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"
28 #include "block_int.h"
31 #include "qemu-coroutine.h"
32 #include "qmp-commands.h"
33 #include "qemu-timer.h"
36 #include <sys/types.h>
38 #include <sys/ioctl.h>
39 #include <sys/queue.h>
49 #define NOT_DONE 0x7fffffff /* used while emulated sync operation in progress */
51 static void bdrv_dev_change_media_cb(BlockDriverState
*bs
, bool load
);
52 static BlockDriverAIOCB
*bdrv_aio_readv_em(BlockDriverState
*bs
,
53 int64_t sector_num
, QEMUIOVector
*qiov
, int nb_sectors
,
54 BlockDriverCompletionFunc
*cb
, void *opaque
);
55 static BlockDriverAIOCB
*bdrv_aio_writev_em(BlockDriverState
*bs
,
56 int64_t sector_num
, QEMUIOVector
*qiov
, int nb_sectors
,
57 BlockDriverCompletionFunc
*cb
, void *opaque
);
58 static int coroutine_fn
bdrv_co_readv_em(BlockDriverState
*bs
,
59 int64_t sector_num
, int nb_sectors
,
61 static int coroutine_fn
bdrv_co_writev_em(BlockDriverState
*bs
,
62 int64_t sector_num
, int nb_sectors
,
64 static int coroutine_fn
bdrv_co_do_readv(BlockDriverState
*bs
,
65 int64_t sector_num
, int nb_sectors
, QEMUIOVector
*qiov
);
66 static int coroutine_fn
bdrv_co_do_writev(BlockDriverState
*bs
,
67 int64_t sector_num
, int nb_sectors
, QEMUIOVector
*qiov
);
68 static BlockDriverAIOCB
*bdrv_co_aio_rw_vector(BlockDriverState
*bs
,
72 BlockDriverCompletionFunc
*cb
,
75 static void coroutine_fn
bdrv_co_do_rw(void *opaque
);
77 static bool bdrv_exceed_bps_limits(BlockDriverState
*bs
, int nb_sectors
,
78 bool is_write
, double elapsed_time
, uint64_t *wait
);
79 static bool bdrv_exceed_iops_limits(BlockDriverState
*bs
, bool is_write
,
80 double elapsed_time
, uint64_t *wait
);
81 static bool bdrv_exceed_io_limits(BlockDriverState
*bs
, int nb_sectors
,
82 bool is_write
, int64_t *wait
);
84 static QTAILQ_HEAD(, BlockDriverState
) bdrv_states
=
85 QTAILQ_HEAD_INITIALIZER(bdrv_states
);
87 static QLIST_HEAD(, BlockDriver
) bdrv_drivers
=
88 QLIST_HEAD_INITIALIZER(bdrv_drivers
);
90 /* The device to use for VM snapshots */
91 static BlockDriverState
*bs_snapshots
;
93 /* If non-zero, use only whitelisted block drivers */
94 static int use_bdrv_whitelist
;
97 static int is_windows_drive_prefix(const char *filename
)
99 return (((filename
[0] >= 'a' && filename
[0] <= 'z') ||
100 (filename
[0] >= 'A' && filename
[0] <= 'Z')) &&
104 int is_windows_drive(const char *filename
)
106 if (is_windows_drive_prefix(filename
) &&
109 if (strstart(filename
, "\\\\.\\", NULL
) ||
110 strstart(filename
, "//./", NULL
))
116 /* throttling disk I/O limits */
117 void bdrv_io_limits_disable(BlockDriverState
*bs
)
119 bs
->io_limits_enabled
= false;
121 while (qemu_co_queue_next(&bs
->throttled_reqs
));
123 if (bs
->block_timer
) {
124 qemu_del_timer(bs
->block_timer
);
125 qemu_free_timer(bs
->block_timer
);
126 bs
->block_timer
= NULL
;
132 memset(&bs
->io_base
, 0, sizeof(bs
->io_base
));
135 static void bdrv_block_timer(void *opaque
)
137 BlockDriverState
*bs
= opaque
;
139 qemu_co_queue_next(&bs
->throttled_reqs
);
142 void bdrv_io_limits_enable(BlockDriverState
*bs
)
144 qemu_co_queue_init(&bs
->throttled_reqs
);
145 bs
->block_timer
= qemu_new_timer_ns(vm_clock
, bdrv_block_timer
, bs
);
146 bs
->slice_time
= 5 * BLOCK_IO_SLICE_TIME
;
147 bs
->slice_start
= qemu_get_clock_ns(vm_clock
);
148 bs
->slice_end
= bs
->slice_start
+ bs
->slice_time
;
149 memset(&bs
->io_base
, 0, sizeof(bs
->io_base
));
150 bs
->io_limits_enabled
= true;
153 bool bdrv_io_limits_enabled(BlockDriverState
*bs
)
155 BlockIOLimit
*io_limits
= &bs
->io_limits
;
156 return io_limits
->bps
[BLOCK_IO_LIMIT_READ
]
157 || io_limits
->bps
[BLOCK_IO_LIMIT_WRITE
]
158 || io_limits
->bps
[BLOCK_IO_LIMIT_TOTAL
]
159 || io_limits
->iops
[BLOCK_IO_LIMIT_READ
]
160 || io_limits
->iops
[BLOCK_IO_LIMIT_WRITE
]
161 || io_limits
->iops
[BLOCK_IO_LIMIT_TOTAL
];
164 static void bdrv_io_limits_intercept(BlockDriverState
*bs
,
165 bool is_write
, int nb_sectors
)
167 int64_t wait_time
= -1;
169 if (!qemu_co_queue_empty(&bs
->throttled_reqs
)) {
170 qemu_co_queue_wait(&bs
->throttled_reqs
);
173 /* In fact, we hope to keep each request's timing, in FIFO mode. The next
174 * throttled requests will not be dequeued until the current request is
175 * allowed to be serviced. So if the current request still exceeds the
176 * limits, it will be inserted to the head. All requests followed it will
177 * be still in throttled_reqs queue.
180 while (bdrv_exceed_io_limits(bs
, nb_sectors
, is_write
, &wait_time
)) {
181 qemu_mod_timer(bs
->block_timer
,
182 wait_time
+ qemu_get_clock_ns(vm_clock
));
183 qemu_co_queue_wait_insert_head(&bs
->throttled_reqs
);
186 qemu_co_queue_next(&bs
->throttled_reqs
);
189 /* check if the path starts with "<protocol>:" */
190 static int path_has_protocol(const char *path
)
193 if (is_windows_drive(path
) ||
194 is_windows_drive_prefix(path
)) {
199 return strchr(path
, ':') != NULL
;
202 int path_is_absolute(const char *path
)
206 /* specific case for names like: "\\.\d:" */
207 if (*path
== '/' || *path
== '\\')
210 p
= strchr(path
, ':');
216 return (*p
== '/' || *p
== '\\');
222 /* if filename is absolute, just copy it to dest. Otherwise, build a
223 path to it by considering it is relative to base_path. URL are
225 void path_combine(char *dest
, int dest_size
,
226 const char *base_path
,
227 const char *filename
)
234 if (path_is_absolute(filename
)) {
235 pstrcpy(dest
, dest_size
, filename
);
237 p
= strchr(base_path
, ':');
242 p1
= strrchr(base_path
, '/');
246 p2
= strrchr(base_path
, '\\');
258 if (len
> dest_size
- 1)
260 memcpy(dest
, base_path
, len
);
262 pstrcat(dest
, dest_size
, filename
);
266 void bdrv_register(BlockDriver
*bdrv
)
268 /* Block drivers without coroutine functions need emulation */
269 if (!bdrv
->bdrv_co_readv
) {
270 bdrv
->bdrv_co_readv
= bdrv_co_readv_em
;
271 bdrv
->bdrv_co_writev
= bdrv_co_writev_em
;
273 /* bdrv_co_readv_em()/brdv_co_writev_em() work in terms of aio, so if
274 * the block driver lacks aio we need to emulate that too.
276 if (!bdrv
->bdrv_aio_readv
) {
277 /* add AIO emulation layer */
278 bdrv
->bdrv_aio_readv
= bdrv_aio_readv_em
;
279 bdrv
->bdrv_aio_writev
= bdrv_aio_writev_em
;
283 QLIST_INSERT_HEAD(&bdrv_drivers
, bdrv
, list
);
286 /* create a new block device (by default it is empty) */
287 BlockDriverState
*bdrv_new(const char *device_name
)
289 BlockDriverState
*bs
;
291 bs
= g_malloc0(sizeof(BlockDriverState
));
292 pstrcpy(bs
->device_name
, sizeof(bs
->device_name
), device_name
);
293 if (device_name
[0] != '\0') {
294 QTAILQ_INSERT_TAIL(&bdrv_states
, bs
, list
);
296 bdrv_iostatus_disable(bs
);
300 BlockDriver
*bdrv_find_format(const char *format_name
)
303 QLIST_FOREACH(drv1
, &bdrv_drivers
, list
) {
304 if (!strcmp(drv1
->format_name
, format_name
)) {
311 static int bdrv_is_whitelisted(BlockDriver
*drv
)
313 static const char *whitelist
[] = {
314 CONFIG_BDRV_WHITELIST
319 return 1; /* no whitelist, anything goes */
321 for (p
= whitelist
; *p
; p
++) {
322 if (!strcmp(drv
->format_name
, *p
)) {
329 BlockDriver
*bdrv_find_whitelisted_format(const char *format_name
)
331 BlockDriver
*drv
= bdrv_find_format(format_name
);
332 return drv
&& bdrv_is_whitelisted(drv
) ? drv
: NULL
;
335 int bdrv_create(BlockDriver
*drv
, const char* filename
,
336 QEMUOptionParameter
*options
)
338 if (!drv
->bdrv_create
)
341 return drv
->bdrv_create(filename
, options
);
344 int bdrv_create_file(const char* filename
, QEMUOptionParameter
*options
)
348 drv
= bdrv_find_protocol(filename
);
353 return bdrv_create(drv
, filename
, options
);
357 void get_tmp_filename(char *filename
, int size
)
359 char temp_dir
[MAX_PATH
];
361 GetTempPath(MAX_PATH
, temp_dir
);
362 GetTempFileName(temp_dir
, "qem", 0, filename
);
365 void get_tmp_filename(char *filename
, int size
)
369 /* XXX: race condition possible */
370 tmpdir
= getenv("TMPDIR");
373 snprintf(filename
, size
, "%s/vl.XXXXXX", tmpdir
);
374 fd
= mkstemp(filename
);
380 * Detect host devices. By convention, /dev/cdrom[N] is always
381 * recognized as a host CDROM.
383 static BlockDriver
*find_hdev_driver(const char *filename
)
385 int score_max
= 0, score
;
386 BlockDriver
*drv
= NULL
, *d
;
388 QLIST_FOREACH(d
, &bdrv_drivers
, list
) {
389 if (d
->bdrv_probe_device
) {
390 score
= d
->bdrv_probe_device(filename
);
391 if (score
> score_max
) {
401 BlockDriver
*bdrv_find_protocol(const char *filename
)
408 /* TODO Drivers without bdrv_file_open must be specified explicitly */
411 * XXX(hch): we really should not let host device detection
412 * override an explicit protocol specification, but moving this
413 * later breaks access to device names with colons in them.
414 * Thanks to the brain-dead persistent naming schemes on udev-
415 * based Linux systems those actually are quite common.
417 drv1
= find_hdev_driver(filename
);
422 if (!path_has_protocol(filename
)) {
423 return bdrv_find_format("file");
425 p
= strchr(filename
, ':');
428 if (len
> sizeof(protocol
) - 1)
429 len
= sizeof(protocol
) - 1;
430 memcpy(protocol
, filename
, len
);
431 protocol
[len
] = '\0';
432 QLIST_FOREACH(drv1
, &bdrv_drivers
, list
) {
433 if (drv1
->protocol_name
&&
434 !strcmp(drv1
->protocol_name
, protocol
)) {
441 static int find_image_format(const char *filename
, BlockDriver
**pdrv
)
443 int ret
, score
, score_max
;
444 BlockDriver
*drv1
, *drv
;
446 BlockDriverState
*bs
;
448 ret
= bdrv_file_open(&bs
, filename
, 0);
454 /* Return the raw BlockDriver * to scsi-generic devices or empty drives */
455 if (bs
->sg
|| !bdrv_is_inserted(bs
)) {
457 drv
= bdrv_find_format("raw");
465 ret
= bdrv_pread(bs
, 0, buf
, sizeof(buf
));
474 QLIST_FOREACH(drv1
, &bdrv_drivers
, list
) {
475 if (drv1
->bdrv_probe
) {
476 score
= drv1
->bdrv_probe(buf
, ret
, filename
);
477 if (score
> score_max
) {
491 * Set the current 'total_sectors' value
493 static int refresh_total_sectors(BlockDriverState
*bs
, int64_t hint
)
495 BlockDriver
*drv
= bs
->drv
;
497 /* Do not attempt drv->bdrv_getlength() on scsi-generic devices */
501 /* query actual device if possible, otherwise just trust the hint */
502 if (drv
->bdrv_getlength
) {
503 int64_t length
= drv
->bdrv_getlength(bs
);
507 hint
= length
>> BDRV_SECTOR_BITS
;
510 bs
->total_sectors
= hint
;
515 * Set open flags for a given cache mode
517 * Return 0 on success, -1 if the cache mode was invalid.
519 int bdrv_parse_cache_flags(const char *mode
, int *flags
)
521 *flags
&= ~BDRV_O_CACHE_MASK
;
523 if (!strcmp(mode
, "off") || !strcmp(mode
, "none")) {
524 *flags
|= BDRV_O_NOCACHE
| BDRV_O_CACHE_WB
;
525 } else if (!strcmp(mode
, "directsync")) {
526 *flags
|= BDRV_O_NOCACHE
;
527 } else if (!strcmp(mode
, "writeback")) {
528 *flags
|= BDRV_O_CACHE_WB
;
529 } else if (!strcmp(mode
, "unsafe")) {
530 *flags
|= BDRV_O_CACHE_WB
;
531 *flags
|= BDRV_O_NO_FLUSH
;
532 } else if (!strcmp(mode
, "writethrough")) {
533 /* this is the default */
542 * Common part for opening disk images and files
544 static int bdrv_open_common(BlockDriverState
*bs
, const char *filename
,
545 int flags
, BlockDriver
*drv
)
551 trace_bdrv_open_common(bs
, filename
, flags
, drv
->format_name
);
554 bs
->total_sectors
= 0;
558 bs
->open_flags
= flags
;
560 bs
->buffer_alignment
= 512;
562 pstrcpy(bs
->filename
, sizeof(bs
->filename
), filename
);
563 bs
->backing_file
[0] = '\0';
565 if (use_bdrv_whitelist
&& !bdrv_is_whitelisted(drv
)) {
570 bs
->opaque
= g_malloc0(drv
->instance_size
);
572 bs
->enable_write_cache
= !!(flags
& BDRV_O_CACHE_WB
);
575 * Clear flags that are internal to the block layer before opening the
578 open_flags
= flags
& ~(BDRV_O_SNAPSHOT
| BDRV_O_NO_BACKING
);
581 * Snapshots should be writable.
583 if (bs
->is_temporary
) {
584 open_flags
|= BDRV_O_RDWR
;
587 bs
->keep_read_only
= bs
->read_only
= !(open_flags
& BDRV_O_RDWR
);
589 /* Open the image, either directly or using a protocol */
590 if (drv
->bdrv_file_open
) {
591 ret
= drv
->bdrv_file_open(bs
, filename
, open_flags
);
593 ret
= bdrv_file_open(&bs
->file
, filename
, open_flags
);
595 ret
= drv
->bdrv_open(bs
, open_flags
);
603 ret
= refresh_total_sectors(bs
, bs
->total_sectors
);
609 if (bs
->is_temporary
) {
617 bdrv_delete(bs
->file
);
627 * Opens a file using a protocol (file, host_device, nbd, ...)
629 int bdrv_file_open(BlockDriverState
**pbs
, const char *filename
, int flags
)
631 BlockDriverState
*bs
;
635 drv
= bdrv_find_protocol(filename
);
641 ret
= bdrv_open_common(bs
, filename
, flags
, drv
);
652 * Opens a disk image (raw, qcow2, vmdk, ...)
654 int bdrv_open(BlockDriverState
*bs
, const char *filename
, int flags
,
658 char tmp_filename
[PATH_MAX
];
660 if (flags
& BDRV_O_SNAPSHOT
) {
661 BlockDriverState
*bs1
;
664 BlockDriver
*bdrv_qcow2
;
665 QEMUOptionParameter
*options
;
666 char backing_filename
[PATH_MAX
];
668 /* if snapshot, we create a temporary backing file and open it
669 instead of opening 'filename' directly */
671 /* if there is a backing file, use it */
673 ret
= bdrv_open(bs1
, filename
, 0, drv
);
678 total_size
= bdrv_getlength(bs1
) & BDRV_SECTOR_MASK
;
680 if (bs1
->drv
&& bs1
->drv
->protocol_name
)
685 get_tmp_filename(tmp_filename
, sizeof(tmp_filename
));
687 /* Real path is meaningless for protocols */
689 snprintf(backing_filename
, sizeof(backing_filename
),
691 else if (!realpath(filename
, backing_filename
))
694 bdrv_qcow2
= bdrv_find_format("qcow2");
695 options
= parse_option_parameters("", bdrv_qcow2
->create_options
, NULL
);
697 set_option_parameter_int(options
, BLOCK_OPT_SIZE
, total_size
);
698 set_option_parameter(options
, BLOCK_OPT_BACKING_FILE
, backing_filename
);
700 set_option_parameter(options
, BLOCK_OPT_BACKING_FMT
,
704 ret
= bdrv_create(bdrv_qcow2
, tmp_filename
, options
);
705 free_option_parameters(options
);
710 filename
= tmp_filename
;
712 bs
->is_temporary
= 1;
715 /* Find the right image format driver */
717 ret
= find_image_format(filename
, &drv
);
721 goto unlink_and_fail
;
725 ret
= bdrv_open_common(bs
, filename
, flags
, drv
);
727 goto unlink_and_fail
;
730 /* If there is a backing file, use it */
731 if ((flags
& BDRV_O_NO_BACKING
) == 0 && bs
->backing_file
[0] != '\0') {
732 char backing_filename
[PATH_MAX
];
734 BlockDriver
*back_drv
= NULL
;
736 bs
->backing_hd
= bdrv_new("");
738 if (path_has_protocol(bs
->backing_file
)) {
739 pstrcpy(backing_filename
, sizeof(backing_filename
),
742 path_combine(backing_filename
, sizeof(backing_filename
),
743 filename
, bs
->backing_file
);
746 if (bs
->backing_format
[0] != '\0') {
747 back_drv
= bdrv_find_format(bs
->backing_format
);
750 /* backing files always opened read-only */
752 flags
& ~(BDRV_O_RDWR
| BDRV_O_SNAPSHOT
| BDRV_O_NO_BACKING
);
754 ret
= bdrv_open(bs
->backing_hd
, backing_filename
, back_flags
, back_drv
);
759 if (bs
->is_temporary
) {
760 bs
->backing_hd
->keep_read_only
= !(flags
& BDRV_O_RDWR
);
762 /* base image inherits from "parent" */
763 bs
->backing_hd
->keep_read_only
= bs
->keep_read_only
;
767 if (!bdrv_key_required(bs
)) {
768 bdrv_dev_change_media_cb(bs
, true);
771 /* throttling disk I/O limits */
772 if (bs
->io_limits_enabled
) {
773 bdrv_io_limits_enable(bs
);
779 if (bs
->is_temporary
) {
785 void bdrv_close(BlockDriverState
*bs
)
788 if (bs
== bs_snapshots
) {
791 if (bs
->backing_hd
) {
792 bdrv_delete(bs
->backing_hd
);
793 bs
->backing_hd
= NULL
;
795 bs
->drv
->bdrv_close(bs
);
798 if (bs
->is_temporary
) {
799 unlink(bs
->filename
);
805 if (bs
->file
!= NULL
) {
806 bdrv_close(bs
->file
);
809 bdrv_dev_change_media_cb(bs
, false);
812 /*throttling disk I/O limits*/
813 if (bs
->io_limits_enabled
) {
814 bdrv_io_limits_disable(bs
);
818 void bdrv_close_all(void)
820 BlockDriverState
*bs
;
822 QTAILQ_FOREACH(bs
, &bdrv_states
, list
) {
827 /* make a BlockDriverState anonymous by removing from bdrv_state list.
828 Also, NULL terminate the device_name to prevent double remove */
829 void bdrv_make_anon(BlockDriverState
*bs
)
831 if (bs
->device_name
[0] != '\0') {
832 QTAILQ_REMOVE(&bdrv_states
, bs
, list
);
834 bs
->device_name
[0] = '\0';
837 void bdrv_delete(BlockDriverState
*bs
)
841 /* remove from list, if necessary */
845 if (bs
->file
!= NULL
) {
846 bdrv_delete(bs
->file
);
849 assert(bs
!= bs_snapshots
);
853 int bdrv_attach_dev(BlockDriverState
*bs
, void *dev
)
854 /* TODO change to DeviceState *dev when all users are qdevified */
860 bdrv_iostatus_reset(bs
);
864 /* TODO qdevified devices don't use this, remove when devices are qdevified */
865 void bdrv_attach_dev_nofail(BlockDriverState
*bs
, void *dev
)
867 if (bdrv_attach_dev(bs
, dev
) < 0) {
872 void bdrv_detach_dev(BlockDriverState
*bs
, void *dev
)
873 /* TODO change to DeviceState *dev when all users are qdevified */
875 assert(bs
->dev
== dev
);
878 bs
->dev_opaque
= NULL
;
879 bs
->buffer_alignment
= 512;
882 /* TODO change to return DeviceState * when all users are qdevified */
883 void *bdrv_get_attached_dev(BlockDriverState
*bs
)
888 void bdrv_set_dev_ops(BlockDriverState
*bs
, const BlockDevOps
*ops
,
892 bs
->dev_opaque
= opaque
;
893 if (bdrv_dev_has_removable_media(bs
) && bs
== bs_snapshots
) {
898 static void bdrv_dev_change_media_cb(BlockDriverState
*bs
, bool load
)
900 if (bs
->dev_ops
&& bs
->dev_ops
->change_media_cb
) {
901 bs
->dev_ops
->change_media_cb(bs
->dev_opaque
, load
);
905 bool bdrv_dev_has_removable_media(BlockDriverState
*bs
)
907 return !bs
->dev
|| (bs
->dev_ops
&& bs
->dev_ops
->change_media_cb
);
910 void bdrv_dev_eject_request(BlockDriverState
*bs
, bool force
)
912 if (bs
->dev_ops
&& bs
->dev_ops
->eject_request_cb
) {
913 bs
->dev_ops
->eject_request_cb(bs
->dev_opaque
, force
);
917 bool bdrv_dev_is_tray_open(BlockDriverState
*bs
)
919 if (bs
->dev_ops
&& bs
->dev_ops
->is_tray_open
) {
920 return bs
->dev_ops
->is_tray_open(bs
->dev_opaque
);
925 static void bdrv_dev_resize_cb(BlockDriverState
*bs
)
927 if (bs
->dev_ops
&& bs
->dev_ops
->resize_cb
) {
928 bs
->dev_ops
->resize_cb(bs
->dev_opaque
);
932 bool bdrv_dev_is_medium_locked(BlockDriverState
*bs
)
934 if (bs
->dev_ops
&& bs
->dev_ops
->is_medium_locked
) {
935 return bs
->dev_ops
->is_medium_locked(bs
->dev_opaque
);
941 * Run consistency checks on an image
943 * Returns 0 if the check could be completed (it doesn't mean that the image is
944 * free of errors) or -errno when an internal error occurred. The results of the
945 * check are stored in res.
947 int bdrv_check(BlockDriverState
*bs
, BdrvCheckResult
*res
)
949 if (bs
->drv
->bdrv_check
== NULL
) {
953 memset(res
, 0, sizeof(*res
));
954 return bs
->drv
->bdrv_check(bs
, res
);
957 #define COMMIT_BUF_SECTORS 2048
959 /* commit COW file into the raw image */
960 int bdrv_commit(BlockDriverState
*bs
)
962 BlockDriver
*drv
= bs
->drv
;
963 BlockDriver
*backing_drv
;
964 int64_t sector
, total_sectors
;
965 int n
, ro
, open_flags
;
966 int ret
= 0, rw_ret
= 0;
969 BlockDriverState
*bs_rw
, *bs_ro
;
974 if (!bs
->backing_hd
) {
978 if (bs
->backing_hd
->keep_read_only
) {
982 backing_drv
= bs
->backing_hd
->drv
;
983 ro
= bs
->backing_hd
->read_only
;
984 strncpy(filename
, bs
->backing_hd
->filename
, sizeof(filename
));
985 open_flags
= bs
->backing_hd
->open_flags
;
989 bdrv_delete(bs
->backing_hd
);
990 bs
->backing_hd
= NULL
;
991 bs_rw
= bdrv_new("");
992 rw_ret
= bdrv_open(bs_rw
, filename
, open_flags
| BDRV_O_RDWR
,
996 /* try to re-open read-only */
997 bs_ro
= bdrv_new("");
998 ret
= bdrv_open(bs_ro
, filename
, open_flags
& ~BDRV_O_RDWR
,
1002 /* drive not functional anymore */
1006 bs
->backing_hd
= bs_ro
;
1009 bs
->backing_hd
= bs_rw
;
1012 total_sectors
= bdrv_getlength(bs
) >> BDRV_SECTOR_BITS
;
1013 buf
= g_malloc(COMMIT_BUF_SECTORS
* BDRV_SECTOR_SIZE
);
1015 for (sector
= 0; sector
< total_sectors
; sector
+= n
) {
1016 if (bdrv_is_allocated(bs
, sector
, COMMIT_BUF_SECTORS
, &n
)) {
1018 if (bdrv_read(bs
, sector
, buf
, n
) != 0) {
1023 if (bdrv_write(bs
->backing_hd
, sector
, buf
, n
) != 0) {
1030 if (drv
->bdrv_make_empty
) {
1031 ret
= drv
->bdrv_make_empty(bs
);
1036 * Make sure all data we wrote to the backing device is actually
1040 bdrv_flush(bs
->backing_hd
);
1047 bdrv_delete(bs
->backing_hd
);
1048 bs
->backing_hd
= NULL
;
1049 bs_ro
= bdrv_new("");
1050 ret
= bdrv_open(bs_ro
, filename
, open_flags
& ~BDRV_O_RDWR
,
1054 /* drive not functional anymore */
1058 bs
->backing_hd
= bs_ro
;
1059 bs
->backing_hd
->keep_read_only
= 0;
1065 void bdrv_commit_all(void)
1067 BlockDriverState
*bs
;
1069 QTAILQ_FOREACH(bs
, &bdrv_states
, list
) {
1077 * -EINVAL - backing format specified, but no file
1078 * -ENOSPC - can't update the backing file because no space is left in the
1080 * -ENOTSUP - format driver doesn't support changing the backing file
1082 int bdrv_change_backing_file(BlockDriverState
*bs
,
1083 const char *backing_file
, const char *backing_fmt
)
1085 BlockDriver
*drv
= bs
->drv
;
1087 if (drv
->bdrv_change_backing_file
!= NULL
) {
1088 return drv
->bdrv_change_backing_file(bs
, backing_file
, backing_fmt
);
1094 static int bdrv_check_byte_request(BlockDriverState
*bs
, int64_t offset
,
1099 if (!bdrv_is_inserted(bs
))
1105 len
= bdrv_getlength(bs
);
1110 if ((offset
> len
) || (len
- offset
< size
))
1116 static int bdrv_check_request(BlockDriverState
*bs
, int64_t sector_num
,
1119 return bdrv_check_byte_request(bs
, sector_num
* BDRV_SECTOR_SIZE
,
1120 nb_sectors
* BDRV_SECTOR_SIZE
);
1123 typedef struct RwCo
{
1124 BlockDriverState
*bs
;
1132 static void coroutine_fn
bdrv_rw_co_entry(void *opaque
)
1134 RwCo
*rwco
= opaque
;
1136 if (!rwco
->is_write
) {
1137 rwco
->ret
= bdrv_co_do_readv(rwco
->bs
, rwco
->sector_num
,
1138 rwco
->nb_sectors
, rwco
->qiov
);
1140 rwco
->ret
= bdrv_co_do_writev(rwco
->bs
, rwco
->sector_num
,
1141 rwco
->nb_sectors
, rwco
->qiov
);
1146 * Process a synchronous request using coroutines
1148 static int bdrv_rw_co(BlockDriverState
*bs
, int64_t sector_num
, uint8_t *buf
,
1149 int nb_sectors
, bool is_write
)
1152 struct iovec iov
= {
1153 .iov_base
= (void *)buf
,
1154 .iov_len
= nb_sectors
* BDRV_SECTOR_SIZE
,
1159 .sector_num
= sector_num
,
1160 .nb_sectors
= nb_sectors
,
1162 .is_write
= is_write
,
1166 qemu_iovec_init_external(&qiov
, &iov
, 1);
1168 if (qemu_in_coroutine()) {
1169 /* Fast-path if already in coroutine context */
1170 bdrv_rw_co_entry(&rwco
);
1172 co
= qemu_coroutine_create(bdrv_rw_co_entry
);
1173 qemu_coroutine_enter(co
, &rwco
);
1174 while (rwco
.ret
== NOT_DONE
) {
1181 /* return < 0 if error. See bdrv_write() for the return codes */
1182 int bdrv_read(BlockDriverState
*bs
, int64_t sector_num
,
1183 uint8_t *buf
, int nb_sectors
)
1185 return bdrv_rw_co(bs
, sector_num
, buf
, nb_sectors
, false);
1188 static void set_dirty_bitmap(BlockDriverState
*bs
, int64_t sector_num
,
1189 int nb_sectors
, int dirty
)
1192 unsigned long val
, idx
, bit
;
1194 start
= sector_num
/ BDRV_SECTORS_PER_DIRTY_CHUNK
;
1195 end
= (sector_num
+ nb_sectors
- 1) / BDRV_SECTORS_PER_DIRTY_CHUNK
;
1197 for (; start
<= end
; start
++) {
1198 idx
= start
/ (sizeof(unsigned long) * 8);
1199 bit
= start
% (sizeof(unsigned long) * 8);
1200 val
= bs
->dirty_bitmap
[idx
];
1202 if (!(val
& (1UL << bit
))) {
1207 if (val
& (1UL << bit
)) {
1209 val
&= ~(1UL << bit
);
1212 bs
->dirty_bitmap
[idx
] = val
;
1216 /* Return < 0 if error. Important errors are:
1217 -EIO generic I/O error (may happen for all errors)
1218 -ENOMEDIUM No media inserted.
1219 -EINVAL Invalid sector number or nb_sectors
1220 -EACCES Trying to write a read-only device
1222 int bdrv_write(BlockDriverState
*bs
, int64_t sector_num
,
1223 const uint8_t *buf
, int nb_sectors
)
1225 return bdrv_rw_co(bs
, sector_num
, (uint8_t *)buf
, nb_sectors
, true);
1228 int bdrv_pread(BlockDriverState
*bs
, int64_t offset
,
1229 void *buf
, int count1
)
1231 uint8_t tmp_buf
[BDRV_SECTOR_SIZE
];
1232 int len
, nb_sectors
, count
;
1237 /* first read to align to sector start */
1238 len
= (BDRV_SECTOR_SIZE
- offset
) & (BDRV_SECTOR_SIZE
- 1);
1241 sector_num
= offset
>> BDRV_SECTOR_BITS
;
1243 if ((ret
= bdrv_read(bs
, sector_num
, tmp_buf
, 1)) < 0)
1245 memcpy(buf
, tmp_buf
+ (offset
& (BDRV_SECTOR_SIZE
- 1)), len
);
1253 /* read the sectors "in place" */
1254 nb_sectors
= count
>> BDRV_SECTOR_BITS
;
1255 if (nb_sectors
> 0) {
1256 if ((ret
= bdrv_read(bs
, sector_num
, buf
, nb_sectors
)) < 0)
1258 sector_num
+= nb_sectors
;
1259 len
= nb_sectors
<< BDRV_SECTOR_BITS
;
1264 /* add data from the last sector */
1266 if ((ret
= bdrv_read(bs
, sector_num
, tmp_buf
, 1)) < 0)
1268 memcpy(buf
, tmp_buf
, count
);
1273 int bdrv_pwrite(BlockDriverState
*bs
, int64_t offset
,
1274 const void *buf
, int count1
)
1276 uint8_t tmp_buf
[BDRV_SECTOR_SIZE
];
1277 int len
, nb_sectors
, count
;
1282 /* first write to align to sector start */
1283 len
= (BDRV_SECTOR_SIZE
- offset
) & (BDRV_SECTOR_SIZE
- 1);
1286 sector_num
= offset
>> BDRV_SECTOR_BITS
;
1288 if ((ret
= bdrv_read(bs
, sector_num
, tmp_buf
, 1)) < 0)
1290 memcpy(tmp_buf
+ (offset
& (BDRV_SECTOR_SIZE
- 1)), buf
, len
);
1291 if ((ret
= bdrv_write(bs
, sector_num
, tmp_buf
, 1)) < 0)
1300 /* write the sectors "in place" */
1301 nb_sectors
= count
>> BDRV_SECTOR_BITS
;
1302 if (nb_sectors
> 0) {
1303 if ((ret
= bdrv_write(bs
, sector_num
, buf
, nb_sectors
)) < 0)
1305 sector_num
+= nb_sectors
;
1306 len
= nb_sectors
<< BDRV_SECTOR_BITS
;
1311 /* add data from the last sector */
1313 if ((ret
= bdrv_read(bs
, sector_num
, tmp_buf
, 1)) < 0)
1315 memcpy(tmp_buf
, buf
, count
);
1316 if ((ret
= bdrv_write(bs
, sector_num
, tmp_buf
, 1)) < 0)
1323 * Writes to the file and ensures that no writes are reordered across this
1324 * request (acts as a barrier)
1326 * Returns 0 on success, -errno in error cases.
1328 int bdrv_pwrite_sync(BlockDriverState
*bs
, int64_t offset
,
1329 const void *buf
, int count
)
1333 ret
= bdrv_pwrite(bs
, offset
, buf
, count
);
1338 /* No flush needed for cache modes that use O_DSYNC */
1339 if ((bs
->open_flags
& BDRV_O_CACHE_WB
) != 0) {
1347 * Handle a read request in coroutine context
1349 static int coroutine_fn
bdrv_co_do_readv(BlockDriverState
*bs
,
1350 int64_t sector_num
, int nb_sectors
, QEMUIOVector
*qiov
)
1352 BlockDriver
*drv
= bs
->drv
;
1357 if (bdrv_check_request(bs
, sector_num
, nb_sectors
)) {
1361 /* throttling disk read I/O */
1362 if (bs
->io_limits_enabled
) {
1363 bdrv_io_limits_intercept(bs
, false, nb_sectors
);
1366 return drv
->bdrv_co_readv(bs
, sector_num
, nb_sectors
, qiov
);
1369 int coroutine_fn
bdrv_co_readv(BlockDriverState
*bs
, int64_t sector_num
,
1370 int nb_sectors
, QEMUIOVector
*qiov
)
1372 trace_bdrv_co_readv(bs
, sector_num
, nb_sectors
);
1374 return bdrv_co_do_readv(bs
, sector_num
, nb_sectors
, qiov
);
1378 * Handle a write request in coroutine context
1380 static int coroutine_fn
bdrv_co_do_writev(BlockDriverState
*bs
,
1381 int64_t sector_num
, int nb_sectors
, QEMUIOVector
*qiov
)
1383 BlockDriver
*drv
= bs
->drv
;
1389 if (bs
->read_only
) {
1392 if (bdrv_check_request(bs
, sector_num
, nb_sectors
)) {
1396 /* throttling disk write I/O */
1397 if (bs
->io_limits_enabled
) {
1398 bdrv_io_limits_intercept(bs
, true, nb_sectors
);
1401 ret
= drv
->bdrv_co_writev(bs
, sector_num
, nb_sectors
, qiov
);
1403 if (bs
->dirty_bitmap
) {
1404 set_dirty_bitmap(bs
, sector_num
, nb_sectors
, 1);
1407 if (bs
->wr_highest_sector
< sector_num
+ nb_sectors
- 1) {
1408 bs
->wr_highest_sector
= sector_num
+ nb_sectors
- 1;
1414 int coroutine_fn
bdrv_co_writev(BlockDriverState
*bs
, int64_t sector_num
,
1415 int nb_sectors
, QEMUIOVector
*qiov
)
1417 trace_bdrv_co_writev(bs
, sector_num
, nb_sectors
);
1419 return bdrv_co_do_writev(bs
, sector_num
, nb_sectors
, qiov
);
1423 * Truncate file to 'offset' bytes (needed only for file protocols)
1425 int bdrv_truncate(BlockDriverState
*bs
, int64_t offset
)
1427 BlockDriver
*drv
= bs
->drv
;
1431 if (!drv
->bdrv_truncate
)
1435 if (bdrv_in_use(bs
))
1437 ret
= drv
->bdrv_truncate(bs
, offset
);
1439 ret
= refresh_total_sectors(bs
, offset
>> BDRV_SECTOR_BITS
);
1440 bdrv_dev_resize_cb(bs
);
1446 * Length of a allocated file in bytes. Sparse files are counted by actual
1447 * allocated space. Return < 0 if error or unknown.
1449 int64_t bdrv_get_allocated_file_size(BlockDriverState
*bs
)
1451 BlockDriver
*drv
= bs
->drv
;
1455 if (drv
->bdrv_get_allocated_file_size
) {
1456 return drv
->bdrv_get_allocated_file_size(bs
);
1459 return bdrv_get_allocated_file_size(bs
->file
);
1465 * Length of a file in bytes. Return < 0 if error or unknown.
1467 int64_t bdrv_getlength(BlockDriverState
*bs
)
1469 BlockDriver
*drv
= bs
->drv
;
1473 if (bs
->growable
|| bdrv_dev_has_removable_media(bs
)) {
1474 if (drv
->bdrv_getlength
) {
1475 return drv
->bdrv_getlength(bs
);
1478 return bs
->total_sectors
* BDRV_SECTOR_SIZE
;
1481 /* return 0 as number of sectors if no device present or error */
1482 void bdrv_get_geometry(BlockDriverState
*bs
, uint64_t *nb_sectors_ptr
)
1485 length
= bdrv_getlength(bs
);
1489 length
= length
>> BDRV_SECTOR_BITS
;
1490 *nb_sectors_ptr
= length
;
1494 uint8_t boot_ind
; /* 0x80 - active */
1495 uint8_t head
; /* starting head */
1496 uint8_t sector
; /* starting sector */
1497 uint8_t cyl
; /* starting cylinder */
1498 uint8_t sys_ind
; /* What partition type */
1499 uint8_t end_head
; /* end head */
1500 uint8_t end_sector
; /* end sector */
1501 uint8_t end_cyl
; /* end cylinder */
1502 uint32_t start_sect
; /* starting sector counting from 0 */
1503 uint32_t nr_sects
; /* nr of sectors in partition */
1506 /* try to guess the disk logical geometry from the MSDOS partition table. Return 0 if OK, -1 if could not guess */
1507 static int guess_disk_lchs(BlockDriverState
*bs
,
1508 int *pcylinders
, int *pheads
, int *psectors
)
1510 uint8_t buf
[BDRV_SECTOR_SIZE
];
1511 int ret
, i
, heads
, sectors
, cylinders
;
1512 struct partition
*p
;
1514 uint64_t nb_sectors
;
1516 bdrv_get_geometry(bs
, &nb_sectors
);
1518 ret
= bdrv_read(bs
, 0, buf
, 1);
1521 /* test msdos magic */
1522 if (buf
[510] != 0x55 || buf
[511] != 0xaa)
1524 for(i
= 0; i
< 4; i
++) {
1525 p
= ((struct partition
*)(buf
+ 0x1be)) + i
;
1526 nr_sects
= le32_to_cpu(p
->nr_sects
);
1527 if (nr_sects
&& p
->end_head
) {
1528 /* We make the assumption that the partition terminates on
1529 a cylinder boundary */
1530 heads
= p
->end_head
+ 1;
1531 sectors
= p
->end_sector
& 63;
1534 cylinders
= nb_sectors
/ (heads
* sectors
);
1535 if (cylinders
< 1 || cylinders
> 16383)
1538 *psectors
= sectors
;
1539 *pcylinders
= cylinders
;
1541 printf("guessed geometry: LCHS=%d %d %d\n",
1542 cylinders
, heads
, sectors
);
1550 void bdrv_guess_geometry(BlockDriverState
*bs
, int *pcyls
, int *pheads
, int *psecs
)
1552 int translation
, lba_detected
= 0;
1553 int cylinders
, heads
, secs
;
1554 uint64_t nb_sectors
;
1556 /* if a geometry hint is available, use it */
1557 bdrv_get_geometry(bs
, &nb_sectors
);
1558 bdrv_get_geometry_hint(bs
, &cylinders
, &heads
, &secs
);
1559 translation
= bdrv_get_translation_hint(bs
);
1560 if (cylinders
!= 0) {
1565 if (guess_disk_lchs(bs
, &cylinders
, &heads
, &secs
) == 0) {
1567 /* if heads > 16, it means that a BIOS LBA
1568 translation was active, so the default
1569 hardware geometry is OK */
1571 goto default_geometry
;
1576 /* disable any translation to be in sync with
1577 the logical geometry */
1578 if (translation
== BIOS_ATA_TRANSLATION_AUTO
) {
1579 bdrv_set_translation_hint(bs
,
1580 BIOS_ATA_TRANSLATION_NONE
);
1585 /* if no geometry, use a standard physical disk geometry */
1586 cylinders
= nb_sectors
/ (16 * 63);
1588 if (cylinders
> 16383)
1590 else if (cylinders
< 2)
1595 if ((lba_detected
== 1) && (translation
== BIOS_ATA_TRANSLATION_AUTO
)) {
1596 if ((*pcyls
* *pheads
) <= 131072) {
1597 bdrv_set_translation_hint(bs
,
1598 BIOS_ATA_TRANSLATION_LARGE
);
1600 bdrv_set_translation_hint(bs
,
1601 BIOS_ATA_TRANSLATION_LBA
);
1605 bdrv_set_geometry_hint(bs
, *pcyls
, *pheads
, *psecs
);
1609 void bdrv_set_geometry_hint(BlockDriverState
*bs
,
1610 int cyls
, int heads
, int secs
)
1617 void bdrv_set_translation_hint(BlockDriverState
*bs
, int translation
)
1619 bs
->translation
= translation
;
1622 void bdrv_get_geometry_hint(BlockDriverState
*bs
,
1623 int *pcyls
, int *pheads
, int *psecs
)
1626 *pheads
= bs
->heads
;
1630 /* throttling disk io limits */
1631 void bdrv_set_io_limits(BlockDriverState
*bs
,
1632 BlockIOLimit
*io_limits
)
1634 bs
->io_limits
= *io_limits
;
1635 bs
->io_limits_enabled
= bdrv_io_limits_enabled(bs
);
1638 /* Recognize floppy formats */
1639 typedef struct FDFormat
{
1646 static const FDFormat fd_formats
[] = {
1647 /* First entry is default format */
1648 /* 1.44 MB 3"1/2 floppy disks */
1649 { FDRIVE_DRV_144
, 18, 80, 1, },
1650 { FDRIVE_DRV_144
, 20, 80, 1, },
1651 { FDRIVE_DRV_144
, 21, 80, 1, },
1652 { FDRIVE_DRV_144
, 21, 82, 1, },
1653 { FDRIVE_DRV_144
, 21, 83, 1, },
1654 { FDRIVE_DRV_144
, 22, 80, 1, },
1655 { FDRIVE_DRV_144
, 23, 80, 1, },
1656 { FDRIVE_DRV_144
, 24, 80, 1, },
1657 /* 2.88 MB 3"1/2 floppy disks */
1658 { FDRIVE_DRV_288
, 36, 80, 1, },
1659 { FDRIVE_DRV_288
, 39, 80, 1, },
1660 { FDRIVE_DRV_288
, 40, 80, 1, },
1661 { FDRIVE_DRV_288
, 44, 80, 1, },
1662 { FDRIVE_DRV_288
, 48, 80, 1, },
1663 /* 720 kB 3"1/2 floppy disks */
1664 { FDRIVE_DRV_144
, 9, 80, 1, },
1665 { FDRIVE_DRV_144
, 10, 80, 1, },
1666 { FDRIVE_DRV_144
, 10, 82, 1, },
1667 { FDRIVE_DRV_144
, 10, 83, 1, },
1668 { FDRIVE_DRV_144
, 13, 80, 1, },
1669 { FDRIVE_DRV_144
, 14, 80, 1, },
1670 /* 1.2 MB 5"1/4 floppy disks */
1671 { FDRIVE_DRV_120
, 15, 80, 1, },
1672 { FDRIVE_DRV_120
, 18, 80, 1, },
1673 { FDRIVE_DRV_120
, 18, 82, 1, },
1674 { FDRIVE_DRV_120
, 18, 83, 1, },
1675 { FDRIVE_DRV_120
, 20, 80, 1, },
1676 /* 720 kB 5"1/4 floppy disks */
1677 { FDRIVE_DRV_120
, 9, 80, 1, },
1678 { FDRIVE_DRV_120
, 11, 80, 1, },
1679 /* 360 kB 5"1/4 floppy disks */
1680 { FDRIVE_DRV_120
, 9, 40, 1, },
1681 { FDRIVE_DRV_120
, 9, 40, 0, },
1682 { FDRIVE_DRV_120
, 10, 41, 1, },
1683 { FDRIVE_DRV_120
, 10, 42, 1, },
1684 /* 320 kB 5"1/4 floppy disks */
1685 { FDRIVE_DRV_120
, 8, 40, 1, },
1686 { FDRIVE_DRV_120
, 8, 40, 0, },
1687 /* 360 kB must match 5"1/4 better than 3"1/2... */
1688 { FDRIVE_DRV_144
, 9, 80, 0, },
1690 { FDRIVE_DRV_NONE
, -1, -1, 0, },
1693 void bdrv_get_floppy_geometry_hint(BlockDriverState
*bs
, int *nb_heads
,
1694 int *max_track
, int *last_sect
,
1695 FDriveType drive_in
, FDriveType
*drive
)
1697 const FDFormat
*parse
;
1698 uint64_t nb_sectors
, size
;
1699 int i
, first_match
, match
;
1701 bdrv_get_geometry_hint(bs
, nb_heads
, max_track
, last_sect
);
1702 if (*nb_heads
!= 0 && *max_track
!= 0 && *last_sect
!= 0) {
1703 /* User defined disk */
1705 bdrv_get_geometry(bs
, &nb_sectors
);
1708 for (i
= 0; ; i
++) {
1709 parse
= &fd_formats
[i
];
1710 if (parse
->drive
== FDRIVE_DRV_NONE
) {
1713 if (drive_in
== parse
->drive
||
1714 drive_in
== FDRIVE_DRV_NONE
) {
1715 size
= (parse
->max_head
+ 1) * parse
->max_track
*
1717 if (nb_sectors
== size
) {
1721 if (first_match
== -1) {
1727 if (first_match
== -1) {
1730 match
= first_match
;
1732 parse
= &fd_formats
[match
];
1734 *nb_heads
= parse
->max_head
+ 1;
1735 *max_track
= parse
->max_track
;
1736 *last_sect
= parse
->last_sect
;
1737 *drive
= parse
->drive
;
1741 int bdrv_get_translation_hint(BlockDriverState
*bs
)
1743 return bs
->translation
;
1746 void bdrv_set_on_error(BlockDriverState
*bs
, BlockErrorAction on_read_error
,
1747 BlockErrorAction on_write_error
)
1749 bs
->on_read_error
= on_read_error
;
1750 bs
->on_write_error
= on_write_error
;
1753 BlockErrorAction
bdrv_get_on_error(BlockDriverState
*bs
, int is_read
)
1755 return is_read
? bs
->on_read_error
: bs
->on_write_error
;
1758 int bdrv_is_read_only(BlockDriverState
*bs
)
1760 return bs
->read_only
;
1763 int bdrv_is_sg(BlockDriverState
*bs
)
1768 int bdrv_enable_write_cache(BlockDriverState
*bs
)
1770 return bs
->enable_write_cache
;
1773 int bdrv_is_encrypted(BlockDriverState
*bs
)
1775 if (bs
->backing_hd
&& bs
->backing_hd
->encrypted
)
1777 return bs
->encrypted
;
1780 int bdrv_key_required(BlockDriverState
*bs
)
1782 BlockDriverState
*backing_hd
= bs
->backing_hd
;
1784 if (backing_hd
&& backing_hd
->encrypted
&& !backing_hd
->valid_key
)
1786 return (bs
->encrypted
&& !bs
->valid_key
);
1789 int bdrv_set_key(BlockDriverState
*bs
, const char *key
)
1792 if (bs
->backing_hd
&& bs
->backing_hd
->encrypted
) {
1793 ret
= bdrv_set_key(bs
->backing_hd
, key
);
1799 if (!bs
->encrypted
) {
1801 } else if (!bs
->drv
|| !bs
->drv
->bdrv_set_key
) {
1804 ret
= bs
->drv
->bdrv_set_key(bs
, key
);
1807 } else if (!bs
->valid_key
) {
1809 /* call the change callback now, we skipped it on open */
1810 bdrv_dev_change_media_cb(bs
, true);
1815 void bdrv_get_format(BlockDriverState
*bs
, char *buf
, int buf_size
)
1820 pstrcpy(buf
, buf_size
, bs
->drv
->format_name
);
1824 void bdrv_iterate_format(void (*it
)(void *opaque
, const char *name
),
1829 QLIST_FOREACH(drv
, &bdrv_drivers
, list
) {
1830 it(opaque
, drv
->format_name
);
1834 BlockDriverState
*bdrv_find(const char *name
)
1836 BlockDriverState
*bs
;
1838 QTAILQ_FOREACH(bs
, &bdrv_states
, list
) {
1839 if (!strcmp(name
, bs
->device_name
)) {
1846 BlockDriverState
*bdrv_next(BlockDriverState
*bs
)
1849 return QTAILQ_FIRST(&bdrv_states
);
1851 return QTAILQ_NEXT(bs
, list
);
1854 void bdrv_iterate(void (*it
)(void *opaque
, BlockDriverState
*bs
), void *opaque
)
1856 BlockDriverState
*bs
;
1858 QTAILQ_FOREACH(bs
, &bdrv_states
, list
) {
1863 const char *bdrv_get_device_name(BlockDriverState
*bs
)
1865 return bs
->device_name
;
1868 void bdrv_flush_all(void)
1870 BlockDriverState
*bs
;
1872 QTAILQ_FOREACH(bs
, &bdrv_states
, list
) {
1873 if (!bdrv_is_read_only(bs
) && bdrv_is_inserted(bs
)) {
1879 int bdrv_has_zero_init(BlockDriverState
*bs
)
1883 if (bs
->drv
->bdrv_has_zero_init
) {
1884 return bs
->drv
->bdrv_has_zero_init(bs
);
1890 typedef struct BdrvCoIsAllocatedData
{
1891 BlockDriverState
*bs
;
1897 } BdrvCoIsAllocatedData
;
1899 /* Coroutine wrapper for bdrv_is_allocated() */
1900 static void coroutine_fn
bdrv_is_allocated_co_entry(void *opaque
)
1902 BdrvCoIsAllocatedData
*data
= opaque
;
1903 BlockDriverState
*bs
= data
->bs
;
1905 data
->ret
= bs
->drv
->bdrv_co_is_allocated(bs
, data
->sector_num
,
1906 data
->nb_sectors
, data
->pnum
);
1911 * Returns true iff the specified sector is present in the disk image. Drivers
1912 * not implementing the functionality are assumed to not support backing files,
1913 * hence all their sectors are reported as allocated.
1915 * 'pnum' is set to the number of sectors (including and immediately following
1916 * the specified sector) that are known to be in the same
1917 * allocated/unallocated state.
1919 * 'nb_sectors' is the max value 'pnum' should be set to.
1921 int bdrv_is_allocated(BlockDriverState
*bs
, int64_t sector_num
, int nb_sectors
,
1925 if (bs
->drv
->bdrv_co_is_allocated
) {
1927 BdrvCoIsAllocatedData data
= {
1929 .sector_num
= sector_num
,
1930 .nb_sectors
= nb_sectors
,
1935 co
= qemu_coroutine_create(bdrv_is_allocated_co_entry
);
1936 qemu_coroutine_enter(co
, &data
);
1937 while (!data
.done
) {
1942 if (!bs
->drv
->bdrv_is_allocated
) {
1943 if (sector_num
>= bs
->total_sectors
) {
1947 n
= bs
->total_sectors
- sector_num
;
1948 *pnum
= (n
< nb_sectors
) ? (n
) : (nb_sectors
);
1951 return bs
->drv
->bdrv_is_allocated(bs
, sector_num
, nb_sectors
, pnum
);
1954 void bdrv_mon_event(const BlockDriverState
*bdrv
,
1955 BlockMonEventAction action
, int is_read
)
1958 const char *action_str
;
1961 case BDRV_ACTION_REPORT
:
1962 action_str
= "report";
1964 case BDRV_ACTION_IGNORE
:
1965 action_str
= "ignore";
1967 case BDRV_ACTION_STOP
:
1968 action_str
= "stop";
1974 data
= qobject_from_jsonf("{ 'device': %s, 'action': %s, 'operation': %s }",
1977 is_read
? "read" : "write");
1978 monitor_protocol_event(QEVENT_BLOCK_IO_ERROR
, data
);
1980 qobject_decref(data
);
1983 BlockInfoList
*qmp_query_block(Error
**errp
)
1985 BlockInfoList
*head
= NULL
, *cur_item
= NULL
;
1986 BlockDriverState
*bs
;
1988 QTAILQ_FOREACH(bs
, &bdrv_states
, list
) {
1989 BlockInfoList
*info
= g_malloc0(sizeof(*info
));
1991 info
->value
= g_malloc0(sizeof(*info
->value
));
1992 info
->value
->device
= g_strdup(bs
->device_name
);
1993 info
->value
->type
= g_strdup("unknown");
1994 info
->value
->locked
= bdrv_dev_is_medium_locked(bs
);
1995 info
->value
->removable
= bdrv_dev_has_removable_media(bs
);
1997 if (bdrv_dev_has_removable_media(bs
)) {
1998 info
->value
->has_tray_open
= true;
1999 info
->value
->tray_open
= bdrv_dev_is_tray_open(bs
);
2002 if (bdrv_iostatus_is_enabled(bs
)) {
2003 info
->value
->has_io_status
= true;
2004 info
->value
->io_status
= bs
->iostatus
;
2008 info
->value
->has_inserted
= true;
2009 info
->value
->inserted
= g_malloc0(sizeof(*info
->value
->inserted
));
2010 info
->value
->inserted
->file
= g_strdup(bs
->filename
);
2011 info
->value
->inserted
->ro
= bs
->read_only
;
2012 info
->value
->inserted
->drv
= g_strdup(bs
->drv
->format_name
);
2013 info
->value
->inserted
->encrypted
= bs
->encrypted
;
2014 if (bs
->backing_file
[0]) {
2015 info
->value
->inserted
->has_backing_file
= true;
2016 info
->value
->inserted
->backing_file
= g_strdup(bs
->backing_file
);
2019 if (bs
->io_limits_enabled
) {
2020 info
->value
->inserted
->bps
=
2021 bs
->io_limits
.bps
[BLOCK_IO_LIMIT_TOTAL
];
2022 info
->value
->inserted
->bps_rd
=
2023 bs
->io_limits
.bps
[BLOCK_IO_LIMIT_READ
];
2024 info
->value
->inserted
->bps_wr
=
2025 bs
->io_limits
.bps
[BLOCK_IO_LIMIT_WRITE
];
2026 info
->value
->inserted
->iops
=
2027 bs
->io_limits
.iops
[BLOCK_IO_LIMIT_TOTAL
];
2028 info
->value
->inserted
->iops_rd
=
2029 bs
->io_limits
.iops
[BLOCK_IO_LIMIT_READ
];
2030 info
->value
->inserted
->iops_wr
=
2031 bs
->io_limits
.iops
[BLOCK_IO_LIMIT_WRITE
];
2035 /* XXX: waiting for the qapi to support GSList */
2037 head
= cur_item
= info
;
2039 cur_item
->next
= info
;
2047 /* Consider exposing this as a full fledged QMP command */
2048 static BlockStats
*qmp_query_blockstat(const BlockDriverState
*bs
, Error
**errp
)
2052 s
= g_malloc0(sizeof(*s
));
2054 if (bs
->device_name
[0]) {
2055 s
->has_device
= true;
2056 s
->device
= g_strdup(bs
->device_name
);
2059 s
->stats
= g_malloc0(sizeof(*s
->stats
));
2060 s
->stats
->rd_bytes
= bs
->nr_bytes
[BDRV_ACCT_READ
];
2061 s
->stats
->wr_bytes
= bs
->nr_bytes
[BDRV_ACCT_WRITE
];
2062 s
->stats
->rd_operations
= bs
->nr_ops
[BDRV_ACCT_READ
];
2063 s
->stats
->wr_operations
= bs
->nr_ops
[BDRV_ACCT_WRITE
];
2064 s
->stats
->wr_highest_offset
= bs
->wr_highest_sector
* BDRV_SECTOR_SIZE
;
2065 s
->stats
->flush_operations
= bs
->nr_ops
[BDRV_ACCT_FLUSH
];
2066 s
->stats
->wr_total_time_ns
= bs
->total_time_ns
[BDRV_ACCT_WRITE
];
2067 s
->stats
->rd_total_time_ns
= bs
->total_time_ns
[BDRV_ACCT_READ
];
2068 s
->stats
->flush_total_time_ns
= bs
->total_time_ns
[BDRV_ACCT_FLUSH
];
2071 s
->has_parent
= true;
2072 s
->parent
= qmp_query_blockstat(bs
->file
, NULL
);
2078 BlockStatsList
*qmp_query_blockstats(Error
**errp
)
2080 BlockStatsList
*head
= NULL
, *cur_item
= NULL
;
2081 BlockDriverState
*bs
;
2083 QTAILQ_FOREACH(bs
, &bdrv_states
, list
) {
2084 BlockStatsList
*info
= g_malloc0(sizeof(*info
));
2085 info
->value
= qmp_query_blockstat(bs
, NULL
);
2087 /* XXX: waiting for the qapi to support GSList */
2089 head
= cur_item
= info
;
2091 cur_item
->next
= info
;
2099 const char *bdrv_get_encrypted_filename(BlockDriverState
*bs
)
2101 if (bs
->backing_hd
&& bs
->backing_hd
->encrypted
)
2102 return bs
->backing_file
;
2103 else if (bs
->encrypted
)
2104 return bs
->filename
;
2109 void bdrv_get_backing_filename(BlockDriverState
*bs
,
2110 char *filename
, int filename_size
)
2112 pstrcpy(filename
, filename_size
, bs
->backing_file
);
2115 int bdrv_write_compressed(BlockDriverState
*bs
, int64_t sector_num
,
2116 const uint8_t *buf
, int nb_sectors
)
2118 BlockDriver
*drv
= bs
->drv
;
2121 if (!drv
->bdrv_write_compressed
)
2123 if (bdrv_check_request(bs
, sector_num
, nb_sectors
))
2126 if (bs
->dirty_bitmap
) {
2127 set_dirty_bitmap(bs
, sector_num
, nb_sectors
, 1);
2130 return drv
->bdrv_write_compressed(bs
, sector_num
, buf
, nb_sectors
);
2133 int bdrv_get_info(BlockDriverState
*bs
, BlockDriverInfo
*bdi
)
2135 BlockDriver
*drv
= bs
->drv
;
2138 if (!drv
->bdrv_get_info
)
2140 memset(bdi
, 0, sizeof(*bdi
));
2141 return drv
->bdrv_get_info(bs
, bdi
);
2144 int bdrv_save_vmstate(BlockDriverState
*bs
, const uint8_t *buf
,
2145 int64_t pos
, int size
)
2147 BlockDriver
*drv
= bs
->drv
;
2150 if (drv
->bdrv_save_vmstate
)
2151 return drv
->bdrv_save_vmstate(bs
, buf
, pos
, size
);
2153 return bdrv_save_vmstate(bs
->file
, buf
, pos
, size
);
2157 int bdrv_load_vmstate(BlockDriverState
*bs
, uint8_t *buf
,
2158 int64_t pos
, int size
)
2160 BlockDriver
*drv
= bs
->drv
;
2163 if (drv
->bdrv_load_vmstate
)
2164 return drv
->bdrv_load_vmstate(bs
, buf
, pos
, size
);
2166 return bdrv_load_vmstate(bs
->file
, buf
, pos
, size
);
2170 void bdrv_debug_event(BlockDriverState
*bs
, BlkDebugEvent event
)
2172 BlockDriver
*drv
= bs
->drv
;
2174 if (!drv
|| !drv
->bdrv_debug_event
) {
2178 return drv
->bdrv_debug_event(bs
, event
);
2182 /**************************************************************/
2183 /* handling of snapshots */
2185 int bdrv_can_snapshot(BlockDriverState
*bs
)
2187 BlockDriver
*drv
= bs
->drv
;
2188 if (!drv
|| !bdrv_is_inserted(bs
) || bdrv_is_read_only(bs
)) {
2192 if (!drv
->bdrv_snapshot_create
) {
2193 if (bs
->file
!= NULL
) {
2194 return bdrv_can_snapshot(bs
->file
);
2202 int bdrv_is_snapshot(BlockDriverState
*bs
)
2204 return !!(bs
->open_flags
& BDRV_O_SNAPSHOT
);
2207 BlockDriverState
*bdrv_snapshots(void)
2209 BlockDriverState
*bs
;
2212 return bs_snapshots
;
2216 while ((bs
= bdrv_next(bs
))) {
2217 if (bdrv_can_snapshot(bs
)) {
2225 int bdrv_snapshot_create(BlockDriverState
*bs
,
2226 QEMUSnapshotInfo
*sn_info
)
2228 BlockDriver
*drv
= bs
->drv
;
2231 if (drv
->bdrv_snapshot_create
)
2232 return drv
->bdrv_snapshot_create(bs
, sn_info
);
2234 return bdrv_snapshot_create(bs
->file
, sn_info
);
2238 int bdrv_snapshot_goto(BlockDriverState
*bs
,
2239 const char *snapshot_id
)
2241 BlockDriver
*drv
= bs
->drv
;
2246 if (drv
->bdrv_snapshot_goto
)
2247 return drv
->bdrv_snapshot_goto(bs
, snapshot_id
);
2250 drv
->bdrv_close(bs
);
2251 ret
= bdrv_snapshot_goto(bs
->file
, snapshot_id
);
2252 open_ret
= drv
->bdrv_open(bs
, bs
->open_flags
);
2254 bdrv_delete(bs
->file
);
2264 int bdrv_snapshot_delete(BlockDriverState
*bs
, const char *snapshot_id
)
2266 BlockDriver
*drv
= bs
->drv
;
2269 if (drv
->bdrv_snapshot_delete
)
2270 return drv
->bdrv_snapshot_delete(bs
, snapshot_id
);
2272 return bdrv_snapshot_delete(bs
->file
, snapshot_id
);
2276 int bdrv_snapshot_list(BlockDriverState
*bs
,
2277 QEMUSnapshotInfo
**psn_info
)
2279 BlockDriver
*drv
= bs
->drv
;
2282 if (drv
->bdrv_snapshot_list
)
2283 return drv
->bdrv_snapshot_list(bs
, psn_info
);
2285 return bdrv_snapshot_list(bs
->file
, psn_info
);
2289 int bdrv_snapshot_load_tmp(BlockDriverState
*bs
,
2290 const char *snapshot_name
)
2292 BlockDriver
*drv
= bs
->drv
;
2296 if (!bs
->read_only
) {
2299 if (drv
->bdrv_snapshot_load_tmp
) {
2300 return drv
->bdrv_snapshot_load_tmp(bs
, snapshot_name
);
2305 #define NB_SUFFIXES 4
2307 char *get_human_readable_size(char *buf
, int buf_size
, int64_t size
)
2309 static const char suffixes
[NB_SUFFIXES
] = "KMGT";
2314 snprintf(buf
, buf_size
, "%" PRId64
, size
);
2317 for(i
= 0; i
< NB_SUFFIXES
; i
++) {
2318 if (size
< (10 * base
)) {
2319 snprintf(buf
, buf_size
, "%0.1f%c",
2320 (double)size
/ base
,
2323 } else if (size
< (1000 * base
) || i
== (NB_SUFFIXES
- 1)) {
2324 snprintf(buf
, buf_size
, "%" PRId64
"%c",
2325 ((size
+ (base
>> 1)) / base
),
2335 char *bdrv_snapshot_dump(char *buf
, int buf_size
, QEMUSnapshotInfo
*sn
)
2337 char buf1
[128], date_buf
[128], clock_buf
[128];
2347 snprintf(buf
, buf_size
,
2348 "%-10s%-20s%7s%20s%15s",
2349 "ID", "TAG", "VM SIZE", "DATE", "VM CLOCK");
2353 ptm
= localtime(&ti
);
2354 strftime(date_buf
, sizeof(date_buf
),
2355 "%Y-%m-%d %H:%M:%S", ptm
);
2357 localtime_r(&ti
, &tm
);
2358 strftime(date_buf
, sizeof(date_buf
),
2359 "%Y-%m-%d %H:%M:%S", &tm
);
2361 secs
= sn
->vm_clock_nsec
/ 1000000000;
2362 snprintf(clock_buf
, sizeof(clock_buf
),
2363 "%02d:%02d:%02d.%03d",
2365 (int)((secs
/ 60) % 60),
2367 (int)((sn
->vm_clock_nsec
/ 1000000) % 1000));
2368 snprintf(buf
, buf_size
,
2369 "%-10s%-20s%7s%20s%15s",
2370 sn
->id_str
, sn
->name
,
2371 get_human_readable_size(buf1
, sizeof(buf1
), sn
->vm_state_size
),
2378 /**************************************************************/
2381 BlockDriverAIOCB
*bdrv_aio_readv(BlockDriverState
*bs
, int64_t sector_num
,
2382 QEMUIOVector
*qiov
, int nb_sectors
,
2383 BlockDriverCompletionFunc
*cb
, void *opaque
)
2385 trace_bdrv_aio_readv(bs
, sector_num
, nb_sectors
, opaque
);
2387 return bdrv_co_aio_rw_vector(bs
, sector_num
, qiov
, nb_sectors
,
2391 BlockDriverAIOCB
*bdrv_aio_writev(BlockDriverState
*bs
, int64_t sector_num
,
2392 QEMUIOVector
*qiov
, int nb_sectors
,
2393 BlockDriverCompletionFunc
*cb
, void *opaque
)
2395 trace_bdrv_aio_writev(bs
, sector_num
, nb_sectors
, opaque
);
2397 return bdrv_co_aio_rw_vector(bs
, sector_num
, qiov
, nb_sectors
,
2402 typedef struct MultiwriteCB
{
2407 BlockDriverCompletionFunc
*cb
;
2409 QEMUIOVector
*free_qiov
;
2414 static void multiwrite_user_cb(MultiwriteCB
*mcb
)
2418 for (i
= 0; i
< mcb
->num_callbacks
; i
++) {
2419 mcb
->callbacks
[i
].cb(mcb
->callbacks
[i
].opaque
, mcb
->error
);
2420 if (mcb
->callbacks
[i
].free_qiov
) {
2421 qemu_iovec_destroy(mcb
->callbacks
[i
].free_qiov
);
2423 g_free(mcb
->callbacks
[i
].free_qiov
);
2424 qemu_vfree(mcb
->callbacks
[i
].free_buf
);
2428 static void multiwrite_cb(void *opaque
, int ret
)
2430 MultiwriteCB
*mcb
= opaque
;
2432 trace_multiwrite_cb(mcb
, ret
);
2434 if (ret
< 0 && !mcb
->error
) {
2438 mcb
->num_requests
--;
2439 if (mcb
->num_requests
== 0) {
2440 multiwrite_user_cb(mcb
);
2445 static int multiwrite_req_compare(const void *a
, const void *b
)
2447 const BlockRequest
*req1
= a
, *req2
= b
;
2450 * Note that we can't simply subtract req2->sector from req1->sector
2451 * here as that could overflow the return value.
2453 if (req1
->sector
> req2
->sector
) {
2455 } else if (req1
->sector
< req2
->sector
) {
2463 * Takes a bunch of requests and tries to merge them. Returns the number of
2464 * requests that remain after merging.
2466 static int multiwrite_merge(BlockDriverState
*bs
, BlockRequest
*reqs
,
2467 int num_reqs
, MultiwriteCB
*mcb
)
2471 // Sort requests by start sector
2472 qsort(reqs
, num_reqs
, sizeof(*reqs
), &multiwrite_req_compare
);
2474 // Check if adjacent requests touch the same clusters. If so, combine them,
2475 // filling up gaps with zero sectors.
2477 for (i
= 1; i
< num_reqs
; i
++) {
2479 int64_t oldreq_last
= reqs
[outidx
].sector
+ reqs
[outidx
].nb_sectors
;
2481 // This handles the cases that are valid for all block drivers, namely
2482 // exactly sequential writes and overlapping writes.
2483 if (reqs
[i
].sector
<= oldreq_last
) {
2487 // The block driver may decide that it makes sense to combine requests
2488 // even if there is a gap of some sectors between them. In this case,
2489 // the gap is filled with zeros (therefore only applicable for yet
2490 // unused space in format like qcow2).
2491 if (!merge
&& bs
->drv
->bdrv_merge_requests
) {
2492 merge
= bs
->drv
->bdrv_merge_requests(bs
, &reqs
[outidx
], &reqs
[i
]);
2495 if (reqs
[outidx
].qiov
->niov
+ reqs
[i
].qiov
->niov
+ 1 > IOV_MAX
) {
2501 QEMUIOVector
*qiov
= g_malloc0(sizeof(*qiov
));
2502 qemu_iovec_init(qiov
,
2503 reqs
[outidx
].qiov
->niov
+ reqs
[i
].qiov
->niov
+ 1);
2505 // Add the first request to the merged one. If the requests are
2506 // overlapping, drop the last sectors of the first request.
2507 size
= (reqs
[i
].sector
- reqs
[outidx
].sector
) << 9;
2508 qemu_iovec_concat(qiov
, reqs
[outidx
].qiov
, size
);
2510 // We might need to add some zeros between the two requests
2511 if (reqs
[i
].sector
> oldreq_last
) {
2512 size_t zero_bytes
= (reqs
[i
].sector
- oldreq_last
) << 9;
2513 uint8_t *buf
= qemu_blockalign(bs
, zero_bytes
);
2514 memset(buf
, 0, zero_bytes
);
2515 qemu_iovec_add(qiov
, buf
, zero_bytes
);
2516 mcb
->callbacks
[i
].free_buf
= buf
;
2519 // Add the second request
2520 qemu_iovec_concat(qiov
, reqs
[i
].qiov
, reqs
[i
].qiov
->size
);
2522 reqs
[outidx
].nb_sectors
= qiov
->size
>> 9;
2523 reqs
[outidx
].qiov
= qiov
;
2525 mcb
->callbacks
[i
].free_qiov
= reqs
[outidx
].qiov
;
2528 reqs
[outidx
].sector
= reqs
[i
].sector
;
2529 reqs
[outidx
].nb_sectors
= reqs
[i
].nb_sectors
;
2530 reqs
[outidx
].qiov
= reqs
[i
].qiov
;
2538 * Submit multiple AIO write requests at once.
2540 * On success, the function returns 0 and all requests in the reqs array have
2541 * been submitted. In error case this function returns -1, and any of the
2542 * requests may or may not be submitted yet. In particular, this means that the
2543 * callback will be called for some of the requests, for others it won't. The
2544 * caller must check the error field of the BlockRequest to wait for the right
2545 * callbacks (if error != 0, no callback will be called).
2547 * The implementation may modify the contents of the reqs array, e.g. to merge
2548 * requests. However, the fields opaque and error are left unmodified as they
2549 * are used to signal failure for a single request to the caller.
2551 int bdrv_aio_multiwrite(BlockDriverState
*bs
, BlockRequest
*reqs
, int num_reqs
)
2553 BlockDriverAIOCB
*acb
;
2557 /* don't submit writes if we don't have a medium */
2558 if (bs
->drv
== NULL
) {
2559 for (i
= 0; i
< num_reqs
; i
++) {
2560 reqs
[i
].error
= -ENOMEDIUM
;
2565 if (num_reqs
== 0) {
2569 // Create MultiwriteCB structure
2570 mcb
= g_malloc0(sizeof(*mcb
) + num_reqs
* sizeof(*mcb
->callbacks
));
2571 mcb
->num_requests
= 0;
2572 mcb
->num_callbacks
= num_reqs
;
2574 for (i
= 0; i
< num_reqs
; i
++) {
2575 mcb
->callbacks
[i
].cb
= reqs
[i
].cb
;
2576 mcb
->callbacks
[i
].opaque
= reqs
[i
].opaque
;
2579 // Check for mergable requests
2580 num_reqs
= multiwrite_merge(bs
, reqs
, num_reqs
, mcb
);
2582 trace_bdrv_aio_multiwrite(mcb
, mcb
->num_callbacks
, num_reqs
);
2585 * Run the aio requests. As soon as one request can't be submitted
2586 * successfully, fail all requests that are not yet submitted (we must
2587 * return failure for all requests anyway)
2589 * num_requests cannot be set to the right value immediately: If
2590 * bdrv_aio_writev fails for some request, num_requests would be too high
2591 * and therefore multiwrite_cb() would never recognize the multiwrite
2592 * request as completed. We also cannot use the loop variable i to set it
2593 * when the first request fails because the callback may already have been
2594 * called for previously submitted requests. Thus, num_requests must be
2595 * incremented for each request that is submitted.
2597 * The problem that callbacks may be called early also means that we need
2598 * to take care that num_requests doesn't become 0 before all requests are
2599 * submitted - multiwrite_cb() would consider the multiwrite request
2600 * completed. A dummy request that is "completed" by a manual call to
2601 * multiwrite_cb() takes care of this.
2603 mcb
->num_requests
= 1;
2605 // Run the aio requests
2606 for (i
= 0; i
< num_reqs
; i
++) {
2607 mcb
->num_requests
++;
2608 acb
= bdrv_aio_writev(bs
, reqs
[i
].sector
, reqs
[i
].qiov
,
2609 reqs
[i
].nb_sectors
, multiwrite_cb
, mcb
);
2612 // We can only fail the whole thing if no request has been
2613 // submitted yet. Otherwise we'll wait for the submitted AIOs to
2614 // complete and report the error in the callback.
2616 trace_bdrv_aio_multiwrite_earlyfail(mcb
);
2619 trace_bdrv_aio_multiwrite_latefail(mcb
, i
);
2620 multiwrite_cb(mcb
, -EIO
);
2626 /* Complete the dummy request */
2627 multiwrite_cb(mcb
, 0);
2632 for (i
= 0; i
< mcb
->num_callbacks
; i
++) {
2633 reqs
[i
].error
= -EIO
;
2639 void bdrv_aio_cancel(BlockDriverAIOCB
*acb
)
2641 acb
->pool
->cancel(acb
);
2644 /* block I/O throttling */
2645 static bool bdrv_exceed_bps_limits(BlockDriverState
*bs
, int nb_sectors
,
2646 bool is_write
, double elapsed_time
, uint64_t *wait
)
2648 uint64_t bps_limit
= 0;
2649 double bytes_limit
, bytes_base
, bytes_res
;
2650 double slice_time
, wait_time
;
2652 if (bs
->io_limits
.bps
[BLOCK_IO_LIMIT_TOTAL
]) {
2653 bps_limit
= bs
->io_limits
.bps
[BLOCK_IO_LIMIT_TOTAL
];
2654 } else if (bs
->io_limits
.bps
[is_write
]) {
2655 bps_limit
= bs
->io_limits
.bps
[is_write
];
2664 slice_time
= bs
->slice_end
- bs
->slice_start
;
2665 slice_time
/= (NANOSECONDS_PER_SECOND
);
2666 bytes_limit
= bps_limit
* slice_time
;
2667 bytes_base
= bs
->nr_bytes
[is_write
] - bs
->io_base
.bytes
[is_write
];
2668 if (bs
->io_limits
.bps
[BLOCK_IO_LIMIT_TOTAL
]) {
2669 bytes_base
+= bs
->nr_bytes
[!is_write
] - bs
->io_base
.bytes
[!is_write
];
2672 /* bytes_base: the bytes of data which have been read/written; and
2673 * it is obtained from the history statistic info.
2674 * bytes_res: the remaining bytes of data which need to be read/written.
2675 * (bytes_base + bytes_res) / bps_limit: used to calcuate
2676 * the total time for completing reading/writting all data.
2678 bytes_res
= (unsigned) nb_sectors
* BDRV_SECTOR_SIZE
;
2680 if (bytes_base
+ bytes_res
<= bytes_limit
) {
2688 /* Calc approx time to dispatch */
2689 wait_time
= (bytes_base
+ bytes_res
) / bps_limit
- elapsed_time
;
2691 /* When the I/O rate at runtime exceeds the limits,
2692 * bs->slice_end need to be extended in order that the current statistic
2693 * info can be kept until the timer fire, so it is increased and tuned
2694 * based on the result of experiment.
2696 bs
->slice_time
= wait_time
* BLOCK_IO_SLICE_TIME
* 10;
2697 bs
->slice_end
+= bs
->slice_time
- 3 * BLOCK_IO_SLICE_TIME
;
2699 *wait
= wait_time
* BLOCK_IO_SLICE_TIME
* 10;
2705 static bool bdrv_exceed_iops_limits(BlockDriverState
*bs
, bool is_write
,
2706 double elapsed_time
, uint64_t *wait
)
2708 uint64_t iops_limit
= 0;
2709 double ios_limit
, ios_base
;
2710 double slice_time
, wait_time
;
2712 if (bs
->io_limits
.iops
[BLOCK_IO_LIMIT_TOTAL
]) {
2713 iops_limit
= bs
->io_limits
.iops
[BLOCK_IO_LIMIT_TOTAL
];
2714 } else if (bs
->io_limits
.iops
[is_write
]) {
2715 iops_limit
= bs
->io_limits
.iops
[is_write
];
2724 slice_time
= bs
->slice_end
- bs
->slice_start
;
2725 slice_time
/= (NANOSECONDS_PER_SECOND
);
2726 ios_limit
= iops_limit
* slice_time
;
2727 ios_base
= bs
->nr_ops
[is_write
] - bs
->io_base
.ios
[is_write
];
2728 if (bs
->io_limits
.iops
[BLOCK_IO_LIMIT_TOTAL
]) {
2729 ios_base
+= bs
->nr_ops
[!is_write
] - bs
->io_base
.ios
[!is_write
];
2732 if (ios_base
+ 1 <= ios_limit
) {
2740 /* Calc approx time to dispatch */
2741 wait_time
= (ios_base
+ 1) / iops_limit
;
2742 if (wait_time
> elapsed_time
) {
2743 wait_time
= wait_time
- elapsed_time
;
2748 bs
->slice_time
= wait_time
* BLOCK_IO_SLICE_TIME
* 10;
2749 bs
->slice_end
+= bs
->slice_time
- 3 * BLOCK_IO_SLICE_TIME
;
2751 *wait
= wait_time
* BLOCK_IO_SLICE_TIME
* 10;
2757 static bool bdrv_exceed_io_limits(BlockDriverState
*bs
, int nb_sectors
,
2758 bool is_write
, int64_t *wait
)
2760 int64_t now
, max_wait
;
2761 uint64_t bps_wait
= 0, iops_wait
= 0;
2762 double elapsed_time
;
2763 int bps_ret
, iops_ret
;
2765 now
= qemu_get_clock_ns(vm_clock
);
2766 if ((bs
->slice_start
< now
)
2767 && (bs
->slice_end
> now
)) {
2768 bs
->slice_end
= now
+ bs
->slice_time
;
2770 bs
->slice_time
= 5 * BLOCK_IO_SLICE_TIME
;
2771 bs
->slice_start
= now
;
2772 bs
->slice_end
= now
+ bs
->slice_time
;
2774 bs
->io_base
.bytes
[is_write
] = bs
->nr_bytes
[is_write
];
2775 bs
->io_base
.bytes
[!is_write
] = bs
->nr_bytes
[!is_write
];
2777 bs
->io_base
.ios
[is_write
] = bs
->nr_ops
[is_write
];
2778 bs
->io_base
.ios
[!is_write
] = bs
->nr_ops
[!is_write
];
2781 elapsed_time
= now
- bs
->slice_start
;
2782 elapsed_time
/= (NANOSECONDS_PER_SECOND
);
2784 bps_ret
= bdrv_exceed_bps_limits(bs
, nb_sectors
,
2785 is_write
, elapsed_time
, &bps_wait
);
2786 iops_ret
= bdrv_exceed_iops_limits(bs
, is_write
,
2787 elapsed_time
, &iops_wait
);
2788 if (bps_ret
|| iops_ret
) {
2789 max_wait
= bps_wait
> iops_wait
? bps_wait
: iops_wait
;
2794 now
= qemu_get_clock_ns(vm_clock
);
2795 if (bs
->slice_end
< now
+ max_wait
) {
2796 bs
->slice_end
= now
+ max_wait
;
2809 /**************************************************************/
2810 /* async block device emulation */
2812 typedef struct BlockDriverAIOCBSync
{
2813 BlockDriverAIOCB common
;
2816 /* vector translation state */
2820 } BlockDriverAIOCBSync
;
2822 static void bdrv_aio_cancel_em(BlockDriverAIOCB
*blockacb
)
2824 BlockDriverAIOCBSync
*acb
=
2825 container_of(blockacb
, BlockDriverAIOCBSync
, common
);
2826 qemu_bh_delete(acb
->bh
);
2828 qemu_aio_release(acb
);
2831 static AIOPool bdrv_em_aio_pool
= {
2832 .aiocb_size
= sizeof(BlockDriverAIOCBSync
),
2833 .cancel
= bdrv_aio_cancel_em
,
2836 static void bdrv_aio_bh_cb(void *opaque
)
2838 BlockDriverAIOCBSync
*acb
= opaque
;
2841 qemu_iovec_from_buffer(acb
->qiov
, acb
->bounce
, acb
->qiov
->size
);
2842 qemu_vfree(acb
->bounce
);
2843 acb
->common
.cb(acb
->common
.opaque
, acb
->ret
);
2844 qemu_bh_delete(acb
->bh
);
2846 qemu_aio_release(acb
);
2849 static BlockDriverAIOCB
*bdrv_aio_rw_vector(BlockDriverState
*bs
,
2853 BlockDriverCompletionFunc
*cb
,
2858 BlockDriverAIOCBSync
*acb
;
2860 acb
= qemu_aio_get(&bdrv_em_aio_pool
, bs
, cb
, opaque
);
2861 acb
->is_write
= is_write
;
2863 acb
->bounce
= qemu_blockalign(bs
, qiov
->size
);
2866 acb
->bh
= qemu_bh_new(bdrv_aio_bh_cb
, acb
);
2869 qemu_iovec_to_buffer(acb
->qiov
, acb
->bounce
);
2870 acb
->ret
= bs
->drv
->bdrv_write(bs
, sector_num
, acb
->bounce
, nb_sectors
);
2872 acb
->ret
= bs
->drv
->bdrv_read(bs
, sector_num
, acb
->bounce
, nb_sectors
);
2875 qemu_bh_schedule(acb
->bh
);
2877 return &acb
->common
;
2880 static BlockDriverAIOCB
*bdrv_aio_readv_em(BlockDriverState
*bs
,
2881 int64_t sector_num
, QEMUIOVector
*qiov
, int nb_sectors
,
2882 BlockDriverCompletionFunc
*cb
, void *opaque
)
2884 return bdrv_aio_rw_vector(bs
, sector_num
, qiov
, nb_sectors
, cb
, opaque
, 0);
2887 static BlockDriverAIOCB
*bdrv_aio_writev_em(BlockDriverState
*bs
,
2888 int64_t sector_num
, QEMUIOVector
*qiov
, int nb_sectors
,
2889 BlockDriverCompletionFunc
*cb
, void *opaque
)
2891 return bdrv_aio_rw_vector(bs
, sector_num
, qiov
, nb_sectors
, cb
, opaque
, 1);
2895 typedef struct BlockDriverAIOCBCoroutine
{
2896 BlockDriverAIOCB common
;
2900 } BlockDriverAIOCBCoroutine
;
2902 static void bdrv_aio_co_cancel_em(BlockDriverAIOCB
*blockacb
)
2907 static AIOPool bdrv_em_co_aio_pool
= {
2908 .aiocb_size
= sizeof(BlockDriverAIOCBCoroutine
),
2909 .cancel
= bdrv_aio_co_cancel_em
,
2912 static void bdrv_co_em_bh(void *opaque
)
2914 BlockDriverAIOCBCoroutine
*acb
= opaque
;
2916 acb
->common
.cb(acb
->common
.opaque
, acb
->req
.error
);
2917 qemu_bh_delete(acb
->bh
);
2918 qemu_aio_release(acb
);
2921 /* Invoke bdrv_co_do_readv/bdrv_co_do_writev */
2922 static void coroutine_fn
bdrv_co_do_rw(void *opaque
)
2924 BlockDriverAIOCBCoroutine
*acb
= opaque
;
2925 BlockDriverState
*bs
= acb
->common
.bs
;
2927 if (!acb
->is_write
) {
2928 acb
->req
.error
= bdrv_co_do_readv(bs
, acb
->req
.sector
,
2929 acb
->req
.nb_sectors
, acb
->req
.qiov
);
2931 acb
->req
.error
= bdrv_co_do_writev(bs
, acb
->req
.sector
,
2932 acb
->req
.nb_sectors
, acb
->req
.qiov
);
2935 acb
->bh
= qemu_bh_new(bdrv_co_em_bh
, acb
);
2936 qemu_bh_schedule(acb
->bh
);
2939 static BlockDriverAIOCB
*bdrv_co_aio_rw_vector(BlockDriverState
*bs
,
2943 BlockDriverCompletionFunc
*cb
,
2948 BlockDriverAIOCBCoroutine
*acb
;
2950 acb
= qemu_aio_get(&bdrv_em_co_aio_pool
, bs
, cb
, opaque
);
2951 acb
->req
.sector
= sector_num
;
2952 acb
->req
.nb_sectors
= nb_sectors
;
2953 acb
->req
.qiov
= qiov
;
2954 acb
->is_write
= is_write
;
2956 co
= qemu_coroutine_create(bdrv_co_do_rw
);
2957 qemu_coroutine_enter(co
, acb
);
2959 return &acb
->common
;
2962 static void coroutine_fn
bdrv_aio_flush_co_entry(void *opaque
)
2964 BlockDriverAIOCBCoroutine
*acb
= opaque
;
2965 BlockDriverState
*bs
= acb
->common
.bs
;
2967 acb
->req
.error
= bdrv_co_flush(bs
);
2968 acb
->bh
= qemu_bh_new(bdrv_co_em_bh
, acb
);
2969 qemu_bh_schedule(acb
->bh
);
2972 BlockDriverAIOCB
*bdrv_aio_flush(BlockDriverState
*bs
,
2973 BlockDriverCompletionFunc
*cb
, void *opaque
)
2975 trace_bdrv_aio_flush(bs
, opaque
);
2978 BlockDriverAIOCBCoroutine
*acb
;
2980 acb
= qemu_aio_get(&bdrv_em_co_aio_pool
, bs
, cb
, opaque
);
2981 co
= qemu_coroutine_create(bdrv_aio_flush_co_entry
);
2982 qemu_coroutine_enter(co
, acb
);
2984 return &acb
->common
;
2987 static void coroutine_fn
bdrv_aio_discard_co_entry(void *opaque
)
2989 BlockDriverAIOCBCoroutine
*acb
= opaque
;
2990 BlockDriverState
*bs
= acb
->common
.bs
;
2992 acb
->req
.error
= bdrv_co_discard(bs
, acb
->req
.sector
, acb
->req
.nb_sectors
);
2993 acb
->bh
= qemu_bh_new(bdrv_co_em_bh
, acb
);
2994 qemu_bh_schedule(acb
->bh
);
2997 BlockDriverAIOCB
*bdrv_aio_discard(BlockDriverState
*bs
,
2998 int64_t sector_num
, int nb_sectors
,
2999 BlockDriverCompletionFunc
*cb
, void *opaque
)
3002 BlockDriverAIOCBCoroutine
*acb
;
3004 trace_bdrv_aio_discard(bs
, sector_num
, nb_sectors
, opaque
);
3006 acb
= qemu_aio_get(&bdrv_em_co_aio_pool
, bs
, cb
, opaque
);
3007 acb
->req
.sector
= sector_num
;
3008 acb
->req
.nb_sectors
= nb_sectors
;
3009 co
= qemu_coroutine_create(bdrv_aio_discard_co_entry
);
3010 qemu_coroutine_enter(co
, acb
);
3012 return &acb
->common
;
3015 void bdrv_init(void)
3017 module_call_init(MODULE_INIT_BLOCK
);
3020 void bdrv_init_with_whitelist(void)
3022 use_bdrv_whitelist
= 1;
3026 void *qemu_aio_get(AIOPool
*pool
, BlockDriverState
*bs
,
3027 BlockDriverCompletionFunc
*cb
, void *opaque
)
3029 BlockDriverAIOCB
*acb
;
3031 if (pool
->free_aiocb
) {
3032 acb
= pool
->free_aiocb
;
3033 pool
->free_aiocb
= acb
->next
;
3035 acb
= g_malloc0(pool
->aiocb_size
);
3040 acb
->opaque
= opaque
;
3044 void qemu_aio_release(void *p
)
3046 BlockDriverAIOCB
*acb
= (BlockDriverAIOCB
*)p
;
3047 AIOPool
*pool
= acb
->pool
;
3048 acb
->next
= pool
->free_aiocb
;
3049 pool
->free_aiocb
= acb
;
3052 /**************************************************************/
3053 /* Coroutine block device emulation */
3055 typedef struct CoroutineIOCompletion
{
3056 Coroutine
*coroutine
;
3058 } CoroutineIOCompletion
;
3060 static void bdrv_co_io_em_complete(void *opaque
, int ret
)
3062 CoroutineIOCompletion
*co
= opaque
;
3065 qemu_coroutine_enter(co
->coroutine
, NULL
);
3068 static int coroutine_fn
bdrv_co_io_em(BlockDriverState
*bs
, int64_t sector_num
,
3069 int nb_sectors
, QEMUIOVector
*iov
,
3072 CoroutineIOCompletion co
= {
3073 .coroutine
= qemu_coroutine_self(),
3075 BlockDriverAIOCB
*acb
;
3078 acb
= bs
->drv
->bdrv_aio_writev(bs
, sector_num
, iov
, nb_sectors
,
3079 bdrv_co_io_em_complete
, &co
);
3081 acb
= bs
->drv
->bdrv_aio_readv(bs
, sector_num
, iov
, nb_sectors
,
3082 bdrv_co_io_em_complete
, &co
);
3085 trace_bdrv_co_io_em(bs
, sector_num
, nb_sectors
, is_write
, acb
);
3089 qemu_coroutine_yield();
3094 static int coroutine_fn
bdrv_co_readv_em(BlockDriverState
*bs
,
3095 int64_t sector_num
, int nb_sectors
,
3098 return bdrv_co_io_em(bs
, sector_num
, nb_sectors
, iov
, false);
3101 static int coroutine_fn
bdrv_co_writev_em(BlockDriverState
*bs
,
3102 int64_t sector_num
, int nb_sectors
,
3105 return bdrv_co_io_em(bs
, sector_num
, nb_sectors
, iov
, true);
3108 static void coroutine_fn
bdrv_flush_co_entry(void *opaque
)
3110 RwCo
*rwco
= opaque
;
3112 rwco
->ret
= bdrv_co_flush(rwco
->bs
);
3115 int coroutine_fn
bdrv_co_flush(BlockDriverState
*bs
)
3123 /* Write back cached data to the OS even with cache=unsafe */
3124 if (bs
->drv
->bdrv_co_flush_to_os
) {
3125 ret
= bs
->drv
->bdrv_co_flush_to_os(bs
);
3131 /* But don't actually force it to the disk with cache=unsafe */
3132 if (bs
->open_flags
& BDRV_O_NO_FLUSH
) {
3136 if (bs
->drv
->bdrv_co_flush_to_disk
) {
3137 return bs
->drv
->bdrv_co_flush_to_disk(bs
);
3138 } else if (bs
->drv
->bdrv_aio_flush
) {
3139 BlockDriverAIOCB
*acb
;
3140 CoroutineIOCompletion co
= {
3141 .coroutine
= qemu_coroutine_self(),
3144 acb
= bs
->drv
->bdrv_aio_flush(bs
, bdrv_co_io_em_complete
, &co
);
3148 qemu_coroutine_yield();
3153 * Some block drivers always operate in either writethrough or unsafe
3154 * mode and don't support bdrv_flush therefore. Usually qemu doesn't
3155 * know how the server works (because the behaviour is hardcoded or
3156 * depends on server-side configuration), so we can't ensure that
3157 * everything is safe on disk. Returning an error doesn't work because
3158 * that would break guests even if the server operates in writethrough
3161 * Let's hope the user knows what he's doing.
3167 void bdrv_invalidate_cache(BlockDriverState
*bs
)
3169 if (bs
->drv
&& bs
->drv
->bdrv_invalidate_cache
) {
3170 bs
->drv
->bdrv_invalidate_cache(bs
);
3174 void bdrv_invalidate_cache_all(void)
3176 BlockDriverState
*bs
;
3178 QTAILQ_FOREACH(bs
, &bdrv_states
, list
) {
3179 bdrv_invalidate_cache(bs
);
3183 int bdrv_flush(BlockDriverState
*bs
)
3191 if (qemu_in_coroutine()) {
3192 /* Fast-path if already in coroutine context */
3193 bdrv_flush_co_entry(&rwco
);
3195 co
= qemu_coroutine_create(bdrv_flush_co_entry
);
3196 qemu_coroutine_enter(co
, &rwco
);
3197 while (rwco
.ret
== NOT_DONE
) {
3205 static void coroutine_fn
bdrv_discard_co_entry(void *opaque
)
3207 RwCo
*rwco
= opaque
;
3209 rwco
->ret
= bdrv_co_discard(rwco
->bs
, rwco
->sector_num
, rwco
->nb_sectors
);
3212 int coroutine_fn
bdrv_co_discard(BlockDriverState
*bs
, int64_t sector_num
,
3217 } else if (bdrv_check_request(bs
, sector_num
, nb_sectors
)) {
3219 } else if (bs
->read_only
) {
3221 } else if (bs
->drv
->bdrv_co_discard
) {
3222 return bs
->drv
->bdrv_co_discard(bs
, sector_num
, nb_sectors
);
3223 } else if (bs
->drv
->bdrv_aio_discard
) {
3224 BlockDriverAIOCB
*acb
;
3225 CoroutineIOCompletion co
= {
3226 .coroutine
= qemu_coroutine_self(),
3229 acb
= bs
->drv
->bdrv_aio_discard(bs
, sector_num
, nb_sectors
,
3230 bdrv_co_io_em_complete
, &co
);
3234 qemu_coroutine_yield();
3242 int bdrv_discard(BlockDriverState
*bs
, int64_t sector_num
, int nb_sectors
)
3247 .sector_num
= sector_num
,
3248 .nb_sectors
= nb_sectors
,
3252 if (qemu_in_coroutine()) {
3253 /* Fast-path if already in coroutine context */
3254 bdrv_discard_co_entry(&rwco
);
3256 co
= qemu_coroutine_create(bdrv_discard_co_entry
);
3257 qemu_coroutine_enter(co
, &rwco
);
3258 while (rwco
.ret
== NOT_DONE
) {
3266 /**************************************************************/
3267 /* removable device support */
3270 * Return TRUE if the media is present
3272 int bdrv_is_inserted(BlockDriverState
*bs
)
3274 BlockDriver
*drv
= bs
->drv
;
3278 if (!drv
->bdrv_is_inserted
)
3280 return drv
->bdrv_is_inserted(bs
);
3284 * Return whether the media changed since the last call to this
3285 * function, or -ENOTSUP if we don't know. Most drivers don't know.
3287 int bdrv_media_changed(BlockDriverState
*bs
)
3289 BlockDriver
*drv
= bs
->drv
;
3291 if (drv
&& drv
->bdrv_media_changed
) {
3292 return drv
->bdrv_media_changed(bs
);
3298 * If eject_flag is TRUE, eject the media. Otherwise, close the tray
3300 void bdrv_eject(BlockDriverState
*bs
, int eject_flag
)
3302 BlockDriver
*drv
= bs
->drv
;
3304 if (drv
&& drv
->bdrv_eject
) {
3305 drv
->bdrv_eject(bs
, eject_flag
);
3310 * Lock or unlock the media (if it is locked, the user won't be able
3311 * to eject it manually).
3313 void bdrv_lock_medium(BlockDriverState
*bs
, bool locked
)
3315 BlockDriver
*drv
= bs
->drv
;
3317 trace_bdrv_lock_medium(bs
, locked
);
3319 if (drv
&& drv
->bdrv_lock_medium
) {
3320 drv
->bdrv_lock_medium(bs
, locked
);
3324 /* needed for generic scsi interface */
3326 int bdrv_ioctl(BlockDriverState
*bs
, unsigned long int req
, void *buf
)
3328 BlockDriver
*drv
= bs
->drv
;
3330 if (drv
&& drv
->bdrv_ioctl
)
3331 return drv
->bdrv_ioctl(bs
, req
, buf
);
3335 BlockDriverAIOCB
*bdrv_aio_ioctl(BlockDriverState
*bs
,
3336 unsigned long int req
, void *buf
,
3337 BlockDriverCompletionFunc
*cb
, void *opaque
)
3339 BlockDriver
*drv
= bs
->drv
;
3341 if (drv
&& drv
->bdrv_aio_ioctl
)
3342 return drv
->bdrv_aio_ioctl(bs
, req
, buf
, cb
, opaque
);
3346 void bdrv_set_buffer_alignment(BlockDriverState
*bs
, int align
)
3348 bs
->buffer_alignment
= align
;
3351 void *qemu_blockalign(BlockDriverState
*bs
, size_t size
)
3353 return qemu_memalign((bs
&& bs
->buffer_alignment
) ? bs
->buffer_alignment
: 512, size
);
3356 void bdrv_set_dirty_tracking(BlockDriverState
*bs
, int enable
)
3358 int64_t bitmap_size
;
3360 bs
->dirty_count
= 0;
3362 if (!bs
->dirty_bitmap
) {
3363 bitmap_size
= (bdrv_getlength(bs
) >> BDRV_SECTOR_BITS
) +
3364 BDRV_SECTORS_PER_DIRTY_CHUNK
* 8 - 1;
3365 bitmap_size
/= BDRV_SECTORS_PER_DIRTY_CHUNK
* 8;
3367 bs
->dirty_bitmap
= g_malloc0(bitmap_size
);
3370 if (bs
->dirty_bitmap
) {
3371 g_free(bs
->dirty_bitmap
);
3372 bs
->dirty_bitmap
= NULL
;
3377 int bdrv_get_dirty(BlockDriverState
*bs
, int64_t sector
)
3379 int64_t chunk
= sector
/ (int64_t)BDRV_SECTORS_PER_DIRTY_CHUNK
;
3381 if (bs
->dirty_bitmap
&&
3382 (sector
<< BDRV_SECTOR_BITS
) < bdrv_getlength(bs
)) {
3383 return !!(bs
->dirty_bitmap
[chunk
/ (sizeof(unsigned long) * 8)] &
3384 (1UL << (chunk
% (sizeof(unsigned long) * 8))));
3390 void bdrv_reset_dirty(BlockDriverState
*bs
, int64_t cur_sector
,
3393 set_dirty_bitmap(bs
, cur_sector
, nr_sectors
, 0);
3396 int64_t bdrv_get_dirty_count(BlockDriverState
*bs
)
3398 return bs
->dirty_count
;
3401 void bdrv_set_in_use(BlockDriverState
*bs
, int in_use
)
3403 assert(bs
->in_use
!= in_use
);
3404 bs
->in_use
= in_use
;
3407 int bdrv_in_use(BlockDriverState
*bs
)
3412 void bdrv_iostatus_enable(BlockDriverState
*bs
)
3414 bs
->iostatus_enabled
= true;
3415 bs
->iostatus
= BLOCK_DEVICE_IO_STATUS_OK
;
3418 /* The I/O status is only enabled if the drive explicitly
3419 * enables it _and_ the VM is configured to stop on errors */
3420 bool bdrv_iostatus_is_enabled(const BlockDriverState
*bs
)
3422 return (bs
->iostatus_enabled
&&
3423 (bs
->on_write_error
== BLOCK_ERR_STOP_ENOSPC
||
3424 bs
->on_write_error
== BLOCK_ERR_STOP_ANY
||
3425 bs
->on_read_error
== BLOCK_ERR_STOP_ANY
));
3428 void bdrv_iostatus_disable(BlockDriverState
*bs
)
3430 bs
->iostatus_enabled
= false;
3433 void bdrv_iostatus_reset(BlockDriverState
*bs
)
3435 if (bdrv_iostatus_is_enabled(bs
)) {
3436 bs
->iostatus
= BLOCK_DEVICE_IO_STATUS_OK
;
3440 /* XXX: Today this is set by device models because it makes the implementation
3441 quite simple. However, the block layer knows about the error, so it's
3442 possible to implement this without device models being involved */
3443 void bdrv_iostatus_set_err(BlockDriverState
*bs
, int error
)
3445 if (bdrv_iostatus_is_enabled(bs
) &&
3446 bs
->iostatus
== BLOCK_DEVICE_IO_STATUS_OK
) {
3448 bs
->iostatus
= error
== ENOSPC
? BLOCK_DEVICE_IO_STATUS_NOSPACE
:
3449 BLOCK_DEVICE_IO_STATUS_FAILED
;
3454 bdrv_acct_start(BlockDriverState
*bs
, BlockAcctCookie
*cookie
, int64_t bytes
,
3455 enum BlockAcctType type
)
3457 assert(type
< BDRV_MAX_IOTYPE
);
3459 cookie
->bytes
= bytes
;
3460 cookie
->start_time_ns
= get_clock();
3461 cookie
->type
= type
;
3465 bdrv_acct_done(BlockDriverState
*bs
, BlockAcctCookie
*cookie
)
3467 assert(cookie
->type
< BDRV_MAX_IOTYPE
);
3469 bs
->nr_bytes
[cookie
->type
] += cookie
->bytes
;
3470 bs
->nr_ops
[cookie
->type
]++;
3471 bs
->total_time_ns
[cookie
->type
] += get_clock() - cookie
->start_time_ns
;
3474 int bdrv_img_create(const char *filename
, const char *fmt
,
3475 const char *base_filename
, const char *base_fmt
,
3476 char *options
, uint64_t img_size
, int flags
)
3478 QEMUOptionParameter
*param
= NULL
, *create_options
= NULL
;
3479 QEMUOptionParameter
*backing_fmt
, *backing_file
, *size
;
3480 BlockDriverState
*bs
= NULL
;
3481 BlockDriver
*drv
, *proto_drv
;
3482 BlockDriver
*backing_drv
= NULL
;
3485 /* Find driver and parse its options */
3486 drv
= bdrv_find_format(fmt
);
3488 error_report("Unknown file format '%s'", fmt
);
3493 proto_drv
= bdrv_find_protocol(filename
);
3495 error_report("Unknown protocol '%s'", filename
);
3500 create_options
= append_option_parameters(create_options
,
3501 drv
->create_options
);
3502 create_options
= append_option_parameters(create_options
,
3503 proto_drv
->create_options
);
3505 /* Create parameter list with default values */
3506 param
= parse_option_parameters("", create_options
, param
);
3508 set_option_parameter_int(param
, BLOCK_OPT_SIZE
, img_size
);
3510 /* Parse -o options */
3512 param
= parse_option_parameters(options
, create_options
, param
);
3513 if (param
== NULL
) {
3514 error_report("Invalid options for file format '%s'.", fmt
);
3520 if (base_filename
) {
3521 if (set_option_parameter(param
, BLOCK_OPT_BACKING_FILE
,
3523 error_report("Backing file not supported for file format '%s'",
3531 if (set_option_parameter(param
, BLOCK_OPT_BACKING_FMT
, base_fmt
)) {
3532 error_report("Backing file format not supported for file "
3533 "format '%s'", fmt
);
3539 backing_file
= get_option_parameter(param
, BLOCK_OPT_BACKING_FILE
);
3540 if (backing_file
&& backing_file
->value
.s
) {
3541 if (!strcmp(filename
, backing_file
->value
.s
)) {
3542 error_report("Error: Trying to create an image with the "
3543 "same filename as the backing file");
3549 backing_fmt
= get_option_parameter(param
, BLOCK_OPT_BACKING_FMT
);
3550 if (backing_fmt
&& backing_fmt
->value
.s
) {
3551 backing_drv
= bdrv_find_format(backing_fmt
->value
.s
);
3553 error_report("Unknown backing file format '%s'",
3554 backing_fmt
->value
.s
);
3560 // The size for the image must always be specified, with one exception:
3561 // If we are using a backing file, we can obtain the size from there
3562 size
= get_option_parameter(param
, BLOCK_OPT_SIZE
);
3563 if (size
&& size
->value
.n
== -1) {
3564 if (backing_file
&& backing_file
->value
.s
) {
3570 ret
= bdrv_open(bs
, backing_file
->value
.s
, flags
, backing_drv
);
3572 error_report("Could not open '%s'", backing_file
->value
.s
);
3575 bdrv_get_geometry(bs
, &size
);
3578 snprintf(buf
, sizeof(buf
), "%" PRId64
, size
);
3579 set_option_parameter(param
, BLOCK_OPT_SIZE
, buf
);
3581 error_report("Image creation needs a size parameter");
3587 printf("Formatting '%s', fmt=%s ", filename
, fmt
);
3588 print_option_parameters(param
);
3591 ret
= bdrv_create(drv
, filename
, param
);
3594 if (ret
== -ENOTSUP
) {
3595 error_report("Formatting or formatting option not supported for "
3596 "file format '%s'", fmt
);
3597 } else if (ret
== -EFBIG
) {
3598 error_report("The image size is too large for file format '%s'",
3601 error_report("%s: error while creating %s: %s", filename
, fmt
,
3607 free_option_parameters(create_options
);
3608 free_option_parameters(param
);