2 * QEMU disk image utility
4 * Copyright (c) 2003-2008 Fabrice Bellard
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:
13 * The above copyright notice and this permission notice shall be included in
14 * all copies or substantial portions of the Software.
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
25 #include "qemu/osdep.h"
28 #include "qemu-common.h"
29 #include "qemu-version.h"
30 #include "qapi/error.h"
31 #include "qapi/qapi-commands-block-core.h"
32 #include "qapi/qapi-visit-block-core.h"
33 #include "qapi/qobject-output-visitor.h"
34 #include "qapi/qmp/qjson.h"
35 #include "qapi/qmp/qdict.h"
36 #include "qemu/cutils.h"
37 #include "qemu/config-file.h"
38 #include "qemu/option.h"
39 #include "qemu/error-report.h"
41 #include "qemu/main-loop.h"
42 #include "qemu/module.h"
43 #include "qemu/sockets.h"
44 #include "qemu/units.h"
45 #include "qom/object_interfaces.h"
46 #include "sysemu/block-backend.h"
47 #include "block/block_int.h"
48 #include "block/blockjob.h"
49 #include "block/qapi.h"
50 #include "crypto/init.h"
51 #include "trace/control.h"
52 #include "qemu/throttle.h"
53 #include "block/throttle-groups.h"
55 #define QEMU_IMG_VERSION "qemu-img version " QEMU_FULL_VERSION \
56 "\n" QEMU_COPYRIGHT "\n"
58 typedef struct img_cmd_t {
60 int (*handler)(int argc, char **argv);
65 OPTION_BACKING_CHAIN = 257,
67 OPTION_IMAGE_OPTS = 259,
69 OPTION_FLUSH_INTERVAL = 261,
70 OPTION_NO_DRAIN = 262,
71 OPTION_TARGET_IMAGE_OPTS = 263,
73 OPTION_PREALLOCATION = 265,
76 OPTION_TARGET_IS_ZERO = 268,
87 typedef enum OutputFormat {
92 /* Default to cache=writeback as data integrity is not important for qemu-img */
93 #define BDRV_DEFAULT_CACHE "writeback"
95 static void format_print(void *opaque, const char *name)
100 static void QEMU_NORETURN GCC_FMT_ATTR(1, 2) error_exit(const char *fmt, ...)
105 error_vreport(fmt, ap);
108 error_printf("Try 'qemu-img --help' for more information\n");
112 static void QEMU_NORETURN missing_argument(const char *option)
114 error_exit("missing argument for option '%s'", option);
117 static void QEMU_NORETURN unrecognized_option(const char *option)
119 error_exit("unrecognized option '%s'", option);
122 /* Please keep in synch with docs/tools/qemu-img.rst */
123 static void QEMU_NORETURN help(void)
125 const char *help_msg =
127 "usage: qemu-img [standard options] command [command options]\n"
128 "QEMU disk image utility\n"
130 " '-h', '--help' display this help and exit\n"
131 " '-V', '--version' output version information and exit\n"
132 " '-T', '--trace' [[enable=]<pattern>][,events=<file>][,file=<file>]\n"
133 " specify tracing options\n"
136 #define DEF(option, callback, arg_string) \
138 #include "qemu-img-cmds.h"
141 "Command parameters:\n"
142 " 'filename' is a disk image filename\n"
143 " 'objectdef' is a QEMU user creatable object definition. See the qemu(1)\n"
144 " manual page for a description of the object properties. The most common\n"
145 " object type is a 'secret', which is used to supply passwords and/or\n"
146 " encryption keys.\n"
147 " 'fmt' is the disk image format. It is guessed automatically in most cases\n"
148 " 'cache' is the cache mode used to write the output disk image, the valid\n"
149 " options are: 'none', 'writeback' (default, except for convert), 'writethrough',\n"
150 " 'directsync' and 'unsafe' (default for convert)\n"
151 " 'src_cache' is the cache mode used to read input disk images, the valid\n"
152 " options are the same as for the 'cache' option\n"
153 " 'size' is the disk image size in bytes. Optional suffixes\n"
154 " 'k' or 'K' (kilobyte, 1024), 'M' (megabyte, 1024k), 'G' (gigabyte, 1024M),\n"
155 " 'T' (terabyte, 1024G), 'P' (petabyte, 1024T) and 'E' (exabyte, 1024P) are\n"
156 " supported. 'b' is ignored.\n"
157 " 'output_filename' is the destination disk image filename\n"
158 " 'output_fmt' is the destination format\n"
159 " 'options' is a comma separated list of format specific options in a\n"
160 " name=value format. Use -o ? for an overview of the options supported by the\n"
162 " 'snapshot_param' is param used for internal snapshot, format\n"
163 " is 'snapshot.id=[ID],snapshot.name=[NAME]', or\n"
165 " '-c' indicates that target image must be compressed (qcow format only)\n"
166 " '-u' allows unsafe backing chains. For rebasing, it is assumed that old and\n"
167 " new backing file match exactly. The image doesn't need a working\n"
168 " backing file before rebasing in this case (useful for renaming the\n"
169 " backing file). For image creation, allow creating without attempting\n"
170 " to open the backing file.\n"
171 " '-h' with or without a command shows this help and lists the supported formats\n"
172 " '-p' show progress of command (only certain commands)\n"
173 " '-q' use Quiet mode - do not print any output (except errors)\n"
174 " '-S' indicates the consecutive number of bytes (defaults to 4k) that must\n"
175 " contain only zeros for qemu-img to create a sparse image during\n"
176 " conversion. If the number of bytes is 0, the source will not be scanned for\n"
177 " unallocated or zero sectors, and the destination image will always be\n"
179 " '--output' takes the format in which the output must be done (human or json)\n"
180 " '-n' skips the target volume creation (useful if the volume is created\n"
181 " prior to running qemu-img)\n"
183 "Parameters to bitmap subcommand:\n"
184 " 'bitmap' is the name of the bitmap to manipulate, through one or more\n"
185 " actions from '--add', '--remove', '--clear', '--enable', '--disable',\n"
186 " or '--merge source'\n"
187 " '-g granularity' sets the granularity for '--add' actions\n"
188 " '-b source' and '-F src_fmt' tell '--merge' actions to find the source\n"
189 " bitmaps from an alternative file\n"
191 "Parameters to check subcommand:\n"
192 " '-r' tries to repair any inconsistencies that are found during the check.\n"
193 " '-r leaks' repairs only cluster leaks, whereas '-r all' fixes all\n"
194 " kinds of errors, with a higher risk of choosing the wrong fix or\n"
195 " hiding corruption that has already occurred.\n"
197 "Parameters to convert subcommand:\n"
198 " '--bitmaps' copies all top-level persistent bitmaps to destination\n"
199 " '-m' specifies how many coroutines work in parallel during the convert\n"
200 " process (defaults to 8)\n"
201 " '-W' allow to write to the target out of order rather than sequential\n"
203 "Parameters to snapshot subcommand:\n"
204 " 'snapshot' is the name of the snapshot to create, apply or delete\n"
205 " '-a' applies a snapshot (revert disk to saved state)\n"
206 " '-c' creates a snapshot\n"
207 " '-d' deletes a snapshot\n"
208 " '-l' lists all snapshots in the given image\n"
210 "Parameters to compare subcommand:\n"
211 " '-f' first image format\n"
212 " '-F' second image format\n"
213 " '-s' run in Strict mode - fail on different image size or sector allocation\n"
215 "Parameters to dd subcommand:\n"
216 " 'bs=BYTES' read and write up to BYTES bytes at a time "
218 " 'count=N' copy only N input blocks\n"
219 " 'if=FILE' read from FILE\n"
220 " 'of=FILE' write to FILE\n"
221 " 'skip=N' skip N bs-sized blocks at the start of input\n";
223 printf("%s\nSupported formats:", help_msg);
224 bdrv_iterate_format(format_print, NULL, false);
225 printf("\n\n" QEMU_HELP_BOTTOM "\n");
230 * Is @optarg safe for accumulate_options()?
231 * It is when multiple of them can be joined together separated by ','.
232 * To make that work, @optarg must not start with ',' (or else a
233 * separating ',' preceding it gets escaped), and it must not end with
234 * an odd number of ',' (or else a separating ',' following it gets
235 * escaped), or be empty (or else a separating ',' preceding it can
236 * escape a separating ',' following it).
239 static bool is_valid_option_list(const char *optarg)
241 size_t len = strlen(optarg);
244 if (!optarg[0] || optarg[0] == ',') {
248 for (i = len; i > 0 && optarg[i - 1] == ','; i--) {
257 static int accumulate_options(char **options, char *optarg)
261 if (!is_valid_option_list(optarg)) {
262 error_report("Invalid option list: %s", optarg);
267 *options = g_strdup(optarg);
269 new_options = g_strdup_printf("%s,%s", *options, optarg);
271 *options = new_options;
276 static QemuOptsList qemu_source_opts = {
278 .implied_opt_name = "file",
279 .head = QTAILQ_HEAD_INITIALIZER(qemu_source_opts.head),
285 static int GCC_FMT_ATTR(2, 3) qprintf(bool quiet, const char *fmt, ...)
291 ret = vprintf(fmt, args);
298 static int print_block_option_help(const char *filename, const char *fmt)
300 BlockDriver *drv, *proto_drv;
301 QemuOptsList *create_opts = NULL;
302 Error *local_err = NULL;
304 /* Find driver and parse its options */
305 drv = bdrv_find_format(fmt);
307 error_report("Unknown file format '%s'", fmt);
311 if (!drv->create_opts) {
312 error_report("Format driver '%s' does not support image creation", fmt);
316 create_opts = qemu_opts_append(create_opts, drv->create_opts);
318 proto_drv = bdrv_find_protocol(filename, true, &local_err);
320 error_report_err(local_err);
321 qemu_opts_free(create_opts);
324 if (!proto_drv->create_opts) {
325 error_report("Protocol driver '%s' does not support image creation",
326 proto_drv->format_name);
327 qemu_opts_free(create_opts);
330 create_opts = qemu_opts_append(create_opts, proto_drv->create_opts);
334 printf("Supported options:\n");
336 printf("Supported %s options:\n", fmt);
338 qemu_opts_print_help(create_opts, false);
339 qemu_opts_free(create_opts);
343 "The protocol level may support further options.\n"
344 "Specify the target filename to include those options.\n");
351 static BlockBackend *img_open_opts(const char *optstr,
352 QemuOpts *opts, int flags, bool writethrough,
353 bool quiet, bool force_share)
356 Error *local_err = NULL;
358 options = qemu_opts_to_qdict(opts, NULL);
360 if (qdict_haskey(options, BDRV_OPT_FORCE_SHARE)
361 && strcmp(qdict_get_str(options, BDRV_OPT_FORCE_SHARE), "on")) {
362 error_report("--force-share/-U conflicts with image options");
363 qobject_unref(options);
366 qdict_put_str(options, BDRV_OPT_FORCE_SHARE, "on");
368 blk = blk_new_open(NULL, NULL, options, flags, &local_err);
370 error_reportf_err(local_err, "Could not open '%s': ", optstr);
373 blk_set_enable_write_cache(blk, !writethrough);
378 static BlockBackend *img_open_file(const char *filename,
380 const char *fmt, int flags,
381 bool writethrough, bool quiet,
385 Error *local_err = NULL;
388 options = qdict_new();
391 qdict_put_str(options, "driver", fmt);
395 qdict_put_bool(options, BDRV_OPT_FORCE_SHARE, true);
397 blk = blk_new_open(filename, NULL, options, flags, &local_err);
399 error_reportf_err(local_err, "Could not open '%s': ", filename);
402 blk_set_enable_write_cache(blk, !writethrough);
408 static int img_add_key_secrets(void *opaque,
409 const char *name, const char *value,
412 QDict *options = opaque;
414 if (g_str_has_suffix(name, "key-secret")) {
415 qdict_put_str(options, name, value);
422 static BlockBackend *img_open(bool image_opts,
423 const char *filename,
424 const char *fmt, int flags, bool writethrough,
425 bool quiet, bool force_share)
431 error_report("--image-opts and --format are mutually exclusive");
434 opts = qemu_opts_parse_noisily(qemu_find_opts("source"),
439 blk = img_open_opts(filename, opts, flags, writethrough, quiet,
442 blk = img_open_file(filename, NULL, fmt, flags, writethrough, quiet,
449 static int add_old_style_options(const char *fmt, QemuOpts *opts,
450 const char *base_filename,
451 const char *base_fmt)
454 if (!qemu_opt_set(opts, BLOCK_OPT_BACKING_FILE, base_filename,
456 error_report("Backing file not supported for file format '%s'",
462 if (!qemu_opt_set(opts, BLOCK_OPT_BACKING_FMT, base_fmt, NULL)) {
463 error_report("Backing file format not supported for file "
471 static int64_t cvtnum_full(const char *name, const char *value, int64_t min,
477 err = qemu_strtosz(value, NULL, &res);
478 if (err < 0 && err != -ERANGE) {
479 error_report("Invalid %s specified. You may use "
480 "k, M, G, T, P or E suffixes for", name);
481 error_report("kilobytes, megabytes, gigabytes, terabytes, "
482 "petabytes and exabytes.");
485 if (err == -ERANGE || res > max || res < min) {
486 error_report("Invalid %s specified. Must be between %" PRId64
487 " and %" PRId64 ".", name, min, max);
493 static int64_t cvtnum(const char *name, const char *value)
495 return cvtnum_full(name, value, 0, INT64_MAX);
498 static int img_create(int argc, char **argv)
501 uint64_t img_size = -1;
502 const char *fmt = "raw";
503 const char *base_fmt = NULL;
504 const char *filename;
505 const char *base_filename = NULL;
506 char *options = NULL;
507 Error *local_err = NULL;
512 static const struct option long_options[] = {
513 {"help", no_argument, 0, 'h'},
514 {"object", required_argument, 0, OPTION_OBJECT},
517 c = getopt_long(argc, argv, ":F:b:f:ho:qu",
524 missing_argument(argv[optind - 1]);
527 unrecognized_option(argv[optind - 1]);
536 base_filename = optarg;
542 if (accumulate_options(&options, optarg) < 0) {
550 flags |= BDRV_O_NO_BACKING;
553 user_creatable_process_cmdline(optarg);
558 /* Get the filename */
559 filename = (optind < argc) ? argv[optind] : NULL;
560 if (options && has_help_option(options)) {
562 return print_block_option_help(filename, fmt);
565 if (optind >= argc) {
566 error_exit("Expecting image file name");
570 /* Get image size, if specified */
574 sval = cvtnum("image size", argv[optind++]);
578 img_size = (uint64_t)sval;
580 if (optind != argc) {
581 error_exit("Unexpected argument: %s", argv[optind]);
584 bdrv_img_create(filename, fmt, base_filename, base_fmt,
585 options, img_size, flags, quiet, &local_err);
587 error_reportf_err(local_err, "%s: ", filename);
599 static void dump_json_image_check(ImageCheck *check, bool quiet)
603 Visitor *v = qobject_output_visitor_new(&obj);
605 visit_type_ImageCheck(v, NULL, &check, &error_abort);
606 visit_complete(v, &obj);
607 str = qobject_to_json_pretty(obj, true);
609 qprintf(quiet, "%s\n", str->str);
612 g_string_free(str, true);
615 static void dump_human_image_check(ImageCheck *check, bool quiet)
617 if (!(check->corruptions || check->leaks || check->check_errors)) {
618 qprintf(quiet, "No errors were found on the image.\n");
620 if (check->corruptions) {
621 qprintf(quiet, "\n%" PRId64 " errors were found on the image.\n"
622 "Data may be corrupted, or further writes to the image "
629 "\n%" PRId64 " leaked clusters were found on the image.\n"
630 "This means waste of disk space, but no harm to data.\n",
634 if (check->check_errors) {
637 " internal errors have occurred during the check.\n",
638 check->check_errors);
642 if (check->total_clusters != 0 && check->allocated_clusters != 0) {
643 qprintf(quiet, "%" PRId64 "/%" PRId64 " = %0.2f%% allocated, "
644 "%0.2f%% fragmented, %0.2f%% compressed clusters\n",
645 check->allocated_clusters, check->total_clusters,
646 check->allocated_clusters * 100.0 / check->total_clusters,
647 check->fragmented_clusters * 100.0 / check->allocated_clusters,
648 check->compressed_clusters * 100.0 /
649 check->allocated_clusters);
652 if (check->image_end_offset) {
654 "Image end offset: %" PRId64 "\n", check->image_end_offset);
658 static int collect_image_check(BlockDriverState *bs,
660 const char *filename,
665 BdrvCheckResult result;
667 ret = bdrv_check(bs, &result, fix);
672 check->filename = g_strdup(filename);
673 check->format = g_strdup(bdrv_get_format_name(bs));
674 check->check_errors = result.check_errors;
675 check->corruptions = result.corruptions;
676 check->has_corruptions = result.corruptions != 0;
677 check->leaks = result.leaks;
678 check->has_leaks = result.leaks != 0;
679 check->corruptions_fixed = result.corruptions_fixed;
680 check->has_corruptions_fixed = result.corruptions_fixed != 0;
681 check->leaks_fixed = result.leaks_fixed;
682 check->has_leaks_fixed = result.leaks_fixed != 0;
683 check->image_end_offset = result.image_end_offset;
684 check->has_image_end_offset = result.image_end_offset != 0;
685 check->total_clusters = result.bfi.total_clusters;
686 check->has_total_clusters = result.bfi.total_clusters != 0;
687 check->allocated_clusters = result.bfi.allocated_clusters;
688 check->has_allocated_clusters = result.bfi.allocated_clusters != 0;
689 check->fragmented_clusters = result.bfi.fragmented_clusters;
690 check->has_fragmented_clusters = result.bfi.fragmented_clusters != 0;
691 check->compressed_clusters = result.bfi.compressed_clusters;
692 check->has_compressed_clusters = result.bfi.compressed_clusters != 0;
698 * Checks an image for consistency. Exit codes:
700 * 0 - Check completed, image is good
701 * 1 - Check not completed because of internal errors
702 * 2 - Check completed, image is corrupted
703 * 3 - Check completed, image has leaked clusters, but is good otherwise
704 * 63 - Checks are not supported by the image format
706 static int img_check(int argc, char **argv)
709 OutputFormat output_format = OFORMAT_HUMAN;
710 const char *filename, *fmt, *output, *cache;
712 BlockDriverState *bs;
714 int flags = BDRV_O_CHECK;
718 bool image_opts = false;
719 bool force_share = false;
723 cache = BDRV_DEFAULT_CACHE;
726 int option_index = 0;
727 static const struct option long_options[] = {
728 {"help", no_argument, 0, 'h'},
729 {"format", required_argument, 0, 'f'},
730 {"repair", required_argument, 0, 'r'},
731 {"output", required_argument, 0, OPTION_OUTPUT},
732 {"object", required_argument, 0, OPTION_OBJECT},
733 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
734 {"force-share", no_argument, 0, 'U'},
737 c = getopt_long(argc, argv, ":hf:r:T:qU",
738 long_options, &option_index);
744 missing_argument(argv[optind - 1]);
747 unrecognized_option(argv[optind - 1]);
756 flags |= BDRV_O_RDWR;
758 if (!strcmp(optarg, "leaks")) {
759 fix = BDRV_FIX_LEAKS;
760 } else if (!strcmp(optarg, "all")) {
761 fix = BDRV_FIX_LEAKS | BDRV_FIX_ERRORS;
763 error_exit("Unknown option value for -r "
764 "(expecting 'leaks' or 'all'): %s", optarg);
780 user_creatable_process_cmdline(optarg);
782 case OPTION_IMAGE_OPTS:
787 if (optind != argc - 1) {
788 error_exit("Expecting one image file name");
790 filename = argv[optind++];
792 if (output && !strcmp(output, "json")) {
793 output_format = OFORMAT_JSON;
794 } else if (output && !strcmp(output, "human")) {
795 output_format = OFORMAT_HUMAN;
797 error_report("--output must be used with human or json as argument.");
801 ret = bdrv_parse_cache_mode(cache, &flags, &writethrough);
803 error_report("Invalid source cache option: %s", cache);
807 blk = img_open(image_opts, filename, fmt, flags, writethrough, quiet,
814 check = g_new0(ImageCheck, 1);
815 ret = collect_image_check(bs, check, filename, fmt, fix);
817 if (ret == -ENOTSUP) {
818 error_report("This image format does not support checks");
823 if (check->corruptions_fixed || check->leaks_fixed) {
824 int corruptions_fixed, leaks_fixed;
825 bool has_leaks_fixed, has_corruptions_fixed;
827 leaks_fixed = check->leaks_fixed;
828 has_leaks_fixed = check->has_leaks_fixed;
829 corruptions_fixed = check->corruptions_fixed;
830 has_corruptions_fixed = check->has_corruptions_fixed;
832 if (output_format == OFORMAT_HUMAN) {
834 "The following inconsistencies were found and repaired:\n\n"
835 " %" PRId64 " leaked clusters\n"
836 " %" PRId64 " corruptions\n\n"
837 "Double checking the fixed image now...\n",
839 check->corruptions_fixed);
842 qapi_free_ImageCheck(check);
843 check = g_new0(ImageCheck, 1);
844 ret = collect_image_check(bs, check, filename, fmt, 0);
846 check->leaks_fixed = leaks_fixed;
847 check->has_leaks_fixed = has_leaks_fixed;
848 check->corruptions_fixed = corruptions_fixed;
849 check->has_corruptions_fixed = has_corruptions_fixed;
853 switch (output_format) {
855 dump_human_image_check(check, quiet);
858 dump_json_image_check(check, quiet);
863 if (ret || check->check_errors) {
865 error_report("Check failed: %s", strerror(-ret));
867 error_report("Check failed");
873 if (check->corruptions) {
875 } else if (check->leaks) {
882 qapi_free_ImageCheck(check);
887 typedef struct CommonBlockJobCBInfo {
888 BlockDriverState *bs;
890 } CommonBlockJobCBInfo;
892 static void common_block_job_cb(void *opaque, int ret)
894 CommonBlockJobCBInfo *cbi = opaque;
897 error_setg_errno(cbi->errp, -ret, "Block job failed");
901 static void run_block_job(BlockJob *job, Error **errp)
903 uint64_t progress_current, progress_total;
904 AioContext *aio_context = blk_get_aio_context(job->blk);
907 aio_context_acquire(aio_context);
910 float progress = 0.0f;
911 aio_poll(aio_context, true);
913 progress_get_snapshot(&job->job.progress, &progress_current,
915 if (progress_total) {
916 progress = (float)progress_current / progress_total * 100.f;
918 qemu_progress_print(progress, 0);
919 } while (!job_is_ready(&job->job) && !job_is_completed(&job->job));
921 if (!job_is_completed(&job->job)) {
922 ret = job_complete_sync(&job->job, errp);
926 job_unref(&job->job);
927 aio_context_release(aio_context);
929 /* publish completion progress only when success */
931 qemu_progress_print(100.f, 0);
935 static int img_commit(int argc, char **argv)
938 const char *filename, *fmt, *cache, *base;
940 BlockDriverState *bs, *base_bs;
942 bool progress = false, quiet = false, drop = false;
944 Error *local_err = NULL;
945 CommonBlockJobCBInfo cbi;
946 bool image_opts = false;
947 AioContext *aio_context;
948 int64_t rate_limit = 0;
951 cache = BDRV_DEFAULT_CACHE;
954 static const struct option long_options[] = {
955 {"help", no_argument, 0, 'h'},
956 {"object", required_argument, 0, OPTION_OBJECT},
957 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
960 c = getopt_long(argc, argv, ":f:ht:b:dpqr:",
967 missing_argument(argv[optind - 1]);
970 unrecognized_option(argv[optind - 1]);
996 rate_limit = cvtnum("rate limit", optarg);
997 if (rate_limit < 0) {
1002 user_creatable_process_cmdline(optarg);
1004 case OPTION_IMAGE_OPTS:
1010 /* Progress is not shown in Quiet mode */
1015 if (optind != argc - 1) {
1016 error_exit("Expecting one image file name");
1018 filename = argv[optind++];
1020 flags = BDRV_O_RDWR | BDRV_O_UNMAP;
1021 ret = bdrv_parse_cache_mode(cache, &flags, &writethrough);
1023 error_report("Invalid cache option: %s", cache);
1027 blk = img_open(image_opts, filename, fmt, flags, writethrough, quiet,
1034 qemu_progress_init(progress, 1.f);
1035 qemu_progress_print(0.f, 100);
1038 base_bs = bdrv_find_backing_image(bs, base);
1040 error_setg(&local_err,
1041 "Did not find '%s' in the backing chain of '%s'",
1046 /* This is different from QMP, which by default uses the deepest file in
1047 * the backing chain (i.e., the very base); however, the traditional
1048 * behavior of qemu-img commit is using the immediate backing file. */
1049 base_bs = bdrv_backing_chain_next(bs);
1051 error_setg(&local_err, "Image does not have a backing file");
1056 cbi = (CommonBlockJobCBInfo){
1061 aio_context = bdrv_get_aio_context(bs);
1062 aio_context_acquire(aio_context);
1063 commit_active_start("commit", bs, base_bs, JOB_DEFAULT, rate_limit,
1064 BLOCKDEV_ON_ERROR_REPORT, NULL, common_block_job_cb,
1065 &cbi, false, &local_err);
1066 aio_context_release(aio_context);
1071 /* When the block job completes, the BlockBackend reference will point to
1072 * the old backing file. In order to avoid that the top image is already
1073 * deleted, so we can still empty it afterwards, increment the reference
1074 * counter here preemptively. */
1079 job = block_job_get("commit");
1081 run_block_job(job, &local_err);
1087 BlockBackend *old_backing_blk;
1089 old_backing_blk = blk_new_with_bs(bs, BLK_PERM_WRITE, BLK_PERM_ALL,
1091 if (!old_backing_blk) {
1094 ret = blk_make_empty(old_backing_blk, &local_err);
1095 blk_unref(old_backing_blk);
1096 if (ret == -ENOTSUP) {
1097 error_free(local_err);
1099 } else if (ret < 0) {
1110 qemu_progress_end();
1115 error_report_err(local_err);
1119 qprintf(quiet, "Image committed.\n");
1124 * Returns -1 if 'buf' contains only zeroes, otherwise the byte index
1125 * of the first sector boundary within buf where the sector contains a
1126 * non-zero byte. This function is robust to a buffer that is not
1129 static int64_t find_nonzero(const uint8_t *buf, int64_t n)
1132 int64_t end = QEMU_ALIGN_DOWN(n, BDRV_SECTOR_SIZE);
1134 for (i = 0; i < end; i += BDRV_SECTOR_SIZE) {
1135 if (!buffer_is_zero(buf + i, BDRV_SECTOR_SIZE)) {
1139 if (i < n && !buffer_is_zero(buf + i, n - end)) {
1146 * Returns true iff the first sector pointed to by 'buf' contains at least
1149 * 'pnum' is set to the number of sectors (including and immediately following
1150 * the first one) that are known to be in the same allocated/unallocated state.
1151 * The function will try to align the end offset to alignment boundaries so
1152 * that the request will at least end aligned and consecutive requests will
1153 * also start at an aligned offset.
1155 static int is_allocated_sectors(const uint8_t *buf, int n, int *pnum,
1156 int64_t sector_num, int alignment)
1165 is_zero = buffer_is_zero(buf, BDRV_SECTOR_SIZE);
1166 for(i = 1; i < n; i++) {
1167 buf += BDRV_SECTOR_SIZE;
1168 if (is_zero != buffer_is_zero(buf, BDRV_SECTOR_SIZE)) {
1173 tail = (sector_num + i) & (alignment - 1);
1175 if (is_zero && i <= tail) {
1176 /* treat unallocated areas which only consist
1177 * of a small tail as allocated. */
1181 /* align up end offset of allocated areas. */
1182 i += alignment - tail;
1185 /* align down end offset of zero areas. */
1194 * Like is_allocated_sectors, but if the buffer starts with a used sector,
1195 * up to 'min' consecutive sectors containing zeros are ignored. This avoids
1196 * breaking up write requests for only small sparse areas.
1198 static int is_allocated_sectors_min(const uint8_t *buf, int n, int *pnum,
1199 int min, int64_t sector_num, int alignment)
1202 int num_checked, num_used;
1208 ret = is_allocated_sectors(buf, n, pnum, sector_num, alignment);
1214 buf += BDRV_SECTOR_SIZE * *pnum;
1216 sector_num += *pnum;
1217 num_checked = num_used;
1220 ret = is_allocated_sectors(buf, n, pnum, sector_num, alignment);
1222 buf += BDRV_SECTOR_SIZE * *pnum;
1224 sector_num += *pnum;
1225 num_checked += *pnum;
1227 num_used = num_checked;
1228 } else if (*pnum >= min) {
1238 * Compares two buffers sector by sector. Returns 0 if the first
1239 * sector of each buffer matches, non-zero otherwise.
1241 * pnum is set to the sector-aligned size of the buffer prefix that
1242 * has the same matching status as the first sector.
1244 static int compare_buffers(const uint8_t *buf1, const uint8_t *buf2,
1245 int64_t bytes, int64_t *pnum)
1248 int64_t i = MIN(bytes, BDRV_SECTOR_SIZE);
1252 res = !!memcmp(buf1, buf2, i);
1254 int64_t len = MIN(bytes - i, BDRV_SECTOR_SIZE);
1256 if (!!memcmp(buf1 + i, buf2 + i, len) != res) {
1266 #define IO_BUF_SIZE (2 * MiB)
1269 * Check if passed sectors are empty (not allocated or contain only 0 bytes)
1271 * Intended for use by 'qemu-img compare': Returns 0 in case sectors are
1272 * filled with 0, 1 if sectors contain non-zero data (this is a comparison
1273 * failure), and 4 on error (the exit status for read errors), after emitting
1276 * @param blk: BlockBackend for the image
1277 * @param offset: Starting offset to check
1278 * @param bytes: Number of bytes to check
1279 * @param filename: Name of disk file we are checking (logging purpose)
1280 * @param buffer: Allocated buffer for storing read data
1281 * @param quiet: Flag for quiet mode
1283 static int check_empty_sectors(BlockBackend *blk, int64_t offset,
1284 int64_t bytes, const char *filename,
1285 uint8_t *buffer, bool quiet)
1290 ret = blk_pread(blk, offset, buffer, bytes);
1292 error_report("Error while reading offset %" PRId64 " of %s: %s",
1293 offset, filename, strerror(-ret));
1296 idx = find_nonzero(buffer, bytes);
1298 qprintf(quiet, "Content mismatch at offset %" PRId64 "!\n",
1307 * Compares two images. Exit codes:
1309 * 0 - Images are identical or the requested help was printed
1311 * >1 - Error occurred
1313 static int img_compare(int argc, char **argv)
1315 const char *fmt1 = NULL, *fmt2 = NULL, *cache, *filename1, *filename2;
1316 BlockBackend *blk1, *blk2;
1317 BlockDriverState *bs1, *bs2;
1318 int64_t total_size1, total_size2;
1319 uint8_t *buf1 = NULL, *buf2 = NULL;
1320 int64_t pnum1, pnum2;
1321 int allocated1, allocated2;
1322 int ret = 0; /* return value - 0 Ident, 1 Different, >1 Error */
1323 bool progress = false, quiet = false, strict = false;
1330 uint64_t progress_base;
1331 bool image_opts = false;
1332 bool force_share = false;
1334 cache = BDRV_DEFAULT_CACHE;
1336 static const struct option long_options[] = {
1337 {"help", no_argument, 0, 'h'},
1338 {"object", required_argument, 0, OPTION_OBJECT},
1339 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
1340 {"force-share", no_argument, 0, 'U'},
1343 c = getopt_long(argc, argv, ":hf:F:T:pqsU",
1344 long_options, NULL);
1350 missing_argument(argv[optind - 1]);
1353 unrecognized_option(argv[optind - 1]);
1381 Error *local_err = NULL;
1383 if (!user_creatable_add_from_str(optarg, &local_err)) {
1385 error_report_err(local_err);
1388 /* Help was printed */
1394 case OPTION_IMAGE_OPTS:
1400 /* Progress is not shown in Quiet mode */
1406 if (optind != argc - 2) {
1407 error_exit("Expecting two image file names");
1409 filename1 = argv[optind++];
1410 filename2 = argv[optind++];
1412 /* Initialize before goto out */
1413 qemu_progress_init(progress, 2.0);
1416 ret = bdrv_parse_cache_mode(cache, &flags, &writethrough);
1418 error_report("Invalid source cache option: %s", cache);
1423 blk1 = img_open(image_opts, filename1, fmt1, flags, writethrough, quiet,
1430 blk2 = img_open(image_opts, filename2, fmt2, flags, writethrough, quiet,
1439 buf1 = blk_blockalign(blk1, IO_BUF_SIZE);
1440 buf2 = blk_blockalign(blk2, IO_BUF_SIZE);
1441 total_size1 = blk_getlength(blk1);
1442 if (total_size1 < 0) {
1443 error_report("Can't get size of %s: %s",
1444 filename1, strerror(-total_size1));
1448 total_size2 = blk_getlength(blk2);
1449 if (total_size2 < 0) {
1450 error_report("Can't get size of %s: %s",
1451 filename2, strerror(-total_size2));
1455 total_size = MIN(total_size1, total_size2);
1456 progress_base = MAX(total_size1, total_size2);
1458 qemu_progress_print(0, 100);
1460 if (strict && total_size1 != total_size2) {
1462 qprintf(quiet, "Strict mode: Image size mismatch!\n");
1466 while (offset < total_size) {
1467 int status1, status2;
1469 status1 = bdrv_block_status_above(bs1, NULL, offset,
1470 total_size1 - offset, &pnum1, NULL,
1474 error_report("Sector allocation test failed for %s", filename1);
1477 allocated1 = status1 & BDRV_BLOCK_ALLOCATED;
1479 status2 = bdrv_block_status_above(bs2, NULL, offset,
1480 total_size2 - offset, &pnum2, NULL,
1484 error_report("Sector allocation test failed for %s", filename2);
1487 allocated2 = status2 & BDRV_BLOCK_ALLOCATED;
1489 assert(pnum1 && pnum2);
1490 chunk = MIN(pnum1, pnum2);
1493 if (status1 != status2) {
1495 qprintf(quiet, "Strict mode: Offset %" PRId64
1496 " block status mismatch!\n", offset);
1500 if ((status1 & BDRV_BLOCK_ZERO) && (status2 & BDRV_BLOCK_ZERO)) {
1502 } else if (allocated1 == allocated2) {
1506 chunk = MIN(chunk, IO_BUF_SIZE);
1507 ret = blk_pread(blk1, offset, buf1, chunk);
1509 error_report("Error while reading offset %" PRId64
1511 offset, filename1, strerror(-ret));
1515 ret = blk_pread(blk2, offset, buf2, chunk);
1517 error_report("Error while reading offset %" PRId64
1519 offset, filename2, strerror(-ret));
1523 ret = compare_buffers(buf1, buf2, chunk, &pnum);
1524 if (ret || pnum != chunk) {
1525 qprintf(quiet, "Content mismatch at offset %" PRId64 "!\n",
1526 offset + (ret ? 0 : pnum));
1532 chunk = MIN(chunk, IO_BUF_SIZE);
1534 ret = check_empty_sectors(blk1, offset, chunk,
1535 filename1, buf1, quiet);
1537 ret = check_empty_sectors(blk2, offset, chunk,
1538 filename2, buf1, quiet);
1545 qemu_progress_print(((float) chunk / progress_base) * 100, 100);
1548 if (total_size1 != total_size2) {
1549 BlockBackend *blk_over;
1550 const char *filename_over;
1552 qprintf(quiet, "Warning: Image size mismatch!\n");
1553 if (total_size1 > total_size2) {
1555 filename_over = filename1;
1558 filename_over = filename2;
1561 while (offset < progress_base) {
1562 ret = bdrv_block_status_above(blk_bs(blk_over), NULL, offset,
1563 progress_base - offset, &chunk,
1567 error_report("Sector allocation test failed for %s",
1572 if (ret & BDRV_BLOCK_ALLOCATED && !(ret & BDRV_BLOCK_ZERO)) {
1573 chunk = MIN(chunk, IO_BUF_SIZE);
1574 ret = check_empty_sectors(blk_over, offset, chunk,
1575 filename_over, buf1, quiet);
1581 qemu_progress_print(((float) chunk / progress_base) * 100, 100);
1585 qprintf(quiet, "Images are identical.\n");
1595 qemu_progress_end();
1599 /* Convenience wrapper around qmp_block_dirty_bitmap_merge */
1600 static void do_dirty_bitmap_merge(const char *dst_node, const char *dst_name,
1601 const char *src_node, const char *src_name,
1604 BlockDirtyBitmapMergeSource *merge_src;
1605 BlockDirtyBitmapMergeSourceList *list = NULL;
1607 merge_src = g_new0(BlockDirtyBitmapMergeSource, 1);
1608 merge_src->type = QTYPE_QDICT;
1609 merge_src->u.external.node = g_strdup(src_node);
1610 merge_src->u.external.name = g_strdup(src_name);
1611 QAPI_LIST_PREPEND(list, merge_src);
1612 qmp_block_dirty_bitmap_merge(dst_node, dst_name, list, errp);
1613 qapi_free_BlockDirtyBitmapMergeSourceList(list);
1616 enum ImgConvertBlockStatus {
1622 #define MAX_COROUTINES 16
1623 #define CONVERT_THROTTLE_GROUP "img_convert"
1625 typedef struct ImgConvertState {
1627 int64_t *src_sectors;
1630 int64_t total_sectors;
1631 int64_t allocated_sectors;
1632 int64_t allocated_done;
1635 enum ImgConvertBlockStatus status;
1636 int64_t sector_next_status;
1637 BlockBackend *target;
1641 bool target_has_backing;
1642 int64_t target_backing_sectors; /* negative if unknown */
1649 size_t cluster_sectors;
1651 long num_coroutines;
1652 int running_coroutines;
1653 Coroutine *co[MAX_COROUTINES];
1654 int64_t wait_sector_num[MAX_COROUTINES];
1659 static void convert_select_part(ImgConvertState *s, int64_t sector_num,
1660 int *src_cur, int64_t *src_cur_offset)
1663 *src_cur_offset = 0;
1664 while (sector_num - *src_cur_offset >= s->src_sectors[*src_cur]) {
1665 *src_cur_offset += s->src_sectors[*src_cur];
1667 assert(*src_cur < s->src_num);
1671 static int convert_iteration_sectors(ImgConvertState *s, int64_t sector_num)
1673 int64_t src_cur_offset;
1674 int ret, n, src_cur;
1675 bool post_backing_zero = false;
1677 convert_select_part(s, sector_num, &src_cur, &src_cur_offset);
1679 assert(s->total_sectors > sector_num);
1680 n = MIN(s->total_sectors - sector_num, BDRV_REQUEST_MAX_SECTORS);
1682 if (s->target_backing_sectors >= 0) {
1683 if (sector_num >= s->target_backing_sectors) {
1684 post_backing_zero = true;
1685 } else if (sector_num + n > s->target_backing_sectors) {
1686 /* Split requests around target_backing_sectors (because
1687 * starting from there, zeros are handled differently) */
1688 n = s->target_backing_sectors - sector_num;
1692 if (s->sector_next_status <= sector_num) {
1693 uint64_t offset = (sector_num - src_cur_offset) * BDRV_SECTOR_SIZE;
1696 BlockDriverState *src_bs = blk_bs(s->src[src_cur]);
1697 BlockDriverState *base;
1699 if (s->target_has_backing) {
1700 base = bdrv_cow_bs(bdrv_skip_filters(src_bs));
1706 count = n * BDRV_SECTOR_SIZE;
1708 ret = bdrv_block_status_above(src_bs, base, offset, count, &count,
1715 warn_report("error while reading block status at "
1716 "offset %" PRIu64 ": %s", offset,
1719 /* Just try to read the data, then */
1720 ret = BDRV_BLOCK_DATA;
1721 count = BDRV_SECTOR_SIZE;
1723 /* Retry on a shorter range */
1724 n = DIV_ROUND_UP(n, 4);
1727 error_report("error while reading block status at offset "
1728 "%" PRIu64 ": %s", offset, strerror(-ret));
1734 n = DIV_ROUND_UP(count, BDRV_SECTOR_SIZE);
1737 * Avoid that s->sector_next_status becomes unaligned to the source
1738 * request alignment and/or cluster size to avoid unnecessary read
1741 tail = (sector_num - src_cur_offset + n) % s->src_alignment[src_cur];
1746 if (ret & BDRV_BLOCK_ZERO) {
1747 s->status = post_backing_zero ? BLK_BACKING_FILE : BLK_ZERO;
1748 } else if (ret & BDRV_BLOCK_DATA) {
1749 s->status = BLK_DATA;
1751 s->status = s->target_has_backing ? BLK_BACKING_FILE : BLK_DATA;
1754 s->sector_next_status = sector_num + n;
1757 n = MIN(n, s->sector_next_status - sector_num);
1758 if (s->status == BLK_DATA) {
1759 n = MIN(n, s->buf_sectors);
1762 /* We need to write complete clusters for compressed images, so if an
1763 * unallocated area is shorter than that, we must consider the whole
1764 * cluster allocated. */
1765 if (s->compressed) {
1766 if (n < s->cluster_sectors) {
1767 n = MIN(s->cluster_sectors, s->total_sectors - sector_num);
1768 s->status = BLK_DATA;
1770 n = QEMU_ALIGN_DOWN(n, s->cluster_sectors);
1777 static int coroutine_fn convert_co_read(ImgConvertState *s, int64_t sector_num,
1778 int nb_sectors, uint8_t *buf)
1780 uint64_t single_read_until = 0;
1783 assert(nb_sectors <= s->buf_sectors);
1784 while (nb_sectors > 0) {
1787 int64_t bs_sectors, src_cur_offset;
1790 /* In the case of compression with multiple source files, we can get a
1791 * nb_sectors that spreads into the next part. So we must be able to
1792 * read across multiple BDSes for one convert_read() call. */
1793 convert_select_part(s, sector_num, &src_cur, &src_cur_offset);
1794 blk = s->src[src_cur];
1795 bs_sectors = s->src_sectors[src_cur];
1797 offset = (sector_num - src_cur_offset) << BDRV_SECTOR_BITS;
1799 n = MIN(nb_sectors, bs_sectors - (sector_num - src_cur_offset));
1800 if (single_read_until > offset) {
1804 ret = blk_co_pread(blk, offset, n << BDRV_SECTOR_BITS, buf, 0);
1808 single_read_until = offset + (n << BDRV_SECTOR_BITS);
1812 warn_report("error while reading offset %" PRIu64
1813 ": %s", offset, strerror(-ret));
1815 memset(buf, 0, BDRV_SECTOR_SIZE);
1824 buf += n * BDRV_SECTOR_SIZE;
1831 static int coroutine_fn convert_co_write(ImgConvertState *s, int64_t sector_num,
1832 int nb_sectors, uint8_t *buf,
1833 enum ImgConvertBlockStatus status)
1837 while (nb_sectors > 0) {
1839 BdrvRequestFlags flags = s->compressed ? BDRV_REQ_WRITE_COMPRESSED : 0;
1842 case BLK_BACKING_FILE:
1843 /* If we have a backing file, leave clusters unallocated that are
1844 * unallocated in the source image, so that the backing file is
1845 * visible at the respective offset. */
1846 assert(s->target_has_backing);
1850 /* If we're told to keep the target fully allocated (-S 0) or there
1851 * is real non-zero data, we must write it. Otherwise we can treat
1852 * it as zero sectors.
1853 * Compressed clusters need to be written as a whole, so in that
1854 * case we can only save the write if the buffer is completely
1856 if (!s->min_sparse ||
1858 is_allocated_sectors_min(buf, n, &n, s->min_sparse,
1859 sector_num, s->alignment)) ||
1861 !buffer_is_zero(buf, n * BDRV_SECTOR_SIZE)))
1863 ret = blk_co_pwrite(s->target, sector_num << BDRV_SECTOR_BITS,
1864 n << BDRV_SECTOR_BITS, buf, flags);
1873 if (s->has_zero_init) {
1874 assert(!s->target_has_backing);
1877 ret = blk_co_pwrite_zeroes(s->target,
1878 sector_num << BDRV_SECTOR_BITS,
1879 n << BDRV_SECTOR_BITS,
1880 BDRV_REQ_MAY_UNMAP);
1889 buf += n * BDRV_SECTOR_SIZE;
1895 static int coroutine_fn convert_co_copy_range(ImgConvertState *s, int64_t sector_num,
1900 while (nb_sectors > 0) {
1903 int64_t bs_sectors, src_cur_offset;
1906 convert_select_part(s, sector_num, &src_cur, &src_cur_offset);
1907 offset = (sector_num - src_cur_offset) << BDRV_SECTOR_BITS;
1908 blk = s->src[src_cur];
1909 bs_sectors = s->src_sectors[src_cur];
1911 n = MIN(nb_sectors, bs_sectors - (sector_num - src_cur_offset));
1913 ret = blk_co_copy_range(blk, offset, s->target,
1914 sector_num << BDRV_SECTOR_BITS,
1915 n << BDRV_SECTOR_BITS, 0, 0);
1926 static void coroutine_fn convert_co_do_copy(void *opaque)
1928 ImgConvertState *s = opaque;
1929 uint8_t *buf = NULL;
1933 for (i = 0; i < s->num_coroutines; i++) {
1934 if (s->co[i] == qemu_coroutine_self()) {
1941 s->running_coroutines++;
1942 buf = blk_blockalign(s->target, s->buf_sectors * BDRV_SECTOR_SIZE);
1947 enum ImgConvertBlockStatus status;
1950 qemu_co_mutex_lock(&s->lock);
1951 if (s->ret != -EINPROGRESS || s->sector_num >= s->total_sectors) {
1952 qemu_co_mutex_unlock(&s->lock);
1955 n = convert_iteration_sectors(s, s->sector_num);
1957 qemu_co_mutex_unlock(&s->lock);
1961 /* save current sector and allocation status to local variables */
1962 sector_num = s->sector_num;
1964 if (!s->min_sparse && s->status == BLK_ZERO) {
1965 n = MIN(n, s->buf_sectors);
1967 /* increment global sector counter so that other coroutines can
1968 * already continue reading beyond this request */
1970 qemu_co_mutex_unlock(&s->lock);
1972 if (status == BLK_DATA || (!s->min_sparse && status == BLK_ZERO)) {
1973 s->allocated_done += n;
1974 qemu_progress_print(100.0 * s->allocated_done /
1975 s->allocated_sectors, 0);
1979 copy_range = s->copy_range && s->status == BLK_DATA;
1980 if (status == BLK_DATA && !copy_range) {
1981 ret = convert_co_read(s, sector_num, n, buf);
1983 error_report("error while reading at byte %lld: %s",
1984 sector_num * BDRV_SECTOR_SIZE, strerror(-ret));
1987 } else if (!s->min_sparse && status == BLK_ZERO) {
1989 memset(buf, 0x00, n * BDRV_SECTOR_SIZE);
1992 if (s->wr_in_order) {
1993 /* keep writes in order */
1994 while (s->wr_offs != sector_num && s->ret == -EINPROGRESS) {
1995 s->wait_sector_num[index] = sector_num;
1996 qemu_coroutine_yield();
1998 s->wait_sector_num[index] = -1;
2001 if (s->ret == -EINPROGRESS) {
2003 ret = convert_co_copy_range(s, sector_num, n);
2005 s->copy_range = false;
2009 ret = convert_co_write(s, sector_num, n, buf, status);
2012 error_report("error while writing at byte %lld: %s",
2013 sector_num * BDRV_SECTOR_SIZE, strerror(-ret));
2018 if (s->wr_in_order) {
2019 /* reenter the coroutine that might have waited
2020 * for this write to complete */
2021 s->wr_offs = sector_num + n;
2022 for (i = 0; i < s->num_coroutines; i++) {
2023 if (s->co[i] && s->wait_sector_num[i] == s->wr_offs) {
2025 * A -> B -> A cannot occur because A has
2026 * s->wait_sector_num[i] == -1 during A -> B. Therefore
2027 * B will never enter A during this time window.
2029 qemu_coroutine_enter(s->co[i]);
2037 s->co[index] = NULL;
2038 s->running_coroutines--;
2039 if (!s->running_coroutines && s->ret == -EINPROGRESS) {
2040 /* the convert job finished successfully */
2045 static int convert_do_copy(ImgConvertState *s)
2048 int64_t sector_num = 0;
2050 /* Check whether we have zero initialisation or can get it efficiently */
2051 if (!s->has_zero_init && s->target_is_new && s->min_sparse &&
2052 !s->target_has_backing) {
2053 s->has_zero_init = bdrv_has_zero_init(blk_bs(s->target));
2056 /* Allocate buffer for copied data. For compressed images, only one cluster
2057 * can be copied at a time. */
2058 if (s->compressed) {
2059 if (s->cluster_sectors <= 0 || s->cluster_sectors > s->buf_sectors) {
2060 error_report("invalid cluster size");
2063 s->buf_sectors = s->cluster_sectors;
2066 while (sector_num < s->total_sectors) {
2067 n = convert_iteration_sectors(s, sector_num);
2071 if (s->status == BLK_DATA || (!s->min_sparse && s->status == BLK_ZERO))
2073 s->allocated_sectors += n;
2079 s->sector_next_status = 0;
2080 s->ret = -EINPROGRESS;
2082 qemu_co_mutex_init(&s->lock);
2083 for (i = 0; i < s->num_coroutines; i++) {
2084 s->co[i] = qemu_coroutine_create(convert_co_do_copy, s);
2085 s->wait_sector_num[i] = -1;
2086 qemu_coroutine_enter(s->co[i]);
2089 while (s->running_coroutines) {
2090 main_loop_wait(false);
2093 if (s->compressed && !s->ret) {
2094 /* signal EOF to align */
2095 ret = blk_pwrite_compressed(s->target, 0, NULL, 0);
2104 static int convert_copy_bitmaps(BlockDriverState *src, BlockDriverState *dst)
2106 BdrvDirtyBitmap *bm;
2109 FOR_EACH_DIRTY_BITMAP(src, bm) {
2112 if (!bdrv_dirty_bitmap_get_persistence(bm)) {
2115 name = bdrv_dirty_bitmap_name(bm);
2116 qmp_block_dirty_bitmap_add(dst->node_name, name,
2117 true, bdrv_dirty_bitmap_granularity(bm),
2119 true, !bdrv_dirty_bitmap_enabled(bm),
2122 error_reportf_err(err, "Failed to create bitmap %s: ", name);
2126 do_dirty_bitmap_merge(dst->node_name, name, src->node_name, name,
2129 error_reportf_err(err, "Failed to populate bitmap %s: ", name);
2137 #define MAX_BUF_SECTORS 32768
2139 static void set_rate_limit(BlockBackend *blk, int64_t rate_limit)
2143 throttle_config_init(&cfg);
2144 cfg.buckets[THROTTLE_BPS_WRITE].avg = rate_limit;
2146 blk_io_limits_enable(blk, CONVERT_THROTTLE_GROUP);
2147 blk_set_io_limits(blk, &cfg);
2150 static int img_convert(int argc, char **argv)
2152 int c, bs_i, flags, src_flags = BDRV_O_NO_SHARE;
2153 const char *fmt = NULL, *out_fmt = NULL, *cache = "unsafe",
2154 *src_cache = BDRV_DEFAULT_CACHE, *out_baseimg = NULL,
2155 *out_filename, *out_baseimg_param, *snapshot_name = NULL;
2156 BlockDriver *drv = NULL, *proto_drv = NULL;
2157 BlockDriverInfo bdi;
2158 BlockDriverState *out_bs;
2159 QemuOpts *opts = NULL, *sn_opts = NULL;
2160 QemuOptsList *create_opts = NULL;
2161 QDict *open_opts = NULL;
2162 char *options = NULL;
2163 Error *local_err = NULL;
2164 bool writethrough, src_writethrough, image_opts = false,
2165 skip_create = false, progress = false, tgt_image_opts = false;
2166 int64_t ret = -EINVAL;
2167 bool force_share = false;
2168 bool explict_min_sparse = false;
2169 bool bitmaps = false;
2170 int64_t rate_limit = 0;
2172 ImgConvertState s = (ImgConvertState) {
2173 /* Need at least 4k of zeros for sparse detection */
2175 .copy_range = false,
2176 .buf_sectors = IO_BUF_SIZE / BDRV_SECTOR_SIZE,
2177 .wr_in_order = true,
2178 .num_coroutines = 8,
2182 static const struct option long_options[] = {
2183 {"help", no_argument, 0, 'h'},
2184 {"object", required_argument, 0, OPTION_OBJECT},
2185 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
2186 {"force-share", no_argument, 0, 'U'},
2187 {"target-image-opts", no_argument, 0, OPTION_TARGET_IMAGE_OPTS},
2188 {"salvage", no_argument, 0, OPTION_SALVAGE},
2189 {"target-is-zero", no_argument, 0, OPTION_TARGET_IS_ZERO},
2190 {"bitmaps", no_argument, 0, OPTION_BITMAPS},
2193 c = getopt_long(argc, argv, ":hf:O:B:Cco:l:S:pt:T:qnm:WUr:",
2194 long_options, NULL);
2200 missing_argument(argv[optind - 1]);
2203 unrecognized_option(argv[optind - 1]);
2215 out_baseimg = optarg;
2218 s.copy_range = true;
2221 s.compressed = true;
2224 if (accumulate_options(&options, optarg) < 0) {
2229 if (strstart(optarg, SNAPSHOT_OPT_BASE, NULL)) {
2230 sn_opts = qemu_opts_parse_noisily(&internal_snapshot_opts,
2233 error_report("Failed in parsing snapshot param '%s'",
2238 snapshot_name = optarg;
2245 sval = cvtnum("buffer size for sparse output", optarg);
2248 } else if (!QEMU_IS_ALIGNED(sval, BDRV_SECTOR_SIZE) ||
2249 sval / BDRV_SECTOR_SIZE > MAX_BUF_SECTORS) {
2250 error_report("Invalid buffer size for sparse output specified. "
2251 "Valid sizes are multiples of %llu up to %llu. Select "
2252 "0 to disable sparse detection (fully allocates output).",
2253 BDRV_SECTOR_SIZE, MAX_BUF_SECTORS * BDRV_SECTOR_SIZE);
2257 s.min_sparse = sval / BDRV_SECTOR_SIZE;
2258 explict_min_sparse = true;
2277 if (qemu_strtol(optarg, NULL, 0, &s.num_coroutines) ||
2278 s.num_coroutines < 1 || s.num_coroutines > MAX_COROUTINES) {
2279 error_report("Invalid number of coroutines. Allowed number of"
2280 " coroutines is between 1 and %d", MAX_COROUTINES);
2285 s.wr_in_order = false;
2291 rate_limit = cvtnum("rate limit", optarg);
2292 if (rate_limit < 0) {
2297 user_creatable_process_cmdline(optarg);
2299 case OPTION_IMAGE_OPTS:
2302 case OPTION_SALVAGE:
2305 case OPTION_TARGET_IMAGE_OPTS:
2306 tgt_image_opts = true;
2308 case OPTION_TARGET_IS_ZERO:
2310 * The user asserting that the target is blank has the
2311 * same effect as the target driver supporting zero
2314 s.has_zero_init = true;
2316 case OPTION_BITMAPS:
2322 if (!out_fmt && !tgt_image_opts) {
2326 if (s.compressed && s.copy_range) {
2327 error_report("Cannot enable copy offloading when -c is used");
2331 if (explict_min_sparse && s.copy_range) {
2332 error_report("Cannot enable copy offloading when -S is used");
2336 if (s.copy_range && s.salvage) {
2337 error_report("Cannot use copy offloading in salvaging mode");
2341 if (tgt_image_opts && !skip_create) {
2342 error_report("--target-image-opts requires use of -n flag");
2346 if (skip_create && options) {
2347 error_report("-o has no effect when skipping image creation");
2351 if (s.has_zero_init && !skip_create) {
2352 error_report("--target-is-zero requires use of -n flag");
2356 s.src_num = argc - optind - 1;
2357 out_filename = s.src_num >= 1 ? argv[argc - 1] : NULL;
2359 if (options && has_help_option(options)) {
2361 ret = print_block_option_help(out_filename, out_fmt);
2364 error_report("Option help requires a format be specified");
2369 if (s.src_num < 1) {
2370 error_report("Must specify image file name");
2374 /* ret is still -EINVAL until here */
2375 ret = bdrv_parse_cache_mode(src_cache, &src_flags, &src_writethrough);
2377 error_report("Invalid source cache option: %s", src_cache);
2381 /* Initialize before goto out */
2385 qemu_progress_init(progress, 1.0);
2386 qemu_progress_print(0, 100);
2388 s.src = g_new0(BlockBackend *, s.src_num);
2389 s.src_sectors = g_new(int64_t, s.src_num);
2390 s.src_alignment = g_new(int, s.src_num);
2392 for (bs_i = 0; bs_i < s.src_num; bs_i++) {
2393 BlockDriverState *src_bs;
2394 s.src[bs_i] = img_open(image_opts, argv[optind + bs_i],
2395 fmt, src_flags, src_writethrough, s.quiet,
2401 s.src_sectors[bs_i] = blk_nb_sectors(s.src[bs_i]);
2402 if (s.src_sectors[bs_i] < 0) {
2403 error_report("Could not get size of %s: %s",
2404 argv[optind + bs_i], strerror(-s.src_sectors[bs_i]));
2408 src_bs = blk_bs(s.src[bs_i]);
2409 s.src_alignment[bs_i] = DIV_ROUND_UP(src_bs->bl.request_alignment,
2411 if (!bdrv_get_info(src_bs, &bdi)) {
2412 s.src_alignment[bs_i] = MAX(s.src_alignment[bs_i],
2413 bdi.cluster_size / BDRV_SECTOR_SIZE);
2415 s.total_sectors += s.src_sectors[bs_i];
2419 bdrv_snapshot_load_tmp(blk_bs(s.src[0]),
2420 qemu_opt_get(sn_opts, SNAPSHOT_OPT_ID),
2421 qemu_opt_get(sn_opts, SNAPSHOT_OPT_NAME),
2423 } else if (snapshot_name != NULL) {
2424 if (s.src_num > 1) {
2425 error_report("No support for concatenating multiple snapshot");
2430 bdrv_snapshot_load_tmp_by_id_or_name(blk_bs(s.src[0]), snapshot_name,
2434 error_reportf_err(local_err, "Failed to load snapshot: ");
2440 /* Find driver and parse its options */
2441 drv = bdrv_find_format(out_fmt);
2443 error_report("Unknown file format '%s'", out_fmt);
2448 proto_drv = bdrv_find_protocol(out_filename, true, &local_err);
2450 error_report_err(local_err);
2455 if (!drv->create_opts) {
2456 error_report("Format driver '%s' does not support image creation",
2462 if (!proto_drv->create_opts) {
2463 error_report("Protocol driver '%s' does not support image creation",
2464 proto_drv->format_name);
2469 create_opts = qemu_opts_append(create_opts, drv->create_opts);
2470 create_opts = qemu_opts_append(create_opts, proto_drv->create_opts);
2472 opts = qemu_opts_create(create_opts, NULL, 0, &error_abort);
2474 if (!qemu_opts_do_parse(opts, options, NULL, &local_err)) {
2475 error_report_err(local_err);
2481 qemu_opt_set_number(opts, BLOCK_OPT_SIZE,
2482 s.total_sectors * BDRV_SECTOR_SIZE, &error_abort);
2483 ret = add_old_style_options(out_fmt, opts, out_baseimg, NULL);
2489 /* Get backing file name if -o backing_file was used */
2490 out_baseimg_param = qemu_opt_get(opts, BLOCK_OPT_BACKING_FILE);
2491 if (out_baseimg_param) {
2492 out_baseimg = out_baseimg_param;
2494 s.target_has_backing = (bool) out_baseimg;
2496 if (s.has_zero_init && s.target_has_backing) {
2497 error_report("Cannot use --target-is-zero when the destination "
2498 "image has a backing file");
2502 if (s.src_num > 1 && out_baseimg) {
2503 error_report("Having a backing file for the target makes no sense when "
2504 "concatenating multiple input images");
2509 if (out_baseimg_param) {
2510 if (!qemu_opt_get(opts, BLOCK_OPT_BACKING_FMT)) {
2511 warn_report("Deprecated use of backing file without explicit "
2516 /* Check if compression is supported */
2519 qemu_opt_get_bool(opts, BLOCK_OPT_ENCRYPT, false);
2520 const char *encryptfmt =
2521 qemu_opt_get(opts, BLOCK_OPT_ENCRYPT_FORMAT);
2522 const char *preallocation =
2523 qemu_opt_get(opts, BLOCK_OPT_PREALLOC);
2525 if (drv && !block_driver_can_compress(drv)) {
2526 error_report("Compression not supported for this file format");
2531 if (encryption || encryptfmt) {
2532 error_report("Compression and encryption not supported at "
2539 && strcmp(preallocation, "off"))
2541 error_report("Compression and preallocation not supported at "
2548 /* Determine if bitmaps need copying */
2550 if (s.src_num > 1) {
2551 error_report("Copying bitmaps only possible with single source");
2555 if (!bdrv_supports_persistent_dirty_bitmap(blk_bs(s.src[0]))) {
2556 error_report("Source lacks bitmap support");
2563 * The later open call will need any decryption secrets, and
2564 * bdrv_create() will purge "opts", so extract them now before
2568 open_opts = qdict_new();
2569 qemu_opt_foreach(opts, img_add_key_secrets, open_opts, &error_abort);
2571 /* Create the new image */
2572 ret = bdrv_create(drv, out_filename, opts, &local_err);
2574 error_reportf_err(local_err, "%s: error while converting %s: ",
2575 out_filename, out_fmt);
2580 s.target_is_new = !skip_create;
2582 flags = s.min_sparse ? (BDRV_O_RDWR | BDRV_O_UNMAP) : BDRV_O_RDWR;
2583 ret = bdrv_parse_cache_mode(cache, &flags, &writethrough);
2585 error_report("Invalid cache option: %s", cache);
2590 s.target = img_open(tgt_image_opts, out_filename, out_fmt,
2591 flags, writethrough, s.quiet, false);
2593 /* TODO ultimately we should allow --target-image-opts
2594 * to be used even when -n is not given.
2595 * That has to wait for bdrv_create to be improved
2596 * to allow filenames in option syntax
2598 s.target = img_open_file(out_filename, open_opts, out_fmt,
2599 flags, writethrough, s.quiet, false);
2600 open_opts = NULL; /* blk_new_open will have freed it */
2606 out_bs = blk_bs(s.target);
2608 if (bitmaps && !bdrv_supports_persistent_dirty_bitmap(out_bs)) {
2609 error_report("Format driver '%s' does not support bitmaps",
2610 out_bs->drv->format_name);
2615 if (s.compressed && !block_driver_can_compress(out_bs->drv)) {
2616 error_report("Compression not supported for this file format");
2621 /* increase bufsectors from the default 4096 (2M) if opt_transfer
2622 * or discard_alignment of the out_bs is greater. Limit to
2623 * MAX_BUF_SECTORS as maximum which is currently 32768 (16MB). */
2624 s.buf_sectors = MIN(MAX_BUF_SECTORS,
2626 MAX(out_bs->bl.opt_transfer >> BDRV_SECTOR_BITS,
2627 out_bs->bl.pdiscard_alignment >>
2628 BDRV_SECTOR_BITS)));
2630 /* try to align the write requests to the destination to avoid unnecessary
2632 s.alignment = MAX(pow2floor(s.min_sparse),
2633 DIV_ROUND_UP(out_bs->bl.request_alignment,
2635 assert(is_power_of_2(s.alignment));
2638 int64_t output_sectors = blk_nb_sectors(s.target);
2639 if (output_sectors < 0) {
2640 error_report("unable to get output image length: %s",
2641 strerror(-output_sectors));
2644 } else if (output_sectors < s.total_sectors) {
2645 error_report("output file is smaller than input file");
2651 if (s.target_has_backing && s.target_is_new) {
2652 /* Errors are treated as "backing length unknown" (which means
2653 * s.target_backing_sectors has to be negative, which it will
2654 * be automatically). The backing file length is used only
2655 * for optimizations, so such a case is not fatal. */
2656 s.target_backing_sectors =
2657 bdrv_nb_sectors(bdrv_backing_chain_next(out_bs));
2659 s.target_backing_sectors = -1;
2662 ret = bdrv_get_info(out_bs, &bdi);
2665 error_report("could not get block driver info");
2669 s.compressed = s.compressed || bdi.needs_compressed_writes;
2670 s.cluster_sectors = bdi.cluster_size / BDRV_SECTOR_SIZE;
2674 set_rate_limit(s.target, rate_limit);
2677 ret = convert_do_copy(&s);
2679 /* Now copy the bitmaps */
2680 if (bitmaps && ret == 0) {
2681 ret = convert_copy_bitmaps(blk_bs(s.src[0]), out_bs);
2686 qemu_progress_print(100, 0);
2688 qemu_progress_end();
2689 qemu_opts_del(opts);
2690 qemu_opts_free(create_opts);
2691 qobject_unref(open_opts);
2692 blk_unref(s.target);
2694 for (bs_i = 0; bs_i < s.src_num; bs_i++) {
2695 blk_unref(s.src[bs_i]);
2699 g_free(s.src_sectors);
2700 g_free(s.src_alignment);
2702 qemu_opts_del(sn_opts);
2709 static void dump_snapshots(BlockDriverState *bs)
2711 QEMUSnapshotInfo *sn_tab, *sn;
2714 nb_sns = bdrv_snapshot_list(bs, &sn_tab);
2717 printf("Snapshot list:\n");
2718 bdrv_snapshot_dump(NULL);
2720 for(i = 0; i < nb_sns; i++) {
2722 bdrv_snapshot_dump(sn);
2728 static void dump_json_image_info_list(ImageInfoList *list)
2732 Visitor *v = qobject_output_visitor_new(&obj);
2734 visit_type_ImageInfoList(v, NULL, &list, &error_abort);
2735 visit_complete(v, &obj);
2736 str = qobject_to_json_pretty(obj, true);
2737 assert(str != NULL);
2738 printf("%s\n", str->str);
2741 g_string_free(str, true);
2744 static void dump_json_image_info(ImageInfo *info)
2748 Visitor *v = qobject_output_visitor_new(&obj);
2750 visit_type_ImageInfo(v, NULL, &info, &error_abort);
2751 visit_complete(v, &obj);
2752 str = qobject_to_json_pretty(obj, true);
2753 assert(str != NULL);
2754 printf("%s\n", str->str);
2757 g_string_free(str, true);
2760 static void dump_human_image_info_list(ImageInfoList *list)
2762 ImageInfoList *elem;
2765 for (elem = list; elem; elem = elem->next) {
2771 bdrv_image_info_dump(elem->value);
2775 static gboolean str_equal_func(gconstpointer a, gconstpointer b)
2777 return strcmp(a, b) == 0;
2781 * Open an image file chain and return an ImageInfoList
2783 * @filename: topmost image filename
2784 * @fmt: topmost image format (may be NULL to autodetect)
2785 * @chain: true - enumerate entire backing file chain
2786 * false - only topmost image file
2788 * Returns a list of ImageInfo objects or NULL if there was an error opening an
2789 * image file. If there was an error a message will have been printed to
2792 static ImageInfoList *collect_image_info_list(bool image_opts,
2793 const char *filename,
2795 bool chain, bool force_share)
2797 ImageInfoList *head = NULL;
2798 ImageInfoList **tail = &head;
2799 GHashTable *filenames;
2802 filenames = g_hash_table_new_full(g_str_hash, str_equal_func, NULL, NULL);
2806 BlockDriverState *bs;
2809 if (g_hash_table_lookup_extended(filenames, filename, NULL, NULL)) {
2810 error_report("Backing file '%s' creates an infinite loop.",
2814 g_hash_table_insert(filenames, (gpointer)filename, NULL);
2816 blk = img_open(image_opts, filename, fmt,
2817 BDRV_O_NO_BACKING | BDRV_O_NO_IO, false, false,
2824 bdrv_query_image_info(bs, &info, &err);
2826 error_report_err(err);
2831 QAPI_LIST_APPEND(tail, info);
2835 /* Clear parameters that only apply to the topmost image */
2836 filename = fmt = NULL;
2840 if (info->has_full_backing_filename) {
2841 filename = info->full_backing_filename;
2842 } else if (info->has_backing_filename) {
2843 error_report("Could not determine absolute backing filename,"
2844 " but backing filename '%s' present",
2845 info->backing_filename);
2848 if (info->has_backing_filename_format) {
2849 fmt = info->backing_filename_format;
2853 g_hash_table_destroy(filenames);
2857 qapi_free_ImageInfoList(head);
2858 g_hash_table_destroy(filenames);
2862 static int img_info(int argc, char **argv)
2865 OutputFormat output_format = OFORMAT_HUMAN;
2867 const char *filename, *fmt, *output;
2868 ImageInfoList *list;
2869 bool image_opts = false;
2870 bool force_share = false;
2875 int option_index = 0;
2876 static const struct option long_options[] = {
2877 {"help", no_argument, 0, 'h'},
2878 {"format", required_argument, 0, 'f'},
2879 {"output", required_argument, 0, OPTION_OUTPUT},
2880 {"backing-chain", no_argument, 0, OPTION_BACKING_CHAIN},
2881 {"object", required_argument, 0, OPTION_OBJECT},
2882 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
2883 {"force-share", no_argument, 0, 'U'},
2886 c = getopt_long(argc, argv, ":f:hU",
2887 long_options, &option_index);
2893 missing_argument(argv[optind - 1]);
2896 unrecognized_option(argv[optind - 1]);
2910 case OPTION_BACKING_CHAIN:
2914 user_creatable_process_cmdline(optarg);
2916 case OPTION_IMAGE_OPTS:
2921 if (optind != argc - 1) {
2922 error_exit("Expecting one image file name");
2924 filename = argv[optind++];
2926 if (output && !strcmp(output, "json")) {
2927 output_format = OFORMAT_JSON;
2928 } else if (output && !strcmp(output, "human")) {
2929 output_format = OFORMAT_HUMAN;
2930 } else if (output) {
2931 error_report("--output must be used with human or json as argument.");
2935 list = collect_image_info_list(image_opts, filename, fmt, chain,
2941 switch (output_format) {
2943 dump_human_image_info_list(list);
2947 dump_json_image_info_list(list);
2949 dump_json_image_info(list->value);
2954 qapi_free_ImageInfoList(list);
2958 static int dump_map_entry(OutputFormat output_format, MapEntry *e,
2961 switch (output_format) {
2963 if (e->data && !e->has_offset) {
2964 error_report("File contains external, encrypted or compressed clusters.");
2967 if (e->data && !e->zero) {
2968 printf("%#-16"PRIx64"%#-16"PRIx64"%#-16"PRIx64"%s\n",
2969 e->start, e->length,
2970 e->has_offset ? e->offset : 0,
2971 e->has_filename ? e->filename : "");
2973 /* This format ignores the distinction between 0, ZERO and ZERO|DATA.
2974 * Modify the flags here to allow more coalescing.
2976 if (next && (!next->data || next->zero)) {
2982 printf("{ \"start\": %"PRId64", \"length\": %"PRId64","
2983 " \"depth\": %"PRId64", \"zero\": %s, \"data\": %s",
2984 e->start, e->length, e->depth,
2985 e->zero ? "true" : "false",
2986 e->data ? "true" : "false");
2987 if (e->has_offset) {
2988 printf(", \"offset\": %"PRId64"", e->offset);
3000 static int get_block_status(BlockDriverState *bs, int64_t offset,
3001 int64_t bytes, MapEntry *e)
3005 BlockDriverState *file;
3008 char *filename = NULL;
3010 /* As an optimization, we could cache the current range of unallocated
3011 * clusters in each file of the chain, and avoid querying the same
3017 bs = bdrv_skip_filters(bs);
3018 ret = bdrv_block_status(bs, offset, bytes, &bytes, &map, &file);
3023 if (ret & (BDRV_BLOCK_ZERO|BDRV_BLOCK_DATA)) {
3026 bs = bdrv_cow_bs(bs);
3035 has_offset = !!(ret & BDRV_BLOCK_OFFSET_VALID);
3037 if (file && has_offset) {
3038 bdrv_refresh_filename(file);
3039 filename = file->filename;
3045 .data = !!(ret & BDRV_BLOCK_DATA),
3046 .zero = !!(ret & BDRV_BLOCK_ZERO),
3048 .has_offset = has_offset,
3050 .has_filename = filename,
3051 .filename = filename,
3057 static inline bool entry_mergeable(const MapEntry *curr, const MapEntry *next)
3059 if (curr->length == 0) {
3062 if (curr->zero != next->zero ||
3063 curr->data != next->data ||
3064 curr->depth != next->depth ||
3065 curr->has_filename != next->has_filename ||
3066 curr->has_offset != next->has_offset) {
3069 if (curr->has_filename && strcmp(curr->filename, next->filename)) {
3072 if (curr->has_offset && curr->offset + curr->length != next->offset) {
3078 static int img_map(int argc, char **argv)
3081 OutputFormat output_format = OFORMAT_HUMAN;
3083 BlockDriverState *bs;
3084 const char *filename, *fmt, *output;
3086 MapEntry curr = { .length = 0 }, next;
3088 bool image_opts = false;
3089 bool force_share = false;
3090 int64_t start_offset = 0;
3091 int64_t max_length = -1;
3096 int option_index = 0;
3097 static const struct option long_options[] = {
3098 {"help", no_argument, 0, 'h'},
3099 {"format", required_argument, 0, 'f'},
3100 {"output", required_argument, 0, OPTION_OUTPUT},
3101 {"object", required_argument, 0, OPTION_OBJECT},
3102 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
3103 {"force-share", no_argument, 0, 'U'},
3104 {"start-offset", required_argument, 0, 's'},
3105 {"max-length", required_argument, 0, 'l'},
3108 c = getopt_long(argc, argv, ":f:s:l:hU",
3109 long_options, &option_index);
3115 missing_argument(argv[optind - 1]);
3118 unrecognized_option(argv[optind - 1]);
3133 start_offset = cvtnum("start offset", optarg);
3134 if (start_offset < 0) {
3139 max_length = cvtnum("max length", optarg);
3140 if (max_length < 0) {
3145 user_creatable_process_cmdline(optarg);
3147 case OPTION_IMAGE_OPTS:
3152 if (optind != argc - 1) {
3153 error_exit("Expecting one image file name");
3155 filename = argv[optind];
3157 if (output && !strcmp(output, "json")) {
3158 output_format = OFORMAT_JSON;
3159 } else if (output && !strcmp(output, "human")) {
3160 output_format = OFORMAT_HUMAN;
3161 } else if (output) {
3162 error_report("--output must be used with human or json as argument.");
3166 blk = img_open(image_opts, filename, fmt, 0, false, false, force_share);
3172 if (output_format == OFORMAT_HUMAN) {
3173 printf("%-16s%-16s%-16s%s\n", "Offset", "Length", "Mapped to", "File");
3174 } else if (output_format == OFORMAT_JSON) {
3178 length = blk_getlength(blk);
3180 error_report("Failed to get size for '%s'", filename);
3183 if (max_length != -1) {
3184 length = MIN(start_offset + max_length, length);
3187 curr.start = start_offset;
3188 while (curr.start + curr.length < length) {
3189 int64_t offset = curr.start + curr.length;
3190 int64_t n = length - offset;
3192 ret = get_block_status(bs, offset, n, &next);
3194 error_report("Could not read file metadata: %s", strerror(-ret));
3198 if (entry_mergeable(&curr, &next)) {
3199 curr.length += next.length;
3203 if (curr.length > 0) {
3204 ret = dump_map_entry(output_format, &curr, &next);
3212 ret = dump_map_entry(output_format, &curr, NULL);
3213 if (output_format == OFORMAT_JSON) {
3222 #define SNAPSHOT_LIST 1
3223 #define SNAPSHOT_CREATE 2
3224 #define SNAPSHOT_APPLY 3
3225 #define SNAPSHOT_DELETE 4
3227 static int img_snapshot(int argc, char **argv)
3230 BlockDriverState *bs;
3231 QEMUSnapshotInfo sn;
3232 char *filename, *snapshot_name = NULL;
3233 int c, ret = 0, bdrv_oflags;
3238 bool image_opts = false;
3239 bool force_share = false;
3241 bdrv_oflags = BDRV_O_RDWR;
3242 /* Parse commandline parameters */
3244 static const struct option long_options[] = {
3245 {"help", no_argument, 0, 'h'},
3246 {"object", required_argument, 0, OPTION_OBJECT},
3247 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
3248 {"force-share", no_argument, 0, 'U'},
3251 c = getopt_long(argc, argv, ":la:c:d:hqU",
3252 long_options, NULL);
3258 missing_argument(argv[optind - 1]);
3261 unrecognized_option(argv[optind - 1]);
3268 error_exit("Cannot mix '-l', '-a', '-c', '-d'");
3271 action = SNAPSHOT_LIST;
3272 bdrv_oflags &= ~BDRV_O_RDWR; /* no need for RW */
3276 error_exit("Cannot mix '-l', '-a', '-c', '-d'");
3279 action = SNAPSHOT_APPLY;
3280 snapshot_name = optarg;
3284 error_exit("Cannot mix '-l', '-a', '-c', '-d'");
3287 action = SNAPSHOT_CREATE;
3288 snapshot_name = optarg;
3292 error_exit("Cannot mix '-l', '-a', '-c', '-d'");
3295 action = SNAPSHOT_DELETE;
3296 snapshot_name = optarg;
3305 user_creatable_process_cmdline(optarg);
3307 case OPTION_IMAGE_OPTS:
3313 if (optind != argc - 1) {
3314 error_exit("Expecting one image file name");
3316 filename = argv[optind++];
3318 /* Open the image */
3319 blk = img_open(image_opts, filename, NULL, bdrv_oflags, false, quiet,
3326 /* Perform the requested action */
3332 case SNAPSHOT_CREATE:
3333 memset(&sn, 0, sizeof(sn));
3334 pstrcpy(sn.name, sizeof(sn.name), snapshot_name);
3336 qemu_gettimeofday(&tv);
3337 sn.date_sec = tv.tv_sec;
3338 sn.date_nsec = tv.tv_usec * 1000;
3340 ret = bdrv_snapshot_create(bs, &sn);
3342 error_report("Could not create snapshot '%s': %d (%s)",
3343 snapshot_name, ret, strerror(-ret));
3347 case SNAPSHOT_APPLY:
3348 ret = bdrv_snapshot_goto(bs, snapshot_name, &err);
3350 error_reportf_err(err, "Could not apply snapshot '%s': ",
3355 case SNAPSHOT_DELETE:
3356 ret = bdrv_snapshot_find(bs, &sn, snapshot_name);
3358 error_report("Could not delete snapshot '%s': snapshot not "
3359 "found", snapshot_name);
3362 ret = bdrv_snapshot_delete(bs, sn.id_str, sn.name, &err);
3364 error_reportf_err(err, "Could not delete snapshot '%s': ",
3380 static int img_rebase(int argc, char **argv)
3382 BlockBackend *blk = NULL, *blk_old_backing = NULL, *blk_new_backing = NULL;
3383 uint8_t *buf_old = NULL;
3384 uint8_t *buf_new = NULL;
3385 BlockDriverState *bs = NULL, *prefix_chain_bs = NULL;
3386 BlockDriverState *unfiltered_bs;
3388 const char *fmt, *cache, *src_cache, *out_basefmt, *out_baseimg;
3389 int c, flags, src_flags, ret;
3390 bool writethrough, src_writethrough;
3392 bool force_share = false;
3395 Error *local_err = NULL;
3396 bool image_opts = false;
3398 /* Parse commandline parameters */
3400 cache = BDRV_DEFAULT_CACHE;
3401 src_cache = BDRV_DEFAULT_CACHE;
3405 static const struct option long_options[] = {
3406 {"help", no_argument, 0, 'h'},
3407 {"object", required_argument, 0, OPTION_OBJECT},
3408 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
3409 {"force-share", no_argument, 0, 'U'},
3412 c = getopt_long(argc, argv, ":hf:F:b:upt:T:qU",
3413 long_options, NULL);
3419 missing_argument(argv[optind - 1]);
3422 unrecognized_option(argv[optind - 1]);
3431 out_basefmt = optarg;
3434 out_baseimg = optarg;
3452 user_creatable_process_cmdline(optarg);
3454 case OPTION_IMAGE_OPTS:
3467 if (optind != argc - 1) {
3468 error_exit("Expecting one image file name");
3470 if (!unsafe && !out_baseimg) {
3471 error_exit("Must specify backing file (-b) or use unsafe mode (-u)");
3473 filename = argv[optind++];
3475 qemu_progress_init(progress, 2.0);
3476 qemu_progress_print(0, 100);
3478 flags = BDRV_O_RDWR | (unsafe ? BDRV_O_NO_BACKING : 0);
3479 ret = bdrv_parse_cache_mode(cache, &flags, &writethrough);
3481 error_report("Invalid cache option: %s", cache);
3486 ret = bdrv_parse_cache_mode(src_cache, &src_flags, &src_writethrough);
3488 error_report("Invalid source cache option: %s", src_cache);
3492 /* The source files are opened read-only, don't care about WCE */
3493 assert((src_flags & BDRV_O_RDWR) == 0);
3494 (void) src_writethrough;
3499 * Ignore the old backing file for unsafe rebase in case we want to correct
3500 * the reference to a renamed or moved backing file.
3502 blk = img_open(image_opts, filename, fmt, flags, writethrough, quiet,
3510 unfiltered_bs = bdrv_skip_filters(bs);
3512 if (out_basefmt != NULL) {
3513 if (bdrv_find_format(out_basefmt) == NULL) {
3514 error_report("Invalid format name: '%s'", out_basefmt);
3520 /* For safe rebasing we need to compare old and new backing file */
3522 QDict *options = NULL;
3523 BlockDriverState *base_bs = bdrv_cow_bs(unfiltered_bs);
3526 blk_old_backing = blk_new(qemu_get_aio_context(),
3527 BLK_PERM_CONSISTENT_READ,
3529 ret = blk_insert_bs(blk_old_backing, base_bs,
3532 error_reportf_err(local_err,
3533 "Could not reuse old backing file '%s': ",
3538 blk_old_backing = NULL;
3541 if (out_baseimg[0]) {
3542 const char *overlay_filename;
3543 char *out_real_path;
3545 options = qdict_new();
3547 qdict_put_str(options, "driver", out_basefmt);
3550 qdict_put_bool(options, BDRV_OPT_FORCE_SHARE, true);
3553 bdrv_refresh_filename(bs);
3554 overlay_filename = bs->exact_filename[0] ? bs->exact_filename
3557 bdrv_get_full_backing_filename_from_filename(overlay_filename,
3561 qobject_unref(options);
3562 error_reportf_err(local_err,
3563 "Could not resolve backing filename: ");
3569 * Find out whether we rebase an image on top of a previous image
3572 prefix_chain_bs = bdrv_find_backing_image(bs, out_real_path);
3573 if (prefix_chain_bs) {
3574 qobject_unref(options);
3575 g_free(out_real_path);
3577 blk_new_backing = blk_new(qemu_get_aio_context(),
3578 BLK_PERM_CONSISTENT_READ,
3580 ret = blk_insert_bs(blk_new_backing, prefix_chain_bs,
3583 error_reportf_err(local_err,
3584 "Could not reuse backing file '%s': ",
3589 blk_new_backing = blk_new_open(out_real_path, NULL,
3590 options, src_flags, &local_err);
3591 g_free(out_real_path);
3592 if (!blk_new_backing) {
3593 error_reportf_err(local_err,
3594 "Could not open new backing file '%s': ",
3604 * Check each unallocated cluster in the COW file. If it is unallocated,
3605 * accesses go to the backing file. We must therefore compare this cluster
3606 * in the old and new backing file, and if they differ we need to copy it
3607 * from the old backing file into the COW file.
3609 * If qemu-img crashes during this step, no harm is done. The content of
3610 * the image is the same as the original one at any time.
3614 int64_t old_backing_size = 0;
3615 int64_t new_backing_size = 0;
3618 float local_progress = 0;
3620 buf_old = blk_blockalign(blk, IO_BUF_SIZE);
3621 buf_new = blk_blockalign(blk, IO_BUF_SIZE);
3623 size = blk_getlength(blk);
3625 error_report("Could not get size of '%s': %s",
3626 filename, strerror(-size));
3630 if (blk_old_backing) {
3631 old_backing_size = blk_getlength(blk_old_backing);
3632 if (old_backing_size < 0) {
3633 char backing_name[PATH_MAX];
3635 bdrv_get_backing_filename(bs, backing_name,
3636 sizeof(backing_name));
3637 error_report("Could not get size of '%s': %s",
3638 backing_name, strerror(-old_backing_size));
3643 if (blk_new_backing) {
3644 new_backing_size = blk_getlength(blk_new_backing);
3645 if (new_backing_size < 0) {
3646 error_report("Could not get size of '%s': %s",
3647 out_baseimg, strerror(-new_backing_size));
3654 local_progress = (float)100 / (size / MIN(size, IO_BUF_SIZE));
3657 for (offset = 0; offset < size; offset += n) {
3658 bool buf_old_is_zero = false;
3660 /* How many bytes can we handle with the next read? */
3661 n = MIN(IO_BUF_SIZE, size - offset);
3663 /* If the cluster is allocated, we don't need to take action */
3664 ret = bdrv_is_allocated(unfiltered_bs, offset, n, &n);
3666 error_report("error while reading image metadata: %s",
3674 if (prefix_chain_bs) {
3676 * If cluster wasn't changed since prefix_chain, we don't need
3679 ret = bdrv_is_allocated_above(bdrv_cow_bs(unfiltered_bs),
3680 prefix_chain_bs, false,
3683 error_report("error while reading image metadata: %s",
3693 * Read old and new backing file and take into consideration that
3694 * backing files may be smaller than the COW image.
3696 if (offset >= old_backing_size) {
3697 memset(buf_old, 0, n);
3698 buf_old_is_zero = true;
3700 if (offset + n > old_backing_size) {
3701 n = old_backing_size - offset;
3704 ret = blk_pread(blk_old_backing, offset, buf_old, n);
3706 error_report("error while reading from old backing file");
3711 if (offset >= new_backing_size || !blk_new_backing) {
3712 memset(buf_new, 0, n);
3714 if (offset + n > new_backing_size) {
3715 n = new_backing_size - offset;
3718 ret = blk_pread(blk_new_backing, offset, buf_new, n);
3720 error_report("error while reading from new backing file");
3725 /* If they differ, we need to write to the COW file */
3726 uint64_t written = 0;
3728 while (written < n) {
3731 if (compare_buffers(buf_old + written, buf_new + written,
3732 n - written, &pnum))
3734 if (buf_old_is_zero) {
3735 ret = blk_pwrite_zeroes(blk, offset + written, pnum, 0);
3737 ret = blk_pwrite(blk, offset + written,
3738 buf_old + written, pnum, 0);
3741 error_report("Error while writing to COW image: %s",
3749 qemu_progress_print(local_progress, 100);
3754 * Change the backing file. All clusters that are different from the old
3755 * backing file are overwritten in the COW file now, so the visible content
3756 * doesn't change when we switch the backing file.
3758 if (out_baseimg && *out_baseimg) {
3759 ret = bdrv_change_backing_file(unfiltered_bs, out_baseimg, out_basefmt,
3762 ret = bdrv_change_backing_file(unfiltered_bs, NULL, NULL, false);
3765 if (ret == -ENOSPC) {
3766 error_report("Could not change the backing file to '%s': No "
3767 "space left in the file header", out_baseimg);
3768 } else if (ret < 0) {
3769 error_report("Could not change the backing file to '%s': %s",
3770 out_baseimg, strerror(-ret));
3773 qemu_progress_print(100, 0);
3775 * TODO At this point it is possible to check if any clusters that are
3776 * allocated in the COW file are the same in the backing file. If so, they
3777 * could be dropped from the COW file. Don't do this before switching the
3778 * backing file, in case of a crash this would lead to corruption.
3781 qemu_progress_end();
3784 blk_unref(blk_old_backing);
3785 blk_unref(blk_new_backing);
3787 qemu_vfree(buf_old);
3788 qemu_vfree(buf_new);
3797 static int img_resize(int argc, char **argv)
3800 int c, ret, relative;
3801 const char *filename, *fmt, *size;
3802 int64_t n, total_size, current_size;
3804 BlockBackend *blk = NULL;
3805 PreallocMode prealloc = PREALLOC_MODE_OFF;
3808 static QemuOptsList resize_options = {
3809 .name = "resize_options",
3810 .head = QTAILQ_HEAD_INITIALIZER(resize_options.head),
3813 .name = BLOCK_OPT_SIZE,
3814 .type = QEMU_OPT_SIZE,
3815 .help = "Virtual disk size"
3821 bool image_opts = false;
3822 bool shrink = false;
3824 /* Remove size from argv manually so that negative numbers are not treated
3825 * as options by getopt. */
3827 error_exit("Not enough arguments");
3831 size = argv[--argc];
3833 /* Parse getopt arguments */
3836 static const struct option long_options[] = {
3837 {"help", no_argument, 0, 'h'},
3838 {"object", required_argument, 0, OPTION_OBJECT},
3839 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
3840 {"preallocation", required_argument, 0, OPTION_PREALLOCATION},
3841 {"shrink", no_argument, 0, OPTION_SHRINK},
3844 c = getopt_long(argc, argv, ":f:hq",
3845 long_options, NULL);
3851 missing_argument(argv[optind - 1]);
3854 unrecognized_option(argv[optind - 1]);
3866 user_creatable_process_cmdline(optarg);
3868 case OPTION_IMAGE_OPTS:
3871 case OPTION_PREALLOCATION:
3872 prealloc = qapi_enum_parse(&PreallocMode_lookup, optarg,
3873 PREALLOC_MODE__MAX, NULL);
3874 if (prealloc == PREALLOC_MODE__MAX) {
3875 error_report("Invalid preallocation mode '%s'", optarg);
3884 if (optind != argc - 1) {
3885 error_exit("Expecting image file name and size");
3887 filename = argv[optind++];
3889 /* Choose grow, shrink, or absolute resize mode */
3905 param = qemu_opts_create(&resize_options, NULL, 0, &error_abort);
3906 if (!qemu_opt_set(param, BLOCK_OPT_SIZE, size, &err)) {
3907 error_report_err(err);
3909 qemu_opts_del(param);
3912 n = qemu_opt_get_size(param, BLOCK_OPT_SIZE, 0);
3913 qemu_opts_del(param);
3915 blk = img_open(image_opts, filename, fmt,
3916 BDRV_O_RDWR | BDRV_O_RESIZE, false, quiet,
3923 current_size = blk_getlength(blk);
3924 if (current_size < 0) {
3925 error_report("Failed to inquire current image length: %s",
3926 strerror(-current_size));
3932 total_size = current_size + n * relative;
3936 if (total_size <= 0) {
3937 error_report("New image size must be positive");
3942 if (total_size <= current_size && prealloc != PREALLOC_MODE_OFF) {
3943 error_report("Preallocation can only be used for growing images");
3948 if (total_size < current_size && !shrink) {
3949 error_report("Use the --shrink option to perform a shrink operation.");
3950 warn_report("Shrinking an image will delete all data beyond the "
3951 "shrunken image's end. Before performing such an "
3952 "operation, make sure there is no important data there.");
3958 * The user expects the image to have the desired size after
3959 * resizing, so pass @exact=true. It is of no use to report
3960 * success when the image has not actually been resized.
3962 ret = blk_truncate(blk, total_size, true, prealloc, 0, &err);
3964 qprintf(quiet, "Image resized.\n");
3966 error_report_err(err);
3976 static void amend_status_cb(BlockDriverState *bs,
3977 int64_t offset, int64_t total_work_size,
3980 qemu_progress_print(100.f * offset / total_work_size, 0);
3983 static int print_amend_option_help(const char *format)
3987 /* Find driver and parse its options */
3988 drv = bdrv_find_format(format);
3990 error_report("Unknown file format '%s'", format);
3994 if (!drv->bdrv_amend_options) {
3995 error_report("Format driver '%s' does not support option amendment",
4000 /* Every driver supporting amendment must have amend_opts */
4001 assert(drv->amend_opts);
4003 printf("Amend options for '%s':\n", format);
4004 qemu_opts_print_help(drv->amend_opts, false);
4008 static int img_amend(int argc, char **argv)
4012 char *options = NULL;
4013 QemuOptsList *amend_opts = NULL;
4014 QemuOpts *opts = NULL;
4015 const char *fmt = NULL, *filename, *cache;
4018 bool quiet = false, progress = false;
4019 BlockBackend *blk = NULL;
4020 BlockDriverState *bs = NULL;
4021 bool image_opts = false;
4024 cache = BDRV_DEFAULT_CACHE;
4026 static const struct option long_options[] = {
4027 {"help", no_argument, 0, 'h'},
4028 {"object", required_argument, 0, OPTION_OBJECT},
4029 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
4030 {"force", no_argument, 0, OPTION_FORCE},
4033 c = getopt_long(argc, argv, ":ho:f:t:pq",
4034 long_options, NULL);
4041 missing_argument(argv[optind - 1]);
4044 unrecognized_option(argv[optind - 1]);
4050 if (accumulate_options(&options, optarg) < 0) {
4052 goto out_no_progress;
4068 user_creatable_process_cmdline(optarg);
4070 case OPTION_IMAGE_OPTS:
4080 error_exit("Must specify options (-o)");
4086 qemu_progress_init(progress, 1.0);
4088 filename = (optind == argc - 1) ? argv[argc - 1] : NULL;
4089 if (fmt && has_help_option(options)) {
4090 /* If a format is explicitly specified (and possibly no filename is
4091 * given), print option help here */
4092 ret = print_amend_option_help(fmt);
4096 if (optind != argc - 1) {
4097 error_report("Expecting one image file name");
4102 flags = BDRV_O_RDWR;
4103 ret = bdrv_parse_cache_mode(cache, &flags, &writethrough);
4105 error_report("Invalid cache option: %s", cache);
4109 blk = img_open(image_opts, filename, fmt, flags, writethrough, quiet,
4117 fmt = bs->drv->format_name;
4119 if (has_help_option(options)) {
4120 /* If the format was auto-detected, print option help here */
4121 ret = print_amend_option_help(fmt);
4125 if (!bs->drv->bdrv_amend_options) {
4126 error_report("Format driver '%s' does not support option amendment",
4132 /* Every driver supporting amendment must have amend_opts */
4133 assert(bs->drv->amend_opts);
4135 amend_opts = qemu_opts_append(amend_opts, bs->drv->amend_opts);
4136 opts = qemu_opts_create(amend_opts, NULL, 0, &error_abort);
4137 if (!qemu_opts_do_parse(opts, options, NULL, &err)) {
4138 /* Try to parse options using the create options */
4139 amend_opts = qemu_opts_append(amend_opts, bs->drv->create_opts);
4140 qemu_opts_del(opts);
4141 opts = qemu_opts_create(amend_opts, NULL, 0, &error_abort);
4142 if (qemu_opts_do_parse(opts, options, NULL, NULL)) {
4143 error_append_hint(&err,
4144 "This option is only supported for image creation\n");
4147 error_report_err(err);
4152 /* In case the driver does not call amend_status_cb() */
4153 qemu_progress_print(0.f, 0);
4154 ret = bdrv_amend_options(bs, opts, &amend_status_cb, NULL, force, &err);
4155 qemu_progress_print(100.f, 0);
4157 error_report_err(err);
4162 qemu_progress_end();
4166 qemu_opts_del(opts);
4167 qemu_opts_free(amend_opts);
4176 typedef struct BenchData {
4178 uint64_t image_size;
4185 bool drain_on_flush;
4194 static void bench_undrained_flush_cb(void *opaque, int ret)
4197 error_report("Failed flush request: %s", strerror(-ret));
4202 static void bench_cb(void *opaque, int ret)
4204 BenchData *b = opaque;
4208 error_report("Failed request: %s", strerror(-ret));
4213 /* Just finished a flush with drained queue: Start next requests */
4214 assert(b->in_flight == 0);
4215 b->in_flush = false;
4216 } else if (b->in_flight > 0) {
4217 int remaining = b->n - b->in_flight;
4222 /* Time for flush? Drain queue if requested, then flush */
4223 if (b->flush_interval && remaining % b->flush_interval == 0) {
4224 if (!b->in_flight || !b->drain_on_flush) {
4225 BlockCompletionFunc *cb;
4227 if (b->drain_on_flush) {
4231 cb = bench_undrained_flush_cb;
4234 acb = blk_aio_flush(b->blk, cb, b);
4236 error_report("Failed to issue flush request");
4240 if (b->drain_on_flush) {
4246 while (b->n > b->in_flight && b->in_flight < b->nrreq) {
4247 int64_t offset = b->offset;
4248 /* blk_aio_* might look for completed I/Os and kick bench_cb
4249 * again, so make sure this operation is counted by in_flight
4250 * and b->offset is ready for the next submission.
4253 b->offset += b->step;
4254 b->offset %= b->image_size;
4256 acb = blk_aio_pwritev(b->blk, offset, b->qiov, 0, bench_cb, b);
4258 acb = blk_aio_preadv(b->blk, offset, b->qiov, 0, bench_cb, b);
4261 error_report("Failed to issue request");
4267 static int img_bench(int argc, char **argv)
4270 const char *fmt = NULL, *filename;
4272 bool image_opts = false;
4273 bool is_write = false;
4277 size_t bufsize = 4096;
4280 int flush_interval = 0;
4281 bool drain_on_flush = true;
4283 BlockBackend *blk = NULL;
4284 BenchData data = {};
4286 bool writethrough = false;
4287 struct timeval t1, t2;
4289 bool force_share = false;
4293 static const struct option long_options[] = {
4294 {"help", no_argument, 0, 'h'},
4295 {"flush-interval", required_argument, 0, OPTION_FLUSH_INTERVAL},
4296 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
4297 {"pattern", required_argument, 0, OPTION_PATTERN},
4298 {"no-drain", no_argument, 0, OPTION_NO_DRAIN},
4299 {"force-share", no_argument, 0, 'U'},
4302 c = getopt_long(argc, argv, ":hc:d:f:ni:o:qs:S:t:wU", long_options,
4310 missing_argument(argv[optind - 1]);
4313 unrecognized_option(argv[optind - 1]);
4322 if (qemu_strtoul(optarg, NULL, 0, &res) < 0 || res > INT_MAX) {
4323 error_report("Invalid request count specified");
4333 if (qemu_strtoul(optarg, NULL, 0, &res) < 0 || res > INT_MAX) {
4334 error_report("Invalid queue depth specified");
4344 flags |= BDRV_O_NATIVE_AIO;
4347 ret = bdrv_parse_aio(optarg, &flags);
4349 error_report("Invalid aio option: %s", optarg);
4356 offset = cvtnum("offset", optarg);
4370 sval = cvtnum_full("buffer size", optarg, 0, INT_MAX);
4382 sval = cvtnum_full("step_size", optarg, 0, INT_MAX);
4391 ret = bdrv_parse_cache_mode(optarg, &flags, &writethrough);
4393 error_report("Invalid cache mode");
4399 flags |= BDRV_O_RDWR;
4405 case OPTION_PATTERN:
4409 if (qemu_strtoul(optarg, NULL, 0, &res) < 0 || res > 0xff) {
4410 error_report("Invalid pattern byte specified");
4416 case OPTION_FLUSH_INTERVAL:
4420 if (qemu_strtoul(optarg, NULL, 0, &res) < 0 || res > INT_MAX) {
4421 error_report("Invalid flush interval specified");
4424 flush_interval = res;
4427 case OPTION_NO_DRAIN:
4428 drain_on_flush = false;
4430 case OPTION_IMAGE_OPTS:
4436 if (optind != argc - 1) {
4437 error_exit("Expecting one image file name");
4439 filename = argv[argc - 1];
4441 if (!is_write && flush_interval) {
4442 error_report("--flush-interval is only available in write tests");
4446 if (flush_interval && flush_interval < depth) {
4447 error_report("Flush interval can't be smaller than depth");
4452 blk = img_open(image_opts, filename, fmt, flags, writethrough, quiet,
4459 image_size = blk_getlength(blk);
4460 if (image_size < 0) {
4465 data = (BenchData) {
4467 .image_size = image_size,
4469 .step = step ?: bufsize,
4474 .flush_interval = flush_interval,
4475 .drain_on_flush = drain_on_flush,
4477 printf("Sending %d %s requests, %d bytes each, %d in parallel "
4478 "(starting at offset %" PRId64 ", step size %d)\n",
4479 data.n, data.write ? "write" : "read", data.bufsize, data.nrreq,
4480 data.offset, data.step);
4481 if (flush_interval) {
4482 printf("Sending flush every %d requests\n", flush_interval);
4485 buf_size = data.nrreq * data.bufsize;
4486 data.buf = blk_blockalign(blk, buf_size);
4487 memset(data.buf, pattern, data.nrreq * data.bufsize);
4489 blk_register_buf(blk, data.buf, buf_size);
4491 data.qiov = g_new(QEMUIOVector, data.nrreq);
4492 for (i = 0; i < data.nrreq; i++) {
4493 qemu_iovec_init(&data.qiov[i], 1);
4494 qemu_iovec_add(&data.qiov[i],
4495 data.buf + i * data.bufsize, data.bufsize);
4498 gettimeofday(&t1, NULL);
4501 while (data.n > 0) {
4502 main_loop_wait(false);
4504 gettimeofday(&t2, NULL);
4506 printf("Run completed in %3.3f seconds.\n",
4507 (t2.tv_sec - t1.tv_sec)
4508 + ((double)(t2.tv_usec - t1.tv_usec) / 1000000));
4512 blk_unregister_buf(blk, data.buf);
4514 qemu_vfree(data.buf);
4531 typedef struct ImgBitmapAction {
4532 enum ImgBitmapAct act;
4533 const char *src; /* only used for merge */
4534 QSIMPLEQ_ENTRY(ImgBitmapAction) next;
4537 static int img_bitmap(int argc, char **argv)
4541 QemuOpts *opts = NULL;
4542 const char *fmt = NULL, *src_fmt = NULL, *src_filename = NULL;
4543 const char *filename, *bitmap;
4544 BlockBackend *blk = NULL, *src = NULL;
4545 BlockDriverState *bs = NULL, *src_bs = NULL;
4546 bool image_opts = false;
4547 int64_t granularity = 0;
4548 bool add = false, merge = false;
4549 QSIMPLEQ_HEAD(, ImgBitmapAction) actions;
4550 ImgBitmapAction *act, *act_next;
4553 QSIMPLEQ_INIT(&actions);
4556 static const struct option long_options[] = {
4557 {"help", no_argument, 0, 'h'},
4558 {"object", required_argument, 0, OPTION_OBJECT},
4559 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
4560 {"add", no_argument, 0, OPTION_ADD},
4561 {"remove", no_argument, 0, OPTION_REMOVE},
4562 {"clear", no_argument, 0, OPTION_CLEAR},
4563 {"enable", no_argument, 0, OPTION_ENABLE},
4564 {"disable", no_argument, 0, OPTION_DISABLE},
4565 {"merge", required_argument, 0, OPTION_MERGE},
4566 {"granularity", required_argument, 0, 'g'},
4567 {"source-file", required_argument, 0, 'b'},
4568 {"source-format", required_argument, 0, 'F'},
4571 c = getopt_long(argc, argv, ":b:f:F:g:h", long_options, NULL);
4578 missing_argument(argv[optind - 1]);
4581 unrecognized_option(argv[optind - 1]);
4587 src_filename = optarg;
4596 granularity = cvtnum("granularity", optarg);
4597 if (granularity < 0) {
4602 act = g_new0(ImgBitmapAction, 1);
4603 act->act = BITMAP_ADD;
4604 QSIMPLEQ_INSERT_TAIL(&actions, act, next);
4608 act = g_new0(ImgBitmapAction, 1);
4609 act->act = BITMAP_REMOVE;
4610 QSIMPLEQ_INSERT_TAIL(&actions, act, next);
4613 act = g_new0(ImgBitmapAction, 1);
4614 act->act = BITMAP_CLEAR;
4615 QSIMPLEQ_INSERT_TAIL(&actions, act, next);
4618 act = g_new0(ImgBitmapAction, 1);
4619 act->act = BITMAP_ENABLE;
4620 QSIMPLEQ_INSERT_TAIL(&actions, act, next);
4622 case OPTION_DISABLE:
4623 act = g_new0(ImgBitmapAction, 1);
4624 act->act = BITMAP_DISABLE;
4625 QSIMPLEQ_INSERT_TAIL(&actions, act, next);
4628 act = g_new0(ImgBitmapAction, 1);
4629 act->act = BITMAP_MERGE;
4631 QSIMPLEQ_INSERT_TAIL(&actions, act, next);
4635 user_creatable_process_cmdline(optarg);
4637 case OPTION_IMAGE_OPTS:
4643 if (QSIMPLEQ_EMPTY(&actions)) {
4644 error_report("Need at least one of --add, --remove, --clear, "
4645 "--enable, --disable, or --merge");
4649 if (granularity && !add) {
4650 error_report("granularity only supported with --add");
4653 if (src_fmt && !src_filename) {
4654 error_report("-F only supported with -b");
4657 if (src_filename && !merge) {
4658 error_report("Merge bitmap source file only supported with "
4663 if (optind != argc - 2) {
4664 error_report("Expecting filename and bitmap name");
4668 filename = argv[optind];
4669 bitmap = argv[optind + 1];
4672 * No need to open backing chains; we will be manipulating bitmaps
4673 * directly in this image without reference to image contents.
4675 blk = img_open(image_opts, filename, fmt, BDRV_O_RDWR | BDRV_O_NO_BACKING,
4676 false, false, false);
4682 src = img_open(false, src_filename, src_fmt, BDRV_O_NO_BACKING,
4683 false, false, false);
4687 src_bs = blk_bs(src);
4692 QSIMPLEQ_FOREACH_SAFE(act, &actions, next, act_next) {
4695 qmp_block_dirty_bitmap_add(bs->node_name, bitmap,
4696 !!granularity, granularity, true, true,
4697 false, false, &err);
4701 qmp_block_dirty_bitmap_remove(bs->node_name, bitmap, &err);
4705 qmp_block_dirty_bitmap_clear(bs->node_name, bitmap, &err);
4709 qmp_block_dirty_bitmap_enable(bs->node_name, bitmap, &err);
4712 case BITMAP_DISABLE:
4713 qmp_block_dirty_bitmap_disable(bs->node_name, bitmap, &err);
4717 do_dirty_bitmap_merge(bs->node_name, bitmap, src_bs->node_name,
4722 g_assert_not_reached();
4726 error_reportf_err(err, "Operation %s on bitmap %s failed: ",
4738 qemu_opts_del(opts);
4754 int bsz; /* Block size */
4762 int (*f)(const char *, struct DdIo *, struct DdIo *, struct DdInfo *);
4766 static int img_dd_bs(const char *arg,
4767 struct DdIo *in, struct DdIo *out,
4772 res = cvtnum_full("bs", arg, 1, INT_MAX);
4777 in->bsz = out->bsz = res;
4782 static int img_dd_count(const char *arg,
4783 struct DdIo *in, struct DdIo *out,
4786 dd->count = cvtnum("count", arg);
4788 if (dd->count < 0) {
4795 static int img_dd_if(const char *arg,
4796 struct DdIo *in, struct DdIo *out,
4799 in->filename = g_strdup(arg);
4804 static int img_dd_of(const char *arg,
4805 struct DdIo *in, struct DdIo *out,
4808 out->filename = g_strdup(arg);
4813 static int img_dd_skip(const char *arg,
4814 struct DdIo *in, struct DdIo *out,
4817 in->offset = cvtnum("skip", arg);
4819 if (in->offset < 0) {
4826 static int img_dd(int argc, char **argv)
4831 BlockDriver *drv = NULL, *proto_drv = NULL;
4832 BlockBackend *blk1 = NULL, *blk2 = NULL;
4833 QemuOpts *opts = NULL;
4834 QemuOptsList *create_opts = NULL;
4835 Error *local_err = NULL;
4836 bool image_opts = false;
4838 const char *out_fmt = "raw";
4839 const char *fmt = NULL;
4841 int64_t block_count = 0, out_pos, in_pos;
4842 bool force_share = false;
4843 struct DdInfo dd = {
4848 .bsz = 512, /* Block size is by default 512 bytes */
4860 const struct DdOpts options[] = {
4861 { "bs", img_dd_bs, C_BS },
4862 { "count", img_dd_count, C_COUNT },
4863 { "if", img_dd_if, C_IF },
4864 { "of", img_dd_of, C_OF },
4865 { "skip", img_dd_skip, C_SKIP },
4868 const struct option long_options[] = {
4869 { "help", no_argument, 0, 'h'},
4870 { "object", required_argument, 0, OPTION_OBJECT},
4871 { "image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
4872 { "force-share", no_argument, 0, 'U'},
4876 while ((c = getopt_long(argc, argv, ":hf:O:U", long_options, NULL))) {
4888 missing_argument(argv[optind - 1]);
4891 unrecognized_option(argv[optind - 1]);
4900 user_creatable_process_cmdline(optarg);
4902 case OPTION_IMAGE_OPTS:
4908 for (i = optind; i < argc; i++) {
4910 arg = g_strdup(argv[i]);
4912 tmp = strchr(arg, '=');
4914 error_report("unrecognized operand %s", arg);
4921 for (j = 0; options[j].name != NULL; j++) {
4922 if (!strcmp(arg, options[j].name)) {
4926 if (options[j].name == NULL) {
4927 error_report("unrecognized operand %s", arg);
4932 if (options[j].f(tmp, &in, &out, &dd) != 0) {
4936 dd.flags |= options[j].flag;
4941 if (!(dd.flags & C_IF && dd.flags & C_OF)) {
4942 error_report("Must specify both input and output files");
4947 blk1 = img_open(image_opts, in.filename, fmt, 0, false, false,
4955 drv = bdrv_find_format(out_fmt);
4957 error_report("Unknown file format");
4961 proto_drv = bdrv_find_protocol(out.filename, true, &local_err);
4964 error_report_err(local_err);
4968 if (!drv->create_opts) {
4969 error_report("Format driver '%s' does not support image creation",
4974 if (!proto_drv->create_opts) {
4975 error_report("Protocol driver '%s' does not support image creation",
4976 proto_drv->format_name);
4980 create_opts = qemu_opts_append(create_opts, drv->create_opts);
4981 create_opts = qemu_opts_append(create_opts, proto_drv->create_opts);
4983 opts = qemu_opts_create(create_opts, NULL, 0, &error_abort);
4985 size = blk_getlength(blk1);
4987 error_report("Failed to get size for '%s'", in.filename);
4992 if (dd.flags & C_COUNT && dd.count <= INT64_MAX / in.bsz &&
4993 dd.count * in.bsz < size) {
4994 size = dd.count * in.bsz;
4997 /* Overflow means the specified offset is beyond input image's size */
4998 if (dd.flags & C_SKIP && (in.offset > INT64_MAX / in.bsz ||
4999 size < in.bsz * in.offset)) {
5000 qemu_opt_set_number(opts, BLOCK_OPT_SIZE, 0, &error_abort);
5002 qemu_opt_set_number(opts, BLOCK_OPT_SIZE,
5003 size - in.bsz * in.offset, &error_abort);
5006 ret = bdrv_create(drv, out.filename, opts, &local_err);
5008 error_reportf_err(local_err,
5009 "%s: error while creating output image: ",
5015 /* TODO, we can't honour --image-opts for the target,
5016 * since it needs to be given in a format compatible
5017 * with the bdrv_create() call above which does not
5018 * support image-opts style.
5020 blk2 = img_open_file(out.filename, NULL, out_fmt, BDRV_O_RDWR,
5021 false, false, false);
5028 if (dd.flags & C_SKIP && (in.offset > INT64_MAX / in.bsz ||
5029 size < in.offset * in.bsz)) {
5030 /* We give a warning if the skip option is bigger than the input
5031 * size and create an empty output disk image (i.e. like dd(1)).
5033 error_report("%s: cannot skip to specified offset", in.filename);
5036 in_pos = in.offset * in.bsz;
5039 in.buf = g_new(uint8_t, in.bsz);
5041 for (out_pos = 0; in_pos < size; block_count++) {
5042 int in_ret, out_ret;
5044 if (in_pos + in.bsz > size) {
5045 in_ret = blk_pread(blk1, in_pos, in.buf, size - in_pos);
5047 in_ret = blk_pread(blk1, in_pos, in.buf, in.bsz);
5050 error_report("error while reading from input image file: %s",
5057 out_ret = blk_pwrite(blk2, out_pos, in.buf, in_ret, 0);
5060 error_report("error while writing to output image file: %s",
5061 strerror(-out_ret));
5070 qemu_opts_del(opts);
5071 qemu_opts_free(create_opts);
5074 g_free(in.filename);
5075 g_free(out.filename);
5085 static void dump_json_block_measure_info(BlockMeasureInfo *info)
5089 Visitor *v = qobject_output_visitor_new(&obj);
5091 visit_type_BlockMeasureInfo(v, NULL, &info, &error_abort);
5092 visit_complete(v, &obj);
5093 str = qobject_to_json_pretty(obj, true);
5094 assert(str != NULL);
5095 printf("%s\n", str->str);
5098 g_string_free(str, true);
5101 static int img_measure(int argc, char **argv)
5103 static const struct option long_options[] = {
5104 {"help", no_argument, 0, 'h'},
5105 {"image-opts", no_argument, 0, OPTION_IMAGE_OPTS},
5106 {"object", required_argument, 0, OPTION_OBJECT},
5107 {"output", required_argument, 0, OPTION_OUTPUT},
5108 {"size", required_argument, 0, OPTION_SIZE},
5109 {"force-share", no_argument, 0, 'U'},
5112 OutputFormat output_format = OFORMAT_HUMAN;
5113 BlockBackend *in_blk = NULL;
5115 const char *filename = NULL;
5116 const char *fmt = NULL;
5117 const char *out_fmt = "raw";
5118 char *options = NULL;
5119 char *snapshot_name = NULL;
5120 bool force_share = false;
5121 QemuOpts *opts = NULL;
5122 QemuOpts *object_opts = NULL;
5123 QemuOpts *sn_opts = NULL;
5124 QemuOptsList *create_opts = NULL;
5125 bool image_opts = false;
5126 uint64_t img_size = UINT64_MAX;
5127 BlockMeasureInfo *info = NULL;
5128 Error *local_err = NULL;
5132 while ((c = getopt_long(argc, argv, "hf:O:o:l:U",
5133 long_options, NULL)) != -1) {
5146 if (accumulate_options(&options, optarg) < 0) {
5151 if (strstart(optarg, SNAPSHOT_OPT_BASE, NULL)) {
5152 sn_opts = qemu_opts_parse_noisily(&internal_snapshot_opts,
5155 error_report("Failed in parsing snapshot param '%s'",
5160 snapshot_name = optarg;
5167 user_creatable_process_cmdline(optarg);
5169 case OPTION_IMAGE_OPTS:
5173 if (!strcmp(optarg, "json")) {
5174 output_format = OFORMAT_JSON;
5175 } else if (!strcmp(optarg, "human")) {
5176 output_format = OFORMAT_HUMAN;
5178 error_report("--output must be used with human or json "
5187 sval = cvtnum("image size", optarg);
5191 img_size = (uint64_t)sval;
5197 if (argc - optind > 1) {
5198 error_report("At most one filename argument is allowed.");
5200 } else if (argc - optind == 1) {
5201 filename = argv[optind];
5204 if (!filename && (image_opts || fmt || snapshot_name || sn_opts)) {
5205 error_report("--image-opts, -f, and -l require a filename argument.");
5208 if (filename && img_size != UINT64_MAX) {
5209 error_report("--size N cannot be used together with a filename.");
5212 if (!filename && img_size == UINT64_MAX) {
5213 error_report("Either --size N or one filename must be specified.");
5218 in_blk = img_open(image_opts, filename, fmt, 0,
5219 false, false, force_share);
5225 bdrv_snapshot_load_tmp(blk_bs(in_blk),
5226 qemu_opt_get(sn_opts, SNAPSHOT_OPT_ID),
5227 qemu_opt_get(sn_opts, SNAPSHOT_OPT_NAME),
5229 } else if (snapshot_name != NULL) {
5230 bdrv_snapshot_load_tmp_by_id_or_name(blk_bs(in_blk),
5231 snapshot_name, &local_err);
5234 error_reportf_err(local_err, "Failed to load snapshot: ");
5239 drv = bdrv_find_format(out_fmt);
5241 error_report("Unknown file format '%s'", out_fmt);
5244 if (!drv->create_opts) {
5245 error_report("Format driver '%s' does not support image creation",
5250 create_opts = qemu_opts_append(create_opts, drv->create_opts);
5251 create_opts = qemu_opts_append(create_opts, bdrv_file.create_opts);
5252 opts = qemu_opts_create(create_opts, NULL, 0, &error_abort);
5254 if (!qemu_opts_do_parse(opts, options, NULL, &local_err)) {
5255 error_report_err(local_err);
5256 error_report("Invalid options for file format '%s'", out_fmt);
5260 if (img_size != UINT64_MAX) {
5261 qemu_opt_set_number(opts, BLOCK_OPT_SIZE, img_size, &error_abort);
5264 info = bdrv_measure(drv, opts, in_blk ? blk_bs(in_blk) : NULL, &local_err);
5266 error_report_err(local_err);
5270 if (output_format == OFORMAT_HUMAN) {
5271 printf("required size: %" PRIu64 "\n", info->required);
5272 printf("fully allocated size: %" PRIu64 "\n", info->fully_allocated);
5273 if (info->has_bitmaps) {
5274 printf("bitmaps size: %" PRIu64 "\n", info->bitmaps);
5277 dump_json_block_measure_info(info);
5283 qapi_free_BlockMeasureInfo(info);
5284 qemu_opts_del(object_opts);
5285 qemu_opts_del(opts);
5286 qemu_opts_del(sn_opts);
5287 qemu_opts_free(create_opts);
5293 static const img_cmd_t img_cmds[] = {
5294 #define DEF(option, callback, arg_string) \
5295 { option, callback },
5296 #include "qemu-img-cmds.h"
5301 int main(int argc, char **argv)
5303 const img_cmd_t *cmd;
5304 const char *cmdname;
5305 Error *local_error = NULL;
5307 static const struct option long_options[] = {
5308 {"help", no_argument, 0, 'h'},
5309 {"version", no_argument, 0, 'V'},
5310 {"trace", required_argument, NULL, 'T'},
5315 signal(SIGPIPE, SIG_IGN);
5319 error_init(argv[0]);
5320 module_call_init(MODULE_INIT_TRACE);
5321 qemu_init_exec_dir(argv[0]);
5323 if (qemu_init_main_loop(&local_error)) {
5324 error_report_err(local_error);
5328 qcrypto_init(&error_fatal);
5330 module_call_init(MODULE_INIT_QOM);
5333 error_exit("Not enough arguments");
5336 qemu_add_opts(&qemu_source_opts);
5337 qemu_add_opts(&qemu_trace_opts);
5339 while ((c = getopt_long(argc, argv, "+:hVT:", long_options, NULL)) != -1) {
5342 missing_argument(argv[optind - 1]);
5345 unrecognized_option(argv[optind - 1]);
5351 printf(QEMU_IMG_VERSION);
5354 trace_opt_parse(optarg);
5359 cmdname = argv[optind];
5361 /* reset getopt_long scanning */
5367 qemu_reset_optind();
5369 if (!trace_init_backends()) {
5373 qemu_set_log(LOG_TRACE);
5375 /* find the command */
5376 for (cmd = img_cmds; cmd->name != NULL; cmd++) {
5377 if (!strcmp(cmdname, cmd->name)) {
5378 return cmd->handler(argc, argv);
5383 error_exit("Command not found: %s", cmdname);