2 * QEMU System Emulator block driver
4 * Copyright (c) 2003 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
24 #include "qemu/osdep.h"
26 #include "block/block_int.h"
27 #include "block/blockjob.h"
28 #include "block/nbd.h"
29 #include "qemu/error-report.h"
30 #include "module_block.h"
31 #include "qemu/module.h"
32 #include "qapi/qmp/qerror.h"
33 #include "qapi/qmp/qbool.h"
34 #include "qapi/qmp/qjson.h"
35 #include "sysemu/block-backend.h"
36 #include "sysemu/sysemu.h"
37 #include "qemu/notify.h"
38 #include "qemu/coroutine.h"
39 #include "block/qapi.h"
40 #include "qmp-commands.h"
41 #include "qemu/timer.h"
42 #include "qapi-event.h"
43 #include "qemu/cutils.h"
47 #include <sys/ioctl.h>
48 #include <sys/queue.h>
58 #define NOT_DONE 0x7fffffff /* used while emulated sync operation in progress */
60 static QTAILQ_HEAD(, BlockDriverState) graph_bdrv_states =
61 QTAILQ_HEAD_INITIALIZER(graph_bdrv_states);
63 static QTAILQ_HEAD(, BlockDriverState) all_bdrv_states =
64 QTAILQ_HEAD_INITIALIZER(all_bdrv_states);
66 static QLIST_HEAD(, BlockDriver) bdrv_drivers =
67 QLIST_HEAD_INITIALIZER(bdrv_drivers);
69 static BlockDriverState *bdrv_open_inherit(const char *filename,
70 const char *reference,
71 QDict *options, int flags,
72 BlockDriverState *parent,
73 const BdrvChildRole *child_role,
76 /* If non-zero, use only whitelisted block drivers */
77 static int use_bdrv_whitelist;
80 static int is_windows_drive_prefix(const char *filename)
82 return (((filename[0] >= 'a' && filename[0] <= 'z') ||
83 (filename[0] >= 'A' && filename[0] <= 'Z')) &&
87 int is_windows_drive(const char *filename)
89 if (is_windows_drive_prefix(filename) &&
92 if (strstart(filename, "\\\\.\\", NULL) ||
93 strstart(filename, "//./", NULL))
99 size_t bdrv_opt_mem_align(BlockDriverState *bs)
101 if (!bs || !bs->drv) {
102 /* page size or 4k (hdd sector size) should be on the safe side */
103 return MAX(4096, getpagesize());
106 return bs->bl.opt_mem_alignment;
109 size_t bdrv_min_mem_align(BlockDriverState *bs)
111 if (!bs || !bs->drv) {
112 /* page size or 4k (hdd sector size) should be on the safe side */
113 return MAX(4096, getpagesize());
116 return bs->bl.min_mem_alignment;
119 /* check if the path starts with "<protocol>:" */
120 int path_has_protocol(const char *path)
125 if (is_windows_drive(path) ||
126 is_windows_drive_prefix(path)) {
129 p = path + strcspn(path, ":/\\");
131 p = path + strcspn(path, ":/");
137 int path_is_absolute(const char *path)
140 /* specific case for names like: "\\.\d:" */
141 if (is_windows_drive(path) || is_windows_drive_prefix(path)) {
144 return (*path == '/' || *path == '\\');
146 return (*path == '/');
150 /* if filename is absolute, just copy it to dest. Otherwise, build a
151 path to it by considering it is relative to base_path. URL are
153 void path_combine(char *dest, int dest_size,
154 const char *base_path,
155 const char *filename)
162 if (path_is_absolute(filename)) {
163 pstrcpy(dest, dest_size, filename);
165 p = strchr(base_path, ':');
170 p1 = strrchr(base_path, '/');
174 p2 = strrchr(base_path, '\\');
186 if (len > dest_size - 1)
188 memcpy(dest, base_path, len);
190 pstrcat(dest, dest_size, filename);
194 void bdrv_get_full_backing_filename_from_filename(const char *backed,
196 char *dest, size_t sz,
199 if (backing[0] == '\0' || path_has_protocol(backing) ||
200 path_is_absolute(backing))
202 pstrcpy(dest, sz, backing);
203 } else if (backed[0] == '\0' || strstart(backed, "json:", NULL)) {
204 error_setg(errp, "Cannot use relative backing file names for '%s'",
207 path_combine(dest, sz, backed, backing);
211 void bdrv_get_full_backing_filename(BlockDriverState *bs, char *dest, size_t sz,
214 char *backed = bs->exact_filename[0] ? bs->exact_filename : bs->filename;
216 bdrv_get_full_backing_filename_from_filename(backed, bs->backing_file,
220 void bdrv_register(BlockDriver *bdrv)
222 QLIST_INSERT_HEAD(&bdrv_drivers, bdrv, list);
225 BlockDriverState *bdrv_new(void)
227 BlockDriverState *bs;
230 bs = g_new0(BlockDriverState, 1);
231 QLIST_INIT(&bs->dirty_bitmaps);
232 for (i = 0; i < BLOCK_OP_TYPE_MAX; i++) {
233 QLIST_INIT(&bs->op_blockers[i]);
235 notifier_with_return_list_init(&bs->before_write_notifiers);
237 bs->aio_context = qemu_get_aio_context();
239 qemu_co_queue_init(&bs->flush_queue);
241 QTAILQ_INSERT_TAIL(&all_bdrv_states, bs, bs_list);
246 static BlockDriver *bdrv_do_find_format(const char *format_name)
250 QLIST_FOREACH(drv1, &bdrv_drivers, list) {
251 if (!strcmp(drv1->format_name, format_name)) {
259 BlockDriver *bdrv_find_format(const char *format_name)
264 drv1 = bdrv_do_find_format(format_name);
269 /* The driver isn't registered, maybe we need to load a module */
270 for (i = 0; i < (int)ARRAY_SIZE(block_driver_modules); ++i) {
271 if (!strcmp(block_driver_modules[i].format_name, format_name)) {
272 block_module_load_one(block_driver_modules[i].library_name);
277 return bdrv_do_find_format(format_name);
280 static int bdrv_is_whitelisted(BlockDriver *drv, bool read_only)
282 static const char *whitelist_rw[] = {
283 CONFIG_BDRV_RW_WHITELIST
285 static const char *whitelist_ro[] = {
286 CONFIG_BDRV_RO_WHITELIST
290 if (!whitelist_rw[0] && !whitelist_ro[0]) {
291 return 1; /* no whitelist, anything goes */
294 for (p = whitelist_rw; *p; p++) {
295 if (!strcmp(drv->format_name, *p)) {
300 for (p = whitelist_ro; *p; p++) {
301 if (!strcmp(drv->format_name, *p)) {
309 bool bdrv_uses_whitelist(void)
311 return use_bdrv_whitelist;
314 typedef struct CreateCo {
322 static void coroutine_fn bdrv_create_co_entry(void *opaque)
324 Error *local_err = NULL;
327 CreateCo *cco = opaque;
330 ret = cco->drv->bdrv_create(cco->filename, cco->opts, &local_err);
331 error_propagate(&cco->err, local_err);
335 int bdrv_create(BlockDriver *drv, const char* filename,
336 QemuOpts *opts, Error **errp)
343 .filename = g_strdup(filename),
349 if (!drv->bdrv_create) {
350 error_setg(errp, "Driver '%s' does not support image creation", drv->format_name);
355 if (qemu_in_coroutine()) {
356 /* Fast-path if already in coroutine context */
357 bdrv_create_co_entry(&cco);
359 co = qemu_coroutine_create(bdrv_create_co_entry, &cco);
360 qemu_coroutine_enter(co);
361 while (cco.ret == NOT_DONE) {
362 aio_poll(qemu_get_aio_context(), true);
369 error_propagate(errp, cco.err);
371 error_setg_errno(errp, -ret, "Could not create image");
376 g_free(cco.filename);
380 int bdrv_create_file(const char *filename, QemuOpts *opts, Error **errp)
383 Error *local_err = NULL;
386 drv = bdrv_find_protocol(filename, true, errp);
391 ret = bdrv_create(drv, filename, opts, &local_err);
392 error_propagate(errp, local_err);
397 * Try to get @bs's logical and physical block size.
398 * On success, store them in @bsz struct and return 0.
399 * On failure return -errno.
400 * @bs must not be empty.
402 int bdrv_probe_blocksizes(BlockDriverState *bs, BlockSizes *bsz)
404 BlockDriver *drv = bs->drv;
406 if (drv && drv->bdrv_probe_blocksizes) {
407 return drv->bdrv_probe_blocksizes(bs, bsz);
414 * Try to get @bs's geometry (cyls, heads, sectors).
415 * On success, store them in @geo struct and return 0.
416 * On failure return -errno.
417 * @bs must not be empty.
419 int bdrv_probe_geometry(BlockDriverState *bs, HDGeometry *geo)
421 BlockDriver *drv = bs->drv;
423 if (drv && drv->bdrv_probe_geometry) {
424 return drv->bdrv_probe_geometry(bs, geo);
431 * Create a uniquely-named empty temporary file.
432 * Return 0 upon success, otherwise a negative errno value.
434 int get_tmp_filename(char *filename, int size)
437 char temp_dir[MAX_PATH];
438 /* GetTempFileName requires that its output buffer (4th param)
439 have length MAX_PATH or greater. */
440 assert(size >= MAX_PATH);
441 return (GetTempPath(MAX_PATH, temp_dir)
442 && GetTempFileName(temp_dir, "qem", 0, filename)
443 ? 0 : -GetLastError());
447 tmpdir = getenv("TMPDIR");
451 if (snprintf(filename, size, "%s/vl.XXXXXX", tmpdir) >= size) {
454 fd = mkstemp(filename);
458 if (close(fd) != 0) {
467 * Detect host devices. By convention, /dev/cdrom[N] is always
468 * recognized as a host CDROM.
470 static BlockDriver *find_hdev_driver(const char *filename)
472 int score_max = 0, score;
473 BlockDriver *drv = NULL, *d;
475 QLIST_FOREACH(d, &bdrv_drivers, list) {
476 if (d->bdrv_probe_device) {
477 score = d->bdrv_probe_device(filename);
478 if (score > score_max) {
488 static BlockDriver *bdrv_do_find_protocol(const char *protocol)
492 QLIST_FOREACH(drv1, &bdrv_drivers, list) {
493 if (drv1->protocol_name && !strcmp(drv1->protocol_name, protocol)) {
501 BlockDriver *bdrv_find_protocol(const char *filename,
502 bool allow_protocol_prefix,
511 /* TODO Drivers without bdrv_file_open must be specified explicitly */
514 * XXX(hch): we really should not let host device detection
515 * override an explicit protocol specification, but moving this
516 * later breaks access to device names with colons in them.
517 * Thanks to the brain-dead persistent naming schemes on udev-
518 * based Linux systems those actually are quite common.
520 drv1 = find_hdev_driver(filename);
525 if (!path_has_protocol(filename) || !allow_protocol_prefix) {
529 p = strchr(filename, ':');
532 if (len > sizeof(protocol) - 1)
533 len = sizeof(protocol) - 1;
534 memcpy(protocol, filename, len);
535 protocol[len] = '\0';
537 drv1 = bdrv_do_find_protocol(protocol);
542 for (i = 0; i < (int)ARRAY_SIZE(block_driver_modules); ++i) {
543 if (block_driver_modules[i].protocol_name &&
544 !strcmp(block_driver_modules[i].protocol_name, protocol)) {
545 block_module_load_one(block_driver_modules[i].library_name);
550 drv1 = bdrv_do_find_protocol(protocol);
552 error_setg(errp, "Unknown protocol '%s'", protocol);
558 * Guess image format by probing its contents.
559 * This is not a good idea when your image is raw (CVE-2008-2004), but
560 * we do it anyway for backward compatibility.
562 * @buf contains the image's first @buf_size bytes.
563 * @buf_size is the buffer size in bytes (generally BLOCK_PROBE_BUF_SIZE,
564 * but can be smaller if the image file is smaller)
565 * @filename is its filename.
567 * For all block drivers, call the bdrv_probe() method to get its
569 * Return the first block driver with the highest probing score.
571 BlockDriver *bdrv_probe_all(const uint8_t *buf, int buf_size,
572 const char *filename)
574 int score_max = 0, score;
575 BlockDriver *drv = NULL, *d;
577 QLIST_FOREACH(d, &bdrv_drivers, list) {
579 score = d->bdrv_probe(buf, buf_size, filename);
580 if (score > score_max) {
590 static int find_image_format(BdrvChild *file, const char *filename,
591 BlockDriver **pdrv, Error **errp)
593 BlockDriverState *bs = file->bs;
595 uint8_t buf[BLOCK_PROBE_BUF_SIZE];
598 /* Return the raw BlockDriver * to scsi-generic devices or empty drives */
599 if (bdrv_is_sg(bs) || !bdrv_is_inserted(bs) || bdrv_getlength(bs) == 0) {
604 ret = bdrv_pread(file, 0, buf, sizeof(buf));
606 error_setg_errno(errp, -ret, "Could not read image for determining its "
612 drv = bdrv_probe_all(buf, ret, filename);
614 error_setg(errp, "Could not determine image format: No compatible "
623 * Set the current 'total_sectors' value
624 * Return 0 on success, -errno on error.
626 static int refresh_total_sectors(BlockDriverState *bs, int64_t hint)
628 BlockDriver *drv = bs->drv;
630 /* Do not attempt drv->bdrv_getlength() on scsi-generic devices */
634 /* query actual device if possible, otherwise just trust the hint */
635 if (drv->bdrv_getlength) {
636 int64_t length = drv->bdrv_getlength(bs);
640 hint = DIV_ROUND_UP(length, BDRV_SECTOR_SIZE);
643 bs->total_sectors = hint;
648 * Combines a QDict of new block driver @options with any missing options taken
649 * from @old_options, so that leaving out an option defaults to its old value.
651 static void bdrv_join_options(BlockDriverState *bs, QDict *options,
654 if (bs->drv && bs->drv->bdrv_join_options) {
655 bs->drv->bdrv_join_options(options, old_options);
657 qdict_join(options, old_options, false);
662 * Set open flags for a given discard mode
664 * Return 0 on success, -1 if the discard mode was invalid.
666 int bdrv_parse_discard_flags(const char *mode, int *flags)
668 *flags &= ~BDRV_O_UNMAP;
670 if (!strcmp(mode, "off") || !strcmp(mode, "ignore")) {
672 } else if (!strcmp(mode, "on") || !strcmp(mode, "unmap")) {
673 *flags |= BDRV_O_UNMAP;
682 * Set open flags for a given cache mode
684 * Return 0 on success, -1 if the cache mode was invalid.
686 int bdrv_parse_cache_mode(const char *mode, int *flags, bool *writethrough)
688 *flags &= ~BDRV_O_CACHE_MASK;
690 if (!strcmp(mode, "off") || !strcmp(mode, "none")) {
691 *writethrough = false;
692 *flags |= BDRV_O_NOCACHE;
693 } else if (!strcmp(mode, "directsync")) {
694 *writethrough = true;
695 *flags |= BDRV_O_NOCACHE;
696 } else if (!strcmp(mode, "writeback")) {
697 *writethrough = false;
698 } else if (!strcmp(mode, "unsafe")) {
699 *writethrough = false;
700 *flags |= BDRV_O_NO_FLUSH;
701 } else if (!strcmp(mode, "writethrough")) {
702 *writethrough = true;
710 static void bdrv_child_cb_drained_begin(BdrvChild *child)
712 BlockDriverState *bs = child->opaque;
713 bdrv_drained_begin(bs);
716 static void bdrv_child_cb_drained_end(BdrvChild *child)
718 BlockDriverState *bs = child->opaque;
719 bdrv_drained_end(bs);
723 * Returns the options and flags that a temporary snapshot should get, based on
724 * the originally requested flags (the originally requested image will have
725 * flags like a backing file)
727 static void bdrv_temp_snapshot_options(int *child_flags, QDict *child_options,
728 int parent_flags, QDict *parent_options)
730 *child_flags = (parent_flags & ~BDRV_O_SNAPSHOT) | BDRV_O_TEMPORARY;
732 /* For temporary files, unconditional cache=unsafe is fine */
733 qdict_set_default_str(child_options, BDRV_OPT_CACHE_DIRECT, "off");
734 qdict_set_default_str(child_options, BDRV_OPT_CACHE_NO_FLUSH, "on");
736 /* Copy the read-only option from the parent */
737 qdict_copy_default(child_options, parent_options, BDRV_OPT_READ_ONLY);
739 /* aio=native doesn't work for cache.direct=off, so disable it for the
740 * temporary snapshot */
741 *child_flags &= ~BDRV_O_NATIVE_AIO;
745 * Returns the options and flags that bs->file should get if a protocol driver
746 * is expected, based on the given options and flags for the parent BDS
748 static void bdrv_inherited_options(int *child_flags, QDict *child_options,
749 int parent_flags, QDict *parent_options)
751 int flags = parent_flags;
753 /* Enable protocol handling, disable format probing for bs->file */
754 flags |= BDRV_O_PROTOCOL;
756 /* If the cache mode isn't explicitly set, inherit direct and no-flush from
758 qdict_copy_default(child_options, parent_options, BDRV_OPT_CACHE_DIRECT);
759 qdict_copy_default(child_options, parent_options, BDRV_OPT_CACHE_NO_FLUSH);
761 /* Inherit the read-only option from the parent if it's not set */
762 qdict_copy_default(child_options, parent_options, BDRV_OPT_READ_ONLY);
764 /* Our block drivers take care to send flushes and respect unmap policy,
765 * so we can default to enable both on lower layers regardless of the
766 * corresponding parent options. */
767 flags |= BDRV_O_UNMAP;
769 /* Clear flags that only apply to the top layer */
770 flags &= ~(BDRV_O_SNAPSHOT | BDRV_O_NO_BACKING | BDRV_O_COPY_ON_READ |
773 *child_flags = flags;
776 const BdrvChildRole child_file = {
777 .inherit_options = bdrv_inherited_options,
778 .drained_begin = bdrv_child_cb_drained_begin,
779 .drained_end = bdrv_child_cb_drained_end,
783 * Returns the options and flags that bs->file should get if the use of formats
784 * (and not only protocols) is permitted for it, based on the given options and
785 * flags for the parent BDS
787 static void bdrv_inherited_fmt_options(int *child_flags, QDict *child_options,
788 int parent_flags, QDict *parent_options)
790 child_file.inherit_options(child_flags, child_options,
791 parent_flags, parent_options);
793 *child_flags &= ~(BDRV_O_PROTOCOL | BDRV_O_NO_IO);
796 const BdrvChildRole child_format = {
797 .inherit_options = bdrv_inherited_fmt_options,
798 .drained_begin = bdrv_child_cb_drained_begin,
799 .drained_end = bdrv_child_cb_drained_end,
803 * Returns the options and flags that bs->backing should get, based on the
804 * given options and flags for the parent BDS
806 static void bdrv_backing_options(int *child_flags, QDict *child_options,
807 int parent_flags, QDict *parent_options)
809 int flags = parent_flags;
811 /* The cache mode is inherited unmodified for backing files; except WCE,
812 * which is only applied on the top level (BlockBackend) */
813 qdict_copy_default(child_options, parent_options, BDRV_OPT_CACHE_DIRECT);
814 qdict_copy_default(child_options, parent_options, BDRV_OPT_CACHE_NO_FLUSH);
816 /* backing files always opened read-only */
817 qdict_set_default_str(child_options, BDRV_OPT_READ_ONLY, "on");
818 flags &= ~BDRV_O_COPY_ON_READ;
820 /* snapshot=on is handled on the top layer */
821 flags &= ~(BDRV_O_SNAPSHOT | BDRV_O_TEMPORARY);
823 *child_flags = flags;
826 static const BdrvChildRole child_backing = {
827 .inherit_options = bdrv_backing_options,
828 .drained_begin = bdrv_child_cb_drained_begin,
829 .drained_end = bdrv_child_cb_drained_end,
832 static int bdrv_open_flags(BlockDriverState *bs, int flags)
834 int open_flags = flags;
837 * Clear flags that are internal to the block layer before opening the
840 open_flags &= ~(BDRV_O_SNAPSHOT | BDRV_O_NO_BACKING | BDRV_O_PROTOCOL);
843 * Snapshots should be writable.
845 if (flags & BDRV_O_TEMPORARY) {
846 open_flags |= BDRV_O_RDWR;
852 static void update_flags_from_options(int *flags, QemuOpts *opts)
854 *flags &= ~BDRV_O_CACHE_MASK;
856 assert(qemu_opt_find(opts, BDRV_OPT_CACHE_NO_FLUSH));
857 if (qemu_opt_get_bool(opts, BDRV_OPT_CACHE_NO_FLUSH, false)) {
858 *flags |= BDRV_O_NO_FLUSH;
861 assert(qemu_opt_find(opts, BDRV_OPT_CACHE_DIRECT));
862 if (qemu_opt_get_bool(opts, BDRV_OPT_CACHE_DIRECT, false)) {
863 *flags |= BDRV_O_NOCACHE;
866 *flags &= ~BDRV_O_RDWR;
868 assert(qemu_opt_find(opts, BDRV_OPT_READ_ONLY));
869 if (!qemu_opt_get_bool(opts, BDRV_OPT_READ_ONLY, false)) {
870 *flags |= BDRV_O_RDWR;
875 static void update_options_from_flags(QDict *options, int flags)
877 if (!qdict_haskey(options, BDRV_OPT_CACHE_DIRECT)) {
878 qdict_put(options, BDRV_OPT_CACHE_DIRECT,
879 qbool_from_bool(flags & BDRV_O_NOCACHE));
881 if (!qdict_haskey(options, BDRV_OPT_CACHE_NO_FLUSH)) {
882 qdict_put(options, BDRV_OPT_CACHE_NO_FLUSH,
883 qbool_from_bool(flags & BDRV_O_NO_FLUSH));
885 if (!qdict_haskey(options, BDRV_OPT_READ_ONLY)) {
886 qdict_put(options, BDRV_OPT_READ_ONLY,
887 qbool_from_bool(!(flags & BDRV_O_RDWR)));
891 static void bdrv_assign_node_name(BlockDriverState *bs,
892 const char *node_name,
895 char *gen_node_name = NULL;
898 node_name = gen_node_name = id_generate(ID_BLOCK);
899 } else if (!id_wellformed(node_name)) {
901 * Check for empty string or invalid characters, but not if it is
902 * generated (generated names use characters not available to the user)
904 error_setg(errp, "Invalid node name");
908 /* takes care of avoiding namespaces collisions */
909 if (blk_by_name(node_name)) {
910 error_setg(errp, "node-name=%s is conflicting with a device id",
915 /* takes care of avoiding duplicates node names */
916 if (bdrv_find_node(node_name)) {
917 error_setg(errp, "Duplicate node name");
921 /* copy node name into the bs and insert it into the graph list */
922 pstrcpy(bs->node_name, sizeof(bs->node_name), node_name);
923 QTAILQ_INSERT_TAIL(&graph_bdrv_states, bs, node_list);
925 g_free(gen_node_name);
928 static QemuOptsList bdrv_runtime_opts = {
929 .name = "bdrv_common",
930 .head = QTAILQ_HEAD_INITIALIZER(bdrv_runtime_opts.head),
934 .type = QEMU_OPT_STRING,
935 .help = "Node name of the block device node",
939 .type = QEMU_OPT_STRING,
940 .help = "Block driver to use for the node",
943 .name = BDRV_OPT_CACHE_DIRECT,
944 .type = QEMU_OPT_BOOL,
945 .help = "Bypass software writeback cache on the host",
948 .name = BDRV_OPT_CACHE_NO_FLUSH,
949 .type = QEMU_OPT_BOOL,
950 .help = "Ignore flush requests",
953 .name = BDRV_OPT_READ_ONLY,
954 .type = QEMU_OPT_BOOL,
955 .help = "Node is opened in read-only mode",
957 { /* end of list */ }
962 * Common part for opening disk images and files
964 * Removes all processed options from *options.
966 static int bdrv_open_common(BlockDriverState *bs, BdrvChild *file,
967 QDict *options, Error **errp)
970 const char *filename;
971 const char *driver_name = NULL;
972 const char *node_name = NULL;
975 Error *local_err = NULL;
977 assert(bs->file == NULL);
978 assert(options != NULL && bs->options != options);
980 opts = qemu_opts_create(&bdrv_runtime_opts, NULL, 0, &error_abort);
981 qemu_opts_absorb_qdict(opts, options, &local_err);
983 error_propagate(errp, local_err);
988 update_flags_from_options(&bs->open_flags, opts);
990 driver_name = qemu_opt_get(opts, "driver");
991 drv = bdrv_find_format(driver_name);
995 filename = file->bs->filename;
997 filename = qdict_get_try_str(options, "filename");
1000 if (drv->bdrv_needs_filename && !filename) {
1001 error_setg(errp, "The '%s' block driver requires a file name",
1007 trace_bdrv_open_common(bs, filename ?: "", bs->open_flags,
1010 node_name = qemu_opt_get(opts, "node-name");
1011 bdrv_assign_node_name(bs, node_name, &local_err);
1013 error_propagate(errp, local_err);
1018 bs->read_only = !(bs->open_flags & BDRV_O_RDWR);
1020 if (use_bdrv_whitelist && !bdrv_is_whitelisted(drv, bs->read_only)) {
1022 !bs->read_only && bdrv_is_whitelisted(drv, true)
1023 ? "Driver '%s' can only be used for read-only devices"
1024 : "Driver '%s' is not whitelisted",
1030 assert(bs->copy_on_read == 0); /* bdrv_new() and bdrv_close() make it so */
1031 if (bs->open_flags & BDRV_O_COPY_ON_READ) {
1032 if (!bs->read_only) {
1033 bdrv_enable_copy_on_read(bs);
1035 error_setg(errp, "Can't use copy-on-read on read-only device");
1041 if (filename != NULL) {
1042 pstrcpy(bs->filename, sizeof(bs->filename), filename);
1044 bs->filename[0] = '\0';
1046 pstrcpy(bs->exact_filename, sizeof(bs->exact_filename), bs->filename);
1049 bs->opaque = g_malloc0(drv->instance_size);
1051 /* Open the image, either directly or using a protocol */
1052 open_flags = bdrv_open_flags(bs, bs->open_flags);
1053 if (drv->bdrv_file_open) {
1054 assert(file == NULL);
1055 assert(!drv->bdrv_needs_filename || filename != NULL);
1056 ret = drv->bdrv_file_open(bs, options, open_flags, &local_err);
1059 error_setg(errp, "Can't use '%s' as a block driver for the "
1060 "protocol level", drv->format_name);
1065 ret = drv->bdrv_open(bs, options, open_flags, &local_err);
1070 error_propagate(errp, local_err);
1071 } else if (bs->filename[0]) {
1072 error_setg_errno(errp, -ret, "Could not open '%s'", bs->filename);
1074 error_setg_errno(errp, -ret, "Could not open image");
1079 ret = refresh_total_sectors(bs, bs->total_sectors);
1081 error_setg_errno(errp, -ret, "Could not refresh total sector count");
1085 bdrv_refresh_limits(bs, &local_err);
1087 error_propagate(errp, local_err);
1092 assert(bdrv_opt_mem_align(bs) != 0);
1093 assert(bdrv_min_mem_align(bs) != 0);
1094 assert(is_power_of_2(bs->bl.request_alignment));
1096 qemu_opts_del(opts);
1105 qemu_opts_del(opts);
1109 static QDict *parse_json_filename(const char *filename, Error **errp)
1111 QObject *options_obj;
1115 ret = strstart(filename, "json:", &filename);
1118 options_obj = qobject_from_json(filename);
1120 error_setg(errp, "Could not parse the JSON options");
1124 if (qobject_type(options_obj) != QTYPE_QDICT) {
1125 qobject_decref(options_obj);
1126 error_setg(errp, "Invalid JSON object given");
1130 options = qobject_to_qdict(options_obj);
1131 qdict_flatten(options);
1136 static void parse_json_protocol(QDict *options, const char **pfilename,
1139 QDict *json_options;
1140 Error *local_err = NULL;
1142 /* Parse json: pseudo-protocol */
1143 if (!*pfilename || !g_str_has_prefix(*pfilename, "json:")) {
1147 json_options = parse_json_filename(*pfilename, &local_err);
1149 error_propagate(errp, local_err);
1153 /* Options given in the filename have lower priority than options
1154 * specified directly */
1155 qdict_join(options, json_options, false);
1156 QDECREF(json_options);
1161 * Fills in default options for opening images and converts the legacy
1162 * filename/flags pair to option QDict entries.
1163 * The BDRV_O_PROTOCOL flag in *flags will be set or cleared accordingly if a
1164 * block driver has been specified explicitly.
1166 static int bdrv_fill_options(QDict **options, const char *filename,
1167 int *flags, Error **errp)
1169 const char *drvname;
1170 bool protocol = *flags & BDRV_O_PROTOCOL;
1171 bool parse_filename = false;
1172 BlockDriver *drv = NULL;
1173 Error *local_err = NULL;
1175 drvname = qdict_get_try_str(*options, "driver");
1177 drv = bdrv_find_format(drvname);
1179 error_setg(errp, "Unknown driver '%s'", drvname);
1182 /* If the user has explicitly specified the driver, this choice should
1183 * override the BDRV_O_PROTOCOL flag */
1184 protocol = drv->bdrv_file_open;
1188 *flags |= BDRV_O_PROTOCOL;
1190 *flags &= ~BDRV_O_PROTOCOL;
1193 /* Translate cache options from flags into options */
1194 update_options_from_flags(*options, *flags);
1196 /* Fetch the file name from the options QDict if necessary */
1197 if (protocol && filename) {
1198 if (!qdict_haskey(*options, "filename")) {
1199 qdict_put(*options, "filename", qstring_from_str(filename));
1200 parse_filename = true;
1202 error_setg(errp, "Can't specify 'file' and 'filename' options at "
1208 /* Find the right block driver */
1209 filename = qdict_get_try_str(*options, "filename");
1211 if (!drvname && protocol) {
1213 drv = bdrv_find_protocol(filename, parse_filename, errp);
1218 drvname = drv->format_name;
1219 qdict_put(*options, "driver", qstring_from_str(drvname));
1221 error_setg(errp, "Must specify either driver or file");
1226 assert(drv || !protocol);
1228 /* Driver-specific filename parsing */
1229 if (drv && drv->bdrv_parse_filename && parse_filename) {
1230 drv->bdrv_parse_filename(filename, *options, &local_err);
1232 error_propagate(errp, local_err);
1236 if (!drv->bdrv_needs_filename) {
1237 qdict_del(*options, "filename");
1244 static void bdrv_replace_child(BdrvChild *child, BlockDriverState *new_bs)
1246 BlockDriverState *old_bs = child->bs;
1249 if (old_bs->quiesce_counter && child->role->drained_end) {
1250 child->role->drained_end(child);
1252 QLIST_REMOVE(child, next_parent);
1258 QLIST_INSERT_HEAD(&new_bs->parents, child, next_parent);
1259 if (new_bs->quiesce_counter && child->role->drained_begin) {
1260 child->role->drained_begin(child);
1265 BdrvChild *bdrv_root_attach_child(BlockDriverState *child_bs,
1266 const char *child_name,
1267 const BdrvChildRole *child_role,
1270 BdrvChild *child = g_new(BdrvChild, 1);
1271 *child = (BdrvChild) {
1273 .name = g_strdup(child_name),
1278 bdrv_replace_child(child, child_bs);
1283 BdrvChild *bdrv_attach_child(BlockDriverState *parent_bs,
1284 BlockDriverState *child_bs,
1285 const char *child_name,
1286 const BdrvChildRole *child_role)
1288 BdrvChild *child = bdrv_root_attach_child(child_bs, child_name, child_role,
1290 QLIST_INSERT_HEAD(&parent_bs->children, child, next);
1294 static void bdrv_detach_child(BdrvChild *child)
1296 if (child->next.le_prev) {
1297 QLIST_REMOVE(child, next);
1298 child->next.le_prev = NULL;
1301 bdrv_replace_child(child, NULL);
1303 g_free(child->name);
1307 void bdrv_root_unref_child(BdrvChild *child)
1309 BlockDriverState *child_bs;
1311 child_bs = child->bs;
1312 bdrv_detach_child(child);
1313 bdrv_unref(child_bs);
1316 void bdrv_unref_child(BlockDriverState *parent, BdrvChild *child)
1318 if (child == NULL) {
1322 if (child->bs->inherits_from == parent) {
1323 child->bs->inherits_from = NULL;
1326 bdrv_root_unref_child(child);
1330 static void bdrv_parent_cb_change_media(BlockDriverState *bs, bool load)
1333 QLIST_FOREACH(c, &bs->parents, next_parent) {
1334 if (c->role->change_media) {
1335 c->role->change_media(c, load);
1340 static void bdrv_parent_cb_resize(BlockDriverState *bs)
1343 QLIST_FOREACH(c, &bs->parents, next_parent) {
1344 if (c->role->resize) {
1351 * Sets the backing file link of a BDS. A new reference is created; callers
1352 * which don't need their own reference any more must call bdrv_unref().
1354 void bdrv_set_backing_hd(BlockDriverState *bs, BlockDriverState *backing_hd)
1357 bdrv_ref(backing_hd);
1361 assert(bs->backing_blocker);
1362 bdrv_op_unblock_all(bs->backing->bs, bs->backing_blocker);
1363 bdrv_unref_child(bs, bs->backing);
1364 } else if (backing_hd) {
1365 error_setg(&bs->backing_blocker,
1366 "node is used as backing hd of '%s'",
1367 bdrv_get_device_or_node_name(bs));
1371 error_free(bs->backing_blocker);
1372 bs->backing_blocker = NULL;
1376 bs->backing = bdrv_attach_child(bs, backing_hd, "backing", &child_backing);
1377 bs->open_flags &= ~BDRV_O_NO_BACKING;
1378 pstrcpy(bs->backing_file, sizeof(bs->backing_file), backing_hd->filename);
1379 pstrcpy(bs->backing_format, sizeof(bs->backing_format),
1380 backing_hd->drv ? backing_hd->drv->format_name : "");
1382 bdrv_op_block_all(backing_hd, bs->backing_blocker);
1383 /* Otherwise we won't be able to commit due to check in bdrv_commit */
1384 bdrv_op_unblock(backing_hd, BLOCK_OP_TYPE_COMMIT_TARGET,
1385 bs->backing_blocker);
1387 * We do backup in 3 ways:
1389 * The target bs is new opened, and the source is top BDS
1390 * 2. blockdev backup
1391 * Both the source and the target are top BDSes.
1392 * 3. internal backup(used for block replication)
1393 * Both the source and the target are backing file
1395 * In case 1 and 2, neither the source nor the target is the backing file.
1396 * In case 3, we will block the top BDS, so there is only one block job
1397 * for the top BDS and its backing chain.
1399 bdrv_op_unblock(backing_hd, BLOCK_OP_TYPE_BACKUP_SOURCE,
1400 bs->backing_blocker);
1401 bdrv_op_unblock(backing_hd, BLOCK_OP_TYPE_BACKUP_TARGET,
1402 bs->backing_blocker);
1404 bdrv_refresh_limits(bs, NULL);
1408 * Opens the backing file for a BlockDriverState if not yet open
1410 * bdref_key specifies the key for the image's BlockdevRef in the options QDict.
1411 * That QDict has to be flattened; therefore, if the BlockdevRef is a QDict
1412 * itself, all options starting with "${bdref_key}." are considered part of the
1415 * TODO Can this be unified with bdrv_open_image()?
1417 int bdrv_open_backing_file(BlockDriverState *bs, QDict *parent_options,
1418 const char *bdref_key, Error **errp)
1420 char *backing_filename = g_malloc0(PATH_MAX);
1421 char *bdref_key_dot;
1422 const char *reference = NULL;
1424 BlockDriverState *backing_hd;
1426 QDict *tmp_parent_options = NULL;
1427 Error *local_err = NULL;
1429 if (bs->backing != NULL) {
1433 /* NULL means an empty set of options */
1434 if (parent_options == NULL) {
1435 tmp_parent_options = qdict_new();
1436 parent_options = tmp_parent_options;
1439 bs->open_flags &= ~BDRV_O_NO_BACKING;
1441 bdref_key_dot = g_strdup_printf("%s.", bdref_key);
1442 qdict_extract_subqdict(parent_options, &options, bdref_key_dot);
1443 g_free(bdref_key_dot);
1445 reference = qdict_get_try_str(parent_options, bdref_key);
1446 if (reference || qdict_haskey(options, "file.filename")) {
1447 backing_filename[0] = '\0';
1448 } else if (bs->backing_file[0] == '\0' && qdict_size(options) == 0) {
1452 bdrv_get_full_backing_filename(bs, backing_filename, PATH_MAX,
1456 error_propagate(errp, local_err);
1462 if (!bs->drv || !bs->drv->supports_backing) {
1464 error_setg(errp, "Driver doesn't support backing files");
1469 if (bs->backing_format[0] != '\0' && !qdict_haskey(options, "driver")) {
1470 qdict_put(options, "driver", qstring_from_str(bs->backing_format));
1473 backing_hd = bdrv_open_inherit(*backing_filename ? backing_filename : NULL,
1474 reference, options, 0, bs, &child_backing,
1477 bs->open_flags |= BDRV_O_NO_BACKING;
1478 error_prepend(errp, "Could not open backing file: ");
1483 /* Hook up the backing file link; drop our reference, bs owns the
1484 * backing_hd reference now */
1485 bdrv_set_backing_hd(bs, backing_hd);
1486 bdrv_unref(backing_hd);
1488 qdict_del(parent_options, bdref_key);
1491 g_free(backing_filename);
1492 QDECREF(tmp_parent_options);
1497 * Opens a disk image whose options are given as BlockdevRef in another block
1500 * If allow_none is true, no image will be opened if filename is false and no
1501 * BlockdevRef is given. NULL will be returned, but errp remains unset.
1503 * bdrev_key specifies the key for the image's BlockdevRef in the options QDict.
1504 * That QDict has to be flattened; therefore, if the BlockdevRef is a QDict
1505 * itself, all options starting with "${bdref_key}." are considered part of the
1508 * The BlockdevRef will be removed from the options QDict.
1510 BdrvChild *bdrv_open_child(const char *filename,
1511 QDict *options, const char *bdref_key,
1512 BlockDriverState* parent,
1513 const BdrvChildRole *child_role,
1514 bool allow_none, Error **errp)
1516 BdrvChild *c = NULL;
1517 BlockDriverState *bs;
1518 QDict *image_options;
1519 char *bdref_key_dot;
1520 const char *reference;
1522 assert(child_role != NULL);
1524 bdref_key_dot = g_strdup_printf("%s.", bdref_key);
1525 qdict_extract_subqdict(options, &image_options, bdref_key_dot);
1526 g_free(bdref_key_dot);
1528 reference = qdict_get_try_str(options, bdref_key);
1529 if (!filename && !reference && !qdict_size(image_options)) {
1531 error_setg(errp, "A block device must be specified for \"%s\"",
1534 QDECREF(image_options);
1538 bs = bdrv_open_inherit(filename, reference, image_options, 0,
1539 parent, child_role, errp);
1544 c = bdrv_attach_child(parent, bs, bdref_key, child_role);
1547 qdict_del(options, bdref_key);
1551 static BlockDriverState *bdrv_append_temp_snapshot(BlockDriverState *bs,
1553 QDict *snapshot_options,
1556 /* TODO: extra byte is a hack to ensure MAX_PATH space on Windows. */
1557 char *tmp_filename = g_malloc0(PATH_MAX + 1);
1559 QemuOpts *opts = NULL;
1560 BlockDriverState *bs_snapshot;
1563 /* if snapshot, we create a temporary backing file and open it
1564 instead of opening 'filename' directly */
1566 /* Get the required size from the image */
1567 total_size = bdrv_getlength(bs);
1568 if (total_size < 0) {
1569 error_setg_errno(errp, -total_size, "Could not get image size");
1573 /* Create the temporary image */
1574 ret = get_tmp_filename(tmp_filename, PATH_MAX + 1);
1576 error_setg_errno(errp, -ret, "Could not get temporary filename");
1580 opts = qemu_opts_create(bdrv_qcow2.create_opts, NULL, 0,
1582 qemu_opt_set_number(opts, BLOCK_OPT_SIZE, total_size, &error_abort);
1583 ret = bdrv_create(&bdrv_qcow2, tmp_filename, opts, errp);
1584 qemu_opts_del(opts);
1586 error_prepend(errp, "Could not create temporary overlay '%s': ",
1591 /* Prepare options QDict for the temporary file */
1592 qdict_put(snapshot_options, "file.driver",
1593 qstring_from_str("file"));
1594 qdict_put(snapshot_options, "file.filename",
1595 qstring_from_str(tmp_filename));
1596 qdict_put(snapshot_options, "driver",
1597 qstring_from_str("qcow2"));
1599 bs_snapshot = bdrv_open(NULL, NULL, snapshot_options, flags, errp);
1600 snapshot_options = NULL;
1606 /* bdrv_append() consumes a strong reference to bs_snapshot (i.e. it will
1607 * call bdrv_unref() on it), so in order to be able to return one, we have
1608 * to increase bs_snapshot's refcount here */
1609 bdrv_ref(bs_snapshot);
1610 bdrv_append(bs_snapshot, bs);
1612 g_free(tmp_filename);
1616 QDECREF(snapshot_options);
1617 g_free(tmp_filename);
1622 * Opens a disk image (raw, qcow2, vmdk, ...)
1624 * options is a QDict of options to pass to the block drivers, or NULL for an
1625 * empty set of options. The reference to the QDict belongs to the block layer
1626 * after the call (even on failure), so if the caller intends to reuse the
1627 * dictionary, it needs to use QINCREF() before calling bdrv_open.
1629 * If *pbs is NULL, a new BDS will be created with a pointer to it stored there.
1630 * If it is not NULL, the referenced BDS will be reused.
1632 * The reference parameter may be used to specify an existing block device which
1633 * should be opened. If specified, neither options nor a filename may be given,
1634 * nor can an existing BDS be reused (that is, *pbs has to be NULL).
1636 static BlockDriverState *bdrv_open_inherit(const char *filename,
1637 const char *reference,
1638 QDict *options, int flags,
1639 BlockDriverState *parent,
1640 const BdrvChildRole *child_role,
1644 BdrvChild *file = NULL;
1645 BlockDriverState *bs;
1646 BlockDriver *drv = NULL;
1647 const char *drvname;
1648 const char *backing;
1649 Error *local_err = NULL;
1650 QDict *snapshot_options = NULL;
1651 int snapshot_flags = 0;
1653 assert(!child_role || !flags);
1654 assert(!child_role == !parent);
1657 bool options_non_empty = options ? qdict_size(options) : false;
1660 if (filename || options_non_empty) {
1661 error_setg(errp, "Cannot reference an existing block device with "
1662 "additional options or a new filename");
1666 bs = bdrv_lookup_bs(reference, reference, errp);
1677 /* NULL means an empty set of options */
1678 if (options == NULL) {
1679 options = qdict_new();
1682 /* json: syntax counts as explicit options, as if in the QDict */
1683 parse_json_protocol(options, &filename, &local_err);
1688 bs->explicit_options = qdict_clone_shallow(options);
1691 bs->inherits_from = parent;
1692 child_role->inherit_options(&flags, options,
1693 parent->open_flags, parent->options);
1696 ret = bdrv_fill_options(&options, filename, &flags, &local_err);
1701 /* Set the BDRV_O_RDWR and BDRV_O_ALLOW_RDWR flags.
1702 * FIXME: we're parsing the QDict to avoid having to create a
1703 * QemuOpts just for this, but neither option is optimal. */
1704 if (g_strcmp0(qdict_get_try_str(options, BDRV_OPT_READ_ONLY), "on") &&
1705 !qdict_get_try_bool(options, BDRV_OPT_READ_ONLY, false)) {
1706 flags |= (BDRV_O_RDWR | BDRV_O_ALLOW_RDWR);
1708 flags &= ~BDRV_O_RDWR;
1711 if (flags & BDRV_O_SNAPSHOT) {
1712 snapshot_options = qdict_new();
1713 bdrv_temp_snapshot_options(&snapshot_flags, snapshot_options,
1715 /* Let bdrv_backing_options() override "read-only" */
1716 qdict_del(options, BDRV_OPT_READ_ONLY);
1717 bdrv_backing_options(&flags, options, flags, options);
1720 bs->open_flags = flags;
1721 bs->options = options;
1722 options = qdict_clone_shallow(options);
1724 /* Find the right image format driver */
1725 drvname = qdict_get_try_str(options, "driver");
1727 drv = bdrv_find_format(drvname);
1729 error_setg(errp, "Unknown driver: '%s'", drvname);
1734 assert(drvname || !(flags & BDRV_O_PROTOCOL));
1736 backing = qdict_get_try_str(options, "backing");
1737 if (backing && *backing == '\0') {
1738 flags |= BDRV_O_NO_BACKING;
1739 qdict_del(options, "backing");
1742 /* Open image file without format layer */
1743 if ((flags & BDRV_O_PROTOCOL) == 0) {
1744 file = bdrv_open_child(filename, options, "file", bs,
1745 &child_file, true, &local_err);
1751 /* Image format probing */
1754 ret = find_image_format(file, filename, &drv, &local_err);
1759 * This option update would logically belong in bdrv_fill_options(),
1760 * but we first need to open bs->file for the probing to work, while
1761 * opening bs->file already requires the (mostly) final set of options
1762 * so that cache mode etc. can be inherited.
1764 * Adding the driver later is somewhat ugly, but it's not an option
1765 * that would ever be inherited, so it's correct. We just need to make
1766 * sure to update both bs->options (which has the full effective
1767 * options for bs) and options (which has file.* already removed).
1769 qdict_put(bs->options, "driver", qstring_from_str(drv->format_name));
1770 qdict_put(options, "driver", qstring_from_str(drv->format_name));
1772 error_setg(errp, "Must specify either driver or file");
1776 /* BDRV_O_PROTOCOL must be set iff a protocol BDS is about to be created */
1777 assert(!!(flags & BDRV_O_PROTOCOL) == !!drv->bdrv_file_open);
1778 /* file must be NULL if a protocol BDS is about to be created
1779 * (the inverse results in an error message from bdrv_open_common()) */
1780 assert(!(flags & BDRV_O_PROTOCOL) || !file);
1782 /* Open the image */
1783 ret = bdrv_open_common(bs, file, options, &local_err);
1788 if (file && (bs->file != file)) {
1789 bdrv_unref_child(bs, file);
1793 /* If there is a backing file, use it */
1794 if ((flags & BDRV_O_NO_BACKING) == 0) {
1795 ret = bdrv_open_backing_file(bs, options, "backing", &local_err);
1797 goto close_and_fail;
1801 bdrv_refresh_filename(bs);
1803 /* Check if any unknown options were used */
1804 if (options && (qdict_size(options) != 0)) {
1805 const QDictEntry *entry = qdict_first(options);
1806 if (flags & BDRV_O_PROTOCOL) {
1807 error_setg(errp, "Block protocol '%s' doesn't support the option "
1808 "'%s'", drv->format_name, entry->key);
1811 "Block format '%s' does not support the option '%s'",
1812 drv->format_name, entry->key);
1815 goto close_and_fail;
1818 if (!bdrv_key_required(bs)) {
1819 bdrv_parent_cb_change_media(bs, true);
1820 } else if (!runstate_check(RUN_STATE_PRELAUNCH)
1821 && !runstate_check(RUN_STATE_INMIGRATE)
1822 && !runstate_check(RUN_STATE_PAUSED)) { /* HACK */
1824 "Guest must be stopped for opening of encrypted image");
1825 goto close_and_fail;
1830 /* For snapshot=on, create a temporary qcow2 overlay. bs points to the
1831 * temporary snapshot afterwards. */
1832 if (snapshot_flags) {
1833 BlockDriverState *snapshot_bs;
1834 snapshot_bs = bdrv_append_temp_snapshot(bs, snapshot_flags,
1835 snapshot_options, &local_err);
1836 snapshot_options = NULL;
1838 goto close_and_fail;
1840 /* We are not going to return bs but the overlay on top of it
1841 * (snapshot_bs); thus, we have to drop the strong reference to bs
1842 * (which we obtained by calling bdrv_new()). bs will not be deleted,
1843 * though, because the overlay still has a reference to it. */
1852 bdrv_unref_child(bs, file);
1854 QDECREF(snapshot_options);
1855 QDECREF(bs->explicit_options);
1856 QDECREF(bs->options);
1860 error_propagate(errp, local_err);
1865 QDECREF(snapshot_options);
1867 error_propagate(errp, local_err);
1871 BlockDriverState *bdrv_open(const char *filename, const char *reference,
1872 QDict *options, int flags, Error **errp)
1874 return bdrv_open_inherit(filename, reference, options, flags, NULL,
1878 typedef struct BlockReopenQueueEntry {
1880 BDRVReopenState state;
1881 QSIMPLEQ_ENTRY(BlockReopenQueueEntry) entry;
1882 } BlockReopenQueueEntry;
1885 * Adds a BlockDriverState to a simple queue for an atomic, transactional
1886 * reopen of multiple devices.
1888 * bs_queue can either be an existing BlockReopenQueue that has had QSIMPLE_INIT
1889 * already performed, or alternatively may be NULL a new BlockReopenQueue will
1890 * be created and initialized. This newly created BlockReopenQueue should be
1891 * passed back in for subsequent calls that are intended to be of the same
1894 * bs is the BlockDriverState to add to the reopen queue.
1896 * options contains the changed options for the associated bs
1897 * (the BlockReopenQueue takes ownership)
1899 * flags contains the open flags for the associated bs
1901 * returns a pointer to bs_queue, which is either the newly allocated
1902 * bs_queue, or the existing bs_queue being used.
1905 static BlockReopenQueue *bdrv_reopen_queue_child(BlockReopenQueue *bs_queue,
1906 BlockDriverState *bs,
1909 const BdrvChildRole *role,
1910 QDict *parent_options,
1915 BlockReopenQueueEntry *bs_entry;
1917 QDict *old_options, *explicit_options;
1919 if (bs_queue == NULL) {
1920 bs_queue = g_new0(BlockReopenQueue, 1);
1921 QSIMPLEQ_INIT(bs_queue);
1925 options = qdict_new();
1929 * Precedence of options:
1930 * 1. Explicitly passed in options (highest)
1931 * 2. Set in flags (only for top level)
1932 * 3. Retained from explicitly set options of bs
1933 * 4. Inherited from parent node
1934 * 5. Retained from effective options of bs
1937 if (!parent_options) {
1939 * Any setting represented by flags is always updated. If the
1940 * corresponding QDict option is set, it takes precedence. Otherwise
1941 * the flag is translated into a QDict option. The old setting of bs is
1944 update_options_from_flags(options, flags);
1947 /* Old explicitly set values (don't overwrite by inherited value) */
1948 old_options = qdict_clone_shallow(bs->explicit_options);
1949 bdrv_join_options(bs, options, old_options);
1950 QDECREF(old_options);
1952 explicit_options = qdict_clone_shallow(options);
1954 /* Inherit from parent node */
1955 if (parent_options) {
1957 role->inherit_options(&flags, options, parent_flags, parent_options);
1960 /* Old values are used for options that aren't set yet */
1961 old_options = qdict_clone_shallow(bs->options);
1962 bdrv_join_options(bs, options, old_options);
1963 QDECREF(old_options);
1965 /* bdrv_open() masks this flag out */
1966 flags &= ~BDRV_O_PROTOCOL;
1968 QLIST_FOREACH(child, &bs->children, next) {
1969 QDict *new_child_options;
1970 char *child_key_dot;
1972 /* reopen can only change the options of block devices that were
1973 * implicitly created and inherited options. For other (referenced)
1974 * block devices, a syntax like "backing.foo" results in an error. */
1975 if (child->bs->inherits_from != bs) {
1979 child_key_dot = g_strdup_printf("%s.", child->name);
1980 qdict_extract_subqdict(options, &new_child_options, child_key_dot);
1981 g_free(child_key_dot);
1983 bdrv_reopen_queue_child(bs_queue, child->bs, new_child_options, 0,
1984 child->role, options, flags);
1987 bs_entry = g_new0(BlockReopenQueueEntry, 1);
1988 QSIMPLEQ_INSERT_TAIL(bs_queue, bs_entry, entry);
1990 bs_entry->state.bs = bs;
1991 bs_entry->state.options = options;
1992 bs_entry->state.explicit_options = explicit_options;
1993 bs_entry->state.flags = flags;
1998 BlockReopenQueue *bdrv_reopen_queue(BlockReopenQueue *bs_queue,
1999 BlockDriverState *bs,
2000 QDict *options, int flags)
2002 return bdrv_reopen_queue_child(bs_queue, bs, options, flags,
2007 * Reopen multiple BlockDriverStates atomically & transactionally.
2009 * The queue passed in (bs_queue) must have been built up previous
2010 * via bdrv_reopen_queue().
2012 * Reopens all BDS specified in the queue, with the appropriate
2013 * flags. All devices are prepared for reopen, and failure of any
2014 * device will cause all device changes to be abandonded, and intermediate
2017 * If all devices prepare successfully, then the changes are committed
2021 int bdrv_reopen_multiple(BlockReopenQueue *bs_queue, Error **errp)
2024 BlockReopenQueueEntry *bs_entry, *next;
2025 Error *local_err = NULL;
2027 assert(bs_queue != NULL);
2031 QSIMPLEQ_FOREACH(bs_entry, bs_queue, entry) {
2032 if (bdrv_reopen_prepare(&bs_entry->state, bs_queue, &local_err)) {
2033 error_propagate(errp, local_err);
2036 bs_entry->prepared = true;
2039 /* If we reach this point, we have success and just need to apply the
2042 QSIMPLEQ_FOREACH(bs_entry, bs_queue, entry) {
2043 bdrv_reopen_commit(&bs_entry->state);
2049 QSIMPLEQ_FOREACH_SAFE(bs_entry, bs_queue, entry, next) {
2050 if (ret && bs_entry->prepared) {
2051 bdrv_reopen_abort(&bs_entry->state);
2053 QDECREF(bs_entry->state.explicit_options);
2055 QDECREF(bs_entry->state.options);
2063 /* Reopen a single BlockDriverState with the specified flags. */
2064 int bdrv_reopen(BlockDriverState *bs, int bdrv_flags, Error **errp)
2067 Error *local_err = NULL;
2068 BlockReopenQueue *queue = bdrv_reopen_queue(NULL, bs, NULL, bdrv_flags);
2070 ret = bdrv_reopen_multiple(queue, &local_err);
2071 if (local_err != NULL) {
2072 error_propagate(errp, local_err);
2079 * Prepares a BlockDriverState for reopen. All changes are staged in the
2080 * 'opaque' field of the BDRVReopenState, which is used and allocated by
2081 * the block driver layer .bdrv_reopen_prepare()
2083 * bs is the BlockDriverState to reopen
2084 * flags are the new open flags
2085 * queue is the reopen queue
2087 * Returns 0 on success, non-zero on error. On error errp will be set
2090 * On failure, bdrv_reopen_abort() will be called to clean up any data.
2091 * It is the responsibility of the caller to then call the abort() or
2092 * commit() for any other BDS that have been left in a prepare() state
2095 int bdrv_reopen_prepare(BDRVReopenState *reopen_state, BlockReopenQueue *queue,
2099 Error *local_err = NULL;
2104 assert(reopen_state != NULL);
2105 assert(reopen_state->bs->drv != NULL);
2106 drv = reopen_state->bs->drv;
2108 /* Process generic block layer options */
2109 opts = qemu_opts_create(&bdrv_runtime_opts, NULL, 0, &error_abort);
2110 qemu_opts_absorb_qdict(opts, reopen_state->options, &local_err);
2112 error_propagate(errp, local_err);
2117 update_flags_from_options(&reopen_state->flags, opts);
2119 /* node-name and driver must be unchanged. Put them back into the QDict, so
2120 * that they are checked at the end of this function. */
2121 value = qemu_opt_get(opts, "node-name");
2123 qdict_put(reopen_state->options, "node-name", qstring_from_str(value));
2126 value = qemu_opt_get(opts, "driver");
2128 qdict_put(reopen_state->options, "driver", qstring_from_str(value));
2131 /* if we are to stay read-only, do not allow permission change
2133 if (!(reopen_state->bs->open_flags & BDRV_O_ALLOW_RDWR) &&
2134 reopen_state->flags & BDRV_O_RDWR) {
2135 error_setg(errp, "Node '%s' is read only",
2136 bdrv_get_device_or_node_name(reopen_state->bs));
2141 ret = bdrv_flush(reopen_state->bs);
2143 error_setg_errno(errp, -ret, "Error flushing drive");
2147 if (drv->bdrv_reopen_prepare) {
2148 ret = drv->bdrv_reopen_prepare(reopen_state, queue, &local_err);
2150 if (local_err != NULL) {
2151 error_propagate(errp, local_err);
2153 error_setg(errp, "failed while preparing to reopen image '%s'",
2154 reopen_state->bs->filename);
2159 /* It is currently mandatory to have a bdrv_reopen_prepare()
2160 * handler for each supported drv. */
2161 error_setg(errp, "Block format '%s' used by node '%s' "
2162 "does not support reopening files", drv->format_name,
2163 bdrv_get_device_or_node_name(reopen_state->bs));
2168 /* Options that are not handled are only okay if they are unchanged
2169 * compared to the old state. It is expected that some options are only
2170 * used for the initial open, but not reopen (e.g. filename) */
2171 if (qdict_size(reopen_state->options)) {
2172 const QDictEntry *entry = qdict_first(reopen_state->options);
2175 QString *new_obj = qobject_to_qstring(entry->value);
2176 const char *new = qstring_get_str(new_obj);
2177 const char *old = qdict_get_try_str(reopen_state->bs->options,
2180 if (!old || strcmp(new, old)) {
2181 error_setg(errp, "Cannot change the option '%s'", entry->key);
2185 } while ((entry = qdict_next(reopen_state->options, entry)));
2191 qemu_opts_del(opts);
2196 * Takes the staged changes for the reopen from bdrv_reopen_prepare(), and
2197 * makes them final by swapping the staging BlockDriverState contents into
2198 * the active BlockDriverState contents.
2200 void bdrv_reopen_commit(BDRVReopenState *reopen_state)
2204 assert(reopen_state != NULL);
2205 drv = reopen_state->bs->drv;
2206 assert(drv != NULL);
2208 /* If there are any driver level actions to take */
2209 if (drv->bdrv_reopen_commit) {
2210 drv->bdrv_reopen_commit(reopen_state);
2213 /* set BDS specific flags now */
2214 QDECREF(reopen_state->bs->explicit_options);
2216 reopen_state->bs->explicit_options = reopen_state->explicit_options;
2217 reopen_state->bs->open_flags = reopen_state->flags;
2218 reopen_state->bs->read_only = !(reopen_state->flags & BDRV_O_RDWR);
2220 bdrv_refresh_limits(reopen_state->bs, NULL);
2224 * Abort the reopen, and delete and free the staged changes in
2227 void bdrv_reopen_abort(BDRVReopenState *reopen_state)
2231 assert(reopen_state != NULL);
2232 drv = reopen_state->bs->drv;
2233 assert(drv != NULL);
2235 if (drv->bdrv_reopen_abort) {
2236 drv->bdrv_reopen_abort(reopen_state);
2239 QDECREF(reopen_state->explicit_options);
2243 static void bdrv_close(BlockDriverState *bs)
2245 BdrvAioNotifier *ban, *ban_next;
2248 assert(!bs->refcnt);
2250 bdrv_drained_begin(bs); /* complete I/O */
2252 bdrv_drain(bs); /* in case flush left pending I/O */
2254 bdrv_release_named_dirty_bitmaps(bs);
2255 assert(QLIST_EMPTY(&bs->dirty_bitmaps));
2258 BdrvChild *child, *next;
2260 bs->drv->bdrv_close(bs);
2263 bdrv_set_backing_hd(bs, NULL);
2265 if (bs->file != NULL) {
2266 bdrv_unref_child(bs, bs->file);
2270 QLIST_FOREACH_SAFE(child, &bs->children, next, next) {
2271 /* TODO Remove bdrv_unref() from drivers' close function and use
2272 * bdrv_unref_child() here */
2273 if (child->bs->inherits_from == bs) {
2274 child->bs->inherits_from = NULL;
2276 bdrv_detach_child(child);
2281 bs->copy_on_read = 0;
2282 bs->backing_file[0] = '\0';
2283 bs->backing_format[0] = '\0';
2284 bs->total_sectors = 0;
2285 bs->encrypted = false;
2286 bs->valid_key = false;
2288 QDECREF(bs->options);
2289 QDECREF(bs->explicit_options);
2291 QDECREF(bs->full_open_options);
2292 bs->full_open_options = NULL;
2295 QLIST_FOREACH_SAFE(ban, &bs->aio_notifiers, list, ban_next) {
2298 QLIST_INIT(&bs->aio_notifiers);
2299 bdrv_drained_end(bs);
2302 void bdrv_close_all(void)
2304 block_job_cancel_sync_all();
2305 nbd_export_close_all();
2307 /* Drop references from requests still in flight, such as canceled block
2308 * jobs whose AIO context has not been polled yet */
2311 blk_remove_all_bs();
2312 blockdev_close_all_bdrv_states();
2314 assert(QTAILQ_EMPTY(&all_bdrv_states));
2317 static void change_parent_backing_link(BlockDriverState *from,
2318 BlockDriverState *to)
2320 BdrvChild *c, *next, *to_c;
2322 QLIST_FOREACH_SAFE(c, &from->parents, next_parent, next) {
2323 if (c->role == &child_backing) {
2324 /* @from is generally not allowed to be a backing file, except for
2325 * when @to is the overlay. In that case, @from may not be replaced
2326 * by @to as @to's backing node. */
2327 QLIST_FOREACH(to_c, &to->children, next) {
2337 assert(c->role != &child_backing);
2339 bdrv_replace_child(c, to);
2345 * Add new bs contents at the top of an image chain while the chain is
2346 * live, while keeping required fields on the top layer.
2348 * This will modify the BlockDriverState fields, and swap contents
2349 * between bs_new and bs_top. Both bs_new and bs_top are modified.
2351 * bs_new must not be attached to a BlockBackend.
2353 * This function does not create any image files.
2355 * bdrv_append() takes ownership of a bs_new reference and unrefs it because
2356 * that's what the callers commonly need. bs_new will be referenced by the old
2357 * parents of bs_top after bdrv_append() returns. If the caller needs to keep a
2358 * reference of its own, it must call bdrv_ref().
2360 void bdrv_append(BlockDriverState *bs_new, BlockDriverState *bs_top)
2362 assert(!bdrv_requests_pending(bs_top));
2363 assert(!bdrv_requests_pending(bs_new));
2367 change_parent_backing_link(bs_top, bs_new);
2368 bdrv_set_backing_hd(bs_new, bs_top);
2371 /* bs_new is now referenced by its new parents, we don't need the
2372 * additional reference any more. */
2376 void bdrv_replace_in_backing_chain(BlockDriverState *old, BlockDriverState *new)
2378 assert(!bdrv_requests_pending(old));
2379 assert(!bdrv_requests_pending(new));
2383 change_parent_backing_link(old, new);
2388 static void bdrv_delete(BlockDriverState *bs)
2391 assert(bdrv_op_blocker_is_empty(bs));
2392 assert(!bs->refcnt);
2396 /* remove from list, if necessary */
2397 if (bs->node_name[0] != '\0') {
2398 QTAILQ_REMOVE(&graph_bdrv_states, bs, node_list);
2400 QTAILQ_REMOVE(&all_bdrv_states, bs, bs_list);
2406 * Run consistency checks on an image
2408 * Returns 0 if the check could be completed (it doesn't mean that the image is
2409 * free of errors) or -errno when an internal error occurred. The results of the
2410 * check are stored in res.
2412 int bdrv_check(BlockDriverState *bs, BdrvCheckResult *res, BdrvCheckMode fix)
2414 if (bs->drv == NULL) {
2417 if (bs->drv->bdrv_check == NULL) {
2421 memset(res, 0, sizeof(*res));
2422 return bs->drv->bdrv_check(bs, res, fix);
2428 * -EINVAL - backing format specified, but no file
2429 * -ENOSPC - can't update the backing file because no space is left in the
2431 * -ENOTSUP - format driver doesn't support changing the backing file
2433 int bdrv_change_backing_file(BlockDriverState *bs,
2434 const char *backing_file, const char *backing_fmt)
2436 BlockDriver *drv = bs->drv;
2439 /* Backing file format doesn't make sense without a backing file */
2440 if (backing_fmt && !backing_file) {
2444 if (drv->bdrv_change_backing_file != NULL) {
2445 ret = drv->bdrv_change_backing_file(bs, backing_file, backing_fmt);
2451 pstrcpy(bs->backing_file, sizeof(bs->backing_file), backing_file ?: "");
2452 pstrcpy(bs->backing_format, sizeof(bs->backing_format), backing_fmt ?: "");
2458 * Finds the image layer in the chain that has 'bs' as its backing file.
2460 * active is the current topmost image.
2462 * Returns NULL if bs is not found in active's image chain,
2463 * or if active == bs.
2465 * Returns the bottommost base image if bs == NULL.
2467 BlockDriverState *bdrv_find_overlay(BlockDriverState *active,
2468 BlockDriverState *bs)
2470 while (active && bs != backing_bs(active)) {
2471 active = backing_bs(active);
2477 /* Given a BDS, searches for the base layer. */
2478 BlockDriverState *bdrv_find_base(BlockDriverState *bs)
2480 return bdrv_find_overlay(bs, NULL);
2484 * Drops images above 'base' up to and including 'top', and sets the image
2485 * above 'top' to have base as its backing file.
2487 * Requires that the overlay to 'top' is opened r/w, so that the backing file
2488 * information in 'bs' can be properly updated.
2490 * E.g., this will convert the following chain:
2491 * bottom <- base <- intermediate <- top <- active
2495 * bottom <- base <- active
2497 * It is allowed for bottom==base, in which case it converts:
2499 * base <- intermediate <- top <- active
2505 * If backing_file_str is non-NULL, it will be used when modifying top's
2506 * overlay image metadata.
2509 * if active == top, that is considered an error
2512 int bdrv_drop_intermediate(BlockDriverState *active, BlockDriverState *top,
2513 BlockDriverState *base, const char *backing_file_str)
2515 BlockDriverState *new_top_bs = NULL;
2518 if (!top->drv || !base->drv) {
2522 new_top_bs = bdrv_find_overlay(active, top);
2524 if (new_top_bs == NULL) {
2525 /* we could not find the image above 'top', this is an error */
2529 /* special case of new_top_bs->backing->bs already pointing to base - nothing
2530 * to do, no intermediate images */
2531 if (backing_bs(new_top_bs) == base) {
2536 /* Make sure that base is in the backing chain of top */
2537 if (!bdrv_chain_contains(top, base)) {
2541 /* success - we can delete the intermediate states, and link top->base */
2542 backing_file_str = backing_file_str ? backing_file_str : base->filename;
2543 ret = bdrv_change_backing_file(new_top_bs, backing_file_str,
2544 base->drv ? base->drv->format_name : "");
2548 bdrv_set_backing_hd(new_top_bs, base);
2556 * Truncate file to 'offset' bytes (needed only for file protocols)
2558 int bdrv_truncate(BlockDriverState *bs, int64_t offset)
2560 BlockDriver *drv = bs->drv;
2564 if (!drv->bdrv_truncate)
2569 ret = drv->bdrv_truncate(bs, offset);
2571 ret = refresh_total_sectors(bs, offset >> BDRV_SECTOR_BITS);
2572 bdrv_dirty_bitmap_truncate(bs);
2573 bdrv_parent_cb_resize(bs);
2580 * Length of a allocated file in bytes. Sparse files are counted by actual
2581 * allocated space. Return < 0 if error or unknown.
2583 int64_t bdrv_get_allocated_file_size(BlockDriverState *bs)
2585 BlockDriver *drv = bs->drv;
2589 if (drv->bdrv_get_allocated_file_size) {
2590 return drv->bdrv_get_allocated_file_size(bs);
2593 return bdrv_get_allocated_file_size(bs->file->bs);
2599 * Return number of sectors on success, -errno on error.
2601 int64_t bdrv_nb_sectors(BlockDriverState *bs)
2603 BlockDriver *drv = bs->drv;
2608 if (drv->has_variable_length) {
2609 int ret = refresh_total_sectors(bs, bs->total_sectors);
2614 return bs->total_sectors;
2618 * Return length in bytes on success, -errno on error.
2619 * The length is always a multiple of BDRV_SECTOR_SIZE.
2621 int64_t bdrv_getlength(BlockDriverState *bs)
2623 int64_t ret = bdrv_nb_sectors(bs);
2625 ret = ret > INT64_MAX / BDRV_SECTOR_SIZE ? -EFBIG : ret;
2626 return ret < 0 ? ret : ret * BDRV_SECTOR_SIZE;
2629 /* return 0 as number of sectors if no device present or error */
2630 void bdrv_get_geometry(BlockDriverState *bs, uint64_t *nb_sectors_ptr)
2632 int64_t nb_sectors = bdrv_nb_sectors(bs);
2634 *nb_sectors_ptr = nb_sectors < 0 ? 0 : nb_sectors;
2637 bool bdrv_is_read_only(BlockDriverState *bs)
2639 return bs->read_only;
2642 bool bdrv_is_sg(BlockDriverState *bs)
2647 bool bdrv_is_encrypted(BlockDriverState *bs)
2649 if (bs->backing && bs->backing->bs->encrypted) {
2652 return bs->encrypted;
2655 bool bdrv_key_required(BlockDriverState *bs)
2657 BdrvChild *backing = bs->backing;
2659 if (backing && backing->bs->encrypted && !backing->bs->valid_key) {
2662 return (bs->encrypted && !bs->valid_key);
2665 int bdrv_set_key(BlockDriverState *bs, const char *key)
2668 if (bs->backing && bs->backing->bs->encrypted) {
2669 ret = bdrv_set_key(bs->backing->bs, key);
2675 if (!bs->encrypted) {
2677 } else if (!bs->drv || !bs->drv->bdrv_set_key) {
2680 ret = bs->drv->bdrv_set_key(bs, key);
2682 bs->valid_key = false;
2683 } else if (!bs->valid_key) {
2684 /* call the change callback now, we skipped it on open */
2685 bs->valid_key = true;
2686 bdrv_parent_cb_change_media(bs, true);
2692 * Provide an encryption key for @bs.
2693 * If @key is non-null:
2694 * If @bs is not encrypted, fail.
2695 * Else if the key is invalid, fail.
2696 * Else set @bs's key to @key, replacing the existing key, if any.
2698 * If @bs is encrypted and still lacks a key, fail.
2700 * On failure, store an error object through @errp if non-null.
2702 void bdrv_add_key(BlockDriverState *bs, const char *key, Error **errp)
2705 if (!bdrv_is_encrypted(bs)) {
2706 error_setg(errp, "Node '%s' is not encrypted",
2707 bdrv_get_device_or_node_name(bs));
2708 } else if (bdrv_set_key(bs, key) < 0) {
2709 error_setg(errp, QERR_INVALID_PASSWORD);
2712 if (bdrv_key_required(bs)) {
2713 error_set(errp, ERROR_CLASS_DEVICE_ENCRYPTED,
2714 "'%s' (%s) is encrypted",
2715 bdrv_get_device_or_node_name(bs),
2716 bdrv_get_encrypted_filename(bs));
2721 const char *bdrv_get_format_name(BlockDriverState *bs)
2723 return bs->drv ? bs->drv->format_name : NULL;
2726 static int qsort_strcmp(const void *a, const void *b)
2728 return strcmp(a, b);
2731 void bdrv_iterate_format(void (*it)(void *opaque, const char *name),
2737 const char **formats = NULL;
2739 QLIST_FOREACH(drv, &bdrv_drivers, list) {
2740 if (drv->format_name) {
2743 while (formats && i && !found) {
2744 found = !strcmp(formats[--i], drv->format_name);
2748 formats = g_renew(const char *, formats, count + 1);
2749 formats[count++] = drv->format_name;
2754 qsort(formats, count, sizeof(formats[0]), qsort_strcmp);
2756 for (i = 0; i < count; i++) {
2757 it(opaque, formats[i]);
2763 /* This function is to find a node in the bs graph */
2764 BlockDriverState *bdrv_find_node(const char *node_name)
2766 BlockDriverState *bs;
2770 QTAILQ_FOREACH(bs, &graph_bdrv_states, node_list) {
2771 if (!strcmp(node_name, bs->node_name)) {
2778 /* Put this QMP function here so it can access the static graph_bdrv_states. */
2779 BlockDeviceInfoList *bdrv_named_nodes_list(Error **errp)
2781 BlockDeviceInfoList *list, *entry;
2782 BlockDriverState *bs;
2785 QTAILQ_FOREACH(bs, &graph_bdrv_states, node_list) {
2786 BlockDeviceInfo *info = bdrv_block_device_info(NULL, bs, errp);
2788 qapi_free_BlockDeviceInfoList(list);
2791 entry = g_malloc0(sizeof(*entry));
2792 entry->value = info;
2800 BlockDriverState *bdrv_lookup_bs(const char *device,
2801 const char *node_name,
2805 BlockDriverState *bs;
2808 blk = blk_by_name(device);
2813 error_setg(errp, "Device '%s' has no medium", device);
2821 bs = bdrv_find_node(node_name);
2828 error_setg(errp, "Cannot find device=%s nor node_name=%s",
2829 device ? device : "",
2830 node_name ? node_name : "");
2834 /* If 'base' is in the same chain as 'top', return true. Otherwise,
2835 * return false. If either argument is NULL, return false. */
2836 bool bdrv_chain_contains(BlockDriverState *top, BlockDriverState *base)
2838 while (top && top != base) {
2839 top = backing_bs(top);
2845 BlockDriverState *bdrv_next_node(BlockDriverState *bs)
2848 return QTAILQ_FIRST(&graph_bdrv_states);
2850 return QTAILQ_NEXT(bs, node_list);
2853 const char *bdrv_get_node_name(const BlockDriverState *bs)
2855 return bs->node_name;
2858 const char *bdrv_get_parent_name(const BlockDriverState *bs)
2863 /* If multiple parents have a name, just pick the first one. */
2864 QLIST_FOREACH(c, &bs->parents, next_parent) {
2865 if (c->role->get_name) {
2866 name = c->role->get_name(c);
2867 if (name && *name) {
2876 /* TODO check what callers really want: bs->node_name or blk_name() */
2877 const char *bdrv_get_device_name(const BlockDriverState *bs)
2879 return bdrv_get_parent_name(bs) ?: "";
2882 /* This can be used to identify nodes that might not have a device
2883 * name associated. Since node and device names live in the same
2884 * namespace, the result is unambiguous. The exception is if both are
2885 * absent, then this returns an empty (non-null) string. */
2886 const char *bdrv_get_device_or_node_name(const BlockDriverState *bs)
2888 return bdrv_get_parent_name(bs) ?: bs->node_name;
2891 int bdrv_get_flags(BlockDriverState *bs)
2893 return bs->open_flags;
2896 int bdrv_has_zero_init_1(BlockDriverState *bs)
2901 int bdrv_has_zero_init(BlockDriverState *bs)
2905 /* If BS is a copy on write image, it is initialized to
2906 the contents of the base image, which may not be zeroes. */
2910 if (bs->drv->bdrv_has_zero_init) {
2911 return bs->drv->bdrv_has_zero_init(bs);
2918 bool bdrv_unallocated_blocks_are_zero(BlockDriverState *bs)
2920 BlockDriverInfo bdi;
2926 if (bdrv_get_info(bs, &bdi) == 0) {
2927 return bdi.unallocated_blocks_are_zero;
2933 bool bdrv_can_write_zeroes_with_unmap(BlockDriverState *bs)
2935 BlockDriverInfo bdi;
2937 if (!(bs->open_flags & BDRV_O_UNMAP)) {
2941 if (bdrv_get_info(bs, &bdi) == 0) {
2942 return bdi.can_write_zeroes_with_unmap;
2948 const char *bdrv_get_encrypted_filename(BlockDriverState *bs)
2950 if (bs->backing && bs->backing->bs->encrypted)
2951 return bs->backing_file;
2952 else if (bs->encrypted)
2953 return bs->filename;
2958 void bdrv_get_backing_filename(BlockDriverState *bs,
2959 char *filename, int filename_size)
2961 pstrcpy(filename, filename_size, bs->backing_file);
2964 int bdrv_get_info(BlockDriverState *bs, BlockDriverInfo *bdi)
2966 BlockDriver *drv = bs->drv;
2969 if (!drv->bdrv_get_info)
2971 memset(bdi, 0, sizeof(*bdi));
2972 return drv->bdrv_get_info(bs, bdi);
2975 ImageInfoSpecific *bdrv_get_specific_info(BlockDriverState *bs)
2977 BlockDriver *drv = bs->drv;
2978 if (drv && drv->bdrv_get_specific_info) {
2979 return drv->bdrv_get_specific_info(bs);
2984 void bdrv_debug_event(BlockDriverState *bs, BlkdebugEvent event)
2986 if (!bs || !bs->drv || !bs->drv->bdrv_debug_event) {
2990 bs->drv->bdrv_debug_event(bs, event);
2993 int bdrv_debug_breakpoint(BlockDriverState *bs, const char *event,
2996 while (bs && bs->drv && !bs->drv->bdrv_debug_breakpoint) {
2997 bs = bs->file ? bs->file->bs : NULL;
3000 if (bs && bs->drv && bs->drv->bdrv_debug_breakpoint) {
3001 return bs->drv->bdrv_debug_breakpoint(bs, event, tag);
3007 int bdrv_debug_remove_breakpoint(BlockDriverState *bs, const char *tag)
3009 while (bs && bs->drv && !bs->drv->bdrv_debug_remove_breakpoint) {
3010 bs = bs->file ? bs->file->bs : NULL;
3013 if (bs && bs->drv && bs->drv->bdrv_debug_remove_breakpoint) {
3014 return bs->drv->bdrv_debug_remove_breakpoint(bs, tag);
3020 int bdrv_debug_resume(BlockDriverState *bs, const char *tag)
3022 while (bs && (!bs->drv || !bs->drv->bdrv_debug_resume)) {
3023 bs = bs->file ? bs->file->bs : NULL;
3026 if (bs && bs->drv && bs->drv->bdrv_debug_resume) {
3027 return bs->drv->bdrv_debug_resume(bs, tag);
3033 bool bdrv_debug_is_suspended(BlockDriverState *bs, const char *tag)
3035 while (bs && bs->drv && !bs->drv->bdrv_debug_is_suspended) {
3036 bs = bs->file ? bs->file->bs : NULL;
3039 if (bs && bs->drv && bs->drv->bdrv_debug_is_suspended) {
3040 return bs->drv->bdrv_debug_is_suspended(bs, tag);
3046 /* backing_file can either be relative, or absolute, or a protocol. If it is
3047 * relative, it must be relative to the chain. So, passing in bs->filename
3048 * from a BDS as backing_file should not be done, as that may be relative to
3049 * the CWD rather than the chain. */
3050 BlockDriverState *bdrv_find_backing_image(BlockDriverState *bs,
3051 const char *backing_file)
3053 char *filename_full = NULL;
3054 char *backing_file_full = NULL;
3055 char *filename_tmp = NULL;
3056 int is_protocol = 0;
3057 BlockDriverState *curr_bs = NULL;
3058 BlockDriverState *retval = NULL;
3060 if (!bs || !bs->drv || !backing_file) {
3064 filename_full = g_malloc(PATH_MAX);
3065 backing_file_full = g_malloc(PATH_MAX);
3066 filename_tmp = g_malloc(PATH_MAX);
3068 is_protocol = path_has_protocol(backing_file);
3070 for (curr_bs = bs; curr_bs->backing; curr_bs = curr_bs->backing->bs) {
3072 /* If either of the filename paths is actually a protocol, then
3073 * compare unmodified paths; otherwise make paths relative */
3074 if (is_protocol || path_has_protocol(curr_bs->backing_file)) {
3075 if (strcmp(backing_file, curr_bs->backing_file) == 0) {
3076 retval = curr_bs->backing->bs;
3080 /* If not an absolute filename path, make it relative to the current
3081 * image's filename path */
3082 path_combine(filename_tmp, PATH_MAX, curr_bs->filename,
3085 /* We are going to compare absolute pathnames */
3086 if (!realpath(filename_tmp, filename_full)) {
3090 /* We need to make sure the backing filename we are comparing against
3091 * is relative to the current image filename (or absolute) */
3092 path_combine(filename_tmp, PATH_MAX, curr_bs->filename,
3093 curr_bs->backing_file);
3095 if (!realpath(filename_tmp, backing_file_full)) {
3099 if (strcmp(backing_file_full, filename_full) == 0) {
3100 retval = curr_bs->backing->bs;
3106 g_free(filename_full);
3107 g_free(backing_file_full);
3108 g_free(filename_tmp);
3112 int bdrv_get_backing_file_depth(BlockDriverState *bs)
3122 return 1 + bdrv_get_backing_file_depth(bs->backing->bs);
3125 void bdrv_init(void)
3127 module_call_init(MODULE_INIT_BLOCK);
3130 void bdrv_init_with_whitelist(void)
3132 use_bdrv_whitelist = 1;
3136 void bdrv_invalidate_cache(BlockDriverState *bs, Error **errp)
3139 Error *local_err = NULL;
3146 if (!(bs->open_flags & BDRV_O_INACTIVE)) {
3149 bs->open_flags &= ~BDRV_O_INACTIVE;
3151 if (bs->drv->bdrv_invalidate_cache) {
3152 bs->drv->bdrv_invalidate_cache(bs, &local_err);
3154 bs->open_flags |= BDRV_O_INACTIVE;
3155 error_propagate(errp, local_err);
3160 QLIST_FOREACH(child, &bs->children, next) {
3161 bdrv_invalidate_cache(child->bs, &local_err);
3163 bs->open_flags |= BDRV_O_INACTIVE;
3164 error_propagate(errp, local_err);
3169 ret = refresh_total_sectors(bs, bs->total_sectors);
3171 bs->open_flags |= BDRV_O_INACTIVE;
3172 error_setg_errno(errp, -ret, "Could not refresh total sector count");
3177 void bdrv_invalidate_cache_all(Error **errp)
3179 BlockDriverState *bs;
3180 Error *local_err = NULL;
3181 BdrvNextIterator it;
3183 for (bs = bdrv_first(&it); bs; bs = bdrv_next(&it)) {
3184 AioContext *aio_context = bdrv_get_aio_context(bs);
3186 aio_context_acquire(aio_context);
3187 bdrv_invalidate_cache(bs, &local_err);
3188 aio_context_release(aio_context);
3190 error_propagate(errp, local_err);
3196 static int bdrv_inactivate_recurse(BlockDriverState *bs,
3202 if (!setting_flag && bs->drv->bdrv_inactivate) {
3203 ret = bs->drv->bdrv_inactivate(bs);
3209 QLIST_FOREACH(child, &bs->children, next) {
3210 ret = bdrv_inactivate_recurse(child->bs, setting_flag);
3217 bs->open_flags |= BDRV_O_INACTIVE;
3222 int bdrv_inactivate_all(void)
3224 BlockDriverState *bs = NULL;
3225 BdrvNextIterator it;
3229 for (bs = bdrv_first(&it); bs; bs = bdrv_next(&it)) {
3230 aio_context_acquire(bdrv_get_aio_context(bs));
3233 /* We do two passes of inactivation. The first pass calls to drivers'
3234 * .bdrv_inactivate callbacks recursively so all cache is flushed to disk;
3235 * the second pass sets the BDRV_O_INACTIVE flag so that no further write
3237 for (pass = 0; pass < 2; pass++) {
3238 for (bs = bdrv_first(&it); bs; bs = bdrv_next(&it)) {
3239 ret = bdrv_inactivate_recurse(bs, pass);
3247 for (bs = bdrv_first(&it); bs; bs = bdrv_next(&it)) {
3248 aio_context_release(bdrv_get_aio_context(bs));
3254 /**************************************************************/
3255 /* removable device support */
3258 * Return TRUE if the media is present
3260 bool bdrv_is_inserted(BlockDriverState *bs)
3262 BlockDriver *drv = bs->drv;
3268 if (drv->bdrv_is_inserted) {
3269 return drv->bdrv_is_inserted(bs);
3271 QLIST_FOREACH(child, &bs->children, next) {
3272 if (!bdrv_is_inserted(child->bs)) {
3280 * Return whether the media changed since the last call to this
3281 * function, or -ENOTSUP if we don't know. Most drivers don't know.
3283 int bdrv_media_changed(BlockDriverState *bs)
3285 BlockDriver *drv = bs->drv;
3287 if (drv && drv->bdrv_media_changed) {
3288 return drv->bdrv_media_changed(bs);
3294 * If eject_flag is TRUE, eject the media. Otherwise, close the tray
3296 void bdrv_eject(BlockDriverState *bs, bool eject_flag)
3298 BlockDriver *drv = bs->drv;
3299 const char *device_name;
3301 if (drv && drv->bdrv_eject) {
3302 drv->bdrv_eject(bs, eject_flag);
3305 device_name = bdrv_get_device_name(bs);
3306 if (device_name[0] != '\0') {
3307 qapi_event_send_device_tray_moved(device_name,
3308 eject_flag, &error_abort);
3313 * Lock or unlock the media (if it is locked, the user won't be able
3314 * to eject it manually).
3316 void bdrv_lock_medium(BlockDriverState *bs, bool locked)
3318 BlockDriver *drv = bs->drv;
3320 trace_bdrv_lock_medium(bs, locked);
3322 if (drv && drv->bdrv_lock_medium) {
3323 drv->bdrv_lock_medium(bs, locked);
3327 /* Get a reference to bs */
3328 void bdrv_ref(BlockDriverState *bs)
3333 /* Release a previously grabbed reference to bs.
3334 * If after releasing, reference count is zero, the BlockDriverState is
3336 void bdrv_unref(BlockDriverState *bs)
3341 assert(bs->refcnt > 0);
3342 if (--bs->refcnt == 0) {
3347 struct BdrvOpBlocker {
3349 QLIST_ENTRY(BdrvOpBlocker) list;
3352 bool bdrv_op_is_blocked(BlockDriverState *bs, BlockOpType op, Error **errp)
3354 BdrvOpBlocker *blocker;
3355 assert((int) op >= 0 && op < BLOCK_OP_TYPE_MAX);
3356 if (!QLIST_EMPTY(&bs->op_blockers[op])) {
3357 blocker = QLIST_FIRST(&bs->op_blockers[op]);
3359 *errp = error_copy(blocker->reason);
3360 error_prepend(errp, "Node '%s' is busy: ",
3361 bdrv_get_device_or_node_name(bs));
3368 void bdrv_op_block(BlockDriverState *bs, BlockOpType op, Error *reason)
3370 BdrvOpBlocker *blocker;
3371 assert((int) op >= 0 && op < BLOCK_OP_TYPE_MAX);
3373 blocker = g_new0(BdrvOpBlocker, 1);
3374 blocker->reason = reason;
3375 QLIST_INSERT_HEAD(&bs->op_blockers[op], blocker, list);
3378 void bdrv_op_unblock(BlockDriverState *bs, BlockOpType op, Error *reason)
3380 BdrvOpBlocker *blocker, *next;
3381 assert((int) op >= 0 && op < BLOCK_OP_TYPE_MAX);
3382 QLIST_FOREACH_SAFE(blocker, &bs->op_blockers[op], list, next) {
3383 if (blocker->reason == reason) {
3384 QLIST_REMOVE(blocker, list);
3390 void bdrv_op_block_all(BlockDriverState *bs, Error *reason)
3393 for (i = 0; i < BLOCK_OP_TYPE_MAX; i++) {
3394 bdrv_op_block(bs, i, reason);
3398 void bdrv_op_unblock_all(BlockDriverState *bs, Error *reason)
3401 for (i = 0; i < BLOCK_OP_TYPE_MAX; i++) {
3402 bdrv_op_unblock(bs, i, reason);
3406 bool bdrv_op_blocker_is_empty(BlockDriverState *bs)
3410 for (i = 0; i < BLOCK_OP_TYPE_MAX; i++) {
3411 if (!QLIST_EMPTY(&bs->op_blockers[i])) {
3418 void bdrv_img_create(const char *filename, const char *fmt,
3419 const char *base_filename, const char *base_fmt,
3420 char *options, uint64_t img_size, int flags,
3421 Error **errp, bool quiet)
3423 QemuOptsList *create_opts = NULL;
3424 QemuOpts *opts = NULL;
3425 const char *backing_fmt, *backing_file;
3427 BlockDriver *drv, *proto_drv;
3428 Error *local_err = NULL;
3431 /* Find driver and parse its options */
3432 drv = bdrv_find_format(fmt);
3434 error_setg(errp, "Unknown file format '%s'", fmt);
3438 proto_drv = bdrv_find_protocol(filename, true, errp);
3443 if (!drv->create_opts) {
3444 error_setg(errp, "Format driver '%s' does not support image creation",
3449 if (!proto_drv->create_opts) {
3450 error_setg(errp, "Protocol driver '%s' does not support image creation",
3451 proto_drv->format_name);
3455 create_opts = qemu_opts_append(create_opts, drv->create_opts);
3456 create_opts = qemu_opts_append(create_opts, proto_drv->create_opts);
3458 /* Create parameter list with default values */
3459 opts = qemu_opts_create(create_opts, NULL, 0, &error_abort);
3460 qemu_opt_set_number(opts, BLOCK_OPT_SIZE, img_size, &error_abort);
3462 /* Parse -o options */
3464 qemu_opts_do_parse(opts, options, NULL, &local_err);
3466 error_report_err(local_err);
3468 error_setg(errp, "Invalid options for file format '%s'", fmt);
3473 if (base_filename) {
3474 qemu_opt_set(opts, BLOCK_OPT_BACKING_FILE, base_filename, &local_err);
3476 error_setg(errp, "Backing file not supported for file format '%s'",
3483 qemu_opt_set(opts, BLOCK_OPT_BACKING_FMT, base_fmt, &local_err);
3485 error_setg(errp, "Backing file format not supported for file "
3486 "format '%s'", fmt);
3491 backing_file = qemu_opt_get(opts, BLOCK_OPT_BACKING_FILE);
3493 if (!strcmp(filename, backing_file)) {
3494 error_setg(errp, "Error: Trying to create an image with the "
3495 "same filename as the backing file");
3500 backing_fmt = qemu_opt_get(opts, BLOCK_OPT_BACKING_FMT);
3502 // The size for the image must always be specified, with one exception:
3503 // If we are using a backing file, we can obtain the size from there
3504 size = qemu_opt_get_size(opts, BLOCK_OPT_SIZE, 0);
3507 BlockDriverState *bs;
3508 char *full_backing = g_new0(char, PATH_MAX);
3511 QDict *backing_options = NULL;
3513 bdrv_get_full_backing_filename_from_filename(filename, backing_file,
3514 full_backing, PATH_MAX,
3517 g_free(full_backing);
3521 /* backing files always opened read-only */
3523 back_flags &= ~(BDRV_O_RDWR | BDRV_O_SNAPSHOT | BDRV_O_NO_BACKING);
3526 backing_options = qdict_new();
3527 qdict_put(backing_options, "driver",
3528 qstring_from_str(backing_fmt));
3531 bs = bdrv_open(full_backing, NULL, backing_options, back_flags,
3533 g_free(full_backing);
3537 size = bdrv_getlength(bs);
3539 error_setg_errno(errp, -size, "Could not get size of '%s'",
3545 qemu_opt_set_number(opts, BLOCK_OPT_SIZE, size, &error_abort);
3549 error_setg(errp, "Image creation needs a size parameter");
3555 printf("Formatting '%s', fmt=%s ", filename, fmt);
3556 qemu_opts_print(opts, " ");
3560 ret = bdrv_create(drv, filename, opts, &local_err);
3562 if (ret == -EFBIG) {
3563 /* This is generally a better message than whatever the driver would
3564 * deliver (especially because of the cluster_size_hint), since that
3565 * is most probably not much different from "image too large". */
3566 const char *cluster_size_hint = "";
3567 if (qemu_opt_get_size(opts, BLOCK_OPT_CLUSTER_SIZE, 0)) {
3568 cluster_size_hint = " (try using a larger cluster size)";
3570 error_setg(errp, "The image size is too large for file format '%s'"
3571 "%s", fmt, cluster_size_hint);
3572 error_free(local_err);
3577 qemu_opts_del(opts);
3578 qemu_opts_free(create_opts);
3579 error_propagate(errp, local_err);
3582 AioContext *bdrv_get_aio_context(BlockDriverState *bs)
3584 return bs->aio_context;
3587 static void bdrv_do_remove_aio_context_notifier(BdrvAioNotifier *ban)
3589 QLIST_REMOVE(ban, list);
3593 void bdrv_detach_aio_context(BlockDriverState *bs)
3595 BdrvAioNotifier *baf, *baf_tmp;
3602 assert(!bs->walking_aio_notifiers);
3603 bs->walking_aio_notifiers = true;
3604 QLIST_FOREACH_SAFE(baf, &bs->aio_notifiers, list, baf_tmp) {
3606 bdrv_do_remove_aio_context_notifier(baf);
3608 baf->detach_aio_context(baf->opaque);
3611 /* Never mind iterating again to check for ->deleted. bdrv_close() will
3612 * remove remaining aio notifiers if we aren't called again.
3614 bs->walking_aio_notifiers = false;
3616 if (bs->drv->bdrv_detach_aio_context) {
3617 bs->drv->bdrv_detach_aio_context(bs);
3619 QLIST_FOREACH(child, &bs->children, next) {
3620 bdrv_detach_aio_context(child->bs);
3623 bs->aio_context = NULL;
3626 void bdrv_attach_aio_context(BlockDriverState *bs,
3627 AioContext *new_context)
3629 BdrvAioNotifier *ban, *ban_tmp;
3636 bs->aio_context = new_context;
3638 QLIST_FOREACH(child, &bs->children, next) {
3639 bdrv_attach_aio_context(child->bs, new_context);
3641 if (bs->drv->bdrv_attach_aio_context) {
3642 bs->drv->bdrv_attach_aio_context(bs, new_context);
3645 assert(!bs->walking_aio_notifiers);
3646 bs->walking_aio_notifiers = true;
3647 QLIST_FOREACH_SAFE(ban, &bs->aio_notifiers, list, ban_tmp) {
3649 bdrv_do_remove_aio_context_notifier(ban);
3651 ban->attached_aio_context(new_context, ban->opaque);
3654 bs->walking_aio_notifiers = false;
3657 void bdrv_set_aio_context(BlockDriverState *bs, AioContext *new_context)
3659 bdrv_drain(bs); /* ensure there are no in-flight requests */
3661 bdrv_detach_aio_context(bs);
3663 /* This function executes in the old AioContext so acquire the new one in
3664 * case it runs in a different thread.
3666 aio_context_acquire(new_context);
3667 bdrv_attach_aio_context(bs, new_context);
3668 aio_context_release(new_context);
3671 void bdrv_add_aio_context_notifier(BlockDriverState *bs,
3672 void (*attached_aio_context)(AioContext *new_context, void *opaque),
3673 void (*detach_aio_context)(void *opaque), void *opaque)
3675 BdrvAioNotifier *ban = g_new(BdrvAioNotifier, 1);
3676 *ban = (BdrvAioNotifier){
3677 .attached_aio_context = attached_aio_context,
3678 .detach_aio_context = detach_aio_context,
3682 QLIST_INSERT_HEAD(&bs->aio_notifiers, ban, list);
3685 void bdrv_remove_aio_context_notifier(BlockDriverState *bs,
3686 void (*attached_aio_context)(AioContext *,
3688 void (*detach_aio_context)(void *),
3691 BdrvAioNotifier *ban, *ban_next;
3693 QLIST_FOREACH_SAFE(ban, &bs->aio_notifiers, list, ban_next) {
3694 if (ban->attached_aio_context == attached_aio_context &&
3695 ban->detach_aio_context == detach_aio_context &&
3696 ban->opaque == opaque &&
3697 ban->deleted == false)
3699 if (bs->walking_aio_notifiers) {
3700 ban->deleted = true;
3702 bdrv_do_remove_aio_context_notifier(ban);
3711 int bdrv_amend_options(BlockDriverState *bs, QemuOpts *opts,
3712 BlockDriverAmendStatusCB *status_cb, void *cb_opaque)
3714 if (!bs->drv->bdrv_amend_options) {
3717 return bs->drv->bdrv_amend_options(bs, opts, status_cb, cb_opaque);
3720 /* This function will be called by the bdrv_recurse_is_first_non_filter method
3721 * of block filter and by bdrv_is_first_non_filter.
3722 * It is used to test if the given bs is the candidate or recurse more in the
3725 bool bdrv_recurse_is_first_non_filter(BlockDriverState *bs,
3726 BlockDriverState *candidate)
3728 /* return false if basic checks fails */
3729 if (!bs || !bs->drv) {
3733 /* the code reached a non block filter driver -> check if the bs is
3734 * the same as the candidate. It's the recursion termination condition.
3736 if (!bs->drv->is_filter) {
3737 return bs == candidate;
3739 /* Down this path the driver is a block filter driver */
3741 /* If the block filter recursion method is defined use it to recurse down
3744 if (bs->drv->bdrv_recurse_is_first_non_filter) {
3745 return bs->drv->bdrv_recurse_is_first_non_filter(bs, candidate);
3748 /* the driver is a block filter but don't allow to recurse -> return false
3753 /* This function checks if the candidate is the first non filter bs down it's
3754 * bs chain. Since we don't have pointers to parents it explore all bs chains
3755 * from the top. Some filters can choose not to pass down the recursion.
3757 bool bdrv_is_first_non_filter(BlockDriverState *candidate)
3759 BlockDriverState *bs;
3760 BdrvNextIterator it;
3762 /* walk down the bs forest recursively */
3763 for (bs = bdrv_first(&it); bs; bs = bdrv_next(&it)) {
3766 /* try to recurse in this top level bs */
3767 perm = bdrv_recurse_is_first_non_filter(bs, candidate);
3769 /* candidate is the first non filter */
3778 BlockDriverState *check_to_replace_node(BlockDriverState *parent_bs,
3779 const char *node_name, Error **errp)
3781 BlockDriverState *to_replace_bs = bdrv_find_node(node_name);
3782 AioContext *aio_context;
3784 if (!to_replace_bs) {
3785 error_setg(errp, "Node name '%s' not found", node_name);
3789 aio_context = bdrv_get_aio_context(to_replace_bs);
3790 aio_context_acquire(aio_context);
3792 if (bdrv_op_is_blocked(to_replace_bs, BLOCK_OP_TYPE_REPLACE, errp)) {
3793 to_replace_bs = NULL;
3797 /* We don't want arbitrary node of the BDS chain to be replaced only the top
3798 * most non filter in order to prevent data corruption.
3799 * Another benefit is that this tests exclude backing files which are
3800 * blocked by the backing blockers.
3802 if (!bdrv_recurse_is_first_non_filter(parent_bs, to_replace_bs)) {
3803 error_setg(errp, "Only top most non filter can be replaced");
3804 to_replace_bs = NULL;
3809 aio_context_release(aio_context);
3810 return to_replace_bs;
3813 static bool append_open_options(QDict *d, BlockDriverState *bs)
3815 const QDictEntry *entry;
3818 bool found_any = false;
3821 for (entry = qdict_first(bs->options); entry;
3822 entry = qdict_next(bs->options, entry))
3824 /* Exclude options for children */
3825 QLIST_FOREACH(child, &bs->children, next) {
3826 if (strstart(qdict_entry_key(entry), child->name, &p)
3827 && (!*p || *p == '.'))
3836 /* And exclude all non-driver-specific options */
3837 for (desc = bdrv_runtime_opts.desc; desc->name; desc++) {
3838 if (!strcmp(qdict_entry_key(entry), desc->name)) {
3846 qobject_incref(qdict_entry_value(entry));
3847 qdict_put_obj(d, qdict_entry_key(entry), qdict_entry_value(entry));
3854 /* Updates the following BDS fields:
3855 * - exact_filename: A filename which may be used for opening a block device
3856 * which (mostly) equals the given BDS (even without any
3857 * other options; so reading and writing must return the same
3858 * results, but caching etc. may be different)
3859 * - full_open_options: Options which, when given when opening a block device
3860 * (without a filename), result in a BDS (mostly)
3861 * equalling the given one
3862 * - filename: If exact_filename is set, it is copied here. Otherwise,
3863 * full_open_options is converted to a JSON object, prefixed with
3864 * "json:" (for use through the JSON pseudo protocol) and put here.
3866 void bdrv_refresh_filename(BlockDriverState *bs)
3868 BlockDriver *drv = bs->drv;
3875 /* This BDS's file name will most probably depend on its file's name, so
3876 * refresh that first */
3878 bdrv_refresh_filename(bs->file->bs);
3881 if (drv->bdrv_refresh_filename) {
3882 /* Obsolete information is of no use here, so drop the old file name
3883 * information before refreshing it */
3884 bs->exact_filename[0] = '\0';
3885 if (bs->full_open_options) {
3886 QDECREF(bs->full_open_options);
3887 bs->full_open_options = NULL;
3891 append_open_options(opts, bs);
3892 drv->bdrv_refresh_filename(bs, opts);
3894 } else if (bs->file) {
3895 /* Try to reconstruct valid information from the underlying file */
3896 bool has_open_options;
3898 bs->exact_filename[0] = '\0';
3899 if (bs->full_open_options) {
3900 QDECREF(bs->full_open_options);
3901 bs->full_open_options = NULL;
3905 has_open_options = append_open_options(opts, bs);
3907 /* If no specific options have been given for this BDS, the filename of
3908 * the underlying file should suffice for this one as well */
3909 if (bs->file->bs->exact_filename[0] && !has_open_options) {
3910 strcpy(bs->exact_filename, bs->file->bs->exact_filename);
3912 /* Reconstructing the full options QDict is simple for most format block
3913 * drivers, as long as the full options are known for the underlying
3914 * file BDS. The full options QDict of that file BDS should somehow
3915 * contain a representation of the filename, therefore the following
3916 * suffices without querying the (exact_)filename of this BDS. */
3917 if (bs->file->bs->full_open_options) {
3918 qdict_put_obj(opts, "driver",
3919 QOBJECT(qstring_from_str(drv->format_name)));
3920 QINCREF(bs->file->bs->full_open_options);
3921 qdict_put_obj(opts, "file",
3922 QOBJECT(bs->file->bs->full_open_options));
3924 bs->full_open_options = opts;
3928 } else if (!bs->full_open_options && qdict_size(bs->options)) {
3929 /* There is no underlying file BDS (at least referenced by BDS.file),
3930 * so the full options QDict should be equal to the options given
3931 * specifically for this block device when it was opened (plus the
3932 * driver specification).
3933 * Because those options don't change, there is no need to update
3934 * full_open_options when it's already set. */
3937 append_open_options(opts, bs);
3938 qdict_put_obj(opts, "driver",
3939 QOBJECT(qstring_from_str(drv->format_name)));
3941 if (bs->exact_filename[0]) {
3942 /* This may not work for all block protocol drivers (some may
3943 * require this filename to be parsed), but we have to find some
3944 * default solution here, so just include it. If some block driver
3945 * does not support pure options without any filename at all or
3946 * needs some special format of the options QDict, it needs to
3947 * implement the driver-specific bdrv_refresh_filename() function.
3949 qdict_put_obj(opts, "filename",
3950 QOBJECT(qstring_from_str(bs->exact_filename)));
3953 bs->full_open_options = opts;
3956 if (bs->exact_filename[0]) {
3957 pstrcpy(bs->filename, sizeof(bs->filename), bs->exact_filename);
3958 } else if (bs->full_open_options) {
3959 QString *json = qobject_to_json(QOBJECT(bs->full_open_options));
3960 snprintf(bs->filename, sizeof(bs->filename), "json:%s",
3961 qstring_get_str(json));
3967 * Hot add/remove a BDS's child. So the user can take a child offline when
3968 * it is broken and take a new child online
3970 void bdrv_add_child(BlockDriverState *parent_bs, BlockDriverState *child_bs,
3974 if (!parent_bs->drv || !parent_bs->drv->bdrv_add_child) {
3975 error_setg(errp, "The node %s does not support adding a child",
3976 bdrv_get_device_or_node_name(parent_bs));
3980 if (!QLIST_EMPTY(&child_bs->parents)) {
3981 error_setg(errp, "The node %s already has a parent",
3982 child_bs->node_name);
3986 parent_bs->drv->bdrv_add_child(parent_bs, child_bs, errp);
3989 void bdrv_del_child(BlockDriverState *parent_bs, BdrvChild *child, Error **errp)
3993 if (!parent_bs->drv || !parent_bs->drv->bdrv_del_child) {
3994 error_setg(errp, "The node %s does not support removing a child",
3995 bdrv_get_device_or_node_name(parent_bs));
3999 QLIST_FOREACH(tmp, &parent_bs->children, next) {
4006 error_setg(errp, "The node %s does not have a child named %s",
4007 bdrv_get_device_or_node_name(parent_bs),
4008 bdrv_get_device_or_node_name(child->bs));
4012 parent_bs->drv->bdrv_del_child(parent_bs, child, errp);