1 /* 2 * QEMU disk image utility 3 * 4 * Copyright (c) 2003-2008 Fabrice Bellard 5 * 6 * Permission is hereby granted, free of charge, to any person obtaining a copy 7 * of this software and associated documentation files (the "Software"), to deal 8 * in the Software without restriction, including without limitation the rights 9 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell 10 * copies of the Software, and to permit persons to whom the Software is 11 * furnished to do so, subject to the following conditions: 12 * 13 * The above copyright notice and this permission notice shall be included in 14 * all copies or substantial portions of the Software. 15 * 16 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR 17 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, 18 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL 19 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER 20 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, 21 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN 22 * THE SOFTWARE. 23 */ 24 #include "qemu/osdep.h" 25 #include "qemu-version.h" 26 #include "qapi/error.h" 27 #include "qapi-visit.h" 28 #include "qapi/qobject-output-visitor.h" 29 #include "qapi/qmp/qerror.h" 30 #include "qapi/qmp/qjson.h" 31 #include "qemu/cutils.h" 32 #include "qemu/config-file.h" 33 #include "qemu/option.h" 34 #include "qemu/error-report.h" 35 #include "qemu/log.h" 36 #include "qom/object_interfaces.h" 37 #include "sysemu/sysemu.h" 38 #include "sysemu/block-backend.h" 39 #include "block/block_int.h" 40 #include "block/blockjob.h" 41 #include "block/qapi.h" 42 #include "crypto/init.h" 43 #include "trace/control.h" 44 #include <getopt.h> 45 46 #define QEMU_IMG_VERSION "qemu-img version " QEMU_VERSION QEMU_PKGVERSION \ 47 "\n" QEMU_COPYRIGHT "\n" 48 49 typedef struct img_cmd_t { 50 const char *name; 51 int (*handler)(int argc, char **argv); 52 } img_cmd_t; 53 54 enum { 55 OPTION_OUTPUT = 256, 56 OPTION_BACKING_CHAIN = 257, 57 OPTION_OBJECT = 258, 58 OPTION_IMAGE_OPTS = 259, 59 OPTION_PATTERN = 260, 60 OPTION_FLUSH_INTERVAL = 261, 61 OPTION_NO_DRAIN = 262, 62 }; 63 64 typedef enum OutputFormat { 65 OFORMAT_JSON, 66 OFORMAT_HUMAN, 67 } OutputFormat; 68 69 /* Default to cache=writeback as data integrity is not important for qemu-img */ 70 #define BDRV_DEFAULT_CACHE "writeback" 71 72 static void format_print(void *opaque, const char *name) 73 { 74 printf(" %s", name); 75 } 76 77 static void QEMU_NORETURN GCC_FMT_ATTR(1, 2) error_exit(const char *fmt, ...) 78 { 79 va_list ap; 80 81 error_printf("qemu-img: "); 82 83 va_start(ap, fmt); 84 error_vprintf(fmt, ap); 85 va_end(ap); 86 87 error_printf("\nTry 'qemu-img --help' for more information\n"); 88 exit(EXIT_FAILURE); 89 } 90 91 static void QEMU_NORETURN missing_argument(const char *option) 92 { 93 error_exit("missing argument for option '%s'", option); 94 } 95 96 static void QEMU_NORETURN unrecognized_option(const char *option) 97 { 98 error_exit("unrecognized option '%s'", option); 99 } 100 101 /* Please keep in synch with qemu-img.texi */ 102 static void QEMU_NORETURN help(void) 103 { 104 const char *help_msg = 105 QEMU_IMG_VERSION 106 "usage: qemu-img [standard options] command [command options]\n" 107 "QEMU disk image utility\n" 108 "\n" 109 " '-h', '--help' display this help and exit\n" 110 " '-V', '--version' output version information and exit\n" 111 " '-T', '--trace' [[enable=]<pattern>][,events=<file>][,file=<file>]\n" 112 " specify tracing options\n" 113 "\n" 114 "Command syntax:\n" 115 #define DEF(option, callback, arg_string) \ 116 " " arg_string "\n" 117 #include "qemu-img-cmds.h" 118 #undef DEF 119 #undef GEN_DOCS 120 "\n" 121 "Command parameters:\n" 122 " 'filename' is a disk image filename\n" 123 " 'objectdef' is a QEMU user creatable object definition. See the qemu(1)\n" 124 " manual page for a description of the object properties. The most common\n" 125 " object type is a 'secret', which is used to supply passwords and/or\n" 126 " encryption keys.\n" 127 " 'fmt' is the disk image format. It is guessed automatically in most cases\n" 128 " 'cache' is the cache mode used to write the output disk image, the valid\n" 129 " options are: 'none', 'writeback' (default, except for convert), 'writethrough',\n" 130 " 'directsync' and 'unsafe' (default for convert)\n" 131 " 'src_cache' is the cache mode used to read input disk images, the valid\n" 132 " options are the same as for the 'cache' option\n" 133 " 'size' is the disk image size in bytes. Optional suffixes\n" 134 " 'k' or 'K' (kilobyte, 1024), 'M' (megabyte, 1024k), 'G' (gigabyte, 1024M),\n" 135 " 'T' (terabyte, 1024G), 'P' (petabyte, 1024T) and 'E' (exabyte, 1024P) are\n" 136 " supported. 'b' is ignored.\n" 137 " 'output_filename' is the destination disk image filename\n" 138 " 'output_fmt' is the destination format\n" 139 " 'options' is a comma separated list of format specific options in a\n" 140 " name=value format. Use -o ? for an overview of the options supported by the\n" 141 " used format\n" 142 " 'snapshot_param' is param used for internal snapshot, format\n" 143 " is 'snapshot.id=[ID],snapshot.name=[NAME]', or\n" 144 " '[ID_OR_NAME]'\n" 145 " 'snapshot_id_or_name' is deprecated, use 'snapshot_param'\n" 146 " instead\n" 147 " '-c' indicates that target image must be compressed (qcow format only)\n" 148 " '-u' enables unsafe rebasing. It is assumed that old and new backing file\n" 149 " match exactly. The image doesn't need a working backing file before\n" 150 " rebasing in this case (useful for renaming the backing file)\n" 151 " '-h' with or without a command shows this help and lists the supported formats\n" 152 " '-p' show progress of command (only certain commands)\n" 153 " '-q' use Quiet mode - do not print any output (except errors)\n" 154 " '-S' indicates the consecutive number of bytes (defaults to 4k) that must\n" 155 " contain only zeros for qemu-img to create a sparse image during\n" 156 " conversion. If the number of bytes is 0, the source will not be scanned for\n" 157 " unallocated or zero sectors, and the destination image will always be\n" 158 " fully allocated\n" 159 " '--output' takes the format in which the output must be done (human or json)\n" 160 " '-n' skips the target volume creation (useful if the volume is created\n" 161 " prior to running qemu-img)\n" 162 "\n" 163 "Parameters to check subcommand:\n" 164 " '-r' tries to repair any inconsistencies that are found during the check.\n" 165 " '-r leaks' repairs only cluster leaks, whereas '-r all' fixes all\n" 166 " kinds of errors, with a higher risk of choosing the wrong fix or\n" 167 " hiding corruption that has already occurred.\n" 168 "\n" 169 "Parameters to convert subcommand:\n" 170 " '-m' specifies how many coroutines work in parallel during the convert\n" 171 " process (defaults to 8)\n" 172 " '-W' allow to write to the target out of order rather than sequential\n" 173 "\n" 174 "Parameters to snapshot subcommand:\n" 175 " 'snapshot' is the name of the snapshot to create, apply or delete\n" 176 " '-a' applies a snapshot (revert disk to saved state)\n" 177 " '-c' creates a snapshot\n" 178 " '-d' deletes a snapshot\n" 179 " '-l' lists all snapshots in the given image\n" 180 "\n" 181 "Parameters to compare subcommand:\n" 182 " '-f' first image format\n" 183 " '-F' second image format\n" 184 " '-s' run in Strict mode - fail on different image size or sector allocation\n" 185 "\n" 186 "Parameters to dd subcommand:\n" 187 " 'bs=BYTES' read and write up to BYTES bytes at a time " 188 "(default: 512)\n" 189 " 'count=N' copy only N input blocks\n" 190 " 'if=FILE' read from FILE\n" 191 " 'of=FILE' write to FILE\n" 192 " 'skip=N' skip N bs-sized blocks at the start of input\n"; 193 194 printf("%s\nSupported formats:", help_msg); 195 bdrv_iterate_format(format_print, NULL); 196 printf("\n"); 197 exit(EXIT_SUCCESS); 198 } 199 200 static QemuOptsList qemu_object_opts = { 201 .name = "object", 202 .implied_opt_name = "qom-type", 203 .head = QTAILQ_HEAD_INITIALIZER(qemu_object_opts.head), 204 .desc = { 205 { } 206 }, 207 }; 208 209 static QemuOptsList qemu_source_opts = { 210 .name = "source", 211 .implied_opt_name = "file", 212 .head = QTAILQ_HEAD_INITIALIZER(qemu_source_opts.head), 213 .desc = { 214 { } 215 }, 216 }; 217 218 static int GCC_FMT_ATTR(2, 3) qprintf(bool quiet, const char *fmt, ...) 219 { 220 int ret = 0; 221 if (!quiet) { 222 va_list args; 223 va_start(args, fmt); 224 ret = vprintf(fmt, args); 225 va_end(args); 226 } 227 return ret; 228 } 229 230 231 static int print_block_option_help(const char *filename, const char *fmt) 232 { 233 BlockDriver *drv, *proto_drv; 234 QemuOptsList *create_opts = NULL; 235 Error *local_err = NULL; 236 237 /* Find driver and parse its options */ 238 drv = bdrv_find_format(fmt); 239 if (!drv) { 240 error_report("Unknown file format '%s'", fmt); 241 return 1; 242 } 243 244 create_opts = qemu_opts_append(create_opts, drv->create_opts); 245 if (filename) { 246 proto_drv = bdrv_find_protocol(filename, true, &local_err); 247 if (!proto_drv) { 248 error_report_err(local_err); 249 qemu_opts_free(create_opts); 250 return 1; 251 } 252 create_opts = qemu_opts_append(create_opts, proto_drv->create_opts); 253 } 254 255 qemu_opts_print_help(create_opts); 256 qemu_opts_free(create_opts); 257 return 0; 258 } 259 260 261 static int img_open_password(BlockBackend *blk, const char *filename, 262 int flags, bool quiet) 263 { 264 BlockDriverState *bs; 265 char password[256]; 266 267 bs = blk_bs(blk); 268 if (bdrv_is_encrypted(bs) && bdrv_key_required(bs) && 269 !(flags & BDRV_O_NO_IO)) { 270 qprintf(quiet, "Disk image '%s' is encrypted.\n", filename); 271 if (qemu_read_password(password, sizeof(password)) < 0) { 272 error_report("No password given"); 273 return -1; 274 } 275 if (bdrv_set_key(bs, password) < 0) { 276 error_report("invalid password"); 277 return -1; 278 } 279 } 280 return 0; 281 } 282 283 284 static BlockBackend *img_open_opts(const char *optstr, 285 QemuOpts *opts, int flags, bool writethrough, 286 bool quiet) 287 { 288 QDict *options; 289 Error *local_err = NULL; 290 BlockBackend *blk; 291 options = qemu_opts_to_qdict(opts, NULL); 292 blk = blk_new_open(NULL, NULL, options, flags, &local_err); 293 if (!blk) { 294 error_reportf_err(local_err, "Could not open '%s': ", optstr); 295 return NULL; 296 } 297 blk_set_enable_write_cache(blk, !writethrough); 298 299 if (img_open_password(blk, optstr, flags, quiet) < 0) { 300 blk_unref(blk); 301 return NULL; 302 } 303 return blk; 304 } 305 306 static BlockBackend *img_open_file(const char *filename, 307 const char *fmt, int flags, 308 bool writethrough, bool quiet) 309 { 310 BlockBackend *blk; 311 Error *local_err = NULL; 312 QDict *options = NULL; 313 314 if (fmt) { 315 options = qdict_new(); 316 qdict_put(options, "driver", qstring_from_str(fmt)); 317 } 318 319 blk = blk_new_open(filename, NULL, options, flags, &local_err); 320 if (!blk) { 321 error_reportf_err(local_err, "Could not open '%s': ", filename); 322 return NULL; 323 } 324 blk_set_enable_write_cache(blk, !writethrough); 325 326 if (img_open_password(blk, filename, flags, quiet) < 0) { 327 blk_unref(blk); 328 return NULL; 329 } 330 return blk; 331 } 332 333 334 static BlockBackend *img_open(bool image_opts, 335 const char *filename, 336 const char *fmt, int flags, bool writethrough, 337 bool quiet) 338 { 339 BlockBackend *blk; 340 if (image_opts) { 341 QemuOpts *opts; 342 if (fmt) { 343 error_report("--image-opts and --format are mutually exclusive"); 344 return NULL; 345 } 346 opts = qemu_opts_parse_noisily(qemu_find_opts("source"), 347 filename, true); 348 if (!opts) { 349 return NULL; 350 } 351 blk = img_open_opts(filename, opts, flags, writethrough, quiet); 352 } else { 353 blk = img_open_file(filename, fmt, flags, writethrough, quiet); 354 } 355 return blk; 356 } 357 358 359 static int add_old_style_options(const char *fmt, QemuOpts *opts, 360 const char *base_filename, 361 const char *base_fmt) 362 { 363 Error *err = NULL; 364 365 if (base_filename) { 366 qemu_opt_set(opts, BLOCK_OPT_BACKING_FILE, base_filename, &err); 367 if (err) { 368 error_report("Backing file not supported for file format '%s'", 369 fmt); 370 error_free(err); 371 return -1; 372 } 373 } 374 if (base_fmt) { 375 qemu_opt_set(opts, BLOCK_OPT_BACKING_FMT, base_fmt, &err); 376 if (err) { 377 error_report("Backing file format not supported for file " 378 "format '%s'", fmt); 379 error_free(err); 380 return -1; 381 } 382 } 383 return 0; 384 } 385 386 static int64_t cvtnum(const char *s) 387 { 388 int err; 389 uint64_t value; 390 391 err = qemu_strtosz(s, NULL, &value); 392 if (err < 0) { 393 return err; 394 } 395 if (value > INT64_MAX) { 396 return -ERANGE; 397 } 398 return value; 399 } 400 401 static int img_create(int argc, char **argv) 402 { 403 int c; 404 uint64_t img_size = -1; 405 const char *fmt = "raw"; 406 const char *base_fmt = NULL; 407 const char *filename; 408 const char *base_filename = NULL; 409 char *options = NULL; 410 Error *local_err = NULL; 411 bool quiet = false; 412 413 for(;;) { 414 static const struct option long_options[] = { 415 {"help", no_argument, 0, 'h'}, 416 {"object", required_argument, 0, OPTION_OBJECT}, 417 {0, 0, 0, 0} 418 }; 419 c = getopt_long(argc, argv, ":F:b:f:he6o:q", 420 long_options, NULL); 421 if (c == -1) { 422 break; 423 } 424 switch(c) { 425 case ':': 426 missing_argument(argv[optind - 1]); 427 break; 428 case '?': 429 unrecognized_option(argv[optind - 1]); 430 break; 431 case 'h': 432 help(); 433 break; 434 case 'F': 435 base_fmt = optarg; 436 break; 437 case 'b': 438 base_filename = optarg; 439 break; 440 case 'f': 441 fmt = optarg; 442 break; 443 case 'e': 444 error_report("option -e is deprecated, please use \'-o " 445 "encryption\' instead!"); 446 goto fail; 447 case '6': 448 error_report("option -6 is deprecated, please use \'-o " 449 "compat6\' instead!"); 450 goto fail; 451 case 'o': 452 if (!is_valid_option_list(optarg)) { 453 error_report("Invalid option list: %s", optarg); 454 goto fail; 455 } 456 if (!options) { 457 options = g_strdup(optarg); 458 } else { 459 char *old_options = options; 460 options = g_strdup_printf("%s,%s", options, optarg); 461 g_free(old_options); 462 } 463 break; 464 case 'q': 465 quiet = true; 466 break; 467 case OPTION_OBJECT: { 468 QemuOpts *opts; 469 opts = qemu_opts_parse_noisily(&qemu_object_opts, 470 optarg, true); 471 if (!opts) { 472 goto fail; 473 } 474 } break; 475 } 476 } 477 478 /* Get the filename */ 479 filename = (optind < argc) ? argv[optind] : NULL; 480 if (options && has_help_option(options)) { 481 g_free(options); 482 return print_block_option_help(filename, fmt); 483 } 484 485 if (optind >= argc) { 486 error_exit("Expecting image file name"); 487 } 488 optind++; 489 490 if (qemu_opts_foreach(&qemu_object_opts, 491 user_creatable_add_opts_foreach, 492 NULL, NULL)) { 493 goto fail; 494 } 495 496 /* Get image size, if specified */ 497 if (optind < argc) { 498 int64_t sval; 499 500 sval = cvtnum(argv[optind++]); 501 if (sval < 0) { 502 if (sval == -ERANGE) { 503 error_report("Image size must be less than 8 EiB!"); 504 } else { 505 error_report("Invalid image size specified! You may use k, M, " 506 "G, T, P or E suffixes for "); 507 error_report("kilobytes, megabytes, gigabytes, terabytes, " 508 "petabytes and exabytes."); 509 } 510 goto fail; 511 } 512 img_size = (uint64_t)sval; 513 } 514 if (optind != argc) { 515 error_exit("Unexpected argument: %s", argv[optind]); 516 } 517 518 bdrv_img_create(filename, fmt, base_filename, base_fmt, 519 options, img_size, 0, &local_err, quiet); 520 if (local_err) { 521 error_reportf_err(local_err, "%s: ", filename); 522 goto fail; 523 } 524 525 g_free(options); 526 return 0; 527 528 fail: 529 g_free(options); 530 return 1; 531 } 532 533 static void dump_json_image_check(ImageCheck *check, bool quiet) 534 { 535 QString *str; 536 QObject *obj; 537 Visitor *v = qobject_output_visitor_new(&obj); 538 539 visit_type_ImageCheck(v, NULL, &check, &error_abort); 540 visit_complete(v, &obj); 541 str = qobject_to_json_pretty(obj); 542 assert(str != NULL); 543 qprintf(quiet, "%s\n", qstring_get_str(str)); 544 qobject_decref(obj); 545 visit_free(v); 546 QDECREF(str); 547 } 548 549 static void dump_human_image_check(ImageCheck *check, bool quiet) 550 { 551 if (!(check->corruptions || check->leaks || check->check_errors)) { 552 qprintf(quiet, "No errors were found on the image.\n"); 553 } else { 554 if (check->corruptions) { 555 qprintf(quiet, "\n%" PRId64 " errors were found on the image.\n" 556 "Data may be corrupted, or further writes to the image " 557 "may corrupt it.\n", 558 check->corruptions); 559 } 560 561 if (check->leaks) { 562 qprintf(quiet, 563 "\n%" PRId64 " leaked clusters were found on the image.\n" 564 "This means waste of disk space, but no harm to data.\n", 565 check->leaks); 566 } 567 568 if (check->check_errors) { 569 qprintf(quiet, 570 "\n%" PRId64 571 " internal errors have occurred during the check.\n", 572 check->check_errors); 573 } 574 } 575 576 if (check->total_clusters != 0 && check->allocated_clusters != 0) { 577 qprintf(quiet, "%" PRId64 "/%" PRId64 " = %0.2f%% allocated, " 578 "%0.2f%% fragmented, %0.2f%% compressed clusters\n", 579 check->allocated_clusters, check->total_clusters, 580 check->allocated_clusters * 100.0 / check->total_clusters, 581 check->fragmented_clusters * 100.0 / check->allocated_clusters, 582 check->compressed_clusters * 100.0 / 583 check->allocated_clusters); 584 } 585 586 if (check->image_end_offset) { 587 qprintf(quiet, 588 "Image end offset: %" PRId64 "\n", check->image_end_offset); 589 } 590 } 591 592 static int collect_image_check(BlockDriverState *bs, 593 ImageCheck *check, 594 const char *filename, 595 const char *fmt, 596 int fix) 597 { 598 int ret; 599 BdrvCheckResult result; 600 601 ret = bdrv_check(bs, &result, fix); 602 if (ret < 0) { 603 return ret; 604 } 605 606 check->filename = g_strdup(filename); 607 check->format = g_strdup(bdrv_get_format_name(bs)); 608 check->check_errors = result.check_errors; 609 check->corruptions = result.corruptions; 610 check->has_corruptions = result.corruptions != 0; 611 check->leaks = result.leaks; 612 check->has_leaks = result.leaks != 0; 613 check->corruptions_fixed = result.corruptions_fixed; 614 check->has_corruptions_fixed = result.corruptions != 0; 615 check->leaks_fixed = result.leaks_fixed; 616 check->has_leaks_fixed = result.leaks != 0; 617 check->image_end_offset = result.image_end_offset; 618 check->has_image_end_offset = result.image_end_offset != 0; 619 check->total_clusters = result.bfi.total_clusters; 620 check->has_total_clusters = result.bfi.total_clusters != 0; 621 check->allocated_clusters = result.bfi.allocated_clusters; 622 check->has_allocated_clusters = result.bfi.allocated_clusters != 0; 623 check->fragmented_clusters = result.bfi.fragmented_clusters; 624 check->has_fragmented_clusters = result.bfi.fragmented_clusters != 0; 625 check->compressed_clusters = result.bfi.compressed_clusters; 626 check->has_compressed_clusters = result.bfi.compressed_clusters != 0; 627 628 return 0; 629 } 630 631 /* 632 * Checks an image for consistency. Exit codes: 633 * 634 * 0 - Check completed, image is good 635 * 1 - Check not completed because of internal errors 636 * 2 - Check completed, image is corrupted 637 * 3 - Check completed, image has leaked clusters, but is good otherwise 638 * 63 - Checks are not supported by the image format 639 */ 640 static int img_check(int argc, char **argv) 641 { 642 int c, ret; 643 OutputFormat output_format = OFORMAT_HUMAN; 644 const char *filename, *fmt, *output, *cache; 645 BlockBackend *blk; 646 BlockDriverState *bs; 647 int fix = 0; 648 int flags = BDRV_O_CHECK; 649 bool writethrough; 650 ImageCheck *check; 651 bool quiet = false; 652 bool image_opts = false; 653 654 fmt = NULL; 655 output = NULL; 656 cache = BDRV_DEFAULT_CACHE; 657 658 for(;;) { 659 int option_index = 0; 660 static const struct option long_options[] = { 661 {"help", no_argument, 0, 'h'}, 662 {"format", required_argument, 0, 'f'}, 663 {"repair", required_argument, 0, 'r'}, 664 {"output", required_argument, 0, OPTION_OUTPUT}, 665 {"object", required_argument, 0, OPTION_OBJECT}, 666 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS}, 667 {0, 0, 0, 0} 668 }; 669 c = getopt_long(argc, argv, ":hf:r:T:q", 670 long_options, &option_index); 671 if (c == -1) { 672 break; 673 } 674 switch(c) { 675 case ':': 676 missing_argument(argv[optind - 1]); 677 break; 678 case '?': 679 unrecognized_option(argv[optind - 1]); 680 break; 681 case 'h': 682 help(); 683 break; 684 case 'f': 685 fmt = optarg; 686 break; 687 case 'r': 688 flags |= BDRV_O_RDWR; 689 690 if (!strcmp(optarg, "leaks")) { 691 fix = BDRV_FIX_LEAKS; 692 } else if (!strcmp(optarg, "all")) { 693 fix = BDRV_FIX_LEAKS | BDRV_FIX_ERRORS; 694 } else { 695 error_exit("Unknown option value for -r " 696 "(expecting 'leaks' or 'all'): %s", optarg); 697 } 698 break; 699 case OPTION_OUTPUT: 700 output = optarg; 701 break; 702 case 'T': 703 cache = optarg; 704 break; 705 case 'q': 706 quiet = true; 707 break; 708 case OPTION_OBJECT: { 709 QemuOpts *opts; 710 opts = qemu_opts_parse_noisily(&qemu_object_opts, 711 optarg, true); 712 if (!opts) { 713 return 1; 714 } 715 } break; 716 case OPTION_IMAGE_OPTS: 717 image_opts = true; 718 break; 719 } 720 } 721 if (optind != argc - 1) { 722 error_exit("Expecting one image file name"); 723 } 724 filename = argv[optind++]; 725 726 if (output && !strcmp(output, "json")) { 727 output_format = OFORMAT_JSON; 728 } else if (output && !strcmp(output, "human")) { 729 output_format = OFORMAT_HUMAN; 730 } else if (output) { 731 error_report("--output must be used with human or json as argument."); 732 return 1; 733 } 734 735 if (qemu_opts_foreach(&qemu_object_opts, 736 user_creatable_add_opts_foreach, 737 NULL, NULL)) { 738 return 1; 739 } 740 741 ret = bdrv_parse_cache_mode(cache, &flags, &writethrough); 742 if (ret < 0) { 743 error_report("Invalid source cache option: %s", cache); 744 return 1; 745 } 746 747 blk = img_open(image_opts, filename, fmt, flags, writethrough, quiet); 748 if (!blk) { 749 return 1; 750 } 751 bs = blk_bs(blk); 752 753 check = g_new0(ImageCheck, 1); 754 ret = collect_image_check(bs, check, filename, fmt, fix); 755 756 if (ret == -ENOTSUP) { 757 error_report("This image format does not support checks"); 758 ret = 63; 759 goto fail; 760 } 761 762 if (check->corruptions_fixed || check->leaks_fixed) { 763 int corruptions_fixed, leaks_fixed; 764 765 leaks_fixed = check->leaks_fixed; 766 corruptions_fixed = check->corruptions_fixed; 767 768 if (output_format == OFORMAT_HUMAN) { 769 qprintf(quiet, 770 "The following inconsistencies were found and repaired:\n\n" 771 " %" PRId64 " leaked clusters\n" 772 " %" PRId64 " corruptions\n\n" 773 "Double checking the fixed image now...\n", 774 check->leaks_fixed, 775 check->corruptions_fixed); 776 } 777 778 ret = collect_image_check(bs, check, filename, fmt, 0); 779 780 check->leaks_fixed = leaks_fixed; 781 check->corruptions_fixed = corruptions_fixed; 782 } 783 784 if (!ret) { 785 switch (output_format) { 786 case OFORMAT_HUMAN: 787 dump_human_image_check(check, quiet); 788 break; 789 case OFORMAT_JSON: 790 dump_json_image_check(check, quiet); 791 break; 792 } 793 } 794 795 if (ret || check->check_errors) { 796 if (ret) { 797 error_report("Check failed: %s", strerror(-ret)); 798 } else { 799 error_report("Check failed"); 800 } 801 ret = 1; 802 goto fail; 803 } 804 805 if (check->corruptions) { 806 ret = 2; 807 } else if (check->leaks) { 808 ret = 3; 809 } else { 810 ret = 0; 811 } 812 813 fail: 814 qapi_free_ImageCheck(check); 815 blk_unref(blk); 816 return ret; 817 } 818 819 typedef struct CommonBlockJobCBInfo { 820 BlockDriverState *bs; 821 Error **errp; 822 } CommonBlockJobCBInfo; 823 824 static void common_block_job_cb(void *opaque, int ret) 825 { 826 CommonBlockJobCBInfo *cbi = opaque; 827 828 if (ret < 0) { 829 error_setg_errno(cbi->errp, -ret, "Block job failed"); 830 } 831 } 832 833 static void run_block_job(BlockJob *job, Error **errp) 834 { 835 AioContext *aio_context = blk_get_aio_context(job->blk); 836 837 /* FIXME In error cases, the job simply goes away and we access a dangling 838 * pointer below. */ 839 aio_context_acquire(aio_context); 840 do { 841 aio_poll(aio_context, true); 842 qemu_progress_print(job->len ? 843 ((float)job->offset / job->len * 100.f) : 0.0f, 0); 844 } while (!job->ready); 845 846 block_job_complete_sync(job, errp); 847 aio_context_release(aio_context); 848 849 /* A block job may finish instantaneously without publishing any progress, 850 * so just signal completion here */ 851 qemu_progress_print(100.f, 0); 852 } 853 854 static int img_commit(int argc, char **argv) 855 { 856 int c, ret, flags; 857 const char *filename, *fmt, *cache, *base; 858 BlockBackend *blk; 859 BlockDriverState *bs, *base_bs; 860 BlockJob *job; 861 bool progress = false, quiet = false, drop = false; 862 bool writethrough; 863 Error *local_err = NULL; 864 CommonBlockJobCBInfo cbi; 865 bool image_opts = false; 866 AioContext *aio_context; 867 868 fmt = NULL; 869 cache = BDRV_DEFAULT_CACHE; 870 base = NULL; 871 for(;;) { 872 static const struct option long_options[] = { 873 {"help", no_argument, 0, 'h'}, 874 {"object", required_argument, 0, OPTION_OBJECT}, 875 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS}, 876 {0, 0, 0, 0} 877 }; 878 c = getopt_long(argc, argv, ":f:ht:b:dpq", 879 long_options, NULL); 880 if (c == -1) { 881 break; 882 } 883 switch(c) { 884 case ':': 885 missing_argument(argv[optind - 1]); 886 break; 887 case '?': 888 unrecognized_option(argv[optind - 1]); 889 break; 890 case 'h': 891 help(); 892 break; 893 case 'f': 894 fmt = optarg; 895 break; 896 case 't': 897 cache = optarg; 898 break; 899 case 'b': 900 base = optarg; 901 /* -b implies -d */ 902 drop = true; 903 break; 904 case 'd': 905 drop = true; 906 break; 907 case 'p': 908 progress = true; 909 break; 910 case 'q': 911 quiet = true; 912 break; 913 case OPTION_OBJECT: { 914 QemuOpts *opts; 915 opts = qemu_opts_parse_noisily(&qemu_object_opts, 916 optarg, true); 917 if (!opts) { 918 return 1; 919 } 920 } break; 921 case OPTION_IMAGE_OPTS: 922 image_opts = true; 923 break; 924 } 925 } 926 927 /* Progress is not shown in Quiet mode */ 928 if (quiet) { 929 progress = false; 930 } 931 932 if (optind != argc - 1) { 933 error_exit("Expecting one image file name"); 934 } 935 filename = argv[optind++]; 936 937 if (qemu_opts_foreach(&qemu_object_opts, 938 user_creatable_add_opts_foreach, 939 NULL, NULL)) { 940 return 1; 941 } 942 943 flags = BDRV_O_RDWR | BDRV_O_UNMAP; 944 ret = bdrv_parse_cache_mode(cache, &flags, &writethrough); 945 if (ret < 0) { 946 error_report("Invalid cache option: %s", cache); 947 return 1; 948 } 949 950 blk = img_open(image_opts, filename, fmt, flags, writethrough, quiet); 951 if (!blk) { 952 return 1; 953 } 954 bs = blk_bs(blk); 955 956 qemu_progress_init(progress, 1.f); 957 qemu_progress_print(0.f, 100); 958 959 if (base) { 960 base_bs = bdrv_find_backing_image(bs, base); 961 if (!base_bs) { 962 error_setg(&local_err, 963 "Did not find '%s' in the backing chain of '%s'", 964 base, filename); 965 goto done; 966 } 967 } else { 968 /* This is different from QMP, which by default uses the deepest file in 969 * the backing chain (i.e., the very base); however, the traditional 970 * behavior of qemu-img commit is using the immediate backing file. */ 971 base_bs = backing_bs(bs); 972 if (!base_bs) { 973 error_setg(&local_err, "Image does not have a backing file"); 974 goto done; 975 } 976 } 977 978 cbi = (CommonBlockJobCBInfo){ 979 .errp = &local_err, 980 .bs = bs, 981 }; 982 983 aio_context = bdrv_get_aio_context(bs); 984 aio_context_acquire(aio_context); 985 commit_active_start("commit", bs, base_bs, BLOCK_JOB_DEFAULT, 0, 986 BLOCKDEV_ON_ERROR_REPORT, NULL, common_block_job_cb, 987 &cbi, &local_err, false); 988 aio_context_release(aio_context); 989 if (local_err) { 990 goto done; 991 } 992 993 /* When the block job completes, the BlockBackend reference will point to 994 * the old backing file. In order to avoid that the top image is already 995 * deleted, so we can still empty it afterwards, increment the reference 996 * counter here preemptively. */ 997 if (!drop) { 998 bdrv_ref(bs); 999 } 1000 1001 job = block_job_get("commit"); 1002 run_block_job(job, &local_err); 1003 if (local_err) { 1004 goto unref_backing; 1005 } 1006 1007 if (!drop && bs->drv->bdrv_make_empty) { 1008 ret = bs->drv->bdrv_make_empty(bs); 1009 if (ret) { 1010 error_setg_errno(&local_err, -ret, "Could not empty %s", 1011 filename); 1012 goto unref_backing; 1013 } 1014 } 1015 1016 unref_backing: 1017 if (!drop) { 1018 bdrv_unref(bs); 1019 } 1020 1021 done: 1022 qemu_progress_end(); 1023 1024 blk_unref(blk); 1025 1026 if (local_err) { 1027 error_report_err(local_err); 1028 return 1; 1029 } 1030 1031 qprintf(quiet, "Image committed.\n"); 1032 return 0; 1033 } 1034 1035 /* 1036 * Returns true iff the first sector pointed to by 'buf' contains at least 1037 * a non-NUL byte. 1038 * 1039 * 'pnum' is set to the number of sectors (including and immediately following 1040 * the first one) that are known to be in the same allocated/unallocated state. 1041 */ 1042 static int is_allocated_sectors(const uint8_t *buf, int n, int *pnum) 1043 { 1044 bool is_zero; 1045 int i; 1046 1047 if (n <= 0) { 1048 *pnum = 0; 1049 return 0; 1050 } 1051 is_zero = buffer_is_zero(buf, 512); 1052 for(i = 1; i < n; i++) { 1053 buf += 512; 1054 if (is_zero != buffer_is_zero(buf, 512)) { 1055 break; 1056 } 1057 } 1058 *pnum = i; 1059 return !is_zero; 1060 } 1061 1062 /* 1063 * Like is_allocated_sectors, but if the buffer starts with a used sector, 1064 * up to 'min' consecutive sectors containing zeros are ignored. This avoids 1065 * breaking up write requests for only small sparse areas. 1066 */ 1067 static int is_allocated_sectors_min(const uint8_t *buf, int n, int *pnum, 1068 int min) 1069 { 1070 int ret; 1071 int num_checked, num_used; 1072 1073 if (n < min) { 1074 min = n; 1075 } 1076 1077 ret = is_allocated_sectors(buf, n, pnum); 1078 if (!ret) { 1079 return ret; 1080 } 1081 1082 num_used = *pnum; 1083 buf += BDRV_SECTOR_SIZE * *pnum; 1084 n -= *pnum; 1085 num_checked = num_used; 1086 1087 while (n > 0) { 1088 ret = is_allocated_sectors(buf, n, pnum); 1089 1090 buf += BDRV_SECTOR_SIZE * *pnum; 1091 n -= *pnum; 1092 num_checked += *pnum; 1093 if (ret) { 1094 num_used = num_checked; 1095 } else if (*pnum >= min) { 1096 break; 1097 } 1098 } 1099 1100 *pnum = num_used; 1101 return 1; 1102 } 1103 1104 /* 1105 * Compares two buffers sector by sector. Returns 0 if the first sector of both 1106 * buffers matches, non-zero otherwise. 1107 * 1108 * pnum is set to the number of sectors (including and immediately following 1109 * the first one) that are known to have the same comparison result 1110 */ 1111 static int compare_sectors(const uint8_t *buf1, const uint8_t *buf2, int n, 1112 int *pnum) 1113 { 1114 bool res; 1115 int i; 1116 1117 if (n <= 0) { 1118 *pnum = 0; 1119 return 0; 1120 } 1121 1122 res = !!memcmp(buf1, buf2, 512); 1123 for(i = 1; i < n; i++) { 1124 buf1 += 512; 1125 buf2 += 512; 1126 1127 if (!!memcmp(buf1, buf2, 512) != res) { 1128 break; 1129 } 1130 } 1131 1132 *pnum = i; 1133 return res; 1134 } 1135 1136 #define IO_BUF_SIZE (2 * 1024 * 1024) 1137 1138 static int64_t sectors_to_bytes(int64_t sectors) 1139 { 1140 return sectors << BDRV_SECTOR_BITS; 1141 } 1142 1143 static int64_t sectors_to_process(int64_t total, int64_t from) 1144 { 1145 return MIN(total - from, IO_BUF_SIZE >> BDRV_SECTOR_BITS); 1146 } 1147 1148 /* 1149 * Check if passed sectors are empty (not allocated or contain only 0 bytes) 1150 * 1151 * Returns 0 in case sectors are filled with 0, 1 if sectors contain non-zero 1152 * data and negative value on error. 1153 * 1154 * @param blk: BlockBackend for the image 1155 * @param sect_num: Number of first sector to check 1156 * @param sect_count: Number of sectors to check 1157 * @param filename: Name of disk file we are checking (logging purpose) 1158 * @param buffer: Allocated buffer for storing read data 1159 * @param quiet: Flag for quiet mode 1160 */ 1161 static int check_empty_sectors(BlockBackend *blk, int64_t sect_num, 1162 int sect_count, const char *filename, 1163 uint8_t *buffer, bool quiet) 1164 { 1165 int pnum, ret = 0; 1166 ret = blk_pread(blk, sect_num << BDRV_SECTOR_BITS, buffer, 1167 sect_count << BDRV_SECTOR_BITS); 1168 if (ret < 0) { 1169 error_report("Error while reading offset %" PRId64 " of %s: %s", 1170 sectors_to_bytes(sect_num), filename, strerror(-ret)); 1171 return ret; 1172 } 1173 ret = is_allocated_sectors(buffer, sect_count, &pnum); 1174 if (ret || pnum != sect_count) { 1175 qprintf(quiet, "Content mismatch at offset %" PRId64 "!\n", 1176 sectors_to_bytes(ret ? sect_num : sect_num + pnum)); 1177 return 1; 1178 } 1179 1180 return 0; 1181 } 1182 1183 /* 1184 * Compares two images. Exit codes: 1185 * 1186 * 0 - Images are identical 1187 * 1 - Images differ 1188 * >1 - Error occurred 1189 */ 1190 static int img_compare(int argc, char **argv) 1191 { 1192 const char *fmt1 = NULL, *fmt2 = NULL, *cache, *filename1, *filename2; 1193 BlockBackend *blk1, *blk2; 1194 BlockDriverState *bs1, *bs2; 1195 int64_t total_sectors1, total_sectors2; 1196 uint8_t *buf1 = NULL, *buf2 = NULL; 1197 int pnum1, pnum2; 1198 int allocated1, allocated2; 1199 int ret = 0; /* return value - 0 Ident, 1 Different, >1 Error */ 1200 bool progress = false, quiet = false, strict = false; 1201 int flags; 1202 bool writethrough; 1203 int64_t total_sectors; 1204 int64_t sector_num = 0; 1205 int64_t nb_sectors; 1206 int c, pnum; 1207 uint64_t progress_base; 1208 bool image_opts = false; 1209 1210 cache = BDRV_DEFAULT_CACHE; 1211 for (;;) { 1212 static const struct option long_options[] = { 1213 {"help", no_argument, 0, 'h'}, 1214 {"object", required_argument, 0, OPTION_OBJECT}, 1215 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS}, 1216 {0, 0, 0, 0} 1217 }; 1218 c = getopt_long(argc, argv, ":hf:F:T:pqs", 1219 long_options, NULL); 1220 if (c == -1) { 1221 break; 1222 } 1223 switch (c) { 1224 case ':': 1225 missing_argument(argv[optind - 1]); 1226 break; 1227 case '?': 1228 unrecognized_option(argv[optind - 1]); 1229 break; 1230 case 'h': 1231 help(); 1232 break; 1233 case 'f': 1234 fmt1 = optarg; 1235 break; 1236 case 'F': 1237 fmt2 = optarg; 1238 break; 1239 case 'T': 1240 cache = optarg; 1241 break; 1242 case 'p': 1243 progress = true; 1244 break; 1245 case 'q': 1246 quiet = true; 1247 break; 1248 case 's': 1249 strict = true; 1250 break; 1251 case OPTION_OBJECT: { 1252 QemuOpts *opts; 1253 opts = qemu_opts_parse_noisily(&qemu_object_opts, 1254 optarg, true); 1255 if (!opts) { 1256 ret = 2; 1257 goto out4; 1258 } 1259 } break; 1260 case OPTION_IMAGE_OPTS: 1261 image_opts = true; 1262 break; 1263 } 1264 } 1265 1266 /* Progress is not shown in Quiet mode */ 1267 if (quiet) { 1268 progress = false; 1269 } 1270 1271 1272 if (optind != argc - 2) { 1273 error_exit("Expecting two image file names"); 1274 } 1275 filename1 = argv[optind++]; 1276 filename2 = argv[optind++]; 1277 1278 if (qemu_opts_foreach(&qemu_object_opts, 1279 user_creatable_add_opts_foreach, 1280 NULL, NULL)) { 1281 ret = 2; 1282 goto out4; 1283 } 1284 1285 /* Initialize before goto out */ 1286 qemu_progress_init(progress, 2.0); 1287 1288 flags = 0; 1289 ret = bdrv_parse_cache_mode(cache, &flags, &writethrough); 1290 if (ret < 0) { 1291 error_report("Invalid source cache option: %s", cache); 1292 ret = 2; 1293 goto out3; 1294 } 1295 1296 blk1 = img_open(image_opts, filename1, fmt1, flags, writethrough, quiet); 1297 if (!blk1) { 1298 ret = 2; 1299 goto out3; 1300 } 1301 1302 blk2 = img_open(image_opts, filename2, fmt2, flags, writethrough, quiet); 1303 if (!blk2) { 1304 ret = 2; 1305 goto out2; 1306 } 1307 bs1 = blk_bs(blk1); 1308 bs2 = blk_bs(blk2); 1309 1310 buf1 = blk_blockalign(blk1, IO_BUF_SIZE); 1311 buf2 = blk_blockalign(blk2, IO_BUF_SIZE); 1312 total_sectors1 = blk_nb_sectors(blk1); 1313 if (total_sectors1 < 0) { 1314 error_report("Can't get size of %s: %s", 1315 filename1, strerror(-total_sectors1)); 1316 ret = 4; 1317 goto out; 1318 } 1319 total_sectors2 = blk_nb_sectors(blk2); 1320 if (total_sectors2 < 0) { 1321 error_report("Can't get size of %s: %s", 1322 filename2, strerror(-total_sectors2)); 1323 ret = 4; 1324 goto out; 1325 } 1326 total_sectors = MIN(total_sectors1, total_sectors2); 1327 progress_base = MAX(total_sectors1, total_sectors2); 1328 1329 qemu_progress_print(0, 100); 1330 1331 if (strict && total_sectors1 != total_sectors2) { 1332 ret = 1; 1333 qprintf(quiet, "Strict mode: Image size mismatch!\n"); 1334 goto out; 1335 } 1336 1337 for (;;) { 1338 int64_t status1, status2; 1339 BlockDriverState *file; 1340 1341 nb_sectors = sectors_to_process(total_sectors, sector_num); 1342 if (nb_sectors <= 0) { 1343 break; 1344 } 1345 status1 = bdrv_get_block_status_above(bs1, NULL, sector_num, 1346 total_sectors1 - sector_num, 1347 &pnum1, &file); 1348 if (status1 < 0) { 1349 ret = 3; 1350 error_report("Sector allocation test failed for %s", filename1); 1351 goto out; 1352 } 1353 allocated1 = status1 & BDRV_BLOCK_ALLOCATED; 1354 1355 status2 = bdrv_get_block_status_above(bs2, NULL, sector_num, 1356 total_sectors2 - sector_num, 1357 &pnum2, &file); 1358 if (status2 < 0) { 1359 ret = 3; 1360 error_report("Sector allocation test failed for %s", filename2); 1361 goto out; 1362 } 1363 allocated2 = status2 & BDRV_BLOCK_ALLOCATED; 1364 if (pnum1) { 1365 nb_sectors = MIN(nb_sectors, pnum1); 1366 } 1367 if (pnum2) { 1368 nb_sectors = MIN(nb_sectors, pnum2); 1369 } 1370 1371 if (strict) { 1372 if ((status1 & ~BDRV_BLOCK_OFFSET_MASK) != 1373 (status2 & ~BDRV_BLOCK_OFFSET_MASK)) { 1374 ret = 1; 1375 qprintf(quiet, "Strict mode: Offset %" PRId64 1376 " block status mismatch!\n", 1377 sectors_to_bytes(sector_num)); 1378 goto out; 1379 } 1380 } 1381 if ((status1 & BDRV_BLOCK_ZERO) && (status2 & BDRV_BLOCK_ZERO)) { 1382 nb_sectors = MIN(pnum1, pnum2); 1383 } else if (allocated1 == allocated2) { 1384 if (allocated1) { 1385 ret = blk_pread(blk1, sector_num << BDRV_SECTOR_BITS, buf1, 1386 nb_sectors << BDRV_SECTOR_BITS); 1387 if (ret < 0) { 1388 error_report("Error while reading offset %" PRId64 " of %s:" 1389 " %s", sectors_to_bytes(sector_num), filename1, 1390 strerror(-ret)); 1391 ret = 4; 1392 goto out; 1393 } 1394 ret = blk_pread(blk2, sector_num << BDRV_SECTOR_BITS, buf2, 1395 nb_sectors << BDRV_SECTOR_BITS); 1396 if (ret < 0) { 1397 error_report("Error while reading offset %" PRId64 1398 " of %s: %s", sectors_to_bytes(sector_num), 1399 filename2, strerror(-ret)); 1400 ret = 4; 1401 goto out; 1402 } 1403 ret = compare_sectors(buf1, buf2, nb_sectors, &pnum); 1404 if (ret || pnum != nb_sectors) { 1405 qprintf(quiet, "Content mismatch at offset %" PRId64 "!\n", 1406 sectors_to_bytes( 1407 ret ? sector_num : sector_num + pnum)); 1408 ret = 1; 1409 goto out; 1410 } 1411 } 1412 } else { 1413 1414 if (allocated1) { 1415 ret = check_empty_sectors(blk1, sector_num, nb_sectors, 1416 filename1, buf1, quiet); 1417 } else { 1418 ret = check_empty_sectors(blk2, sector_num, nb_sectors, 1419 filename2, buf1, quiet); 1420 } 1421 if (ret) { 1422 if (ret < 0) { 1423 error_report("Error while reading offset %" PRId64 ": %s", 1424 sectors_to_bytes(sector_num), strerror(-ret)); 1425 ret = 4; 1426 } 1427 goto out; 1428 } 1429 } 1430 sector_num += nb_sectors; 1431 qemu_progress_print(((float) nb_sectors / progress_base)*100, 100); 1432 } 1433 1434 if (total_sectors1 != total_sectors2) { 1435 BlockBackend *blk_over; 1436 int64_t total_sectors_over; 1437 const char *filename_over; 1438 1439 qprintf(quiet, "Warning: Image size mismatch!\n"); 1440 if (total_sectors1 > total_sectors2) { 1441 total_sectors_over = total_sectors1; 1442 blk_over = blk1; 1443 filename_over = filename1; 1444 } else { 1445 total_sectors_over = total_sectors2; 1446 blk_over = blk2; 1447 filename_over = filename2; 1448 } 1449 1450 for (;;) { 1451 nb_sectors = sectors_to_process(total_sectors_over, sector_num); 1452 if (nb_sectors <= 0) { 1453 break; 1454 } 1455 ret = bdrv_is_allocated_above(blk_bs(blk_over), NULL, sector_num, 1456 nb_sectors, &pnum); 1457 if (ret < 0) { 1458 ret = 3; 1459 error_report("Sector allocation test failed for %s", 1460 filename_over); 1461 goto out; 1462 1463 } 1464 nb_sectors = pnum; 1465 if (ret) { 1466 ret = check_empty_sectors(blk_over, sector_num, nb_sectors, 1467 filename_over, buf1, quiet); 1468 if (ret) { 1469 if (ret < 0) { 1470 error_report("Error while reading offset %" PRId64 1471 " of %s: %s", sectors_to_bytes(sector_num), 1472 filename_over, strerror(-ret)); 1473 ret = 4; 1474 } 1475 goto out; 1476 } 1477 } 1478 sector_num += nb_sectors; 1479 qemu_progress_print(((float) nb_sectors / progress_base)*100, 100); 1480 } 1481 } 1482 1483 qprintf(quiet, "Images are identical.\n"); 1484 ret = 0; 1485 1486 out: 1487 qemu_vfree(buf1); 1488 qemu_vfree(buf2); 1489 blk_unref(blk2); 1490 out2: 1491 blk_unref(blk1); 1492 out3: 1493 qemu_progress_end(); 1494 out4: 1495 return ret; 1496 } 1497 1498 enum ImgConvertBlockStatus { 1499 BLK_DATA, 1500 BLK_ZERO, 1501 BLK_BACKING_FILE, 1502 }; 1503 1504 #define MAX_COROUTINES 16 1505 1506 typedef struct ImgConvertState { 1507 BlockBackend **src; 1508 int64_t *src_sectors; 1509 int src_num; 1510 int64_t total_sectors; 1511 int64_t allocated_sectors; 1512 int64_t allocated_done; 1513 int64_t sector_num; 1514 int64_t wr_offs; 1515 enum ImgConvertBlockStatus status; 1516 int64_t sector_next_status; 1517 BlockBackend *target; 1518 bool has_zero_init; 1519 bool compressed; 1520 bool target_has_backing; 1521 bool wr_in_order; 1522 int min_sparse; 1523 size_t cluster_sectors; 1524 size_t buf_sectors; 1525 int num_coroutines; 1526 int running_coroutines; 1527 Coroutine *co[MAX_COROUTINES]; 1528 int64_t wait_sector_num[MAX_COROUTINES]; 1529 CoMutex lock; 1530 int ret; 1531 } ImgConvertState; 1532 1533 static void convert_select_part(ImgConvertState *s, int64_t sector_num, 1534 int *src_cur, int64_t *src_cur_offset) 1535 { 1536 *src_cur = 0; 1537 *src_cur_offset = 0; 1538 while (sector_num - *src_cur_offset >= s->src_sectors[*src_cur]) { 1539 *src_cur_offset += s->src_sectors[*src_cur]; 1540 (*src_cur)++; 1541 assert(*src_cur < s->src_num); 1542 } 1543 } 1544 1545 static int convert_iteration_sectors(ImgConvertState *s, int64_t sector_num) 1546 { 1547 int64_t ret, src_cur_offset; 1548 int n, src_cur; 1549 1550 convert_select_part(s, sector_num, &src_cur, &src_cur_offset); 1551 1552 assert(s->total_sectors > sector_num); 1553 n = MIN(s->total_sectors - sector_num, BDRV_REQUEST_MAX_SECTORS); 1554 1555 if (s->sector_next_status <= sector_num) { 1556 BlockDriverState *file; 1557 ret = bdrv_get_block_status(blk_bs(s->src[src_cur]), 1558 sector_num - src_cur_offset, 1559 n, &n, &file); 1560 if (ret < 0) { 1561 return ret; 1562 } 1563 1564 if (ret & BDRV_BLOCK_ZERO) { 1565 s->status = BLK_ZERO; 1566 } else if (ret & BDRV_BLOCK_DATA) { 1567 s->status = BLK_DATA; 1568 } else if (!s->target_has_backing) { 1569 /* Without a target backing file we must copy over the contents of 1570 * the backing file as well. */ 1571 /* Check block status of the backing file chain to avoid 1572 * needlessly reading zeroes and limiting the iteration to the 1573 * buffer size */ 1574 ret = bdrv_get_block_status_above(blk_bs(s->src[src_cur]), NULL, 1575 sector_num - src_cur_offset, 1576 n, &n, &file); 1577 if (ret < 0) { 1578 return ret; 1579 } 1580 1581 if (ret & BDRV_BLOCK_ZERO) { 1582 s->status = BLK_ZERO; 1583 } else { 1584 s->status = BLK_DATA; 1585 } 1586 } else { 1587 s->status = BLK_BACKING_FILE; 1588 } 1589 1590 s->sector_next_status = sector_num + n; 1591 } 1592 1593 n = MIN(n, s->sector_next_status - sector_num); 1594 if (s->status == BLK_DATA) { 1595 n = MIN(n, s->buf_sectors); 1596 } 1597 1598 /* We need to write complete clusters for compressed images, so if an 1599 * unallocated area is shorter than that, we must consider the whole 1600 * cluster allocated. */ 1601 if (s->compressed) { 1602 if (n < s->cluster_sectors) { 1603 n = MIN(s->cluster_sectors, s->total_sectors - sector_num); 1604 s->status = BLK_DATA; 1605 } else { 1606 n = QEMU_ALIGN_DOWN(n, s->cluster_sectors); 1607 } 1608 } 1609 1610 return n; 1611 } 1612 1613 static int coroutine_fn convert_co_read(ImgConvertState *s, int64_t sector_num, 1614 int nb_sectors, uint8_t *buf) 1615 { 1616 int n, ret; 1617 QEMUIOVector qiov; 1618 struct iovec iov; 1619 1620 assert(nb_sectors <= s->buf_sectors); 1621 while (nb_sectors > 0) { 1622 BlockBackend *blk; 1623 int src_cur; 1624 int64_t bs_sectors, src_cur_offset; 1625 1626 /* In the case of compression with multiple source files, we can get a 1627 * nb_sectors that spreads into the next part. So we must be able to 1628 * read across multiple BDSes for one convert_read() call. */ 1629 convert_select_part(s, sector_num, &src_cur, &src_cur_offset); 1630 blk = s->src[src_cur]; 1631 bs_sectors = s->src_sectors[src_cur]; 1632 1633 n = MIN(nb_sectors, bs_sectors - (sector_num - src_cur_offset)); 1634 iov.iov_base = buf; 1635 iov.iov_len = n << BDRV_SECTOR_BITS; 1636 qemu_iovec_init_external(&qiov, &iov, 1); 1637 1638 ret = blk_co_preadv( 1639 blk, (sector_num - src_cur_offset) << BDRV_SECTOR_BITS, 1640 n << BDRV_SECTOR_BITS, &qiov, 0); 1641 if (ret < 0) { 1642 return ret; 1643 } 1644 1645 sector_num += n; 1646 nb_sectors -= n; 1647 buf += n * BDRV_SECTOR_SIZE; 1648 } 1649 1650 return 0; 1651 } 1652 1653 1654 static int coroutine_fn convert_co_write(ImgConvertState *s, int64_t sector_num, 1655 int nb_sectors, uint8_t *buf, 1656 enum ImgConvertBlockStatus status) 1657 { 1658 int ret; 1659 QEMUIOVector qiov; 1660 struct iovec iov; 1661 1662 while (nb_sectors > 0) { 1663 int n = nb_sectors; 1664 switch (status) { 1665 case BLK_BACKING_FILE: 1666 /* If we have a backing file, leave clusters unallocated that are 1667 * unallocated in the source image, so that the backing file is 1668 * visible at the respective offset. */ 1669 assert(s->target_has_backing); 1670 break; 1671 1672 case BLK_DATA: 1673 /* We must always write compressed clusters as a whole, so don't 1674 * try to find zeroed parts in the buffer. We can only save the 1675 * write if the buffer is completely zeroed and we're allowed to 1676 * keep the target sparse. */ 1677 if (s->compressed) { 1678 if (s->has_zero_init && s->min_sparse && 1679 buffer_is_zero(buf, n * BDRV_SECTOR_SIZE)) 1680 { 1681 assert(!s->target_has_backing); 1682 break; 1683 } 1684 1685 iov.iov_base = buf; 1686 iov.iov_len = n << BDRV_SECTOR_BITS; 1687 qemu_iovec_init_external(&qiov, &iov, 1); 1688 1689 ret = blk_co_pwritev(s->target, sector_num << BDRV_SECTOR_BITS, 1690 n << BDRV_SECTOR_BITS, &qiov, 1691 BDRV_REQ_WRITE_COMPRESSED); 1692 if (ret < 0) { 1693 return ret; 1694 } 1695 break; 1696 } 1697 1698 /* If there is real non-zero data or we're told to keep the target 1699 * fully allocated (-S 0), we must write it. Otherwise we can treat 1700 * it as zero sectors. */ 1701 if (!s->min_sparse || 1702 is_allocated_sectors_min(buf, n, &n, s->min_sparse)) 1703 { 1704 iov.iov_base = buf; 1705 iov.iov_len = n << BDRV_SECTOR_BITS; 1706 qemu_iovec_init_external(&qiov, &iov, 1); 1707 1708 ret = blk_co_pwritev(s->target, sector_num << BDRV_SECTOR_BITS, 1709 n << BDRV_SECTOR_BITS, &qiov, 0); 1710 if (ret < 0) { 1711 return ret; 1712 } 1713 break; 1714 } 1715 /* fall-through */ 1716 1717 case BLK_ZERO: 1718 if (s->has_zero_init) { 1719 break; 1720 } 1721 ret = blk_co_pwrite_zeroes(s->target, 1722 sector_num << BDRV_SECTOR_BITS, 1723 n << BDRV_SECTOR_BITS, 0); 1724 if (ret < 0) { 1725 return ret; 1726 } 1727 break; 1728 } 1729 1730 sector_num += n; 1731 nb_sectors -= n; 1732 buf += n * BDRV_SECTOR_SIZE; 1733 } 1734 1735 return 0; 1736 } 1737 1738 static void coroutine_fn convert_co_do_copy(void *opaque) 1739 { 1740 ImgConvertState *s = opaque; 1741 uint8_t *buf = NULL; 1742 int ret, i; 1743 int index = -1; 1744 1745 for (i = 0; i < s->num_coroutines; i++) { 1746 if (s->co[i] == qemu_coroutine_self()) { 1747 index = i; 1748 break; 1749 } 1750 } 1751 assert(index >= 0); 1752 1753 s->running_coroutines++; 1754 buf = blk_blockalign(s->target, s->buf_sectors * BDRV_SECTOR_SIZE); 1755 1756 while (1) { 1757 int n; 1758 int64_t sector_num; 1759 enum ImgConvertBlockStatus status; 1760 1761 qemu_co_mutex_lock(&s->lock); 1762 if (s->ret != -EINPROGRESS || s->sector_num >= s->total_sectors) { 1763 qemu_co_mutex_unlock(&s->lock); 1764 goto out; 1765 } 1766 n = convert_iteration_sectors(s, s->sector_num); 1767 if (n < 0) { 1768 qemu_co_mutex_unlock(&s->lock); 1769 s->ret = n; 1770 goto out; 1771 } 1772 /* save current sector and allocation status to local variables */ 1773 sector_num = s->sector_num; 1774 status = s->status; 1775 if (!s->min_sparse && s->status == BLK_ZERO) { 1776 n = MIN(n, s->buf_sectors); 1777 } 1778 /* increment global sector counter so that other coroutines can 1779 * already continue reading beyond this request */ 1780 s->sector_num += n; 1781 qemu_co_mutex_unlock(&s->lock); 1782 1783 if (status == BLK_DATA || (!s->min_sparse && status == BLK_ZERO)) { 1784 s->allocated_done += n; 1785 qemu_progress_print(100.0 * s->allocated_done / 1786 s->allocated_sectors, 0); 1787 } 1788 1789 if (status == BLK_DATA) { 1790 ret = convert_co_read(s, sector_num, n, buf); 1791 if (ret < 0) { 1792 error_report("error while reading sector %" PRId64 1793 ": %s", sector_num, strerror(-ret)); 1794 s->ret = ret; 1795 goto out; 1796 } 1797 } else if (!s->min_sparse && status == BLK_ZERO) { 1798 status = BLK_DATA; 1799 memset(buf, 0x00, n * BDRV_SECTOR_SIZE); 1800 } 1801 1802 if (s->wr_in_order) { 1803 /* keep writes in order */ 1804 while (s->wr_offs != sector_num) { 1805 if (s->ret != -EINPROGRESS) { 1806 goto out; 1807 } 1808 s->wait_sector_num[index] = sector_num; 1809 qemu_coroutine_yield(); 1810 } 1811 s->wait_sector_num[index] = -1; 1812 } 1813 1814 ret = convert_co_write(s, sector_num, n, buf, status); 1815 if (ret < 0) { 1816 error_report("error while writing sector %" PRId64 1817 ": %s", sector_num, strerror(-ret)); 1818 s->ret = ret; 1819 goto out; 1820 } 1821 1822 if (s->wr_in_order) { 1823 /* reenter the coroutine that might have waited 1824 * for this write to complete */ 1825 s->wr_offs = sector_num + n; 1826 for (i = 0; i < s->num_coroutines; i++) { 1827 if (s->co[i] && s->wait_sector_num[i] == s->wr_offs) { 1828 /* 1829 * A -> B -> A cannot occur because A has 1830 * s->wait_sector_num[i] == -1 during A -> B. Therefore 1831 * B will never enter A during this time window. 1832 */ 1833 qemu_coroutine_enter(s->co[i]); 1834 break; 1835 } 1836 } 1837 } 1838 } 1839 1840 out: 1841 qemu_vfree(buf); 1842 s->co[index] = NULL; 1843 s->running_coroutines--; 1844 if (!s->running_coroutines && s->ret == -EINPROGRESS) { 1845 /* the convert job finished successfully */ 1846 s->ret = 0; 1847 } 1848 } 1849 1850 static int convert_do_copy(ImgConvertState *s) 1851 { 1852 int ret, i, n; 1853 int64_t sector_num = 0; 1854 1855 /* Check whether we have zero initialisation or can get it efficiently */ 1856 s->has_zero_init = s->min_sparse && !s->target_has_backing 1857 ? bdrv_has_zero_init(blk_bs(s->target)) 1858 : false; 1859 1860 if (!s->has_zero_init && !s->target_has_backing && 1861 bdrv_can_write_zeroes_with_unmap(blk_bs(s->target))) 1862 { 1863 ret = blk_make_zero(s->target, BDRV_REQ_MAY_UNMAP); 1864 if (ret == 0) { 1865 s->has_zero_init = true; 1866 } 1867 } 1868 1869 /* Allocate buffer for copied data. For compressed images, only one cluster 1870 * can be copied at a time. */ 1871 if (s->compressed) { 1872 if (s->cluster_sectors <= 0 || s->cluster_sectors > s->buf_sectors) { 1873 error_report("invalid cluster size"); 1874 return -EINVAL; 1875 } 1876 s->buf_sectors = s->cluster_sectors; 1877 } 1878 1879 while (sector_num < s->total_sectors) { 1880 n = convert_iteration_sectors(s, sector_num); 1881 if (n < 0) { 1882 return n; 1883 } 1884 if (s->status == BLK_DATA || (!s->min_sparse && s->status == BLK_ZERO)) 1885 { 1886 s->allocated_sectors += n; 1887 } 1888 sector_num += n; 1889 } 1890 1891 /* Do the copy */ 1892 s->sector_next_status = 0; 1893 s->ret = -EINPROGRESS; 1894 1895 qemu_co_mutex_init(&s->lock); 1896 for (i = 0; i < s->num_coroutines; i++) { 1897 s->co[i] = qemu_coroutine_create(convert_co_do_copy, s); 1898 s->wait_sector_num[i] = -1; 1899 qemu_coroutine_enter(s->co[i]); 1900 } 1901 1902 while (s->ret == -EINPROGRESS) { 1903 main_loop_wait(false); 1904 } 1905 1906 if (s->compressed && !s->ret) { 1907 /* signal EOF to align */ 1908 ret = blk_pwrite_compressed(s->target, 0, NULL, 0); 1909 if (ret < 0) { 1910 return ret; 1911 } 1912 } 1913 1914 return s->ret; 1915 } 1916 1917 static int img_convert(int argc, char **argv) 1918 { 1919 int c, bs_n, bs_i, compress, cluster_sectors, skip_create; 1920 int64_t ret = 0; 1921 int progress = 0, flags, src_flags; 1922 bool writethrough, src_writethrough; 1923 const char *fmt, *out_fmt, *cache, *src_cache, *out_baseimg, *out_filename; 1924 BlockDriver *drv, *proto_drv; 1925 BlockBackend **blk = NULL, *out_blk = NULL; 1926 BlockDriverState **bs = NULL, *out_bs = NULL; 1927 int64_t total_sectors; 1928 int64_t *bs_sectors = NULL; 1929 size_t bufsectors = IO_BUF_SIZE / BDRV_SECTOR_SIZE; 1930 BlockDriverInfo bdi; 1931 QemuOpts *opts = NULL; 1932 QemuOptsList *create_opts = NULL; 1933 const char *out_baseimg_param; 1934 char *options = NULL; 1935 const char *snapshot_name = NULL; 1936 int min_sparse = 8; /* Need at least 4k of zeros for sparse detection */ 1937 bool quiet = false; 1938 Error *local_err = NULL; 1939 QemuOpts *sn_opts = NULL; 1940 ImgConvertState state; 1941 bool image_opts = false; 1942 bool wr_in_order = true; 1943 long num_coroutines = 8; 1944 1945 fmt = NULL; 1946 out_fmt = "raw"; 1947 cache = "unsafe"; 1948 src_cache = BDRV_DEFAULT_CACHE; 1949 out_baseimg = NULL; 1950 compress = 0; 1951 skip_create = 0; 1952 for(;;) { 1953 static const struct option long_options[] = { 1954 {"help", no_argument, 0, 'h'}, 1955 {"object", required_argument, 0, OPTION_OBJECT}, 1956 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS}, 1957 {0, 0, 0, 0} 1958 }; 1959 c = getopt_long(argc, argv, ":hf:O:B:ce6o:s:l:S:pt:T:qnm:W", 1960 long_options, NULL); 1961 if (c == -1) { 1962 break; 1963 } 1964 switch(c) { 1965 case ':': 1966 missing_argument(argv[optind - 1]); 1967 break; 1968 case '?': 1969 unrecognized_option(argv[optind - 1]); 1970 break; 1971 case 'h': 1972 help(); 1973 break; 1974 case 'f': 1975 fmt = optarg; 1976 break; 1977 case 'O': 1978 out_fmt = optarg; 1979 break; 1980 case 'B': 1981 out_baseimg = optarg; 1982 break; 1983 case 'c': 1984 compress = 1; 1985 break; 1986 case 'e': 1987 error_report("option -e is deprecated, please use \'-o " 1988 "encryption\' instead!"); 1989 ret = -1; 1990 goto fail_getopt; 1991 case '6': 1992 error_report("option -6 is deprecated, please use \'-o " 1993 "compat6\' instead!"); 1994 ret = -1; 1995 goto fail_getopt; 1996 case 'o': 1997 if (!is_valid_option_list(optarg)) { 1998 error_report("Invalid option list: %s", optarg); 1999 ret = -1; 2000 goto fail_getopt; 2001 } 2002 if (!options) { 2003 options = g_strdup(optarg); 2004 } else { 2005 char *old_options = options; 2006 options = g_strdup_printf("%s,%s", options, optarg); 2007 g_free(old_options); 2008 } 2009 break; 2010 case 's': 2011 snapshot_name = optarg; 2012 break; 2013 case 'l': 2014 if (strstart(optarg, SNAPSHOT_OPT_BASE, NULL)) { 2015 sn_opts = qemu_opts_parse_noisily(&internal_snapshot_opts, 2016 optarg, false); 2017 if (!sn_opts) { 2018 error_report("Failed in parsing snapshot param '%s'", 2019 optarg); 2020 ret = -1; 2021 goto fail_getopt; 2022 } 2023 } else { 2024 snapshot_name = optarg; 2025 } 2026 break; 2027 case 'S': 2028 { 2029 int64_t sval; 2030 2031 sval = cvtnum(optarg); 2032 if (sval < 0) { 2033 error_report("Invalid minimum zero buffer size for sparse output specified"); 2034 ret = -1; 2035 goto fail_getopt; 2036 } 2037 2038 min_sparse = sval / BDRV_SECTOR_SIZE; 2039 break; 2040 } 2041 case 'p': 2042 progress = 1; 2043 break; 2044 case 't': 2045 cache = optarg; 2046 break; 2047 case 'T': 2048 src_cache = optarg; 2049 break; 2050 case 'q': 2051 quiet = true; 2052 break; 2053 case 'n': 2054 skip_create = 1; 2055 break; 2056 case 'm': 2057 if (qemu_strtol(optarg, NULL, 0, &num_coroutines) || 2058 num_coroutines < 1 || num_coroutines > MAX_COROUTINES) { 2059 error_report("Invalid number of coroutines. Allowed number of" 2060 " coroutines is between 1 and %d", MAX_COROUTINES); 2061 ret = -1; 2062 goto fail_getopt; 2063 } 2064 break; 2065 case 'W': 2066 wr_in_order = false; 2067 break; 2068 case OPTION_OBJECT: 2069 opts = qemu_opts_parse_noisily(&qemu_object_opts, 2070 optarg, true); 2071 if (!opts) { 2072 goto fail_getopt; 2073 } 2074 break; 2075 case OPTION_IMAGE_OPTS: 2076 image_opts = true; 2077 break; 2078 } 2079 } 2080 2081 if (qemu_opts_foreach(&qemu_object_opts, 2082 user_creatable_add_opts_foreach, 2083 NULL, NULL)) { 2084 goto fail_getopt; 2085 } 2086 2087 if (!wr_in_order && compress) { 2088 error_report("Out of order write and compress are mutually exclusive"); 2089 ret = -1; 2090 goto fail_getopt; 2091 } 2092 2093 /* Initialize before goto out */ 2094 if (quiet) { 2095 progress = 0; 2096 } 2097 qemu_progress_init(progress, 1.0); 2098 2099 bs_n = argc - optind - 1; 2100 out_filename = bs_n >= 1 ? argv[argc - 1] : NULL; 2101 2102 if (options && has_help_option(options)) { 2103 ret = print_block_option_help(out_filename, out_fmt); 2104 goto out; 2105 } 2106 2107 if (bs_n < 1) { 2108 error_exit("Must specify image file name"); 2109 } 2110 2111 2112 if (bs_n > 1 && out_baseimg) { 2113 error_report("-B makes no sense when concatenating multiple input " 2114 "images"); 2115 ret = -1; 2116 goto out; 2117 } 2118 2119 src_flags = 0; 2120 ret = bdrv_parse_cache_mode(src_cache, &src_flags, &src_writethrough); 2121 if (ret < 0) { 2122 error_report("Invalid source cache option: %s", src_cache); 2123 goto out; 2124 } 2125 2126 qemu_progress_print(0, 100); 2127 2128 blk = g_new0(BlockBackend *, bs_n); 2129 bs = g_new0(BlockDriverState *, bs_n); 2130 bs_sectors = g_new(int64_t, bs_n); 2131 2132 total_sectors = 0; 2133 for (bs_i = 0; bs_i < bs_n; bs_i++) { 2134 blk[bs_i] = img_open(image_opts, argv[optind + bs_i], 2135 fmt, src_flags, src_writethrough, quiet); 2136 if (!blk[bs_i]) { 2137 ret = -1; 2138 goto out; 2139 } 2140 bs[bs_i] = blk_bs(blk[bs_i]); 2141 bs_sectors[bs_i] = blk_nb_sectors(blk[bs_i]); 2142 if (bs_sectors[bs_i] < 0) { 2143 error_report("Could not get size of %s: %s", 2144 argv[optind + bs_i], strerror(-bs_sectors[bs_i])); 2145 ret = -1; 2146 goto out; 2147 } 2148 total_sectors += bs_sectors[bs_i]; 2149 } 2150 2151 if (sn_opts) { 2152 bdrv_snapshot_load_tmp(bs[0], 2153 qemu_opt_get(sn_opts, SNAPSHOT_OPT_ID), 2154 qemu_opt_get(sn_opts, SNAPSHOT_OPT_NAME), 2155 &local_err); 2156 } else if (snapshot_name != NULL) { 2157 if (bs_n > 1) { 2158 error_report("No support for concatenating multiple snapshot"); 2159 ret = -1; 2160 goto out; 2161 } 2162 2163 bdrv_snapshot_load_tmp_by_id_or_name(bs[0], snapshot_name, &local_err); 2164 } 2165 if (local_err) { 2166 error_reportf_err(local_err, "Failed to load snapshot: "); 2167 ret = -1; 2168 goto out; 2169 } 2170 2171 /* Find driver and parse its options */ 2172 drv = bdrv_find_format(out_fmt); 2173 if (!drv) { 2174 error_report("Unknown file format '%s'", out_fmt); 2175 ret = -1; 2176 goto out; 2177 } 2178 2179 proto_drv = bdrv_find_protocol(out_filename, true, &local_err); 2180 if (!proto_drv) { 2181 error_report_err(local_err); 2182 ret = -1; 2183 goto out; 2184 } 2185 2186 if (!skip_create) { 2187 if (!drv->create_opts) { 2188 error_report("Format driver '%s' does not support image creation", 2189 drv->format_name); 2190 ret = -1; 2191 goto out; 2192 } 2193 2194 if (!proto_drv->create_opts) { 2195 error_report("Protocol driver '%s' does not support image creation", 2196 proto_drv->format_name); 2197 ret = -1; 2198 goto out; 2199 } 2200 2201 create_opts = qemu_opts_append(create_opts, drv->create_opts); 2202 create_opts = qemu_opts_append(create_opts, proto_drv->create_opts); 2203 2204 opts = qemu_opts_create(create_opts, NULL, 0, &error_abort); 2205 if (options) { 2206 qemu_opts_do_parse(opts, options, NULL, &local_err); 2207 if (local_err) { 2208 error_report_err(local_err); 2209 ret = -1; 2210 goto out; 2211 } 2212 } 2213 2214 qemu_opt_set_number(opts, BLOCK_OPT_SIZE, total_sectors * 512, 2215 &error_abort); 2216 ret = add_old_style_options(out_fmt, opts, out_baseimg, NULL); 2217 if (ret < 0) { 2218 goto out; 2219 } 2220 } 2221 2222 /* Get backing file name if -o backing_file was used */ 2223 out_baseimg_param = qemu_opt_get(opts, BLOCK_OPT_BACKING_FILE); 2224 if (out_baseimg_param) { 2225 out_baseimg = out_baseimg_param; 2226 } 2227 2228 /* Check if compression is supported */ 2229 if (compress) { 2230 bool encryption = 2231 qemu_opt_get_bool(opts, BLOCK_OPT_ENCRYPT, false); 2232 const char *preallocation = 2233 qemu_opt_get(opts, BLOCK_OPT_PREALLOC); 2234 2235 if (!drv->bdrv_co_pwritev_compressed) { 2236 error_report("Compression not supported for this file format"); 2237 ret = -1; 2238 goto out; 2239 } 2240 2241 if (encryption) { 2242 error_report("Compression and encryption not supported at " 2243 "the same time"); 2244 ret = -1; 2245 goto out; 2246 } 2247 2248 if (preallocation 2249 && strcmp(preallocation, "off")) 2250 { 2251 error_report("Compression and preallocation not supported at " 2252 "the same time"); 2253 ret = -1; 2254 goto out; 2255 } 2256 } 2257 2258 if (!skip_create) { 2259 /* Create the new image */ 2260 ret = bdrv_create(drv, out_filename, opts, &local_err); 2261 if (ret < 0) { 2262 error_reportf_err(local_err, "%s: error while converting %s: ", 2263 out_filename, out_fmt); 2264 goto out; 2265 } 2266 } 2267 2268 flags = min_sparse ? (BDRV_O_RDWR | BDRV_O_UNMAP) : BDRV_O_RDWR; 2269 ret = bdrv_parse_cache_mode(cache, &flags, &writethrough); 2270 if (ret < 0) { 2271 error_report("Invalid cache option: %s", cache); 2272 goto out; 2273 } 2274 2275 /* XXX we should allow --image-opts to trigger use of 2276 * img_open() here, but then we have trouble with 2277 * the bdrv_create() call which takes different params. 2278 * Not critical right now, so fix can wait... 2279 */ 2280 out_blk = img_open_file(out_filename, out_fmt, flags, writethrough, quiet); 2281 if (!out_blk) { 2282 ret = -1; 2283 goto out; 2284 } 2285 out_bs = blk_bs(out_blk); 2286 2287 /* increase bufsectors from the default 4096 (2M) if opt_transfer 2288 * or discard_alignment of the out_bs is greater. Limit to 32768 (16MB) 2289 * as maximum. */ 2290 bufsectors = MIN(32768, 2291 MAX(bufsectors, 2292 MAX(out_bs->bl.opt_transfer >> BDRV_SECTOR_BITS, 2293 out_bs->bl.pdiscard_alignment >> 2294 BDRV_SECTOR_BITS))); 2295 2296 if (skip_create) { 2297 int64_t output_sectors = blk_nb_sectors(out_blk); 2298 if (output_sectors < 0) { 2299 error_report("unable to get output image length: %s", 2300 strerror(-output_sectors)); 2301 ret = -1; 2302 goto out; 2303 } else if (output_sectors < total_sectors) { 2304 error_report("output file is smaller than input file"); 2305 ret = -1; 2306 goto out; 2307 } 2308 } 2309 2310 cluster_sectors = 0; 2311 ret = bdrv_get_info(out_bs, &bdi); 2312 if (ret < 0) { 2313 if (compress) { 2314 error_report("could not get block driver info"); 2315 goto out; 2316 } 2317 } else { 2318 compress = compress || bdi.needs_compressed_writes; 2319 cluster_sectors = bdi.cluster_size / BDRV_SECTOR_SIZE; 2320 } 2321 2322 state = (ImgConvertState) { 2323 .src = blk, 2324 .src_sectors = bs_sectors, 2325 .src_num = bs_n, 2326 .total_sectors = total_sectors, 2327 .target = out_blk, 2328 .compressed = compress, 2329 .target_has_backing = (bool) out_baseimg, 2330 .min_sparse = min_sparse, 2331 .cluster_sectors = cluster_sectors, 2332 .buf_sectors = bufsectors, 2333 .wr_in_order = wr_in_order, 2334 .num_coroutines = num_coroutines, 2335 }; 2336 ret = convert_do_copy(&state); 2337 2338 out: 2339 if (!ret) { 2340 qemu_progress_print(100, 0); 2341 } 2342 qemu_progress_end(); 2343 qemu_opts_del(opts); 2344 qemu_opts_free(create_opts); 2345 qemu_opts_del(sn_opts); 2346 blk_unref(out_blk); 2347 g_free(bs); 2348 if (blk) { 2349 for (bs_i = 0; bs_i < bs_n; bs_i++) { 2350 blk_unref(blk[bs_i]); 2351 } 2352 g_free(blk); 2353 } 2354 g_free(bs_sectors); 2355 fail_getopt: 2356 g_free(options); 2357 2358 if (ret) { 2359 return 1; 2360 } 2361 return 0; 2362 } 2363 2364 2365 static void dump_snapshots(BlockDriverState *bs) 2366 { 2367 QEMUSnapshotInfo *sn_tab, *sn; 2368 int nb_sns, i; 2369 2370 nb_sns = bdrv_snapshot_list(bs, &sn_tab); 2371 if (nb_sns <= 0) 2372 return; 2373 printf("Snapshot list:\n"); 2374 bdrv_snapshot_dump(fprintf, stdout, NULL); 2375 printf("\n"); 2376 for(i = 0; i < nb_sns; i++) { 2377 sn = &sn_tab[i]; 2378 bdrv_snapshot_dump(fprintf, stdout, sn); 2379 printf("\n"); 2380 } 2381 g_free(sn_tab); 2382 } 2383 2384 static void dump_json_image_info_list(ImageInfoList *list) 2385 { 2386 QString *str; 2387 QObject *obj; 2388 Visitor *v = qobject_output_visitor_new(&obj); 2389 2390 visit_type_ImageInfoList(v, NULL, &list, &error_abort); 2391 visit_complete(v, &obj); 2392 str = qobject_to_json_pretty(obj); 2393 assert(str != NULL); 2394 printf("%s\n", qstring_get_str(str)); 2395 qobject_decref(obj); 2396 visit_free(v); 2397 QDECREF(str); 2398 } 2399 2400 static void dump_json_image_info(ImageInfo *info) 2401 { 2402 QString *str; 2403 QObject *obj; 2404 Visitor *v = qobject_output_visitor_new(&obj); 2405 2406 visit_type_ImageInfo(v, NULL, &info, &error_abort); 2407 visit_complete(v, &obj); 2408 str = qobject_to_json_pretty(obj); 2409 assert(str != NULL); 2410 printf("%s\n", qstring_get_str(str)); 2411 qobject_decref(obj); 2412 visit_free(v); 2413 QDECREF(str); 2414 } 2415 2416 static void dump_human_image_info_list(ImageInfoList *list) 2417 { 2418 ImageInfoList *elem; 2419 bool delim = false; 2420 2421 for (elem = list; elem; elem = elem->next) { 2422 if (delim) { 2423 printf("\n"); 2424 } 2425 delim = true; 2426 2427 bdrv_image_info_dump(fprintf, stdout, elem->value); 2428 } 2429 } 2430 2431 static gboolean str_equal_func(gconstpointer a, gconstpointer b) 2432 { 2433 return strcmp(a, b) == 0; 2434 } 2435 2436 /** 2437 * Open an image file chain and return an ImageInfoList 2438 * 2439 * @filename: topmost image filename 2440 * @fmt: topmost image format (may be NULL to autodetect) 2441 * @chain: true - enumerate entire backing file chain 2442 * false - only topmost image file 2443 * 2444 * Returns a list of ImageInfo objects or NULL if there was an error opening an 2445 * image file. If there was an error a message will have been printed to 2446 * stderr. 2447 */ 2448 static ImageInfoList *collect_image_info_list(bool image_opts, 2449 const char *filename, 2450 const char *fmt, 2451 bool chain) 2452 { 2453 ImageInfoList *head = NULL; 2454 ImageInfoList **last = &head; 2455 GHashTable *filenames; 2456 Error *err = NULL; 2457 2458 filenames = g_hash_table_new_full(g_str_hash, str_equal_func, NULL, NULL); 2459 2460 while (filename) { 2461 BlockBackend *blk; 2462 BlockDriverState *bs; 2463 ImageInfo *info; 2464 ImageInfoList *elem; 2465 2466 if (g_hash_table_lookup_extended(filenames, filename, NULL, NULL)) { 2467 error_report("Backing file '%s' creates an infinite loop.", 2468 filename); 2469 goto err; 2470 } 2471 g_hash_table_insert(filenames, (gpointer)filename, NULL); 2472 2473 blk = img_open(image_opts, filename, fmt, 2474 BDRV_O_NO_BACKING | BDRV_O_NO_IO, false, false); 2475 if (!blk) { 2476 goto err; 2477 } 2478 bs = blk_bs(blk); 2479 2480 bdrv_query_image_info(bs, &info, &err); 2481 if (err) { 2482 error_report_err(err); 2483 blk_unref(blk); 2484 goto err; 2485 } 2486 2487 elem = g_new0(ImageInfoList, 1); 2488 elem->value = info; 2489 *last = elem; 2490 last = &elem->next; 2491 2492 blk_unref(blk); 2493 2494 filename = fmt = NULL; 2495 if (chain) { 2496 if (info->has_full_backing_filename) { 2497 filename = info->full_backing_filename; 2498 } else if (info->has_backing_filename) { 2499 error_report("Could not determine absolute backing filename," 2500 " but backing filename '%s' present", 2501 info->backing_filename); 2502 goto err; 2503 } 2504 if (info->has_backing_filename_format) { 2505 fmt = info->backing_filename_format; 2506 } 2507 } 2508 } 2509 g_hash_table_destroy(filenames); 2510 return head; 2511 2512 err: 2513 qapi_free_ImageInfoList(head); 2514 g_hash_table_destroy(filenames); 2515 return NULL; 2516 } 2517 2518 static int img_info(int argc, char **argv) 2519 { 2520 int c; 2521 OutputFormat output_format = OFORMAT_HUMAN; 2522 bool chain = false; 2523 const char *filename, *fmt, *output; 2524 ImageInfoList *list; 2525 bool image_opts = false; 2526 2527 fmt = NULL; 2528 output = NULL; 2529 for(;;) { 2530 int option_index = 0; 2531 static const struct option long_options[] = { 2532 {"help", no_argument, 0, 'h'}, 2533 {"format", required_argument, 0, 'f'}, 2534 {"output", required_argument, 0, OPTION_OUTPUT}, 2535 {"backing-chain", no_argument, 0, OPTION_BACKING_CHAIN}, 2536 {"object", required_argument, 0, OPTION_OBJECT}, 2537 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS}, 2538 {0, 0, 0, 0} 2539 }; 2540 c = getopt_long(argc, argv, ":f:h", 2541 long_options, &option_index); 2542 if (c == -1) { 2543 break; 2544 } 2545 switch(c) { 2546 case ':': 2547 missing_argument(argv[optind - 1]); 2548 break; 2549 case '?': 2550 unrecognized_option(argv[optind - 1]); 2551 break; 2552 case 'h': 2553 help(); 2554 break; 2555 case 'f': 2556 fmt = optarg; 2557 break; 2558 case OPTION_OUTPUT: 2559 output = optarg; 2560 break; 2561 case OPTION_BACKING_CHAIN: 2562 chain = true; 2563 break; 2564 case OPTION_OBJECT: { 2565 QemuOpts *opts; 2566 opts = qemu_opts_parse_noisily(&qemu_object_opts, 2567 optarg, true); 2568 if (!opts) { 2569 return 1; 2570 } 2571 } break; 2572 case OPTION_IMAGE_OPTS: 2573 image_opts = true; 2574 break; 2575 } 2576 } 2577 if (optind != argc - 1) { 2578 error_exit("Expecting one image file name"); 2579 } 2580 filename = argv[optind++]; 2581 2582 if (output && !strcmp(output, "json")) { 2583 output_format = OFORMAT_JSON; 2584 } else if (output && !strcmp(output, "human")) { 2585 output_format = OFORMAT_HUMAN; 2586 } else if (output) { 2587 error_report("--output must be used with human or json as argument."); 2588 return 1; 2589 } 2590 2591 if (qemu_opts_foreach(&qemu_object_opts, 2592 user_creatable_add_opts_foreach, 2593 NULL, NULL)) { 2594 return 1; 2595 } 2596 2597 list = collect_image_info_list(image_opts, filename, fmt, chain); 2598 if (!list) { 2599 return 1; 2600 } 2601 2602 switch (output_format) { 2603 case OFORMAT_HUMAN: 2604 dump_human_image_info_list(list); 2605 break; 2606 case OFORMAT_JSON: 2607 if (chain) { 2608 dump_json_image_info_list(list); 2609 } else { 2610 dump_json_image_info(list->value); 2611 } 2612 break; 2613 } 2614 2615 qapi_free_ImageInfoList(list); 2616 return 0; 2617 } 2618 2619 static void dump_map_entry(OutputFormat output_format, MapEntry *e, 2620 MapEntry *next) 2621 { 2622 switch (output_format) { 2623 case OFORMAT_HUMAN: 2624 if (e->data && !e->has_offset) { 2625 error_report("File contains external, encrypted or compressed clusters."); 2626 exit(1); 2627 } 2628 if (e->data && !e->zero) { 2629 printf("%#-16"PRIx64"%#-16"PRIx64"%#-16"PRIx64"%s\n", 2630 e->start, e->length, 2631 e->has_offset ? e->offset : 0, 2632 e->has_filename ? e->filename : ""); 2633 } 2634 /* This format ignores the distinction between 0, ZERO and ZERO|DATA. 2635 * Modify the flags here to allow more coalescing. 2636 */ 2637 if (next && (!next->data || next->zero)) { 2638 next->data = false; 2639 next->zero = true; 2640 } 2641 break; 2642 case OFORMAT_JSON: 2643 printf("%s{ \"start\": %"PRId64", \"length\": %"PRId64"," 2644 " \"depth\": %"PRId64", \"zero\": %s, \"data\": %s", 2645 (e->start == 0 ? "[" : ",\n"), 2646 e->start, e->length, e->depth, 2647 e->zero ? "true" : "false", 2648 e->data ? "true" : "false"); 2649 if (e->has_offset) { 2650 printf(", \"offset\": %"PRId64"", e->offset); 2651 } 2652 putchar('}'); 2653 2654 if (!next) { 2655 printf("]\n"); 2656 } 2657 break; 2658 } 2659 } 2660 2661 static int get_block_status(BlockDriverState *bs, int64_t sector_num, 2662 int nb_sectors, MapEntry *e) 2663 { 2664 int64_t ret; 2665 int depth; 2666 BlockDriverState *file; 2667 bool has_offset; 2668 2669 /* As an optimization, we could cache the current range of unallocated 2670 * clusters in each file of the chain, and avoid querying the same 2671 * range repeatedly. 2672 */ 2673 2674 depth = 0; 2675 for (;;) { 2676 ret = bdrv_get_block_status(bs, sector_num, nb_sectors, &nb_sectors, 2677 &file); 2678 if (ret < 0) { 2679 return ret; 2680 } 2681 assert(nb_sectors); 2682 if (ret & (BDRV_BLOCK_ZERO|BDRV_BLOCK_DATA)) { 2683 break; 2684 } 2685 bs = backing_bs(bs); 2686 if (bs == NULL) { 2687 ret = 0; 2688 break; 2689 } 2690 2691 depth++; 2692 } 2693 2694 has_offset = !!(ret & BDRV_BLOCK_OFFSET_VALID); 2695 2696 *e = (MapEntry) { 2697 .start = sector_num * BDRV_SECTOR_SIZE, 2698 .length = nb_sectors * BDRV_SECTOR_SIZE, 2699 .data = !!(ret & BDRV_BLOCK_DATA), 2700 .zero = !!(ret & BDRV_BLOCK_ZERO), 2701 .offset = ret & BDRV_BLOCK_OFFSET_MASK, 2702 .has_offset = has_offset, 2703 .depth = depth, 2704 .has_filename = file && has_offset, 2705 .filename = file && has_offset ? file->filename : NULL, 2706 }; 2707 2708 return 0; 2709 } 2710 2711 static inline bool entry_mergeable(const MapEntry *curr, const MapEntry *next) 2712 { 2713 if (curr->length == 0) { 2714 return false; 2715 } 2716 if (curr->zero != next->zero || 2717 curr->data != next->data || 2718 curr->depth != next->depth || 2719 curr->has_filename != next->has_filename || 2720 curr->has_offset != next->has_offset) { 2721 return false; 2722 } 2723 if (curr->has_filename && strcmp(curr->filename, next->filename)) { 2724 return false; 2725 } 2726 if (curr->has_offset && curr->offset + curr->length != next->offset) { 2727 return false; 2728 } 2729 return true; 2730 } 2731 2732 static int img_map(int argc, char **argv) 2733 { 2734 int c; 2735 OutputFormat output_format = OFORMAT_HUMAN; 2736 BlockBackend *blk; 2737 BlockDriverState *bs; 2738 const char *filename, *fmt, *output; 2739 int64_t length; 2740 MapEntry curr = { .length = 0 }, next; 2741 int ret = 0; 2742 bool image_opts = false; 2743 2744 fmt = NULL; 2745 output = NULL; 2746 for (;;) { 2747 int option_index = 0; 2748 static const struct option long_options[] = { 2749 {"help", no_argument, 0, 'h'}, 2750 {"format", required_argument, 0, 'f'}, 2751 {"output", required_argument, 0, OPTION_OUTPUT}, 2752 {"object", required_argument, 0, OPTION_OBJECT}, 2753 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS}, 2754 {0, 0, 0, 0} 2755 }; 2756 c = getopt_long(argc, argv, ":f:h", 2757 long_options, &option_index); 2758 if (c == -1) { 2759 break; 2760 } 2761 switch (c) { 2762 case ':': 2763 missing_argument(argv[optind - 1]); 2764 break; 2765 case '?': 2766 unrecognized_option(argv[optind - 1]); 2767 break; 2768 case 'h': 2769 help(); 2770 break; 2771 case 'f': 2772 fmt = optarg; 2773 break; 2774 case OPTION_OUTPUT: 2775 output = optarg; 2776 break; 2777 case OPTION_OBJECT: { 2778 QemuOpts *opts; 2779 opts = qemu_opts_parse_noisily(&qemu_object_opts, 2780 optarg, true); 2781 if (!opts) { 2782 return 1; 2783 } 2784 } break; 2785 case OPTION_IMAGE_OPTS: 2786 image_opts = true; 2787 break; 2788 } 2789 } 2790 if (optind != argc - 1) { 2791 error_exit("Expecting one image file name"); 2792 } 2793 filename = argv[optind]; 2794 2795 if (output && !strcmp(output, "json")) { 2796 output_format = OFORMAT_JSON; 2797 } else if (output && !strcmp(output, "human")) { 2798 output_format = OFORMAT_HUMAN; 2799 } else if (output) { 2800 error_report("--output must be used with human or json as argument."); 2801 return 1; 2802 } 2803 2804 if (qemu_opts_foreach(&qemu_object_opts, 2805 user_creatable_add_opts_foreach, 2806 NULL, NULL)) { 2807 return 1; 2808 } 2809 2810 blk = img_open(image_opts, filename, fmt, 0, false, false); 2811 if (!blk) { 2812 return 1; 2813 } 2814 bs = blk_bs(blk); 2815 2816 if (output_format == OFORMAT_HUMAN) { 2817 printf("%-16s%-16s%-16s%s\n", "Offset", "Length", "Mapped to", "File"); 2818 } 2819 2820 length = blk_getlength(blk); 2821 while (curr.start + curr.length < length) { 2822 int64_t nsectors_left; 2823 int64_t sector_num; 2824 int n; 2825 2826 sector_num = (curr.start + curr.length) >> BDRV_SECTOR_BITS; 2827 2828 /* Probe up to 1 GiB at a time. */ 2829 nsectors_left = DIV_ROUND_UP(length, BDRV_SECTOR_SIZE) - sector_num; 2830 n = MIN(1 << (30 - BDRV_SECTOR_BITS), nsectors_left); 2831 ret = get_block_status(bs, sector_num, n, &next); 2832 2833 if (ret < 0) { 2834 error_report("Could not read file metadata: %s", strerror(-ret)); 2835 goto out; 2836 } 2837 2838 if (entry_mergeable(&curr, &next)) { 2839 curr.length += next.length; 2840 continue; 2841 } 2842 2843 if (curr.length > 0) { 2844 dump_map_entry(output_format, &curr, &next); 2845 } 2846 curr = next; 2847 } 2848 2849 dump_map_entry(output_format, &curr, NULL); 2850 2851 out: 2852 blk_unref(blk); 2853 return ret < 0; 2854 } 2855 2856 #define SNAPSHOT_LIST 1 2857 #define SNAPSHOT_CREATE 2 2858 #define SNAPSHOT_APPLY 3 2859 #define SNAPSHOT_DELETE 4 2860 2861 static int img_snapshot(int argc, char **argv) 2862 { 2863 BlockBackend *blk; 2864 BlockDriverState *bs; 2865 QEMUSnapshotInfo sn; 2866 char *filename, *snapshot_name = NULL; 2867 int c, ret = 0, bdrv_oflags; 2868 int action = 0; 2869 qemu_timeval tv; 2870 bool quiet = false; 2871 Error *err = NULL; 2872 bool image_opts = false; 2873 2874 bdrv_oflags = BDRV_O_RDWR; 2875 /* Parse commandline parameters */ 2876 for(;;) { 2877 static const struct option long_options[] = { 2878 {"help", no_argument, 0, 'h'}, 2879 {"object", required_argument, 0, OPTION_OBJECT}, 2880 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS}, 2881 {0, 0, 0, 0} 2882 }; 2883 c = getopt_long(argc, argv, ":la:c:d:hq", 2884 long_options, NULL); 2885 if (c == -1) { 2886 break; 2887 } 2888 switch(c) { 2889 case ':': 2890 missing_argument(argv[optind - 1]); 2891 break; 2892 case '?': 2893 unrecognized_option(argv[optind - 1]); 2894 break; 2895 case 'h': 2896 help(); 2897 return 0; 2898 case 'l': 2899 if (action) { 2900 error_exit("Cannot mix '-l', '-a', '-c', '-d'"); 2901 return 0; 2902 } 2903 action = SNAPSHOT_LIST; 2904 bdrv_oflags &= ~BDRV_O_RDWR; /* no need for RW */ 2905 break; 2906 case 'a': 2907 if (action) { 2908 error_exit("Cannot mix '-l', '-a', '-c', '-d'"); 2909 return 0; 2910 } 2911 action = SNAPSHOT_APPLY; 2912 snapshot_name = optarg; 2913 break; 2914 case 'c': 2915 if (action) { 2916 error_exit("Cannot mix '-l', '-a', '-c', '-d'"); 2917 return 0; 2918 } 2919 action = SNAPSHOT_CREATE; 2920 snapshot_name = optarg; 2921 break; 2922 case 'd': 2923 if (action) { 2924 error_exit("Cannot mix '-l', '-a', '-c', '-d'"); 2925 return 0; 2926 } 2927 action = SNAPSHOT_DELETE; 2928 snapshot_name = optarg; 2929 break; 2930 case 'q': 2931 quiet = true; 2932 break; 2933 case OPTION_OBJECT: { 2934 QemuOpts *opts; 2935 opts = qemu_opts_parse_noisily(&qemu_object_opts, 2936 optarg, true); 2937 if (!opts) { 2938 return 1; 2939 } 2940 } break; 2941 case OPTION_IMAGE_OPTS: 2942 image_opts = true; 2943 break; 2944 } 2945 } 2946 2947 if (optind != argc - 1) { 2948 error_exit("Expecting one image file name"); 2949 } 2950 filename = argv[optind++]; 2951 2952 if (qemu_opts_foreach(&qemu_object_opts, 2953 user_creatable_add_opts_foreach, 2954 NULL, NULL)) { 2955 return 1; 2956 } 2957 2958 /* Open the image */ 2959 blk = img_open(image_opts, filename, NULL, bdrv_oflags, false, quiet); 2960 if (!blk) { 2961 return 1; 2962 } 2963 bs = blk_bs(blk); 2964 2965 /* Perform the requested action */ 2966 switch(action) { 2967 case SNAPSHOT_LIST: 2968 dump_snapshots(bs); 2969 break; 2970 2971 case SNAPSHOT_CREATE: 2972 memset(&sn, 0, sizeof(sn)); 2973 pstrcpy(sn.name, sizeof(sn.name), snapshot_name); 2974 2975 qemu_gettimeofday(&tv); 2976 sn.date_sec = tv.tv_sec; 2977 sn.date_nsec = tv.tv_usec * 1000; 2978 2979 ret = bdrv_snapshot_create(bs, &sn); 2980 if (ret) { 2981 error_report("Could not create snapshot '%s': %d (%s)", 2982 snapshot_name, ret, strerror(-ret)); 2983 } 2984 break; 2985 2986 case SNAPSHOT_APPLY: 2987 ret = bdrv_snapshot_goto(bs, snapshot_name); 2988 if (ret) { 2989 error_report("Could not apply snapshot '%s': %d (%s)", 2990 snapshot_name, ret, strerror(-ret)); 2991 } 2992 break; 2993 2994 case SNAPSHOT_DELETE: 2995 bdrv_snapshot_delete_by_id_or_name(bs, snapshot_name, &err); 2996 if (err) { 2997 error_reportf_err(err, "Could not delete snapshot '%s': ", 2998 snapshot_name); 2999 ret = 1; 3000 } 3001 break; 3002 } 3003 3004 /* Cleanup */ 3005 blk_unref(blk); 3006 if (ret) { 3007 return 1; 3008 } 3009 return 0; 3010 } 3011 3012 static int img_rebase(int argc, char **argv) 3013 { 3014 BlockBackend *blk = NULL, *blk_old_backing = NULL, *blk_new_backing = NULL; 3015 uint8_t *buf_old = NULL; 3016 uint8_t *buf_new = NULL; 3017 BlockDriverState *bs = NULL; 3018 char *filename; 3019 const char *fmt, *cache, *src_cache, *out_basefmt, *out_baseimg; 3020 int c, flags, src_flags, ret; 3021 bool writethrough, src_writethrough; 3022 int unsafe = 0; 3023 int progress = 0; 3024 bool quiet = false; 3025 Error *local_err = NULL; 3026 bool image_opts = false; 3027 3028 /* Parse commandline parameters */ 3029 fmt = NULL; 3030 cache = BDRV_DEFAULT_CACHE; 3031 src_cache = BDRV_DEFAULT_CACHE; 3032 out_baseimg = NULL; 3033 out_basefmt = NULL; 3034 for(;;) { 3035 static const struct option long_options[] = { 3036 {"help", no_argument, 0, 'h'}, 3037 {"object", required_argument, 0, OPTION_OBJECT}, 3038 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS}, 3039 {0, 0, 0, 0} 3040 }; 3041 c = getopt_long(argc, argv, ":hf:F:b:upt:T:q", 3042 long_options, NULL); 3043 if (c == -1) { 3044 break; 3045 } 3046 switch(c) { 3047 case ':': 3048 missing_argument(argv[optind - 1]); 3049 break; 3050 case '?': 3051 unrecognized_option(argv[optind - 1]); 3052 break; 3053 case 'h': 3054 help(); 3055 return 0; 3056 case 'f': 3057 fmt = optarg; 3058 break; 3059 case 'F': 3060 out_basefmt = optarg; 3061 break; 3062 case 'b': 3063 out_baseimg = optarg; 3064 break; 3065 case 'u': 3066 unsafe = 1; 3067 break; 3068 case 'p': 3069 progress = 1; 3070 break; 3071 case 't': 3072 cache = optarg; 3073 break; 3074 case 'T': 3075 src_cache = optarg; 3076 break; 3077 case 'q': 3078 quiet = true; 3079 break; 3080 case OPTION_OBJECT: { 3081 QemuOpts *opts; 3082 opts = qemu_opts_parse_noisily(&qemu_object_opts, 3083 optarg, true); 3084 if (!opts) { 3085 return 1; 3086 } 3087 } break; 3088 case OPTION_IMAGE_OPTS: 3089 image_opts = true; 3090 break; 3091 } 3092 } 3093 3094 if (quiet) { 3095 progress = 0; 3096 } 3097 3098 if (optind != argc - 1) { 3099 error_exit("Expecting one image file name"); 3100 } 3101 if (!unsafe && !out_baseimg) { 3102 error_exit("Must specify backing file (-b) or use unsafe mode (-u)"); 3103 } 3104 filename = argv[optind++]; 3105 3106 if (qemu_opts_foreach(&qemu_object_opts, 3107 user_creatable_add_opts_foreach, 3108 NULL, NULL)) { 3109 return 1; 3110 } 3111 3112 qemu_progress_init(progress, 2.0); 3113 qemu_progress_print(0, 100); 3114 3115 flags = BDRV_O_RDWR | (unsafe ? BDRV_O_NO_BACKING : 0); 3116 ret = bdrv_parse_cache_mode(cache, &flags, &writethrough); 3117 if (ret < 0) { 3118 error_report("Invalid cache option: %s", cache); 3119 goto out; 3120 } 3121 3122 src_flags = 0; 3123 ret = bdrv_parse_cache_mode(src_cache, &src_flags, &src_writethrough); 3124 if (ret < 0) { 3125 error_report("Invalid source cache option: %s", src_cache); 3126 goto out; 3127 } 3128 3129 /* The source files are opened read-only, don't care about WCE */ 3130 assert((src_flags & BDRV_O_RDWR) == 0); 3131 (void) src_writethrough; 3132 3133 /* 3134 * Open the images. 3135 * 3136 * Ignore the old backing file for unsafe rebase in case we want to correct 3137 * the reference to a renamed or moved backing file. 3138 */ 3139 blk = img_open(image_opts, filename, fmt, flags, writethrough, quiet); 3140 if (!blk) { 3141 ret = -1; 3142 goto out; 3143 } 3144 bs = blk_bs(blk); 3145 3146 if (out_basefmt != NULL) { 3147 if (bdrv_find_format(out_basefmt) == NULL) { 3148 error_report("Invalid format name: '%s'", out_basefmt); 3149 ret = -1; 3150 goto out; 3151 } 3152 } 3153 3154 /* For safe rebasing we need to compare old and new backing file */ 3155 if (!unsafe) { 3156 char backing_name[PATH_MAX]; 3157 QDict *options = NULL; 3158 3159 if (bs->backing_format[0] != '\0') { 3160 options = qdict_new(); 3161 qdict_put(options, "driver", qstring_from_str(bs->backing_format)); 3162 } 3163 3164 bdrv_get_backing_filename(bs, backing_name, sizeof(backing_name)); 3165 blk_old_backing = blk_new_open(backing_name, NULL, 3166 options, src_flags, &local_err); 3167 if (!blk_old_backing) { 3168 error_reportf_err(local_err, 3169 "Could not open old backing file '%s': ", 3170 backing_name); 3171 ret = -1; 3172 goto out; 3173 } 3174 3175 if (out_baseimg[0]) { 3176 if (out_basefmt) { 3177 options = qdict_new(); 3178 qdict_put(options, "driver", qstring_from_str(out_basefmt)); 3179 } else { 3180 options = NULL; 3181 } 3182 3183 blk_new_backing = blk_new_open(out_baseimg, NULL, 3184 options, src_flags, &local_err); 3185 if (!blk_new_backing) { 3186 error_reportf_err(local_err, 3187 "Could not open new backing file '%s': ", 3188 out_baseimg); 3189 ret = -1; 3190 goto out; 3191 } 3192 } 3193 } 3194 3195 /* 3196 * Check each unallocated cluster in the COW file. If it is unallocated, 3197 * accesses go to the backing file. We must therefore compare this cluster 3198 * in the old and new backing file, and if they differ we need to copy it 3199 * from the old backing file into the COW file. 3200 * 3201 * If qemu-img crashes during this step, no harm is done. The content of 3202 * the image is the same as the original one at any time. 3203 */ 3204 if (!unsafe) { 3205 int64_t num_sectors; 3206 int64_t old_backing_num_sectors; 3207 int64_t new_backing_num_sectors = 0; 3208 uint64_t sector; 3209 int n; 3210 float local_progress = 0; 3211 3212 buf_old = blk_blockalign(blk, IO_BUF_SIZE); 3213 buf_new = blk_blockalign(blk, IO_BUF_SIZE); 3214 3215 num_sectors = blk_nb_sectors(blk); 3216 if (num_sectors < 0) { 3217 error_report("Could not get size of '%s': %s", 3218 filename, strerror(-num_sectors)); 3219 ret = -1; 3220 goto out; 3221 } 3222 old_backing_num_sectors = blk_nb_sectors(blk_old_backing); 3223 if (old_backing_num_sectors < 0) { 3224 char backing_name[PATH_MAX]; 3225 3226 bdrv_get_backing_filename(bs, backing_name, sizeof(backing_name)); 3227 error_report("Could not get size of '%s': %s", 3228 backing_name, strerror(-old_backing_num_sectors)); 3229 ret = -1; 3230 goto out; 3231 } 3232 if (blk_new_backing) { 3233 new_backing_num_sectors = blk_nb_sectors(blk_new_backing); 3234 if (new_backing_num_sectors < 0) { 3235 error_report("Could not get size of '%s': %s", 3236 out_baseimg, strerror(-new_backing_num_sectors)); 3237 ret = -1; 3238 goto out; 3239 } 3240 } 3241 3242 if (num_sectors != 0) { 3243 local_progress = (float)100 / 3244 (num_sectors / MIN(num_sectors, IO_BUF_SIZE / 512)); 3245 } 3246 3247 for (sector = 0; sector < num_sectors; sector += n) { 3248 3249 /* How many sectors can we handle with the next read? */ 3250 if (sector + (IO_BUF_SIZE / 512) <= num_sectors) { 3251 n = (IO_BUF_SIZE / 512); 3252 } else { 3253 n = num_sectors - sector; 3254 } 3255 3256 /* If the cluster is allocated, we don't need to take action */ 3257 ret = bdrv_is_allocated(bs, sector, n, &n); 3258 if (ret < 0) { 3259 error_report("error while reading image metadata: %s", 3260 strerror(-ret)); 3261 goto out; 3262 } 3263 if (ret) { 3264 continue; 3265 } 3266 3267 /* 3268 * Read old and new backing file and take into consideration that 3269 * backing files may be smaller than the COW image. 3270 */ 3271 if (sector >= old_backing_num_sectors) { 3272 memset(buf_old, 0, n * BDRV_SECTOR_SIZE); 3273 } else { 3274 if (sector + n > old_backing_num_sectors) { 3275 n = old_backing_num_sectors - sector; 3276 } 3277 3278 ret = blk_pread(blk_old_backing, sector << BDRV_SECTOR_BITS, 3279 buf_old, n << BDRV_SECTOR_BITS); 3280 if (ret < 0) { 3281 error_report("error while reading from old backing file"); 3282 goto out; 3283 } 3284 } 3285 3286 if (sector >= new_backing_num_sectors || !blk_new_backing) { 3287 memset(buf_new, 0, n * BDRV_SECTOR_SIZE); 3288 } else { 3289 if (sector + n > new_backing_num_sectors) { 3290 n = new_backing_num_sectors - sector; 3291 } 3292 3293 ret = blk_pread(blk_new_backing, sector << BDRV_SECTOR_BITS, 3294 buf_new, n << BDRV_SECTOR_BITS); 3295 if (ret < 0) { 3296 error_report("error while reading from new backing file"); 3297 goto out; 3298 } 3299 } 3300 3301 /* If they differ, we need to write to the COW file */ 3302 uint64_t written = 0; 3303 3304 while (written < n) { 3305 int pnum; 3306 3307 if (compare_sectors(buf_old + written * 512, 3308 buf_new + written * 512, n - written, &pnum)) 3309 { 3310 ret = blk_pwrite(blk, 3311 (sector + written) << BDRV_SECTOR_BITS, 3312 buf_old + written * 512, 3313 pnum << BDRV_SECTOR_BITS, 0); 3314 if (ret < 0) { 3315 error_report("Error while writing to COW image: %s", 3316 strerror(-ret)); 3317 goto out; 3318 } 3319 } 3320 3321 written += pnum; 3322 } 3323 qemu_progress_print(local_progress, 100); 3324 } 3325 } 3326 3327 /* 3328 * Change the backing file. All clusters that are different from the old 3329 * backing file are overwritten in the COW file now, so the visible content 3330 * doesn't change when we switch the backing file. 3331 */ 3332 if (out_baseimg && *out_baseimg) { 3333 ret = bdrv_change_backing_file(bs, out_baseimg, out_basefmt); 3334 } else { 3335 ret = bdrv_change_backing_file(bs, NULL, NULL); 3336 } 3337 3338 if (ret == -ENOSPC) { 3339 error_report("Could not change the backing file to '%s': No " 3340 "space left in the file header", out_baseimg); 3341 } else if (ret < 0) { 3342 error_report("Could not change the backing file to '%s': %s", 3343 out_baseimg, strerror(-ret)); 3344 } 3345 3346 qemu_progress_print(100, 0); 3347 /* 3348 * TODO At this point it is possible to check if any clusters that are 3349 * allocated in the COW file are the same in the backing file. If so, they 3350 * could be dropped from the COW file. Don't do this before switching the 3351 * backing file, in case of a crash this would lead to corruption. 3352 */ 3353 out: 3354 qemu_progress_end(); 3355 /* Cleanup */ 3356 if (!unsafe) { 3357 blk_unref(blk_old_backing); 3358 blk_unref(blk_new_backing); 3359 } 3360 qemu_vfree(buf_old); 3361 qemu_vfree(buf_new); 3362 3363 blk_unref(blk); 3364 if (ret) { 3365 return 1; 3366 } 3367 return 0; 3368 } 3369 3370 static int img_resize(int argc, char **argv) 3371 { 3372 Error *err = NULL; 3373 int c, ret, relative; 3374 const char *filename, *fmt, *size; 3375 int64_t n, total_size; 3376 bool quiet = false; 3377 BlockBackend *blk = NULL; 3378 QemuOpts *param; 3379 3380 static QemuOptsList resize_options = { 3381 .name = "resize_options", 3382 .head = QTAILQ_HEAD_INITIALIZER(resize_options.head), 3383 .desc = { 3384 { 3385 .name = BLOCK_OPT_SIZE, 3386 .type = QEMU_OPT_SIZE, 3387 .help = "Virtual disk size" 3388 }, { 3389 /* end of list */ 3390 } 3391 }, 3392 }; 3393 bool image_opts = false; 3394 3395 /* Remove size from argv manually so that negative numbers are not treated 3396 * as options by getopt. */ 3397 if (argc < 3) { 3398 error_exit("Not enough arguments"); 3399 return 1; 3400 } 3401 3402 size = argv[--argc]; 3403 3404 /* Parse getopt arguments */ 3405 fmt = NULL; 3406 for(;;) { 3407 static const struct option long_options[] = { 3408 {"help", no_argument, 0, 'h'}, 3409 {"object", required_argument, 0, OPTION_OBJECT}, 3410 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS}, 3411 {0, 0, 0, 0} 3412 }; 3413 c = getopt_long(argc, argv, ":f:hq", 3414 long_options, NULL); 3415 if (c == -1) { 3416 break; 3417 } 3418 switch(c) { 3419 case ':': 3420 missing_argument(argv[optind - 1]); 3421 break; 3422 case '?': 3423 unrecognized_option(argv[optind - 1]); 3424 break; 3425 case 'h': 3426 help(); 3427 break; 3428 case 'f': 3429 fmt = optarg; 3430 break; 3431 case 'q': 3432 quiet = true; 3433 break; 3434 case OPTION_OBJECT: { 3435 QemuOpts *opts; 3436 opts = qemu_opts_parse_noisily(&qemu_object_opts, 3437 optarg, true); 3438 if (!opts) { 3439 return 1; 3440 } 3441 } break; 3442 case OPTION_IMAGE_OPTS: 3443 image_opts = true; 3444 break; 3445 } 3446 } 3447 if (optind != argc - 1) { 3448 error_exit("Expecting one image file name"); 3449 } 3450 filename = argv[optind++]; 3451 3452 if (qemu_opts_foreach(&qemu_object_opts, 3453 user_creatable_add_opts_foreach, 3454 NULL, NULL)) { 3455 return 1; 3456 } 3457 3458 /* Choose grow, shrink, or absolute resize mode */ 3459 switch (size[0]) { 3460 case '+': 3461 relative = 1; 3462 size++; 3463 break; 3464 case '-': 3465 relative = -1; 3466 size++; 3467 break; 3468 default: 3469 relative = 0; 3470 break; 3471 } 3472 3473 /* Parse size */ 3474 param = qemu_opts_create(&resize_options, NULL, 0, &error_abort); 3475 qemu_opt_set(param, BLOCK_OPT_SIZE, size, &err); 3476 if (err) { 3477 error_report_err(err); 3478 ret = -1; 3479 qemu_opts_del(param); 3480 goto out; 3481 } 3482 n = qemu_opt_get_size(param, BLOCK_OPT_SIZE, 0); 3483 qemu_opts_del(param); 3484 3485 blk = img_open(image_opts, filename, fmt, 3486 BDRV_O_RDWR | BDRV_O_RESIZE, false, quiet); 3487 if (!blk) { 3488 ret = -1; 3489 goto out; 3490 } 3491 3492 if (relative) { 3493 total_size = blk_getlength(blk) + n * relative; 3494 } else { 3495 total_size = n; 3496 } 3497 if (total_size <= 0) { 3498 error_report("New image size must be positive"); 3499 ret = -1; 3500 goto out; 3501 } 3502 3503 ret = blk_truncate(blk, total_size); 3504 switch (ret) { 3505 case 0: 3506 qprintf(quiet, "Image resized.\n"); 3507 break; 3508 case -ENOTSUP: 3509 error_report("This image does not support resize"); 3510 break; 3511 case -EACCES: 3512 error_report("Image is read-only"); 3513 break; 3514 default: 3515 error_report("Error resizing image: %s", strerror(-ret)); 3516 break; 3517 } 3518 out: 3519 blk_unref(blk); 3520 if (ret) { 3521 return 1; 3522 } 3523 return 0; 3524 } 3525 3526 static void amend_status_cb(BlockDriverState *bs, 3527 int64_t offset, int64_t total_work_size, 3528 void *opaque) 3529 { 3530 qemu_progress_print(100.f * offset / total_work_size, 0); 3531 } 3532 3533 static int img_amend(int argc, char **argv) 3534 { 3535 Error *err = NULL; 3536 int c, ret = 0; 3537 char *options = NULL; 3538 QemuOptsList *create_opts = NULL; 3539 QemuOpts *opts = NULL; 3540 const char *fmt = NULL, *filename, *cache; 3541 int flags; 3542 bool writethrough; 3543 bool quiet = false, progress = false; 3544 BlockBackend *blk = NULL; 3545 BlockDriverState *bs = NULL; 3546 bool image_opts = false; 3547 3548 cache = BDRV_DEFAULT_CACHE; 3549 for (;;) { 3550 static const struct option long_options[] = { 3551 {"help", no_argument, 0, 'h'}, 3552 {"object", required_argument, 0, OPTION_OBJECT}, 3553 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS}, 3554 {0, 0, 0, 0} 3555 }; 3556 c = getopt_long(argc, argv, ":ho:f:t:pq", 3557 long_options, NULL); 3558 if (c == -1) { 3559 break; 3560 } 3561 3562 switch (c) { 3563 case ':': 3564 missing_argument(argv[optind - 1]); 3565 break; 3566 case '?': 3567 unrecognized_option(argv[optind - 1]); 3568 break; 3569 case 'h': 3570 help(); 3571 break; 3572 case 'o': 3573 if (!is_valid_option_list(optarg)) { 3574 error_report("Invalid option list: %s", optarg); 3575 ret = -1; 3576 goto out_no_progress; 3577 } 3578 if (!options) { 3579 options = g_strdup(optarg); 3580 } else { 3581 char *old_options = options; 3582 options = g_strdup_printf("%s,%s", options, optarg); 3583 g_free(old_options); 3584 } 3585 break; 3586 case 'f': 3587 fmt = optarg; 3588 break; 3589 case 't': 3590 cache = optarg; 3591 break; 3592 case 'p': 3593 progress = true; 3594 break; 3595 case 'q': 3596 quiet = true; 3597 break; 3598 case OPTION_OBJECT: 3599 opts = qemu_opts_parse_noisily(&qemu_object_opts, 3600 optarg, true); 3601 if (!opts) { 3602 ret = -1; 3603 goto out_no_progress; 3604 } 3605 break; 3606 case OPTION_IMAGE_OPTS: 3607 image_opts = true; 3608 break; 3609 } 3610 } 3611 3612 if (!options) { 3613 error_exit("Must specify options (-o)"); 3614 } 3615 3616 if (qemu_opts_foreach(&qemu_object_opts, 3617 user_creatable_add_opts_foreach, 3618 NULL, NULL)) { 3619 ret = -1; 3620 goto out_no_progress; 3621 } 3622 3623 if (quiet) { 3624 progress = false; 3625 } 3626 qemu_progress_init(progress, 1.0); 3627 3628 filename = (optind == argc - 1) ? argv[argc - 1] : NULL; 3629 if (fmt && has_help_option(options)) { 3630 /* If a format is explicitly specified (and possibly no filename is 3631 * given), print option help here */ 3632 ret = print_block_option_help(filename, fmt); 3633 goto out; 3634 } 3635 3636 if (optind != argc - 1) { 3637 error_report("Expecting one image file name"); 3638 ret = -1; 3639 goto out; 3640 } 3641 3642 flags = BDRV_O_RDWR; 3643 ret = bdrv_parse_cache_mode(cache, &flags, &writethrough); 3644 if (ret < 0) { 3645 error_report("Invalid cache option: %s", cache); 3646 goto out; 3647 } 3648 3649 blk = img_open(image_opts, filename, fmt, flags, writethrough, quiet); 3650 if (!blk) { 3651 ret = -1; 3652 goto out; 3653 } 3654 bs = blk_bs(blk); 3655 3656 fmt = bs->drv->format_name; 3657 3658 if (has_help_option(options)) { 3659 /* If the format was auto-detected, print option help here */ 3660 ret = print_block_option_help(filename, fmt); 3661 goto out; 3662 } 3663 3664 if (!bs->drv->create_opts) { 3665 error_report("Format driver '%s' does not support any options to amend", 3666 fmt); 3667 ret = -1; 3668 goto out; 3669 } 3670 3671 create_opts = qemu_opts_append(create_opts, bs->drv->create_opts); 3672 opts = qemu_opts_create(create_opts, NULL, 0, &error_abort); 3673 qemu_opts_do_parse(opts, options, NULL, &err); 3674 if (err) { 3675 error_report_err(err); 3676 ret = -1; 3677 goto out; 3678 } 3679 3680 /* In case the driver does not call amend_status_cb() */ 3681 qemu_progress_print(0.f, 0); 3682 ret = bdrv_amend_options(bs, opts, &amend_status_cb, NULL); 3683 qemu_progress_print(100.f, 0); 3684 if (ret < 0) { 3685 error_report("Error while amending options: %s", strerror(-ret)); 3686 goto out; 3687 } 3688 3689 out: 3690 qemu_progress_end(); 3691 3692 out_no_progress: 3693 blk_unref(blk); 3694 qemu_opts_del(opts); 3695 qemu_opts_free(create_opts); 3696 g_free(options); 3697 3698 if (ret) { 3699 return 1; 3700 } 3701 return 0; 3702 } 3703 3704 typedef struct BenchData { 3705 BlockBackend *blk; 3706 uint64_t image_size; 3707 bool write; 3708 int bufsize; 3709 int step; 3710 int nrreq; 3711 int n; 3712 int flush_interval; 3713 bool drain_on_flush; 3714 uint8_t *buf; 3715 QEMUIOVector *qiov; 3716 3717 int in_flight; 3718 bool in_flush; 3719 uint64_t offset; 3720 } BenchData; 3721 3722 static void bench_undrained_flush_cb(void *opaque, int ret) 3723 { 3724 if (ret < 0) { 3725 error_report("Failed flush request: %s", strerror(-ret)); 3726 exit(EXIT_FAILURE); 3727 } 3728 } 3729 3730 static void bench_cb(void *opaque, int ret) 3731 { 3732 BenchData *b = opaque; 3733 BlockAIOCB *acb; 3734 3735 if (ret < 0) { 3736 error_report("Failed request: %s", strerror(-ret)); 3737 exit(EXIT_FAILURE); 3738 } 3739 3740 if (b->in_flush) { 3741 /* Just finished a flush with drained queue: Start next requests */ 3742 assert(b->in_flight == 0); 3743 b->in_flush = false; 3744 } else if (b->in_flight > 0) { 3745 int remaining = b->n - b->in_flight; 3746 3747 b->n--; 3748 b->in_flight--; 3749 3750 /* Time for flush? Drain queue if requested, then flush */ 3751 if (b->flush_interval && remaining % b->flush_interval == 0) { 3752 if (!b->in_flight || !b->drain_on_flush) { 3753 BlockCompletionFunc *cb; 3754 3755 if (b->drain_on_flush) { 3756 b->in_flush = true; 3757 cb = bench_cb; 3758 } else { 3759 cb = bench_undrained_flush_cb; 3760 } 3761 3762 acb = blk_aio_flush(b->blk, cb, b); 3763 if (!acb) { 3764 error_report("Failed to issue flush request"); 3765 exit(EXIT_FAILURE); 3766 } 3767 } 3768 if (b->drain_on_flush) { 3769 return; 3770 } 3771 } 3772 } 3773 3774 while (b->n > b->in_flight && b->in_flight < b->nrreq) { 3775 int64_t offset = b->offset; 3776 /* blk_aio_* might look for completed I/Os and kick bench_cb 3777 * again, so make sure this operation is counted by in_flight 3778 * and b->offset is ready for the next submission. 3779 */ 3780 b->in_flight++; 3781 b->offset += b->step; 3782 b->offset %= b->image_size; 3783 if (b->write) { 3784 acb = blk_aio_pwritev(b->blk, offset, b->qiov, 0, bench_cb, b); 3785 } else { 3786 acb = blk_aio_preadv(b->blk, offset, b->qiov, 0, bench_cb, b); 3787 } 3788 if (!acb) { 3789 error_report("Failed to issue request"); 3790 exit(EXIT_FAILURE); 3791 } 3792 } 3793 } 3794 3795 static int img_bench(int argc, char **argv) 3796 { 3797 int c, ret = 0; 3798 const char *fmt = NULL, *filename; 3799 bool quiet = false; 3800 bool image_opts = false; 3801 bool is_write = false; 3802 int count = 75000; 3803 int depth = 64; 3804 int64_t offset = 0; 3805 size_t bufsize = 4096; 3806 int pattern = 0; 3807 size_t step = 0; 3808 int flush_interval = 0; 3809 bool drain_on_flush = true; 3810 int64_t image_size; 3811 BlockBackend *blk = NULL; 3812 BenchData data = {}; 3813 int flags = 0; 3814 bool writethrough = false; 3815 struct timeval t1, t2; 3816 int i; 3817 3818 for (;;) { 3819 static const struct option long_options[] = { 3820 {"help", no_argument, 0, 'h'}, 3821 {"flush-interval", required_argument, 0, OPTION_FLUSH_INTERVAL}, 3822 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS}, 3823 {"pattern", required_argument, 0, OPTION_PATTERN}, 3824 {"no-drain", no_argument, 0, OPTION_NO_DRAIN}, 3825 {0, 0, 0, 0} 3826 }; 3827 c = getopt_long(argc, argv, ":hc:d:f:no:qs:S:t:w", long_options, NULL); 3828 if (c == -1) { 3829 break; 3830 } 3831 3832 switch (c) { 3833 case ':': 3834 missing_argument(argv[optind - 1]); 3835 break; 3836 case '?': 3837 unrecognized_option(argv[optind - 1]); 3838 break; 3839 case 'h': 3840 help(); 3841 break; 3842 case 'c': 3843 { 3844 unsigned long res; 3845 3846 if (qemu_strtoul(optarg, NULL, 0, &res) < 0 || res > INT_MAX) { 3847 error_report("Invalid request count specified"); 3848 return 1; 3849 } 3850 count = res; 3851 break; 3852 } 3853 case 'd': 3854 { 3855 unsigned long res; 3856 3857 if (qemu_strtoul(optarg, NULL, 0, &res) < 0 || res > INT_MAX) { 3858 error_report("Invalid queue depth specified"); 3859 return 1; 3860 } 3861 depth = res; 3862 break; 3863 } 3864 case 'f': 3865 fmt = optarg; 3866 break; 3867 case 'n': 3868 flags |= BDRV_O_NATIVE_AIO; 3869 break; 3870 case 'o': 3871 { 3872 offset = cvtnum(optarg); 3873 if (offset < 0) { 3874 error_report("Invalid offset specified"); 3875 return 1; 3876 } 3877 break; 3878 } 3879 break; 3880 case 'q': 3881 quiet = true; 3882 break; 3883 case 's': 3884 { 3885 int64_t sval; 3886 3887 sval = cvtnum(optarg); 3888 if (sval < 0 || sval > INT_MAX) { 3889 error_report("Invalid buffer size specified"); 3890 return 1; 3891 } 3892 3893 bufsize = sval; 3894 break; 3895 } 3896 case 'S': 3897 { 3898 int64_t sval; 3899 3900 sval = cvtnum(optarg); 3901 if (sval < 0 || sval > INT_MAX) { 3902 error_report("Invalid step size specified"); 3903 return 1; 3904 } 3905 3906 step = sval; 3907 break; 3908 } 3909 case 't': 3910 ret = bdrv_parse_cache_mode(optarg, &flags, &writethrough); 3911 if (ret < 0) { 3912 error_report("Invalid cache mode"); 3913 ret = -1; 3914 goto out; 3915 } 3916 break; 3917 case 'w': 3918 flags |= BDRV_O_RDWR; 3919 is_write = true; 3920 break; 3921 case OPTION_PATTERN: 3922 { 3923 unsigned long res; 3924 3925 if (qemu_strtoul(optarg, NULL, 0, &res) < 0 || res > 0xff) { 3926 error_report("Invalid pattern byte specified"); 3927 return 1; 3928 } 3929 pattern = res; 3930 break; 3931 } 3932 case OPTION_FLUSH_INTERVAL: 3933 { 3934 unsigned long res; 3935 3936 if (qemu_strtoul(optarg, NULL, 0, &res) < 0 || res > INT_MAX) { 3937 error_report("Invalid flush interval specified"); 3938 return 1; 3939 } 3940 flush_interval = res; 3941 break; 3942 } 3943 case OPTION_NO_DRAIN: 3944 drain_on_flush = false; 3945 break; 3946 case OPTION_IMAGE_OPTS: 3947 image_opts = true; 3948 break; 3949 } 3950 } 3951 3952 if (optind != argc - 1) { 3953 error_exit("Expecting one image file name"); 3954 } 3955 filename = argv[argc - 1]; 3956 3957 if (!is_write && flush_interval) { 3958 error_report("--flush-interval is only available in write tests"); 3959 ret = -1; 3960 goto out; 3961 } 3962 if (flush_interval && flush_interval < depth) { 3963 error_report("Flush interval can't be smaller than depth"); 3964 ret = -1; 3965 goto out; 3966 } 3967 3968 blk = img_open(image_opts, filename, fmt, flags, writethrough, quiet); 3969 if (!blk) { 3970 ret = -1; 3971 goto out; 3972 } 3973 3974 image_size = blk_getlength(blk); 3975 if (image_size < 0) { 3976 ret = image_size; 3977 goto out; 3978 } 3979 3980 data = (BenchData) { 3981 .blk = blk, 3982 .image_size = image_size, 3983 .bufsize = bufsize, 3984 .step = step ?: bufsize, 3985 .nrreq = depth, 3986 .n = count, 3987 .offset = offset, 3988 .write = is_write, 3989 .flush_interval = flush_interval, 3990 .drain_on_flush = drain_on_flush, 3991 }; 3992 printf("Sending %d %s requests, %d bytes each, %d in parallel " 3993 "(starting at offset %" PRId64 ", step size %d)\n", 3994 data.n, data.write ? "write" : "read", data.bufsize, data.nrreq, 3995 data.offset, data.step); 3996 if (flush_interval) { 3997 printf("Sending flush every %d requests\n", flush_interval); 3998 } 3999 4000 data.buf = blk_blockalign(blk, data.nrreq * data.bufsize); 4001 memset(data.buf, pattern, data.nrreq * data.bufsize); 4002 4003 data.qiov = g_new(QEMUIOVector, data.nrreq); 4004 for (i = 0; i < data.nrreq; i++) { 4005 qemu_iovec_init(&data.qiov[i], 1); 4006 qemu_iovec_add(&data.qiov[i], 4007 data.buf + i * data.bufsize, data.bufsize); 4008 } 4009 4010 gettimeofday(&t1, NULL); 4011 bench_cb(&data, 0); 4012 4013 while (data.n > 0) { 4014 main_loop_wait(false); 4015 } 4016 gettimeofday(&t2, NULL); 4017 4018 printf("Run completed in %3.3f seconds.\n", 4019 (t2.tv_sec - t1.tv_sec) 4020 + ((double)(t2.tv_usec - t1.tv_usec) / 1000000)); 4021 4022 out: 4023 qemu_vfree(data.buf); 4024 blk_unref(blk); 4025 4026 if (ret) { 4027 return 1; 4028 } 4029 return 0; 4030 } 4031 4032 #define C_BS 01 4033 #define C_COUNT 02 4034 #define C_IF 04 4035 #define C_OF 010 4036 #define C_SKIP 020 4037 4038 struct DdInfo { 4039 unsigned int flags; 4040 int64_t count; 4041 }; 4042 4043 struct DdIo { 4044 int bsz; /* Block size */ 4045 char *filename; 4046 uint8_t *buf; 4047 int64_t offset; 4048 }; 4049 4050 struct DdOpts { 4051 const char *name; 4052 int (*f)(const char *, struct DdIo *, struct DdIo *, struct DdInfo *); 4053 unsigned int flag; 4054 }; 4055 4056 static int img_dd_bs(const char *arg, 4057 struct DdIo *in, struct DdIo *out, 4058 struct DdInfo *dd) 4059 { 4060 int64_t res; 4061 4062 res = cvtnum(arg); 4063 4064 if (res <= 0 || res > INT_MAX) { 4065 error_report("invalid number: '%s'", arg); 4066 return 1; 4067 } 4068 in->bsz = out->bsz = res; 4069 4070 return 0; 4071 } 4072 4073 static int img_dd_count(const char *arg, 4074 struct DdIo *in, struct DdIo *out, 4075 struct DdInfo *dd) 4076 { 4077 dd->count = cvtnum(arg); 4078 4079 if (dd->count < 0) { 4080 error_report("invalid number: '%s'", arg); 4081 return 1; 4082 } 4083 4084 return 0; 4085 } 4086 4087 static int img_dd_if(const char *arg, 4088 struct DdIo *in, struct DdIo *out, 4089 struct DdInfo *dd) 4090 { 4091 in->filename = g_strdup(arg); 4092 4093 return 0; 4094 } 4095 4096 static int img_dd_of(const char *arg, 4097 struct DdIo *in, struct DdIo *out, 4098 struct DdInfo *dd) 4099 { 4100 out->filename = g_strdup(arg); 4101 4102 return 0; 4103 } 4104 4105 static int img_dd_skip(const char *arg, 4106 struct DdIo *in, struct DdIo *out, 4107 struct DdInfo *dd) 4108 { 4109 in->offset = cvtnum(arg); 4110 4111 if (in->offset < 0) { 4112 error_report("invalid number: '%s'", arg); 4113 return 1; 4114 } 4115 4116 return 0; 4117 } 4118 4119 static int img_dd(int argc, char **argv) 4120 { 4121 int ret = 0; 4122 char *arg = NULL; 4123 char *tmp; 4124 BlockDriver *drv = NULL, *proto_drv = NULL; 4125 BlockBackend *blk1 = NULL, *blk2 = NULL; 4126 QemuOpts *opts = NULL; 4127 QemuOptsList *create_opts = NULL; 4128 Error *local_err = NULL; 4129 bool image_opts = false; 4130 int c, i; 4131 const char *out_fmt = "raw"; 4132 const char *fmt = NULL; 4133 int64_t size = 0; 4134 int64_t block_count = 0, out_pos, in_pos; 4135 struct DdInfo dd = { 4136 .flags = 0, 4137 .count = 0, 4138 }; 4139 struct DdIo in = { 4140 .bsz = 512, /* Block size is by default 512 bytes */ 4141 .filename = NULL, 4142 .buf = NULL, 4143 .offset = 0 4144 }; 4145 struct DdIo out = { 4146 .bsz = 512, 4147 .filename = NULL, 4148 .buf = NULL, 4149 .offset = 0 4150 }; 4151 4152 const struct DdOpts options[] = { 4153 { "bs", img_dd_bs, C_BS }, 4154 { "count", img_dd_count, C_COUNT }, 4155 { "if", img_dd_if, C_IF }, 4156 { "of", img_dd_of, C_OF }, 4157 { "skip", img_dd_skip, C_SKIP }, 4158 { NULL, NULL, 0 } 4159 }; 4160 const struct option long_options[] = { 4161 { "help", no_argument, 0, 'h'}, 4162 { "image-opts", no_argument, 0, OPTION_IMAGE_OPTS}, 4163 { 0, 0, 0, 0 } 4164 }; 4165 4166 while ((c = getopt_long(argc, argv, ":hf:O:", long_options, NULL))) { 4167 if (c == EOF) { 4168 break; 4169 } 4170 switch (c) { 4171 case 'O': 4172 out_fmt = optarg; 4173 break; 4174 case 'f': 4175 fmt = optarg; 4176 break; 4177 case ':': 4178 missing_argument(argv[optind - 1]); 4179 break; 4180 case '?': 4181 unrecognized_option(argv[optind - 1]); 4182 break; 4183 case 'h': 4184 help(); 4185 break; 4186 case OPTION_IMAGE_OPTS: 4187 image_opts = true; 4188 break; 4189 } 4190 } 4191 4192 for (i = optind; i < argc; i++) { 4193 int j; 4194 arg = g_strdup(argv[i]); 4195 4196 tmp = strchr(arg, '='); 4197 if (tmp == NULL) { 4198 error_report("unrecognized operand %s", arg); 4199 ret = -1; 4200 goto out; 4201 } 4202 4203 *tmp++ = '\0'; 4204 4205 for (j = 0; options[j].name != NULL; j++) { 4206 if (!strcmp(arg, options[j].name)) { 4207 break; 4208 } 4209 } 4210 if (options[j].name == NULL) { 4211 error_report("unrecognized operand %s", arg); 4212 ret = -1; 4213 goto out; 4214 } 4215 4216 if (options[j].f(tmp, &in, &out, &dd) != 0) { 4217 ret = -1; 4218 goto out; 4219 } 4220 dd.flags |= options[j].flag; 4221 g_free(arg); 4222 arg = NULL; 4223 } 4224 4225 if (!(dd.flags & C_IF && dd.flags & C_OF)) { 4226 error_report("Must specify both input and output files"); 4227 ret = -1; 4228 goto out; 4229 } 4230 blk1 = img_open(image_opts, in.filename, fmt, 0, false, false); 4231 4232 if (!blk1) { 4233 ret = -1; 4234 goto out; 4235 } 4236 4237 drv = bdrv_find_format(out_fmt); 4238 if (!drv) { 4239 error_report("Unknown file format"); 4240 ret = -1; 4241 goto out; 4242 } 4243 proto_drv = bdrv_find_protocol(out.filename, true, &local_err); 4244 4245 if (!proto_drv) { 4246 error_report_err(local_err); 4247 ret = -1; 4248 goto out; 4249 } 4250 if (!drv->create_opts) { 4251 error_report("Format driver '%s' does not support image creation", 4252 drv->format_name); 4253 ret = -1; 4254 goto out; 4255 } 4256 if (!proto_drv->create_opts) { 4257 error_report("Protocol driver '%s' does not support image creation", 4258 proto_drv->format_name); 4259 ret = -1; 4260 goto out; 4261 } 4262 create_opts = qemu_opts_append(create_opts, drv->create_opts); 4263 create_opts = qemu_opts_append(create_opts, proto_drv->create_opts); 4264 4265 opts = qemu_opts_create(create_opts, NULL, 0, &error_abort); 4266 4267 size = blk_getlength(blk1); 4268 if (size < 0) { 4269 error_report("Failed to get size for '%s'", in.filename); 4270 ret = -1; 4271 goto out; 4272 } 4273 4274 if (dd.flags & C_COUNT && dd.count <= INT64_MAX / in.bsz && 4275 dd.count * in.bsz < size) { 4276 size = dd.count * in.bsz; 4277 } 4278 4279 /* Overflow means the specified offset is beyond input image's size */ 4280 if (dd.flags & C_SKIP && (in.offset > INT64_MAX / in.bsz || 4281 size < in.bsz * in.offset)) { 4282 qemu_opt_set_number(opts, BLOCK_OPT_SIZE, 0, &error_abort); 4283 } else { 4284 qemu_opt_set_number(opts, BLOCK_OPT_SIZE, 4285 size - in.bsz * in.offset, &error_abort); 4286 } 4287 4288 ret = bdrv_create(drv, out.filename, opts, &local_err); 4289 if (ret < 0) { 4290 error_reportf_err(local_err, 4291 "%s: error while creating output image: ", 4292 out.filename); 4293 ret = -1; 4294 goto out; 4295 } 4296 4297 blk2 = img_open(image_opts, out.filename, out_fmt, BDRV_O_RDWR, 4298 false, false); 4299 4300 if (!blk2) { 4301 ret = -1; 4302 goto out; 4303 } 4304 4305 if (dd.flags & C_SKIP && (in.offset > INT64_MAX / in.bsz || 4306 size < in.offset * in.bsz)) { 4307 /* We give a warning if the skip option is bigger than the input 4308 * size and create an empty output disk image (i.e. like dd(1)). 4309 */ 4310 error_report("%s: cannot skip to specified offset", in.filename); 4311 in_pos = size; 4312 } else { 4313 in_pos = in.offset * in.bsz; 4314 } 4315 4316 in.buf = g_new(uint8_t, in.bsz); 4317 4318 for (out_pos = 0; in_pos < size; block_count++) { 4319 int in_ret, out_ret; 4320 4321 if (in_pos + in.bsz > size) { 4322 in_ret = blk_pread(blk1, in_pos, in.buf, size - in_pos); 4323 } else { 4324 in_ret = blk_pread(blk1, in_pos, in.buf, in.bsz); 4325 } 4326 if (in_ret < 0) { 4327 error_report("error while reading from input image file: %s", 4328 strerror(-in_ret)); 4329 ret = -1; 4330 goto out; 4331 } 4332 in_pos += in_ret; 4333 4334 out_ret = blk_pwrite(blk2, out_pos, in.buf, in_ret, 0); 4335 4336 if (out_ret < 0) { 4337 error_report("error while writing to output image file: %s", 4338 strerror(-out_ret)); 4339 ret = -1; 4340 goto out; 4341 } 4342 out_pos += out_ret; 4343 } 4344 4345 out: 4346 g_free(arg); 4347 qemu_opts_del(opts); 4348 qemu_opts_free(create_opts); 4349 blk_unref(blk1); 4350 blk_unref(blk2); 4351 g_free(in.filename); 4352 g_free(out.filename); 4353 g_free(in.buf); 4354 g_free(out.buf); 4355 4356 if (ret) { 4357 return 1; 4358 } 4359 return 0; 4360 } 4361 4362 4363 static const img_cmd_t img_cmds[] = { 4364 #define DEF(option, callback, arg_string) \ 4365 { option, callback }, 4366 #include "qemu-img-cmds.h" 4367 #undef DEF 4368 #undef GEN_DOCS 4369 { NULL, NULL, }, 4370 }; 4371 4372 int main(int argc, char **argv) 4373 { 4374 const img_cmd_t *cmd; 4375 const char *cmdname; 4376 Error *local_error = NULL; 4377 char *trace_file = NULL; 4378 int c; 4379 static const struct option long_options[] = { 4380 {"help", no_argument, 0, 'h'}, 4381 {"version", no_argument, 0, 'V'}, 4382 {"trace", required_argument, NULL, 'T'}, 4383 {0, 0, 0, 0} 4384 }; 4385 4386 #ifdef CONFIG_POSIX 4387 signal(SIGPIPE, SIG_IGN); 4388 #endif 4389 4390 module_call_init(MODULE_INIT_TRACE); 4391 error_set_progname(argv[0]); 4392 qemu_init_exec_dir(argv[0]); 4393 4394 if (qemu_init_main_loop(&local_error)) { 4395 error_report_err(local_error); 4396 exit(EXIT_FAILURE); 4397 } 4398 4399 qcrypto_init(&error_fatal); 4400 4401 module_call_init(MODULE_INIT_QOM); 4402 bdrv_init(); 4403 if (argc < 2) { 4404 error_exit("Not enough arguments"); 4405 } 4406 4407 qemu_add_opts(&qemu_object_opts); 4408 qemu_add_opts(&qemu_source_opts); 4409 qemu_add_opts(&qemu_trace_opts); 4410 4411 while ((c = getopt_long(argc, argv, "+:hVT:", long_options, NULL)) != -1) { 4412 switch (c) { 4413 case ':': 4414 missing_argument(argv[optind - 1]); 4415 return 0; 4416 case '?': 4417 unrecognized_option(argv[optind - 1]); 4418 return 0; 4419 case 'h': 4420 help(); 4421 return 0; 4422 case 'V': 4423 printf(QEMU_IMG_VERSION); 4424 return 0; 4425 case 'T': 4426 g_free(trace_file); 4427 trace_file = trace_opt_parse(optarg); 4428 break; 4429 } 4430 } 4431 4432 cmdname = argv[optind]; 4433 4434 /* reset getopt_long scanning */ 4435 argc -= optind; 4436 if (argc < 1) { 4437 return 0; 4438 } 4439 argv += optind; 4440 optind = 0; 4441 4442 if (!trace_init_backends()) { 4443 exit(1); 4444 } 4445 trace_init_file(trace_file); 4446 qemu_set_log(LOG_TRACE); 4447 4448 /* find the command */ 4449 for (cmd = img_cmds; cmd->name != NULL; cmd++) { 4450 if (!strcmp(cmdname, cmd->name)) { 4451 return cmd->handler(argc, argv); 4452 } 4453 } 4454 4455 /* not found */ 4456 error_exit("Command not found: %s", cmdname); 4457 } 4458