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/cutils.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 <IOKit/storage/IODVDMedia.h>
49 #include <CoreFoundation/CoreFoundation.h>
53 #define _POSIX_PTHREAD_SEMANTICS 1
57 #include <sys/ioctl.h>
58 #include <sys/param.h>
59 #include <linux/cdrom.h>
62 #include <linux/hdreg.h>
68 #define FS_NOCOW_FL 0x00800000 /* Do not cow file */
71 #if defined(CONFIG_FALLOCATE_PUNCH_HOLE) || defined(CONFIG_FALLOCATE_ZERO_RANGE)
72 #include <linux/falloc.h>
74 #if defined (__FreeBSD__) || defined(__FreeBSD_kernel__)
80 #include <sys/ioctl.h>
81 #include <sys/disklabel.h>
86 #include <sys/ioctl.h>
87 #include <sys/disklabel.h>
93 #include <sys/ioctl.h>
94 #include <sys/diskslice.h>
101 //#define DEBUG_BLOCK
104 # define DEBUG_BLOCK_PRINT 1
106 # define DEBUG_BLOCK_PRINT 0
108 #define DPRINTF(fmt, ...) \
110 if (DEBUG_BLOCK_PRINT) { \
111 printf(fmt, ## __VA_ARGS__); \
115 /* OS X does not have O_DSYNC */
118 #define O_DSYNC O_SYNC
119 #elif defined(O_FSYNC)
120 #define O_DSYNC O_FSYNC
124 /* Approximate O_DIRECT with O_DSYNC if O_DIRECT isn't available */
126 #define O_DIRECT O_DSYNC
132 #define MAX_BLOCKSIZE 4096
134 typedef struct BDRVRawState
{
140 #ifdef CONFIG_LINUX_AIO
142 LinuxAioState
*aio_ctx
;
148 bool has_write_zeroes
:1;
149 bool discard_zeroes
:1;
151 bool needs_alignment
;
154 typedef struct BDRVRawReopenState
{
157 #ifdef CONFIG_LINUX_AIO
160 } BDRVRawReopenState
;
162 static int fd_open(BlockDriverState
*bs
);
163 static int64_t raw_getlength(BlockDriverState
*bs
);
165 typedef struct RawPosixAIOData
{
166 BlockDriverState
*bs
;
169 struct iovec
*aio_iov
;
174 #define aio_ioctl_cmd aio_nbytes /* for QEMU_AIO_IOCTL */
179 #if defined(__FreeBSD__) || defined(__FreeBSD_kernel__)
180 static int cdrom_reopen(BlockDriverState
*bs
);
183 #if defined(__NetBSD__)
184 static int raw_normalize_devicepath(const char **filename
)
186 static char namebuf
[PATH_MAX
];
187 const char *dp
, *fname
;
191 dp
= strrchr(fname
, '/');
192 if (lstat(fname
, &sb
) < 0) {
193 fprintf(stderr
, "%s: stat failed: %s\n",
194 fname
, strerror(errno
));
198 if (!S_ISBLK(sb
.st_mode
)) {
203 snprintf(namebuf
, PATH_MAX
, "r%s", fname
);
205 snprintf(namebuf
, PATH_MAX
, "%.*s/r%s",
206 (int)(dp
- fname
), fname
, dp
+ 1);
208 fprintf(stderr
, "%s is a block device", fname
);
210 fprintf(stderr
, ", using %s\n", *filename
);
215 static int raw_normalize_devicepath(const char **filename
)
222 * Get logical block size via ioctl. On success store it in @sector_size_p.
224 static int probe_logical_blocksize(int fd
, unsigned int *sector_size_p
)
226 unsigned int sector_size
;
227 bool success
= false;
231 /* Try a few ioctls to get the right size */
233 if (ioctl(fd
, BLKSSZGET
, §or_size
) >= 0) {
234 *sector_size_p
= sector_size
;
238 #ifdef DKIOCGETBLOCKSIZE
239 if (ioctl(fd
, DKIOCGETBLOCKSIZE
, §or_size
) >= 0) {
240 *sector_size_p
= sector_size
;
244 #ifdef DIOCGSECTORSIZE
245 if (ioctl(fd
, DIOCGSECTORSIZE
, §or_size
) >= 0) {
246 *sector_size_p
= sector_size
;
251 return success
? 0 : -errno
;
255 * Get physical block size of @fd.
256 * On success, store it in @blk_size and return 0.
257 * On failure, return -errno.
259 static int probe_physical_blocksize(int fd
, unsigned int *blk_size
)
262 if (ioctl(fd
, BLKPBSZGET
, blk_size
) < 0) {
271 /* Check if read is allowed with given memory buffer and length.
273 * This function is used to check O_DIRECT memory buffer and request alignment.
275 static bool raw_is_io_aligned(int fd
, void *buf
, size_t len
)
277 ssize_t ret
= pread(fd
, buf
, len
, 0);
284 /* The Linux kernel returns EINVAL for misaligned O_DIRECT reads. Ignore
285 * other errors (e.g. real I/O error), which could happen on a failed
286 * drive, since we only care about probing alignment.
288 if (errno
!= EINVAL
) {
296 static void raw_probe_alignment(BlockDriverState
*bs
, int fd
, Error
**errp
)
298 BDRVRawState
*s
= bs
->opaque
;
300 size_t max_align
= MAX(MAX_BLOCKSIZE
, getpagesize());
302 /* For SCSI generic devices the alignment is not really used.
303 With buffered I/O, we don't have any restrictions. */
304 if (bdrv_is_sg(bs
) || !s
->needs_alignment
) {
305 bs
->request_alignment
= 1;
310 bs
->request_alignment
= 0;
312 /* Let's try to use the logical blocksize for the alignment. */
313 if (probe_logical_blocksize(fd
, &bs
->request_alignment
) < 0) {
314 bs
->request_alignment
= 0;
319 if (xfsctl(NULL
, fd
, XFS_IOC_DIOINFO
, &da
) >= 0) {
320 bs
->request_alignment
= da
.d_miniosz
;
321 /* The kernel returns wrong information for d_mem */
322 /* s->buf_align = da.d_mem; */
327 /* If we could not get the sizes so far, we can only guess them */
330 buf
= qemu_memalign(max_align
, 2 * max_align
);
331 for (align
= 512; align
<= max_align
; align
<<= 1) {
332 if (raw_is_io_aligned(fd
, buf
+ align
, max_align
)) {
333 s
->buf_align
= align
;
340 if (!bs
->request_alignment
) {
342 buf
= qemu_memalign(s
->buf_align
, max_align
);
343 for (align
= 512; align
<= max_align
; align
<<= 1) {
344 if (raw_is_io_aligned(fd
, buf
, align
)) {
345 bs
->request_alignment
= align
;
352 if (!s
->buf_align
|| !bs
->request_alignment
) {
353 error_setg(errp
, "Could not find working O_DIRECT alignment. "
354 "Try cache.direct=off.");
358 static void raw_parse_flags(int bdrv_flags
, int *open_flags
)
360 assert(open_flags
!= NULL
);
362 *open_flags
|= O_BINARY
;
363 *open_flags
&= ~O_ACCMODE
;
364 if (bdrv_flags
& BDRV_O_RDWR
) {
365 *open_flags
|= O_RDWR
;
367 *open_flags
|= O_RDONLY
;
370 /* Use O_DSYNC for write-through caching, no flags for write-back caching,
371 * and O_DIRECT for no caching. */
372 if ((bdrv_flags
& BDRV_O_NOCACHE
)) {
373 *open_flags
|= O_DIRECT
;
377 static void raw_detach_aio_context(BlockDriverState
*bs
)
379 #ifdef CONFIG_LINUX_AIO
380 BDRVRawState
*s
= bs
->opaque
;
383 laio_detach_aio_context(s
->aio_ctx
, bdrv_get_aio_context(bs
));
388 static void raw_attach_aio_context(BlockDriverState
*bs
,
389 AioContext
*new_context
)
391 #ifdef CONFIG_LINUX_AIO
392 BDRVRawState
*s
= bs
->opaque
;
395 laio_attach_aio_context(s
->aio_ctx
, new_context
);
400 #ifdef CONFIG_LINUX_AIO
401 static int raw_set_aio(LinuxAioState
**aio_ctx
, int *use_aio
, int bdrv_flags
)
404 assert(aio_ctx
!= NULL
);
405 assert(use_aio
!= NULL
);
407 * Currently Linux do AIO only for files opened with O_DIRECT
408 * specified so check NOCACHE flag too
410 if ((bdrv_flags
& (BDRV_O_NOCACHE
|BDRV_O_NATIVE_AIO
)) ==
411 (BDRV_O_NOCACHE
|BDRV_O_NATIVE_AIO
)) {
413 /* if non-NULL, laio_init() has already been run */
414 if (*aio_ctx
== NULL
) {
415 *aio_ctx
= laio_init();
432 static void raw_parse_filename(const char *filename
, QDict
*options
,
435 /* The filename does not have to be prefixed by the protocol name, since
436 * "file" is the default protocol; therefore, the return value of this
437 * function call can be ignored. */
438 strstart(filename
, "file:", &filename
);
440 qdict_put_obj(options
, "filename", QOBJECT(qstring_from_str(filename
)));
443 static QemuOptsList raw_runtime_opts
= {
445 .head
= QTAILQ_HEAD_INITIALIZER(raw_runtime_opts
.head
),
449 .type
= QEMU_OPT_STRING
,
450 .help
= "File name of the image",
452 { /* end of list */ }
456 static int raw_open_common(BlockDriverState
*bs
, QDict
*options
,
457 int bdrv_flags
, int open_flags
, Error
**errp
)
459 BDRVRawState
*s
= bs
->opaque
;
461 Error
*local_err
= NULL
;
462 const char *filename
= NULL
;
466 opts
= qemu_opts_create(&raw_runtime_opts
, NULL
, 0, &error_abort
);
467 qemu_opts_absorb_qdict(opts
, options
, &local_err
);
469 error_propagate(errp
, local_err
);
474 filename
= qemu_opt_get(opts
, "filename");
476 ret
= raw_normalize_devicepath(&filename
);
478 error_setg_errno(errp
, -ret
, "Could not normalize device path");
482 s
->open_flags
= open_flags
;
483 raw_parse_flags(bdrv_flags
, &s
->open_flags
);
486 fd
= qemu_open(filename
, s
->open_flags
, 0644);
496 #ifdef CONFIG_LINUX_AIO
497 if (raw_set_aio(&s
->aio_ctx
, &s
->use_aio
, bdrv_flags
)) {
500 error_setg_errno(errp
, -ret
, "Could not set AIO state");
503 if (!s
->use_aio
&& (bdrv_flags
& BDRV_O_NATIVE_AIO
)) {
504 error_setg(errp
, "aio=native was specified, but it requires "
505 "cache.direct=on, which was not specified.");
510 if (bdrv_flags
& BDRV_O_NATIVE_AIO
) {
511 error_setg(errp
, "aio=native was specified, but is not supported "
516 #endif /* !defined(CONFIG_LINUX_AIO) */
518 s
->has_discard
= true;
519 s
->has_write_zeroes
= true;
520 bs
->supported_zero_flags
= BDRV_REQ_MAY_UNMAP
;
521 if ((bs
->open_flags
& BDRV_O_NOCACHE
) != 0) {
522 s
->needs_alignment
= true;
525 if (fstat(s
->fd
, &st
) < 0) {
527 error_setg_errno(errp
, errno
, "Could not stat file");
530 if (S_ISREG(st
.st_mode
)) {
531 s
->discard_zeroes
= true;
532 s
->has_fallocate
= true;
534 if (S_ISBLK(st
.st_mode
)) {
535 #ifdef BLKDISCARDZEROES
537 if (ioctl(s
->fd
, BLKDISCARDZEROES
, &arg
) == 0 && arg
) {
538 s
->discard_zeroes
= true;
542 /* On Linux 3.10, BLKDISCARD leaves stale data in the page cache. Do
543 * not rely on the contents of discarded blocks unless using O_DIRECT.
544 * Same for BLKZEROOUT.
546 if (!(bs
->open_flags
& BDRV_O_NOCACHE
)) {
547 s
->discard_zeroes
= false;
548 s
->has_write_zeroes
= false;
553 if (S_ISCHR(st
.st_mode
)) {
555 * The file is a char device (disk), which on FreeBSD isn't behind
556 * a pager, so force all requests to be aligned. This is needed
557 * so QEMU makes sure all IO operations on the device are aligned
558 * to sector size, or else FreeBSD will reject them with EINVAL.
560 s
->needs_alignment
= true;
565 if (platform_test_xfs_fd(s
->fd
)) {
570 raw_attach_aio_context(bs
, bdrv_get_aio_context(bs
));
574 if (filename
&& (bdrv_flags
& BDRV_O_TEMPORARY
)) {
581 static int raw_open(BlockDriverState
*bs
, QDict
*options
, int flags
,
584 BDRVRawState
*s
= bs
->opaque
;
585 Error
*local_err
= NULL
;
588 s
->type
= FTYPE_FILE
;
589 ret
= raw_open_common(bs
, options
, flags
, 0, &local_err
);
591 error_propagate(errp
, local_err
);
596 static int raw_reopen_prepare(BDRVReopenState
*state
,
597 BlockReopenQueue
*queue
, Error
**errp
)
600 BDRVRawReopenState
*raw_s
;
602 Error
*local_err
= NULL
;
604 assert(state
!= NULL
);
605 assert(state
->bs
!= NULL
);
607 s
= state
->bs
->opaque
;
609 state
->opaque
= g_new0(BDRVRawReopenState
, 1);
610 raw_s
= state
->opaque
;
612 #ifdef CONFIG_LINUX_AIO
613 raw_s
->use_aio
= s
->use_aio
;
615 /* we can use s->aio_ctx instead of a copy, because the use_aio flag is
616 * valid in the 'false' condition even if aio_ctx is set, and raw_set_aio()
617 * won't override aio_ctx if aio_ctx is non-NULL */
618 if (raw_set_aio(&s
->aio_ctx
, &raw_s
->use_aio
, state
->flags
)) {
619 error_setg(errp
, "Could not set AIO state");
624 if (s
->type
== FTYPE_CD
) {
625 raw_s
->open_flags
|= O_NONBLOCK
;
628 raw_parse_flags(state
->flags
, &raw_s
->open_flags
);
632 int fcntl_flags
= O_APPEND
| O_NONBLOCK
;
634 fcntl_flags
|= O_NOATIME
;
638 /* Not all operating systems have O_ASYNC, and those that don't
639 * will not let us track the state into raw_s->open_flags (typically
640 * you achieve the same effect with an ioctl, for example I_SETSIG
641 * on Solaris). But we do not use O_ASYNC, so that's fine.
643 assert((s
->open_flags
& O_ASYNC
) == 0);
646 if ((raw_s
->open_flags
& ~fcntl_flags
) == (s
->open_flags
& ~fcntl_flags
)) {
647 /* dup the original fd */
648 /* TODO: use qemu fcntl wrapper */
649 #ifdef F_DUPFD_CLOEXEC
650 raw_s
->fd
= fcntl(s
->fd
, F_DUPFD_CLOEXEC
, 0);
652 raw_s
->fd
= dup(s
->fd
);
653 if (raw_s
->fd
!= -1) {
654 qemu_set_cloexec(raw_s
->fd
);
657 if (raw_s
->fd
>= 0) {
658 ret
= fcntl_setfl(raw_s
->fd
, raw_s
->open_flags
);
660 qemu_close(raw_s
->fd
);
666 /* If we cannot use fcntl, or fcntl failed, fall back to qemu_open() */
667 if (raw_s
->fd
== -1) {
668 const char *normalized_filename
= state
->bs
->filename
;
669 ret
= raw_normalize_devicepath(&normalized_filename
);
671 error_setg_errno(errp
, -ret
, "Could not normalize device path");
673 assert(!(raw_s
->open_flags
& O_CREAT
));
674 raw_s
->fd
= qemu_open(normalized_filename
, raw_s
->open_flags
);
675 if (raw_s
->fd
== -1) {
676 error_setg_errno(errp
, errno
, "Could not reopen file");
682 /* Fail already reopen_prepare() if we can't get a working O_DIRECT
683 * alignment with the new fd. */
684 if (raw_s
->fd
!= -1) {
685 raw_probe_alignment(state
->bs
, raw_s
->fd
, &local_err
);
687 qemu_close(raw_s
->fd
);
689 error_propagate(errp
, local_err
);
697 static void raw_reopen_commit(BDRVReopenState
*state
)
699 BDRVRawReopenState
*raw_s
= state
->opaque
;
700 BDRVRawState
*s
= state
->bs
->opaque
;
702 s
->open_flags
= raw_s
->open_flags
;
706 #ifdef CONFIG_LINUX_AIO
707 s
->use_aio
= raw_s
->use_aio
;
710 g_free(state
->opaque
);
711 state
->opaque
= NULL
;
715 static void raw_reopen_abort(BDRVReopenState
*state
)
717 BDRVRawReopenState
*raw_s
= state
->opaque
;
719 /* nothing to do if NULL, we didn't get far enough */
724 if (raw_s
->fd
>= 0) {
725 qemu_close(raw_s
->fd
);
728 g_free(state
->opaque
);
729 state
->opaque
= NULL
;
732 static int hdev_get_max_transfer_length(int fd
)
736 if (ioctl(fd
, BLKSECTGET
, &max_sectors
) == 0) {
746 static void raw_refresh_limits(BlockDriverState
*bs
, Error
**errp
)
748 BDRVRawState
*s
= bs
->opaque
;
751 if (!fstat(s
->fd
, &st
)) {
752 if (S_ISBLK(st
.st_mode
)) {
753 int ret
= hdev_get_max_transfer_length(s
->fd
);
755 bs
->bl
.max_transfer_length
= ret
;
760 raw_probe_alignment(bs
, s
->fd
, errp
);
761 bs
->bl
.min_mem_alignment
= s
->buf_align
;
762 bs
->bl
.opt_mem_alignment
= MAX(s
->buf_align
, getpagesize());
765 static int check_for_dasd(int fd
)
768 struct dasd_information2_t info
= {0};
770 return ioctl(fd
, BIODASDINFO2
, &info
);
777 * Try to get @bs's logical and physical block size.
778 * On success, store them in @bsz and return zero.
779 * On failure, return negative errno.
781 static int hdev_probe_blocksizes(BlockDriverState
*bs
, BlockSizes
*bsz
)
783 BDRVRawState
*s
= bs
->opaque
;
786 /* If DASD, get blocksizes */
787 if (check_for_dasd(s
->fd
) < 0) {
790 ret
= probe_logical_blocksize(s
->fd
, &bsz
->log
);
794 return probe_physical_blocksize(s
->fd
, &bsz
->phys
);
798 * Try to get @bs's geometry: cyls, heads, sectors.
799 * On success, store them in @geo and return 0.
800 * On failure return -errno.
801 * (Allows block driver to assign default geometry values that guest sees)
804 static int hdev_probe_geometry(BlockDriverState
*bs
, HDGeometry
*geo
)
806 BDRVRawState
*s
= bs
->opaque
;
807 struct hd_geometry ioctl_geo
= {0};
809 /* If DASD, get its geometry */
810 if (check_for_dasd(s
->fd
) < 0) {
813 if (ioctl(s
->fd
, HDIO_GETGEO
, &ioctl_geo
) < 0) {
816 /* HDIO_GETGEO may return success even though geo contains zeros
817 (e.g. certain multipath setups) */
818 if (!ioctl_geo
.heads
|| !ioctl_geo
.sectors
|| !ioctl_geo
.cylinders
) {
821 /* Do not return a geometry for partition */
822 if (ioctl_geo
.start
!= 0) {
825 geo
->heads
= ioctl_geo
.heads
;
826 geo
->sectors
= ioctl_geo
.sectors
;
827 geo
->cylinders
= ioctl_geo
.cylinders
;
831 #else /* __linux__ */
832 static int hdev_probe_geometry(BlockDriverState
*bs
, HDGeometry
*geo
)
838 static ssize_t
handle_aiocb_ioctl(RawPosixAIOData
*aiocb
)
842 ret
= ioctl(aiocb
->aio_fildes
, aiocb
->aio_ioctl_cmd
, aiocb
->aio_ioctl_buf
);
850 static ssize_t
handle_aiocb_flush(RawPosixAIOData
*aiocb
)
854 ret
= qemu_fdatasync(aiocb
->aio_fildes
);
863 static bool preadv_present
= true;
866 qemu_preadv(int fd
, const struct iovec
*iov
, int nr_iov
, off_t offset
)
868 return preadv(fd
, iov
, nr_iov
, offset
);
872 qemu_pwritev(int fd
, const struct iovec
*iov
, int nr_iov
, off_t offset
)
874 return pwritev(fd
, iov
, nr_iov
, offset
);
879 static bool preadv_present
= false;
882 qemu_preadv(int fd
, const struct iovec
*iov
, int nr_iov
, off_t offset
)
888 qemu_pwritev(int fd
, const struct iovec
*iov
, int nr_iov
, off_t offset
)
895 static ssize_t
handle_aiocb_rw_vector(RawPosixAIOData
*aiocb
)
900 if (aiocb
->aio_type
& QEMU_AIO_WRITE
)
901 len
= qemu_pwritev(aiocb
->aio_fildes
,
906 len
= qemu_preadv(aiocb
->aio_fildes
,
910 } while (len
== -1 && errno
== EINTR
);
919 * Read/writes the data to/from a given linear buffer.
921 * Returns the number of bytes handles or -errno in case of an error. Short
922 * reads are only returned if the end of the file is reached.
924 static ssize_t
handle_aiocb_rw_linear(RawPosixAIOData
*aiocb
, char *buf
)
929 while (offset
< aiocb
->aio_nbytes
) {
930 if (aiocb
->aio_type
& QEMU_AIO_WRITE
) {
931 len
= pwrite(aiocb
->aio_fildes
,
932 (const char *)buf
+ offset
,
933 aiocb
->aio_nbytes
- offset
,
934 aiocb
->aio_offset
+ offset
);
936 len
= pread(aiocb
->aio_fildes
,
938 aiocb
->aio_nbytes
- offset
,
939 aiocb
->aio_offset
+ offset
);
941 if (len
== -1 && errno
== EINTR
) {
943 } else if (len
== -1 && errno
== EINVAL
&&
944 (aiocb
->bs
->open_flags
& BDRV_O_NOCACHE
) &&
945 !(aiocb
->aio_type
& QEMU_AIO_WRITE
) &&
947 /* O_DIRECT pread() may fail with EINVAL when offset is unaligned
948 * after a short read. Assume that O_DIRECT short reads only occur
949 * at EOF. Therefore this is a short read, not an I/O error.
952 } else if (len
== -1) {
955 } else if (len
== 0) {
964 static ssize_t
handle_aiocb_rw(RawPosixAIOData
*aiocb
)
969 if (!(aiocb
->aio_type
& QEMU_AIO_MISALIGNED
)) {
971 * If there is just a single buffer, and it is properly aligned
972 * we can just use plain pread/pwrite without any problems.
974 if (aiocb
->aio_niov
== 1) {
975 return handle_aiocb_rw_linear(aiocb
, aiocb
->aio_iov
->iov_base
);
978 * We have more than one iovec, and all are properly aligned.
980 * Try preadv/pwritev first and fall back to linearizing the
981 * buffer if it's not supported.
983 if (preadv_present
) {
984 nbytes
= handle_aiocb_rw_vector(aiocb
);
985 if (nbytes
== aiocb
->aio_nbytes
||
986 (nbytes
< 0 && nbytes
!= -ENOSYS
)) {
989 preadv_present
= false;
993 * XXX(hch): short read/write. no easy way to handle the reminder
994 * using these interfaces. For now retry using plain
1000 * Ok, we have to do it the hard way, copy all segments into
1001 * a single aligned buffer.
1003 buf
= qemu_try_blockalign(aiocb
->bs
, aiocb
->aio_nbytes
);
1008 if (aiocb
->aio_type
& QEMU_AIO_WRITE
) {
1012 for (i
= 0; i
< aiocb
->aio_niov
; ++i
) {
1013 memcpy(p
, aiocb
->aio_iov
[i
].iov_base
, aiocb
->aio_iov
[i
].iov_len
);
1014 p
+= aiocb
->aio_iov
[i
].iov_len
;
1016 assert(p
- buf
== aiocb
->aio_nbytes
);
1019 nbytes
= handle_aiocb_rw_linear(aiocb
, buf
);
1020 if (!(aiocb
->aio_type
& QEMU_AIO_WRITE
)) {
1022 size_t count
= aiocb
->aio_nbytes
, copy
;
1025 for (i
= 0; i
< aiocb
->aio_niov
&& count
; ++i
) {
1027 if (copy
> aiocb
->aio_iov
[i
].iov_len
) {
1028 copy
= aiocb
->aio_iov
[i
].iov_len
;
1030 memcpy(aiocb
->aio_iov
[i
].iov_base
, p
, copy
);
1031 assert(count
>= copy
);
1043 static int xfs_write_zeroes(BDRVRawState
*s
, int64_t offset
, uint64_t bytes
)
1045 struct xfs_flock64 fl
;
1048 memset(&fl
, 0, sizeof(fl
));
1049 fl
.l_whence
= SEEK_SET
;
1050 fl
.l_start
= offset
;
1053 if (xfsctl(NULL
, s
->fd
, XFS_IOC_ZERO_RANGE
, &fl
) < 0) {
1055 DPRINTF("cannot write zero range (%s)\n", strerror(errno
));
1062 static int xfs_discard(BDRVRawState
*s
, int64_t offset
, uint64_t bytes
)
1064 struct xfs_flock64 fl
;
1067 memset(&fl
, 0, sizeof(fl
));
1068 fl
.l_whence
= SEEK_SET
;
1069 fl
.l_start
= offset
;
1072 if (xfsctl(NULL
, s
->fd
, XFS_IOC_UNRESVSP64
, &fl
) < 0) {
1074 DPRINTF("cannot punch hole (%s)\n", strerror(errno
));
1082 static int translate_err(int err
)
1084 if (err
== -ENODEV
|| err
== -ENOSYS
|| err
== -EOPNOTSUPP
||
1091 #ifdef CONFIG_FALLOCATE
1092 static int do_fallocate(int fd
, int mode
, off_t offset
, off_t len
)
1095 if (fallocate(fd
, mode
, offset
, len
) == 0) {
1098 } while (errno
== EINTR
);
1099 return translate_err(-errno
);
1103 static ssize_t
handle_aiocb_write_zeroes_block(RawPosixAIOData
*aiocb
)
1106 BDRVRawState
*s
= aiocb
->bs
->opaque
;
1108 if (!s
->has_write_zeroes
) {
1114 uint64_t range
[2] = { aiocb
->aio_offset
, aiocb
->aio_nbytes
};
1115 if (ioctl(aiocb
->aio_fildes
, BLKZEROOUT
, range
) == 0) {
1118 } while (errno
== EINTR
);
1120 ret
= translate_err(-errno
);
1123 if (ret
== -ENOTSUP
) {
1124 s
->has_write_zeroes
= false;
1129 static ssize_t
handle_aiocb_write_zeroes(RawPosixAIOData
*aiocb
)
1131 #if defined(CONFIG_FALLOCATE) || defined(CONFIG_XFS)
1132 BDRVRawState
*s
= aiocb
->bs
->opaque
;
1135 if (aiocb
->aio_type
& QEMU_AIO_BLKDEV
) {
1136 return handle_aiocb_write_zeroes_block(aiocb
);
1141 return xfs_write_zeroes(s
, aiocb
->aio_offset
, aiocb
->aio_nbytes
);
1145 #ifdef CONFIG_FALLOCATE_ZERO_RANGE
1146 if (s
->has_write_zeroes
) {
1147 int ret
= do_fallocate(s
->fd
, FALLOC_FL_ZERO_RANGE
,
1148 aiocb
->aio_offset
, aiocb
->aio_nbytes
);
1149 if (ret
== 0 || ret
!= -ENOTSUP
) {
1152 s
->has_write_zeroes
= false;
1156 #ifdef CONFIG_FALLOCATE_PUNCH_HOLE
1157 if (s
->has_discard
&& s
->has_fallocate
) {
1158 int ret
= do_fallocate(s
->fd
,
1159 FALLOC_FL_PUNCH_HOLE
| FALLOC_FL_KEEP_SIZE
,
1160 aiocb
->aio_offset
, aiocb
->aio_nbytes
);
1162 ret
= do_fallocate(s
->fd
, 0, aiocb
->aio_offset
, aiocb
->aio_nbytes
);
1163 if (ret
== 0 || ret
!= -ENOTSUP
) {
1166 s
->has_fallocate
= false;
1167 } else if (ret
!= -ENOTSUP
) {
1170 s
->has_discard
= false;
1175 #ifdef CONFIG_FALLOCATE
1176 if (s
->has_fallocate
&& aiocb
->aio_offset
>= bdrv_getlength(aiocb
->bs
)) {
1177 int ret
= do_fallocate(s
->fd
, 0, aiocb
->aio_offset
, aiocb
->aio_nbytes
);
1178 if (ret
== 0 || ret
!= -ENOTSUP
) {
1181 s
->has_fallocate
= false;
1188 static ssize_t
handle_aiocb_discard(RawPosixAIOData
*aiocb
)
1190 int ret
= -EOPNOTSUPP
;
1191 BDRVRawState
*s
= aiocb
->bs
->opaque
;
1193 if (!s
->has_discard
) {
1197 if (aiocb
->aio_type
& QEMU_AIO_BLKDEV
) {
1200 uint64_t range
[2] = { aiocb
->aio_offset
, aiocb
->aio_nbytes
};
1201 if (ioctl(aiocb
->aio_fildes
, BLKDISCARD
, range
) == 0) {
1204 } while (errno
== EINTR
);
1211 return xfs_discard(s
, aiocb
->aio_offset
, aiocb
->aio_nbytes
);
1215 #ifdef CONFIG_FALLOCATE_PUNCH_HOLE
1216 ret
= do_fallocate(s
->fd
, FALLOC_FL_PUNCH_HOLE
| FALLOC_FL_KEEP_SIZE
,
1217 aiocb
->aio_offset
, aiocb
->aio_nbytes
);
1221 ret
= translate_err(ret
);
1222 if (ret
== -ENOTSUP
) {
1223 s
->has_discard
= false;
1228 static int aio_worker(void *arg
)
1230 RawPosixAIOData
*aiocb
= arg
;
1233 switch (aiocb
->aio_type
& QEMU_AIO_TYPE_MASK
) {
1235 ret
= handle_aiocb_rw(aiocb
);
1236 if (ret
>= 0 && ret
< aiocb
->aio_nbytes
) {
1237 iov_memset(aiocb
->aio_iov
, aiocb
->aio_niov
, ret
,
1238 0, aiocb
->aio_nbytes
- ret
);
1240 ret
= aiocb
->aio_nbytes
;
1242 if (ret
== aiocb
->aio_nbytes
) {
1244 } else if (ret
>= 0 && ret
< aiocb
->aio_nbytes
) {
1248 case QEMU_AIO_WRITE
:
1249 ret
= handle_aiocb_rw(aiocb
);
1250 if (ret
== aiocb
->aio_nbytes
) {
1252 } else if (ret
>= 0 && ret
< aiocb
->aio_nbytes
) {
1256 case QEMU_AIO_FLUSH
:
1257 ret
= handle_aiocb_flush(aiocb
);
1259 case QEMU_AIO_IOCTL
:
1260 ret
= handle_aiocb_ioctl(aiocb
);
1262 case QEMU_AIO_DISCARD
:
1263 ret
= handle_aiocb_discard(aiocb
);
1265 case QEMU_AIO_WRITE_ZEROES
:
1266 ret
= handle_aiocb_write_zeroes(aiocb
);
1269 fprintf(stderr
, "invalid aio request (0x%x)\n", aiocb
->aio_type
);
1278 static int paio_submit_co(BlockDriverState
*bs
, int fd
,
1279 int64_t offset
, QEMUIOVector
*qiov
,
1280 int count
, int type
)
1282 RawPosixAIOData
*acb
= g_new(RawPosixAIOData
, 1);
1286 acb
->aio_type
= type
;
1287 acb
->aio_fildes
= fd
;
1289 acb
->aio_nbytes
= count
;
1290 acb
->aio_offset
= offset
;
1293 acb
->aio_iov
= qiov
->iov
;
1294 acb
->aio_niov
= qiov
->niov
;
1295 assert(qiov
->size
== count
);
1298 trace_paio_submit_co(offset
, count
, type
);
1299 pool
= aio_get_thread_pool(bdrv_get_aio_context(bs
));
1300 return thread_pool_submit_co(pool
, aio_worker
, acb
);
1303 static BlockAIOCB
*paio_submit(BlockDriverState
*bs
, int fd
,
1304 int64_t sector_num
, QEMUIOVector
*qiov
, int nb_sectors
,
1305 BlockCompletionFunc
*cb
, void *opaque
, int type
)
1307 RawPosixAIOData
*acb
= g_new(RawPosixAIOData
, 1);
1311 acb
->aio_type
= type
;
1312 acb
->aio_fildes
= fd
;
1314 acb
->aio_nbytes
= nb_sectors
* BDRV_SECTOR_SIZE
;
1315 acb
->aio_offset
= sector_num
* BDRV_SECTOR_SIZE
;
1318 acb
->aio_iov
= qiov
->iov
;
1319 acb
->aio_niov
= qiov
->niov
;
1320 assert(qiov
->size
== acb
->aio_nbytes
);
1323 trace_paio_submit(acb
, opaque
, sector_num
, nb_sectors
, type
);
1324 pool
= aio_get_thread_pool(bdrv_get_aio_context(bs
));
1325 return thread_pool_submit_aio(pool
, aio_worker
, acb
, cb
, opaque
);
1328 static int coroutine_fn
raw_co_prw(BlockDriverState
*bs
, uint64_t offset
,
1329 uint64_t bytes
, QEMUIOVector
*qiov
, int type
)
1331 BDRVRawState
*s
= bs
->opaque
;
1333 if (fd_open(bs
) < 0)
1337 * Check if the underlying device requires requests to be aligned,
1338 * and if the request we are trying to submit is aligned or not.
1339 * If this is the case tell the low-level driver that it needs
1340 * to copy the buffer.
1342 if (s
->needs_alignment
) {
1343 if (!bdrv_qiov_is_aligned(bs
, qiov
)) {
1344 type
|= QEMU_AIO_MISALIGNED
;
1345 #ifdef CONFIG_LINUX_AIO
1346 } else if (s
->use_aio
) {
1347 assert(qiov
->size
== bytes
);
1348 return laio_co_submit(bs
, s
->aio_ctx
, s
->fd
, offset
, qiov
, type
);
1353 return paio_submit_co(bs
, s
->fd
, offset
, qiov
, bytes
, type
);
1356 static int coroutine_fn
raw_co_preadv(BlockDriverState
*bs
, uint64_t offset
,
1357 uint64_t bytes
, QEMUIOVector
*qiov
,
1360 return raw_co_prw(bs
, offset
, bytes
, qiov
, QEMU_AIO_READ
);
1363 static int coroutine_fn
raw_co_pwritev(BlockDriverState
*bs
, uint64_t offset
,
1364 uint64_t bytes
, QEMUIOVector
*qiov
,
1368 return raw_co_prw(bs
, offset
, bytes
, qiov
, QEMU_AIO_WRITE
);
1371 static void raw_aio_plug(BlockDriverState
*bs
)
1373 #ifdef CONFIG_LINUX_AIO
1374 BDRVRawState
*s
= bs
->opaque
;
1376 laio_io_plug(bs
, s
->aio_ctx
);
1381 static void raw_aio_unplug(BlockDriverState
*bs
)
1383 #ifdef CONFIG_LINUX_AIO
1384 BDRVRawState
*s
= bs
->opaque
;
1386 laio_io_unplug(bs
, s
->aio_ctx
);
1391 static BlockAIOCB
*raw_aio_flush(BlockDriverState
*bs
,
1392 BlockCompletionFunc
*cb
, void *opaque
)
1394 BDRVRawState
*s
= bs
->opaque
;
1396 if (fd_open(bs
) < 0)
1399 return paio_submit(bs
, s
->fd
, 0, NULL
, 0, cb
, opaque
, QEMU_AIO_FLUSH
);
1402 static void raw_close(BlockDriverState
*bs
)
1404 BDRVRawState
*s
= bs
->opaque
;
1406 raw_detach_aio_context(bs
);
1408 #ifdef CONFIG_LINUX_AIO
1410 laio_cleanup(s
->aio_ctx
);
1419 static int raw_truncate(BlockDriverState
*bs
, int64_t offset
)
1421 BDRVRawState
*s
= bs
->opaque
;
1424 if (fstat(s
->fd
, &st
)) {
1428 if (S_ISREG(st
.st_mode
)) {
1429 if (ftruncate(s
->fd
, offset
) < 0) {
1432 } else if (S_ISCHR(st
.st_mode
) || S_ISBLK(st
.st_mode
)) {
1433 if (offset
> raw_getlength(bs
)) {
1444 static int64_t raw_getlength(BlockDriverState
*bs
)
1446 BDRVRawState
*s
= bs
->opaque
;
1452 if (S_ISCHR(st
.st_mode
) || S_ISBLK(st
.st_mode
)) {
1453 struct disklabel dl
;
1455 if (ioctl(fd
, DIOCGDINFO
, &dl
))
1457 return (uint64_t)dl
.d_secsize
*
1458 dl
.d_partitions
[DISKPART(st
.st_rdev
)].p_size
;
1462 #elif defined(__NetBSD__)
1463 static int64_t raw_getlength(BlockDriverState
*bs
)
1465 BDRVRawState
*s
= bs
->opaque
;
1471 if (S_ISCHR(st
.st_mode
) || S_ISBLK(st
.st_mode
)) {
1472 struct dkwedge_info dkw
;
1474 if (ioctl(fd
, DIOCGWEDGEINFO
, &dkw
) != -1) {
1475 return dkw
.dkw_size
* 512;
1477 struct disklabel dl
;
1479 if (ioctl(fd
, DIOCGDINFO
, &dl
))
1481 return (uint64_t)dl
.d_secsize
*
1482 dl
.d_partitions
[DISKPART(st
.st_rdev
)].p_size
;
1487 #elif defined(__sun__)
1488 static int64_t raw_getlength(BlockDriverState
*bs
)
1490 BDRVRawState
*s
= bs
->opaque
;
1491 struct dk_minfo minfo
;
1501 * Use the DKIOCGMEDIAINFO ioctl to read the size.
1503 ret
= ioctl(s
->fd
, DKIOCGMEDIAINFO
, &minfo
);
1505 return minfo
.dki_lbsize
* minfo
.dki_capacity
;
1509 * There are reports that lseek on some devices fails, but
1510 * irc discussion said that contingency on contingency was overkill.
1512 size
= lseek(s
->fd
, 0, SEEK_END
);
1518 #elif defined(CONFIG_BSD)
1519 static int64_t raw_getlength(BlockDriverState
*bs
)
1521 BDRVRawState
*s
= bs
->opaque
;
1525 #if defined (__FreeBSD__) || defined(__FreeBSD_kernel__)
1534 #if defined (__FreeBSD__) || defined(__FreeBSD_kernel__)
1537 if (!fstat(fd
, &sb
) && (S_IFCHR
& sb
.st_mode
)) {
1538 #ifdef DIOCGMEDIASIZE
1539 if (ioctl(fd
, DIOCGMEDIASIZE
, (off_t
*)&size
))
1540 #elif defined(DIOCGPART)
1543 if (ioctl(fd
, DIOCGPART
, &pi
) == 0)
1544 size
= pi
.media_size
;
1550 #if defined(__APPLE__) && defined(__MACH__)
1552 uint64_t sectors
= 0;
1553 uint32_t sector_size
= 0;
1555 if (ioctl(fd
, DKIOCGETBLOCKCOUNT
, §ors
) == 0
1556 && ioctl(fd
, DKIOCGETBLOCKSIZE
, §or_size
) == 0) {
1557 size
= sectors
* sector_size
;
1559 size
= lseek(fd
, 0LL, SEEK_END
);
1566 size
= lseek(fd
, 0LL, SEEK_END
);
1571 #if defined(__FreeBSD__) || defined(__FreeBSD_kernel__)
1574 /* XXX FreeBSD acd returns UINT_MAX sectors for an empty drive */
1575 if (size
== 2048LL * (unsigned)-1)
1577 /* XXX no disc? maybe we need to reopen... */
1578 if (size
<= 0 && !reopened
&& cdrom_reopen(bs
) >= 0) {
1585 size
= lseek(fd
, 0, SEEK_END
);
1593 static int64_t raw_getlength(BlockDriverState
*bs
)
1595 BDRVRawState
*s
= bs
->opaque
;
1604 size
= lseek(s
->fd
, 0, SEEK_END
);
1612 static int64_t raw_get_allocated_file_size(BlockDriverState
*bs
)
1615 BDRVRawState
*s
= bs
->opaque
;
1617 if (fstat(s
->fd
, &st
) < 0) {
1620 return (int64_t)st
.st_blocks
* 512;
1623 static int raw_create(const char *filename
, QemuOpts
*opts
, Error
**errp
)
1627 int64_t total_size
= 0;
1629 PreallocMode prealloc
;
1631 Error
*local_err
= NULL
;
1633 strstart(filename
, "file:", &filename
);
1635 /* Read out options */
1636 total_size
= ROUND_UP(qemu_opt_get_size_del(opts
, BLOCK_OPT_SIZE
, 0),
1638 nocow
= qemu_opt_get_bool(opts
, BLOCK_OPT_NOCOW
, false);
1639 buf
= qemu_opt_get_del(opts
, BLOCK_OPT_PREALLOC
);
1640 prealloc
= qapi_enum_parse(PreallocMode_lookup
, buf
,
1641 PREALLOC_MODE__MAX
, PREALLOC_MODE_OFF
,
1645 error_propagate(errp
, local_err
);
1650 fd
= qemu_open(filename
, O_RDWR
| O_CREAT
| O_TRUNC
| O_BINARY
,
1654 error_setg_errno(errp
, -result
, "Could not create file");
1660 /* Set NOCOW flag to solve performance issue on fs like btrfs.
1661 * This is an optimisation. The FS_IOC_SETFLAGS ioctl return value
1662 * will be ignored since any failure of this operation should not
1663 * block the left work.
1666 if (ioctl(fd
, FS_IOC_GETFLAGS
, &attr
) == 0) {
1667 attr
|= FS_NOCOW_FL
;
1668 ioctl(fd
, FS_IOC_SETFLAGS
, &attr
);
1673 if (ftruncate(fd
, total_size
) != 0) {
1675 error_setg_errno(errp
, -result
, "Could not resize file");
1680 #ifdef CONFIG_POSIX_FALLOCATE
1681 case PREALLOC_MODE_FALLOC
:
1682 /* posix_fallocate() doesn't set errno. */
1683 result
= -posix_fallocate(fd
, 0, total_size
);
1685 error_setg_errno(errp
, -result
,
1686 "Could not preallocate data for the new file");
1690 case PREALLOC_MODE_FULL
:
1692 int64_t num
= 0, left
= total_size
;
1693 buf
= g_malloc0(65536);
1696 num
= MIN(left
, 65536);
1697 result
= write(fd
, buf
, num
);
1700 error_setg_errno(errp
, -result
,
1701 "Could not write to the new file");
1710 error_setg_errno(errp
, -result
,
1711 "Could not flush new file to disk");
1717 case PREALLOC_MODE_OFF
:
1721 error_setg(errp
, "Unsupported preallocation mode: %s",
1722 PreallocMode_lookup
[prealloc
]);
1727 if (qemu_close(fd
) != 0 && result
== 0) {
1729 error_setg_errno(errp
, -result
, "Could not close the new file");
1736 * Find allocation range in @bs around offset @start.
1737 * May change underlying file descriptor's file offset.
1738 * If @start is not in a hole, store @start in @data, and the
1739 * beginning of the next hole in @hole, and return 0.
1740 * If @start is in a non-trailing hole, store @start in @hole and the
1741 * beginning of the next non-hole in @data, and return 0.
1742 * If @start is in a trailing hole or beyond EOF, return -ENXIO.
1743 * If we can't find out, return a negative errno other than -ENXIO.
1745 static int find_allocation(BlockDriverState
*bs
, off_t start
,
1746 off_t
*data
, off_t
*hole
)
1748 #if defined SEEK_HOLE && defined SEEK_DATA
1749 BDRVRawState
*s
= bs
->opaque
;
1754 * D1. offs == start: start is in data
1755 * D2. offs > start: start is in a hole, next data at offs
1756 * D3. offs < 0, errno = ENXIO: either start is in a trailing hole
1757 * or start is beyond EOF
1758 * If the latter happens, the file has been truncated behind
1759 * our back since we opened it. All bets are off then.
1760 * Treating like a trailing hole is simplest.
1761 * D4. offs < 0, errno != ENXIO: we learned nothing
1763 offs
= lseek(s
->fd
, start
, SEEK_DATA
);
1765 return -errno
; /* D3 or D4 */
1767 assert(offs
>= start
);
1770 /* D2: in hole, next data at offs */
1776 /* D1: in data, end not yet known */
1780 * H1. offs == start: start is in a hole
1781 * If this happens here, a hole has been dug behind our back
1782 * since the previous lseek().
1783 * H2. offs > start: either start is in data, next hole at offs,
1784 * or start is in trailing hole, EOF at offs
1785 * Linux treats trailing holes like any other hole: offs ==
1786 * start. Solaris seeks to EOF instead: offs > start (blech).
1787 * If that happens here, a hole has been dug behind our back
1788 * since the previous lseek().
1789 * H3. offs < 0, errno = ENXIO: start is beyond EOF
1790 * If this happens, the file has been truncated behind our
1791 * back since we opened it. Treat it like a trailing hole.
1792 * H4. offs < 0, errno != ENXIO: we learned nothing
1793 * Pretend we know nothing at all, i.e. "forget" about D1.
1795 offs
= lseek(s
->fd
, start
, SEEK_HOLE
);
1797 return -errno
; /* D1 and (H3 or H4) */
1799 assert(offs
>= start
);
1803 * D1 and H2: either in data, next hole at offs, or it was in
1804 * data but is now in a trailing hole. In the latter case,
1805 * all bets are off. Treating it as if it there was data all
1806 * the way to EOF is safe, so simply do that.
1821 * Returns the allocation status of the specified sectors.
1823 * If 'sector_num' is beyond the end of the disk image the return value is 0
1824 * and 'pnum' is set to 0.
1826 * 'pnum' is set to the number of sectors (including and immediately following
1827 * the specified sector) that are known to be in the same
1828 * allocated/unallocated state.
1830 * 'nb_sectors' is the max value 'pnum' should be set to. If nb_sectors goes
1831 * beyond the end of the disk image it will be clamped.
1833 static int64_t coroutine_fn
raw_co_get_block_status(BlockDriverState
*bs
,
1835 int nb_sectors
, int *pnum
,
1836 BlockDriverState
**file
)
1838 off_t start
, data
= 0, hole
= 0;
1847 start
= sector_num
* BDRV_SECTOR_SIZE
;
1848 total_size
= bdrv_getlength(bs
);
1849 if (total_size
< 0) {
1851 } else if (start
>= total_size
) {
1854 } else if (start
+ nb_sectors
* BDRV_SECTOR_SIZE
> total_size
) {
1855 nb_sectors
= DIV_ROUND_UP(total_size
- start
, BDRV_SECTOR_SIZE
);
1858 ret
= find_allocation(bs
, start
, &data
, &hole
);
1859 if (ret
== -ENXIO
) {
1862 ret
= BDRV_BLOCK_ZERO
;
1863 } else if (ret
< 0) {
1864 /* No info available, so pretend there are no holes */
1866 ret
= BDRV_BLOCK_DATA
;
1867 } else if (data
== start
) {
1868 /* On a data extent, compute sectors to the end of the extent,
1869 * possibly including a partial sector at EOF. */
1870 *pnum
= MIN(nb_sectors
, DIV_ROUND_UP(hole
- start
, BDRV_SECTOR_SIZE
));
1871 ret
= BDRV_BLOCK_DATA
;
1873 /* On a hole, compute sectors to the beginning of the next extent. */
1874 assert(hole
== start
);
1875 *pnum
= MIN(nb_sectors
, (data
- start
) / BDRV_SECTOR_SIZE
);
1876 ret
= BDRV_BLOCK_ZERO
;
1879 return ret
| BDRV_BLOCK_OFFSET_VALID
| start
;
1882 static coroutine_fn BlockAIOCB
*raw_aio_discard(BlockDriverState
*bs
,
1883 int64_t sector_num
, int nb_sectors
,
1884 BlockCompletionFunc
*cb
, void *opaque
)
1886 BDRVRawState
*s
= bs
->opaque
;
1888 return paio_submit(bs
, s
->fd
, sector_num
, NULL
, nb_sectors
,
1889 cb
, opaque
, QEMU_AIO_DISCARD
);
1892 static int coroutine_fn
raw_co_pwrite_zeroes(
1893 BlockDriverState
*bs
, int64_t offset
,
1894 int count
, BdrvRequestFlags flags
)
1896 BDRVRawState
*s
= bs
->opaque
;
1898 if (!(flags
& BDRV_REQ_MAY_UNMAP
)) {
1899 return paio_submit_co(bs
, s
->fd
, offset
, NULL
, count
,
1900 QEMU_AIO_WRITE_ZEROES
);
1901 } else if (s
->discard_zeroes
) {
1902 return paio_submit_co(bs
, s
->fd
, offset
, NULL
, count
,
1908 static int raw_get_info(BlockDriverState
*bs
, BlockDriverInfo
*bdi
)
1910 BDRVRawState
*s
= bs
->opaque
;
1912 bdi
->unallocated_blocks_are_zero
= s
->discard_zeroes
;
1913 bdi
->can_write_zeroes_with_unmap
= s
->discard_zeroes
;
1917 static QemuOptsList raw_create_opts
= {
1918 .name
= "raw-create-opts",
1919 .head
= QTAILQ_HEAD_INITIALIZER(raw_create_opts
.head
),
1922 .name
= BLOCK_OPT_SIZE
,
1923 .type
= QEMU_OPT_SIZE
,
1924 .help
= "Virtual disk size"
1927 .name
= BLOCK_OPT_NOCOW
,
1928 .type
= QEMU_OPT_BOOL
,
1929 .help
= "Turn off copy-on-write (valid only on btrfs)"
1932 .name
= BLOCK_OPT_PREALLOC
,
1933 .type
= QEMU_OPT_STRING
,
1934 .help
= "Preallocation mode (allowed values: off, falloc, full)"
1936 { /* end of list */ }
1940 BlockDriver bdrv_file
= {
1941 .format_name
= "file",
1942 .protocol_name
= "file",
1943 .instance_size
= sizeof(BDRVRawState
),
1944 .bdrv_needs_filename
= true,
1945 .bdrv_probe
= NULL
, /* no probe for protocols */
1946 .bdrv_parse_filename
= raw_parse_filename
,
1947 .bdrv_file_open
= raw_open
,
1948 .bdrv_reopen_prepare
= raw_reopen_prepare
,
1949 .bdrv_reopen_commit
= raw_reopen_commit
,
1950 .bdrv_reopen_abort
= raw_reopen_abort
,
1951 .bdrv_close
= raw_close
,
1952 .bdrv_create
= raw_create
,
1953 .bdrv_has_zero_init
= bdrv_has_zero_init_1
,
1954 .bdrv_co_get_block_status
= raw_co_get_block_status
,
1955 .bdrv_co_pwrite_zeroes
= raw_co_pwrite_zeroes
,
1957 .bdrv_co_preadv
= raw_co_preadv
,
1958 .bdrv_co_pwritev
= raw_co_pwritev
,
1959 .bdrv_aio_flush
= raw_aio_flush
,
1960 .bdrv_aio_discard
= raw_aio_discard
,
1961 .bdrv_refresh_limits
= raw_refresh_limits
,
1962 .bdrv_io_plug
= raw_aio_plug
,
1963 .bdrv_io_unplug
= raw_aio_unplug
,
1965 .bdrv_truncate
= raw_truncate
,
1966 .bdrv_getlength
= raw_getlength
,
1967 .bdrv_get_info
= raw_get_info
,
1968 .bdrv_get_allocated_file_size
1969 = raw_get_allocated_file_size
,
1971 .bdrv_detach_aio_context
= raw_detach_aio_context
,
1972 .bdrv_attach_aio_context
= raw_attach_aio_context
,
1974 .create_opts
= &raw_create_opts
,
1977 /***********************************************/
1980 #if defined(__APPLE__) && defined(__MACH__)
1981 static kern_return_t
GetBSDPath(io_iterator_t mediaIterator
, char *bsdPath
,
1982 CFIndex maxPathSize
, int flags
);
1983 static char *FindEjectableOpticalMedia(io_iterator_t
*mediaIterator
)
1985 kern_return_t kernResult
= KERN_FAILURE
;
1986 mach_port_t masterPort
;
1987 CFMutableDictionaryRef classesToMatch
;
1988 const char *matching_array
[] = {kIODVDMediaClass
, kIOCDMediaClass
};
1989 char *mediaType
= NULL
;
1991 kernResult
= IOMasterPort( MACH_PORT_NULL
, &masterPort
);
1992 if ( KERN_SUCCESS
!= kernResult
) {
1993 printf( "IOMasterPort returned %d\n", kernResult
);
1997 for (index
= 0; index
< ARRAY_SIZE(matching_array
); index
++) {
1998 classesToMatch
= IOServiceMatching(matching_array
[index
]);
1999 if (classesToMatch
== NULL
) {
2000 error_report("IOServiceMatching returned NULL for %s",
2001 matching_array
[index
]);
2004 CFDictionarySetValue(classesToMatch
, CFSTR(kIOMediaEjectableKey
),
2006 kernResult
= IOServiceGetMatchingServices(masterPort
, classesToMatch
,
2008 if (kernResult
!= KERN_SUCCESS
) {
2009 error_report("Note: IOServiceGetMatchingServices returned %d",
2014 /* If a match was found, leave the loop */
2015 if (*mediaIterator
!= 0) {
2016 DPRINTF("Matching using %s\n", matching_array
[index
]);
2017 mediaType
= g_strdup(matching_array
[index
]);
2024 kern_return_t
GetBSDPath(io_iterator_t mediaIterator
, char *bsdPath
,
2025 CFIndex maxPathSize
, int flags
)
2027 io_object_t nextMedia
;
2028 kern_return_t kernResult
= KERN_FAILURE
;
2030 nextMedia
= IOIteratorNext( mediaIterator
);
2033 CFTypeRef bsdPathAsCFString
;
2034 bsdPathAsCFString
= IORegistryEntryCreateCFProperty( nextMedia
, CFSTR( kIOBSDNameKey
), kCFAllocatorDefault
, 0 );
2035 if ( bsdPathAsCFString
) {
2036 size_t devPathLength
;
2037 strcpy( bsdPath
, _PATH_DEV
);
2038 if (flags
& BDRV_O_NOCACHE
) {
2039 strcat(bsdPath
, "r");
2041 devPathLength
= strlen( bsdPath
);
2042 if ( CFStringGetCString( bsdPathAsCFString
, bsdPath
+ devPathLength
, maxPathSize
- devPathLength
, kCFStringEncodingASCII
) ) {
2043 kernResult
= KERN_SUCCESS
;
2045 CFRelease( bsdPathAsCFString
);
2047 IOObjectRelease( nextMedia
);
2053 /* Sets up a real cdrom for use in QEMU */
2054 static bool setup_cdrom(char *bsd_path
, Error
**errp
)
2056 int index
, num_of_test_partitions
= 2, fd
;
2057 char test_partition
[MAXPATHLEN
];
2058 bool partition_found
= false;
2060 /* look for a working partition */
2061 for (index
= 0; index
< num_of_test_partitions
; index
++) {
2062 snprintf(test_partition
, sizeof(test_partition
), "%ss%d", bsd_path
,
2064 fd
= qemu_open(test_partition
, O_RDONLY
| O_BINARY
| O_LARGEFILE
);
2066 partition_found
= true;
2072 /* if a working partition on the device was not found */
2073 if (partition_found
== false) {
2074 error_setg(errp
, "Failed to find a working partition on disc");
2076 DPRINTF("Using %s as optical disc\n", test_partition
);
2077 pstrcpy(bsd_path
, MAXPATHLEN
, test_partition
);
2079 return partition_found
;
2082 /* Prints directions on mounting and unmounting a device */
2083 static void print_unmounting_directions(const char *file_name
)
2085 error_report("If device %s is mounted on the desktop, unmount"
2086 " it first before using it in QEMU", file_name
);
2087 error_report("Command to unmount device: diskutil unmountDisk %s",
2089 error_report("Command to mount device: diskutil mountDisk %s", file_name
);
2092 #endif /* defined(__APPLE__) && defined(__MACH__) */
2094 static int hdev_probe_device(const char *filename
)
2098 /* allow a dedicated CD-ROM driver to match with a higher priority */
2099 if (strstart(filename
, "/dev/cdrom", NULL
))
2102 if (stat(filename
, &st
) >= 0 &&
2103 (S_ISCHR(st
.st_mode
) || S_ISBLK(st
.st_mode
))) {
2110 static int check_hdev_writable(BDRVRawState
*s
)
2112 #if defined(BLKROGET)
2113 /* Linux block devices can be configured "read-only" using blockdev(8).
2114 * This is independent of device node permissions and therefore open(2)
2115 * with O_RDWR succeeds. Actual writes fail with EPERM.
2117 * bdrv_open() is supposed to fail if the disk is read-only. Explicitly
2118 * check for read-only block devices so that Linux block devices behave
2124 if (fstat(s
->fd
, &st
)) {
2128 if (!S_ISBLK(st
.st_mode
)) {
2132 if (ioctl(s
->fd
, BLKROGET
, &readonly
) < 0) {
2139 #endif /* defined(BLKROGET) */
2143 static void hdev_parse_filename(const char *filename
, QDict
*options
,
2146 /* The prefix is optional, just as for "file". */
2147 strstart(filename
, "host_device:", &filename
);
2149 qdict_put_obj(options
, "filename", QOBJECT(qstring_from_str(filename
)));
2152 static bool hdev_is_sg(BlockDriverState
*bs
)
2155 #if defined(__linux__)
2158 struct sg_scsi_id scsiid
;
2161 if (stat(bs
->filename
, &st
) >= 0 && S_ISCHR(st
.st_mode
) &&
2162 !bdrv_ioctl(bs
, SG_GET_VERSION_NUM
, &sg_version
) &&
2163 !bdrv_ioctl(bs
, SG_GET_SCSI_ID
, &scsiid
)) {
2164 DPRINTF("SG device found: type=%d, version=%d\n",
2165 scsiid
.scsi_type
, sg_version
);
2174 static int hdev_open(BlockDriverState
*bs
, QDict
*options
, int flags
,
2177 BDRVRawState
*s
= bs
->opaque
;
2178 Error
*local_err
= NULL
;
2181 #if defined(__APPLE__) && defined(__MACH__)
2182 const char *filename
= qdict_get_str(options
, "filename");
2183 char bsd_path
[MAXPATHLEN
] = "";
2184 bool error_occurred
= false;
2186 /* If using a real cdrom */
2187 if (strcmp(filename
, "/dev/cdrom") == 0) {
2188 char *mediaType
= NULL
;
2189 kern_return_t ret_val
;
2190 io_iterator_t mediaIterator
= 0;
2192 mediaType
= FindEjectableOpticalMedia(&mediaIterator
);
2193 if (mediaType
== NULL
) {
2194 error_setg(errp
, "Please make sure your CD/DVD is in the optical"
2196 error_occurred
= true;
2197 goto hdev_open_Mac_error
;
2200 ret_val
= GetBSDPath(mediaIterator
, bsd_path
, sizeof(bsd_path
), flags
);
2201 if (ret_val
!= KERN_SUCCESS
) {
2202 error_setg(errp
, "Could not get BSD path for optical drive");
2203 error_occurred
= true;
2204 goto hdev_open_Mac_error
;
2207 /* If a real optical drive was not found */
2208 if (bsd_path
[0] == '\0') {
2209 error_setg(errp
, "Failed to obtain bsd path for optical drive");
2210 error_occurred
= true;
2211 goto hdev_open_Mac_error
;
2214 /* If using a cdrom disc and finding a partition on the disc failed */
2215 if (strncmp(mediaType
, kIOCDMediaClass
, 9) == 0 &&
2216 setup_cdrom(bsd_path
, errp
) == false) {
2217 print_unmounting_directions(bsd_path
);
2218 error_occurred
= true;
2219 goto hdev_open_Mac_error
;
2222 qdict_put(options
, "filename", qstring_from_str(bsd_path
));
2224 hdev_open_Mac_error
:
2226 if (mediaIterator
) {
2227 IOObjectRelease(mediaIterator
);
2229 if (error_occurred
) {
2233 #endif /* defined(__APPLE__) && defined(__MACH__) */
2235 s
->type
= FTYPE_FILE
;
2237 ret
= raw_open_common(bs
, options
, flags
, 0, &local_err
);
2240 error_propagate(errp
, local_err
);
2242 #if defined(__APPLE__) && defined(__MACH__)
2244 filename
= bsd_path
;
2246 /* if a physical device experienced an error while being opened */
2247 if (strncmp(filename
, "/dev/", 5) == 0) {
2248 print_unmounting_directions(filename
);
2250 #endif /* defined(__APPLE__) && defined(__MACH__) */
2254 /* Since this does ioctl the device must be already opened */
2255 bs
->sg
= hdev_is_sg(bs
);
2257 if (flags
& BDRV_O_RDWR
) {
2258 ret
= check_hdev_writable(s
);
2261 error_setg_errno(errp
, -ret
, "The device is not writable");
2269 #if defined(__linux__)
2271 static BlockAIOCB
*hdev_aio_ioctl(BlockDriverState
*bs
,
2272 unsigned long int req
, void *buf
,
2273 BlockCompletionFunc
*cb
, void *opaque
)
2275 BDRVRawState
*s
= bs
->opaque
;
2276 RawPosixAIOData
*acb
;
2279 if (fd_open(bs
) < 0)
2282 acb
= g_new(RawPosixAIOData
, 1);
2284 acb
->aio_type
= QEMU_AIO_IOCTL
;
2285 acb
->aio_fildes
= s
->fd
;
2286 acb
->aio_offset
= 0;
2287 acb
->aio_ioctl_buf
= buf
;
2288 acb
->aio_ioctl_cmd
= req
;
2289 pool
= aio_get_thread_pool(bdrv_get_aio_context(bs
));
2290 return thread_pool_submit_aio(pool
, aio_worker
, acb
, cb
, opaque
);
2294 static int fd_open(BlockDriverState
*bs
)
2296 BDRVRawState
*s
= bs
->opaque
;
2298 /* this is just to ensure s->fd is sane (its called by io ops) */
2304 static coroutine_fn BlockAIOCB
*hdev_aio_discard(BlockDriverState
*bs
,
2305 int64_t sector_num
, int nb_sectors
,
2306 BlockCompletionFunc
*cb
, void *opaque
)
2308 BDRVRawState
*s
= bs
->opaque
;
2310 if (fd_open(bs
) < 0) {
2313 return paio_submit(bs
, s
->fd
, sector_num
, NULL
, nb_sectors
,
2314 cb
, opaque
, QEMU_AIO_DISCARD
|QEMU_AIO_BLKDEV
);
2317 static coroutine_fn
int hdev_co_pwrite_zeroes(BlockDriverState
*bs
,
2318 int64_t offset
, int count
, BdrvRequestFlags flags
)
2320 BDRVRawState
*s
= bs
->opaque
;
2327 if (!(flags
& BDRV_REQ_MAY_UNMAP
)) {
2328 return paio_submit_co(bs
, s
->fd
, offset
, NULL
, count
,
2329 QEMU_AIO_WRITE_ZEROES
|QEMU_AIO_BLKDEV
);
2330 } else if (s
->discard_zeroes
) {
2331 return paio_submit_co(bs
, s
->fd
, offset
, NULL
, count
,
2332 QEMU_AIO_DISCARD
|QEMU_AIO_BLKDEV
);
2337 static int hdev_create(const char *filename
, QemuOpts
*opts
,
2342 struct stat stat_buf
;
2343 int64_t total_size
= 0;
2346 /* This function is used by both protocol block drivers and therefore either
2347 * of these prefixes may be given.
2348 * The return value has to be stored somewhere, otherwise this is an error
2349 * due to -Werror=unused-value. */
2351 strstart(filename
, "host_device:", &filename
) ||
2352 strstart(filename
, "host_cdrom:" , &filename
);
2356 ret
= raw_normalize_devicepath(&filename
);
2358 error_setg_errno(errp
, -ret
, "Could not normalize device path");
2362 /* Read out options */
2363 total_size
= ROUND_UP(qemu_opt_get_size_del(opts
, BLOCK_OPT_SIZE
, 0),
2366 fd
= qemu_open(filename
, O_WRONLY
| O_BINARY
);
2369 error_setg_errno(errp
, -ret
, "Could not open device");
2373 if (fstat(fd
, &stat_buf
) < 0) {
2375 error_setg_errno(errp
, -ret
, "Could not stat device");
2376 } else if (!S_ISBLK(stat_buf
.st_mode
) && !S_ISCHR(stat_buf
.st_mode
)) {
2378 "The given file is neither a block nor a character device");
2380 } else if (lseek(fd
, 0, SEEK_END
) < total_size
) {
2381 error_setg(errp
, "Device is too small");
2389 static BlockDriver bdrv_host_device
= {
2390 .format_name
= "host_device",
2391 .protocol_name
= "host_device",
2392 .instance_size
= sizeof(BDRVRawState
),
2393 .bdrv_needs_filename
= true,
2394 .bdrv_probe_device
= hdev_probe_device
,
2395 .bdrv_parse_filename
= hdev_parse_filename
,
2396 .bdrv_file_open
= hdev_open
,
2397 .bdrv_close
= raw_close
,
2398 .bdrv_reopen_prepare
= raw_reopen_prepare
,
2399 .bdrv_reopen_commit
= raw_reopen_commit
,
2400 .bdrv_reopen_abort
= raw_reopen_abort
,
2401 .bdrv_create
= hdev_create
,
2402 .create_opts
= &raw_create_opts
,
2403 .bdrv_co_pwrite_zeroes
= hdev_co_pwrite_zeroes
,
2405 .bdrv_co_preadv
= raw_co_preadv
,
2406 .bdrv_co_pwritev
= raw_co_pwritev
,
2407 .bdrv_aio_flush
= raw_aio_flush
,
2408 .bdrv_aio_discard
= hdev_aio_discard
,
2409 .bdrv_refresh_limits
= raw_refresh_limits
,
2410 .bdrv_io_plug
= raw_aio_plug
,
2411 .bdrv_io_unplug
= raw_aio_unplug
,
2413 .bdrv_truncate
= raw_truncate
,
2414 .bdrv_getlength
= raw_getlength
,
2415 .bdrv_get_info
= raw_get_info
,
2416 .bdrv_get_allocated_file_size
2417 = raw_get_allocated_file_size
,
2418 .bdrv_probe_blocksizes
= hdev_probe_blocksizes
,
2419 .bdrv_probe_geometry
= hdev_probe_geometry
,
2421 .bdrv_detach_aio_context
= raw_detach_aio_context
,
2422 .bdrv_attach_aio_context
= raw_attach_aio_context
,
2424 /* generic scsi device */
2426 .bdrv_aio_ioctl
= hdev_aio_ioctl
,
2430 #if defined(__linux__) || defined(__FreeBSD__) || defined(__FreeBSD_kernel__)
2431 static void cdrom_parse_filename(const char *filename
, QDict
*options
,
2434 /* The prefix is optional, just as for "file". */
2435 strstart(filename
, "host_cdrom:", &filename
);
2437 qdict_put_obj(options
, "filename", QOBJECT(qstring_from_str(filename
)));
2442 static int cdrom_open(BlockDriverState
*bs
, QDict
*options
, int flags
,
2445 BDRVRawState
*s
= bs
->opaque
;
2446 Error
*local_err
= NULL
;
2451 /* open will not fail even if no CD is inserted, so add O_NONBLOCK */
2452 ret
= raw_open_common(bs
, options
, flags
, O_NONBLOCK
, &local_err
);
2454 error_propagate(errp
, local_err
);
2459 static int cdrom_probe_device(const char *filename
)
2465 fd
= qemu_open(filename
, O_RDONLY
| O_NONBLOCK
);
2469 ret
= fstat(fd
, &st
);
2470 if (ret
== -1 || !S_ISBLK(st
.st_mode
)) {
2474 /* Attempt to detect via a CDROM specific ioctl */
2475 ret
= ioctl(fd
, CDROM_DRIVE_STATUS
, CDSL_CURRENT
);
2485 static bool cdrom_is_inserted(BlockDriverState
*bs
)
2487 BDRVRawState
*s
= bs
->opaque
;
2490 ret
= ioctl(s
->fd
, CDROM_DRIVE_STATUS
, CDSL_CURRENT
);
2491 return ret
== CDS_DISC_OK
;
2494 static void cdrom_eject(BlockDriverState
*bs
, bool eject_flag
)
2496 BDRVRawState
*s
= bs
->opaque
;
2499 if (ioctl(s
->fd
, CDROMEJECT
, NULL
) < 0)
2500 perror("CDROMEJECT");
2502 if (ioctl(s
->fd
, CDROMCLOSETRAY
, NULL
) < 0)
2503 perror("CDROMEJECT");
2507 static void cdrom_lock_medium(BlockDriverState
*bs
, bool locked
)
2509 BDRVRawState
*s
= bs
->opaque
;
2511 if (ioctl(s
->fd
, CDROM_LOCKDOOR
, locked
) < 0) {
2513 * Note: an error can happen if the distribution automatically
2516 /* perror("CDROM_LOCKDOOR"); */
2520 static BlockDriver bdrv_host_cdrom
= {
2521 .format_name
= "host_cdrom",
2522 .protocol_name
= "host_cdrom",
2523 .instance_size
= sizeof(BDRVRawState
),
2524 .bdrv_needs_filename
= true,
2525 .bdrv_probe_device
= cdrom_probe_device
,
2526 .bdrv_parse_filename
= cdrom_parse_filename
,
2527 .bdrv_file_open
= cdrom_open
,
2528 .bdrv_close
= raw_close
,
2529 .bdrv_reopen_prepare
= raw_reopen_prepare
,
2530 .bdrv_reopen_commit
= raw_reopen_commit
,
2531 .bdrv_reopen_abort
= raw_reopen_abort
,
2532 .bdrv_create
= hdev_create
,
2533 .create_opts
= &raw_create_opts
,
2536 .bdrv_co_preadv
= raw_co_preadv
,
2537 .bdrv_co_pwritev
= raw_co_pwritev
,
2538 .bdrv_aio_flush
= raw_aio_flush
,
2539 .bdrv_refresh_limits
= raw_refresh_limits
,
2540 .bdrv_io_plug
= raw_aio_plug
,
2541 .bdrv_io_unplug
= raw_aio_unplug
,
2543 .bdrv_truncate
= raw_truncate
,
2544 .bdrv_getlength
= raw_getlength
,
2545 .has_variable_length
= true,
2546 .bdrv_get_allocated_file_size
2547 = raw_get_allocated_file_size
,
2549 .bdrv_detach_aio_context
= raw_detach_aio_context
,
2550 .bdrv_attach_aio_context
= raw_attach_aio_context
,
2552 /* removable device support */
2553 .bdrv_is_inserted
= cdrom_is_inserted
,
2554 .bdrv_eject
= cdrom_eject
,
2555 .bdrv_lock_medium
= cdrom_lock_medium
,
2557 /* generic scsi device */
2558 .bdrv_aio_ioctl
= hdev_aio_ioctl
,
2560 #endif /* __linux__ */
2562 #if defined (__FreeBSD__) || defined(__FreeBSD_kernel__)
2563 static int cdrom_open(BlockDriverState
*bs
, QDict
*options
, int flags
,
2566 BDRVRawState
*s
= bs
->opaque
;
2567 Error
*local_err
= NULL
;
2572 ret
= raw_open_common(bs
, options
, flags
, 0, &local_err
);
2575 error_propagate(errp
, local_err
);
2580 /* make sure the door isn't locked at this time */
2581 ioctl(s
->fd
, CDIOCALLOW
);
2585 static int cdrom_probe_device(const char *filename
)
2587 if (strstart(filename
, "/dev/cd", NULL
) ||
2588 strstart(filename
, "/dev/acd", NULL
))
2593 static int cdrom_reopen(BlockDriverState
*bs
)
2595 BDRVRawState
*s
= bs
->opaque
;
2599 * Force reread of possibly changed/newly loaded disc,
2600 * FreeBSD seems to not notice sometimes...
2604 fd
= qemu_open(bs
->filename
, s
->open_flags
, 0644);
2611 /* make sure the door isn't locked at this time */
2612 ioctl(s
->fd
, CDIOCALLOW
);
2616 static bool cdrom_is_inserted(BlockDriverState
*bs
)
2618 return raw_getlength(bs
) > 0;
2621 static void cdrom_eject(BlockDriverState
*bs
, bool eject_flag
)
2623 BDRVRawState
*s
= bs
->opaque
;
2628 (void) ioctl(s
->fd
, CDIOCALLOW
);
2631 if (ioctl(s
->fd
, CDIOCEJECT
) < 0)
2632 perror("CDIOCEJECT");
2634 if (ioctl(s
->fd
, CDIOCCLOSE
) < 0)
2635 perror("CDIOCCLOSE");
2641 static void cdrom_lock_medium(BlockDriverState
*bs
, bool locked
)
2643 BDRVRawState
*s
= bs
->opaque
;
2647 if (ioctl(s
->fd
, (locked
? CDIOCPREVENT
: CDIOCALLOW
)) < 0) {
2649 * Note: an error can happen if the distribution automatically
2652 /* perror("CDROM_LOCKDOOR"); */
2656 static BlockDriver bdrv_host_cdrom
= {
2657 .format_name
= "host_cdrom",
2658 .protocol_name
= "host_cdrom",
2659 .instance_size
= sizeof(BDRVRawState
),
2660 .bdrv_needs_filename
= true,
2661 .bdrv_probe_device
= cdrom_probe_device
,
2662 .bdrv_parse_filename
= cdrom_parse_filename
,
2663 .bdrv_file_open
= cdrom_open
,
2664 .bdrv_close
= raw_close
,
2665 .bdrv_reopen_prepare
= raw_reopen_prepare
,
2666 .bdrv_reopen_commit
= raw_reopen_commit
,
2667 .bdrv_reopen_abort
= raw_reopen_abort
,
2668 .bdrv_create
= hdev_create
,
2669 .create_opts
= &raw_create_opts
,
2671 .bdrv_co_preadv
= raw_co_preadv
,
2672 .bdrv_co_pwritev
= raw_co_pwritev
,
2673 .bdrv_aio_flush
= raw_aio_flush
,
2674 .bdrv_refresh_limits
= raw_refresh_limits
,
2675 .bdrv_io_plug
= raw_aio_plug
,
2676 .bdrv_io_unplug
= raw_aio_unplug
,
2678 .bdrv_truncate
= raw_truncate
,
2679 .bdrv_getlength
= raw_getlength
,
2680 .has_variable_length
= true,
2681 .bdrv_get_allocated_file_size
2682 = raw_get_allocated_file_size
,
2684 .bdrv_detach_aio_context
= raw_detach_aio_context
,
2685 .bdrv_attach_aio_context
= raw_attach_aio_context
,
2687 /* removable device support */
2688 .bdrv_is_inserted
= cdrom_is_inserted
,
2689 .bdrv_eject
= cdrom_eject
,
2690 .bdrv_lock_medium
= cdrom_lock_medium
,
2692 #endif /* __FreeBSD__ */
2694 static void bdrv_file_init(void)
2697 * Register all the drivers. Note that order is important, the driver
2698 * registered last will get probed first.
2700 bdrv_register(&bdrv_file
);
2701 bdrv_register(&bdrv_host_device
);
2703 bdrv_register(&bdrv_host_cdrom
);
2705 #if defined(__FreeBSD__) || defined(__FreeBSD_kernel__)
2706 bdrv_register(&bdrv_host_cdrom
);
2710 block_init(bdrv_file_init
);