2 * Block driver for RAW files (posix)
4 * Copyright (c) 2006 Fabrice Bellard
6 * Permission is hereby granted, free of charge, to any person obtaining a copy
7 * of this software and associated documentation files (the "Software"), to deal
8 * in the Software without restriction, including without limitation the rights
9 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
10 * copies of the Software, and to permit persons to whom the Software is
11 * furnished to do so, subject to the following conditions:
13 * The above copyright notice and this permission notice shall be included in
14 * all copies or substantial portions of the Software.
16 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
17 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
18 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
19 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
20 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
21 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
24 #include "qemu/osdep.h"
25 #include "qapi/error.h"
26 #include "qemu-common.h"
27 #include "qemu/error-report.h"
28 #include "qemu/timer.h"
30 #include "block/block_int.h"
31 #include "qemu/module.h"
33 #include "block/thread-pool.h"
36 #include "qapi/util.h"
37 #include "qapi/qmp/qstring.h"
39 #if defined(__APPLE__) && (__MACH__)
41 #include <sys/param.h>
42 #include <IOKit/IOKitLib.h>
43 #include <IOKit/IOBSD.h>
44 #include <IOKit/storage/IOMediaBSDClient.h>
45 #include <IOKit/storage/IOMedia.h>
46 #include <IOKit/storage/IOCDMedia.h>
47 //#include <IOKit/storage/IOCDTypes.h>
48 #include <CoreFoundation/CoreFoundation.h>
52 #define _POSIX_PTHREAD_SEMANTICS 1
56 #include <sys/ioctl.h>
57 #include <sys/param.h>
58 #include <linux/cdrom.h>
61 #include <linux/hdreg.h>
67 #define FS_NOCOW_FL 0x00800000 /* Do not cow file */
70 #if defined(CONFIG_FALLOCATE_PUNCH_HOLE) || defined(CONFIG_FALLOCATE_ZERO_RANGE)
71 #include <linux/falloc.h>
73 #if defined (__FreeBSD__) || defined(__FreeBSD_kernel__)
79 #include <sys/ioctl.h>
80 #include <sys/disklabel.h>
85 #include <sys/ioctl.h>
86 #include <sys/disklabel.h>
92 #include <sys/ioctl.h>
93 #include <sys/diskslice.h>
100 //#define DEBUG_BLOCK
103 # define DEBUG_BLOCK_PRINT 1
105 # define DEBUG_BLOCK_PRINT 0
107 #define DPRINTF(fmt, ...) \
109 if (DEBUG_BLOCK_PRINT) { \
110 printf(fmt, ## __VA_ARGS__); \
114 /* OS X does not have O_DSYNC */
117 #define O_DSYNC O_SYNC
118 #elif defined(O_FSYNC)
119 #define O_DSYNC O_FSYNC
123 /* Approximate O_DIRECT with O_DSYNC if O_DIRECT isn't available */
125 #define O_DIRECT O_DSYNC
131 #define MAX_BLOCKSIZE 4096
133 typedef struct BDRVRawState
{
139 #ifdef CONFIG_LINUX_AIO
147 bool has_write_zeroes
:1;
148 bool discard_zeroes
:1;
150 bool needs_alignment
;
153 typedef struct BDRVRawReopenState
{
156 #ifdef CONFIG_LINUX_AIO
159 } BDRVRawReopenState
;
161 static int fd_open(BlockDriverState
*bs
);
162 static int64_t raw_getlength(BlockDriverState
*bs
);
164 typedef struct RawPosixAIOData
{
165 BlockDriverState
*bs
;
168 struct iovec
*aio_iov
;
173 #define aio_ioctl_cmd aio_nbytes /* for QEMU_AIO_IOCTL */
178 #if defined(__FreeBSD__) || defined(__FreeBSD_kernel__)
179 static int cdrom_reopen(BlockDriverState
*bs
);
182 #if defined(__NetBSD__)
183 static int raw_normalize_devicepath(const char **filename
)
185 static char namebuf
[PATH_MAX
];
186 const char *dp
, *fname
;
190 dp
= strrchr(fname
, '/');
191 if (lstat(fname
, &sb
) < 0) {
192 fprintf(stderr
, "%s: stat failed: %s\n",
193 fname
, strerror(errno
));
197 if (!S_ISBLK(sb
.st_mode
)) {
202 snprintf(namebuf
, PATH_MAX
, "r%s", fname
);
204 snprintf(namebuf
, PATH_MAX
, "%.*s/r%s",
205 (int)(dp
- fname
), fname
, dp
+ 1);
207 fprintf(stderr
, "%s is a block device", fname
);
209 fprintf(stderr
, ", using %s\n", *filename
);
214 static int raw_normalize_devicepath(const char **filename
)
221 * Get logical block size via ioctl. On success store it in @sector_size_p.
223 static int probe_logical_blocksize(int fd
, unsigned int *sector_size_p
)
225 unsigned int sector_size
;
226 bool success
= false;
230 /* Try a few ioctls to get the right size */
232 if (ioctl(fd
, BLKSSZGET
, §or_size
) >= 0) {
233 *sector_size_p
= sector_size
;
237 #ifdef DKIOCGETBLOCKSIZE
238 if (ioctl(fd
, DKIOCGETBLOCKSIZE
, §or_size
) >= 0) {
239 *sector_size_p
= sector_size
;
243 #ifdef DIOCGSECTORSIZE
244 if (ioctl(fd
, DIOCGSECTORSIZE
, §or_size
) >= 0) {
245 *sector_size_p
= sector_size
;
250 return success
? 0 : -errno
;
254 * Get physical block size of @fd.
255 * On success, store it in @blk_size and return 0.
256 * On failure, return -errno.
258 static int probe_physical_blocksize(int fd
, unsigned int *blk_size
)
261 if (ioctl(fd
, BLKPBSZGET
, blk_size
) < 0) {
270 /* Check if read is allowed with given memory buffer and length.
272 * This function is used to check O_DIRECT memory buffer and request alignment.
274 static bool raw_is_io_aligned(int fd
, void *buf
, size_t len
)
276 ssize_t ret
= pread(fd
, buf
, len
, 0);
283 /* The Linux kernel returns EINVAL for misaligned O_DIRECT reads. Ignore
284 * other errors (e.g. real I/O error), which could happen on a failed
285 * drive, since we only care about probing alignment.
287 if (errno
!= EINVAL
) {
295 static void raw_probe_alignment(BlockDriverState
*bs
, int fd
, Error
**errp
)
297 BDRVRawState
*s
= bs
->opaque
;
299 size_t max_align
= MAX(MAX_BLOCKSIZE
, getpagesize());
301 /* For SCSI generic devices the alignment is not really used.
302 With buffered I/O, we don't have any restrictions. */
303 if (bdrv_is_sg(bs
) || !s
->needs_alignment
) {
304 bs
->request_alignment
= 1;
309 bs
->request_alignment
= 0;
311 /* Let's try to use the logical blocksize for the alignment. */
312 if (probe_logical_blocksize(fd
, &bs
->request_alignment
) < 0) {
313 bs
->request_alignment
= 0;
318 if (xfsctl(NULL
, fd
, XFS_IOC_DIOINFO
, &da
) >= 0) {
319 bs
->request_alignment
= da
.d_miniosz
;
320 /* The kernel returns wrong information for d_mem */
321 /* s->buf_align = da.d_mem; */
326 /* If we could not get the sizes so far, we can only guess them */
329 buf
= qemu_memalign(max_align
, 2 * max_align
);
330 for (align
= 512; align
<= max_align
; align
<<= 1) {
331 if (raw_is_io_aligned(fd
, buf
+ align
, max_align
)) {
332 s
->buf_align
= align
;
339 if (!bs
->request_alignment
) {
341 buf
= qemu_memalign(s
->buf_align
, max_align
);
342 for (align
= 512; align
<= max_align
; align
<<= 1) {
343 if (raw_is_io_aligned(fd
, buf
, align
)) {
344 bs
->request_alignment
= align
;
351 if (!s
->buf_align
|| !bs
->request_alignment
) {
352 error_setg(errp
, "Could not find working O_DIRECT alignment. "
353 "Try cache.direct=off.");
357 static void raw_parse_flags(int bdrv_flags
, int *open_flags
)
359 assert(open_flags
!= NULL
);
361 *open_flags
|= O_BINARY
;
362 *open_flags
&= ~O_ACCMODE
;
363 if (bdrv_flags
& BDRV_O_RDWR
) {
364 *open_flags
|= O_RDWR
;
366 *open_flags
|= O_RDONLY
;
369 /* Use O_DSYNC for write-through caching, no flags for write-back caching,
370 * and O_DIRECT for no caching. */
371 if ((bdrv_flags
& BDRV_O_NOCACHE
)) {
372 *open_flags
|= O_DIRECT
;
376 static void raw_detach_aio_context(BlockDriverState
*bs
)
378 #ifdef CONFIG_LINUX_AIO
379 BDRVRawState
*s
= bs
->opaque
;
382 laio_detach_aio_context(s
->aio_ctx
, bdrv_get_aio_context(bs
));
387 static void raw_attach_aio_context(BlockDriverState
*bs
,
388 AioContext
*new_context
)
390 #ifdef CONFIG_LINUX_AIO
391 BDRVRawState
*s
= bs
->opaque
;
394 laio_attach_aio_context(s
->aio_ctx
, new_context
);
399 #ifdef CONFIG_LINUX_AIO
400 static int raw_set_aio(void **aio_ctx
, int *use_aio
, int bdrv_flags
)
403 assert(aio_ctx
!= NULL
);
404 assert(use_aio
!= NULL
);
406 * Currently Linux do AIO only for files opened with O_DIRECT
407 * specified so check NOCACHE flag too
409 if ((bdrv_flags
& (BDRV_O_NOCACHE
|BDRV_O_NATIVE_AIO
)) ==
410 (BDRV_O_NOCACHE
|BDRV_O_NATIVE_AIO
)) {
412 /* if non-NULL, laio_init() has already been run */
413 if (*aio_ctx
== NULL
) {
414 *aio_ctx
= laio_init();
431 static void raw_parse_filename(const char *filename
, QDict
*options
,
434 /* The filename does not have to be prefixed by the protocol name, since
435 * "file" is the default protocol; therefore, the return value of this
436 * function call can be ignored. */
437 strstart(filename
, "file:", &filename
);
439 qdict_put_obj(options
, "filename", QOBJECT(qstring_from_str(filename
)));
442 static QemuOptsList raw_runtime_opts
= {
444 .head
= QTAILQ_HEAD_INITIALIZER(raw_runtime_opts
.head
),
448 .type
= QEMU_OPT_STRING
,
449 .help
= "File name of the image",
451 { /* end of list */ }
455 static int raw_open_common(BlockDriverState
*bs
, QDict
*options
,
456 int bdrv_flags
, int open_flags
, Error
**errp
)
458 BDRVRawState
*s
= bs
->opaque
;
460 Error
*local_err
= NULL
;
461 const char *filename
= NULL
;
465 opts
= qemu_opts_create(&raw_runtime_opts
, NULL
, 0, &error_abort
);
466 qemu_opts_absorb_qdict(opts
, options
, &local_err
);
468 error_propagate(errp
, local_err
);
473 filename
= qemu_opt_get(opts
, "filename");
475 ret
= raw_normalize_devicepath(&filename
);
477 error_setg_errno(errp
, -ret
, "Could not normalize device path");
481 s
->open_flags
= open_flags
;
482 raw_parse_flags(bdrv_flags
, &s
->open_flags
);
485 fd
= qemu_open(filename
, s
->open_flags
, 0644);
495 #ifdef CONFIG_LINUX_AIO
496 if (raw_set_aio(&s
->aio_ctx
, &s
->use_aio
, bdrv_flags
)) {
499 error_setg_errno(errp
, -ret
, "Could not set AIO state");
502 if (!s
->use_aio
&& (bdrv_flags
& BDRV_O_NATIVE_AIO
)) {
503 error_setg(errp
, "aio=native was specified, but it requires "
504 "cache.direct=on, which was not specified.");
509 if (bdrv_flags
& BDRV_O_NATIVE_AIO
) {
510 error_setg(errp
, "aio=native was specified, but is not supported "
515 #endif /* !defined(CONFIG_LINUX_AIO) */
517 s
->has_discard
= true;
518 s
->has_write_zeroes
= true;
519 if ((bs
->open_flags
& BDRV_O_NOCACHE
) != 0) {
520 s
->needs_alignment
= true;
523 if (fstat(s
->fd
, &st
) < 0) {
525 error_setg_errno(errp
, errno
, "Could not stat file");
528 if (S_ISREG(st
.st_mode
)) {
529 s
->discard_zeroes
= true;
530 s
->has_fallocate
= true;
532 if (S_ISBLK(st
.st_mode
)) {
533 #ifdef BLKDISCARDZEROES
535 if (ioctl(s
->fd
, BLKDISCARDZEROES
, &arg
) == 0 && arg
) {
536 s
->discard_zeroes
= true;
540 /* On Linux 3.10, BLKDISCARD leaves stale data in the page cache. Do
541 * not rely on the contents of discarded blocks unless using O_DIRECT.
542 * Same for BLKZEROOUT.
544 if (!(bs
->open_flags
& BDRV_O_NOCACHE
)) {
545 s
->discard_zeroes
= false;
546 s
->has_write_zeroes
= false;
551 if (S_ISCHR(st
.st_mode
)) {
553 * The file is a char device (disk), which on FreeBSD isn't behind
554 * a pager, so force all requests to be aligned. This is needed
555 * so QEMU makes sure all IO operations on the device are aligned
556 * to sector size, or else FreeBSD will reject them with EINVAL.
558 s
->needs_alignment
= true;
563 if (platform_test_xfs_fd(s
->fd
)) {
568 raw_attach_aio_context(bs
, bdrv_get_aio_context(bs
));
572 if (filename
&& (bdrv_flags
& BDRV_O_TEMPORARY
)) {
579 static int raw_open(BlockDriverState
*bs
, QDict
*options
, int flags
,
582 BDRVRawState
*s
= bs
->opaque
;
583 Error
*local_err
= NULL
;
586 s
->type
= FTYPE_FILE
;
587 ret
= raw_open_common(bs
, options
, flags
, 0, &local_err
);
589 error_propagate(errp
, local_err
);
594 static int raw_reopen_prepare(BDRVReopenState
*state
,
595 BlockReopenQueue
*queue
, Error
**errp
)
598 BDRVRawReopenState
*raw_s
;
600 Error
*local_err
= NULL
;
602 assert(state
!= NULL
);
603 assert(state
->bs
!= NULL
);
605 s
= state
->bs
->opaque
;
607 state
->opaque
= g_new0(BDRVRawReopenState
, 1);
608 raw_s
= state
->opaque
;
610 #ifdef CONFIG_LINUX_AIO
611 raw_s
->use_aio
= s
->use_aio
;
613 /* we can use s->aio_ctx instead of a copy, because the use_aio flag is
614 * valid in the 'false' condition even if aio_ctx is set, and raw_set_aio()
615 * won't override aio_ctx if aio_ctx is non-NULL */
616 if (raw_set_aio(&s
->aio_ctx
, &raw_s
->use_aio
, state
->flags
)) {
617 error_setg(errp
, "Could not set AIO state");
622 if (s
->type
== FTYPE_CD
) {
623 raw_s
->open_flags
|= O_NONBLOCK
;
626 raw_parse_flags(state
->flags
, &raw_s
->open_flags
);
630 int fcntl_flags
= O_APPEND
| O_NONBLOCK
;
632 fcntl_flags
|= O_NOATIME
;
636 /* Not all operating systems have O_ASYNC, and those that don't
637 * will not let us track the state into raw_s->open_flags (typically
638 * you achieve the same effect with an ioctl, for example I_SETSIG
639 * on Solaris). But we do not use O_ASYNC, so that's fine.
641 assert((s
->open_flags
& O_ASYNC
) == 0);
644 if ((raw_s
->open_flags
& ~fcntl_flags
) == (s
->open_flags
& ~fcntl_flags
)) {
645 /* dup the original fd */
646 /* TODO: use qemu fcntl wrapper */
647 #ifdef F_DUPFD_CLOEXEC
648 raw_s
->fd
= fcntl(s
->fd
, F_DUPFD_CLOEXEC
, 0);
650 raw_s
->fd
= dup(s
->fd
);
651 if (raw_s
->fd
!= -1) {
652 qemu_set_cloexec(raw_s
->fd
);
655 if (raw_s
->fd
>= 0) {
656 ret
= fcntl_setfl(raw_s
->fd
, raw_s
->open_flags
);
658 qemu_close(raw_s
->fd
);
664 /* If we cannot use fcntl, or fcntl failed, fall back to qemu_open() */
665 if (raw_s
->fd
== -1) {
666 const char *normalized_filename
= state
->bs
->filename
;
667 ret
= raw_normalize_devicepath(&normalized_filename
);
669 error_setg_errno(errp
, -ret
, "Could not normalize device path");
671 assert(!(raw_s
->open_flags
& O_CREAT
));
672 raw_s
->fd
= qemu_open(normalized_filename
, raw_s
->open_flags
);
673 if (raw_s
->fd
== -1) {
674 error_setg_errno(errp
, errno
, "Could not reopen file");
680 /* Fail already reopen_prepare() if we can't get a working O_DIRECT
681 * alignment with the new fd. */
682 if (raw_s
->fd
!= -1) {
683 raw_probe_alignment(state
->bs
, raw_s
->fd
, &local_err
);
685 qemu_close(raw_s
->fd
);
687 error_propagate(errp
, local_err
);
695 static void raw_reopen_commit(BDRVReopenState
*state
)
697 BDRVRawReopenState
*raw_s
= state
->opaque
;
698 BDRVRawState
*s
= state
->bs
->opaque
;
700 s
->open_flags
= raw_s
->open_flags
;
704 #ifdef CONFIG_LINUX_AIO
705 s
->use_aio
= raw_s
->use_aio
;
708 g_free(state
->opaque
);
709 state
->opaque
= NULL
;
713 static void raw_reopen_abort(BDRVReopenState
*state
)
715 BDRVRawReopenState
*raw_s
= state
->opaque
;
717 /* nothing to do if NULL, we didn't get far enough */
722 if (raw_s
->fd
>= 0) {
723 qemu_close(raw_s
->fd
);
726 g_free(state
->opaque
);
727 state
->opaque
= NULL
;
730 static void raw_refresh_limits(BlockDriverState
*bs
, Error
**errp
)
732 BDRVRawState
*s
= bs
->opaque
;
734 raw_probe_alignment(bs
, s
->fd
, errp
);
735 bs
->bl
.min_mem_alignment
= s
->buf_align
;
736 bs
->bl
.opt_mem_alignment
= MAX(s
->buf_align
, getpagesize());
739 static int check_for_dasd(int fd
)
742 struct dasd_information2_t info
= {0};
744 return ioctl(fd
, BIODASDINFO2
, &info
);
751 * Try to get @bs's logical and physical block size.
752 * On success, store them in @bsz and return zero.
753 * On failure, return negative errno.
755 static int hdev_probe_blocksizes(BlockDriverState
*bs
, BlockSizes
*bsz
)
757 BDRVRawState
*s
= bs
->opaque
;
760 /* If DASD, get blocksizes */
761 if (check_for_dasd(s
->fd
) < 0) {
764 ret
= probe_logical_blocksize(s
->fd
, &bsz
->log
);
768 return probe_physical_blocksize(s
->fd
, &bsz
->phys
);
772 * Try to get @bs's geometry: cyls, heads, sectors.
773 * On success, store them in @geo and return 0.
774 * On failure return -errno.
775 * (Allows block driver to assign default geometry values that guest sees)
778 static int hdev_probe_geometry(BlockDriverState
*bs
, HDGeometry
*geo
)
780 BDRVRawState
*s
= bs
->opaque
;
781 struct hd_geometry ioctl_geo
= {0};
783 /* If DASD, get its geometry */
784 if (check_for_dasd(s
->fd
) < 0) {
787 if (ioctl(s
->fd
, HDIO_GETGEO
, &ioctl_geo
) < 0) {
790 /* HDIO_GETGEO may return success even though geo contains zeros
791 (e.g. certain multipath setups) */
792 if (!ioctl_geo
.heads
|| !ioctl_geo
.sectors
|| !ioctl_geo
.cylinders
) {
795 /* Do not return a geometry for partition */
796 if (ioctl_geo
.start
!= 0) {
799 geo
->heads
= ioctl_geo
.heads
;
800 geo
->sectors
= ioctl_geo
.sectors
;
801 geo
->cylinders
= ioctl_geo
.cylinders
;
805 #else /* __linux__ */
806 static int hdev_probe_geometry(BlockDriverState
*bs
, HDGeometry
*geo
)
812 static ssize_t
handle_aiocb_ioctl(RawPosixAIOData
*aiocb
)
816 ret
= ioctl(aiocb
->aio_fildes
, aiocb
->aio_ioctl_cmd
, aiocb
->aio_ioctl_buf
);
824 static ssize_t
handle_aiocb_flush(RawPosixAIOData
*aiocb
)
828 ret
= qemu_fdatasync(aiocb
->aio_fildes
);
837 static bool preadv_present
= true;
840 qemu_preadv(int fd
, const struct iovec
*iov
, int nr_iov
, off_t offset
)
842 return preadv(fd
, iov
, nr_iov
, offset
);
846 qemu_pwritev(int fd
, const struct iovec
*iov
, int nr_iov
, off_t offset
)
848 return pwritev(fd
, iov
, nr_iov
, offset
);
853 static bool preadv_present
= false;
856 qemu_preadv(int fd
, const struct iovec
*iov
, int nr_iov
, off_t offset
)
862 qemu_pwritev(int fd
, const struct iovec
*iov
, int nr_iov
, off_t offset
)
869 static ssize_t
handle_aiocb_rw_vector(RawPosixAIOData
*aiocb
)
874 if (aiocb
->aio_type
& QEMU_AIO_WRITE
)
875 len
= qemu_pwritev(aiocb
->aio_fildes
,
880 len
= qemu_preadv(aiocb
->aio_fildes
,
884 } while (len
== -1 && errno
== EINTR
);
893 * Read/writes the data to/from a given linear buffer.
895 * Returns the number of bytes handles or -errno in case of an error. Short
896 * reads are only returned if the end of the file is reached.
898 static ssize_t
handle_aiocb_rw_linear(RawPosixAIOData
*aiocb
, char *buf
)
903 while (offset
< aiocb
->aio_nbytes
) {
904 if (aiocb
->aio_type
& QEMU_AIO_WRITE
) {
905 len
= pwrite(aiocb
->aio_fildes
,
906 (const char *)buf
+ offset
,
907 aiocb
->aio_nbytes
- offset
,
908 aiocb
->aio_offset
+ offset
);
910 len
= pread(aiocb
->aio_fildes
,
912 aiocb
->aio_nbytes
- offset
,
913 aiocb
->aio_offset
+ offset
);
915 if (len
== -1 && errno
== EINTR
) {
917 } else if (len
== -1 && errno
== EINVAL
&&
918 (aiocb
->bs
->open_flags
& BDRV_O_NOCACHE
) &&
919 !(aiocb
->aio_type
& QEMU_AIO_WRITE
) &&
921 /* O_DIRECT pread() may fail with EINVAL when offset is unaligned
922 * after a short read. Assume that O_DIRECT short reads only occur
923 * at EOF. Therefore this is a short read, not an I/O error.
926 } else if (len
== -1) {
929 } else if (len
== 0) {
938 static ssize_t
handle_aiocb_rw(RawPosixAIOData
*aiocb
)
943 if (!(aiocb
->aio_type
& QEMU_AIO_MISALIGNED
)) {
945 * If there is just a single buffer, and it is properly aligned
946 * we can just use plain pread/pwrite without any problems.
948 if (aiocb
->aio_niov
== 1) {
949 return handle_aiocb_rw_linear(aiocb
, aiocb
->aio_iov
->iov_base
);
952 * We have more than one iovec, and all are properly aligned.
954 * Try preadv/pwritev first and fall back to linearizing the
955 * buffer if it's not supported.
957 if (preadv_present
) {
958 nbytes
= handle_aiocb_rw_vector(aiocb
);
959 if (nbytes
== aiocb
->aio_nbytes
||
960 (nbytes
< 0 && nbytes
!= -ENOSYS
)) {
963 preadv_present
= false;
967 * XXX(hch): short read/write. no easy way to handle the reminder
968 * using these interfaces. For now retry using plain
974 * Ok, we have to do it the hard way, copy all segments into
975 * a single aligned buffer.
977 buf
= qemu_try_blockalign(aiocb
->bs
, aiocb
->aio_nbytes
);
982 if (aiocb
->aio_type
& QEMU_AIO_WRITE
) {
986 for (i
= 0; i
< aiocb
->aio_niov
; ++i
) {
987 memcpy(p
, aiocb
->aio_iov
[i
].iov_base
, aiocb
->aio_iov
[i
].iov_len
);
988 p
+= aiocb
->aio_iov
[i
].iov_len
;
990 assert(p
- buf
== aiocb
->aio_nbytes
);
993 nbytes
= handle_aiocb_rw_linear(aiocb
, buf
);
994 if (!(aiocb
->aio_type
& QEMU_AIO_WRITE
)) {
996 size_t count
= aiocb
->aio_nbytes
, copy
;
999 for (i
= 0; i
< aiocb
->aio_niov
&& count
; ++i
) {
1001 if (copy
> aiocb
->aio_iov
[i
].iov_len
) {
1002 copy
= aiocb
->aio_iov
[i
].iov_len
;
1004 memcpy(aiocb
->aio_iov
[i
].iov_base
, p
, copy
);
1005 assert(count
>= copy
);
1017 static int xfs_write_zeroes(BDRVRawState
*s
, int64_t offset
, uint64_t bytes
)
1019 struct xfs_flock64 fl
;
1022 memset(&fl
, 0, sizeof(fl
));
1023 fl
.l_whence
= SEEK_SET
;
1024 fl
.l_start
= offset
;
1027 if (xfsctl(NULL
, s
->fd
, XFS_IOC_ZERO_RANGE
, &fl
) < 0) {
1029 DPRINTF("cannot write zero range (%s)\n", strerror(errno
));
1036 static int xfs_discard(BDRVRawState
*s
, int64_t offset
, uint64_t bytes
)
1038 struct xfs_flock64 fl
;
1041 memset(&fl
, 0, sizeof(fl
));
1042 fl
.l_whence
= SEEK_SET
;
1043 fl
.l_start
= offset
;
1046 if (xfsctl(NULL
, s
->fd
, XFS_IOC_UNRESVSP64
, &fl
) < 0) {
1048 DPRINTF("cannot punch hole (%s)\n", strerror(errno
));
1056 static int translate_err(int err
)
1058 if (err
== -ENODEV
|| err
== -ENOSYS
|| err
== -EOPNOTSUPP
||
1065 #ifdef CONFIG_FALLOCATE
1066 static int do_fallocate(int fd
, int mode
, off_t offset
, off_t len
)
1069 if (fallocate(fd
, mode
, offset
, len
) == 0) {
1072 } while (errno
== EINTR
);
1073 return translate_err(-errno
);
1077 static ssize_t
handle_aiocb_write_zeroes_block(RawPosixAIOData
*aiocb
)
1080 BDRVRawState
*s
= aiocb
->bs
->opaque
;
1082 if (!s
->has_write_zeroes
) {
1088 uint64_t range
[2] = { aiocb
->aio_offset
, aiocb
->aio_nbytes
};
1089 if (ioctl(aiocb
->aio_fildes
, BLKZEROOUT
, range
) == 0) {
1092 } while (errno
== EINTR
);
1094 ret
= translate_err(-errno
);
1097 if (ret
== -ENOTSUP
) {
1098 s
->has_write_zeroes
= false;
1103 static ssize_t
handle_aiocb_write_zeroes(RawPosixAIOData
*aiocb
)
1105 #if defined(CONFIG_FALLOCATE) || defined(CONFIG_XFS)
1106 BDRVRawState
*s
= aiocb
->bs
->opaque
;
1109 if (aiocb
->aio_type
& QEMU_AIO_BLKDEV
) {
1110 return handle_aiocb_write_zeroes_block(aiocb
);
1115 return xfs_write_zeroes(s
, aiocb
->aio_offset
, aiocb
->aio_nbytes
);
1119 #ifdef CONFIG_FALLOCATE_ZERO_RANGE
1120 if (s
->has_write_zeroes
) {
1121 int ret
= do_fallocate(s
->fd
, FALLOC_FL_ZERO_RANGE
,
1122 aiocb
->aio_offset
, aiocb
->aio_nbytes
);
1123 if (ret
== 0 || ret
!= -ENOTSUP
) {
1126 s
->has_write_zeroes
= false;
1130 #ifdef CONFIG_FALLOCATE_PUNCH_HOLE
1131 if (s
->has_discard
&& s
->has_fallocate
) {
1132 int ret
= do_fallocate(s
->fd
,
1133 FALLOC_FL_PUNCH_HOLE
| FALLOC_FL_KEEP_SIZE
,
1134 aiocb
->aio_offset
, aiocb
->aio_nbytes
);
1136 ret
= do_fallocate(s
->fd
, 0, aiocb
->aio_offset
, aiocb
->aio_nbytes
);
1137 if (ret
== 0 || ret
!= -ENOTSUP
) {
1140 s
->has_fallocate
= false;
1141 } else if (ret
!= -ENOTSUP
) {
1144 s
->has_discard
= false;
1149 #ifdef CONFIG_FALLOCATE
1150 if (s
->has_fallocate
&& aiocb
->aio_offset
>= bdrv_getlength(aiocb
->bs
)) {
1151 int ret
= do_fallocate(s
->fd
, 0, aiocb
->aio_offset
, aiocb
->aio_nbytes
);
1152 if (ret
== 0 || ret
!= -ENOTSUP
) {
1155 s
->has_fallocate
= false;
1162 static ssize_t
handle_aiocb_discard(RawPosixAIOData
*aiocb
)
1164 int ret
= -EOPNOTSUPP
;
1165 BDRVRawState
*s
= aiocb
->bs
->opaque
;
1167 if (!s
->has_discard
) {
1171 if (aiocb
->aio_type
& QEMU_AIO_BLKDEV
) {
1174 uint64_t range
[2] = { aiocb
->aio_offset
, aiocb
->aio_nbytes
};
1175 if (ioctl(aiocb
->aio_fildes
, BLKDISCARD
, range
) == 0) {
1178 } while (errno
== EINTR
);
1185 return xfs_discard(s
, aiocb
->aio_offset
, aiocb
->aio_nbytes
);
1189 #ifdef CONFIG_FALLOCATE_PUNCH_HOLE
1190 ret
= do_fallocate(s
->fd
, FALLOC_FL_PUNCH_HOLE
| FALLOC_FL_KEEP_SIZE
,
1191 aiocb
->aio_offset
, aiocb
->aio_nbytes
);
1195 ret
= translate_err(ret
);
1196 if (ret
== -ENOTSUP
) {
1197 s
->has_discard
= false;
1202 static int aio_worker(void *arg
)
1204 RawPosixAIOData
*aiocb
= arg
;
1207 switch (aiocb
->aio_type
& QEMU_AIO_TYPE_MASK
) {
1209 ret
= handle_aiocb_rw(aiocb
);
1210 if (ret
>= 0 && ret
< aiocb
->aio_nbytes
) {
1211 iov_memset(aiocb
->aio_iov
, aiocb
->aio_niov
, ret
,
1212 0, aiocb
->aio_nbytes
- ret
);
1214 ret
= aiocb
->aio_nbytes
;
1216 if (ret
== aiocb
->aio_nbytes
) {
1218 } else if (ret
>= 0 && ret
< aiocb
->aio_nbytes
) {
1222 case QEMU_AIO_WRITE
:
1223 ret
= handle_aiocb_rw(aiocb
);
1224 if (ret
== aiocb
->aio_nbytes
) {
1226 } else if (ret
>= 0 && ret
< aiocb
->aio_nbytes
) {
1230 case QEMU_AIO_FLUSH
:
1231 ret
= handle_aiocb_flush(aiocb
);
1233 case QEMU_AIO_IOCTL
:
1234 ret
= handle_aiocb_ioctl(aiocb
);
1236 case QEMU_AIO_DISCARD
:
1237 ret
= handle_aiocb_discard(aiocb
);
1239 case QEMU_AIO_WRITE_ZEROES
:
1240 ret
= handle_aiocb_write_zeroes(aiocb
);
1243 fprintf(stderr
, "invalid aio request (0x%x)\n", aiocb
->aio_type
);
1252 static int paio_submit_co(BlockDriverState
*bs
, int fd
,
1253 int64_t sector_num
, QEMUIOVector
*qiov
, int nb_sectors
,
1256 RawPosixAIOData
*acb
= g_new(RawPosixAIOData
, 1);
1260 acb
->aio_type
= type
;
1261 acb
->aio_fildes
= fd
;
1263 acb
->aio_nbytes
= nb_sectors
* BDRV_SECTOR_SIZE
;
1264 acb
->aio_offset
= sector_num
* BDRV_SECTOR_SIZE
;
1267 acb
->aio_iov
= qiov
->iov
;
1268 acb
->aio_niov
= qiov
->niov
;
1269 assert(qiov
->size
== acb
->aio_nbytes
);
1272 trace_paio_submit_co(sector_num
, nb_sectors
, type
);
1273 pool
= aio_get_thread_pool(bdrv_get_aio_context(bs
));
1274 return thread_pool_submit_co(pool
, aio_worker
, acb
);
1277 static BlockAIOCB
*paio_submit(BlockDriverState
*bs
, int fd
,
1278 int64_t sector_num
, QEMUIOVector
*qiov
, int nb_sectors
,
1279 BlockCompletionFunc
*cb
, void *opaque
, int type
)
1281 RawPosixAIOData
*acb
= g_new(RawPosixAIOData
, 1);
1285 acb
->aio_type
= type
;
1286 acb
->aio_fildes
= fd
;
1288 acb
->aio_nbytes
= nb_sectors
* BDRV_SECTOR_SIZE
;
1289 acb
->aio_offset
= sector_num
* BDRV_SECTOR_SIZE
;
1292 acb
->aio_iov
= qiov
->iov
;
1293 acb
->aio_niov
= qiov
->niov
;
1294 assert(qiov
->size
== acb
->aio_nbytes
);
1297 trace_paio_submit(acb
, opaque
, sector_num
, nb_sectors
, type
);
1298 pool
= aio_get_thread_pool(bdrv_get_aio_context(bs
));
1299 return thread_pool_submit_aio(pool
, aio_worker
, acb
, cb
, opaque
);
1302 static BlockAIOCB
*raw_aio_submit(BlockDriverState
*bs
,
1303 int64_t sector_num
, QEMUIOVector
*qiov
, int nb_sectors
,
1304 BlockCompletionFunc
*cb
, void *opaque
, int type
)
1306 BDRVRawState
*s
= bs
->opaque
;
1308 if (fd_open(bs
) < 0)
1312 * Check if the underlying device requires requests to be aligned,
1313 * and if the request we are trying to submit is aligned or not.
1314 * If this is the case tell the low-level driver that it needs
1315 * to copy the buffer.
1317 if (s
->needs_alignment
) {
1318 if (!bdrv_qiov_is_aligned(bs
, qiov
)) {
1319 type
|= QEMU_AIO_MISALIGNED
;
1320 #ifdef CONFIG_LINUX_AIO
1321 } else if (s
->use_aio
) {
1322 return laio_submit(bs
, s
->aio_ctx
, s
->fd
, sector_num
, qiov
,
1323 nb_sectors
, cb
, opaque
, type
);
1328 return paio_submit(bs
, s
->fd
, sector_num
, qiov
, nb_sectors
,
1332 static void raw_aio_plug(BlockDriverState
*bs
)
1334 #ifdef CONFIG_LINUX_AIO
1335 BDRVRawState
*s
= bs
->opaque
;
1337 laio_io_plug(bs
, s
->aio_ctx
);
1342 static void raw_aio_unplug(BlockDriverState
*bs
)
1344 #ifdef CONFIG_LINUX_AIO
1345 BDRVRawState
*s
= bs
->opaque
;
1347 laio_io_unplug(bs
, s
->aio_ctx
, true);
1352 static void raw_aio_flush_io_queue(BlockDriverState
*bs
)
1354 #ifdef CONFIG_LINUX_AIO
1355 BDRVRawState
*s
= bs
->opaque
;
1357 laio_io_unplug(bs
, s
->aio_ctx
, false);
1362 static BlockAIOCB
*raw_aio_readv(BlockDriverState
*bs
,
1363 int64_t sector_num
, QEMUIOVector
*qiov
, int nb_sectors
,
1364 BlockCompletionFunc
*cb
, void *opaque
)
1366 return raw_aio_submit(bs
, sector_num
, qiov
, nb_sectors
,
1367 cb
, opaque
, QEMU_AIO_READ
);
1370 static BlockAIOCB
*raw_aio_writev(BlockDriverState
*bs
,
1371 int64_t sector_num
, QEMUIOVector
*qiov
, int nb_sectors
,
1372 BlockCompletionFunc
*cb
, void *opaque
)
1374 return raw_aio_submit(bs
, sector_num
, qiov
, nb_sectors
,
1375 cb
, opaque
, QEMU_AIO_WRITE
);
1378 static BlockAIOCB
*raw_aio_flush(BlockDriverState
*bs
,
1379 BlockCompletionFunc
*cb
, void *opaque
)
1381 BDRVRawState
*s
= bs
->opaque
;
1383 if (fd_open(bs
) < 0)
1386 return paio_submit(bs
, s
->fd
, 0, NULL
, 0, cb
, opaque
, QEMU_AIO_FLUSH
);
1389 static void raw_close(BlockDriverState
*bs
)
1391 BDRVRawState
*s
= bs
->opaque
;
1393 raw_detach_aio_context(bs
);
1395 #ifdef CONFIG_LINUX_AIO
1397 laio_cleanup(s
->aio_ctx
);
1406 static int raw_truncate(BlockDriverState
*bs
, int64_t offset
)
1408 BDRVRawState
*s
= bs
->opaque
;
1411 if (fstat(s
->fd
, &st
)) {
1415 if (S_ISREG(st
.st_mode
)) {
1416 if (ftruncate(s
->fd
, offset
) < 0) {
1419 } else if (S_ISCHR(st
.st_mode
) || S_ISBLK(st
.st_mode
)) {
1420 if (offset
> raw_getlength(bs
)) {
1431 static int64_t raw_getlength(BlockDriverState
*bs
)
1433 BDRVRawState
*s
= bs
->opaque
;
1439 if (S_ISCHR(st
.st_mode
) || S_ISBLK(st
.st_mode
)) {
1440 struct disklabel dl
;
1442 if (ioctl(fd
, DIOCGDINFO
, &dl
))
1444 return (uint64_t)dl
.d_secsize
*
1445 dl
.d_partitions
[DISKPART(st
.st_rdev
)].p_size
;
1449 #elif defined(__NetBSD__)
1450 static int64_t raw_getlength(BlockDriverState
*bs
)
1452 BDRVRawState
*s
= bs
->opaque
;
1458 if (S_ISCHR(st
.st_mode
) || S_ISBLK(st
.st_mode
)) {
1459 struct dkwedge_info dkw
;
1461 if (ioctl(fd
, DIOCGWEDGEINFO
, &dkw
) != -1) {
1462 return dkw
.dkw_size
* 512;
1464 struct disklabel dl
;
1466 if (ioctl(fd
, DIOCGDINFO
, &dl
))
1468 return (uint64_t)dl
.d_secsize
*
1469 dl
.d_partitions
[DISKPART(st
.st_rdev
)].p_size
;
1474 #elif defined(__sun__)
1475 static int64_t raw_getlength(BlockDriverState
*bs
)
1477 BDRVRawState
*s
= bs
->opaque
;
1478 struct dk_minfo minfo
;
1488 * Use the DKIOCGMEDIAINFO ioctl to read the size.
1490 ret
= ioctl(s
->fd
, DKIOCGMEDIAINFO
, &minfo
);
1492 return minfo
.dki_lbsize
* minfo
.dki_capacity
;
1496 * There are reports that lseek on some devices fails, but
1497 * irc discussion said that contingency on contingency was overkill.
1499 size
= lseek(s
->fd
, 0, SEEK_END
);
1505 #elif defined(CONFIG_BSD)
1506 static int64_t raw_getlength(BlockDriverState
*bs
)
1508 BDRVRawState
*s
= bs
->opaque
;
1512 #if defined (__FreeBSD__) || defined(__FreeBSD_kernel__)
1521 #if defined (__FreeBSD__) || defined(__FreeBSD_kernel__)
1524 if (!fstat(fd
, &sb
) && (S_IFCHR
& sb
.st_mode
)) {
1525 #ifdef DIOCGMEDIASIZE
1526 if (ioctl(fd
, DIOCGMEDIASIZE
, (off_t
*)&size
))
1527 #elif defined(DIOCGPART)
1530 if (ioctl(fd
, DIOCGPART
, &pi
) == 0)
1531 size
= pi
.media_size
;
1537 #if defined(__APPLE__) && defined(__MACH__)
1539 uint64_t sectors
= 0;
1540 uint32_t sector_size
= 0;
1542 if (ioctl(fd
, DKIOCGETBLOCKCOUNT
, §ors
) == 0
1543 && ioctl(fd
, DKIOCGETBLOCKSIZE
, §or_size
) == 0) {
1544 size
= sectors
* sector_size
;
1546 size
= lseek(fd
, 0LL, SEEK_END
);
1553 size
= lseek(fd
, 0LL, SEEK_END
);
1558 #if defined(__FreeBSD__) || defined(__FreeBSD_kernel__)
1561 /* XXX FreeBSD acd returns UINT_MAX sectors for an empty drive */
1562 if (size
== 2048LL * (unsigned)-1)
1564 /* XXX no disc? maybe we need to reopen... */
1565 if (size
<= 0 && !reopened
&& cdrom_reopen(bs
) >= 0) {
1572 size
= lseek(fd
, 0, SEEK_END
);
1580 static int64_t raw_getlength(BlockDriverState
*bs
)
1582 BDRVRawState
*s
= bs
->opaque
;
1591 size
= lseek(s
->fd
, 0, SEEK_END
);
1599 static int64_t raw_get_allocated_file_size(BlockDriverState
*bs
)
1602 BDRVRawState
*s
= bs
->opaque
;
1604 if (fstat(s
->fd
, &st
) < 0) {
1607 return (int64_t)st
.st_blocks
* 512;
1610 static int raw_create(const char *filename
, QemuOpts
*opts
, Error
**errp
)
1614 int64_t total_size
= 0;
1616 PreallocMode prealloc
;
1618 Error
*local_err
= NULL
;
1620 strstart(filename
, "file:", &filename
);
1622 /* Read out options */
1623 total_size
= ROUND_UP(qemu_opt_get_size_del(opts
, BLOCK_OPT_SIZE
, 0),
1625 nocow
= qemu_opt_get_bool(opts
, BLOCK_OPT_NOCOW
, false);
1626 buf
= qemu_opt_get_del(opts
, BLOCK_OPT_PREALLOC
);
1627 prealloc
= qapi_enum_parse(PreallocMode_lookup
, buf
,
1628 PREALLOC_MODE__MAX
, PREALLOC_MODE_OFF
,
1632 error_propagate(errp
, local_err
);
1637 fd
= qemu_open(filename
, O_RDWR
| O_CREAT
| O_TRUNC
| O_BINARY
,
1641 error_setg_errno(errp
, -result
, "Could not create file");
1647 /* Set NOCOW flag to solve performance issue on fs like btrfs.
1648 * This is an optimisation. The FS_IOC_SETFLAGS ioctl return value
1649 * will be ignored since any failure of this operation should not
1650 * block the left work.
1653 if (ioctl(fd
, FS_IOC_GETFLAGS
, &attr
) == 0) {
1654 attr
|= FS_NOCOW_FL
;
1655 ioctl(fd
, FS_IOC_SETFLAGS
, &attr
);
1660 if (ftruncate(fd
, total_size
) != 0) {
1662 error_setg_errno(errp
, -result
, "Could not resize file");
1667 #ifdef CONFIG_POSIX_FALLOCATE
1668 case PREALLOC_MODE_FALLOC
:
1669 /* posix_fallocate() doesn't set errno. */
1670 result
= -posix_fallocate(fd
, 0, total_size
);
1672 error_setg_errno(errp
, -result
,
1673 "Could not preallocate data for the new file");
1677 case PREALLOC_MODE_FULL
:
1679 int64_t num
= 0, left
= total_size
;
1680 buf
= g_malloc0(65536);
1683 num
= MIN(left
, 65536);
1684 result
= write(fd
, buf
, num
);
1687 error_setg_errno(errp
, -result
,
1688 "Could not write to the new file");
1697 error_setg_errno(errp
, -result
,
1698 "Could not flush new file to disk");
1704 case PREALLOC_MODE_OFF
:
1708 error_setg(errp
, "Unsupported preallocation mode: %s",
1709 PreallocMode_lookup
[prealloc
]);
1714 if (qemu_close(fd
) != 0 && result
== 0) {
1716 error_setg_errno(errp
, -result
, "Could not close the new file");
1723 * Find allocation range in @bs around offset @start.
1724 * May change underlying file descriptor's file offset.
1725 * If @start is not in a hole, store @start in @data, and the
1726 * beginning of the next hole in @hole, and return 0.
1727 * If @start is in a non-trailing hole, store @start in @hole and the
1728 * beginning of the next non-hole in @data, and return 0.
1729 * If @start is in a trailing hole or beyond EOF, return -ENXIO.
1730 * If we can't find out, return a negative errno other than -ENXIO.
1732 static int find_allocation(BlockDriverState
*bs
, off_t start
,
1733 off_t
*data
, off_t
*hole
)
1735 #if defined SEEK_HOLE && defined SEEK_DATA
1736 BDRVRawState
*s
= bs
->opaque
;
1741 * D1. offs == start: start is in data
1742 * D2. offs > start: start is in a hole, next data at offs
1743 * D3. offs < 0, errno = ENXIO: either start is in a trailing hole
1744 * or start is beyond EOF
1745 * If the latter happens, the file has been truncated behind
1746 * our back since we opened it. All bets are off then.
1747 * Treating like a trailing hole is simplest.
1748 * D4. offs < 0, errno != ENXIO: we learned nothing
1750 offs
= lseek(s
->fd
, start
, SEEK_DATA
);
1752 return -errno
; /* D3 or D4 */
1754 assert(offs
>= start
);
1757 /* D2: in hole, next data at offs */
1763 /* D1: in data, end not yet known */
1767 * H1. offs == start: start is in a hole
1768 * If this happens here, a hole has been dug behind our back
1769 * since the previous lseek().
1770 * H2. offs > start: either start is in data, next hole at offs,
1771 * or start is in trailing hole, EOF at offs
1772 * Linux treats trailing holes like any other hole: offs ==
1773 * start. Solaris seeks to EOF instead: offs > start (blech).
1774 * If that happens here, a hole has been dug behind our back
1775 * since the previous lseek().
1776 * H3. offs < 0, errno = ENXIO: start is beyond EOF
1777 * If this happens, the file has been truncated behind our
1778 * back since we opened it. Treat it like a trailing hole.
1779 * H4. offs < 0, errno != ENXIO: we learned nothing
1780 * Pretend we know nothing at all, i.e. "forget" about D1.
1782 offs
= lseek(s
->fd
, start
, SEEK_HOLE
);
1784 return -errno
; /* D1 and (H3 or H4) */
1786 assert(offs
>= start
);
1790 * D1 and H2: either in data, next hole at offs, or it was in
1791 * data but is now in a trailing hole. In the latter case,
1792 * all bets are off. Treating it as if it there was data all
1793 * the way to EOF is safe, so simply do that.
1808 * Returns the allocation status of the specified sectors.
1810 * If 'sector_num' is beyond the end of the disk image the return value is 0
1811 * and 'pnum' is set to 0.
1813 * 'pnum' is set to the number of sectors (including and immediately following
1814 * the specified sector) that are known to be in the same
1815 * allocated/unallocated state.
1817 * 'nb_sectors' is the max value 'pnum' should be set to. If nb_sectors goes
1818 * beyond the end of the disk image it will be clamped.
1820 static int64_t coroutine_fn
raw_co_get_block_status(BlockDriverState
*bs
,
1822 int nb_sectors
, int *pnum
,
1823 BlockDriverState
**file
)
1825 off_t start
, data
= 0, hole
= 0;
1834 start
= sector_num
* BDRV_SECTOR_SIZE
;
1835 total_size
= bdrv_getlength(bs
);
1836 if (total_size
< 0) {
1838 } else if (start
>= total_size
) {
1841 } else if (start
+ nb_sectors
* BDRV_SECTOR_SIZE
> total_size
) {
1842 nb_sectors
= DIV_ROUND_UP(total_size
- start
, BDRV_SECTOR_SIZE
);
1845 ret
= find_allocation(bs
, start
, &data
, &hole
);
1846 if (ret
== -ENXIO
) {
1849 ret
= BDRV_BLOCK_ZERO
;
1850 } else if (ret
< 0) {
1851 /* No info available, so pretend there are no holes */
1853 ret
= BDRV_BLOCK_DATA
;
1854 } else if (data
== start
) {
1855 /* On a data extent, compute sectors to the end of the extent,
1856 * possibly including a partial sector at EOF. */
1857 *pnum
= MIN(nb_sectors
, DIV_ROUND_UP(hole
- start
, BDRV_SECTOR_SIZE
));
1858 ret
= BDRV_BLOCK_DATA
;
1860 /* On a hole, compute sectors to the beginning of the next extent. */
1861 assert(hole
== start
);
1862 *pnum
= MIN(nb_sectors
, (data
- start
) / BDRV_SECTOR_SIZE
);
1863 ret
= BDRV_BLOCK_ZERO
;
1866 return ret
| BDRV_BLOCK_OFFSET_VALID
| start
;
1869 static coroutine_fn BlockAIOCB
*raw_aio_discard(BlockDriverState
*bs
,
1870 int64_t sector_num
, int nb_sectors
,
1871 BlockCompletionFunc
*cb
, void *opaque
)
1873 BDRVRawState
*s
= bs
->opaque
;
1875 return paio_submit(bs
, s
->fd
, sector_num
, NULL
, nb_sectors
,
1876 cb
, opaque
, QEMU_AIO_DISCARD
);
1879 static int coroutine_fn
raw_co_write_zeroes(
1880 BlockDriverState
*bs
, int64_t sector_num
,
1881 int nb_sectors
, BdrvRequestFlags flags
)
1883 BDRVRawState
*s
= bs
->opaque
;
1885 if (!(flags
& BDRV_REQ_MAY_UNMAP
)) {
1886 return paio_submit_co(bs
, s
->fd
, sector_num
, NULL
, nb_sectors
,
1887 QEMU_AIO_WRITE_ZEROES
);
1888 } else if (s
->discard_zeroes
) {
1889 return paio_submit_co(bs
, s
->fd
, sector_num
, NULL
, nb_sectors
,
1895 static int raw_get_info(BlockDriverState
*bs
, BlockDriverInfo
*bdi
)
1897 BDRVRawState
*s
= bs
->opaque
;
1899 bdi
->unallocated_blocks_are_zero
= s
->discard_zeroes
;
1900 bdi
->can_write_zeroes_with_unmap
= s
->discard_zeroes
;
1904 static QemuOptsList raw_create_opts
= {
1905 .name
= "raw-create-opts",
1906 .head
= QTAILQ_HEAD_INITIALIZER(raw_create_opts
.head
),
1909 .name
= BLOCK_OPT_SIZE
,
1910 .type
= QEMU_OPT_SIZE
,
1911 .help
= "Virtual disk size"
1914 .name
= BLOCK_OPT_NOCOW
,
1915 .type
= QEMU_OPT_BOOL
,
1916 .help
= "Turn off copy-on-write (valid only on btrfs)"
1919 .name
= BLOCK_OPT_PREALLOC
,
1920 .type
= QEMU_OPT_STRING
,
1921 .help
= "Preallocation mode (allowed values: off, falloc, full)"
1923 { /* end of list */ }
1927 BlockDriver bdrv_file
= {
1928 .format_name
= "file",
1929 .protocol_name
= "file",
1930 .instance_size
= sizeof(BDRVRawState
),
1931 .bdrv_needs_filename
= true,
1932 .bdrv_probe
= NULL
, /* no probe for protocols */
1933 .bdrv_parse_filename
= raw_parse_filename
,
1934 .bdrv_file_open
= raw_open
,
1935 .bdrv_reopen_prepare
= raw_reopen_prepare
,
1936 .bdrv_reopen_commit
= raw_reopen_commit
,
1937 .bdrv_reopen_abort
= raw_reopen_abort
,
1938 .bdrv_close
= raw_close
,
1939 .bdrv_create
= raw_create
,
1940 .bdrv_has_zero_init
= bdrv_has_zero_init_1
,
1941 .bdrv_co_get_block_status
= raw_co_get_block_status
,
1942 .bdrv_co_write_zeroes
= raw_co_write_zeroes
,
1944 .bdrv_aio_readv
= raw_aio_readv
,
1945 .bdrv_aio_writev
= raw_aio_writev
,
1946 .bdrv_aio_flush
= raw_aio_flush
,
1947 .bdrv_aio_discard
= raw_aio_discard
,
1948 .bdrv_refresh_limits
= raw_refresh_limits
,
1949 .bdrv_io_plug
= raw_aio_plug
,
1950 .bdrv_io_unplug
= raw_aio_unplug
,
1951 .bdrv_flush_io_queue
= raw_aio_flush_io_queue
,
1953 .bdrv_truncate
= raw_truncate
,
1954 .bdrv_getlength
= raw_getlength
,
1955 .bdrv_get_info
= raw_get_info
,
1956 .bdrv_get_allocated_file_size
1957 = raw_get_allocated_file_size
,
1959 .bdrv_detach_aio_context
= raw_detach_aio_context
,
1960 .bdrv_attach_aio_context
= raw_attach_aio_context
,
1962 .create_opts
= &raw_create_opts
,
1965 /***********************************************/
1968 #if defined(__APPLE__) && defined(__MACH__)
1969 static kern_return_t
FindEjectableCDMedia( io_iterator_t
*mediaIterator
);
1970 static kern_return_t
GetBSDPath(io_iterator_t mediaIterator
, char *bsdPath
,
1971 CFIndex maxPathSize
, int flags
);
1972 kern_return_t
FindEjectableCDMedia( io_iterator_t
*mediaIterator
)
1974 kern_return_t kernResult
;
1975 mach_port_t masterPort
;
1976 CFMutableDictionaryRef classesToMatch
;
1978 kernResult
= IOMasterPort( MACH_PORT_NULL
, &masterPort
);
1979 if ( KERN_SUCCESS
!= kernResult
) {
1980 printf( "IOMasterPort returned %d\n", kernResult
);
1983 classesToMatch
= IOServiceMatching( kIOCDMediaClass
);
1984 if ( classesToMatch
== NULL
) {
1985 printf( "IOServiceMatching returned a NULL dictionary.\n" );
1987 CFDictionarySetValue( classesToMatch
, CFSTR( kIOMediaEjectableKey
), kCFBooleanTrue
);
1989 kernResult
= IOServiceGetMatchingServices( masterPort
, classesToMatch
, mediaIterator
);
1990 if ( KERN_SUCCESS
!= kernResult
)
1992 printf( "IOServiceGetMatchingServices returned %d\n", kernResult
);
1998 kern_return_t
GetBSDPath(io_iterator_t mediaIterator
, char *bsdPath
,
1999 CFIndex maxPathSize
, int flags
)
2001 io_object_t nextMedia
;
2002 kern_return_t kernResult
= KERN_FAILURE
;
2004 nextMedia
= IOIteratorNext( mediaIterator
);
2007 CFTypeRef bsdPathAsCFString
;
2008 bsdPathAsCFString
= IORegistryEntryCreateCFProperty( nextMedia
, CFSTR( kIOBSDNameKey
), kCFAllocatorDefault
, 0 );
2009 if ( bsdPathAsCFString
) {
2010 size_t devPathLength
;
2011 strcpy( bsdPath
, _PATH_DEV
);
2012 if (flags
& BDRV_O_NOCACHE
) {
2013 strcat(bsdPath
, "r");
2015 devPathLength
= strlen( bsdPath
);
2016 if ( CFStringGetCString( bsdPathAsCFString
, bsdPath
+ devPathLength
, maxPathSize
- devPathLength
, kCFStringEncodingASCII
) ) {
2017 kernResult
= KERN_SUCCESS
;
2019 CFRelease( bsdPathAsCFString
);
2021 IOObjectRelease( nextMedia
);
2029 static int hdev_probe_device(const char *filename
)
2033 /* allow a dedicated CD-ROM driver to match with a higher priority */
2034 if (strstart(filename
, "/dev/cdrom", NULL
))
2037 if (stat(filename
, &st
) >= 0 &&
2038 (S_ISCHR(st
.st_mode
) || S_ISBLK(st
.st_mode
))) {
2045 static int check_hdev_writable(BDRVRawState
*s
)
2047 #if defined(BLKROGET)
2048 /* Linux block devices can be configured "read-only" using blockdev(8).
2049 * This is independent of device node permissions and therefore open(2)
2050 * with O_RDWR succeeds. Actual writes fail with EPERM.
2052 * bdrv_open() is supposed to fail if the disk is read-only. Explicitly
2053 * check for read-only block devices so that Linux block devices behave
2059 if (fstat(s
->fd
, &st
)) {
2063 if (!S_ISBLK(st
.st_mode
)) {
2067 if (ioctl(s
->fd
, BLKROGET
, &readonly
) < 0) {
2074 #endif /* defined(BLKROGET) */
2078 static void hdev_parse_filename(const char *filename
, QDict
*options
,
2081 /* The prefix is optional, just as for "file". */
2082 strstart(filename
, "host_device:", &filename
);
2084 qdict_put_obj(options
, "filename", QOBJECT(qstring_from_str(filename
)));
2087 static bool hdev_is_sg(BlockDriverState
*bs
)
2090 #if defined(__linux__)
2093 struct sg_scsi_id scsiid
;
2096 if (stat(bs
->filename
, &st
) >= 0 && S_ISCHR(st
.st_mode
) &&
2097 !bdrv_ioctl(bs
, SG_GET_VERSION_NUM
, &sg_version
) &&
2098 !bdrv_ioctl(bs
, SG_GET_SCSI_ID
, &scsiid
)) {
2099 DPRINTF("SG device found: type=%d, version=%d\n",
2100 scsiid
.scsi_type
, sg_version
);
2109 static int hdev_open(BlockDriverState
*bs
, QDict
*options
, int flags
,
2112 BDRVRawState
*s
= bs
->opaque
;
2113 Error
*local_err
= NULL
;
2116 #if defined(__APPLE__) && defined(__MACH__)
2117 const char *filename
= qdict_get_str(options
, "filename");
2119 if (strstart(filename
, "/dev/cdrom", NULL
)) {
2120 kern_return_t kernResult
;
2121 io_iterator_t mediaIterator
;
2122 char bsdPath
[ MAXPATHLEN
];
2125 kernResult
= FindEjectableCDMedia( &mediaIterator
);
2126 kernResult
= GetBSDPath(mediaIterator
, bsdPath
, sizeof(bsdPath
),
2128 if ( bsdPath
[ 0 ] != '\0' ) {
2129 strcat(bsdPath
,"s0");
2130 /* some CDs don't have a partition 0 */
2131 fd
= qemu_open(bsdPath
, O_RDONLY
| O_BINARY
| O_LARGEFILE
);
2133 bsdPath
[strlen(bsdPath
)-1] = '1';
2138 qdict_put(options
, "filename", qstring_from_str(filename
));
2141 if ( mediaIterator
)
2142 IOObjectRelease( mediaIterator
);
2146 s
->type
= FTYPE_FILE
;
2148 ret
= raw_open_common(bs
, options
, flags
, 0, &local_err
);
2151 error_propagate(errp
, local_err
);
2156 /* Since this does ioctl the device must be already opened */
2157 bs
->sg
= hdev_is_sg(bs
);
2159 if (flags
& BDRV_O_RDWR
) {
2160 ret
= check_hdev_writable(s
);
2163 error_setg_errno(errp
, -ret
, "The device is not writable");
2171 #if defined(__linux__)
2173 static BlockAIOCB
*hdev_aio_ioctl(BlockDriverState
*bs
,
2174 unsigned long int req
, void *buf
,
2175 BlockCompletionFunc
*cb
, void *opaque
)
2177 BDRVRawState
*s
= bs
->opaque
;
2178 RawPosixAIOData
*acb
;
2181 if (fd_open(bs
) < 0)
2184 acb
= g_new(RawPosixAIOData
, 1);
2186 acb
->aio_type
= QEMU_AIO_IOCTL
;
2187 acb
->aio_fildes
= s
->fd
;
2188 acb
->aio_offset
= 0;
2189 acb
->aio_ioctl_buf
= buf
;
2190 acb
->aio_ioctl_cmd
= req
;
2191 pool
= aio_get_thread_pool(bdrv_get_aio_context(bs
));
2192 return thread_pool_submit_aio(pool
, aio_worker
, acb
, cb
, opaque
);
2196 static int fd_open(BlockDriverState
*bs
)
2198 BDRVRawState
*s
= bs
->opaque
;
2200 /* this is just to ensure s->fd is sane (its called by io ops) */
2206 static coroutine_fn BlockAIOCB
*hdev_aio_discard(BlockDriverState
*bs
,
2207 int64_t sector_num
, int nb_sectors
,
2208 BlockCompletionFunc
*cb
, void *opaque
)
2210 BDRVRawState
*s
= bs
->opaque
;
2212 if (fd_open(bs
) < 0) {
2215 return paio_submit(bs
, s
->fd
, sector_num
, NULL
, nb_sectors
,
2216 cb
, opaque
, QEMU_AIO_DISCARD
|QEMU_AIO_BLKDEV
);
2219 static coroutine_fn
int hdev_co_write_zeroes(BlockDriverState
*bs
,
2220 int64_t sector_num
, int nb_sectors
, BdrvRequestFlags flags
)
2222 BDRVRawState
*s
= bs
->opaque
;
2229 if (!(flags
& BDRV_REQ_MAY_UNMAP
)) {
2230 return paio_submit_co(bs
, s
->fd
, sector_num
, NULL
, nb_sectors
,
2231 QEMU_AIO_WRITE_ZEROES
|QEMU_AIO_BLKDEV
);
2232 } else if (s
->discard_zeroes
) {
2233 return paio_submit_co(bs
, s
->fd
, sector_num
, NULL
, nb_sectors
,
2234 QEMU_AIO_DISCARD
|QEMU_AIO_BLKDEV
);
2239 static int hdev_create(const char *filename
, QemuOpts
*opts
,
2244 struct stat stat_buf
;
2245 int64_t total_size
= 0;
2248 /* This function is used by both protocol block drivers and therefore either
2249 * of these prefixes may be given.
2250 * The return value has to be stored somewhere, otherwise this is an error
2251 * due to -Werror=unused-value. */
2253 strstart(filename
, "host_device:", &filename
) ||
2254 strstart(filename
, "host_cdrom:" , &filename
);
2258 ret
= raw_normalize_devicepath(&filename
);
2260 error_setg_errno(errp
, -ret
, "Could not normalize device path");
2264 /* Read out options */
2265 total_size
= ROUND_UP(qemu_opt_get_size_del(opts
, BLOCK_OPT_SIZE
, 0),
2268 fd
= qemu_open(filename
, O_WRONLY
| O_BINARY
);
2271 error_setg_errno(errp
, -ret
, "Could not open device");
2275 if (fstat(fd
, &stat_buf
) < 0) {
2277 error_setg_errno(errp
, -ret
, "Could not stat device");
2278 } else if (!S_ISBLK(stat_buf
.st_mode
) && !S_ISCHR(stat_buf
.st_mode
)) {
2280 "The given file is neither a block nor a character device");
2282 } else if (lseek(fd
, 0, SEEK_END
) < total_size
) {
2283 error_setg(errp
, "Device is too small");
2291 static BlockDriver bdrv_host_device
= {
2292 .format_name
= "host_device",
2293 .protocol_name
= "host_device",
2294 .instance_size
= sizeof(BDRVRawState
),
2295 .bdrv_needs_filename
= true,
2296 .bdrv_probe_device
= hdev_probe_device
,
2297 .bdrv_parse_filename
= hdev_parse_filename
,
2298 .bdrv_file_open
= hdev_open
,
2299 .bdrv_close
= raw_close
,
2300 .bdrv_reopen_prepare
= raw_reopen_prepare
,
2301 .bdrv_reopen_commit
= raw_reopen_commit
,
2302 .bdrv_reopen_abort
= raw_reopen_abort
,
2303 .bdrv_create
= hdev_create
,
2304 .create_opts
= &raw_create_opts
,
2305 .bdrv_co_write_zeroes
= hdev_co_write_zeroes
,
2307 .bdrv_aio_readv
= raw_aio_readv
,
2308 .bdrv_aio_writev
= raw_aio_writev
,
2309 .bdrv_aio_flush
= raw_aio_flush
,
2310 .bdrv_aio_discard
= hdev_aio_discard
,
2311 .bdrv_refresh_limits
= raw_refresh_limits
,
2312 .bdrv_io_plug
= raw_aio_plug
,
2313 .bdrv_io_unplug
= raw_aio_unplug
,
2314 .bdrv_flush_io_queue
= raw_aio_flush_io_queue
,
2316 .bdrv_truncate
= raw_truncate
,
2317 .bdrv_getlength
= raw_getlength
,
2318 .bdrv_get_info
= raw_get_info
,
2319 .bdrv_get_allocated_file_size
2320 = raw_get_allocated_file_size
,
2321 .bdrv_probe_blocksizes
= hdev_probe_blocksizes
,
2322 .bdrv_probe_geometry
= hdev_probe_geometry
,
2324 .bdrv_detach_aio_context
= raw_detach_aio_context
,
2325 .bdrv_attach_aio_context
= raw_attach_aio_context
,
2327 /* generic scsi device */
2329 .bdrv_aio_ioctl
= hdev_aio_ioctl
,
2333 #if defined(__linux__) || defined(__FreeBSD__) || defined(__FreeBSD_kernel__)
2334 static void cdrom_parse_filename(const char *filename
, QDict
*options
,
2337 /* The prefix is optional, just as for "file". */
2338 strstart(filename
, "host_cdrom:", &filename
);
2340 qdict_put_obj(options
, "filename", QOBJECT(qstring_from_str(filename
)));
2345 static int cdrom_open(BlockDriverState
*bs
, QDict
*options
, int flags
,
2348 BDRVRawState
*s
= bs
->opaque
;
2349 Error
*local_err
= NULL
;
2354 /* open will not fail even if no CD is inserted, so add O_NONBLOCK */
2355 ret
= raw_open_common(bs
, options
, flags
, O_NONBLOCK
, &local_err
);
2357 error_propagate(errp
, local_err
);
2362 static int cdrom_probe_device(const char *filename
)
2368 fd
= qemu_open(filename
, O_RDONLY
| O_NONBLOCK
);
2372 ret
= fstat(fd
, &st
);
2373 if (ret
== -1 || !S_ISBLK(st
.st_mode
)) {
2377 /* Attempt to detect via a CDROM specific ioctl */
2378 ret
= ioctl(fd
, CDROM_DRIVE_STATUS
, CDSL_CURRENT
);
2388 static bool cdrom_is_inserted(BlockDriverState
*bs
)
2390 BDRVRawState
*s
= bs
->opaque
;
2393 ret
= ioctl(s
->fd
, CDROM_DRIVE_STATUS
, CDSL_CURRENT
);
2394 return ret
== CDS_DISC_OK
;
2397 static void cdrom_eject(BlockDriverState
*bs
, bool eject_flag
)
2399 BDRVRawState
*s
= bs
->opaque
;
2402 if (ioctl(s
->fd
, CDROMEJECT
, NULL
) < 0)
2403 perror("CDROMEJECT");
2405 if (ioctl(s
->fd
, CDROMCLOSETRAY
, NULL
) < 0)
2406 perror("CDROMEJECT");
2410 static void cdrom_lock_medium(BlockDriverState
*bs
, bool locked
)
2412 BDRVRawState
*s
= bs
->opaque
;
2414 if (ioctl(s
->fd
, CDROM_LOCKDOOR
, locked
) < 0) {
2416 * Note: an error can happen if the distribution automatically
2419 /* perror("CDROM_LOCKDOOR"); */
2423 static BlockDriver bdrv_host_cdrom
= {
2424 .format_name
= "host_cdrom",
2425 .protocol_name
= "host_cdrom",
2426 .instance_size
= sizeof(BDRVRawState
),
2427 .bdrv_needs_filename
= true,
2428 .bdrv_probe_device
= cdrom_probe_device
,
2429 .bdrv_parse_filename
= cdrom_parse_filename
,
2430 .bdrv_file_open
= cdrom_open
,
2431 .bdrv_close
= raw_close
,
2432 .bdrv_reopen_prepare
= raw_reopen_prepare
,
2433 .bdrv_reopen_commit
= raw_reopen_commit
,
2434 .bdrv_reopen_abort
= raw_reopen_abort
,
2435 .bdrv_create
= hdev_create
,
2436 .create_opts
= &raw_create_opts
,
2438 .bdrv_aio_readv
= raw_aio_readv
,
2439 .bdrv_aio_writev
= raw_aio_writev
,
2440 .bdrv_aio_flush
= raw_aio_flush
,
2441 .bdrv_refresh_limits
= raw_refresh_limits
,
2442 .bdrv_io_plug
= raw_aio_plug
,
2443 .bdrv_io_unplug
= raw_aio_unplug
,
2444 .bdrv_flush_io_queue
= raw_aio_flush_io_queue
,
2446 .bdrv_truncate
= raw_truncate
,
2447 .bdrv_getlength
= raw_getlength
,
2448 .has_variable_length
= true,
2449 .bdrv_get_allocated_file_size
2450 = raw_get_allocated_file_size
,
2452 .bdrv_detach_aio_context
= raw_detach_aio_context
,
2453 .bdrv_attach_aio_context
= raw_attach_aio_context
,
2455 /* removable device support */
2456 .bdrv_is_inserted
= cdrom_is_inserted
,
2457 .bdrv_eject
= cdrom_eject
,
2458 .bdrv_lock_medium
= cdrom_lock_medium
,
2460 /* generic scsi device */
2461 .bdrv_aio_ioctl
= hdev_aio_ioctl
,
2463 #endif /* __linux__ */
2465 #if defined (__FreeBSD__) || defined(__FreeBSD_kernel__)
2466 static int cdrom_open(BlockDriverState
*bs
, QDict
*options
, int flags
,
2469 BDRVRawState
*s
= bs
->opaque
;
2470 Error
*local_err
= NULL
;
2475 ret
= raw_open_common(bs
, options
, flags
, 0, &local_err
);
2478 error_propagate(errp
, local_err
);
2483 /* make sure the door isn't locked at this time */
2484 ioctl(s
->fd
, CDIOCALLOW
);
2488 static int cdrom_probe_device(const char *filename
)
2490 if (strstart(filename
, "/dev/cd", NULL
) ||
2491 strstart(filename
, "/dev/acd", NULL
))
2496 static int cdrom_reopen(BlockDriverState
*bs
)
2498 BDRVRawState
*s
= bs
->opaque
;
2502 * Force reread of possibly changed/newly loaded disc,
2503 * FreeBSD seems to not notice sometimes...
2507 fd
= qemu_open(bs
->filename
, s
->open_flags
, 0644);
2514 /* make sure the door isn't locked at this time */
2515 ioctl(s
->fd
, CDIOCALLOW
);
2519 static bool cdrom_is_inserted(BlockDriverState
*bs
)
2521 return raw_getlength(bs
) > 0;
2524 static void cdrom_eject(BlockDriverState
*bs
, bool eject_flag
)
2526 BDRVRawState
*s
= bs
->opaque
;
2531 (void) ioctl(s
->fd
, CDIOCALLOW
);
2534 if (ioctl(s
->fd
, CDIOCEJECT
) < 0)
2535 perror("CDIOCEJECT");
2537 if (ioctl(s
->fd
, CDIOCCLOSE
) < 0)
2538 perror("CDIOCCLOSE");
2544 static void cdrom_lock_medium(BlockDriverState
*bs
, bool locked
)
2546 BDRVRawState
*s
= bs
->opaque
;
2550 if (ioctl(s
->fd
, (locked
? CDIOCPREVENT
: CDIOCALLOW
)) < 0) {
2552 * Note: an error can happen if the distribution automatically
2555 /* perror("CDROM_LOCKDOOR"); */
2559 static BlockDriver bdrv_host_cdrom
= {
2560 .format_name
= "host_cdrom",
2561 .protocol_name
= "host_cdrom",
2562 .instance_size
= sizeof(BDRVRawState
),
2563 .bdrv_needs_filename
= true,
2564 .bdrv_probe_device
= cdrom_probe_device
,
2565 .bdrv_parse_filename
= cdrom_parse_filename
,
2566 .bdrv_file_open
= cdrom_open
,
2567 .bdrv_close
= raw_close
,
2568 .bdrv_reopen_prepare
= raw_reopen_prepare
,
2569 .bdrv_reopen_commit
= raw_reopen_commit
,
2570 .bdrv_reopen_abort
= raw_reopen_abort
,
2571 .bdrv_create
= hdev_create
,
2572 .create_opts
= &raw_create_opts
,
2574 .bdrv_aio_readv
= raw_aio_readv
,
2575 .bdrv_aio_writev
= raw_aio_writev
,
2576 .bdrv_aio_flush
= raw_aio_flush
,
2577 .bdrv_refresh_limits
= raw_refresh_limits
,
2578 .bdrv_io_plug
= raw_aio_plug
,
2579 .bdrv_io_unplug
= raw_aio_unplug
,
2580 .bdrv_flush_io_queue
= raw_aio_flush_io_queue
,
2582 .bdrv_truncate
= raw_truncate
,
2583 .bdrv_getlength
= raw_getlength
,
2584 .has_variable_length
= true,
2585 .bdrv_get_allocated_file_size
2586 = raw_get_allocated_file_size
,
2588 .bdrv_detach_aio_context
= raw_detach_aio_context
,
2589 .bdrv_attach_aio_context
= raw_attach_aio_context
,
2591 /* removable device support */
2592 .bdrv_is_inserted
= cdrom_is_inserted
,
2593 .bdrv_eject
= cdrom_eject
,
2594 .bdrv_lock_medium
= cdrom_lock_medium
,
2596 #endif /* __FreeBSD__ */
2598 static void bdrv_file_init(void)
2601 * Register all the drivers. Note that order is important, the driver
2602 * registered last will get probed first.
2604 bdrv_register(&bdrv_file
);
2605 bdrv_register(&bdrv_host_device
);
2607 bdrv_register(&bdrv_host_cdrom
);
2609 #if defined(__FreeBSD__) || defined(__FreeBSD_kernel__)
2610 bdrv_register(&bdrv_host_cdrom
);
2614 block_init(bdrv_file_init
);