1 /* 2 * QEMU disk image utility 3 * 4 * Copyright (c) 2003-2008 Fabrice Bellard 5 * 6 * Permission is hereby granted, free of charge, to any person obtaining a copy 7 * of this software and associated documentation files (the "Software"), to deal 8 * in the Software without restriction, including without limitation the rights 9 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell 10 * copies of the Software, and to permit persons to whom the Software is 11 * furnished to do so, subject to the following conditions: 12 * 13 * The above copyright notice and this permission notice shall be included in 14 * all copies or substantial portions of the Software. 15 * 16 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR 17 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, 18 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL 19 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER 20 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, 21 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN 22 * THE SOFTWARE. 23 */ 24 25 #include "qemu/osdep.h" 26 #include <getopt.h> 27 28 #include "qemu/help-texts.h" 29 #include "qemu/qemu-progress.h" 30 #include "qemu-version.h" 31 #include "qapi/error.h" 32 #include "qapi/qapi-commands-block-core.h" 33 #include "qapi/qapi-visit-block-core.h" 34 #include "qapi/qobject-output-visitor.h" 35 #include "qapi/qmp/qjson.h" 36 #include "qapi/qmp/qdict.h" 37 #include "qemu/cutils.h" 38 #include "qemu/config-file.h" 39 #include "qemu/option.h" 40 #include "qemu/error-report.h" 41 #include "qemu/log.h" 42 #include "qemu/main-loop.h" 43 #include "qemu/module.h" 44 #include "qemu/sockets.h" 45 #include "qemu/units.h" 46 #include "qemu/memalign.h" 47 #include "qom/object_interfaces.h" 48 #include "sysemu/block-backend.h" 49 #include "block/block_int.h" 50 #include "block/blockjob.h" 51 #include "block/dirty-bitmap.h" 52 #include "block/qapi.h" 53 #include "crypto/init.h" 54 #include "trace/control.h" 55 #include "qemu/throttle.h" 56 #include "block/throttle-groups.h" 57 58 #define QEMU_IMG_VERSION "qemu-img version " QEMU_FULL_VERSION \ 59 "\n" QEMU_COPYRIGHT "\n" 60 61 typedef struct img_cmd_t { 62 const char *name; 63 int (*handler)(int argc, char **argv); 64 } img_cmd_t; 65 66 enum { 67 OPTION_OUTPUT = 256, 68 OPTION_BACKING_CHAIN = 257, 69 OPTION_OBJECT = 258, 70 OPTION_IMAGE_OPTS = 259, 71 OPTION_PATTERN = 260, 72 OPTION_FLUSH_INTERVAL = 261, 73 OPTION_NO_DRAIN = 262, 74 OPTION_TARGET_IMAGE_OPTS = 263, 75 OPTION_SIZE = 264, 76 OPTION_PREALLOCATION = 265, 77 OPTION_SHRINK = 266, 78 OPTION_SALVAGE = 267, 79 OPTION_TARGET_IS_ZERO = 268, 80 OPTION_ADD = 269, 81 OPTION_REMOVE = 270, 82 OPTION_CLEAR = 271, 83 OPTION_ENABLE = 272, 84 OPTION_DISABLE = 273, 85 OPTION_MERGE = 274, 86 OPTION_BITMAPS = 275, 87 OPTION_FORCE = 276, 88 OPTION_SKIP_BROKEN = 277, 89 }; 90 91 typedef enum OutputFormat { 92 OFORMAT_JSON, 93 OFORMAT_HUMAN, 94 } OutputFormat; 95 96 /* Default to cache=writeback as data integrity is not important for qemu-img */ 97 #define BDRV_DEFAULT_CACHE "writeback" 98 99 static void format_print(void *opaque, const char *name) 100 { 101 printf(" %s", name); 102 } 103 104 static G_NORETURN G_GNUC_PRINTF(1, 2) 105 void error_exit(const char *fmt, ...) 106 { 107 va_list ap; 108 109 va_start(ap, fmt); 110 error_vreport(fmt, ap); 111 va_end(ap); 112 113 error_printf("Try 'qemu-img --help' for more information\n"); 114 exit(EXIT_FAILURE); 115 } 116 117 static G_NORETURN 118 void missing_argument(const char *option) 119 { 120 error_exit("missing argument for option '%s'", option); 121 } 122 123 static G_NORETURN 124 void unrecognized_option(const char *option) 125 { 126 error_exit("unrecognized option '%s'", option); 127 } 128 129 /* Please keep in synch with docs/tools/qemu-img.rst */ 130 static G_NORETURN 131 void help(void) 132 { 133 const char *help_msg = 134 QEMU_IMG_VERSION 135 "usage: qemu-img [standard options] command [command options]\n" 136 "QEMU disk image utility\n" 137 "\n" 138 " '-h', '--help' display this help and exit\n" 139 " '-V', '--version' output version information and exit\n" 140 " '-T', '--trace' [[enable=]<pattern>][,events=<file>][,file=<file>]\n" 141 " specify tracing options\n" 142 "\n" 143 "Command syntax:\n" 144 #define DEF(option, callback, arg_string) \ 145 " " arg_string "\n" 146 #include "qemu-img-cmds.h" 147 #undef DEF 148 "\n" 149 "Command parameters:\n" 150 " 'filename' is a disk image filename\n" 151 " 'objectdef' is a QEMU user creatable object definition. See the qemu(1)\n" 152 " manual page for a description of the object properties. The most common\n" 153 " object type is a 'secret', which is used to supply passwords and/or\n" 154 " encryption keys.\n" 155 " 'fmt' is the disk image format. It is guessed automatically in most cases\n" 156 " 'cache' is the cache mode used to write the output disk image, the valid\n" 157 " options are: 'none', 'writeback' (default, except for convert), 'writethrough',\n" 158 " 'directsync' and 'unsafe' (default for convert)\n" 159 " 'src_cache' is the cache mode used to read input disk images, the valid\n" 160 " options are the same as for the 'cache' option\n" 161 " 'size' is the disk image size in bytes. Optional suffixes\n" 162 " 'k' or 'K' (kilobyte, 1024), 'M' (megabyte, 1024k), 'G' (gigabyte, 1024M),\n" 163 " 'T' (terabyte, 1024G), 'P' (petabyte, 1024T) and 'E' (exabyte, 1024P) are\n" 164 " supported. 'b' is ignored.\n" 165 " 'output_filename' is the destination disk image filename\n" 166 " 'output_fmt' is the destination format\n" 167 " 'options' is a comma separated list of format specific options in a\n" 168 " name=value format. Use -o help for an overview of the options supported by\n" 169 " the used format\n" 170 " 'snapshot_param' is param used for internal snapshot, format\n" 171 " is 'snapshot.id=[ID],snapshot.name=[NAME]', or\n" 172 " '[ID_OR_NAME]'\n" 173 " '-c' indicates that target image must be compressed (qcow format only)\n" 174 " '-u' allows unsafe backing chains. For rebasing, it is assumed that old and\n" 175 " new backing file match exactly. The image doesn't need a working\n" 176 " backing file before rebasing in this case (useful for renaming the\n" 177 " backing file). For image creation, allow creating without attempting\n" 178 " to open the backing file.\n" 179 " '-h' with or without a command shows this help and lists the supported formats\n" 180 " '-p' show progress of command (only certain commands)\n" 181 " '-q' use Quiet mode - do not print any output (except errors)\n" 182 " '-S' indicates the consecutive number of bytes (defaults to 4k) that must\n" 183 " contain only zeros for qemu-img to create a sparse image during\n" 184 " conversion. If the number of bytes is 0, the source will not be scanned for\n" 185 " unallocated or zero sectors, and the destination image will always be\n" 186 " fully allocated\n" 187 " '--output' takes the format in which the output must be done (human or json)\n" 188 " '-n' skips the target volume creation (useful if the volume is created\n" 189 " prior to running qemu-img)\n" 190 "\n" 191 "Parameters to bitmap subcommand:\n" 192 " 'bitmap' is the name of the bitmap to manipulate, through one or more\n" 193 " actions from '--add', '--remove', '--clear', '--enable', '--disable',\n" 194 " or '--merge source'\n" 195 " '-g granularity' sets the granularity for '--add' actions\n" 196 " '-b source' and '-F src_fmt' tell '--merge' actions to find the source\n" 197 " bitmaps from an alternative file\n" 198 "\n" 199 "Parameters to check subcommand:\n" 200 " '-r' tries to repair any inconsistencies that are found during the check.\n" 201 " '-r leaks' repairs only cluster leaks, whereas '-r all' fixes all\n" 202 " kinds of errors, with a higher risk of choosing the wrong fix or\n" 203 " hiding corruption that has already occurred.\n" 204 "\n" 205 "Parameters to convert subcommand:\n" 206 " '--bitmaps' copies all top-level persistent bitmaps to destination\n" 207 " '-m' specifies how many coroutines work in parallel during the convert\n" 208 " process (defaults to 8)\n" 209 " '-W' allow to write to the target out of order rather than sequential\n" 210 "\n" 211 "Parameters to snapshot subcommand:\n" 212 " 'snapshot' is the name of the snapshot to create, apply or delete\n" 213 " '-a' applies a snapshot (revert disk to saved state)\n" 214 " '-c' creates a snapshot\n" 215 " '-d' deletes a snapshot\n" 216 " '-l' lists all snapshots in the given image\n" 217 "\n" 218 "Parameters to compare subcommand:\n" 219 " '-f' first image format\n" 220 " '-F' second image format\n" 221 " '-s' run in Strict mode - fail on different image size or sector allocation\n" 222 "\n" 223 "Parameters to dd subcommand:\n" 224 " 'bs=BYTES' read and write up to BYTES bytes at a time " 225 "(default: 512)\n" 226 " 'count=N' copy only N input blocks\n" 227 " 'if=FILE' read from FILE\n" 228 " 'of=FILE' write to FILE\n" 229 " 'skip=N' skip N bs-sized blocks at the start of input\n"; 230 231 printf("%s\nSupported formats:", help_msg); 232 bdrv_iterate_format(format_print, NULL, false); 233 printf("\n\n" QEMU_HELP_BOTTOM "\n"); 234 exit(EXIT_SUCCESS); 235 } 236 237 /* 238 * Is @list safe for accumulate_options()? 239 * It is when multiple of them can be joined together separated by ','. 240 * To make that work, @list must not start with ',' (or else a 241 * separating ',' preceding it gets escaped), and it must not end with 242 * an odd number of ',' (or else a separating ',' following it gets 243 * escaped), or be empty (or else a separating ',' preceding it can 244 * escape a separating ',' following it). 245 * 246 */ 247 static bool is_valid_option_list(const char *list) 248 { 249 size_t len = strlen(list); 250 size_t i; 251 252 if (!list[0] || list[0] == ',') { 253 return false; 254 } 255 256 for (i = len; i > 0 && list[i - 1] == ','; i--) { 257 } 258 if ((len - i) % 2) { 259 return false; 260 } 261 262 return true; 263 } 264 265 static int accumulate_options(char **options, char *list) 266 { 267 char *new_options; 268 269 if (!is_valid_option_list(list)) { 270 error_report("Invalid option list: %s", list); 271 return -1; 272 } 273 274 if (!*options) { 275 *options = g_strdup(list); 276 } else { 277 new_options = g_strdup_printf("%s,%s", *options, list); 278 g_free(*options); 279 *options = new_options; 280 } 281 return 0; 282 } 283 284 static QemuOptsList qemu_source_opts = { 285 .name = "source", 286 .implied_opt_name = "file", 287 .head = QTAILQ_HEAD_INITIALIZER(qemu_source_opts.head), 288 .desc = { 289 { } 290 }, 291 }; 292 293 static int G_GNUC_PRINTF(2, 3) qprintf(bool quiet, const char *fmt, ...) 294 { 295 int ret = 0; 296 if (!quiet) { 297 va_list args; 298 va_start(args, fmt); 299 ret = vprintf(fmt, args); 300 va_end(args); 301 } 302 return ret; 303 } 304 305 306 static int print_block_option_help(const char *filename, const char *fmt) 307 { 308 BlockDriver *drv, *proto_drv; 309 QemuOptsList *create_opts = NULL; 310 Error *local_err = NULL; 311 312 /* Find driver and parse its options */ 313 drv = bdrv_find_format(fmt); 314 if (!drv) { 315 error_report("Unknown file format '%s'", fmt); 316 return 1; 317 } 318 319 if (!drv->create_opts) { 320 error_report("Format driver '%s' does not support image creation", fmt); 321 return 1; 322 } 323 324 create_opts = qemu_opts_append(create_opts, drv->create_opts); 325 if (filename) { 326 proto_drv = bdrv_find_protocol(filename, true, &local_err); 327 if (!proto_drv) { 328 error_report_err(local_err); 329 qemu_opts_free(create_opts); 330 return 1; 331 } 332 if (!proto_drv->create_opts) { 333 error_report("Protocol driver '%s' does not support image creation", 334 proto_drv->format_name); 335 qemu_opts_free(create_opts); 336 return 1; 337 } 338 create_opts = qemu_opts_append(create_opts, proto_drv->create_opts); 339 } 340 341 if (filename) { 342 printf("Supported options:\n"); 343 } else { 344 printf("Supported %s options:\n", fmt); 345 } 346 qemu_opts_print_help(create_opts, false); 347 qemu_opts_free(create_opts); 348 349 if (!filename) { 350 printf("\n" 351 "The protocol level may support further options.\n" 352 "Specify the target filename to include those options.\n"); 353 } 354 355 return 0; 356 } 357 358 359 static BlockBackend *img_open_opts(const char *optstr, 360 QemuOpts *opts, int flags, bool writethrough, 361 bool quiet, bool force_share) 362 { 363 QDict *options; 364 Error *local_err = NULL; 365 BlockBackend *blk; 366 options = qemu_opts_to_qdict(opts, NULL); 367 if (force_share) { 368 if (qdict_haskey(options, BDRV_OPT_FORCE_SHARE) 369 && strcmp(qdict_get_str(options, BDRV_OPT_FORCE_SHARE), "on")) { 370 error_report("--force-share/-U conflicts with image options"); 371 qobject_unref(options); 372 return NULL; 373 } 374 qdict_put_str(options, BDRV_OPT_FORCE_SHARE, "on"); 375 } 376 blk = blk_new_open(NULL, NULL, options, flags, &local_err); 377 if (!blk) { 378 error_reportf_err(local_err, "Could not open '%s': ", optstr); 379 return NULL; 380 } 381 blk_set_enable_write_cache(blk, !writethrough); 382 383 return blk; 384 } 385 386 static BlockBackend *img_open_file(const char *filename, 387 QDict *options, 388 const char *fmt, int flags, 389 bool writethrough, bool quiet, 390 bool force_share) 391 { 392 BlockBackend *blk; 393 Error *local_err = NULL; 394 395 if (!options) { 396 options = qdict_new(); 397 } 398 if (fmt) { 399 qdict_put_str(options, "driver", fmt); 400 } 401 402 if (force_share) { 403 qdict_put_bool(options, BDRV_OPT_FORCE_SHARE, true); 404 } 405 blk = blk_new_open(filename, NULL, options, flags, &local_err); 406 if (!blk) { 407 error_reportf_err(local_err, "Could not open '%s': ", filename); 408 return NULL; 409 } 410 blk_set_enable_write_cache(blk, !writethrough); 411 412 return blk; 413 } 414 415 416 static int img_add_key_secrets(void *opaque, 417 const char *name, const char *value, 418 Error **errp) 419 { 420 QDict *options = opaque; 421 422 if (g_str_has_suffix(name, "key-secret")) { 423 qdict_put_str(options, name, value); 424 } 425 426 return 0; 427 } 428 429 430 static BlockBackend *img_open(bool image_opts, 431 const char *filename, 432 const char *fmt, int flags, bool writethrough, 433 bool quiet, bool force_share) 434 { 435 BlockBackend *blk; 436 if (image_opts) { 437 QemuOpts *opts; 438 if (fmt) { 439 error_report("--image-opts and --format are mutually exclusive"); 440 return NULL; 441 } 442 opts = qemu_opts_parse_noisily(qemu_find_opts("source"), 443 filename, true); 444 if (!opts) { 445 return NULL; 446 } 447 blk = img_open_opts(filename, opts, flags, writethrough, quiet, 448 force_share); 449 } else { 450 blk = img_open_file(filename, NULL, fmt, flags, writethrough, quiet, 451 force_share); 452 } 453 454 if (blk) { 455 blk_set_force_allow_inactivate(blk); 456 } 457 458 return blk; 459 } 460 461 462 static int add_old_style_options(const char *fmt, QemuOpts *opts, 463 const char *base_filename, 464 const char *base_fmt) 465 { 466 if (base_filename) { 467 if (!qemu_opt_set(opts, BLOCK_OPT_BACKING_FILE, base_filename, 468 NULL)) { 469 error_report("Backing file not supported for file format '%s'", 470 fmt); 471 return -1; 472 } 473 } 474 if (base_fmt) { 475 if (!qemu_opt_set(opts, BLOCK_OPT_BACKING_FMT, base_fmt, NULL)) { 476 error_report("Backing file format not supported for file " 477 "format '%s'", fmt); 478 return -1; 479 } 480 } 481 return 0; 482 } 483 484 static int64_t cvtnum_full(const char *name, const char *value, int64_t min, 485 int64_t max) 486 { 487 int err; 488 uint64_t res; 489 490 err = qemu_strtosz(value, NULL, &res); 491 if (err < 0 && err != -ERANGE) { 492 error_report("Invalid %s specified. You may use " 493 "k, M, G, T, P or E suffixes for", name); 494 error_report("kilobytes, megabytes, gigabytes, terabytes, " 495 "petabytes and exabytes."); 496 return err; 497 } 498 if (err == -ERANGE || res > max || res < min) { 499 error_report("Invalid %s specified. Must be between %" PRId64 500 " and %" PRId64 ".", name, min, max); 501 return -ERANGE; 502 } 503 return res; 504 } 505 506 static int64_t cvtnum(const char *name, const char *value) 507 { 508 return cvtnum_full(name, value, 0, INT64_MAX); 509 } 510 511 static int img_create(int argc, char **argv) 512 { 513 int c; 514 uint64_t img_size = -1; 515 const char *fmt = "raw"; 516 const char *base_fmt = NULL; 517 const char *filename; 518 const char *base_filename = NULL; 519 char *options = NULL; 520 Error *local_err = NULL; 521 bool quiet = false; 522 int flags = 0; 523 524 for(;;) { 525 static const struct option long_options[] = { 526 {"help", no_argument, 0, 'h'}, 527 {"object", required_argument, 0, OPTION_OBJECT}, 528 {0, 0, 0, 0} 529 }; 530 c = getopt_long(argc, argv, ":F:b:f:ho:qu", 531 long_options, NULL); 532 if (c == -1) { 533 break; 534 } 535 switch(c) { 536 case ':': 537 missing_argument(argv[optind - 1]); 538 break; 539 case '?': 540 unrecognized_option(argv[optind - 1]); 541 break; 542 case 'h': 543 help(); 544 break; 545 case 'F': 546 base_fmt = optarg; 547 break; 548 case 'b': 549 base_filename = optarg; 550 break; 551 case 'f': 552 fmt = optarg; 553 break; 554 case 'o': 555 if (accumulate_options(&options, optarg) < 0) { 556 goto fail; 557 } 558 break; 559 case 'q': 560 quiet = true; 561 break; 562 case 'u': 563 flags |= BDRV_O_NO_BACKING; 564 break; 565 case OPTION_OBJECT: 566 user_creatable_process_cmdline(optarg); 567 break; 568 } 569 } 570 571 /* Get the filename */ 572 filename = (optind < argc) ? argv[optind] : NULL; 573 if (options && has_help_option(options)) { 574 g_free(options); 575 return print_block_option_help(filename, fmt); 576 } 577 578 if (optind >= argc) { 579 error_exit("Expecting image file name"); 580 } 581 optind++; 582 583 /* Get image size, if specified */ 584 if (optind < argc) { 585 int64_t sval; 586 587 sval = cvtnum("image size", argv[optind++]); 588 if (sval < 0) { 589 goto fail; 590 } 591 img_size = (uint64_t)sval; 592 } 593 if (optind != argc) { 594 error_exit("Unexpected argument: %s", argv[optind]); 595 } 596 597 bdrv_img_create(filename, fmt, base_filename, base_fmt, 598 options, img_size, flags, quiet, &local_err); 599 if (local_err) { 600 error_reportf_err(local_err, "%s: ", filename); 601 goto fail; 602 } 603 604 g_free(options); 605 return 0; 606 607 fail: 608 g_free(options); 609 return 1; 610 } 611 612 static void dump_json_image_check(ImageCheck *check, bool quiet) 613 { 614 GString *str; 615 QObject *obj; 616 Visitor *v = qobject_output_visitor_new(&obj); 617 618 visit_type_ImageCheck(v, NULL, &check, &error_abort); 619 visit_complete(v, &obj); 620 str = qobject_to_json_pretty(obj, true); 621 assert(str != NULL); 622 qprintf(quiet, "%s\n", str->str); 623 qobject_unref(obj); 624 visit_free(v); 625 g_string_free(str, true); 626 } 627 628 static void dump_human_image_check(ImageCheck *check, bool quiet) 629 { 630 if (!(check->corruptions || check->leaks || check->check_errors)) { 631 qprintf(quiet, "No errors were found on the image.\n"); 632 } else { 633 if (check->corruptions) { 634 qprintf(quiet, "\n%" PRId64 " errors were found on the image.\n" 635 "Data may be corrupted, or further writes to the image " 636 "may corrupt it.\n", 637 check->corruptions); 638 } 639 640 if (check->leaks) { 641 qprintf(quiet, 642 "\n%" PRId64 " leaked clusters were found on the image.\n" 643 "This means waste of disk space, but no harm to data.\n", 644 check->leaks); 645 } 646 647 if (check->check_errors) { 648 qprintf(quiet, 649 "\n%" PRId64 650 " internal errors have occurred during the check.\n", 651 check->check_errors); 652 } 653 } 654 655 if (check->total_clusters != 0 && check->allocated_clusters != 0) { 656 qprintf(quiet, "%" PRId64 "/%" PRId64 " = %0.2f%% allocated, " 657 "%0.2f%% fragmented, %0.2f%% compressed clusters\n", 658 check->allocated_clusters, check->total_clusters, 659 check->allocated_clusters * 100.0 / check->total_clusters, 660 check->fragmented_clusters * 100.0 / check->allocated_clusters, 661 check->compressed_clusters * 100.0 / 662 check->allocated_clusters); 663 } 664 665 if (check->image_end_offset) { 666 qprintf(quiet, 667 "Image end offset: %" PRId64 "\n", check->image_end_offset); 668 } 669 } 670 671 static int collect_image_check(BlockDriverState *bs, 672 ImageCheck *check, 673 const char *filename, 674 const char *fmt, 675 int fix) 676 { 677 int ret; 678 BdrvCheckResult result; 679 680 ret = bdrv_check(bs, &result, fix); 681 if (ret < 0) { 682 return ret; 683 } 684 685 check->filename = g_strdup(filename); 686 check->format = g_strdup(bdrv_get_format_name(bs)); 687 check->check_errors = result.check_errors; 688 check->corruptions = result.corruptions; 689 check->has_corruptions = result.corruptions != 0; 690 check->leaks = result.leaks; 691 check->has_leaks = result.leaks != 0; 692 check->corruptions_fixed = result.corruptions_fixed; 693 check->has_corruptions_fixed = result.corruptions_fixed != 0; 694 check->leaks_fixed = result.leaks_fixed; 695 check->has_leaks_fixed = result.leaks_fixed != 0; 696 check->image_end_offset = result.image_end_offset; 697 check->has_image_end_offset = result.image_end_offset != 0; 698 check->total_clusters = result.bfi.total_clusters; 699 check->has_total_clusters = result.bfi.total_clusters != 0; 700 check->allocated_clusters = result.bfi.allocated_clusters; 701 check->has_allocated_clusters = result.bfi.allocated_clusters != 0; 702 check->fragmented_clusters = result.bfi.fragmented_clusters; 703 check->has_fragmented_clusters = result.bfi.fragmented_clusters != 0; 704 check->compressed_clusters = result.bfi.compressed_clusters; 705 check->has_compressed_clusters = result.bfi.compressed_clusters != 0; 706 707 return 0; 708 } 709 710 /* 711 * Checks an image for consistency. Exit codes: 712 * 713 * 0 - Check completed, image is good 714 * 1 - Check not completed because of internal errors 715 * 2 - Check completed, image is corrupted 716 * 3 - Check completed, image has leaked clusters, but is good otherwise 717 * 63 - Checks are not supported by the image format 718 */ 719 static int img_check(int argc, char **argv) 720 { 721 int c, ret; 722 OutputFormat output_format = OFORMAT_HUMAN; 723 const char *filename, *fmt, *output, *cache; 724 BlockBackend *blk; 725 BlockDriverState *bs; 726 int fix = 0; 727 int flags = BDRV_O_CHECK; 728 bool writethrough; 729 ImageCheck *check; 730 bool quiet = false; 731 bool image_opts = false; 732 bool force_share = false; 733 734 fmt = NULL; 735 output = NULL; 736 cache = BDRV_DEFAULT_CACHE; 737 738 for(;;) { 739 int option_index = 0; 740 static const struct option long_options[] = { 741 {"help", no_argument, 0, 'h'}, 742 {"format", required_argument, 0, 'f'}, 743 {"repair", required_argument, 0, 'r'}, 744 {"output", required_argument, 0, OPTION_OUTPUT}, 745 {"object", required_argument, 0, OPTION_OBJECT}, 746 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS}, 747 {"force-share", no_argument, 0, 'U'}, 748 {0, 0, 0, 0} 749 }; 750 c = getopt_long(argc, argv, ":hf:r:T:qU", 751 long_options, &option_index); 752 if (c == -1) { 753 break; 754 } 755 switch(c) { 756 case ':': 757 missing_argument(argv[optind - 1]); 758 break; 759 case '?': 760 unrecognized_option(argv[optind - 1]); 761 break; 762 case 'h': 763 help(); 764 break; 765 case 'f': 766 fmt = optarg; 767 break; 768 case 'r': 769 flags |= BDRV_O_RDWR; 770 771 if (!strcmp(optarg, "leaks")) { 772 fix = BDRV_FIX_LEAKS; 773 } else if (!strcmp(optarg, "all")) { 774 fix = BDRV_FIX_LEAKS | BDRV_FIX_ERRORS; 775 } else { 776 error_exit("Unknown option value for -r " 777 "(expecting 'leaks' or 'all'): %s", optarg); 778 } 779 break; 780 case OPTION_OUTPUT: 781 output = optarg; 782 break; 783 case 'T': 784 cache = optarg; 785 break; 786 case 'q': 787 quiet = true; 788 break; 789 case 'U': 790 force_share = true; 791 break; 792 case OPTION_OBJECT: 793 user_creatable_process_cmdline(optarg); 794 break; 795 case OPTION_IMAGE_OPTS: 796 image_opts = true; 797 break; 798 } 799 } 800 if (optind != argc - 1) { 801 error_exit("Expecting one image file name"); 802 } 803 filename = argv[optind++]; 804 805 if (output && !strcmp(output, "json")) { 806 output_format = OFORMAT_JSON; 807 } else if (output && !strcmp(output, "human")) { 808 output_format = OFORMAT_HUMAN; 809 } else if (output) { 810 error_report("--output must be used with human or json as argument."); 811 return 1; 812 } 813 814 ret = bdrv_parse_cache_mode(cache, &flags, &writethrough); 815 if (ret < 0) { 816 error_report("Invalid source cache option: %s", cache); 817 return 1; 818 } 819 820 blk = img_open(image_opts, filename, fmt, flags, writethrough, quiet, 821 force_share); 822 if (!blk) { 823 return 1; 824 } 825 bs = blk_bs(blk); 826 827 check = g_new0(ImageCheck, 1); 828 ret = collect_image_check(bs, check, filename, fmt, fix); 829 830 if (ret == -ENOTSUP) { 831 error_report("This image format does not support checks"); 832 ret = 63; 833 goto fail; 834 } 835 836 if (check->corruptions_fixed || check->leaks_fixed) { 837 int corruptions_fixed, leaks_fixed; 838 bool has_leaks_fixed, has_corruptions_fixed; 839 840 leaks_fixed = check->leaks_fixed; 841 has_leaks_fixed = check->has_leaks_fixed; 842 corruptions_fixed = check->corruptions_fixed; 843 has_corruptions_fixed = check->has_corruptions_fixed; 844 845 if (output_format == OFORMAT_HUMAN) { 846 qprintf(quiet, 847 "The following inconsistencies were found and repaired:\n\n" 848 " %" PRId64 " leaked clusters\n" 849 " %" PRId64 " corruptions\n\n" 850 "Double checking the fixed image now...\n", 851 check->leaks_fixed, 852 check->corruptions_fixed); 853 } 854 855 qapi_free_ImageCheck(check); 856 check = g_new0(ImageCheck, 1); 857 ret = collect_image_check(bs, check, filename, fmt, 0); 858 859 check->leaks_fixed = leaks_fixed; 860 check->has_leaks_fixed = has_leaks_fixed; 861 check->corruptions_fixed = corruptions_fixed; 862 check->has_corruptions_fixed = has_corruptions_fixed; 863 } 864 865 if (!ret) { 866 switch (output_format) { 867 case OFORMAT_HUMAN: 868 dump_human_image_check(check, quiet); 869 break; 870 case OFORMAT_JSON: 871 dump_json_image_check(check, quiet); 872 break; 873 } 874 } 875 876 if (ret || check->check_errors) { 877 if (ret) { 878 error_report("Check failed: %s", strerror(-ret)); 879 } else { 880 error_report("Check failed"); 881 } 882 ret = 1; 883 goto fail; 884 } 885 886 if (check->corruptions) { 887 ret = 2; 888 } else if (check->leaks) { 889 ret = 3; 890 } else { 891 ret = 0; 892 } 893 894 fail: 895 qapi_free_ImageCheck(check); 896 blk_unref(blk); 897 return ret; 898 } 899 900 typedef struct CommonBlockJobCBInfo { 901 BlockDriverState *bs; 902 Error **errp; 903 } CommonBlockJobCBInfo; 904 905 static void common_block_job_cb(void *opaque, int ret) 906 { 907 CommonBlockJobCBInfo *cbi = opaque; 908 909 if (ret < 0) { 910 error_setg_errno(cbi->errp, -ret, "Block job failed"); 911 } 912 } 913 914 static void run_block_job(BlockJob *job, Error **errp) 915 { 916 uint64_t progress_current, progress_total; 917 AioContext *aio_context = block_job_get_aio_context(job); 918 int ret = 0; 919 920 job_lock(); 921 job_ref_locked(&job->job); 922 do { 923 float progress = 0.0f; 924 job_unlock(); 925 aio_poll(aio_context, true); 926 927 progress_get_snapshot(&job->job.progress, &progress_current, 928 &progress_total); 929 if (progress_total) { 930 progress = (float)progress_current / progress_total * 100.f; 931 } 932 qemu_progress_print(progress, 0); 933 job_lock(); 934 } while (!job_is_ready_locked(&job->job) && 935 !job_is_completed_locked(&job->job)); 936 937 if (!job_is_completed_locked(&job->job)) { 938 ret = job_complete_sync_locked(&job->job, errp); 939 } else { 940 ret = job->job.ret; 941 } 942 job_unref_locked(&job->job); 943 job_unlock(); 944 945 /* publish completion progress only when success */ 946 if (!ret) { 947 qemu_progress_print(100.f, 0); 948 } 949 } 950 951 static int img_commit(int argc, char **argv) 952 { 953 int c, ret, flags; 954 const char *filename, *fmt, *cache, *base; 955 BlockBackend *blk; 956 BlockDriverState *bs, *base_bs; 957 BlockJob *job; 958 bool progress = false, quiet = false, drop = false; 959 bool writethrough; 960 Error *local_err = NULL; 961 CommonBlockJobCBInfo cbi; 962 bool image_opts = false; 963 AioContext *aio_context; 964 int64_t rate_limit = 0; 965 966 fmt = NULL; 967 cache = BDRV_DEFAULT_CACHE; 968 base = NULL; 969 for(;;) { 970 static const struct option long_options[] = { 971 {"help", no_argument, 0, 'h'}, 972 {"object", required_argument, 0, OPTION_OBJECT}, 973 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS}, 974 {0, 0, 0, 0} 975 }; 976 c = getopt_long(argc, argv, ":f:ht:b:dpqr:", 977 long_options, NULL); 978 if (c == -1) { 979 break; 980 } 981 switch(c) { 982 case ':': 983 missing_argument(argv[optind - 1]); 984 break; 985 case '?': 986 unrecognized_option(argv[optind - 1]); 987 break; 988 case 'h': 989 help(); 990 break; 991 case 'f': 992 fmt = optarg; 993 break; 994 case 't': 995 cache = optarg; 996 break; 997 case 'b': 998 base = optarg; 999 /* -b implies -d */ 1000 drop = true; 1001 break; 1002 case 'd': 1003 drop = true; 1004 break; 1005 case 'p': 1006 progress = true; 1007 break; 1008 case 'q': 1009 quiet = true; 1010 break; 1011 case 'r': 1012 rate_limit = cvtnum("rate limit", optarg); 1013 if (rate_limit < 0) { 1014 return 1; 1015 } 1016 break; 1017 case OPTION_OBJECT: 1018 user_creatable_process_cmdline(optarg); 1019 break; 1020 case OPTION_IMAGE_OPTS: 1021 image_opts = true; 1022 break; 1023 } 1024 } 1025 1026 /* Progress is not shown in Quiet mode */ 1027 if (quiet) { 1028 progress = false; 1029 } 1030 1031 if (optind != argc - 1) { 1032 error_exit("Expecting one image file name"); 1033 } 1034 filename = argv[optind++]; 1035 1036 flags = BDRV_O_RDWR | BDRV_O_UNMAP; 1037 ret = bdrv_parse_cache_mode(cache, &flags, &writethrough); 1038 if (ret < 0) { 1039 error_report("Invalid cache option: %s", cache); 1040 return 1; 1041 } 1042 1043 blk = img_open(image_opts, filename, fmt, flags, writethrough, quiet, 1044 false); 1045 if (!blk) { 1046 return 1; 1047 } 1048 bs = blk_bs(blk); 1049 1050 qemu_progress_init(progress, 1.f); 1051 qemu_progress_print(0.f, 100); 1052 1053 if (base) { 1054 base_bs = bdrv_find_backing_image(bs, base); 1055 if (!base_bs) { 1056 error_setg(&local_err, 1057 "Did not find '%s' in the backing chain of '%s'", 1058 base, filename); 1059 goto done; 1060 } 1061 } else { 1062 /* This is different from QMP, which by default uses the deepest file in 1063 * the backing chain (i.e., the very base); however, the traditional 1064 * behavior of qemu-img commit is using the immediate backing file. */ 1065 base_bs = bdrv_backing_chain_next(bs); 1066 if (!base_bs) { 1067 error_setg(&local_err, "Image does not have a backing file"); 1068 goto done; 1069 } 1070 } 1071 1072 cbi = (CommonBlockJobCBInfo){ 1073 .errp = &local_err, 1074 .bs = bs, 1075 }; 1076 1077 aio_context = bdrv_get_aio_context(bs); 1078 aio_context_acquire(aio_context); 1079 commit_active_start("commit", bs, base_bs, JOB_DEFAULT, rate_limit, 1080 BLOCKDEV_ON_ERROR_REPORT, NULL, common_block_job_cb, 1081 &cbi, false, &local_err); 1082 aio_context_release(aio_context); 1083 if (local_err) { 1084 goto done; 1085 } 1086 1087 /* When the block job completes, the BlockBackend reference will point to 1088 * the old backing file. In order to avoid that the top image is already 1089 * deleted, so we can still empty it afterwards, increment the reference 1090 * counter here preemptively. */ 1091 if (!drop) { 1092 bdrv_ref(bs); 1093 } 1094 1095 job = block_job_get("commit"); 1096 assert(job); 1097 run_block_job(job, &local_err); 1098 if (local_err) { 1099 goto unref_backing; 1100 } 1101 1102 if (!drop) { 1103 BlockBackend *old_backing_blk; 1104 1105 old_backing_blk = blk_new_with_bs(bs, BLK_PERM_WRITE, BLK_PERM_ALL, 1106 &local_err); 1107 if (!old_backing_blk) { 1108 goto unref_backing; 1109 } 1110 ret = blk_make_empty(old_backing_blk, &local_err); 1111 blk_unref(old_backing_blk); 1112 if (ret == -ENOTSUP) { 1113 error_free(local_err); 1114 local_err = NULL; 1115 } else if (ret < 0) { 1116 goto unref_backing; 1117 } 1118 } 1119 1120 unref_backing: 1121 if (!drop) { 1122 bdrv_unref(bs); 1123 } 1124 1125 done: 1126 qemu_progress_end(); 1127 1128 /* 1129 * Manually inactivate the image first because this way we can know whether 1130 * an error occurred. blk_unref() doesn't tell us about failures. 1131 */ 1132 ret = bdrv_inactivate_all(); 1133 if (ret < 0 && !local_err) { 1134 error_setg_errno(&local_err, -ret, "Error while closing the image"); 1135 } 1136 blk_unref(blk); 1137 1138 if (local_err) { 1139 error_report_err(local_err); 1140 return 1; 1141 } 1142 1143 qprintf(quiet, "Image committed.\n"); 1144 return 0; 1145 } 1146 1147 /* 1148 * Returns -1 if 'buf' contains only zeroes, otherwise the byte index 1149 * of the first sector boundary within buf where the sector contains a 1150 * non-zero byte. This function is robust to a buffer that is not 1151 * sector-aligned. 1152 */ 1153 static int64_t find_nonzero(const uint8_t *buf, int64_t n) 1154 { 1155 int64_t i; 1156 int64_t end = QEMU_ALIGN_DOWN(n, BDRV_SECTOR_SIZE); 1157 1158 for (i = 0; i < end; i += BDRV_SECTOR_SIZE) { 1159 if (!buffer_is_zero(buf + i, BDRV_SECTOR_SIZE)) { 1160 return i; 1161 } 1162 } 1163 if (i < n && !buffer_is_zero(buf + i, n - end)) { 1164 return i; 1165 } 1166 return -1; 1167 } 1168 1169 /* 1170 * Returns true iff the first sector pointed to by 'buf' contains at least 1171 * a non-NUL byte. 1172 * 1173 * 'pnum' is set to the number of sectors (including and immediately following 1174 * the first one) that are known to be in the same allocated/unallocated state. 1175 * The function will try to align the end offset to alignment boundaries so 1176 * that the request will at least end aligned and consecutive requests will 1177 * also start at an aligned offset. 1178 */ 1179 static int is_allocated_sectors(const uint8_t *buf, int n, int *pnum, 1180 int64_t sector_num, int alignment) 1181 { 1182 bool is_zero; 1183 int i, tail; 1184 1185 if (n <= 0) { 1186 *pnum = 0; 1187 return 0; 1188 } 1189 is_zero = buffer_is_zero(buf, BDRV_SECTOR_SIZE); 1190 for(i = 1; i < n; i++) { 1191 buf += BDRV_SECTOR_SIZE; 1192 if (is_zero != buffer_is_zero(buf, BDRV_SECTOR_SIZE)) { 1193 break; 1194 } 1195 } 1196 1197 if (i == n) { 1198 /* 1199 * The whole buf is the same. 1200 * No reason to split it into chunks, so return now. 1201 */ 1202 *pnum = i; 1203 return !is_zero; 1204 } 1205 1206 tail = (sector_num + i) & (alignment - 1); 1207 if (tail) { 1208 if (is_zero && i <= tail) { 1209 /* 1210 * For sure next sector after i is data, and it will rewrite this 1211 * tail anyway due to RMW. So, let's just write data now. 1212 */ 1213 is_zero = false; 1214 } 1215 if (!is_zero) { 1216 /* If possible, align up end offset of allocated areas. */ 1217 i += alignment - tail; 1218 i = MIN(i, n); 1219 } else { 1220 /* 1221 * For sure next sector after i is data, and it will rewrite this 1222 * tail anyway due to RMW. Better is avoid RMW and write zeroes up 1223 * to aligned bound. 1224 */ 1225 i -= tail; 1226 } 1227 } 1228 *pnum = i; 1229 return !is_zero; 1230 } 1231 1232 /* 1233 * Like is_allocated_sectors, but if the buffer starts with a used sector, 1234 * up to 'min' consecutive sectors containing zeros are ignored. This avoids 1235 * breaking up write requests for only small sparse areas. 1236 */ 1237 static int is_allocated_sectors_min(const uint8_t *buf, int n, int *pnum, 1238 int min, int64_t sector_num, int alignment) 1239 { 1240 int ret; 1241 int num_checked, num_used; 1242 1243 if (n < min) { 1244 min = n; 1245 } 1246 1247 ret = is_allocated_sectors(buf, n, pnum, sector_num, alignment); 1248 if (!ret) { 1249 return ret; 1250 } 1251 1252 num_used = *pnum; 1253 buf += BDRV_SECTOR_SIZE * *pnum; 1254 n -= *pnum; 1255 sector_num += *pnum; 1256 num_checked = num_used; 1257 1258 while (n > 0) { 1259 ret = is_allocated_sectors(buf, n, pnum, sector_num, alignment); 1260 1261 buf += BDRV_SECTOR_SIZE * *pnum; 1262 n -= *pnum; 1263 sector_num += *pnum; 1264 num_checked += *pnum; 1265 if (ret) { 1266 num_used = num_checked; 1267 } else if (*pnum >= min) { 1268 break; 1269 } 1270 } 1271 1272 *pnum = num_used; 1273 return 1; 1274 } 1275 1276 /* 1277 * Compares two buffers chunk by chunk, where @chsize is the chunk size. 1278 * If @chsize is 0, default chunk size of BDRV_SECTOR_SIZE is used. 1279 * Returns 0 if the first chunk of each buffer matches, non-zero otherwise. 1280 * 1281 * @pnum is set to the size of the buffer prefix aligned to @chsize that 1282 * has the same matching status as the first chunk. 1283 */ 1284 static int compare_buffers(const uint8_t *buf1, const uint8_t *buf2, 1285 int64_t bytes, uint64_t chsize, int64_t *pnum) 1286 { 1287 bool res; 1288 int64_t i; 1289 1290 assert(bytes > 0); 1291 1292 if (!chsize) { 1293 chsize = BDRV_SECTOR_SIZE; 1294 } 1295 i = MIN(bytes, chsize); 1296 1297 res = !!memcmp(buf1, buf2, i); 1298 while (i < bytes) { 1299 int64_t len = MIN(bytes - i, chsize); 1300 1301 if (!!memcmp(buf1 + i, buf2 + i, len) != res) { 1302 break; 1303 } 1304 i += len; 1305 } 1306 1307 *pnum = i; 1308 return res; 1309 } 1310 1311 #define IO_BUF_SIZE (2 * MiB) 1312 1313 /* 1314 * Check if passed sectors are empty (not allocated or contain only 0 bytes) 1315 * 1316 * Intended for use by 'qemu-img compare': Returns 0 in case sectors are 1317 * filled with 0, 1 if sectors contain non-zero data (this is a comparison 1318 * failure), and 4 on error (the exit status for read errors), after emitting 1319 * an error message. 1320 * 1321 * @param blk: BlockBackend for the image 1322 * @param offset: Starting offset to check 1323 * @param bytes: Number of bytes to check 1324 * @param filename: Name of disk file we are checking (logging purpose) 1325 * @param buffer: Allocated buffer for storing read data 1326 * @param quiet: Flag for quiet mode 1327 */ 1328 static int check_empty_sectors(BlockBackend *blk, int64_t offset, 1329 int64_t bytes, const char *filename, 1330 uint8_t *buffer, bool quiet) 1331 { 1332 int ret = 0; 1333 int64_t idx; 1334 1335 ret = blk_pread(blk, offset, bytes, buffer, 0); 1336 if (ret < 0) { 1337 error_report("Error while reading offset %" PRId64 " of %s: %s", 1338 offset, filename, strerror(-ret)); 1339 return 4; 1340 } 1341 idx = find_nonzero(buffer, bytes); 1342 if (idx >= 0) { 1343 qprintf(quiet, "Content mismatch at offset %" PRId64 "!\n", 1344 offset + idx); 1345 return 1; 1346 } 1347 1348 return 0; 1349 } 1350 1351 /* 1352 * Compares two images. Exit codes: 1353 * 1354 * 0 - Images are identical or the requested help was printed 1355 * 1 - Images differ 1356 * >1 - Error occurred 1357 */ 1358 static int img_compare(int argc, char **argv) 1359 { 1360 const char *fmt1 = NULL, *fmt2 = NULL, *cache, *filename1, *filename2; 1361 BlockBackend *blk1, *blk2; 1362 BlockDriverState *bs1, *bs2; 1363 int64_t total_size1, total_size2; 1364 uint8_t *buf1 = NULL, *buf2 = NULL; 1365 int64_t pnum1, pnum2; 1366 int allocated1, allocated2; 1367 int ret = 0; /* return value - 0 Ident, 1 Different, >1 Error */ 1368 bool progress = false, quiet = false, strict = false; 1369 int flags; 1370 bool writethrough; 1371 int64_t total_size; 1372 int64_t offset = 0; 1373 int64_t chunk; 1374 int c; 1375 uint64_t progress_base; 1376 bool image_opts = false; 1377 bool force_share = false; 1378 1379 cache = BDRV_DEFAULT_CACHE; 1380 for (;;) { 1381 static const struct option long_options[] = { 1382 {"help", no_argument, 0, 'h'}, 1383 {"object", required_argument, 0, OPTION_OBJECT}, 1384 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS}, 1385 {"force-share", no_argument, 0, 'U'}, 1386 {0, 0, 0, 0} 1387 }; 1388 c = getopt_long(argc, argv, ":hf:F:T:pqsU", 1389 long_options, NULL); 1390 if (c == -1) { 1391 break; 1392 } 1393 switch (c) { 1394 case ':': 1395 missing_argument(argv[optind - 1]); 1396 break; 1397 case '?': 1398 unrecognized_option(argv[optind - 1]); 1399 break; 1400 case 'h': 1401 help(); 1402 break; 1403 case 'f': 1404 fmt1 = optarg; 1405 break; 1406 case 'F': 1407 fmt2 = optarg; 1408 break; 1409 case 'T': 1410 cache = optarg; 1411 break; 1412 case 'p': 1413 progress = true; 1414 break; 1415 case 'q': 1416 quiet = true; 1417 break; 1418 case 's': 1419 strict = true; 1420 break; 1421 case 'U': 1422 force_share = true; 1423 break; 1424 case OPTION_OBJECT: 1425 { 1426 Error *local_err = NULL; 1427 1428 if (!user_creatable_add_from_str(optarg, &local_err)) { 1429 if (local_err) { 1430 error_report_err(local_err); 1431 exit(2); 1432 } else { 1433 /* Help was printed */ 1434 exit(EXIT_SUCCESS); 1435 } 1436 } 1437 break; 1438 } 1439 case OPTION_IMAGE_OPTS: 1440 image_opts = true; 1441 break; 1442 } 1443 } 1444 1445 /* Progress is not shown in Quiet mode */ 1446 if (quiet) { 1447 progress = false; 1448 } 1449 1450 1451 if (optind != argc - 2) { 1452 error_exit("Expecting two image file names"); 1453 } 1454 filename1 = argv[optind++]; 1455 filename2 = argv[optind++]; 1456 1457 /* Initialize before goto out */ 1458 qemu_progress_init(progress, 2.0); 1459 1460 flags = 0; 1461 ret = bdrv_parse_cache_mode(cache, &flags, &writethrough); 1462 if (ret < 0) { 1463 error_report("Invalid source cache option: %s", cache); 1464 ret = 2; 1465 goto out3; 1466 } 1467 1468 blk1 = img_open(image_opts, filename1, fmt1, flags, writethrough, quiet, 1469 force_share); 1470 if (!blk1) { 1471 ret = 2; 1472 goto out3; 1473 } 1474 1475 blk2 = img_open(image_opts, filename2, fmt2, flags, writethrough, quiet, 1476 force_share); 1477 if (!blk2) { 1478 ret = 2; 1479 goto out2; 1480 } 1481 bs1 = blk_bs(blk1); 1482 bs2 = blk_bs(blk2); 1483 1484 buf1 = blk_blockalign(blk1, IO_BUF_SIZE); 1485 buf2 = blk_blockalign(blk2, IO_BUF_SIZE); 1486 total_size1 = blk_getlength(blk1); 1487 if (total_size1 < 0) { 1488 error_report("Can't get size of %s: %s", 1489 filename1, strerror(-total_size1)); 1490 ret = 4; 1491 goto out; 1492 } 1493 total_size2 = blk_getlength(blk2); 1494 if (total_size2 < 0) { 1495 error_report("Can't get size of %s: %s", 1496 filename2, strerror(-total_size2)); 1497 ret = 4; 1498 goto out; 1499 } 1500 total_size = MIN(total_size1, total_size2); 1501 progress_base = MAX(total_size1, total_size2); 1502 1503 qemu_progress_print(0, 100); 1504 1505 if (strict && total_size1 != total_size2) { 1506 ret = 1; 1507 qprintf(quiet, "Strict mode: Image size mismatch!\n"); 1508 goto out; 1509 } 1510 1511 while (offset < total_size) { 1512 int status1, status2; 1513 1514 status1 = bdrv_block_status_above(bs1, NULL, offset, 1515 total_size1 - offset, &pnum1, NULL, 1516 NULL); 1517 if (status1 < 0) { 1518 ret = 3; 1519 error_report("Sector allocation test failed for %s", filename1); 1520 goto out; 1521 } 1522 allocated1 = status1 & BDRV_BLOCK_ALLOCATED; 1523 1524 status2 = bdrv_block_status_above(bs2, NULL, offset, 1525 total_size2 - offset, &pnum2, NULL, 1526 NULL); 1527 if (status2 < 0) { 1528 ret = 3; 1529 error_report("Sector allocation test failed for %s", filename2); 1530 goto out; 1531 } 1532 allocated2 = status2 & BDRV_BLOCK_ALLOCATED; 1533 1534 assert(pnum1 && pnum2); 1535 chunk = MIN(pnum1, pnum2); 1536 1537 if (strict) { 1538 if (status1 != status2) { 1539 ret = 1; 1540 qprintf(quiet, "Strict mode: Offset %" PRId64 1541 " block status mismatch!\n", offset); 1542 goto out; 1543 } 1544 } 1545 if ((status1 & BDRV_BLOCK_ZERO) && (status2 & BDRV_BLOCK_ZERO)) { 1546 /* nothing to do */ 1547 } else if (allocated1 == allocated2) { 1548 if (allocated1) { 1549 int64_t pnum; 1550 1551 chunk = MIN(chunk, IO_BUF_SIZE); 1552 ret = blk_pread(blk1, offset, chunk, buf1, 0); 1553 if (ret < 0) { 1554 error_report("Error while reading offset %" PRId64 1555 " of %s: %s", 1556 offset, filename1, strerror(-ret)); 1557 ret = 4; 1558 goto out; 1559 } 1560 ret = blk_pread(blk2, offset, chunk, buf2, 0); 1561 if (ret < 0) { 1562 error_report("Error while reading offset %" PRId64 1563 " of %s: %s", 1564 offset, filename2, strerror(-ret)); 1565 ret = 4; 1566 goto out; 1567 } 1568 ret = compare_buffers(buf1, buf2, chunk, 0, &pnum); 1569 if (ret || pnum != chunk) { 1570 qprintf(quiet, "Content mismatch at offset %" PRId64 "!\n", 1571 offset + (ret ? 0 : pnum)); 1572 ret = 1; 1573 goto out; 1574 } 1575 } 1576 } else { 1577 chunk = MIN(chunk, IO_BUF_SIZE); 1578 if (allocated1) { 1579 ret = check_empty_sectors(blk1, offset, chunk, 1580 filename1, buf1, quiet); 1581 } else { 1582 ret = check_empty_sectors(blk2, offset, chunk, 1583 filename2, buf1, quiet); 1584 } 1585 if (ret) { 1586 goto out; 1587 } 1588 } 1589 offset += chunk; 1590 qemu_progress_print(((float) chunk / progress_base) * 100, 100); 1591 } 1592 1593 if (total_size1 != total_size2) { 1594 BlockBackend *blk_over; 1595 const char *filename_over; 1596 1597 qprintf(quiet, "Warning: Image size mismatch!\n"); 1598 if (total_size1 > total_size2) { 1599 blk_over = blk1; 1600 filename_over = filename1; 1601 } else { 1602 blk_over = blk2; 1603 filename_over = filename2; 1604 } 1605 1606 while (offset < progress_base) { 1607 ret = bdrv_block_status_above(blk_bs(blk_over), NULL, offset, 1608 progress_base - offset, &chunk, 1609 NULL, NULL); 1610 if (ret < 0) { 1611 ret = 3; 1612 error_report("Sector allocation test failed for %s", 1613 filename_over); 1614 goto out; 1615 1616 } 1617 if (ret & BDRV_BLOCK_ALLOCATED && !(ret & BDRV_BLOCK_ZERO)) { 1618 chunk = MIN(chunk, IO_BUF_SIZE); 1619 ret = check_empty_sectors(blk_over, offset, chunk, 1620 filename_over, buf1, quiet); 1621 if (ret) { 1622 goto out; 1623 } 1624 } 1625 offset += chunk; 1626 qemu_progress_print(((float) chunk / progress_base) * 100, 100); 1627 } 1628 } 1629 1630 qprintf(quiet, "Images are identical.\n"); 1631 ret = 0; 1632 1633 out: 1634 qemu_vfree(buf1); 1635 qemu_vfree(buf2); 1636 blk_unref(blk2); 1637 out2: 1638 blk_unref(blk1); 1639 out3: 1640 qemu_progress_end(); 1641 return ret; 1642 } 1643 1644 /* Convenience wrapper around qmp_block_dirty_bitmap_merge */ 1645 static void do_dirty_bitmap_merge(const char *dst_node, const char *dst_name, 1646 const char *src_node, const char *src_name, 1647 Error **errp) 1648 { 1649 BlockDirtyBitmapOrStr *merge_src; 1650 BlockDirtyBitmapOrStrList *list = NULL; 1651 1652 merge_src = g_new0(BlockDirtyBitmapOrStr, 1); 1653 merge_src->type = QTYPE_QDICT; 1654 merge_src->u.external.node = g_strdup(src_node); 1655 merge_src->u.external.name = g_strdup(src_name); 1656 QAPI_LIST_PREPEND(list, merge_src); 1657 qmp_block_dirty_bitmap_merge(dst_node, dst_name, list, errp); 1658 qapi_free_BlockDirtyBitmapOrStrList(list); 1659 } 1660 1661 enum ImgConvertBlockStatus { 1662 BLK_DATA, 1663 BLK_ZERO, 1664 BLK_BACKING_FILE, 1665 }; 1666 1667 #define MAX_COROUTINES 16 1668 #define CONVERT_THROTTLE_GROUP "img_convert" 1669 1670 typedef struct ImgConvertState { 1671 BlockBackend **src; 1672 int64_t *src_sectors; 1673 int *src_alignment; 1674 int src_num; 1675 int64_t total_sectors; 1676 int64_t allocated_sectors; 1677 int64_t allocated_done; 1678 int64_t sector_num; 1679 int64_t wr_offs; 1680 enum ImgConvertBlockStatus status; 1681 int64_t sector_next_status; 1682 BlockBackend *target; 1683 bool has_zero_init; 1684 bool compressed; 1685 bool target_is_new; 1686 bool target_has_backing; 1687 int64_t target_backing_sectors; /* negative if unknown */ 1688 bool wr_in_order; 1689 bool copy_range; 1690 bool salvage; 1691 bool quiet; 1692 int min_sparse; 1693 int alignment; 1694 size_t cluster_sectors; 1695 size_t buf_sectors; 1696 long num_coroutines; 1697 int running_coroutines; 1698 Coroutine *co[MAX_COROUTINES]; 1699 int64_t wait_sector_num[MAX_COROUTINES]; 1700 CoMutex lock; 1701 int ret; 1702 } ImgConvertState; 1703 1704 static void convert_select_part(ImgConvertState *s, int64_t sector_num, 1705 int *src_cur, int64_t *src_cur_offset) 1706 { 1707 *src_cur = 0; 1708 *src_cur_offset = 0; 1709 while (sector_num - *src_cur_offset >= s->src_sectors[*src_cur]) { 1710 *src_cur_offset += s->src_sectors[*src_cur]; 1711 (*src_cur)++; 1712 assert(*src_cur < s->src_num); 1713 } 1714 } 1715 1716 static int convert_iteration_sectors(ImgConvertState *s, int64_t sector_num) 1717 { 1718 int64_t src_cur_offset; 1719 int ret, n, src_cur; 1720 bool post_backing_zero = false; 1721 1722 convert_select_part(s, sector_num, &src_cur, &src_cur_offset); 1723 1724 assert(s->total_sectors > sector_num); 1725 n = MIN(s->total_sectors - sector_num, BDRV_REQUEST_MAX_SECTORS); 1726 1727 if (s->target_backing_sectors >= 0) { 1728 if (sector_num >= s->target_backing_sectors) { 1729 post_backing_zero = true; 1730 } else if (sector_num + n > s->target_backing_sectors) { 1731 /* Split requests around target_backing_sectors (because 1732 * starting from there, zeros are handled differently) */ 1733 n = s->target_backing_sectors - sector_num; 1734 } 1735 } 1736 1737 if (s->sector_next_status <= sector_num) { 1738 uint64_t offset = (sector_num - src_cur_offset) * BDRV_SECTOR_SIZE; 1739 int64_t count; 1740 int tail; 1741 BlockDriverState *src_bs = blk_bs(s->src[src_cur]); 1742 BlockDriverState *base; 1743 1744 if (s->target_has_backing) { 1745 base = bdrv_cow_bs(bdrv_skip_filters(src_bs)); 1746 } else { 1747 base = NULL; 1748 } 1749 1750 do { 1751 count = n * BDRV_SECTOR_SIZE; 1752 1753 ret = bdrv_block_status_above(src_bs, base, offset, count, &count, 1754 NULL, NULL); 1755 1756 if (ret < 0) { 1757 if (s->salvage) { 1758 if (n == 1) { 1759 if (!s->quiet) { 1760 warn_report("error while reading block status at " 1761 "offset %" PRIu64 ": %s", offset, 1762 strerror(-ret)); 1763 } 1764 /* Just try to read the data, then */ 1765 ret = BDRV_BLOCK_DATA; 1766 count = BDRV_SECTOR_SIZE; 1767 } else { 1768 /* Retry on a shorter range */ 1769 n = DIV_ROUND_UP(n, 4); 1770 } 1771 } else { 1772 error_report("error while reading block status at offset " 1773 "%" PRIu64 ": %s", offset, strerror(-ret)); 1774 return ret; 1775 } 1776 } 1777 } while (ret < 0); 1778 1779 n = DIV_ROUND_UP(count, BDRV_SECTOR_SIZE); 1780 1781 /* 1782 * Avoid that s->sector_next_status becomes unaligned to the source 1783 * request alignment and/or cluster size to avoid unnecessary read 1784 * cycles. 1785 */ 1786 tail = (sector_num - src_cur_offset + n) % s->src_alignment[src_cur]; 1787 if (n > tail) { 1788 n -= tail; 1789 } 1790 1791 if (ret & BDRV_BLOCK_ZERO) { 1792 s->status = post_backing_zero ? BLK_BACKING_FILE : BLK_ZERO; 1793 } else if (ret & BDRV_BLOCK_DATA) { 1794 s->status = BLK_DATA; 1795 } else { 1796 s->status = s->target_has_backing ? BLK_BACKING_FILE : BLK_DATA; 1797 } 1798 1799 s->sector_next_status = sector_num + n; 1800 } 1801 1802 n = MIN(n, s->sector_next_status - sector_num); 1803 if (s->status == BLK_DATA) { 1804 n = MIN(n, s->buf_sectors); 1805 } 1806 1807 /* We need to write complete clusters for compressed images, so if an 1808 * unallocated area is shorter than that, we must consider the whole 1809 * cluster allocated. */ 1810 if (s->compressed) { 1811 if (n < s->cluster_sectors) { 1812 n = MIN(s->cluster_sectors, s->total_sectors - sector_num); 1813 s->status = BLK_DATA; 1814 } else { 1815 n = QEMU_ALIGN_DOWN(n, s->cluster_sectors); 1816 } 1817 } 1818 1819 return n; 1820 } 1821 1822 static int coroutine_fn convert_co_read(ImgConvertState *s, int64_t sector_num, 1823 int nb_sectors, uint8_t *buf) 1824 { 1825 uint64_t single_read_until = 0; 1826 int n, ret; 1827 1828 assert(nb_sectors <= s->buf_sectors); 1829 while (nb_sectors > 0) { 1830 BlockBackend *blk; 1831 int src_cur; 1832 int64_t bs_sectors, src_cur_offset; 1833 uint64_t offset; 1834 1835 /* In the case of compression with multiple source files, we can get a 1836 * nb_sectors that spreads into the next part. So we must be able to 1837 * read across multiple BDSes for one convert_read() call. */ 1838 convert_select_part(s, sector_num, &src_cur, &src_cur_offset); 1839 blk = s->src[src_cur]; 1840 bs_sectors = s->src_sectors[src_cur]; 1841 1842 offset = (sector_num - src_cur_offset) << BDRV_SECTOR_BITS; 1843 1844 n = MIN(nb_sectors, bs_sectors - (sector_num - src_cur_offset)); 1845 if (single_read_until > offset) { 1846 n = 1; 1847 } 1848 1849 ret = blk_co_pread(blk, offset, n << BDRV_SECTOR_BITS, buf, 0); 1850 if (ret < 0) { 1851 if (s->salvage) { 1852 if (n > 1) { 1853 single_read_until = offset + (n << BDRV_SECTOR_BITS); 1854 continue; 1855 } else { 1856 if (!s->quiet) { 1857 warn_report("error while reading offset %" PRIu64 1858 ": %s", offset, strerror(-ret)); 1859 } 1860 memset(buf, 0, BDRV_SECTOR_SIZE); 1861 } 1862 } else { 1863 return ret; 1864 } 1865 } 1866 1867 sector_num += n; 1868 nb_sectors -= n; 1869 buf += n * BDRV_SECTOR_SIZE; 1870 } 1871 1872 return 0; 1873 } 1874 1875 1876 static int coroutine_fn convert_co_write(ImgConvertState *s, int64_t sector_num, 1877 int nb_sectors, uint8_t *buf, 1878 enum ImgConvertBlockStatus status) 1879 { 1880 int ret; 1881 1882 while (nb_sectors > 0) { 1883 int n = nb_sectors; 1884 BdrvRequestFlags flags = s->compressed ? BDRV_REQ_WRITE_COMPRESSED : 0; 1885 1886 switch (status) { 1887 case BLK_BACKING_FILE: 1888 /* If we have a backing file, leave clusters unallocated that are 1889 * unallocated in the source image, so that the backing file is 1890 * visible at the respective offset. */ 1891 assert(s->target_has_backing); 1892 break; 1893 1894 case BLK_DATA: 1895 /* If we're told to keep the target fully allocated (-S 0) or there 1896 * is real non-zero data, we must write it. Otherwise we can treat 1897 * it as zero sectors. 1898 * Compressed clusters need to be written as a whole, so in that 1899 * case we can only save the write if the buffer is completely 1900 * zeroed. */ 1901 if (!s->min_sparse || 1902 (!s->compressed && 1903 is_allocated_sectors_min(buf, n, &n, s->min_sparse, 1904 sector_num, s->alignment)) || 1905 (s->compressed && 1906 !buffer_is_zero(buf, n * BDRV_SECTOR_SIZE))) 1907 { 1908 ret = blk_co_pwrite(s->target, sector_num << BDRV_SECTOR_BITS, 1909 n << BDRV_SECTOR_BITS, buf, flags); 1910 if (ret < 0) { 1911 return ret; 1912 } 1913 break; 1914 } 1915 /* fall-through */ 1916 1917 case BLK_ZERO: 1918 if (s->has_zero_init) { 1919 assert(!s->target_has_backing); 1920 break; 1921 } 1922 ret = blk_co_pwrite_zeroes(s->target, 1923 sector_num << BDRV_SECTOR_BITS, 1924 n << BDRV_SECTOR_BITS, 1925 BDRV_REQ_MAY_UNMAP); 1926 if (ret < 0) { 1927 return ret; 1928 } 1929 break; 1930 } 1931 1932 sector_num += n; 1933 nb_sectors -= n; 1934 buf += n * BDRV_SECTOR_SIZE; 1935 } 1936 1937 return 0; 1938 } 1939 1940 static int coroutine_fn convert_co_copy_range(ImgConvertState *s, int64_t sector_num, 1941 int nb_sectors) 1942 { 1943 int n, ret; 1944 1945 while (nb_sectors > 0) { 1946 BlockBackend *blk; 1947 int src_cur; 1948 int64_t bs_sectors, src_cur_offset; 1949 int64_t offset; 1950 1951 convert_select_part(s, sector_num, &src_cur, &src_cur_offset); 1952 offset = (sector_num - src_cur_offset) << BDRV_SECTOR_BITS; 1953 blk = s->src[src_cur]; 1954 bs_sectors = s->src_sectors[src_cur]; 1955 1956 n = MIN(nb_sectors, bs_sectors - (sector_num - src_cur_offset)); 1957 1958 ret = blk_co_copy_range(blk, offset, s->target, 1959 sector_num << BDRV_SECTOR_BITS, 1960 n << BDRV_SECTOR_BITS, 0, 0); 1961 if (ret < 0) { 1962 return ret; 1963 } 1964 1965 sector_num += n; 1966 nb_sectors -= n; 1967 } 1968 return 0; 1969 } 1970 1971 static void coroutine_fn convert_co_do_copy(void *opaque) 1972 { 1973 ImgConvertState *s = opaque; 1974 uint8_t *buf = NULL; 1975 int ret, i; 1976 int index = -1; 1977 1978 for (i = 0; i < s->num_coroutines; i++) { 1979 if (s->co[i] == qemu_coroutine_self()) { 1980 index = i; 1981 break; 1982 } 1983 } 1984 assert(index >= 0); 1985 1986 s->running_coroutines++; 1987 buf = blk_blockalign(s->target, s->buf_sectors * BDRV_SECTOR_SIZE); 1988 1989 while (1) { 1990 int n; 1991 int64_t sector_num; 1992 enum ImgConvertBlockStatus status; 1993 bool copy_range; 1994 1995 qemu_co_mutex_lock(&s->lock); 1996 if (s->ret != -EINPROGRESS || s->sector_num >= s->total_sectors) { 1997 qemu_co_mutex_unlock(&s->lock); 1998 break; 1999 } 2000 WITH_GRAPH_RDLOCK_GUARD() { 2001 n = convert_iteration_sectors(s, s->sector_num); 2002 } 2003 if (n < 0) { 2004 qemu_co_mutex_unlock(&s->lock); 2005 s->ret = n; 2006 break; 2007 } 2008 /* save current sector and allocation status to local variables */ 2009 sector_num = s->sector_num; 2010 status = s->status; 2011 if (!s->min_sparse && s->status == BLK_ZERO) { 2012 n = MIN(n, s->buf_sectors); 2013 } 2014 /* increment global sector counter so that other coroutines can 2015 * already continue reading beyond this request */ 2016 s->sector_num += n; 2017 qemu_co_mutex_unlock(&s->lock); 2018 2019 if (status == BLK_DATA || (!s->min_sparse && status == BLK_ZERO)) { 2020 s->allocated_done += n; 2021 qemu_progress_print(100.0 * s->allocated_done / 2022 s->allocated_sectors, 0); 2023 } 2024 2025 retry: 2026 copy_range = s->copy_range && s->status == BLK_DATA; 2027 if (status == BLK_DATA && !copy_range) { 2028 ret = convert_co_read(s, sector_num, n, buf); 2029 if (ret < 0) { 2030 error_report("error while reading at byte %lld: %s", 2031 sector_num * BDRV_SECTOR_SIZE, strerror(-ret)); 2032 s->ret = ret; 2033 } 2034 } else if (!s->min_sparse && status == BLK_ZERO) { 2035 status = BLK_DATA; 2036 memset(buf, 0x00, n * BDRV_SECTOR_SIZE); 2037 } 2038 2039 if (s->wr_in_order) { 2040 /* keep writes in order */ 2041 while (s->wr_offs != sector_num && s->ret == -EINPROGRESS) { 2042 s->wait_sector_num[index] = sector_num; 2043 qemu_coroutine_yield(); 2044 } 2045 s->wait_sector_num[index] = -1; 2046 } 2047 2048 if (s->ret == -EINPROGRESS) { 2049 if (copy_range) { 2050 WITH_GRAPH_RDLOCK_GUARD() { 2051 ret = convert_co_copy_range(s, sector_num, n); 2052 } 2053 if (ret) { 2054 s->copy_range = false; 2055 goto retry; 2056 } 2057 } else { 2058 ret = convert_co_write(s, sector_num, n, buf, status); 2059 } 2060 if (ret < 0) { 2061 error_report("error while writing at byte %lld: %s", 2062 sector_num * BDRV_SECTOR_SIZE, strerror(-ret)); 2063 s->ret = ret; 2064 } 2065 } 2066 2067 if (s->wr_in_order) { 2068 /* reenter the coroutine that might have waited 2069 * for this write to complete */ 2070 s->wr_offs = sector_num + n; 2071 for (i = 0; i < s->num_coroutines; i++) { 2072 if (s->co[i] && s->wait_sector_num[i] == s->wr_offs) { 2073 /* 2074 * A -> B -> A cannot occur because A has 2075 * s->wait_sector_num[i] == -1 during A -> B. Therefore 2076 * B will never enter A during this time window. 2077 */ 2078 qemu_coroutine_enter(s->co[i]); 2079 break; 2080 } 2081 } 2082 } 2083 } 2084 2085 qemu_vfree(buf); 2086 s->co[index] = NULL; 2087 s->running_coroutines--; 2088 if (!s->running_coroutines && s->ret == -EINPROGRESS) { 2089 /* the convert job finished successfully */ 2090 s->ret = 0; 2091 } 2092 } 2093 2094 static int convert_do_copy(ImgConvertState *s) 2095 { 2096 int ret, i, n; 2097 int64_t sector_num = 0; 2098 2099 /* Check whether we have zero initialisation or can get it efficiently */ 2100 if (!s->has_zero_init && s->target_is_new && s->min_sparse && 2101 !s->target_has_backing) { 2102 bdrv_graph_rdlock_main_loop(); 2103 s->has_zero_init = bdrv_has_zero_init(blk_bs(s->target)); 2104 bdrv_graph_rdunlock_main_loop(); 2105 } 2106 2107 /* Allocate buffer for copied data. For compressed images, only one cluster 2108 * can be copied at a time. */ 2109 if (s->compressed) { 2110 if (s->cluster_sectors <= 0 || s->cluster_sectors > s->buf_sectors) { 2111 error_report("invalid cluster size"); 2112 return -EINVAL; 2113 } 2114 s->buf_sectors = s->cluster_sectors; 2115 } 2116 2117 while (sector_num < s->total_sectors) { 2118 n = convert_iteration_sectors(s, sector_num); 2119 if (n < 0) { 2120 return n; 2121 } 2122 if (s->status == BLK_DATA || (!s->min_sparse && s->status == BLK_ZERO)) 2123 { 2124 s->allocated_sectors += n; 2125 } 2126 sector_num += n; 2127 } 2128 2129 /* Do the copy */ 2130 s->sector_next_status = 0; 2131 s->ret = -EINPROGRESS; 2132 2133 qemu_co_mutex_init(&s->lock); 2134 for (i = 0; i < s->num_coroutines; i++) { 2135 s->co[i] = qemu_coroutine_create(convert_co_do_copy, s); 2136 s->wait_sector_num[i] = -1; 2137 qemu_coroutine_enter(s->co[i]); 2138 } 2139 2140 while (s->running_coroutines) { 2141 main_loop_wait(false); 2142 } 2143 2144 if (s->compressed && !s->ret) { 2145 /* signal EOF to align */ 2146 ret = blk_pwrite_compressed(s->target, 0, 0, NULL); 2147 if (ret < 0) { 2148 return ret; 2149 } 2150 } 2151 2152 return s->ret; 2153 } 2154 2155 /* Check that bitmaps can be copied, or output an error */ 2156 static int convert_check_bitmaps(BlockDriverState *src, bool skip_broken) 2157 { 2158 BdrvDirtyBitmap *bm; 2159 2160 if (!bdrv_supports_persistent_dirty_bitmap(src)) { 2161 error_report("Source lacks bitmap support"); 2162 return -1; 2163 } 2164 FOR_EACH_DIRTY_BITMAP(src, bm) { 2165 if (!bdrv_dirty_bitmap_get_persistence(bm)) { 2166 continue; 2167 } 2168 if (!skip_broken && bdrv_dirty_bitmap_inconsistent(bm)) { 2169 error_report("Cannot copy inconsistent bitmap '%s'", 2170 bdrv_dirty_bitmap_name(bm)); 2171 error_printf("Try --skip-broken-bitmaps, or " 2172 "use 'qemu-img bitmap --remove' to delete it\n"); 2173 return -1; 2174 } 2175 } 2176 return 0; 2177 } 2178 2179 static int convert_copy_bitmaps(BlockDriverState *src, BlockDriverState *dst, 2180 bool skip_broken) 2181 { 2182 BdrvDirtyBitmap *bm; 2183 Error *err = NULL; 2184 2185 FOR_EACH_DIRTY_BITMAP(src, bm) { 2186 const char *name; 2187 2188 if (!bdrv_dirty_bitmap_get_persistence(bm)) { 2189 continue; 2190 } 2191 name = bdrv_dirty_bitmap_name(bm); 2192 if (skip_broken && bdrv_dirty_bitmap_inconsistent(bm)) { 2193 warn_report("Skipping inconsistent bitmap '%s'", name); 2194 continue; 2195 } 2196 qmp_block_dirty_bitmap_add(dst->node_name, name, 2197 true, bdrv_dirty_bitmap_granularity(bm), 2198 true, true, 2199 true, !bdrv_dirty_bitmap_enabled(bm), 2200 &err); 2201 if (err) { 2202 error_reportf_err(err, "Failed to create bitmap %s: ", name); 2203 return -1; 2204 } 2205 2206 do_dirty_bitmap_merge(dst->node_name, name, src->node_name, name, 2207 &err); 2208 if (err) { 2209 error_reportf_err(err, "Failed to populate bitmap %s: ", name); 2210 qmp_block_dirty_bitmap_remove(dst->node_name, name, NULL); 2211 return -1; 2212 } 2213 } 2214 2215 return 0; 2216 } 2217 2218 #define MAX_BUF_SECTORS 32768 2219 2220 static void set_rate_limit(BlockBackend *blk, int64_t rate_limit) 2221 { 2222 ThrottleConfig cfg; 2223 2224 throttle_config_init(&cfg); 2225 cfg.buckets[THROTTLE_BPS_WRITE].avg = rate_limit; 2226 2227 blk_io_limits_enable(blk, CONVERT_THROTTLE_GROUP); 2228 blk_set_io_limits(blk, &cfg); 2229 } 2230 2231 static int img_convert(int argc, char **argv) 2232 { 2233 int c, bs_i, flags, src_flags = BDRV_O_NO_SHARE; 2234 const char *fmt = NULL, *out_fmt = NULL, *cache = "unsafe", 2235 *src_cache = BDRV_DEFAULT_CACHE, *out_baseimg = NULL, 2236 *out_filename, *out_baseimg_param, *snapshot_name = NULL, 2237 *backing_fmt = NULL; 2238 BlockDriver *drv = NULL, *proto_drv = NULL; 2239 BlockDriverInfo bdi; 2240 BlockDriverState *out_bs; 2241 QemuOpts *opts = NULL, *sn_opts = NULL; 2242 QemuOptsList *create_opts = NULL; 2243 QDict *open_opts = NULL; 2244 char *options = NULL; 2245 Error *local_err = NULL; 2246 bool writethrough, src_writethrough, image_opts = false, 2247 skip_create = false, progress = false, tgt_image_opts = false; 2248 int64_t ret = -EINVAL; 2249 bool force_share = false; 2250 bool explict_min_sparse = false; 2251 bool bitmaps = false; 2252 bool skip_broken = false; 2253 int64_t rate_limit = 0; 2254 2255 ImgConvertState s = (ImgConvertState) { 2256 /* Need at least 4k of zeros for sparse detection */ 2257 .min_sparse = 8, 2258 .copy_range = false, 2259 .buf_sectors = IO_BUF_SIZE / BDRV_SECTOR_SIZE, 2260 .wr_in_order = true, 2261 .num_coroutines = 8, 2262 }; 2263 2264 for(;;) { 2265 static const struct option long_options[] = { 2266 {"help", no_argument, 0, 'h'}, 2267 {"object", required_argument, 0, OPTION_OBJECT}, 2268 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS}, 2269 {"force-share", no_argument, 0, 'U'}, 2270 {"target-image-opts", no_argument, 0, OPTION_TARGET_IMAGE_OPTS}, 2271 {"salvage", no_argument, 0, OPTION_SALVAGE}, 2272 {"target-is-zero", no_argument, 0, OPTION_TARGET_IS_ZERO}, 2273 {"bitmaps", no_argument, 0, OPTION_BITMAPS}, 2274 {"skip-broken-bitmaps", no_argument, 0, OPTION_SKIP_BROKEN}, 2275 {0, 0, 0, 0} 2276 }; 2277 c = getopt_long(argc, argv, ":hf:O:B:CcF:o:l:S:pt:T:qnm:WUr:", 2278 long_options, NULL); 2279 if (c == -1) { 2280 break; 2281 } 2282 switch(c) { 2283 case ':': 2284 missing_argument(argv[optind - 1]); 2285 break; 2286 case '?': 2287 unrecognized_option(argv[optind - 1]); 2288 break; 2289 case 'h': 2290 help(); 2291 break; 2292 case 'f': 2293 fmt = optarg; 2294 break; 2295 case 'O': 2296 out_fmt = optarg; 2297 break; 2298 case 'B': 2299 out_baseimg = optarg; 2300 break; 2301 case 'C': 2302 s.copy_range = true; 2303 break; 2304 case 'c': 2305 s.compressed = true; 2306 break; 2307 case 'F': 2308 backing_fmt = optarg; 2309 break; 2310 case 'o': 2311 if (accumulate_options(&options, optarg) < 0) { 2312 goto fail_getopt; 2313 } 2314 break; 2315 case 'l': 2316 if (strstart(optarg, SNAPSHOT_OPT_BASE, NULL)) { 2317 sn_opts = qemu_opts_parse_noisily(&internal_snapshot_opts, 2318 optarg, false); 2319 if (!sn_opts) { 2320 error_report("Failed in parsing snapshot param '%s'", 2321 optarg); 2322 goto fail_getopt; 2323 } 2324 } else { 2325 snapshot_name = optarg; 2326 } 2327 break; 2328 case 'S': 2329 { 2330 int64_t sval; 2331 2332 sval = cvtnum("buffer size for sparse output", optarg); 2333 if (sval < 0) { 2334 goto fail_getopt; 2335 } else if (!QEMU_IS_ALIGNED(sval, BDRV_SECTOR_SIZE) || 2336 sval / BDRV_SECTOR_SIZE > MAX_BUF_SECTORS) { 2337 error_report("Invalid buffer size for sparse output specified. " 2338 "Valid sizes are multiples of %llu up to %llu. Select " 2339 "0 to disable sparse detection (fully allocates output).", 2340 BDRV_SECTOR_SIZE, MAX_BUF_SECTORS * BDRV_SECTOR_SIZE); 2341 goto fail_getopt; 2342 } 2343 2344 s.min_sparse = sval / BDRV_SECTOR_SIZE; 2345 explict_min_sparse = true; 2346 break; 2347 } 2348 case 'p': 2349 progress = true; 2350 break; 2351 case 't': 2352 cache = optarg; 2353 break; 2354 case 'T': 2355 src_cache = optarg; 2356 break; 2357 case 'q': 2358 s.quiet = true; 2359 break; 2360 case 'n': 2361 skip_create = true; 2362 break; 2363 case 'm': 2364 if (qemu_strtol(optarg, NULL, 0, &s.num_coroutines) || 2365 s.num_coroutines < 1 || s.num_coroutines > MAX_COROUTINES) { 2366 error_report("Invalid number of coroutines. Allowed number of" 2367 " coroutines is between 1 and %d", MAX_COROUTINES); 2368 goto fail_getopt; 2369 } 2370 break; 2371 case 'W': 2372 s.wr_in_order = false; 2373 break; 2374 case 'U': 2375 force_share = true; 2376 break; 2377 case 'r': 2378 rate_limit = cvtnum("rate limit", optarg); 2379 if (rate_limit < 0) { 2380 goto fail_getopt; 2381 } 2382 break; 2383 case OPTION_OBJECT: 2384 user_creatable_process_cmdline(optarg); 2385 break; 2386 case OPTION_IMAGE_OPTS: 2387 image_opts = true; 2388 break; 2389 case OPTION_SALVAGE: 2390 s.salvage = true; 2391 break; 2392 case OPTION_TARGET_IMAGE_OPTS: 2393 tgt_image_opts = true; 2394 break; 2395 case OPTION_TARGET_IS_ZERO: 2396 /* 2397 * The user asserting that the target is blank has the 2398 * same effect as the target driver supporting zero 2399 * initialisation. 2400 */ 2401 s.has_zero_init = true; 2402 break; 2403 case OPTION_BITMAPS: 2404 bitmaps = true; 2405 break; 2406 case OPTION_SKIP_BROKEN: 2407 skip_broken = true; 2408 break; 2409 } 2410 } 2411 2412 if (!out_fmt && !tgt_image_opts) { 2413 out_fmt = "raw"; 2414 } 2415 2416 if (skip_broken && !bitmaps) { 2417 error_report("Use of --skip-broken-bitmaps requires --bitmaps"); 2418 goto fail_getopt; 2419 } 2420 2421 if (s.compressed && s.copy_range) { 2422 error_report("Cannot enable copy offloading when -c is used"); 2423 goto fail_getopt; 2424 } 2425 2426 if (explict_min_sparse && s.copy_range) { 2427 error_report("Cannot enable copy offloading when -S is used"); 2428 goto fail_getopt; 2429 } 2430 2431 if (s.copy_range && s.salvage) { 2432 error_report("Cannot use copy offloading in salvaging mode"); 2433 goto fail_getopt; 2434 } 2435 2436 if (tgt_image_opts && !skip_create) { 2437 error_report("--target-image-opts requires use of -n flag"); 2438 goto fail_getopt; 2439 } 2440 2441 if (skip_create && options) { 2442 error_report("-o has no effect when skipping image creation"); 2443 goto fail_getopt; 2444 } 2445 2446 if (s.has_zero_init && !skip_create) { 2447 error_report("--target-is-zero requires use of -n flag"); 2448 goto fail_getopt; 2449 } 2450 2451 s.src_num = argc - optind - 1; 2452 out_filename = s.src_num >= 1 ? argv[argc - 1] : NULL; 2453 2454 if (options && has_help_option(options)) { 2455 if (out_fmt) { 2456 ret = print_block_option_help(out_filename, out_fmt); 2457 goto fail_getopt; 2458 } else { 2459 error_report("Option help requires a format be specified"); 2460 goto fail_getopt; 2461 } 2462 } 2463 2464 if (s.src_num < 1) { 2465 error_report("Must specify image file name"); 2466 goto fail_getopt; 2467 } 2468 2469 /* ret is still -EINVAL until here */ 2470 ret = bdrv_parse_cache_mode(src_cache, &src_flags, &src_writethrough); 2471 if (ret < 0) { 2472 error_report("Invalid source cache option: %s", src_cache); 2473 goto fail_getopt; 2474 } 2475 2476 /* Initialize before goto out */ 2477 if (s.quiet) { 2478 progress = false; 2479 } 2480 qemu_progress_init(progress, 1.0); 2481 qemu_progress_print(0, 100); 2482 2483 s.src = g_new0(BlockBackend *, s.src_num); 2484 s.src_sectors = g_new(int64_t, s.src_num); 2485 s.src_alignment = g_new(int, s.src_num); 2486 2487 for (bs_i = 0; bs_i < s.src_num; bs_i++) { 2488 BlockDriverState *src_bs; 2489 s.src[bs_i] = img_open(image_opts, argv[optind + bs_i], 2490 fmt, src_flags, src_writethrough, s.quiet, 2491 force_share); 2492 if (!s.src[bs_i]) { 2493 ret = -1; 2494 goto out; 2495 } 2496 s.src_sectors[bs_i] = blk_nb_sectors(s.src[bs_i]); 2497 if (s.src_sectors[bs_i] < 0) { 2498 error_report("Could not get size of %s: %s", 2499 argv[optind + bs_i], strerror(-s.src_sectors[bs_i])); 2500 ret = -1; 2501 goto out; 2502 } 2503 src_bs = blk_bs(s.src[bs_i]); 2504 s.src_alignment[bs_i] = DIV_ROUND_UP(src_bs->bl.request_alignment, 2505 BDRV_SECTOR_SIZE); 2506 if (!bdrv_get_info(src_bs, &bdi)) { 2507 s.src_alignment[bs_i] = MAX(s.src_alignment[bs_i], 2508 bdi.cluster_size / BDRV_SECTOR_SIZE); 2509 } 2510 s.total_sectors += s.src_sectors[bs_i]; 2511 } 2512 2513 if (sn_opts) { 2514 bdrv_snapshot_load_tmp(blk_bs(s.src[0]), 2515 qemu_opt_get(sn_opts, SNAPSHOT_OPT_ID), 2516 qemu_opt_get(sn_opts, SNAPSHOT_OPT_NAME), 2517 &local_err); 2518 } else if (snapshot_name != NULL) { 2519 if (s.src_num > 1) { 2520 error_report("No support for concatenating multiple snapshot"); 2521 ret = -1; 2522 goto out; 2523 } 2524 2525 bdrv_snapshot_load_tmp_by_id_or_name(blk_bs(s.src[0]), snapshot_name, 2526 &local_err); 2527 } 2528 if (local_err) { 2529 error_reportf_err(local_err, "Failed to load snapshot: "); 2530 ret = -1; 2531 goto out; 2532 } 2533 2534 if (!skip_create) { 2535 /* Find driver and parse its options */ 2536 drv = bdrv_find_format(out_fmt); 2537 if (!drv) { 2538 error_report("Unknown file format '%s'", out_fmt); 2539 ret = -1; 2540 goto out; 2541 } 2542 2543 proto_drv = bdrv_find_protocol(out_filename, true, &local_err); 2544 if (!proto_drv) { 2545 error_report_err(local_err); 2546 ret = -1; 2547 goto out; 2548 } 2549 2550 if (!drv->create_opts) { 2551 error_report("Format driver '%s' does not support image creation", 2552 drv->format_name); 2553 ret = -1; 2554 goto out; 2555 } 2556 2557 if (!proto_drv->create_opts) { 2558 error_report("Protocol driver '%s' does not support image creation", 2559 proto_drv->format_name); 2560 ret = -1; 2561 goto out; 2562 } 2563 2564 create_opts = qemu_opts_append(create_opts, drv->create_opts); 2565 create_opts = qemu_opts_append(create_opts, proto_drv->create_opts); 2566 2567 opts = qemu_opts_create(create_opts, NULL, 0, &error_abort); 2568 if (options) { 2569 if (!qemu_opts_do_parse(opts, options, NULL, &local_err)) { 2570 error_report_err(local_err); 2571 ret = -1; 2572 goto out; 2573 } 2574 } 2575 2576 qemu_opt_set_number(opts, BLOCK_OPT_SIZE, 2577 s.total_sectors * BDRV_SECTOR_SIZE, &error_abort); 2578 ret = add_old_style_options(out_fmt, opts, out_baseimg, backing_fmt); 2579 if (ret < 0) { 2580 goto out; 2581 } 2582 } 2583 2584 /* Get backing file name if -o backing_file was used */ 2585 out_baseimg_param = qemu_opt_get(opts, BLOCK_OPT_BACKING_FILE); 2586 if (out_baseimg_param) { 2587 out_baseimg = out_baseimg_param; 2588 } 2589 s.target_has_backing = (bool) out_baseimg; 2590 2591 if (s.has_zero_init && s.target_has_backing) { 2592 error_report("Cannot use --target-is-zero when the destination " 2593 "image has a backing file"); 2594 goto out; 2595 } 2596 2597 if (s.src_num > 1 && out_baseimg) { 2598 error_report("Having a backing file for the target makes no sense when " 2599 "concatenating multiple input images"); 2600 ret = -1; 2601 goto out; 2602 } 2603 2604 if (out_baseimg_param) { 2605 if (!qemu_opt_get(opts, BLOCK_OPT_BACKING_FMT)) { 2606 error_report("Use of backing file requires explicit " 2607 "backing format"); 2608 ret = -1; 2609 goto out; 2610 } 2611 } 2612 2613 /* Check if compression is supported */ 2614 if (s.compressed) { 2615 bool encryption = 2616 qemu_opt_get_bool(opts, BLOCK_OPT_ENCRYPT, false); 2617 const char *encryptfmt = 2618 qemu_opt_get(opts, BLOCK_OPT_ENCRYPT_FORMAT); 2619 const char *preallocation = 2620 qemu_opt_get(opts, BLOCK_OPT_PREALLOC); 2621 2622 if (drv && !block_driver_can_compress(drv)) { 2623 error_report("Compression not supported for this file format"); 2624 ret = -1; 2625 goto out; 2626 } 2627 2628 if (encryption || encryptfmt) { 2629 error_report("Compression and encryption not supported at " 2630 "the same time"); 2631 ret = -1; 2632 goto out; 2633 } 2634 2635 if (preallocation 2636 && strcmp(preallocation, "off")) 2637 { 2638 error_report("Compression and preallocation not supported at " 2639 "the same time"); 2640 ret = -1; 2641 goto out; 2642 } 2643 } 2644 2645 /* Determine if bitmaps need copying */ 2646 if (bitmaps) { 2647 if (s.src_num > 1) { 2648 error_report("Copying bitmaps only possible with single source"); 2649 ret = -1; 2650 goto out; 2651 } 2652 ret = convert_check_bitmaps(blk_bs(s.src[0]), skip_broken); 2653 if (ret < 0) { 2654 goto out; 2655 } 2656 } 2657 2658 /* 2659 * The later open call will need any decryption secrets, and 2660 * bdrv_create() will purge "opts", so extract them now before 2661 * they are lost. 2662 */ 2663 if (!skip_create) { 2664 open_opts = qdict_new(); 2665 qemu_opt_foreach(opts, img_add_key_secrets, open_opts, &error_abort); 2666 2667 /* Create the new image */ 2668 ret = bdrv_create(drv, out_filename, opts, &local_err); 2669 if (ret < 0) { 2670 error_reportf_err(local_err, "%s: error while converting %s: ", 2671 out_filename, out_fmt); 2672 goto out; 2673 } 2674 } 2675 2676 s.target_is_new = !skip_create; 2677 2678 flags = s.min_sparse ? (BDRV_O_RDWR | BDRV_O_UNMAP) : BDRV_O_RDWR; 2679 ret = bdrv_parse_cache_mode(cache, &flags, &writethrough); 2680 if (ret < 0) { 2681 error_report("Invalid cache option: %s", cache); 2682 goto out; 2683 } 2684 2685 if (flags & BDRV_O_NOCACHE) { 2686 /* 2687 * If we open the target with O_DIRECT, it may be necessary to 2688 * extend its size to align to the physical sector size. 2689 */ 2690 flags |= BDRV_O_RESIZE; 2691 } 2692 2693 if (skip_create) { 2694 s.target = img_open(tgt_image_opts, out_filename, out_fmt, 2695 flags, writethrough, s.quiet, false); 2696 } else { 2697 /* TODO ultimately we should allow --target-image-opts 2698 * to be used even when -n is not given. 2699 * That has to wait for bdrv_create to be improved 2700 * to allow filenames in option syntax 2701 */ 2702 s.target = img_open_file(out_filename, open_opts, out_fmt, 2703 flags, writethrough, s.quiet, false); 2704 open_opts = NULL; /* blk_new_open will have freed it */ 2705 } 2706 if (!s.target) { 2707 ret = -1; 2708 goto out; 2709 } 2710 out_bs = blk_bs(s.target); 2711 2712 if (bitmaps && !bdrv_supports_persistent_dirty_bitmap(out_bs)) { 2713 error_report("Format driver '%s' does not support bitmaps", 2714 out_bs->drv->format_name); 2715 ret = -1; 2716 goto out; 2717 } 2718 2719 if (s.compressed && !block_driver_can_compress(out_bs->drv)) { 2720 error_report("Compression not supported for this file format"); 2721 ret = -1; 2722 goto out; 2723 } 2724 2725 /* increase bufsectors from the default 4096 (2M) if opt_transfer 2726 * or discard_alignment of the out_bs is greater. Limit to 2727 * MAX_BUF_SECTORS as maximum which is currently 32768 (16MB). */ 2728 s.buf_sectors = MIN(MAX_BUF_SECTORS, 2729 MAX(s.buf_sectors, 2730 MAX(out_bs->bl.opt_transfer >> BDRV_SECTOR_BITS, 2731 out_bs->bl.pdiscard_alignment >> 2732 BDRV_SECTOR_BITS))); 2733 2734 /* try to align the write requests to the destination to avoid unnecessary 2735 * RMW cycles. */ 2736 s.alignment = MAX(pow2floor(s.min_sparse), 2737 DIV_ROUND_UP(out_bs->bl.request_alignment, 2738 BDRV_SECTOR_SIZE)); 2739 assert(is_power_of_2(s.alignment)); 2740 2741 if (skip_create) { 2742 int64_t output_sectors = blk_nb_sectors(s.target); 2743 if (output_sectors < 0) { 2744 error_report("unable to get output image length: %s", 2745 strerror(-output_sectors)); 2746 ret = -1; 2747 goto out; 2748 } else if (output_sectors < s.total_sectors) { 2749 error_report("output file is smaller than input file"); 2750 ret = -1; 2751 goto out; 2752 } 2753 } 2754 2755 if (s.target_has_backing && s.target_is_new) { 2756 /* Errors are treated as "backing length unknown" (which means 2757 * s.target_backing_sectors has to be negative, which it will 2758 * be automatically). The backing file length is used only 2759 * for optimizations, so such a case is not fatal. */ 2760 s.target_backing_sectors = 2761 bdrv_nb_sectors(bdrv_backing_chain_next(out_bs)); 2762 } else { 2763 s.target_backing_sectors = -1; 2764 } 2765 2766 ret = bdrv_get_info(out_bs, &bdi); 2767 if (ret < 0) { 2768 if (s.compressed) { 2769 error_report("could not get block driver info"); 2770 goto out; 2771 } 2772 } else { 2773 s.compressed = s.compressed || bdi.needs_compressed_writes; 2774 s.cluster_sectors = bdi.cluster_size / BDRV_SECTOR_SIZE; 2775 } 2776 2777 if (rate_limit) { 2778 set_rate_limit(s.target, rate_limit); 2779 } 2780 2781 ret = convert_do_copy(&s); 2782 2783 /* Now copy the bitmaps */ 2784 if (bitmaps && ret == 0) { 2785 ret = convert_copy_bitmaps(blk_bs(s.src[0]), out_bs, skip_broken); 2786 } 2787 2788 out: 2789 if (!ret) { 2790 qemu_progress_print(100, 0); 2791 } 2792 qemu_progress_end(); 2793 qemu_opts_del(opts); 2794 qemu_opts_free(create_opts); 2795 qobject_unref(open_opts); 2796 blk_unref(s.target); 2797 if (s.src) { 2798 for (bs_i = 0; bs_i < s.src_num; bs_i++) { 2799 blk_unref(s.src[bs_i]); 2800 } 2801 g_free(s.src); 2802 } 2803 g_free(s.src_sectors); 2804 g_free(s.src_alignment); 2805 fail_getopt: 2806 qemu_opts_del(sn_opts); 2807 g_free(options); 2808 2809 return !!ret; 2810 } 2811 2812 2813 static void dump_snapshots(BlockDriverState *bs) 2814 { 2815 QEMUSnapshotInfo *sn_tab, *sn; 2816 int nb_sns, i; 2817 2818 nb_sns = bdrv_snapshot_list(bs, &sn_tab); 2819 if (nb_sns <= 0) 2820 return; 2821 printf("Snapshot list:\n"); 2822 bdrv_snapshot_dump(NULL); 2823 printf("\n"); 2824 for(i = 0; i < nb_sns; i++) { 2825 sn = &sn_tab[i]; 2826 bdrv_snapshot_dump(sn); 2827 printf("\n"); 2828 } 2829 g_free(sn_tab); 2830 } 2831 2832 static void dump_json_block_graph_info_list(BlockGraphInfoList *list) 2833 { 2834 GString *str; 2835 QObject *obj; 2836 Visitor *v = qobject_output_visitor_new(&obj); 2837 2838 visit_type_BlockGraphInfoList(v, NULL, &list, &error_abort); 2839 visit_complete(v, &obj); 2840 str = qobject_to_json_pretty(obj, true); 2841 assert(str != NULL); 2842 printf("%s\n", str->str); 2843 qobject_unref(obj); 2844 visit_free(v); 2845 g_string_free(str, true); 2846 } 2847 2848 static void dump_json_block_graph_info(BlockGraphInfo *info) 2849 { 2850 GString *str; 2851 QObject *obj; 2852 Visitor *v = qobject_output_visitor_new(&obj); 2853 2854 visit_type_BlockGraphInfo(v, NULL, &info, &error_abort); 2855 visit_complete(v, &obj); 2856 str = qobject_to_json_pretty(obj, true); 2857 assert(str != NULL); 2858 printf("%s\n", str->str); 2859 qobject_unref(obj); 2860 visit_free(v); 2861 g_string_free(str, true); 2862 } 2863 2864 static void dump_human_image_info(BlockGraphInfo *info, int indentation, 2865 const char *path) 2866 { 2867 BlockChildInfoList *children_list; 2868 2869 bdrv_node_info_dump(qapi_BlockGraphInfo_base(info), indentation, 2870 info->children == NULL); 2871 2872 for (children_list = info->children; children_list; 2873 children_list = children_list->next) 2874 { 2875 BlockChildInfo *child = children_list->value; 2876 g_autofree char *child_path = NULL; 2877 2878 printf("%*sChild node '%s%s':\n", 2879 indentation * 4, "", path, child->name); 2880 child_path = g_strdup_printf("%s%s/", path, child->name); 2881 dump_human_image_info(child->info, indentation + 1, child_path); 2882 } 2883 } 2884 2885 static void dump_human_image_info_list(BlockGraphInfoList *list) 2886 { 2887 BlockGraphInfoList *elem; 2888 bool delim = false; 2889 2890 for (elem = list; elem; elem = elem->next) { 2891 if (delim) { 2892 printf("\n"); 2893 } 2894 delim = true; 2895 2896 dump_human_image_info(elem->value, 0, "/"); 2897 } 2898 } 2899 2900 static gboolean str_equal_func(gconstpointer a, gconstpointer b) 2901 { 2902 return strcmp(a, b) == 0; 2903 } 2904 2905 /** 2906 * Open an image file chain and return an BlockGraphInfoList 2907 * 2908 * @filename: topmost image filename 2909 * @fmt: topmost image format (may be NULL to autodetect) 2910 * @chain: true - enumerate entire backing file chain 2911 * false - only topmost image file 2912 * 2913 * Returns a list of BlockNodeInfo objects or NULL if there was an error 2914 * opening an image file. If there was an error a message will have been 2915 * printed to stderr. 2916 */ 2917 static BlockGraphInfoList *collect_image_info_list(bool image_opts, 2918 const char *filename, 2919 const char *fmt, 2920 bool chain, bool force_share) 2921 { 2922 BlockGraphInfoList *head = NULL; 2923 BlockGraphInfoList **tail = &head; 2924 GHashTable *filenames; 2925 Error *err = NULL; 2926 2927 filenames = g_hash_table_new_full(g_str_hash, str_equal_func, NULL, NULL); 2928 2929 while (filename) { 2930 BlockBackend *blk; 2931 BlockDriverState *bs; 2932 BlockGraphInfo *info; 2933 2934 if (g_hash_table_lookup_extended(filenames, filename, NULL, NULL)) { 2935 error_report("Backing file '%s' creates an infinite loop.", 2936 filename); 2937 goto err; 2938 } 2939 g_hash_table_insert(filenames, (gpointer)filename, NULL); 2940 2941 blk = img_open(image_opts, filename, fmt, 2942 BDRV_O_NO_BACKING | BDRV_O_NO_IO, false, false, 2943 force_share); 2944 if (!blk) { 2945 goto err; 2946 } 2947 bs = blk_bs(blk); 2948 2949 /* 2950 * Note that the returned BlockGraphInfo object will not have 2951 * information about this image's backing node, because we have opened 2952 * it with BDRV_O_NO_BACKING. Printing this object will therefore not 2953 * duplicate the backing chain information that we obtain by walking 2954 * the chain manually here. 2955 */ 2956 bdrv_graph_rdlock_main_loop(); 2957 bdrv_query_block_graph_info(bs, &info, &err); 2958 bdrv_graph_rdunlock_main_loop(); 2959 2960 if (err) { 2961 error_report_err(err); 2962 blk_unref(blk); 2963 goto err; 2964 } 2965 2966 QAPI_LIST_APPEND(tail, info); 2967 2968 blk_unref(blk); 2969 2970 /* Clear parameters that only apply to the topmost image */ 2971 filename = fmt = NULL; 2972 image_opts = false; 2973 2974 if (chain) { 2975 if (info->full_backing_filename) { 2976 filename = info->full_backing_filename; 2977 } else if (info->backing_filename) { 2978 error_report("Could not determine absolute backing filename," 2979 " but backing filename '%s' present", 2980 info->backing_filename); 2981 goto err; 2982 } 2983 if (info->backing_filename_format) { 2984 fmt = info->backing_filename_format; 2985 } 2986 } 2987 } 2988 g_hash_table_destroy(filenames); 2989 return head; 2990 2991 err: 2992 qapi_free_BlockGraphInfoList(head); 2993 g_hash_table_destroy(filenames); 2994 return NULL; 2995 } 2996 2997 static int img_info(int argc, char **argv) 2998 { 2999 int c; 3000 OutputFormat output_format = OFORMAT_HUMAN; 3001 bool chain = false; 3002 const char *filename, *fmt, *output; 3003 BlockGraphInfoList *list; 3004 bool image_opts = false; 3005 bool force_share = false; 3006 3007 fmt = NULL; 3008 output = NULL; 3009 for(;;) { 3010 int option_index = 0; 3011 static const struct option long_options[] = { 3012 {"help", no_argument, 0, 'h'}, 3013 {"format", required_argument, 0, 'f'}, 3014 {"output", required_argument, 0, OPTION_OUTPUT}, 3015 {"backing-chain", no_argument, 0, OPTION_BACKING_CHAIN}, 3016 {"object", required_argument, 0, OPTION_OBJECT}, 3017 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS}, 3018 {"force-share", no_argument, 0, 'U'}, 3019 {0, 0, 0, 0} 3020 }; 3021 c = getopt_long(argc, argv, ":f:hU", 3022 long_options, &option_index); 3023 if (c == -1) { 3024 break; 3025 } 3026 switch(c) { 3027 case ':': 3028 missing_argument(argv[optind - 1]); 3029 break; 3030 case '?': 3031 unrecognized_option(argv[optind - 1]); 3032 break; 3033 case 'h': 3034 help(); 3035 break; 3036 case 'f': 3037 fmt = optarg; 3038 break; 3039 case 'U': 3040 force_share = true; 3041 break; 3042 case OPTION_OUTPUT: 3043 output = optarg; 3044 break; 3045 case OPTION_BACKING_CHAIN: 3046 chain = true; 3047 break; 3048 case OPTION_OBJECT: 3049 user_creatable_process_cmdline(optarg); 3050 break; 3051 case OPTION_IMAGE_OPTS: 3052 image_opts = true; 3053 break; 3054 } 3055 } 3056 if (optind != argc - 1) { 3057 error_exit("Expecting one image file name"); 3058 } 3059 filename = argv[optind++]; 3060 3061 if (output && !strcmp(output, "json")) { 3062 output_format = OFORMAT_JSON; 3063 } else if (output && !strcmp(output, "human")) { 3064 output_format = OFORMAT_HUMAN; 3065 } else if (output) { 3066 error_report("--output must be used with human or json as argument."); 3067 return 1; 3068 } 3069 3070 list = collect_image_info_list(image_opts, filename, fmt, chain, 3071 force_share); 3072 if (!list) { 3073 return 1; 3074 } 3075 3076 switch (output_format) { 3077 case OFORMAT_HUMAN: 3078 dump_human_image_info_list(list); 3079 break; 3080 case OFORMAT_JSON: 3081 if (chain) { 3082 dump_json_block_graph_info_list(list); 3083 } else { 3084 dump_json_block_graph_info(list->value); 3085 } 3086 break; 3087 } 3088 3089 qapi_free_BlockGraphInfoList(list); 3090 return 0; 3091 } 3092 3093 static int dump_map_entry(OutputFormat output_format, MapEntry *e, 3094 MapEntry *next) 3095 { 3096 switch (output_format) { 3097 case OFORMAT_HUMAN: 3098 if (e->data && !e->has_offset) { 3099 error_report("File contains external, encrypted or compressed clusters."); 3100 return -1; 3101 } 3102 if (e->data && !e->zero) { 3103 printf("%#-16"PRIx64"%#-16"PRIx64"%#-16"PRIx64"%s\n", 3104 e->start, e->length, 3105 e->has_offset ? e->offset : 0, 3106 e->filename ?: ""); 3107 } 3108 /* This format ignores the distinction between 0, ZERO and ZERO|DATA. 3109 * Modify the flags here to allow more coalescing. 3110 */ 3111 if (next && (!next->data || next->zero)) { 3112 next->data = false; 3113 next->zero = true; 3114 } 3115 break; 3116 case OFORMAT_JSON: 3117 printf("{ \"start\": %"PRId64", \"length\": %"PRId64"," 3118 " \"depth\": %"PRId64", \"present\": %s, \"zero\": %s," 3119 " \"data\": %s, \"compressed\": %s", 3120 e->start, e->length, e->depth, 3121 e->present ? "true" : "false", 3122 e->zero ? "true" : "false", 3123 e->data ? "true" : "false", 3124 e->compressed ? "true" : "false"); 3125 if (e->has_offset) { 3126 printf(", \"offset\": %"PRId64"", e->offset); 3127 } 3128 putchar('}'); 3129 3130 if (next) { 3131 puts(","); 3132 } 3133 break; 3134 } 3135 return 0; 3136 } 3137 3138 static int get_block_status(BlockDriverState *bs, int64_t offset, 3139 int64_t bytes, MapEntry *e) 3140 { 3141 int ret; 3142 int depth; 3143 BlockDriverState *file; 3144 bool has_offset; 3145 int64_t map; 3146 char *filename = NULL; 3147 3148 /* As an optimization, we could cache the current range of unallocated 3149 * clusters in each file of the chain, and avoid querying the same 3150 * range repeatedly. 3151 */ 3152 3153 depth = 0; 3154 for (;;) { 3155 bs = bdrv_skip_filters(bs); 3156 ret = bdrv_block_status(bs, offset, bytes, &bytes, &map, &file); 3157 if (ret < 0) { 3158 return ret; 3159 } 3160 assert(bytes); 3161 if (ret & (BDRV_BLOCK_ZERO|BDRV_BLOCK_DATA)) { 3162 break; 3163 } 3164 bs = bdrv_cow_bs(bs); 3165 if (bs == NULL) { 3166 ret = 0; 3167 break; 3168 } 3169 3170 depth++; 3171 } 3172 3173 has_offset = !!(ret & BDRV_BLOCK_OFFSET_VALID); 3174 3175 if (file && has_offset) { 3176 bdrv_graph_rdlock_main_loop(); 3177 bdrv_refresh_filename(file); 3178 bdrv_graph_rdunlock_main_loop(); 3179 filename = file->filename; 3180 } 3181 3182 *e = (MapEntry) { 3183 .start = offset, 3184 .length = bytes, 3185 .data = !!(ret & BDRV_BLOCK_DATA), 3186 .zero = !!(ret & BDRV_BLOCK_ZERO), 3187 .compressed = !!(ret & BDRV_BLOCK_COMPRESSED), 3188 .offset = map, 3189 .has_offset = has_offset, 3190 .depth = depth, 3191 .present = !!(ret & BDRV_BLOCK_ALLOCATED), 3192 .filename = filename, 3193 }; 3194 3195 return 0; 3196 } 3197 3198 static inline bool entry_mergeable(const MapEntry *curr, const MapEntry *next) 3199 { 3200 if (curr->length == 0) { 3201 return false; 3202 } 3203 if (curr->zero != next->zero || 3204 curr->data != next->data || 3205 curr->compressed != next->compressed || 3206 curr->depth != next->depth || 3207 curr->present != next->present || 3208 !curr->filename != !next->filename || 3209 curr->has_offset != next->has_offset) { 3210 return false; 3211 } 3212 if (curr->filename && strcmp(curr->filename, next->filename)) { 3213 return false; 3214 } 3215 if (curr->has_offset && curr->offset + curr->length != next->offset) { 3216 return false; 3217 } 3218 return true; 3219 } 3220 3221 static int img_map(int argc, char **argv) 3222 { 3223 int c; 3224 OutputFormat output_format = OFORMAT_HUMAN; 3225 BlockBackend *blk; 3226 BlockDriverState *bs; 3227 const char *filename, *fmt, *output; 3228 int64_t length; 3229 MapEntry curr = { .length = 0 }, next; 3230 int ret = 0; 3231 bool image_opts = false; 3232 bool force_share = false; 3233 int64_t start_offset = 0; 3234 int64_t max_length = -1; 3235 3236 fmt = NULL; 3237 output = NULL; 3238 for (;;) { 3239 int option_index = 0; 3240 static const struct option long_options[] = { 3241 {"help", no_argument, 0, 'h'}, 3242 {"format", required_argument, 0, 'f'}, 3243 {"output", required_argument, 0, OPTION_OUTPUT}, 3244 {"object", required_argument, 0, OPTION_OBJECT}, 3245 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS}, 3246 {"force-share", no_argument, 0, 'U'}, 3247 {"start-offset", required_argument, 0, 's'}, 3248 {"max-length", required_argument, 0, 'l'}, 3249 {0, 0, 0, 0} 3250 }; 3251 c = getopt_long(argc, argv, ":f:s:l:hU", 3252 long_options, &option_index); 3253 if (c == -1) { 3254 break; 3255 } 3256 switch (c) { 3257 case ':': 3258 missing_argument(argv[optind - 1]); 3259 break; 3260 case '?': 3261 unrecognized_option(argv[optind - 1]); 3262 break; 3263 case 'h': 3264 help(); 3265 break; 3266 case 'f': 3267 fmt = optarg; 3268 break; 3269 case 'U': 3270 force_share = true; 3271 break; 3272 case OPTION_OUTPUT: 3273 output = optarg; 3274 break; 3275 case 's': 3276 start_offset = cvtnum("start offset", optarg); 3277 if (start_offset < 0) { 3278 return 1; 3279 } 3280 break; 3281 case 'l': 3282 max_length = cvtnum("max length", optarg); 3283 if (max_length < 0) { 3284 return 1; 3285 } 3286 break; 3287 case OPTION_OBJECT: 3288 user_creatable_process_cmdline(optarg); 3289 break; 3290 case OPTION_IMAGE_OPTS: 3291 image_opts = true; 3292 break; 3293 } 3294 } 3295 if (optind != argc - 1) { 3296 error_exit("Expecting one image file name"); 3297 } 3298 filename = argv[optind]; 3299 3300 if (output && !strcmp(output, "json")) { 3301 output_format = OFORMAT_JSON; 3302 } else if (output && !strcmp(output, "human")) { 3303 output_format = OFORMAT_HUMAN; 3304 } else if (output) { 3305 error_report("--output must be used with human or json as argument."); 3306 return 1; 3307 } 3308 3309 blk = img_open(image_opts, filename, fmt, 0, false, false, force_share); 3310 if (!blk) { 3311 return 1; 3312 } 3313 bs = blk_bs(blk); 3314 3315 if (output_format == OFORMAT_HUMAN) { 3316 printf("%-16s%-16s%-16s%s\n", "Offset", "Length", "Mapped to", "File"); 3317 } else if (output_format == OFORMAT_JSON) { 3318 putchar('['); 3319 } 3320 3321 length = blk_getlength(blk); 3322 if (length < 0) { 3323 error_report("Failed to get size for '%s'", filename); 3324 return 1; 3325 } 3326 if (max_length != -1) { 3327 length = MIN(start_offset + max_length, length); 3328 } 3329 3330 curr.start = start_offset; 3331 while (curr.start + curr.length < length) { 3332 int64_t offset = curr.start + curr.length; 3333 int64_t n = length - offset; 3334 3335 ret = get_block_status(bs, offset, n, &next); 3336 if (ret < 0) { 3337 error_report("Could not read file metadata: %s", strerror(-ret)); 3338 goto out; 3339 } 3340 3341 if (entry_mergeable(&curr, &next)) { 3342 curr.length += next.length; 3343 continue; 3344 } 3345 3346 if (curr.length > 0) { 3347 ret = dump_map_entry(output_format, &curr, &next); 3348 if (ret < 0) { 3349 goto out; 3350 } 3351 } 3352 curr = next; 3353 } 3354 3355 ret = dump_map_entry(output_format, &curr, NULL); 3356 if (output_format == OFORMAT_JSON) { 3357 puts("]"); 3358 } 3359 3360 out: 3361 blk_unref(blk); 3362 return ret < 0; 3363 } 3364 3365 #define SNAPSHOT_LIST 1 3366 #define SNAPSHOT_CREATE 2 3367 #define SNAPSHOT_APPLY 3 3368 #define SNAPSHOT_DELETE 4 3369 3370 static int img_snapshot(int argc, char **argv) 3371 { 3372 BlockBackend *blk; 3373 BlockDriverState *bs; 3374 QEMUSnapshotInfo sn; 3375 char *filename, *snapshot_name = NULL; 3376 int c, ret = 0, bdrv_oflags; 3377 int action = 0; 3378 bool quiet = false; 3379 Error *err = NULL; 3380 bool image_opts = false; 3381 bool force_share = false; 3382 int64_t rt; 3383 3384 bdrv_oflags = BDRV_O_RDWR; 3385 /* Parse commandline parameters */ 3386 for(;;) { 3387 static const struct option long_options[] = { 3388 {"help", no_argument, 0, 'h'}, 3389 {"object", required_argument, 0, OPTION_OBJECT}, 3390 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS}, 3391 {"force-share", no_argument, 0, 'U'}, 3392 {0, 0, 0, 0} 3393 }; 3394 c = getopt_long(argc, argv, ":la:c:d:hqU", 3395 long_options, NULL); 3396 if (c == -1) { 3397 break; 3398 } 3399 switch(c) { 3400 case ':': 3401 missing_argument(argv[optind - 1]); 3402 break; 3403 case '?': 3404 unrecognized_option(argv[optind - 1]); 3405 break; 3406 case 'h': 3407 help(); 3408 return 0; 3409 case 'l': 3410 if (action) { 3411 error_exit("Cannot mix '-l', '-a', '-c', '-d'"); 3412 return 0; 3413 } 3414 action = SNAPSHOT_LIST; 3415 bdrv_oflags &= ~BDRV_O_RDWR; /* no need for RW */ 3416 break; 3417 case 'a': 3418 if (action) { 3419 error_exit("Cannot mix '-l', '-a', '-c', '-d'"); 3420 return 0; 3421 } 3422 action = SNAPSHOT_APPLY; 3423 snapshot_name = optarg; 3424 break; 3425 case 'c': 3426 if (action) { 3427 error_exit("Cannot mix '-l', '-a', '-c', '-d'"); 3428 return 0; 3429 } 3430 action = SNAPSHOT_CREATE; 3431 snapshot_name = optarg; 3432 break; 3433 case 'd': 3434 if (action) { 3435 error_exit("Cannot mix '-l', '-a', '-c', '-d'"); 3436 return 0; 3437 } 3438 action = SNAPSHOT_DELETE; 3439 snapshot_name = optarg; 3440 break; 3441 case 'q': 3442 quiet = true; 3443 break; 3444 case 'U': 3445 force_share = true; 3446 break; 3447 case OPTION_OBJECT: 3448 user_creatable_process_cmdline(optarg); 3449 break; 3450 case OPTION_IMAGE_OPTS: 3451 image_opts = true; 3452 break; 3453 } 3454 } 3455 3456 if (optind != argc - 1) { 3457 error_exit("Expecting one image file name"); 3458 } 3459 filename = argv[optind++]; 3460 3461 /* Open the image */ 3462 blk = img_open(image_opts, filename, NULL, bdrv_oflags, false, quiet, 3463 force_share); 3464 if (!blk) { 3465 return 1; 3466 } 3467 bs = blk_bs(blk); 3468 3469 /* Perform the requested action */ 3470 switch(action) { 3471 case SNAPSHOT_LIST: 3472 dump_snapshots(bs); 3473 break; 3474 3475 case SNAPSHOT_CREATE: 3476 memset(&sn, 0, sizeof(sn)); 3477 pstrcpy(sn.name, sizeof(sn.name), snapshot_name); 3478 3479 rt = g_get_real_time(); 3480 sn.date_sec = rt / G_USEC_PER_SEC; 3481 sn.date_nsec = (rt % G_USEC_PER_SEC) * 1000; 3482 3483 bdrv_graph_rdlock_main_loop(); 3484 ret = bdrv_snapshot_create(bs, &sn); 3485 bdrv_graph_rdunlock_main_loop(); 3486 3487 if (ret) { 3488 error_report("Could not create snapshot '%s': %s", 3489 snapshot_name, strerror(-ret)); 3490 } 3491 break; 3492 3493 case SNAPSHOT_APPLY: 3494 ret = bdrv_snapshot_goto(bs, snapshot_name, &err); 3495 if (ret) { 3496 error_reportf_err(err, "Could not apply snapshot '%s': ", 3497 snapshot_name); 3498 } 3499 break; 3500 3501 case SNAPSHOT_DELETE: 3502 bdrv_graph_rdlock_main_loop(); 3503 ret = bdrv_snapshot_find(bs, &sn, snapshot_name); 3504 if (ret < 0) { 3505 error_report("Could not delete snapshot '%s': snapshot not " 3506 "found", snapshot_name); 3507 ret = 1; 3508 } else { 3509 ret = bdrv_snapshot_delete(bs, sn.id_str, sn.name, &err); 3510 if (ret < 0) { 3511 error_reportf_err(err, "Could not delete snapshot '%s': ", 3512 snapshot_name); 3513 ret = 1; 3514 } 3515 } 3516 bdrv_graph_rdunlock_main_loop(); 3517 break; 3518 } 3519 3520 /* Cleanup */ 3521 blk_unref(blk); 3522 if (ret) { 3523 return 1; 3524 } 3525 return 0; 3526 } 3527 3528 static int img_rebase(int argc, char **argv) 3529 { 3530 BlockBackend *blk = NULL, *blk_old_backing = NULL, *blk_new_backing = NULL; 3531 uint8_t *buf_old = NULL; 3532 uint8_t *buf_new = NULL; 3533 BlockDriverState *bs = NULL, *prefix_chain_bs = NULL; 3534 BlockDriverState *unfiltered_bs; 3535 BlockDriverInfo bdi = {0}; 3536 char *filename; 3537 const char *fmt, *cache, *src_cache, *out_basefmt, *out_baseimg; 3538 int c, flags, src_flags, ret; 3539 BdrvRequestFlags write_flags = 0; 3540 bool writethrough, src_writethrough; 3541 int unsafe = 0; 3542 bool force_share = false; 3543 int progress = 0; 3544 bool quiet = false; 3545 bool compress = false; 3546 Error *local_err = NULL; 3547 bool image_opts = false; 3548 int64_t write_align; 3549 3550 /* Parse commandline parameters */ 3551 fmt = NULL; 3552 cache = BDRV_DEFAULT_CACHE; 3553 src_cache = BDRV_DEFAULT_CACHE; 3554 out_baseimg = NULL; 3555 out_basefmt = NULL; 3556 for(;;) { 3557 static const struct option long_options[] = { 3558 {"help", no_argument, 0, 'h'}, 3559 {"object", required_argument, 0, OPTION_OBJECT}, 3560 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS}, 3561 {"force-share", no_argument, 0, 'U'}, 3562 {"compress", no_argument, 0, 'c'}, 3563 {0, 0, 0, 0} 3564 }; 3565 c = getopt_long(argc, argv, ":hf:F:b:upt:T:qUc", 3566 long_options, NULL); 3567 if (c == -1) { 3568 break; 3569 } 3570 switch(c) { 3571 case ':': 3572 missing_argument(argv[optind - 1]); 3573 break; 3574 case '?': 3575 unrecognized_option(argv[optind - 1]); 3576 break; 3577 case 'h': 3578 help(); 3579 return 0; 3580 case 'f': 3581 fmt = optarg; 3582 break; 3583 case 'F': 3584 out_basefmt = optarg; 3585 break; 3586 case 'b': 3587 out_baseimg = optarg; 3588 break; 3589 case 'u': 3590 unsafe = 1; 3591 break; 3592 case 'p': 3593 progress = 1; 3594 break; 3595 case 't': 3596 cache = optarg; 3597 break; 3598 case 'T': 3599 src_cache = optarg; 3600 break; 3601 case 'q': 3602 quiet = true; 3603 break; 3604 case OPTION_OBJECT: 3605 user_creatable_process_cmdline(optarg); 3606 break; 3607 case OPTION_IMAGE_OPTS: 3608 image_opts = true; 3609 break; 3610 case 'U': 3611 force_share = true; 3612 break; 3613 case 'c': 3614 compress = true; 3615 break; 3616 } 3617 } 3618 3619 if (quiet) { 3620 progress = 0; 3621 } 3622 3623 if (optind != argc - 1) { 3624 error_exit("Expecting one image file name"); 3625 } 3626 if (!unsafe && !out_baseimg) { 3627 error_exit("Must specify backing file (-b) or use unsafe mode (-u)"); 3628 } 3629 filename = argv[optind++]; 3630 3631 qemu_progress_init(progress, 2.0); 3632 qemu_progress_print(0, 100); 3633 3634 flags = BDRV_O_RDWR | (unsafe ? BDRV_O_NO_BACKING : 0); 3635 ret = bdrv_parse_cache_mode(cache, &flags, &writethrough); 3636 if (ret < 0) { 3637 error_report("Invalid cache option: %s", cache); 3638 goto out; 3639 } 3640 3641 src_flags = 0; 3642 ret = bdrv_parse_cache_mode(src_cache, &src_flags, &src_writethrough); 3643 if (ret < 0) { 3644 error_report("Invalid source cache option: %s", src_cache); 3645 goto out; 3646 } 3647 3648 /* The source files are opened read-only, don't care about WCE */ 3649 assert((src_flags & BDRV_O_RDWR) == 0); 3650 (void) src_writethrough; 3651 3652 /* 3653 * Open the images. 3654 * 3655 * Ignore the old backing file for unsafe rebase in case we want to correct 3656 * the reference to a renamed or moved backing file. 3657 */ 3658 blk = img_open(image_opts, filename, fmt, flags, writethrough, quiet, 3659 false); 3660 if (!blk) { 3661 ret = -1; 3662 goto out; 3663 } 3664 bs = blk_bs(blk); 3665 3666 unfiltered_bs = bdrv_skip_filters(bs); 3667 3668 if (compress && !block_driver_can_compress(unfiltered_bs->drv)) { 3669 error_report("Compression not supported for this file format"); 3670 ret = -1; 3671 goto out; 3672 } else if (compress) { 3673 write_flags |= BDRV_REQ_WRITE_COMPRESSED; 3674 } 3675 3676 if (out_basefmt != NULL) { 3677 if (bdrv_find_format(out_basefmt) == NULL) { 3678 error_report("Invalid format name: '%s'", out_basefmt); 3679 ret = -1; 3680 goto out; 3681 } 3682 } 3683 3684 /* 3685 * We need overlay subcluster size (or cluster size in case writes are 3686 * compressed) to make sure write requests are aligned. 3687 */ 3688 ret = bdrv_get_info(unfiltered_bs, &bdi); 3689 if (ret < 0) { 3690 error_report("could not get block driver info"); 3691 goto out; 3692 } else if (bdi.subcluster_size == 0) { 3693 bdi.cluster_size = bdi.subcluster_size = 1; 3694 } 3695 3696 write_align = compress ? bdi.cluster_size : bdi.subcluster_size; 3697 3698 /* For safe rebasing we need to compare old and new backing file */ 3699 if (!unsafe) { 3700 QDict *options = NULL; 3701 BlockDriverState *base_bs = bdrv_cow_bs(unfiltered_bs); 3702 3703 if (base_bs) { 3704 blk_old_backing = blk_new(qemu_get_aio_context(), 3705 BLK_PERM_CONSISTENT_READ, 3706 BLK_PERM_ALL); 3707 ret = blk_insert_bs(blk_old_backing, base_bs, 3708 &local_err); 3709 if (ret < 0) { 3710 error_reportf_err(local_err, 3711 "Could not reuse old backing file '%s': ", 3712 base_bs->filename); 3713 goto out; 3714 } 3715 } else { 3716 blk_old_backing = NULL; 3717 } 3718 3719 if (out_baseimg[0]) { 3720 const char *overlay_filename; 3721 char *out_real_path; 3722 3723 options = qdict_new(); 3724 if (out_basefmt) { 3725 qdict_put_str(options, "driver", out_basefmt); 3726 } 3727 if (force_share) { 3728 qdict_put_bool(options, BDRV_OPT_FORCE_SHARE, true); 3729 } 3730 3731 bdrv_graph_rdlock_main_loop(); 3732 bdrv_refresh_filename(bs); 3733 bdrv_graph_rdunlock_main_loop(); 3734 overlay_filename = bs->exact_filename[0] ? bs->exact_filename 3735 : bs->filename; 3736 out_real_path = 3737 bdrv_get_full_backing_filename_from_filename(overlay_filename, 3738 out_baseimg, 3739 &local_err); 3740 if (local_err) { 3741 qobject_unref(options); 3742 error_reportf_err(local_err, 3743 "Could not resolve backing filename: "); 3744 ret = -1; 3745 goto out; 3746 } 3747 3748 /* 3749 * Find out whether we rebase an image on top of a previous image 3750 * in its chain. 3751 */ 3752 prefix_chain_bs = bdrv_find_backing_image(bs, out_real_path); 3753 if (prefix_chain_bs) { 3754 qobject_unref(options); 3755 g_free(out_real_path); 3756 3757 blk_new_backing = blk_new(qemu_get_aio_context(), 3758 BLK_PERM_CONSISTENT_READ, 3759 BLK_PERM_ALL); 3760 ret = blk_insert_bs(blk_new_backing, prefix_chain_bs, 3761 &local_err); 3762 if (ret < 0) { 3763 error_reportf_err(local_err, 3764 "Could not reuse backing file '%s': ", 3765 out_baseimg); 3766 goto out; 3767 } 3768 } else { 3769 blk_new_backing = blk_new_open(out_real_path, NULL, 3770 options, src_flags, &local_err); 3771 g_free(out_real_path); 3772 if (!blk_new_backing) { 3773 error_reportf_err(local_err, 3774 "Could not open new backing file '%s': ", 3775 out_baseimg); 3776 ret = -1; 3777 goto out; 3778 } 3779 } 3780 } 3781 } 3782 3783 /* 3784 * Check each unallocated cluster in the COW file. If it is unallocated, 3785 * accesses go to the backing file. We must therefore compare this cluster 3786 * in the old and new backing file, and if they differ we need to copy it 3787 * from the old backing file into the COW file. 3788 * 3789 * If qemu-img crashes during this step, no harm is done. The content of 3790 * the image is the same as the original one at any time. 3791 */ 3792 if (!unsafe) { 3793 int64_t size; 3794 int64_t old_backing_size = 0; 3795 int64_t new_backing_size = 0; 3796 uint64_t offset; 3797 int64_t n, n_old = 0, n_new = 0; 3798 float local_progress = 0; 3799 3800 if (blk_old_backing && bdrv_opt_mem_align(blk_bs(blk_old_backing)) > 3801 bdrv_opt_mem_align(blk_bs(blk))) { 3802 buf_old = blk_blockalign(blk_old_backing, IO_BUF_SIZE); 3803 } else { 3804 buf_old = blk_blockalign(blk, IO_BUF_SIZE); 3805 } 3806 buf_new = blk_blockalign(blk_new_backing, IO_BUF_SIZE); 3807 3808 size = blk_getlength(blk); 3809 if (size < 0) { 3810 error_report("Could not get size of '%s': %s", 3811 filename, strerror(-size)); 3812 ret = -1; 3813 goto out; 3814 } 3815 if (blk_old_backing) { 3816 old_backing_size = blk_getlength(blk_old_backing); 3817 if (old_backing_size < 0) { 3818 char backing_name[PATH_MAX]; 3819 3820 bdrv_get_backing_filename(bs, backing_name, 3821 sizeof(backing_name)); 3822 error_report("Could not get size of '%s': %s", 3823 backing_name, strerror(-old_backing_size)); 3824 ret = -1; 3825 goto out; 3826 } 3827 } 3828 if (blk_new_backing) { 3829 new_backing_size = blk_getlength(blk_new_backing); 3830 if (new_backing_size < 0) { 3831 error_report("Could not get size of '%s': %s", 3832 out_baseimg, strerror(-new_backing_size)); 3833 ret = -1; 3834 goto out; 3835 } 3836 } 3837 3838 if (size != 0) { 3839 local_progress = (float)100 / (size / MIN(size, IO_BUF_SIZE)); 3840 } 3841 3842 for (offset = 0; offset < size; offset += n) { 3843 bool old_backing_eof = false; 3844 int64_t n_alloc; 3845 3846 /* How many bytes can we handle with the next read? */ 3847 n = MIN(IO_BUF_SIZE, size - offset); 3848 3849 /* If the cluster is allocated, we don't need to take action */ 3850 ret = bdrv_is_allocated(unfiltered_bs, offset, n, &n); 3851 if (ret < 0) { 3852 error_report("error while reading image metadata: %s", 3853 strerror(-ret)); 3854 goto out; 3855 } 3856 if (ret) { 3857 continue; 3858 } 3859 3860 if (prefix_chain_bs) { 3861 uint64_t bytes = n; 3862 3863 /* 3864 * If cluster wasn't changed since prefix_chain, we don't need 3865 * to take action 3866 */ 3867 ret = bdrv_is_allocated_above(bdrv_cow_bs(unfiltered_bs), 3868 prefix_chain_bs, false, 3869 offset, n, &n); 3870 if (ret < 0) { 3871 error_report("error while reading image metadata: %s", 3872 strerror(-ret)); 3873 goto out; 3874 } 3875 if (!ret && n) { 3876 continue; 3877 } 3878 if (!n) { 3879 /* 3880 * If we've reached EOF of the old backing, it means that 3881 * offsets beyond the old backing size were read as zeroes. 3882 * Now we will need to explicitly zero the cluster in 3883 * order to preserve that state after the rebase. 3884 */ 3885 n = bytes; 3886 } 3887 } 3888 3889 /* 3890 * At this point we know that the region [offset; offset + n) 3891 * is unallocated within the target image. This region might be 3892 * unaligned to the target image's (sub)cluster boundaries, as 3893 * old backing may have smaller clusters (or have subclusters). 3894 * We extend it to the aligned boundaries to avoid CoW on 3895 * partial writes in blk_pwrite(), 3896 */ 3897 n += offset - QEMU_ALIGN_DOWN(offset, write_align); 3898 offset = QEMU_ALIGN_DOWN(offset, write_align); 3899 n += QEMU_ALIGN_UP(offset + n, write_align) - (offset + n); 3900 n = MIN(n, size - offset); 3901 assert(!bdrv_is_allocated(unfiltered_bs, offset, n, &n_alloc) && 3902 n_alloc == n); 3903 3904 /* 3905 * Much like with the target image, we'll try to read as much 3906 * of the old and new backings as we can. 3907 */ 3908 n_old = MIN(n, MAX(0, old_backing_size - (int64_t) offset)); 3909 n_new = MIN(n, MAX(0, new_backing_size - (int64_t) offset)); 3910 3911 /* 3912 * Read old and new backing file and take into consideration that 3913 * backing files may be smaller than the COW image. 3914 */ 3915 memset(buf_old + n_old, 0, n - n_old); 3916 if (!n_old) { 3917 old_backing_eof = true; 3918 } else { 3919 ret = blk_pread(blk_old_backing, offset, n_old, buf_old, 0); 3920 if (ret < 0) { 3921 error_report("error while reading from old backing file"); 3922 goto out; 3923 } 3924 } 3925 3926 memset(buf_new + n_new, 0, n - n_new); 3927 if (n_new) { 3928 ret = blk_pread(blk_new_backing, offset, n_new, buf_new, 0); 3929 if (ret < 0) { 3930 error_report("error while reading from new backing file"); 3931 goto out; 3932 } 3933 } 3934 3935 /* If they differ, we need to write to the COW file */ 3936 uint64_t written = 0; 3937 3938 while (written < n) { 3939 int64_t pnum; 3940 3941 if (compare_buffers(buf_old + written, buf_new + written, 3942 n - written, write_align, &pnum)) 3943 { 3944 if (old_backing_eof) { 3945 ret = blk_pwrite_zeroes(blk, offset + written, pnum, 0); 3946 } else { 3947 assert(written + pnum <= IO_BUF_SIZE); 3948 ret = blk_pwrite(blk, offset + written, pnum, 3949 buf_old + written, write_flags); 3950 } 3951 if (ret < 0) { 3952 error_report("Error while writing to COW image: %s", 3953 strerror(-ret)); 3954 goto out; 3955 } 3956 } 3957 3958 written += pnum; 3959 if (offset + written >= old_backing_size) { 3960 old_backing_eof = true; 3961 } 3962 } 3963 qemu_progress_print(local_progress, 100); 3964 } 3965 } 3966 3967 /* 3968 * Change the backing file. All clusters that are different from the old 3969 * backing file are overwritten in the COW file now, so the visible content 3970 * doesn't change when we switch the backing file. 3971 */ 3972 if (out_baseimg && *out_baseimg) { 3973 ret = bdrv_change_backing_file(unfiltered_bs, out_baseimg, out_basefmt, 3974 true); 3975 } else { 3976 ret = bdrv_change_backing_file(unfiltered_bs, NULL, NULL, false); 3977 } 3978 3979 if (ret == -ENOSPC) { 3980 error_report("Could not change the backing file to '%s': No " 3981 "space left in the file header", out_baseimg); 3982 } else if (ret == -EINVAL && out_baseimg && !out_basefmt) { 3983 error_report("Could not change the backing file to '%s': backing " 3984 "format must be specified", out_baseimg); 3985 } else if (ret < 0) { 3986 error_report("Could not change the backing file to '%s': %s", 3987 out_baseimg, strerror(-ret)); 3988 } 3989 3990 qemu_progress_print(100, 0); 3991 /* 3992 * TODO At this point it is possible to check if any clusters that are 3993 * allocated in the COW file are the same in the backing file. If so, they 3994 * could be dropped from the COW file. Don't do this before switching the 3995 * backing file, in case of a crash this would lead to corruption. 3996 */ 3997 out: 3998 qemu_progress_end(); 3999 /* Cleanup */ 4000 if (!unsafe) { 4001 blk_unref(blk_old_backing); 4002 blk_unref(blk_new_backing); 4003 } 4004 qemu_vfree(buf_old); 4005 qemu_vfree(buf_new); 4006 4007 blk_unref(blk); 4008 if (ret) { 4009 return 1; 4010 } 4011 return 0; 4012 } 4013 4014 static int img_resize(int argc, char **argv) 4015 { 4016 Error *err = NULL; 4017 int c, ret, relative; 4018 const char *filename, *fmt, *size; 4019 int64_t n, total_size, current_size; 4020 bool quiet = false; 4021 BlockBackend *blk = NULL; 4022 PreallocMode prealloc = PREALLOC_MODE_OFF; 4023 QemuOpts *param; 4024 4025 static QemuOptsList resize_options = { 4026 .name = "resize_options", 4027 .head = QTAILQ_HEAD_INITIALIZER(resize_options.head), 4028 .desc = { 4029 { 4030 .name = BLOCK_OPT_SIZE, 4031 .type = QEMU_OPT_SIZE, 4032 .help = "Virtual disk size" 4033 }, { 4034 /* end of list */ 4035 } 4036 }, 4037 }; 4038 bool image_opts = false; 4039 bool shrink = false; 4040 4041 /* Remove size from argv manually so that negative numbers are not treated 4042 * as options by getopt. */ 4043 if (argc < 3) { 4044 error_exit("Not enough arguments"); 4045 return 1; 4046 } 4047 4048 size = argv[--argc]; 4049 4050 /* Parse getopt arguments */ 4051 fmt = NULL; 4052 for(;;) { 4053 static const struct option long_options[] = { 4054 {"help", no_argument, 0, 'h'}, 4055 {"object", required_argument, 0, OPTION_OBJECT}, 4056 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS}, 4057 {"preallocation", required_argument, 0, OPTION_PREALLOCATION}, 4058 {"shrink", no_argument, 0, OPTION_SHRINK}, 4059 {0, 0, 0, 0} 4060 }; 4061 c = getopt_long(argc, argv, ":f:hq", 4062 long_options, NULL); 4063 if (c == -1) { 4064 break; 4065 } 4066 switch(c) { 4067 case ':': 4068 missing_argument(argv[optind - 1]); 4069 break; 4070 case '?': 4071 unrecognized_option(argv[optind - 1]); 4072 break; 4073 case 'h': 4074 help(); 4075 break; 4076 case 'f': 4077 fmt = optarg; 4078 break; 4079 case 'q': 4080 quiet = true; 4081 break; 4082 case OPTION_OBJECT: 4083 user_creatable_process_cmdline(optarg); 4084 break; 4085 case OPTION_IMAGE_OPTS: 4086 image_opts = true; 4087 break; 4088 case OPTION_PREALLOCATION: 4089 prealloc = qapi_enum_parse(&PreallocMode_lookup, optarg, 4090 PREALLOC_MODE__MAX, NULL); 4091 if (prealloc == PREALLOC_MODE__MAX) { 4092 error_report("Invalid preallocation mode '%s'", optarg); 4093 return 1; 4094 } 4095 break; 4096 case OPTION_SHRINK: 4097 shrink = true; 4098 break; 4099 } 4100 } 4101 if (optind != argc - 1) { 4102 error_exit("Expecting image file name and size"); 4103 } 4104 filename = argv[optind++]; 4105 4106 /* Choose grow, shrink, or absolute resize mode */ 4107 switch (size[0]) { 4108 case '+': 4109 relative = 1; 4110 size++; 4111 break; 4112 case '-': 4113 relative = -1; 4114 size++; 4115 break; 4116 default: 4117 relative = 0; 4118 break; 4119 } 4120 4121 /* Parse size */ 4122 param = qemu_opts_create(&resize_options, NULL, 0, &error_abort); 4123 if (!qemu_opt_set(param, BLOCK_OPT_SIZE, size, &err)) { 4124 error_report_err(err); 4125 ret = -1; 4126 qemu_opts_del(param); 4127 goto out; 4128 } 4129 n = qemu_opt_get_size(param, BLOCK_OPT_SIZE, 0); 4130 qemu_opts_del(param); 4131 4132 blk = img_open(image_opts, filename, fmt, 4133 BDRV_O_RDWR | BDRV_O_RESIZE, false, quiet, 4134 false); 4135 if (!blk) { 4136 ret = -1; 4137 goto out; 4138 } 4139 4140 current_size = blk_getlength(blk); 4141 if (current_size < 0) { 4142 error_report("Failed to inquire current image length: %s", 4143 strerror(-current_size)); 4144 ret = -1; 4145 goto out; 4146 } 4147 4148 if (relative) { 4149 total_size = current_size + n * relative; 4150 } else { 4151 total_size = n; 4152 } 4153 if (total_size <= 0) { 4154 error_report("New image size must be positive"); 4155 ret = -1; 4156 goto out; 4157 } 4158 4159 if (total_size <= current_size && prealloc != PREALLOC_MODE_OFF) { 4160 error_report("Preallocation can only be used for growing images"); 4161 ret = -1; 4162 goto out; 4163 } 4164 4165 if (total_size < current_size && !shrink) { 4166 error_report("Use the --shrink option to perform a shrink operation."); 4167 warn_report("Shrinking an image will delete all data beyond the " 4168 "shrunken image's end. Before performing such an " 4169 "operation, make sure there is no important data there."); 4170 ret = -1; 4171 goto out; 4172 } 4173 4174 /* 4175 * The user expects the image to have the desired size after 4176 * resizing, so pass @exact=true. It is of no use to report 4177 * success when the image has not actually been resized. 4178 */ 4179 ret = blk_truncate(blk, total_size, true, prealloc, 0, &err); 4180 if (!ret) { 4181 qprintf(quiet, "Image resized.\n"); 4182 } else { 4183 error_report_err(err); 4184 } 4185 out: 4186 blk_unref(blk); 4187 if (ret) { 4188 return 1; 4189 } 4190 return 0; 4191 } 4192 4193 static void amend_status_cb(BlockDriverState *bs, 4194 int64_t offset, int64_t total_work_size, 4195 void *opaque) 4196 { 4197 qemu_progress_print(100.f * offset / total_work_size, 0); 4198 } 4199 4200 static int print_amend_option_help(const char *format) 4201 { 4202 BlockDriver *drv; 4203 4204 GRAPH_RDLOCK_GUARD_MAINLOOP(); 4205 4206 /* Find driver and parse its options */ 4207 drv = bdrv_find_format(format); 4208 if (!drv) { 4209 error_report("Unknown file format '%s'", format); 4210 return 1; 4211 } 4212 4213 if (!drv->bdrv_amend_options) { 4214 error_report("Format driver '%s' does not support option amendment", 4215 format); 4216 return 1; 4217 } 4218 4219 /* Every driver supporting amendment must have amend_opts */ 4220 assert(drv->amend_opts); 4221 4222 printf("Amend options for '%s':\n", format); 4223 qemu_opts_print_help(drv->amend_opts, false); 4224 return 0; 4225 } 4226 4227 static int img_amend(int argc, char **argv) 4228 { 4229 Error *err = NULL; 4230 int c, ret = 0; 4231 char *options = NULL; 4232 QemuOptsList *amend_opts = NULL; 4233 QemuOpts *opts = NULL; 4234 const char *fmt = NULL, *filename, *cache; 4235 int flags; 4236 bool writethrough; 4237 bool quiet = false, progress = false; 4238 BlockBackend *blk = NULL; 4239 BlockDriverState *bs = NULL; 4240 bool image_opts = false; 4241 bool force = false; 4242 4243 cache = BDRV_DEFAULT_CACHE; 4244 for (;;) { 4245 static const struct option long_options[] = { 4246 {"help", no_argument, 0, 'h'}, 4247 {"object", required_argument, 0, OPTION_OBJECT}, 4248 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS}, 4249 {"force", no_argument, 0, OPTION_FORCE}, 4250 {0, 0, 0, 0} 4251 }; 4252 c = getopt_long(argc, argv, ":ho:f:t:pq", 4253 long_options, NULL); 4254 if (c == -1) { 4255 break; 4256 } 4257 4258 switch (c) { 4259 case ':': 4260 missing_argument(argv[optind - 1]); 4261 break; 4262 case '?': 4263 unrecognized_option(argv[optind - 1]); 4264 break; 4265 case 'h': 4266 help(); 4267 break; 4268 case 'o': 4269 if (accumulate_options(&options, optarg) < 0) { 4270 ret = -1; 4271 goto out_no_progress; 4272 } 4273 break; 4274 case 'f': 4275 fmt = optarg; 4276 break; 4277 case 't': 4278 cache = optarg; 4279 break; 4280 case 'p': 4281 progress = true; 4282 break; 4283 case 'q': 4284 quiet = true; 4285 break; 4286 case OPTION_OBJECT: 4287 user_creatable_process_cmdline(optarg); 4288 break; 4289 case OPTION_IMAGE_OPTS: 4290 image_opts = true; 4291 break; 4292 case OPTION_FORCE: 4293 force = true; 4294 break; 4295 } 4296 } 4297 4298 if (!options) { 4299 error_exit("Must specify options (-o)"); 4300 } 4301 4302 if (quiet) { 4303 progress = false; 4304 } 4305 qemu_progress_init(progress, 1.0); 4306 4307 filename = (optind == argc - 1) ? argv[argc - 1] : NULL; 4308 if (fmt && has_help_option(options)) { 4309 /* If a format is explicitly specified (and possibly no filename is 4310 * given), print option help here */ 4311 ret = print_amend_option_help(fmt); 4312 goto out; 4313 } 4314 4315 if (optind != argc - 1) { 4316 error_report("Expecting one image file name"); 4317 ret = -1; 4318 goto out; 4319 } 4320 4321 flags = BDRV_O_RDWR; 4322 ret = bdrv_parse_cache_mode(cache, &flags, &writethrough); 4323 if (ret < 0) { 4324 error_report("Invalid cache option: %s", cache); 4325 goto out; 4326 } 4327 4328 blk = img_open(image_opts, filename, fmt, flags, writethrough, quiet, 4329 false); 4330 if (!blk) { 4331 ret = -1; 4332 goto out; 4333 } 4334 bs = blk_bs(blk); 4335 4336 fmt = bs->drv->format_name; 4337 4338 if (has_help_option(options)) { 4339 /* If the format was auto-detected, print option help here */ 4340 ret = print_amend_option_help(fmt); 4341 goto out; 4342 } 4343 4344 bdrv_graph_rdlock_main_loop(); 4345 if (!bs->drv->bdrv_amend_options) { 4346 error_report("Format driver '%s' does not support option amendment", 4347 fmt); 4348 bdrv_graph_rdunlock_main_loop(); 4349 ret = -1; 4350 goto out; 4351 } 4352 4353 /* Every driver supporting amendment must have amend_opts */ 4354 assert(bs->drv->amend_opts); 4355 4356 amend_opts = qemu_opts_append(amend_opts, bs->drv->amend_opts); 4357 opts = qemu_opts_create(amend_opts, NULL, 0, &error_abort); 4358 if (!qemu_opts_do_parse(opts, options, NULL, &err)) { 4359 /* Try to parse options using the create options */ 4360 amend_opts = qemu_opts_append(amend_opts, bs->drv->create_opts); 4361 qemu_opts_del(opts); 4362 opts = qemu_opts_create(amend_opts, NULL, 0, &error_abort); 4363 if (qemu_opts_do_parse(opts, options, NULL, NULL)) { 4364 error_append_hint(&err, 4365 "This option is only supported for image creation\n"); 4366 } 4367 4368 bdrv_graph_rdunlock_main_loop(); 4369 error_report_err(err); 4370 ret = -1; 4371 goto out; 4372 } 4373 4374 /* In case the driver does not call amend_status_cb() */ 4375 qemu_progress_print(0.f, 0); 4376 ret = bdrv_amend_options(bs, opts, &amend_status_cb, NULL, force, &err); 4377 qemu_progress_print(100.f, 0); 4378 bdrv_graph_rdunlock_main_loop(); 4379 4380 if (ret < 0) { 4381 error_report_err(err); 4382 goto out; 4383 } 4384 4385 out: 4386 qemu_progress_end(); 4387 4388 out_no_progress: 4389 blk_unref(blk); 4390 qemu_opts_del(opts); 4391 qemu_opts_free(amend_opts); 4392 g_free(options); 4393 4394 if (ret) { 4395 return 1; 4396 } 4397 return 0; 4398 } 4399 4400 typedef struct BenchData { 4401 BlockBackend *blk; 4402 uint64_t image_size; 4403 bool write; 4404 int bufsize; 4405 int step; 4406 int nrreq; 4407 int n; 4408 int flush_interval; 4409 bool drain_on_flush; 4410 uint8_t *buf; 4411 QEMUIOVector *qiov; 4412 4413 int in_flight; 4414 bool in_flush; 4415 uint64_t offset; 4416 } BenchData; 4417 4418 static void bench_undrained_flush_cb(void *opaque, int ret) 4419 { 4420 if (ret < 0) { 4421 error_report("Failed flush request: %s", strerror(-ret)); 4422 exit(EXIT_FAILURE); 4423 } 4424 } 4425 4426 static void bench_cb(void *opaque, int ret) 4427 { 4428 BenchData *b = opaque; 4429 BlockAIOCB *acb; 4430 4431 if (ret < 0) { 4432 error_report("Failed request: %s", strerror(-ret)); 4433 exit(EXIT_FAILURE); 4434 } 4435 4436 if (b->in_flush) { 4437 /* Just finished a flush with drained queue: Start next requests */ 4438 assert(b->in_flight == 0); 4439 b->in_flush = false; 4440 } else if (b->in_flight > 0) { 4441 int remaining = b->n - b->in_flight; 4442 4443 b->n--; 4444 b->in_flight--; 4445 4446 /* Time for flush? Drain queue if requested, then flush */ 4447 if (b->flush_interval && remaining % b->flush_interval == 0) { 4448 if (!b->in_flight || !b->drain_on_flush) { 4449 BlockCompletionFunc *cb; 4450 4451 if (b->drain_on_flush) { 4452 b->in_flush = true; 4453 cb = bench_cb; 4454 } else { 4455 cb = bench_undrained_flush_cb; 4456 } 4457 4458 acb = blk_aio_flush(b->blk, cb, b); 4459 if (!acb) { 4460 error_report("Failed to issue flush request"); 4461 exit(EXIT_FAILURE); 4462 } 4463 } 4464 if (b->drain_on_flush) { 4465 return; 4466 } 4467 } 4468 } 4469 4470 while (b->n > b->in_flight && b->in_flight < b->nrreq) { 4471 int64_t offset = b->offset; 4472 /* blk_aio_* might look for completed I/Os and kick bench_cb 4473 * again, so make sure this operation is counted by in_flight 4474 * and b->offset is ready for the next submission. 4475 */ 4476 b->in_flight++; 4477 b->offset += b->step; 4478 b->offset %= b->image_size; 4479 if (b->write) { 4480 acb = blk_aio_pwritev(b->blk, offset, b->qiov, 0, bench_cb, b); 4481 } else { 4482 acb = blk_aio_preadv(b->blk, offset, b->qiov, 0, bench_cb, b); 4483 } 4484 if (!acb) { 4485 error_report("Failed to issue request"); 4486 exit(EXIT_FAILURE); 4487 } 4488 } 4489 } 4490 4491 static int img_bench(int argc, char **argv) 4492 { 4493 int c, ret = 0; 4494 const char *fmt = NULL, *filename; 4495 bool quiet = false; 4496 bool image_opts = false; 4497 bool is_write = false; 4498 int count = 75000; 4499 int depth = 64; 4500 int64_t offset = 0; 4501 size_t bufsize = 4096; 4502 int pattern = 0; 4503 size_t step = 0; 4504 int flush_interval = 0; 4505 bool drain_on_flush = true; 4506 int64_t image_size; 4507 BlockBackend *blk = NULL; 4508 BenchData data = {}; 4509 int flags = 0; 4510 bool writethrough = false; 4511 struct timeval t1, t2; 4512 int i; 4513 bool force_share = false; 4514 size_t buf_size = 0; 4515 4516 for (;;) { 4517 static const struct option long_options[] = { 4518 {"help", no_argument, 0, 'h'}, 4519 {"flush-interval", required_argument, 0, OPTION_FLUSH_INTERVAL}, 4520 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS}, 4521 {"pattern", required_argument, 0, OPTION_PATTERN}, 4522 {"no-drain", no_argument, 0, OPTION_NO_DRAIN}, 4523 {"force-share", no_argument, 0, 'U'}, 4524 {0, 0, 0, 0} 4525 }; 4526 c = getopt_long(argc, argv, ":hc:d:f:ni:o:qs:S:t:wU", long_options, 4527 NULL); 4528 if (c == -1) { 4529 break; 4530 } 4531 4532 switch (c) { 4533 case ':': 4534 missing_argument(argv[optind - 1]); 4535 break; 4536 case '?': 4537 unrecognized_option(argv[optind - 1]); 4538 break; 4539 case 'h': 4540 help(); 4541 break; 4542 case 'c': 4543 { 4544 unsigned long res; 4545 4546 if (qemu_strtoul(optarg, NULL, 0, &res) < 0 || res > INT_MAX) { 4547 error_report("Invalid request count specified"); 4548 return 1; 4549 } 4550 count = res; 4551 break; 4552 } 4553 case 'd': 4554 { 4555 unsigned long res; 4556 4557 if (qemu_strtoul(optarg, NULL, 0, &res) < 0 || res > INT_MAX) { 4558 error_report("Invalid queue depth specified"); 4559 return 1; 4560 } 4561 depth = res; 4562 break; 4563 } 4564 case 'f': 4565 fmt = optarg; 4566 break; 4567 case 'n': 4568 flags |= BDRV_O_NATIVE_AIO; 4569 break; 4570 case 'i': 4571 ret = bdrv_parse_aio(optarg, &flags); 4572 if (ret < 0) { 4573 error_report("Invalid aio option: %s", optarg); 4574 ret = -1; 4575 goto out; 4576 } 4577 break; 4578 case 'o': 4579 { 4580 offset = cvtnum("offset", optarg); 4581 if (offset < 0) { 4582 return 1; 4583 } 4584 break; 4585 } 4586 break; 4587 case 'q': 4588 quiet = true; 4589 break; 4590 case 's': 4591 { 4592 int64_t sval; 4593 4594 sval = cvtnum_full("buffer size", optarg, 0, INT_MAX); 4595 if (sval < 0) { 4596 return 1; 4597 } 4598 4599 bufsize = sval; 4600 break; 4601 } 4602 case 'S': 4603 { 4604 int64_t sval; 4605 4606 sval = cvtnum_full("step_size", optarg, 0, INT_MAX); 4607 if (sval < 0) { 4608 return 1; 4609 } 4610 4611 step = sval; 4612 break; 4613 } 4614 case 't': 4615 ret = bdrv_parse_cache_mode(optarg, &flags, &writethrough); 4616 if (ret < 0) { 4617 error_report("Invalid cache mode"); 4618 ret = -1; 4619 goto out; 4620 } 4621 break; 4622 case 'w': 4623 flags |= BDRV_O_RDWR; 4624 is_write = true; 4625 break; 4626 case 'U': 4627 force_share = true; 4628 break; 4629 case OPTION_PATTERN: 4630 { 4631 unsigned long res; 4632 4633 if (qemu_strtoul(optarg, NULL, 0, &res) < 0 || res > 0xff) { 4634 error_report("Invalid pattern byte specified"); 4635 return 1; 4636 } 4637 pattern = res; 4638 break; 4639 } 4640 case OPTION_FLUSH_INTERVAL: 4641 { 4642 unsigned long res; 4643 4644 if (qemu_strtoul(optarg, NULL, 0, &res) < 0 || res > INT_MAX) { 4645 error_report("Invalid flush interval specified"); 4646 return 1; 4647 } 4648 flush_interval = res; 4649 break; 4650 } 4651 case OPTION_NO_DRAIN: 4652 drain_on_flush = false; 4653 break; 4654 case OPTION_IMAGE_OPTS: 4655 image_opts = true; 4656 break; 4657 } 4658 } 4659 4660 if (optind != argc - 1) { 4661 error_exit("Expecting one image file name"); 4662 } 4663 filename = argv[argc - 1]; 4664 4665 if (!is_write && flush_interval) { 4666 error_report("--flush-interval is only available in write tests"); 4667 ret = -1; 4668 goto out; 4669 } 4670 if (flush_interval && flush_interval < depth) { 4671 error_report("Flush interval can't be smaller than depth"); 4672 ret = -1; 4673 goto out; 4674 } 4675 4676 blk = img_open(image_opts, filename, fmt, flags, writethrough, quiet, 4677 force_share); 4678 if (!blk) { 4679 ret = -1; 4680 goto out; 4681 } 4682 4683 image_size = blk_getlength(blk); 4684 if (image_size < 0) { 4685 ret = image_size; 4686 goto out; 4687 } 4688 4689 data = (BenchData) { 4690 .blk = blk, 4691 .image_size = image_size, 4692 .bufsize = bufsize, 4693 .step = step ?: bufsize, 4694 .nrreq = depth, 4695 .n = count, 4696 .offset = offset, 4697 .write = is_write, 4698 .flush_interval = flush_interval, 4699 .drain_on_flush = drain_on_flush, 4700 }; 4701 printf("Sending %d %s requests, %d bytes each, %d in parallel " 4702 "(starting at offset %" PRId64 ", step size %d)\n", 4703 data.n, data.write ? "write" : "read", data.bufsize, data.nrreq, 4704 data.offset, data.step); 4705 if (flush_interval) { 4706 printf("Sending flush every %d requests\n", flush_interval); 4707 } 4708 4709 buf_size = data.nrreq * data.bufsize; 4710 data.buf = blk_blockalign(blk, buf_size); 4711 memset(data.buf, pattern, data.nrreq * data.bufsize); 4712 4713 blk_register_buf(blk, data.buf, buf_size, &error_fatal); 4714 4715 data.qiov = g_new(QEMUIOVector, data.nrreq); 4716 for (i = 0; i < data.nrreq; i++) { 4717 qemu_iovec_init(&data.qiov[i], 1); 4718 qemu_iovec_add(&data.qiov[i], 4719 data.buf + i * data.bufsize, data.bufsize); 4720 } 4721 4722 gettimeofday(&t1, NULL); 4723 bench_cb(&data, 0); 4724 4725 while (data.n > 0) { 4726 main_loop_wait(false); 4727 } 4728 gettimeofday(&t2, NULL); 4729 4730 printf("Run completed in %3.3f seconds.\n", 4731 (t2.tv_sec - t1.tv_sec) 4732 + ((double)(t2.tv_usec - t1.tv_usec) / 1000000)); 4733 4734 out: 4735 if (data.buf) { 4736 blk_unregister_buf(blk, data.buf, buf_size); 4737 } 4738 qemu_vfree(data.buf); 4739 blk_unref(blk); 4740 4741 if (ret) { 4742 return 1; 4743 } 4744 return 0; 4745 } 4746 4747 enum ImgBitmapAct { 4748 BITMAP_ADD, 4749 BITMAP_REMOVE, 4750 BITMAP_CLEAR, 4751 BITMAP_ENABLE, 4752 BITMAP_DISABLE, 4753 BITMAP_MERGE, 4754 }; 4755 typedef struct ImgBitmapAction { 4756 enum ImgBitmapAct act; 4757 const char *src; /* only used for merge */ 4758 QSIMPLEQ_ENTRY(ImgBitmapAction) next; 4759 } ImgBitmapAction; 4760 4761 static int img_bitmap(int argc, char **argv) 4762 { 4763 Error *err = NULL; 4764 int c, ret = 1; 4765 QemuOpts *opts = NULL; 4766 const char *fmt = NULL, *src_fmt = NULL, *src_filename = NULL; 4767 const char *filename, *bitmap; 4768 BlockBackend *blk = NULL, *src = NULL; 4769 BlockDriverState *bs = NULL, *src_bs = NULL; 4770 bool image_opts = false; 4771 int64_t granularity = 0; 4772 bool add = false, merge = false; 4773 QSIMPLEQ_HEAD(, ImgBitmapAction) actions; 4774 ImgBitmapAction *act, *act_next; 4775 const char *op; 4776 int inactivate_ret; 4777 4778 QSIMPLEQ_INIT(&actions); 4779 4780 for (;;) { 4781 static const struct option long_options[] = { 4782 {"help", no_argument, 0, 'h'}, 4783 {"object", required_argument, 0, OPTION_OBJECT}, 4784 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS}, 4785 {"add", no_argument, 0, OPTION_ADD}, 4786 {"remove", no_argument, 0, OPTION_REMOVE}, 4787 {"clear", no_argument, 0, OPTION_CLEAR}, 4788 {"enable", no_argument, 0, OPTION_ENABLE}, 4789 {"disable", no_argument, 0, OPTION_DISABLE}, 4790 {"merge", required_argument, 0, OPTION_MERGE}, 4791 {"granularity", required_argument, 0, 'g'}, 4792 {"source-file", required_argument, 0, 'b'}, 4793 {"source-format", required_argument, 0, 'F'}, 4794 {0, 0, 0, 0} 4795 }; 4796 c = getopt_long(argc, argv, ":b:f:F:g:h", long_options, NULL); 4797 if (c == -1) { 4798 break; 4799 } 4800 4801 switch (c) { 4802 case ':': 4803 missing_argument(argv[optind - 1]); 4804 break; 4805 case '?': 4806 unrecognized_option(argv[optind - 1]); 4807 break; 4808 case 'h': 4809 help(); 4810 break; 4811 case 'b': 4812 src_filename = optarg; 4813 break; 4814 case 'f': 4815 fmt = optarg; 4816 break; 4817 case 'F': 4818 src_fmt = optarg; 4819 break; 4820 case 'g': 4821 granularity = cvtnum("granularity", optarg); 4822 if (granularity < 0) { 4823 return 1; 4824 } 4825 break; 4826 case OPTION_ADD: 4827 act = g_new0(ImgBitmapAction, 1); 4828 act->act = BITMAP_ADD; 4829 QSIMPLEQ_INSERT_TAIL(&actions, act, next); 4830 add = true; 4831 break; 4832 case OPTION_REMOVE: 4833 act = g_new0(ImgBitmapAction, 1); 4834 act->act = BITMAP_REMOVE; 4835 QSIMPLEQ_INSERT_TAIL(&actions, act, next); 4836 break; 4837 case OPTION_CLEAR: 4838 act = g_new0(ImgBitmapAction, 1); 4839 act->act = BITMAP_CLEAR; 4840 QSIMPLEQ_INSERT_TAIL(&actions, act, next); 4841 break; 4842 case OPTION_ENABLE: 4843 act = g_new0(ImgBitmapAction, 1); 4844 act->act = BITMAP_ENABLE; 4845 QSIMPLEQ_INSERT_TAIL(&actions, act, next); 4846 break; 4847 case OPTION_DISABLE: 4848 act = g_new0(ImgBitmapAction, 1); 4849 act->act = BITMAP_DISABLE; 4850 QSIMPLEQ_INSERT_TAIL(&actions, act, next); 4851 break; 4852 case OPTION_MERGE: 4853 act = g_new0(ImgBitmapAction, 1); 4854 act->act = BITMAP_MERGE; 4855 act->src = optarg; 4856 QSIMPLEQ_INSERT_TAIL(&actions, act, next); 4857 merge = true; 4858 break; 4859 case OPTION_OBJECT: 4860 user_creatable_process_cmdline(optarg); 4861 break; 4862 case OPTION_IMAGE_OPTS: 4863 image_opts = true; 4864 break; 4865 } 4866 } 4867 4868 if (QSIMPLEQ_EMPTY(&actions)) { 4869 error_report("Need at least one of --add, --remove, --clear, " 4870 "--enable, --disable, or --merge"); 4871 goto out; 4872 } 4873 4874 if (granularity && !add) { 4875 error_report("granularity only supported with --add"); 4876 goto out; 4877 } 4878 if (src_fmt && !src_filename) { 4879 error_report("-F only supported with -b"); 4880 goto out; 4881 } 4882 if (src_filename && !merge) { 4883 error_report("Merge bitmap source file only supported with " 4884 "--merge"); 4885 goto out; 4886 } 4887 4888 if (optind != argc - 2) { 4889 error_report("Expecting filename and bitmap name"); 4890 goto out; 4891 } 4892 4893 filename = argv[optind]; 4894 bitmap = argv[optind + 1]; 4895 4896 /* 4897 * No need to open backing chains; we will be manipulating bitmaps 4898 * directly in this image without reference to image contents. 4899 */ 4900 blk = img_open(image_opts, filename, fmt, BDRV_O_RDWR | BDRV_O_NO_BACKING, 4901 false, false, false); 4902 if (!blk) { 4903 goto out; 4904 } 4905 bs = blk_bs(blk); 4906 if (src_filename) { 4907 src = img_open(false, src_filename, src_fmt, BDRV_O_NO_BACKING, 4908 false, false, false); 4909 if (!src) { 4910 goto out; 4911 } 4912 src_bs = blk_bs(src); 4913 } else { 4914 src_bs = bs; 4915 } 4916 4917 QSIMPLEQ_FOREACH_SAFE(act, &actions, next, act_next) { 4918 switch (act->act) { 4919 case BITMAP_ADD: 4920 qmp_block_dirty_bitmap_add(bs->node_name, bitmap, 4921 !!granularity, granularity, true, true, 4922 false, false, &err); 4923 op = "add"; 4924 break; 4925 case BITMAP_REMOVE: 4926 qmp_block_dirty_bitmap_remove(bs->node_name, bitmap, &err); 4927 op = "remove"; 4928 break; 4929 case BITMAP_CLEAR: 4930 qmp_block_dirty_bitmap_clear(bs->node_name, bitmap, &err); 4931 op = "clear"; 4932 break; 4933 case BITMAP_ENABLE: 4934 qmp_block_dirty_bitmap_enable(bs->node_name, bitmap, &err); 4935 op = "enable"; 4936 break; 4937 case BITMAP_DISABLE: 4938 qmp_block_dirty_bitmap_disable(bs->node_name, bitmap, &err); 4939 op = "disable"; 4940 break; 4941 case BITMAP_MERGE: 4942 do_dirty_bitmap_merge(bs->node_name, bitmap, src_bs->node_name, 4943 act->src, &err); 4944 op = "merge"; 4945 break; 4946 default: 4947 g_assert_not_reached(); 4948 } 4949 4950 if (err) { 4951 error_reportf_err(err, "Operation %s on bitmap %s failed: ", 4952 op, bitmap); 4953 goto out; 4954 } 4955 g_free(act); 4956 } 4957 4958 ret = 0; 4959 4960 out: 4961 /* 4962 * Manually inactivate the images first because this way we can know whether 4963 * an error occurred. blk_unref() doesn't tell us about failures. 4964 */ 4965 inactivate_ret = bdrv_inactivate_all(); 4966 if (inactivate_ret < 0) { 4967 error_report("Error while closing the image: %s", strerror(-inactivate_ret)); 4968 ret = 1; 4969 } 4970 4971 blk_unref(src); 4972 blk_unref(blk); 4973 qemu_opts_del(opts); 4974 return ret; 4975 } 4976 4977 #define C_BS 01 4978 #define C_COUNT 02 4979 #define C_IF 04 4980 #define C_OF 010 4981 #define C_SKIP 020 4982 4983 struct DdInfo { 4984 unsigned int flags; 4985 int64_t count; 4986 }; 4987 4988 struct DdIo { 4989 int bsz; /* Block size */ 4990 char *filename; 4991 uint8_t *buf; 4992 int64_t offset; 4993 }; 4994 4995 struct DdOpts { 4996 const char *name; 4997 int (*f)(const char *, struct DdIo *, struct DdIo *, struct DdInfo *); 4998 unsigned int flag; 4999 }; 5000 5001 static int img_dd_bs(const char *arg, 5002 struct DdIo *in, struct DdIo *out, 5003 struct DdInfo *dd) 5004 { 5005 int64_t res; 5006 5007 res = cvtnum_full("bs", arg, 1, INT_MAX); 5008 5009 if (res < 0) { 5010 return 1; 5011 } 5012 in->bsz = out->bsz = res; 5013 5014 return 0; 5015 } 5016 5017 static int img_dd_count(const char *arg, 5018 struct DdIo *in, struct DdIo *out, 5019 struct DdInfo *dd) 5020 { 5021 dd->count = cvtnum("count", arg); 5022 5023 if (dd->count < 0) { 5024 return 1; 5025 } 5026 5027 return 0; 5028 } 5029 5030 static int img_dd_if(const char *arg, 5031 struct DdIo *in, struct DdIo *out, 5032 struct DdInfo *dd) 5033 { 5034 in->filename = g_strdup(arg); 5035 5036 return 0; 5037 } 5038 5039 static int img_dd_of(const char *arg, 5040 struct DdIo *in, struct DdIo *out, 5041 struct DdInfo *dd) 5042 { 5043 out->filename = g_strdup(arg); 5044 5045 return 0; 5046 } 5047 5048 static int img_dd_skip(const char *arg, 5049 struct DdIo *in, struct DdIo *out, 5050 struct DdInfo *dd) 5051 { 5052 in->offset = cvtnum("skip", arg); 5053 5054 if (in->offset < 0) { 5055 return 1; 5056 } 5057 5058 return 0; 5059 } 5060 5061 static int img_dd(int argc, char **argv) 5062 { 5063 int ret = 0; 5064 char *arg = NULL; 5065 char *tmp; 5066 BlockDriver *drv = NULL, *proto_drv = NULL; 5067 BlockBackend *blk1 = NULL, *blk2 = NULL; 5068 QemuOpts *opts = NULL; 5069 QemuOptsList *create_opts = NULL; 5070 Error *local_err = NULL; 5071 bool image_opts = false; 5072 int c, i; 5073 const char *out_fmt = "raw"; 5074 const char *fmt = NULL; 5075 int64_t size = 0; 5076 int64_t out_pos, in_pos; 5077 bool force_share = false; 5078 struct DdInfo dd = { 5079 .flags = 0, 5080 .count = 0, 5081 }; 5082 struct DdIo in = { 5083 .bsz = 512, /* Block size is by default 512 bytes */ 5084 .filename = NULL, 5085 .buf = NULL, 5086 .offset = 0 5087 }; 5088 struct DdIo out = { 5089 .bsz = 512, 5090 .filename = NULL, 5091 .buf = NULL, 5092 .offset = 0 5093 }; 5094 5095 const struct DdOpts options[] = { 5096 { "bs", img_dd_bs, C_BS }, 5097 { "count", img_dd_count, C_COUNT }, 5098 { "if", img_dd_if, C_IF }, 5099 { "of", img_dd_of, C_OF }, 5100 { "skip", img_dd_skip, C_SKIP }, 5101 { NULL, NULL, 0 } 5102 }; 5103 const struct option long_options[] = { 5104 { "help", no_argument, 0, 'h'}, 5105 { "object", required_argument, 0, OPTION_OBJECT}, 5106 { "image-opts", no_argument, 0, OPTION_IMAGE_OPTS}, 5107 { "force-share", no_argument, 0, 'U'}, 5108 { 0, 0, 0, 0 } 5109 }; 5110 5111 while ((c = getopt_long(argc, argv, ":hf:O:U", long_options, NULL))) { 5112 if (c == EOF) { 5113 break; 5114 } 5115 switch (c) { 5116 case 'O': 5117 out_fmt = optarg; 5118 break; 5119 case 'f': 5120 fmt = optarg; 5121 break; 5122 case ':': 5123 missing_argument(argv[optind - 1]); 5124 break; 5125 case '?': 5126 unrecognized_option(argv[optind - 1]); 5127 break; 5128 case 'h': 5129 help(); 5130 break; 5131 case 'U': 5132 force_share = true; 5133 break; 5134 case OPTION_OBJECT: 5135 user_creatable_process_cmdline(optarg); 5136 break; 5137 case OPTION_IMAGE_OPTS: 5138 image_opts = true; 5139 break; 5140 } 5141 } 5142 5143 for (i = optind; i < argc; i++) { 5144 int j; 5145 arg = g_strdup(argv[i]); 5146 5147 tmp = strchr(arg, '='); 5148 if (tmp == NULL) { 5149 error_report("unrecognized operand %s", arg); 5150 ret = -1; 5151 goto out; 5152 } 5153 5154 *tmp++ = '\0'; 5155 5156 for (j = 0; options[j].name != NULL; j++) { 5157 if (!strcmp(arg, options[j].name)) { 5158 break; 5159 } 5160 } 5161 if (options[j].name == NULL) { 5162 error_report("unrecognized operand %s", arg); 5163 ret = -1; 5164 goto out; 5165 } 5166 5167 if (options[j].f(tmp, &in, &out, &dd) != 0) { 5168 ret = -1; 5169 goto out; 5170 } 5171 dd.flags |= options[j].flag; 5172 g_free(arg); 5173 arg = NULL; 5174 } 5175 5176 if (!(dd.flags & C_IF && dd.flags & C_OF)) { 5177 error_report("Must specify both input and output files"); 5178 ret = -1; 5179 goto out; 5180 } 5181 5182 blk1 = img_open(image_opts, in.filename, fmt, 0, false, false, 5183 force_share); 5184 5185 if (!blk1) { 5186 ret = -1; 5187 goto out; 5188 } 5189 5190 drv = bdrv_find_format(out_fmt); 5191 if (!drv) { 5192 error_report("Unknown file format"); 5193 ret = -1; 5194 goto out; 5195 } 5196 proto_drv = bdrv_find_protocol(out.filename, true, &local_err); 5197 5198 if (!proto_drv) { 5199 error_report_err(local_err); 5200 ret = -1; 5201 goto out; 5202 } 5203 if (!drv->create_opts) { 5204 error_report("Format driver '%s' does not support image creation", 5205 drv->format_name); 5206 ret = -1; 5207 goto out; 5208 } 5209 if (!proto_drv->create_opts) { 5210 error_report("Protocol driver '%s' does not support image creation", 5211 proto_drv->format_name); 5212 ret = -1; 5213 goto out; 5214 } 5215 create_opts = qemu_opts_append(create_opts, drv->create_opts); 5216 create_opts = qemu_opts_append(create_opts, proto_drv->create_opts); 5217 5218 opts = qemu_opts_create(create_opts, NULL, 0, &error_abort); 5219 5220 size = blk_getlength(blk1); 5221 if (size < 0) { 5222 error_report("Failed to get size for '%s'", in.filename); 5223 ret = -1; 5224 goto out; 5225 } 5226 5227 if (dd.flags & C_COUNT && dd.count <= INT64_MAX / in.bsz && 5228 dd.count * in.bsz < size) { 5229 size = dd.count * in.bsz; 5230 } 5231 5232 /* Overflow means the specified offset is beyond input image's size */ 5233 if (dd.flags & C_SKIP && (in.offset > INT64_MAX / in.bsz || 5234 size < in.bsz * in.offset)) { 5235 qemu_opt_set_number(opts, BLOCK_OPT_SIZE, 0, &error_abort); 5236 } else { 5237 qemu_opt_set_number(opts, BLOCK_OPT_SIZE, 5238 size - in.bsz * in.offset, &error_abort); 5239 } 5240 5241 ret = bdrv_create(drv, out.filename, opts, &local_err); 5242 if (ret < 0) { 5243 error_reportf_err(local_err, 5244 "%s: error while creating output image: ", 5245 out.filename); 5246 ret = -1; 5247 goto out; 5248 } 5249 5250 /* TODO, we can't honour --image-opts for the target, 5251 * since it needs to be given in a format compatible 5252 * with the bdrv_create() call above which does not 5253 * support image-opts style. 5254 */ 5255 blk2 = img_open_file(out.filename, NULL, out_fmt, BDRV_O_RDWR, 5256 false, false, false); 5257 5258 if (!blk2) { 5259 ret = -1; 5260 goto out; 5261 } 5262 5263 if (dd.flags & C_SKIP && (in.offset > INT64_MAX / in.bsz || 5264 size < in.offset * in.bsz)) { 5265 /* We give a warning if the skip option is bigger than the input 5266 * size and create an empty output disk image (i.e. like dd(1)). 5267 */ 5268 error_report("%s: cannot skip to specified offset", in.filename); 5269 in_pos = size; 5270 } else { 5271 in_pos = in.offset * in.bsz; 5272 } 5273 5274 in.buf = g_new(uint8_t, in.bsz); 5275 5276 for (out_pos = 0; in_pos < size; ) { 5277 int bytes = (in_pos + in.bsz > size) ? size - in_pos : in.bsz; 5278 5279 ret = blk_pread(blk1, in_pos, bytes, in.buf, 0); 5280 if (ret < 0) { 5281 error_report("error while reading from input image file: %s", 5282 strerror(-ret)); 5283 goto out; 5284 } 5285 in_pos += bytes; 5286 5287 ret = blk_pwrite(blk2, out_pos, bytes, in.buf, 0); 5288 if (ret < 0) { 5289 error_report("error while writing to output image file: %s", 5290 strerror(-ret)); 5291 goto out; 5292 } 5293 out_pos += bytes; 5294 } 5295 5296 out: 5297 g_free(arg); 5298 qemu_opts_del(opts); 5299 qemu_opts_free(create_opts); 5300 blk_unref(blk1); 5301 blk_unref(blk2); 5302 g_free(in.filename); 5303 g_free(out.filename); 5304 g_free(in.buf); 5305 g_free(out.buf); 5306 5307 if (ret) { 5308 return 1; 5309 } 5310 return 0; 5311 } 5312 5313 static void dump_json_block_measure_info(BlockMeasureInfo *info) 5314 { 5315 GString *str; 5316 QObject *obj; 5317 Visitor *v = qobject_output_visitor_new(&obj); 5318 5319 visit_type_BlockMeasureInfo(v, NULL, &info, &error_abort); 5320 visit_complete(v, &obj); 5321 str = qobject_to_json_pretty(obj, true); 5322 assert(str != NULL); 5323 printf("%s\n", str->str); 5324 qobject_unref(obj); 5325 visit_free(v); 5326 g_string_free(str, true); 5327 } 5328 5329 static int img_measure(int argc, char **argv) 5330 { 5331 static const struct option long_options[] = { 5332 {"help", no_argument, 0, 'h'}, 5333 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS}, 5334 {"object", required_argument, 0, OPTION_OBJECT}, 5335 {"output", required_argument, 0, OPTION_OUTPUT}, 5336 {"size", required_argument, 0, OPTION_SIZE}, 5337 {"force-share", no_argument, 0, 'U'}, 5338 {0, 0, 0, 0} 5339 }; 5340 OutputFormat output_format = OFORMAT_HUMAN; 5341 BlockBackend *in_blk = NULL; 5342 BlockDriver *drv; 5343 const char *filename = NULL; 5344 const char *fmt = NULL; 5345 const char *out_fmt = "raw"; 5346 char *options = NULL; 5347 char *snapshot_name = NULL; 5348 bool force_share = false; 5349 QemuOpts *opts = NULL; 5350 QemuOpts *object_opts = NULL; 5351 QemuOpts *sn_opts = NULL; 5352 QemuOptsList *create_opts = NULL; 5353 bool image_opts = false; 5354 uint64_t img_size = UINT64_MAX; 5355 BlockMeasureInfo *info = NULL; 5356 Error *local_err = NULL; 5357 int ret = 1; 5358 int c; 5359 5360 while ((c = getopt_long(argc, argv, "hf:O:o:l:U", 5361 long_options, NULL)) != -1) { 5362 switch (c) { 5363 case '?': 5364 case 'h': 5365 help(); 5366 break; 5367 case 'f': 5368 fmt = optarg; 5369 break; 5370 case 'O': 5371 out_fmt = optarg; 5372 break; 5373 case 'o': 5374 if (accumulate_options(&options, optarg) < 0) { 5375 goto out; 5376 } 5377 break; 5378 case 'l': 5379 if (strstart(optarg, SNAPSHOT_OPT_BASE, NULL)) { 5380 sn_opts = qemu_opts_parse_noisily(&internal_snapshot_opts, 5381 optarg, false); 5382 if (!sn_opts) { 5383 error_report("Failed in parsing snapshot param '%s'", 5384 optarg); 5385 goto out; 5386 } 5387 } else { 5388 snapshot_name = optarg; 5389 } 5390 break; 5391 case 'U': 5392 force_share = true; 5393 break; 5394 case OPTION_OBJECT: 5395 user_creatable_process_cmdline(optarg); 5396 break; 5397 case OPTION_IMAGE_OPTS: 5398 image_opts = true; 5399 break; 5400 case OPTION_OUTPUT: 5401 if (!strcmp(optarg, "json")) { 5402 output_format = OFORMAT_JSON; 5403 } else if (!strcmp(optarg, "human")) { 5404 output_format = OFORMAT_HUMAN; 5405 } else { 5406 error_report("--output must be used with human or json " 5407 "as argument."); 5408 goto out; 5409 } 5410 break; 5411 case OPTION_SIZE: 5412 { 5413 int64_t sval; 5414 5415 sval = cvtnum("image size", optarg); 5416 if (sval < 0) { 5417 goto out; 5418 } 5419 img_size = (uint64_t)sval; 5420 } 5421 break; 5422 } 5423 } 5424 5425 if (argc - optind > 1) { 5426 error_report("At most one filename argument is allowed."); 5427 goto out; 5428 } else if (argc - optind == 1) { 5429 filename = argv[optind]; 5430 } 5431 5432 if (!filename && (image_opts || fmt || snapshot_name || sn_opts)) { 5433 error_report("--image-opts, -f, and -l require a filename argument."); 5434 goto out; 5435 } 5436 if (filename && img_size != UINT64_MAX) { 5437 error_report("--size N cannot be used together with a filename."); 5438 goto out; 5439 } 5440 if (!filename && img_size == UINT64_MAX) { 5441 error_report("Either --size N or one filename must be specified."); 5442 goto out; 5443 } 5444 5445 if (filename) { 5446 in_blk = img_open(image_opts, filename, fmt, 0, 5447 false, false, force_share); 5448 if (!in_blk) { 5449 goto out; 5450 } 5451 5452 if (sn_opts) { 5453 bdrv_snapshot_load_tmp(blk_bs(in_blk), 5454 qemu_opt_get(sn_opts, SNAPSHOT_OPT_ID), 5455 qemu_opt_get(sn_opts, SNAPSHOT_OPT_NAME), 5456 &local_err); 5457 } else if (snapshot_name != NULL) { 5458 bdrv_snapshot_load_tmp_by_id_or_name(blk_bs(in_blk), 5459 snapshot_name, &local_err); 5460 } 5461 if (local_err) { 5462 error_reportf_err(local_err, "Failed to load snapshot: "); 5463 goto out; 5464 } 5465 } 5466 5467 drv = bdrv_find_format(out_fmt); 5468 if (!drv) { 5469 error_report("Unknown file format '%s'", out_fmt); 5470 goto out; 5471 } 5472 if (!drv->create_opts) { 5473 error_report("Format driver '%s' does not support image creation", 5474 drv->format_name); 5475 goto out; 5476 } 5477 5478 create_opts = qemu_opts_append(create_opts, drv->create_opts); 5479 create_opts = qemu_opts_append(create_opts, bdrv_file.create_opts); 5480 opts = qemu_opts_create(create_opts, NULL, 0, &error_abort); 5481 if (options) { 5482 if (!qemu_opts_do_parse(opts, options, NULL, &local_err)) { 5483 error_report_err(local_err); 5484 error_report("Invalid options for file format '%s'", out_fmt); 5485 goto out; 5486 } 5487 } 5488 if (img_size != UINT64_MAX) { 5489 qemu_opt_set_number(opts, BLOCK_OPT_SIZE, img_size, &error_abort); 5490 } 5491 5492 info = bdrv_measure(drv, opts, in_blk ? blk_bs(in_blk) : NULL, &local_err); 5493 if (local_err) { 5494 error_report_err(local_err); 5495 goto out; 5496 } 5497 5498 if (output_format == OFORMAT_HUMAN) { 5499 printf("required size: %" PRIu64 "\n", info->required); 5500 printf("fully allocated size: %" PRIu64 "\n", info->fully_allocated); 5501 if (info->has_bitmaps) { 5502 printf("bitmaps size: %" PRIu64 "\n", info->bitmaps); 5503 } 5504 } else { 5505 dump_json_block_measure_info(info); 5506 } 5507 5508 ret = 0; 5509 5510 out: 5511 qapi_free_BlockMeasureInfo(info); 5512 qemu_opts_del(object_opts); 5513 qemu_opts_del(opts); 5514 qemu_opts_del(sn_opts); 5515 qemu_opts_free(create_opts); 5516 g_free(options); 5517 blk_unref(in_blk); 5518 return ret; 5519 } 5520 5521 static const img_cmd_t img_cmds[] = { 5522 #define DEF(option, callback, arg_string) \ 5523 { option, callback }, 5524 #include "qemu-img-cmds.h" 5525 #undef DEF 5526 { NULL, NULL, }, 5527 }; 5528 5529 int main(int argc, char **argv) 5530 { 5531 const img_cmd_t *cmd; 5532 const char *cmdname; 5533 int c; 5534 static const struct option long_options[] = { 5535 {"help", no_argument, 0, 'h'}, 5536 {"version", no_argument, 0, 'V'}, 5537 {"trace", required_argument, NULL, 'T'}, 5538 {0, 0, 0, 0} 5539 }; 5540 5541 #ifdef CONFIG_POSIX 5542 signal(SIGPIPE, SIG_IGN); 5543 #endif 5544 5545 socket_init(); 5546 error_init(argv[0]); 5547 module_call_init(MODULE_INIT_TRACE); 5548 qemu_init_exec_dir(argv[0]); 5549 5550 qemu_init_main_loop(&error_fatal); 5551 5552 qcrypto_init(&error_fatal); 5553 5554 module_call_init(MODULE_INIT_QOM); 5555 bdrv_init(); 5556 if (argc < 2) { 5557 error_exit("Not enough arguments"); 5558 } 5559 5560 qemu_add_opts(&qemu_source_opts); 5561 qemu_add_opts(&qemu_trace_opts); 5562 5563 while ((c = getopt_long(argc, argv, "+:hVT:", long_options, NULL)) != -1) { 5564 switch (c) { 5565 case ':': 5566 missing_argument(argv[optind - 1]); 5567 return 0; 5568 case '?': 5569 unrecognized_option(argv[optind - 1]); 5570 return 0; 5571 case 'h': 5572 help(); 5573 return 0; 5574 case 'V': 5575 printf(QEMU_IMG_VERSION); 5576 return 0; 5577 case 'T': 5578 trace_opt_parse(optarg); 5579 break; 5580 } 5581 } 5582 5583 cmdname = argv[optind]; 5584 5585 /* reset getopt_long scanning */ 5586 argc -= optind; 5587 if (argc < 1) { 5588 return 0; 5589 } 5590 argv += optind; 5591 qemu_reset_optind(); 5592 5593 if (!trace_init_backends()) { 5594 exit(1); 5595 } 5596 trace_init_file(); 5597 qemu_set_log(LOG_TRACE, &error_fatal); 5598 5599 /* find the command */ 5600 for (cmd = img_cmds; cmd->name != NULL; cmd++) { 5601 if (!strcmp(cmdname, cmd->name)) { 5602 return cmd->handler(argc, argv); 5603 } 5604 } 5605 5606 /* not found */ 5607 error_exit("Command not found: %s", cmdname); 5608 } 5609