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 "qobject/qjson.h" 36 #include "qobject/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 "system/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)(const struct img_cmd_t *ccmd, int argc, char **argv); 64 const char *description; 65 } img_cmd_t; 66 67 enum { 68 OPTION_OUTPUT = 256, 69 OPTION_BACKING_CHAIN = 257, 70 OPTION_OBJECT = 258, 71 OPTION_IMAGE_OPTS = 259, 72 OPTION_PATTERN = 260, 73 OPTION_FLUSH_INTERVAL = 261, 74 OPTION_NO_DRAIN = 262, 75 OPTION_TARGET_IMAGE_OPTS = 263, 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 OPTION_LIMITS = 278, 90 }; 91 92 typedef enum OutputFormat { 93 OFORMAT_JSON, 94 OFORMAT_HUMAN, 95 } OutputFormat; 96 97 /* Default to cache=writeback as data integrity is not important for qemu-img */ 98 #define BDRV_DEFAULT_CACHE "writeback" 99 100 static G_NORETURN 101 void tryhelp(const char *argv0) 102 { 103 error_printf("Try '%s --help' for more information\n", argv0); 104 exit(EXIT_FAILURE); 105 } 106 107 static G_NORETURN G_GNUC_PRINTF(2, 3) 108 void error_exit(const char *argv0, const char *fmt, ...) 109 { 110 va_list ap; 111 112 va_start(ap, fmt); 113 error_vreport(fmt, ap); 114 va_end(ap); 115 116 tryhelp(argv0); 117 } 118 119 /* 120 * Print --help output for a command and exit. 121 * @syntax and @description are multi-line with trailing EOL 122 * (to allow easy extending of the text) 123 * @syntax has each subsequent line indented by 8 chars. 124 * @description is indented by 2 chars for argument on each own line, 125 * and with 5 chars for argument description (like -h arg below). 126 */ 127 static G_NORETURN 128 void cmd_help(const img_cmd_t *ccmd, 129 const char *syntax, const char *arguments) 130 { 131 printf( 132 "Usage:\n" 133 " %s %s %s\n" 134 "%s.\n" 135 "\n" 136 "Arguments:\n" 137 " -h, --help\n" 138 " print this help and exit\n" 139 "%s\n", 140 "qemu-img", ccmd->name, syntax, ccmd->description, arguments); 141 exit(EXIT_SUCCESS); 142 } 143 144 static OutputFormat parse_output_format(const char *argv0, const char *arg) 145 { 146 if (!strcmp(arg, "json")) { 147 return OFORMAT_JSON; 148 } else if (!strcmp(arg, "human")) { 149 return OFORMAT_HUMAN; 150 } else { 151 error_exit(argv0, "--output expects 'human' or 'json', not '%s'", arg); 152 } 153 } 154 155 /* 156 * Is @list safe for accumulate_options()? 157 * It is when multiple of them can be joined together separated by ','. 158 * To make that work, @list must not start with ',' (or else a 159 * separating ',' preceding it gets escaped), and it must not end with 160 * an odd number of ',' (or else a separating ',' following it gets 161 * escaped), or be empty (or else a separating ',' preceding it can 162 * escape a separating ',' following it). 163 * 164 */ 165 static bool is_valid_option_list(const char *list) 166 { 167 size_t len = strlen(list); 168 size_t i; 169 170 if (!list[0] || list[0] == ',') { 171 return false; 172 } 173 174 for (i = len; i > 0 && list[i - 1] == ','; i--) { 175 } 176 if ((len - i) % 2) { 177 return false; 178 } 179 180 return true; 181 } 182 183 static int accumulate_options(char **options, char *list) 184 { 185 char *new_options; 186 187 if (!is_valid_option_list(list)) { 188 error_report("Invalid option list: %s", list); 189 return -1; 190 } 191 192 if (!*options) { 193 *options = g_strdup(list); 194 } else { 195 new_options = g_strdup_printf("%s,%s", *options, list); 196 g_free(*options); 197 *options = new_options; 198 } 199 return 0; 200 } 201 202 static QemuOptsList qemu_source_opts = { 203 .name = "source", 204 .implied_opt_name = "file", 205 .head = QTAILQ_HEAD_INITIALIZER(qemu_source_opts.head), 206 .desc = { 207 { } 208 }, 209 }; 210 211 static int G_GNUC_PRINTF(2, 3) qprintf(bool quiet, const char *fmt, ...) 212 { 213 int ret = 0; 214 if (!quiet) { 215 va_list args; 216 va_start(args, fmt); 217 ret = vprintf(fmt, args); 218 va_end(args); 219 } 220 return ret; 221 } 222 223 224 static int print_block_option_help(const char *filename, const char *fmt) 225 { 226 BlockDriver *drv, *proto_drv; 227 QemuOptsList *create_opts = NULL; 228 Error *local_err = NULL; 229 230 /* Find driver and parse its options */ 231 drv = bdrv_find_format(fmt); 232 if (!drv) { 233 error_report("Unknown file format '%s'", fmt); 234 return 1; 235 } 236 237 if (!drv->create_opts) { 238 error_report("Format driver '%s' does not support image creation", fmt); 239 return 1; 240 } 241 242 create_opts = qemu_opts_append(create_opts, drv->create_opts); 243 if (filename) { 244 proto_drv = bdrv_find_protocol(filename, true, &local_err); 245 if (!proto_drv) { 246 error_report_err(local_err); 247 qemu_opts_free(create_opts); 248 return 1; 249 } 250 if (!proto_drv->create_opts) { 251 error_report("Protocol driver '%s' does not support image creation", 252 proto_drv->format_name); 253 qemu_opts_free(create_opts); 254 return 1; 255 } 256 create_opts = qemu_opts_append(create_opts, proto_drv->create_opts); 257 } 258 259 if (filename) { 260 printf("Supported options:\n"); 261 } else { 262 printf("Supported %s options:\n", fmt); 263 } 264 qemu_opts_print_help(create_opts, false); 265 qemu_opts_free(create_opts); 266 267 if (!filename) { 268 printf("\n" 269 "The protocol level may support further options.\n" 270 "Specify the target filename to include those options.\n"); 271 } 272 273 return 0; 274 } 275 276 277 static BlockBackend *img_open_opts(const char *optstr, 278 QemuOpts *opts, int flags, bool writethrough, 279 bool quiet, bool force_share) 280 { 281 QDict *options; 282 Error *local_err = NULL; 283 BlockBackend *blk; 284 options = qemu_opts_to_qdict(opts, NULL); 285 if (force_share) { 286 if (qdict_haskey(options, BDRV_OPT_FORCE_SHARE) 287 && strcmp(qdict_get_str(options, BDRV_OPT_FORCE_SHARE), "on")) { 288 error_report("--force-share/-U conflicts with image options"); 289 qobject_unref(options); 290 return NULL; 291 } 292 qdict_put_str(options, BDRV_OPT_FORCE_SHARE, "on"); 293 } 294 blk = blk_new_open(NULL, NULL, options, flags, &local_err); 295 if (!blk) { 296 error_reportf_err(local_err, "Could not open '%s': ", optstr); 297 return NULL; 298 } 299 blk_set_enable_write_cache(blk, !writethrough); 300 301 return blk; 302 } 303 304 static BlockBackend *img_open_file(const char *filename, 305 QDict *options, 306 const char *fmt, int flags, 307 bool writethrough, bool quiet, 308 bool force_share) 309 { 310 BlockBackend *blk; 311 Error *local_err = NULL; 312 313 if (!options) { 314 options = qdict_new(); 315 } 316 if (fmt) { 317 qdict_put_str(options, "driver", fmt); 318 } 319 320 if (force_share) { 321 qdict_put_bool(options, BDRV_OPT_FORCE_SHARE, true); 322 } 323 blk = blk_new_open(filename, NULL, options, flags, &local_err); 324 if (!blk) { 325 error_reportf_err(local_err, "Could not open '%s': ", filename); 326 return NULL; 327 } 328 blk_set_enable_write_cache(blk, !writethrough); 329 330 return blk; 331 } 332 333 334 static int img_add_key_secrets(void *opaque, 335 const char *name, const char *value, 336 Error **errp) 337 { 338 QDict *options = opaque; 339 340 if (g_str_has_suffix(name, "key-secret")) { 341 qdict_put_str(options, name, value); 342 } 343 344 return 0; 345 } 346 347 348 static BlockBackend *img_open(bool image_opts, 349 const char *filename, 350 const char *fmt, int flags, bool writethrough, 351 bool quiet, bool force_share) 352 { 353 BlockBackend *blk; 354 if (image_opts) { 355 QemuOpts *opts; 356 if (fmt) { 357 error_report("--image-opts and --format are mutually exclusive"); 358 return NULL; 359 } 360 opts = qemu_opts_parse_noisily(qemu_find_opts("source"), 361 filename, true); 362 if (!opts) { 363 return NULL; 364 } 365 blk = img_open_opts(filename, opts, flags, writethrough, quiet, 366 force_share); 367 } else { 368 blk = img_open_file(filename, NULL, fmt, flags, writethrough, quiet, 369 force_share); 370 } 371 372 if (blk) { 373 blk_set_force_allow_inactivate(blk); 374 } 375 376 return blk; 377 } 378 379 380 static int add_old_style_options(const char *fmt, QemuOpts *opts, 381 const char *base_filename, 382 const char *base_fmt) 383 { 384 if (base_filename) { 385 if (!qemu_opt_set(opts, BLOCK_OPT_BACKING_FILE, base_filename, 386 NULL)) { 387 error_report("Backing file not supported for file format '%s'", 388 fmt); 389 return -1; 390 } 391 } 392 if (base_fmt) { 393 if (!qemu_opt_set(opts, BLOCK_OPT_BACKING_FMT, base_fmt, NULL)) { 394 error_report("Backing file format not supported for file " 395 "format '%s'", fmt); 396 return -1; 397 } 398 } 399 return 0; 400 } 401 402 static int64_t cvtnum_full(const char *name, const char *value, 403 bool is_size, int64_t min, int64_t max) 404 { 405 int err; 406 uint64_t res; 407 408 err = is_size ? qemu_strtosz(value, NULL, &res) : 409 qemu_strtou64(value, NULL, 0, &res); 410 if (err < 0 && err != -ERANGE) { 411 error_report("Invalid %s specified: '%s'", name, value); 412 return err; 413 } 414 if (err == -ERANGE || res > max || res < min) { 415 error_report("Invalid %s specified. Must be between %" PRId64 416 " and %" PRId64 ".", name, min, max); 417 return -ERANGE; 418 } 419 return res; 420 } 421 422 static int64_t cvtnum(const char *name, const char *value, bool is_size) 423 { 424 return cvtnum_full(name, value, is_size, 0, INT64_MAX); 425 } 426 427 static int img_create(const img_cmd_t *ccmd, int argc, char **argv) 428 { 429 int c; 430 int64_t img_size = -1; 431 const char *fmt = "raw"; 432 const char *base_fmt = NULL; 433 const char *filename; 434 const char *base_filename = NULL; 435 char *options = NULL; 436 Error *local_err = NULL; 437 bool quiet = false; 438 int flags = 0; 439 440 for(;;) { 441 static const struct option long_options[] = { 442 {"help", no_argument, 0, 'h'}, 443 {"format", required_argument, 0, 'f'}, 444 {"options", required_argument, 0, 'o'}, 445 {"backing", required_argument, 0, 'b'}, 446 {"backing-format", required_argument, 0, 'B'}, /* was -F in 10.0 */ 447 {"backing-unsafe", no_argument, 0, 'u'}, 448 {"quiet", no_argument, 0, 'q'}, 449 {"object", required_argument, 0, OPTION_OBJECT}, 450 {0, 0, 0, 0} 451 }; 452 c = getopt_long(argc, argv, "hf:o:b:F:B:uq", 453 long_options, NULL); 454 if (c == -1) { 455 break; 456 } 457 switch(c) { 458 case 'h': 459 cmd_help(ccmd, "[-f FMT] [-o FMT_OPTS]\n" 460 " [-b BACKING_FILE [-B BACKING_FMT]] [-u]\n" 461 " [-q] [--object OBJDEF] FILE [SIZE]\n" 462 , 463 " -f, --format FMT\n" 464 " specifies the format of the new image (default: raw)\n" 465 " -o, --options FMT_OPTS\n" 466 " format-specific options (specify '-o help' for help)\n" 467 " -b, --backing BACKING_FILE\n" 468 " create target image to be a CoW on top of BACKING_FILE\n" 469 " -B, --backing-format BACKING_FMT (was -F in <= 10.0)\n" 470 " specifies the format of BACKING_FILE (default: probing is used)\n" 471 " -u, --backing-unsafe\n" 472 " do not fail if BACKING_FILE can not be read\n" 473 " -q, --quiet\n" 474 " quiet mode (produce only error messages if any)\n" 475 " --object OBJDEF\n" 476 " defines QEMU user-creatable object\n" 477 " FILE\n" 478 " name of the image file to create (will be overritten if already exists)\n" 479 " SIZE[bKMGTPE]\n" 480 " image size with optional multiplier suffix (powers of 1024)\n" 481 " (required unless BACKING_FILE is specified)\n" 482 ); 483 break; 484 case 'f': 485 fmt = optarg; 486 break; 487 case 'o': 488 if (accumulate_options(&options, optarg) < 0) { 489 goto fail; 490 } 491 break; 492 case 'b': 493 base_filename = optarg; 494 break; 495 case 'F': /* <=10.0 */ 496 case 'B': 497 base_fmt = optarg; 498 break; 499 case 'u': 500 flags |= BDRV_O_NO_BACKING; 501 break; 502 case 'q': 503 quiet = true; 504 break; 505 case OPTION_OBJECT: 506 user_creatable_process_cmdline(optarg); 507 break; 508 default: 509 tryhelp(argv[0]); 510 } 511 } 512 513 /* Get the filename */ 514 filename = (optind < argc) ? argv[optind] : NULL; 515 if (options && has_help_option(options)) { 516 g_free(options); 517 return print_block_option_help(filename, fmt); 518 } 519 520 if (optind >= argc) { 521 error_exit(argv[0], "Expecting image file name"); 522 } 523 optind++; 524 525 /* Get image size, if specified */ 526 if (optind < argc) { 527 img_size = cvtnum("image size", argv[optind++], true); 528 if (img_size < 0) { 529 goto fail; 530 } 531 } 532 if (optind != argc) { 533 error_exit(argv[0], "Unexpected argument: %s", argv[optind]); 534 } 535 536 bdrv_img_create(filename, fmt, base_filename, base_fmt, 537 options, img_size, flags, quiet, &local_err); 538 if (local_err) { 539 error_reportf_err(local_err, "%s: ", filename); 540 goto fail; 541 } 542 543 g_free(options); 544 return 0; 545 546 fail: 547 g_free(options); 548 return 1; 549 } 550 551 static void dump_json_image_check(ImageCheck *check, bool quiet) 552 { 553 GString *str; 554 QObject *obj; 555 Visitor *v = qobject_output_visitor_new(&obj); 556 557 visit_type_ImageCheck(v, NULL, &check, &error_abort); 558 visit_complete(v, &obj); 559 str = qobject_to_json_pretty(obj, true); 560 assert(str != NULL); 561 qprintf(quiet, "%s\n", str->str); 562 qobject_unref(obj); 563 visit_free(v); 564 g_string_free(str, true); 565 } 566 567 static void dump_human_image_check(ImageCheck *check, bool quiet) 568 { 569 if (!(check->corruptions || check->leaks || check->check_errors)) { 570 qprintf(quiet, "No errors were found on the image.\n"); 571 } else { 572 if (check->corruptions) { 573 qprintf(quiet, "\n%" PRId64 " errors were found on the image.\n" 574 "Data may be corrupted, or further writes to the image " 575 "may corrupt it.\n", 576 check->corruptions); 577 } 578 579 if (check->leaks) { 580 qprintf(quiet, 581 "\n%" PRId64 " leaked clusters were found on the image.\n" 582 "This means waste of disk space, but no harm to data.\n", 583 check->leaks); 584 } 585 586 if (check->check_errors) { 587 qprintf(quiet, 588 "\n%" PRId64 589 " internal errors have occurred during the check.\n", 590 check->check_errors); 591 } 592 } 593 594 if (check->total_clusters != 0 && check->allocated_clusters != 0) { 595 qprintf(quiet, "%" PRId64 "/%" PRId64 " = %0.2f%% allocated, " 596 "%0.2f%% fragmented, %0.2f%% compressed clusters\n", 597 check->allocated_clusters, check->total_clusters, 598 check->allocated_clusters * 100.0 / check->total_clusters, 599 check->fragmented_clusters * 100.0 / check->allocated_clusters, 600 check->compressed_clusters * 100.0 / 601 check->allocated_clusters); 602 } 603 604 if (check->image_end_offset) { 605 qprintf(quiet, 606 "Image end offset: %" PRId64 "\n", check->image_end_offset); 607 } 608 } 609 610 static int collect_image_check(BlockDriverState *bs, 611 ImageCheck *check, 612 const char *filename, 613 const char *fmt, 614 int fix) 615 { 616 int ret; 617 BdrvCheckResult result; 618 619 ret = bdrv_check(bs, &result, fix); 620 if (ret < 0) { 621 return ret; 622 } 623 624 check->filename = g_strdup(filename); 625 check->format = g_strdup(bdrv_get_format_name(bs)); 626 check->check_errors = result.check_errors; 627 check->corruptions = result.corruptions; 628 check->has_corruptions = result.corruptions != 0; 629 check->leaks = result.leaks; 630 check->has_leaks = result.leaks != 0; 631 check->corruptions_fixed = result.corruptions_fixed; 632 check->has_corruptions_fixed = result.corruptions_fixed != 0; 633 check->leaks_fixed = result.leaks_fixed; 634 check->has_leaks_fixed = result.leaks_fixed != 0; 635 check->image_end_offset = result.image_end_offset; 636 check->has_image_end_offset = result.image_end_offset != 0; 637 check->total_clusters = result.bfi.total_clusters; 638 check->has_total_clusters = result.bfi.total_clusters != 0; 639 check->allocated_clusters = result.bfi.allocated_clusters; 640 check->has_allocated_clusters = result.bfi.allocated_clusters != 0; 641 check->fragmented_clusters = result.bfi.fragmented_clusters; 642 check->has_fragmented_clusters = result.bfi.fragmented_clusters != 0; 643 check->compressed_clusters = result.bfi.compressed_clusters; 644 check->has_compressed_clusters = result.bfi.compressed_clusters != 0; 645 646 return 0; 647 } 648 649 /* 650 * Checks an image for consistency. Exit codes: 651 * 652 * 0 - Check completed, image is good 653 * 1 - Check not completed because of internal errors 654 * 2 - Check completed, image is corrupted 655 * 3 - Check completed, image has leaked clusters, but is good otherwise 656 * 63 - Checks are not supported by the image format 657 */ 658 static int img_check(const img_cmd_t *ccmd, int argc, char **argv) 659 { 660 int c, ret; 661 OutputFormat output_format = OFORMAT_HUMAN; 662 const char *filename, *fmt, *cache; 663 BlockBackend *blk; 664 BlockDriverState *bs; 665 int fix = 0; 666 int flags = BDRV_O_CHECK; 667 bool writethrough; 668 ImageCheck *check; 669 bool quiet = false; 670 bool image_opts = false; 671 bool force_share = false; 672 673 fmt = NULL; 674 cache = BDRV_DEFAULT_CACHE; 675 676 for(;;) { 677 int option_index = 0; 678 static const struct option long_options[] = { 679 {"help", no_argument, 0, 'h'}, 680 {"format", required_argument, 0, 'f'}, 681 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS}, 682 {"cache", required_argument, 0, 'T'}, 683 {"repair", required_argument, 0, 'r'}, 684 {"force-share", no_argument, 0, 'U'}, 685 {"output", required_argument, 0, OPTION_OUTPUT}, 686 {"quiet", no_argument, 0, 'q'}, 687 {"object", required_argument, 0, OPTION_OBJECT}, 688 {0, 0, 0, 0} 689 }; 690 c = getopt_long(argc, argv, "hf:T:r:Uq", 691 long_options, &option_index); 692 if (c == -1) { 693 break; 694 } 695 switch(c) { 696 case 'h': 697 cmd_help(ccmd, "[-f FMT | --image-opts] [-T CACHE_MODE] [-r leaks|all]\n" 698 " [-U] [--output human|json] [-q] [--object OBJDEF] FILE\n" 699 , 700 " -f, --format FMT\n" 701 " specifies the format of the image explicitly (default: probing is used)\n" 702 " --image-opts\n" 703 " treat FILE as an option string (key=value,..), not a file name\n" 704 " (incompatible with -f|--format)\n" 705 " -T, --cache CACHE_MODE\n" /* why not -t ? */ 706 " cache mode (default: " BDRV_DEFAULT_CACHE ")\n" 707 " -r, --repair leaks|all\n" 708 " repair errors of the given category in the image (image will be\n" 709 " opened in read-write mode, incompatible with -U|--force-share)\n" 710 " -U, --force-share\n" 711 " open image in shared mode for concurrent access\n" 712 " --output human|json\n" 713 " output format (default: human)\n" 714 " -q, --quiet\n" 715 " quiet mode (produce only error messages if any)\n" 716 " --object OBJDEF\n" 717 " defines QEMU user-creatable object\n" 718 " FILE\n" 719 " name of the image file, or an option string (key=value,..)\n" 720 " with --image-opts, to operate on\n" 721 ); 722 break; 723 case 'f': 724 fmt = optarg; 725 break; 726 case OPTION_IMAGE_OPTS: 727 image_opts = true; 728 break; 729 case 'T': 730 cache = optarg; 731 break; 732 case 'r': 733 flags |= BDRV_O_RDWR; 734 735 if (!strcmp(optarg, "leaks")) { 736 fix = BDRV_FIX_LEAKS; 737 } else if (!strcmp(optarg, "all")) { 738 fix = BDRV_FIX_LEAKS | BDRV_FIX_ERRORS; 739 } else { 740 error_exit(argv[0], 741 "--repair (-r) expects 'leaks' or 'all', not '%s'", 742 optarg); 743 } 744 break; 745 case 'U': 746 force_share = true; 747 break; 748 case OPTION_OUTPUT: 749 output_format = parse_output_format(argv[0], optarg); 750 break; 751 case 'q': 752 quiet = true; 753 break; 754 case OPTION_OBJECT: 755 user_creatable_process_cmdline(optarg); 756 break; 757 default: 758 tryhelp(argv[0]); 759 } 760 } 761 if (optind != argc - 1) { 762 error_exit(argv[0], "Expecting one image file name"); 763 } 764 filename = argv[optind++]; 765 766 ret = bdrv_parse_cache_mode(cache, &flags, &writethrough); 767 if (ret < 0) { 768 error_report("Invalid source cache option: %s", cache); 769 return 1; 770 } 771 772 blk = img_open(image_opts, filename, fmt, flags, writethrough, quiet, 773 force_share); 774 if (!blk) { 775 return 1; 776 } 777 bs = blk_bs(blk); 778 779 check = g_new0(ImageCheck, 1); 780 ret = collect_image_check(bs, check, filename, fmt, fix); 781 782 if (ret == -ENOTSUP) { 783 error_report("This image format does not support checks"); 784 ret = 63; 785 goto fail; 786 } 787 788 if (check->corruptions_fixed || check->leaks_fixed) { 789 int corruptions_fixed, leaks_fixed; 790 bool has_leaks_fixed, has_corruptions_fixed; 791 792 leaks_fixed = check->leaks_fixed; 793 has_leaks_fixed = check->has_leaks_fixed; 794 corruptions_fixed = check->corruptions_fixed; 795 has_corruptions_fixed = check->has_corruptions_fixed; 796 797 if (output_format == OFORMAT_HUMAN) { 798 qprintf(quiet, 799 "The following inconsistencies were found and repaired:\n\n" 800 " %" PRId64 " leaked clusters\n" 801 " %" PRId64 " corruptions\n\n" 802 "Double checking the fixed image now...\n", 803 check->leaks_fixed, 804 check->corruptions_fixed); 805 } 806 807 qapi_free_ImageCheck(check); 808 check = g_new0(ImageCheck, 1); 809 ret = collect_image_check(bs, check, filename, fmt, 0); 810 811 check->leaks_fixed = leaks_fixed; 812 check->has_leaks_fixed = has_leaks_fixed; 813 check->corruptions_fixed = corruptions_fixed; 814 check->has_corruptions_fixed = has_corruptions_fixed; 815 } 816 817 if (!ret) { 818 switch (output_format) { 819 case OFORMAT_HUMAN: 820 dump_human_image_check(check, quiet); 821 break; 822 case OFORMAT_JSON: 823 dump_json_image_check(check, quiet); 824 break; 825 } 826 } 827 828 if (ret || check->check_errors) { 829 if (ret) { 830 error_report("Check failed: %s", strerror(-ret)); 831 } else { 832 error_report("Check failed"); 833 } 834 ret = 1; 835 goto fail; 836 } 837 838 if (check->corruptions) { 839 ret = 2; 840 } else if (check->leaks) { 841 ret = 3; 842 } else { 843 ret = 0; 844 } 845 846 fail: 847 qapi_free_ImageCheck(check); 848 blk_unref(blk); 849 return ret; 850 } 851 852 typedef struct CommonBlockJobCBInfo { 853 BlockDriverState *bs; 854 Error **errp; 855 } CommonBlockJobCBInfo; 856 857 static void common_block_job_cb(void *opaque, int ret) 858 { 859 CommonBlockJobCBInfo *cbi = opaque; 860 861 if (ret < 0) { 862 error_setg_errno(cbi->errp, -ret, "Block job failed"); 863 } 864 } 865 866 static void run_block_job(BlockJob *job, Error **errp) 867 { 868 uint64_t progress_current, progress_total; 869 AioContext *aio_context = block_job_get_aio_context(job); 870 int ret = 0; 871 872 job_lock(); 873 job_ref_locked(&job->job); 874 do { 875 float progress = 0.0f; 876 job_unlock(); 877 aio_poll(aio_context, true); 878 879 progress_get_snapshot(&job->job.progress, &progress_current, 880 &progress_total); 881 if (progress_total) { 882 progress = (float)progress_current / progress_total * 100.f; 883 } 884 qemu_progress_print(progress, 0); 885 job_lock(); 886 } while (!job_is_ready_locked(&job->job) && 887 !job_is_completed_locked(&job->job)); 888 889 if (!job_is_completed_locked(&job->job)) { 890 ret = job_complete_sync_locked(&job->job, errp); 891 } else { 892 ret = job->job.ret; 893 } 894 job_unref_locked(&job->job); 895 job_unlock(); 896 897 /* publish completion progress only when success */ 898 if (!ret) { 899 qemu_progress_print(100.f, 0); 900 } 901 } 902 903 static int img_commit(const img_cmd_t *ccmd, int argc, char **argv) 904 { 905 int c, ret, flags; 906 const char *filename, *fmt, *cache, *base; 907 BlockBackend *blk; 908 BlockDriverState *bs, *base_bs; 909 BlockJob *job; 910 bool progress = false, quiet = false, drop = false; 911 bool writethrough; 912 Error *local_err = NULL; 913 CommonBlockJobCBInfo cbi; 914 bool image_opts = false; 915 int64_t rate_limit = 0; 916 917 fmt = NULL; 918 cache = BDRV_DEFAULT_CACHE; 919 base = NULL; 920 for(;;) { 921 static const struct option long_options[] = { 922 {"help", no_argument, 0, 'h'}, 923 {"format", required_argument, 0, 'f'}, 924 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS}, 925 {"cache", required_argument, 0, 't'}, 926 {"drop", no_argument, 0, 'd'}, 927 {"base", required_argument, 0, 'b'}, 928 {"rate-limit", required_argument, 0, 'r'}, 929 {"progress", no_argument, 0, 'p'}, 930 {"quiet", no_argument, 0, 'q'}, 931 {"object", required_argument, 0, OPTION_OBJECT}, 932 {0, 0, 0, 0} 933 }; 934 c = getopt_long(argc, argv, "hf:t:db:r:pq", 935 long_options, NULL); 936 if (c == -1) { 937 break; 938 } 939 switch(c) { 940 case 'h': 941 cmd_help(ccmd, "[-f FMT | --image-opts] [-t CACHE_MODE] [-b BASE_IMG]\n" 942 " [-d] [-r RATE] [-q] [--object OBJDEF] FILE\n" 943 , 944 " -f, --format FMT\n" 945 " specify FILE image format explicitly (default: probing is used)\n" 946 " --image-opts\n" 947 " treat FILE as an option string (key=value,..), not a file name\n" 948 " (incompatible with -f|--format)\n" 949 " -t, --cache CACHE_MODE image cache mode (default: " BDRV_DEFAULT_CACHE ")\n" 950 " -d, --drop\n" 951 " skip emptying FILE on completion\n" 952 " -b, --base BASE_IMG\n" 953 " image in the backing chain to commit change to\n" 954 " (default: immediate backing file; implies --drop)\n" 955 " -r, --rate-limit RATE\n" 956 " I/O rate limit, in bytes per second\n" 957 " -p, --progress\n" 958 " display progress information\n" 959 " -q, --quiet\n" 960 " quiet mode (produce only error messages if any)\n" 961 " --object OBJDEF\n" 962 " defines QEMU user-creatable object\n" 963 " FILE\n" 964 " name of the image file, or an option string (key=value,..)\n" 965 " with --image-opts, to operate on\n" 966 ); 967 break; 968 case 'f': 969 fmt = optarg; 970 break; 971 case OPTION_IMAGE_OPTS: 972 image_opts = true; 973 break; 974 case 't': 975 cache = optarg; 976 break; 977 case 'd': 978 drop = true; 979 break; 980 case 'b': 981 base = optarg; 982 /* -b implies -d */ 983 drop = true; 984 break; 985 case 'r': 986 rate_limit = cvtnum("rate limit", optarg, true); 987 if (rate_limit < 0) { 988 return 1; 989 } 990 break; 991 case 'p': 992 progress = true; 993 break; 994 case 'q': 995 quiet = true; 996 break; 997 case OPTION_OBJECT: 998 user_creatable_process_cmdline(optarg); 999 break; 1000 default: 1001 tryhelp(argv[0]); 1002 } 1003 } 1004 1005 /* Progress is not shown in Quiet mode */ 1006 if (quiet) { 1007 progress = false; 1008 } 1009 1010 if (optind != argc - 1) { 1011 error_exit(argv[0], "Expecting one image file name"); 1012 } 1013 filename = argv[optind++]; 1014 1015 flags = BDRV_O_RDWR | BDRV_O_UNMAP; 1016 ret = bdrv_parse_cache_mode(cache, &flags, &writethrough); 1017 if (ret < 0) { 1018 error_report("Invalid cache option: %s", cache); 1019 return 1; 1020 } 1021 1022 blk = img_open(image_opts, filename, fmt, flags, writethrough, quiet, 1023 false); 1024 if (!blk) { 1025 return 1; 1026 } 1027 bs = blk_bs(blk); 1028 1029 qemu_progress_init(progress, 1.f); 1030 qemu_progress_print(0.f, 100); 1031 1032 bdrv_graph_rdlock_main_loop(); 1033 if (base) { 1034 base_bs = bdrv_find_backing_image(bs, base); 1035 if (!base_bs) { 1036 error_setg(&local_err, 1037 "Did not find '%s' in the backing chain of '%s'", 1038 base, filename); 1039 bdrv_graph_rdunlock_main_loop(); 1040 goto done; 1041 } 1042 } else { 1043 /* This is different from QMP, which by default uses the deepest file in 1044 * the backing chain (i.e., the very base); however, the traditional 1045 * behavior of qemu-img commit is using the immediate backing file. */ 1046 base_bs = bdrv_backing_chain_next(bs); 1047 if (!base_bs) { 1048 error_setg(&local_err, "Image does not have a backing file"); 1049 bdrv_graph_rdunlock_main_loop(); 1050 goto done; 1051 } 1052 } 1053 bdrv_graph_rdunlock_main_loop(); 1054 1055 cbi = (CommonBlockJobCBInfo){ 1056 .errp = &local_err, 1057 .bs = bs, 1058 }; 1059 1060 commit_active_start("commit", bs, base_bs, JOB_DEFAULT, rate_limit, 1061 BLOCKDEV_ON_ERROR_REPORT, NULL, common_block_job_cb, 1062 &cbi, false, &local_err); 1063 if (local_err) { 1064 goto done; 1065 } 1066 1067 /* When the block job completes, the BlockBackend reference will point to 1068 * the old backing file. In order to avoid that the top image is already 1069 * deleted, so we can still empty it afterwards, increment the reference 1070 * counter here preemptively. */ 1071 if (!drop) { 1072 bdrv_ref(bs); 1073 } 1074 1075 job = block_job_get("commit"); 1076 assert(job); 1077 run_block_job(job, &local_err); 1078 if (local_err) { 1079 goto unref_backing; 1080 } 1081 1082 if (!drop) { 1083 BlockBackend *old_backing_blk; 1084 1085 old_backing_blk = blk_new_with_bs(bs, BLK_PERM_WRITE, BLK_PERM_ALL, 1086 &local_err); 1087 if (!old_backing_blk) { 1088 goto unref_backing; 1089 } 1090 ret = blk_make_empty(old_backing_blk, &local_err); 1091 blk_unref(old_backing_blk); 1092 if (ret == -ENOTSUP) { 1093 error_free(local_err); 1094 local_err = NULL; 1095 } else if (ret < 0) { 1096 goto unref_backing; 1097 } 1098 } 1099 1100 unref_backing: 1101 if (!drop) { 1102 bdrv_unref(bs); 1103 } 1104 1105 done: 1106 qemu_progress_end(); 1107 1108 /* 1109 * Manually inactivate the image first because this way we can know whether 1110 * an error occurred. blk_unref() doesn't tell us about failures. 1111 */ 1112 ret = bdrv_inactivate_all(); 1113 if (ret < 0 && !local_err) { 1114 error_setg_errno(&local_err, -ret, "Error while closing the image"); 1115 } 1116 blk_unref(blk); 1117 1118 if (local_err) { 1119 error_report_err(local_err); 1120 return 1; 1121 } 1122 1123 qprintf(quiet, "Image committed.\n"); 1124 return 0; 1125 } 1126 1127 /* 1128 * Returns -1 if 'buf' contains only zeroes, otherwise the byte index 1129 * of the first sector boundary within buf where the sector contains a 1130 * non-zero byte. This function is robust to a buffer that is not 1131 * sector-aligned. 1132 */ 1133 static int64_t find_nonzero(const uint8_t *buf, int64_t n) 1134 { 1135 int64_t i; 1136 int64_t end = QEMU_ALIGN_DOWN(n, BDRV_SECTOR_SIZE); 1137 1138 for (i = 0; i < end; i += BDRV_SECTOR_SIZE) { 1139 if (!buffer_is_zero(buf + i, BDRV_SECTOR_SIZE)) { 1140 return i; 1141 } 1142 } 1143 if (i < n && !buffer_is_zero(buf + i, n - end)) { 1144 return i; 1145 } 1146 return -1; 1147 } 1148 1149 /* 1150 * Returns true iff the first sector pointed to by 'buf' contains at least 1151 * a non-NUL byte. 1152 * 1153 * 'pnum' is set to the number of sectors (including and immediately following 1154 * the first one) that are known to be in the same allocated/unallocated state. 1155 * The function will try to align the end offset to alignment boundaries so 1156 * that the request will at least end aligned and consecutive requests will 1157 * also start at an aligned offset. 1158 */ 1159 static int is_allocated_sectors(const uint8_t *buf, int n, int *pnum, 1160 int64_t sector_num, int alignment) 1161 { 1162 bool is_zero; 1163 int i, tail; 1164 1165 if (n <= 0) { 1166 *pnum = 0; 1167 return 0; 1168 } 1169 is_zero = buffer_is_zero(buf, BDRV_SECTOR_SIZE); 1170 for(i = 1; i < n; i++) { 1171 buf += BDRV_SECTOR_SIZE; 1172 if (is_zero != buffer_is_zero(buf, BDRV_SECTOR_SIZE)) { 1173 break; 1174 } 1175 } 1176 1177 if (i == n) { 1178 /* 1179 * The whole buf is the same. 1180 * No reason to split it into chunks, so return now. 1181 */ 1182 *pnum = i; 1183 return !is_zero; 1184 } 1185 1186 tail = (sector_num + i) & (alignment - 1); 1187 if (tail) { 1188 if (is_zero && i <= tail) { 1189 /* 1190 * For sure next sector after i is data, and it will rewrite this 1191 * tail anyway due to RMW. So, let's just write data now. 1192 */ 1193 is_zero = false; 1194 } 1195 if (!is_zero) { 1196 /* If possible, align up end offset of allocated areas. */ 1197 i += alignment - tail; 1198 i = MIN(i, n); 1199 } else { 1200 /* 1201 * For sure next sector after i is data, and it will rewrite this 1202 * tail anyway due to RMW. Better is avoid RMW and write zeroes up 1203 * to aligned bound. 1204 */ 1205 i -= tail; 1206 } 1207 } 1208 *pnum = i; 1209 return !is_zero; 1210 } 1211 1212 /* 1213 * Like is_allocated_sectors, but if the buffer starts with a used sector, 1214 * up to 'min' consecutive sectors containing zeros are ignored. This avoids 1215 * breaking up write requests for only small sparse areas. 1216 */ 1217 static int is_allocated_sectors_min(const uint8_t *buf, int n, int *pnum, 1218 int min, int64_t sector_num, int alignment) 1219 { 1220 int ret; 1221 int num_checked, num_used; 1222 1223 if (n < min) { 1224 min = n; 1225 } 1226 1227 ret = is_allocated_sectors(buf, n, pnum, sector_num, alignment); 1228 if (!ret) { 1229 return ret; 1230 } 1231 1232 num_used = *pnum; 1233 buf += BDRV_SECTOR_SIZE * *pnum; 1234 n -= *pnum; 1235 sector_num += *pnum; 1236 num_checked = num_used; 1237 1238 while (n > 0) { 1239 ret = is_allocated_sectors(buf, n, pnum, sector_num, alignment); 1240 1241 buf += BDRV_SECTOR_SIZE * *pnum; 1242 n -= *pnum; 1243 sector_num += *pnum; 1244 num_checked += *pnum; 1245 if (ret) { 1246 num_used = num_checked; 1247 } else if (*pnum >= min) { 1248 break; 1249 } 1250 } 1251 1252 *pnum = num_used; 1253 return 1; 1254 } 1255 1256 /* 1257 * Compares two buffers chunk by chunk, where @chsize is the chunk size. 1258 * If @chsize is 0, default chunk size of BDRV_SECTOR_SIZE is used. 1259 * Returns 0 if the first chunk of each buffer matches, non-zero otherwise. 1260 * 1261 * @pnum is set to the size of the buffer prefix aligned to @chsize that 1262 * has the same matching status as the first chunk. 1263 */ 1264 static int compare_buffers(const uint8_t *buf1, const uint8_t *buf2, 1265 int64_t bytes, uint64_t chsize, int64_t *pnum) 1266 { 1267 bool res; 1268 int64_t i; 1269 1270 assert(bytes > 0); 1271 1272 if (!chsize) { 1273 chsize = BDRV_SECTOR_SIZE; 1274 } 1275 i = MIN(bytes, chsize); 1276 1277 res = !!memcmp(buf1, buf2, i); 1278 while (i < bytes) { 1279 int64_t len = MIN(bytes - i, chsize); 1280 1281 if (!!memcmp(buf1 + i, buf2 + i, len) != res) { 1282 break; 1283 } 1284 i += len; 1285 } 1286 1287 *pnum = i; 1288 return res; 1289 } 1290 1291 #define IO_BUF_SIZE (2 * MiB) 1292 1293 /* 1294 * Check if passed sectors are empty (not allocated or contain only 0 bytes) 1295 * 1296 * Intended for use by 'qemu-img compare': Returns 0 in case sectors are 1297 * filled with 0, 1 if sectors contain non-zero data (this is a comparison 1298 * failure), and 4 on error (the exit status for read errors), after emitting 1299 * an error message. 1300 * 1301 * @param blk: BlockBackend for the image 1302 * @param offset: Starting offset to check 1303 * @param bytes: Number of bytes to check 1304 * @param filename: Name of disk file we are checking (logging purpose) 1305 * @param buffer: Allocated buffer for storing read data 1306 * @param quiet: Flag for quiet mode 1307 */ 1308 static int check_empty_sectors(BlockBackend *blk, int64_t offset, 1309 int64_t bytes, const char *filename, 1310 uint8_t *buffer, bool quiet) 1311 { 1312 int ret = 0; 1313 int64_t idx; 1314 1315 ret = blk_pread(blk, offset, bytes, buffer, 0); 1316 if (ret < 0) { 1317 error_report("Error while reading offset %" PRId64 " of %s: %s", 1318 offset, filename, strerror(-ret)); 1319 return 4; 1320 } 1321 idx = find_nonzero(buffer, bytes); 1322 if (idx >= 0) { 1323 qprintf(quiet, "Content mismatch at offset %" PRId64 "!\n", 1324 offset + idx); 1325 return 1; 1326 } 1327 1328 return 0; 1329 } 1330 1331 /* 1332 * Compares two images. Exit codes: 1333 * 1334 * 0 - Images are identical or the requested help was printed 1335 * 1 - Images differ 1336 * >1 - Error occurred 1337 */ 1338 static int img_compare(const img_cmd_t *ccmd, int argc, char **argv) 1339 { 1340 const char *fmt1 = NULL, *fmt2 = NULL, *cache, *filename1, *filename2; 1341 BlockBackend *blk1, *blk2; 1342 BlockDriverState *bs1, *bs2; 1343 int64_t total_size1, total_size2; 1344 uint8_t *buf1 = NULL, *buf2 = NULL; 1345 int64_t pnum1, pnum2; 1346 int allocated1, allocated2; 1347 int ret = 0; /* return value - 0 Ident, 1 Different, >1 Error */ 1348 bool progress = false, quiet = false, strict = false; 1349 int flags; 1350 bool writethrough; 1351 int64_t total_size; 1352 int64_t offset = 0; 1353 int64_t chunk; 1354 int c; 1355 uint64_t progress_base; 1356 bool image_opts = false; 1357 bool force_share = false; 1358 1359 cache = BDRV_DEFAULT_CACHE; 1360 for (;;) { 1361 static const struct option long_options[] = { 1362 {"help", no_argument, 0, 'h'}, 1363 {"a-format", required_argument, 0, 'f'}, 1364 {"b-format", required_argument, 0, 'F'}, 1365 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS}, 1366 {"strict", no_argument, 0, 's'}, 1367 {"cache", required_argument, 0, 'T'}, 1368 {"force-share", no_argument, 0, 'U'}, 1369 {"progress", no_argument, 0, 'p'}, 1370 {"quiet", no_argument, 0, 'q'}, 1371 {"object", required_argument, 0, OPTION_OBJECT}, 1372 {0, 0, 0, 0} 1373 }; 1374 c = getopt_long(argc, argv, "hf:F:sT:Upq", 1375 long_options, NULL); 1376 if (c == -1) { 1377 break; 1378 } 1379 switch (c) { 1380 case 'h': 1381 cmd_help(ccmd, 1382 "[[-f FMT] [-F FMT] | --image-opts] [-s] [-T CACHE]\n" 1383 " [-U] [-p] [-q] [--object OBJDEF] FILE1 FILE2\n" 1384 , 1385 " -f, --a-format FMT\n" 1386 " specify FILE1 image format explicitly (default: probing is used)\n" 1387 " -F, --b-format FMT\n" 1388 " specify FILE2 image format explicitly (default: probing is used)\n" 1389 " --image-opts\n" 1390 " treat FILE1 and FILE2 as option strings (key=value,..), not file names\n" 1391 " (incompatible with -f|--a-format and -F|--b-format)\n" 1392 " -s, --strict\n" 1393 " strict mode, also check if sizes are equal\n" 1394 " -T, --cache CACHE_MODE\n" 1395 " images caching mode (default: " BDRV_DEFAULT_CACHE ")\n" 1396 " -U, --force-share\n" 1397 " open images in shared mode for concurrent access\n" 1398 " -p, --progress\n" 1399 " display progress information\n" 1400 " -q, --quiet\n" 1401 " quiet mode (produce only error messages if any)\n" 1402 " --object OBJDEF\n" 1403 " defines QEMU user-creatable object\n" 1404 " FILE1, FILE2\n" 1405 " names of the image files, or option strings (key=value,..)\n" 1406 " with --image-opts, to compare\n" 1407 ); 1408 break; 1409 case 'f': 1410 fmt1 = optarg; 1411 break; 1412 case 'F': 1413 fmt2 = optarg; 1414 break; 1415 case OPTION_IMAGE_OPTS: 1416 image_opts = true; 1417 break; 1418 case 's': 1419 strict = true; 1420 break; 1421 case 'T': 1422 cache = optarg; 1423 break; 1424 case 'U': 1425 force_share = true; 1426 break; 1427 case 'p': 1428 progress = true; 1429 break; 1430 case 'q': 1431 quiet = true; 1432 break; 1433 case OPTION_OBJECT: 1434 user_creatable_process_cmdline(optarg); 1435 break; 1436 default: 1437 tryhelp(argv[0]); 1438 } 1439 } 1440 1441 /* Progress is not shown in Quiet mode */ 1442 if (quiet) { 1443 progress = false; 1444 } 1445 1446 1447 if (optind != argc - 2) { 1448 error_exit(argv[0], "Expecting two image file names"); 1449 } 1450 filename1 = argv[optind++]; 1451 filename2 = argv[optind++]; 1452 1453 /* Initialize before goto out */ 1454 qemu_progress_init(progress, 2.0); 1455 1456 flags = 0; 1457 ret = bdrv_parse_cache_mode(cache, &flags, &writethrough); 1458 if (ret < 0) { 1459 error_report("Invalid source cache option: %s", cache); 1460 ret = 2; 1461 goto out3; 1462 } 1463 1464 blk1 = img_open(image_opts, filename1, fmt1, flags, writethrough, quiet, 1465 force_share); 1466 if (!blk1) { 1467 ret = 2; 1468 goto out3; 1469 } 1470 1471 blk2 = img_open(image_opts, filename2, fmt2, flags, writethrough, quiet, 1472 force_share); 1473 if (!blk2) { 1474 ret = 2; 1475 goto out2; 1476 } 1477 bs1 = blk_bs(blk1); 1478 bs2 = blk_bs(blk2); 1479 1480 buf1 = blk_blockalign(blk1, IO_BUF_SIZE); 1481 buf2 = blk_blockalign(blk2, IO_BUF_SIZE); 1482 total_size1 = blk_getlength(blk1); 1483 if (total_size1 < 0) { 1484 error_report("Can't get size of %s: %s", 1485 filename1, strerror(-total_size1)); 1486 ret = 4; 1487 goto out; 1488 } 1489 total_size2 = blk_getlength(blk2); 1490 if (total_size2 < 0) { 1491 error_report("Can't get size of %s: %s", 1492 filename2, strerror(-total_size2)); 1493 ret = 4; 1494 goto out; 1495 } 1496 total_size = MIN(total_size1, total_size2); 1497 progress_base = MAX(total_size1, total_size2); 1498 1499 qemu_progress_print(0, 100); 1500 1501 if (strict && total_size1 != total_size2) { 1502 ret = 1; 1503 qprintf(quiet, "Strict mode: Image size mismatch!\n"); 1504 goto out; 1505 } 1506 1507 while (offset < total_size) { 1508 int status1, status2; 1509 1510 status1 = bdrv_block_status_above(bs1, NULL, offset, 1511 total_size1 - offset, &pnum1, NULL, 1512 NULL); 1513 if (status1 < 0) { 1514 ret = 3; 1515 error_report("Sector allocation test failed for %s", filename1); 1516 goto out; 1517 } 1518 allocated1 = status1 & BDRV_BLOCK_ALLOCATED; 1519 1520 status2 = bdrv_block_status_above(bs2, NULL, offset, 1521 total_size2 - offset, &pnum2, NULL, 1522 NULL); 1523 if (status2 < 0) { 1524 ret = 3; 1525 error_report("Sector allocation test failed for %s", filename2); 1526 goto out; 1527 } 1528 allocated2 = status2 & BDRV_BLOCK_ALLOCATED; 1529 1530 assert(pnum1 && pnum2); 1531 chunk = MIN(pnum1, pnum2); 1532 1533 if (strict) { 1534 if (status1 != status2) { 1535 ret = 1; 1536 qprintf(quiet, "Strict mode: Offset %" PRId64 1537 " block status mismatch!\n", offset); 1538 goto out; 1539 } 1540 } 1541 if ((status1 & BDRV_BLOCK_ZERO) && (status2 & BDRV_BLOCK_ZERO)) { 1542 /* nothing to do */ 1543 } else if (allocated1 == allocated2) { 1544 if (allocated1) { 1545 int64_t pnum; 1546 1547 chunk = MIN(chunk, IO_BUF_SIZE); 1548 ret = blk_pread(blk1, offset, chunk, buf1, 0); 1549 if (ret < 0) { 1550 error_report("Error while reading offset %" PRId64 1551 " of %s: %s", 1552 offset, filename1, strerror(-ret)); 1553 ret = 4; 1554 goto out; 1555 } 1556 ret = blk_pread(blk2, offset, chunk, buf2, 0); 1557 if (ret < 0) { 1558 error_report("Error while reading offset %" PRId64 1559 " of %s: %s", 1560 offset, filename2, strerror(-ret)); 1561 ret = 4; 1562 goto out; 1563 } 1564 ret = compare_buffers(buf1, buf2, chunk, 0, &pnum); 1565 if (ret || pnum != chunk) { 1566 qprintf(quiet, "Content mismatch at offset %" PRId64 "!\n", 1567 offset + (ret ? 0 : pnum)); 1568 ret = 1; 1569 goto out; 1570 } 1571 } 1572 } else { 1573 chunk = MIN(chunk, IO_BUF_SIZE); 1574 if (allocated1) { 1575 ret = check_empty_sectors(blk1, offset, chunk, 1576 filename1, buf1, quiet); 1577 } else { 1578 ret = check_empty_sectors(blk2, offset, chunk, 1579 filename2, buf1, quiet); 1580 } 1581 if (ret) { 1582 goto out; 1583 } 1584 } 1585 offset += chunk; 1586 qemu_progress_print(((float) chunk / progress_base) * 100, 100); 1587 } 1588 1589 if (total_size1 != total_size2) { 1590 BlockBackend *blk_over; 1591 const char *filename_over; 1592 1593 qprintf(quiet, "Warning: Image size mismatch!\n"); 1594 if (total_size1 > total_size2) { 1595 blk_over = blk1; 1596 filename_over = filename1; 1597 } else { 1598 blk_over = blk2; 1599 filename_over = filename2; 1600 } 1601 1602 while (offset < progress_base) { 1603 ret = bdrv_block_status_above(blk_bs(blk_over), NULL, offset, 1604 progress_base - offset, &chunk, 1605 NULL, NULL); 1606 if (ret < 0) { 1607 ret = 3; 1608 error_report("Sector allocation test failed for %s", 1609 filename_over); 1610 goto out; 1611 1612 } 1613 if (ret & BDRV_BLOCK_ALLOCATED && !(ret & BDRV_BLOCK_ZERO)) { 1614 chunk = MIN(chunk, IO_BUF_SIZE); 1615 ret = check_empty_sectors(blk_over, offset, chunk, 1616 filename_over, buf1, quiet); 1617 if (ret) { 1618 goto out; 1619 } 1620 } 1621 offset += chunk; 1622 qemu_progress_print(((float) chunk / progress_base) * 100, 100); 1623 } 1624 } 1625 1626 qprintf(quiet, "Images are identical.\n"); 1627 ret = 0; 1628 1629 out: 1630 qemu_vfree(buf1); 1631 qemu_vfree(buf2); 1632 blk_unref(blk2); 1633 out2: 1634 blk_unref(blk1); 1635 out3: 1636 qemu_progress_end(); 1637 return ret; 1638 } 1639 1640 /* Convenience wrapper around qmp_block_dirty_bitmap_merge */ 1641 static void do_dirty_bitmap_merge(const char *dst_node, const char *dst_name, 1642 const char *src_node, const char *src_name, 1643 Error **errp) 1644 { 1645 BlockDirtyBitmapOrStr *merge_src; 1646 BlockDirtyBitmapOrStrList *list = NULL; 1647 1648 merge_src = g_new0(BlockDirtyBitmapOrStr, 1); 1649 merge_src->type = QTYPE_QDICT; 1650 merge_src->u.external.node = g_strdup(src_node); 1651 merge_src->u.external.name = g_strdup(src_name); 1652 QAPI_LIST_PREPEND(list, merge_src); 1653 qmp_block_dirty_bitmap_merge(dst_node, dst_name, list, errp); 1654 qapi_free_BlockDirtyBitmapOrStrList(list); 1655 } 1656 1657 enum ImgConvertBlockStatus { 1658 BLK_DATA, 1659 BLK_ZERO, 1660 BLK_BACKING_FILE, 1661 }; 1662 1663 #define MAX_COROUTINES 16 1664 #define CONVERT_THROTTLE_GROUP "img_convert" 1665 1666 typedef struct ImgConvertState { 1667 BlockBackend **src; 1668 int64_t *src_sectors; 1669 int *src_alignment; 1670 int src_num; 1671 int64_t total_sectors; 1672 int64_t allocated_sectors; 1673 int64_t allocated_done; 1674 int64_t sector_num; 1675 int64_t wr_offs; 1676 enum ImgConvertBlockStatus status; 1677 int64_t sector_next_status; 1678 BlockBackend *target; 1679 bool has_zero_init; 1680 bool compressed; 1681 bool target_is_new; 1682 bool target_has_backing; 1683 int64_t target_backing_sectors; /* negative if unknown */ 1684 bool wr_in_order; 1685 bool copy_range; 1686 bool salvage; 1687 bool quiet; 1688 int min_sparse; 1689 int alignment; 1690 size_t cluster_sectors; 1691 size_t buf_sectors; 1692 long num_coroutines; 1693 int running_coroutines; 1694 Coroutine *co[MAX_COROUTINES]; 1695 int64_t wait_sector_num[MAX_COROUTINES]; 1696 CoMutex lock; 1697 int ret; 1698 } ImgConvertState; 1699 1700 static void convert_select_part(ImgConvertState *s, int64_t sector_num, 1701 int *src_cur, int64_t *src_cur_offset) 1702 { 1703 *src_cur = 0; 1704 *src_cur_offset = 0; 1705 while (sector_num - *src_cur_offset >= s->src_sectors[*src_cur]) { 1706 *src_cur_offset += s->src_sectors[*src_cur]; 1707 (*src_cur)++; 1708 assert(*src_cur < s->src_num); 1709 } 1710 } 1711 1712 static int coroutine_mixed_fn GRAPH_RDLOCK 1713 convert_iteration_sectors(ImgConvertState *s, int64_t sector_num) 1714 { 1715 int64_t src_cur_offset; 1716 int ret, n, src_cur; 1717 bool post_backing_zero = false; 1718 1719 convert_select_part(s, sector_num, &src_cur, &src_cur_offset); 1720 1721 assert(s->total_sectors > sector_num); 1722 n = MIN(s->total_sectors - sector_num, BDRV_REQUEST_MAX_SECTORS); 1723 1724 if (s->target_backing_sectors >= 0) { 1725 if (sector_num >= s->target_backing_sectors) { 1726 post_backing_zero = true; 1727 } else if (sector_num + n > s->target_backing_sectors) { 1728 /* Split requests around target_backing_sectors (because 1729 * starting from there, zeros are handled differently) */ 1730 n = s->target_backing_sectors - sector_num; 1731 } 1732 } 1733 1734 if (s->sector_next_status <= sector_num) { 1735 uint64_t offset = (sector_num - src_cur_offset) * BDRV_SECTOR_SIZE; 1736 int64_t count; 1737 int tail; 1738 BlockDriverState *src_bs = blk_bs(s->src[src_cur]); 1739 BlockDriverState *base; 1740 1741 if (s->target_has_backing) { 1742 base = bdrv_cow_bs(bdrv_skip_filters(src_bs)); 1743 } else { 1744 base = NULL; 1745 } 1746 1747 do { 1748 count = n * BDRV_SECTOR_SIZE; 1749 1750 ret = bdrv_block_status_above(src_bs, base, offset, count, &count, 1751 NULL, NULL); 1752 1753 if (ret < 0) { 1754 if (s->salvage) { 1755 if (n == 1) { 1756 if (!s->quiet) { 1757 warn_report("error while reading block status at " 1758 "offset %" PRIu64 ": %s", offset, 1759 strerror(-ret)); 1760 } 1761 /* Just try to read the data, then */ 1762 ret = BDRV_BLOCK_DATA; 1763 count = BDRV_SECTOR_SIZE; 1764 } else { 1765 /* Retry on a shorter range */ 1766 n = DIV_ROUND_UP(n, 4); 1767 } 1768 } else { 1769 error_report("error while reading block status at offset " 1770 "%" PRIu64 ": %s", offset, strerror(-ret)); 1771 return ret; 1772 } 1773 } 1774 } while (ret < 0); 1775 1776 n = DIV_ROUND_UP(count, BDRV_SECTOR_SIZE); 1777 1778 /* 1779 * Avoid that s->sector_next_status becomes unaligned to the source 1780 * request alignment and/or cluster size to avoid unnecessary read 1781 * cycles. 1782 */ 1783 tail = (sector_num - src_cur_offset + n) % s->src_alignment[src_cur]; 1784 if (n > tail) { 1785 n -= tail; 1786 } 1787 1788 if (ret & BDRV_BLOCK_ZERO) { 1789 s->status = post_backing_zero ? BLK_BACKING_FILE : BLK_ZERO; 1790 } else if (ret & BDRV_BLOCK_DATA) { 1791 s->status = BLK_DATA; 1792 } else { 1793 s->status = s->target_has_backing ? BLK_BACKING_FILE : BLK_DATA; 1794 } 1795 1796 s->sector_next_status = sector_num + n; 1797 } 1798 1799 n = MIN(n, s->sector_next_status - sector_num); 1800 if (s->status == BLK_DATA) { 1801 n = MIN(n, s->buf_sectors); 1802 } 1803 1804 /* We need to write complete clusters for compressed images, so if an 1805 * unallocated area is shorter than that, we must consider the whole 1806 * cluster allocated. */ 1807 if (s->compressed) { 1808 if (n < s->cluster_sectors) { 1809 n = MIN(s->cluster_sectors, s->total_sectors - sector_num); 1810 s->status = BLK_DATA; 1811 } else { 1812 n = QEMU_ALIGN_DOWN(n, s->cluster_sectors); 1813 } 1814 } 1815 1816 return n; 1817 } 1818 1819 static int coroutine_fn convert_co_read(ImgConvertState *s, int64_t sector_num, 1820 int nb_sectors, uint8_t *buf) 1821 { 1822 uint64_t single_read_until = 0; 1823 int n, ret; 1824 1825 assert(nb_sectors <= s->buf_sectors); 1826 while (nb_sectors > 0) { 1827 BlockBackend *blk; 1828 int src_cur; 1829 int64_t bs_sectors, src_cur_offset; 1830 uint64_t offset; 1831 1832 /* In the case of compression with multiple source files, we can get a 1833 * nb_sectors that spreads into the next part. So we must be able to 1834 * read across multiple BDSes for one convert_read() call. */ 1835 convert_select_part(s, sector_num, &src_cur, &src_cur_offset); 1836 blk = s->src[src_cur]; 1837 bs_sectors = s->src_sectors[src_cur]; 1838 1839 offset = (sector_num - src_cur_offset) << BDRV_SECTOR_BITS; 1840 1841 n = MIN(nb_sectors, bs_sectors - (sector_num - src_cur_offset)); 1842 if (single_read_until > offset) { 1843 n = 1; 1844 } 1845 1846 ret = blk_co_pread(blk, offset, n << BDRV_SECTOR_BITS, buf, 0); 1847 if (ret < 0) { 1848 if (s->salvage) { 1849 if (n > 1) { 1850 single_read_until = offset + (n << BDRV_SECTOR_BITS); 1851 continue; 1852 } else { 1853 if (!s->quiet) { 1854 warn_report("error while reading offset %" PRIu64 1855 ": %s", offset, strerror(-ret)); 1856 } 1857 memset(buf, 0, BDRV_SECTOR_SIZE); 1858 } 1859 } else { 1860 return ret; 1861 } 1862 } 1863 1864 sector_num += n; 1865 nb_sectors -= n; 1866 buf += n * BDRV_SECTOR_SIZE; 1867 } 1868 1869 return 0; 1870 } 1871 1872 1873 static int coroutine_fn convert_co_write(ImgConvertState *s, int64_t sector_num, 1874 int nb_sectors, uint8_t *buf, 1875 enum ImgConvertBlockStatus status) 1876 { 1877 int ret; 1878 1879 while (nb_sectors > 0) { 1880 int n = nb_sectors; 1881 BdrvRequestFlags flags = s->compressed ? BDRV_REQ_WRITE_COMPRESSED : 0; 1882 1883 switch (status) { 1884 case BLK_BACKING_FILE: 1885 /* If we have a backing file, leave clusters unallocated that are 1886 * unallocated in the source image, so that the backing file is 1887 * visible at the respective offset. */ 1888 assert(s->target_has_backing); 1889 break; 1890 1891 case BLK_DATA: 1892 /* If we're told to keep the target fully allocated (-S 0) or there 1893 * is real non-zero data, we must write it. Otherwise we can treat 1894 * it as zero sectors. 1895 * Compressed clusters need to be written as a whole, so in that 1896 * case we can only save the write if the buffer is completely 1897 * zeroed. */ 1898 if (!s->min_sparse || 1899 (!s->compressed && 1900 is_allocated_sectors_min(buf, n, &n, s->min_sparse, 1901 sector_num, s->alignment)) || 1902 (s->compressed && 1903 !buffer_is_zero(buf, n * BDRV_SECTOR_SIZE))) 1904 { 1905 ret = blk_co_pwrite(s->target, sector_num << BDRV_SECTOR_BITS, 1906 n << BDRV_SECTOR_BITS, buf, flags); 1907 if (ret < 0) { 1908 return ret; 1909 } 1910 break; 1911 } 1912 /* fall-through */ 1913 1914 case BLK_ZERO: 1915 if (s->has_zero_init) { 1916 assert(!s->target_has_backing); 1917 break; 1918 } 1919 ret = blk_co_pwrite_zeroes(s->target, 1920 sector_num << BDRV_SECTOR_BITS, 1921 n << BDRV_SECTOR_BITS, 1922 BDRV_REQ_MAY_UNMAP); 1923 if (ret < 0) { 1924 return ret; 1925 } 1926 break; 1927 } 1928 1929 sector_num += n; 1930 nb_sectors -= n; 1931 buf += n * BDRV_SECTOR_SIZE; 1932 } 1933 1934 return 0; 1935 } 1936 1937 static int coroutine_fn convert_co_copy_range(ImgConvertState *s, int64_t sector_num, 1938 int nb_sectors) 1939 { 1940 int n, ret; 1941 1942 while (nb_sectors > 0) { 1943 BlockBackend *blk; 1944 int src_cur; 1945 int64_t bs_sectors, src_cur_offset; 1946 int64_t offset; 1947 1948 convert_select_part(s, sector_num, &src_cur, &src_cur_offset); 1949 offset = (sector_num - src_cur_offset) << BDRV_SECTOR_BITS; 1950 blk = s->src[src_cur]; 1951 bs_sectors = s->src_sectors[src_cur]; 1952 1953 n = MIN(nb_sectors, bs_sectors - (sector_num - src_cur_offset)); 1954 1955 ret = blk_co_copy_range(blk, offset, s->target, 1956 sector_num << BDRV_SECTOR_BITS, 1957 n << BDRV_SECTOR_BITS, 0, 0); 1958 if (ret < 0) { 1959 return ret; 1960 } 1961 1962 sector_num += n; 1963 nb_sectors -= n; 1964 } 1965 return 0; 1966 } 1967 1968 static void coroutine_fn convert_co_do_copy(void *opaque) 1969 { 1970 ImgConvertState *s = opaque; 1971 uint8_t *buf = NULL; 1972 int ret, i; 1973 int index = -1; 1974 1975 for (i = 0; i < s->num_coroutines; i++) { 1976 if (s->co[i] == qemu_coroutine_self()) { 1977 index = i; 1978 break; 1979 } 1980 } 1981 assert(index >= 0); 1982 1983 s->running_coroutines++; 1984 buf = blk_blockalign(s->target, s->buf_sectors * BDRV_SECTOR_SIZE); 1985 1986 while (1) { 1987 int n; 1988 int64_t sector_num; 1989 enum ImgConvertBlockStatus status; 1990 bool copy_range; 1991 1992 qemu_co_mutex_lock(&s->lock); 1993 if (s->ret != -EINPROGRESS || s->sector_num >= s->total_sectors) { 1994 qemu_co_mutex_unlock(&s->lock); 1995 break; 1996 } 1997 WITH_GRAPH_RDLOCK_GUARD() { 1998 n = convert_iteration_sectors(s, s->sector_num); 1999 } 2000 if (n < 0) { 2001 qemu_co_mutex_unlock(&s->lock); 2002 s->ret = n; 2003 break; 2004 } 2005 /* save current sector and allocation status to local variables */ 2006 sector_num = s->sector_num; 2007 status = s->status; 2008 if (!s->min_sparse && s->status == BLK_ZERO) { 2009 n = MIN(n, s->buf_sectors); 2010 } 2011 /* increment global sector counter so that other coroutines can 2012 * already continue reading beyond this request */ 2013 s->sector_num += n; 2014 qemu_co_mutex_unlock(&s->lock); 2015 2016 if (status == BLK_DATA || (!s->min_sparse && status == BLK_ZERO)) { 2017 s->allocated_done += n; 2018 qemu_progress_print(100.0 * s->allocated_done / 2019 s->allocated_sectors, 0); 2020 } 2021 2022 retry: 2023 copy_range = s->copy_range && s->status == BLK_DATA; 2024 if (status == BLK_DATA && !copy_range) { 2025 ret = convert_co_read(s, sector_num, n, buf); 2026 if (ret < 0) { 2027 error_report("error while reading at byte %lld: %s", 2028 sector_num * BDRV_SECTOR_SIZE, strerror(-ret)); 2029 s->ret = ret; 2030 } 2031 } else if (!s->min_sparse && status == BLK_ZERO) { 2032 status = BLK_DATA; 2033 memset(buf, 0x00, n * BDRV_SECTOR_SIZE); 2034 } 2035 2036 if (s->wr_in_order) { 2037 /* keep writes in order */ 2038 while (s->wr_offs != sector_num && s->ret == -EINPROGRESS) { 2039 s->wait_sector_num[index] = sector_num; 2040 qemu_coroutine_yield(); 2041 } 2042 s->wait_sector_num[index] = -1; 2043 } 2044 2045 if (s->ret == -EINPROGRESS) { 2046 if (copy_range) { 2047 WITH_GRAPH_RDLOCK_GUARD() { 2048 ret = convert_co_copy_range(s, sector_num, n); 2049 } 2050 if (ret) { 2051 s->copy_range = false; 2052 goto retry; 2053 } 2054 } else { 2055 ret = convert_co_write(s, sector_num, n, buf, status); 2056 } 2057 if (ret < 0) { 2058 error_report("error while writing at byte %lld: %s", 2059 sector_num * BDRV_SECTOR_SIZE, strerror(-ret)); 2060 s->ret = ret; 2061 } 2062 } 2063 2064 if (s->wr_in_order) { 2065 /* reenter the coroutine that might have waited 2066 * for this write to complete */ 2067 s->wr_offs = sector_num + n; 2068 for (i = 0; i < s->num_coroutines; i++) { 2069 if (s->co[i] && s->wait_sector_num[i] == s->wr_offs) { 2070 /* 2071 * A -> B -> A cannot occur because A has 2072 * s->wait_sector_num[i] == -1 during A -> B. Therefore 2073 * B will never enter A during this time window. 2074 */ 2075 qemu_coroutine_enter(s->co[i]); 2076 break; 2077 } 2078 } 2079 } 2080 } 2081 2082 qemu_vfree(buf); 2083 s->co[index] = NULL; 2084 s->running_coroutines--; 2085 if (!s->running_coroutines && s->ret == -EINPROGRESS) { 2086 /* the convert job finished successfully */ 2087 s->ret = 0; 2088 } 2089 } 2090 2091 static int convert_do_copy(ImgConvertState *s) 2092 { 2093 int ret, i, n; 2094 int64_t sector_num = 0; 2095 2096 /* Check whether we have zero initialisation or can get it efficiently */ 2097 if (!s->has_zero_init && s->target_is_new && s->min_sparse && 2098 !s->target_has_backing) { 2099 bdrv_graph_rdlock_main_loop(); 2100 s->has_zero_init = bdrv_has_zero_init(blk_bs(s->target)); 2101 bdrv_graph_rdunlock_main_loop(); 2102 } 2103 2104 /* Allocate buffer for copied data. For compressed images, only one cluster 2105 * can be copied at a time. */ 2106 if (s->compressed) { 2107 if (s->cluster_sectors <= 0 || s->cluster_sectors > s->buf_sectors) { 2108 error_report("invalid cluster size"); 2109 return -EINVAL; 2110 } 2111 s->buf_sectors = s->cluster_sectors; 2112 } 2113 2114 while (sector_num < s->total_sectors) { 2115 bdrv_graph_rdlock_main_loop(); 2116 n = convert_iteration_sectors(s, sector_num); 2117 bdrv_graph_rdunlock_main_loop(); 2118 if (n < 0) { 2119 return n; 2120 } 2121 if (s->status == BLK_DATA || (!s->min_sparse && s->status == BLK_ZERO)) 2122 { 2123 s->allocated_sectors += n; 2124 } 2125 sector_num += n; 2126 } 2127 2128 /* Do the copy */ 2129 s->sector_next_status = 0; 2130 s->ret = -EINPROGRESS; 2131 2132 qemu_co_mutex_init(&s->lock); 2133 for (i = 0; i < s->num_coroutines; i++) { 2134 s->co[i] = qemu_coroutine_create(convert_co_do_copy, s); 2135 s->wait_sector_num[i] = -1; 2136 qemu_coroutine_enter(s->co[i]); 2137 } 2138 2139 while (s->running_coroutines) { 2140 main_loop_wait(false); 2141 } 2142 2143 if (s->compressed && !s->ret) { 2144 /* signal EOF to align */ 2145 ret = blk_pwrite_compressed(s->target, 0, 0, NULL); 2146 if (ret < 0) { 2147 return ret; 2148 } 2149 } 2150 2151 return s->ret; 2152 } 2153 2154 /* Check that bitmaps can be copied, or output an error */ 2155 static int convert_check_bitmaps(BlockDriverState *src, bool skip_broken) 2156 { 2157 BdrvDirtyBitmap *bm; 2158 2159 if (!bdrv_supports_persistent_dirty_bitmap(src)) { 2160 error_report("Source lacks bitmap support"); 2161 return -1; 2162 } 2163 FOR_EACH_DIRTY_BITMAP(src, bm) { 2164 if (!bdrv_dirty_bitmap_get_persistence(bm)) { 2165 continue; 2166 } 2167 if (!skip_broken && bdrv_dirty_bitmap_inconsistent(bm)) { 2168 error_report("Cannot copy inconsistent bitmap '%s'", 2169 bdrv_dirty_bitmap_name(bm)); 2170 error_printf("Try --skip-broken-bitmaps, or " 2171 "use 'qemu-img bitmap --remove' to delete it\n"); 2172 return -1; 2173 } 2174 } 2175 return 0; 2176 } 2177 2178 static int convert_copy_bitmaps(BlockDriverState *src, BlockDriverState *dst, 2179 bool skip_broken) 2180 { 2181 BdrvDirtyBitmap *bm; 2182 Error *err = NULL; 2183 2184 FOR_EACH_DIRTY_BITMAP(src, bm) { 2185 const char *name; 2186 2187 if (!bdrv_dirty_bitmap_get_persistence(bm)) { 2188 continue; 2189 } 2190 name = bdrv_dirty_bitmap_name(bm); 2191 if (skip_broken && bdrv_dirty_bitmap_inconsistent(bm)) { 2192 warn_report("Skipping inconsistent bitmap '%s'", name); 2193 continue; 2194 } 2195 qmp_block_dirty_bitmap_add(dst->node_name, name, 2196 true, bdrv_dirty_bitmap_granularity(bm), 2197 true, true, 2198 true, !bdrv_dirty_bitmap_enabled(bm), 2199 &err); 2200 if (err) { 2201 error_reportf_err(err, "Failed to create bitmap %s: ", name); 2202 return -1; 2203 } 2204 2205 do_dirty_bitmap_merge(dst->node_name, name, src->node_name, name, 2206 &err); 2207 if (err) { 2208 error_reportf_err(err, "Failed to populate bitmap %s: ", name); 2209 qmp_block_dirty_bitmap_remove(dst->node_name, name, NULL); 2210 return -1; 2211 } 2212 } 2213 2214 return 0; 2215 } 2216 2217 #define MAX_BUF_SECTORS 32768 2218 2219 static void set_rate_limit(BlockBackend *blk, int64_t rate_limit) 2220 { 2221 ThrottleConfig cfg; 2222 2223 throttle_config_init(&cfg); 2224 cfg.buckets[THROTTLE_BPS_WRITE].avg = rate_limit; 2225 2226 blk_io_limits_enable(blk, CONVERT_THROTTLE_GROUP); 2227 blk_set_io_limits(blk, &cfg); 2228 } 2229 2230 static int img_convert(const img_cmd_t *ccmd, int argc, char **argv) 2231 { 2232 int c, bs_i, flags, src_flags = BDRV_O_NO_SHARE; 2233 const char *fmt = NULL, *out_fmt = NULL, *cache = "unsafe", 2234 *src_cache = BDRV_DEFAULT_CACHE, *out_baseimg = NULL, 2235 *out_filename, *out_baseimg_param, *snapshot_name = NULL, 2236 *backing_fmt = NULL; 2237 BlockDriver *drv = NULL, *proto_drv = NULL; 2238 BlockDriverInfo bdi; 2239 BlockDriverState *out_bs; 2240 QemuOpts *opts = NULL, *sn_opts = NULL; 2241 QemuOptsList *create_opts = NULL; 2242 QDict *open_opts = NULL; 2243 char *options = NULL; 2244 Error *local_err = NULL; 2245 bool writethrough, src_writethrough, image_opts = false, 2246 skip_create = false, progress = false, tgt_image_opts = false; 2247 int64_t ret = -EINVAL; 2248 bool force_share = false; 2249 bool explict_min_sparse = false; 2250 bool bitmaps = false; 2251 bool skip_broken = false; 2252 int64_t rate_limit = 0; 2253 2254 ImgConvertState s = (ImgConvertState) { 2255 /* Need at least 4k of zeros for sparse detection */ 2256 .min_sparse = 8, 2257 .copy_range = false, 2258 .buf_sectors = IO_BUF_SIZE / BDRV_SECTOR_SIZE, 2259 .wr_in_order = true, 2260 .num_coroutines = 8, 2261 }; 2262 2263 for(;;) { 2264 static const struct option long_options[] = { 2265 {"help", no_argument, 0, 'h'}, 2266 {"source-format", required_argument, 0, 'f'}, 2267 /* 2268 * XXX: historic --image-opts acts on source file only, 2269 * it seems better to have it affect both source and target, 2270 * and have separate --source-image-opts for source, 2271 * but this might break existing setups. 2272 */ 2273 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS}, 2274 {"source-cache", required_argument, 0, 'T'}, 2275 {"snapshot", required_argument, 0, 'l'}, 2276 {"bitmaps", no_argument, 0, OPTION_BITMAPS}, 2277 {"skip-broken-bitmaps", no_argument, 0, OPTION_SKIP_BROKEN}, 2278 {"salvage", no_argument, 0, OPTION_SALVAGE}, 2279 {"target-format", required_argument, 0, 'O'}, 2280 {"target-image-opts", no_argument, 0, OPTION_TARGET_IMAGE_OPTS}, 2281 {"target-format-options", required_argument, 0, 'o'}, 2282 {"target-cache", required_argument, 0, 't'}, 2283 {"backing", required_argument, 0, 'b'}, 2284 {"backing-format", required_argument, 0, 'F'}, 2285 {"sparse-size", required_argument, 0, 'S'}, 2286 {"no-create", no_argument, 0, 'n'}, 2287 {"target-is-zero", no_argument, 0, OPTION_TARGET_IS_ZERO}, 2288 {"force-share", no_argument, 0, 'U'}, 2289 {"rate-limit", required_argument, 0, 'r'}, 2290 {"parallel", required_argument, 0, 'm'}, 2291 {"oob-writes", no_argument, 0, 'W'}, 2292 {"copy-range-offloading", no_argument, 0, 'C'}, 2293 {"progress", no_argument, 0, 'p'}, 2294 {"quiet", no_argument, 0, 'q'}, 2295 {"object", required_argument, 0, OPTION_OBJECT}, 2296 {0, 0, 0, 0} 2297 }; 2298 c = getopt_long(argc, argv, "hf:O:b:B:CcF:o:l:S:pt:T:nm:WUr:q", 2299 long_options, NULL); 2300 if (c == -1) { 2301 break; 2302 } 2303 switch (c) { 2304 case 'h': 2305 cmd_help(ccmd, "[-f SRC_FMT | --image-opts] [-T SRC_CACHE]\n" 2306 " [-l SNAPSHOT] [--bitmaps [--skip-broken-bitmaps]] [--salvage]\n" 2307 " [-O TGT_FMT | --target-image-opts] [-o TGT_FMT_OPTS] [-t TGT_CACHE]\n" 2308 " [-b BACKING_FILE [-F BACKING_FMT]] [-S SPARSE_SIZE]\n" 2309 " [-n] [--target-is-zero] [-c]\n" 2310 " [-U] [-r RATE] [-m NUM_PARALLEL] [-W] [-C] [-p] [-q] [--object OBJDEF]\n" 2311 " SRC_FILE [SRC_FILE2...] TGT_FILE\n" 2312 , 2313 " -f, --source-format SRC_FMT\n" 2314 " specify format of all SRC_FILEs explicitly (default: probing is used)\n" 2315 " --image-opts\n" 2316 " treat each SRC_FILE as an option string (key=value,...), not a file name\n" 2317 " (incompatible with -f|--source-format)\n" 2318 " -T, --source-cache SRC_CACHE\n" 2319 " source image(s) cache mode (" BDRV_DEFAULT_CACHE ")\n" 2320 " -l, --snapshot SNAPSHOT\n" 2321 " specify source snapshot\n" 2322 " --bitmaps\n" 2323 " also copy any persistent bitmaps present in source\n" 2324 " --skip-broken-bitmaps\n" 2325 " skip (do not error out) any broken bitmaps\n" 2326 " --salvage\n" 2327 " ignore errors on input (convert unreadable areas to zeros)\n" 2328 " -O, --target-format TGT_FMT\n" 2329 " specify TGT_FILE image format (default: raw)\n" 2330 " --target-image-opts\n" 2331 " treat TGT_FILE as an option string (key=value,...), not a file name\n" 2332 " (incompatible with -O|--target-format)\n" 2333 " -o, --target-format-options TGT_FMT_OPTS\n" 2334 " TGT_FMT-specific options\n" 2335 " -t, --target-cache TGT_CACHE\n" 2336 " cache mode when opening output image (default: unsafe)\n" 2337 " -b, --backing BACKING_FILE (was -B in <= 10.0)\n" 2338 " create target image to be a CoW on top of BACKING_FILE\n" 2339 " -F, --backing-format BACKING_FMT\n" /* -B used for -b in <=10.0 */ 2340 " specify BACKING_FILE image format explicitly (default: probing is used)\n" 2341 " -S, --sparse-size SPARSE_SIZE[bkKMGTPE]\n" 2342 " specify number of consecutive zero bytes to treat as a gap on output\n" 2343 " (rounded down to nearest 512 bytes), with optional multiplier suffix\n" 2344 " -n, --no-create\n" 2345 " omit target volume creation (e.g. on rbd)\n" 2346 " --target-is-zero\n" 2347 " indicates that the target volume is pre-zeroed\n" 2348 " -c, --compress\n" 2349 " create compressed output image (qcow and qcow2 formats only)\n" 2350 " -U, --force-share\n" 2351 " open images in shared mode for concurrent access\n" 2352 " -r, --rate-limit RATE\n" 2353 " I/O rate limit, in bytes per second\n" 2354 " -m, --parallel NUM_PARALLEL\n" 2355 " specify parallelism (default: 8)\n" 2356 " -C, --copy-range-offloading\n" 2357 " try to use copy offloading\n" 2358 " -W, --oob-writes\n" 2359 " enable out-of-order writes to improve performance\n" 2360 " -p, --progress\n" 2361 " display progress information\n" 2362 " -q, --quiet\n" 2363 " quiet mode (produce only error messages if any)\n" 2364 " --object OBJDEF\n" 2365 " defines QEMU user-creatable object\n" 2366 " SRC_FILE...\n" 2367 " one or more source image file names,\n" 2368 " or option strings (key=value,..) with --source-image-opts\n" 2369 " TGT_FILE\n" 2370 " target (output) image file name,\n" 2371 " or option string (key=value,..) with --target-image-opts\n" 2372 ); 2373 break; 2374 case 'f': 2375 fmt = optarg; 2376 break; 2377 case OPTION_IMAGE_OPTS: 2378 image_opts = true; 2379 break; 2380 case 'T': 2381 src_cache = optarg; 2382 break; 2383 case 'l': 2384 if (strstart(optarg, SNAPSHOT_OPT_BASE, NULL)) { 2385 sn_opts = qemu_opts_parse_noisily(&internal_snapshot_opts, 2386 optarg, false); 2387 if (!sn_opts) { 2388 error_report("Failed in parsing snapshot param '%s'", 2389 optarg); 2390 goto fail_getopt; 2391 } 2392 } else { 2393 snapshot_name = optarg; 2394 } 2395 break; 2396 case OPTION_BITMAPS: 2397 bitmaps = true; 2398 break; 2399 case OPTION_SKIP_BROKEN: 2400 skip_broken = true; 2401 break; 2402 case OPTION_SALVAGE: 2403 s.salvage = true; 2404 break; 2405 case 'O': 2406 out_fmt = optarg; 2407 break; 2408 case OPTION_TARGET_IMAGE_OPTS: 2409 tgt_image_opts = true; 2410 break; 2411 case 'o': 2412 if (accumulate_options(&options, optarg) < 0) { 2413 goto fail_getopt; 2414 } 2415 break; 2416 case 't': 2417 cache = optarg; 2418 break; 2419 case 'B': /* <=10.0 */ 2420 case 'b': 2421 out_baseimg = optarg; 2422 break; 2423 case 'F': /* can't use -B as it used as -b in <=10.0 */ 2424 backing_fmt = optarg; 2425 break; 2426 case 'S': 2427 { 2428 int64_t sval; 2429 2430 sval = cvtnum("buffer size for sparse output", optarg, true); 2431 if (sval < 0) { 2432 goto fail_getopt; 2433 } else if (!QEMU_IS_ALIGNED(sval, BDRV_SECTOR_SIZE) || 2434 sval / BDRV_SECTOR_SIZE > MAX_BUF_SECTORS) { 2435 error_report("Invalid buffer size for sparse output specified. " 2436 "Valid sizes are multiples of %llu up to %llu. Select " 2437 "0 to disable sparse detection (fully allocates output).", 2438 BDRV_SECTOR_SIZE, MAX_BUF_SECTORS * BDRV_SECTOR_SIZE); 2439 goto fail_getopt; 2440 } 2441 2442 s.min_sparse = sval / BDRV_SECTOR_SIZE; 2443 explict_min_sparse = true; 2444 break; 2445 } 2446 case 'n': 2447 skip_create = true; 2448 break; 2449 case OPTION_TARGET_IS_ZERO: 2450 /* 2451 * The user asserting that the target is blank has the 2452 * same effect as the target driver supporting zero 2453 * initialisation. 2454 */ 2455 s.has_zero_init = true; 2456 break; 2457 case 'c': 2458 s.compressed = true; 2459 break; 2460 case 'U': 2461 force_share = true; 2462 break; 2463 case 'r': 2464 rate_limit = cvtnum("rate limit", optarg, true); 2465 if (rate_limit < 0) { 2466 goto fail_getopt; 2467 } 2468 break; 2469 case 'm': 2470 s.num_coroutines = cvtnum_full("number of coroutines", optarg, 2471 false, 1, MAX_COROUTINES); 2472 if (s.num_coroutines < 0) { 2473 goto fail_getopt; 2474 } 2475 break; 2476 case 'W': 2477 s.wr_in_order = false; 2478 break; 2479 case 'C': 2480 s.copy_range = true; 2481 break; 2482 case 'p': 2483 progress = true; 2484 break; 2485 case 'q': 2486 s.quiet = true; 2487 break; 2488 case OPTION_OBJECT: 2489 user_creatable_process_cmdline(optarg); 2490 break; 2491 default: 2492 tryhelp(argv[0]); 2493 } 2494 } 2495 2496 if (!out_fmt && !tgt_image_opts) { 2497 out_fmt = "raw"; 2498 } 2499 2500 if (skip_broken && !bitmaps) { 2501 error_report("Use of --skip-broken-bitmaps requires --bitmaps"); 2502 goto fail_getopt; 2503 } 2504 2505 if (s.compressed && s.copy_range) { 2506 error_report("Cannot enable copy offloading when -c is used"); 2507 goto fail_getopt; 2508 } 2509 2510 if (explict_min_sparse && s.copy_range) { 2511 error_report("Cannot enable copy offloading when -S is used"); 2512 goto fail_getopt; 2513 } 2514 2515 if (s.copy_range && s.salvage) { 2516 error_report("Cannot use copy offloading in salvaging mode"); 2517 goto fail_getopt; 2518 } 2519 2520 if (tgt_image_opts && !skip_create) { 2521 error_report("--target-image-opts requires use of -n flag"); 2522 goto fail_getopt; 2523 } 2524 2525 if (skip_create && options) { 2526 error_report("-o has no effect when skipping image creation"); 2527 goto fail_getopt; 2528 } 2529 2530 if (s.has_zero_init && !skip_create) { 2531 error_report("--target-is-zero requires use of -n flag"); 2532 goto fail_getopt; 2533 } 2534 2535 s.src_num = argc - optind - 1; 2536 out_filename = s.src_num >= 1 ? argv[argc - 1] : NULL; 2537 2538 if (options && has_help_option(options)) { 2539 if (out_fmt) { 2540 ret = print_block_option_help(out_filename, out_fmt); 2541 goto fail_getopt; 2542 } else { 2543 error_report("Option help requires a format be specified"); 2544 goto fail_getopt; 2545 } 2546 } 2547 2548 if (s.src_num < 1) { 2549 error_report("Must specify image file name"); 2550 goto fail_getopt; 2551 } 2552 2553 /* ret is still -EINVAL until here */ 2554 ret = bdrv_parse_cache_mode(src_cache, &src_flags, &src_writethrough); 2555 if (ret < 0) { 2556 error_report("Invalid source cache option: %s", src_cache); 2557 goto fail_getopt; 2558 } 2559 2560 /* Initialize before goto out */ 2561 if (s.quiet) { 2562 progress = false; 2563 } 2564 qemu_progress_init(progress, 1.0); 2565 qemu_progress_print(0, 100); 2566 2567 s.src = g_new0(BlockBackend *, s.src_num); 2568 s.src_sectors = g_new(int64_t, s.src_num); 2569 s.src_alignment = g_new(int, s.src_num); 2570 2571 for (bs_i = 0; bs_i < s.src_num; bs_i++) { 2572 BlockDriverState *src_bs; 2573 s.src[bs_i] = img_open(image_opts, argv[optind + bs_i], 2574 fmt, src_flags, src_writethrough, s.quiet, 2575 force_share); 2576 if (!s.src[bs_i]) { 2577 ret = -1; 2578 goto out; 2579 } 2580 s.src_sectors[bs_i] = blk_nb_sectors(s.src[bs_i]); 2581 if (s.src_sectors[bs_i] < 0) { 2582 error_report("Could not get size of %s: %s", 2583 argv[optind + bs_i], strerror(-s.src_sectors[bs_i])); 2584 ret = -1; 2585 goto out; 2586 } 2587 src_bs = blk_bs(s.src[bs_i]); 2588 s.src_alignment[bs_i] = DIV_ROUND_UP(src_bs->bl.request_alignment, 2589 BDRV_SECTOR_SIZE); 2590 if (!bdrv_get_info(src_bs, &bdi)) { 2591 s.src_alignment[bs_i] = MAX(s.src_alignment[bs_i], 2592 bdi.cluster_size / BDRV_SECTOR_SIZE); 2593 } 2594 s.total_sectors += s.src_sectors[bs_i]; 2595 } 2596 2597 if (sn_opts) { 2598 bdrv_snapshot_load_tmp(blk_bs(s.src[0]), 2599 qemu_opt_get(sn_opts, SNAPSHOT_OPT_ID), 2600 qemu_opt_get(sn_opts, SNAPSHOT_OPT_NAME), 2601 &local_err); 2602 } else if (snapshot_name != NULL) { 2603 if (s.src_num > 1) { 2604 error_report("No support for concatenating multiple snapshot"); 2605 ret = -1; 2606 goto out; 2607 } 2608 2609 bdrv_snapshot_load_tmp_by_id_or_name(blk_bs(s.src[0]), snapshot_name, 2610 &local_err); 2611 } 2612 if (local_err) { 2613 error_reportf_err(local_err, "Failed to load snapshot: "); 2614 ret = -1; 2615 goto out; 2616 } 2617 2618 if (!skip_create) { 2619 /* Find driver and parse its options */ 2620 drv = bdrv_find_format(out_fmt); 2621 if (!drv) { 2622 error_report("Unknown file format '%s'", out_fmt); 2623 ret = -1; 2624 goto out; 2625 } 2626 2627 proto_drv = bdrv_find_protocol(out_filename, true, &local_err); 2628 if (!proto_drv) { 2629 error_report_err(local_err); 2630 ret = -1; 2631 goto out; 2632 } 2633 2634 if (!drv->create_opts) { 2635 error_report("Format driver '%s' does not support image creation", 2636 drv->format_name); 2637 ret = -1; 2638 goto out; 2639 } 2640 2641 if (!proto_drv->create_opts) { 2642 error_report("Protocol driver '%s' does not support image creation", 2643 proto_drv->format_name); 2644 ret = -1; 2645 goto out; 2646 } 2647 2648 create_opts = qemu_opts_append(create_opts, drv->create_opts); 2649 create_opts = qemu_opts_append(create_opts, proto_drv->create_opts); 2650 2651 opts = qemu_opts_create(create_opts, NULL, 0, &error_abort); 2652 if (options) { 2653 if (!qemu_opts_do_parse(opts, options, NULL, &local_err)) { 2654 error_report_err(local_err); 2655 ret = -1; 2656 goto out; 2657 } 2658 } 2659 2660 qemu_opt_set_number(opts, BLOCK_OPT_SIZE, 2661 s.total_sectors * BDRV_SECTOR_SIZE, &error_abort); 2662 ret = add_old_style_options(out_fmt, opts, out_baseimg, backing_fmt); 2663 if (ret < 0) { 2664 goto out; 2665 } 2666 } 2667 2668 /* Get backing file name if -o backing_file was used */ 2669 out_baseimg_param = qemu_opt_get(opts, BLOCK_OPT_BACKING_FILE); 2670 if (out_baseimg_param) { 2671 out_baseimg = out_baseimg_param; 2672 } 2673 s.target_has_backing = (bool) out_baseimg; 2674 2675 if (s.has_zero_init && s.target_has_backing) { 2676 error_report("Cannot use --target-is-zero when the destination " 2677 "image has a backing file"); 2678 goto out; 2679 } 2680 2681 if (s.src_num > 1 && out_baseimg) { 2682 error_report("Having a backing file for the target makes no sense when " 2683 "concatenating multiple input images"); 2684 ret = -1; 2685 goto out; 2686 } 2687 2688 if (out_baseimg_param) { 2689 if (!qemu_opt_get(opts, BLOCK_OPT_BACKING_FMT)) { 2690 error_report("Use of backing file requires explicit " 2691 "backing format"); 2692 ret = -1; 2693 goto out; 2694 } 2695 } 2696 2697 /* Check if compression is supported */ 2698 if (s.compressed) { 2699 bool encryption = 2700 qemu_opt_get_bool(opts, BLOCK_OPT_ENCRYPT, false); 2701 const char *encryptfmt = 2702 qemu_opt_get(opts, BLOCK_OPT_ENCRYPT_FORMAT); 2703 const char *preallocation = 2704 qemu_opt_get(opts, BLOCK_OPT_PREALLOC); 2705 2706 if (drv && !block_driver_can_compress(drv)) { 2707 error_report("Compression not supported for this file format"); 2708 ret = -1; 2709 goto out; 2710 } 2711 2712 if (encryption || encryptfmt) { 2713 error_report("Compression and encryption not supported at " 2714 "the same time"); 2715 ret = -1; 2716 goto out; 2717 } 2718 2719 if (preallocation 2720 && strcmp(preallocation, "off")) 2721 { 2722 error_report("Compression and preallocation not supported at " 2723 "the same time"); 2724 ret = -1; 2725 goto out; 2726 } 2727 } 2728 2729 /* Determine if bitmaps need copying */ 2730 if (bitmaps) { 2731 if (s.src_num > 1) { 2732 error_report("Copying bitmaps only possible with single source"); 2733 ret = -1; 2734 goto out; 2735 } 2736 ret = convert_check_bitmaps(blk_bs(s.src[0]), skip_broken); 2737 if (ret < 0) { 2738 goto out; 2739 } 2740 } 2741 2742 /* 2743 * The later open call will need any decryption secrets, and 2744 * bdrv_create() will purge "opts", so extract them now before 2745 * they are lost. 2746 */ 2747 if (!skip_create) { 2748 open_opts = qdict_new(); 2749 qemu_opt_foreach(opts, img_add_key_secrets, open_opts, &error_abort); 2750 2751 /* Create the new image */ 2752 ret = bdrv_create(drv, out_filename, opts, &local_err); 2753 if (ret < 0) { 2754 error_reportf_err(local_err, "%s: error while converting %s: ", 2755 out_filename, out_fmt); 2756 goto out; 2757 } 2758 } 2759 2760 s.target_is_new = !skip_create; 2761 2762 flags = s.min_sparse ? (BDRV_O_RDWR | BDRV_O_UNMAP) : BDRV_O_RDWR; 2763 ret = bdrv_parse_cache_mode(cache, &flags, &writethrough); 2764 if (ret < 0) { 2765 error_report("Invalid cache option: %s", cache); 2766 goto out; 2767 } 2768 2769 if (flags & BDRV_O_NOCACHE) { 2770 /* 2771 * If we open the target with O_DIRECT, it may be necessary to 2772 * extend its size to align to the physical sector size. 2773 */ 2774 flags |= BDRV_O_RESIZE; 2775 } 2776 2777 if (skip_create) { 2778 s.target = img_open(tgt_image_opts, out_filename, out_fmt, 2779 flags, writethrough, s.quiet, false); 2780 } else { 2781 /* TODO ultimately we should allow --target-image-opts 2782 * to be used even when -n is not given. 2783 * That has to wait for bdrv_create to be improved 2784 * to allow filenames in option syntax 2785 */ 2786 s.target = img_open_file(out_filename, open_opts, out_fmt, 2787 flags, writethrough, s.quiet, false); 2788 open_opts = NULL; /* blk_new_open will have freed it */ 2789 } 2790 if (!s.target) { 2791 ret = -1; 2792 goto out; 2793 } 2794 out_bs = blk_bs(s.target); 2795 2796 if (bitmaps && !bdrv_supports_persistent_dirty_bitmap(out_bs)) { 2797 error_report("Format driver '%s' does not support bitmaps", 2798 out_bs->drv->format_name); 2799 ret = -1; 2800 goto out; 2801 } 2802 2803 if (s.compressed && !block_driver_can_compress(out_bs->drv)) { 2804 error_report("Compression not supported for this file format"); 2805 ret = -1; 2806 goto out; 2807 } 2808 2809 /* increase bufsectors from the default 4096 (2M) if opt_transfer 2810 * or discard_alignment of the out_bs is greater. Limit to 2811 * MAX_BUF_SECTORS as maximum which is currently 32768 (16MB). */ 2812 s.buf_sectors = MIN(MAX_BUF_SECTORS, 2813 MAX(s.buf_sectors, 2814 MAX(out_bs->bl.opt_transfer >> BDRV_SECTOR_BITS, 2815 out_bs->bl.pdiscard_alignment >> 2816 BDRV_SECTOR_BITS))); 2817 2818 /* try to align the write requests to the destination to avoid unnecessary 2819 * RMW cycles. */ 2820 s.alignment = MAX(pow2floor(s.min_sparse), 2821 DIV_ROUND_UP(out_bs->bl.request_alignment, 2822 BDRV_SECTOR_SIZE)); 2823 assert(is_power_of_2(s.alignment)); 2824 2825 if (skip_create) { 2826 int64_t output_sectors = blk_nb_sectors(s.target); 2827 if (output_sectors < 0) { 2828 error_report("unable to get output image length: %s", 2829 strerror(-output_sectors)); 2830 ret = -1; 2831 goto out; 2832 } else if (output_sectors < s.total_sectors) { 2833 error_report("output file is smaller than input file"); 2834 ret = -1; 2835 goto out; 2836 } 2837 } 2838 2839 if (s.target_has_backing && s.target_is_new) { 2840 /* Errors are treated as "backing length unknown" (which means 2841 * s.target_backing_sectors has to be negative, which it will 2842 * be automatically). The backing file length is used only 2843 * for optimizations, so such a case is not fatal. */ 2844 bdrv_graph_rdlock_main_loop(); 2845 s.target_backing_sectors = 2846 bdrv_nb_sectors(bdrv_backing_chain_next(out_bs)); 2847 bdrv_graph_rdunlock_main_loop(); 2848 } else { 2849 s.target_backing_sectors = -1; 2850 } 2851 2852 ret = bdrv_get_info(out_bs, &bdi); 2853 if (ret < 0) { 2854 if (s.compressed) { 2855 error_report("could not get block driver info"); 2856 goto out; 2857 } 2858 } else { 2859 s.compressed = s.compressed || bdi.needs_compressed_writes; 2860 s.cluster_sectors = bdi.cluster_size / BDRV_SECTOR_SIZE; 2861 } 2862 2863 if (rate_limit) { 2864 set_rate_limit(s.target, rate_limit); 2865 } 2866 2867 ret = convert_do_copy(&s); 2868 2869 /* Now copy the bitmaps */ 2870 if (bitmaps && ret == 0) { 2871 ret = convert_copy_bitmaps(blk_bs(s.src[0]), out_bs, skip_broken); 2872 } 2873 2874 out: 2875 if (!ret) { 2876 qemu_progress_print(100, 0); 2877 } 2878 qemu_progress_end(); 2879 qemu_opts_del(opts); 2880 qemu_opts_free(create_opts); 2881 qobject_unref(open_opts); 2882 blk_unref(s.target); 2883 if (s.src) { 2884 for (bs_i = 0; bs_i < s.src_num; bs_i++) { 2885 blk_unref(s.src[bs_i]); 2886 } 2887 g_free(s.src); 2888 } 2889 g_free(s.src_sectors); 2890 g_free(s.src_alignment); 2891 fail_getopt: 2892 qemu_opts_del(sn_opts); 2893 g_free(options); 2894 2895 return !!ret; 2896 } 2897 2898 2899 static void dump_snapshots(BlockDriverState *bs) 2900 { 2901 QEMUSnapshotInfo *sn_tab, *sn; 2902 int nb_sns, i; 2903 2904 nb_sns = bdrv_snapshot_list(bs, &sn_tab); 2905 if (nb_sns <= 0) 2906 return; 2907 printf("Snapshot list:\n"); 2908 bdrv_snapshot_dump(NULL); 2909 printf("\n"); 2910 for(i = 0; i < nb_sns; i++) { 2911 sn = &sn_tab[i]; 2912 bdrv_snapshot_dump(sn); 2913 printf("\n"); 2914 } 2915 g_free(sn_tab); 2916 } 2917 2918 static void dump_json_block_graph_info_list(BlockGraphInfoList *list) 2919 { 2920 GString *str; 2921 QObject *obj; 2922 Visitor *v = qobject_output_visitor_new(&obj); 2923 2924 visit_type_BlockGraphInfoList(v, NULL, &list, &error_abort); 2925 visit_complete(v, &obj); 2926 str = qobject_to_json_pretty(obj, true); 2927 assert(str != NULL); 2928 printf("%s\n", str->str); 2929 qobject_unref(obj); 2930 visit_free(v); 2931 g_string_free(str, true); 2932 } 2933 2934 static void dump_json_block_graph_info(BlockGraphInfo *info) 2935 { 2936 GString *str; 2937 QObject *obj; 2938 Visitor *v = qobject_output_visitor_new(&obj); 2939 2940 visit_type_BlockGraphInfo(v, NULL, &info, &error_abort); 2941 visit_complete(v, &obj); 2942 str = qobject_to_json_pretty(obj, true); 2943 assert(str != NULL); 2944 printf("%s\n", str->str); 2945 qobject_unref(obj); 2946 visit_free(v); 2947 g_string_free(str, true); 2948 } 2949 2950 static void dump_human_image_info(BlockGraphInfo *info, int indentation, 2951 const char *path) 2952 { 2953 BlockChildInfoList *children_list; 2954 2955 bdrv_node_info_dump(qapi_BlockGraphInfo_base(info), indentation, 2956 info->children == NULL); 2957 2958 for (children_list = info->children; children_list; 2959 children_list = children_list->next) 2960 { 2961 BlockChildInfo *child = children_list->value; 2962 g_autofree char *child_path = NULL; 2963 2964 printf("%*sChild node '%s%s':\n", 2965 indentation * 4, "", path, child->name); 2966 child_path = g_strdup_printf("%s%s/", path, child->name); 2967 dump_human_image_info(child->info, indentation + 1, child_path); 2968 } 2969 } 2970 2971 static void dump_human_image_info_list(BlockGraphInfoList *list) 2972 { 2973 BlockGraphInfoList *elem; 2974 bool delim = false; 2975 2976 for (elem = list; elem; elem = elem->next) { 2977 if (delim) { 2978 printf("\n"); 2979 } 2980 delim = true; 2981 2982 dump_human_image_info(elem->value, 0, "/"); 2983 } 2984 } 2985 2986 static gboolean str_equal_func(gconstpointer a, gconstpointer b) 2987 { 2988 return strcmp(a, b) == 0; 2989 } 2990 2991 /** 2992 * Open an image file chain and return an BlockGraphInfoList 2993 * 2994 * @filename: topmost image filename 2995 * @fmt: topmost image format (may be NULL to autodetect) 2996 * @chain: true - enumerate entire backing file chain 2997 * false - only topmost image file 2998 * 2999 * Returns a list of BlockNodeInfo objects or NULL if there was an error 3000 * opening an image file. If there was an error a message will have been 3001 * printed to stderr. 3002 */ 3003 static BlockGraphInfoList *collect_image_info_list(bool image_opts, 3004 const char *filename, 3005 const char *fmt, 3006 const char *cache, 3007 bool chain, bool limits, 3008 bool force_share) 3009 { 3010 BlockGraphInfoList *head = NULL; 3011 BlockGraphInfoList **tail = &head; 3012 GHashTable *filenames; 3013 Error *err = NULL; 3014 int cache_flags = 0; 3015 bool writethrough = false; 3016 int ret; 3017 3018 ret = bdrv_parse_cache_mode(cache, &cache_flags, &writethrough); 3019 if (ret < 0) { 3020 error_report("Invalid cache option: %s", cache); 3021 return NULL; 3022 } 3023 3024 filenames = g_hash_table_new_full(g_str_hash, str_equal_func, NULL, NULL); 3025 3026 while (filename) { 3027 BlockBackend *blk; 3028 BlockDriverState *bs; 3029 BlockGraphInfo *info; 3030 3031 if (g_hash_table_lookup_extended(filenames, filename, NULL, NULL)) { 3032 error_report("Backing file '%s' creates an infinite loop.", 3033 filename); 3034 goto err; 3035 } 3036 g_hash_table_insert(filenames, (gpointer)filename, NULL); 3037 3038 blk = img_open(image_opts, filename, fmt, 3039 BDRV_O_NO_BACKING | BDRV_O_NO_IO | cache_flags, 3040 writethrough, false, force_share); 3041 if (!blk) { 3042 goto err; 3043 } 3044 bs = blk_bs(blk); 3045 3046 /* 3047 * Note that the returned BlockGraphInfo object will not have 3048 * information about this image's backing node, because we have opened 3049 * it with BDRV_O_NO_BACKING. Printing this object will therefore not 3050 * duplicate the backing chain information that we obtain by walking 3051 * the chain manually here. 3052 */ 3053 bdrv_graph_rdlock_main_loop(); 3054 bdrv_query_block_graph_info(bs, &info, limits, &err); 3055 bdrv_graph_rdunlock_main_loop(); 3056 3057 if (err) { 3058 error_report_err(err); 3059 blk_unref(blk); 3060 goto err; 3061 } 3062 3063 QAPI_LIST_APPEND(tail, info); 3064 3065 blk_unref(blk); 3066 3067 /* Clear parameters that only apply to the topmost image */ 3068 filename = fmt = NULL; 3069 image_opts = false; 3070 3071 if (chain) { 3072 if (info->full_backing_filename) { 3073 filename = info->full_backing_filename; 3074 } else if (info->backing_filename) { 3075 error_report("Could not determine absolute backing filename," 3076 " but backing filename '%s' present", 3077 info->backing_filename); 3078 goto err; 3079 } 3080 if (info->backing_filename_format) { 3081 fmt = info->backing_filename_format; 3082 } 3083 } 3084 } 3085 g_hash_table_destroy(filenames); 3086 return head; 3087 3088 err: 3089 qapi_free_BlockGraphInfoList(head); 3090 g_hash_table_destroy(filenames); 3091 return NULL; 3092 } 3093 3094 static int img_info(const img_cmd_t *ccmd, int argc, char **argv) 3095 { 3096 int c; 3097 OutputFormat output_format = OFORMAT_HUMAN; 3098 bool chain = false; 3099 const char *filename, *fmt; 3100 const char *cache = BDRV_DEFAULT_CACHE; 3101 BlockGraphInfoList *list; 3102 bool image_opts = false; 3103 bool force_share = false; 3104 bool limits = false; 3105 3106 fmt = NULL; 3107 for(;;) { 3108 static const struct option long_options[] = { 3109 {"help", no_argument, 0, 'h'}, 3110 {"format", required_argument, 0, 'f'}, 3111 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS}, 3112 {"backing-chain", no_argument, 0, OPTION_BACKING_CHAIN}, 3113 {"cache", required_argument, 0, 't'}, 3114 {"force-share", no_argument, 0, 'U'}, 3115 {"limits", no_argument, 0, OPTION_LIMITS}, 3116 {"output", required_argument, 0, OPTION_OUTPUT}, 3117 {"object", required_argument, 0, OPTION_OBJECT}, 3118 {0, 0, 0, 0} 3119 }; 3120 c = getopt_long(argc, argv, "hf:t:U", long_options, NULL); 3121 if (c == -1) { 3122 break; 3123 } 3124 switch(c) { 3125 case 'h': 3126 cmd_help(ccmd, "[-f FMT | --image-opts] [--backing-chain] [-U]\n" 3127 " [--output human|json] [--object OBJDEF] FILE\n" 3128 , 3129 " -f, --format FMT\n" 3130 " specify FILE image format explicitly (default: probing is used)\n" 3131 " --image-opts\n" 3132 " treat FILE as an option string (key=value,..), not a file name\n" 3133 " (incompatible with -f|--format)\n" 3134 " --backing-chain\n" 3135 " display information about the backing chain for copy-on-write overlays\n" 3136 " -t, --cache CACHE\n" 3137 " cache mode for FILE (default: " BDRV_DEFAULT_CACHE ")\n" 3138 " -U, --force-share\n" 3139 " open image in shared mode for concurrent access\n" 3140 " --limits\n" 3141 " show detected block limits (may depend on options, e.g. cache mode)\n" 3142 " --output human|json\n" 3143 " specify output format (default: human)\n" 3144 " --object OBJDEF\n" 3145 " defines QEMU user-creatable object\n" 3146 " FILE\n" 3147 " name of the image file, or option string (key=value,..)\n" 3148 " with --image-opts, to operate on\n" 3149 ); 3150 break; 3151 case 'f': 3152 fmt = optarg; 3153 break; 3154 case OPTION_IMAGE_OPTS: 3155 image_opts = true; 3156 break; 3157 case OPTION_BACKING_CHAIN: 3158 chain = true; 3159 break; 3160 case 't': 3161 cache = optarg; 3162 break; 3163 case 'U': 3164 force_share = true; 3165 break; 3166 case OPTION_LIMITS: 3167 limits = true; 3168 break; 3169 case OPTION_OUTPUT: 3170 output_format = parse_output_format(argv[0], optarg); 3171 break; 3172 case OPTION_OBJECT: 3173 user_creatable_process_cmdline(optarg); 3174 break; 3175 default: 3176 tryhelp(argv[0]); 3177 } 3178 } 3179 if (optind != argc - 1) { 3180 error_exit(argv[0], "Expecting one image file name"); 3181 } 3182 filename = argv[optind++]; 3183 3184 list = collect_image_info_list(image_opts, filename, fmt, cache, chain, 3185 limits, force_share); 3186 if (!list) { 3187 return 1; 3188 } 3189 3190 switch (output_format) { 3191 case OFORMAT_HUMAN: 3192 dump_human_image_info_list(list); 3193 break; 3194 case OFORMAT_JSON: 3195 if (chain) { 3196 dump_json_block_graph_info_list(list); 3197 } else { 3198 dump_json_block_graph_info(list->value); 3199 } 3200 break; 3201 } 3202 3203 qapi_free_BlockGraphInfoList(list); 3204 return 0; 3205 } 3206 3207 static int dump_map_entry(OutputFormat output_format, MapEntry *e, 3208 MapEntry *next) 3209 { 3210 switch (output_format) { 3211 case OFORMAT_HUMAN: 3212 if (e->data && !e->has_offset) { 3213 error_report("File contains external, encrypted or compressed clusters."); 3214 return -1; 3215 } 3216 if (e->data && !e->zero) { 3217 printf("%#-16"PRIx64"%#-16"PRIx64"%#-16"PRIx64"%s\n", 3218 e->start, e->length, 3219 e->has_offset ? e->offset : 0, 3220 e->filename ?: ""); 3221 } 3222 /* This format ignores the distinction between 0, ZERO and ZERO|DATA. 3223 * Modify the flags here to allow more coalescing. 3224 */ 3225 if (next && (!next->data || next->zero)) { 3226 next->data = false; 3227 next->zero = true; 3228 } 3229 break; 3230 case OFORMAT_JSON: 3231 printf("{ \"start\": %"PRId64", \"length\": %"PRId64"," 3232 " \"depth\": %"PRId64", \"present\": %s, \"zero\": %s," 3233 " \"data\": %s, \"compressed\": %s", 3234 e->start, e->length, e->depth, 3235 e->present ? "true" : "false", 3236 e->zero ? "true" : "false", 3237 e->data ? "true" : "false", 3238 e->compressed ? "true" : "false"); 3239 if (e->has_offset) { 3240 printf(", \"offset\": %"PRId64"", e->offset); 3241 } 3242 putchar('}'); 3243 3244 if (next) { 3245 puts(","); 3246 } 3247 break; 3248 } 3249 return 0; 3250 } 3251 3252 static int get_block_status(BlockDriverState *bs, int64_t offset, 3253 int64_t bytes, MapEntry *e) 3254 { 3255 int ret; 3256 int depth; 3257 BlockDriverState *file; 3258 bool has_offset; 3259 int64_t map; 3260 char *filename = NULL; 3261 3262 GLOBAL_STATE_CODE(); 3263 GRAPH_RDLOCK_GUARD_MAINLOOP(); 3264 3265 /* As an optimization, we could cache the current range of unallocated 3266 * clusters in each file of the chain, and avoid querying the same 3267 * range repeatedly. 3268 */ 3269 3270 depth = 0; 3271 for (;;) { 3272 bs = bdrv_skip_filters(bs); 3273 ret = bdrv_block_status(bs, offset, bytes, &bytes, &map, &file); 3274 if (ret < 0) { 3275 return ret; 3276 } 3277 assert(bytes); 3278 if (ret & (BDRV_BLOCK_ZERO|BDRV_BLOCK_DATA)) { 3279 break; 3280 } 3281 bs = bdrv_cow_bs(bs); 3282 if (bs == NULL) { 3283 ret = 0; 3284 break; 3285 } 3286 3287 depth++; 3288 } 3289 3290 has_offset = !!(ret & BDRV_BLOCK_OFFSET_VALID); 3291 3292 if (file && has_offset) { 3293 bdrv_refresh_filename(file); 3294 filename = file->filename; 3295 } 3296 3297 *e = (MapEntry) { 3298 .start = offset, 3299 .length = bytes, 3300 .data = !!(ret & BDRV_BLOCK_DATA), 3301 .zero = !!(ret & BDRV_BLOCK_ZERO), 3302 .compressed = !!(ret & BDRV_BLOCK_COMPRESSED), 3303 .offset = map, 3304 .has_offset = has_offset, 3305 .depth = depth, 3306 .present = !!(ret & BDRV_BLOCK_ALLOCATED), 3307 .filename = filename, 3308 }; 3309 3310 return 0; 3311 } 3312 3313 static inline bool entry_mergeable(const MapEntry *curr, const MapEntry *next) 3314 { 3315 if (curr->length == 0) { 3316 return false; 3317 } 3318 if (curr->zero != next->zero || 3319 curr->data != next->data || 3320 curr->compressed != next->compressed || 3321 curr->depth != next->depth || 3322 curr->present != next->present || 3323 !curr->filename != !next->filename || 3324 curr->has_offset != next->has_offset) { 3325 return false; 3326 } 3327 if (curr->filename && strcmp(curr->filename, next->filename)) { 3328 return false; 3329 } 3330 if (curr->has_offset && curr->offset + curr->length != next->offset) { 3331 return false; 3332 } 3333 return true; 3334 } 3335 3336 static int img_map(const img_cmd_t *ccmd, int argc, char **argv) 3337 { 3338 int c; 3339 OutputFormat output_format = OFORMAT_HUMAN; 3340 BlockBackend *blk; 3341 BlockDriverState *bs; 3342 const char *filename, *fmt; 3343 int64_t length; 3344 MapEntry curr = { .length = 0 }, next; 3345 int ret = 0; 3346 bool image_opts = false; 3347 bool force_share = false; 3348 int64_t start_offset = 0; 3349 int64_t max_length = -1; 3350 3351 fmt = NULL; 3352 for (;;) { 3353 static const struct option long_options[] = { 3354 {"help", no_argument, 0, 'h'}, 3355 {"format", required_argument, 0, 'f'}, 3356 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS}, 3357 {"start-offset", required_argument, 0, 's'}, 3358 {"max-length", required_argument, 0, 'l'}, 3359 {"force-share", no_argument, 0, 'U'}, 3360 {"output", required_argument, 0, OPTION_OUTPUT}, 3361 {"object", required_argument, 0, OPTION_OBJECT}, 3362 {0, 0, 0, 0} 3363 }; 3364 c = getopt_long(argc, argv, "hf:s:l:U", 3365 long_options, NULL); 3366 if (c == -1) { 3367 break; 3368 } 3369 switch (c) { 3370 case 'h': 3371 cmd_help(ccmd, "[-f FMT | --image-opts]\n" 3372 " [--start-offset OFFSET] [--max-length LENGTH]\n" 3373 " [--output human|json] [-U] [--object OBJDEF] FILE\n" 3374 , 3375 " -f, --format FMT\n" 3376 " specify FILE image format explicitly (default: probing is used)\n" 3377 " --image-opts\n" 3378 " treat FILE as an option string (key=value,..), not a file name\n" 3379 " (incompatible with -f|--format)\n" 3380 " -s, --start-offset OFFSET\n" 3381 " start at the given OFFSET in the image, not at the beginning\n" 3382 " -l, --max-length LENGTH\n" 3383 " process at most LENGTH bytes instead of up to the end of the image\n" 3384 " --output human|json\n" 3385 " specify output format name (default: human)\n" 3386 " -U, --force-share\n" 3387 " open image in shared mode for concurrent access\n" 3388 " --object OBJDEF\n" 3389 " defines QEMU user-creatable object\n" 3390 " FILE\n" 3391 " the image file name, or option string (key=value,..)\n" 3392 " with --image-opts, to operate on\n" 3393 ); 3394 break; 3395 case 'f': 3396 fmt = optarg; 3397 break; 3398 case OPTION_IMAGE_OPTS: 3399 image_opts = true; 3400 break; 3401 case 's': 3402 start_offset = cvtnum("start offset", optarg, true); 3403 if (start_offset < 0) { 3404 return 1; 3405 } 3406 break; 3407 case 'l': 3408 max_length = cvtnum("max length", optarg, true); 3409 if (max_length < 0) { 3410 return 1; 3411 } 3412 break; 3413 case OPTION_OUTPUT: 3414 output_format = parse_output_format(argv[0], optarg); 3415 break; 3416 case 'U': 3417 force_share = true; 3418 break; 3419 case OPTION_OBJECT: 3420 user_creatable_process_cmdline(optarg); 3421 break; 3422 default: 3423 tryhelp(argv[0]); 3424 } 3425 } 3426 if (optind != argc - 1) { 3427 error_exit(argv[0], "Expecting one image file name"); 3428 } 3429 filename = argv[optind]; 3430 3431 blk = img_open(image_opts, filename, fmt, 0, false, false, force_share); 3432 if (!blk) { 3433 return 1; 3434 } 3435 bs = blk_bs(blk); 3436 3437 if (output_format == OFORMAT_HUMAN) { 3438 printf("%-16s%-16s%-16s%s\n", "Offset", "Length", "Mapped to", "File"); 3439 } else if (output_format == OFORMAT_JSON) { 3440 putchar('['); 3441 } 3442 3443 length = blk_getlength(blk); 3444 if (length < 0) { 3445 error_report("Failed to get size for '%s'", filename); 3446 return 1; 3447 } 3448 if (max_length != -1) { 3449 length = MIN(start_offset + max_length, length); 3450 } 3451 3452 curr.start = start_offset; 3453 while (curr.start + curr.length < length) { 3454 int64_t offset = curr.start + curr.length; 3455 int64_t n = length - offset; 3456 3457 ret = get_block_status(bs, offset, n, &next); 3458 if (ret < 0) { 3459 error_report("Could not read file metadata: %s", strerror(-ret)); 3460 goto out; 3461 } 3462 3463 if (entry_mergeable(&curr, &next)) { 3464 curr.length += next.length; 3465 continue; 3466 } 3467 3468 if (curr.length > 0) { 3469 ret = dump_map_entry(output_format, &curr, &next); 3470 if (ret < 0) { 3471 goto out; 3472 } 3473 } 3474 curr = next; 3475 } 3476 3477 ret = dump_map_entry(output_format, &curr, NULL); 3478 if (output_format == OFORMAT_JSON) { 3479 puts("]"); 3480 } 3481 3482 out: 3483 blk_unref(blk); 3484 return ret < 0; 3485 } 3486 3487 /* the same as options */ 3488 #define SNAPSHOT_LIST 'l' 3489 #define SNAPSHOT_CREATE 'c' 3490 #define SNAPSHOT_APPLY 'a' 3491 #define SNAPSHOT_DELETE 'd' 3492 3493 static int img_snapshot(const img_cmd_t *ccmd, int argc, char **argv) 3494 { 3495 BlockBackend *blk; 3496 BlockDriverState *bs; 3497 QEMUSnapshotInfo sn; 3498 char *filename, *fmt = NULL, *snapshot_name = NULL; 3499 int c, ret = 0; 3500 int action = 0; 3501 bool quiet = false; 3502 Error *err = NULL; 3503 bool image_opts = false; 3504 bool force_share = false; 3505 int64_t rt; 3506 3507 /* Parse commandline parameters */ 3508 for(;;) { 3509 static const struct option long_options[] = { 3510 {"help", no_argument, 0, 'h'}, 3511 {"format", required_argument, 0, 'f'}, 3512 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS}, 3513 {"list", no_argument, 0, SNAPSHOT_LIST}, 3514 {"apply", required_argument, 0, SNAPSHOT_APPLY}, 3515 {"create", required_argument, 0, SNAPSHOT_CREATE}, 3516 {"delete", required_argument, 0, SNAPSHOT_DELETE}, 3517 {"force-share", no_argument, 0, 'U'}, 3518 {"quiet", no_argument, 0, 'q'}, 3519 {"object", required_argument, 0, OPTION_OBJECT}, 3520 {0, 0, 0, 0} 3521 }; 3522 c = getopt_long(argc, argv, "hf:la:c:d:Uq", 3523 long_options, NULL); 3524 if (c == -1) { 3525 break; 3526 } 3527 switch(c) { 3528 case 'h': 3529 cmd_help(ccmd, "[-f FMT | --image-opts] [-l | -a|-c|-d SNAPSHOT]\n" 3530 " [-U] [-q] [--object OBJDEF] FILE\n" 3531 , 3532 " -f, --format FMT\n" 3533 " specify FILE format explicitly (default: probing is used)\n" 3534 " --image-opts\n" 3535 " treat FILE as an option string (key=value,..), not a file name\n" 3536 " (incompatible with -f|--format)\n" 3537 " -l, --list\n" 3538 " list snapshots in FILE (default action if no -l|-c|-a|-d is given)\n" 3539 " -c, --create SNAPSHOT\n" 3540 " create named snapshot\n" 3541 " -a, --apply SNAPSHOT\n" 3542 " apply named snapshot to the base\n" 3543 " -d, --delete SNAPSHOT\n" 3544 " delete named snapshot\n" 3545 " (only one of -l|-c|-a|-d can be specified)\n" 3546 " -U, --force-share\n" 3547 " open image in shared mode for concurrent access\n" 3548 " -q, --quiet\n" 3549 " quiet mode (produce only error messages if any)\n" 3550 " --object OBJDEF\n" 3551 " defines QEMU user-creatable object\n" 3552 " FILE\n" 3553 " name of the image file, or option string (key=value,..)\n" 3554 " with --image-opts) to operate on\n" 3555 ); 3556 break; 3557 case 'f': 3558 fmt = optarg; 3559 break; 3560 case OPTION_IMAGE_OPTS: 3561 image_opts = true; 3562 break; 3563 case SNAPSHOT_LIST: 3564 case SNAPSHOT_APPLY: 3565 case SNAPSHOT_CREATE: 3566 case SNAPSHOT_DELETE: 3567 if (action) { 3568 error_exit(argv[0], "Cannot mix '-l', '-a', '-c', '-d'"); 3569 return 0; 3570 } 3571 action = c; 3572 snapshot_name = optarg; 3573 break; 3574 case 'U': 3575 force_share = true; 3576 break; 3577 case 'q': 3578 quiet = true; 3579 break; 3580 case OPTION_OBJECT: 3581 user_creatable_process_cmdline(optarg); 3582 break; 3583 default: 3584 tryhelp(argv[0]); 3585 } 3586 } 3587 3588 if (optind != argc - 1) { 3589 error_exit(argv[0], "Expecting one image file name"); 3590 } 3591 filename = argv[optind++]; 3592 3593 if (!action) { 3594 action = SNAPSHOT_LIST; 3595 } 3596 3597 /* Open the image */ 3598 blk = img_open(image_opts, filename, fmt, 3599 action == SNAPSHOT_LIST ? 0 : BDRV_O_RDWR, 3600 false, quiet, force_share); 3601 if (!blk) { 3602 return 1; 3603 } 3604 bs = blk_bs(blk); 3605 3606 /* Perform the requested action */ 3607 switch(action) { 3608 case SNAPSHOT_LIST: 3609 dump_snapshots(bs); 3610 break; 3611 3612 case SNAPSHOT_CREATE: 3613 memset(&sn, 0, sizeof(sn)); 3614 pstrcpy(sn.name, sizeof(sn.name), snapshot_name); 3615 3616 rt = g_get_real_time(); 3617 sn.date_sec = rt / G_USEC_PER_SEC; 3618 sn.date_nsec = (rt % G_USEC_PER_SEC) * 1000; 3619 3620 bdrv_graph_rdlock_main_loop(); 3621 ret = bdrv_snapshot_create(bs, &sn); 3622 bdrv_graph_rdunlock_main_loop(); 3623 3624 if (ret) { 3625 error_report("Could not create snapshot '%s': %s", 3626 snapshot_name, strerror(-ret)); 3627 } 3628 break; 3629 3630 case SNAPSHOT_APPLY: 3631 ret = bdrv_snapshot_goto(bs, snapshot_name, &err); 3632 if (ret) { 3633 error_reportf_err(err, "Could not apply snapshot '%s': ", 3634 snapshot_name); 3635 } 3636 break; 3637 3638 case SNAPSHOT_DELETE: 3639 bdrv_drain_all_begin(); 3640 bdrv_graph_rdlock_main_loop(); 3641 ret = bdrv_snapshot_find(bs, &sn, snapshot_name); 3642 if (ret < 0) { 3643 error_report("Could not delete snapshot '%s': snapshot not " 3644 "found", snapshot_name); 3645 ret = 1; 3646 } else { 3647 ret = bdrv_snapshot_delete(bs, sn.id_str, sn.name, &err); 3648 if (ret < 0) { 3649 error_reportf_err(err, "Could not delete snapshot '%s': ", 3650 snapshot_name); 3651 ret = 1; 3652 } 3653 } 3654 bdrv_graph_rdunlock_main_loop(); 3655 bdrv_drain_all_end(); 3656 break; 3657 } 3658 3659 /* Cleanup */ 3660 blk_unref(blk); 3661 if (ret) { 3662 return 1; 3663 } 3664 return 0; 3665 } 3666 3667 static int img_rebase(const img_cmd_t *ccmd, int argc, char **argv) 3668 { 3669 BlockBackend *blk = NULL, *blk_old_backing = NULL, *blk_new_backing = NULL; 3670 uint8_t *buf_old = NULL; 3671 uint8_t *buf_new = NULL; 3672 BlockDriverState *bs = NULL, *prefix_chain_bs = NULL; 3673 BlockDriverState *unfiltered_bs, *unfiltered_bs_cow; 3674 BlockDriverInfo bdi = {0}; 3675 char *filename; 3676 const char *fmt, *cache, *src_cache, *out_basefmt, *out_baseimg; 3677 int c, flags, src_flags, ret; 3678 BdrvRequestFlags write_flags = 0; 3679 bool writethrough, src_writethrough; 3680 int unsafe = 0; 3681 bool force_share = false; 3682 int progress = 0; 3683 bool quiet = false; 3684 bool compress = false; 3685 Error *local_err = NULL; 3686 bool image_opts = false; 3687 int64_t write_align; 3688 3689 /* Parse commandline parameters */ 3690 fmt = NULL; 3691 cache = BDRV_DEFAULT_CACHE; 3692 src_cache = BDRV_DEFAULT_CACHE; 3693 out_baseimg = NULL; 3694 out_basefmt = NULL; 3695 for(;;) { 3696 static const struct option long_options[] = { 3697 {"help", no_argument, 0, 'h'}, 3698 {"format", required_argument, 0, 'f'}, 3699 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS}, 3700 {"cache", required_argument, 0, 't'}, 3701 {"compress", no_argument, 0, 'c'}, 3702 {"backing", required_argument, 0, 'b'}, 3703 {"backing-format", required_argument, 0, 'B'}, 3704 {"backing-cache", required_argument, 0, 'T'}, 3705 {"backing-unsafe", no_argument, 0, 'u'}, 3706 {"force-share", no_argument, 0, 'U'}, 3707 {"progress", no_argument, 0, 'p'}, 3708 {"quiet", no_argument, 0, 'q'}, 3709 {"object", required_argument, 0, OPTION_OBJECT}, 3710 {0, 0, 0, 0} 3711 }; 3712 c = getopt_long(argc, argv, "hf:t:cb:F:B:T:uUpq", 3713 long_options, NULL); 3714 if (c == -1) { 3715 break; 3716 } 3717 switch (c) { 3718 case 'h': 3719 cmd_help(ccmd, "[-f FMT | --image-opts] [-t CACHE]\n" 3720 " [-b BACKING_FILE [-B BACKING_FMT] [-T BACKING_CACHE]] [-u]\n" 3721 " [-c] [-U] [-p] [-q] [--object OBJDEF] FILE\n" 3722 , 3723 " -f, --format FMT\n" 3724 " specify FILE format explicitly (default: probing is used)\n" 3725 " --image-opts\n" 3726 " treat FILE as an option string (key=value,..), not a file name\n" 3727 " (incompatible with -f|--format)\n" 3728 " -t, --cache CACHE\n" 3729 " cache mode for FILE (default: " BDRV_DEFAULT_CACHE ")\n" 3730 " -b, --backing BACKING_FILE|\"\"\n" 3731 " rebase onto this file (specify empty name for no backing file)\n" 3732 " -B, --backing-format BACKING_FMT (was -F in <=10.0)\n" 3733 " specify format for BACKING_FILE explicitly (default: probing is used)\n" 3734 " -T, --backing-cache CACHE\n" 3735 " BACKING_FILE cache mode (default: " BDRV_DEFAULT_CACHE ")\n" 3736 " -u, --backing-unsafe\n" 3737 " do not fail if BACKING_FILE can not be read\n" 3738 " -c, --compress\n" 3739 " compress image (when image supports this)\n" 3740 " -U, --force-share\n" 3741 " open image in shared mode for concurrent access\n" 3742 " -p, --progress\n" 3743 " display progress information\n" 3744 " -q, --quiet\n" 3745 " quiet mode (produce only error messages if any)\n" 3746 " --object OBJDEF\n" 3747 " defines QEMU user-creatable object\n" 3748 " FILE\n" 3749 " name of the image file, or option string (key=value,..)\n" 3750 " with --image-opts, to operate on\n" 3751 ); 3752 return 0; 3753 case 'f': 3754 fmt = optarg; 3755 break; 3756 case OPTION_IMAGE_OPTS: 3757 image_opts = true; 3758 break; 3759 case 't': 3760 cache = optarg; 3761 break; 3762 case 'b': 3763 out_baseimg = optarg; 3764 break; 3765 case 'F': /* <=10.0 */ 3766 case 'B': 3767 out_basefmt = optarg; 3768 break; 3769 case 'u': 3770 unsafe = 1; 3771 break; 3772 case 'c': 3773 compress = true; 3774 break; 3775 case 'U': 3776 force_share = true; 3777 break; 3778 case 'p': 3779 progress = 1; 3780 break; 3781 case 'T': 3782 src_cache = optarg; 3783 break; 3784 case 'q': 3785 quiet = true; 3786 break; 3787 case OPTION_OBJECT: 3788 user_creatable_process_cmdline(optarg); 3789 break; 3790 default: 3791 tryhelp(argv[0]); 3792 } 3793 } 3794 3795 if (quiet) { 3796 progress = 0; 3797 } 3798 3799 if (optind != argc - 1) { 3800 error_exit(argv[0], "Expecting one image file name"); 3801 } 3802 if (!unsafe && !out_baseimg) { 3803 error_exit(argv[0], 3804 "Must specify backing file (-b) or use unsafe mode (-u)"); 3805 } 3806 filename = argv[optind++]; 3807 3808 qemu_progress_init(progress, 2.0); 3809 qemu_progress_print(0, 100); 3810 3811 flags = BDRV_O_RDWR | (unsafe ? BDRV_O_NO_BACKING : 0); 3812 ret = bdrv_parse_cache_mode(cache, &flags, &writethrough); 3813 if (ret < 0) { 3814 error_report("Invalid cache option: %s", cache); 3815 goto out; 3816 } 3817 3818 src_flags = 0; 3819 ret = bdrv_parse_cache_mode(src_cache, &src_flags, &src_writethrough); 3820 if (ret < 0) { 3821 error_report("Invalid source cache option: %s", src_cache); 3822 goto out; 3823 } 3824 3825 /* The source files are opened read-only, don't care about WCE */ 3826 assert((src_flags & BDRV_O_RDWR) == 0); 3827 (void) src_writethrough; 3828 3829 /* 3830 * Open the images. 3831 * 3832 * Ignore the old backing file for unsafe rebase in case we want to correct 3833 * the reference to a renamed or moved backing file. 3834 */ 3835 blk = img_open(image_opts, filename, fmt, flags, writethrough, quiet, 3836 false); 3837 if (!blk) { 3838 ret = -1; 3839 goto out; 3840 } 3841 bs = blk_bs(blk); 3842 3843 bdrv_graph_rdlock_main_loop(); 3844 unfiltered_bs = bdrv_skip_filters(bs); 3845 unfiltered_bs_cow = bdrv_cow_bs(unfiltered_bs); 3846 bdrv_graph_rdunlock_main_loop(); 3847 3848 if (compress && !block_driver_can_compress(unfiltered_bs->drv)) { 3849 error_report("Compression not supported for this file format"); 3850 ret = -1; 3851 goto out; 3852 } else if (compress) { 3853 write_flags |= BDRV_REQ_WRITE_COMPRESSED; 3854 } 3855 3856 if (out_basefmt != NULL) { 3857 if (bdrv_find_format(out_basefmt) == NULL) { 3858 error_report("Invalid format name: '%s'", out_basefmt); 3859 ret = -1; 3860 goto out; 3861 } 3862 } 3863 3864 /* 3865 * We need overlay subcluster size (or cluster size in case writes are 3866 * compressed) to make sure write requests are aligned. 3867 */ 3868 ret = bdrv_get_info(unfiltered_bs, &bdi); 3869 if (ret < 0) { 3870 error_report("could not get block driver info"); 3871 goto out; 3872 } else if (bdi.subcluster_size == 0) { 3873 bdi.cluster_size = bdi.subcluster_size = 1; 3874 } 3875 3876 write_align = compress ? bdi.cluster_size : bdi.subcluster_size; 3877 3878 /* For safe rebasing we need to compare old and new backing file */ 3879 if (!unsafe) { 3880 QDict *options = NULL; 3881 BlockDriverState *base_bs; 3882 3883 bdrv_graph_rdlock_main_loop(); 3884 base_bs = bdrv_cow_bs(unfiltered_bs); 3885 bdrv_graph_rdunlock_main_loop(); 3886 3887 if (base_bs) { 3888 blk_old_backing = blk_new(qemu_get_aio_context(), 3889 BLK_PERM_CONSISTENT_READ, 3890 BLK_PERM_ALL); 3891 ret = blk_insert_bs(blk_old_backing, base_bs, 3892 &local_err); 3893 if (ret < 0) { 3894 error_reportf_err(local_err, 3895 "Could not reuse old backing file '%s': ", 3896 base_bs->filename); 3897 goto out; 3898 } 3899 } else { 3900 blk_old_backing = NULL; 3901 } 3902 3903 if (out_baseimg[0]) { 3904 const char *overlay_filename; 3905 char *out_real_path; 3906 3907 options = qdict_new(); 3908 if (out_basefmt) { 3909 qdict_put_str(options, "driver", out_basefmt); 3910 } 3911 if (force_share) { 3912 qdict_put_bool(options, BDRV_OPT_FORCE_SHARE, true); 3913 } 3914 3915 bdrv_graph_rdlock_main_loop(); 3916 bdrv_refresh_filename(bs); 3917 bdrv_graph_rdunlock_main_loop(); 3918 overlay_filename = bs->exact_filename[0] ? bs->exact_filename 3919 : bs->filename; 3920 out_real_path = 3921 bdrv_get_full_backing_filename_from_filename(overlay_filename, 3922 out_baseimg, 3923 &local_err); 3924 if (local_err) { 3925 qobject_unref(options); 3926 error_reportf_err(local_err, 3927 "Could not resolve backing filename: "); 3928 ret = -1; 3929 goto out; 3930 } 3931 3932 /* 3933 * Find out whether we rebase an image on top of a previous image 3934 * in its chain. 3935 */ 3936 prefix_chain_bs = bdrv_find_backing_image(bs, out_real_path); 3937 if (prefix_chain_bs) { 3938 qobject_unref(options); 3939 g_free(out_real_path); 3940 3941 blk_new_backing = blk_new(qemu_get_aio_context(), 3942 BLK_PERM_CONSISTENT_READ, 3943 BLK_PERM_ALL); 3944 ret = blk_insert_bs(blk_new_backing, prefix_chain_bs, 3945 &local_err); 3946 if (ret < 0) { 3947 error_reportf_err(local_err, 3948 "Could not reuse backing file '%s': ", 3949 out_baseimg); 3950 goto out; 3951 } 3952 } else { 3953 blk_new_backing = blk_new_open(out_real_path, NULL, 3954 options, src_flags, &local_err); 3955 g_free(out_real_path); 3956 if (!blk_new_backing) { 3957 error_reportf_err(local_err, 3958 "Could not open new backing file '%s': ", 3959 out_baseimg); 3960 ret = -1; 3961 goto out; 3962 } 3963 } 3964 } 3965 } 3966 3967 /* 3968 * Check each unallocated cluster in the COW file. If it is unallocated, 3969 * accesses go to the backing file. We must therefore compare this cluster 3970 * in the old and new backing file, and if they differ we need to copy it 3971 * from the old backing file into the COW file. 3972 * 3973 * If qemu-img crashes during this step, no harm is done. The content of 3974 * the image is the same as the original one at any time. 3975 */ 3976 if (!unsafe) { 3977 int64_t size; 3978 int64_t old_backing_size = 0; 3979 int64_t new_backing_size = 0; 3980 uint64_t offset; 3981 int64_t n, n_old = 0, n_new = 0; 3982 float local_progress = 0; 3983 3984 if (blk_old_backing && bdrv_opt_mem_align(blk_bs(blk_old_backing)) > 3985 bdrv_opt_mem_align(blk_bs(blk))) { 3986 buf_old = blk_blockalign(blk_old_backing, IO_BUF_SIZE); 3987 } else { 3988 buf_old = blk_blockalign(blk, IO_BUF_SIZE); 3989 } 3990 buf_new = blk_blockalign(blk_new_backing, IO_BUF_SIZE); 3991 3992 size = blk_getlength(blk); 3993 if (size < 0) { 3994 error_report("Could not get size of '%s': %s", 3995 filename, strerror(-size)); 3996 ret = -1; 3997 goto out; 3998 } 3999 if (blk_old_backing) { 4000 old_backing_size = blk_getlength(blk_old_backing); 4001 if (old_backing_size < 0) { 4002 char backing_name[PATH_MAX]; 4003 4004 bdrv_get_backing_filename(bs, backing_name, 4005 sizeof(backing_name)); 4006 error_report("Could not get size of '%s': %s", 4007 backing_name, strerror(-old_backing_size)); 4008 ret = -1; 4009 goto out; 4010 } 4011 } 4012 if (blk_new_backing) { 4013 new_backing_size = blk_getlength(blk_new_backing); 4014 if (new_backing_size < 0) { 4015 error_report("Could not get size of '%s': %s", 4016 out_baseimg, strerror(-new_backing_size)); 4017 ret = -1; 4018 goto out; 4019 } 4020 } 4021 4022 if (size != 0) { 4023 local_progress = (float)100 / (size / MIN(size, IO_BUF_SIZE)); 4024 } 4025 4026 for (offset = 0; offset < size; offset += n) { 4027 bool old_backing_eof = false; 4028 int64_t n_alloc; 4029 4030 /* How many bytes can we handle with the next read? */ 4031 n = MIN(IO_BUF_SIZE, size - offset); 4032 4033 /* If the cluster is allocated, we don't need to take action */ 4034 ret = bdrv_is_allocated(unfiltered_bs, offset, n, &n); 4035 if (ret < 0) { 4036 error_report("error while reading image metadata: %s", 4037 strerror(-ret)); 4038 goto out; 4039 } 4040 if (ret) { 4041 continue; 4042 } 4043 4044 if (prefix_chain_bs) { 4045 uint64_t bytes = n; 4046 4047 /* 4048 * If cluster wasn't changed since prefix_chain, we don't need 4049 * to take action 4050 */ 4051 ret = bdrv_is_allocated_above(unfiltered_bs_cow, 4052 prefix_chain_bs, false, 4053 offset, n, &n); 4054 if (ret < 0) { 4055 error_report("error while reading image metadata: %s", 4056 strerror(-ret)); 4057 goto out; 4058 } 4059 if (!ret && n) { 4060 continue; 4061 } 4062 if (!n) { 4063 /* 4064 * If we've reached EOF of the old backing, it means that 4065 * offsets beyond the old backing size were read as zeroes. 4066 * Now we will need to explicitly zero the cluster in 4067 * order to preserve that state after the rebase. 4068 */ 4069 n = bytes; 4070 } 4071 } 4072 4073 /* 4074 * At this point we know that the region [offset; offset + n) 4075 * is unallocated within the target image. This region might be 4076 * unaligned to the target image's (sub)cluster boundaries, as 4077 * old backing may have smaller clusters (or have subclusters). 4078 * We extend it to the aligned boundaries to avoid CoW on 4079 * partial writes in blk_pwrite(), 4080 */ 4081 n += offset - QEMU_ALIGN_DOWN(offset, write_align); 4082 offset = QEMU_ALIGN_DOWN(offset, write_align); 4083 n += QEMU_ALIGN_UP(offset + n, write_align) - (offset + n); 4084 n = MIN(n, MIN(size - offset, IO_BUF_SIZE)); 4085 assert(!bdrv_is_allocated(unfiltered_bs, offset, n, &n_alloc) && 4086 n_alloc == n); 4087 4088 /* 4089 * Much like with the target image, we'll try to read as much 4090 * of the old and new backings as we can. 4091 */ 4092 n_old = MIN(n, MAX(0, old_backing_size - (int64_t) offset)); 4093 n_new = MIN(n, MAX(0, new_backing_size - (int64_t) offset)); 4094 4095 /* 4096 * Read old and new backing file and take into consideration that 4097 * backing files may be smaller than the COW image. 4098 */ 4099 memset(buf_old + n_old, 0, n - n_old); 4100 if (!n_old) { 4101 old_backing_eof = true; 4102 } else { 4103 ret = blk_pread(blk_old_backing, offset, n_old, buf_old, 0); 4104 if (ret < 0) { 4105 error_report("error while reading from old backing file"); 4106 goto out; 4107 } 4108 } 4109 4110 memset(buf_new + n_new, 0, n - n_new); 4111 if (n_new) { 4112 ret = blk_pread(blk_new_backing, offset, n_new, buf_new, 0); 4113 if (ret < 0) { 4114 error_report("error while reading from new backing file"); 4115 goto out; 4116 } 4117 } 4118 4119 /* If they differ, we need to write to the COW file */ 4120 uint64_t written = 0; 4121 4122 while (written < n) { 4123 int64_t pnum; 4124 4125 if (compare_buffers(buf_old + written, buf_new + written, 4126 n - written, write_align, &pnum)) 4127 { 4128 if (old_backing_eof) { 4129 ret = blk_pwrite_zeroes(blk, offset + written, pnum, 0); 4130 } else { 4131 assert(written + pnum <= IO_BUF_SIZE); 4132 ret = blk_pwrite(blk, offset + written, pnum, 4133 buf_old + written, write_flags); 4134 } 4135 if (ret < 0) { 4136 error_report("Error while writing to COW image: %s", 4137 strerror(-ret)); 4138 goto out; 4139 } 4140 } 4141 4142 written += pnum; 4143 if (offset + written >= old_backing_size) { 4144 old_backing_eof = true; 4145 } 4146 } 4147 qemu_progress_print(local_progress, 100); 4148 } 4149 } 4150 4151 /* 4152 * Change the backing file. All clusters that are different from the old 4153 * backing file are overwritten in the COW file now, so the visible content 4154 * doesn't change when we switch the backing file. 4155 */ 4156 if (out_baseimg && *out_baseimg) { 4157 ret = bdrv_change_backing_file(unfiltered_bs, out_baseimg, out_basefmt, 4158 true); 4159 } else { 4160 ret = bdrv_change_backing_file(unfiltered_bs, NULL, NULL, false); 4161 } 4162 4163 if (ret == -ENOSPC) { 4164 error_report("Could not change the backing file to '%s': No " 4165 "space left in the file header", out_baseimg); 4166 } else if (ret == -EINVAL && out_baseimg && !out_basefmt) { 4167 error_report("Could not change the backing file to '%s': backing " 4168 "format must be specified", out_baseimg); 4169 } else if (ret < 0) { 4170 error_report("Could not change the backing file to '%s': %s", 4171 out_baseimg, strerror(-ret)); 4172 } 4173 4174 qemu_progress_print(100, 0); 4175 /* 4176 * TODO At this point it is possible to check if any clusters that are 4177 * allocated in the COW file are the same in the backing file. If so, they 4178 * could be dropped from the COW file. Don't do this before switching the 4179 * backing file, in case of a crash this would lead to corruption. 4180 */ 4181 out: 4182 qemu_progress_end(); 4183 /* Cleanup */ 4184 if (!unsafe) { 4185 blk_unref(blk_old_backing); 4186 blk_unref(blk_new_backing); 4187 } 4188 qemu_vfree(buf_old); 4189 qemu_vfree(buf_new); 4190 4191 blk_unref(blk); 4192 if (ret) { 4193 return 1; 4194 } 4195 return 0; 4196 } 4197 4198 static int img_resize(const img_cmd_t *ccmd, int argc, char **argv) 4199 { 4200 Error *err = NULL; 4201 int c, ret, relative; 4202 const char *filename = NULL, *fmt = NULL, *size = NULL; 4203 int64_t n, total_size, current_size; 4204 bool quiet = false; 4205 BlockBackend *blk = NULL; 4206 PreallocMode prealloc = PREALLOC_MODE_OFF; 4207 QemuOpts *param; 4208 4209 static QemuOptsList resize_options = { 4210 .name = "resize_options", 4211 .head = QTAILQ_HEAD_INITIALIZER(resize_options.head), 4212 .desc = { 4213 { 4214 .name = BLOCK_OPT_SIZE, 4215 .type = QEMU_OPT_SIZE, 4216 .help = "Virtual disk size" 4217 }, { 4218 /* end of list */ 4219 } 4220 }, 4221 }; 4222 bool image_opts = false; 4223 bool shrink = false; 4224 4225 /* Parse getopt arguments */ 4226 for(;;) { 4227 static const struct option long_options[] = { 4228 {"help", no_argument, 0, 'h'}, 4229 {"format", required_argument, 0, 'f'}, 4230 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS}, 4231 {"preallocation", required_argument, 0, OPTION_PREALLOCATION}, 4232 {"shrink", no_argument, 0, OPTION_SHRINK}, 4233 {"quiet", no_argument, 0, 'q'}, 4234 {"object", required_argument, 0, OPTION_OBJECT}, 4235 {0, 0, 0, 0} 4236 }; 4237 c = getopt_long(argc, argv, "-hf:q", 4238 long_options, NULL); 4239 if (c == -1) { 4240 break; 4241 } 4242 switch(c) { 4243 case 'h': 4244 cmd_help(ccmd, "[-f FMT | --image-opts] [--preallocation PREALLOC] [--shrink]\n" 4245 " [-q] [--object OBJDEF] FILE [+-]SIZE[bkKMGTPE]\n" 4246 , 4247 " -f, --format FMT\n" 4248 " specify FILE format explicitly (default: probing is used)\n" 4249 " --image-opts\n" 4250 " treat FILE as an option string (key=value,...), not a file name\n" 4251 " (incompatible with -f|--format)\n" 4252 " --shrink\n" 4253 " allow operation when the new size is smaller than the original\n" 4254 " --preallocation PREALLOC\n" 4255 " specify FMT-specific preallocation type for the new areas\n" 4256 " -q, --quiet\n" 4257 " quiet mode (produce only error messages if any)\n" 4258 " --object OBJDEF\n" 4259 " defines QEMU user-creatable object\n" 4260 " FILE\n" 4261 " name of the image file, or option string (key=value,..)\n" 4262 " with --image-opts, to operate on\n" 4263 " [+-]SIZE[bkKMGTPE]\n" 4264 " new image size or amount by which to shrink (-)/grow (+),\n" 4265 " with optional multiplier suffix (powers of 1024, default is bytes)\n" 4266 ); 4267 return 0; 4268 case 'f': 4269 fmt = optarg; 4270 break; 4271 case OPTION_IMAGE_OPTS: 4272 image_opts = true; 4273 break; 4274 case OPTION_PREALLOCATION: 4275 prealloc = qapi_enum_parse(&PreallocMode_lookup, optarg, 4276 PREALLOC_MODE__MAX, NULL); 4277 if (prealloc == PREALLOC_MODE__MAX) { 4278 error_report("Invalid preallocation mode '%s'", optarg); 4279 return 1; 4280 } 4281 break; 4282 case OPTION_SHRINK: 4283 shrink = true; 4284 break; 4285 case 'q': 4286 quiet = true; 4287 break; 4288 case OPTION_OBJECT: 4289 user_creatable_process_cmdline(optarg); 4290 break; 4291 case 1: /* a non-optional argument */ 4292 if (!filename) { 4293 filename = optarg; 4294 /* see if we have -size (number) next to filename */ 4295 if (optind < argc) { 4296 size = argv[optind]; 4297 if (size[0] == '-' && size[1] >= '0' && size[1] <= '9') { 4298 ++optind; 4299 } else { 4300 size = NULL; 4301 } 4302 } 4303 } else if (!size) { 4304 size = optarg; 4305 } else { 4306 error_exit(argv[0], "Extra argument(s) in command line"); 4307 } 4308 break; 4309 default: 4310 tryhelp(argv[0]); 4311 } 4312 } 4313 if (!filename && optind < argc) { 4314 filename = argv[optind++]; 4315 } 4316 if (!size && optind < argc) { 4317 size = argv[optind++]; 4318 } 4319 if (!filename || !size || optind < argc) { 4320 error_exit(argv[0], "Expecting image file name and size"); 4321 } 4322 4323 /* Choose grow, shrink, or absolute resize mode */ 4324 switch (size[0]) { 4325 case '+': 4326 relative = 1; 4327 size++; 4328 break; 4329 case '-': 4330 relative = -1; 4331 size++; 4332 break; 4333 default: 4334 relative = 0; 4335 break; 4336 } 4337 4338 /* Parse size */ 4339 param = qemu_opts_create(&resize_options, NULL, 0, &error_abort); 4340 if (!qemu_opt_set(param, BLOCK_OPT_SIZE, size, &err)) { 4341 error_report_err(err); 4342 ret = -1; 4343 qemu_opts_del(param); 4344 goto out; 4345 } 4346 n = qemu_opt_get_size(param, BLOCK_OPT_SIZE, 0); 4347 qemu_opts_del(param); 4348 4349 blk = img_open(image_opts, filename, fmt, 4350 BDRV_O_RDWR | BDRV_O_RESIZE, false, quiet, 4351 false); 4352 if (!blk) { 4353 ret = -1; 4354 goto out; 4355 } 4356 4357 current_size = blk_getlength(blk); 4358 if (current_size < 0) { 4359 error_report("Failed to inquire current image length: %s", 4360 strerror(-current_size)); 4361 ret = -1; 4362 goto out; 4363 } 4364 4365 if (relative) { 4366 total_size = current_size + n * relative; 4367 } else { 4368 total_size = n; 4369 } 4370 if (total_size <= 0) { 4371 error_report("New image size must be positive"); 4372 ret = -1; 4373 goto out; 4374 } 4375 4376 if (total_size <= current_size && prealloc != PREALLOC_MODE_OFF) { 4377 error_report("Preallocation can only be used for growing images"); 4378 ret = -1; 4379 goto out; 4380 } 4381 4382 if (total_size < current_size && !shrink) { 4383 error_report("Use the --shrink option to perform a shrink operation."); 4384 warn_report("Shrinking an image will delete all data beyond the " 4385 "shrunken image's end. Before performing such an " 4386 "operation, make sure there is no important data there."); 4387 ret = -1; 4388 goto out; 4389 } 4390 4391 /* 4392 * The user expects the image to have the desired size after 4393 * resizing, so pass @exact=true. It is of no use to report 4394 * success when the image has not actually been resized. 4395 */ 4396 ret = blk_truncate(blk, total_size, true, prealloc, 0, &err); 4397 if (!ret) { 4398 qprintf(quiet, "Image resized.\n"); 4399 } else { 4400 error_report_err(err); 4401 } 4402 out: 4403 blk_unref(blk); 4404 if (ret) { 4405 return 1; 4406 } 4407 return 0; 4408 } 4409 4410 static void amend_status_cb(BlockDriverState *bs, 4411 int64_t offset, int64_t total_work_size, 4412 void *opaque) 4413 { 4414 qemu_progress_print(100.f * offset / total_work_size, 0); 4415 } 4416 4417 static int print_amend_option_help(const char *format) 4418 { 4419 BlockDriver *drv; 4420 4421 GRAPH_RDLOCK_GUARD_MAINLOOP(); 4422 4423 /* Find driver and parse its options */ 4424 drv = bdrv_find_format(format); 4425 if (!drv) { 4426 error_report("Unknown file format '%s'", format); 4427 return 1; 4428 } 4429 4430 if (!drv->bdrv_amend_options) { 4431 error_report("Format driver '%s' does not support option amendment", 4432 format); 4433 return 1; 4434 } 4435 4436 /* Every driver supporting amendment must have amend_opts */ 4437 assert(drv->amend_opts); 4438 4439 printf("Amend options for '%s':\n", format); 4440 qemu_opts_print_help(drv->amend_opts, false); 4441 return 0; 4442 } 4443 4444 static int img_amend(const img_cmd_t *ccmd, int argc, char **argv) 4445 { 4446 Error *err = NULL; 4447 int c, ret = 0; 4448 char *options = NULL; 4449 QemuOptsList *amend_opts = NULL; 4450 QemuOpts *opts = NULL; 4451 const char *fmt = NULL, *filename, *cache; 4452 int flags; 4453 bool writethrough; 4454 bool quiet = false, progress = false; 4455 BlockBackend *blk = NULL; 4456 BlockDriverState *bs = NULL; 4457 bool image_opts = false; 4458 bool force = false; 4459 4460 cache = BDRV_DEFAULT_CACHE; 4461 for (;;) { 4462 static const struct option long_options[] = { 4463 {"help", no_argument, 0, 'h'}, 4464 {"options", required_argument, 0, 'o'}, 4465 {"format", required_argument, 0, 'f'}, 4466 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS}, 4467 {"cache", required_argument, 0, 't'}, 4468 {"force", no_argument, 0, OPTION_FORCE}, 4469 {"progress", no_argument, 0, 'p'}, 4470 {"quiet", no_argument, 0, 'q'}, 4471 {"object", required_argument, 0, OPTION_OBJECT}, 4472 {0, 0, 0, 0} 4473 }; 4474 c = getopt_long(argc, argv, "ho:f:t:pq", 4475 long_options, NULL); 4476 if (c == -1) { 4477 break; 4478 } 4479 4480 switch (c) { 4481 case 'h': 4482 cmd_help(ccmd, "-o FMT_OPTS [-f FMT | --image-opts]\n" 4483 " [-t CACHE] [--force] [-p] [-q] [--object OBJDEF] FILE\n" 4484 , 4485 " -o, --options FMT_OPTS\n" 4486 " FMT-specfic format options (required)\n" 4487 " -f, --format FMT\n" 4488 " specify FILE format explicitly (default: probing is used)\n" 4489 " --image-opts\n" 4490 " treat FILE as an option string (key=value,..), not a file name\n" 4491 " (incompatible with -f|--format)\n" 4492 " -t, --cache CACHE\n" 4493 " cache mode for FILE (default: " BDRV_DEFAULT_CACHE ")\n" 4494 " --force\n" 4495 " allow certain unsafe operations\n" 4496 " -p, --progres\n" 4497 " show operation progress\n" 4498 " -q, --quiet\n" 4499 " quiet mode (produce only error messages if any)\n" 4500 " --object OBJDEF\n" 4501 " defines QEMU user-creatable object\n" 4502 " FILE\n" 4503 " name of the image file, or option string (key=value,..)\n" 4504 " with --image-opts, to operate on\n" 4505 ); 4506 break; 4507 case 'o': 4508 if (accumulate_options(&options, optarg) < 0) { 4509 ret = -1; 4510 goto out_no_progress; 4511 } 4512 break; 4513 case 'f': 4514 fmt = optarg; 4515 break; 4516 case OPTION_IMAGE_OPTS: 4517 image_opts = true; 4518 break; 4519 case 't': 4520 cache = optarg; 4521 break; 4522 case OPTION_FORCE: 4523 force = true; 4524 break; 4525 case 'p': 4526 progress = true; 4527 break; 4528 case 'q': 4529 quiet = true; 4530 break; 4531 case OPTION_OBJECT: 4532 user_creatable_process_cmdline(optarg); 4533 break; 4534 default: 4535 tryhelp(argv[0]); 4536 } 4537 } 4538 4539 if (!options) { 4540 error_exit(argv[0], "Must specify options (-o)"); 4541 } 4542 4543 if (quiet) { 4544 progress = false; 4545 } 4546 qemu_progress_init(progress, 1.0); 4547 4548 filename = (optind == argc - 1) ? argv[argc - 1] : NULL; 4549 if (fmt && has_help_option(options)) { 4550 /* If a format is explicitly specified (and possibly no filename is 4551 * given), print option help here */ 4552 ret = print_amend_option_help(fmt); 4553 goto out; 4554 } 4555 4556 if (optind != argc - 1) { 4557 error_report("Expecting one image file name"); 4558 ret = -1; 4559 goto out; 4560 } 4561 4562 flags = BDRV_O_RDWR; 4563 ret = bdrv_parse_cache_mode(cache, &flags, &writethrough); 4564 if (ret < 0) { 4565 error_report("Invalid cache option: %s", cache); 4566 goto out; 4567 } 4568 4569 blk = img_open(image_opts, filename, fmt, flags, writethrough, quiet, 4570 false); 4571 if (!blk) { 4572 ret = -1; 4573 goto out; 4574 } 4575 bs = blk_bs(blk); 4576 4577 fmt = bs->drv->format_name; 4578 4579 if (has_help_option(options)) { 4580 /* If the format was auto-detected, print option help here */ 4581 ret = print_amend_option_help(fmt); 4582 goto out; 4583 } 4584 4585 bdrv_graph_rdlock_main_loop(); 4586 if (!bs->drv->bdrv_amend_options) { 4587 error_report("Format driver '%s' does not support option amendment", 4588 fmt); 4589 bdrv_graph_rdunlock_main_loop(); 4590 ret = -1; 4591 goto out; 4592 } 4593 4594 /* Every driver supporting amendment must have amend_opts */ 4595 assert(bs->drv->amend_opts); 4596 4597 amend_opts = qemu_opts_append(amend_opts, bs->drv->amend_opts); 4598 opts = qemu_opts_create(amend_opts, NULL, 0, &error_abort); 4599 if (!qemu_opts_do_parse(opts, options, NULL, &err)) { 4600 qemu_opts_del(opts); 4601 /* Try to parse options using the create options */ 4602 amend_opts = qemu_opts_append(amend_opts, bs->drv->create_opts); 4603 opts = qemu_opts_create(amend_opts, NULL, 0, &error_abort); 4604 if (qemu_opts_do_parse(opts, options, NULL, NULL)) { 4605 error_append_hint(&err, 4606 "This option is only supported for image creation\n"); 4607 } 4608 4609 bdrv_graph_rdunlock_main_loop(); 4610 error_report_err(err); 4611 ret = -1; 4612 goto out; 4613 } 4614 4615 /* In case the driver does not call amend_status_cb() */ 4616 qemu_progress_print(0.f, 0); 4617 ret = bdrv_amend_options(bs, opts, &amend_status_cb, NULL, force, &err); 4618 qemu_progress_print(100.f, 0); 4619 bdrv_graph_rdunlock_main_loop(); 4620 4621 if (ret < 0) { 4622 error_report_err(err); 4623 goto out; 4624 } 4625 4626 out: 4627 qemu_progress_end(); 4628 4629 out_no_progress: 4630 blk_unref(blk); 4631 qemu_opts_del(opts); 4632 qemu_opts_free(amend_opts); 4633 g_free(options); 4634 4635 if (ret) { 4636 return 1; 4637 } 4638 return 0; 4639 } 4640 4641 typedef struct BenchData { 4642 BlockBackend *blk; 4643 uint64_t image_size; 4644 bool write; 4645 int bufsize; 4646 int step; 4647 int nrreq; 4648 int n; 4649 int flush_interval; 4650 bool drain_on_flush; 4651 uint8_t *buf; 4652 QEMUIOVector *qiov; 4653 4654 int in_flight; 4655 bool in_flush; 4656 uint64_t offset; 4657 } BenchData; 4658 4659 static void bench_undrained_flush_cb(void *opaque, int ret) 4660 { 4661 if (ret < 0) { 4662 error_report("Failed flush request: %s", strerror(-ret)); 4663 exit(EXIT_FAILURE); 4664 } 4665 } 4666 4667 static void bench_cb(void *opaque, int ret) 4668 { 4669 BenchData *b = opaque; 4670 BlockAIOCB *acb; 4671 4672 if (ret < 0) { 4673 error_report("Failed request: %s", strerror(-ret)); 4674 exit(EXIT_FAILURE); 4675 } 4676 4677 if (b->in_flush) { 4678 /* Just finished a flush with drained queue: Start next requests */ 4679 assert(b->in_flight == 0); 4680 b->in_flush = false; 4681 } else if (b->in_flight > 0) { 4682 int remaining = b->n - b->in_flight; 4683 4684 b->n--; 4685 b->in_flight--; 4686 4687 /* Time for flush? Drain queue if requested, then flush */ 4688 if (b->flush_interval && remaining % b->flush_interval == 0) { 4689 if (!b->in_flight || !b->drain_on_flush) { 4690 BlockCompletionFunc *cb; 4691 4692 if (b->drain_on_flush) { 4693 b->in_flush = true; 4694 cb = bench_cb; 4695 } else { 4696 cb = bench_undrained_flush_cb; 4697 } 4698 4699 acb = blk_aio_flush(b->blk, cb, b); 4700 if (!acb) { 4701 error_report("Failed to issue flush request"); 4702 exit(EXIT_FAILURE); 4703 } 4704 } 4705 if (b->drain_on_flush) { 4706 return; 4707 } 4708 } 4709 } 4710 4711 while (b->n > b->in_flight && b->in_flight < b->nrreq) { 4712 int64_t offset = b->offset; 4713 /* blk_aio_* might look for completed I/Os and kick bench_cb 4714 * again, so make sure this operation is counted by in_flight 4715 * and b->offset is ready for the next submission. 4716 */ 4717 b->in_flight++; 4718 b->offset += b->step; 4719 if (b->image_size <= b->bufsize) { 4720 b->offset = 0; 4721 } else { 4722 b->offset %= b->image_size - b->bufsize; 4723 } 4724 if (b->write) { 4725 acb = blk_aio_pwritev(b->blk, offset, b->qiov, 0, bench_cb, b); 4726 } else { 4727 acb = blk_aio_preadv(b->blk, offset, b->qiov, 0, bench_cb, b); 4728 } 4729 if (!acb) { 4730 error_report("Failed to issue request"); 4731 exit(EXIT_FAILURE); 4732 } 4733 } 4734 } 4735 4736 static int img_bench(const img_cmd_t *ccmd, int argc, char **argv) 4737 { 4738 int c, ret = 0; 4739 const char *fmt = NULL, *filename; 4740 bool quiet = false; 4741 bool image_opts = false; 4742 bool is_write = false; 4743 int count = 75000; 4744 int depth = 64; 4745 int64_t offset = 0; 4746 ssize_t bufsize = 4096; 4747 int pattern = 0; 4748 ssize_t step = 0; 4749 int flush_interval = 0; 4750 bool drain_on_flush = true; 4751 int64_t image_size; 4752 BlockBackend *blk = NULL; 4753 BenchData data = {}; 4754 int flags = 0; 4755 bool writethrough = false; 4756 struct timeval t1, t2; 4757 int i; 4758 bool force_share = false; 4759 size_t buf_size = 0; 4760 4761 for (;;) { 4762 static const struct option long_options[] = { 4763 {"help", no_argument, 0, 'h'}, 4764 {"format", required_argument, 0, 'f'}, 4765 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS}, 4766 {"cache", required_argument, 0, 't'}, 4767 {"count", required_argument, 0, 'c'}, 4768 {"depth", required_argument, 0, 'd'}, 4769 {"offset", required_argument, 0, 'o'}, 4770 {"buffer-size", required_argument, 0, 's'}, 4771 {"step-size", required_argument, 0, 'S'}, 4772 {"write", no_argument, 0, 'w'}, 4773 {"pattern", required_argument, 0, OPTION_PATTERN}, 4774 {"flush-interval", required_argument, 0, OPTION_FLUSH_INTERVAL}, 4775 {"no-drain", no_argument, 0, OPTION_NO_DRAIN}, 4776 {"aio", required_argument, 0, 'i'}, 4777 {"native", no_argument, 0, 'n'}, 4778 {"force-share", no_argument, 0, 'U'}, 4779 {"quiet", no_argument, 0, 'q'}, 4780 {"object", required_argument, 0, OPTION_OBJECT}, 4781 {0, 0, 0, 0} 4782 }; 4783 c = getopt_long(argc, argv, "hf:t:c:d:o:s:S:wi:nUq", 4784 long_options, NULL); 4785 if (c == -1) { 4786 break; 4787 } 4788 4789 switch (c) { 4790 case 'h': 4791 cmd_help(ccmd, "[-f FMT | --image-opts] [-t CACHE]\n" 4792 " [-c COUNT] [-d DEPTH] [-o OFFSET] [-s BUFFER_SIZE] [-S STEP_SIZE]\n" 4793 " [-w [--pattern PATTERN] [--flush-interval INTERVAL [--no-drain]]]\n" 4794 " [-i AIO] [-n] [-U] [-q] FILE\n" 4795 , 4796 " -f, --format FMT\n" 4797 " specify FILE format explicitly\n" 4798 " --image-opts\n" 4799 " indicates that FILE is a complete image specification\n" 4800 " instead of a file name (incompatible with --format)\n" 4801 " -t, --cache CACHE\n" 4802 " cache mode for FILE (default: " BDRV_DEFAULT_CACHE ")\n" 4803 " -c, --count COUNT\n" 4804 " number of I/O requests to perform\n" 4805 " -d, --depth DEPTH\n" 4806 " number of requests to perform in parallel\n" 4807 " -o, --offset OFFSET\n" 4808 " start first request at this OFFSET\n" 4809 " -s, --buffer-size BUFFER_SIZE[bkKMGTPE]\n" 4810 " size of each I/O request, with optional multiplier suffix\n" 4811 " (powers of 1024, default is 4K)\n" 4812 " -S, --step-size STEP_SIZE[bkKMGTPE]\n" 4813 " each next request offset increment, with optional multiplier suffix\n" 4814 " (powers of 1024, default is the same as BUFFER_SIZE)\n" 4815 " -w, --write\n" 4816 " perform write test (default is read)\n" 4817 " --pattern PATTERN\n" 4818 " write this pattern byte instead of zero\n" 4819 " --flush-interval FLUSH_INTERVAL\n" 4820 " issue flush after this number of requests\n" 4821 " --no-drain\n" 4822 " do not wait when flushing pending requests\n" 4823 " -i, --aio AIO\n" 4824 " async-io backend (threads, native, io_uring)\n" 4825 " -n, --native\n" 4826 " use native AIO backend if possible\n" 4827 " -U, --force-share\n" 4828 " open images in shared mode for concurrent access\n" 4829 " -q, --quiet\n" 4830 " quiet mode (produce only error messages if any)\n" 4831 " --object OBJDEF\n" 4832 " defines QEMU user-creatable object\n" 4833 " FILE\n" 4834 " name of the image file, or option string (key=value,..)\n" 4835 " with --image-opts, to operate on\n" 4836 ); 4837 break; 4838 case 'f': 4839 fmt = optarg; 4840 break; 4841 case OPTION_IMAGE_OPTS: 4842 image_opts = true; 4843 break; 4844 case 't': 4845 ret = bdrv_parse_cache_mode(optarg, &flags, &writethrough); 4846 if (ret < 0) { 4847 error_report("Invalid cache mode"); 4848 ret = -1; 4849 goto out; 4850 } 4851 break; 4852 case 'c': 4853 count = cvtnum_full("request count", optarg, false, 1, INT_MAX); 4854 if (count < 0) { 4855 return 1; 4856 } 4857 break; 4858 case 'd': 4859 depth = cvtnum_full("queue depth", optarg, false, 1, INT_MAX); 4860 if (depth < 0) { 4861 return 1; 4862 } 4863 break; 4864 case 'n': 4865 flags |= BDRV_O_NATIVE_AIO; 4866 break; 4867 case 'i': 4868 ret = bdrv_parse_aio(optarg, &flags); 4869 if (ret < 0) { 4870 error_report("Invalid aio option: %s", optarg); 4871 ret = -1; 4872 goto out; 4873 } 4874 break; 4875 case 'o': 4876 offset = cvtnum("offset", optarg, true); 4877 if (offset < 0) { 4878 return 1; 4879 } 4880 break; 4881 case 's': 4882 bufsize = cvtnum_full("buffer size", optarg, true, 1, INT_MAX); 4883 if (bufsize < 0) { 4884 return 1; 4885 } 4886 break; 4887 case 'S': 4888 step = cvtnum_full("step size", optarg, true, 0, INT_MAX); 4889 if (step < 0) { 4890 return 1; 4891 } 4892 break; 4893 case 'w': 4894 flags |= BDRV_O_RDWR; 4895 is_write = true; 4896 break; 4897 case OPTION_PATTERN: 4898 pattern = cvtnum_full("pattern byte", optarg, false, 0, 0xff); 4899 if (pattern < 0) { 4900 return 1; 4901 } 4902 break; 4903 case OPTION_FLUSH_INTERVAL: 4904 flush_interval = cvtnum_full("flush interval", optarg, 4905 false, 0, INT_MAX); 4906 if (flush_interval < 0) { 4907 return 1; 4908 } 4909 break; 4910 case OPTION_NO_DRAIN: 4911 drain_on_flush = false; 4912 break; 4913 case 'U': 4914 force_share = true; 4915 break; 4916 case 'q': 4917 quiet = true; 4918 break; 4919 case OPTION_OBJECT: 4920 user_creatable_process_cmdline(optarg); 4921 break; 4922 default: 4923 tryhelp(argv[0]); 4924 } 4925 } 4926 4927 if (optind != argc - 1) { 4928 error_exit(argv[0], "Expecting one image file name"); 4929 } 4930 filename = argv[argc - 1]; 4931 4932 if (!is_write && flush_interval) { 4933 error_report("--flush-interval is only available in write tests"); 4934 ret = -1; 4935 goto out; 4936 } 4937 if (flush_interval && flush_interval < depth) { 4938 error_report("Flush interval can't be smaller than depth"); 4939 ret = -1; 4940 goto out; 4941 } 4942 4943 blk = img_open(image_opts, filename, fmt, flags, writethrough, quiet, 4944 force_share); 4945 if (!blk) { 4946 ret = -1; 4947 goto out; 4948 } 4949 4950 image_size = blk_getlength(blk); 4951 if (image_size < 0) { 4952 ret = image_size; 4953 goto out; 4954 } 4955 4956 data = (BenchData) { 4957 .blk = blk, 4958 .image_size = image_size, 4959 .bufsize = bufsize, 4960 .step = step ?: bufsize, 4961 .nrreq = depth, 4962 .n = count, 4963 .offset = offset, 4964 .write = is_write, 4965 .flush_interval = flush_interval, 4966 .drain_on_flush = drain_on_flush, 4967 }; 4968 printf("Sending %d %s requests, %d bytes each, %d in parallel " 4969 "(starting at offset %" PRId64 ", step size %d)\n", 4970 data.n, data.write ? "write" : "read", data.bufsize, data.nrreq, 4971 data.offset, data.step); 4972 if (flush_interval) { 4973 printf("Sending flush every %d requests\n", flush_interval); 4974 } 4975 4976 buf_size = data.nrreq * data.bufsize; 4977 data.buf = blk_blockalign(blk, buf_size); 4978 memset(data.buf, pattern, data.nrreq * data.bufsize); 4979 4980 blk_register_buf(blk, data.buf, buf_size, &error_fatal); 4981 4982 data.qiov = g_new(QEMUIOVector, data.nrreq); 4983 for (i = 0; i < data.nrreq; i++) { 4984 qemu_iovec_init(&data.qiov[i], 1); 4985 qemu_iovec_add(&data.qiov[i], 4986 data.buf + i * data.bufsize, data.bufsize); 4987 } 4988 4989 gettimeofday(&t1, NULL); 4990 bench_cb(&data, 0); 4991 4992 while (data.n > 0) { 4993 main_loop_wait(false); 4994 } 4995 gettimeofday(&t2, NULL); 4996 4997 printf("Run completed in %3.3f seconds.\n", 4998 (t2.tv_sec - t1.tv_sec) 4999 + ((double)(t2.tv_usec - t1.tv_usec) / 1000000)); 5000 5001 out: 5002 if (data.buf) { 5003 blk_unregister_buf(blk, data.buf, buf_size); 5004 } 5005 qemu_vfree(data.buf); 5006 blk_unref(blk); 5007 5008 if (ret) { 5009 return 1; 5010 } 5011 return 0; 5012 } 5013 5014 enum ImgBitmapAct { 5015 BITMAP_ADD, 5016 BITMAP_REMOVE, 5017 BITMAP_CLEAR, 5018 BITMAP_ENABLE, 5019 BITMAP_DISABLE, 5020 BITMAP_MERGE, 5021 }; 5022 typedef struct ImgBitmapAction { 5023 enum ImgBitmapAct act; 5024 const char *src; /* only used for merge */ 5025 QSIMPLEQ_ENTRY(ImgBitmapAction) next; 5026 } ImgBitmapAction; 5027 5028 static int img_bitmap(const img_cmd_t *ccmd, int argc, char **argv) 5029 { 5030 Error *err = NULL; 5031 int c, ret = 1; 5032 QemuOpts *opts = NULL; 5033 const char *fmt = NULL, *src_fmt = NULL, *src_filename = NULL; 5034 const char *filename, *bitmap; 5035 BlockBackend *blk = NULL, *src = NULL; 5036 BlockDriverState *bs = NULL, *src_bs = NULL; 5037 bool image_opts = false; 5038 int64_t granularity = 0; 5039 bool add = false, merge = false; 5040 QSIMPLEQ_HEAD(, ImgBitmapAction) actions; 5041 ImgBitmapAction *act, *act_next; 5042 const char *op; 5043 int inactivate_ret; 5044 5045 QSIMPLEQ_INIT(&actions); 5046 5047 for (;;) { 5048 static const struct option long_options[] = { 5049 {"help", no_argument, 0, 'h'}, 5050 {"format", required_argument, 0, 'f'}, 5051 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS}, 5052 {"add", no_argument, 0, OPTION_ADD}, 5053 {"granularity", required_argument, 0, 'g'}, 5054 {"remove", no_argument, 0, OPTION_REMOVE}, 5055 {"clear", no_argument, 0, OPTION_CLEAR}, 5056 {"enable", no_argument, 0, OPTION_ENABLE}, 5057 {"disable", no_argument, 0, OPTION_DISABLE}, 5058 {"merge", required_argument, 0, OPTION_MERGE}, 5059 {"source-file", required_argument, 0, 'b'}, 5060 {"source-format", required_argument, 0, 'F'}, 5061 {"object", required_argument, 0, OPTION_OBJECT}, 5062 {0, 0, 0, 0} 5063 }; 5064 c = getopt_long(argc, argv, "hf:g:b:F:", 5065 long_options, NULL); 5066 if (c == -1) { 5067 break; 5068 } 5069 5070 switch (c) { 5071 case 'h': 5072 cmd_help(ccmd, "[-f FMT | --image-opts]\n" 5073 " ( --add [-g SIZE] | --remove | --clear | --enable | --disable |\n" 5074 " --merge SOURCE [-b SRC_FILE [-F SRC_FMT]] )..\n" 5075 " [--object OBJDEF] FILE BITMAP\n" 5076 , 5077 " -f, --format FMT\n" 5078 " specify FILE format explicitly (default: probing is used)\n" 5079 " --image-opts\n" 5080 " treat FILE as an option string (key=value,..), not a file name\n" 5081 " (incompatible with -f|--format)\n" 5082 " --add\n" 5083 " creates BITMAP in FILE, enables to record future edits\n" 5084 " -g, --granularity SIZE[bKMGTPE]\n" 5085 " sets non-default granularity for the bitmap being added,\n" 5086 " with optional multiplier suffix (in powers of 1024)\n" 5087 " --remove\n" 5088 " removes BITMAP from FILE\n" 5089 " --clear\n" 5090 " clears BITMAP in FILE\n" 5091 " --enable, --disable\n" 5092 " starts and stops recording future edits to BITMAP in FILE\n" 5093 " --merge SOURCE\n" 5094 " merges contents of the SOURCE bitmap into BITMAP in FILE\n" 5095 " -b, --source-file SRC_FILE\n" 5096 " select alternative source file for --merge\n" 5097 " -F, --source-format SRC_FMT\n" 5098 " specify format for SRC_FILE explicitly\n" 5099 " --object OBJDEF\n" 5100 " defines QEMU user-creatable object\n" 5101 " FILE\n" 5102 " name of the image file, or option string (key=value,..)\n" 5103 " with --image-opts, to operate on\n" 5104 " BITMAP\n" 5105 " name of the bitmap to add, remove, clear, enable, disable or merge to\n" 5106 ); 5107 break; 5108 case 'f': 5109 fmt = optarg; 5110 break; 5111 case OPTION_IMAGE_OPTS: 5112 image_opts = true; 5113 break; 5114 case OPTION_ADD: 5115 act = g_new0(ImgBitmapAction, 1); 5116 act->act = BITMAP_ADD; 5117 QSIMPLEQ_INSERT_TAIL(&actions, act, next); 5118 add = true; 5119 break; 5120 case 'g': 5121 granularity = cvtnum("granularity", optarg, true); 5122 if (granularity < 0) { 5123 return 1; 5124 } 5125 break; 5126 case OPTION_REMOVE: 5127 act = g_new0(ImgBitmapAction, 1); 5128 act->act = BITMAP_REMOVE; 5129 QSIMPLEQ_INSERT_TAIL(&actions, act, next); 5130 break; 5131 case OPTION_CLEAR: 5132 act = g_new0(ImgBitmapAction, 1); 5133 act->act = BITMAP_CLEAR; 5134 QSIMPLEQ_INSERT_TAIL(&actions, act, next); 5135 break; 5136 case OPTION_ENABLE: 5137 act = g_new0(ImgBitmapAction, 1); 5138 act->act = BITMAP_ENABLE; 5139 QSIMPLEQ_INSERT_TAIL(&actions, act, next); 5140 break; 5141 case OPTION_DISABLE: 5142 act = g_new0(ImgBitmapAction, 1); 5143 act->act = BITMAP_DISABLE; 5144 QSIMPLEQ_INSERT_TAIL(&actions, act, next); 5145 break; 5146 case OPTION_MERGE: 5147 act = g_new0(ImgBitmapAction, 1); 5148 act->act = BITMAP_MERGE; 5149 act->src = optarg; 5150 QSIMPLEQ_INSERT_TAIL(&actions, act, next); 5151 merge = true; 5152 break; 5153 case 'b': 5154 src_filename = optarg; 5155 break; 5156 case 'F': 5157 src_fmt = optarg; 5158 break; 5159 case OPTION_OBJECT: 5160 user_creatable_process_cmdline(optarg); 5161 break; 5162 default: 5163 tryhelp(argv[0]); 5164 } 5165 } 5166 5167 if (QSIMPLEQ_EMPTY(&actions)) { 5168 error_report("Need at least one of --add, --remove, --clear, " 5169 "--enable, --disable, or --merge"); 5170 goto out; 5171 } 5172 5173 if (granularity && !add) { 5174 error_report("granularity only supported with --add"); 5175 goto out; 5176 } 5177 if (src_fmt && !src_filename) { 5178 error_report("-F only supported with -b"); 5179 goto out; 5180 } 5181 if (src_filename && !merge) { 5182 error_report("Merge bitmap source file only supported with " 5183 "--merge"); 5184 goto out; 5185 } 5186 5187 if (optind != argc - 2) { 5188 error_report("Expecting filename and bitmap name"); 5189 goto out; 5190 } 5191 5192 filename = argv[optind]; 5193 bitmap = argv[optind + 1]; 5194 5195 /* 5196 * No need to open backing chains; we will be manipulating bitmaps 5197 * directly in this image without reference to image contents. 5198 */ 5199 blk = img_open(image_opts, filename, fmt, BDRV_O_RDWR | BDRV_O_NO_BACKING, 5200 false, false, false); 5201 if (!blk) { 5202 goto out; 5203 } 5204 bs = blk_bs(blk); 5205 if (src_filename) { 5206 src = img_open(false, src_filename, src_fmt, BDRV_O_NO_BACKING, 5207 false, false, false); 5208 if (!src) { 5209 goto out; 5210 } 5211 src_bs = blk_bs(src); 5212 } else { 5213 src_bs = bs; 5214 } 5215 5216 QSIMPLEQ_FOREACH_SAFE(act, &actions, next, act_next) { 5217 switch (act->act) { 5218 case BITMAP_ADD: 5219 qmp_block_dirty_bitmap_add(bs->node_name, bitmap, 5220 !!granularity, granularity, true, true, 5221 false, false, &err); 5222 op = "add"; 5223 break; 5224 case BITMAP_REMOVE: 5225 qmp_block_dirty_bitmap_remove(bs->node_name, bitmap, &err); 5226 op = "remove"; 5227 break; 5228 case BITMAP_CLEAR: 5229 qmp_block_dirty_bitmap_clear(bs->node_name, bitmap, &err); 5230 op = "clear"; 5231 break; 5232 case BITMAP_ENABLE: 5233 qmp_block_dirty_bitmap_enable(bs->node_name, bitmap, &err); 5234 op = "enable"; 5235 break; 5236 case BITMAP_DISABLE: 5237 qmp_block_dirty_bitmap_disable(bs->node_name, bitmap, &err); 5238 op = "disable"; 5239 break; 5240 case BITMAP_MERGE: 5241 do_dirty_bitmap_merge(bs->node_name, bitmap, src_bs->node_name, 5242 act->src, &err); 5243 op = "merge"; 5244 break; 5245 default: 5246 g_assert_not_reached(); 5247 } 5248 5249 if (err) { 5250 error_reportf_err(err, "Operation %s on bitmap %s failed: ", 5251 op, bitmap); 5252 goto out; 5253 } 5254 g_free(act); 5255 } 5256 5257 ret = 0; 5258 5259 out: 5260 /* 5261 * Manually inactivate the images first because this way we can know whether 5262 * an error occurred. blk_unref() doesn't tell us about failures. 5263 */ 5264 inactivate_ret = bdrv_inactivate_all(); 5265 if (inactivate_ret < 0) { 5266 error_report("Error while closing the image: %s", strerror(-inactivate_ret)); 5267 ret = 1; 5268 } 5269 5270 blk_unref(src); 5271 blk_unref(blk); 5272 qemu_opts_del(opts); 5273 return ret; 5274 } 5275 5276 #define C_BS 01 5277 #define C_COUNT 02 5278 #define C_IF 04 5279 #define C_OF 010 5280 #define C_SKIP 020 5281 5282 struct DdInfo { 5283 unsigned int flags; 5284 int64_t count; 5285 }; 5286 5287 struct DdIo { 5288 int bsz; /* Block size */ 5289 char *filename; 5290 uint8_t *buf; 5291 int64_t offset; 5292 }; 5293 5294 struct DdOpts { 5295 const char *name; 5296 int (*f)(const char *, struct DdIo *, struct DdIo *, struct DdInfo *); 5297 unsigned int flag; 5298 }; 5299 5300 static int img_dd_bs(const char *arg, 5301 struct DdIo *in, struct DdIo *out, 5302 struct DdInfo *dd) 5303 { 5304 int64_t res; 5305 5306 res = cvtnum_full("bs", arg, true, 1, INT_MAX); 5307 5308 if (res < 0) { 5309 return 1; 5310 } 5311 in->bsz = out->bsz = res; 5312 5313 return 0; 5314 } 5315 5316 static int img_dd_count(const char *arg, 5317 struct DdIo *in, struct DdIo *out, 5318 struct DdInfo *dd) 5319 { 5320 dd->count = cvtnum("count", arg, true); 5321 5322 if (dd->count < 0) { 5323 return 1; 5324 } 5325 5326 return 0; 5327 } 5328 5329 static int img_dd_if(const char *arg, 5330 struct DdIo *in, struct DdIo *out, 5331 struct DdInfo *dd) 5332 { 5333 in->filename = g_strdup(arg); 5334 5335 return 0; 5336 } 5337 5338 static int img_dd_of(const char *arg, 5339 struct DdIo *in, struct DdIo *out, 5340 struct DdInfo *dd) 5341 { 5342 out->filename = g_strdup(arg); 5343 5344 return 0; 5345 } 5346 5347 static int img_dd_skip(const char *arg, 5348 struct DdIo *in, struct DdIo *out, 5349 struct DdInfo *dd) 5350 { 5351 in->offset = cvtnum("skip", arg, true); 5352 5353 if (in->offset < 0) { 5354 return 1; 5355 } 5356 5357 return 0; 5358 } 5359 5360 static int img_dd(const img_cmd_t *ccmd, int argc, char **argv) 5361 { 5362 int ret = 0; 5363 char *arg = NULL; 5364 char *tmp; 5365 BlockDriver *drv = NULL, *proto_drv = NULL; 5366 BlockBackend *blk1 = NULL, *blk2 = NULL; 5367 QemuOpts *opts = NULL; 5368 QemuOptsList *create_opts = NULL; 5369 Error *local_err = NULL; 5370 bool image_opts = false; 5371 int c, i; 5372 const char *out_fmt = "raw"; 5373 const char *fmt = NULL; 5374 int64_t size = 0; 5375 int64_t out_pos, in_pos; 5376 bool force_share = false; 5377 struct DdInfo dd = { 5378 .flags = 0, 5379 .count = 0, 5380 }; 5381 struct DdIo in = { 5382 .bsz = 512, /* Block size is by default 512 bytes */ 5383 .filename = NULL, 5384 .buf = NULL, 5385 .offset = 0 5386 }; 5387 struct DdIo out = { 5388 .bsz = 512, 5389 .filename = NULL, 5390 .buf = NULL, 5391 .offset = 0 5392 }; 5393 5394 const struct DdOpts options[] = { 5395 { "bs", img_dd_bs, C_BS }, 5396 { "count", img_dd_count, C_COUNT }, 5397 { "if", img_dd_if, C_IF }, 5398 { "of", img_dd_of, C_OF }, 5399 { "skip", img_dd_skip, C_SKIP }, 5400 { NULL, NULL, 0 } 5401 }; 5402 const struct option long_options[] = { 5403 { "help", no_argument, 0, 'h'}, 5404 { "format", required_argument, 0, 'f'}, 5405 { "image-opts", no_argument, 0, OPTION_IMAGE_OPTS}, 5406 { "output-format", required_argument, 0, 'O'}, 5407 { "force-share", no_argument, 0, 'U'}, 5408 { "object", required_argument, 0, OPTION_OBJECT}, 5409 { 0, 0, 0, 0 } 5410 }; 5411 5412 while ((c = getopt_long(argc, argv, "hf:O:U", long_options, NULL))) { 5413 if (c == EOF) { 5414 break; 5415 } 5416 switch (c) { 5417 case 'h': 5418 cmd_help(ccmd, "[-f FMT|--image-opts] [-O OUTPUT_FMT] [-U]\n" 5419 " [--object OBJDEF] [bs=BLOCK_SIZE] [count=BLOCKS] if=INPUT of=OUTPUT\n" 5420 , 5421 " -f, --format FMT\n" 5422 " specify format for INPUT explicitly (default: probing is used)\n" 5423 " --image-opts\n" 5424 " treat INPUT as an option string (key=value,..), not a file name\n" 5425 " (incompatible with -f|--format)\n" 5426 " -O, --output-format OUTPUT_FMT\n" 5427 " format of the OUTPUT (default: raw)\n" 5428 " -U, --force-share\n" 5429 " open images in shared mode for concurrent access\n" 5430 " --object OBJDEF\n" 5431 " defines QEMU user-creatable object\n" 5432 " bs=BLOCK_SIZE[bKMGTP]\n" 5433 " size of the I/O block, with optional multiplier suffix (powers of 1024)\n" 5434 " (default: 512)\n" 5435 " count=COUNT\n" 5436 " number of blocks to convert (default whole INPUT)\n" 5437 " if=INPUT\n" 5438 " name of the file, or option string (key=value,..)\n" 5439 " with --image-opts, to use for input\n" 5440 " of=OUTPUT\n" 5441 " output file name to create (will be overridden if alrady exists)\n" 5442 ); 5443 break; 5444 case 'f': 5445 fmt = optarg; 5446 break; 5447 case OPTION_IMAGE_OPTS: 5448 image_opts = true; 5449 break; 5450 case 'O': 5451 out_fmt = optarg; 5452 break; 5453 case 'U': 5454 force_share = true; 5455 break; 5456 case OPTION_OBJECT: 5457 user_creatable_process_cmdline(optarg); 5458 break; 5459 default: 5460 tryhelp(argv[0]); 5461 } 5462 } 5463 5464 for (i = optind; i < argc; i++) { 5465 int j; 5466 arg = g_strdup(argv[i]); 5467 5468 tmp = strchr(arg, '='); 5469 if (tmp == NULL) { 5470 error_report("unrecognized operand %s", arg); 5471 ret = -1; 5472 goto out; 5473 } 5474 5475 *tmp++ = '\0'; 5476 5477 for (j = 0; options[j].name != NULL; j++) { 5478 if (!strcmp(arg, options[j].name)) { 5479 break; 5480 } 5481 } 5482 if (options[j].name == NULL) { 5483 error_report("unrecognized operand %s", arg); 5484 ret = -1; 5485 goto out; 5486 } 5487 5488 if (options[j].f(tmp, &in, &out, &dd) != 0) { 5489 ret = -1; 5490 goto out; 5491 } 5492 dd.flags |= options[j].flag; 5493 g_free(arg); 5494 arg = NULL; 5495 } 5496 5497 if (!(dd.flags & C_IF && dd.flags & C_OF)) { 5498 error_report("Must specify both input and output files"); 5499 ret = -1; 5500 goto out; 5501 } 5502 5503 blk1 = img_open(image_opts, in.filename, fmt, 0, false, false, 5504 force_share); 5505 5506 if (!blk1) { 5507 ret = -1; 5508 goto out; 5509 } 5510 5511 drv = bdrv_find_format(out_fmt); 5512 if (!drv) { 5513 error_report("Unknown file format"); 5514 ret = -1; 5515 goto out; 5516 } 5517 proto_drv = bdrv_find_protocol(out.filename, true, &local_err); 5518 5519 if (!proto_drv) { 5520 error_report_err(local_err); 5521 ret = -1; 5522 goto out; 5523 } 5524 if (!drv->create_opts) { 5525 error_report("Format driver '%s' does not support image creation", 5526 drv->format_name); 5527 ret = -1; 5528 goto out; 5529 } 5530 if (!proto_drv->create_opts) { 5531 error_report("Protocol driver '%s' does not support image creation", 5532 proto_drv->format_name); 5533 ret = -1; 5534 goto out; 5535 } 5536 create_opts = qemu_opts_append(create_opts, drv->create_opts); 5537 create_opts = qemu_opts_append(create_opts, proto_drv->create_opts); 5538 5539 opts = qemu_opts_create(create_opts, NULL, 0, &error_abort); 5540 5541 size = blk_getlength(blk1); 5542 if (size < 0) { 5543 error_report("Failed to get size for '%s'", in.filename); 5544 ret = -1; 5545 goto out; 5546 } 5547 5548 if (dd.flags & C_COUNT && dd.count <= INT64_MAX / in.bsz && 5549 dd.count * in.bsz < size) { 5550 size = dd.count * in.bsz; 5551 } 5552 5553 /* Overflow means the specified offset is beyond input image's size */ 5554 if (dd.flags & C_SKIP && (in.offset > INT64_MAX / in.bsz || 5555 size < in.bsz * in.offset)) { 5556 qemu_opt_set_number(opts, BLOCK_OPT_SIZE, 0, &error_abort); 5557 } else { 5558 qemu_opt_set_number(opts, BLOCK_OPT_SIZE, 5559 size - in.bsz * in.offset, &error_abort); 5560 } 5561 5562 ret = bdrv_create(drv, out.filename, opts, &local_err); 5563 if (ret < 0) { 5564 error_reportf_err(local_err, 5565 "%s: error while creating output image: ", 5566 out.filename); 5567 ret = -1; 5568 goto out; 5569 } 5570 5571 /* TODO, we can't honour --image-opts for the target, 5572 * since it needs to be given in a format compatible 5573 * with the bdrv_create() call above which does not 5574 * support image-opts style. 5575 */ 5576 blk2 = img_open_file(out.filename, NULL, out_fmt, BDRV_O_RDWR, 5577 false, false, false); 5578 5579 if (!blk2) { 5580 ret = -1; 5581 goto out; 5582 } 5583 5584 if (dd.flags & C_SKIP && (in.offset > INT64_MAX / in.bsz || 5585 size < in.offset * in.bsz)) { 5586 /* We give a warning if the skip option is bigger than the input 5587 * size and create an empty output disk image (i.e. like dd(1)). 5588 */ 5589 error_report("%s: cannot skip to specified offset", in.filename); 5590 in_pos = size; 5591 } else { 5592 in_pos = in.offset * in.bsz; 5593 } 5594 5595 in.buf = g_new(uint8_t, in.bsz); 5596 5597 for (out_pos = 0; in_pos < size; ) { 5598 int bytes = (in_pos + in.bsz > size) ? size - in_pos : in.bsz; 5599 5600 ret = blk_pread(blk1, in_pos, bytes, in.buf, 0); 5601 if (ret < 0) { 5602 error_report("error while reading from input image file: %s", 5603 strerror(-ret)); 5604 goto out; 5605 } 5606 in_pos += bytes; 5607 5608 ret = blk_pwrite(blk2, out_pos, bytes, in.buf, 0); 5609 if (ret < 0) { 5610 error_report("error while writing to output image file: %s", 5611 strerror(-ret)); 5612 goto out; 5613 } 5614 out_pos += bytes; 5615 } 5616 5617 out: 5618 g_free(arg); 5619 qemu_opts_del(opts); 5620 qemu_opts_free(create_opts); 5621 blk_unref(blk1); 5622 blk_unref(blk2); 5623 g_free(in.filename); 5624 g_free(out.filename); 5625 g_free(in.buf); 5626 g_free(out.buf); 5627 5628 if (ret) { 5629 return 1; 5630 } 5631 return 0; 5632 } 5633 5634 static void dump_json_block_measure_info(BlockMeasureInfo *info) 5635 { 5636 GString *str; 5637 QObject *obj; 5638 Visitor *v = qobject_output_visitor_new(&obj); 5639 5640 visit_type_BlockMeasureInfo(v, NULL, &info, &error_abort); 5641 visit_complete(v, &obj); 5642 str = qobject_to_json_pretty(obj, true); 5643 assert(str != NULL); 5644 printf("%s\n", str->str); 5645 qobject_unref(obj); 5646 visit_free(v); 5647 g_string_free(str, true); 5648 } 5649 5650 static int img_measure(const img_cmd_t *ccmd, int argc, char **argv) 5651 { 5652 OutputFormat output_format = OFORMAT_HUMAN; 5653 BlockBackend *in_blk = NULL; 5654 BlockDriver *drv; 5655 const char *filename = NULL; 5656 const char *fmt = NULL; 5657 const char *out_fmt = "raw"; 5658 char *options = NULL; 5659 char *snapshot_name = NULL; 5660 bool force_share = false; 5661 QemuOpts *opts = NULL; 5662 QemuOpts *object_opts = NULL; 5663 QemuOpts *sn_opts = NULL; 5664 QemuOptsList *create_opts = NULL; 5665 bool image_opts = false; 5666 int64_t img_size = -1; 5667 BlockMeasureInfo *info = NULL; 5668 Error *local_err = NULL; 5669 int ret = 1; 5670 int c; 5671 5672 static const struct option long_options[] = { 5673 {"help", no_argument, 0, 'h'}, 5674 {"source-format", required_argument, 0, 'f'}, /* img_convert */ 5675 {"format", required_argument, 0, 'f'}, 5676 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS}, 5677 {"source-image-opts", no_argument, 0, OPTION_IMAGE_OPTS}, /* img_convert */ 5678 {"snapshot", required_argument, 0, 'l'}, 5679 {"target-format", required_argument, 0, 'O'}, 5680 {"target-format-options", required_argument, 0, 'o'}, /* img_convert */ 5681 {"options", required_argument, 0, 'o'}, 5682 {"force-share", no_argument, 0, 'U'}, 5683 {"output", required_argument, 0, OPTION_OUTPUT}, 5684 {"object", required_argument, 0, OPTION_OBJECT}, 5685 {"size", required_argument, 0, 's'}, 5686 {0, 0, 0, 0} 5687 }; 5688 5689 while ((c = getopt_long(argc, argv, "hf:l:O:o:Us:", 5690 long_options, NULL)) != -1) { 5691 switch (c) { 5692 case 'h': 5693 cmd_help(ccmd, "[-f FMT|--image-opts] [-l SNAPSHOT]\n" 5694 " [-O TARGET_FMT] [-o TARGET_FMT_OPTS] [--output human|json]\n" 5695 " [--object OBJDEF] (--size SIZE | FILE)\n" 5696 , 5697 " -f, --format\n" 5698 " specify format of FILE explicitly (default: probing is used)\n" 5699 " --image-opts\n" 5700 " indicates that FILE is a complete image specification\n" 5701 " instead of a file name (incompatible with --format)\n" 5702 " -l, --snapshot SNAPSHOT\n" 5703 " use this snapshot in FILE as source\n" 5704 " -O, --target-format TARGET_FMT\n" 5705 " desired target/output image format (default: raw)\n" 5706 " -o TARGET_FMT_OPTS\n" 5707 " options specific to TARGET_FMT\n" 5708 " --output human|json\n" 5709 " output format (default: human)\n" 5710 " -U, --force-share\n" 5711 " open images in shared mode for concurrent access\n" 5712 " --object OBJDEF\n" 5713 " defines QEMU user-creatable object\n" 5714 " -s, --size SIZE[bKMGTPE]\n" 5715 " measure file size for given image size,\n" 5716 " with optional multiplier suffix (powers of 1024)\n" 5717 " FILE\n" 5718 " measure file size required to convert from FILE (either a file name\n" 5719 " or an option string (key=value,..) with --image-options)\n" 5720 ); 5721 break; 5722 case 'f': 5723 fmt = optarg; 5724 break; 5725 case OPTION_IMAGE_OPTS: 5726 image_opts = true; 5727 break; 5728 case 'l': 5729 if (strstart(optarg, SNAPSHOT_OPT_BASE, NULL)) { 5730 sn_opts = qemu_opts_parse_noisily(&internal_snapshot_opts, 5731 optarg, false); 5732 if (!sn_opts) { 5733 error_report("Failed in parsing snapshot param '%s'", 5734 optarg); 5735 goto out; 5736 } 5737 } else { 5738 snapshot_name = optarg; 5739 } 5740 break; 5741 case 'O': 5742 out_fmt = optarg; 5743 break; 5744 case 'o': 5745 if (accumulate_options(&options, optarg) < 0) { 5746 goto out; 5747 } 5748 break; 5749 case 'U': 5750 force_share = true; 5751 break; 5752 case OPTION_OUTPUT: 5753 output_format = parse_output_format(argv[0], optarg); 5754 break; 5755 case OPTION_OBJECT: 5756 user_creatable_process_cmdline(optarg); 5757 break; 5758 case 's': 5759 img_size = cvtnum("image size", optarg, true); 5760 if (img_size < 0) { 5761 goto out; 5762 } 5763 break; 5764 default: 5765 tryhelp(argv[0]); 5766 } 5767 } 5768 5769 if (argc - optind > 1) { 5770 error_report("At most one filename argument is allowed."); 5771 goto out; 5772 } else if (argc - optind == 1) { 5773 filename = argv[optind]; 5774 } 5775 5776 if (!filename && (image_opts || fmt || snapshot_name || sn_opts)) { 5777 error_report("--image-opts, -f, and -l require a filename argument."); 5778 goto out; 5779 } 5780 if (filename && img_size != -1) { 5781 error_report("--size N cannot be used together with a filename."); 5782 goto out; 5783 } 5784 if (!filename && img_size == -1) { 5785 error_report("Either --size N or one filename must be specified."); 5786 goto out; 5787 } 5788 5789 if (filename) { 5790 in_blk = img_open(image_opts, filename, fmt, 0, 5791 false, false, force_share); 5792 if (!in_blk) { 5793 goto out; 5794 } 5795 5796 if (sn_opts) { 5797 bdrv_snapshot_load_tmp(blk_bs(in_blk), 5798 qemu_opt_get(sn_opts, SNAPSHOT_OPT_ID), 5799 qemu_opt_get(sn_opts, SNAPSHOT_OPT_NAME), 5800 &local_err); 5801 } else if (snapshot_name != NULL) { 5802 bdrv_snapshot_load_tmp_by_id_or_name(blk_bs(in_blk), 5803 snapshot_name, &local_err); 5804 } 5805 if (local_err) { 5806 error_reportf_err(local_err, "Failed to load snapshot: "); 5807 goto out; 5808 } 5809 } 5810 5811 drv = bdrv_find_format(out_fmt); 5812 if (!drv) { 5813 error_report("Unknown file format '%s'", out_fmt); 5814 goto out; 5815 } 5816 if (!drv->create_opts) { 5817 error_report("Format driver '%s' does not support image creation", 5818 drv->format_name); 5819 goto out; 5820 } 5821 5822 create_opts = qemu_opts_append(create_opts, drv->create_opts); 5823 create_opts = qemu_opts_append(create_opts, bdrv_file.create_opts); 5824 opts = qemu_opts_create(create_opts, NULL, 0, &error_abort); 5825 if (options) { 5826 if (!qemu_opts_do_parse(opts, options, NULL, &local_err)) { 5827 error_report_err(local_err); 5828 error_report("Invalid options for file format '%s'", out_fmt); 5829 goto out; 5830 } 5831 } 5832 if (img_size != -1) { 5833 qemu_opt_set_number(opts, BLOCK_OPT_SIZE, img_size, &error_abort); 5834 } 5835 5836 info = bdrv_measure(drv, opts, in_blk ? blk_bs(in_blk) : NULL, &local_err); 5837 if (local_err) { 5838 error_report_err(local_err); 5839 goto out; 5840 } 5841 5842 if (output_format == OFORMAT_HUMAN) { 5843 printf("required size: %" PRIu64 "\n", info->required); 5844 printf("fully allocated size: %" PRIu64 "\n", info->fully_allocated); 5845 if (info->has_bitmaps) { 5846 printf("bitmaps size: %" PRIu64 "\n", info->bitmaps); 5847 } 5848 } else { 5849 dump_json_block_measure_info(info); 5850 } 5851 5852 ret = 0; 5853 5854 out: 5855 qapi_free_BlockMeasureInfo(info); 5856 qemu_opts_del(object_opts); 5857 qemu_opts_del(opts); 5858 qemu_opts_del(sn_opts); 5859 qemu_opts_free(create_opts); 5860 g_free(options); 5861 blk_unref(in_blk); 5862 return ret; 5863 } 5864 5865 static const img_cmd_t img_cmds[] = { 5866 { "amend", img_amend, 5867 "Update format-specific options of the image" }, 5868 { "bench", img_bench, 5869 "Run a simple image benchmark" }, 5870 { "bitmap", img_bitmap, 5871 "Perform modifications of the persistent bitmap in the image" }, 5872 { "check", img_check, 5873 "Check basic image integrity" }, 5874 { "commit", img_commit, 5875 "Commit image to its backing file" }, 5876 { "compare", img_compare, 5877 "Check if two images have the same contents" }, 5878 { "convert", img_convert, 5879 "Copy one or more images to another with optional format conversion" }, 5880 { "create", img_create, 5881 "Create and format a new image file" }, 5882 { "dd", img_dd, 5883 "Copy input to output with optional format conversion" }, 5884 { "info", img_info, 5885 "Display information about the image" }, 5886 { "map", img_map, 5887 "Dump image metadata" }, 5888 { "measure", img_measure, 5889 "Calculate the file size required for a new image" }, 5890 { "rebase", img_rebase, 5891 "Change the backing file of the image" }, 5892 { "resize", img_resize, 5893 "Resize the image" }, 5894 { "snapshot", img_snapshot, 5895 "List or manipulate snapshots in the image" }, 5896 { NULL, NULL, }, 5897 }; 5898 5899 static void format_print(void *opaque, const char *name) 5900 { 5901 int *np = opaque; 5902 if (*np + strlen(name) > 75) { 5903 printf("\n "); 5904 *np = 1; 5905 } 5906 *np += printf(" %s", name); 5907 } 5908 5909 int main(int argc, char **argv) 5910 { 5911 const img_cmd_t *cmd; 5912 const char *cmdname; 5913 int c; 5914 static const struct option long_options[] = { 5915 {"help", no_argument, 0, 'h'}, 5916 {"version", no_argument, 0, 'V'}, 5917 {"trace", required_argument, NULL, 'T'}, 5918 {0, 0, 0, 0} 5919 }; 5920 5921 #ifdef CONFIG_POSIX 5922 signal(SIGPIPE, SIG_IGN); 5923 #endif 5924 5925 socket_init(); 5926 error_init(argv[0]); 5927 module_call_init(MODULE_INIT_TRACE); 5928 qemu_init_exec_dir(argv[0]); 5929 5930 qemu_init_main_loop(&error_fatal); 5931 5932 qcrypto_init(&error_fatal); 5933 5934 module_call_init(MODULE_INIT_QOM); 5935 bdrv_init(); 5936 5937 qemu_add_opts(&qemu_source_opts); 5938 qemu_add_opts(&qemu_trace_opts); 5939 5940 while ((c = getopt_long(argc, argv, "+hVT:", long_options, NULL)) != -1) { 5941 switch (c) { 5942 case 'h': 5943 printf( 5944 QEMU_IMG_VERSION 5945 "QEMU disk image utility. Usage:\n" 5946 "\n" 5947 " qemu-img [standard options] COMMAND [--help | command options]\n" 5948 "\n" 5949 "Standard options:\n" 5950 " -h, --help\n" 5951 " display this help and exit\n" 5952 " -V, --version\n" 5953 " display version info and exit\n" 5954 " -T,--trace TRACE\n" 5955 " specify tracing options:\n" 5956 " [[enable=]<pattern>][,events=<file>][,file=<file>]\n" 5957 "\n" 5958 "Recognized commands (run qemu-img COMMAND --help for command-specific help):\n\n"); 5959 for (cmd = img_cmds; cmd->name != NULL; cmd++) { 5960 printf(" %s - %s\n", cmd->name, cmd->description); 5961 } 5962 printf("\nSupported image formats:\n"); 5963 c = 99; /* force a newline */ 5964 bdrv_iterate_format(format_print, &c, false); 5965 if (c) { 5966 printf("\n"); 5967 } 5968 printf("\n" QEMU_HELP_BOTTOM "\n"); 5969 return 0; 5970 case 'V': 5971 printf(QEMU_IMG_VERSION); 5972 return 0; 5973 case 'T': 5974 trace_opt_parse(optarg); 5975 break; 5976 default: 5977 tryhelp(argv[0]); 5978 } 5979 } 5980 5981 if (optind >= argc) { 5982 error_exit(argv[0], "Not enough arguments"); 5983 } 5984 5985 cmdname = argv[optind]; 5986 5987 if (!trace_init_backends()) { 5988 exit(1); 5989 } 5990 trace_init_file(); 5991 qemu_set_log(LOG_TRACE, &error_fatal); 5992 5993 /* find the command */ 5994 for (cmd = img_cmds; cmd->name != NULL; cmd++) { 5995 if (!strcmp(cmdname, cmd->name)) { 5996 g_autofree char *argv0 = g_strdup_printf("%s %s", argv[0], cmdname); 5997 /* reset options and getopt processing (incl return order) */ 5998 argv += optind; 5999 argc -= optind; 6000 qemu_reset_optind(); 6001 argv[0] = argv0; 6002 return cmd->handler(cmd, argc, argv); 6003 } 6004 } 6005 6006 /* not found */ 6007 error_exit(argv[0], "Command not found: %s", cmdname); 6008 } 6009