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-common.h"
25 #include "qemu/error-report.h"
26 #include "qemu/timer.h"
28 #include "block/block_int.h"
29 #include "qemu/module.h"
31 #include "block/thread-pool.h"
34 #include "qapi/util.h"
35 #include "qapi/qmp/qstring.h"
37 #if defined(__APPLE__) && (__MACH__)
39 #include <sys/param.h>
40 #include <IOKit/IOKitLib.h>
41 #include <IOKit/IOBSD.h>
42 #include <IOKit/storage/IOMediaBSDClient.h>
43 #include <IOKit/storage/IOMedia.h>
44 #include <IOKit/storage/IOCDMedia.h>
45 //#include <IOKit/storage/IOCDTypes.h>
46 #include <CoreFoundation/CoreFoundation.h>
50 #define _POSIX_PTHREAD_SEMANTICS 1
54 #include <sys/types.h>
56 #include <sys/ioctl.h>
57 #include <sys/param.h>
58 #include <linux/cdrom.h>
61 #include <linux/hdreg.h>
66 #define FS_NOCOW_FL 0x00800000 /* Do not cow file */
69 #if defined(CONFIG_FALLOCATE_PUNCH_HOLE) || defined(CONFIG_FALLOCATE_ZERO_RANGE)
70 #include <linux/falloc.h>
72 #if defined (__FreeBSD__) || defined(__FreeBSD_kernel__)
78 #include <sys/ioctl.h>
79 #include <sys/disklabel.h>
84 #include <sys/ioctl.h>
85 #include <sys/disklabel.h>
91 #include <sys/ioctl.h>
92 #include <sys/diskslice.h>
99 //#define DEBUG_FLOPPY
101 //#define DEBUG_BLOCK
102 #if defined(DEBUG_BLOCK)
103 #define DEBUG_BLOCK_PRINT(formatCstr, ...) do { if (qemu_log_enabled()) \
104 { qemu_log(formatCstr, ## __VA_ARGS__); qemu_log_flush(); } } while (0)
106 #define DEBUG_BLOCK_PRINT(formatCstr, ...)
109 /* OS X does not have O_DSYNC */
112 #define O_DSYNC O_SYNC
113 #elif defined(O_FSYNC)
114 #define O_DSYNC O_FSYNC
118 /* Approximate O_DIRECT with O_DSYNC if O_DIRECT isn't available */
120 #define O_DIRECT O_DSYNC
127 /* if the FD is not accessed during that time (in ns), we try to
128 reopen it to see if the disk has been changed */
129 #define FD_OPEN_TIMEOUT (1000000000)
131 #define MAX_BLOCKSIZE 4096
133 typedef struct BDRVRawState
{
139 #if defined(__linux__)
140 /* linux floppy specific */
141 int64_t fd_open_time
;
142 int64_t fd_error_time
;
144 int fd_media_changed
;
146 #ifdef CONFIG_LINUX_AIO
154 bool has_write_zeroes
:1;
155 bool discard_zeroes
:1;
157 bool needs_alignment
;
160 typedef struct BDRVRawReopenState
{
163 #ifdef CONFIG_LINUX_AIO
166 } BDRVRawReopenState
;
168 static int fd_open(BlockDriverState
*bs
);
169 static int64_t raw_getlength(BlockDriverState
*bs
);
171 typedef struct RawPosixAIOData
{
172 BlockDriverState
*bs
;
175 struct iovec
*aio_iov
;
180 #define aio_ioctl_cmd aio_nbytes /* for QEMU_AIO_IOCTL */
185 #if defined(__FreeBSD__) || defined(__FreeBSD_kernel__)
186 static int cdrom_reopen(BlockDriverState
*bs
);
189 #if defined(__NetBSD__)
190 static int raw_normalize_devicepath(const char **filename
)
192 static char namebuf
[PATH_MAX
];
193 const char *dp
, *fname
;
197 dp
= strrchr(fname
, '/');
198 if (lstat(fname
, &sb
) < 0) {
199 fprintf(stderr
, "%s: stat failed: %s\n",
200 fname
, strerror(errno
));
204 if (!S_ISBLK(sb
.st_mode
)) {
209 snprintf(namebuf
, PATH_MAX
, "r%s", fname
);
211 snprintf(namebuf
, PATH_MAX
, "%.*s/r%s",
212 (int)(dp
- fname
), fname
, dp
+ 1);
214 fprintf(stderr
, "%s is a block device", fname
);
216 fprintf(stderr
, ", using %s\n", *filename
);
221 static int raw_normalize_devicepath(const char **filename
)
228 * Get logical block size via ioctl. On success store it in @sector_size_p.
230 static int probe_logical_blocksize(int fd
, unsigned int *sector_size_p
)
232 unsigned int sector_size
;
233 bool success
= false;
237 /* Try a few ioctls to get the right size */
239 if (ioctl(fd
, BLKSSZGET
, §or_size
) >= 0) {
240 *sector_size_p
= sector_size
;
244 #ifdef DKIOCGETBLOCKSIZE
245 if (ioctl(fd
, DKIOCGETBLOCKSIZE
, §or_size
) >= 0) {
246 *sector_size_p
= sector_size
;
250 #ifdef DIOCGSECTORSIZE
251 if (ioctl(fd
, DIOCGSECTORSIZE
, §or_size
) >= 0) {
252 *sector_size_p
= sector_size
;
257 return success
? 0 : -errno
;
261 * Get physical block size of @fd.
262 * On success, store it in @blk_size and return 0.
263 * On failure, return -errno.
265 static int probe_physical_blocksize(int fd
, unsigned int *blk_size
)
268 if (ioctl(fd
, BLKPBSZGET
, blk_size
) < 0) {
277 /* Check if read is allowed with given memory buffer and length.
279 * This function is used to check O_DIRECT memory buffer and request alignment.
281 static bool raw_is_io_aligned(int fd
, void *buf
, size_t len
)
283 ssize_t ret
= pread(fd
, buf
, len
, 0);
290 /* The Linux kernel returns EINVAL for misaligned O_DIRECT reads. Ignore
291 * other errors (e.g. real I/O error), which could happen on a failed
292 * drive, since we only care about probing alignment.
294 if (errno
!= EINVAL
) {
302 static void raw_probe_alignment(BlockDriverState
*bs
, int fd
, Error
**errp
)
304 BDRVRawState
*s
= bs
->opaque
;
306 size_t max_align
= MAX(MAX_BLOCKSIZE
, getpagesize());
308 /* For /dev/sg devices the alignment is not really used.
309 With buffered I/O, we don't have any restrictions. */
310 if (bs
->sg
|| !s
->needs_alignment
) {
311 bs
->request_alignment
= 1;
316 bs
->request_alignment
= 0;
318 /* Let's try to use the logical blocksize for the alignment. */
319 if (probe_logical_blocksize(fd
, &bs
->request_alignment
) < 0) {
320 bs
->request_alignment
= 0;
325 if (xfsctl(NULL
, fd
, XFS_IOC_DIOINFO
, &da
) >= 0) {
326 bs
->request_alignment
= da
.d_miniosz
;
327 /* The kernel returns wrong information for d_mem */
328 /* s->buf_align = da.d_mem; */
333 /* If we could not get the sizes so far, we can only guess them */
336 buf
= qemu_memalign(max_align
, 2 * max_align
);
337 for (align
= 512; align
<= max_align
; align
<<= 1) {
338 if (raw_is_io_aligned(fd
, buf
+ align
, max_align
)) {
339 s
->buf_align
= align
;
346 if (!bs
->request_alignment
) {
348 buf
= qemu_memalign(s
->buf_align
, max_align
);
349 for (align
= 512; align
<= max_align
; align
<<= 1) {
350 if (raw_is_io_aligned(fd
, buf
, align
)) {
351 bs
->request_alignment
= align
;
358 if (!s
->buf_align
|| !bs
->request_alignment
) {
359 error_setg(errp
, "Could not find working O_DIRECT alignment. "
360 "Try cache.direct=off.");
364 static void raw_parse_flags(int bdrv_flags
, int *open_flags
)
366 assert(open_flags
!= NULL
);
368 *open_flags
|= O_BINARY
;
369 *open_flags
&= ~O_ACCMODE
;
370 if (bdrv_flags
& BDRV_O_RDWR
) {
371 *open_flags
|= O_RDWR
;
373 *open_flags
|= O_RDONLY
;
376 /* Use O_DSYNC for write-through caching, no flags for write-back caching,
377 * and O_DIRECT for no caching. */
378 if ((bdrv_flags
& BDRV_O_NOCACHE
)) {
379 *open_flags
|= O_DIRECT
;
383 static void raw_detach_aio_context(BlockDriverState
*bs
)
385 #ifdef CONFIG_LINUX_AIO
386 BDRVRawState
*s
= bs
->opaque
;
389 laio_detach_aio_context(s
->aio_ctx
, bdrv_get_aio_context(bs
));
394 static void raw_attach_aio_context(BlockDriverState
*bs
,
395 AioContext
*new_context
)
397 #ifdef CONFIG_LINUX_AIO
398 BDRVRawState
*s
= bs
->opaque
;
401 laio_attach_aio_context(s
->aio_ctx
, new_context
);
406 #ifdef CONFIG_LINUX_AIO
407 static int raw_set_aio(void **aio_ctx
, int *use_aio
, int bdrv_flags
)
410 assert(aio_ctx
!= NULL
);
411 assert(use_aio
!= NULL
);
413 * Currently Linux do AIO only for files opened with O_DIRECT
414 * specified so check NOCACHE flag too
416 if ((bdrv_flags
& (BDRV_O_NOCACHE
|BDRV_O_NATIVE_AIO
)) ==
417 (BDRV_O_NOCACHE
|BDRV_O_NATIVE_AIO
)) {
419 /* if non-NULL, laio_init() has already been run */
420 if (*aio_ctx
== NULL
) {
421 *aio_ctx
= laio_init();
438 static void raw_parse_filename(const char *filename
, QDict
*options
,
441 /* The filename does not have to be prefixed by the protocol name, since
442 * "file" is the default protocol; therefore, the return value of this
443 * function call can be ignored. */
444 strstart(filename
, "file:", &filename
);
446 qdict_put_obj(options
, "filename", QOBJECT(qstring_from_str(filename
)));
449 static QemuOptsList raw_runtime_opts
= {
451 .head
= QTAILQ_HEAD_INITIALIZER(raw_runtime_opts
.head
),
455 .type
= QEMU_OPT_STRING
,
456 .help
= "File name of the image",
458 { /* end of list */ }
462 static int raw_open_common(BlockDriverState
*bs
, QDict
*options
,
463 int bdrv_flags
, int open_flags
, Error
**errp
)
465 BDRVRawState
*s
= bs
->opaque
;
467 Error
*local_err
= NULL
;
468 const char *filename
= NULL
;
472 opts
= qemu_opts_create(&raw_runtime_opts
, NULL
, 0, &error_abort
);
473 qemu_opts_absorb_qdict(opts
, options
, &local_err
);
475 error_propagate(errp
, local_err
);
480 filename
= qemu_opt_get(opts
, "filename");
482 ret
= raw_normalize_devicepath(&filename
);
484 error_setg_errno(errp
, -ret
, "Could not normalize device path");
488 s
->open_flags
= open_flags
;
489 raw_parse_flags(bdrv_flags
, &s
->open_flags
);
492 fd
= qemu_open(filename
, s
->open_flags
, 0644);
502 #ifdef CONFIG_LINUX_AIO
503 if (raw_set_aio(&s
->aio_ctx
, &s
->use_aio
, bdrv_flags
)) {
506 error_setg_errno(errp
, -ret
, "Could not set AIO state");
509 if (!s
->use_aio
&& (bdrv_flags
& BDRV_O_NATIVE_AIO
)) {
510 error_printf("WARNING: aio=native was specified for '%s', but "
511 "it requires cache.direct=on, which was not "
512 "specified. Falling back to aio=threads.\n"
513 " This will become an error condition in "
514 "future QEMU versions.\n",
519 s
->has_discard
= true;
520 s
->has_write_zeroes
= true;
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_FD
|| 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 assert(!(raw_s
->open_flags
& O_CREAT
));
669 raw_s
->fd
= qemu_open(state
->bs
->filename
, raw_s
->open_flags
);
670 if (raw_s
->fd
== -1) {
671 error_setg_errno(errp
, errno
, "Could not reopen file");
676 /* Fail already reopen_prepare() if we can't get a working O_DIRECT
677 * alignment with the new fd. */
678 if (raw_s
->fd
!= -1) {
679 raw_probe_alignment(state
->bs
, raw_s
->fd
, &local_err
);
681 qemu_close(raw_s
->fd
);
683 error_propagate(errp
, local_err
);
691 static void raw_reopen_commit(BDRVReopenState
*state
)
693 BDRVRawReopenState
*raw_s
= state
->opaque
;
694 BDRVRawState
*s
= state
->bs
->opaque
;
696 s
->open_flags
= raw_s
->open_flags
;
700 #ifdef CONFIG_LINUX_AIO
701 s
->use_aio
= raw_s
->use_aio
;
704 g_free(state
->opaque
);
705 state
->opaque
= NULL
;
709 static void raw_reopen_abort(BDRVReopenState
*state
)
711 BDRVRawReopenState
*raw_s
= state
->opaque
;
713 /* nothing to do if NULL, we didn't get far enough */
718 if (raw_s
->fd
>= 0) {
719 qemu_close(raw_s
->fd
);
722 g_free(state
->opaque
);
723 state
->opaque
= NULL
;
726 static void raw_refresh_limits(BlockDriverState
*bs
, Error
**errp
)
728 BDRVRawState
*s
= bs
->opaque
;
730 raw_probe_alignment(bs
, s
->fd
, errp
);
731 bs
->bl
.min_mem_alignment
= s
->buf_align
;
732 bs
->bl
.opt_mem_alignment
= MAX(s
->buf_align
, getpagesize());
735 static int check_for_dasd(int fd
)
738 struct dasd_information2_t info
= {0};
740 return ioctl(fd
, BIODASDINFO2
, &info
);
747 * Try to get @bs's logical and physical block size.
748 * On success, store them in @bsz and return zero.
749 * On failure, return negative errno.
751 static int hdev_probe_blocksizes(BlockDriverState
*bs
, BlockSizes
*bsz
)
753 BDRVRawState
*s
= bs
->opaque
;
756 /* If DASD, get blocksizes */
757 if (check_for_dasd(s
->fd
) < 0) {
760 ret
= probe_logical_blocksize(s
->fd
, &bsz
->log
);
764 return probe_physical_blocksize(s
->fd
, &bsz
->phys
);
768 * Try to get @bs's geometry: cyls, heads, sectors.
769 * On success, store them in @geo and return 0.
770 * On failure return -errno.
771 * (Allows block driver to assign default geometry values that guest sees)
774 static int hdev_probe_geometry(BlockDriverState
*bs
, HDGeometry
*geo
)
776 BDRVRawState
*s
= bs
->opaque
;
777 struct hd_geometry ioctl_geo
= {0};
780 /* If DASD, get its geometry */
781 if (check_for_dasd(s
->fd
) < 0) {
784 if (ioctl(s
->fd
, HDIO_GETGEO
, &ioctl_geo
) < 0) {
787 /* HDIO_GETGEO may return success even though geo contains zeros
788 (e.g. certain multipath setups) */
789 if (!ioctl_geo
.heads
|| !ioctl_geo
.sectors
|| !ioctl_geo
.cylinders
) {
792 /* Do not return a geometry for partition */
793 if (ioctl_geo
.start
!= 0) {
796 geo
->heads
= ioctl_geo
.heads
;
797 geo
->sectors
= ioctl_geo
.sectors
;
798 if (!probe_physical_blocksize(s
->fd
, &blksize
)) {
799 /* overwrite cyls: HDIO_GETGEO result is incorrect for big drives */
800 geo
->cylinders
= bdrv_nb_sectors(bs
) / (blksize
/ BDRV_SECTOR_SIZE
)
801 / (geo
->heads
* geo
->sectors
);
804 geo
->cylinders
= ioctl_geo
.cylinders
;
808 #else /* __linux__ */
809 static int hdev_probe_geometry(BlockDriverState
*bs
, HDGeometry
*geo
)
815 static ssize_t
handle_aiocb_ioctl(RawPosixAIOData
*aiocb
)
819 ret
= ioctl(aiocb
->aio_fildes
, aiocb
->aio_ioctl_cmd
, aiocb
->aio_ioctl_buf
);
827 static ssize_t
handle_aiocb_flush(RawPosixAIOData
*aiocb
)
831 ret
= qemu_fdatasync(aiocb
->aio_fildes
);
840 static bool preadv_present
= true;
843 qemu_preadv(int fd
, const struct iovec
*iov
, int nr_iov
, off_t offset
)
845 return preadv(fd
, iov
, nr_iov
, offset
);
849 qemu_pwritev(int fd
, const struct iovec
*iov
, int nr_iov
, off_t offset
)
851 return pwritev(fd
, iov
, nr_iov
, offset
);
856 static bool preadv_present
= false;
859 qemu_preadv(int fd
, const struct iovec
*iov
, int nr_iov
, off_t offset
)
865 qemu_pwritev(int fd
, const struct iovec
*iov
, int nr_iov
, off_t offset
)
872 static ssize_t
handle_aiocb_rw_vector(RawPosixAIOData
*aiocb
)
877 if (aiocb
->aio_type
& QEMU_AIO_WRITE
)
878 len
= qemu_pwritev(aiocb
->aio_fildes
,
883 len
= qemu_preadv(aiocb
->aio_fildes
,
887 } while (len
== -1 && errno
== EINTR
);
896 * Read/writes the data to/from a given linear buffer.
898 * Returns the number of bytes handles or -errno in case of an error. Short
899 * reads are only returned if the end of the file is reached.
901 static ssize_t
handle_aiocb_rw_linear(RawPosixAIOData
*aiocb
, char *buf
)
906 while (offset
< aiocb
->aio_nbytes
) {
907 if (aiocb
->aio_type
& QEMU_AIO_WRITE
) {
908 len
= pwrite(aiocb
->aio_fildes
,
909 (const char *)buf
+ offset
,
910 aiocb
->aio_nbytes
- offset
,
911 aiocb
->aio_offset
+ offset
);
913 len
= pread(aiocb
->aio_fildes
,
915 aiocb
->aio_nbytes
- offset
,
916 aiocb
->aio_offset
+ offset
);
918 if (len
== -1 && errno
== EINTR
) {
920 } else if (len
== -1 && errno
== EINVAL
&&
921 (aiocb
->bs
->open_flags
& BDRV_O_NOCACHE
) &&
922 !(aiocb
->aio_type
& QEMU_AIO_WRITE
) &&
924 /* O_DIRECT pread() may fail with EINVAL when offset is unaligned
925 * after a short read. Assume that O_DIRECT short reads only occur
926 * at EOF. Therefore this is a short read, not an I/O error.
929 } else if (len
== -1) {
932 } else if (len
== 0) {
941 static ssize_t
handle_aiocb_rw(RawPosixAIOData
*aiocb
)
946 if (!(aiocb
->aio_type
& QEMU_AIO_MISALIGNED
)) {
948 * If there is just a single buffer, and it is properly aligned
949 * we can just use plain pread/pwrite without any problems.
951 if (aiocb
->aio_niov
== 1) {
952 return handle_aiocb_rw_linear(aiocb
, aiocb
->aio_iov
->iov_base
);
955 * We have more than one iovec, and all are properly aligned.
957 * Try preadv/pwritev first and fall back to linearizing the
958 * buffer if it's not supported.
960 if (preadv_present
) {
961 nbytes
= handle_aiocb_rw_vector(aiocb
);
962 if (nbytes
== aiocb
->aio_nbytes
||
963 (nbytes
< 0 && nbytes
!= -ENOSYS
)) {
966 preadv_present
= false;
970 * XXX(hch): short read/write. no easy way to handle the reminder
971 * using these interfaces. For now retry using plain
977 * Ok, we have to do it the hard way, copy all segments into
978 * a single aligned buffer.
980 buf
= qemu_try_blockalign(aiocb
->bs
, aiocb
->aio_nbytes
);
985 if (aiocb
->aio_type
& QEMU_AIO_WRITE
) {
989 for (i
= 0; i
< aiocb
->aio_niov
; ++i
) {
990 memcpy(p
, aiocb
->aio_iov
[i
].iov_base
, aiocb
->aio_iov
[i
].iov_len
);
991 p
+= aiocb
->aio_iov
[i
].iov_len
;
993 assert(p
- buf
== aiocb
->aio_nbytes
);
996 nbytes
= handle_aiocb_rw_linear(aiocb
, buf
);
997 if (!(aiocb
->aio_type
& QEMU_AIO_WRITE
)) {
999 size_t count
= aiocb
->aio_nbytes
, copy
;
1002 for (i
= 0; i
< aiocb
->aio_niov
&& count
; ++i
) {
1004 if (copy
> aiocb
->aio_iov
[i
].iov_len
) {
1005 copy
= aiocb
->aio_iov
[i
].iov_len
;
1007 memcpy(aiocb
->aio_iov
[i
].iov_base
, p
, copy
);
1008 assert(count
>= copy
);
1020 static int xfs_write_zeroes(BDRVRawState
*s
, int64_t offset
, uint64_t bytes
)
1022 struct xfs_flock64 fl
;
1024 memset(&fl
, 0, sizeof(fl
));
1025 fl
.l_whence
= SEEK_SET
;
1026 fl
.l_start
= offset
;
1029 if (xfsctl(NULL
, s
->fd
, XFS_IOC_ZERO_RANGE
, &fl
) < 0) {
1030 DEBUG_BLOCK_PRINT("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
;
1041 memset(&fl
, 0, sizeof(fl
));
1042 fl
.l_whence
= SEEK_SET
;
1043 fl
.l_start
= offset
;
1046 if (xfsctl(NULL
, s
->fd
, XFS_IOC_UNRESVSP64
, &fl
) < 0) {
1047 DEBUG_BLOCK_PRINT("cannot punch hole (%s)\n", strerror(errno
));
1055 static int translate_err(int err
)
1057 if (err
== -ENODEV
|| err
== -ENOSYS
|| err
== -EOPNOTSUPP
||
1064 #ifdef CONFIG_FALLOCATE
1065 static int do_fallocate(int fd
, int mode
, off_t offset
, off_t len
)
1068 if (fallocate(fd
, mode
, offset
, len
) == 0) {
1071 } while (errno
== EINTR
);
1072 return translate_err(-errno
);
1076 static ssize_t
handle_aiocb_write_zeroes_block(RawPosixAIOData
*aiocb
)
1079 BDRVRawState
*s
= aiocb
->bs
->opaque
;
1081 if (!s
->has_write_zeroes
) {
1087 uint64_t range
[2] = { aiocb
->aio_offset
, aiocb
->aio_nbytes
};
1088 if (ioctl(aiocb
->aio_fildes
, BLKZEROOUT
, range
) == 0) {
1091 } while (errno
== EINTR
);
1093 ret
= translate_err(-errno
);
1096 if (ret
== -ENOTSUP
) {
1097 s
->has_write_zeroes
= false;
1102 static ssize_t
handle_aiocb_write_zeroes(RawPosixAIOData
*aiocb
)
1104 #if defined(CONFIG_FALLOCATE) || defined(CONFIG_XFS)
1105 BDRVRawState
*s
= aiocb
->bs
->opaque
;
1108 if (aiocb
->aio_type
& QEMU_AIO_BLKDEV
) {
1109 return handle_aiocb_write_zeroes_block(aiocb
);
1114 return xfs_write_zeroes(s
, aiocb
->aio_offset
, aiocb
->aio_nbytes
);
1118 #ifdef CONFIG_FALLOCATE_ZERO_RANGE
1119 if (s
->has_write_zeroes
) {
1120 int ret
= do_fallocate(s
->fd
, FALLOC_FL_ZERO_RANGE
,
1121 aiocb
->aio_offset
, aiocb
->aio_nbytes
);
1122 if (ret
== 0 || ret
!= -ENOTSUP
) {
1125 s
->has_write_zeroes
= false;
1129 #ifdef CONFIG_FALLOCATE_PUNCH_HOLE
1130 if (s
->has_discard
&& s
->has_fallocate
) {
1131 int ret
= do_fallocate(s
->fd
,
1132 FALLOC_FL_PUNCH_HOLE
| FALLOC_FL_KEEP_SIZE
,
1133 aiocb
->aio_offset
, aiocb
->aio_nbytes
);
1135 ret
= do_fallocate(s
->fd
, 0, aiocb
->aio_offset
, aiocb
->aio_nbytes
);
1136 if (ret
== 0 || ret
!= -ENOTSUP
) {
1139 s
->has_fallocate
= false;
1140 } else if (ret
!= -ENOTSUP
) {
1143 s
->has_discard
= false;
1148 #ifdef CONFIG_FALLOCATE
1149 if (s
->has_fallocate
&& aiocb
->aio_offset
>= bdrv_getlength(aiocb
->bs
)) {
1150 int ret
= do_fallocate(s
->fd
, 0, aiocb
->aio_offset
, aiocb
->aio_nbytes
);
1151 if (ret
== 0 || ret
!= -ENOTSUP
) {
1154 s
->has_fallocate
= false;
1161 static ssize_t
handle_aiocb_discard(RawPosixAIOData
*aiocb
)
1163 int ret
= -EOPNOTSUPP
;
1164 BDRVRawState
*s
= aiocb
->bs
->opaque
;
1166 if (!s
->has_discard
) {
1170 if (aiocb
->aio_type
& QEMU_AIO_BLKDEV
) {
1173 uint64_t range
[2] = { aiocb
->aio_offset
, aiocb
->aio_nbytes
};
1174 if (ioctl(aiocb
->aio_fildes
, BLKDISCARD
, range
) == 0) {
1177 } while (errno
== EINTR
);
1184 return xfs_discard(s
, aiocb
->aio_offset
, aiocb
->aio_nbytes
);
1188 #ifdef CONFIG_FALLOCATE_PUNCH_HOLE
1189 ret
= do_fallocate(s
->fd
, FALLOC_FL_PUNCH_HOLE
| FALLOC_FL_KEEP_SIZE
,
1190 aiocb
->aio_offset
, aiocb
->aio_nbytes
);
1194 ret
= translate_err(ret
);
1195 if (ret
== -ENOTSUP
) {
1196 s
->has_discard
= false;
1201 static int aio_worker(void *arg
)
1203 RawPosixAIOData
*aiocb
= arg
;
1206 switch (aiocb
->aio_type
& QEMU_AIO_TYPE_MASK
) {
1208 ret
= handle_aiocb_rw(aiocb
);
1209 if (ret
>= 0 && ret
< aiocb
->aio_nbytes
) {
1210 iov_memset(aiocb
->aio_iov
, aiocb
->aio_niov
, ret
,
1211 0, aiocb
->aio_nbytes
- ret
);
1213 ret
= aiocb
->aio_nbytes
;
1215 if (ret
== aiocb
->aio_nbytes
) {
1217 } else if (ret
>= 0 && ret
< aiocb
->aio_nbytes
) {
1221 case QEMU_AIO_WRITE
:
1222 ret
= handle_aiocb_rw(aiocb
);
1223 if (ret
== aiocb
->aio_nbytes
) {
1225 } else if (ret
>= 0 && ret
< aiocb
->aio_nbytes
) {
1229 case QEMU_AIO_FLUSH
:
1230 ret
= handle_aiocb_flush(aiocb
);
1232 case QEMU_AIO_IOCTL
:
1233 ret
= handle_aiocb_ioctl(aiocb
);
1235 case QEMU_AIO_DISCARD
:
1236 ret
= handle_aiocb_discard(aiocb
);
1238 case QEMU_AIO_WRITE_ZEROES
:
1239 ret
= handle_aiocb_write_zeroes(aiocb
);
1242 fprintf(stderr
, "invalid aio request (0x%x)\n", aiocb
->aio_type
);
1247 g_slice_free(RawPosixAIOData
, aiocb
);
1251 static int paio_submit_co(BlockDriverState
*bs
, int fd
,
1252 int64_t sector_num
, QEMUIOVector
*qiov
, int nb_sectors
,
1255 RawPosixAIOData
*acb
= g_slice_new(RawPosixAIOData
);
1259 acb
->aio_type
= type
;
1260 acb
->aio_fildes
= fd
;
1262 acb
->aio_nbytes
= nb_sectors
* BDRV_SECTOR_SIZE
;
1263 acb
->aio_offset
= sector_num
* BDRV_SECTOR_SIZE
;
1266 acb
->aio_iov
= qiov
->iov
;
1267 acb
->aio_niov
= qiov
->niov
;
1268 assert(qiov
->size
== acb
->aio_nbytes
);
1271 trace_paio_submit_co(sector_num
, nb_sectors
, type
);
1272 pool
= aio_get_thread_pool(bdrv_get_aio_context(bs
));
1273 return thread_pool_submit_co(pool
, aio_worker
, acb
);
1276 static BlockAIOCB
*paio_submit(BlockDriverState
*bs
, int fd
,
1277 int64_t sector_num
, QEMUIOVector
*qiov
, int nb_sectors
,
1278 BlockCompletionFunc
*cb
, void *opaque
, int type
)
1280 RawPosixAIOData
*acb
= g_slice_new(RawPosixAIOData
);
1284 acb
->aio_type
= type
;
1285 acb
->aio_fildes
= fd
;
1287 acb
->aio_nbytes
= nb_sectors
* BDRV_SECTOR_SIZE
;
1288 acb
->aio_offset
= sector_num
* BDRV_SECTOR_SIZE
;
1291 acb
->aio_iov
= qiov
->iov
;
1292 acb
->aio_niov
= qiov
->niov
;
1293 assert(qiov
->size
== acb
->aio_nbytes
);
1296 trace_paio_submit(acb
, opaque
, sector_num
, nb_sectors
, type
);
1297 pool
= aio_get_thread_pool(bdrv_get_aio_context(bs
));
1298 return thread_pool_submit_aio(pool
, aio_worker
, acb
, cb
, opaque
);
1301 static BlockAIOCB
*raw_aio_submit(BlockDriverState
*bs
,
1302 int64_t sector_num
, QEMUIOVector
*qiov
, int nb_sectors
,
1303 BlockCompletionFunc
*cb
, void *opaque
, int type
)
1305 BDRVRawState
*s
= bs
->opaque
;
1307 if (fd_open(bs
) < 0)
1311 * Check if the underlying device requires requests to be aligned,
1312 * and if the request we are trying to submit is aligned or not.
1313 * If this is the case tell the low-level driver that it needs
1314 * to copy the buffer.
1316 if (s
->needs_alignment
) {
1317 if (!bdrv_qiov_is_aligned(bs
, qiov
)) {
1318 type
|= QEMU_AIO_MISALIGNED
;
1319 #ifdef CONFIG_LINUX_AIO
1320 } else if (s
->use_aio
) {
1321 return laio_submit(bs
, s
->aio_ctx
, s
->fd
, sector_num
, qiov
,
1322 nb_sectors
, cb
, opaque
, type
);
1327 return paio_submit(bs
, s
->fd
, sector_num
, qiov
, nb_sectors
,
1331 static void raw_aio_plug(BlockDriverState
*bs
)
1333 #ifdef CONFIG_LINUX_AIO
1334 BDRVRawState
*s
= bs
->opaque
;
1336 laio_io_plug(bs
, s
->aio_ctx
);
1341 static void raw_aio_unplug(BlockDriverState
*bs
)
1343 #ifdef CONFIG_LINUX_AIO
1344 BDRVRawState
*s
= bs
->opaque
;
1346 laio_io_unplug(bs
, s
->aio_ctx
, true);
1351 static void raw_aio_flush_io_queue(BlockDriverState
*bs
)
1353 #ifdef CONFIG_LINUX_AIO
1354 BDRVRawState
*s
= bs
->opaque
;
1356 laio_io_unplug(bs
, s
->aio_ctx
, false);
1361 static BlockAIOCB
*raw_aio_readv(BlockDriverState
*bs
,
1362 int64_t sector_num
, QEMUIOVector
*qiov
, int nb_sectors
,
1363 BlockCompletionFunc
*cb
, void *opaque
)
1365 return raw_aio_submit(bs
, sector_num
, qiov
, nb_sectors
,
1366 cb
, opaque
, QEMU_AIO_READ
);
1369 static BlockAIOCB
*raw_aio_writev(BlockDriverState
*bs
,
1370 int64_t sector_num
, QEMUIOVector
*qiov
, int nb_sectors
,
1371 BlockCompletionFunc
*cb
, void *opaque
)
1373 return raw_aio_submit(bs
, sector_num
, qiov
, nb_sectors
,
1374 cb
, opaque
, QEMU_AIO_WRITE
);
1377 static BlockAIOCB
*raw_aio_flush(BlockDriverState
*bs
,
1378 BlockCompletionFunc
*cb
, void *opaque
)
1380 BDRVRawState
*s
= bs
->opaque
;
1382 if (fd_open(bs
) < 0)
1385 return paio_submit(bs
, s
->fd
, 0, NULL
, 0, cb
, opaque
, QEMU_AIO_FLUSH
);
1388 static void raw_close(BlockDriverState
*bs
)
1390 BDRVRawState
*s
= bs
->opaque
;
1392 raw_detach_aio_context(bs
);
1394 #ifdef CONFIG_LINUX_AIO
1396 laio_cleanup(s
->aio_ctx
);
1405 static int raw_truncate(BlockDriverState
*bs
, int64_t offset
)
1407 BDRVRawState
*s
= bs
->opaque
;
1410 if (fstat(s
->fd
, &st
)) {
1414 if (S_ISREG(st
.st_mode
)) {
1415 if (ftruncate(s
->fd
, offset
) < 0) {
1418 } else if (S_ISCHR(st
.st_mode
) || S_ISBLK(st
.st_mode
)) {
1419 if (offset
> raw_getlength(bs
)) {
1430 static int64_t raw_getlength(BlockDriverState
*bs
)
1432 BDRVRawState
*s
= bs
->opaque
;
1438 if (S_ISCHR(st
.st_mode
) || S_ISBLK(st
.st_mode
)) {
1439 struct disklabel dl
;
1441 if (ioctl(fd
, DIOCGDINFO
, &dl
))
1443 return (uint64_t)dl
.d_secsize
*
1444 dl
.d_partitions
[DISKPART(st
.st_rdev
)].p_size
;
1448 #elif defined(__NetBSD__)
1449 static int64_t raw_getlength(BlockDriverState
*bs
)
1451 BDRVRawState
*s
= bs
->opaque
;
1457 if (S_ISCHR(st
.st_mode
) || S_ISBLK(st
.st_mode
)) {
1458 struct dkwedge_info dkw
;
1460 if (ioctl(fd
, DIOCGWEDGEINFO
, &dkw
) != -1) {
1461 return dkw
.dkw_size
* 512;
1463 struct disklabel dl
;
1465 if (ioctl(fd
, DIOCGDINFO
, &dl
))
1467 return (uint64_t)dl
.d_secsize
*
1468 dl
.d_partitions
[DISKPART(st
.st_rdev
)].p_size
;
1473 #elif defined(__sun__)
1474 static int64_t raw_getlength(BlockDriverState
*bs
)
1476 BDRVRawState
*s
= bs
->opaque
;
1477 struct dk_minfo minfo
;
1487 * Use the DKIOCGMEDIAINFO ioctl to read the size.
1489 ret
= ioctl(s
->fd
, DKIOCGMEDIAINFO
, &minfo
);
1491 return minfo
.dki_lbsize
* minfo
.dki_capacity
;
1495 * There are reports that lseek on some devices fails, but
1496 * irc discussion said that contingency on contingency was overkill.
1498 size
= lseek(s
->fd
, 0, SEEK_END
);
1504 #elif defined(CONFIG_BSD)
1505 static int64_t raw_getlength(BlockDriverState
*bs
)
1507 BDRVRawState
*s
= bs
->opaque
;
1511 #if defined (__FreeBSD__) || defined(__FreeBSD_kernel__)
1520 #if defined (__FreeBSD__) || defined(__FreeBSD_kernel__)
1523 if (!fstat(fd
, &sb
) && (S_IFCHR
& sb
.st_mode
)) {
1524 #ifdef DIOCGMEDIASIZE
1525 if (ioctl(fd
, DIOCGMEDIASIZE
, (off_t
*)&size
))
1526 #elif defined(DIOCGPART)
1529 if (ioctl(fd
, DIOCGPART
, &pi
) == 0)
1530 size
= pi
.media_size
;
1536 #if defined(__APPLE__) && defined(__MACH__)
1538 uint64_t sectors
= 0;
1539 uint32_t sector_size
= 0;
1541 if (ioctl(fd
, DKIOCGETBLOCKCOUNT
, §ors
) == 0
1542 && ioctl(fd
, DKIOCGETBLOCKSIZE
, §or_size
) == 0) {
1543 size
= sectors
* sector_size
;
1545 size
= lseek(fd
, 0LL, SEEK_END
);
1552 size
= lseek(fd
, 0LL, SEEK_END
);
1557 #if defined(__FreeBSD__) || defined(__FreeBSD_kernel__)
1560 /* XXX FreeBSD acd returns UINT_MAX sectors for an empty drive */
1561 if (size
== 2048LL * (unsigned)-1)
1563 /* XXX no disc? maybe we need to reopen... */
1564 if (size
<= 0 && !reopened
&& cdrom_reopen(bs
) >= 0) {
1571 size
= lseek(fd
, 0, SEEK_END
);
1579 static int64_t raw_getlength(BlockDriverState
*bs
)
1581 BDRVRawState
*s
= bs
->opaque
;
1590 size
= lseek(s
->fd
, 0, SEEK_END
);
1598 static int64_t raw_get_allocated_file_size(BlockDriverState
*bs
)
1601 BDRVRawState
*s
= bs
->opaque
;
1603 if (fstat(s
->fd
, &st
) < 0) {
1606 return (int64_t)st
.st_blocks
* 512;
1609 static int raw_create(const char *filename
, QemuOpts
*opts
, Error
**errp
)
1613 int64_t total_size
= 0;
1615 PreallocMode prealloc
;
1617 Error
*local_err
= NULL
;
1619 strstart(filename
, "file:", &filename
);
1621 /* Read out options */
1622 total_size
= ROUND_UP(qemu_opt_get_size_del(opts
, BLOCK_OPT_SIZE
, 0),
1624 nocow
= qemu_opt_get_bool(opts
, BLOCK_OPT_NOCOW
, false);
1625 buf
= qemu_opt_get_del(opts
, BLOCK_OPT_PREALLOC
);
1626 prealloc
= qapi_enum_parse(PreallocMode_lookup
, buf
,
1627 PREALLOC_MODE_MAX
, PREALLOC_MODE_OFF
,
1631 error_propagate(errp
, local_err
);
1636 fd
= qemu_open(filename
, O_WRONLY
| O_CREAT
| O_TRUNC
| O_BINARY
,
1640 error_setg_errno(errp
, -result
, "Could not create file");
1646 /* Set NOCOW flag to solve performance issue on fs like btrfs.
1647 * This is an optimisation. The FS_IOC_SETFLAGS ioctl return value
1648 * will be ignored since any failure of this operation should not
1649 * block the left work.
1652 if (ioctl(fd
, FS_IOC_GETFLAGS
, &attr
) == 0) {
1653 attr
|= FS_NOCOW_FL
;
1654 ioctl(fd
, FS_IOC_SETFLAGS
, &attr
);
1659 if (ftruncate(fd
, total_size
) != 0) {
1661 error_setg_errno(errp
, -result
, "Could not resize file");
1666 #ifdef CONFIG_POSIX_FALLOCATE
1667 case PREALLOC_MODE_FALLOC
:
1668 /* posix_fallocate() doesn't set errno. */
1669 result
= -posix_fallocate(fd
, 0, total_size
);
1671 error_setg_errno(errp
, -result
,
1672 "Could not preallocate data for the new file");
1676 case PREALLOC_MODE_FULL
:
1678 int64_t num
= 0, left
= total_size
;
1679 buf
= g_malloc0(65536);
1682 num
= MIN(left
, 65536);
1683 result
= write(fd
, buf
, num
);
1686 error_setg_errno(errp
, -result
,
1687 "Could not write to the new file");
1696 error_setg_errno(errp
, -result
,
1697 "Could not flush new file to disk");
1703 case PREALLOC_MODE_OFF
:
1707 error_setg(errp
, "Unsupported preallocation mode: %s",
1708 PreallocMode_lookup
[prealloc
]);
1713 if (qemu_close(fd
) != 0 && result
== 0) {
1715 error_setg_errno(errp
, -result
, "Could not close the new file");
1722 * Find allocation range in @bs around offset @start.
1723 * May change underlying file descriptor's file offset.
1724 * If @start is not in a hole, store @start in @data, and the
1725 * beginning of the next hole in @hole, and return 0.
1726 * If @start is in a non-trailing hole, store @start in @hole and the
1727 * beginning of the next non-hole in @data, and return 0.
1728 * If @start is in a trailing hole or beyond EOF, return -ENXIO.
1729 * If we can't find out, return a negative errno other than -ENXIO.
1731 static int find_allocation(BlockDriverState
*bs
, off_t start
,
1732 off_t
*data
, off_t
*hole
)
1734 #if defined SEEK_HOLE && defined SEEK_DATA
1735 BDRVRawState
*s
= bs
->opaque
;
1740 * D1. offs == start: start is in data
1741 * D2. offs > start: start is in a hole, next data at offs
1742 * D3. offs < 0, errno = ENXIO: either start is in a trailing hole
1743 * or start is beyond EOF
1744 * If the latter happens, the file has been truncated behind
1745 * our back since we opened it. All bets are off then.
1746 * Treating like a trailing hole is simplest.
1747 * D4. offs < 0, errno != ENXIO: we learned nothing
1749 offs
= lseek(s
->fd
, start
, SEEK_DATA
);
1751 return -errno
; /* D3 or D4 */
1753 assert(offs
>= start
);
1756 /* D2: in hole, next data at offs */
1762 /* D1: in data, end not yet known */
1766 * H1. offs == start: start is in a hole
1767 * If this happens here, a hole has been dug behind our back
1768 * since the previous lseek().
1769 * H2. offs > start: either start is in data, next hole at offs,
1770 * or start is in trailing hole, EOF at offs
1771 * Linux treats trailing holes like any other hole: offs ==
1772 * start. Solaris seeks to EOF instead: offs > start (blech).
1773 * If that happens here, a hole has been dug behind our back
1774 * since the previous lseek().
1775 * H3. offs < 0, errno = ENXIO: start is beyond EOF
1776 * If this happens, the file has been truncated behind our
1777 * back since we opened it. Treat it like a trailing hole.
1778 * H4. offs < 0, errno != ENXIO: we learned nothing
1779 * Pretend we know nothing at all, i.e. "forget" about D1.
1781 offs
= lseek(s
->fd
, start
, SEEK_HOLE
);
1783 return -errno
; /* D1 and (H3 or H4) */
1785 assert(offs
>= start
);
1789 * D1 and H2: either in data, next hole at offs, or it was in
1790 * data but is now in a trailing hole. In the latter case,
1791 * all bets are off. Treating it as if it there was data all
1792 * the way to EOF is safe, so simply do that.
1807 * Returns the allocation status of the specified sectors.
1809 * If 'sector_num' is beyond the end of the disk image the return value is 0
1810 * and 'pnum' is set to 0.
1812 * 'pnum' is set to the number of sectors (including and immediately following
1813 * the specified sector) that are known to be in the same
1814 * allocated/unallocated state.
1816 * 'nb_sectors' is the max value 'pnum' should be set to. If nb_sectors goes
1817 * beyond the end of the disk image it will be clamped.
1819 static int64_t coroutine_fn
raw_co_get_block_status(BlockDriverState
*bs
,
1821 int nb_sectors
, int *pnum
)
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
;
1863 return ret
| BDRV_BLOCK_OFFSET_VALID
| start
;
1866 static coroutine_fn BlockAIOCB
*raw_aio_discard(BlockDriverState
*bs
,
1867 int64_t sector_num
, int nb_sectors
,
1868 BlockCompletionFunc
*cb
, void *opaque
)
1870 BDRVRawState
*s
= bs
->opaque
;
1872 return paio_submit(bs
, s
->fd
, sector_num
, NULL
, nb_sectors
,
1873 cb
, opaque
, QEMU_AIO_DISCARD
);
1876 static int coroutine_fn
raw_co_write_zeroes(
1877 BlockDriverState
*bs
, int64_t sector_num
,
1878 int nb_sectors
, BdrvRequestFlags flags
)
1880 BDRVRawState
*s
= bs
->opaque
;
1882 if (!(flags
& BDRV_REQ_MAY_UNMAP
)) {
1883 return paio_submit_co(bs
, s
->fd
, sector_num
, NULL
, nb_sectors
,
1884 QEMU_AIO_WRITE_ZEROES
);
1885 } else if (s
->discard_zeroes
) {
1886 return paio_submit_co(bs
, s
->fd
, sector_num
, NULL
, nb_sectors
,
1892 static int raw_get_info(BlockDriverState
*bs
, BlockDriverInfo
*bdi
)
1894 BDRVRawState
*s
= bs
->opaque
;
1896 bdi
->unallocated_blocks_are_zero
= s
->discard_zeroes
;
1897 bdi
->can_write_zeroes_with_unmap
= s
->discard_zeroes
;
1901 static QemuOptsList raw_create_opts
= {
1902 .name
= "raw-create-opts",
1903 .head
= QTAILQ_HEAD_INITIALIZER(raw_create_opts
.head
),
1906 .name
= BLOCK_OPT_SIZE
,
1907 .type
= QEMU_OPT_SIZE
,
1908 .help
= "Virtual disk size"
1911 .name
= BLOCK_OPT_NOCOW
,
1912 .type
= QEMU_OPT_BOOL
,
1913 .help
= "Turn off copy-on-write (valid only on btrfs)"
1916 .name
= BLOCK_OPT_PREALLOC
,
1917 .type
= QEMU_OPT_STRING
,
1918 .help
= "Preallocation mode (allowed values: off, falloc, full)"
1920 { /* end of list */ }
1924 BlockDriver bdrv_file
= {
1925 .format_name
= "file",
1926 .protocol_name
= "file",
1927 .instance_size
= sizeof(BDRVRawState
),
1928 .bdrv_needs_filename
= true,
1929 .bdrv_probe
= NULL
, /* no probe for protocols */
1930 .bdrv_parse_filename
= raw_parse_filename
,
1931 .bdrv_file_open
= raw_open
,
1932 .bdrv_reopen_prepare
= raw_reopen_prepare
,
1933 .bdrv_reopen_commit
= raw_reopen_commit
,
1934 .bdrv_reopen_abort
= raw_reopen_abort
,
1935 .bdrv_close
= raw_close
,
1936 .bdrv_create
= raw_create
,
1937 .bdrv_has_zero_init
= bdrv_has_zero_init_1
,
1938 .bdrv_co_get_block_status
= raw_co_get_block_status
,
1939 .bdrv_co_write_zeroes
= raw_co_write_zeroes
,
1941 .bdrv_aio_readv
= raw_aio_readv
,
1942 .bdrv_aio_writev
= raw_aio_writev
,
1943 .bdrv_aio_flush
= raw_aio_flush
,
1944 .bdrv_aio_discard
= raw_aio_discard
,
1945 .bdrv_refresh_limits
= raw_refresh_limits
,
1946 .bdrv_io_plug
= raw_aio_plug
,
1947 .bdrv_io_unplug
= raw_aio_unplug
,
1948 .bdrv_flush_io_queue
= raw_aio_flush_io_queue
,
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 .bdrv_detach_aio_context
= raw_detach_aio_context
,
1957 .bdrv_attach_aio_context
= raw_attach_aio_context
,
1959 .create_opts
= &raw_create_opts
,
1962 /***********************************************/
1965 #if defined(__APPLE__) && defined(__MACH__)
1966 static kern_return_t
FindEjectableCDMedia( io_iterator_t
*mediaIterator
);
1967 static kern_return_t
GetBSDPath( io_iterator_t mediaIterator
, char *bsdPath
, CFIndex maxPathSize
);
1969 kern_return_t
FindEjectableCDMedia( io_iterator_t
*mediaIterator
)
1971 kern_return_t kernResult
;
1972 mach_port_t masterPort
;
1973 CFMutableDictionaryRef classesToMatch
;
1975 kernResult
= IOMasterPort( MACH_PORT_NULL
, &masterPort
);
1976 if ( KERN_SUCCESS
!= kernResult
) {
1977 printf( "IOMasterPort returned %d\n", kernResult
);
1980 classesToMatch
= IOServiceMatching( kIOCDMediaClass
);
1981 if ( classesToMatch
== NULL
) {
1982 printf( "IOServiceMatching returned a NULL dictionary.\n" );
1984 CFDictionarySetValue( classesToMatch
, CFSTR( kIOMediaEjectableKey
), kCFBooleanTrue
);
1986 kernResult
= IOServiceGetMatchingServices( masterPort
, classesToMatch
, mediaIterator
);
1987 if ( KERN_SUCCESS
!= kernResult
)
1989 printf( "IOServiceGetMatchingServices returned %d\n", kernResult
);
1995 kern_return_t
GetBSDPath( io_iterator_t mediaIterator
, char *bsdPath
, CFIndex maxPathSize
)
1997 io_object_t nextMedia
;
1998 kern_return_t kernResult
= KERN_FAILURE
;
2000 nextMedia
= IOIteratorNext( mediaIterator
);
2003 CFTypeRef bsdPathAsCFString
;
2004 bsdPathAsCFString
= IORegistryEntryCreateCFProperty( nextMedia
, CFSTR( kIOBSDNameKey
), kCFAllocatorDefault
, 0 );
2005 if ( bsdPathAsCFString
) {
2006 size_t devPathLength
;
2007 strcpy( bsdPath
, _PATH_DEV
);
2008 strcat( bsdPath
, "r" );
2009 devPathLength
= strlen( bsdPath
);
2010 if ( CFStringGetCString( bsdPathAsCFString
, bsdPath
+ devPathLength
, maxPathSize
- devPathLength
, kCFStringEncodingASCII
) ) {
2011 kernResult
= KERN_SUCCESS
;
2013 CFRelease( bsdPathAsCFString
);
2015 IOObjectRelease( nextMedia
);
2023 static int hdev_probe_device(const char *filename
)
2027 /* allow a dedicated CD-ROM driver to match with a higher priority */
2028 if (strstart(filename
, "/dev/cdrom", NULL
))
2031 if (stat(filename
, &st
) >= 0 &&
2032 (S_ISCHR(st
.st_mode
) || S_ISBLK(st
.st_mode
))) {
2039 static int check_hdev_writable(BDRVRawState
*s
)
2041 #if defined(BLKROGET)
2042 /* Linux block devices can be configured "read-only" using blockdev(8).
2043 * This is independent of device node permissions and therefore open(2)
2044 * with O_RDWR succeeds. Actual writes fail with EPERM.
2046 * bdrv_open() is supposed to fail if the disk is read-only. Explicitly
2047 * check for read-only block devices so that Linux block devices behave
2053 if (fstat(s
->fd
, &st
)) {
2057 if (!S_ISBLK(st
.st_mode
)) {
2061 if (ioctl(s
->fd
, BLKROGET
, &readonly
) < 0) {
2068 #endif /* defined(BLKROGET) */
2072 static void hdev_parse_filename(const char *filename
, QDict
*options
,
2075 /* The prefix is optional, just as for "file". */
2076 strstart(filename
, "host_device:", &filename
);
2078 qdict_put_obj(options
, "filename", QOBJECT(qstring_from_str(filename
)));
2081 static int hdev_open(BlockDriverState
*bs
, QDict
*options
, int flags
,
2084 BDRVRawState
*s
= bs
->opaque
;
2085 Error
*local_err
= NULL
;
2087 const char *filename
= qdict_get_str(options
, "filename");
2089 #if defined(__APPLE__) && defined(__MACH__)
2090 if (strstart(filename
, "/dev/cdrom", NULL
)) {
2091 kern_return_t kernResult
;
2092 io_iterator_t mediaIterator
;
2093 char bsdPath
[ MAXPATHLEN
];
2096 kernResult
= FindEjectableCDMedia( &mediaIterator
);
2097 kernResult
= GetBSDPath( mediaIterator
, bsdPath
, sizeof( bsdPath
) );
2099 if ( bsdPath
[ 0 ] != '\0' ) {
2100 strcat(bsdPath
,"s0");
2101 /* some CDs don't have a partition 0 */
2102 fd
= qemu_open(bsdPath
, O_RDONLY
| O_BINARY
| O_LARGEFILE
);
2104 bsdPath
[strlen(bsdPath
)-1] = '1';
2109 qdict_put(options
, "filename", qstring_from_str(filename
));
2112 if ( mediaIterator
)
2113 IOObjectRelease( mediaIterator
);
2117 s
->type
= FTYPE_FILE
;
2118 #if defined(__linux__)
2120 char resolved_path
[ MAXPATHLEN
], *temp
;
2122 temp
= realpath(filename
, resolved_path
);
2123 if (temp
&& strstart(temp
, "/dev/sg", NULL
)) {
2129 ret
= raw_open_common(bs
, options
, flags
, 0, &local_err
);
2132 error_propagate(errp
, local_err
);
2137 if (flags
& BDRV_O_RDWR
) {
2138 ret
= check_hdev_writable(s
);
2141 error_setg_errno(errp
, -ret
, "The device is not writable");
2149 #if defined(__linux__)
2150 /* Note: we do not have a reliable method to detect if the floppy is
2151 present. The current method is to try to open the floppy at every
2152 I/O and to keep it opened during a few hundreds of ms. */
2153 static int fd_open(BlockDriverState
*bs
)
2155 BDRVRawState
*s
= bs
->opaque
;
2156 int last_media_present
;
2158 if (s
->type
!= FTYPE_FD
)
2160 last_media_present
= (s
->fd
>= 0);
2162 (qemu_clock_get_ns(QEMU_CLOCK_REALTIME
) - s
->fd_open_time
) >= FD_OPEN_TIMEOUT
) {
2166 printf("Floppy closed\n");
2170 if (s
->fd_got_error
&&
2171 (qemu_clock_get_ns(QEMU_CLOCK_REALTIME
) - s
->fd_error_time
) < FD_OPEN_TIMEOUT
) {
2173 printf("No floppy (open delayed)\n");
2177 s
->fd
= qemu_open(bs
->filename
, s
->open_flags
& ~O_NONBLOCK
);
2179 s
->fd_error_time
= qemu_clock_get_ns(QEMU_CLOCK_REALTIME
);
2180 s
->fd_got_error
= 1;
2181 if (last_media_present
)
2182 s
->fd_media_changed
= 1;
2184 printf("No floppy\n");
2189 printf("Floppy opened\n");
2192 if (!last_media_present
)
2193 s
->fd_media_changed
= 1;
2194 s
->fd_open_time
= qemu_clock_get_ns(QEMU_CLOCK_REALTIME
);
2195 s
->fd_got_error
= 0;
2199 static int hdev_ioctl(BlockDriverState
*bs
, unsigned long int req
, void *buf
)
2201 BDRVRawState
*s
= bs
->opaque
;
2203 return ioctl(s
->fd
, req
, buf
);
2206 static BlockAIOCB
*hdev_aio_ioctl(BlockDriverState
*bs
,
2207 unsigned long int req
, void *buf
,
2208 BlockCompletionFunc
*cb
, void *opaque
)
2210 BDRVRawState
*s
= bs
->opaque
;
2211 RawPosixAIOData
*acb
;
2214 if (fd_open(bs
) < 0)
2217 acb
= g_slice_new(RawPosixAIOData
);
2219 acb
->aio_type
= QEMU_AIO_IOCTL
;
2220 acb
->aio_fildes
= s
->fd
;
2221 acb
->aio_offset
= 0;
2222 acb
->aio_ioctl_buf
= buf
;
2223 acb
->aio_ioctl_cmd
= req
;
2224 pool
= aio_get_thread_pool(bdrv_get_aio_context(bs
));
2225 return thread_pool_submit_aio(pool
, aio_worker
, acb
, cb
, opaque
);
2228 #elif defined(__FreeBSD__) || defined(__FreeBSD_kernel__)
2229 static int fd_open(BlockDriverState
*bs
)
2231 BDRVRawState
*s
= bs
->opaque
;
2233 /* this is just to ensure s->fd is sane (its called by io ops) */
2238 #else /* !linux && !FreeBSD */
2240 static int fd_open(BlockDriverState
*bs
)
2245 #endif /* !linux && !FreeBSD */
2247 static coroutine_fn BlockAIOCB
*hdev_aio_discard(BlockDriverState
*bs
,
2248 int64_t sector_num
, int nb_sectors
,
2249 BlockCompletionFunc
*cb
, void *opaque
)
2251 BDRVRawState
*s
= bs
->opaque
;
2253 if (fd_open(bs
) < 0) {
2256 return paio_submit(bs
, s
->fd
, sector_num
, NULL
, nb_sectors
,
2257 cb
, opaque
, QEMU_AIO_DISCARD
|QEMU_AIO_BLKDEV
);
2260 static coroutine_fn
int hdev_co_write_zeroes(BlockDriverState
*bs
,
2261 int64_t sector_num
, int nb_sectors
, BdrvRequestFlags flags
)
2263 BDRVRawState
*s
= bs
->opaque
;
2270 if (!(flags
& BDRV_REQ_MAY_UNMAP
)) {
2271 return paio_submit_co(bs
, s
->fd
, sector_num
, NULL
, nb_sectors
,
2272 QEMU_AIO_WRITE_ZEROES
|QEMU_AIO_BLKDEV
);
2273 } else if (s
->discard_zeroes
) {
2274 return paio_submit_co(bs
, s
->fd
, sector_num
, NULL
, nb_sectors
,
2275 QEMU_AIO_DISCARD
|QEMU_AIO_BLKDEV
);
2280 static int hdev_create(const char *filename
, QemuOpts
*opts
,
2285 struct stat stat_buf
;
2286 int64_t total_size
= 0;
2289 /* This function is used by all three protocol block drivers and therefore
2290 * any of these three prefixes may be given.
2291 * The return value has to be stored somewhere, otherwise this is an error
2292 * due to -Werror=unused-value. */
2294 strstart(filename
, "host_device:", &filename
) ||
2295 strstart(filename
, "host_cdrom:" , &filename
) ||
2296 strstart(filename
, "host_floppy:", &filename
);
2300 /* Read out options */
2301 total_size
= ROUND_UP(qemu_opt_get_size_del(opts
, BLOCK_OPT_SIZE
, 0),
2304 fd
= qemu_open(filename
, O_WRONLY
| O_BINARY
);
2307 error_setg_errno(errp
, -ret
, "Could not open device");
2311 if (fstat(fd
, &stat_buf
) < 0) {
2313 error_setg_errno(errp
, -ret
, "Could not stat device");
2314 } else if (!S_ISBLK(stat_buf
.st_mode
) && !S_ISCHR(stat_buf
.st_mode
)) {
2316 "The given file is neither a block nor a character device");
2318 } else if (lseek(fd
, 0, SEEK_END
) < total_size
) {
2319 error_setg(errp
, "Device is too small");
2327 static BlockDriver bdrv_host_device
= {
2328 .format_name
= "host_device",
2329 .protocol_name
= "host_device",
2330 .instance_size
= sizeof(BDRVRawState
),
2331 .bdrv_needs_filename
= true,
2332 .bdrv_probe_device
= hdev_probe_device
,
2333 .bdrv_parse_filename
= hdev_parse_filename
,
2334 .bdrv_file_open
= hdev_open
,
2335 .bdrv_close
= raw_close
,
2336 .bdrv_reopen_prepare
= raw_reopen_prepare
,
2337 .bdrv_reopen_commit
= raw_reopen_commit
,
2338 .bdrv_reopen_abort
= raw_reopen_abort
,
2339 .bdrv_create
= hdev_create
,
2340 .create_opts
= &raw_create_opts
,
2341 .bdrv_co_write_zeroes
= hdev_co_write_zeroes
,
2343 .bdrv_aio_readv
= raw_aio_readv
,
2344 .bdrv_aio_writev
= raw_aio_writev
,
2345 .bdrv_aio_flush
= raw_aio_flush
,
2346 .bdrv_aio_discard
= hdev_aio_discard
,
2347 .bdrv_refresh_limits
= raw_refresh_limits
,
2348 .bdrv_io_plug
= raw_aio_plug
,
2349 .bdrv_io_unplug
= raw_aio_unplug
,
2350 .bdrv_flush_io_queue
= raw_aio_flush_io_queue
,
2352 .bdrv_truncate
= raw_truncate
,
2353 .bdrv_getlength
= raw_getlength
,
2354 .bdrv_get_info
= raw_get_info
,
2355 .bdrv_get_allocated_file_size
2356 = raw_get_allocated_file_size
,
2357 .bdrv_probe_blocksizes
= hdev_probe_blocksizes
,
2358 .bdrv_probe_geometry
= hdev_probe_geometry
,
2360 .bdrv_detach_aio_context
= raw_detach_aio_context
,
2361 .bdrv_attach_aio_context
= raw_attach_aio_context
,
2363 /* generic scsi device */
2365 .bdrv_ioctl
= hdev_ioctl
,
2366 .bdrv_aio_ioctl
= hdev_aio_ioctl
,
2371 static void floppy_parse_filename(const char *filename
, QDict
*options
,
2374 /* The prefix is optional, just as for "file". */
2375 strstart(filename
, "host_floppy:", &filename
);
2377 qdict_put_obj(options
, "filename", QOBJECT(qstring_from_str(filename
)));
2380 static int floppy_open(BlockDriverState
*bs
, QDict
*options
, int flags
,
2383 BDRVRawState
*s
= bs
->opaque
;
2384 Error
*local_err
= NULL
;
2389 /* open will not fail even if no floppy is inserted, so add O_NONBLOCK */
2390 ret
= raw_open_common(bs
, options
, flags
, O_NONBLOCK
, &local_err
);
2393 error_propagate(errp
, local_err
);
2398 /* close fd so that we can reopen it as needed */
2401 s
->fd_media_changed
= 1;
2403 error_report("Host floppy pass-through is deprecated");
2404 error_printf("Support for it will be removed in a future release.\n");
2408 static int floppy_probe_device(const char *filename
)
2412 struct floppy_struct fdparam
;
2415 if (strstart(filename
, "/dev/fd", NULL
) &&
2416 !strstart(filename
, "/dev/fdset/", NULL
)) {
2420 fd
= qemu_open(filename
, O_RDONLY
| O_NONBLOCK
);
2424 ret
= fstat(fd
, &st
);
2425 if (ret
== -1 || !S_ISBLK(st
.st_mode
)) {
2429 /* Attempt to detect via a floppy specific ioctl */
2430 ret
= ioctl(fd
, FDGETPRM
, &fdparam
);
2441 static int floppy_is_inserted(BlockDriverState
*bs
)
2443 return fd_open(bs
) >= 0;
2446 static int floppy_media_changed(BlockDriverState
*bs
)
2448 BDRVRawState
*s
= bs
->opaque
;
2452 * XXX: we do not have a true media changed indication.
2453 * It does not work if the floppy is changed without trying to read it.
2456 ret
= s
->fd_media_changed
;
2457 s
->fd_media_changed
= 0;
2459 printf("Floppy changed=%d\n", ret
);
2464 static void floppy_eject(BlockDriverState
*bs
, bool eject_flag
)
2466 BDRVRawState
*s
= bs
->opaque
;
2473 fd
= qemu_open(bs
->filename
, s
->open_flags
| O_NONBLOCK
);
2475 if (ioctl(fd
, FDEJECT
, 0) < 0)
2481 static BlockDriver bdrv_host_floppy
= {
2482 .format_name
= "host_floppy",
2483 .protocol_name
= "host_floppy",
2484 .instance_size
= sizeof(BDRVRawState
),
2485 .bdrv_needs_filename
= true,
2486 .bdrv_probe_device
= floppy_probe_device
,
2487 .bdrv_parse_filename
= floppy_parse_filename
,
2488 .bdrv_file_open
= floppy_open
,
2489 .bdrv_close
= raw_close
,
2490 .bdrv_reopen_prepare
= raw_reopen_prepare
,
2491 .bdrv_reopen_commit
= raw_reopen_commit
,
2492 .bdrv_reopen_abort
= raw_reopen_abort
,
2493 .bdrv_create
= hdev_create
,
2494 .create_opts
= &raw_create_opts
,
2496 .bdrv_aio_readv
= raw_aio_readv
,
2497 .bdrv_aio_writev
= raw_aio_writev
,
2498 .bdrv_aio_flush
= raw_aio_flush
,
2499 .bdrv_refresh_limits
= raw_refresh_limits
,
2500 .bdrv_io_plug
= raw_aio_plug
,
2501 .bdrv_io_unplug
= raw_aio_unplug
,
2502 .bdrv_flush_io_queue
= raw_aio_flush_io_queue
,
2504 .bdrv_truncate
= raw_truncate
,
2505 .bdrv_getlength
= raw_getlength
,
2506 .has_variable_length
= true,
2507 .bdrv_get_allocated_file_size
2508 = raw_get_allocated_file_size
,
2510 .bdrv_detach_aio_context
= raw_detach_aio_context
,
2511 .bdrv_attach_aio_context
= raw_attach_aio_context
,
2513 /* removable device support */
2514 .bdrv_is_inserted
= floppy_is_inserted
,
2515 .bdrv_media_changed
= floppy_media_changed
,
2516 .bdrv_eject
= floppy_eject
,
2520 #if defined(__linux__) || defined(__FreeBSD__) || defined(__FreeBSD_kernel__)
2521 static void cdrom_parse_filename(const char *filename
, QDict
*options
,
2524 /* The prefix is optional, just as for "file". */
2525 strstart(filename
, "host_cdrom:", &filename
);
2527 qdict_put_obj(options
, "filename", QOBJECT(qstring_from_str(filename
)));
2532 static int cdrom_open(BlockDriverState
*bs
, QDict
*options
, int flags
,
2535 BDRVRawState
*s
= bs
->opaque
;
2536 Error
*local_err
= NULL
;
2541 /* open will not fail even if no CD is inserted, so add O_NONBLOCK */
2542 ret
= raw_open_common(bs
, options
, flags
, O_NONBLOCK
, &local_err
);
2544 error_propagate(errp
, local_err
);
2549 static int cdrom_probe_device(const char *filename
)
2555 fd
= qemu_open(filename
, O_RDONLY
| O_NONBLOCK
);
2559 ret
= fstat(fd
, &st
);
2560 if (ret
== -1 || !S_ISBLK(st
.st_mode
)) {
2564 /* Attempt to detect via a CDROM specific ioctl */
2565 ret
= ioctl(fd
, CDROM_DRIVE_STATUS
, CDSL_CURRENT
);
2575 static int cdrom_is_inserted(BlockDriverState
*bs
)
2577 BDRVRawState
*s
= bs
->opaque
;
2580 ret
= ioctl(s
->fd
, CDROM_DRIVE_STATUS
, CDSL_CURRENT
);
2581 if (ret
== CDS_DISC_OK
)
2586 static void cdrom_eject(BlockDriverState
*bs
, bool eject_flag
)
2588 BDRVRawState
*s
= bs
->opaque
;
2591 if (ioctl(s
->fd
, CDROMEJECT
, NULL
) < 0)
2592 perror("CDROMEJECT");
2594 if (ioctl(s
->fd
, CDROMCLOSETRAY
, NULL
) < 0)
2595 perror("CDROMEJECT");
2599 static void cdrom_lock_medium(BlockDriverState
*bs
, bool locked
)
2601 BDRVRawState
*s
= bs
->opaque
;
2603 if (ioctl(s
->fd
, CDROM_LOCKDOOR
, locked
) < 0) {
2605 * Note: an error can happen if the distribution automatically
2608 /* perror("CDROM_LOCKDOOR"); */
2612 static BlockDriver bdrv_host_cdrom
= {
2613 .format_name
= "host_cdrom",
2614 .protocol_name
= "host_cdrom",
2615 .instance_size
= sizeof(BDRVRawState
),
2616 .bdrv_needs_filename
= true,
2617 .bdrv_probe_device
= cdrom_probe_device
,
2618 .bdrv_parse_filename
= cdrom_parse_filename
,
2619 .bdrv_file_open
= cdrom_open
,
2620 .bdrv_close
= raw_close
,
2621 .bdrv_reopen_prepare
= raw_reopen_prepare
,
2622 .bdrv_reopen_commit
= raw_reopen_commit
,
2623 .bdrv_reopen_abort
= raw_reopen_abort
,
2624 .bdrv_create
= hdev_create
,
2625 .create_opts
= &raw_create_opts
,
2627 .bdrv_aio_readv
= raw_aio_readv
,
2628 .bdrv_aio_writev
= raw_aio_writev
,
2629 .bdrv_aio_flush
= raw_aio_flush
,
2630 .bdrv_refresh_limits
= raw_refresh_limits
,
2631 .bdrv_io_plug
= raw_aio_plug
,
2632 .bdrv_io_unplug
= raw_aio_unplug
,
2633 .bdrv_flush_io_queue
= raw_aio_flush_io_queue
,
2635 .bdrv_truncate
= raw_truncate
,
2636 .bdrv_getlength
= raw_getlength
,
2637 .has_variable_length
= true,
2638 .bdrv_get_allocated_file_size
2639 = raw_get_allocated_file_size
,
2641 .bdrv_detach_aio_context
= raw_detach_aio_context
,
2642 .bdrv_attach_aio_context
= raw_attach_aio_context
,
2644 /* removable device support */
2645 .bdrv_is_inserted
= cdrom_is_inserted
,
2646 .bdrv_eject
= cdrom_eject
,
2647 .bdrv_lock_medium
= cdrom_lock_medium
,
2649 /* generic scsi device */
2650 .bdrv_ioctl
= hdev_ioctl
,
2651 .bdrv_aio_ioctl
= hdev_aio_ioctl
,
2653 #endif /* __linux__ */
2655 #if defined (__FreeBSD__) || defined(__FreeBSD_kernel__)
2656 static int cdrom_open(BlockDriverState
*bs
, QDict
*options
, int flags
,
2659 BDRVRawState
*s
= bs
->opaque
;
2660 Error
*local_err
= NULL
;
2665 ret
= raw_open_common(bs
, options
, flags
, 0, &local_err
);
2668 error_propagate(errp
, local_err
);
2673 /* make sure the door isn't locked at this time */
2674 ioctl(s
->fd
, CDIOCALLOW
);
2678 static int cdrom_probe_device(const char *filename
)
2680 if (strstart(filename
, "/dev/cd", NULL
) ||
2681 strstart(filename
, "/dev/acd", NULL
))
2686 static int cdrom_reopen(BlockDriverState
*bs
)
2688 BDRVRawState
*s
= bs
->opaque
;
2692 * Force reread of possibly changed/newly loaded disc,
2693 * FreeBSD seems to not notice sometimes...
2697 fd
= qemu_open(bs
->filename
, s
->open_flags
, 0644);
2704 /* make sure the door isn't locked at this time */
2705 ioctl(s
->fd
, CDIOCALLOW
);
2709 static int cdrom_is_inserted(BlockDriverState
*bs
)
2711 return raw_getlength(bs
) > 0;
2714 static void cdrom_eject(BlockDriverState
*bs
, bool eject_flag
)
2716 BDRVRawState
*s
= bs
->opaque
;
2721 (void) ioctl(s
->fd
, CDIOCALLOW
);
2724 if (ioctl(s
->fd
, CDIOCEJECT
) < 0)
2725 perror("CDIOCEJECT");
2727 if (ioctl(s
->fd
, CDIOCCLOSE
) < 0)
2728 perror("CDIOCCLOSE");
2734 static void cdrom_lock_medium(BlockDriverState
*bs
, bool locked
)
2736 BDRVRawState
*s
= bs
->opaque
;
2740 if (ioctl(s
->fd
, (locked
? CDIOCPREVENT
: CDIOCALLOW
)) < 0) {
2742 * Note: an error can happen if the distribution automatically
2745 /* perror("CDROM_LOCKDOOR"); */
2749 static BlockDriver bdrv_host_cdrom
= {
2750 .format_name
= "host_cdrom",
2751 .protocol_name
= "host_cdrom",
2752 .instance_size
= sizeof(BDRVRawState
),
2753 .bdrv_needs_filename
= true,
2754 .bdrv_probe_device
= cdrom_probe_device
,
2755 .bdrv_parse_filename
= cdrom_parse_filename
,
2756 .bdrv_file_open
= cdrom_open
,
2757 .bdrv_close
= raw_close
,
2758 .bdrv_reopen_prepare
= raw_reopen_prepare
,
2759 .bdrv_reopen_commit
= raw_reopen_commit
,
2760 .bdrv_reopen_abort
= raw_reopen_abort
,
2761 .bdrv_create
= hdev_create
,
2762 .create_opts
= &raw_create_opts
,
2764 .bdrv_aio_readv
= raw_aio_readv
,
2765 .bdrv_aio_writev
= raw_aio_writev
,
2766 .bdrv_aio_flush
= raw_aio_flush
,
2767 .bdrv_refresh_limits
= raw_refresh_limits
,
2768 .bdrv_io_plug
= raw_aio_plug
,
2769 .bdrv_io_unplug
= raw_aio_unplug
,
2770 .bdrv_flush_io_queue
= raw_aio_flush_io_queue
,
2772 .bdrv_truncate
= raw_truncate
,
2773 .bdrv_getlength
= raw_getlength
,
2774 .has_variable_length
= true,
2775 .bdrv_get_allocated_file_size
2776 = raw_get_allocated_file_size
,
2778 .bdrv_detach_aio_context
= raw_detach_aio_context
,
2779 .bdrv_attach_aio_context
= raw_attach_aio_context
,
2781 /* removable device support */
2782 .bdrv_is_inserted
= cdrom_is_inserted
,
2783 .bdrv_eject
= cdrom_eject
,
2784 .bdrv_lock_medium
= cdrom_lock_medium
,
2786 #endif /* __FreeBSD__ */
2788 static void bdrv_file_init(void)
2791 * Register all the drivers. Note that order is important, the driver
2792 * registered last will get probed first.
2794 bdrv_register(&bdrv_file
);
2795 bdrv_register(&bdrv_host_device
);
2797 bdrv_register(&bdrv_host_floppy
);
2798 bdrv_register(&bdrv_host_cdrom
);
2800 #if defined(__FreeBSD__) || defined(__FreeBSD_kernel__)
2801 bdrv_register(&bdrv_host_cdrom
);
2805 block_init(bdrv_file_init
);