid source cache option: %s", src_cache);
2068 goto fail_getopt;
2069 }
2070
2071 /* Initialize before goto out */
2072 if (quiet) {
2073 progress = false;
2074 }
2075 qemu_progress_init(progress, 1.0);
2076 qemu_progress_print(0, 100);
2077
2078 s.src = g_new0(BlockBackend *, s.src_num);
2079 s.src_sectors = g_new(int64_t, s.src_num);
2080
2081 for (bs_i = 0; bs_i < s.src_num; bs_i++) {
2082 s.src[bs_i] = img_open(image_opts, argv[optind + bs_i],
2083 fmt, src_flags, src_writethrough, quiet);
2084 if (!s.src[bs_i]) {
2085 ret = -1;
2086 goto out;
2087 }
2088 s.src_sectors[bs_i] = blk_nb_sectors(s.src[bs_i]);
2089 if (s.src_sectors[bs_i] < 0) {
2090 error_report("Could not get size of %s: %s",
2091 argv[optind + bs_i], strerror(-s.src_sectors[bs_i]));
2092 ret = -1;
2093 goto out;
2094 }
2095 s.total_sectors += s.src_sectors[bs_i];
2096 }
2097
2098 if (sn_opts) {
2099 bdrv_snapshot_load_tmp(blk_bs(s.src[0]),
2100 qemu_opt_get(sn_opts, SNAPSHOT_OPT_ID),
2101 qemu_opt_get(sn_opts, SNAPSHOT_OPT_NAME),
2102 &local_err);
2103 } else if (snapshot_name != NULL) {
2104 if (s.src_num > 1) {
2105 error_report("No support for concatenating multiple snapshot");
2106 ret = -1;
2107 goto out;
2108 }
2109
2110 bdrv_snapshot_load_tmp_by_id_or_name(blk_bs(s.src[0]), snapshot_name,
2111 &local_err);
2112 }
2113 if (local_err) {
2114 error_reportf_err(local_err, "Failed to load snapshot: ");
2115 ret = -1;
2116 goto out;
2117 }
2118
2119 /* Find driver and parse its options */
2120 drv = bdrv_find_format(out_fmt);
2121 if (!drv) {
2122 error_report("Unknown file format '%s'", out_fmt);
2123 ret = -1;
2124 goto out;
2125 }
2126
2127 proto_drv = bdrv_find_protocol(out_filename, true, &local_err);
2128 if (!proto_drv) {
2129 error_report_err(local_err);
2130 ret = -1;
2131 goto out;
2132 }
2133
2134 if (!skip_create) {
2135 if (!drv->create_opts) {
2136 error_report("Format driver '%s' does not support image creation",
2137 drv->format_name);
2138 ret = -1;
2139 goto out;
2140 }
2141
2142 if (!proto_drv->create_opts) {
2143 error_report("Protocol driver '%s' does not support image creation",
2144 proto_drv->format_name);
2145 ret = -1;
2146 goto out;
2147 }
2148
2149 create_opts = qemu_opts_append(create_opts, drv->create_opts);
2150 create_opts = qemu_opts_append(create_opts, proto_drv->create_opts);
2151
2152 opts = qemu_opts_create(create_opts, NULL, 0, &error_abort);
2153 if (options) {
2154 qemu_opts_do_parse(opts, options, NULL, &local_err);
2155 if (local_err) {
2156 error_report_err(local_err);
2157 ret = -1;
2158 goto out;
2159 }
2160 }
2161
2162 qemu_opt_set_number(opts, BLOCK_OPT_SIZE, s.total_sectors * 512,
2163 &error_abort);
2164 ret = add_old_style_options(out_fmt, opts, out_baseimg, NULL);
2165 if (ret < 0) {
2166 goto out;
2167 }
2168 }
2169
2170 /* Get backing file name if -o backing_file was used */
2171 out_baseimg_param = qemu_opt_get(opts, BLOCK_OPT_BACKING_FILE);
2172 if (out_baseimg_param) {
2173 out_baseimg = out_baseimg_param;
2174 }
2175 s.target_has_backing = (bool) out_baseimg;
2176
2177 if (s.src_num > 1 && out_baseimg) {
2178 error_report("Having a backing file for the target makes no sense when "
2179 "concatenating multiple input images");
2180 ret = -1;
2181 goto out;
2182 }
2183
2184 /* Check if compression is supported */
2185 if (s.compressed) {
2186 bool encryption =
2187 qemu_opt_get_bool(opts, BLOCK_OPT_ENCRYPT, false);
2188 const char *preallocation =
2189 qemu_opt_get(opts, BLOCK_OPT_PREALLOC);
2190
2191 if (!drv->bdrv_co_pwritev_compressed) {
2192 error_report("Compression not supported for this file format");
2193 ret = -1;
2194 goto out;
2195 }
2196
2197 if (encryption) {
2198 error_report("Compression and encryption not supported at "
2199 "the same time");
2200 ret = -1;
2201 goto out;
2202 }
2203
2204 if (preallocation
2205 && strcmp(preallocation, "off"))
2206 {
2207 error_report("Compression and preallocation not supported at "
2208 "the same time");
2209 ret = -1;
2210 goto out;
2211 }
2212 }
2213
2214 if (!skip_create) {
2215 /* Create the new image */
2216 ret = bdrv_create(drv, out_filename, opts, &local_err);
2217 if (ret < 0) {
2218 error_reportf_err(local_err, "%s: error while converting %s: ",
2219 out_filename, out_fmt);
2220 goto out;
2221 }
2222 }
2223
2224 flags = s.min_sparse ? (BDRV_O_RDWR | BDRV_O_UNMAP) : BDRV_O_RDWR;
2225 ret = bdrv_parse_cache_mode(cache, &flags, &writethrough);
2226 if (ret < 0) {
2227 error_report("Invalid cache option: %s", cache);
2228 goto out;
2229 }
2230
2231 /* XXX we should allow --image-opts to trigger use of
2232 * img_open() here, but then we have trouble with
2233 * the bdrv_create() call which takes different params.
2234 * Not critical right now, so fix can wait...
2235 */
2236 s.target = img_open_file(out_filename, out_fmt, flags, writethrough, quiet);
2237 if (!s.target) {
2238 ret = -1;
2239 goto out;
2240 }
2241 out_bs = blk_bs(s.target);
2242
2243 /* increase bufsectors from the default 4096 (2M) if opt_transfer
2244 * or discard_alignment of the out_bs is greater. Limit to 32768 (16MB)
2245 * as maximum. */
2246 s.buf_sectors = MIN(32768,
2247 MAX(s.buf_sectors,
2248 MAX(out_bs->bl.opt_transfer >> BDRV_SECTOR_BITS,
2249 out_bs->bl.pdiscard_alignment >>
2250 BDRV_SECTOR_BITS)));
2251
2252 if (skip_create) {
2253 int64_t output_sectors = blk_nb_sectors(s.target);
2254 if (output_sectors < 0) {
2255 error_report("unable to get output image length: %s",
2256 strerror(-output_sectors));
2257 ret = -1;
2258 goto out;
2259 } else if (output_sectors < s.total_sectors) {
2260 error_report("output file is smaller than input file");
2261 ret = -1;
2262 goto out;
2263 }
2264 }
2265
2266 ret = bdrv_get_info(out_bs, &bdi);
2267 if (ret < 0) {
2268 if (s.compressed) {
2269 error_report("could not get block driver info");
2270 goto out;
2271 }
2272 } else {
2273 s.compressed = s.compressed || bdi.needs_compressed_writes;
2274 s.cluster_sectors = bdi.cluster_size / BDRV_SECTOR_SIZE;
2275 }
2276
2277 ret = convert_do_copy(&s);
2278 out:
2279 if (!ret) {
2280 qemu_progress_print(100, 0);
2281 }
2282 qemu_progress_end();
2283 qemu_opts_del(opts);
2284 qemu_opts_free(create_opts);
2285 qemu_opts_del(sn_opts);
2286 blk_unref(s.target);
2287 if (s.src) {
2288 for (bs_i = 0; bs_i < s.src_num; bs_i++) {
2289 blk_unref(s.src[bs_i]);
2290 }
2291 g_free(s.src);
2292 }
2293 g_free(s.src_sectors);
2294 fail_getopt:
2295 g_free(options);
2296
2297 return !!ret;
2298 }
2299
2300
2301 static void dump_snapshots(BlockDriverState *bs)
2302 {
2303 QEMUSnapshotInfo *sn_tab, *sn;
2304 int nb_sns, i;
2305
2306 nb_sns = bdrv_snapshot_list(bs, &sn_tab);
2307 if (nb_sns <= 0)
2308 return;
2309 printf("Snapshot list:\n");
2310 bdrv_snapshot_dump(fprintf, stdout, NULL);
2311 printf("\n");
2312 for(i = 0; i < nb_sns; i++) {
2313 sn = &sn_tab[i];
2314 bdrv_snapshot_dump(fprintf, stdout, sn);
2315 printf("\n");
2316 }
2317 g_free(sn_tab);
2318 }
2319
2320 static void dump_json_image_info_list(ImageInfoList *list)
2321 {
2322 QString *str;
2323 QObject *obj;
2324 Visitor *v = qobject_output_visitor_new(&obj);
2325
2326 visit_type_ImageInfoList(v, NULL, &list, &error_abort);
2327 visit_complete(v, &obj);
2328 str = qobject_to_json_pretty(obj);
2329 assert(str != NULL);
2330 printf("%s\n", qstring_get_str(str));
2331 qobject_decref(obj);
2332 visit_free(v);
2333 QDECREF(str);
2334 }
2335
2336 static void dump_json_image_info(ImageInfo *info)
2337 {
2338 QString *str;
2339 QObject *obj;
2340 Visitor *v = qobject_output_visitor_new(&obj);
2341
2342 visit_type_ImageInfo(v, NULL, &info, &error_abort);
2343 visit_complete(v, &obj);
2344 str = qobject_to_json_pretty(obj);
2345 assert(str != NULL);
2346 printf("%s\n", qstring_get_str(str));
2347 qobject_decref(obj);
2348 visit_free(v);
2349 QDECREF(str);
2350 }
2351
2352 static void dump_human_image_info_list(ImageInfoList *list)
2353 {
2354 ImageInfoList *elem;
2355 bool delim = false;
2356
2357 for (elem = list; elem; elem = elem->next) {
2358 if (delim) {
2359 printf("\n");
2360 }
2361 delim = true;
2362
2363 bdrv_image_info_dump(fprintf, stdout, elem->value);
2364 }
2365 }
2366
2367 static gboolean str_equal_func(gconstpointer a, gconstpointer b)
2368 {
2369 return strcmp(a, b) == 0;
2370 }
2371
2372 /**
2373 * Open an image file chain and return an ImageInfoList
2374 *
2375 * @filename: topmost image filename
2376 * @fmt: topmost image format (may be NULL to autodetect)
2377 * @chain: true - enumerate entire backing file chain
2378 * false - only topmost image file
2379 *
2380 * Returns a list of ImageInfo objects or NULL if there was an error opening an
2381 * image file. If there was an error a message will have been printed to
2382 * stderr.
2383 */
2384 static ImageInfoList *collect_image_info_list(bool image_opts,
2385 const char *filename,
2386 const char *fmt,
2387 bool chain)
2388 {
2389 ImageInfoList *head = NULL;
2390 ImageInfoList **last = &head;
2391 GHashTable *filenames;
2392 Error *err = NULL;
2393
2394 filenames = g_hash_table_new_full(g_str_hash, str_equal_func, NULL, NULL);
2395
2396 while (filename) {
2397 BlockBackend *blk;
2398 BlockDriverState *bs;
2399 ImageInfo *info;
2400 ImageInfoList *elem;
2401
2402 if (g_hash_table_lookup_extended(filenames, filename, NULL, NULL)) {
2403 error_report("Backing file '%s' creates an infinite loop.",
2404 filename);
2405 goto err;
2406 }
2407 g_hash_table_insert(filenames, (gpointer)filename, NULL);
2408
2409 blk = img_open(image_opts, filename, fmt,
2410 BDRV_O_NO_BACKING | BDRV_O_NO_IO, false, false);
2411 if (!blk) {
2412 goto err;
2413 }
2414 bs = blk_bs(blk);
2415
2416 bdrv_query_image_info(bs, &info, &err);
2417 if (err) {
2418 error_report_err(err);
2419 blk_unref(blk);
2420 goto err;
2421 }
2422
2423 elem = g_new0(ImageInfoList, 1);
2424 elem->value = info;
2425 *last = elem;
2426 last = &elem->next;
2427
2428 blk_unref(blk);
2429
2430 filename = fmt = NULL;
2431 if (chain) {
2432 if (info->has_full_backing_filename) {
2433 filename = info->full_backing_filename;
2434 } else if (info->has_backing_filename) {
2435 error_report("Could not determine absolute backing filename,"
2436 " but backing filename '%s' present",
2437 info->backing_filename);
2438 goto err;
2439 }
2440 if (info->has_backing_filename_format) {
2441 fmt = info->backing_filename_format;
2442 }
2443 }
2444 }
2445 g_hash_table_destroy(filenames);
2446 return head;
2447
2448 err:
2449 qapi_free_ImageInfoList(head);
2450 g_hash_table_destroy(filenames);
2451 return NULL;
2452 }
2453
2454 static int img_info(int argc, char **argv)
2455 {
2456 int c;
2457 OutputFormat output_format = OFORMAT_HUMAN;
2458 bool chain = false;
2459 const char *filename, *fmt, *output;
2460 ImageInfoList *list;
2461 bool image_opts = false;
2462
2463 fmt = NULL;
2464 output = NULL;
2465 for(;;) {
2466 int option_index = 0;
2467 static const struct option long_options[] = {
2468 {"help", no_argument, 0, 'h'},
2469 {"format", required_argument, 0, 'f'},
2470 {"output", required_argument, 0, OPTION_OUTPUT},
2471 {"backing-chain", no_argument, 0, OPTION_BACKING_CHAIN},
2472 {"object", required_argument, 0, OPTION_OBJECT},
2473 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
2474 {0, 0, 0, 0}
2475 };
2476 c = getopt_long(argc, argv, ":f:h",
2477 long_options, &option_index);
2478 if (c == -1) {
2479 break;
2480 }
2481 switch(c) {
2482 case ':':
2483 missing_argument(argv[optind - 1]);
2484 break;
2485 case '?':
2486 unrecognized_option(argv[optind - 1]);
2487 break;
2488 case 'h':
2489 help();
2490 break;
2491 case 'f':
2492 fmt = optarg;
2493 break;
2494 case OPTION_OUTPUT:
2495 output = optarg;
2496 break;
2497 case OPTION_BACKING_CHAIN:
2498 chain = true;
2499 break;
2500 case OPTION_OBJECT: {
2501 QemuOpts *opts;
2502 opts = qemu_opts_parse_noisily(&qemu_object_opts,
2503 optarg, true);
2504 if (!opts) {
2505 return 1;
2506 }
2507 } break;
2508 case OPTION_IMAGE_OPTS:
2509 image_opts = true;
2510 break;
2511 }
2512 }
2513 if (optind != argc - 1) {
2514 error_exit("Expecting one image file name");
2515 }
2516 filename = argv[optind++];
2517
2518 if (output && !strcmp(output, "json")) {
2519 output_format = OFORMAT_JSON;
2520 } else if (output && !strcmp(output, "human")) {
2521 output_format = OFORMAT_HUMAN;
2522 } else if (output) {
2523 error_report("--output must be used with human or json as argument.");
2524 return 1;
2525 }
2526
2527 if (qemu_opts_foreach(&qemu_object_opts,
2528 user_creatable_add_opts_foreach,
2529 NULL, NULL)) {
2530 return 1;
2531 }
2532
2533 list = collect_image_info_list(image_opts, filename, fmt, chain);
2534 if (!list) {
2535 return 1;
2536 }
2537
2538 switch (output_format) {
2539 case OFORMAT_HUMAN:
2540 dump_human_image_info_list(list);
2541 break;
2542 case OFORMAT_JSON:
2543 if (chain) {
2544 dump_json_image_info_list(list);
2545 } else {
2546 dump_json_image_info(list->value);
2547 }
2548 break;
2549 }
2550
2551 qapi_free_ImageInfoList(list);
2552 return 0;
2553 }
2554
2555 static void dump_map_entry(OutputFormat output_format, MapEntry *e,
2556 MapEntry *next)
2557 {
2558 switch (output_format) {
2559 case OFORMAT_HUMAN:
2560 if (e->data && !e->has_offset) {
2561 error_report("File contains external, encrypted or compressed clusters.");
2562 exit(1);
2563 }
2564 if (e->data && !e->zero) {
2565 printf("%#-16"PRIx64"%#-16"PRIx64"%#-16"PRIx64"%s\n",
2566 e->start, e->length,
2567 e->has_offset ? e->offset : 0,
2568 e->has_filename ? e->filename : "");
2569 }
2570 /* This format ignores the distinction between 0, ZERO and ZERO|DATA.
2571 * Modify the flags here to allow more coalescing.
2572 */
2573 if (next && (!next->data || next->zero)) {
2574 next->data = false;
2575 next->zero = true;
2576 }
2577 break;
2578 case OFORMAT_JSON:
2579 printf("%s{ \"start\": %"PRId64", \"length\": %"PRId64","
2580 " \"depth\": %"PRId64", \"zero\": %s, \"data\": %s",
2581 (e->start == 0 ? "[" : ",\n"),
2582 e->start, e->length, e->depth,
2583 e->zero ? "true" : "false",
2584 e->data ? "true" : "false");
2585 if (e->has_offset) {
2586 printf(", \"offset\": %"PRId64"", e->offset);
2587 }
2588 putchar('}');
2589
2590 if (!next) {
2591 printf("]\n");
2592 }
2593 break;
2594 }
2595 }
2596
2597 static int get_block_status(BlockDriverState *bs, int64_t sector_num,
2598 int nb_sectors, MapEntry *e)
2599 {
2600 int64_t ret;
2601 int depth;
2602 BlockDriverState *file;
2603 bool has_offset;
2604
2605 /* As an optimization, we could cache the current range of unallocated
2606 * clusters in each file of the chain, and avoid querying the same
2607 * range repeatedly.
2608 */
2609
2610 depth = 0;
2611 for (;;) {
2612 ret = bdrv_get_block_status(bs, sector_num, nb_sectors, &nb_sectors,
2613 &file);
2614 if (ret < 0) {
2615 return ret;
2616 }
2617 assert(nb_sectors);
2618 if (ret & (BDRV_BLOCK_ZERO|BDRV_BLOCK_DATA)) {
2619 break;
2620 }
2621 bs = backing_bs(bs);
2622 if (bs == NULL) {
2623 ret = 0;
2624 break;
2625 }
2626
2627 depth++;
2628 }
2629
2630 has_offset = !!(ret & BDRV_BLOCK_OFFSET_VALID);
2631
2632 *e = (MapEntry) {
2633 .start = sector_num * BDRV_SECTOR_SIZE,
2634 .length = nb_sectors * BDRV_SECTOR_SIZE,
2635 .data = !!(ret & BDRV_BLOCK_DATA),
2636 .zero = !!(ret & BDRV_BLOCK_ZERO),
2637 .offset = ret & BDRV_BLOCK_OFFSET_MASK,
2638 .has_offset = has_offset,
2639 .depth = depth,
2640 .has_filename = file && has_offset,
2641 .filename = file && has_offset ? file->filename : NULL,
2642 };
2643
2644 return 0;
2645 }
2646
2647 static inline bool entry_mergeable(const MapEntry *curr, const MapEntry *next)
2648 {
2649 if (curr->length == 0) {
2650 return false;
2651 }
2652 if (curr->zero != next->zero ||
2653 curr->data != next->data ||
2654 curr->depth != next->depth ||
2655 curr->has_filename != next->has_filename ||
2656 curr->has_offset != next->has_offset) {
2657 return false;
2658 }
2659 if (curr->has_filename && strcmp(curr->filename, next->filename)) {
2660 return false;
2661 }
2662 if (curr->has_offset && curr->offset + curr->length != next->offset) {
2663 return false;
2664 }
2665 return true;
2666 }
2667
2668 static int img_map(int argc, char **argv)
2669 {
2670 int c;
2671 OutputFormat output_format = OFORMAT_HUMAN;
2672 BlockBackend *blk;
2673 BlockDriverState *bs;
2674 const char *filename, *fmt, *output;
2675 int64_t length;
2676 MapEntry curr = { .length = 0 }, next;
2677 int ret = 0;
2678 bool image_opts = false;
2679
2680 fmt = NULL;
2681 output = NULL;
2682 for (;;) {
2683 int option_index = 0;
2684 static const struct option long_options[] = {
2685 {"help", no_argument, 0, 'h'},
2686 {"format", required_argument, 0, 'f'},
2687 {"output", required_argument, 0, OPTION_OUTPUT},
2688 {"object", required_argument, 0, OPTION_OBJECT},
2689 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
2690 {0, 0, 0, 0}
2691 };
2692 c = getopt_long(argc, argv, ":f:h",
2693 long_options, &option_index);
2694 if (c == -1) {
2695 break;
2696 }
2697 switch (c) {
2698 case ':':
2699 missing_argument(argv[optind - 1]);
2700 break;
2701 case '?':
2702 unrecognized_option(argv[optind - 1]);
2703 break;
2704 case 'h':
2705 help();
2706 break;
2707 case 'f':
2708 fmt = optarg;
2709 break;
2710 case OPTION_OUTPUT:
2711 output = optarg;
2712 break;
2713 case OPTION_OBJECT: {
2714 QemuOpts *opts;
2715 opts = qemu_opts_parse_noisily(&qemu_object_opts,
2716 optarg, true);
2717 if (!opts) {
2718 return 1;
2719 }
2720 } break;
2721 case OPTION_IMAGE_OPTS:
2722 image_opts = true;
2723 break;
2724 }
2725 }
2726 if (optind != argc - 1) {
2727 error_exit("Expecting one image file name");
2728 }
2729 filename = argv[optind];
2730
2731 if (output && !strcmp(output, "json")) {
2732 output_format = OFORMAT_JSON;
2733 } else if (output && !strcmp(output, "human")) {
2734 output_format = OFORMAT_HUMAN;
2735 } else if (output) {
2736 error_report("--output must be used with human or json as argument.");
2737 return 1;
2738 }
2739
2740 if (qemu_opts_foreach(&qemu_object_opts,
2741 user_creatable_add_opts_foreach,
2742 NULL, NULL)) {
2743 return 1;
2744 }
2745
2746 blk = img_open(image_opts, filename, fmt, 0, false, false);
2747 if (!blk) {
2748 return 1;
2749 }
2750 bs = blk_bs(blk);
2751
2752 if (output_format == OFORMAT_HUMAN) {
2753 printf("%-16s%-16s%-16s%s\n", "Offset", "Length", "Mapped to", "File");
2754 }
2755
2756 length = blk_getlength(blk);
2757 while (curr.start + curr.length < length) {
2758 int64_t nsectors_left;
2759 int64_t sector_num;
2760 int n;
2761
2762 sector_num = (curr.start + curr.length) >> BDRV_SECTOR_BITS;
2763
2764 /* Probe up to 1 GiB at a time. */
2765 nsectors_left = DIV_ROUND_UP(length, BDRV_SECTOR_SIZE) - sector_num;
2766 n = MIN(1 << (30 - BDRV_SECTOR_BITS), nsectors_left);
2767 ret = get_block_status(bs, sector_num, n, &next);
2768
2769 if (ret < 0) {
2770 error_report("Could not read file metadata: %s", strerror(-ret));
2771 goto out;
2772 }
2773
2774 if (entry_mergeable(&curr, &next)) {
2775 curr.length += next.length;
2776 continue;
2777 }
2778
2779 if (curr.length > 0) {
2780 dump_map_entry(output_format, &curr, &next);
2781 }
2782 curr = next;
2783 }
2784
2785 dump_map_entry(output_format, &curr, NULL);
2786
2787 out:
2788 blk_unref(blk);
2789 return ret < 0;
2790 }
2791
2792 #define SNAPSHOT_LIST 1
2793 #define SNAPSHOT_CREATE 2
2794 #define SNAPSHOT_APPLY 3
2795 #define SNAPSHOT_DELETE 4
2796
2797 static int img_snapshot(int argc, char **argv)
2798 {
2799 BlockBackend *blk;
2800 BlockDriverState *bs;
2801 QEMUSnapshotInfo sn;
2802 char *filename, *snapshot_name = NULL;
2803 int c, ret = 0, bdrv_oflags;
2804 int action = 0;
2805 qemu_timeval tv;
2806 bool quiet = false;
2807 Error *err = NULL;
2808 bool image_opts = false;
2809
2810 bdrv_oflags = BDRV_O_RDWR;
2811 /* Parse commandline parameters */
2812 for(;;) {
2813 static const struct option long_options[] = {
2814 {"help", no_argument, 0, 'h'},
2815 {"object", required_argument, 0, OPTION_OBJECT},
2816 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
2817 {0, 0, 0, 0}
2818 };
2819 c = getopt_long(argc, argv, ":la:c:d:hq",
2820 long_options, NULL);
2821 if (c == -1) {
2822 break;
2823 }
2824 switch(c) {
2825 case ':':
2826 missing_argument(argv[optind - 1]);
2827 break;
2828 case '?':
2829 unrecognized_option(argv[optind - 1]);
2830 break;
2831 case 'h':
2832 help();
2833 return 0;
2834 case 'l':
2835 if (action) {
2836 error_exit("Cannot mix '-l', '-a', '-c', '-d'");
2837 return 0;
2838 }
2839 action = SNAPSHOT_LIST;
2840 bdrv_oflags &= ~BDRV_O_RDWR; /* no need for RW */
2841 break;
2842 case 'a':
2843 if (action) {
2844 error_exit("Cannot mix '-l', '-a', '-c', '-d'");
2845 return 0;
2846 }
2847 action = SNAPSHOT_APPLY;
2848 snapshot_name = optarg;
2849 break;
2850 case 'c':
2851 if (action) {
2852 error_exit("Cannot mix '-l', '-a', '-c', '-d'");
2853 return 0;
2854 }
2855 action = SNAPSHOT_CREATE;
2856 snapshot_name = optarg;
2857 break;
2858 case 'd':
2859 if (action) {
2860 error_exit("Cannot mix '-l', '-a', '-c', '-d'");
2861 return 0;
2862 }
2863 action = SNAPSHOT_DELETE;
2864 snapshot_name = optarg;
2865 break;
2866 case 'q':
2867 quiet = true;
2868 break;
2869 case OPTION_OBJECT: {
2870 QemuOpts *opts;
2871 opts = qemu_opts_parse_noisily(&qemu_object_opts,
2872 optarg, true);
2873 if (!opts) {
2874 return 1;
2875 }
2876 } break;
2877 case OPTION_IMAGE_OPTS:
2878 image_opts = true;
2879 break;
2880 }
2881 }
2882
2883 if (optind != argc - 1) {
2884 error_exit("Expecting one image file name");
2885 }
2886 filename = argv[optind++];
2887
2888 if (qemu_opts_foreach(&qemu_object_opts,
2889 user_creatable_add_opts_foreach,
2890 NULL, NULL)) {
2891 return 1;
2892 }
2893
2894 /* Open the image */
2895 blk = img_open(image_opts, filename, NULL, bdrv_oflags, false, quiet);
2896 if (!blk) {
2897 return 1;
2898 }
2899 bs = blk_bs(blk);
2900
2901 /* Perform the requested action */
2902 switch(action) {
2903 case SNAPSHOT_LIST:
2904 dump_snapshots(bs);
2905 break;
2906
2907 case SNAPSHOT_CREATE:
2908 memset(&sn, 0, sizeof(sn));
2909 pstrcpy(sn.name, sizeof(sn.name), snapshot_name);
2910
2911 qemu_gettimeofday(&tv);
2912 sn.date_sec = tv.tv_sec;
2913 sn.date_nsec = tv.tv_usec * 1000;
2914
2915 ret = bdrv_snapshot_create(bs, &sn);
2916 if (ret) {
2917 error_report("Could not create snapshot '%s': %d (%s)",
2918 snapshot_name, ret, strerror(-ret));
2919 }
2920 break;
2921
2922 case SNAPSHOT_APPLY:
2923 ret = bdrv_snapshot_goto(bs, snapshot_name);
2924 if (ret) {
2925 error_report("Could not apply snapshot '%s': %d (%s)",
2926 snapshot_name, ret, strerror(-ret));
2927 }
2928 break;
2929
2930 case SNAPSHOT_DELETE:
2931 bdrv_snapshot_delete_by_id_or_name(bs, snapshot_name, &err);
2932 if (err) {
2933 error_reportf_err(err, "Could not delete snapshot '%s': ",
2934 snapshot_name);
2935 ret = 1;
2936 }
2937 break;
2938 }
2939
2940 /* Cleanup */
2941 blk_unref(blk);
2942 if (ret) {
2943 return 1;
2944 }
2945 return 0;
2946 }
2947
2948 static int img_rebase(int argc, char **argv)
2949 {
2950 BlockBackend *blk = NULL, *blk_old_backing = NULL, *blk_new_backing = NULL;
2951 uint8_t *buf_old = NULL;
2952 uint8_t *buf_new = NULL;
2953 BlockDriverState *bs = NULL;
2954 char *filename;
2955 const char *fmt, *cache, *src_cache, *out_basefmt, *out_baseimg;
2956 int c, flags, src_flags, ret;
2957 bool writethrough, src_writethrough;
2958 int unsafe = 0;
2959 int progress = 0;
2960 bool quiet = false;
2961 Error *local_err = NULL;
2962 bool image_opts = false;
2963
2964 /* Parse commandline parameters */
2965 fmt = NULL;
2966 cache = BDRV_DEFAULT_CACHE;
2967 src_cache = BDRV_DEFAULT_CACHE;
2968 out_baseimg = NULL;
2969 out_basefmt = NULL;
2970 for(;;) {
2971 static const struct option long_options[] = {
2972 {"help", no_argument, 0, 'h'},
2973 {"object", required_argument, 0, OPTION_OBJECT},
2974 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
2975 {0, 0, 0, 0}
2976 };
2977 c = getopt_long(argc, argv, ":hf:F:b:upt:T:q",
2978 long_options, NULL);
2979 if (c == -1) {
2980 break;
2981 }
2982 switch(c) {
2983 case ':':
2984 missing_argument(argv[optind - 1]);
2985 break;
2986 case '?':
2987 unrecognized_option(argv[optind - 1]);
2988 break;
2989 case 'h':
2990 help();
2991 return 0;
2992 case 'f':
2993 fmt = optarg;
2994 break;
2995 case 'F':
2996 out_basefmt = optarg;
2997 break;
2998 case 'b':
2999 out_baseimg = optarg;
3000 break;
3001 case 'u':
3002 unsafe = 1;
3003 break;
3004 case 'p':
3005 progress = 1;
3006 break;
3007 case 't':
3008 cache = optarg;
3009 break;
3010 case 'T':
3011 src_cache = optarg;
3012 break;
3013 case 'q':
3014 quiet = true;
3015 break;
3016 case OPTION_OBJECT: {
3017 QemuOpts *opts;
3018 opts = qemu_opts_parse_noisily(&qemu_object_opts,
3019 optarg, true);
3020 if (!opts) {
3021 return 1;
3022 }
3023 } break;
3024 case OPTION_IMAGE_OPTS:
3025 image_opts = true;
3026 break;
3027 }
3028 }
3029
3030 if (quiet) {
3031 progress = 0;
3032 }
3033
3034 if (optind != argc - 1) {
3035 error_exit("Expecting one image file name");
3036 }
3037 if (!unsafe && !out_baseimg) {
3038 error_exit("Must specify backing file (-b) or use unsafe mode (-u)");
3039 }
3040 filename = argv[optind++];
3041
3042 if (qemu_opts_foreach(&qemu_object_opts,
3043 user_creatable_add_opts_foreach,
3044 NULL, NULL)) {
3045 return 1;
3046 }
3047
3048 qemu_progress_init(progress, 2.0);
3049 qemu_progress_print(0, 100);
3050
3051 flags = BDRV_O_RDWR | (unsafe ? BDRV_O_NO_BACKING : 0);
3052 ret = bdrv_parse_cache_mode(cache, &flags, &writethrough);
3053 if (ret < 0) {
3054 error_report("Invalid cache option: %s", cache);
3055 goto out;
3056 }
3057
3058 src_flags = 0;
3059 ret = bdrv_parse_cache_mode(src_cache, &src_flags, &src_writethrough);
3060 if (ret < 0) {
3061 error_report("Invalid source cache option: %s", src_cache);
3062 goto out;
3063 }
3064
3065 /* The source files are opened read-only, don't care about WCE */
3066 assert((src_flags & BDRV_O_RDWR) == 0);
3067 (void) src_writethrough;
3068
3069 /*
3070 * Open the images.
3071 *
3072 * Ignore the old backing file for unsafe rebase in case we want to correct
3073 * the reference to a renamed or moved backing file.
3074 */
3075 blk = img_open(image_opts, filename, fmt, flags, writethrough, quiet);
3076 if (!blk) {
3077 ret = -1;
3078 goto out;
3079 }
3080 bs = blk_bs(blk);
3081
3082 if (out_basefmt != NULL) {
3083 if (bdrv_find_format(out_basefmt) == NULL) {
3084 error_report("Invalid format name: '%s'", out_basefmt);
3085 ret = -1;
3086 goto out;
3087 }
3088 }
3089
3090 /* For safe rebasing we need to compare old and new backing file */
3091 if (!unsafe) {
3092 char backing_name[PATH_MAX];
3093 QDict *options = NULL;
3094
3095 if (bs->backing_format[0] != '\0') {
3096 options = qdict_new();
3097 qdict_put(options, "driver", qstring_from_str(bs->backing_format));
3098 }
3099
3100 bdrv_get_backing_filename(bs, backing_name, sizeof(backing_name));
3101 blk_old_backing = blk_new_open(backing_name, NULL,
3102 options, src_flags, &local_err);
3103 if (!blk_old_backing) {
3104 error_reportf_err(local_err,
3105 "Could not open old backing file '%s': ",
3106 backing_name);
3107 ret = -1;
3108 goto out;
3109 }
3110
3111 if (out_baseimg[0]) {
3112 if (out_basefmt) {
3113 options = qdict_new();
3114 qdict_put(options, "driver", qstring_from_str(out_basefmt));
3115 } else {
3116 options = NULL;
3117 }
3118
3119 blk_new_backing = blk_new_open(out_baseimg, NULL,
3120 options, src_flags, &local_err);
3121 if (!blk_new_backing) {
3122 error_reportf_err(local_err,
3123 "Could not open new backing file '%s': ",
3124 out_baseimg);
3125 ret = -1;
3126 goto out;
3127 }
3128 }
3129 }
3130
3131 /*
3132 * Check each unallocated cluster in the COW file. If it is unallocated,
3133 * accesses go to the backing file. We must therefore compare this cluster
3134 * in the old and new backing file, and if they differ we need to copy it
3135 * from the old backing file into the COW file.
3136 *
3137 * If qemu-img crashes during this step, no harm is done. The content of
3138 * the image is the same as the original one at any time.
3139 */
3140 if (!unsafe) {
3141 int64_t num_sectors;
3142 int64_t old_backing_num_sectors;
3143 int64_t new_backing_num_sectors = 0;
3144 uint64_t sector;
3145 int n;
3146 float local_progress = 0;
3147
3148 buf_old = blk_blockalign(blk, IO_BUF_SIZE);
3149 buf_new = blk_blockalign(blk, IO_BUF_SIZE);
3150
3151 num_sectors = blk_nb_sectors(blk);
3152 if (num_sectors < 0) {
3153 error_report("Could not get size of '%s': %s",
3154 filename, strerror(-num_sectors));
3155 ret = -1;
3156 goto out;
3157 }
3158 old_backing_num_sectors = blk_nb_sectors(blk_old_backing);
3159 if (old_backing_num_sectors < 0) {
3160 char backing_name[PATH_MAX];
3161
3162 bdrv_get_backing_filename(bs, backing_name, sizeof(backing_name));
3163 error_report("Could not get size of '%s': %s",
3164 backing_name, strerror(-old_backing_num_sectors));
3165 ret = -1;
3166 goto out;
3167 }
3168 if (blk_new_backing) {
3169 new_backing_num_sectors = blk_nb_sectors(blk_new_backing);
3170 if (new_backing_num_sectors < 0) {
3171 error_report("Could not get size of '%s': %s",
3172 out_baseimg, strerror(-new_backing_num_sectors));
3173 ret = -1;
3174 goto out;
3175 }
3176 }
3177
3178 if (num_sectors != 0) {
3179 local_progress = (float)100 /
3180 (num_sectors / MIN(num_sectors, IO_BUF_SIZE / 512));
3181 }
3182
3183 for (sector = 0; sector < num_sectors; sector += n) {
3184
3185 /* How many sectors can we handle with the next read? */
3186 if (sector + (IO_BUF_SIZE / 512) <= num_sectors) {
3187 n = (IO_BUF_SIZE / 512);
3188 } else {
3189 n = num_sectors - sector;
3190 }
3191
3192 /* If the cluster is allocated, we don't need to take action */
3193 ret = bdrv_is_allocated(bs, sector, n, &n);
3194 if (ret < 0) {
3195 error_report("error while reading image metadata: %s",
3196 strerror(-ret));
3197 goto out;
3198 }
3199 if (ret) {
3200 continue;
3201 }
3202
3203 /*
3204 * Read old and new backing file and take into consideration that
3205 * backing files may be smaller than the COW image.
3206 */
3207 if (sector >= old_backing_num_sectors) {
3208 memset(buf_old, 0, n * BDRV_SECTOR_SIZE);
3209 } else {
3210 if (sector + n > old_backing_num_sectors) {
3211 n = old_backing_num_sectors - sector;
3212 }
3213
3214 ret = blk_pread(blk_old_backing, sector << BDRV_SECTOR_BITS,
3215 buf_old, n << BDRV_SECTOR_BITS);
3216 if (ret < 0) {
3217 error_report("error while reading from old backing file");
3218 goto out;
3219 }
3220 }
3221
3222 if (sector >= new_backing_num_sectors || !blk_new_backing) {
3223 memset(buf_new, 0, n * BDRV_SECTOR_SIZE);
3224 } else {
3225 if (sector + n > new_backing_num_sectors) {
3226 n = new_backing_num_sectors - sector;
3227 }
3228
3229 ret = blk_pread(blk_new_backing, sector << BDRV_SECTOR_BITS,
3230 buf_new, n << BDRV_SECTOR_BITS);
3231 if (ret < 0) {
3232 error_report("error while reading from new backing file");
3233 goto out;
3234 }
3235 }
3236
3237 /* If they differ, we need to write to the COW file */
3238 uint64_t written = 0;
3239
3240 while (written < n) {
3241 int pnum;
3242
3243 if (compare_sectors(buf_old + written * 512,
3244 buf_new + written * 512, n - written, &pnum))
3245 {
3246 ret = blk_pwrite(blk,
3247 (sector + written) << BDRV_SECTOR_BITS,
3248 buf_old + written * 512,
3249 pnum << BDRV_SECTOR_BITS, 0);
3250 if (ret < 0) {
3251 error_report("Error while writing to COW image: %s",
3252 strerror(-ret));
3253 goto out;
3254 }
3255 }
3256
3257 written += pnum;
3258 }
3259 qemu_progress_print(local_progress, 100);
3260 }
3261 }
3262
3263 /*
3264 * Change the backing file. All clusters that are different from the old
3265 * backing file are overwritten in the COW file now, so the visible content
3266 * doesn't change when we switch the backing file.
3267 */
3268 if (out_baseimg && *out_baseimg) {
3269 ret = bdrv_change_backing_file(bs, out_baseimg, out_basefmt);
3270 } else {
3271 ret = bdrv_change_backing_file(bs, NULL, NULL);
3272 }
3273
3274 if (ret == -ENOSPC) {
3275 error_report("Could not change the backing file to '%s': No "
3276 "space left in the file header", out_baseimg);
3277 } else if (ret < 0) {
3278 error_report("Could not change the backing file to '%s': %s",
3279 out_baseimg, strerror(-ret));
3280 }
3281
3282 qemu_progress_print(100, 0);
3283 /*
3284 * TODO At this point it is possible to check if any clusters that are
3285 * allocated in the COW file are the same in the backing file. If so, they
3286 * could be dropped from the COW file. Don't do this before switching the
3287 * backing file, in case of a crash this would lead to corruption.
3288 */
3289 out:
3290 qemu_progress_end();
3291 /* Cleanup */
3292 if (!unsafe) {
3293 blk_unref(blk_old_backing);
3294 blk_unref(blk_new_backing);
3295 }
3296 qemu_vfree(buf_old);
3297 qemu_vfree(buf_new);
3298
3299 blk_unref(blk);
3300 if (ret) {
3301 return 1;
3302 }
3303 return 0;
3304 }
3305
3306 static int img_resize(int argc, char **argv)
3307 {
3308 Error *err = NULL;
3309 int c, ret, relative;
3310 const char *filename, *fmt, *size;
3311 int64_t n, total_size;
3312 bool quiet = false;
3313 BlockBackend *blk = NULL;
3314 QemuOpts *param;
3315
3316 static QemuOptsList resize_options = {
3317 .name = "resize_options",
3318 .head = QTAILQ_HEAD_INITIALIZER(resize_options.head),
3319 .desc = {
3320 {
3321 .name = BLOCK_OPT_SIZE,
3322 .type = QEMU_OPT_SIZE,
3323 .help = "Virtual disk size"
3324 }, {
3325 /* end of list */
3326 }
3327 },
3328 };
3329 bool image_opts = false;
3330
3331 /* Remove size from argv manually so that negative numbers are not treated
3332 * as options by getopt. */
3333 if (argc < 3) {
3334 error_exit("Not enough arguments");
3335 return 1;
3336 }
3337
3338 size = argv[--argc];
3339
3340 /* Parse getopt arguments */
3341 fmt = NULL;
3342 for(;;) {
3343 static const struct option long_options[] = {
3344 {"help", no_argument, 0, 'h'},
3345 {"object", required_argument, 0, OPTION_OBJECT},
3346 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
3347 {0, 0, 0, 0}
3348 };
3349 c = getopt_long(argc, argv, ":f:hq",
3350 long_options, NULL);
3351 if (c == -1) {
3352 break;
3353 }
3354 switch(c) {
3355 case ':':
3356 missing_argument(argv[optind - 1]);
3357 break;
3358 case '?':
3359 unrecognized_option(argv[optind - 1]);
3360 break;
3361 case 'h':
3362 help();
3363 break;
3364 case 'f':
3365 fmt = optarg;
3366 break;
3367 case 'q':
3368 quiet = true;
3369 break;
3370 case OPTION_OBJECT: {
3371 QemuOpts *opts;
3372 opts = qemu_opts_parse_noisily(&qemu_object_opts,
3373 optarg, true);
3374 if (!opts) {
3375 return 1;
3376 }
3377 } break;
3378 case OPTION_IMAGE_OPTS:
3379 image_opts = true;
3380 break;
3381 }
3382 }
3383 if (optind != argc - 1) {
3384 error_exit("Expecting one image file name");
3385 }
3386 filename = argv[optind++];
3387
3388 if (qemu_opts_foreach(&qemu_object_opts,
3389 user_creatable_add_opts_foreach,
3390 NULL, NULL)) {
3391 return 1;
3392 }
3393
3394 /* Choose grow, shrink, or absolute resize mode */
3395 switch (size[0]) {
3396 case '+':
3397 relative = 1;
3398 size++;
3399 break;
3400 case '-':
3401 relative = -1;
3402 size++;
3403 break;
3404 default:
3405 relative = 0;
3406 break;
3407 }
3408
3409 /* Parse size */
3410 param = qemu_opts_create(&resize_options, NULL, 0, &error_abort);
3411 qemu_opt_set(param, BLOCK_OPT_SIZE, size, &err);
3412 if (err) {
3413 error_report_err(err);
3414 ret = -1;
3415 qemu_opts_del(param);
3416 goto out;
3417 }
3418 n = qemu_opt_get_size(param, BLOCK_OPT_SIZE, 0);
3419 qemu_opts_del(param);
3420
3421 blk = img_open(image_opts, filename, fmt,
3422 BDRV_O_RDWR | BDRV_O_RESIZE, false, quiet);
3423 if (!blk) {
3424 ret = -1;
3425 goto out;
3426 }
3427
3428 if (relative) {
3429 total_size = blk_getlength(blk) + n * relative;
3430 } else {
3431 total_size = n;
3432 }
3433 if (total_size <= 0) {
3434 error_report("New image size must be positive");
3435 ret = -1;
3436 goto out;
3437 }
3438
3439 ret = blk_truncate(blk, total_size, &err);
3440 if (!ret) {
3441 qprintf(quiet, "Image resized.\n");
3442 } else {
3443 error_report_err(err);
3444 }
3445 out:
3446 blk_unref(blk);
3447 if (ret) {
3448 return 1;
3449 }
3450 return 0;
3451 }
3452
3453 static void amend_status_cb(BlockDriverState *bs,
3454 int64_t offset, int64_t total_work_size,
3455 void *opaque)
3456 {
3457 qemu_progress_print(100.f * offset / total_work_size, 0);
3458 }
3459
3460 static int img_amend(int argc, char **argv)
3461 {
3462 Error *err = NULL;
3463 int c, ret = 0;
3464 char *options = NULL;
3465 QemuOptsList *create_opts = NULL;
3466 QemuOpts *opts = NULL;
3467 const char *fmt = NULL, *filename, *cache;
3468 int flags;
3469 bool writethrough;
3470 bool quiet = false, progress = false;
3471 BlockBackend *blk = NULL;
3472 BlockDriverState *bs = NULL;
3473 bool image_opts = false;
3474
3475 cache = BDRV_DEFAULT_CACHE;
3476 for (;;) {
3477 static const struct option long_options[] = {
3478 {"help", no_argument, 0, 'h'},
3479 {"object", required_argument, 0, OPTION_OBJECT},
3480 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
3481 {0, 0, 0, 0}
3482 };
3483 c = getopt_long(argc, argv, ":ho:f:t:pq",
3484 long_options, NULL);
3485 if (c == -1) {
3486 break;
3487 }
3488
3489 switch (c) {
3490 case ':':
3491 missing_argument(argv[optind - 1]);
3492 break;
3493 case '?':
3494 unrecognized_option(argv[optind - 1]);
3495 break;
3496 case 'h':
3497 help();
3498 break;
3499 case 'o':
3500 if (!is_valid_option_list(optarg)) {
3501 error_report("Invalid option list: %s", optarg);
3502 ret = -1;
3503 goto out_no_progress;
3504 }
3505 if (!options) {
3506 options = g_strdup(optarg);
3507 } else {
3508 char *old_options = options;
3509 options = g_strdup_printf("%s,%s", options, optarg);
3510 g_free(old_options);
3511 }
3512 break;
3513 case 'f':
3514 fmt = optarg;
3515 break;
3516 case 't':
3517 cache = optarg;
3518 break;
3519 case 'p':
3520 progress = true;
3521 break;
3522 case 'q':
3523 quiet = true;
3524 break;
3525 case OPTION_OBJECT:
3526 opts = qemu_opts_parse_noisily(&qemu_object_opts,
3527 optarg, true);
3528 if (!opts) {
3529 ret = -1;
3530 goto out_no_progress;
3531 }
3532 break;
3533 case OPTION_IMAGE_OPTS:
3534 image_opts = true;
3535 break;
3536 }
3537 }
3538
3539 if (!options) {
3540 error_exit("Must specify options (-o)");
3541 }
3542
3543 if (qemu_opts_foreach(&qemu_object_opts,
3544 user_creatable_add_opts_foreach,
3545 NULL, NULL)) {
3546 ret = -1;
3547 goto out_no_progress;
3548 }
3549
3550 if (quiet) {
3551 progress = false;
3552 }
3553 qemu_progress_init(progress, 1.0);
3554
3555 filename = (optind == argc - 1) ? argv[argc - 1] : NULL;
3556 if (fmt && has_help_option(options)) {
3557 /* If a format is explicitly specified (and possibly no filename is
3558 * given), print option help here */
3559 ret = print_block_option_help(filename, fmt);
3560 goto out;
3561 }
3562
3563 if (optind != argc - 1) {
3564 error_report("Expecting one image file name");
3565 ret = -1;
3566 goto out;
3567 }
3568
3569 flags = BDRV_O_RDWR;
3570 ret = bdrv_parse_cache_mode(cache, &flags, &writethrough);
3571 if (ret < 0) {
3572 error_report("Invalid cache option: %s", cache);
3573 goto out;
3574 }
3575
3576 blk = img_open(image_opts, filename, fmt, flags, writethrough, quiet);
3577 if (!blk) {
3578 ret = -1;
3579 goto out;
3580 }
3581 bs = blk_bs(blk);
3582
3583 fmt = bs->drv->format_name;
3584
3585 if (has_help_option(options)) {
3586 /* If the format was auto-detected, print option help here */
3587 ret = print_block_option_help(filename, fmt);
3588 goto out;
3589 }
3590
3591 if (!bs->drv->create_opts) {
3592 error_report("Format driver '%s' does not support any options to amend",
3593 fmt);
3594 ret = -1;
3595 goto out;
3596 }
3597
3598 create_opts = qemu_opts_append(create_opts, bs->drv->create_opts);
3599 opts = qemu_opts_create(create_opts, NULL, 0, &error_abort);
3600 qemu_opts_do_parse(opts, options, NULL, &err);
3601 if (err) {
3602 error_report_err(err);
3603 ret = -1;
3604 goto out;
3605 }
3606
3607 /* In case the driver does not call amend_status_cb() */
3608 qemu_progress_print(0.f, 0);
3609 ret = bdrv_amend_options(bs, opts, &amend_status_cb, NULL);
3610 qemu_progress_print(100.f, 0);
3611 if (ret < 0) {
3612 error_report("Error while amending options: %s", strerror(-ret));
3613 goto out;
3614 }
3615
3616 out:
3617 qemu_progress_end();
3618
3619 out_no_progress:
3620 blk_unref(blk);
3621 qemu_opts_del(opts);
3622 qemu_opts_free(create_opts);
3623 g_free(options);
3624
3625 if (ret) {
3626 return 1;
3627 }
3628 return 0;
3629 }
3630
3631 typedef struct BenchData {
3632 BlockBackend *blk;
3633 uint64_t image_size;
3634 bool write;
3635 int bufsize;
3636 int step;
3637 int nrreq;
3638 int n;
3639 int flush_interval;
3640 bool drain_on_flush;
3641 uint8_t *buf;
3642 QEMUIOVector *qiov;
3643
3644 int in_flight;
3645 bool in_flush;
3646 uint64_t offset;
3647 } BenchData;
3648
3649 static void bench_undrained_flush_cb(void *opaque, int ret)
3650 {
3651 if (ret < 0) {
3652 error_report("Failed flush request: %s", strerror(-ret));
3653 exit(EXIT_FAILURE);
3654 }
3655 }
3656
3657 static void bench_cb(void *opaque, int ret)
3658 {
3659 BenchData *b = opaque;
3660 BlockAIOCB *acb;
3661
3662 if (ret < 0) {
3663 error_report("Failed request: %s", strerror(-ret));
3664 exit(EXIT_FAILURE);
3665 }
3666
3667 if (b->in_flush) {
3668 /* Just finished a flush with drained queue: Start next requests */
3669 assert(b->in_flight == 0);
3670 b->in_flush = false;
3671 } else if (b->in_flight > 0) {
3672 int remaining = b->n - b->in_flight;
3673
3674 b->n--;
3675 b->in_flight--;
3676
3677 /* Time for flush? Drain queue if requested, then flush */
3678 if (b->flush_interval && remaining % b->flush_interval == 0) {
3679 if (!b->in_flight || !b->drain_on_flush) {
3680 BlockCompletionFunc *cb;
3681
3682 if (b->drain_on_flush) {
3683 b->in_flush = true;
3684 cb = bench_cb;
3685 } else {
3686 cb = bench_undrained_flush_cb;
3687 }
3688
3689 acb = blk_aio_flush(b->blk, cb, b);
3690 if (!acb) {
3691 error_report("Failed to issue flush request");
3692 exit(EXIT_FAILURE);
3693 }
3694 }
3695 if (b->drain_on_flush) {
3696 return;
3697 }
3698 }
3699 }
3700
3701 while (b->n > b->in_flight && b->in_flight < b->nrreq) {
3702 int64_t offset = b->offset;
3703 /* blk_aio_* might look for completed I/Os and kick bench_cb
3704 * again, so make sure this operation is counted by in_flight
3705 * and b->offset is ready for the next submission.
3706 */
3707 b->in_flight++;
3708 b->offset += b->step;
3709 b->offset %= b->image_size;
3710 if (b->write) {
3711 acb = blk_aio_pwritev(b->blk, offset, b->qiov, 0, bench_cb, b);
3712 } else {
3713 acb = blk_aio_preadv(b->blk, offset, b->qiov, 0, bench_cb, b);
3714 }
3715 if (!acb) {
3716 error_report("Failed to issue request");
3717 exit(EXIT_FAILURE);
3718 }
3719 }
3720 }
3721
3722 static int img_bench(int argc, char **argv)
3723 {
3724 int c, ret = 0;
3725 const char *fmt = NULL, *filename;
3726 bool quiet = false;
3727 bool image_opts = false;
3728 bool is_write = false;
3729 int count = 75000;
3730 int depth = 64;
3731 int64_t offset = 0;
3732 size_t bufsize = 4096;
3733 int pattern = 0;
3734 size_t step = 0;
3735 int flush_interval = 0;
3736 bool drain_on_flush = true;
3737 int64_t image_size;
3738 BlockBackend *blk = NULL;
3739 BenchData data = {};
3740 int flags = 0;
3741 bool writethrough = false;
3742 struct timeval t1, t2;
3743 int i;
3744
3745 for (;;) {
3746 static const struct option long_options[] = {
3747 {"help", no_argument, 0, 'h'},
3748 {"flush-interval", required_argument, 0, OPTION_FLUSH_INTERVAL},
3749 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
3750 {"pattern", required_argument, 0, OPTION_PATTERN},
3751 {"no-drain", no_argument, 0, OPTION_NO_DRAIN},
3752 {0, 0, 0, 0}
3753 };
3754 c = getopt_long(argc, argv, ":hc:d:f:no:qs:S:t:w", long_options, NULL);
3755 if (c == -1) {
3756 break;
3757 }
3758
3759 switch (c) {
3760 case ':':
3761 missing_argument(argv[optind - 1]);
3762 break;
3763 case '?':
3764 unrecognized_option(argv[optind - 1]);
3765 break;
3766 case 'h':
3767 help();
3768 break;
3769 case 'c':
3770 {
3771 unsigned long res;
3772
3773 if (qemu_strtoul(optarg, NULL, 0, &res) < 0 || res > INT_MAX) {
3774 error_report("Invalid request count specified");
3775 return 1;
3776 }
3777 count = res;
3778 break;
3779 }
3780 case 'd':
3781 {
3782 unsigned long res;
3783
3784 if (qemu_strtoul(optarg, NULL, 0, &res) < 0 || res > INT_MAX) {
3785 error_report("Invalid queue depth specified");
3786 return 1;
3787 }
3788 depth = res;
3789 break;
3790 }
3791 case 'f':
3792 fmt = optarg;
3793 break;
3794 case 'n':
3795 flags |= BDRV_O_NATIVE_AIO;
3796 break;
3797 case 'o':
3798 {
3799 offset = cvtnum(optarg);
3800 if (offset < 0) {
3801 error_report("Invalid offset specified");
3802 return 1;
3803 }
3804 break;
3805 }
3806 break;
3807 case 'q':
3808 quiet = true;
3809 break;
3810 case 's':
3811 {
3812 int64_t sval;
3813
3814 sval = cvtnum(optarg);
3815 if (sval < 0 || sval > INT_MAX) {
3816 error_report("Invalid buffer size specified");
3817 return 1;
3818 }
3819
3820 bufsize = sval;
3821 break;
3822 }
3823 case 'S':
3824 {
3825 int64_t sval;
3826
3827 sval = cvtnum(optarg);
3828 if (sval < 0 || sval > INT_MAX) {
3829 error_report("Invalid step size specified");
3830 return 1;
3831 }
3832
3833 step = sval;
3834 break;
3835 }
3836 case 't':
3837 ret = bdrv_parse_cache_mode(optarg, &flags, &writethrough);
3838 if (ret < 0) {
3839 error_report("Invalid cache mode");
3840 ret = -1;
3841 goto out;
3842 }
3843 break;
3844 case 'w':
3845 flags |= BDRV_O_RDWR;
3846 is_write = true;
3847 break;
3848 case OPTION_PATTERN:
3849 {
3850 unsigned long res;
3851
3852 if (qemu_strtoul(optarg, NULL, 0, &res) < 0 || res > 0xff) {
3853 error_report("Invalid pattern byte specified");
3854 return 1;
3855 }
3856 pattern = res;
3857 break;
3858 }
3859 case OPTION_FLUSH_INTERVAL:
3860 {
3861 unsigned long res;
3862
3863 if (qemu_strtoul(optarg, NULL, 0, &res) < 0 || res > INT_MAX) {
3864 error_report("Invalid flush interval specified");
3865 return 1;
3866 }
3867 flush_interval = res;
3868 break;
3869 }
3870 case OPTION_NO_DRAIN:
3871 drain_on_flush = false;
3872 break;
3873 case OPTION_IMAGE_OPTS:
3874 image_opts = true;
3875 break;
3876 }
3877 }
3878
3879 if (optind != argc - 1) {
3880 error_exit("Expecting one image file name");
3881 }
3882 filename = argv[argc - 1];
3883
3884 if (!is_write && flush_interval) {
3885 error_report("--flush-interval is only available in write tests");
3886 ret = -1;
3887 goto out;
3888 }
3889 if (flush_interval && flush_interval < depth) {
3890 error_report("Flush interval can't be smaller than depth");
3891 ret = -1;
3892 goto out;
3893 }
3894
3895 blk = img_open(image_opts, filename, fmt, flags, writethrough, quiet);
3896 if (!blk) {
3897 ret = -1;
3898 goto out;
3899 }
3900
3901 image_size = blk_getlength(blk);
3902 if (image_size < 0) {
3903 ret = image_size;
3904 goto out;
3905 }
3906
3907 data = (BenchData) {
3908 .blk = blk,
3909 .image_size = image_size,
3910 .bufsize = bufsize,
3911 .step = step ?: bufsize,
3912 .nrreq = depth,
3913 .n = count,
3914 .offset = offset,
3915 .write = is_write,
3916 .flush_interval = flush_interval,
3917 .drain_on_flush = drain_on_flush,
3918 };
3919 printf("Sending %d %s requests, %d bytes each, %d in parallel "
3920 "(starting at offset %" PRId64 ", step size %d)\n",
3921 data.n, data.write ? "write" : "read", data.bufsize, data.nrreq,
3922 data.offset, data.step);
3923 if (flush_interval) {
3924 printf("Sending flush every %d requests\n", flush_interval);
3925 }
3926
3927 data.buf = blk_blockalign(blk, data.nrreq * data.bufsize);
3928 memset(data.buf, pattern, data.nrreq * data.bufsize);
3929
3930 data.qiov = g_new(QEMUIOVector, data.nrreq);
3931 for (i = 0; i < data.nrreq; i++) {
3932 qemu_iovec_init(&data.qiov[i], 1);
3933 qemu_iovec_add(&data.qiov[i],
3934 data.buf + i * data.bufsize, data.bufsize);
3935 }
3936
3937 gettimeofday(&t1, NULL);
3938 bench_cb(&data, 0);
3939
3940 while (data.n > 0) {
3941 main_loop_wait(false);
3942 }
3943 gettimeofday(&t2, NULL);
3944
3945 printf("Run completed in %3.3f seconds.\n",
3946 (t2.tv_sec - t1.tv_sec)
3947 + ((double)(t2.tv_usec - t1.tv_usec) / 1000000));
3948
3949 out:
3950 qemu_vfree(data.buf);
3951 blk_unref(blk);
3952
3953 if (ret) {
3954 return 1;
3955 }
3956 return 0;
3957 }
3958
3959 #define C_BS 01
3960 #define C_COUNT 02
3961 #define C_IF 04
3962 #define C_OF 010
3963 #define C_SKIP 020
3964
3965 struct DdInfo {
3966 unsigned int flags;
3967 int64_t count;
3968 };
3969
3970 struct DdIo {
3971 int bsz; /* Block size */
3972 char *filename;
3973 uint8_t *buf;
3974 int64_t offset;
3975 };
3976
3977 struct DdOpts {
3978 const char *name;
3979 int (*f)(const char *, struct DdIo *, struct DdIo *, struct DdInfo *);
3980 unsigned int flag;
3981 };
3982
3983 static int img_dd_bs(const char *arg,
3984 struct DdIo *in, struct DdIo *out,
3985 struct DdInfo *dd)
3986 {
3987 int64_t res;
3988
3989 res = cvtnum(arg);
3990
3991 if (res <= 0 || res > INT_MAX) {
3992 error_report("invalid number: '%s'", arg);
3993 return 1;
3994 }
3995 in->bsz = out->bsz = res;
3996
3997 return 0;
3998 }
3999
4000 static int img_dd_count(const char *arg,
4001 struct DdIo *in, struct DdIo *out,
4002 struct DdInfo *dd)
4003 {
4004 dd->count = cvtnum(arg);
4005
4006 if (dd->count < 0) {
4007 error_report("invalid number: '%s'", arg);
4008 return 1;
4009 }
4010
4011 return 0;
4012 }
4013
4014 static int img_dd_if(const char *arg,
4015 struct DdIo *in, struct DdIo *out,
4016 struct DdInfo *dd)
4017 {
4018 in->filename = g_strdup(arg);
4019
4020 return 0;
4021 }
4022
4023 static int img_dd_of(const char *arg,
4024 struct DdIo *in, struct DdIo *out,
4025 struct DdInfo *dd)
4026 {
4027 out->filename = g_strdup(arg);
4028
4029 return 0;
4030 }
4031
4032 static int img_dd_skip(const char *arg,
4033 struct DdIo *in, struct DdIo *out,
4034 struct DdInfo *dd)
4035 {
4036 in->offset = cvtnum(arg);
4037
4038 if (in->offset < 0) {
4039 error_report("invalid number: '%s'", arg);
4040 return 1;
4041 }
4042
4043 return 0;
4044 }
4045
4046 static int img_dd(int argc, char **argv)
4047 {
4048 int ret = 0;
4049 char *arg = NULL;
4050 char *tmp;
4051 BlockDriver *drv = NULL, *proto_drv = NULL;
4052 BlockBackend *blk1 = NULL, *blk2 = NULL;
4053 QemuOpts *opts = NULL;
4054 QemuOptsList *create_opts = NULL;
4055 Error *local_err = NULL;
4056 bool image_opts = false;
4057 int c, i;
4058 const char *out_fmt = "raw";
4059 const char *fmt = NULL;
4060 int64_t size = 0;
4061 int64_t block_count = 0, out_pos, in_pos;
4062 struct DdInfo dd = {
4063 .flags = 0,
4064 .count = 0,
4065 };
4066 struct DdIo in = {
4067 .bsz = 512, /* Block size is by default 512 bytes */
4068 .filename = NULL,
4069 .buf = NULL,
4070 .offset = 0
4071 };
4072 struct DdIo out = {
4073 .bsz = 512,
4074 .filename = NULL,
4075 .buf = NULL,
4076 .offset = 0
4077 };
4078
4079 const struct DdOpts options[] = {
4080 { "bs", img_dd_bs, C_BS },
4081 { "count", img_dd_count, C_COUNT },
4082 { "if", img_dd_if, C_IF },
4083 { "of", img_dd_of, C_OF },
4084 { "skip", img_dd_skip, C_SKIP },
4085 { NULL, NULL, 0 }
4086 };
4087 const struct option long_options[] = {
4088 { "help", no_argument, 0, 'h'},
4089 { "image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
4090 { 0, 0, 0, 0 }
4091 };
4092
4093 while ((c = getopt_long(argc, argv, ":hf:O:", long_options, NULL))) {
4094 if (c == EOF) {
4095 break;
4096 }
4097 switch (c) {
4098 case 'O':
4099 out_fmt = optarg;
4100 break;
4101 case 'f':
4102 fmt = optarg;
4103 break;
4104 case ':':
4105 missing_argument(argv[optind - 1]);
4106 break;
4107 case '?':
4108 unrecognized_option(argv[optind - 1]);
4109 break;
4110 case 'h':
4111 help();
4112 break;
4113 case OPTION_IMAGE_OPTS:
4114 image_opts = true;
4115 break;
4116 }
4117 }
4118
4119 for (i = optind; i < argc; i++) {
4120 int j;
4121 arg = g_strdup(argv[i]);
4122
4123 tmp = strchr(arg, '=');
4124 if (tmp == NULL) {
4125 error_report("unrecognized operand %s", arg);
4126 ret = -1;
4127 goto out;
4128 }
4129
4130 *tmp++ = '\0';
4131
4132 for (j = 0; options[j].name != NULL; j++) {
4133 if (!strcmp(arg, options[j].name)) {
4134 break;
4135 }
4136 }
4137 if (options[j].name == NULL) {
4138 error_report("unrecognized operand %s", arg);
4139 ret = -1;
4140 goto out;
4141 }
4142
4143 if (options[j].f(tmp, &in, &out, &dd) != 0) {
4144 ret = -1;
4145 goto out;
4146 }
4147 dd.flags |= options[j].flag;
4148 g_free(arg);
4149 arg = NULL;
4150 }
4151
4152 if (!(dd.flags & C_IF && dd.flags & C_OF)) {
4153 error_report("Must specify both input and output files");
4154 ret = -1;
4155 goto out;
4156 }
4157 blk1 = img_open(image_opts, in.filename, fmt, 0, false, false);
4158
4159 if (!blk1) {
4160 ret = -1;
4161 goto out;
4162 }
4163
4164 drv = bdrv_find_format(out_fmt);
4165 if (!drv) {
4166 error_report("Unknown file format");
4167 ret = -1;
4168 goto out;
4169 }
4170 proto_drv = bdrv_find_protocol(out.filename, true, &local_err);
4171
4172 if (!proto_drv) {
4173 error_report_err(local_err);
4174 ret = -1;
4175 goto out;
4176 }
4177 if (!drv->create_opts) {
4178 error_report("Format driver '%s' does not support image creation",
4179 drv->format_name);
4180 ret = -1;
4181 goto out;
4182 }
4183 if (!proto_drv->create_opts) {
4184 error_report("Protocol driver '%s' does not support image creation",
4185 proto_drv->format_name);
4186 ret = -1;
4187 goto out;
4188 }
4189 create_opts = qemu_opts_append(create_opts, drv->create_opts);
4190 create_opts = qemu_opts_append(create_opts, proto_drv->create_opts);
4191
4192 opts = qemu_opts_create(create_opts, NULL, 0, &error_abort);
4193
4194 size = blk_getlength(blk1);
4195 if (size < 0) {
4196 error_report("Failed to get size for '%s'", in.filename);
4197 ret = -1;
4198 goto out;
4199 }
4200
4201 if (dd.flags & C_COUNT && dd.count <= INT64_MAX / in.bsz &&
4202 dd.count * in.bsz < size) {
4203 size = dd.count * in.bsz;
4204 }
4205
4206 /* Overflow means the specified offset is beyond input image's size */
4207 if (dd.flags & C_SKIP && (in.offset > INT64_MAX / in.bsz ||
4208 size < in.bsz * in.offset)) {
4209 qemu_opt_set_number(opts, BLOCK_OPT_SIZE, 0, &error_abort);
4210 } else {
4211 qemu_opt_set_number(opts, BLOCK_OPT_SIZE,
4212 size - in.bsz * in.offset, &error_abort);
4213 }
4214
4215 ret = bdrv_create(drv, out.filename, opts, &local_err);
4216 if (ret < 0) {
4217 error_reportf_err(local_err,
4218 "%s: error while creating output image: ",
4219 out.filename);
4220 ret = -1;
4221 goto out;
4222 }
4223
4224 blk2 = img_open(image_opts, out.filename, out_fmt, BDRV_O_RDWR,
4225 false, false);
4226
4227 if (!blk2) {
4228 ret = -1;
4229 goto out;
4230 }
4231
4232 if (dd.flags & C_SKIP && (in.offset > INT64_MAX / in.bsz ||
4233 size < in.offset * in.bsz)) {
4234 /* We give a warning if the skip option is bigger than the input
4235 * size and create an empty output disk image (i.e. like dd(1)).
4236 */
4237 error_report("%s: cannot skip to specified offset", in.filename);
4238 in_pos = size;
4239 } else {
4240 in_pos = in.offset * in.bsz;
4241 }
4242
4243 in.buf = g_new(uint8_t, in.bsz);
4244
4245 for (out_pos = 0; in_pos < size; block_count++) {
4246 int in_ret, out_ret;
4247
4248 if (in_pos + in.bsz > size) {
4249 in_ret = blk_pread(blk1, in_pos, in.buf, size - in_pos);
4250 } else {
4251 in_ret = blk_pread(blk1, in_pos, in.buf, in.bsz);
4252 }
4253 if (in_ret < 0) {
4254 error_report("error while reading from input image file: %s",
4255 strerror(-in_ret));
4256 ret = -1;
4257 goto out;
4258 }
4259 in_pos += in_ret;
4260
4261 out_ret = blk_pwrite(blk2, out_pos, in.buf, in_ret, 0);
4262
4263 if (out_ret < 0) {
4264 error_report("error while writing to output image file: %s",
4265 strerror(-out_ret));
4266 ret = -1;
4267 goto out;
4268 }
4269 out_pos += out_ret;
4270 }
4271
4272 out:
4273 g_free(arg);
4274 qemu_opts_del(opts);
4275 qemu_opts_free(create_opts);
4276 blk_unref(blk1);
4277 blk_unref(blk2);
4278 g_free(in.filename);
4279 g_free(out.filename);
4280 g_free(in.buf);
4281 g_free(out.buf);
4282
4283 if (ret) {
4284 return 1;
4285 }
4286 return 0;
4287 }
4288
4289
4290 static const img_cmd_t img_cmds[] = {
4291 #define DEF(option, callback, arg_string) \
4292 { option, callback },
4293 #include "qemu-img-cmds.h"
4294 #undef DEF
4295 #undef GEN_DOCS
4296 { NULL, NULL, },
4297 };
4298
4299 int main(int argc, char **argv)
4300 {
4301 const img_cmd_t *cmd;
4302 const char *cmdname;
4303 Error *local_error = NULL;
4304 char *trace_file = NULL;
4305 int c;
4306 static const struct option long_options[] = {
4307 {"help", no_argument, 0, 'h'},
4308 {"version", no_argument, 0, 'V'},
4309 {"trace", required_argument, NULL, 'T'},
4310 {0, 0, 0, 0}
4311 };
4312
4313 #ifdef CONFIG_POSIX
4314 signal(SIGPIPE, SIG_IGN);
4315 #endif
4316
4317 module_call_init(MODULE_INIT_TRACE);
4318 error_set_progname(argv[0]);
4319 qemu_init_exec_dir(argv[0]);
4320
4321 if (qemu_init_main_loop(&local_error)) {
4322 error_report_err(local_error);
4323 exit(EXIT_FAILURE);
4324 }
4325
4326 qcrypto_init(&error_fatal);
4327
4328 module_call_init(MODULE_INIT_QOM);
4329 bdrv_init();
4330 if (argc < 2) {
4331 error_exit("Not enough arguments");
4332 }
4333
4334 qemu_add_opts(&qemu_object_opts);
4335 qemu_add_opts(&qemu_source_opts);
4336 qemu_add_opts(&qemu_trace_opts);
4337
4338 while ((c = getopt_long(argc, argv, "+:hVT:", long_options, NULL)) != -1) {
4339 switch (c) {
4340 case ':':
4341 missing_argument(argv[optind - 1]);
4342 return 0;
4343 case '?':
4344 unrecognized_option(argv[optind - 1]);
4345 return 0;
4346 case 'h':
4347 help();
4348 return 0;
4349 case 'V':
4350 printf(QEMU_IMG_VERSION);
4351 return 0;
4352 case 'T':
4353 g_free(trace_file);
4354 trace_file = trace_opt_parse(optarg);
4355 break;
4356 }
4357 }
4358
4359 cmdname = argv[optind];
4360
4361 /* reset getopt_long scanning */
4362 argc -= optind;
4363 if (argc < 1) {
4364 return 0;
4365 }
4366 argv += optind;
4367 optind = 0;
4368
4369 if (!trace_init_backends()) {
4370 exit(1);
4371 }
4372 trace_init_file(trace_file);
4373 qemu_set_log(LOG_TRACE);
4374
4375 /* find the command */
4376 for (cmd = img_cmds; cmd->name != NULL; cmd++) {
4377 if (!strcmp(cmdname, cmd->name)) {
4378 return cmd->handler(argc, argv);
4379 }
4380 }
4381
4382 /* not found */
4383 error_exit("Command not found: %s", cmdname);
4384 }