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"
35 #include "block/raw-aio.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
{
144 bool has_write_zeroes
:1;
145 bool discard_zeroes
:1;
146 bool use_linux_aio
:1;
148 bool needs_alignment
;
151 typedef struct BDRVRawReopenState
{
154 } BDRVRawReopenState
;
156 static int fd_open(BlockDriverState
*bs
);
157 static int64_t raw_getlength(BlockDriverState
*bs
);
159 typedef struct RawPosixAIOData
{
160 BlockDriverState
*bs
;
163 struct iovec
*aio_iov
;
168 #define aio_ioctl_cmd aio_nbytes /* for QEMU_AIO_IOCTL */
173 #if defined(__FreeBSD__) || defined(__FreeBSD_kernel__)
174 static int cdrom_reopen(BlockDriverState
*bs
);
177 #if defined(__NetBSD__)
178 static int raw_normalize_devicepath(const char **filename
)
180 static char namebuf
[PATH_MAX
];
181 const char *dp
, *fname
;
185 dp
= strrchr(fname
, '/');
186 if (lstat(fname
, &sb
) < 0) {
187 fprintf(stderr
, "%s: stat failed: %s\n",
188 fname
, strerror(errno
));
192 if (!S_ISBLK(sb
.st_mode
)) {
197 snprintf(namebuf
, PATH_MAX
, "r%s", fname
);
199 snprintf(namebuf
, PATH_MAX
, "%.*s/r%s",
200 (int)(dp
- fname
), fname
, dp
+ 1);
202 fprintf(stderr
, "%s is a block device", fname
);
204 fprintf(stderr
, ", using %s\n", *filename
);
209 static int raw_normalize_devicepath(const char **filename
)
216 * Get logical block size via ioctl. On success store it in @sector_size_p.
218 static int probe_logical_blocksize(int fd
, unsigned int *sector_size_p
)
220 unsigned int sector_size
;
221 bool success
= false;
225 /* Try a few ioctls to get the right size */
227 if (ioctl(fd
, BLKSSZGET
, §or_size
) >= 0) {
228 *sector_size_p
= sector_size
;
232 #ifdef DKIOCGETBLOCKSIZE
233 if (ioctl(fd
, DKIOCGETBLOCKSIZE
, §or_size
) >= 0) {
234 *sector_size_p
= sector_size
;
238 #ifdef DIOCGSECTORSIZE
239 if (ioctl(fd
, DIOCGSECTORSIZE
, §or_size
) >= 0) {
240 *sector_size_p
= sector_size
;
245 return success
? 0 : -errno
;
249 * Get physical block size of @fd.
250 * On success, store it in @blk_size and return 0.
251 * On failure, return -errno.
253 static int probe_physical_blocksize(int fd
, unsigned int *blk_size
)
256 if (ioctl(fd
, BLKPBSZGET
, blk_size
) < 0) {
265 /* Check if read is allowed with given memory buffer and length.
267 * This function is used to check O_DIRECT memory buffer and request alignment.
269 static bool raw_is_io_aligned(int fd
, void *buf
, size_t len
)
271 ssize_t ret
= pread(fd
, buf
, len
, 0);
278 /* The Linux kernel returns EINVAL for misaligned O_DIRECT reads. Ignore
279 * other errors (e.g. real I/O error), which could happen on a failed
280 * drive, since we only care about probing alignment.
282 if (errno
!= EINVAL
) {
290 static void raw_probe_alignment(BlockDriverState
*bs
, int fd
, Error
**errp
)
292 BDRVRawState
*s
= bs
->opaque
;
294 size_t max_align
= MAX(MAX_BLOCKSIZE
, getpagesize());
296 /* For SCSI generic devices the alignment is not really used.
297 With buffered I/O, we don't have any restrictions. */
298 if (bdrv_is_sg(bs
) || !s
->needs_alignment
) {
299 bs
->bl
.request_alignment
= 1;
304 bs
->bl
.request_alignment
= 0;
306 /* Let's try to use the logical blocksize for the alignment. */
307 if (probe_logical_blocksize(fd
, &bs
->bl
.request_alignment
) < 0) {
308 bs
->bl
.request_alignment
= 0;
313 if (xfsctl(NULL
, fd
, XFS_IOC_DIOINFO
, &da
) >= 0) {
314 bs
->bl
.request_alignment
= da
.d_miniosz
;
315 /* The kernel returns wrong information for d_mem */
316 /* s->buf_align = da.d_mem; */
321 /* If we could not get the sizes so far, we can only guess them */
324 buf
= qemu_memalign(max_align
, 2 * max_align
);
325 for (align
= 512; align
<= max_align
; align
<<= 1) {
326 if (raw_is_io_aligned(fd
, buf
+ align
, max_align
)) {
327 s
->buf_align
= align
;
334 if (!bs
->bl
.request_alignment
) {
336 buf
= qemu_memalign(s
->buf_align
, max_align
);
337 for (align
= 512; align
<= max_align
; align
<<= 1) {
338 if (raw_is_io_aligned(fd
, buf
, align
)) {
339 bs
->bl
.request_alignment
= align
;
346 if (!s
->buf_align
|| !bs
->bl
.request_alignment
) {
347 error_setg(errp
, "Could not find working O_DIRECT alignment");
348 error_append_hint(errp
, "Try cache.direct=off\n");
352 static void raw_parse_flags(int bdrv_flags
, int *open_flags
)
354 assert(open_flags
!= NULL
);
356 *open_flags
|= O_BINARY
;
357 *open_flags
&= ~O_ACCMODE
;
358 if (bdrv_flags
& BDRV_O_RDWR
) {
359 *open_flags
|= O_RDWR
;
361 *open_flags
|= O_RDONLY
;
364 /* Use O_DSYNC for write-through caching, no flags for write-back caching,
365 * and O_DIRECT for no caching. */
366 if ((bdrv_flags
& BDRV_O_NOCACHE
)) {
367 *open_flags
|= O_DIRECT
;
371 static void raw_parse_filename(const char *filename
, QDict
*options
,
374 /* The filename does not have to be prefixed by the protocol name, since
375 * "file" is the default protocol; therefore, the return value of this
376 * function call can be ignored. */
377 strstart(filename
, "file:", &filename
);
379 qdict_put_obj(options
, "filename", QOBJECT(qstring_from_str(filename
)));
382 static QemuOptsList raw_runtime_opts
= {
384 .head
= QTAILQ_HEAD_INITIALIZER(raw_runtime_opts
.head
),
388 .type
= QEMU_OPT_STRING
,
389 .help
= "File name of the image",
393 .type
= QEMU_OPT_STRING
,
394 .help
= "host AIO implementation (threads, native)",
396 { /* end of list */ }
400 static int raw_open_common(BlockDriverState
*bs
, QDict
*options
,
401 int bdrv_flags
, int open_flags
, Error
**errp
)
403 BDRVRawState
*s
= bs
->opaque
;
405 Error
*local_err
= NULL
;
406 const char *filename
= NULL
;
407 BlockdevAioOptions aio
, aio_default
;
411 opts
= qemu_opts_create(&raw_runtime_opts
, NULL
, 0, &error_abort
);
412 qemu_opts_absorb_qdict(opts
, options
, &local_err
);
414 error_propagate(errp
, local_err
);
419 filename
= qemu_opt_get(opts
, "filename");
421 ret
= raw_normalize_devicepath(&filename
);
423 error_setg_errno(errp
, -ret
, "Could not normalize device path");
427 aio_default
= (bdrv_flags
& BDRV_O_NATIVE_AIO
)
428 ? BLOCKDEV_AIO_OPTIONS_NATIVE
429 : BLOCKDEV_AIO_OPTIONS_THREADS
;
430 aio
= qapi_enum_parse(BlockdevAioOptions_lookup
, qemu_opt_get(opts
, "aio"),
431 BLOCKDEV_AIO_OPTIONS__MAX
, aio_default
, &local_err
);
433 error_propagate(errp
, local_err
);
437 s
->use_linux_aio
= (aio
== BLOCKDEV_AIO_OPTIONS_NATIVE
);
439 s
->open_flags
= open_flags
;
440 raw_parse_flags(bdrv_flags
, &s
->open_flags
);
443 fd
= qemu_open(filename
, s
->open_flags
, 0644);
446 error_setg_errno(errp
, errno
, "Could not open '%s'", filename
);
454 #ifdef CONFIG_LINUX_AIO
455 /* Currently Linux does AIO only for files opened with O_DIRECT */
456 if (s
->use_linux_aio
&& !(s
->open_flags
& O_DIRECT
)) {
457 error_setg(errp
, "aio=native was specified, but it requires "
458 "cache.direct=on, which was not specified.");
463 if (s
->use_linux_aio
) {
464 error_setg(errp
, "aio=native was specified, but is not supported "
469 #endif /* !defined(CONFIG_LINUX_AIO) */
471 s
->has_discard
= true;
472 s
->has_write_zeroes
= true;
473 bs
->supported_zero_flags
= BDRV_REQ_MAY_UNMAP
;
474 if ((bs
->open_flags
& BDRV_O_NOCACHE
) != 0) {
475 s
->needs_alignment
= true;
478 if (fstat(s
->fd
, &st
) < 0) {
480 error_setg_errno(errp
, errno
, "Could not stat file");
483 if (S_ISREG(st
.st_mode
)) {
484 s
->discard_zeroes
= true;
485 s
->has_fallocate
= true;
487 if (S_ISBLK(st
.st_mode
)) {
488 #ifdef BLKDISCARDZEROES
490 if (ioctl(s
->fd
, BLKDISCARDZEROES
, &arg
) == 0 && arg
) {
491 s
->discard_zeroes
= true;
495 /* On Linux 3.10, BLKDISCARD leaves stale data in the page cache. Do
496 * not rely on the contents of discarded blocks unless using O_DIRECT.
497 * Same for BLKZEROOUT.
499 if (!(bs
->open_flags
& BDRV_O_NOCACHE
)) {
500 s
->discard_zeroes
= false;
501 s
->has_write_zeroes
= false;
506 if (S_ISCHR(st
.st_mode
)) {
508 * The file is a char device (disk), which on FreeBSD isn't behind
509 * a pager, so force all requests to be aligned. This is needed
510 * so QEMU makes sure all IO operations on the device are aligned
511 * to sector size, or else FreeBSD will reject them with EINVAL.
513 s
->needs_alignment
= true;
518 if (platform_test_xfs_fd(s
->fd
)) {
525 if (filename
&& (bdrv_flags
& BDRV_O_TEMPORARY
)) {
532 static int raw_open(BlockDriverState
*bs
, QDict
*options
, int flags
,
535 BDRVRawState
*s
= bs
->opaque
;
537 s
->type
= FTYPE_FILE
;
538 return raw_open_common(bs
, options
, flags
, 0, errp
);
541 static int raw_reopen_prepare(BDRVReopenState
*state
,
542 BlockReopenQueue
*queue
, Error
**errp
)
545 BDRVRawReopenState
*rs
;
547 Error
*local_err
= NULL
;
549 assert(state
!= NULL
);
550 assert(state
->bs
!= NULL
);
552 s
= state
->bs
->opaque
;
554 state
->opaque
= g_new0(BDRVRawReopenState
, 1);
557 if (s
->type
== FTYPE_CD
) {
558 rs
->open_flags
|= O_NONBLOCK
;
561 raw_parse_flags(state
->flags
, &rs
->open_flags
);
565 int fcntl_flags
= O_APPEND
| O_NONBLOCK
;
567 fcntl_flags
|= O_NOATIME
;
571 /* Not all operating systems have O_ASYNC, and those that don't
572 * will not let us track the state into rs->open_flags (typically
573 * you achieve the same effect with an ioctl, for example I_SETSIG
574 * on Solaris). But we do not use O_ASYNC, so that's fine.
576 assert((s
->open_flags
& O_ASYNC
) == 0);
579 if ((rs
->open_flags
& ~fcntl_flags
) == (s
->open_flags
& ~fcntl_flags
)) {
580 /* dup the original fd */
581 rs
->fd
= qemu_dup(s
->fd
);
583 ret
= fcntl_setfl(rs
->fd
, rs
->open_flags
);
591 /* If we cannot use fcntl, or fcntl failed, fall back to qemu_open() */
593 const char *normalized_filename
= state
->bs
->filename
;
594 ret
= raw_normalize_devicepath(&normalized_filename
);
596 error_setg_errno(errp
, -ret
, "Could not normalize device path");
598 assert(!(rs
->open_flags
& O_CREAT
));
599 rs
->fd
= qemu_open(normalized_filename
, rs
->open_flags
);
601 error_setg_errno(errp
, errno
, "Could not reopen file");
607 /* Fail already reopen_prepare() if we can't get a working O_DIRECT
608 * alignment with the new fd. */
610 raw_probe_alignment(state
->bs
, rs
->fd
, &local_err
);
614 error_propagate(errp
, local_err
);
622 static void raw_reopen_commit(BDRVReopenState
*state
)
624 BDRVRawReopenState
*rs
= state
->opaque
;
625 BDRVRawState
*s
= state
->bs
->opaque
;
627 s
->open_flags
= rs
->open_flags
;
632 g_free(state
->opaque
);
633 state
->opaque
= NULL
;
637 static void raw_reopen_abort(BDRVReopenState
*state
)
639 BDRVRawReopenState
*rs
= state
->opaque
;
641 /* nothing to do if NULL, we didn't get far enough */
650 g_free(state
->opaque
);
651 state
->opaque
= NULL
;
654 static int hdev_get_max_transfer_length(BlockDriverState
*bs
, int fd
)
658 short max_sectors
= 0;
659 if (bs
->sg
&& ioctl(fd
, BLKSECTGET
, &max_bytes
) == 0) {
661 } else if (!bs
->sg
&& ioctl(fd
, BLKSECTGET
, &max_sectors
) == 0) {
662 return max_sectors
<< BDRV_SECTOR_BITS
;
671 static int hdev_get_max_segments(const struct stat
*st
)
681 sysfspath
= g_strdup_printf("/sys/dev/block/%u:%u/queue/max_segments",
682 major(st
->st_rdev
), minor(st
->st_rdev
));
683 fd
= open(sysfspath
, O_RDONLY
);
689 ret
= read(fd
, buf
, sizeof(buf
) - 1);
690 } while (ret
== -1 && errno
== EINTR
);
694 } else if (ret
== 0) {
699 /* The file is ended with '\n', pass 'end' to accept that. */
700 ret
= qemu_strtol(buf
, &end
, 10, &max_segments
);
701 if (ret
== 0 && end
&& *end
== '\n') {
716 static void raw_refresh_limits(BlockDriverState
*bs
, Error
**errp
)
718 BDRVRawState
*s
= bs
->opaque
;
721 if (!fstat(s
->fd
, &st
)) {
722 if (S_ISBLK(st
.st_mode
) || S_ISCHR(st
.st_mode
)) {
723 int ret
= hdev_get_max_transfer_length(bs
, s
->fd
);
724 if (ret
> 0 && ret
<= BDRV_REQUEST_MAX_BYTES
) {
725 bs
->bl
.max_transfer
= pow2floor(ret
);
727 ret
= hdev_get_max_segments(&st
);
729 bs
->bl
.max_transfer
= MIN(bs
->bl
.max_transfer
,
730 ret
* getpagesize());
735 raw_probe_alignment(bs
, s
->fd
, errp
);
736 bs
->bl
.min_mem_alignment
= s
->buf_align
;
737 bs
->bl
.opt_mem_alignment
= MAX(s
->buf_align
, getpagesize());
740 static int check_for_dasd(int fd
)
743 struct dasd_information2_t info
= {0};
745 return ioctl(fd
, BIODASDINFO2
, &info
);
752 * Try to get @bs's logical and physical block size.
753 * On success, store them in @bsz and return zero.
754 * On failure, return negative errno.
756 static int hdev_probe_blocksizes(BlockDriverState
*bs
, BlockSizes
*bsz
)
758 BDRVRawState
*s
= bs
->opaque
;
761 /* If DASD, get blocksizes */
762 if (check_for_dasd(s
->fd
) < 0) {
765 ret
= probe_logical_blocksize(s
->fd
, &bsz
->log
);
769 return probe_physical_blocksize(s
->fd
, &bsz
->phys
);
773 * Try to get @bs's geometry: cyls, heads, sectors.
774 * On success, store them in @geo and return 0.
775 * On failure return -errno.
776 * (Allows block driver to assign default geometry values that guest sees)
779 static int hdev_probe_geometry(BlockDriverState
*bs
, HDGeometry
*geo
)
781 BDRVRawState
*s
= bs
->opaque
;
782 struct hd_geometry ioctl_geo
= {0};
784 /* If DASD, get its geometry */
785 if (check_for_dasd(s
->fd
) < 0) {
788 if (ioctl(s
->fd
, HDIO_GETGEO
, &ioctl_geo
) < 0) {
791 /* HDIO_GETGEO may return success even though geo contains zeros
792 (e.g. certain multipath setups) */
793 if (!ioctl_geo
.heads
|| !ioctl_geo
.sectors
|| !ioctl_geo
.cylinders
) {
796 /* Do not return a geometry for partition */
797 if (ioctl_geo
.start
!= 0) {
800 geo
->heads
= ioctl_geo
.heads
;
801 geo
->sectors
= ioctl_geo
.sectors
;
802 geo
->cylinders
= ioctl_geo
.cylinders
;
806 #else /* __linux__ */
807 static int hdev_probe_geometry(BlockDriverState
*bs
, HDGeometry
*geo
)
813 static ssize_t
handle_aiocb_ioctl(RawPosixAIOData
*aiocb
)
817 ret
= ioctl(aiocb
->aio_fildes
, aiocb
->aio_ioctl_cmd
, aiocb
->aio_ioctl_buf
);
825 static ssize_t
handle_aiocb_flush(RawPosixAIOData
*aiocb
)
829 ret
= qemu_fdatasync(aiocb
->aio_fildes
);
838 static bool preadv_present
= true;
841 qemu_preadv(int fd
, const struct iovec
*iov
, int nr_iov
, off_t offset
)
843 return preadv(fd
, iov
, nr_iov
, offset
);
847 qemu_pwritev(int fd
, const struct iovec
*iov
, int nr_iov
, off_t offset
)
849 return pwritev(fd
, iov
, nr_iov
, offset
);
854 static bool preadv_present
= false;
857 qemu_preadv(int fd
, const struct iovec
*iov
, int nr_iov
, off_t offset
)
863 qemu_pwritev(int fd
, const struct iovec
*iov
, int nr_iov
, off_t offset
)
870 static ssize_t
handle_aiocb_rw_vector(RawPosixAIOData
*aiocb
)
875 if (aiocb
->aio_type
& QEMU_AIO_WRITE
)
876 len
= qemu_pwritev(aiocb
->aio_fildes
,
881 len
= qemu_preadv(aiocb
->aio_fildes
,
885 } while (len
== -1 && errno
== EINTR
);
894 * Read/writes the data to/from a given linear buffer.
896 * Returns the number of bytes handles or -errno in case of an error. Short
897 * reads are only returned if the end of the file is reached.
899 static ssize_t
handle_aiocb_rw_linear(RawPosixAIOData
*aiocb
, char *buf
)
904 while (offset
< aiocb
->aio_nbytes
) {
905 if (aiocb
->aio_type
& QEMU_AIO_WRITE
) {
906 len
= pwrite(aiocb
->aio_fildes
,
907 (const char *)buf
+ offset
,
908 aiocb
->aio_nbytes
- offset
,
909 aiocb
->aio_offset
+ offset
);
911 len
= pread(aiocb
->aio_fildes
,
913 aiocb
->aio_nbytes
- offset
,
914 aiocb
->aio_offset
+ offset
);
916 if (len
== -1 && errno
== EINTR
) {
918 } else if (len
== -1 && errno
== EINVAL
&&
919 (aiocb
->bs
->open_flags
& BDRV_O_NOCACHE
) &&
920 !(aiocb
->aio_type
& QEMU_AIO_WRITE
) &&
922 /* O_DIRECT pread() may fail with EINVAL when offset is unaligned
923 * after a short read. Assume that O_DIRECT short reads only occur
924 * at EOF. Therefore this is a short read, not an I/O error.
927 } else if (len
== -1) {
930 } else if (len
== 0) {
939 static ssize_t
handle_aiocb_rw(RawPosixAIOData
*aiocb
)
944 if (!(aiocb
->aio_type
& QEMU_AIO_MISALIGNED
)) {
946 * If there is just a single buffer, and it is properly aligned
947 * we can just use plain pread/pwrite without any problems.
949 if (aiocb
->aio_niov
== 1) {
950 return handle_aiocb_rw_linear(aiocb
, aiocb
->aio_iov
->iov_base
);
953 * We have more than one iovec, and all are properly aligned.
955 * Try preadv/pwritev first and fall back to linearizing the
956 * buffer if it's not supported.
958 if (preadv_present
) {
959 nbytes
= handle_aiocb_rw_vector(aiocb
);
960 if (nbytes
== aiocb
->aio_nbytes
||
961 (nbytes
< 0 && nbytes
!= -ENOSYS
)) {
964 preadv_present
= false;
968 * XXX(hch): short read/write. no easy way to handle the reminder
969 * using these interfaces. For now retry using plain
975 * Ok, we have to do it the hard way, copy all segments into
976 * a single aligned buffer.
978 buf
= qemu_try_blockalign(aiocb
->bs
, aiocb
->aio_nbytes
);
983 if (aiocb
->aio_type
& QEMU_AIO_WRITE
) {
987 for (i
= 0; i
< aiocb
->aio_niov
; ++i
) {
988 memcpy(p
, aiocb
->aio_iov
[i
].iov_base
, aiocb
->aio_iov
[i
].iov_len
);
989 p
+= aiocb
->aio_iov
[i
].iov_len
;
991 assert(p
- buf
== aiocb
->aio_nbytes
);
994 nbytes
= handle_aiocb_rw_linear(aiocb
, buf
);
995 if (!(aiocb
->aio_type
& QEMU_AIO_WRITE
)) {
997 size_t count
= aiocb
->aio_nbytes
, copy
;
1000 for (i
= 0; i
< aiocb
->aio_niov
&& count
; ++i
) {
1002 if (copy
> aiocb
->aio_iov
[i
].iov_len
) {
1003 copy
= aiocb
->aio_iov
[i
].iov_len
;
1005 memcpy(aiocb
->aio_iov
[i
].iov_base
, p
, copy
);
1006 assert(count
>= copy
);
1018 static int xfs_write_zeroes(BDRVRawState
*s
, int64_t offset
, uint64_t bytes
)
1020 struct xfs_flock64 fl
;
1023 memset(&fl
, 0, sizeof(fl
));
1024 fl
.l_whence
= SEEK_SET
;
1025 fl
.l_start
= offset
;
1028 if (xfsctl(NULL
, s
->fd
, XFS_IOC_ZERO_RANGE
, &fl
) < 0) {
1030 DPRINTF("cannot write zero range (%s)\n", strerror(errno
));
1037 static int xfs_discard(BDRVRawState
*s
, int64_t offset
, uint64_t bytes
)
1039 struct xfs_flock64 fl
;
1042 memset(&fl
, 0, sizeof(fl
));
1043 fl
.l_whence
= SEEK_SET
;
1044 fl
.l_start
= offset
;
1047 if (xfsctl(NULL
, s
->fd
, XFS_IOC_UNRESVSP64
, &fl
) < 0) {
1049 DPRINTF("cannot punch hole (%s)\n", strerror(errno
));
1057 static int translate_err(int err
)
1059 if (err
== -ENODEV
|| err
== -ENOSYS
|| err
== -EOPNOTSUPP
||
1066 #ifdef CONFIG_FALLOCATE
1067 static int do_fallocate(int fd
, int mode
, off_t offset
, off_t len
)
1070 if (fallocate(fd
, mode
, offset
, len
) == 0) {
1073 } while (errno
== EINTR
);
1074 return translate_err(-errno
);
1078 static ssize_t
handle_aiocb_write_zeroes_block(RawPosixAIOData
*aiocb
)
1081 BDRVRawState
*s
= aiocb
->bs
->opaque
;
1083 if (!s
->has_write_zeroes
) {
1089 uint64_t range
[2] = { aiocb
->aio_offset
, aiocb
->aio_nbytes
};
1090 if (ioctl(aiocb
->aio_fildes
, BLKZEROOUT
, range
) == 0) {
1093 } while (errno
== EINTR
);
1095 ret
= translate_err(-errno
);
1098 if (ret
== -ENOTSUP
) {
1099 s
->has_write_zeroes
= false;
1104 static ssize_t
handle_aiocb_write_zeroes(RawPosixAIOData
*aiocb
)
1106 #if defined(CONFIG_FALLOCATE) || defined(CONFIG_XFS)
1107 BDRVRawState
*s
= aiocb
->bs
->opaque
;
1110 if (aiocb
->aio_type
& QEMU_AIO_BLKDEV
) {
1111 return handle_aiocb_write_zeroes_block(aiocb
);
1116 return xfs_write_zeroes(s
, aiocb
->aio_offset
, aiocb
->aio_nbytes
);
1120 #ifdef CONFIG_FALLOCATE_ZERO_RANGE
1121 if (s
->has_write_zeroes
) {
1122 int ret
= do_fallocate(s
->fd
, FALLOC_FL_ZERO_RANGE
,
1123 aiocb
->aio_offset
, aiocb
->aio_nbytes
);
1124 if (ret
== 0 || ret
!= -ENOTSUP
) {
1127 s
->has_write_zeroes
= false;
1131 #ifdef CONFIG_FALLOCATE_PUNCH_HOLE
1132 if (s
->has_discard
&& s
->has_fallocate
) {
1133 int ret
= do_fallocate(s
->fd
,
1134 FALLOC_FL_PUNCH_HOLE
| FALLOC_FL_KEEP_SIZE
,
1135 aiocb
->aio_offset
, aiocb
->aio_nbytes
);
1137 ret
= do_fallocate(s
->fd
, 0, aiocb
->aio_offset
, aiocb
->aio_nbytes
);
1138 if (ret
== 0 || ret
!= -ENOTSUP
) {
1141 s
->has_fallocate
= false;
1142 } else if (ret
!= -ENOTSUP
) {
1145 s
->has_discard
= false;
1150 #ifdef CONFIG_FALLOCATE
1151 if (s
->has_fallocate
&& aiocb
->aio_offset
>= bdrv_getlength(aiocb
->bs
)) {
1152 int ret
= do_fallocate(s
->fd
, 0, aiocb
->aio_offset
, aiocb
->aio_nbytes
);
1153 if (ret
== 0 || ret
!= -ENOTSUP
) {
1156 s
->has_fallocate
= false;
1163 static ssize_t
handle_aiocb_discard(RawPosixAIOData
*aiocb
)
1165 int ret
= -EOPNOTSUPP
;
1166 BDRVRawState
*s
= aiocb
->bs
->opaque
;
1168 if (!s
->has_discard
) {
1172 if (aiocb
->aio_type
& QEMU_AIO_BLKDEV
) {
1175 uint64_t range
[2] = { aiocb
->aio_offset
, aiocb
->aio_nbytes
};
1176 if (ioctl(aiocb
->aio_fildes
, BLKDISCARD
, range
) == 0) {
1179 } while (errno
== EINTR
);
1186 return xfs_discard(s
, aiocb
->aio_offset
, aiocb
->aio_nbytes
);
1190 #ifdef CONFIG_FALLOCATE_PUNCH_HOLE
1191 ret
= do_fallocate(s
->fd
, FALLOC_FL_PUNCH_HOLE
| FALLOC_FL_KEEP_SIZE
,
1192 aiocb
->aio_offset
, aiocb
->aio_nbytes
);
1196 ret
= translate_err(ret
);
1197 if (ret
== -ENOTSUP
) {
1198 s
->has_discard
= false;
1203 static int aio_worker(void *arg
)
1205 RawPosixAIOData
*aiocb
= arg
;
1208 switch (aiocb
->aio_type
& QEMU_AIO_TYPE_MASK
) {
1210 ret
= handle_aiocb_rw(aiocb
);
1211 if (ret
>= 0 && ret
< aiocb
->aio_nbytes
) {
1212 iov_memset(aiocb
->aio_iov
, aiocb
->aio_niov
, ret
,
1213 0, aiocb
->aio_nbytes
- ret
);
1215 ret
= aiocb
->aio_nbytes
;
1217 if (ret
== aiocb
->aio_nbytes
) {
1219 } else if (ret
>= 0 && ret
< aiocb
->aio_nbytes
) {
1223 case QEMU_AIO_WRITE
:
1224 ret
= handle_aiocb_rw(aiocb
);
1225 if (ret
== aiocb
->aio_nbytes
) {
1227 } else if (ret
>= 0 && ret
< aiocb
->aio_nbytes
) {
1231 case QEMU_AIO_FLUSH
:
1232 ret
= handle_aiocb_flush(aiocb
);
1234 case QEMU_AIO_IOCTL
:
1235 ret
= handle_aiocb_ioctl(aiocb
);
1237 case QEMU_AIO_DISCARD
:
1238 ret
= handle_aiocb_discard(aiocb
);
1240 case QEMU_AIO_WRITE_ZEROES
:
1241 ret
= handle_aiocb_write_zeroes(aiocb
);
1244 fprintf(stderr
, "invalid aio request (0x%x)\n", aiocb
->aio_type
);
1253 static int paio_submit_co(BlockDriverState
*bs
, int fd
,
1254 int64_t offset
, QEMUIOVector
*qiov
,
1255 int count
, int type
)
1257 RawPosixAIOData
*acb
= g_new(RawPosixAIOData
, 1);
1261 acb
->aio_type
= type
;
1262 acb
->aio_fildes
= fd
;
1264 acb
->aio_nbytes
= count
;
1265 acb
->aio_offset
= offset
;
1268 acb
->aio_iov
= qiov
->iov
;
1269 acb
->aio_niov
= qiov
->niov
;
1270 assert(qiov
->size
== count
);
1273 trace_paio_submit_co(offset
, count
, type
);
1274 pool
= aio_get_thread_pool(bdrv_get_aio_context(bs
));
1275 return thread_pool_submit_co(pool
, aio_worker
, acb
);
1278 static BlockAIOCB
*paio_submit(BlockDriverState
*bs
, int fd
,
1279 int64_t offset
, QEMUIOVector
*qiov
, int count
,
1280 BlockCompletionFunc
*cb
, void *opaque
, 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
== acb
->aio_nbytes
);
1298 trace_paio_submit(acb
, opaque
, offset
, count
, type
);
1299 pool
= aio_get_thread_pool(bdrv_get_aio_context(bs
));
1300 return thread_pool_submit_aio(pool
, aio_worker
, acb
, cb
, opaque
);
1303 static int coroutine_fn
raw_co_prw(BlockDriverState
*bs
, uint64_t offset
,
1304 uint64_t bytes
, QEMUIOVector
*qiov
, 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_linux_aio
) {
1322 LinuxAioState
*aio
= aio_get_linux_aio(bdrv_get_aio_context(bs
));
1323 assert(qiov
->size
== bytes
);
1324 return laio_co_submit(bs
, aio
, s
->fd
, offset
, qiov
, type
);
1329 return paio_submit_co(bs
, s
->fd
, offset
, qiov
, bytes
, type
);
1332 static int coroutine_fn
raw_co_preadv(BlockDriverState
*bs
, uint64_t offset
,
1333 uint64_t bytes
, QEMUIOVector
*qiov
,
1336 return raw_co_prw(bs
, offset
, bytes
, qiov
, QEMU_AIO_READ
);
1339 static int coroutine_fn
raw_co_pwritev(BlockDriverState
*bs
, uint64_t offset
,
1340 uint64_t bytes
, QEMUIOVector
*qiov
,
1344 return raw_co_prw(bs
, offset
, bytes
, qiov
, QEMU_AIO_WRITE
);
1347 static void raw_aio_plug(BlockDriverState
*bs
)
1349 #ifdef CONFIG_LINUX_AIO
1350 BDRVRawState
*s
= bs
->opaque
;
1351 if (s
->use_linux_aio
) {
1352 LinuxAioState
*aio
= aio_get_linux_aio(bdrv_get_aio_context(bs
));
1353 laio_io_plug(bs
, aio
);
1358 static void raw_aio_unplug(BlockDriverState
*bs
)
1360 #ifdef CONFIG_LINUX_AIO
1361 BDRVRawState
*s
= bs
->opaque
;
1362 if (s
->use_linux_aio
) {
1363 LinuxAioState
*aio
= aio_get_linux_aio(bdrv_get_aio_context(bs
));
1364 laio_io_unplug(bs
, aio
);
1369 static BlockAIOCB
*raw_aio_flush(BlockDriverState
*bs
,
1370 BlockCompletionFunc
*cb
, void *opaque
)
1372 BDRVRawState
*s
= bs
->opaque
;
1374 if (fd_open(bs
) < 0)
1377 return paio_submit(bs
, s
->fd
, 0, NULL
, 0, cb
, opaque
, QEMU_AIO_FLUSH
);
1380 static void raw_close(BlockDriverState
*bs
)
1382 BDRVRawState
*s
= bs
->opaque
;
1390 static int raw_truncate(BlockDriverState
*bs
, int64_t offset
)
1392 BDRVRawState
*s
= bs
->opaque
;
1395 if (fstat(s
->fd
, &st
)) {
1399 if (S_ISREG(st
.st_mode
)) {
1400 if (ftruncate(s
->fd
, offset
) < 0) {
1403 } else if (S_ISCHR(st
.st_mode
) || S_ISBLK(st
.st_mode
)) {
1404 if (offset
> raw_getlength(bs
)) {
1415 static int64_t raw_getlength(BlockDriverState
*bs
)
1417 BDRVRawState
*s
= bs
->opaque
;
1423 if (S_ISCHR(st
.st_mode
) || S_ISBLK(st
.st_mode
)) {
1424 struct disklabel dl
;
1426 if (ioctl(fd
, DIOCGDINFO
, &dl
))
1428 return (uint64_t)dl
.d_secsize
*
1429 dl
.d_partitions
[DISKPART(st
.st_rdev
)].p_size
;
1433 #elif defined(__NetBSD__)
1434 static int64_t raw_getlength(BlockDriverState
*bs
)
1436 BDRVRawState
*s
= bs
->opaque
;
1442 if (S_ISCHR(st
.st_mode
) || S_ISBLK(st
.st_mode
)) {
1443 struct dkwedge_info dkw
;
1445 if (ioctl(fd
, DIOCGWEDGEINFO
, &dkw
) != -1) {
1446 return dkw
.dkw_size
* 512;
1448 struct disklabel dl
;
1450 if (ioctl(fd
, DIOCGDINFO
, &dl
))
1452 return (uint64_t)dl
.d_secsize
*
1453 dl
.d_partitions
[DISKPART(st
.st_rdev
)].p_size
;
1458 #elif defined(__sun__)
1459 static int64_t raw_getlength(BlockDriverState
*bs
)
1461 BDRVRawState
*s
= bs
->opaque
;
1462 struct dk_minfo minfo
;
1472 * Use the DKIOCGMEDIAINFO ioctl to read the size.
1474 ret
= ioctl(s
->fd
, DKIOCGMEDIAINFO
, &minfo
);
1476 return minfo
.dki_lbsize
* minfo
.dki_capacity
;
1480 * There are reports that lseek on some devices fails, but
1481 * irc discussion said that contingency on contingency was overkill.
1483 size
= lseek(s
->fd
, 0, SEEK_END
);
1489 #elif defined(CONFIG_BSD)
1490 static int64_t raw_getlength(BlockDriverState
*bs
)
1492 BDRVRawState
*s
= bs
->opaque
;
1496 #if defined (__FreeBSD__) || defined(__FreeBSD_kernel__)
1505 #if defined (__FreeBSD__) || defined(__FreeBSD_kernel__)
1508 if (!fstat(fd
, &sb
) && (S_IFCHR
& sb
.st_mode
)) {
1509 #ifdef DIOCGMEDIASIZE
1510 if (ioctl(fd
, DIOCGMEDIASIZE
, (off_t
*)&size
))
1511 #elif defined(DIOCGPART)
1514 if (ioctl(fd
, DIOCGPART
, &pi
) == 0)
1515 size
= pi
.media_size
;
1521 #if defined(__APPLE__) && defined(__MACH__)
1523 uint64_t sectors
= 0;
1524 uint32_t sector_size
= 0;
1526 if (ioctl(fd
, DKIOCGETBLOCKCOUNT
, §ors
) == 0
1527 && ioctl(fd
, DKIOCGETBLOCKSIZE
, §or_size
) == 0) {
1528 size
= sectors
* sector_size
;
1530 size
= lseek(fd
, 0LL, SEEK_END
);
1537 size
= lseek(fd
, 0LL, SEEK_END
);
1542 #if defined(__FreeBSD__) || defined(__FreeBSD_kernel__)
1545 /* XXX FreeBSD acd returns UINT_MAX sectors for an empty drive */
1546 if (size
== 2048LL * (unsigned)-1)
1548 /* XXX no disc? maybe we need to reopen... */
1549 if (size
<= 0 && !reopened
&& cdrom_reopen(bs
) >= 0) {
1556 size
= lseek(fd
, 0, SEEK_END
);
1564 static int64_t raw_getlength(BlockDriverState
*bs
)
1566 BDRVRawState
*s
= bs
->opaque
;
1575 size
= lseek(s
->fd
, 0, SEEK_END
);
1583 static int64_t raw_get_allocated_file_size(BlockDriverState
*bs
)
1586 BDRVRawState
*s
= bs
->opaque
;
1588 if (fstat(s
->fd
, &st
) < 0) {
1591 return (int64_t)st
.st_blocks
* 512;
1594 static int raw_create(const char *filename
, QemuOpts
*opts
, Error
**errp
)
1598 int64_t total_size
= 0;
1600 PreallocMode prealloc
;
1602 Error
*local_err
= NULL
;
1604 strstart(filename
, "file:", &filename
);
1606 /* Read out options */
1607 total_size
= ROUND_UP(qemu_opt_get_size_del(opts
, BLOCK_OPT_SIZE
, 0),
1609 nocow
= qemu_opt_get_bool(opts
, BLOCK_OPT_NOCOW
, false);
1610 buf
= qemu_opt_get_del(opts
, BLOCK_OPT_PREALLOC
);
1611 prealloc
= qapi_enum_parse(PreallocMode_lookup
, buf
,
1612 PREALLOC_MODE__MAX
, PREALLOC_MODE_OFF
,
1616 error_propagate(errp
, local_err
);
1621 fd
= qemu_open(filename
, O_RDWR
| O_CREAT
| O_TRUNC
| O_BINARY
,
1625 error_setg_errno(errp
, -result
, "Could not create file");
1631 /* Set NOCOW flag to solve performance issue on fs like btrfs.
1632 * This is an optimisation. The FS_IOC_SETFLAGS ioctl return value
1633 * will be ignored since any failure of this operation should not
1634 * block the left work.
1637 if (ioctl(fd
, FS_IOC_GETFLAGS
, &attr
) == 0) {
1638 attr
|= FS_NOCOW_FL
;
1639 ioctl(fd
, FS_IOC_SETFLAGS
, &attr
);
1645 #ifdef CONFIG_POSIX_FALLOCATE
1646 case PREALLOC_MODE_FALLOC
:
1648 * Truncating before posix_fallocate() makes it about twice slower on
1649 * file systems that do not support fallocate(), trying to check if a
1650 * block is allocated before allocating it, so don't do that here.
1652 result
= -posix_fallocate(fd
, 0, total_size
);
1654 /* posix_fallocate() doesn't set errno. */
1655 error_setg_errno(errp
, -result
,
1656 "Could not preallocate data for the new file");
1660 case PREALLOC_MODE_FULL
:
1663 * Knowing the final size from the beginning could allow the file
1664 * system driver to do less allocations and possibly avoid
1665 * fragmentation of the file.
1667 if (ftruncate(fd
, total_size
) != 0) {
1669 error_setg_errno(errp
, -result
, "Could not resize file");
1673 int64_t num
= 0, left
= total_size
;
1674 buf
= g_malloc0(65536);
1677 num
= MIN(left
, 65536);
1678 result
= write(fd
, buf
, num
);
1681 error_setg_errno(errp
, -result
,
1682 "Could not write to the new file");
1691 error_setg_errno(errp
, -result
,
1692 "Could not flush new file to disk");
1698 case PREALLOC_MODE_OFF
:
1699 if (ftruncate(fd
, total_size
) != 0) {
1701 error_setg_errno(errp
, -result
, "Could not resize file");
1706 error_setg(errp
, "Unsupported preallocation mode: %s",
1707 PreallocMode_lookup
[prealloc
]);
1712 if (qemu_close(fd
) != 0 && result
== 0) {
1714 error_setg_errno(errp
, -result
, "Could not close the new file");
1721 * Find allocation range in @bs around offset @start.
1722 * May change underlying file descriptor's file offset.
1723 * If @start is not in a hole, store @start in @data, and the
1724 * beginning of the next hole in @hole, and return 0.
1725 * If @start is in a non-trailing hole, store @start in @hole and the
1726 * beginning of the next non-hole in @data, and return 0.
1727 * If @start is in a trailing hole or beyond EOF, return -ENXIO.
1728 * If we can't find out, return a negative errno other than -ENXIO.
1730 static int find_allocation(BlockDriverState
*bs
, off_t start
,
1731 off_t
*data
, off_t
*hole
)
1733 #if defined SEEK_HOLE && defined SEEK_DATA
1734 BDRVRawState
*s
= bs
->opaque
;
1739 * D1. offs == start: start is in data
1740 * D2. offs > start: start is in a hole, next data at offs
1741 * D3. offs < 0, errno = ENXIO: either start is in a trailing hole
1742 * or start is beyond EOF
1743 * If the latter happens, the file has been truncated behind
1744 * our back since we opened it. All bets are off then.
1745 * Treating like a trailing hole is simplest.
1746 * D4. offs < 0, errno != ENXIO: we learned nothing
1748 offs
= lseek(s
->fd
, start
, SEEK_DATA
);
1750 return -errno
; /* D3 or D4 */
1752 assert(offs
>= start
);
1755 /* D2: in hole, next data at offs */
1761 /* D1: in data, end not yet known */
1765 * H1. offs == start: start is in a hole
1766 * If this happens here, a hole has been dug behind our back
1767 * since the previous lseek().
1768 * H2. offs > start: either start is in data, next hole at offs,
1769 * or start is in trailing hole, EOF at offs
1770 * Linux treats trailing holes like any other hole: offs ==
1771 * start. Solaris seeks to EOF instead: offs > start (blech).
1772 * If that happens here, a hole has been dug behind our back
1773 * since the previous lseek().
1774 * H3. offs < 0, errno = ENXIO: start is beyond EOF
1775 * If this happens, the file has been truncated behind our
1776 * back since we opened it. Treat it like a trailing hole.
1777 * H4. offs < 0, errno != ENXIO: we learned nothing
1778 * Pretend we know nothing at all, i.e. "forget" about D1.
1780 offs
= lseek(s
->fd
, start
, SEEK_HOLE
);
1782 return -errno
; /* D1 and (H3 or H4) */
1784 assert(offs
>= start
);
1788 * D1 and H2: either in data, next hole at offs, or it was in
1789 * data but is now in a trailing hole. In the latter case,
1790 * all bets are off. Treating it as if it there was data all
1791 * the way to EOF is safe, so simply do that.
1806 * Returns the allocation status of the specified sectors.
1808 * If 'sector_num' is beyond the end of the disk image the return value is 0
1809 * and 'pnum' is set to 0.
1811 * 'pnum' is set to the number of sectors (including and immediately following
1812 * the specified sector) that are known to be in the same
1813 * allocated/unallocated state.
1815 * 'nb_sectors' is the max value 'pnum' should be set to. If nb_sectors goes
1816 * beyond the end of the disk image it will be clamped.
1818 static int64_t coroutine_fn
raw_co_get_block_status(BlockDriverState
*bs
,
1820 int nb_sectors
, int *pnum
,
1821 BlockDriverState
**file
)
1823 off_t start
, data
= 0, hole
= 0;
1832 start
= sector_num
* BDRV_SECTOR_SIZE
;
1833 total_size
= bdrv_getlength(bs
);
1834 if (total_size
< 0) {
1836 } else if (start
>= total_size
) {
1839 } else if (start
+ nb_sectors
* BDRV_SECTOR_SIZE
> total_size
) {
1840 nb_sectors
= DIV_ROUND_UP(total_size
- start
, BDRV_SECTOR_SIZE
);
1843 ret
= find_allocation(bs
, start
, &data
, &hole
);
1844 if (ret
== -ENXIO
) {
1847 ret
= BDRV_BLOCK_ZERO
;
1848 } else if (ret
< 0) {
1849 /* No info available, so pretend there are no holes */
1851 ret
= BDRV_BLOCK_DATA
;
1852 } else if (data
== start
) {
1853 /* On a data extent, compute sectors to the end of the extent,
1854 * possibly including a partial sector at EOF. */
1855 *pnum
= MIN(nb_sectors
, DIV_ROUND_UP(hole
- start
, BDRV_SECTOR_SIZE
));
1856 ret
= BDRV_BLOCK_DATA
;
1858 /* On a hole, compute sectors to the beginning of the next extent. */
1859 assert(hole
== start
);
1860 *pnum
= MIN(nb_sectors
, (data
- start
) / BDRV_SECTOR_SIZE
);
1861 ret
= BDRV_BLOCK_ZERO
;
1864 return ret
| BDRV_BLOCK_OFFSET_VALID
| start
;
1867 static coroutine_fn BlockAIOCB
*raw_aio_pdiscard(BlockDriverState
*bs
,
1868 int64_t offset
, int count
,
1869 BlockCompletionFunc
*cb
, void *opaque
)
1871 BDRVRawState
*s
= bs
->opaque
;
1873 return paio_submit(bs
, s
->fd
, offset
, NULL
, count
,
1874 cb
, opaque
, QEMU_AIO_DISCARD
);
1877 static int coroutine_fn
raw_co_pwrite_zeroes(
1878 BlockDriverState
*bs
, int64_t offset
,
1879 int count
, BdrvRequestFlags flags
)
1881 BDRVRawState
*s
= bs
->opaque
;
1883 if (!(flags
& BDRV_REQ_MAY_UNMAP
)) {
1884 return paio_submit_co(bs
, s
->fd
, offset
, NULL
, count
,
1885 QEMU_AIO_WRITE_ZEROES
);
1886 } else if (s
->discard_zeroes
) {
1887 return paio_submit_co(bs
, s
->fd
, offset
, NULL
, count
,
1893 static int raw_get_info(BlockDriverState
*bs
, BlockDriverInfo
*bdi
)
1895 BDRVRawState
*s
= bs
->opaque
;
1897 bdi
->unallocated_blocks_are_zero
= s
->discard_zeroes
;
1898 bdi
->can_write_zeroes_with_unmap
= s
->discard_zeroes
;
1902 static QemuOptsList raw_create_opts
= {
1903 .name
= "raw-create-opts",
1904 .head
= QTAILQ_HEAD_INITIALIZER(raw_create_opts
.head
),
1907 .name
= BLOCK_OPT_SIZE
,
1908 .type
= QEMU_OPT_SIZE
,
1909 .help
= "Virtual disk size"
1912 .name
= BLOCK_OPT_NOCOW
,
1913 .type
= QEMU_OPT_BOOL
,
1914 .help
= "Turn off copy-on-write (valid only on btrfs)"
1917 .name
= BLOCK_OPT_PREALLOC
,
1918 .type
= QEMU_OPT_STRING
,
1919 .help
= "Preallocation mode (allowed values: off, falloc, full)"
1921 { /* end of list */ }
1925 BlockDriver bdrv_file
= {
1926 .format_name
= "file",
1927 .protocol_name
= "file",
1928 .instance_size
= sizeof(BDRVRawState
),
1929 .bdrv_needs_filename
= true,
1930 .bdrv_probe
= NULL
, /* no probe for protocols */
1931 .bdrv_parse_filename
= raw_parse_filename
,
1932 .bdrv_file_open
= raw_open
,
1933 .bdrv_reopen_prepare
= raw_reopen_prepare
,
1934 .bdrv_reopen_commit
= raw_reopen_commit
,
1935 .bdrv_reopen_abort
= raw_reopen_abort
,
1936 .bdrv_close
= raw_close
,
1937 .bdrv_create
= raw_create
,
1938 .bdrv_has_zero_init
= bdrv_has_zero_init_1
,
1939 .bdrv_co_get_block_status
= raw_co_get_block_status
,
1940 .bdrv_co_pwrite_zeroes
= raw_co_pwrite_zeroes
,
1942 .bdrv_co_preadv
= raw_co_preadv
,
1943 .bdrv_co_pwritev
= raw_co_pwritev
,
1944 .bdrv_aio_flush
= raw_aio_flush
,
1945 .bdrv_aio_pdiscard
= raw_aio_pdiscard
,
1946 .bdrv_refresh_limits
= raw_refresh_limits
,
1947 .bdrv_io_plug
= raw_aio_plug
,
1948 .bdrv_io_unplug
= raw_aio_unplug
,
1950 .bdrv_truncate
= raw_truncate
,
1951 .bdrv_getlength
= raw_getlength
,
1952 .bdrv_get_info
= raw_get_info
,
1953 .bdrv_get_allocated_file_size
1954 = raw_get_allocated_file_size
,
1956 .create_opts
= &raw_create_opts
,
1959 /***********************************************/
1962 #if defined(__APPLE__) && defined(__MACH__)
1963 static kern_return_t
GetBSDPath(io_iterator_t mediaIterator
, char *bsdPath
,
1964 CFIndex maxPathSize
, int flags
);
1965 static char *FindEjectableOpticalMedia(io_iterator_t
*mediaIterator
)
1967 kern_return_t kernResult
= KERN_FAILURE
;
1968 mach_port_t masterPort
;
1969 CFMutableDictionaryRef classesToMatch
;
1970 const char *matching_array
[] = {kIODVDMediaClass
, kIOCDMediaClass
};
1971 char *mediaType
= NULL
;
1973 kernResult
= IOMasterPort( MACH_PORT_NULL
, &masterPort
);
1974 if ( KERN_SUCCESS
!= kernResult
) {
1975 printf( "IOMasterPort returned %d\n", kernResult
);
1979 for (index
= 0; index
< ARRAY_SIZE(matching_array
); index
++) {
1980 classesToMatch
= IOServiceMatching(matching_array
[index
]);
1981 if (classesToMatch
== NULL
) {
1982 error_report("IOServiceMatching returned NULL for %s",
1983 matching_array
[index
]);
1986 CFDictionarySetValue(classesToMatch
, CFSTR(kIOMediaEjectableKey
),
1988 kernResult
= IOServiceGetMatchingServices(masterPort
, classesToMatch
,
1990 if (kernResult
!= KERN_SUCCESS
) {
1991 error_report("Note: IOServiceGetMatchingServices returned %d",
1996 /* If a match was found, leave the loop */
1997 if (*mediaIterator
!= 0) {
1998 DPRINTF("Matching using %s\n", matching_array
[index
]);
1999 mediaType
= g_strdup(matching_array
[index
]);
2006 kern_return_t
GetBSDPath(io_iterator_t mediaIterator
, char *bsdPath
,
2007 CFIndex maxPathSize
, int flags
)
2009 io_object_t nextMedia
;
2010 kern_return_t kernResult
= KERN_FAILURE
;
2012 nextMedia
= IOIteratorNext( mediaIterator
);
2015 CFTypeRef bsdPathAsCFString
;
2016 bsdPathAsCFString
= IORegistryEntryCreateCFProperty( nextMedia
, CFSTR( kIOBSDNameKey
), kCFAllocatorDefault
, 0 );
2017 if ( bsdPathAsCFString
) {
2018 size_t devPathLength
;
2019 strcpy( bsdPath
, _PATH_DEV
);
2020 if (flags
& BDRV_O_NOCACHE
) {
2021 strcat(bsdPath
, "r");
2023 devPathLength
= strlen( bsdPath
);
2024 if ( CFStringGetCString( bsdPathAsCFString
, bsdPath
+ devPathLength
, maxPathSize
- devPathLength
, kCFStringEncodingASCII
) ) {
2025 kernResult
= KERN_SUCCESS
;
2027 CFRelease( bsdPathAsCFString
);
2029 IOObjectRelease( nextMedia
);
2035 /* Sets up a real cdrom for use in QEMU */
2036 static bool setup_cdrom(char *bsd_path
, Error
**errp
)
2038 int index
, num_of_test_partitions
= 2, fd
;
2039 char test_partition
[MAXPATHLEN
];
2040 bool partition_found
= false;
2042 /* look for a working partition */
2043 for (index
= 0; index
< num_of_test_partitions
; index
++) {
2044 snprintf(test_partition
, sizeof(test_partition
), "%ss%d", bsd_path
,
2046 fd
= qemu_open(test_partition
, O_RDONLY
| O_BINARY
| O_LARGEFILE
);
2048 partition_found
= true;
2054 /* if a working partition on the device was not found */
2055 if (partition_found
== false) {
2056 error_setg(errp
, "Failed to find a working partition on disc");
2058 DPRINTF("Using %s as optical disc\n", test_partition
);
2059 pstrcpy(bsd_path
, MAXPATHLEN
, test_partition
);
2061 return partition_found
;
2064 /* Prints directions on mounting and unmounting a device */
2065 static void print_unmounting_directions(const char *file_name
)
2067 error_report("If device %s is mounted on the desktop, unmount"
2068 " it first before using it in QEMU", file_name
);
2069 error_report("Command to unmount device: diskutil unmountDisk %s",
2071 error_report("Command to mount device: diskutil mountDisk %s", file_name
);
2074 #endif /* defined(__APPLE__) && defined(__MACH__) */
2076 static int hdev_probe_device(const char *filename
)
2080 /* allow a dedicated CD-ROM driver to match with a higher priority */
2081 if (strstart(filename
, "/dev/cdrom", NULL
))
2084 if (stat(filename
, &st
) >= 0 &&
2085 (S_ISCHR(st
.st_mode
) || S_ISBLK(st
.st_mode
))) {
2092 static int check_hdev_writable(BDRVRawState
*s
)
2094 #if defined(BLKROGET)
2095 /* Linux block devices can be configured "read-only" using blockdev(8).
2096 * This is independent of device node permissions and therefore open(2)
2097 * with O_RDWR succeeds. Actual writes fail with EPERM.
2099 * bdrv_open() is supposed to fail if the disk is read-only. Explicitly
2100 * check for read-only block devices so that Linux block devices behave
2106 if (fstat(s
->fd
, &st
)) {
2110 if (!S_ISBLK(st
.st_mode
)) {
2114 if (ioctl(s
->fd
, BLKROGET
, &readonly
) < 0) {
2121 #endif /* defined(BLKROGET) */
2125 static void hdev_parse_filename(const char *filename
, QDict
*options
,
2128 /* The prefix is optional, just as for "file". */
2129 strstart(filename
, "host_device:", &filename
);
2131 qdict_put_obj(options
, "filename", QOBJECT(qstring_from_str(filename
)));
2134 static bool hdev_is_sg(BlockDriverState
*bs
)
2137 #if defined(__linux__)
2139 BDRVRawState
*s
= bs
->opaque
;
2141 struct sg_scsi_id scsiid
;
2145 if (stat(bs
->filename
, &st
) < 0 || !S_ISCHR(st
.st_mode
)) {
2149 ret
= ioctl(s
->fd
, SG_GET_VERSION_NUM
, &sg_version
);
2154 ret
= ioctl(s
->fd
, SG_GET_SCSI_ID
, &scsiid
);
2156 DPRINTF("SG device found: type=%d, version=%d\n",
2157 scsiid
.scsi_type
, sg_version
);
2166 static int hdev_open(BlockDriverState
*bs
, QDict
*options
, int flags
,
2169 BDRVRawState
*s
= bs
->opaque
;
2170 Error
*local_err
= NULL
;
2173 #if defined(__APPLE__) && defined(__MACH__)
2174 const char *filename
= qdict_get_str(options
, "filename");
2175 char bsd_path
[MAXPATHLEN
] = "";
2176 bool error_occurred
= false;
2178 /* If using a real cdrom */
2179 if (strcmp(filename
, "/dev/cdrom") == 0) {
2180 char *mediaType
= NULL
;
2181 kern_return_t ret_val
;
2182 io_iterator_t mediaIterator
= 0;
2184 mediaType
= FindEjectableOpticalMedia(&mediaIterator
);
2185 if (mediaType
== NULL
) {
2186 error_setg(errp
, "Please make sure your CD/DVD is in the optical"
2188 error_occurred
= true;
2189 goto hdev_open_Mac_error
;
2192 ret_val
= GetBSDPath(mediaIterator
, bsd_path
, sizeof(bsd_path
), flags
);
2193 if (ret_val
!= KERN_SUCCESS
) {
2194 error_setg(errp
, "Could not get BSD path for optical drive");
2195 error_occurred
= true;
2196 goto hdev_open_Mac_error
;
2199 /* If a real optical drive was not found */
2200 if (bsd_path
[0] == '\0') {
2201 error_setg(errp
, "Failed to obtain bsd path for optical drive");
2202 error_occurred
= true;
2203 goto hdev_open_Mac_error
;
2206 /* If using a cdrom disc and finding a partition on the disc failed */
2207 if (strncmp(mediaType
, kIOCDMediaClass
, 9) == 0 &&
2208 setup_cdrom(bsd_path
, errp
) == false) {
2209 print_unmounting_directions(bsd_path
);
2210 error_occurred
= true;
2211 goto hdev_open_Mac_error
;
2214 qdict_put(options
, "filename", qstring_from_str(bsd_path
));
2216 hdev_open_Mac_error
:
2218 if (mediaIterator
) {
2219 IOObjectRelease(mediaIterator
);
2221 if (error_occurred
) {
2225 #endif /* defined(__APPLE__) && defined(__MACH__) */
2227 s
->type
= FTYPE_FILE
;
2229 ret
= raw_open_common(bs
, options
, flags
, 0, &local_err
);
2231 error_propagate(errp
, local_err
);
2232 #if defined(__APPLE__) && defined(__MACH__)
2234 filename
= bsd_path
;
2236 /* if a physical device experienced an error while being opened */
2237 if (strncmp(filename
, "/dev/", 5) == 0) {
2238 print_unmounting_directions(filename
);
2240 #endif /* defined(__APPLE__) && defined(__MACH__) */
2244 /* Since this does ioctl the device must be already opened */
2245 bs
->sg
= hdev_is_sg(bs
);
2247 if (flags
& BDRV_O_RDWR
) {
2248 ret
= check_hdev_writable(s
);
2251 error_setg_errno(errp
, -ret
, "The device is not writable");
2259 #if defined(__linux__)
2261 static BlockAIOCB
*hdev_aio_ioctl(BlockDriverState
*bs
,
2262 unsigned long int req
, void *buf
,
2263 BlockCompletionFunc
*cb
, void *opaque
)
2265 BDRVRawState
*s
= bs
->opaque
;
2266 RawPosixAIOData
*acb
;
2269 if (fd_open(bs
) < 0)
2272 acb
= g_new(RawPosixAIOData
, 1);
2274 acb
->aio_type
= QEMU_AIO_IOCTL
;
2275 acb
->aio_fildes
= s
->fd
;
2276 acb
->aio_offset
= 0;
2277 acb
->aio_ioctl_buf
= buf
;
2278 acb
->aio_ioctl_cmd
= req
;
2279 pool
= aio_get_thread_pool(bdrv_get_aio_context(bs
));
2280 return thread_pool_submit_aio(pool
, aio_worker
, acb
, cb
, opaque
);
2284 static int fd_open(BlockDriverState
*bs
)
2286 BDRVRawState
*s
= bs
->opaque
;
2288 /* this is just to ensure s->fd is sane (its called by io ops) */
2294 static coroutine_fn BlockAIOCB
*hdev_aio_pdiscard(BlockDriverState
*bs
,
2295 int64_t offset
, int count
,
2296 BlockCompletionFunc
*cb
, void *opaque
)
2298 BDRVRawState
*s
= bs
->opaque
;
2300 if (fd_open(bs
) < 0) {
2303 return paio_submit(bs
, s
->fd
, offset
, NULL
, count
,
2304 cb
, opaque
, QEMU_AIO_DISCARD
|QEMU_AIO_BLKDEV
);
2307 static coroutine_fn
int hdev_co_pwrite_zeroes(BlockDriverState
*bs
,
2308 int64_t offset
, int count
, BdrvRequestFlags flags
)
2310 BDRVRawState
*s
= bs
->opaque
;
2317 if (!(flags
& BDRV_REQ_MAY_UNMAP
)) {
2318 return paio_submit_co(bs
, s
->fd
, offset
, NULL
, count
,
2319 QEMU_AIO_WRITE_ZEROES
|QEMU_AIO_BLKDEV
);
2320 } else if (s
->discard_zeroes
) {
2321 return paio_submit_co(bs
, s
->fd
, offset
, NULL
, count
,
2322 QEMU_AIO_DISCARD
|QEMU_AIO_BLKDEV
);
2327 static int hdev_create(const char *filename
, QemuOpts
*opts
,
2332 struct stat stat_buf
;
2333 int64_t total_size
= 0;
2336 /* This function is used by both protocol block drivers and therefore either
2337 * of these prefixes may be given.
2338 * The return value has to be stored somewhere, otherwise this is an error
2339 * due to -Werror=unused-value. */
2341 strstart(filename
, "host_device:", &filename
) ||
2342 strstart(filename
, "host_cdrom:" , &filename
);
2346 ret
= raw_normalize_devicepath(&filename
);
2348 error_setg_errno(errp
, -ret
, "Could not normalize device path");
2352 /* Read out options */
2353 total_size
= ROUND_UP(qemu_opt_get_size_del(opts
, BLOCK_OPT_SIZE
, 0),
2356 fd
= qemu_open(filename
, O_WRONLY
| O_BINARY
);
2359 error_setg_errno(errp
, -ret
, "Could not open device");
2363 if (fstat(fd
, &stat_buf
) < 0) {
2365 error_setg_errno(errp
, -ret
, "Could not stat device");
2366 } else if (!S_ISBLK(stat_buf
.st_mode
) && !S_ISCHR(stat_buf
.st_mode
)) {
2368 "The given file is neither a block nor a character device");
2370 } else if (lseek(fd
, 0, SEEK_END
) < total_size
) {
2371 error_setg(errp
, "Device is too small");
2379 static BlockDriver bdrv_host_device
= {
2380 .format_name
= "host_device",
2381 .protocol_name
= "host_device",
2382 .instance_size
= sizeof(BDRVRawState
),
2383 .bdrv_needs_filename
= true,
2384 .bdrv_probe_device
= hdev_probe_device
,
2385 .bdrv_parse_filename
= hdev_parse_filename
,
2386 .bdrv_file_open
= hdev_open
,
2387 .bdrv_close
= raw_close
,
2388 .bdrv_reopen_prepare
= raw_reopen_prepare
,
2389 .bdrv_reopen_commit
= raw_reopen_commit
,
2390 .bdrv_reopen_abort
= raw_reopen_abort
,
2391 .bdrv_create
= hdev_create
,
2392 .create_opts
= &raw_create_opts
,
2393 .bdrv_co_pwrite_zeroes
= hdev_co_pwrite_zeroes
,
2395 .bdrv_co_preadv
= raw_co_preadv
,
2396 .bdrv_co_pwritev
= raw_co_pwritev
,
2397 .bdrv_aio_flush
= raw_aio_flush
,
2398 .bdrv_aio_pdiscard
= hdev_aio_pdiscard
,
2399 .bdrv_refresh_limits
= raw_refresh_limits
,
2400 .bdrv_io_plug
= raw_aio_plug
,
2401 .bdrv_io_unplug
= raw_aio_unplug
,
2403 .bdrv_truncate
= raw_truncate
,
2404 .bdrv_getlength
= raw_getlength
,
2405 .bdrv_get_info
= raw_get_info
,
2406 .bdrv_get_allocated_file_size
2407 = raw_get_allocated_file_size
,
2408 .bdrv_probe_blocksizes
= hdev_probe_blocksizes
,
2409 .bdrv_probe_geometry
= hdev_probe_geometry
,
2411 /* generic scsi device */
2413 .bdrv_aio_ioctl
= hdev_aio_ioctl
,
2417 #if defined(__linux__) || defined(__FreeBSD__) || defined(__FreeBSD_kernel__)
2418 static void cdrom_parse_filename(const char *filename
, QDict
*options
,
2421 /* The prefix is optional, just as for "file". */
2422 strstart(filename
, "host_cdrom:", &filename
);
2424 qdict_put_obj(options
, "filename", QOBJECT(qstring_from_str(filename
)));
2429 static int cdrom_open(BlockDriverState
*bs
, QDict
*options
, int flags
,
2432 BDRVRawState
*s
= bs
->opaque
;
2436 /* open will not fail even if no CD is inserted, so add O_NONBLOCK */
2437 return raw_open_common(bs
, options
, flags
, O_NONBLOCK
, errp
);
2440 static int cdrom_probe_device(const char *filename
)
2446 fd
= qemu_open(filename
, O_RDONLY
| O_NONBLOCK
);
2450 ret
= fstat(fd
, &st
);
2451 if (ret
== -1 || !S_ISBLK(st
.st_mode
)) {
2455 /* Attempt to detect via a CDROM specific ioctl */
2456 ret
= ioctl(fd
, CDROM_DRIVE_STATUS
, CDSL_CURRENT
);
2466 static bool cdrom_is_inserted(BlockDriverState
*bs
)
2468 BDRVRawState
*s
= bs
->opaque
;
2471 ret
= ioctl(s
->fd
, CDROM_DRIVE_STATUS
, CDSL_CURRENT
);
2472 return ret
== CDS_DISC_OK
;
2475 static void cdrom_eject(BlockDriverState
*bs
, bool eject_flag
)
2477 BDRVRawState
*s
= bs
->opaque
;
2480 if (ioctl(s
->fd
, CDROMEJECT
, NULL
) < 0)
2481 perror("CDROMEJECT");
2483 if (ioctl(s
->fd
, CDROMCLOSETRAY
, NULL
) < 0)
2484 perror("CDROMEJECT");
2488 static void cdrom_lock_medium(BlockDriverState
*bs
, bool locked
)
2490 BDRVRawState
*s
= bs
->opaque
;
2492 if (ioctl(s
->fd
, CDROM_LOCKDOOR
, locked
) < 0) {
2494 * Note: an error can happen if the distribution automatically
2497 /* perror("CDROM_LOCKDOOR"); */
2501 static BlockDriver bdrv_host_cdrom
= {
2502 .format_name
= "host_cdrom",
2503 .protocol_name
= "host_cdrom",
2504 .instance_size
= sizeof(BDRVRawState
),
2505 .bdrv_needs_filename
= true,
2506 .bdrv_probe_device
= cdrom_probe_device
,
2507 .bdrv_parse_filename
= cdrom_parse_filename
,
2508 .bdrv_file_open
= cdrom_open
,
2509 .bdrv_close
= raw_close
,
2510 .bdrv_reopen_prepare
= raw_reopen_prepare
,
2511 .bdrv_reopen_commit
= raw_reopen_commit
,
2512 .bdrv_reopen_abort
= raw_reopen_abort
,
2513 .bdrv_create
= hdev_create
,
2514 .create_opts
= &raw_create_opts
,
2517 .bdrv_co_preadv
= raw_co_preadv
,
2518 .bdrv_co_pwritev
= raw_co_pwritev
,
2519 .bdrv_aio_flush
= raw_aio_flush
,
2520 .bdrv_refresh_limits
= raw_refresh_limits
,
2521 .bdrv_io_plug
= raw_aio_plug
,
2522 .bdrv_io_unplug
= raw_aio_unplug
,
2524 .bdrv_truncate
= raw_truncate
,
2525 .bdrv_getlength
= raw_getlength
,
2526 .has_variable_length
= true,
2527 .bdrv_get_allocated_file_size
2528 = raw_get_allocated_file_size
,
2530 /* removable device support */
2531 .bdrv_is_inserted
= cdrom_is_inserted
,
2532 .bdrv_eject
= cdrom_eject
,
2533 .bdrv_lock_medium
= cdrom_lock_medium
,
2535 /* generic scsi device */
2536 .bdrv_aio_ioctl
= hdev_aio_ioctl
,
2538 #endif /* __linux__ */
2540 #if defined (__FreeBSD__) || defined(__FreeBSD_kernel__)
2541 static int cdrom_open(BlockDriverState
*bs
, QDict
*options
, int flags
,
2544 BDRVRawState
*s
= bs
->opaque
;
2545 Error
*local_err
= NULL
;
2550 ret
= raw_open_common(bs
, options
, flags
, 0, &local_err
);
2552 error_propagate(errp
, local_err
);
2556 /* make sure the door isn't locked at this time */
2557 ioctl(s
->fd
, CDIOCALLOW
);
2561 static int cdrom_probe_device(const char *filename
)
2563 if (strstart(filename
, "/dev/cd", NULL
) ||
2564 strstart(filename
, "/dev/acd", NULL
))
2569 static int cdrom_reopen(BlockDriverState
*bs
)
2571 BDRVRawState
*s
= bs
->opaque
;
2575 * Force reread of possibly changed/newly loaded disc,
2576 * FreeBSD seems to not notice sometimes...
2580 fd
= qemu_open(bs
->filename
, s
->open_flags
, 0644);
2587 /* make sure the door isn't locked at this time */
2588 ioctl(s
->fd
, CDIOCALLOW
);
2592 static bool cdrom_is_inserted(BlockDriverState
*bs
)
2594 return raw_getlength(bs
) > 0;
2597 static void cdrom_eject(BlockDriverState
*bs
, bool eject_flag
)
2599 BDRVRawState
*s
= bs
->opaque
;
2604 (void) ioctl(s
->fd
, CDIOCALLOW
);
2607 if (ioctl(s
->fd
, CDIOCEJECT
) < 0)
2608 perror("CDIOCEJECT");
2610 if (ioctl(s
->fd
, CDIOCCLOSE
) < 0)
2611 perror("CDIOCCLOSE");
2617 static void cdrom_lock_medium(BlockDriverState
*bs
, bool locked
)
2619 BDRVRawState
*s
= bs
->opaque
;
2623 if (ioctl(s
->fd
, (locked
? CDIOCPREVENT
: CDIOCALLOW
)) < 0) {
2625 * Note: an error can happen if the distribution automatically
2628 /* perror("CDROM_LOCKDOOR"); */
2632 static BlockDriver bdrv_host_cdrom
= {
2633 .format_name
= "host_cdrom",
2634 .protocol_name
= "host_cdrom",
2635 .instance_size
= sizeof(BDRVRawState
),
2636 .bdrv_needs_filename
= true,
2637 .bdrv_probe_device
= cdrom_probe_device
,
2638 .bdrv_parse_filename
= cdrom_parse_filename
,
2639 .bdrv_file_open
= cdrom_open
,
2640 .bdrv_close
= raw_close
,
2641 .bdrv_reopen_prepare
= raw_reopen_prepare
,
2642 .bdrv_reopen_commit
= raw_reopen_commit
,
2643 .bdrv_reopen_abort
= raw_reopen_abort
,
2644 .bdrv_create
= hdev_create
,
2645 .create_opts
= &raw_create_opts
,
2647 .bdrv_co_preadv
= raw_co_preadv
,
2648 .bdrv_co_pwritev
= raw_co_pwritev
,
2649 .bdrv_aio_flush
= raw_aio_flush
,
2650 .bdrv_refresh_limits
= raw_refresh_limits
,
2651 .bdrv_io_plug
= raw_aio_plug
,
2652 .bdrv_io_unplug
= raw_aio_unplug
,
2654 .bdrv_truncate
= raw_truncate
,
2655 .bdrv_getlength
= raw_getlength
,
2656 .has_variable_length
= true,
2657 .bdrv_get_allocated_file_size
2658 = raw_get_allocated_file_size
,
2660 /* removable device support */
2661 .bdrv_is_inserted
= cdrom_is_inserted
,
2662 .bdrv_eject
= cdrom_eject
,
2663 .bdrv_lock_medium
= cdrom_lock_medium
,
2665 #endif /* __FreeBSD__ */
2667 static void bdrv_file_init(void)
2670 * Register all the drivers. Note that order is important, the driver
2671 * registered last will get probed first.
2673 bdrv_register(&bdrv_file
);
2674 bdrv_register(&bdrv_host_device
);
2676 bdrv_register(&bdrv_host_cdrom
);
2678 #if defined(__FreeBSD__) || defined(__FreeBSD_kernel__)
2679 bdrv_register(&bdrv_host_cdrom
);
2683 block_init(bdrv_file_init
);