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"
26 /* include native header before sys-queue.h */
27 #include <sys/queue.h>
30 #include "qemu-common.h"
32 #include "block_int.h"
36 #include <sys/types.h>
38 #include <sys/ioctl.h>
49 #define SECTOR_SIZE (1 << SECTOR_BITS)
51 static BlockDriverAIOCB *bdrv_aio_readv_em(BlockDriverState *bs,
52 int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
53 BlockDriverCompletionFunc *cb, void *opaque);
54 static BlockDriverAIOCB *bdrv_aio_writev_em(BlockDriverState *bs,
55 int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
56 BlockDriverCompletionFunc *cb, void *opaque);
57 static BlockDriverAIOCB *bdrv_aio_flush_em(BlockDriverState *bs,
58 BlockDriverCompletionFunc *cb, void *opaque);
59 static int bdrv_read_em(BlockDriverState *bs, int64_t sector_num,
60 uint8_t *buf, int nb_sectors);
61 static int bdrv_write_em(BlockDriverState *bs, int64_t sector_num,
62 const uint8_t *buf, int nb_sectors);
64 BlockDriverState *bdrv_first;
66 static BlockDriver *first_drv;
68 int path_is_absolute(const char *path)
72 /* specific case for names like: "\\.\d:" */
73 if (*path == '/' || *path == '\\')
76 p = strchr(path, ':');
82 return (*p == '/' || *p == '\\');
88 /* if filename is absolute, just copy it to dest. Otherwise, build a
89 path to it by considering it is relative to base_path. URL are
91 void path_combine(char *dest, int dest_size,
92 const char *base_path,
100 if (path_is_absolute(filename)) {
101 pstrcpy(dest, dest_size, filename);
103 p = strchr(base_path, ':');
108 p1 = strrchr(base_path, '/');
112 p2 = strrchr(base_path, '\\');
124 if (len > dest_size - 1)
126 memcpy(dest, base_path, len);
128 pstrcat(dest, dest_size, filename);
132 void bdrv_register(BlockDriver *bdrv)
134 if (!bdrv->bdrv_aio_readv) {
135 /* add AIO emulation layer */
136 bdrv->bdrv_aio_readv = bdrv_aio_readv_em;
137 bdrv->bdrv_aio_writev = bdrv_aio_writev_em;
138 } else if (!bdrv->bdrv_read) {
139 /* add synchronous IO emulation layer */
140 bdrv->bdrv_read = bdrv_read_em;
141 bdrv->bdrv_write = bdrv_write_em;
144 if (!bdrv->bdrv_aio_flush)
145 bdrv->bdrv_aio_flush = bdrv_aio_flush_em;
147 bdrv->next = first_drv;
151 /* create a new block device (by default it is empty) */
152 BlockDriverState *bdrv_new(const char *device_name)
154 BlockDriverState **pbs, *bs;
156 bs = qemu_mallocz(sizeof(BlockDriverState));
157 pstrcpy(bs->device_name, sizeof(bs->device_name), device_name);
158 if (device_name[0] != '\0') {
159 /* insert at the end */
168 BlockDriver *bdrv_find_format(const char *format_name)
171 for(drv1 = first_drv; drv1 != NULL; drv1 = drv1->next) {
172 if (!strcmp(drv1->format_name, format_name))
178 int bdrv_create(BlockDriver *drv, const char* filename,
179 QEMUOptionParameter *options)
181 if (!drv->bdrv_create)
184 return drv->bdrv_create(filename, options);
188 void get_tmp_filename(char *filename, int size)
190 char temp_dir[MAX_PATH];
192 GetTempPath(MAX_PATH, temp_dir);
193 GetTempFileName(temp_dir, "qem", 0, filename);
196 void get_tmp_filename(char *filename, int size)
200 /* XXX: race condition possible */
201 tmpdir = getenv("TMPDIR");
204 snprintf(filename, size, "%s/vl.XXXXXX", tmpdir);
205 fd = mkstemp(filename);
211 static int is_windows_drive_prefix(const char *filename)
213 return (((filename[0] >= 'a' && filename[0] <= 'z') ||
214 (filename[0] >= 'A' && filename[0] <= 'Z')) &&
218 int is_windows_drive(const char *filename)
220 if (is_windows_drive_prefix(filename) &&
223 if (strstart(filename, "\\\\.\\", NULL) ||
224 strstart(filename, "//./", NULL))
230 static BlockDriver *find_protocol(const char *filename)
238 if (is_windows_drive(filename) ||
239 is_windows_drive_prefix(filename))
240 return bdrv_find_format("raw");
242 p = strchr(filename, ':');
244 return bdrv_find_format("raw");
246 if (len > sizeof(protocol) - 1)
247 len = sizeof(protocol) - 1;
248 memcpy(protocol, filename, len);
249 protocol[len] = '\0';
250 for(drv1 = first_drv; drv1 != NULL; drv1 = drv1->next) {
251 if (drv1->protocol_name &&
252 !strcmp(drv1->protocol_name, protocol))
259 * Detect host devices. By convention, /dev/cdrom[N] is always
260 * recognized as a host CDROM.
262 static BlockDriver *find_hdev_driver(const char *filename)
264 int score_max = 0, score;
265 BlockDriver *drv = NULL, *d;
267 for (d = first_drv; d; d = d->next) {
268 if (d->bdrv_probe_device) {
269 score = d->bdrv_probe_device(filename);
270 if (score > score_max) {
280 static BlockDriver *find_image_format(const char *filename)
282 int ret, score, score_max;
283 BlockDriver *drv1, *drv;
285 BlockDriverState *bs;
287 drv = find_protocol(filename);
288 /* no need to test disk image formats for vvfat */
289 if (drv && strcmp(drv->format_name, "vvfat") == 0)
292 ret = bdrv_file_open(&bs, filename, BDRV_O_RDONLY);
295 ret = bdrv_pread(bs, 0, buf, sizeof(buf));
302 for(drv1 = first_drv; drv1 != NULL; drv1 = drv1->next) {
303 if (drv1->bdrv_probe) {
304 score = drv1->bdrv_probe(buf, ret, filename);
305 if (score > score_max) {
314 int bdrv_file_open(BlockDriverState **pbs, const char *filename, int flags)
316 BlockDriverState *bs;
320 ret = bdrv_open2(bs, filename, flags | BDRV_O_FILE, NULL);
330 int bdrv_open(BlockDriverState *bs, const char *filename, int flags)
332 return bdrv_open2(bs, filename, flags, NULL);
335 int bdrv_open2(BlockDriverState *bs, const char *filename, int flags,
339 char tmp_filename[PATH_MAX];
340 char backing_filename[PATH_MAX];
343 bs->is_temporary = 0;
346 /* buffer_alignment defaulted to 512, drivers can change this value */
347 bs->buffer_alignment = 512;
349 if (flags & BDRV_O_SNAPSHOT) {
350 BlockDriverState *bs1;
353 BlockDriver *bdrv_qcow2;
354 QEMUOptionParameter *options;
356 /* if snapshot, we create a temporary backing file and open it
357 instead of opening 'filename' directly */
359 /* if there is a backing file, use it */
361 ret = bdrv_open2(bs1, filename, 0, drv);
366 total_size = bdrv_getlength(bs1) >> SECTOR_BITS;
368 if (bs1->drv && bs1->drv->protocol_name)
373 get_tmp_filename(tmp_filename, sizeof(tmp_filename));
375 /* Real path is meaningless for protocols */
377 snprintf(backing_filename, sizeof(backing_filename),
380 realpath(filename, backing_filename);
382 bdrv_qcow2 = bdrv_find_format("qcow2");
383 options = parse_option_parameters("", bdrv_qcow2->create_options, NULL);
385 set_option_parameter_int(options, BLOCK_OPT_SIZE, total_size * 512);
386 set_option_parameter(options, BLOCK_OPT_BACKING_FILE, backing_filename);
388 set_option_parameter(options, BLOCK_OPT_BACKING_FMT,
392 ret = bdrv_create(bdrv_qcow2, tmp_filename, options);
397 filename = tmp_filename;
399 bs->is_temporary = 1;
402 pstrcpy(bs->filename, sizeof(bs->filename), filename);
403 if (flags & BDRV_O_FILE) {
404 drv = find_protocol(filename);
406 drv = find_hdev_driver(filename);
408 drv = find_image_format(filename);
413 goto unlink_and_fail;
416 bs->opaque = qemu_mallocz(drv->instance_size);
419 * Yes, BDRV_O_NOCACHE aka O_DIRECT means we have to present a
420 * write cache to the guest. We do need the fdatasync to flush
421 * out transactions for block allocations, and we maybe have a
422 * volatile write cache in our backing device to deal with.
424 if (flags & (BDRV_O_CACHE_WB|BDRV_O_NOCACHE))
425 bs->enable_write_cache = 1;
427 /* Note: for compatibility, we open disk image files as RDWR, and
428 RDONLY as fallback */
429 if (!(flags & BDRV_O_FILE))
430 open_flags = BDRV_O_RDWR |
431 (flags & (BDRV_O_CACHE_MASK|BDRV_O_NATIVE_AIO));
433 open_flags = flags & ~(BDRV_O_FILE | BDRV_O_SNAPSHOT);
434 ret = drv->bdrv_open(bs, filename, open_flags);
435 if ((ret == -EACCES || ret == -EPERM) && !(flags & BDRV_O_FILE)) {
436 ret = drv->bdrv_open(bs, filename, open_flags & ~BDRV_O_RDWR);
440 qemu_free(bs->opaque);
444 if (bs->is_temporary)
448 if (drv->bdrv_getlength) {
449 bs->total_sectors = bdrv_getlength(bs) >> SECTOR_BITS;
452 if (bs->is_temporary) {
456 if (bs->backing_file[0] != '\0') {
457 /* if there is a backing file, use it */
458 BlockDriver *back_drv = NULL;
459 bs->backing_hd = bdrv_new("");
460 path_combine(backing_filename, sizeof(backing_filename),
461 filename, bs->backing_file);
462 if (bs->backing_format[0] != '\0')
463 back_drv = bdrv_find_format(bs->backing_format);
464 ret = bdrv_open2(bs->backing_hd, backing_filename, open_flags,
472 if (!bdrv_key_required(bs)) {
473 /* call the change callback */
474 bs->media_changed = 1;
476 bs->change_cb(bs->change_opaque);
481 void bdrv_close(BlockDriverState *bs)
485 bdrv_delete(bs->backing_hd);
486 bs->drv->bdrv_close(bs);
487 qemu_free(bs->opaque);
489 if (bs->is_temporary) {
490 unlink(bs->filename);
496 /* call the change callback */
497 bs->media_changed = 1;
499 bs->change_cb(bs->change_opaque);
503 void bdrv_delete(BlockDriverState *bs)
505 BlockDriverState **pbs;
508 while (*pbs != bs && *pbs != NULL)
518 * Run consistency checks on an image
520 * Returns the number of errors or -errno when an internal error occurs
522 int bdrv_check(BlockDriverState *bs)
524 if (bs->drv->bdrv_check == NULL) {
528 return bs->drv->bdrv_check(bs);
531 /* commit COW file into the raw image */
532 int bdrv_commit(BlockDriverState *bs)
534 BlockDriver *drv = bs->drv;
535 int64_t i, total_sectors;
537 unsigned char sector[512];
546 if (!bs->backing_hd) {
550 total_sectors = bdrv_getlength(bs) >> SECTOR_BITS;
551 for (i = 0; i < total_sectors;) {
552 if (drv->bdrv_is_allocated(bs, i, 65536, &n)) {
553 for(j = 0; j < n; j++) {
554 if (bdrv_read(bs, i, sector, 1) != 0) {
558 if (bdrv_write(bs->backing_hd, i, sector, 1) != 0) {
568 if (drv->bdrv_make_empty)
569 return drv->bdrv_make_empty(bs);
574 static int bdrv_check_byte_request(BlockDriverState *bs, int64_t offset,
579 if (!bdrv_is_inserted(bs))
585 len = bdrv_getlength(bs);
590 if ((offset > len) || (len - offset < size))
596 static int bdrv_check_request(BlockDriverState *bs, int64_t sector_num,
599 return bdrv_check_byte_request(bs, sector_num * 512, nb_sectors * 512);
602 /* return < 0 if error. See bdrv_write() for the return codes */
603 int bdrv_read(BlockDriverState *bs, int64_t sector_num,
604 uint8_t *buf, int nb_sectors)
606 BlockDriver *drv = bs->drv;
610 if (bdrv_check_request(bs, sector_num, nb_sectors))
613 return drv->bdrv_read(bs, sector_num, buf, nb_sectors);
616 /* Return < 0 if error. Important errors are:
617 -EIO generic I/O error (may happen for all errors)
618 -ENOMEDIUM No media inserted.
619 -EINVAL Invalid sector number or nb_sectors
620 -EACCES Trying to write a read-only device
622 int bdrv_write(BlockDriverState *bs, int64_t sector_num,
623 const uint8_t *buf, int nb_sectors)
625 BlockDriver *drv = bs->drv;
630 if (bdrv_check_request(bs, sector_num, nb_sectors))
633 return drv->bdrv_write(bs, sector_num, buf, nb_sectors);
636 int bdrv_pread(BlockDriverState *bs, int64_t offset,
637 void *buf, int count1)
639 uint8_t tmp_buf[SECTOR_SIZE];
640 int len, nb_sectors, count;
644 /* first read to align to sector start */
645 len = (SECTOR_SIZE - offset) & (SECTOR_SIZE - 1);
648 sector_num = offset >> SECTOR_BITS;
650 if (bdrv_read(bs, sector_num, tmp_buf, 1) < 0)
652 memcpy(buf, tmp_buf + (offset & (SECTOR_SIZE - 1)), len);
660 /* read the sectors "in place" */
661 nb_sectors = count >> SECTOR_BITS;
662 if (nb_sectors > 0) {
663 if (bdrv_read(bs, sector_num, buf, nb_sectors) < 0)
665 sector_num += nb_sectors;
666 len = nb_sectors << SECTOR_BITS;
671 /* add data from the last sector */
673 if (bdrv_read(bs, sector_num, tmp_buf, 1) < 0)
675 memcpy(buf, tmp_buf, count);
680 int bdrv_pwrite(BlockDriverState *bs, int64_t offset,
681 const void *buf, int count1)
683 uint8_t tmp_buf[SECTOR_SIZE];
684 int len, nb_sectors, count;
688 /* first write to align to sector start */
689 len = (SECTOR_SIZE - offset) & (SECTOR_SIZE - 1);
692 sector_num = offset >> SECTOR_BITS;
694 if (bdrv_read(bs, sector_num, tmp_buf, 1) < 0)
696 memcpy(tmp_buf + (offset & (SECTOR_SIZE - 1)), buf, len);
697 if (bdrv_write(bs, sector_num, tmp_buf, 1) < 0)
706 /* write the sectors "in place" */
707 nb_sectors = count >> SECTOR_BITS;
708 if (nb_sectors > 0) {
709 if (bdrv_write(bs, sector_num, buf, nb_sectors) < 0)
711 sector_num += nb_sectors;
712 len = nb_sectors << SECTOR_BITS;
717 /* add data from the last sector */
719 if (bdrv_read(bs, sector_num, tmp_buf, 1) < 0)
721 memcpy(tmp_buf, buf, count);
722 if (bdrv_write(bs, sector_num, tmp_buf, 1) < 0)
729 * Truncate file to 'offset' bytes (needed only for file protocols)
731 int bdrv_truncate(BlockDriverState *bs, int64_t offset)
733 BlockDriver *drv = bs->drv;
736 if (!drv->bdrv_truncate)
738 return drv->bdrv_truncate(bs, offset);
742 * Length of a file in bytes. Return < 0 if error or unknown.
744 int64_t bdrv_getlength(BlockDriverState *bs)
746 BlockDriver *drv = bs->drv;
749 if (!drv->bdrv_getlength) {
751 return bs->total_sectors * SECTOR_SIZE;
753 return drv->bdrv_getlength(bs);
756 /* return 0 as number of sectors if no device present or error */
757 void bdrv_get_geometry(BlockDriverState *bs, uint64_t *nb_sectors_ptr)
760 length = bdrv_getlength(bs);
764 length = length >> SECTOR_BITS;
765 *nb_sectors_ptr = length;
769 uint8_t boot_ind; /* 0x80 - active */
770 uint8_t head; /* starting head */
771 uint8_t sector; /* starting sector */
772 uint8_t cyl; /* starting cylinder */
773 uint8_t sys_ind; /* What partition type */
774 uint8_t end_head; /* end head */
775 uint8_t end_sector; /* end sector */
776 uint8_t end_cyl; /* end cylinder */
777 uint32_t start_sect; /* starting sector counting from 0 */
778 uint32_t nr_sects; /* nr of sectors in partition */
779 } __attribute__((packed));
781 /* try to guess the disk logical geometry from the MSDOS partition table. Return 0 if OK, -1 if could not guess */
782 static int guess_disk_lchs(BlockDriverState *bs,
783 int *pcylinders, int *pheads, int *psectors)
786 int ret, i, heads, sectors, cylinders;
791 bdrv_get_geometry(bs, &nb_sectors);
793 ret = bdrv_read(bs, 0, buf, 1);
796 /* test msdos magic */
797 if (buf[510] != 0x55 || buf[511] != 0xaa)
799 for(i = 0; i < 4; i++) {
800 p = ((struct partition *)(buf + 0x1be)) + i;
801 nr_sects = le32_to_cpu(p->nr_sects);
802 if (nr_sects && p->end_head) {
803 /* We make the assumption that the partition terminates on
804 a cylinder boundary */
805 heads = p->end_head + 1;
806 sectors = p->end_sector & 63;
809 cylinders = nb_sectors / (heads * sectors);
810 if (cylinders < 1 || cylinders > 16383)
814 *pcylinders = cylinders;
816 printf("guessed geometry: LCHS=%d %d %d\n",
817 cylinders, heads, sectors);
825 void bdrv_guess_geometry(BlockDriverState *bs, int *pcyls, int *pheads, int *psecs)
827 int translation, lba_detected = 0;
828 int cylinders, heads, secs;
831 /* if a geometry hint is available, use it */
832 bdrv_get_geometry(bs, &nb_sectors);
833 bdrv_get_geometry_hint(bs, &cylinders, &heads, &secs);
834 translation = bdrv_get_translation_hint(bs);
835 if (cylinders != 0) {
840 if (guess_disk_lchs(bs, &cylinders, &heads, &secs) == 0) {
842 /* if heads > 16, it means that a BIOS LBA
843 translation was active, so the default
844 hardware geometry is OK */
846 goto default_geometry;
851 /* disable any translation to be in sync with
852 the logical geometry */
853 if (translation == BIOS_ATA_TRANSLATION_AUTO) {
854 bdrv_set_translation_hint(bs,
855 BIOS_ATA_TRANSLATION_NONE);
860 /* if no geometry, use a standard physical disk geometry */
861 cylinders = nb_sectors / (16 * 63);
863 if (cylinders > 16383)
865 else if (cylinders < 2)
870 if ((lba_detected == 1) && (translation == BIOS_ATA_TRANSLATION_AUTO)) {
871 if ((*pcyls * *pheads) <= 131072) {
872 bdrv_set_translation_hint(bs,
873 BIOS_ATA_TRANSLATION_LARGE);
875 bdrv_set_translation_hint(bs,
876 BIOS_ATA_TRANSLATION_LBA);
880 bdrv_set_geometry_hint(bs, *pcyls, *pheads, *psecs);
884 void bdrv_set_geometry_hint(BlockDriverState *bs,
885 int cyls, int heads, int secs)
892 void bdrv_set_type_hint(BlockDriverState *bs, int type)
895 bs->removable = ((type == BDRV_TYPE_CDROM ||
896 type == BDRV_TYPE_FLOPPY));
899 void bdrv_set_translation_hint(BlockDriverState *bs, int translation)
901 bs->translation = translation;
904 void bdrv_get_geometry_hint(BlockDriverState *bs,
905 int *pcyls, int *pheads, int *psecs)
912 int bdrv_get_type_hint(BlockDriverState *bs)
917 int bdrv_get_translation_hint(BlockDriverState *bs)
919 return bs->translation;
922 int bdrv_is_removable(BlockDriverState *bs)
924 return bs->removable;
927 int bdrv_is_read_only(BlockDriverState *bs)
929 return bs->read_only;
932 int bdrv_is_sg(BlockDriverState *bs)
937 int bdrv_enable_write_cache(BlockDriverState *bs)
939 return bs->enable_write_cache;
942 /* XXX: no longer used */
943 void bdrv_set_change_cb(BlockDriverState *bs,
944 void (*change_cb)(void *opaque), void *opaque)
946 bs->change_cb = change_cb;
947 bs->change_opaque = opaque;
950 int bdrv_is_encrypted(BlockDriverState *bs)
952 if (bs->backing_hd && bs->backing_hd->encrypted)
954 return bs->encrypted;
957 int bdrv_key_required(BlockDriverState *bs)
959 BlockDriverState *backing_hd = bs->backing_hd;
961 if (backing_hd && backing_hd->encrypted && !backing_hd->valid_key)
963 return (bs->encrypted && !bs->valid_key);
966 int bdrv_set_key(BlockDriverState *bs, const char *key)
969 if (bs->backing_hd && bs->backing_hd->encrypted) {
970 ret = bdrv_set_key(bs->backing_hd, key);
976 if (!bs->encrypted || !bs->drv || !bs->drv->bdrv_set_key)
978 ret = bs->drv->bdrv_set_key(bs, key);
981 } else if (!bs->valid_key) {
983 /* call the change callback now, we skipped it on open */
984 bs->media_changed = 1;
986 bs->change_cb(bs->change_opaque);
991 void bdrv_get_format(BlockDriverState *bs, char *buf, int buf_size)
996 pstrcpy(buf, buf_size, bs->drv->format_name);
1000 void bdrv_iterate_format(void (*it)(void *opaque, const char *name),
1005 for (drv = first_drv; drv != NULL; drv = drv->next) {
1006 it(opaque, drv->format_name);
1010 BlockDriverState *bdrv_find(const char *name)
1012 BlockDriverState *bs;
1014 for (bs = bdrv_first; bs != NULL; bs = bs->next) {
1015 if (!strcmp(name, bs->device_name))
1021 void bdrv_iterate(void (*it)(void *opaque, BlockDriverState *bs), void *opaque)
1023 BlockDriverState *bs;
1025 for (bs = bdrv_first; bs != NULL; bs = bs->next) {
1030 const char *bdrv_get_device_name(BlockDriverState *bs)
1032 return bs->device_name;
1035 void bdrv_flush(BlockDriverState *bs)
1039 if (bs->drv->bdrv_flush)
1040 bs->drv->bdrv_flush(bs);
1042 bdrv_flush(bs->backing_hd);
1045 void bdrv_flush_all(void)
1047 BlockDriverState *bs;
1049 for (bs = bdrv_first; bs != NULL; bs = bs->next)
1050 if (bs->drv && !bdrv_is_read_only(bs) &&
1051 (!bdrv_is_removable(bs) || bdrv_is_inserted(bs)))
1056 * Returns true iff the specified sector is present in the disk image. Drivers
1057 * not implementing the functionality are assumed to not support backing files,
1058 * hence all their sectors are reported as allocated.
1060 * 'pnum' is set to the number of sectors (including and immediately following
1061 * the specified sector) that are known to be in the same
1062 * allocated/unallocated state.
1064 * 'nb_sectors' is the max value 'pnum' should be set to.
1066 int bdrv_is_allocated(BlockDriverState *bs, int64_t sector_num, int nb_sectors,
1070 if (!bs->drv->bdrv_is_allocated) {
1071 if (sector_num >= bs->total_sectors) {
1075 n = bs->total_sectors - sector_num;
1076 *pnum = (n < nb_sectors) ? (n) : (nb_sectors);
1079 return bs->drv->bdrv_is_allocated(bs, sector_num, nb_sectors, pnum);
1082 void bdrv_info(Monitor *mon)
1084 BlockDriverState *bs;
1086 for (bs = bdrv_first; bs != NULL; bs = bs->next) {
1087 monitor_printf(mon, "%s:", bs->device_name);
1088 monitor_printf(mon, " type=");
1091 monitor_printf(mon, "hd");
1093 case BDRV_TYPE_CDROM:
1094 monitor_printf(mon, "cdrom");
1096 case BDRV_TYPE_FLOPPY:
1097 monitor_printf(mon, "floppy");
1100 monitor_printf(mon, " removable=%d", bs->removable);
1101 if (bs->removable) {
1102 monitor_printf(mon, " locked=%d", bs->locked);
1105 monitor_printf(mon, " file=");
1106 monitor_print_filename(mon, bs->filename);
1107 if (bs->backing_file[0] != '\0') {
1108 monitor_printf(mon, " backing_file=");
1109 monitor_print_filename(mon, bs->backing_file);
1111 monitor_printf(mon, " ro=%d", bs->read_only);
1112 monitor_printf(mon, " drv=%s", bs->drv->format_name);
1113 monitor_printf(mon, " encrypted=%d", bdrv_is_encrypted(bs));
1115 monitor_printf(mon, " [not inserted]");
1117 monitor_printf(mon, "\n");
1121 /* The "info blockstats" command. */
1122 void bdrv_info_stats(Monitor *mon)
1124 BlockDriverState *bs;
1126 for (bs = bdrv_first; bs != NULL; bs = bs->next) {
1127 monitor_printf(mon, "%s:"
1128 " rd_bytes=%" PRIu64
1129 " wr_bytes=%" PRIu64
1130 " rd_operations=%" PRIu64
1131 " wr_operations=%" PRIu64
1134 bs->rd_bytes, bs->wr_bytes,
1135 bs->rd_ops, bs->wr_ops);
1139 const char *bdrv_get_encrypted_filename(BlockDriverState *bs)
1141 if (bs->backing_hd && bs->backing_hd->encrypted)
1142 return bs->backing_file;
1143 else if (bs->encrypted)
1144 return bs->filename;
1149 void bdrv_get_backing_filename(BlockDriverState *bs,
1150 char *filename, int filename_size)
1152 if (!bs->backing_hd) {
1153 pstrcpy(filename, filename_size, "");
1155 pstrcpy(filename, filename_size, bs->backing_file);
1159 int bdrv_write_compressed(BlockDriverState *bs, int64_t sector_num,
1160 const uint8_t *buf, int nb_sectors)
1162 BlockDriver *drv = bs->drv;
1165 if (!drv->bdrv_write_compressed)
1167 if (bdrv_check_request(bs, sector_num, nb_sectors))
1169 return drv->bdrv_write_compressed(bs, sector_num, buf, nb_sectors);
1172 int bdrv_get_info(BlockDriverState *bs, BlockDriverInfo *bdi)
1174 BlockDriver *drv = bs->drv;
1177 if (!drv->bdrv_get_info)
1179 memset(bdi, 0, sizeof(*bdi));
1180 return drv->bdrv_get_info(bs, bdi);
1183 int bdrv_save_vmstate(BlockDriverState *bs, const uint8_t *buf,
1184 int64_t pos, int size)
1186 BlockDriver *drv = bs->drv;
1189 if (!drv->bdrv_save_vmstate)
1191 return drv->bdrv_save_vmstate(bs, buf, pos, size);
1194 int bdrv_load_vmstate(BlockDriverState *bs, uint8_t *buf,
1195 int64_t pos, int size)
1197 BlockDriver *drv = bs->drv;
1200 if (!drv->bdrv_load_vmstate)
1202 return drv->bdrv_load_vmstate(bs, buf, pos, size);
1205 /**************************************************************/
1206 /* handling of snapshots */
1208 int bdrv_snapshot_create(BlockDriverState *bs,
1209 QEMUSnapshotInfo *sn_info)
1211 BlockDriver *drv = bs->drv;
1214 if (!drv->bdrv_snapshot_create)
1216 return drv->bdrv_snapshot_create(bs, sn_info);
1219 int bdrv_snapshot_goto(BlockDriverState *bs,
1220 const char *snapshot_id)
1222 BlockDriver *drv = bs->drv;
1225 if (!drv->bdrv_snapshot_goto)
1227 return drv->bdrv_snapshot_goto(bs, snapshot_id);
1230 int bdrv_snapshot_delete(BlockDriverState *bs, const char *snapshot_id)
1232 BlockDriver *drv = bs->drv;
1235 if (!drv->bdrv_snapshot_delete)
1237 return drv->bdrv_snapshot_delete(bs, snapshot_id);
1240 int bdrv_snapshot_list(BlockDriverState *bs,
1241 QEMUSnapshotInfo **psn_info)
1243 BlockDriver *drv = bs->drv;
1246 if (!drv->bdrv_snapshot_list)
1248 return drv->bdrv_snapshot_list(bs, psn_info);
1251 #define NB_SUFFIXES 4
1253 char *get_human_readable_size(char *buf, int buf_size, int64_t size)
1255 static const char suffixes[NB_SUFFIXES] = "KMGT";
1260 snprintf(buf, buf_size, "%" PRId64, size);
1263 for(i = 0; i < NB_SUFFIXES; i++) {
1264 if (size < (10 * base)) {
1265 snprintf(buf, buf_size, "%0.1f%c",
1266 (double)size / base,
1269 } else if (size < (1000 * base) || i == (NB_SUFFIXES - 1)) {
1270 snprintf(buf, buf_size, "%" PRId64 "%c",
1271 ((size + (base >> 1)) / base),
1281 char *bdrv_snapshot_dump(char *buf, int buf_size, QEMUSnapshotInfo *sn)
1283 char buf1[128], date_buf[128], clock_buf[128];
1293 snprintf(buf, buf_size,
1294 "%-10s%-20s%7s%20s%15s",
1295 "ID", "TAG", "VM SIZE", "DATE", "VM CLOCK");
1299 ptm = localtime(&ti);
1300 strftime(date_buf, sizeof(date_buf),
1301 "%Y-%m-%d %H:%M:%S", ptm);
1303 localtime_r(&ti, &tm);
1304 strftime(date_buf, sizeof(date_buf),
1305 "%Y-%m-%d %H:%M:%S", &tm);
1307 secs = sn->vm_clock_nsec / 1000000000;
1308 snprintf(clock_buf, sizeof(clock_buf),
1309 "%02d:%02d:%02d.%03d",
1311 (int)((secs / 60) % 60),
1313 (int)((sn->vm_clock_nsec / 1000000) % 1000));
1314 snprintf(buf, buf_size,
1315 "%-10s%-20s%7s%20s%15s",
1316 sn->id_str, sn->name,
1317 get_human_readable_size(buf1, sizeof(buf1), sn->vm_state_size),
1325 /**************************************************************/
1328 BlockDriverAIOCB *bdrv_aio_readv(BlockDriverState *bs, int64_t sector_num,
1329 QEMUIOVector *qiov, int nb_sectors,
1330 BlockDriverCompletionFunc *cb, void *opaque)
1332 BlockDriver *drv = bs->drv;
1333 BlockDriverAIOCB *ret;
1337 if (bdrv_check_request(bs, sector_num, nb_sectors))
1340 ret = drv->bdrv_aio_readv(bs, sector_num, qiov, nb_sectors,
1344 /* Update stats even though technically transfer has not happened. */
1345 bs->rd_bytes += (unsigned) nb_sectors * SECTOR_SIZE;
1352 BlockDriverAIOCB *bdrv_aio_writev(BlockDriverState *bs, int64_t sector_num,
1353 QEMUIOVector *qiov, int nb_sectors,
1354 BlockDriverCompletionFunc *cb, void *opaque)
1356 BlockDriver *drv = bs->drv;
1357 BlockDriverAIOCB *ret;
1363 if (bdrv_check_request(bs, sector_num, nb_sectors))
1366 ret = drv->bdrv_aio_writev(bs, sector_num, qiov, nb_sectors,
1370 /* Update stats even though technically transfer has not happened. */
1371 bs->wr_bytes += (unsigned) nb_sectors * SECTOR_SIZE;
1379 typedef struct MultiwriteCB {
1384 BlockDriverCompletionFunc *cb;
1386 QEMUIOVector *free_qiov;
1391 static void multiwrite_user_cb(MultiwriteCB *mcb)
1395 for (i = 0; i < mcb->num_callbacks; i++) {
1396 mcb->callbacks[i].cb(mcb->callbacks[i].opaque, mcb->error);
1397 qemu_free(mcb->callbacks[i].free_qiov);
1398 qemu_free(mcb->callbacks[i].free_buf);
1402 static void multiwrite_cb(void *opaque, int ret)
1404 MultiwriteCB *mcb = opaque;
1408 multiwrite_user_cb(mcb);
1411 mcb->num_requests--;
1412 if (mcb->num_requests == 0) {
1413 if (mcb->error == 0) {
1414 multiwrite_user_cb(mcb);
1420 static int multiwrite_req_compare(const void *a, const void *b)
1422 return (((BlockRequest*) a)->sector - ((BlockRequest*) b)->sector);
1426 * Takes a bunch of requests and tries to merge them. Returns the number of
1427 * requests that remain after merging.
1429 static int multiwrite_merge(BlockDriverState *bs, BlockRequest *reqs,
1430 int num_reqs, MultiwriteCB *mcb)
1434 // Sort requests by start sector
1435 qsort(reqs, num_reqs, sizeof(*reqs), &multiwrite_req_compare);
1437 // Check if adjacent requests touch the same clusters. If so, combine them,
1438 // filling up gaps with zero sectors.
1440 for (i = 1; i < num_reqs; i++) {
1442 int64_t oldreq_last = reqs[outidx].sector + reqs[outidx].nb_sectors;
1444 // This handles the cases that are valid for all block drivers, namely
1445 // exactly sequential writes and overlapping writes.
1446 if (reqs[i].sector <= oldreq_last) {
1450 // The block driver may decide that it makes sense to combine requests
1451 // even if there is a gap of some sectors between them. In this case,
1452 // the gap is filled with zeros (therefore only applicable for yet
1453 // unused space in format like qcow2).
1454 if (!merge && bs->drv->bdrv_merge_requests) {
1455 merge = bs->drv->bdrv_merge_requests(bs, &reqs[outidx], &reqs[i]);
1460 QEMUIOVector *qiov = qemu_mallocz(sizeof(*qiov));
1461 qemu_iovec_init(qiov,
1462 reqs[outidx].qiov->niov + reqs[i].qiov->niov + 1);
1464 // Add the first request to the merged one. If the requests are
1465 // overlapping, drop the last sectors of the first request.
1466 size = (reqs[i].sector - reqs[outidx].sector) << 9;
1467 qemu_iovec_concat(qiov, reqs[outidx].qiov, size);
1469 // We might need to add some zeros between the two requests
1470 if (reqs[i].sector > oldreq_last) {
1471 size_t zero_bytes = (reqs[i].sector - oldreq_last) << 9;
1472 uint8_t *buf = qemu_blockalign(bs, zero_bytes);
1473 memset(buf, 0, zero_bytes);
1474 qemu_iovec_add(qiov, buf, zero_bytes);
1475 mcb->callbacks[i].free_buf = buf;
1478 // Add the second request
1479 qemu_iovec_concat(qiov, reqs[i].qiov, reqs[i].qiov->size);
1481 reqs[outidx].nb_sectors += reqs[i].nb_sectors;
1482 reqs[outidx].qiov = qiov;
1484 mcb->callbacks[i].free_qiov = reqs[outidx].qiov;
1487 reqs[outidx].sector = reqs[i].sector;
1488 reqs[outidx].nb_sectors = reqs[i].nb_sectors;
1489 reqs[outidx].qiov = reqs[i].qiov;
1497 * Submit multiple AIO write requests at once.
1499 * On success, the function returns 0 and all requests in the reqs array have
1500 * been submitted. In error case this function returns -1, and any of the
1501 * requests may or may not be submitted yet. In particular, this means that the
1502 * callback will be called for some of the requests, for others it won't. The
1503 * caller must check the error field of the BlockRequest to wait for the right
1504 * callbacks (if error != 0, no callback will be called).
1506 * The implementation may modify the contents of the reqs array, e.g. to merge
1507 * requests. However, the fields opaque and error are left unmodified as they
1508 * are used to signal failure for a single request to the caller.
1510 int bdrv_aio_multiwrite(BlockDriverState *bs, BlockRequest *reqs, int num_reqs)
1512 BlockDriverAIOCB *acb;
1516 if (num_reqs == 0) {
1520 // Create MultiwriteCB structure
1521 mcb = qemu_mallocz(sizeof(*mcb) + num_reqs * sizeof(*mcb->callbacks));
1522 mcb->num_requests = 0;
1523 mcb->num_callbacks = num_reqs;
1525 for (i = 0; i < num_reqs; i++) {
1526 mcb->callbacks[i].cb = reqs[i].cb;
1527 mcb->callbacks[i].opaque = reqs[i].opaque;
1530 // Check for mergable requests
1531 num_reqs = multiwrite_merge(bs, reqs, num_reqs, mcb);
1533 // Run the aio requests
1534 for (i = 0; i < num_reqs; i++) {
1535 acb = bdrv_aio_writev(bs, reqs[i].sector, reqs[i].qiov,
1536 reqs[i].nb_sectors, multiwrite_cb, mcb);
1539 // We can only fail the whole thing if no request has been
1540 // submitted yet. Otherwise we'll wait for the submitted AIOs to
1541 // complete and report the error in the callback.
1542 if (mcb->num_requests == 0) {
1543 reqs[i].error = EIO;
1550 mcb->num_requests++;
1561 BlockDriverAIOCB *bdrv_aio_flush(BlockDriverState *bs,
1562 BlockDriverCompletionFunc *cb, void *opaque)
1564 BlockDriver *drv = bs->drv;
1570 * Note that unlike bdrv_flush the driver is reponsible for flushing a
1571 * backing image if it exists.
1573 return drv->bdrv_aio_flush(bs, cb, opaque);
1576 void bdrv_aio_cancel(BlockDriverAIOCB *acb)
1578 acb->pool->cancel(acb);
1582 /**************************************************************/
1583 /* async block device emulation */
1585 typedef struct BlockDriverAIOCBSync {
1586 BlockDriverAIOCB common;
1589 /* vector translation state */
1593 } BlockDriverAIOCBSync;
1595 static void bdrv_aio_cancel_em(BlockDriverAIOCB *blockacb)
1597 BlockDriverAIOCBSync *acb = (BlockDriverAIOCBSync *)blockacb;
1598 qemu_bh_delete(acb->bh);
1600 qemu_aio_release(acb);
1603 static AIOPool bdrv_em_aio_pool = {
1604 .aiocb_size = sizeof(BlockDriverAIOCBSync),
1605 .cancel = bdrv_aio_cancel_em,
1608 static void bdrv_aio_bh_cb(void *opaque)
1610 BlockDriverAIOCBSync *acb = opaque;
1613 qemu_iovec_from_buffer(acb->qiov, acb->bounce, acb->qiov->size);
1614 qemu_vfree(acb->bounce);
1615 acb->common.cb(acb->common.opaque, acb->ret);
1616 qemu_bh_delete(acb->bh);
1618 qemu_aio_release(acb);
1621 static BlockDriverAIOCB *bdrv_aio_rw_vector(BlockDriverState *bs,
1625 BlockDriverCompletionFunc *cb,
1630 BlockDriverAIOCBSync *acb;
1632 acb = qemu_aio_get(&bdrv_em_aio_pool, bs, cb, opaque);
1633 acb->is_write = is_write;
1635 acb->bounce = qemu_blockalign(bs, qiov->size);
1638 acb->bh = qemu_bh_new(bdrv_aio_bh_cb, acb);
1641 qemu_iovec_to_buffer(acb->qiov, acb->bounce);
1642 acb->ret = bdrv_write(bs, sector_num, acb->bounce, nb_sectors);
1644 acb->ret = bdrv_read(bs, sector_num, acb->bounce, nb_sectors);
1647 qemu_bh_schedule(acb->bh);
1649 return &acb->common;
1652 static BlockDriverAIOCB *bdrv_aio_readv_em(BlockDriverState *bs,
1653 int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
1654 BlockDriverCompletionFunc *cb, void *opaque)
1656 return bdrv_aio_rw_vector(bs, sector_num, qiov, nb_sectors, cb, opaque, 0);
1659 static BlockDriverAIOCB *bdrv_aio_writev_em(BlockDriverState *bs,
1660 int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
1661 BlockDriverCompletionFunc *cb, void *opaque)
1663 return bdrv_aio_rw_vector(bs, sector_num, qiov, nb_sectors, cb, opaque, 1);
1666 static BlockDriverAIOCB *bdrv_aio_flush_em(BlockDriverState *bs,
1667 BlockDriverCompletionFunc *cb, void *opaque)
1669 BlockDriverAIOCBSync *acb;
1671 acb = qemu_aio_get(&bdrv_em_aio_pool, bs, cb, opaque);
1672 acb->is_write = 1; /* don't bounce in the completion hadler */
1678 acb->bh = qemu_bh_new(bdrv_aio_bh_cb, acb);
1681 qemu_bh_schedule(acb->bh);
1682 return &acb->common;
1685 /**************************************************************/
1686 /* sync block device emulation */
1688 static void bdrv_rw_em_cb(void *opaque, int ret)
1690 *(int *)opaque = ret;
1693 #define NOT_DONE 0x7fffffff
1695 static int bdrv_read_em(BlockDriverState *bs, int64_t sector_num,
1696 uint8_t *buf, int nb_sectors)
1699 BlockDriverAIOCB *acb;
1703 async_ret = NOT_DONE;
1704 iov.iov_base = (void *)buf;
1705 iov.iov_len = nb_sectors * 512;
1706 qemu_iovec_init_external(&qiov, &iov, 1);
1707 acb = bdrv_aio_readv(bs, sector_num, &qiov, nb_sectors,
1708 bdrv_rw_em_cb, &async_ret);
1712 while (async_ret == NOT_DONE) {
1719 static int bdrv_write_em(BlockDriverState *bs, int64_t sector_num,
1720 const uint8_t *buf, int nb_sectors)
1723 BlockDriverAIOCB *acb;
1727 async_ret = NOT_DONE;
1728 iov.iov_base = (void *)buf;
1729 iov.iov_len = nb_sectors * 512;
1730 qemu_iovec_init_external(&qiov, &iov, 1);
1731 acb = bdrv_aio_writev(bs, sector_num, &qiov, nb_sectors,
1732 bdrv_rw_em_cb, &async_ret);
1735 while (async_ret == NOT_DONE) {
1741 void bdrv_init(void)
1743 module_call_init(MODULE_INIT_BLOCK);
1746 void *qemu_aio_get(AIOPool *pool, BlockDriverState *bs,
1747 BlockDriverCompletionFunc *cb, void *opaque)
1749 BlockDriverAIOCB *acb;
1751 if (pool->free_aiocb) {
1752 acb = pool->free_aiocb;
1753 pool->free_aiocb = acb->next;
1755 acb = qemu_mallocz(pool->aiocb_size);
1760 acb->opaque = opaque;
1764 void qemu_aio_release(void *p)
1766 BlockDriverAIOCB *acb = (BlockDriverAIOCB *)p;
1767 AIOPool *pool = acb->pool;
1768 acb->next = pool->free_aiocb;
1769 pool->free_aiocb = acb;
1772 /**************************************************************/
1773 /* removable device support */
1776 * Return TRUE if the media is present
1778 int bdrv_is_inserted(BlockDriverState *bs)
1780 BlockDriver *drv = bs->drv;
1784 if (!drv->bdrv_is_inserted)
1786 ret = drv->bdrv_is_inserted(bs);
1791 * Return TRUE if the media changed since the last call to this
1792 * function. It is currently only used for floppy disks
1794 int bdrv_media_changed(BlockDriverState *bs)
1796 BlockDriver *drv = bs->drv;
1799 if (!drv || !drv->bdrv_media_changed)
1802 ret = drv->bdrv_media_changed(bs);
1803 if (ret == -ENOTSUP)
1804 ret = bs->media_changed;
1805 bs->media_changed = 0;
1810 * If eject_flag is TRUE, eject the media. Otherwise, close the tray
1812 int bdrv_eject(BlockDriverState *bs, int eject_flag)
1814 BlockDriver *drv = bs->drv;
1821 if (!drv || !drv->bdrv_eject) {
1824 ret = drv->bdrv_eject(bs, eject_flag);
1826 if (ret == -ENOTSUP) {
1835 int bdrv_is_locked(BlockDriverState *bs)
1841 * Lock or unlock the media (if it is locked, the user won't be able
1842 * to eject it manually).
1844 void bdrv_set_locked(BlockDriverState *bs, int locked)
1846 BlockDriver *drv = bs->drv;
1848 bs->locked = locked;
1849 if (drv && drv->bdrv_set_locked) {
1850 drv->bdrv_set_locked(bs, locked);
1854 /* needed for generic scsi interface */
1856 int bdrv_ioctl(BlockDriverState *bs, unsigned long int req, void *buf)
1858 BlockDriver *drv = bs->drv;
1860 if (drv && drv->bdrv_ioctl)
1861 return drv->bdrv_ioctl(bs, req, buf);
1865 BlockDriverAIOCB *bdrv_aio_ioctl(BlockDriverState *bs,
1866 unsigned long int req, void *buf,
1867 BlockDriverCompletionFunc *cb, void *opaque)
1869 BlockDriver *drv = bs->drv;
1871 if (drv && drv->bdrv_aio_ioctl)
1872 return drv->bdrv_aio_ioctl(bs, req, buf, cb, opaque);
1876 void *qemu_blockalign(BlockDriverState *bs, size_t size)
1878 return qemu_memalign((bs && bs->buffer_alignment) ? bs->buffer_alignment : 512, size);