]> git.proxmox.com Git - qemu.git/blame - block-raw-posix.c
configure sensitive to user locale
[qemu.git] / block-raw-posix.c
CommitLineData
83f64091 1/*
223d4670 2 * Block driver for RAW files (posix)
5fafdf24 3 *
83f64091 4 * Copyright (c) 2006 Fabrice Bellard
5fafdf24 5 *
83f64091
FB
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 */
faf07963 24#include "qemu-common.h"
87ecb68b 25#include "qemu-timer.h"
baf35cb9 26#include "qemu-char.h"
83f64091
FB
27#include "block_int.h"
28#include <assert.h>
414f0dab 29#ifdef CONFIG_AIO
3c529d93 30#include "posix-aio-compat.h"
414f0dab 31#endif
83f64091 32
83f64091
FB
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__
2e9671da
TS
46#define _POSIX_PTHREAD_SEMANTICS 1
47#include <signal.h>
83f64091
FB
48#include <sys/dkio.h>
49#endif
19cb3738
FB
50#ifdef __linux__
51#include <sys/ioctl.h>
52#include <linux/cdrom.h>
53#include <linux/fd.h>
54#endif
1cb6c3fd 55#ifdef __FreeBSD__
543952ca 56#include <signal.h>
1cb6c3fd
TS
57#include <sys/disk.h>
58#endif
83f64091 59
128ab2ff
BS
60#ifdef __OpenBSD__
61#include <sys/ioctl.h>
62#include <sys/disklabel.h>
63#include <sys/dkio.h>
64#endif
65
c5e97233
BS
66#ifdef __DragonFly__
67#include <sys/ioctl.h>
68#include <sys/diskslice.h>
69#endif
70
19cb3738 71//#define DEBUG_FLOPPY
83f64091 72
faf07963 73//#define DEBUG_BLOCK
03ff3ca3 74#if defined(DEBUG_BLOCK)
93fcfe39 75#define DEBUG_BLOCK_PRINT(formatCstr, args...) do { if (qemu_log_enabled()) \
31b1a7b4 76 { qemu_log(formatCstr, ##args); qemu_log_flush(); } } while (0)
8c05dbf9
TS
77#else
78#define DEBUG_BLOCK_PRINT(formatCstr, args...)
79#endif
80
f6465578
AL
81/* OS X does not have O_DSYNC */
82#ifndef O_DSYNC
7ab064d2 83#define O_DSYNC O_SYNC
f6465578
AL
84#endif
85
9f7965c7
AL
86/* Approximate O_DIRECT with O_DSYNC if O_DIRECT isn't available */
87#ifndef O_DIRECT
88#define O_DIRECT O_DSYNC
89#endif
90
19cb3738
FB
91#define FTYPE_FILE 0
92#define FTYPE_CD 1
93#define FTYPE_FD 2
83f64091 94
bed5cc52
FB
95#define ALIGNED_BUFFER_SIZE (32 * 512)
96
19cb3738
FB
97/* if the FD is not accessed during that time (in ms), we try to
98 reopen it to see if the disk has been changed */
99#define FD_OPEN_TIMEOUT 1000
83f64091 100
19cb3738
FB
101typedef struct BDRVRawState {
102 int fd;
103 int type;
8c05dbf9 104 unsigned int lseek_err_cnt;
19cb3738
FB
105#if defined(__linux__)
106 /* linux floppy specific */
6dd2db52 107 int fd_open_flags;
19cb3738
FB
108 int64_t fd_open_time;
109 int64_t fd_error_time;
110 int fd_got_error;
111 int fd_media_changed;
83f64091 112#endif
bed5cc52 113 uint8_t* aligned_buf;
19cb3738
FB
114} BDRVRawState;
115
a76bab49
AL
116static int posix_aio_init(void);
117
19cb3738 118static int fd_open(BlockDriverState *bs);
83f64091
FB
119
120static int raw_open(BlockDriverState *bs, const char *filename, int flags)
121{
122 BDRVRawState *s = bs->opaque;
19cb3738 123 int fd, open_flags, ret;
83f64091 124
a76bab49
AL
125 posix_aio_init();
126
8c05dbf9
TS
127 s->lseek_err_cnt = 0;
128
83f64091
FB
129 open_flags = O_BINARY;
130 if ((flags & BDRV_O_ACCESS) == O_RDWR) {
131 open_flags |= O_RDWR;
132 } else {
133 open_flags |= O_RDONLY;
134 bs->read_only = 1;
135 }
136 if (flags & BDRV_O_CREAT)
137 open_flags |= O_CREAT | O_TRUNC;
9f7965c7
AL
138
139 /* Use O_DSYNC for write-through caching, no flags for write-back caching,
140 * and O_DIRECT for no caching. */
141 if ((flags & BDRV_O_NOCACHE))
33f00271 142 open_flags |= O_DIRECT;
9f7965c7
AL
143 else if (!(flags & BDRV_O_CACHE_WB))
144 open_flags |= O_DSYNC;
83f64091 145
19cb3738
FB
146 s->type = FTYPE_FILE;
147
83f64091 148 fd = open(filename, open_flags, 0644);
19cb3738
FB
149 if (fd < 0) {
150 ret = -errno;
151 if (ret == -EROFS)
152 ret = -EACCES;
153 return ret;
154 }
83f64091 155 s->fd = fd;
bed5cc52 156 s->aligned_buf = NULL;
9f7965c7 157 if ((flags & BDRV_O_NOCACHE)) {
bed5cc52
FB
158 s->aligned_buf = qemu_memalign(512, ALIGNED_BUFFER_SIZE);
159 if (s->aligned_buf == NULL) {
160 ret = -errno;
161 close(fd);
162 return ret;
163 }
164 }
83f64091
FB
165 return 0;
166}
167
168/* XXX: use host sector size if necessary with:
169#ifdef DIOCGSECTORSIZE
170 {
171 unsigned int sectorsize = 512;
172 if (!ioctl(fd, DIOCGSECTORSIZE, &sectorsize) &&
173 sectorsize > bufsize)
174 bufsize = sectorsize;
175 }
176#endif
177#ifdef CONFIG_COCOA
178 u_int32_t blockSize = 512;
179 if ( !ioctl( fd, DKIOCGETBLOCKSIZE, &blockSize ) && blockSize > bufsize) {
180 bufsize = blockSize;
181 }
182#endif
183*/
184
bed5cc52
FB
185/*
186 * offset and count are in bytes, but must be multiples of 512 for files
187 * opened with O_DIRECT. buf must be aligned to 512 bytes then.
188 *
189 * This function may be called without alignment if the caller ensures
190 * that O_DIRECT is not in effect.
191 */
192static int raw_pread_aligned(BlockDriverState *bs, int64_t offset,
83f64091
FB
193 uint8_t *buf, int count)
194{
195 BDRVRawState *s = bs->opaque;
196 int ret;
3b46e624 197
19cb3738
FB
198 ret = fd_open(bs);
199 if (ret < 0)
200 return ret;
201
985a03b0 202 if (offset >= 0 && lseek(s->fd, offset, SEEK_SET) == (off_t)-1) {
8c05dbf9
TS
203 ++(s->lseek_err_cnt);
204 if(s->lseek_err_cnt <= 10) {
92868412
JM
205 DEBUG_BLOCK_PRINT("raw_pread(%d:%s, %" PRId64 ", %p, %d) [%" PRId64
206 "] lseek failed : %d = %s\n",
8c05dbf9
TS
207 s->fd, bs->filename, offset, buf, count,
208 bs->total_sectors, errno, strerror(errno));
209 }
210 return -1;
211 }
212 s->lseek_err_cnt=0;
213
83f64091 214 ret = read(s->fd, buf, count);
8c05dbf9
TS
215 if (ret == count)
216 goto label__raw_read__success;
217
92868412
JM
218 DEBUG_BLOCK_PRINT("raw_pread(%d:%s, %" PRId64 ", %p, %d) [%" PRId64
219 "] read failed %d : %d = %s\n",
8c05dbf9
TS
220 s->fd, bs->filename, offset, buf, count,
221 bs->total_sectors, ret, errno, strerror(errno));
222
223 /* Try harder for CDrom. */
224 if (bs->type == BDRV_TYPE_CDROM) {
225 lseek(s->fd, offset, SEEK_SET);
226 ret = read(s->fd, buf, count);
227 if (ret == count)
228 goto label__raw_read__success;
229 lseek(s->fd, offset, SEEK_SET);
230 ret = read(s->fd, buf, count);
231 if (ret == count)
232 goto label__raw_read__success;
233
92868412
JM
234 DEBUG_BLOCK_PRINT("raw_pread(%d:%s, %" PRId64 ", %p, %d) [%" PRId64
235 "] retry read failed %d : %d = %s\n",
8c05dbf9
TS
236 s->fd, bs->filename, offset, buf, count,
237 bs->total_sectors, ret, errno, strerror(errno));
238 }
239
8c05dbf9
TS
240label__raw_read__success:
241
83f64091
FB
242 return ret;
243}
244
bed5cc52
FB
245/*
246 * offset and count are in bytes, but must be multiples of 512 for files
247 * opened with O_DIRECT. buf must be aligned to 512 bytes then.
248 *
249 * This function may be called without alignment if the caller ensures
250 * that O_DIRECT is not in effect.
251 */
252static int raw_pwrite_aligned(BlockDriverState *bs, int64_t offset,
83f64091
FB
253 const uint8_t *buf, int count)
254{
255 BDRVRawState *s = bs->opaque;
256 int ret;
3b46e624 257
19cb3738
FB
258 ret = fd_open(bs);
259 if (ret < 0)
4141d4c2 260 return -errno;
19cb3738 261
985a03b0 262 if (offset >= 0 && lseek(s->fd, offset, SEEK_SET) == (off_t)-1) {
8c05dbf9
TS
263 ++(s->lseek_err_cnt);
264 if(s->lseek_err_cnt) {
92868412
JM
265 DEBUG_BLOCK_PRINT("raw_pwrite(%d:%s, %" PRId64 ", %p, %d) [%"
266 PRId64 "] lseek failed : %d = %s\n",
8c05dbf9
TS
267 s->fd, bs->filename, offset, buf, count,
268 bs->total_sectors, errno, strerror(errno));
269 }
4141d4c2 270 return -EIO;
8c05dbf9
TS
271 }
272 s->lseek_err_cnt = 0;
273
83f64091 274 ret = write(s->fd, buf, count);
8c05dbf9
TS
275 if (ret == count)
276 goto label__raw_write__success;
277
92868412
JM
278 DEBUG_BLOCK_PRINT("raw_pwrite(%d:%s, %" PRId64 ", %p, %d) [%" PRId64
279 "] write failed %d : %d = %s\n",
8c05dbf9
TS
280 s->fd, bs->filename, offset, buf, count,
281 bs->total_sectors, ret, errno, strerror(errno));
282
8c05dbf9
TS
283label__raw_write__success:
284
4141d4c2 285 return (ret < 0) ? -errno : ret;
83f64091
FB
286}
287
bed5cc52 288
bed5cc52
FB
289/*
290 * offset and count are in bytes and possibly not aligned. For files opened
291 * with O_DIRECT, necessary alignments are ensured before calling
292 * raw_pread_aligned to do the actual read.
293 */
294static int raw_pread(BlockDriverState *bs, int64_t offset,
295 uint8_t *buf, int count)
296{
297 BDRVRawState *s = bs->opaque;
298 int size, ret, shift, sum;
299
300 sum = 0;
301
302 if (s->aligned_buf != NULL) {
303
304 if (offset & 0x1ff) {
305 /* align offset on a 512 bytes boundary */
306
307 shift = offset & 0x1ff;
308 size = (shift + count + 0x1ff) & ~0x1ff;
309 if (size > ALIGNED_BUFFER_SIZE)
310 size = ALIGNED_BUFFER_SIZE;
311 ret = raw_pread_aligned(bs, offset - shift, s->aligned_buf, size);
312 if (ret < 0)
313 return ret;
314
315 size = 512 - shift;
316 if (size > count)
317 size = count;
318 memcpy(buf, s->aligned_buf + shift, size);
319
320 buf += size;
321 offset += size;
322 count -= size;
323 sum += size;
324
325 if (count == 0)
326 return sum;
327 }
328 if (count & 0x1ff || (uintptr_t) buf & 0x1ff) {
329
330 /* read on aligned buffer */
331
332 while (count) {
333
334 size = (count + 0x1ff) & ~0x1ff;
335 if (size > ALIGNED_BUFFER_SIZE)
336 size = ALIGNED_BUFFER_SIZE;
337
338 ret = raw_pread_aligned(bs, offset, s->aligned_buf, size);
339 if (ret < 0)
340 return ret;
341
342 size = ret;
343 if (size > count)
344 size = count;
345
346 memcpy(buf, s->aligned_buf, size);
347
348 buf += size;
349 offset += size;
350 count -= size;
351 sum += size;
352 }
353
354 return sum;
355 }
356 }
357
358 return raw_pread_aligned(bs, offset, buf, count) + sum;
359}
360
361/*
362 * offset and count are in bytes and possibly not aligned. For files opened
363 * with O_DIRECT, necessary alignments are ensured before calling
364 * raw_pwrite_aligned to do the actual write.
365 */
366static int raw_pwrite(BlockDriverState *bs, int64_t offset,
367 const uint8_t *buf, int count)
368{
369 BDRVRawState *s = bs->opaque;
370 int size, ret, shift, sum;
371
372 sum = 0;
373
374 if (s->aligned_buf != NULL) {
375
376 if (offset & 0x1ff) {
377 /* align offset on a 512 bytes boundary */
378 shift = offset & 0x1ff;
379 ret = raw_pread_aligned(bs, offset - shift, s->aligned_buf, 512);
380 if (ret < 0)
381 return ret;
382
383 size = 512 - shift;
384 if (size > count)
385 size = count;
386 memcpy(s->aligned_buf + shift, buf, size);
387
388 ret = raw_pwrite_aligned(bs, offset - shift, s->aligned_buf, 512);
389 if (ret < 0)
390 return ret;
391
392 buf += size;
393 offset += size;
394 count -= size;
395 sum += size;
396
397 if (count == 0)
398 return sum;
399 }
400 if (count & 0x1ff || (uintptr_t) buf & 0x1ff) {
401
402 while ((size = (count & ~0x1ff)) != 0) {
403
404 if (size > ALIGNED_BUFFER_SIZE)
405 size = ALIGNED_BUFFER_SIZE;
406
407 memcpy(s->aligned_buf, buf, size);
408
409 ret = raw_pwrite_aligned(bs, offset, s->aligned_buf, size);
410 if (ret < 0)
411 return ret;
412
413 buf += ret;
414 offset += ret;
415 count -= ret;
416 sum += ret;
417 }
418 /* here, count < 512 because (count & ~0x1ff) == 0 */
419 if (count) {
420 ret = raw_pread_aligned(bs, offset, s->aligned_buf, 512);
421 if (ret < 0)
422 return ret;
423 memcpy(s->aligned_buf, buf, count);
424
425 ret = raw_pwrite_aligned(bs, offset, s->aligned_buf, 512);
426 if (ret < 0)
427 return ret;
428 if (count < ret)
429 ret = count;
430
431 sum += ret;
432 }
433 return sum;
434 }
435 }
436 return raw_pwrite_aligned(bs, offset, buf, count) + sum;
437}
438
414f0dab 439#ifdef CONFIG_AIO
83f64091 440/***********************************************************/
19cb3738 441/* Unix AIO using POSIX AIO */
83f64091
FB
442
443typedef struct RawAIOCB {
ce1a14dc 444 BlockDriverAIOCB common;
3c529d93 445 struct qemu_paiocb aiocb;
ce1a14dc 446 struct RawAIOCB *next;
bed5cc52 447 int ret;
83f64091
FB
448} RawAIOCB;
449
a76bab49
AL
450typedef struct PosixAioState
451{
9e472e10 452 int rfd, wfd;
a76bab49
AL
453 RawAIOCB *first_aio;
454} PosixAioState;
83f64091 455
a76bab49 456static void posix_aio_read(void *opaque)
83f64091 457{
a76bab49 458 PosixAioState *s = opaque;
ce1a14dc 459 RawAIOCB *acb, **pacb;
83f64091 460 int ret;
9e472e10
AL
461 ssize_t len;
462
e20e830b
AL
463 /* read all bytes from signal pipe */
464 for (;;) {
465 char bytes[16];
9e472e10 466
e20e830b 467 len = read(s->rfd, bytes, sizeof(bytes));
2c41a5f9 468 if (len == -1 && errno == EINTR)
e20e830b
AL
469 continue; /* try again */
470 if (len == sizeof(bytes))
471 continue; /* more to read */
472 break;
473 }
83f64091
FB
474
475 for(;;) {
a76bab49 476 pacb = &s->first_aio;
83f64091
FB
477 for(;;) {
478 acb = *pacb;
479 if (!acb)
480 goto the_end;
3c529d93 481 ret = qemu_paio_error(&acb->aiocb);
83f64091
FB
482 if (ret == ECANCELED) {
483 /* remove the request */
ce1a14dc
PB
484 *pacb = acb->next;
485 qemu_aio_release(acb);
83f64091
FB
486 } else if (ret != EINPROGRESS) {
487 /* end of aio */
488 if (ret == 0) {
3c529d93 489 ret = qemu_paio_return(&acb->aiocb);
ce1a14dc 490 if (ret == acb->aiocb.aio_nbytes)
83f64091
FB
491 ret = 0;
492 else
19cb3738 493 ret = -EINVAL;
83f64091
FB
494 } else {
495 ret = -ret;
496 }
497 /* remove the request */
ce1a14dc 498 *pacb = acb->next;
83f64091 499 /* call the callback */
ce1a14dc
PB
500 acb->common.cb(acb->common.opaque, ret);
501 qemu_aio_release(acb);
83f64091
FB
502 break;
503 } else {
ce1a14dc 504 pacb = &acb->next;
83f64091
FB
505 }
506 }
507 }
508 the_end: ;
509}
510
a76bab49 511static int posix_aio_flush(void *opaque)
baf35cb9 512{
a76bab49
AL
513 PosixAioState *s = opaque;
514 return !!s->first_aio;
515}
516
517static PosixAioState *posix_aio_state;
baf35cb9 518
9e472e10
AL
519static void aio_signal_handler(int signum)
520{
521 if (posix_aio_state) {
522 char byte = 0;
523
524 write(posix_aio_state->wfd, &byte, sizeof(byte));
525 }
526
527 qemu_service_io();
528}
529
a76bab49
AL
530static int posix_aio_init(void)
531{
9e472e10 532 struct sigaction act;
a76bab49 533 PosixAioState *s;
9e472e10 534 int fds[2];
3c529d93 535 struct qemu_paioinit ai;
a76bab49
AL
536
537 if (posix_aio_state)
538 return 0;
ad02ad6f 539
a76bab49 540 s = qemu_malloc(sizeof(PosixAioState));
baf35cb9 541
9e472e10
AL
542 sigfillset(&act.sa_mask);
543 act.sa_flags = 0; /* do not restart syscalls to interrupt select() */
544 act.sa_handler = aio_signal_handler;
545 sigaction(SIGUSR2, &act, NULL);
546
a76bab49 547 s->first_aio = NULL;
9e472e10
AL
548 if (pipe(fds) == -1) {
549 fprintf(stderr, "failed to create pipe\n");
27463101
AL
550 return -errno;
551 }
2c41a5f9 552
9e472e10
AL
553 s->rfd = fds[0];
554 s->wfd = fds[1];
555
e20e830b 556 fcntl(s->rfd, F_SETFL, O_NONBLOCK);
9e472e10 557 fcntl(s->wfd, F_SETFL, O_NONBLOCK);
2c41a5f9 558
9e472e10 559 qemu_aio_set_fd_handler(s->rfd, posix_aio_read, NULL, posix_aio_flush, s);
baf35cb9 560
3c529d93
AL
561 memset(&ai, 0, sizeof(ai));
562 ai.aio_threads = 64;
563 ai.aio_num = 64;
564 qemu_paio_init(&ai);
acce87f9 565
a76bab49 566 posix_aio_state = s;
6eb5733a 567
a76bab49 568 return 0;
83f64091
FB
569}
570
ce1a14dc
PB
571static RawAIOCB *raw_aio_setup(BlockDriverState *bs,
572 int64_t sector_num, uint8_t *buf, int nb_sectors,
573 BlockDriverCompletionFunc *cb, void *opaque)
83f64091 574{
ce1a14dc
PB
575 BDRVRawState *s = bs->opaque;
576 RawAIOCB *acb;
577
19cb3738
FB
578 if (fd_open(bs) < 0)
579 return NULL;
580
ce1a14dc
PB
581 acb = qemu_aio_get(bs, cb, opaque);
582 if (!acb)
583 return NULL;
3c529d93 584 acb->aiocb.aio_fildes = s->fd;
55f11ca3 585 acb->aiocb.ev_signo = SIGUSR2;
ce1a14dc 586 acb->aiocb.aio_buf = buf;
985a03b0
TS
587 if (nb_sectors < 0)
588 acb->aiocb.aio_nbytes = -nb_sectors;
589 else
590 acb->aiocb.aio_nbytes = nb_sectors * 512;
ce1a14dc 591 acb->aiocb.aio_offset = sector_num * 512;
a76bab49
AL
592 acb->next = posix_aio_state->first_aio;
593 posix_aio_state->first_aio = acb;
ce1a14dc 594 return acb;
83f64091
FB
595}
596
bed5cc52
FB
597static void raw_aio_em_cb(void* opaque)
598{
599 RawAIOCB *acb = opaque;
600 acb->common.cb(acb->common.opaque, acb->ret);
601 qemu_aio_release(acb);
602}
bed5cc52 603
22bf1458
AL
604static void raw_aio_remove(RawAIOCB *acb)
605{
606 RawAIOCB **pacb;
607
608 /* remove the callback from the queue */
609 pacb = &posix_aio_state->first_aio;
610 for(;;) {
611 if (*pacb == NULL) {
7a11b22e 612 fprintf(stderr, "raw_aio_remove: aio request not found!\n");
22bf1458
AL
613 break;
614 } else if (*pacb == acb) {
615 *pacb = acb->next;
616 qemu_aio_release(acb);
617 break;
618 }
7a11b22e 619 pacb = &(*pacb)->next;
22bf1458
AL
620 }
621}
622
ce1a14dc
PB
623static BlockDriverAIOCB *raw_aio_read(BlockDriverState *bs,
624 int64_t sector_num, uint8_t *buf, int nb_sectors,
625 BlockDriverCompletionFunc *cb, void *opaque)
83f64091 626{
ce1a14dc 627 RawAIOCB *acb;
83f64091 628
bed5cc52
FB
629 /*
630 * If O_DIRECT is used and the buffer is not aligned fall back
631 * to synchronous IO.
632 */
bed5cc52
FB
633 BDRVRawState *s = bs->opaque;
634
635 if (unlikely(s->aligned_buf != NULL && ((uintptr_t) buf % 512))) {
636 QEMUBH *bh;
637 acb = qemu_aio_get(bs, cb, opaque);
638 acb->ret = raw_pread(bs, 512 * sector_num, buf, 512 * nb_sectors);
639 bh = qemu_bh_new(raw_aio_em_cb, acb);
640 qemu_bh_schedule(bh);
641 return &acb->common;
642 }
bed5cc52 643
ce1a14dc
PB
644 acb = raw_aio_setup(bs, sector_num, buf, nb_sectors, cb, opaque);
645 if (!acb)
646 return NULL;
3c529d93 647 if (qemu_paio_read(&acb->aiocb) < 0) {
22bf1458 648 raw_aio_remove(acb);
ce1a14dc 649 return NULL;
5fafdf24 650 }
ce1a14dc 651 return &acb->common;
83f64091
FB
652}
653
ce1a14dc
PB
654static BlockDriverAIOCB *raw_aio_write(BlockDriverState *bs,
655 int64_t sector_num, const uint8_t *buf, int nb_sectors,
656 BlockDriverCompletionFunc *cb, void *opaque)
83f64091 657{
ce1a14dc 658 RawAIOCB *acb;
83f64091 659
bed5cc52
FB
660 /*
661 * If O_DIRECT is used and the buffer is not aligned fall back
662 * to synchronous IO.
663 */
bed5cc52
FB
664 BDRVRawState *s = bs->opaque;
665
666 if (unlikely(s->aligned_buf != NULL && ((uintptr_t) buf % 512))) {
667 QEMUBH *bh;
668 acb = qemu_aio_get(bs, cb, opaque);
669 acb->ret = raw_pwrite(bs, 512 * sector_num, buf, 512 * nb_sectors);
670 bh = qemu_bh_new(raw_aio_em_cb, acb);
671 qemu_bh_schedule(bh);
672 return &acb->common;
673 }
bed5cc52 674
ce1a14dc
PB
675 acb = raw_aio_setup(bs, sector_num, (uint8_t*)buf, nb_sectors, cb, opaque);
676 if (!acb)
677 return NULL;
3c529d93 678 if (qemu_paio_write(&acb->aiocb) < 0) {
22bf1458 679 raw_aio_remove(acb);
ce1a14dc 680 return NULL;
5fafdf24 681 }
ce1a14dc 682 return &acb->common;
83f64091
FB
683}
684
ce1a14dc 685static void raw_aio_cancel(BlockDriverAIOCB *blockacb)
83f64091 686{
83f64091 687 int ret;
ce1a14dc 688 RawAIOCB *acb = (RawAIOCB *)blockacb;
83f64091 689
3c529d93
AL
690 ret = qemu_paio_cancel(acb->aiocb.aio_fildes, &acb->aiocb);
691 if (ret == QEMU_PAIO_NOTCANCELED) {
83f64091
FB
692 /* fail safe: if the aio could not be canceled, we wait for
693 it */
3c529d93 694 while (qemu_paio_error(&acb->aiocb) == EINPROGRESS);
83f64091
FB
695 }
696
22bf1458 697 raw_aio_remove(acb);
83f64091 698}
a76bab49
AL
699#else /* CONFIG_AIO */
700static int posix_aio_init(void)
414f0dab 701{
674a24ac 702 return 0;
414f0dab 703}
414f0dab
BS
704#endif /* CONFIG_AIO */
705
53538725 706
83f64091
FB
707static void raw_close(BlockDriverState *bs)
708{
709 BDRVRawState *s = bs->opaque;
19cb3738
FB
710 if (s->fd >= 0) {
711 close(s->fd);
712 s->fd = -1;
bed5cc52
FB
713 if (s->aligned_buf != NULL)
714 qemu_free(s->aligned_buf);
19cb3738 715 }
83f64091
FB
716}
717
718static int raw_truncate(BlockDriverState *bs, int64_t offset)
719{
720 BDRVRawState *s = bs->opaque;
19cb3738
FB
721 if (s->type != FTYPE_FILE)
722 return -ENOTSUP;
83f64091
FB
723 if (ftruncate(s->fd, offset) < 0)
724 return -errno;
725 return 0;
726}
727
128ab2ff
BS
728#ifdef __OpenBSD__
729static int64_t raw_getlength(BlockDriverState *bs)
730{
731 BDRVRawState *s = bs->opaque;
732 int fd = s->fd;
733 struct stat st;
734
735 if (fstat(fd, &st))
736 return -1;
737 if (S_ISCHR(st.st_mode) || S_ISBLK(st.st_mode)) {
738 struct disklabel dl;
739
740 if (ioctl(fd, DIOCGDINFO, &dl))
741 return -1;
742 return (uint64_t)dl.d_secsize *
743 dl.d_partitions[DISKPART(st.st_rdev)].p_size;
744 } else
745 return st.st_size;
746}
747#else /* !__OpenBSD__ */
83f64091
FB
748static int64_t raw_getlength(BlockDriverState *bs)
749{
750 BDRVRawState *s = bs->opaque;
751 int fd = s->fd;
752 int64_t size;
179a2c19 753#ifdef HOST_BSD
83f64091
FB
754 struct stat sb;
755#endif
756#ifdef __sun__
757 struct dk_minfo minfo;
758 int rv;
759#endif
19cb3738
FB
760 int ret;
761
762 ret = fd_open(bs);
763 if (ret < 0)
764 return ret;
83f64091 765
179a2c19 766#ifdef HOST_BSD
83f64091
FB
767 if (!fstat(fd, &sb) && (S_IFCHR & sb.st_mode)) {
768#ifdef DIOCGMEDIASIZE
769 if (ioctl(fd, DIOCGMEDIASIZE, (off_t *)&size))
c5e97233
BS
770#elif defined(DIOCGPART)
771 {
772 struct partinfo pi;
773 if (ioctl(fd, DIOCGPART, &pi) == 0)
774 size = pi.media_size;
775 else
776 size = 0;
777 }
778 if (size == 0)
83f64091
FB
779#endif
780#ifdef CONFIG_COCOA
781 size = LONG_LONG_MAX;
782#else
783 size = lseek(fd, 0LL, SEEK_END);
784#endif
785 } else
786#endif
787#ifdef __sun__
788 /*
789 * use the DKIOCGMEDIAINFO ioctl to read the size.
790 */
791 rv = ioctl ( fd, DKIOCGMEDIAINFO, &minfo );
792 if ( rv != -1 ) {
793 size = minfo.dki_lbsize * minfo.dki_capacity;
794 } else /* there are reports that lseek on some devices
795 fails, but irc discussion said that contingency
796 on contingency was overkill */
797#endif
798 {
799 size = lseek(fd, 0, SEEK_END);
800 }
83f64091
FB
801 return size;
802}
128ab2ff 803#endif
83f64091
FB
804
805static int raw_create(const char *filename, int64_t total_size,
806 const char *backing_file, int flags)
807{
808 int fd;
809
810 if (flags || backing_file)
811 return -ENOTSUP;
812
5fafdf24 813 fd = open(filename, O_WRONLY | O_CREAT | O_TRUNC | O_BINARY,
83f64091
FB
814 0644);
815 if (fd < 0)
816 return -EIO;
817 ftruncate(fd, total_size * 512);
818 close(fd);
819 return 0;
820}
821
822static void raw_flush(BlockDriverState *bs)
823{
824 BDRVRawState *s = bs->opaque;
825 fsync(s->fd);
826}
827
828BlockDriver bdrv_raw = {
829 "raw",
830 sizeof(BDRVRawState),
831 NULL, /* no probe for protocols */
832 raw_open,
833 NULL,
834 NULL,
835 raw_close,
836 raw_create,
837 raw_flush,
3b46e624 838
414f0dab 839#ifdef CONFIG_AIO
83f64091
FB
840 .bdrv_aio_read = raw_aio_read,
841 .bdrv_aio_write = raw_aio_write,
842 .bdrv_aio_cancel = raw_aio_cancel,
ce1a14dc 843 .aiocb_size = sizeof(RawAIOCB),
414f0dab 844#endif
3c529d93 845
83f64091
FB
846 .bdrv_pread = raw_pread,
847 .bdrv_pwrite = raw_pwrite,
848 .bdrv_truncate = raw_truncate,
849 .bdrv_getlength = raw_getlength,
850};
851
19cb3738
FB
852/***********************************************/
853/* host device */
854
855#ifdef CONFIG_COCOA
856static kern_return_t FindEjectableCDMedia( io_iterator_t *mediaIterator );
857static kern_return_t GetBSDPath( io_iterator_t mediaIterator, char *bsdPath, CFIndex maxPathSize );
858
859kern_return_t FindEjectableCDMedia( io_iterator_t *mediaIterator )
860{
5fafdf24 861 kern_return_t kernResult;
19cb3738
FB
862 mach_port_t masterPort;
863 CFMutableDictionaryRef classesToMatch;
864
865 kernResult = IOMasterPort( MACH_PORT_NULL, &masterPort );
866 if ( KERN_SUCCESS != kernResult ) {
867 printf( "IOMasterPort returned %d\n", kernResult );
868 }
3b46e624 869
5fafdf24 870 classesToMatch = IOServiceMatching( kIOCDMediaClass );
19cb3738
FB
871 if ( classesToMatch == NULL ) {
872 printf( "IOServiceMatching returned a NULL dictionary.\n" );
873 } else {
874 CFDictionarySetValue( classesToMatch, CFSTR( kIOMediaEjectableKey ), kCFBooleanTrue );
875 }
876 kernResult = IOServiceGetMatchingServices( masterPort, classesToMatch, mediaIterator );
877 if ( KERN_SUCCESS != kernResult )
878 {
879 printf( "IOServiceGetMatchingServices returned %d\n", kernResult );
880 }
3b46e624 881
19cb3738
FB
882 return kernResult;
883}
884
885kern_return_t GetBSDPath( io_iterator_t mediaIterator, char *bsdPath, CFIndex maxPathSize )
886{
887 io_object_t nextMedia;
888 kern_return_t kernResult = KERN_FAILURE;
889 *bsdPath = '\0';
890 nextMedia = IOIteratorNext( mediaIterator );
891 if ( nextMedia )
892 {
893 CFTypeRef bsdPathAsCFString;
894 bsdPathAsCFString = IORegistryEntryCreateCFProperty( nextMedia, CFSTR( kIOBSDNameKey ), kCFAllocatorDefault, 0 );
895 if ( bsdPathAsCFString ) {
896 size_t devPathLength;
897 strcpy( bsdPath, _PATH_DEV );
898 strcat( bsdPath, "r" );
899 devPathLength = strlen( bsdPath );
900 if ( CFStringGetCString( bsdPathAsCFString, bsdPath + devPathLength, maxPathSize - devPathLength, kCFStringEncodingASCII ) ) {
901 kernResult = KERN_SUCCESS;
902 }
903 CFRelease( bsdPathAsCFString );
904 }
905 IOObjectRelease( nextMedia );
906 }
3b46e624 907
19cb3738
FB
908 return kernResult;
909}
910
911#endif
912
913static int hdev_open(BlockDriverState *bs, const char *filename, int flags)
914{
915 BDRVRawState *s = bs->opaque;
3c529d93 916 int fd, open_flags, ret;
19cb3738 917
a76bab49
AL
918 posix_aio_init();
919
19cb3738
FB
920#ifdef CONFIG_COCOA
921 if (strstart(filename, "/dev/cdrom", NULL)) {
922 kern_return_t kernResult;
923 io_iterator_t mediaIterator;
924 char bsdPath[ MAXPATHLEN ];
925 int fd;
5fafdf24 926
19cb3738
FB
927 kernResult = FindEjectableCDMedia( &mediaIterator );
928 kernResult = GetBSDPath( mediaIterator, bsdPath, sizeof( bsdPath ) );
3b46e624 929
19cb3738
FB
930 if ( bsdPath[ 0 ] != '\0' ) {
931 strcat(bsdPath,"s0");
932 /* some CDs don't have a partition 0 */
933 fd = open(bsdPath, O_RDONLY | O_BINARY | O_LARGEFILE);
934 if (fd < 0) {
935 bsdPath[strlen(bsdPath)-1] = '1';
936 } else {
937 close(fd);
938 }
939 filename = bsdPath;
940 }
3b46e624 941
19cb3738
FB
942 if ( mediaIterator )
943 IOObjectRelease( mediaIterator );
944 }
945#endif
946 open_flags = O_BINARY;
947 if ((flags & BDRV_O_ACCESS) == O_RDWR) {
948 open_flags |= O_RDWR;
949 } else {
950 open_flags |= O_RDONLY;
951 bs->read_only = 1;
952 }
9f7965c7
AL
953 /* Use O_DSYNC for write-through caching, no flags for write-back caching,
954 * and O_DIRECT for no caching. */
955 if ((flags & BDRV_O_NOCACHE))
33f00271 956 open_flags |= O_DIRECT;
9f7965c7
AL
957 else if (!(flags & BDRV_O_CACHE_WB))
958 open_flags |= O_DSYNC;
19cb3738
FB
959
960 s->type = FTYPE_FILE;
961#if defined(__linux__)
962 if (strstart(filename, "/dev/cd", NULL)) {
963 /* open will not fail even if no CD is inserted */
964 open_flags |= O_NONBLOCK;
965 s->type = FTYPE_CD;
966 } else if (strstart(filename, "/dev/fd", NULL)) {
967 s->type = FTYPE_FD;
6dd2db52 968 s->fd_open_flags = open_flags;
19cb3738
FB
969 /* open will not fail even if no floppy is inserted */
970 open_flags |= O_NONBLOCK;
985a03b0
TS
971 } else if (strstart(filename, "/dev/sg", NULL)) {
972 bs->sg = 1;
19cb3738
FB
973 }
974#endif
975 fd = open(filename, open_flags, 0644);
976 if (fd < 0) {
977 ret = -errno;
978 if (ret == -EROFS)
979 ret = -EACCES;
980 return ret;
981 }
982 s->fd = fd;
983#if defined(__linux__)
984 /* close fd so that we can reopen it as needed */
985 if (s->type == FTYPE_FD) {
986 close(s->fd);
987 s->fd = -1;
988 s->fd_media_changed = 1;
989 }
990#endif
991 return 0;
992}
993
03ff3ca3 994#if defined(__linux__)
19cb3738
FB
995/* Note: we do not have a reliable method to detect if the floppy is
996 present. The current method is to try to open the floppy at every
997 I/O and to keep it opened during a few hundreds of ms. */
998static int fd_open(BlockDriverState *bs)
999{
1000 BDRVRawState *s = bs->opaque;
1001 int last_media_present;
1002
1003 if (s->type != FTYPE_FD)
1004 return 0;
1005 last_media_present = (s->fd >= 0);
5fafdf24 1006 if (s->fd >= 0 &&
19cb3738
FB
1007 (qemu_get_clock(rt_clock) - s->fd_open_time) >= FD_OPEN_TIMEOUT) {
1008 close(s->fd);
1009 s->fd = -1;
1010#ifdef DEBUG_FLOPPY
1011 printf("Floppy closed\n");
1012#endif
1013 }
1014 if (s->fd < 0) {
5fafdf24 1015 if (s->fd_got_error &&
19cb3738
FB
1016 (qemu_get_clock(rt_clock) - s->fd_error_time) < FD_OPEN_TIMEOUT) {
1017#ifdef DEBUG_FLOPPY
1018 printf("No floppy (open delayed)\n");
1019#endif
1020 return -EIO;
1021 }
6dd2db52 1022 s->fd = open(bs->filename, s->fd_open_flags);
19cb3738
FB
1023 if (s->fd < 0) {
1024 s->fd_error_time = qemu_get_clock(rt_clock);
1025 s->fd_got_error = 1;
1026 if (last_media_present)
1027 s->fd_media_changed = 1;
1028#ifdef DEBUG_FLOPPY
1029 printf("No floppy\n");
1030#endif
1031 return -EIO;
1032 }
1033#ifdef DEBUG_FLOPPY
1034 printf("Floppy opened\n");
1035#endif
1036 }
1037 if (!last_media_present)
1038 s->fd_media_changed = 1;
1039 s->fd_open_time = qemu_get_clock(rt_clock);
1040 s->fd_got_error = 0;
1041 return 0;
1042}
19cb3738
FB
1043
1044static int raw_is_inserted(BlockDriverState *bs)
1045{
1046 BDRVRawState *s = bs->opaque;
1047 int ret;
1048
1049 switch(s->type) {
1050 case FTYPE_CD:
1051 ret = ioctl(s->fd, CDROM_DRIVE_STATUS, CDSL_CURRENT);
1052 if (ret == CDS_DISC_OK)
1053 return 1;
1054 else
1055 return 0;
1056 break;
1057 case FTYPE_FD:
1058 ret = fd_open(bs);
1059 return (ret >= 0);
1060 default:
1061 return 1;
1062 }
1063}
1064
1065/* currently only used by fdc.c, but a CD version would be good too */
1066static int raw_media_changed(BlockDriverState *bs)
1067{
1068 BDRVRawState *s = bs->opaque;
1069
1070 switch(s->type) {
1071 case FTYPE_FD:
1072 {
1073 int ret;
1074 /* XXX: we do not have a true media changed indication. It
1075 does not work if the floppy is changed without trying
1076 to read it */
1077 fd_open(bs);
1078 ret = s->fd_media_changed;
1079 s->fd_media_changed = 0;
1080#ifdef DEBUG_FLOPPY
1081 printf("Floppy changed=%d\n", ret);
1082#endif
1083 return ret;
1084 }
1085 default:
1086 return -ENOTSUP;
1087 }
1088}
1089
1090static int raw_eject(BlockDriverState *bs, int eject_flag)
1091{
1092 BDRVRawState *s = bs->opaque;
1093
1094 switch(s->type) {
1095 case FTYPE_CD:
1096 if (eject_flag) {
1097 if (ioctl (s->fd, CDROMEJECT, NULL) < 0)
1098 perror("CDROMEJECT");
1099 } else {
1100 if (ioctl (s->fd, CDROMCLOSETRAY, NULL) < 0)
1101 perror("CDROMEJECT");
1102 }
1103 break;
1104 case FTYPE_FD:
1105 {
1106 int fd;
1107 if (s->fd >= 0) {
1108 close(s->fd);
1109 s->fd = -1;
1110 }
6dd2db52 1111 fd = open(bs->filename, s->fd_open_flags | O_NONBLOCK);
19cb3738
FB
1112 if (fd >= 0) {
1113 if (ioctl(fd, FDEJECT, 0) < 0)
1114 perror("FDEJECT");
1115 close(fd);
1116 }
1117 }
1118 break;
1119 default:
1120 return -ENOTSUP;
1121 }
1122 return 0;
1123}
1124
1125static int raw_set_locked(BlockDriverState *bs, int locked)
1126{
1127 BDRVRawState *s = bs->opaque;
1128
1129 switch(s->type) {
1130 case FTYPE_CD:
1131 if (ioctl (s->fd, CDROM_LOCKDOOR, locked) < 0) {
1132 /* Note: an error can happen if the distribution automatically
1133 mounts the CD-ROM */
1134 // perror("CDROM_LOCKDOOR");
1135 }
1136 break;
1137 default:
1138 return -ENOTSUP;
1139 }
1140 return 0;
1141}
1142
985a03b0
TS
1143static int raw_ioctl(BlockDriverState *bs, unsigned long int req, void *buf)
1144{
1145 BDRVRawState *s = bs->opaque;
1146
1147 return ioctl(s->fd, req, buf);
1148}
19cb3738
FB
1149#else
1150
08af02e2
AL
1151static int fd_open(BlockDriverState *bs)
1152{
1153 return 0;
1154}
1155
19cb3738
FB
1156static int raw_is_inserted(BlockDriverState *bs)
1157{
1158 return 1;
1159}
1160
1161static int raw_media_changed(BlockDriverState *bs)
1162{
1163 return -ENOTSUP;
1164}
1165
1166static int raw_eject(BlockDriverState *bs, int eject_flag)
1167{
1168 return -ENOTSUP;
1169}
1170
1171static int raw_set_locked(BlockDriverState *bs, int locked)
1172{
1173 return -ENOTSUP;
1174}
1175
985a03b0
TS
1176static int raw_ioctl(BlockDriverState *bs, unsigned long int req, void *buf)
1177{
1178 return -ENOTSUP;
1179}
19cb3738
FB
1180#endif /* !linux */
1181
1182BlockDriver bdrv_host_device = {
e60f469c
AJ
1183 .format_name = "host_device",
1184 .instance_size = sizeof(BDRVRawState),
1185 .bdrv_open = hdev_open,
1186 .bdrv_close = raw_close,
1187 .bdrv_flush = raw_flush,
3b46e624 1188
414f0dab 1189#ifdef CONFIG_AIO
e60f469c
AJ
1190 .bdrv_aio_read = raw_aio_read,
1191 .bdrv_aio_write = raw_aio_write,
1192 .bdrv_aio_cancel = raw_aio_cancel,
1193 .aiocb_size = sizeof(RawAIOCB),
414f0dab 1194#endif
3c529d93 1195
e60f469c
AJ
1196 .bdrv_pread = raw_pread,
1197 .bdrv_pwrite = raw_pwrite,
1198 .bdrv_getlength = raw_getlength,
19cb3738
FB
1199
1200 /* removable device support */
e60f469c
AJ
1201 .bdrv_is_inserted = raw_is_inserted,
1202 .bdrv_media_changed = raw_media_changed,
1203 .bdrv_eject = raw_eject,
1204 .bdrv_set_locked = raw_set_locked,
985a03b0 1205 /* generic scsi device */
e60f469c 1206 .bdrv_ioctl = raw_ioctl,
19cb3738 1207};