Fix kvm_tpr_opt_setup() args
[qemu-kvm/fedora.git] / block-raw-posix.c
blob820de243a5869b76ae2178c3cb44071690cd43ef
1 /*
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
22 * THE SOFTWARE.
24 #include "qemu-common.h"
25 #include "qemu-timer.h"
26 #include "qemu-char.h"
27 #include "block_int.h"
28 #include "compatfd.h"
29 #include <assert.h>
30 #ifdef CONFIG_AIO
31 #include "posix-aio-compat.h"
32 #endif
34 #ifdef CONFIG_COCOA
35 #include <paths.h>
36 #include <sys/param.h>
37 #include <IOKit/IOKitLib.h>
38 #include <IOKit/IOBSD.h>
39 #include <IOKit/storage/IOMediaBSDClient.h>
40 #include <IOKit/storage/IOMedia.h>
41 #include <IOKit/storage/IOCDMedia.h>
42 //#include <IOKit/storage/IOCDTypes.h>
43 #include <CoreFoundation/CoreFoundation.h>
44 #endif
46 #ifdef __sun__
47 #define _POSIX_PTHREAD_SEMANTICS 1
48 #include <signal.h>
49 #include <sys/dkio.h>
50 #endif
51 #ifdef __linux__
52 #include <sys/ioctl.h>
53 #include <linux/cdrom.h>
54 #include <linux/fd.h>
55 #endif
56 #ifdef __FreeBSD__
57 #include <signal.h>
58 #include <sys/disk.h>
59 #endif
61 #ifdef __OpenBSD__
62 #include <sys/ioctl.h>
63 #include <sys/disklabel.h>
64 #include <sys/dkio.h>
65 #endif
67 //#define DEBUG_FLOPPY
69 //#define DEBUG_BLOCK
70 #if defined(DEBUG_BLOCK)
71 #define DEBUG_BLOCK_PRINT(formatCstr, args...) do { if (loglevel != 0) \
72 { fprintf(logfile, formatCstr, ##args); fflush(logfile); } } while (0)
73 #else
74 #define DEBUG_BLOCK_PRINT(formatCstr, args...)
75 #endif
77 /* OS X does not have O_DSYNC */
78 #ifndef O_DSYNC
79 #define O_DSYNC O_SYNC
80 #endif
82 /* Approximate O_DIRECT with O_DSYNC if O_DIRECT isn't available */
83 #ifndef O_DIRECT
84 #define O_DIRECT O_DSYNC
85 #endif
87 #define FTYPE_FILE 0
88 #define FTYPE_CD 1
89 #define FTYPE_FD 2
91 #define ALIGNED_BUFFER_SIZE (32 * 512)
93 /* if the FD is not accessed during that time (in ms), we try to
94 reopen it to see if the disk has been changed */
95 #define FD_OPEN_TIMEOUT 1000
97 typedef struct BDRVRawState {
98 int fd;
99 int type;
100 unsigned int lseek_err_cnt;
101 #if defined(__linux__)
102 /* linux floppy specific */
103 int fd_open_flags;
104 int64_t fd_open_time;
105 int64_t fd_error_time;
106 int fd_got_error;
107 int fd_media_changed;
108 #endif
109 uint8_t* aligned_buf;
110 } BDRVRawState;
112 static int posix_aio_init(void);
114 static int fd_open(BlockDriverState *bs);
116 static int raw_open(BlockDriverState *bs, const char *filename, int flags)
118 BDRVRawState *s = bs->opaque;
119 int fd, open_flags, ret;
121 posix_aio_init();
123 s->lseek_err_cnt = 0;
125 open_flags = O_BINARY;
126 if ((flags & BDRV_O_ACCESS) == O_RDWR) {
127 open_flags |= O_RDWR;
128 } else {
129 open_flags |= O_RDONLY;
130 bs->read_only = 1;
132 if (flags & BDRV_O_CREAT)
133 open_flags |= O_CREAT | O_TRUNC;
135 /* Use O_DSYNC for write-through caching, no flags for write-back caching,
136 * and O_DIRECT for no caching. */
137 if ((flags & BDRV_O_NOCACHE))
138 open_flags |= O_DIRECT;
139 else if (!(flags & BDRV_O_CACHE_WB))
140 open_flags |= O_DSYNC;
142 s->type = FTYPE_FILE;
144 fd = open(filename, open_flags, 0644);
145 if (fd < 0) {
146 ret = -errno;
147 if (ret == -EROFS)
148 ret = -EACCES;
149 return ret;
151 s->fd = fd;
152 s->aligned_buf = NULL;
153 if ((flags & BDRV_O_NOCACHE)) {
154 s->aligned_buf = qemu_memalign(512, ALIGNED_BUFFER_SIZE);
155 if (s->aligned_buf == NULL) {
156 ret = -errno;
157 close(fd);
158 return ret;
161 return 0;
164 /* XXX: use host sector size if necessary with:
165 #ifdef DIOCGSECTORSIZE
167 unsigned int sectorsize = 512;
168 if (!ioctl(fd, DIOCGSECTORSIZE, &sectorsize) &&
169 sectorsize > bufsize)
170 bufsize = sectorsize;
172 #endif
173 #ifdef CONFIG_COCOA
174 u_int32_t blockSize = 512;
175 if ( !ioctl( fd, DKIOCGETBLOCKSIZE, &blockSize ) && blockSize > bufsize) {
176 bufsize = blockSize;
178 #endif
182 * offset and count are in bytes, but must be multiples of 512 for files
183 * opened with O_DIRECT. buf must be aligned to 512 bytes then.
185 * This function may be called without alignment if the caller ensures
186 * that O_DIRECT is not in effect.
188 static int raw_pread_aligned(BlockDriverState *bs, int64_t offset,
189 uint8_t *buf, int count)
191 BDRVRawState *s = bs->opaque;
192 int ret;
194 ret = fd_open(bs);
195 if (ret < 0)
196 return ret;
198 if (offset >= 0 && lseek(s->fd, offset, SEEK_SET) == (off_t)-1) {
199 ++(s->lseek_err_cnt);
200 if(s->lseek_err_cnt <= 10) {
201 DEBUG_BLOCK_PRINT("raw_pread(%d:%s, %" PRId64 ", %p, %d) [%" PRId64
202 "] lseek failed : %d = %s\n",
203 s->fd, bs->filename, offset, buf, count,
204 bs->total_sectors, errno, strerror(errno));
206 return -1;
208 s->lseek_err_cnt=0;
210 ret = read(s->fd, buf, count);
211 if (ret == count)
212 goto label__raw_read__success;
214 DEBUG_BLOCK_PRINT("raw_pread(%d:%s, %" PRId64 ", %p, %d) [%" PRId64
215 "] read failed %d : %d = %s\n",
216 s->fd, bs->filename, offset, buf, count,
217 bs->total_sectors, ret, errno, strerror(errno));
219 /* Try harder for CDrom. */
220 if (bs->type == BDRV_TYPE_CDROM) {
221 lseek(s->fd, offset, SEEK_SET);
222 ret = read(s->fd, buf, count);
223 if (ret == count)
224 goto label__raw_read__success;
225 lseek(s->fd, offset, SEEK_SET);
226 ret = read(s->fd, buf, count);
227 if (ret == count)
228 goto label__raw_read__success;
230 DEBUG_BLOCK_PRINT("raw_pread(%d:%s, %" PRId64 ", %p, %d) [%" PRId64
231 "] retry read failed %d : %d = %s\n",
232 s->fd, bs->filename, offset, buf, count,
233 bs->total_sectors, ret, errno, strerror(errno));
236 label__raw_read__success:
238 return ret;
242 * offset and count are in bytes, but must be multiples of 512 for files
243 * opened with O_DIRECT. buf must be aligned to 512 bytes then.
245 * This function may be called without alignment if the caller ensures
246 * that O_DIRECT is not in effect.
248 static int raw_pwrite_aligned(BlockDriverState *bs, int64_t offset,
249 const uint8_t *buf, int count)
251 BDRVRawState *s = bs->opaque;
252 int ret;
254 ret = fd_open(bs);
255 if (ret < 0)
256 return ret;
258 if (offset >= 0 && lseek(s->fd, offset, SEEK_SET) == (off_t)-1) {
259 ++(s->lseek_err_cnt);
260 if(s->lseek_err_cnt) {
261 DEBUG_BLOCK_PRINT("raw_pwrite(%d:%s, %" PRId64 ", %p, %d) [%"
262 PRId64 "] lseek failed : %d = %s\n",
263 s->fd, bs->filename, offset, buf, count,
264 bs->total_sectors, errno, strerror(errno));
266 return -1;
268 s->lseek_err_cnt = 0;
270 ret = write(s->fd, buf, count);
271 if (ret == count)
272 goto label__raw_write__success;
274 DEBUG_BLOCK_PRINT("raw_pwrite(%d:%s, %" PRId64 ", %p, %d) [%" PRId64
275 "] write failed %d : %d = %s\n",
276 s->fd, bs->filename, offset, buf, count,
277 bs->total_sectors, ret, errno, strerror(errno));
279 label__raw_write__success:
281 return ret;
286 * offset and count are in bytes and possibly not aligned. For files opened
287 * with O_DIRECT, necessary alignments are ensured before calling
288 * raw_pread_aligned to do the actual read.
290 static int raw_pread(BlockDriverState *bs, int64_t offset,
291 uint8_t *buf, int count)
293 BDRVRawState *s = bs->opaque;
294 int size, ret, shift, sum;
296 sum = 0;
298 if (s->aligned_buf != NULL) {
300 if (offset & 0x1ff) {
301 /* align offset on a 512 bytes boundary */
303 shift = offset & 0x1ff;
304 size = (shift + count + 0x1ff) & ~0x1ff;
305 if (size > ALIGNED_BUFFER_SIZE)
306 size = ALIGNED_BUFFER_SIZE;
307 ret = raw_pread_aligned(bs, offset - shift, s->aligned_buf, size);
308 if (ret < 0)
309 return ret;
311 size = 512 - shift;
312 if (size > count)
313 size = count;
314 memcpy(buf, s->aligned_buf + shift, size);
316 buf += size;
317 offset += size;
318 count -= size;
319 sum += size;
321 if (count == 0)
322 return sum;
324 if (count & 0x1ff || (uintptr_t) buf & 0x1ff) {
326 /* read on aligned buffer */
328 while (count) {
330 size = (count + 0x1ff) & ~0x1ff;
331 if (size > ALIGNED_BUFFER_SIZE)
332 size = ALIGNED_BUFFER_SIZE;
334 ret = raw_pread_aligned(bs, offset, s->aligned_buf, size);
335 if (ret < 0)
336 return ret;
338 size = ret;
339 if (size > count)
340 size = count;
342 memcpy(buf, s->aligned_buf, size);
344 buf += size;
345 offset += size;
346 count -= size;
347 sum += size;
350 return sum;
354 return raw_pread_aligned(bs, offset, buf, count) + sum;
358 * offset and count are in bytes and possibly not aligned. For files opened
359 * with O_DIRECT, necessary alignments are ensured before calling
360 * raw_pwrite_aligned to do the actual write.
362 static int raw_pwrite(BlockDriverState *bs, int64_t offset,
363 const uint8_t *buf, int count)
365 BDRVRawState *s = bs->opaque;
366 int size, ret, shift, sum;
368 sum = 0;
370 if (s->aligned_buf != NULL) {
372 if (offset & 0x1ff) {
373 /* align offset on a 512 bytes boundary */
374 shift = offset & 0x1ff;
375 ret = raw_pread_aligned(bs, offset - shift, s->aligned_buf, 512);
376 if (ret < 0)
377 return ret;
379 size = 512 - shift;
380 if (size > count)
381 size = count;
382 memcpy(s->aligned_buf + shift, buf, size);
384 ret = raw_pwrite_aligned(bs, offset - shift, s->aligned_buf, 512);
385 if (ret < 0)
386 return ret;
388 buf += size;
389 offset += size;
390 count -= size;
391 sum += size;
393 if (count == 0)
394 return sum;
396 if (count & 0x1ff || (uintptr_t) buf & 0x1ff) {
398 while ((size = (count & ~0x1ff)) != 0) {
400 if (size > ALIGNED_BUFFER_SIZE)
401 size = ALIGNED_BUFFER_SIZE;
403 memcpy(s->aligned_buf, buf, size);
405 ret = raw_pwrite_aligned(bs, offset, s->aligned_buf, size);
406 if (ret < 0)
407 return ret;
409 buf += ret;
410 offset += ret;
411 count -= ret;
412 sum += ret;
414 /* here, count < 512 because (count & ~0x1ff) == 0 */
415 if (count) {
416 ret = raw_pread_aligned(bs, offset, s->aligned_buf, 512);
417 if (ret < 0)
418 return ret;
419 memcpy(s->aligned_buf, buf, count);
421 ret = raw_pwrite_aligned(bs, offset, s->aligned_buf, 512);
422 if (ret < 0)
423 return ret;
424 if (count < ret)
425 ret = count;
427 sum += ret;
429 return sum;
432 return raw_pwrite_aligned(bs, offset, buf, count) + sum;
435 #ifdef CONFIG_AIO
436 /***********************************************************/
437 /* Unix AIO using POSIX AIO */
439 typedef struct RawAIOCB {
440 BlockDriverAIOCB common;
441 struct qemu_paiocb aiocb;
442 struct RawAIOCB *next;
443 int ret;
444 } RawAIOCB;
446 typedef struct PosixAioState
448 int fd;
449 RawAIOCB *first_aio;
450 } PosixAioState;
452 static void posix_aio_read(void *opaque)
454 PosixAioState *s = opaque;
455 RawAIOCB *acb, **pacb;
456 int ret;
457 size_t offset;
458 union {
459 struct qemu_signalfd_siginfo siginfo;
460 char buf[128];
461 } sig;
463 /* try to read from signalfd, don't freak out if we can't read anything */
464 offset = 0;
465 while (offset < 128) {
466 ssize_t len;
468 len = read(s->fd, sig.buf + offset, 128 - offset);
469 if (len == -1 && errno == EINTR)
470 continue;
471 if (len == -1 && errno == EAGAIN) {
472 /* there is no natural reason for this to happen,
473 * so we'll spin hard until we get everything just
474 * to be on the safe side. */
475 if (offset > 0)
476 continue;
479 offset += len;
482 for(;;) {
483 pacb = &s->first_aio;
484 for(;;) {
485 acb = *pacb;
486 if (!acb)
487 goto the_end;
488 ret = qemu_paio_error(&acb->aiocb);
489 if (ret == ECANCELED) {
490 /* remove the request */
491 *pacb = acb->next;
492 qemu_aio_release(acb);
493 } else if (ret != EINPROGRESS) {
494 /* end of aio */
495 if (ret == 0) {
496 ret = qemu_paio_return(&acb->aiocb);
497 if (ret == acb->aiocb.aio_nbytes)
498 ret = 0;
499 else
500 ret = -EINVAL;
501 } else {
502 ret = -ret;
504 /* remove the request */
505 *pacb = acb->next;
506 /* call the callback */
507 acb->common.cb(acb->common.opaque, ret);
508 qemu_aio_release(acb);
509 break;
510 } else {
511 pacb = &acb->next;
515 the_end: ;
518 static int posix_aio_flush(void *opaque)
520 PosixAioState *s = opaque;
521 return !!s->first_aio;
524 static PosixAioState *posix_aio_state;
526 static int posix_aio_init(void)
528 sigset_t mask;
529 PosixAioState *s;
530 int fds[2];
531 struct qemu_paioinit ai;
533 if (posix_aio_state)
534 return 0;
536 s = qemu_malloc(sizeof(PosixAioState));
537 if (s == NULL)
538 return -ENOMEM;
540 /* Make sure to block AIO signal */
541 sigemptyset(&mask);
542 sigaddset(&mask, SIGUSR2);
543 sigprocmask(SIG_BLOCK, &mask, NULL);
545 s->first_aio = NULL;
546 s->fd = qemu_signalfd(&mask);
547 if (s->fd == -1) {
548 fprintf(stderr, "failed to create signalfd\n");
549 return -errno;
552 fcntl(s->fd, F_SETFL, O_NONBLOCK);
554 qemu_aio_set_fd_handler(s->fd, posix_aio_read, NULL, posix_aio_flush, s);
556 memset(&ai, 0, sizeof(ai));
557 ai.aio_threads = 64;
558 ai.aio_num = 64;
559 qemu_paio_init(&ai);
561 posix_aio_state = s;
563 return 0;
566 static RawAIOCB *raw_aio_setup(BlockDriverState *bs,
567 int64_t sector_num, uint8_t *buf, int nb_sectors,
568 BlockDriverCompletionFunc *cb, void *opaque)
570 BDRVRawState *s = bs->opaque;
571 RawAIOCB *acb;
573 if (fd_open(bs) < 0)
574 return NULL;
576 acb = qemu_aio_get(bs, cb, opaque);
577 if (!acb)
578 return NULL;
579 acb->aiocb.aio_fildes = s->fd;
580 acb->aiocb.aio_sigevent.sigev_signo = SIGUSR2;
581 acb->aiocb.aio_sigevent.sigev_notify = SIGEV_SIGNAL;
582 acb->aiocb.aio_buf = buf;
583 if (nb_sectors < 0)
584 acb->aiocb.aio_nbytes = -nb_sectors;
585 else
586 acb->aiocb.aio_nbytes = nb_sectors * 512;
587 acb->aiocb.aio_offset = sector_num * 512;
588 acb->next = posix_aio_state->first_aio;
589 posix_aio_state->first_aio = acb;
590 return acb;
593 static void raw_aio_em_cb(void* opaque)
595 RawAIOCB *acb = opaque;
596 acb->common.cb(acb->common.opaque, acb->ret);
597 qemu_aio_release(acb);
600 static BlockDriverAIOCB *raw_aio_read(BlockDriverState *bs,
601 int64_t sector_num, uint8_t *buf, int nb_sectors,
602 BlockDriverCompletionFunc *cb, void *opaque)
604 RawAIOCB *acb;
607 * If O_DIRECT is used and the buffer is not aligned fall back
608 * to synchronous IO.
610 BDRVRawState *s = bs->opaque;
612 if (unlikely(s->aligned_buf != NULL && ((uintptr_t) buf % 512))) {
613 QEMUBH *bh;
614 acb = qemu_aio_get(bs, cb, opaque);
615 acb->ret = raw_pread(bs, 512 * sector_num, buf, 512 * nb_sectors);
616 bh = qemu_bh_new(raw_aio_em_cb, acb);
617 qemu_bh_schedule(bh);
618 return &acb->common;
621 acb = raw_aio_setup(bs, sector_num, buf, nb_sectors, cb, opaque);
622 if (!acb)
623 return NULL;
624 if (qemu_paio_read(&acb->aiocb) < 0) {
625 qemu_aio_release(acb);
626 return NULL;
628 return &acb->common;
631 static BlockDriverAIOCB *raw_aio_write(BlockDriverState *bs,
632 int64_t sector_num, const uint8_t *buf, int nb_sectors,
633 BlockDriverCompletionFunc *cb, void *opaque)
635 RawAIOCB *acb;
638 * If O_DIRECT is used and the buffer is not aligned fall back
639 * to synchronous IO.
641 BDRVRawState *s = bs->opaque;
643 if (unlikely(s->aligned_buf != NULL && ((uintptr_t) buf % 512))) {
644 QEMUBH *bh;
645 acb = qemu_aio_get(bs, cb, opaque);
646 acb->ret = raw_pwrite(bs, 512 * sector_num, buf, 512 * nb_sectors);
647 bh = qemu_bh_new(raw_aio_em_cb, acb);
648 qemu_bh_schedule(bh);
649 return &acb->common;
652 acb = raw_aio_setup(bs, sector_num, (uint8_t*)buf, nb_sectors, cb, opaque);
653 if (!acb)
654 return NULL;
655 if (qemu_paio_write(&acb->aiocb) < 0) {
656 qemu_aio_release(acb);
657 return NULL;
659 return &acb->common;
662 static void raw_aio_cancel(BlockDriverAIOCB *blockacb)
664 int ret;
665 RawAIOCB *acb = (RawAIOCB *)blockacb;
666 RawAIOCB **pacb;
668 ret = qemu_paio_cancel(acb->aiocb.aio_fildes, &acb->aiocb);
669 if (ret == QEMU_PAIO_NOTCANCELED) {
670 /* fail safe: if the aio could not be canceled, we wait for
671 it */
672 while (qemu_paio_error(&acb->aiocb) == EINPROGRESS);
675 /* remove the callback from the queue */
676 pacb = &posix_aio_state->first_aio;
677 for(;;) {
678 if (*pacb == NULL) {
679 break;
680 } else if (*pacb == acb) {
681 *pacb = acb->next;
682 qemu_aio_release(acb);
683 break;
685 pacb = &acb->next;
688 #else /* CONFIG_AIO */
689 static int posix_aio_init(void)
691 return 0;
693 #endif /* CONFIG_AIO */
696 static void raw_close(BlockDriverState *bs)
698 BDRVRawState *s = bs->opaque;
699 if (s->fd >= 0) {
700 close(s->fd);
701 s->fd = -1;
702 if (s->aligned_buf != NULL)
703 qemu_free(s->aligned_buf);
707 static int raw_truncate(BlockDriverState *bs, int64_t offset)
709 BDRVRawState *s = bs->opaque;
710 if (s->type != FTYPE_FILE)
711 return -ENOTSUP;
712 if (ftruncate(s->fd, offset) < 0)
713 return -errno;
714 return 0;
717 #ifdef __OpenBSD__
718 static int64_t raw_getlength(BlockDriverState *bs)
720 BDRVRawState *s = bs->opaque;
721 int fd = s->fd;
722 struct stat st;
724 if (fstat(fd, &st))
725 return -1;
726 if (S_ISCHR(st.st_mode) || S_ISBLK(st.st_mode)) {
727 struct disklabel dl;
729 if (ioctl(fd, DIOCGDINFO, &dl))
730 return -1;
731 return (uint64_t)dl.d_secsize *
732 dl.d_partitions[DISKPART(st.st_rdev)].p_size;
733 } else
734 return st.st_size;
736 #else /* !__OpenBSD__ */
737 static int64_t raw_getlength(BlockDriverState *bs)
739 BDRVRawState *s = bs->opaque;
740 int fd = s->fd;
741 int64_t size;
742 #ifdef _BSD
743 struct stat sb;
744 #endif
745 #ifdef __sun__
746 struct dk_minfo minfo;
747 int rv;
748 #endif
749 int ret;
751 ret = fd_open(bs);
752 if (ret < 0)
753 return ret;
755 #ifdef _BSD
756 if (!fstat(fd, &sb) && (S_IFCHR & sb.st_mode)) {
757 #ifdef DIOCGMEDIASIZE
758 if (ioctl(fd, DIOCGMEDIASIZE, (off_t *)&size))
759 #endif
760 #ifdef CONFIG_COCOA
761 size = LONG_LONG_MAX;
762 #else
763 size = lseek(fd, 0LL, SEEK_END);
764 #endif
765 } else
766 #endif
767 #ifdef __sun__
769 * use the DKIOCGMEDIAINFO ioctl to read the size.
771 rv = ioctl ( fd, DKIOCGMEDIAINFO, &minfo );
772 if ( rv != -1 ) {
773 size = minfo.dki_lbsize * minfo.dki_capacity;
774 } else /* there are reports that lseek on some devices
775 fails, but irc discussion said that contingency
776 on contingency was overkill */
777 #endif
779 size = lseek(fd, 0, SEEK_END);
781 return size;
783 #endif
785 static int raw_create(const char *filename, int64_t total_size,
786 const char *backing_file, int flags)
788 int fd;
790 if (flags || backing_file)
791 return -ENOTSUP;
793 fd = open(filename, O_WRONLY | O_CREAT | O_TRUNC | O_BINARY,
794 0644);
795 if (fd < 0)
796 return -EIO;
797 ftruncate(fd, total_size * 512);
798 close(fd);
799 return 0;
802 static void raw_flush(BlockDriverState *bs)
804 BDRVRawState *s = bs->opaque;
805 fsync(s->fd);
808 BlockDriver bdrv_raw = {
809 "raw",
810 sizeof(BDRVRawState),
811 NULL, /* no probe for protocols */
812 raw_open,
813 NULL,
814 NULL,
815 raw_close,
816 raw_create,
817 raw_flush,
819 #ifdef CONFIG_AIO
820 .bdrv_aio_read = raw_aio_read,
821 .bdrv_aio_write = raw_aio_write,
822 .bdrv_aio_cancel = raw_aio_cancel,
823 .aiocb_size = sizeof(RawAIOCB),
824 #endif
826 .bdrv_pread = raw_pread,
827 .bdrv_pwrite = raw_pwrite,
828 .bdrv_truncate = raw_truncate,
829 .bdrv_getlength = raw_getlength,
832 /***********************************************/
833 /* host device */
835 #ifdef CONFIG_COCOA
836 static kern_return_t FindEjectableCDMedia( io_iterator_t *mediaIterator );
837 static kern_return_t GetBSDPath( io_iterator_t mediaIterator, char *bsdPath, CFIndex maxPathSize );
839 kern_return_t FindEjectableCDMedia( io_iterator_t *mediaIterator )
841 kern_return_t kernResult;
842 mach_port_t masterPort;
843 CFMutableDictionaryRef classesToMatch;
845 kernResult = IOMasterPort( MACH_PORT_NULL, &masterPort );
846 if ( KERN_SUCCESS != kernResult ) {
847 printf( "IOMasterPort returned %d\n", kernResult );
850 classesToMatch = IOServiceMatching( kIOCDMediaClass );
851 if ( classesToMatch == NULL ) {
852 printf( "IOServiceMatching returned a NULL dictionary.\n" );
853 } else {
854 CFDictionarySetValue( classesToMatch, CFSTR( kIOMediaEjectableKey ), kCFBooleanTrue );
856 kernResult = IOServiceGetMatchingServices( masterPort, classesToMatch, mediaIterator );
857 if ( KERN_SUCCESS != kernResult )
859 printf( "IOServiceGetMatchingServices returned %d\n", kernResult );
862 return kernResult;
865 kern_return_t GetBSDPath( io_iterator_t mediaIterator, char *bsdPath, CFIndex maxPathSize )
867 io_object_t nextMedia;
868 kern_return_t kernResult = KERN_FAILURE;
869 *bsdPath = '\0';
870 nextMedia = IOIteratorNext( mediaIterator );
871 if ( nextMedia )
873 CFTypeRef bsdPathAsCFString;
874 bsdPathAsCFString = IORegistryEntryCreateCFProperty( nextMedia, CFSTR( kIOBSDNameKey ), kCFAllocatorDefault, 0 );
875 if ( bsdPathAsCFString ) {
876 size_t devPathLength;
877 strcpy( bsdPath, _PATH_DEV );
878 strcat( bsdPath, "r" );
879 devPathLength = strlen( bsdPath );
880 if ( CFStringGetCString( bsdPathAsCFString, bsdPath + devPathLength, maxPathSize - devPathLength, kCFStringEncodingASCII ) ) {
881 kernResult = KERN_SUCCESS;
883 CFRelease( bsdPathAsCFString );
885 IOObjectRelease( nextMedia );
888 return kernResult;
891 #endif
893 static int hdev_open(BlockDriverState *bs, const char *filename, int flags)
895 BDRVRawState *s = bs->opaque;
896 int fd, open_flags, ret;
898 posix_aio_init();
900 #ifdef CONFIG_COCOA
901 if (strstart(filename, "/dev/cdrom", NULL)) {
902 kern_return_t kernResult;
903 io_iterator_t mediaIterator;
904 char bsdPath[ MAXPATHLEN ];
905 int fd;
907 kernResult = FindEjectableCDMedia( &mediaIterator );
908 kernResult = GetBSDPath( mediaIterator, bsdPath, sizeof( bsdPath ) );
910 if ( bsdPath[ 0 ] != '\0' ) {
911 strcat(bsdPath,"s0");
912 /* some CDs don't have a partition 0 */
913 fd = open(bsdPath, O_RDONLY | O_BINARY | O_LARGEFILE);
914 if (fd < 0) {
915 bsdPath[strlen(bsdPath)-1] = '1';
916 } else {
917 close(fd);
919 filename = bsdPath;
922 if ( mediaIterator )
923 IOObjectRelease( mediaIterator );
925 #endif
926 open_flags = O_BINARY;
927 if ((flags & BDRV_O_ACCESS) == O_RDWR) {
928 open_flags |= O_RDWR;
929 } else {
930 open_flags |= O_RDONLY;
931 bs->read_only = 1;
933 /* Use O_DSYNC for write-through caching, no flags for write-back caching,
934 * and O_DIRECT for no caching. */
935 if ((flags & BDRV_O_NOCACHE))
936 open_flags |= O_DIRECT;
937 else if (!(flags & BDRV_O_CACHE_WB))
938 open_flags |= O_DSYNC;
940 s->type = FTYPE_FILE;
941 #if defined(__linux__)
942 if (strstart(filename, "/dev/cd", NULL)) {
943 /* open will not fail even if no CD is inserted */
944 open_flags |= O_NONBLOCK;
945 s->type = FTYPE_CD;
946 } else if (strstart(filename, "/dev/fd", NULL)) {
947 s->type = FTYPE_FD;
948 s->fd_open_flags = open_flags;
949 /* open will not fail even if no floppy is inserted */
950 open_flags |= O_NONBLOCK;
951 } else if (strstart(filename, "/dev/sg", NULL)) {
952 bs->sg = 1;
954 #endif
955 fd = open(filename, open_flags, 0644);
956 if (fd < 0) {
957 ret = -errno;
958 if (ret == -EROFS)
959 ret = -EACCES;
960 return ret;
962 s->fd = fd;
963 #if defined(__linux__)
964 /* close fd so that we can reopen it as needed */
965 if (s->type == FTYPE_FD) {
966 close(s->fd);
967 s->fd = -1;
968 s->fd_media_changed = 1;
970 #endif
971 return 0;
974 #if defined(__linux__)
975 /* Note: we do not have a reliable method to detect if the floppy is
976 present. The current method is to try to open the floppy at every
977 I/O and to keep it opened during a few hundreds of ms. */
978 static int fd_open(BlockDriverState *bs)
980 BDRVRawState *s = bs->opaque;
981 int last_media_present;
983 if (s->type != FTYPE_FD)
984 return 0;
985 last_media_present = (s->fd >= 0);
986 if (s->fd >= 0 &&
987 (qemu_get_clock(rt_clock) - s->fd_open_time) >= FD_OPEN_TIMEOUT) {
988 close(s->fd);
989 s->fd = -1;
990 #ifdef DEBUG_FLOPPY
991 printf("Floppy closed\n");
992 #endif
994 if (s->fd < 0) {
995 if (s->fd_got_error &&
996 (qemu_get_clock(rt_clock) - s->fd_error_time) < FD_OPEN_TIMEOUT) {
997 #ifdef DEBUG_FLOPPY
998 printf("No floppy (open delayed)\n");
999 #endif
1000 return -EIO;
1002 s->fd = open(bs->filename, s->fd_open_flags);
1003 if (s->fd < 0) {
1004 s->fd_error_time = qemu_get_clock(rt_clock);
1005 s->fd_got_error = 1;
1006 if (last_media_present)
1007 s->fd_media_changed = 1;
1008 #ifdef DEBUG_FLOPPY
1009 printf("No floppy\n");
1010 #endif
1011 return -EIO;
1013 #ifdef DEBUG_FLOPPY
1014 printf("Floppy opened\n");
1015 #endif
1017 if (!last_media_present)
1018 s->fd_media_changed = 1;
1019 s->fd_open_time = qemu_get_clock(rt_clock);
1020 s->fd_got_error = 0;
1021 return 0;
1024 static int raw_is_inserted(BlockDriverState *bs)
1026 BDRVRawState *s = bs->opaque;
1027 int ret;
1029 switch(s->type) {
1030 case FTYPE_CD:
1031 ret = ioctl(s->fd, CDROM_DRIVE_STATUS, CDSL_CURRENT);
1032 if (ret == CDS_DISC_OK)
1033 return 1;
1034 else
1035 return 0;
1036 break;
1037 case FTYPE_FD:
1038 ret = fd_open(bs);
1039 return (ret >= 0);
1040 default:
1041 return 1;
1045 /* currently only used by fdc.c, but a CD version would be good too */
1046 static int raw_media_changed(BlockDriverState *bs)
1048 BDRVRawState *s = bs->opaque;
1050 switch(s->type) {
1051 case FTYPE_FD:
1053 int ret;
1054 /* XXX: we do not have a true media changed indication. It
1055 does not work if the floppy is changed without trying
1056 to read it */
1057 fd_open(bs);
1058 ret = s->fd_media_changed;
1059 s->fd_media_changed = 0;
1060 #ifdef DEBUG_FLOPPY
1061 printf("Floppy changed=%d\n", ret);
1062 #endif
1063 return ret;
1065 default:
1066 return -ENOTSUP;
1070 static int raw_eject(BlockDriverState *bs, int eject_flag)
1072 BDRVRawState *s = bs->opaque;
1074 switch(s->type) {
1075 case FTYPE_CD:
1076 if (eject_flag) {
1077 if (ioctl (s->fd, CDROMEJECT, NULL) < 0)
1078 perror("CDROMEJECT");
1079 } else {
1080 if (ioctl (s->fd, CDROMCLOSETRAY, NULL) < 0)
1081 perror("CDROMEJECT");
1083 break;
1084 case FTYPE_FD:
1086 int fd;
1087 if (s->fd >= 0) {
1088 close(s->fd);
1089 s->fd = -1;
1091 fd = open(bs->filename, s->fd_open_flags | O_NONBLOCK);
1092 if (fd >= 0) {
1093 if (ioctl(fd, FDEJECT, 0) < 0)
1094 perror("FDEJECT");
1095 close(fd);
1098 break;
1099 default:
1100 return -ENOTSUP;
1102 return 0;
1105 static int raw_set_locked(BlockDriverState *bs, int locked)
1107 BDRVRawState *s = bs->opaque;
1109 switch(s->type) {
1110 case FTYPE_CD:
1111 if (ioctl (s->fd, CDROM_LOCKDOOR, locked) < 0) {
1112 /* Note: an error can happen if the distribution automatically
1113 mounts the CD-ROM */
1114 // perror("CDROM_LOCKDOOR");
1116 break;
1117 default:
1118 return -ENOTSUP;
1120 return 0;
1123 static int raw_ioctl(BlockDriverState *bs, unsigned long int req, void *buf)
1125 BDRVRawState *s = bs->opaque;
1127 return ioctl(s->fd, req, buf);
1129 #else
1131 static int fd_open(BlockDriverState *bs)
1133 return 0;
1136 static int raw_is_inserted(BlockDriverState *bs)
1138 return 1;
1141 static int raw_media_changed(BlockDriverState *bs)
1143 return -ENOTSUP;
1146 static int raw_eject(BlockDriverState *bs, int eject_flag)
1148 return -ENOTSUP;
1151 static int raw_set_locked(BlockDriverState *bs, int locked)
1153 return -ENOTSUP;
1156 static int raw_ioctl(BlockDriverState *bs, unsigned long int req, void *buf)
1158 return -ENOTSUP;
1160 #endif /* !linux */
1162 BlockDriver bdrv_host_device = {
1163 "host_device",
1164 sizeof(BDRVRawState),
1165 NULL, /* no probe for protocols */
1166 hdev_open,
1167 NULL,
1168 NULL,
1169 raw_close,
1170 NULL,
1171 raw_flush,
1173 #ifdef CONFIG_AIO
1174 .bdrv_aio_read = raw_aio_read,
1175 .bdrv_aio_write = raw_aio_write,
1176 .bdrv_aio_cancel = raw_aio_cancel,
1177 .aiocb_size = sizeof(RawAIOCB),
1178 #endif
1180 .bdrv_pread = raw_pread,
1181 .bdrv_pwrite = raw_pwrite,
1182 .bdrv_getlength = raw_getlength,
1184 /* removable device support */
1185 .bdrv_is_inserted = raw_is_inserted,
1186 .bdrv_media_changed = raw_media_changed,
1187 .bdrv_eject = raw_eject,
1188 .bdrv_set_locked = raw_set_locked,
1189 /* generic scsi device */
1190 .bdrv_ioctl = raw_ioctl,