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 "config-host.h"
25 #include "qemu-common.h"
27 #include "block_int.h"
29 #include "qemu-objects.h"
32 #include <sys/types.h>
34 #include <sys/ioctl.h>
35 #include <sys/queue.h>
45 static BlockDriverAIOCB *bdrv_aio_readv_em(BlockDriverState *bs,
46 int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
47 BlockDriverCompletionFunc *cb, void *opaque);
48 static BlockDriverAIOCB *bdrv_aio_writev_em(BlockDriverState *bs,
49 int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
50 BlockDriverCompletionFunc *cb, void *opaque);
51 static BlockDriverAIOCB *bdrv_aio_flush_em(BlockDriverState *bs,
52 BlockDriverCompletionFunc *cb, void *opaque);
53 static int bdrv_read_em(BlockDriverState *bs, int64_t sector_num,
54 uint8_t *buf, int nb_sectors);
55 static int bdrv_write_em(BlockDriverState *bs, int64_t sector_num,
56 const uint8_t *buf, int nb_sectors);
58 BlockDriverState *bdrv_first;
60 static BlockDriver *first_drv;
62 /* If non-zero, use only whitelisted block drivers */
63 static int use_bdrv_whitelist;
65 int path_is_absolute(const char *path)
69 /* specific case for names like: "\\.\d:" */
70 if (*path == '/' || *path == '\\')
73 p = strchr(path, ':');
79 return (*p == '/' || *p == '\\');
85 /* if filename is absolute, just copy it to dest. Otherwise, build a
86 path to it by considering it is relative to base_path. URL are
88 void path_combine(char *dest, int dest_size,
89 const char *base_path,
97 if (path_is_absolute(filename)) {
98 pstrcpy(dest, dest_size, filename);
100 p = strchr(base_path, ':');
105 p1 = strrchr(base_path, '/');
109 p2 = strrchr(base_path, '\\');
121 if (len > dest_size - 1)
123 memcpy(dest, base_path, len);
125 pstrcat(dest, dest_size, filename);
129 void bdrv_register(BlockDriver *bdrv)
131 if (!bdrv->bdrv_aio_readv) {
132 /* add AIO emulation layer */
133 bdrv->bdrv_aio_readv = bdrv_aio_readv_em;
134 bdrv->bdrv_aio_writev = bdrv_aio_writev_em;
135 } else if (!bdrv->bdrv_read) {
136 /* add synchronous IO emulation layer */
137 bdrv->bdrv_read = bdrv_read_em;
138 bdrv->bdrv_write = bdrv_write_em;
141 if (!bdrv->bdrv_aio_flush)
142 bdrv->bdrv_aio_flush = bdrv_aio_flush_em;
144 bdrv->next = first_drv;
148 /* create a new block device (by default it is empty) */
149 BlockDriverState *bdrv_new(const char *device_name)
151 BlockDriverState **pbs, *bs;
153 bs = qemu_mallocz(sizeof(BlockDriverState));
154 pstrcpy(bs->device_name, sizeof(bs->device_name), device_name);
155 if (device_name[0] != '\0') {
156 /* insert at the end */
165 BlockDriver *bdrv_find_format(const char *format_name)
168 for(drv1 = first_drv; drv1 != NULL; drv1 = drv1->next) {
169 if (!strcmp(drv1->format_name, format_name))
175 static int bdrv_is_whitelisted(BlockDriver *drv)
177 static const char *whitelist[] = {
178 CONFIG_BDRV_WHITELIST
183 return 1; /* no whitelist, anything goes */
185 for (p = whitelist; *p; p++) {
186 if (!strcmp(drv->format_name, *p)) {
193 BlockDriver *bdrv_find_whitelisted_format(const char *format_name)
195 BlockDriver *drv = bdrv_find_format(format_name);
196 return drv && bdrv_is_whitelisted(drv) ? drv : NULL;
199 int bdrv_create(BlockDriver *drv, const char* filename,
200 QEMUOptionParameter *options)
202 if (!drv->bdrv_create)
205 return drv->bdrv_create(filename, options);
209 void get_tmp_filename(char *filename, int size)
211 char temp_dir[MAX_PATH];
213 GetTempPath(MAX_PATH, temp_dir);
214 GetTempFileName(temp_dir, "qem", 0, filename);
217 void get_tmp_filename(char *filename, int size)
221 /* XXX: race condition possible */
222 tmpdir = getenv("TMPDIR");
225 snprintf(filename, size, "%s/vl.XXXXXX", tmpdir);
226 fd = mkstemp(filename);
232 static int is_windows_drive_prefix(const char *filename)
234 return (((filename[0] >= 'a' && filename[0] <= 'z') ||
235 (filename[0] >= 'A' && filename[0] <= 'Z')) &&
239 int is_windows_drive(const char *filename)
241 if (is_windows_drive_prefix(filename) &&
244 if (strstart(filename, "\\\\.\\", NULL) ||
245 strstart(filename, "//./", NULL))
251 static BlockDriver *find_protocol(const char *filename)
259 if (is_windows_drive(filename) ||
260 is_windows_drive_prefix(filename))
261 return bdrv_find_format("raw");
263 p = strchr(filename, ':');
265 return bdrv_find_format("raw");
267 if (len > sizeof(protocol) - 1)
268 len = sizeof(protocol) - 1;
269 memcpy(protocol, filename, len);
270 protocol[len] = '\0';
271 for(drv1 = first_drv; drv1 != NULL; drv1 = drv1->next) {
272 if (drv1->protocol_name &&
273 !strcmp(drv1->protocol_name, protocol))
280 * Detect host devices. By convention, /dev/cdrom[N] is always
281 * recognized as a host CDROM.
283 static BlockDriver *find_hdev_driver(const char *filename)
285 int score_max = 0, score;
286 BlockDriver *drv = NULL, *d;
288 for (d = first_drv; d; d = d->next) {
289 if (d->bdrv_probe_device) {
290 score = d->bdrv_probe_device(filename);
291 if (score > score_max) {
301 static BlockDriver *find_image_format(const char *filename)
303 int ret, score, score_max;
304 BlockDriver *drv1, *drv;
306 BlockDriverState *bs;
308 drv = find_protocol(filename);
309 /* no need to test disk image formats for vvfat */
310 if (drv && strcmp(drv->format_name, "vvfat") == 0)
313 ret = bdrv_file_open(&bs, filename, 0);
316 ret = bdrv_pread(bs, 0, buf, sizeof(buf));
323 for(drv1 = first_drv; drv1 != NULL; drv1 = drv1->next) {
324 if (drv1->bdrv_probe) {
325 score = drv1->bdrv_probe(buf, ret, filename);
326 if (score > score_max) {
335 int bdrv_file_open(BlockDriverState **pbs, const char *filename, int flags)
337 BlockDriverState *bs;
341 ret = bdrv_open2(bs, filename, flags | BDRV_O_FILE, NULL);
351 int bdrv_open(BlockDriverState *bs, const char *filename, int flags)
353 return bdrv_open2(bs, filename, flags, NULL);
356 int bdrv_open2(BlockDriverState *bs, const char *filename, int flags,
360 char tmp_filename[PATH_MAX];
361 char backing_filename[PATH_MAX];
363 bs->is_temporary = 0;
366 /* buffer_alignment defaulted to 512, drivers can change this value */
367 bs->buffer_alignment = 512;
369 if (flags & BDRV_O_SNAPSHOT) {
370 BlockDriverState *bs1;
373 BlockDriver *bdrv_qcow2;
374 QEMUOptionParameter *options;
376 /* if snapshot, we create a temporary backing file and open it
377 instead of opening 'filename' directly */
379 /* if there is a backing file, use it */
381 ret = bdrv_open2(bs1, filename, 0, drv);
386 total_size = bdrv_getlength(bs1) >> BDRV_SECTOR_BITS;
388 if (bs1->drv && bs1->drv->protocol_name)
393 get_tmp_filename(tmp_filename, sizeof(tmp_filename));
395 /* Real path is meaningless for protocols */
397 snprintf(backing_filename, sizeof(backing_filename),
399 else if (!realpath(filename, backing_filename))
402 bdrv_qcow2 = bdrv_find_format("qcow2");
403 options = parse_option_parameters("", bdrv_qcow2->create_options, NULL);
405 set_option_parameter_int(options, BLOCK_OPT_SIZE, total_size * 512);
406 set_option_parameter(options, BLOCK_OPT_BACKING_FILE, backing_filename);
408 set_option_parameter(options, BLOCK_OPT_BACKING_FMT,
412 ret = bdrv_create(bdrv_qcow2, tmp_filename, options);
417 filename = tmp_filename;
419 bs->is_temporary = 1;
422 pstrcpy(bs->filename, sizeof(bs->filename), filename);
423 if (flags & BDRV_O_FILE) {
424 drv = find_protocol(filename);
426 drv = find_hdev_driver(filename);
428 drv = find_image_format(filename);
433 goto unlink_and_fail;
436 bs->opaque = qemu_mallocz(drv->instance_size);
439 * Yes, BDRV_O_NOCACHE aka O_DIRECT means we have to present a
440 * write cache to the guest. We do need the fdatasync to flush
441 * out transactions for block allocations, and we maybe have a
442 * volatile write cache in our backing device to deal with.
444 if (flags & (BDRV_O_CACHE_WB|BDRV_O_NOCACHE))
445 bs->enable_write_cache = 1;
447 bs->read_only = (flags & BDRV_O_RDWR) == 0;
448 if (!(flags & BDRV_O_FILE)) {
449 open_flags = (flags & (BDRV_O_RDWR | BDRV_O_CACHE_MASK|BDRV_O_NATIVE_AIO));
450 if (bs->is_temporary) { /* snapshot should be writeable */
451 open_flags |= BDRV_O_RDWR;
454 open_flags = flags & ~(BDRV_O_FILE | BDRV_O_SNAPSHOT);
456 if (use_bdrv_whitelist && !bdrv_is_whitelisted(drv)) {
459 ret = drv->bdrv_open(bs, filename, open_flags);
462 qemu_free(bs->opaque);
466 if (bs->is_temporary)
470 if (drv->bdrv_getlength) {
471 bs->total_sectors = bdrv_getlength(bs) >> BDRV_SECTOR_BITS;
474 if (bs->is_temporary) {
478 if ((flags & BDRV_O_NO_BACKING) == 0 && bs->backing_file[0] != '\0') {
479 /* if there is a backing file, use it */
480 BlockDriver *back_drv = NULL;
481 bs->backing_hd = bdrv_new("");
482 path_combine(backing_filename, sizeof(backing_filename),
483 filename, bs->backing_file);
484 if (bs->backing_format[0] != '\0')
485 back_drv = bdrv_find_format(bs->backing_format);
486 ret = bdrv_open2(bs->backing_hd, backing_filename, open_flags,
488 bs->backing_hd->read_only = (open_flags & BDRV_O_RDWR) == 0;
495 if (!bdrv_key_required(bs)) {
496 /* call the change callback */
497 bs->media_changed = 1;
499 bs->change_cb(bs->change_opaque);
504 void bdrv_close(BlockDriverState *bs)
508 bdrv_delete(bs->backing_hd);
509 bs->drv->bdrv_close(bs);
510 qemu_free(bs->opaque);
512 if (bs->is_temporary) {
513 unlink(bs->filename);
519 /* call the change callback */
520 bs->media_changed = 1;
522 bs->change_cb(bs->change_opaque);
526 void bdrv_delete(BlockDriverState *bs)
528 BlockDriverState **pbs;
531 while (*pbs != bs && *pbs != NULL)
541 * Run consistency checks on an image
543 * Returns the number of errors or -errno when an internal error occurs
545 int bdrv_check(BlockDriverState *bs)
547 if (bs->drv->bdrv_check == NULL) {
551 return bs->drv->bdrv_check(bs);
554 /* commit COW file into the raw image */
555 int bdrv_commit(BlockDriverState *bs)
557 BlockDriver *drv = bs->drv;
558 int64_t i, total_sectors;
560 unsigned char sector[512];
569 if (!bs->backing_hd) {
573 total_sectors = bdrv_getlength(bs) >> BDRV_SECTOR_BITS;
574 for (i = 0; i < total_sectors;) {
575 if (drv->bdrv_is_allocated(bs, i, 65536, &n)) {
576 for(j = 0; j < n; j++) {
577 if (bdrv_read(bs, i, sector, 1) != 0) {
581 if (bdrv_write(bs->backing_hd, i, sector, 1) != 0) {
591 if (drv->bdrv_make_empty)
592 return drv->bdrv_make_empty(bs);
595 * Make sure all data we wrote to the backing device is actually
599 bdrv_flush(bs->backing_hd);
606 * -EINVAL - backing format specified, but no file
607 * -ENOSPC - can't update the backing file because no space is left in the
609 * -ENOTSUP - format driver doesn't support changing the backing file
611 int bdrv_change_backing_file(BlockDriverState *bs,
612 const char *backing_file, const char *backing_fmt)
614 BlockDriver *drv = bs->drv;
616 if (drv->bdrv_change_backing_file != NULL) {
617 return drv->bdrv_change_backing_file(bs, backing_file, backing_fmt);
623 static int bdrv_check_byte_request(BlockDriverState *bs, int64_t offset,
628 if (!bdrv_is_inserted(bs))
634 len = bdrv_getlength(bs);
639 if ((offset > len) || (len - offset < size))
645 static int bdrv_check_request(BlockDriverState *bs, int64_t sector_num,
648 return bdrv_check_byte_request(bs, sector_num * 512, nb_sectors * 512);
651 /* return < 0 if error. See bdrv_write() for the return codes */
652 int bdrv_read(BlockDriverState *bs, int64_t sector_num,
653 uint8_t *buf, int nb_sectors)
655 BlockDriver *drv = bs->drv;
659 if (bdrv_check_request(bs, sector_num, nb_sectors))
662 return drv->bdrv_read(bs, sector_num, buf, nb_sectors);
665 static void set_dirty_bitmap(BlockDriverState *bs, int64_t sector_num,
666 int nb_sectors, int dirty)
669 unsigned long val, idx, bit;
671 start = sector_num / BDRV_SECTORS_PER_DIRTY_CHUNK;
672 end = (sector_num + nb_sectors - 1) / BDRV_SECTORS_PER_DIRTY_CHUNK;
674 for (; start <= end; start++) {
675 idx = start / (sizeof(unsigned long) * 8);
676 bit = start % (sizeof(unsigned long) * 8);
677 val = bs->dirty_bitmap[idx];
683 bs->dirty_bitmap[idx] = val;
687 /* Return < 0 if error. Important errors are:
688 -EIO generic I/O error (may happen for all errors)
689 -ENOMEDIUM No media inserted.
690 -EINVAL Invalid sector number or nb_sectors
691 -EACCES Trying to write a read-only device
693 int bdrv_write(BlockDriverState *bs, int64_t sector_num,
694 const uint8_t *buf, int nb_sectors)
696 BlockDriver *drv = bs->drv;
701 if (bdrv_check_request(bs, sector_num, nb_sectors))
704 if (bs->dirty_bitmap) {
705 set_dirty_bitmap(bs, sector_num, nb_sectors, 1);
708 return drv->bdrv_write(bs, sector_num, buf, nb_sectors);
711 int bdrv_pread(BlockDriverState *bs, int64_t offset,
712 void *buf, int count1)
714 uint8_t tmp_buf[BDRV_SECTOR_SIZE];
715 int len, nb_sectors, count;
719 /* first read to align to sector start */
720 len = (BDRV_SECTOR_SIZE - offset) & (BDRV_SECTOR_SIZE - 1);
723 sector_num = offset >> BDRV_SECTOR_BITS;
725 if (bdrv_read(bs, sector_num, tmp_buf, 1) < 0)
727 memcpy(buf, tmp_buf + (offset & (BDRV_SECTOR_SIZE - 1)), len);
735 /* read the sectors "in place" */
736 nb_sectors = count >> BDRV_SECTOR_BITS;
737 if (nb_sectors > 0) {
738 if (bdrv_read(bs, sector_num, buf, nb_sectors) < 0)
740 sector_num += nb_sectors;
741 len = nb_sectors << BDRV_SECTOR_BITS;
746 /* add data from the last sector */
748 if (bdrv_read(bs, sector_num, tmp_buf, 1) < 0)
750 memcpy(buf, tmp_buf, count);
755 int bdrv_pwrite(BlockDriverState *bs, int64_t offset,
756 const void *buf, int count1)
758 uint8_t tmp_buf[BDRV_SECTOR_SIZE];
759 int len, nb_sectors, count;
763 /* first write to align to sector start */
764 len = (BDRV_SECTOR_SIZE - offset) & (BDRV_SECTOR_SIZE - 1);
767 sector_num = offset >> BDRV_SECTOR_BITS;
769 if (bdrv_read(bs, sector_num, tmp_buf, 1) < 0)
771 memcpy(tmp_buf + (offset & (BDRV_SECTOR_SIZE - 1)), buf, len);
772 if (bdrv_write(bs, sector_num, tmp_buf, 1) < 0)
781 /* write the sectors "in place" */
782 nb_sectors = count >> BDRV_SECTOR_BITS;
783 if (nb_sectors > 0) {
784 if (bdrv_write(bs, sector_num, buf, nb_sectors) < 0)
786 sector_num += nb_sectors;
787 len = nb_sectors << BDRV_SECTOR_BITS;
792 /* add data from the last sector */
794 if (bdrv_read(bs, sector_num, tmp_buf, 1) < 0)
796 memcpy(tmp_buf, buf, count);
797 if (bdrv_write(bs, sector_num, tmp_buf, 1) < 0)
804 * Truncate file to 'offset' bytes (needed only for file protocols)
806 int bdrv_truncate(BlockDriverState *bs, int64_t offset)
808 BlockDriver *drv = bs->drv;
811 if (!drv->bdrv_truncate)
815 return drv->bdrv_truncate(bs, offset);
819 * Length of a file in bytes. Return < 0 if error or unknown.
821 int64_t bdrv_getlength(BlockDriverState *bs)
823 BlockDriver *drv = bs->drv;
826 if (!drv->bdrv_getlength) {
828 return bs->total_sectors * BDRV_SECTOR_SIZE;
830 return drv->bdrv_getlength(bs);
833 /* return 0 as number of sectors if no device present or error */
834 void bdrv_get_geometry(BlockDriverState *bs, uint64_t *nb_sectors_ptr)
837 length = bdrv_getlength(bs);
841 length = length >> BDRV_SECTOR_BITS;
842 *nb_sectors_ptr = length;
846 uint8_t boot_ind; /* 0x80 - active */
847 uint8_t head; /* starting head */
848 uint8_t sector; /* starting sector */
849 uint8_t cyl; /* starting cylinder */
850 uint8_t sys_ind; /* What partition type */
851 uint8_t end_head; /* end head */
852 uint8_t end_sector; /* end sector */
853 uint8_t end_cyl; /* end cylinder */
854 uint32_t start_sect; /* starting sector counting from 0 */
855 uint32_t nr_sects; /* nr of sectors in partition */
856 } __attribute__((packed));
858 /* try to guess the disk logical geometry from the MSDOS partition table. Return 0 if OK, -1 if could not guess */
859 static int guess_disk_lchs(BlockDriverState *bs,
860 int *pcylinders, int *pheads, int *psectors)
863 int ret, i, heads, sectors, cylinders;
868 bdrv_get_geometry(bs, &nb_sectors);
870 ret = bdrv_read(bs, 0, buf, 1);
873 /* test msdos magic */
874 if (buf[510] != 0x55 || buf[511] != 0xaa)
876 for(i = 0; i < 4; i++) {
877 p = ((struct partition *)(buf + 0x1be)) + i;
878 nr_sects = le32_to_cpu(p->nr_sects);
879 if (nr_sects && p->end_head) {
880 /* We make the assumption that the partition terminates on
881 a cylinder boundary */
882 heads = p->end_head + 1;
883 sectors = p->end_sector & 63;
886 cylinders = nb_sectors / (heads * sectors);
887 if (cylinders < 1 || cylinders > 16383)
891 *pcylinders = cylinders;
893 printf("guessed geometry: LCHS=%d %d %d\n",
894 cylinders, heads, sectors);
902 void bdrv_guess_geometry(BlockDriverState *bs, int *pcyls, int *pheads, int *psecs)
904 int translation, lba_detected = 0;
905 int cylinders, heads, secs;
908 /* if a geometry hint is available, use it */
909 bdrv_get_geometry(bs, &nb_sectors);
910 bdrv_get_geometry_hint(bs, &cylinders, &heads, &secs);
911 translation = bdrv_get_translation_hint(bs);
912 if (cylinders != 0) {
917 if (guess_disk_lchs(bs, &cylinders, &heads, &secs) == 0) {
919 /* if heads > 16, it means that a BIOS LBA
920 translation was active, so the default
921 hardware geometry is OK */
923 goto default_geometry;
928 /* disable any translation to be in sync with
929 the logical geometry */
930 if (translation == BIOS_ATA_TRANSLATION_AUTO) {
931 bdrv_set_translation_hint(bs,
932 BIOS_ATA_TRANSLATION_NONE);
937 /* if no geometry, use a standard physical disk geometry */
938 cylinders = nb_sectors / (16 * 63);
940 if (cylinders > 16383)
942 else if (cylinders < 2)
947 if ((lba_detected == 1) && (translation == BIOS_ATA_TRANSLATION_AUTO)) {
948 if ((*pcyls * *pheads) <= 131072) {
949 bdrv_set_translation_hint(bs,
950 BIOS_ATA_TRANSLATION_LARGE);
952 bdrv_set_translation_hint(bs,
953 BIOS_ATA_TRANSLATION_LBA);
957 bdrv_set_geometry_hint(bs, *pcyls, *pheads, *psecs);
961 void bdrv_set_geometry_hint(BlockDriverState *bs,
962 int cyls, int heads, int secs)
969 void bdrv_set_type_hint(BlockDriverState *bs, int type)
972 bs->removable = ((type == BDRV_TYPE_CDROM ||
973 type == BDRV_TYPE_FLOPPY));
976 void bdrv_set_translation_hint(BlockDriverState *bs, int translation)
978 bs->translation = translation;
981 void bdrv_get_geometry_hint(BlockDriverState *bs,
982 int *pcyls, int *pheads, int *psecs)
989 int bdrv_get_type_hint(BlockDriverState *bs)
994 int bdrv_get_translation_hint(BlockDriverState *bs)
996 return bs->translation;
999 int bdrv_is_removable(BlockDriverState *bs)
1001 return bs->removable;
1004 int bdrv_is_read_only(BlockDriverState *bs)
1006 return bs->read_only;
1009 int bdrv_set_read_only(BlockDriverState *bs, int read_only)
1011 int ret = bs->read_only;
1012 bs->read_only = read_only;
1016 int bdrv_is_sg(BlockDriverState *bs)
1021 int bdrv_enable_write_cache(BlockDriverState *bs)
1023 return bs->enable_write_cache;
1026 /* XXX: no longer used */
1027 void bdrv_set_change_cb(BlockDriverState *bs,
1028 void (*change_cb)(void *opaque), void *opaque)
1030 bs->change_cb = change_cb;
1031 bs->change_opaque = opaque;
1034 int bdrv_is_encrypted(BlockDriverState *bs)
1036 if (bs->backing_hd && bs->backing_hd->encrypted)
1038 return bs->encrypted;
1041 int bdrv_key_required(BlockDriverState *bs)
1043 BlockDriverState *backing_hd = bs->backing_hd;
1045 if (backing_hd && backing_hd->encrypted && !backing_hd->valid_key)
1047 return (bs->encrypted && !bs->valid_key);
1050 int bdrv_set_key(BlockDriverState *bs, const char *key)
1053 if (bs->backing_hd && bs->backing_hd->encrypted) {
1054 ret = bdrv_set_key(bs->backing_hd, key);
1060 if (!bs->encrypted || !bs->drv || !bs->drv->bdrv_set_key)
1062 ret = bs->drv->bdrv_set_key(bs, key);
1065 } else if (!bs->valid_key) {
1067 /* call the change callback now, we skipped it on open */
1068 bs->media_changed = 1;
1070 bs->change_cb(bs->change_opaque);
1075 void bdrv_get_format(BlockDriverState *bs, char *buf, int buf_size)
1080 pstrcpy(buf, buf_size, bs->drv->format_name);
1084 void bdrv_iterate_format(void (*it)(void *opaque, const char *name),
1089 for (drv = first_drv; drv != NULL; drv = drv->next) {
1090 it(opaque, drv->format_name);
1094 BlockDriverState *bdrv_find(const char *name)
1096 BlockDriverState *bs;
1098 for (bs = bdrv_first; bs != NULL; bs = bs->next) {
1099 if (!strcmp(name, bs->device_name))
1105 void bdrv_iterate(void (*it)(void *opaque, BlockDriverState *bs), void *opaque)
1107 BlockDriverState *bs;
1109 for (bs = bdrv_first; bs != NULL; bs = bs->next) {
1114 const char *bdrv_get_device_name(BlockDriverState *bs)
1116 return bs->device_name;
1119 void bdrv_flush(BlockDriverState *bs)
1121 if (bs->drv && bs->drv->bdrv_flush)
1122 bs->drv->bdrv_flush(bs);
1125 void bdrv_flush_all(void)
1127 BlockDriverState *bs;
1129 for (bs = bdrv_first; bs != NULL; bs = bs->next)
1130 if (bs->drv && !bdrv_is_read_only(bs) &&
1131 (!bdrv_is_removable(bs) || bdrv_is_inserted(bs)))
1136 * Returns true iff the specified sector is present in the disk image. Drivers
1137 * not implementing the functionality are assumed to not support backing files,
1138 * hence all their sectors are reported as allocated.
1140 * 'pnum' is set to the number of sectors (including and immediately following
1141 * the specified sector) that are known to be in the same
1142 * allocated/unallocated state.
1144 * 'nb_sectors' is the max value 'pnum' should be set to.
1146 int bdrv_is_allocated(BlockDriverState *bs, int64_t sector_num, int nb_sectors,
1150 if (!bs->drv->bdrv_is_allocated) {
1151 if (sector_num >= bs->total_sectors) {
1155 n = bs->total_sectors - sector_num;
1156 *pnum = (n < nb_sectors) ? (n) : (nb_sectors);
1159 return bs->drv->bdrv_is_allocated(bs, sector_num, nb_sectors, pnum);
1162 static void bdrv_print_dict(QObject *obj, void *opaque)
1165 Monitor *mon = opaque;
1167 bs_dict = qobject_to_qdict(obj);
1169 monitor_printf(mon, "%s: type=%s removable=%d",
1170 qdict_get_str(bs_dict, "device"),
1171 qdict_get_str(bs_dict, "type"),
1172 qdict_get_bool(bs_dict, "removable"));
1174 if (qdict_get_bool(bs_dict, "removable")) {
1175 monitor_printf(mon, " locked=%d", qdict_get_bool(bs_dict, "locked"));
1178 if (qdict_haskey(bs_dict, "inserted")) {
1179 QDict *qdict = qobject_to_qdict(qdict_get(bs_dict, "inserted"));
1181 monitor_printf(mon, " file=");
1182 monitor_print_filename(mon, qdict_get_str(qdict, "file"));
1183 if (qdict_haskey(qdict, "backing_file")) {
1184 monitor_printf(mon, " backing_file=");
1185 monitor_print_filename(mon, qdict_get_str(qdict, "backing_file"));
1187 monitor_printf(mon, " ro=%d drv=%s encrypted=%d",
1188 qdict_get_bool(qdict, "ro"),
1189 qdict_get_str(qdict, "drv"),
1190 qdict_get_bool(qdict, "encrypted"));
1192 monitor_printf(mon, " [not inserted]");
1195 monitor_printf(mon, "\n");
1198 void bdrv_info_print(Monitor *mon, const QObject *data)
1200 qlist_iter(qobject_to_qlist(data), bdrv_print_dict, mon);
1204 * bdrv_info(): Block devices information
1206 * Each block device information is stored in a QDict and the
1207 * returned QObject is a QList of all devices.
1209 * The QDict contains the following:
1211 * - "device": device name
1212 * - "type": device type
1213 * - "removable": true if the device is removable, false otherwise
1214 * - "locked": true if the device is locked, false otherwise
1215 * - "inserted": only present if the device is inserted, it is a QDict
1216 * containing the following:
1217 * - "file": device file name
1218 * - "ro": true if read-only, false otherwise
1219 * - "drv": driver format name
1220 * - "backing_file": backing file name if one is used
1221 * - "encrypted": true if encrypted, false otherwise
1225 * [ { "device": "ide0-hd0", "type": "hd", "removable": false, "locked": false,
1226 * "inserted": { "file": "/tmp/foobar", "ro": false, "drv": "qcow2" } },
1227 * { "device": "floppy0", "type": "floppy", "removable": true,
1228 * "locked": false } ]
1230 void bdrv_info(Monitor *mon, QObject **ret_data)
1233 BlockDriverState *bs;
1235 bs_list = qlist_new();
1237 for (bs = bdrv_first; bs != NULL; bs = bs->next) {
1239 const char *type = "unknown";
1245 case BDRV_TYPE_CDROM:
1248 case BDRV_TYPE_FLOPPY:
1253 bs_obj = qobject_from_jsonf("{ 'device': %s, 'type': %s, "
1254 "'removable': %i, 'locked': %i }",
1255 bs->device_name, type, bs->removable,
1257 assert(bs_obj != NULL);
1261 QDict *bs_dict = qobject_to_qdict(bs_obj);
1263 obj = qobject_from_jsonf("{ 'file': %s, 'ro': %i, 'drv': %s, "
1264 "'encrypted': %i }",
1265 bs->filename, bs->read_only,
1266 bs->drv->format_name,
1267 bdrv_is_encrypted(bs));
1268 assert(obj != NULL);
1269 if (bs->backing_file[0] != '\0') {
1270 QDict *qdict = qobject_to_qdict(obj);
1271 qdict_put(qdict, "backing_file",
1272 qstring_from_str(bs->backing_file));
1275 qdict_put_obj(bs_dict, "inserted", obj);
1277 qlist_append_obj(bs_list, bs_obj);
1280 *ret_data = QOBJECT(bs_list);
1283 static void bdrv_stats_iter(QObject *data, void *opaque)
1286 Monitor *mon = opaque;
1288 qdict = qobject_to_qdict(data);
1289 monitor_printf(mon, "%s:", qdict_get_str(qdict, "device"));
1291 qdict = qobject_to_qdict(qdict_get(qdict, "stats"));
1292 monitor_printf(mon, " rd_bytes=%" PRId64
1293 " wr_bytes=%" PRId64
1294 " rd_operations=%" PRId64
1295 " wr_operations=%" PRId64
1297 qdict_get_int(qdict, "rd_bytes"),
1298 qdict_get_int(qdict, "wr_bytes"),
1299 qdict_get_int(qdict, "rd_operations"),
1300 qdict_get_int(qdict, "wr_operations"));
1303 void bdrv_stats_print(Monitor *mon, const QObject *data)
1305 qlist_iter(qobject_to_qlist(data), bdrv_stats_iter, mon);
1309 * bdrv_info_stats(): show block device statistics
1311 * Each device statistic information is stored in a QDict and
1312 * the returned QObject is a QList of all devices.
1314 * The QDict contains the following:
1316 * - "device": device name
1317 * - "stats": A QDict with the statistics information, it contains:
1318 * - "rd_bytes": bytes read
1319 * - "wr_bytes": bytes written
1320 * - "rd_operations": read operations
1321 * - "wr_operations": write operations
1325 * [ { "device": "ide0-hd0",
1326 * "stats": { "rd_bytes": 512,
1328 * "rd_operations": 1,
1329 * "wr_operations": 0 } },
1330 * { "device": "ide1-cd0",
1331 * "stats": { "rd_bytes": 0,
1333 * "rd_operations": 0,
1334 * "wr_operations": 0 } } ]
1336 void bdrv_info_stats(Monitor *mon, QObject **ret_data)
1340 BlockDriverState *bs;
1342 devices = qlist_new();
1344 for (bs = bdrv_first; bs != NULL; bs = bs->next) {
1345 obj = qobject_from_jsonf("{ 'device': %s, 'stats': {"
1346 "'rd_bytes': %" PRId64 ","
1347 "'wr_bytes': %" PRId64 ","
1348 "'rd_operations': %" PRId64 ","
1349 "'wr_operations': %" PRId64
1352 bs->rd_bytes, bs->wr_bytes,
1353 bs->rd_ops, bs->wr_ops);
1354 assert(obj != NULL);
1355 qlist_append_obj(devices, obj);
1358 *ret_data = QOBJECT(devices);
1361 const char *bdrv_get_encrypted_filename(BlockDriverState *bs)
1363 if (bs->backing_hd && bs->backing_hd->encrypted)
1364 return bs->backing_file;
1365 else if (bs->encrypted)
1366 return bs->filename;
1371 void bdrv_get_backing_filename(BlockDriverState *bs,
1372 char *filename, int filename_size)
1374 if (!bs->backing_file) {
1375 pstrcpy(filename, filename_size, "");
1377 pstrcpy(filename, filename_size, bs->backing_file);
1381 int bdrv_write_compressed(BlockDriverState *bs, int64_t sector_num,
1382 const uint8_t *buf, int nb_sectors)
1384 BlockDriver *drv = bs->drv;
1387 if (!drv->bdrv_write_compressed)
1389 if (bdrv_check_request(bs, sector_num, nb_sectors))
1392 if (bs->dirty_bitmap) {
1393 set_dirty_bitmap(bs, sector_num, nb_sectors, 1);
1396 return drv->bdrv_write_compressed(bs, sector_num, buf, nb_sectors);
1399 int bdrv_get_info(BlockDriverState *bs, BlockDriverInfo *bdi)
1401 BlockDriver *drv = bs->drv;
1404 if (!drv->bdrv_get_info)
1406 memset(bdi, 0, sizeof(*bdi));
1407 return drv->bdrv_get_info(bs, bdi);
1410 int bdrv_save_vmstate(BlockDriverState *bs, const uint8_t *buf,
1411 int64_t pos, int size)
1413 BlockDriver *drv = bs->drv;
1416 if (!drv->bdrv_save_vmstate)
1418 return drv->bdrv_save_vmstate(bs, buf, pos, size);
1421 int bdrv_load_vmstate(BlockDriverState *bs, uint8_t *buf,
1422 int64_t pos, int size)
1424 BlockDriver *drv = bs->drv;
1427 if (!drv->bdrv_load_vmstate)
1429 return drv->bdrv_load_vmstate(bs, buf, pos, size);
1432 /**************************************************************/
1433 /* handling of snapshots */
1435 int bdrv_snapshot_create(BlockDriverState *bs,
1436 QEMUSnapshotInfo *sn_info)
1438 BlockDriver *drv = bs->drv;
1441 if (!drv->bdrv_snapshot_create)
1443 return drv->bdrv_snapshot_create(bs, sn_info);
1446 int bdrv_snapshot_goto(BlockDriverState *bs,
1447 const char *snapshot_id)
1449 BlockDriver *drv = bs->drv;
1452 if (!drv->bdrv_snapshot_goto)
1454 return drv->bdrv_snapshot_goto(bs, snapshot_id);
1457 int bdrv_snapshot_delete(BlockDriverState *bs, const char *snapshot_id)
1459 BlockDriver *drv = bs->drv;
1462 if (!drv->bdrv_snapshot_delete)
1464 return drv->bdrv_snapshot_delete(bs, snapshot_id);
1467 int bdrv_snapshot_list(BlockDriverState *bs,
1468 QEMUSnapshotInfo **psn_info)
1470 BlockDriver *drv = bs->drv;
1473 if (!drv->bdrv_snapshot_list)
1475 return drv->bdrv_snapshot_list(bs, psn_info);
1478 #define NB_SUFFIXES 4
1480 char *get_human_readable_size(char *buf, int buf_size, int64_t size)
1482 static const char suffixes[NB_SUFFIXES] = "KMGT";
1487 snprintf(buf, buf_size, "%" PRId64, size);
1490 for(i = 0; i < NB_SUFFIXES; i++) {
1491 if (size < (10 * base)) {
1492 snprintf(buf, buf_size, "%0.1f%c",
1493 (double)size / base,
1496 } else if (size < (1000 * base) || i == (NB_SUFFIXES - 1)) {
1497 snprintf(buf, buf_size, "%" PRId64 "%c",
1498 ((size + (base >> 1)) / base),
1508 char *bdrv_snapshot_dump(char *buf, int buf_size, QEMUSnapshotInfo *sn)
1510 char buf1[128], date_buf[128], clock_buf[128];
1520 snprintf(buf, buf_size,
1521 "%-10s%-20s%7s%20s%15s",
1522 "ID", "TAG", "VM SIZE", "DATE", "VM CLOCK");
1526 ptm = localtime(&ti);
1527 strftime(date_buf, sizeof(date_buf),
1528 "%Y-%m-%d %H:%M:%S", ptm);
1530 localtime_r(&ti, &tm);
1531 strftime(date_buf, sizeof(date_buf),
1532 "%Y-%m-%d %H:%M:%S", &tm);
1534 secs = sn->vm_clock_nsec / 1000000000;
1535 snprintf(clock_buf, sizeof(clock_buf),
1536 "%02d:%02d:%02d.%03d",
1538 (int)((secs / 60) % 60),
1540 (int)((sn->vm_clock_nsec / 1000000) % 1000));
1541 snprintf(buf, buf_size,
1542 "%-10s%-20s%7s%20s%15s",
1543 sn->id_str, sn->name,
1544 get_human_readable_size(buf1, sizeof(buf1), sn->vm_state_size),
1552 /**************************************************************/
1555 BlockDriverAIOCB *bdrv_aio_readv(BlockDriverState *bs, int64_t sector_num,
1556 QEMUIOVector *qiov, int nb_sectors,
1557 BlockDriverCompletionFunc *cb, void *opaque)
1559 BlockDriver *drv = bs->drv;
1560 BlockDriverAIOCB *ret;
1564 if (bdrv_check_request(bs, sector_num, nb_sectors))
1567 ret = drv->bdrv_aio_readv(bs, sector_num, qiov, nb_sectors,
1571 /* Update stats even though technically transfer has not happened. */
1572 bs->rd_bytes += (unsigned) nb_sectors * BDRV_SECTOR_SIZE;
1579 BlockDriverAIOCB *bdrv_aio_writev(BlockDriverState *bs, int64_t sector_num,
1580 QEMUIOVector *qiov, int nb_sectors,
1581 BlockDriverCompletionFunc *cb, void *opaque)
1583 BlockDriver *drv = bs->drv;
1584 BlockDriverAIOCB *ret;
1590 if (bdrv_check_request(bs, sector_num, nb_sectors))
1593 if (bs->dirty_bitmap) {
1594 set_dirty_bitmap(bs, sector_num, nb_sectors, 1);
1597 ret = drv->bdrv_aio_writev(bs, sector_num, qiov, nb_sectors,
1601 /* Update stats even though technically transfer has not happened. */
1602 bs->wr_bytes += (unsigned) nb_sectors * BDRV_SECTOR_SIZE;
1610 typedef struct MultiwriteCB {
1615 BlockDriverCompletionFunc *cb;
1617 QEMUIOVector *free_qiov;
1622 static void multiwrite_user_cb(MultiwriteCB *mcb)
1626 for (i = 0; i < mcb->num_callbacks; i++) {
1627 mcb->callbacks[i].cb(mcb->callbacks[i].opaque, mcb->error);
1628 qemu_free(mcb->callbacks[i].free_qiov);
1629 qemu_free(mcb->callbacks[i].free_buf);
1633 static void multiwrite_cb(void *opaque, int ret)
1635 MultiwriteCB *mcb = opaque;
1639 multiwrite_user_cb(mcb);
1642 mcb->num_requests--;
1643 if (mcb->num_requests == 0) {
1644 if (mcb->error == 0) {
1645 multiwrite_user_cb(mcb);
1651 static int multiwrite_req_compare(const void *a, const void *b)
1653 return (((BlockRequest*) a)->sector - ((BlockRequest*) b)->sector);
1657 * Takes a bunch of requests and tries to merge them. Returns the number of
1658 * requests that remain after merging.
1660 static int multiwrite_merge(BlockDriverState *bs, BlockRequest *reqs,
1661 int num_reqs, MultiwriteCB *mcb)
1665 // Sort requests by start sector
1666 qsort(reqs, num_reqs, sizeof(*reqs), &multiwrite_req_compare);
1668 // Check if adjacent requests touch the same clusters. If so, combine them,
1669 // filling up gaps with zero sectors.
1671 for (i = 1; i < num_reqs; i++) {
1673 int64_t oldreq_last = reqs[outidx].sector + reqs[outidx].nb_sectors;
1675 // This handles the cases that are valid for all block drivers, namely
1676 // exactly sequential writes and overlapping writes.
1677 if (reqs[i].sector <= oldreq_last) {
1681 // The block driver may decide that it makes sense to combine requests
1682 // even if there is a gap of some sectors between them. In this case,
1683 // the gap is filled with zeros (therefore only applicable for yet
1684 // unused space in format like qcow2).
1685 if (!merge && bs->drv->bdrv_merge_requests) {
1686 merge = bs->drv->bdrv_merge_requests(bs, &reqs[outidx], &reqs[i]);
1691 QEMUIOVector *qiov = qemu_mallocz(sizeof(*qiov));
1692 qemu_iovec_init(qiov,
1693 reqs[outidx].qiov->niov + reqs[i].qiov->niov + 1);
1695 // Add the first request to the merged one. If the requests are
1696 // overlapping, drop the last sectors of the first request.
1697 size = (reqs[i].sector - reqs[outidx].sector) << 9;
1698 qemu_iovec_concat(qiov, reqs[outidx].qiov, size);
1700 // We might need to add some zeros between the two requests
1701 if (reqs[i].sector > oldreq_last) {
1702 size_t zero_bytes = (reqs[i].sector - oldreq_last) << 9;
1703 uint8_t *buf = qemu_blockalign(bs, zero_bytes);
1704 memset(buf, 0, zero_bytes);
1705 qemu_iovec_add(qiov, buf, zero_bytes);
1706 mcb->callbacks[i].free_buf = buf;
1709 // Add the second request
1710 qemu_iovec_concat(qiov, reqs[i].qiov, reqs[i].qiov->size);
1712 reqs[outidx].nb_sectors += reqs[i].nb_sectors;
1713 reqs[outidx].qiov = qiov;
1715 mcb->callbacks[i].free_qiov = reqs[outidx].qiov;
1718 reqs[outidx].sector = reqs[i].sector;
1719 reqs[outidx].nb_sectors = reqs[i].nb_sectors;
1720 reqs[outidx].qiov = reqs[i].qiov;
1728 * Submit multiple AIO write requests at once.
1730 * On success, the function returns 0 and all requests in the reqs array have
1731 * been submitted. In error case this function returns -1, and any of the
1732 * requests may or may not be submitted yet. In particular, this means that the
1733 * callback will be called for some of the requests, for others it won't. The
1734 * caller must check the error field of the BlockRequest to wait for the right
1735 * callbacks (if error != 0, no callback will be called).
1737 * The implementation may modify the contents of the reqs array, e.g. to merge
1738 * requests. However, the fields opaque and error are left unmodified as they
1739 * are used to signal failure for a single request to the caller.
1741 int bdrv_aio_multiwrite(BlockDriverState *bs, BlockRequest *reqs, int num_reqs)
1743 BlockDriverAIOCB *acb;
1747 if (num_reqs == 0) {
1751 // Create MultiwriteCB structure
1752 mcb = qemu_mallocz(sizeof(*mcb) + num_reqs * sizeof(*mcb->callbacks));
1753 mcb->num_requests = 0;
1754 mcb->num_callbacks = num_reqs;
1756 for (i = 0; i < num_reqs; i++) {
1757 mcb->callbacks[i].cb = reqs[i].cb;
1758 mcb->callbacks[i].opaque = reqs[i].opaque;
1761 // Check for mergable requests
1762 num_reqs = multiwrite_merge(bs, reqs, num_reqs, mcb);
1764 // Run the aio requests
1765 for (i = 0; i < num_reqs; i++) {
1766 acb = bdrv_aio_writev(bs, reqs[i].sector, reqs[i].qiov,
1767 reqs[i].nb_sectors, multiwrite_cb, mcb);
1770 // We can only fail the whole thing if no request has been
1771 // submitted yet. Otherwise we'll wait for the submitted AIOs to
1772 // complete and report the error in the callback.
1773 if (mcb->num_requests == 0) {
1774 reqs[i].error = EIO;
1781 mcb->num_requests++;
1792 BlockDriverAIOCB *bdrv_aio_flush(BlockDriverState *bs,
1793 BlockDriverCompletionFunc *cb, void *opaque)
1795 BlockDriver *drv = bs->drv;
1799 return drv->bdrv_aio_flush(bs, cb, opaque);
1802 void bdrv_aio_cancel(BlockDriverAIOCB *acb)
1804 acb->pool->cancel(acb);
1808 /**************************************************************/
1809 /* async block device emulation */
1811 typedef struct BlockDriverAIOCBSync {
1812 BlockDriverAIOCB common;
1815 /* vector translation state */
1819 } BlockDriverAIOCBSync;
1821 static void bdrv_aio_cancel_em(BlockDriverAIOCB *blockacb)
1823 BlockDriverAIOCBSync *acb = (BlockDriverAIOCBSync *)blockacb;
1824 qemu_bh_delete(acb->bh);
1826 qemu_aio_release(acb);
1829 static AIOPool bdrv_em_aio_pool = {
1830 .aiocb_size = sizeof(BlockDriverAIOCBSync),
1831 .cancel = bdrv_aio_cancel_em,
1834 static void bdrv_aio_bh_cb(void *opaque)
1836 BlockDriverAIOCBSync *acb = opaque;
1839 qemu_iovec_from_buffer(acb->qiov, acb->bounce, acb->qiov->size);
1840 qemu_vfree(acb->bounce);
1841 acb->common.cb(acb->common.opaque, acb->ret);
1842 qemu_bh_delete(acb->bh);
1844 qemu_aio_release(acb);
1847 static BlockDriverAIOCB *bdrv_aio_rw_vector(BlockDriverState *bs,
1851 BlockDriverCompletionFunc *cb,
1856 BlockDriverAIOCBSync *acb;
1858 acb = qemu_aio_get(&bdrv_em_aio_pool, bs, cb, opaque);
1859 acb->is_write = is_write;
1861 acb->bounce = qemu_blockalign(bs, qiov->size);
1864 acb->bh = qemu_bh_new(bdrv_aio_bh_cb, acb);
1867 qemu_iovec_to_buffer(acb->qiov, acb->bounce);
1868 acb->ret = bdrv_write(bs, sector_num, acb->bounce, nb_sectors);
1870 acb->ret = bdrv_read(bs, sector_num, acb->bounce, nb_sectors);
1873 qemu_bh_schedule(acb->bh);
1875 return &acb->common;
1878 static BlockDriverAIOCB *bdrv_aio_readv_em(BlockDriverState *bs,
1879 int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
1880 BlockDriverCompletionFunc *cb, void *opaque)
1882 return bdrv_aio_rw_vector(bs, sector_num, qiov, nb_sectors, cb, opaque, 0);
1885 static BlockDriverAIOCB *bdrv_aio_writev_em(BlockDriverState *bs,
1886 int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
1887 BlockDriverCompletionFunc *cb, void *opaque)
1889 return bdrv_aio_rw_vector(bs, sector_num, qiov, nb_sectors, cb, opaque, 1);
1892 static BlockDriverAIOCB *bdrv_aio_flush_em(BlockDriverState *bs,
1893 BlockDriverCompletionFunc *cb, void *opaque)
1895 BlockDriverAIOCBSync *acb;
1897 acb = qemu_aio_get(&bdrv_em_aio_pool, bs, cb, opaque);
1898 acb->is_write = 1; /* don't bounce in the completion hadler */
1904 acb->bh = qemu_bh_new(bdrv_aio_bh_cb, acb);
1907 qemu_bh_schedule(acb->bh);
1908 return &acb->common;
1911 /**************************************************************/
1912 /* sync block device emulation */
1914 static void bdrv_rw_em_cb(void *opaque, int ret)
1916 *(int *)opaque = ret;
1919 #define NOT_DONE 0x7fffffff
1921 static int bdrv_read_em(BlockDriverState *bs, int64_t sector_num,
1922 uint8_t *buf, int nb_sectors)
1925 BlockDriverAIOCB *acb;
1929 async_context_push();
1931 async_ret = NOT_DONE;
1932 iov.iov_base = (void *)buf;
1933 iov.iov_len = nb_sectors * 512;
1934 qemu_iovec_init_external(&qiov, &iov, 1);
1935 acb = bdrv_aio_readv(bs, sector_num, &qiov, nb_sectors,
1936 bdrv_rw_em_cb, &async_ret);
1942 while (async_ret == NOT_DONE) {
1948 async_context_pop();
1952 static int bdrv_write_em(BlockDriverState *bs, int64_t sector_num,
1953 const uint8_t *buf, int nb_sectors)
1956 BlockDriverAIOCB *acb;
1960 async_context_push();
1962 async_ret = NOT_DONE;
1963 iov.iov_base = (void *)buf;
1964 iov.iov_len = nb_sectors * 512;
1965 qemu_iovec_init_external(&qiov, &iov, 1);
1966 acb = bdrv_aio_writev(bs, sector_num, &qiov, nb_sectors,
1967 bdrv_rw_em_cb, &async_ret);
1972 while (async_ret == NOT_DONE) {
1977 async_context_pop();
1981 void bdrv_init(void)
1983 module_call_init(MODULE_INIT_BLOCK);
1986 void bdrv_init_with_whitelist(void)
1988 use_bdrv_whitelist = 1;
1992 void *qemu_aio_get(AIOPool *pool, BlockDriverState *bs,
1993 BlockDriverCompletionFunc *cb, void *opaque)
1995 BlockDriverAIOCB *acb;
1997 if (pool->free_aiocb) {
1998 acb = pool->free_aiocb;
1999 pool->free_aiocb = acb->next;
2001 acb = qemu_mallocz(pool->aiocb_size);
2006 acb->opaque = opaque;
2010 void qemu_aio_release(void *p)
2012 BlockDriverAIOCB *acb = (BlockDriverAIOCB *)p;
2013 AIOPool *pool = acb->pool;
2014 acb->next = pool->free_aiocb;
2015 pool->free_aiocb = acb;
2018 /**************************************************************/
2019 /* removable device support */
2022 * Return TRUE if the media is present
2024 int bdrv_is_inserted(BlockDriverState *bs)
2026 BlockDriver *drv = bs->drv;
2030 if (!drv->bdrv_is_inserted)
2032 ret = drv->bdrv_is_inserted(bs);
2037 * Return TRUE if the media changed since the last call to this
2038 * function. It is currently only used for floppy disks
2040 int bdrv_media_changed(BlockDriverState *bs)
2042 BlockDriver *drv = bs->drv;
2045 if (!drv || !drv->bdrv_media_changed)
2048 ret = drv->bdrv_media_changed(bs);
2049 if (ret == -ENOTSUP)
2050 ret = bs->media_changed;
2051 bs->media_changed = 0;
2056 * If eject_flag is TRUE, eject the media. Otherwise, close the tray
2058 int bdrv_eject(BlockDriverState *bs, int eject_flag)
2060 BlockDriver *drv = bs->drv;
2067 if (!drv || !drv->bdrv_eject) {
2070 ret = drv->bdrv_eject(bs, eject_flag);
2072 if (ret == -ENOTSUP) {
2081 int bdrv_is_locked(BlockDriverState *bs)
2087 * Lock or unlock the media (if it is locked, the user won't be able
2088 * to eject it manually).
2090 void bdrv_set_locked(BlockDriverState *bs, int locked)
2092 BlockDriver *drv = bs->drv;
2094 bs->locked = locked;
2095 if (drv && drv->bdrv_set_locked) {
2096 drv->bdrv_set_locked(bs, locked);
2100 /* needed for generic scsi interface */
2102 int bdrv_ioctl(BlockDriverState *bs, unsigned long int req, void *buf)
2104 BlockDriver *drv = bs->drv;
2106 if (drv && drv->bdrv_ioctl)
2107 return drv->bdrv_ioctl(bs, req, buf);
2111 BlockDriverAIOCB *bdrv_aio_ioctl(BlockDriverState *bs,
2112 unsigned long int req, void *buf,
2113 BlockDriverCompletionFunc *cb, void *opaque)
2115 BlockDriver *drv = bs->drv;
2117 if (drv && drv->bdrv_aio_ioctl)
2118 return drv->bdrv_aio_ioctl(bs, req, buf, cb, opaque);
2124 void *qemu_blockalign(BlockDriverState *bs, size_t size)
2126 return qemu_memalign((bs && bs->buffer_alignment) ? bs->buffer_alignment : 512, size);
2129 void bdrv_set_dirty_tracking(BlockDriverState *bs, int enable)
2131 int64_t bitmap_size;
2134 if (!bs->dirty_bitmap) {
2135 bitmap_size = (bdrv_getlength(bs) >> BDRV_SECTOR_BITS) +
2136 BDRV_SECTORS_PER_DIRTY_CHUNK * 8 - 1;
2137 bitmap_size /= BDRV_SECTORS_PER_DIRTY_CHUNK * 8;
2139 bs->dirty_bitmap = qemu_mallocz(bitmap_size);
2142 if (bs->dirty_bitmap) {
2143 qemu_free(bs->dirty_bitmap);
2144 bs->dirty_bitmap = NULL;
2149 int bdrv_get_dirty(BlockDriverState *bs, int64_t sector)
2151 int64_t chunk = sector / (int64_t)BDRV_SECTORS_PER_DIRTY_CHUNK;
2153 if (bs->dirty_bitmap &&
2154 (sector << BDRV_SECTOR_BITS) < bdrv_getlength(bs)) {
2155 return bs->dirty_bitmap[chunk / (sizeof(unsigned long) * 8)] &
2156 (1 << (chunk % (sizeof(unsigned long) * 8)));
2162 void bdrv_reset_dirty(BlockDriverState *bs, int64_t cur_sector,
2165 set_dirty_bitmap(bs, cur_sector, nr_sectors, 0);