]> git.proxmox.com Git - qemu.git/blob - block-raw-posix.c
Use VNC protocol defines (Alexander Graf)
[qemu.git] / block-raw-posix.c
1 /*
2 * Block driver for RAW files (posix)
3 *
4 * Copyright (c) 2006 Fabrice Bellard
5 *
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:
12 *
13 * The above copyright notice and this permission notice shall be included in
14 * all copies or substantial portions of the Software.
15 *
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.
23 */
24 #include "qemu-common.h"
25 #include "qemu-timer.h"
26 #include "qemu-char.h"
27 #include "block_int.h"
28 #include <assert.h>
29 #ifdef CONFIG_AIO
30 #include "posix-aio-compat.h"
31 #endif
32
33 #ifdef CONFIG_COCOA
34 #include <paths.h>
35 #include <sys/param.h>
36 #include <IOKit/IOKitLib.h>
37 #include <IOKit/IOBSD.h>
38 #include <IOKit/storage/IOMediaBSDClient.h>
39 #include <IOKit/storage/IOMedia.h>
40 #include <IOKit/storage/IOCDMedia.h>
41 //#include <IOKit/storage/IOCDTypes.h>
42 #include <CoreFoundation/CoreFoundation.h>
43 #endif
44
45 #ifdef __sun__
46 #define _POSIX_PTHREAD_SEMANTICS 1
47 #include <signal.h>
48 #include <sys/dkio.h>
49 #endif
50 #ifdef __linux__
51 #include <sys/ioctl.h>
52 #include <linux/cdrom.h>
53 #include <linux/fd.h>
54 #endif
55 #ifdef __FreeBSD__
56 #include <signal.h>
57 #include <sys/disk.h>
58 #endif
59
60 #ifdef __OpenBSD__
61 #include <sys/ioctl.h>
62 #include <sys/disklabel.h>
63 #include <sys/dkio.h>
64 #endif
65
66 //#define DEBUG_FLOPPY
67
68 //#define DEBUG_BLOCK
69 #if defined(DEBUG_BLOCK)
70 #define DEBUG_BLOCK_PRINT(formatCstr, args...) do { if (qemu_log_enabled()) \
71 { qemu_log(formatCstr, ##args); qemu_log_flush(); } } while (0)
72 #else
73 #define DEBUG_BLOCK_PRINT(formatCstr, args...)
74 #endif
75
76 /* OS X does not have O_DSYNC */
77 #ifndef O_DSYNC
78 #define O_DSYNC O_SYNC
79 #endif
80
81 /* Approximate O_DIRECT with O_DSYNC if O_DIRECT isn't available */
82 #ifndef O_DIRECT
83 #define O_DIRECT O_DSYNC
84 #endif
85
86 #define FTYPE_FILE 0
87 #define FTYPE_CD 1
88 #define FTYPE_FD 2
89
90 #define ALIGNED_BUFFER_SIZE (32 * 512)
91
92 /* if the FD is not accessed during that time (in ms), we try to
93 reopen it to see if the disk has been changed */
94 #define FD_OPEN_TIMEOUT 1000
95
96 typedef struct BDRVRawState {
97 int fd;
98 int type;
99 unsigned int lseek_err_cnt;
100 #if defined(__linux__)
101 /* linux floppy specific */
102 int fd_open_flags;
103 int64_t fd_open_time;
104 int64_t fd_error_time;
105 int fd_got_error;
106 int fd_media_changed;
107 #endif
108 uint8_t* aligned_buf;
109 } BDRVRawState;
110
111 static int posix_aio_init(void);
112
113 static int fd_open(BlockDriverState *bs);
114
115 static int raw_open(BlockDriverState *bs, const char *filename, int flags)
116 {
117 BDRVRawState *s = bs->opaque;
118 int fd, open_flags, ret;
119
120 posix_aio_init();
121
122 s->lseek_err_cnt = 0;
123
124 open_flags = O_BINARY;
125 if ((flags & BDRV_O_ACCESS) == O_RDWR) {
126 open_flags |= O_RDWR;
127 } else {
128 open_flags |= O_RDONLY;
129 bs->read_only = 1;
130 }
131 if (flags & BDRV_O_CREAT)
132 open_flags |= O_CREAT | O_TRUNC;
133
134 /* Use O_DSYNC for write-through caching, no flags for write-back caching,
135 * and O_DIRECT for no caching. */
136 if ((flags & BDRV_O_NOCACHE))
137 open_flags |= O_DIRECT;
138 else if (!(flags & BDRV_O_CACHE_WB))
139 open_flags |= O_DSYNC;
140
141 s->type = FTYPE_FILE;
142
143 fd = open(filename, open_flags, 0644);
144 if (fd < 0) {
145 ret = -errno;
146 if (ret == -EROFS)
147 ret = -EACCES;
148 return ret;
149 }
150 s->fd = fd;
151 s->aligned_buf = NULL;
152 if ((flags & BDRV_O_NOCACHE)) {
153 s->aligned_buf = qemu_memalign(512, ALIGNED_BUFFER_SIZE);
154 if (s->aligned_buf == NULL) {
155 ret = -errno;
156 close(fd);
157 return ret;
158 }
159 }
160 return 0;
161 }
162
163 /* XXX: use host sector size if necessary with:
164 #ifdef DIOCGSECTORSIZE
165 {
166 unsigned int sectorsize = 512;
167 if (!ioctl(fd, DIOCGSECTORSIZE, &sectorsize) &&
168 sectorsize > bufsize)
169 bufsize = sectorsize;
170 }
171 #endif
172 #ifdef CONFIG_COCOA
173 u_int32_t blockSize = 512;
174 if ( !ioctl( fd, DKIOCGETBLOCKSIZE, &blockSize ) && blockSize > bufsize) {
175 bufsize = blockSize;
176 }
177 #endif
178 */
179
180 /*
181 * offset and count are in bytes, but must be multiples of 512 for files
182 * opened with O_DIRECT. buf must be aligned to 512 bytes then.
183 *
184 * This function may be called without alignment if the caller ensures
185 * that O_DIRECT is not in effect.
186 */
187 static int raw_pread_aligned(BlockDriverState *bs, int64_t offset,
188 uint8_t *buf, int count)
189 {
190 BDRVRawState *s = bs->opaque;
191 int ret;
192
193 ret = fd_open(bs);
194 if (ret < 0)
195 return ret;
196
197 if (offset >= 0 && lseek(s->fd, offset, SEEK_SET) == (off_t)-1) {
198 ++(s->lseek_err_cnt);
199 if(s->lseek_err_cnt <= 10) {
200 DEBUG_BLOCK_PRINT("raw_pread(%d:%s, %" PRId64 ", %p, %d) [%" PRId64
201 "] lseek failed : %d = %s\n",
202 s->fd, bs->filename, offset, buf, count,
203 bs->total_sectors, errno, strerror(errno));
204 }
205 return -1;
206 }
207 s->lseek_err_cnt=0;
208
209 ret = read(s->fd, buf, count);
210 if (ret == count)
211 goto label__raw_read__success;
212
213 DEBUG_BLOCK_PRINT("raw_pread(%d:%s, %" PRId64 ", %p, %d) [%" PRId64
214 "] read failed %d : %d = %s\n",
215 s->fd, bs->filename, offset, buf, count,
216 bs->total_sectors, ret, errno, strerror(errno));
217
218 /* Try harder for CDrom. */
219 if (bs->type == BDRV_TYPE_CDROM) {
220 lseek(s->fd, offset, SEEK_SET);
221 ret = read(s->fd, buf, count);
222 if (ret == count)
223 goto label__raw_read__success;
224 lseek(s->fd, offset, SEEK_SET);
225 ret = read(s->fd, buf, count);
226 if (ret == count)
227 goto label__raw_read__success;
228
229 DEBUG_BLOCK_PRINT("raw_pread(%d:%s, %" PRId64 ", %p, %d) [%" PRId64
230 "] retry read failed %d : %d = %s\n",
231 s->fd, bs->filename, offset, buf, count,
232 bs->total_sectors, ret, errno, strerror(errno));
233 }
234
235 label__raw_read__success:
236
237 return ret;
238 }
239
240 /*
241 * offset and count are in bytes, but must be multiples of 512 for files
242 * opened with O_DIRECT. buf must be aligned to 512 bytes then.
243 *
244 * This function may be called without alignment if the caller ensures
245 * that O_DIRECT is not in effect.
246 */
247 static int raw_pwrite_aligned(BlockDriverState *bs, int64_t offset,
248 const uint8_t *buf, int count)
249 {
250 BDRVRawState *s = bs->opaque;
251 int ret;
252
253 ret = fd_open(bs);
254 if (ret < 0)
255 return -errno;
256
257 if (offset >= 0 && lseek(s->fd, offset, SEEK_SET) == (off_t)-1) {
258 ++(s->lseek_err_cnt);
259 if(s->lseek_err_cnt) {
260 DEBUG_BLOCK_PRINT("raw_pwrite(%d:%s, %" PRId64 ", %p, %d) [%"
261 PRId64 "] lseek failed : %d = %s\n",
262 s->fd, bs->filename, offset, buf, count,
263 bs->total_sectors, errno, strerror(errno));
264 }
265 return -EIO;
266 }
267 s->lseek_err_cnt = 0;
268
269 ret = write(s->fd, buf, count);
270 if (ret == count)
271 goto label__raw_write__success;
272
273 DEBUG_BLOCK_PRINT("raw_pwrite(%d:%s, %" PRId64 ", %p, %d) [%" PRId64
274 "] write failed %d : %d = %s\n",
275 s->fd, bs->filename, offset, buf, count,
276 bs->total_sectors, ret, errno, strerror(errno));
277
278 label__raw_write__success:
279
280 return (ret < 0) ? -errno : ret;
281 }
282
283
284 /*
285 * offset and count are in bytes and possibly not aligned. For files opened
286 * with O_DIRECT, necessary alignments are ensured before calling
287 * raw_pread_aligned to do the actual read.
288 */
289 static int raw_pread(BlockDriverState *bs, int64_t offset,
290 uint8_t *buf, int count)
291 {
292 BDRVRawState *s = bs->opaque;
293 int size, ret, shift, sum;
294
295 sum = 0;
296
297 if (s->aligned_buf != NULL) {
298
299 if (offset & 0x1ff) {
300 /* align offset on a 512 bytes boundary */
301
302 shift = offset & 0x1ff;
303 size = (shift + count + 0x1ff) & ~0x1ff;
304 if (size > ALIGNED_BUFFER_SIZE)
305 size = ALIGNED_BUFFER_SIZE;
306 ret = raw_pread_aligned(bs, offset - shift, s->aligned_buf, size);
307 if (ret < 0)
308 return ret;
309
310 size = 512 - shift;
311 if (size > count)
312 size = count;
313 memcpy(buf, s->aligned_buf + shift, size);
314
315 buf += size;
316 offset += size;
317 count -= size;
318 sum += size;
319
320 if (count == 0)
321 return sum;
322 }
323 if (count & 0x1ff || (uintptr_t) buf & 0x1ff) {
324
325 /* read on aligned buffer */
326
327 while (count) {
328
329 size = (count + 0x1ff) & ~0x1ff;
330 if (size > ALIGNED_BUFFER_SIZE)
331 size = ALIGNED_BUFFER_SIZE;
332
333 ret = raw_pread_aligned(bs, offset, s->aligned_buf, size);
334 if (ret < 0)
335 return ret;
336
337 size = ret;
338 if (size > count)
339 size = count;
340
341 memcpy(buf, s->aligned_buf, size);
342
343 buf += size;
344 offset += size;
345 count -= size;
346 sum += size;
347 }
348
349 return sum;
350 }
351 }
352
353 return raw_pread_aligned(bs, offset, buf, count) + sum;
354 }
355
356 /*
357 * offset and count are in bytes and possibly not aligned. For files opened
358 * with O_DIRECT, necessary alignments are ensured before calling
359 * raw_pwrite_aligned to do the actual write.
360 */
361 static int raw_pwrite(BlockDriverState *bs, int64_t offset,
362 const uint8_t *buf, int count)
363 {
364 BDRVRawState *s = bs->opaque;
365 int size, ret, shift, sum;
366
367 sum = 0;
368
369 if (s->aligned_buf != NULL) {
370
371 if (offset & 0x1ff) {
372 /* align offset on a 512 bytes boundary */
373 shift = offset & 0x1ff;
374 ret = raw_pread_aligned(bs, offset - shift, s->aligned_buf, 512);
375 if (ret < 0)
376 return ret;
377
378 size = 512 - shift;
379 if (size > count)
380 size = count;
381 memcpy(s->aligned_buf + shift, buf, size);
382
383 ret = raw_pwrite_aligned(bs, offset - shift, s->aligned_buf, 512);
384 if (ret < 0)
385 return ret;
386
387 buf += size;
388 offset += size;
389 count -= size;
390 sum += size;
391
392 if (count == 0)
393 return sum;
394 }
395 if (count & 0x1ff || (uintptr_t) buf & 0x1ff) {
396
397 while ((size = (count & ~0x1ff)) != 0) {
398
399 if (size > ALIGNED_BUFFER_SIZE)
400 size = ALIGNED_BUFFER_SIZE;
401
402 memcpy(s->aligned_buf, buf, size);
403
404 ret = raw_pwrite_aligned(bs, offset, s->aligned_buf, size);
405 if (ret < 0)
406 return ret;
407
408 buf += ret;
409 offset += ret;
410 count -= ret;
411 sum += ret;
412 }
413 /* here, count < 512 because (count & ~0x1ff) == 0 */
414 if (count) {
415 ret = raw_pread_aligned(bs, offset, s->aligned_buf, 512);
416 if (ret < 0)
417 return ret;
418 memcpy(s->aligned_buf, buf, count);
419
420 ret = raw_pwrite_aligned(bs, offset, s->aligned_buf, 512);
421 if (ret < 0)
422 return ret;
423 if (count < ret)
424 ret = count;
425
426 sum += ret;
427 }
428 return sum;
429 }
430 }
431 return raw_pwrite_aligned(bs, offset, buf, count) + sum;
432 }
433
434 #ifdef CONFIG_AIO
435 /***********************************************************/
436 /* Unix AIO using POSIX AIO */
437
438 typedef struct RawAIOCB {
439 BlockDriverAIOCB common;
440 struct qemu_paiocb aiocb;
441 struct RawAIOCB *next;
442 int ret;
443 } RawAIOCB;
444
445 typedef struct PosixAioState
446 {
447 int rfd, wfd;
448 RawAIOCB *first_aio;
449 } PosixAioState;
450
451 static void posix_aio_read(void *opaque)
452 {
453 PosixAioState *s = opaque;
454 RawAIOCB *acb, **pacb;
455 int ret;
456 ssize_t len;
457
458 /* read all bytes from signal pipe */
459 for (;;) {
460 char bytes[16];
461
462 len = read(s->rfd, bytes, sizeof(bytes));
463 if (len == -1 && errno == EINTR)
464 continue; /* try again */
465 if (len == sizeof(bytes))
466 continue; /* more to read */
467 break;
468 }
469
470 for(;;) {
471 pacb = &s->first_aio;
472 for(;;) {
473 acb = *pacb;
474 if (!acb)
475 goto the_end;
476 ret = qemu_paio_error(&acb->aiocb);
477 if (ret == ECANCELED) {
478 /* remove the request */
479 *pacb = acb->next;
480 qemu_aio_release(acb);
481 } else if (ret != EINPROGRESS) {
482 /* end of aio */
483 if (ret == 0) {
484 ret = qemu_paio_return(&acb->aiocb);
485 if (ret == acb->aiocb.aio_nbytes)
486 ret = 0;
487 else
488 ret = -EINVAL;
489 } else {
490 ret = -ret;
491 }
492 /* remove the request */
493 *pacb = acb->next;
494 /* call the callback */
495 acb->common.cb(acb->common.opaque, ret);
496 qemu_aio_release(acb);
497 break;
498 } else {
499 pacb = &acb->next;
500 }
501 }
502 }
503 the_end: ;
504 }
505
506 static int posix_aio_flush(void *opaque)
507 {
508 PosixAioState *s = opaque;
509 return !!s->first_aio;
510 }
511
512 static PosixAioState *posix_aio_state;
513
514 static void aio_signal_handler(int signum)
515 {
516 if (posix_aio_state) {
517 char byte = 0;
518
519 write(posix_aio_state->wfd, &byte, sizeof(byte));
520 }
521
522 qemu_service_io();
523 }
524
525 static int posix_aio_init(void)
526 {
527 struct sigaction act;
528 PosixAioState *s;
529 int fds[2];
530 struct qemu_paioinit ai;
531
532 if (posix_aio_state)
533 return 0;
534
535 s = qemu_malloc(sizeof(PosixAioState));
536 if (s == NULL)
537 return -ENOMEM;
538
539 sigfillset(&act.sa_mask);
540 act.sa_flags = 0; /* do not restart syscalls to interrupt select() */
541 act.sa_handler = aio_signal_handler;
542 sigaction(SIGUSR2, &act, NULL);
543
544 s->first_aio = NULL;
545 if (pipe(fds) == -1) {
546 fprintf(stderr, "failed to create pipe\n");
547 return -errno;
548 }
549
550 s->rfd = fds[0];
551 s->wfd = fds[1];
552
553 fcntl(s->rfd, F_SETFL, O_NONBLOCK);
554 fcntl(s->wfd, F_SETFL, O_NONBLOCK);
555
556 qemu_aio_set_fd_handler(s->rfd, posix_aio_read, NULL, posix_aio_flush, s);
557
558 memset(&ai, 0, sizeof(ai));
559 ai.aio_threads = 64;
560 ai.aio_num = 64;
561 qemu_paio_init(&ai);
562
563 posix_aio_state = s;
564
565 return 0;
566 }
567
568 static RawAIOCB *raw_aio_setup(BlockDriverState *bs,
569 int64_t sector_num, uint8_t *buf, int nb_sectors,
570 BlockDriverCompletionFunc *cb, void *opaque)
571 {
572 BDRVRawState *s = bs->opaque;
573 RawAIOCB *acb;
574
575 if (fd_open(bs) < 0)
576 return NULL;
577
578 acb = qemu_aio_get(bs, cb, opaque);
579 if (!acb)
580 return NULL;
581 acb->aiocb.aio_fildes = s->fd;
582 acb->aiocb.ev_signo = SIGUSR2;
583 acb->aiocb.aio_buf = buf;
584 if (nb_sectors < 0)
585 acb->aiocb.aio_nbytes = -nb_sectors;
586 else
587 acb->aiocb.aio_nbytes = nb_sectors * 512;
588 acb->aiocb.aio_offset = sector_num * 512;
589 acb->next = posix_aio_state->first_aio;
590 posix_aio_state->first_aio = acb;
591 return acb;
592 }
593
594 static void raw_aio_em_cb(void* opaque)
595 {
596 RawAIOCB *acb = opaque;
597 acb->common.cb(acb->common.opaque, acb->ret);
598 qemu_aio_release(acb);
599 }
600
601 static void raw_aio_remove(RawAIOCB *acb)
602 {
603 RawAIOCB **pacb;
604
605 /* remove the callback from the queue */
606 pacb = &posix_aio_state->first_aio;
607 for(;;) {
608 if (*pacb == NULL) {
609 break;
610 } else if (*pacb == acb) {
611 *pacb = acb->next;
612 qemu_aio_release(acb);
613 break;
614 }
615 pacb = &acb->next;
616 }
617 }
618
619 static BlockDriverAIOCB *raw_aio_read(BlockDriverState *bs,
620 int64_t sector_num, uint8_t *buf, int nb_sectors,
621 BlockDriverCompletionFunc *cb, void *opaque)
622 {
623 RawAIOCB *acb;
624
625 /*
626 * If O_DIRECT is used and the buffer is not aligned fall back
627 * to synchronous IO.
628 */
629 BDRVRawState *s = bs->opaque;
630
631 if (unlikely(s->aligned_buf != NULL && ((uintptr_t) buf % 512))) {
632 QEMUBH *bh;
633 acb = qemu_aio_get(bs, cb, opaque);
634 acb->ret = raw_pread(bs, 512 * sector_num, buf, 512 * nb_sectors);
635 bh = qemu_bh_new(raw_aio_em_cb, acb);
636 qemu_bh_schedule(bh);
637 return &acb->common;
638 }
639
640 acb = raw_aio_setup(bs, sector_num, buf, nb_sectors, cb, opaque);
641 if (!acb)
642 return NULL;
643 if (qemu_paio_read(&acb->aiocb) < 0) {
644 raw_aio_remove(acb);
645 return NULL;
646 }
647 return &acb->common;
648 }
649
650 static BlockDriverAIOCB *raw_aio_write(BlockDriverState *bs,
651 int64_t sector_num, const uint8_t *buf, int nb_sectors,
652 BlockDriverCompletionFunc *cb, void *opaque)
653 {
654 RawAIOCB *acb;
655
656 /*
657 * If O_DIRECT is used and the buffer is not aligned fall back
658 * to synchronous IO.
659 */
660 BDRVRawState *s = bs->opaque;
661
662 if (unlikely(s->aligned_buf != NULL && ((uintptr_t) buf % 512))) {
663 QEMUBH *bh;
664 acb = qemu_aio_get(bs, cb, opaque);
665 acb->ret = raw_pwrite(bs, 512 * sector_num, buf, 512 * nb_sectors);
666 bh = qemu_bh_new(raw_aio_em_cb, acb);
667 qemu_bh_schedule(bh);
668 return &acb->common;
669 }
670
671 acb = raw_aio_setup(bs, sector_num, (uint8_t*)buf, nb_sectors, cb, opaque);
672 if (!acb)
673 return NULL;
674 if (qemu_paio_write(&acb->aiocb) < 0) {
675 raw_aio_remove(acb);
676 return NULL;
677 }
678 return &acb->common;
679 }
680
681 static void raw_aio_cancel(BlockDriverAIOCB *blockacb)
682 {
683 int ret;
684 RawAIOCB *acb = (RawAIOCB *)blockacb;
685
686 ret = qemu_paio_cancel(acb->aiocb.aio_fildes, &acb->aiocb);
687 if (ret == QEMU_PAIO_NOTCANCELED) {
688 /* fail safe: if the aio could not be canceled, we wait for
689 it */
690 while (qemu_paio_error(&acb->aiocb) == EINPROGRESS);
691 }
692
693 raw_aio_remove(acb);
694 }
695 #else /* CONFIG_AIO */
696 static int posix_aio_init(void)
697 {
698 return 0;
699 }
700 #endif /* CONFIG_AIO */
701
702
703 static void raw_close(BlockDriverState *bs)
704 {
705 BDRVRawState *s = bs->opaque;
706 if (s->fd >= 0) {
707 close(s->fd);
708 s->fd = -1;
709 if (s->aligned_buf != NULL)
710 qemu_free(s->aligned_buf);
711 }
712 }
713
714 static int raw_truncate(BlockDriverState *bs, int64_t offset)
715 {
716 BDRVRawState *s = bs->opaque;
717 if (s->type != FTYPE_FILE)
718 return -ENOTSUP;
719 if (ftruncate(s->fd, offset) < 0)
720 return -errno;
721 return 0;
722 }
723
724 #ifdef __OpenBSD__
725 static int64_t raw_getlength(BlockDriverState *bs)
726 {
727 BDRVRawState *s = bs->opaque;
728 int fd = s->fd;
729 struct stat st;
730
731 if (fstat(fd, &st))
732 return -1;
733 if (S_ISCHR(st.st_mode) || S_ISBLK(st.st_mode)) {
734 struct disklabel dl;
735
736 if (ioctl(fd, DIOCGDINFO, &dl))
737 return -1;
738 return (uint64_t)dl.d_secsize *
739 dl.d_partitions[DISKPART(st.st_rdev)].p_size;
740 } else
741 return st.st_size;
742 }
743 #else /* !__OpenBSD__ */
744 static int64_t raw_getlength(BlockDriverState *bs)
745 {
746 BDRVRawState *s = bs->opaque;
747 int fd = s->fd;
748 int64_t size;
749 #ifdef _BSD
750 struct stat sb;
751 #endif
752 #ifdef __sun__
753 struct dk_minfo minfo;
754 int rv;
755 #endif
756 int ret;
757
758 ret = fd_open(bs);
759 if (ret < 0)
760 return ret;
761
762 #ifdef _BSD
763 if (!fstat(fd, &sb) && (S_IFCHR & sb.st_mode)) {
764 #ifdef DIOCGMEDIASIZE
765 if (ioctl(fd, DIOCGMEDIASIZE, (off_t *)&size))
766 #endif
767 #ifdef CONFIG_COCOA
768 size = LONG_LONG_MAX;
769 #else
770 size = lseek(fd, 0LL, SEEK_END);
771 #endif
772 } else
773 #endif
774 #ifdef __sun__
775 /*
776 * use the DKIOCGMEDIAINFO ioctl to read the size.
777 */
778 rv = ioctl ( fd, DKIOCGMEDIAINFO, &minfo );
779 if ( rv != -1 ) {
780 size = minfo.dki_lbsize * minfo.dki_capacity;
781 } else /* there are reports that lseek on some devices
782 fails, but irc discussion said that contingency
783 on contingency was overkill */
784 #endif
785 {
786 size = lseek(fd, 0, SEEK_END);
787 }
788 return size;
789 }
790 #endif
791
792 static int raw_create(const char *filename, int64_t total_size,
793 const char *backing_file, int flags)
794 {
795 int fd;
796
797 if (flags || backing_file)
798 return -ENOTSUP;
799
800 fd = open(filename, O_WRONLY | O_CREAT | O_TRUNC | O_BINARY,
801 0644);
802 if (fd < 0)
803 return -EIO;
804 ftruncate(fd, total_size * 512);
805 close(fd);
806 return 0;
807 }
808
809 static void raw_flush(BlockDriverState *bs)
810 {
811 BDRVRawState *s = bs->opaque;
812 fsync(s->fd);
813 }
814
815 BlockDriver bdrv_raw = {
816 "raw",
817 sizeof(BDRVRawState),
818 NULL, /* no probe for protocols */
819 raw_open,
820 NULL,
821 NULL,
822 raw_close,
823 raw_create,
824 raw_flush,
825
826 #ifdef CONFIG_AIO
827 .bdrv_aio_read = raw_aio_read,
828 .bdrv_aio_write = raw_aio_write,
829 .bdrv_aio_cancel = raw_aio_cancel,
830 .aiocb_size = sizeof(RawAIOCB),
831 #endif
832
833 .bdrv_pread = raw_pread,
834 .bdrv_pwrite = raw_pwrite,
835 .bdrv_truncate = raw_truncate,
836 .bdrv_getlength = raw_getlength,
837 };
838
839 /***********************************************/
840 /* host device */
841
842 #ifdef CONFIG_COCOA
843 static kern_return_t FindEjectableCDMedia( io_iterator_t *mediaIterator );
844 static kern_return_t GetBSDPath( io_iterator_t mediaIterator, char *bsdPath, CFIndex maxPathSize );
845
846 kern_return_t FindEjectableCDMedia( io_iterator_t *mediaIterator )
847 {
848 kern_return_t kernResult;
849 mach_port_t masterPort;
850 CFMutableDictionaryRef classesToMatch;
851
852 kernResult = IOMasterPort( MACH_PORT_NULL, &masterPort );
853 if ( KERN_SUCCESS != kernResult ) {
854 printf( "IOMasterPort returned %d\n", kernResult );
855 }
856
857 classesToMatch = IOServiceMatching( kIOCDMediaClass );
858 if ( classesToMatch == NULL ) {
859 printf( "IOServiceMatching returned a NULL dictionary.\n" );
860 } else {
861 CFDictionarySetValue( classesToMatch, CFSTR( kIOMediaEjectableKey ), kCFBooleanTrue );
862 }
863 kernResult = IOServiceGetMatchingServices( masterPort, classesToMatch, mediaIterator );
864 if ( KERN_SUCCESS != kernResult )
865 {
866 printf( "IOServiceGetMatchingServices returned %d\n", kernResult );
867 }
868
869 return kernResult;
870 }
871
872 kern_return_t GetBSDPath( io_iterator_t mediaIterator, char *bsdPath, CFIndex maxPathSize )
873 {
874 io_object_t nextMedia;
875 kern_return_t kernResult = KERN_FAILURE;
876 *bsdPath = '\0';
877 nextMedia = IOIteratorNext( mediaIterator );
878 if ( nextMedia )
879 {
880 CFTypeRef bsdPathAsCFString;
881 bsdPathAsCFString = IORegistryEntryCreateCFProperty( nextMedia, CFSTR( kIOBSDNameKey ), kCFAllocatorDefault, 0 );
882 if ( bsdPathAsCFString ) {
883 size_t devPathLength;
884 strcpy( bsdPath, _PATH_DEV );
885 strcat( bsdPath, "r" );
886 devPathLength = strlen( bsdPath );
887 if ( CFStringGetCString( bsdPathAsCFString, bsdPath + devPathLength, maxPathSize - devPathLength, kCFStringEncodingASCII ) ) {
888 kernResult = KERN_SUCCESS;
889 }
890 CFRelease( bsdPathAsCFString );
891 }
892 IOObjectRelease( nextMedia );
893 }
894
895 return kernResult;
896 }
897
898 #endif
899
900 static int hdev_open(BlockDriverState *bs, const char *filename, int flags)
901 {
902 BDRVRawState *s = bs->opaque;
903 int fd, open_flags, ret;
904
905 posix_aio_init();
906
907 #ifdef CONFIG_COCOA
908 if (strstart(filename, "/dev/cdrom", NULL)) {
909 kern_return_t kernResult;
910 io_iterator_t mediaIterator;
911 char bsdPath[ MAXPATHLEN ];
912 int fd;
913
914 kernResult = FindEjectableCDMedia( &mediaIterator );
915 kernResult = GetBSDPath( mediaIterator, bsdPath, sizeof( bsdPath ) );
916
917 if ( bsdPath[ 0 ] != '\0' ) {
918 strcat(bsdPath,"s0");
919 /* some CDs don't have a partition 0 */
920 fd = open(bsdPath, O_RDONLY | O_BINARY | O_LARGEFILE);
921 if (fd < 0) {
922 bsdPath[strlen(bsdPath)-1] = '1';
923 } else {
924 close(fd);
925 }
926 filename = bsdPath;
927 }
928
929 if ( mediaIterator )
930 IOObjectRelease( mediaIterator );
931 }
932 #endif
933 open_flags = O_BINARY;
934 if ((flags & BDRV_O_ACCESS) == O_RDWR) {
935 open_flags |= O_RDWR;
936 } else {
937 open_flags |= O_RDONLY;
938 bs->read_only = 1;
939 }
940 /* Use O_DSYNC for write-through caching, no flags for write-back caching,
941 * and O_DIRECT for no caching. */
942 if ((flags & BDRV_O_NOCACHE))
943 open_flags |= O_DIRECT;
944 else if (!(flags & BDRV_O_CACHE_WB))
945 open_flags |= O_DSYNC;
946
947 s->type = FTYPE_FILE;
948 #if defined(__linux__)
949 if (strstart(filename, "/dev/cd", NULL)) {
950 /* open will not fail even if no CD is inserted */
951 open_flags |= O_NONBLOCK;
952 s->type = FTYPE_CD;
953 } else if (strstart(filename, "/dev/fd", NULL)) {
954 s->type = FTYPE_FD;
955 s->fd_open_flags = open_flags;
956 /* open will not fail even if no floppy is inserted */
957 open_flags |= O_NONBLOCK;
958 } else if (strstart(filename, "/dev/sg", NULL)) {
959 bs->sg = 1;
960 }
961 #endif
962 fd = open(filename, open_flags, 0644);
963 if (fd < 0) {
964 ret = -errno;
965 if (ret == -EROFS)
966 ret = -EACCES;
967 return ret;
968 }
969 s->fd = fd;
970 #if defined(__linux__)
971 /* close fd so that we can reopen it as needed */
972 if (s->type == FTYPE_FD) {
973 close(s->fd);
974 s->fd = -1;
975 s->fd_media_changed = 1;
976 }
977 #endif
978 return 0;
979 }
980
981 #if defined(__linux__)
982 /* Note: we do not have a reliable method to detect if the floppy is
983 present. The current method is to try to open the floppy at every
984 I/O and to keep it opened during a few hundreds of ms. */
985 static int fd_open(BlockDriverState *bs)
986 {
987 BDRVRawState *s = bs->opaque;
988 int last_media_present;
989
990 if (s->type != FTYPE_FD)
991 return 0;
992 last_media_present = (s->fd >= 0);
993 if (s->fd >= 0 &&
994 (qemu_get_clock(rt_clock) - s->fd_open_time) >= FD_OPEN_TIMEOUT) {
995 close(s->fd);
996 s->fd = -1;
997 #ifdef DEBUG_FLOPPY
998 printf("Floppy closed\n");
999 #endif
1000 }
1001 if (s->fd < 0) {
1002 if (s->fd_got_error &&
1003 (qemu_get_clock(rt_clock) - s->fd_error_time) < FD_OPEN_TIMEOUT) {
1004 #ifdef DEBUG_FLOPPY
1005 printf("No floppy (open delayed)\n");
1006 #endif
1007 return -EIO;
1008 }
1009 s->fd = open(bs->filename, s->fd_open_flags);
1010 if (s->fd < 0) {
1011 s->fd_error_time = qemu_get_clock(rt_clock);
1012 s->fd_got_error = 1;
1013 if (last_media_present)
1014 s->fd_media_changed = 1;
1015 #ifdef DEBUG_FLOPPY
1016 printf("No floppy\n");
1017 #endif
1018 return -EIO;
1019 }
1020 #ifdef DEBUG_FLOPPY
1021 printf("Floppy opened\n");
1022 #endif
1023 }
1024 if (!last_media_present)
1025 s->fd_media_changed = 1;
1026 s->fd_open_time = qemu_get_clock(rt_clock);
1027 s->fd_got_error = 0;
1028 return 0;
1029 }
1030
1031 static int raw_is_inserted(BlockDriverState *bs)
1032 {
1033 BDRVRawState *s = bs->opaque;
1034 int ret;
1035
1036 switch(s->type) {
1037 case FTYPE_CD:
1038 ret = ioctl(s->fd, CDROM_DRIVE_STATUS, CDSL_CURRENT);
1039 if (ret == CDS_DISC_OK)
1040 return 1;
1041 else
1042 return 0;
1043 break;
1044 case FTYPE_FD:
1045 ret = fd_open(bs);
1046 return (ret >= 0);
1047 default:
1048 return 1;
1049 }
1050 }
1051
1052 /* currently only used by fdc.c, but a CD version would be good too */
1053 static int raw_media_changed(BlockDriverState *bs)
1054 {
1055 BDRVRawState *s = bs->opaque;
1056
1057 switch(s->type) {
1058 case FTYPE_FD:
1059 {
1060 int ret;
1061 /* XXX: we do not have a true media changed indication. It
1062 does not work if the floppy is changed without trying
1063 to read it */
1064 fd_open(bs);
1065 ret = s->fd_media_changed;
1066 s->fd_media_changed = 0;
1067 #ifdef DEBUG_FLOPPY
1068 printf("Floppy changed=%d\n", ret);
1069 #endif
1070 return ret;
1071 }
1072 default:
1073 return -ENOTSUP;
1074 }
1075 }
1076
1077 static int raw_eject(BlockDriverState *bs, int eject_flag)
1078 {
1079 BDRVRawState *s = bs->opaque;
1080
1081 switch(s->type) {
1082 case FTYPE_CD:
1083 if (eject_flag) {
1084 if (ioctl (s->fd, CDROMEJECT, NULL) < 0)
1085 perror("CDROMEJECT");
1086 } else {
1087 if (ioctl (s->fd, CDROMCLOSETRAY, NULL) < 0)
1088 perror("CDROMEJECT");
1089 }
1090 break;
1091 case FTYPE_FD:
1092 {
1093 int fd;
1094 if (s->fd >= 0) {
1095 close(s->fd);
1096 s->fd = -1;
1097 }
1098 fd = open(bs->filename, s->fd_open_flags | O_NONBLOCK);
1099 if (fd >= 0) {
1100 if (ioctl(fd, FDEJECT, 0) < 0)
1101 perror("FDEJECT");
1102 close(fd);
1103 }
1104 }
1105 break;
1106 default:
1107 return -ENOTSUP;
1108 }
1109 return 0;
1110 }
1111
1112 static int raw_set_locked(BlockDriverState *bs, int locked)
1113 {
1114 BDRVRawState *s = bs->opaque;
1115
1116 switch(s->type) {
1117 case FTYPE_CD:
1118 if (ioctl (s->fd, CDROM_LOCKDOOR, locked) < 0) {
1119 /* Note: an error can happen if the distribution automatically
1120 mounts the CD-ROM */
1121 // perror("CDROM_LOCKDOOR");
1122 }
1123 break;
1124 default:
1125 return -ENOTSUP;
1126 }
1127 return 0;
1128 }
1129
1130 static int raw_ioctl(BlockDriverState *bs, unsigned long int req, void *buf)
1131 {
1132 BDRVRawState *s = bs->opaque;
1133
1134 return ioctl(s->fd, req, buf);
1135 }
1136 #else
1137
1138 static int fd_open(BlockDriverState *bs)
1139 {
1140 return 0;
1141 }
1142
1143 static int raw_is_inserted(BlockDriverState *bs)
1144 {
1145 return 1;
1146 }
1147
1148 static int raw_media_changed(BlockDriverState *bs)
1149 {
1150 return -ENOTSUP;
1151 }
1152
1153 static int raw_eject(BlockDriverState *bs, int eject_flag)
1154 {
1155 return -ENOTSUP;
1156 }
1157
1158 static int raw_set_locked(BlockDriverState *bs, int locked)
1159 {
1160 return -ENOTSUP;
1161 }
1162
1163 static int raw_ioctl(BlockDriverState *bs, unsigned long int req, void *buf)
1164 {
1165 return -ENOTSUP;
1166 }
1167 #endif /* !linux */
1168
1169 BlockDriver bdrv_host_device = {
1170 "host_device",
1171 sizeof(BDRVRawState),
1172 NULL, /* no probe for protocols */
1173 hdev_open,
1174 NULL,
1175 NULL,
1176 raw_close,
1177 NULL,
1178 raw_flush,
1179
1180 #ifdef CONFIG_AIO
1181 .bdrv_aio_read = raw_aio_read,
1182 .bdrv_aio_write = raw_aio_write,
1183 .bdrv_aio_cancel = raw_aio_cancel,
1184 .aiocb_size = sizeof(RawAIOCB),
1185 #endif
1186
1187 .bdrv_pread = raw_pread,
1188 .bdrv_pwrite = raw_pwrite,
1189 .bdrv_getlength = raw_getlength,
1190
1191 /* removable device support */
1192 .bdrv_is_inserted = raw_is_inserted,
1193 .bdrv_media_changed = raw_media_changed,
1194 .bdrv_eject = raw_eject,
1195 .bdrv_set_locked = raw_set_locked,
1196 /* generic scsi device */
1197 .bdrv_ioctl = raw_ioctl,
1198 };