]>
Commit | Line | Data |
---|---|---|
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" |
d49b6836 | 25 | #include "qemu/error-report.h" |
1de7afc9 PB |
26 | #include "qemu/timer.h" |
27 | #include "qemu/log.h" | |
737e150e | 28 | #include "block/block_int.h" |
1de7afc9 | 29 | #include "qemu/module.h" |
de81a169 | 30 | #include "trace.h" |
737e150e | 31 | #include "block/thread-pool.h" |
1de7afc9 | 32 | #include "qemu/iov.h" |
9f8540ec | 33 | #include "raw-aio.h" |
06247428 | 34 | #include "qapi/util.h" |
d49b6836 | 35 | #include "qapi/qmp/qstring.h" |
83f64091 | 36 | |
83affaa6 | 37 | #if defined(__APPLE__) && (__MACH__) |
83f64091 FB |
38 | #include <paths.h> |
39 | #include <sys/param.h> | |
40 | #include <IOKit/IOKitLib.h> | |
41 | #include <IOKit/IOBSD.h> | |
42 | #include <IOKit/storage/IOMediaBSDClient.h> | |
43 | #include <IOKit/storage/IOMedia.h> | |
44 | #include <IOKit/storage/IOCDMedia.h> | |
45 | //#include <IOKit/storage/IOCDTypes.h> | |
46 | #include <CoreFoundation/CoreFoundation.h> | |
47 | #endif | |
48 | ||
49 | #ifdef __sun__ | |
2e9671da | 50 | #define _POSIX_PTHREAD_SEMANTICS 1 |
83f64091 FB |
51 | #include <sys/dkio.h> |
52 | #endif | |
19cb3738 | 53 | #ifdef __linux__ |
343f8568 JS |
54 | #include <sys/types.h> |
55 | #include <sys/stat.h> | |
19cb3738 | 56 | #include <sys/ioctl.h> |
05acda4d | 57 | #include <sys/param.h> |
19cb3738 FB |
58 | #include <linux/cdrom.h> |
59 | #include <linux/fd.h> | |
5500316d | 60 | #include <linux/fs.h> |
1a9335e4 ET |
61 | #include <linux/hdreg.h> |
62 | #ifdef __s390__ | |
63 | #include <asm/dasd.h> | |
64 | #endif | |
4ab15590 CL |
65 | #ifndef FS_NOCOW_FL |
66 | #define FS_NOCOW_FL 0x00800000 /* Do not cow file */ | |
67 | #endif | |
5500316d | 68 | #endif |
b953f075 | 69 | #if defined(CONFIG_FALLOCATE_PUNCH_HOLE) || defined(CONFIG_FALLOCATE_ZERO_RANGE) |
3d4fa43e KK |
70 | #include <linux/falloc.h> |
71 | #endif | |
a167ba50 | 72 | #if defined (__FreeBSD__) || defined(__FreeBSD_kernel__) |
1cb6c3fd | 73 | #include <sys/disk.h> |
9f23011a | 74 | #include <sys/cdio.h> |
1cb6c3fd | 75 | #endif |
83f64091 | 76 | |
128ab2ff BS |
77 | #ifdef __OpenBSD__ |
78 | #include <sys/ioctl.h> | |
79 | #include <sys/disklabel.h> | |
80 | #include <sys/dkio.h> | |
81 | #endif | |
82 | ||
d1f6fd8d CE |
83 | #ifdef __NetBSD__ |
84 | #include <sys/ioctl.h> | |
85 | #include <sys/disklabel.h> | |
86 | #include <sys/dkio.h> | |
87 | #include <sys/disk.h> | |
88 | #endif | |
89 | ||
c5e97233 BS |
90 | #ifdef __DragonFly__ |
91 | #include <sys/ioctl.h> | |
92 | #include <sys/diskslice.h> | |
93 | #endif | |
94 | ||
dce512de CH |
95 | #ifdef CONFIG_XFS |
96 | #include <xfs/xfs.h> | |
97 | #endif | |
98 | ||
faf07963 | 99 | //#define DEBUG_BLOCK |
bcb22555 DA |
100 | |
101 | #ifdef DEBUG_BLOCK | |
102 | # define DEBUG_BLOCK_PRINT 1 | |
8c05dbf9 | 103 | #else |
bcb22555 | 104 | # define DEBUG_BLOCK_PRINT 0 |
8c05dbf9 | 105 | #endif |
bcb22555 DA |
106 | #define DPRINTF(fmt, ...) \ |
107 | do { \ | |
108 | if (DEBUG_BLOCK_PRINT) { \ | |
109 | printf(fmt, ## __VA_ARGS__); \ | |
110 | } \ | |
111 | } while (0) | |
8c05dbf9 | 112 | |
f6465578 AL |
113 | /* OS X does not have O_DSYNC */ |
114 | #ifndef O_DSYNC | |
1c27a8b3 | 115 | #ifdef O_SYNC |
7ab064d2 | 116 | #define O_DSYNC O_SYNC |
1c27a8b3 JA |
117 | #elif defined(O_FSYNC) |
118 | #define O_DSYNC O_FSYNC | |
119 | #endif | |
f6465578 AL |
120 | #endif |
121 | ||
9f7965c7 AL |
122 | /* Approximate O_DIRECT with O_DSYNC if O_DIRECT isn't available */ |
123 | #ifndef O_DIRECT | |
124 | #define O_DIRECT O_DSYNC | |
125 | #endif | |
126 | ||
19cb3738 FB |
127 | #define FTYPE_FILE 0 |
128 | #define FTYPE_CD 1 | |
129 | #define FTYPE_FD 2 | |
83f64091 | 130 | |
c57c846a | 131 | /* if the FD is not accessed during that time (in ns), we try to |
19cb3738 | 132 | reopen it to see if the disk has been changed */ |
c57c846a | 133 | #define FD_OPEN_TIMEOUT (1000000000) |
83f64091 | 134 | |
581b9e29 CH |
135 | #define MAX_BLOCKSIZE 4096 |
136 | ||
19cb3738 FB |
137 | typedef struct BDRVRawState { |
138 | int fd; | |
139 | int type; | |
0e1d8f4c | 140 | int open_flags; |
c25f53b0 PB |
141 | size_t buf_align; |
142 | ||
19cb3738 FB |
143 | #if defined(__linux__) |
144 | /* linux floppy specific */ | |
19cb3738 FB |
145 | int64_t fd_open_time; |
146 | int64_t fd_error_time; | |
147 | int fd_got_error; | |
148 | int fd_media_changed; | |
83f64091 | 149 | #endif |
e44bd6fc | 150 | #ifdef CONFIG_LINUX_AIO |
5c6c3a6c | 151 | int use_aio; |
1e5b9d2f | 152 | void *aio_ctx; |
e44bd6fc | 153 | #endif |
dce512de | 154 | #ifdef CONFIG_XFS |
260a82e5 | 155 | bool is_xfs:1; |
dce512de | 156 | #endif |
260a82e5 | 157 | bool has_discard:1; |
97a2ae34 | 158 | bool has_write_zeroes:1; |
260a82e5 | 159 | bool discard_zeroes:1; |
d50d8222 | 160 | bool has_fallocate; |
3cad8307 | 161 | bool needs_alignment; |
19cb3738 FB |
162 | } BDRVRawState; |
163 | ||
eeb6b45d JC |
164 | typedef struct BDRVRawReopenState { |
165 | int fd; | |
166 | int open_flags; | |
167 | #ifdef CONFIG_LINUX_AIO | |
168 | int use_aio; | |
169 | #endif | |
170 | } BDRVRawReopenState; | |
171 | ||
19cb3738 | 172 | static int fd_open(BlockDriverState *bs); |
22afa7b5 | 173 | static int64_t raw_getlength(BlockDriverState *bs); |
83f64091 | 174 | |
de81a169 PB |
175 | typedef struct RawPosixAIOData { |
176 | BlockDriverState *bs; | |
177 | int aio_fildes; | |
178 | union { | |
179 | struct iovec *aio_iov; | |
180 | void *aio_ioctl_buf; | |
181 | }; | |
182 | int aio_niov; | |
8238010b | 183 | uint64_t aio_nbytes; |
de81a169 PB |
184 | #define aio_ioctl_cmd aio_nbytes /* for QEMU_AIO_IOCTL */ |
185 | off_t aio_offset; | |
186 | int aio_type; | |
187 | } RawPosixAIOData; | |
188 | ||
a167ba50 | 189 | #if defined(__FreeBSD__) || defined(__FreeBSD_kernel__) |
f3a5d3f8 | 190 | static int cdrom_reopen(BlockDriverState *bs); |
9f23011a BS |
191 | #endif |
192 | ||
1de1ae0a CE |
193 | #if defined(__NetBSD__) |
194 | static int raw_normalize_devicepath(const char **filename) | |
195 | { | |
196 | static char namebuf[PATH_MAX]; | |
197 | const char *dp, *fname; | |
198 | struct stat sb; | |
199 | ||
200 | fname = *filename; | |
201 | dp = strrchr(fname, '/'); | |
202 | if (lstat(fname, &sb) < 0) { | |
203 | fprintf(stderr, "%s: stat failed: %s\n", | |
204 | fname, strerror(errno)); | |
205 | return -errno; | |
206 | } | |
207 | ||
208 | if (!S_ISBLK(sb.st_mode)) { | |
209 | return 0; | |
210 | } | |
211 | ||
212 | if (dp == NULL) { | |
213 | snprintf(namebuf, PATH_MAX, "r%s", fname); | |
214 | } else { | |
215 | snprintf(namebuf, PATH_MAX, "%.*s/r%s", | |
216 | (int)(dp - fname), fname, dp + 1); | |
217 | } | |
218 | fprintf(stderr, "%s is a block device", fname); | |
219 | *filename = namebuf; | |
220 | fprintf(stderr, ", using %s\n", *filename); | |
221 | ||
222 | return 0; | |
223 | } | |
224 | #else | |
225 | static int raw_normalize_devicepath(const char **filename) | |
226 | { | |
227 | return 0; | |
228 | } | |
229 | #endif | |
230 | ||
8a4ed0d1 ET |
231 | /* |
232 | * Get logical block size via ioctl. On success store it in @sector_size_p. | |
233 | */ | |
234 | static int probe_logical_blocksize(int fd, unsigned int *sector_size_p) | |
c25f53b0 | 235 | { |
c25f53b0 | 236 | unsigned int sector_size; |
8a4ed0d1 | 237 | bool success = false; |
c25f53b0 | 238 | |
8a4ed0d1 | 239 | errno = ENOTSUP; |
c25f53b0 PB |
240 | |
241 | /* Try a few ioctls to get the right size */ | |
c25f53b0 | 242 | #ifdef BLKSSZGET |
df26a350 | 243 | if (ioctl(fd, BLKSSZGET, §or_size) >= 0) { |
8a4ed0d1 ET |
244 | *sector_size_p = sector_size; |
245 | success = true; | |
c25f53b0 PB |
246 | } |
247 | #endif | |
248 | #ifdef DKIOCGETBLOCKSIZE | |
df26a350 | 249 | if (ioctl(fd, DKIOCGETBLOCKSIZE, §or_size) >= 0) { |
8a4ed0d1 ET |
250 | *sector_size_p = sector_size; |
251 | success = true; | |
c25f53b0 PB |
252 | } |
253 | #endif | |
254 | #ifdef DIOCGSECTORSIZE | |
df26a350 | 255 | if (ioctl(fd, DIOCGSECTORSIZE, §or_size) >= 0) { |
8a4ed0d1 ET |
256 | *sector_size_p = sector_size; |
257 | success = true; | |
c25f53b0 PB |
258 | } |
259 | #endif | |
8a4ed0d1 ET |
260 | |
261 | return success ? 0 : -errno; | |
262 | } | |
263 | ||
1a9335e4 ET |
264 | /** |
265 | * Get physical block size of @fd. | |
266 | * On success, store it in @blk_size and return 0. | |
267 | * On failure, return -errno. | |
268 | */ | |
269 | static int probe_physical_blocksize(int fd, unsigned int *blk_size) | |
270 | { | |
271 | #ifdef BLKPBSZGET | |
272 | if (ioctl(fd, BLKPBSZGET, blk_size) < 0) { | |
273 | return -errno; | |
274 | } | |
275 | return 0; | |
276 | #else | |
277 | return -ENOTSUP; | |
278 | #endif | |
279 | } | |
280 | ||
22d182e8 SH |
281 | /* Check if read is allowed with given memory buffer and length. |
282 | * | |
283 | * This function is used to check O_DIRECT memory buffer and request alignment. | |
284 | */ | |
285 | static bool raw_is_io_aligned(int fd, void *buf, size_t len) | |
286 | { | |
287 | ssize_t ret = pread(fd, buf, len, 0); | |
288 | ||
289 | if (ret >= 0) { | |
290 | return true; | |
291 | } | |
292 | ||
293 | #ifdef __linux__ | |
294 | /* The Linux kernel returns EINVAL for misaligned O_DIRECT reads. Ignore | |
295 | * other errors (e.g. real I/O error), which could happen on a failed | |
296 | * drive, since we only care about probing alignment. | |
297 | */ | |
298 | if (errno != EINVAL) { | |
299 | return true; | |
300 | } | |
301 | #endif | |
302 | ||
303 | return false; | |
304 | } | |
305 | ||
8a4ed0d1 ET |
306 | static void raw_probe_alignment(BlockDriverState *bs, int fd, Error **errp) |
307 | { | |
308 | BDRVRawState *s = bs->opaque; | |
309 | char *buf; | |
459b4e66 | 310 | size_t max_align = MAX(MAX_BLOCKSIZE, getpagesize()); |
8a4ed0d1 | 311 | |
b192af8a | 312 | /* For SCSI generic devices the alignment is not really used. |
8a4ed0d1 | 313 | With buffered I/O, we don't have any restrictions. */ |
b192af8a | 314 | if (bdrv_is_sg(bs) || !s->needs_alignment) { |
8a4ed0d1 ET |
315 | bs->request_alignment = 1; |
316 | s->buf_align = 1; | |
317 | return; | |
318 | } | |
319 | ||
320 | bs->request_alignment = 0; | |
321 | s->buf_align = 0; | |
322 | /* Let's try to use the logical blocksize for the alignment. */ | |
323 | if (probe_logical_blocksize(fd, &bs->request_alignment) < 0) { | |
324 | bs->request_alignment = 0; | |
325 | } | |
c25f53b0 PB |
326 | #ifdef CONFIG_XFS |
327 | if (s->is_xfs) { | |
328 | struct dioattr da; | |
df26a350 | 329 | if (xfsctl(NULL, fd, XFS_IOC_DIOINFO, &da) >= 0) { |
c25f53b0 PB |
330 | bs->request_alignment = da.d_miniosz; |
331 | /* The kernel returns wrong information for d_mem */ | |
332 | /* s->buf_align = da.d_mem; */ | |
333 | } | |
334 | } | |
335 | #endif | |
336 | ||
337 | /* If we could not get the sizes so far, we can only guess them */ | |
338 | if (!s->buf_align) { | |
339 | size_t align; | |
459b4e66 DL |
340 | buf = qemu_memalign(max_align, 2 * max_align); |
341 | for (align = 512; align <= max_align; align <<= 1) { | |
342 | if (raw_is_io_aligned(fd, buf + align, max_align)) { | |
c25f53b0 PB |
343 | s->buf_align = align; |
344 | break; | |
345 | } | |
346 | } | |
347 | qemu_vfree(buf); | |
348 | } | |
349 | ||
350 | if (!bs->request_alignment) { | |
351 | size_t align; | |
459b4e66 DL |
352 | buf = qemu_memalign(s->buf_align, max_align); |
353 | for (align = 512; align <= max_align; align <<= 1) { | |
22d182e8 | 354 | if (raw_is_io_aligned(fd, buf, align)) { |
c25f53b0 PB |
355 | bs->request_alignment = align; |
356 | break; | |
357 | } | |
358 | } | |
359 | qemu_vfree(buf); | |
360 | } | |
df26a350 KW |
361 | |
362 | if (!s->buf_align || !bs->request_alignment) { | |
363 | error_setg(errp, "Could not find working O_DIRECT alignment. " | |
364 | "Try cache.direct=off."); | |
365 | } | |
c25f53b0 PB |
366 | } |
367 | ||
6a8dc042 JC |
368 | static void raw_parse_flags(int bdrv_flags, int *open_flags) |
369 | { | |
370 | assert(open_flags != NULL); | |
371 | ||
372 | *open_flags |= O_BINARY; | |
373 | *open_flags &= ~O_ACCMODE; | |
374 | if (bdrv_flags & BDRV_O_RDWR) { | |
375 | *open_flags |= O_RDWR; | |
376 | } else { | |
377 | *open_flags |= O_RDONLY; | |
378 | } | |
379 | ||
380 | /* Use O_DSYNC for write-through caching, no flags for write-back caching, | |
381 | * and O_DIRECT for no caching. */ | |
382 | if ((bdrv_flags & BDRV_O_NOCACHE)) { | |
383 | *open_flags |= O_DIRECT; | |
384 | } | |
6a8dc042 JC |
385 | } |
386 | ||
c2f3426c SH |
387 | static void raw_detach_aio_context(BlockDriverState *bs) |
388 | { | |
389 | #ifdef CONFIG_LINUX_AIO | |
390 | BDRVRawState *s = bs->opaque; | |
391 | ||
392 | if (s->use_aio) { | |
393 | laio_detach_aio_context(s->aio_ctx, bdrv_get_aio_context(bs)); | |
394 | } | |
395 | #endif | |
396 | } | |
397 | ||
398 | static void raw_attach_aio_context(BlockDriverState *bs, | |
399 | AioContext *new_context) | |
400 | { | |
401 | #ifdef CONFIG_LINUX_AIO | |
402 | BDRVRawState *s = bs->opaque; | |
403 | ||
404 | if (s->use_aio) { | |
405 | laio_attach_aio_context(s->aio_ctx, new_context); | |
406 | } | |
407 | #endif | |
408 | } | |
409 | ||
fc32a72d JC |
410 | #ifdef CONFIG_LINUX_AIO |
411 | static int raw_set_aio(void **aio_ctx, int *use_aio, int bdrv_flags) | |
412 | { | |
413 | int ret = -1; | |
414 | assert(aio_ctx != NULL); | |
415 | assert(use_aio != NULL); | |
416 | /* | |
417 | * Currently Linux do AIO only for files opened with O_DIRECT | |
418 | * specified so check NOCACHE flag too | |
419 | */ | |
420 | if ((bdrv_flags & (BDRV_O_NOCACHE|BDRV_O_NATIVE_AIO)) == | |
421 | (BDRV_O_NOCACHE|BDRV_O_NATIVE_AIO)) { | |
422 | ||
423 | /* if non-NULL, laio_init() has already been run */ | |
424 | if (*aio_ctx == NULL) { | |
425 | *aio_ctx = laio_init(); | |
426 | if (!*aio_ctx) { | |
427 | goto error; | |
428 | } | |
429 | } | |
430 | *use_aio = 1; | |
431 | } else { | |
432 | *use_aio = 0; | |
433 | } | |
434 | ||
435 | ret = 0; | |
436 | ||
437 | error: | |
438 | return ret; | |
439 | } | |
440 | #endif | |
441 | ||
078896a9 HR |
442 | static void raw_parse_filename(const char *filename, QDict *options, |
443 | Error **errp) | |
444 | { | |
445 | /* The filename does not have to be prefixed by the protocol name, since | |
446 | * "file" is the default protocol; therefore, the return value of this | |
447 | * function call can be ignored. */ | |
448 | strstart(filename, "file:", &filename); | |
449 | ||
450 | qdict_put_obj(options, "filename", QOBJECT(qstring_from_str(filename))); | |
451 | } | |
452 | ||
c66a6157 KW |
453 | static QemuOptsList raw_runtime_opts = { |
454 | .name = "raw", | |
455 | .head = QTAILQ_HEAD_INITIALIZER(raw_runtime_opts.head), | |
456 | .desc = { | |
457 | { | |
458 | .name = "filename", | |
459 | .type = QEMU_OPT_STRING, | |
460 | .help = "File name of the image", | |
461 | }, | |
462 | { /* end of list */ } | |
463 | }, | |
464 | }; | |
465 | ||
466 | static int raw_open_common(BlockDriverState *bs, QDict *options, | |
e428e439 | 467 | int bdrv_flags, int open_flags, Error **errp) |
83f64091 FB |
468 | { |
469 | BDRVRawState *s = bs->opaque; | |
c66a6157 KW |
470 | QemuOpts *opts; |
471 | Error *local_err = NULL; | |
8bfea15d | 472 | const char *filename = NULL; |
0e1d8f4c | 473 | int fd, ret; |
260a82e5 | 474 | struct stat st; |
83f64091 | 475 | |
87ea75d5 | 476 | opts = qemu_opts_create(&raw_runtime_opts, NULL, 0, &error_abort); |
c66a6157 | 477 | qemu_opts_absorb_qdict(opts, options, &local_err); |
84d18f06 | 478 | if (local_err) { |
e428e439 | 479 | error_propagate(errp, local_err); |
c66a6157 KW |
480 | ret = -EINVAL; |
481 | goto fail; | |
482 | } | |
483 | ||
484 | filename = qemu_opt_get(opts, "filename"); | |
485 | ||
1de1ae0a CE |
486 | ret = raw_normalize_devicepath(&filename); |
487 | if (ret != 0) { | |
e428e439 | 488 | error_setg_errno(errp, -ret, "Could not normalize device path"); |
c66a6157 | 489 | goto fail; |
1de1ae0a CE |
490 | } |
491 | ||
6a8dc042 JC |
492 | s->open_flags = open_flags; |
493 | raw_parse_flags(bdrv_flags, &s->open_flags); | |
83f64091 | 494 | |
90babde0 | 495 | s->fd = -1; |
40ff6d7e | 496 | fd = qemu_open(filename, s->open_flags, 0644); |
19cb3738 FB |
497 | if (fd < 0) { |
498 | ret = -errno; | |
c66a6157 | 499 | if (ret == -EROFS) { |
19cb3738 | 500 | ret = -EACCES; |
c66a6157 KW |
501 | } |
502 | goto fail; | |
19cb3738 | 503 | } |
83f64091 | 504 | s->fd = fd; |
9ef91a67 | 505 | |
5c6c3a6c | 506 | #ifdef CONFIG_LINUX_AIO |
fc32a72d | 507 | if (raw_set_aio(&s->aio_ctx, &s->use_aio, bdrv_flags)) { |
47e6b251 | 508 | qemu_close(fd); |
c66a6157 | 509 | ret = -errno; |
e428e439 | 510 | error_setg_errno(errp, -ret, "Could not set AIO state"); |
c66a6157 | 511 | goto fail; |
9ef91a67 | 512 | } |
96518254 KW |
513 | if (!s->use_aio && (bdrv_flags & BDRV_O_NATIVE_AIO)) { |
514 | error_printf("WARNING: aio=native was specified for '%s', but " | |
515 | "it requires cache.direct=on, which was not " | |
516 | "specified. Falling back to aio=threads.\n" | |
517 | " This will become an error condition in " | |
518 | "future QEMU versions.\n", | |
519 | bs->filename); | |
520 | } | |
fc32a72d | 521 | #endif |
9ef91a67 | 522 | |
7ce21016 | 523 | s->has_discard = true; |
97a2ae34 | 524 | s->has_write_zeroes = true; |
3cad8307 RPM |
525 | if ((bs->open_flags & BDRV_O_NOCACHE) != 0) { |
526 | s->needs_alignment = true; | |
527 | } | |
260a82e5 PB |
528 | |
529 | if (fstat(s->fd, &st) < 0) { | |
01212d4e | 530 | ret = -errno; |
260a82e5 PB |
531 | error_setg_errno(errp, errno, "Could not stat file"); |
532 | goto fail; | |
533 | } | |
534 | if (S_ISREG(st.st_mode)) { | |
535 | s->discard_zeroes = true; | |
d50d8222 | 536 | s->has_fallocate = true; |
260a82e5 | 537 | } |
d0b4503e PB |
538 | if (S_ISBLK(st.st_mode)) { |
539 | #ifdef BLKDISCARDZEROES | |
540 | unsigned int arg; | |
541 | if (ioctl(s->fd, BLKDISCARDZEROES, &arg) == 0 && arg) { | |
542 | s->discard_zeroes = true; | |
543 | } | |
544 | #endif | |
545 | #ifdef __linux__ | |
546 | /* On Linux 3.10, BLKDISCARD leaves stale data in the page cache. Do | |
547 | * not rely on the contents of discarded blocks unless using O_DIRECT. | |
97a2ae34 | 548 | * Same for BLKZEROOUT. |
d0b4503e PB |
549 | */ |
550 | if (!(bs->open_flags & BDRV_O_NOCACHE)) { | |
551 | s->discard_zeroes = false; | |
97a2ae34 | 552 | s->has_write_zeroes = false; |
d0b4503e PB |
553 | } |
554 | #endif | |
555 | } | |
3cad8307 RPM |
556 | #ifdef __FreeBSD__ |
557 | if (S_ISCHR(st.st_mode)) { | |
558 | /* | |
559 | * The file is a char device (disk), which on FreeBSD isn't behind | |
560 | * a pager, so force all requests to be aligned. This is needed | |
561 | * so QEMU makes sure all IO operations on the device are aligned | |
562 | * to sector size, or else FreeBSD will reject them with EINVAL. | |
563 | */ | |
564 | s->needs_alignment = true; | |
565 | } | |
566 | #endif | |
260a82e5 | 567 | |
dce512de CH |
568 | #ifdef CONFIG_XFS |
569 | if (platform_test_xfs_fd(s->fd)) { | |
7ce21016 | 570 | s->is_xfs = true; |
dce512de CH |
571 | } |
572 | #endif | |
573 | ||
c2f3426c SH |
574 | raw_attach_aio_context(bs, bdrv_get_aio_context(bs)); |
575 | ||
c66a6157 KW |
576 | ret = 0; |
577 | fail: | |
8bfea15d KW |
578 | if (filename && (bdrv_flags & BDRV_O_TEMPORARY)) { |
579 | unlink(filename); | |
580 | } | |
c66a6157 KW |
581 | qemu_opts_del(opts); |
582 | return ret; | |
83f64091 FB |
583 | } |
584 | ||
015a1036 HR |
585 | static int raw_open(BlockDriverState *bs, QDict *options, int flags, |
586 | Error **errp) | |
90babde0 CH |
587 | { |
588 | BDRVRawState *s = bs->opaque; | |
e428e439 HR |
589 | Error *local_err = NULL; |
590 | int ret; | |
90babde0 CH |
591 | |
592 | s->type = FTYPE_FILE; | |
e428e439 | 593 | ret = raw_open_common(bs, options, flags, 0, &local_err); |
84d18f06 | 594 | if (local_err) { |
e428e439 HR |
595 | error_propagate(errp, local_err); |
596 | } | |
597 | return ret; | |
90babde0 CH |
598 | } |
599 | ||
eeb6b45d JC |
600 | static int raw_reopen_prepare(BDRVReopenState *state, |
601 | BlockReopenQueue *queue, Error **errp) | |
602 | { | |
603 | BDRVRawState *s; | |
604 | BDRVRawReopenState *raw_s; | |
605 | int ret = 0; | |
df26a350 | 606 | Error *local_err = NULL; |
eeb6b45d JC |
607 | |
608 | assert(state != NULL); | |
609 | assert(state->bs != NULL); | |
610 | ||
611 | s = state->bs->opaque; | |
612 | ||
5839e53b | 613 | state->opaque = g_new0(BDRVRawReopenState, 1); |
eeb6b45d JC |
614 | raw_s = state->opaque; |
615 | ||
616 | #ifdef CONFIG_LINUX_AIO | |
617 | raw_s->use_aio = s->use_aio; | |
618 | ||
619 | /* we can use s->aio_ctx instead of a copy, because the use_aio flag is | |
620 | * valid in the 'false' condition even if aio_ctx is set, and raw_set_aio() | |
621 | * won't override aio_ctx if aio_ctx is non-NULL */ | |
622 | if (raw_set_aio(&s->aio_ctx, &raw_s->use_aio, state->flags)) { | |
e428e439 | 623 | error_setg(errp, "Could not set AIO state"); |
eeb6b45d JC |
624 | return -1; |
625 | } | |
626 | #endif | |
627 | ||
1bc6b705 JC |
628 | if (s->type == FTYPE_FD || s->type == FTYPE_CD) { |
629 | raw_s->open_flags |= O_NONBLOCK; | |
630 | } | |
631 | ||
eeb6b45d JC |
632 | raw_parse_flags(state->flags, &raw_s->open_flags); |
633 | ||
634 | raw_s->fd = -1; | |
635 | ||
fdf263f6 | 636 | int fcntl_flags = O_APPEND | O_NONBLOCK; |
eeb6b45d JC |
637 | #ifdef O_NOATIME |
638 | fcntl_flags |= O_NOATIME; | |
639 | #endif | |
640 | ||
fdf263f6 AF |
641 | #ifdef O_ASYNC |
642 | /* Not all operating systems have O_ASYNC, and those that don't | |
643 | * will not let us track the state into raw_s->open_flags (typically | |
644 | * you achieve the same effect with an ioctl, for example I_SETSIG | |
645 | * on Solaris). But we do not use O_ASYNC, so that's fine. | |
646 | */ | |
647 | assert((s->open_flags & O_ASYNC) == 0); | |
648 | #endif | |
649 | ||
eeb6b45d JC |
650 | if ((raw_s->open_flags & ~fcntl_flags) == (s->open_flags & ~fcntl_flags)) { |
651 | /* dup the original fd */ | |
652 | /* TODO: use qemu fcntl wrapper */ | |
653 | #ifdef F_DUPFD_CLOEXEC | |
654 | raw_s->fd = fcntl(s->fd, F_DUPFD_CLOEXEC, 0); | |
655 | #else | |
656 | raw_s->fd = dup(s->fd); | |
657 | if (raw_s->fd != -1) { | |
658 | qemu_set_cloexec(raw_s->fd); | |
659 | } | |
660 | #endif | |
661 | if (raw_s->fd >= 0) { | |
662 | ret = fcntl_setfl(raw_s->fd, raw_s->open_flags); | |
663 | if (ret) { | |
664 | qemu_close(raw_s->fd); | |
665 | raw_s->fd = -1; | |
666 | } | |
667 | } | |
668 | } | |
669 | ||
670 | /* If we cannot use fcntl, or fcntl failed, fall back to qemu_open() */ | |
671 | if (raw_s->fd == -1) { | |
672 | assert(!(raw_s->open_flags & O_CREAT)); | |
673 | raw_s->fd = qemu_open(state->bs->filename, raw_s->open_flags); | |
674 | if (raw_s->fd == -1) { | |
e428e439 | 675 | error_setg_errno(errp, errno, "Could not reopen file"); |
eeb6b45d JC |
676 | ret = -1; |
677 | } | |
678 | } | |
df26a350 KW |
679 | |
680 | /* Fail already reopen_prepare() if we can't get a working O_DIRECT | |
681 | * alignment with the new fd. */ | |
682 | if (raw_s->fd != -1) { | |
683 | raw_probe_alignment(state->bs, raw_s->fd, &local_err); | |
684 | if (local_err) { | |
685 | qemu_close(raw_s->fd); | |
686 | raw_s->fd = -1; | |
687 | error_propagate(errp, local_err); | |
688 | ret = -EINVAL; | |
689 | } | |
690 | } | |
691 | ||
eeb6b45d JC |
692 | return ret; |
693 | } | |
694 | ||
eeb6b45d JC |
695 | static void raw_reopen_commit(BDRVReopenState *state) |
696 | { | |
697 | BDRVRawReopenState *raw_s = state->opaque; | |
698 | BDRVRawState *s = state->bs->opaque; | |
699 | ||
700 | s->open_flags = raw_s->open_flags; | |
701 | ||
702 | qemu_close(s->fd); | |
703 | s->fd = raw_s->fd; | |
704 | #ifdef CONFIG_LINUX_AIO | |
705 | s->use_aio = raw_s->use_aio; | |
706 | #endif | |
707 | ||
708 | g_free(state->opaque); | |
709 | state->opaque = NULL; | |
710 | } | |
711 | ||
712 | ||
713 | static void raw_reopen_abort(BDRVReopenState *state) | |
714 | { | |
715 | BDRVRawReopenState *raw_s = state->opaque; | |
716 | ||
717 | /* nothing to do if NULL, we didn't get far enough */ | |
718 | if (raw_s == NULL) { | |
719 | return; | |
720 | } | |
721 | ||
722 | if (raw_s->fd >= 0) { | |
723 | qemu_close(raw_s->fd); | |
724 | raw_s->fd = -1; | |
725 | } | |
726 | g_free(state->opaque); | |
727 | state->opaque = NULL; | |
728 | } | |
729 | ||
3baca891 | 730 | static void raw_refresh_limits(BlockDriverState *bs, Error **errp) |
c25f53b0 PB |
731 | { |
732 | BDRVRawState *s = bs->opaque; | |
eeb6b45d | 733 | |
df26a350 | 734 | raw_probe_alignment(bs, s->fd, errp); |
4196d2f0 | 735 | bs->bl.min_mem_alignment = s->buf_align; |
459b4e66 | 736 | bs->bl.opt_mem_alignment = MAX(s->buf_align, getpagesize()); |
c25f53b0 | 737 | } |
83f64091 | 738 | |
1a9335e4 ET |
739 | static int check_for_dasd(int fd) |
740 | { | |
741 | #ifdef BIODASDINFO2 | |
742 | struct dasd_information2_t info = {0}; | |
743 | ||
744 | return ioctl(fd, BIODASDINFO2, &info); | |
745 | #else | |
746 | return -1; | |
747 | #endif | |
748 | } | |
749 | ||
750 | /** | |
751 | * Try to get @bs's logical and physical block size. | |
752 | * On success, store them in @bsz and return zero. | |
753 | * On failure, return negative errno. | |
754 | */ | |
755 | static int hdev_probe_blocksizes(BlockDriverState *bs, BlockSizes *bsz) | |
756 | { | |
757 | BDRVRawState *s = bs->opaque; | |
758 | int ret; | |
759 | ||
760 | /* If DASD, get blocksizes */ | |
761 | if (check_for_dasd(s->fd) < 0) { | |
762 | return -ENOTSUP; | |
763 | } | |
764 | ret = probe_logical_blocksize(s->fd, &bsz->log); | |
765 | if (ret < 0) { | |
766 | return ret; | |
767 | } | |
768 | return probe_physical_blocksize(s->fd, &bsz->phys); | |
769 | } | |
770 | ||
771 | /** | |
772 | * Try to get @bs's geometry: cyls, heads, sectors. | |
773 | * On success, store them in @geo and return 0. | |
774 | * On failure return -errno. | |
775 | * (Allows block driver to assign default geometry values that guest sees) | |
776 | */ | |
777 | #ifdef __linux__ | |
778 | static int hdev_probe_geometry(BlockDriverState *bs, HDGeometry *geo) | |
779 | { | |
780 | BDRVRawState *s = bs->opaque; | |
781 | struct hd_geometry ioctl_geo = {0}; | |
782 | uint32_t blksize; | |
783 | ||
784 | /* If DASD, get its geometry */ | |
785 | if (check_for_dasd(s->fd) < 0) { | |
786 | return -ENOTSUP; | |
787 | } | |
788 | if (ioctl(s->fd, HDIO_GETGEO, &ioctl_geo) < 0) { | |
789 | return -errno; | |
790 | } | |
791 | /* HDIO_GETGEO may return success even though geo contains zeros | |
792 | (e.g. certain multipath setups) */ | |
793 | if (!ioctl_geo.heads || !ioctl_geo.sectors || !ioctl_geo.cylinders) { | |
794 | return -ENOTSUP; | |
795 | } | |
796 | /* Do not return a geometry for partition */ | |
797 | if (ioctl_geo.start != 0) { | |
798 | return -ENOTSUP; | |
799 | } | |
800 | geo->heads = ioctl_geo.heads; | |
801 | geo->sectors = ioctl_geo.sectors; | |
802 | if (!probe_physical_blocksize(s->fd, &blksize)) { | |
803 | /* overwrite cyls: HDIO_GETGEO result is incorrect for big drives */ | |
804 | geo->cylinders = bdrv_nb_sectors(bs) / (blksize / BDRV_SECTOR_SIZE) | |
805 | / (geo->heads * geo->sectors); | |
806 | return 0; | |
807 | } | |
808 | geo->cylinders = ioctl_geo.cylinders; | |
809 | ||
810 | return 0; | |
811 | } | |
812 | #else /* __linux__ */ | |
813 | static int hdev_probe_geometry(BlockDriverState *bs, HDGeometry *geo) | |
814 | { | |
815 | return -ENOTSUP; | |
816 | } | |
817 | #endif | |
818 | ||
de81a169 PB |
819 | static ssize_t handle_aiocb_ioctl(RawPosixAIOData *aiocb) |
820 | { | |
821 | int ret; | |
822 | ||
823 | ret = ioctl(aiocb->aio_fildes, aiocb->aio_ioctl_cmd, aiocb->aio_ioctl_buf); | |
824 | if (ret == -1) { | |
825 | return -errno; | |
826 | } | |
827 | ||
b608c8dc | 828 | return 0; |
de81a169 PB |
829 | } |
830 | ||
831 | static ssize_t handle_aiocb_flush(RawPosixAIOData *aiocb) | |
832 | { | |
833 | int ret; | |
834 | ||
835 | ret = qemu_fdatasync(aiocb->aio_fildes); | |
836 | if (ret == -1) { | |
837 | return -errno; | |
838 | } | |
839 | return 0; | |
840 | } | |
841 | ||
842 | #ifdef CONFIG_PREADV | |
843 | ||
844 | static bool preadv_present = true; | |
845 | ||
846 | static ssize_t | |
847 | qemu_preadv(int fd, const struct iovec *iov, int nr_iov, off_t offset) | |
848 | { | |
849 | return preadv(fd, iov, nr_iov, offset); | |
850 | } | |
851 | ||
852 | static ssize_t | |
853 | qemu_pwritev(int fd, const struct iovec *iov, int nr_iov, off_t offset) | |
854 | { | |
855 | return pwritev(fd, iov, nr_iov, offset); | |
856 | } | |
857 | ||
858 | #else | |
859 | ||
860 | static bool preadv_present = false; | |
861 | ||
862 | static ssize_t | |
863 | qemu_preadv(int fd, const struct iovec *iov, int nr_iov, off_t offset) | |
864 | { | |
865 | return -ENOSYS; | |
866 | } | |
867 | ||
868 | static ssize_t | |
869 | qemu_pwritev(int fd, const struct iovec *iov, int nr_iov, off_t offset) | |
870 | { | |
871 | return -ENOSYS; | |
872 | } | |
873 | ||
874 | #endif | |
875 | ||
876 | static ssize_t handle_aiocb_rw_vector(RawPosixAIOData *aiocb) | |
877 | { | |
878 | ssize_t len; | |
879 | ||
880 | do { | |
881 | if (aiocb->aio_type & QEMU_AIO_WRITE) | |
882 | len = qemu_pwritev(aiocb->aio_fildes, | |
883 | aiocb->aio_iov, | |
884 | aiocb->aio_niov, | |
885 | aiocb->aio_offset); | |
886 | else | |
887 | len = qemu_preadv(aiocb->aio_fildes, | |
888 | aiocb->aio_iov, | |
889 | aiocb->aio_niov, | |
890 | aiocb->aio_offset); | |
891 | } while (len == -1 && errno == EINTR); | |
892 | ||
893 | if (len == -1) { | |
894 | return -errno; | |
895 | } | |
896 | return len; | |
897 | } | |
898 | ||
899 | /* | |
900 | * Read/writes the data to/from a given linear buffer. | |
901 | * | |
902 | * Returns the number of bytes handles or -errno in case of an error. Short | |
903 | * reads are only returned if the end of the file is reached. | |
904 | */ | |
905 | static ssize_t handle_aiocb_rw_linear(RawPosixAIOData *aiocb, char *buf) | |
906 | { | |
907 | ssize_t offset = 0; | |
908 | ssize_t len; | |
909 | ||
910 | while (offset < aiocb->aio_nbytes) { | |
911 | if (aiocb->aio_type & QEMU_AIO_WRITE) { | |
912 | len = pwrite(aiocb->aio_fildes, | |
913 | (const char *)buf + offset, | |
914 | aiocb->aio_nbytes - offset, | |
915 | aiocb->aio_offset + offset); | |
916 | } else { | |
917 | len = pread(aiocb->aio_fildes, | |
918 | buf + offset, | |
919 | aiocb->aio_nbytes - offset, | |
920 | aiocb->aio_offset + offset); | |
921 | } | |
922 | if (len == -1 && errno == EINTR) { | |
923 | continue; | |
61ed73cf SH |
924 | } else if (len == -1 && errno == EINVAL && |
925 | (aiocb->bs->open_flags & BDRV_O_NOCACHE) && | |
926 | !(aiocb->aio_type & QEMU_AIO_WRITE) && | |
927 | offset > 0) { | |
928 | /* O_DIRECT pread() may fail with EINVAL when offset is unaligned | |
929 | * after a short read. Assume that O_DIRECT short reads only occur | |
930 | * at EOF. Therefore this is a short read, not an I/O error. | |
931 | */ | |
932 | break; | |
de81a169 PB |
933 | } else if (len == -1) { |
934 | offset = -errno; | |
935 | break; | |
936 | } else if (len == 0) { | |
937 | break; | |
938 | } | |
939 | offset += len; | |
940 | } | |
941 | ||
942 | return offset; | |
943 | } | |
944 | ||
945 | static ssize_t handle_aiocb_rw(RawPosixAIOData *aiocb) | |
946 | { | |
947 | ssize_t nbytes; | |
948 | char *buf; | |
949 | ||
950 | if (!(aiocb->aio_type & QEMU_AIO_MISALIGNED)) { | |
951 | /* | |
952 | * If there is just a single buffer, and it is properly aligned | |
953 | * we can just use plain pread/pwrite without any problems. | |
954 | */ | |
955 | if (aiocb->aio_niov == 1) { | |
956 | return handle_aiocb_rw_linear(aiocb, aiocb->aio_iov->iov_base); | |
957 | } | |
958 | /* | |
959 | * We have more than one iovec, and all are properly aligned. | |
960 | * | |
961 | * Try preadv/pwritev first and fall back to linearizing the | |
962 | * buffer if it's not supported. | |
963 | */ | |
964 | if (preadv_present) { | |
965 | nbytes = handle_aiocb_rw_vector(aiocb); | |
966 | if (nbytes == aiocb->aio_nbytes || | |
967 | (nbytes < 0 && nbytes != -ENOSYS)) { | |
968 | return nbytes; | |
969 | } | |
970 | preadv_present = false; | |
971 | } | |
972 | ||
973 | /* | |
974 | * XXX(hch): short read/write. no easy way to handle the reminder | |
975 | * using these interfaces. For now retry using plain | |
976 | * pread/pwrite? | |
977 | */ | |
978 | } | |
979 | ||
980 | /* | |
981 | * Ok, we have to do it the hard way, copy all segments into | |
982 | * a single aligned buffer. | |
983 | */ | |
50d4a858 KW |
984 | buf = qemu_try_blockalign(aiocb->bs, aiocb->aio_nbytes); |
985 | if (buf == NULL) { | |
986 | return -ENOMEM; | |
987 | } | |
988 | ||
de81a169 PB |
989 | if (aiocb->aio_type & QEMU_AIO_WRITE) { |
990 | char *p = buf; | |
991 | int i; | |
992 | ||
993 | for (i = 0; i < aiocb->aio_niov; ++i) { | |
994 | memcpy(p, aiocb->aio_iov[i].iov_base, aiocb->aio_iov[i].iov_len); | |
995 | p += aiocb->aio_iov[i].iov_len; | |
996 | } | |
8eb029c2 | 997 | assert(p - buf == aiocb->aio_nbytes); |
de81a169 PB |
998 | } |
999 | ||
1000 | nbytes = handle_aiocb_rw_linear(aiocb, buf); | |
1001 | if (!(aiocb->aio_type & QEMU_AIO_WRITE)) { | |
1002 | char *p = buf; | |
1003 | size_t count = aiocb->aio_nbytes, copy; | |
1004 | int i; | |
1005 | ||
1006 | for (i = 0; i < aiocb->aio_niov && count; ++i) { | |
1007 | copy = count; | |
1008 | if (copy > aiocb->aio_iov[i].iov_len) { | |
1009 | copy = aiocb->aio_iov[i].iov_len; | |
1010 | } | |
1011 | memcpy(aiocb->aio_iov[i].iov_base, p, copy); | |
8eb029c2 | 1012 | assert(count >= copy); |
de81a169 PB |
1013 | p += copy; |
1014 | count -= copy; | |
1015 | } | |
8eb029c2 | 1016 | assert(count == 0); |
de81a169 PB |
1017 | } |
1018 | qemu_vfree(buf); | |
1019 | ||
1020 | return nbytes; | |
1021 | } | |
1022 | ||
8238010b | 1023 | #ifdef CONFIG_XFS |
97a2ae34 PB |
1024 | static int xfs_write_zeroes(BDRVRawState *s, int64_t offset, uint64_t bytes) |
1025 | { | |
1026 | struct xfs_flock64 fl; | |
bcb22555 | 1027 | int err; |
97a2ae34 PB |
1028 | |
1029 | memset(&fl, 0, sizeof(fl)); | |
1030 | fl.l_whence = SEEK_SET; | |
1031 | fl.l_start = offset; | |
1032 | fl.l_len = bytes; | |
1033 | ||
1034 | if (xfsctl(NULL, s->fd, XFS_IOC_ZERO_RANGE, &fl) < 0) { | |
bcb22555 DA |
1035 | err = errno; |
1036 | DPRINTF("cannot write zero range (%s)\n", strerror(errno)); | |
1037 | return -err; | |
97a2ae34 PB |
1038 | } |
1039 | ||
1040 | return 0; | |
1041 | } | |
1042 | ||
8238010b PB |
1043 | static int xfs_discard(BDRVRawState *s, int64_t offset, uint64_t bytes) |
1044 | { | |
1045 | struct xfs_flock64 fl; | |
bcb22555 | 1046 | int err; |
8238010b PB |
1047 | |
1048 | memset(&fl, 0, sizeof(fl)); | |
1049 | fl.l_whence = SEEK_SET; | |
1050 | fl.l_start = offset; | |
1051 | fl.l_len = bytes; | |
1052 | ||
1053 | if (xfsctl(NULL, s->fd, XFS_IOC_UNRESVSP64, &fl) < 0) { | |
bcb22555 DA |
1054 | err = errno; |
1055 | DPRINTF("cannot punch hole (%s)\n", strerror(errno)); | |
1056 | return -err; | |
8238010b PB |
1057 | } |
1058 | ||
1059 | return 0; | |
1060 | } | |
1061 | #endif | |
1062 | ||
1486df0e DL |
1063 | static int translate_err(int err) |
1064 | { | |
1065 | if (err == -ENODEV || err == -ENOSYS || err == -EOPNOTSUPP || | |
1066 | err == -ENOTTY) { | |
1067 | err = -ENOTSUP; | |
1068 | } | |
1069 | return err; | |
1070 | } | |
1071 | ||
d50d8222 | 1072 | #ifdef CONFIG_FALLOCATE |
0b991712 DL |
1073 | static int do_fallocate(int fd, int mode, off_t offset, off_t len) |
1074 | { | |
1075 | do { | |
1076 | if (fallocate(fd, mode, offset, len) == 0) { | |
1077 | return 0; | |
1078 | } | |
1079 | } while (errno == EINTR); | |
1080 | return translate_err(-errno); | |
1081 | } | |
1082 | #endif | |
1083 | ||
37cc9f7f | 1084 | static ssize_t handle_aiocb_write_zeroes_block(RawPosixAIOData *aiocb) |
97a2ae34 | 1085 | { |
37cc9f7f | 1086 | int ret = -ENOTSUP; |
97a2ae34 PB |
1087 | BDRVRawState *s = aiocb->bs->opaque; |
1088 | ||
37cc9f7f | 1089 | if (!s->has_write_zeroes) { |
97a2ae34 PB |
1090 | return -ENOTSUP; |
1091 | } | |
1092 | ||
97a2ae34 | 1093 | #ifdef BLKZEROOUT |
37cc9f7f DL |
1094 | do { |
1095 | uint64_t range[2] = { aiocb->aio_offset, aiocb->aio_nbytes }; | |
1096 | if (ioctl(aiocb->aio_fildes, BLKZEROOUT, range) == 0) { | |
1097 | return 0; | |
97a2ae34 | 1098 | } |
37cc9f7f DL |
1099 | } while (errno == EINTR); |
1100 | ||
1101 | ret = translate_err(-errno); | |
97a2ae34 | 1102 | #endif |
97a2ae34 | 1103 | |
1486df0e | 1104 | if (ret == -ENOTSUP) { |
97a2ae34 | 1105 | s->has_write_zeroes = false; |
97a2ae34 PB |
1106 | } |
1107 | return ret; | |
1108 | } | |
1109 | ||
37cc9f7f DL |
1110 | static ssize_t handle_aiocb_write_zeroes(RawPosixAIOData *aiocb) |
1111 | { | |
a6dcf097 | 1112 | #if defined(CONFIG_FALLOCATE) || defined(CONFIG_XFS) |
37cc9f7f | 1113 | BDRVRawState *s = aiocb->bs->opaque; |
a6dcf097 | 1114 | #endif |
37cc9f7f DL |
1115 | |
1116 | if (aiocb->aio_type & QEMU_AIO_BLKDEV) { | |
1117 | return handle_aiocb_write_zeroes_block(aiocb); | |
1118 | } | |
1119 | ||
1120 | #ifdef CONFIG_XFS | |
1121 | if (s->is_xfs) { | |
1122 | return xfs_write_zeroes(s, aiocb->aio_offset, aiocb->aio_nbytes); | |
1123 | } | |
1124 | #endif | |
1125 | ||
b953f075 DL |
1126 | #ifdef CONFIG_FALLOCATE_ZERO_RANGE |
1127 | if (s->has_write_zeroes) { | |
1128 | int ret = do_fallocate(s->fd, FALLOC_FL_ZERO_RANGE, | |
1129 | aiocb->aio_offset, aiocb->aio_nbytes); | |
1130 | if (ret == 0 || ret != -ENOTSUP) { | |
1131 | return ret; | |
1132 | } | |
1133 | s->has_write_zeroes = false; | |
1134 | } | |
1135 | #endif | |
1136 | ||
1cdc3239 DL |
1137 | #ifdef CONFIG_FALLOCATE_PUNCH_HOLE |
1138 | if (s->has_discard && s->has_fallocate) { | |
1139 | int ret = do_fallocate(s->fd, | |
1140 | FALLOC_FL_PUNCH_HOLE | FALLOC_FL_KEEP_SIZE, | |
1141 | aiocb->aio_offset, aiocb->aio_nbytes); | |
1142 | if (ret == 0) { | |
1143 | ret = do_fallocate(s->fd, 0, aiocb->aio_offset, aiocb->aio_nbytes); | |
1144 | if (ret == 0 || ret != -ENOTSUP) { | |
1145 | return ret; | |
1146 | } | |
1147 | s->has_fallocate = false; | |
1148 | } else if (ret != -ENOTSUP) { | |
1149 | return ret; | |
1150 | } else { | |
1151 | s->has_discard = false; | |
1152 | } | |
1153 | } | |
1154 | #endif | |
1155 | ||
d50d8222 DL |
1156 | #ifdef CONFIG_FALLOCATE |
1157 | if (s->has_fallocate && aiocb->aio_offset >= bdrv_getlength(aiocb->bs)) { | |
1158 | int ret = do_fallocate(s->fd, 0, aiocb->aio_offset, aiocb->aio_nbytes); | |
1159 | if (ret == 0 || ret != -ENOTSUP) { | |
1160 | return ret; | |
1161 | } | |
1162 | s->has_fallocate = false; | |
1163 | } | |
1164 | #endif | |
1165 | ||
37cc9f7f DL |
1166 | return -ENOTSUP; |
1167 | } | |
1168 | ||
8238010b PB |
1169 | static ssize_t handle_aiocb_discard(RawPosixAIOData *aiocb) |
1170 | { | |
1171 | int ret = -EOPNOTSUPP; | |
1172 | BDRVRawState *s = aiocb->bs->opaque; | |
1173 | ||
7ce21016 PB |
1174 | if (!s->has_discard) { |
1175 | return -ENOTSUP; | |
8238010b PB |
1176 | } |
1177 | ||
1178 | if (aiocb->aio_type & QEMU_AIO_BLKDEV) { | |
1179 | #ifdef BLKDISCARD | |
1180 | do { | |
1181 | uint64_t range[2] = { aiocb->aio_offset, aiocb->aio_nbytes }; | |
1182 | if (ioctl(aiocb->aio_fildes, BLKDISCARD, range) == 0) { | |
1183 | return 0; | |
1184 | } | |
1185 | } while (errno == EINTR); | |
1186 | ||
1187 | ret = -errno; | |
1188 | #endif | |
1189 | } else { | |
1190 | #ifdef CONFIG_XFS | |
1191 | if (s->is_xfs) { | |
1192 | return xfs_discard(s, aiocb->aio_offset, aiocb->aio_nbytes); | |
1193 | } | |
1194 | #endif | |
1195 | ||
1196 | #ifdef CONFIG_FALLOCATE_PUNCH_HOLE | |
0b991712 DL |
1197 | ret = do_fallocate(s->fd, FALLOC_FL_PUNCH_HOLE | FALLOC_FL_KEEP_SIZE, |
1198 | aiocb->aio_offset, aiocb->aio_nbytes); | |
8238010b PB |
1199 | #endif |
1200 | } | |
1201 | ||
1486df0e DL |
1202 | ret = translate_err(ret); |
1203 | if (ret == -ENOTSUP) { | |
7ce21016 | 1204 | s->has_discard = false; |
8238010b PB |
1205 | } |
1206 | return ret; | |
1207 | } | |
1208 | ||
de81a169 PB |
1209 | static int aio_worker(void *arg) |
1210 | { | |
1211 | RawPosixAIOData *aiocb = arg; | |
1212 | ssize_t ret = 0; | |
1213 | ||
1214 | switch (aiocb->aio_type & QEMU_AIO_TYPE_MASK) { | |
1215 | case QEMU_AIO_READ: | |
1216 | ret = handle_aiocb_rw(aiocb); | |
c0191e76 | 1217 | if (ret >= 0 && ret < aiocb->aio_nbytes) { |
de81a169 PB |
1218 | iov_memset(aiocb->aio_iov, aiocb->aio_niov, ret, |
1219 | 0, aiocb->aio_nbytes - ret); | |
1220 | ||
1221 | ret = aiocb->aio_nbytes; | |
1222 | } | |
1223 | if (ret == aiocb->aio_nbytes) { | |
1224 | ret = 0; | |
1225 | } else if (ret >= 0 && ret < aiocb->aio_nbytes) { | |
1226 | ret = -EINVAL; | |
1227 | } | |
1228 | break; | |
1229 | case QEMU_AIO_WRITE: | |
1230 | ret = handle_aiocb_rw(aiocb); | |
1231 | if (ret == aiocb->aio_nbytes) { | |
1232 | ret = 0; | |
1233 | } else if (ret >= 0 && ret < aiocb->aio_nbytes) { | |
1234 | ret = -EINVAL; | |
1235 | } | |
1236 | break; | |
1237 | case QEMU_AIO_FLUSH: | |
1238 | ret = handle_aiocb_flush(aiocb); | |
1239 | break; | |
1240 | case QEMU_AIO_IOCTL: | |
1241 | ret = handle_aiocb_ioctl(aiocb); | |
1242 | break; | |
8238010b PB |
1243 | case QEMU_AIO_DISCARD: |
1244 | ret = handle_aiocb_discard(aiocb); | |
1245 | break; | |
97a2ae34 PB |
1246 | case QEMU_AIO_WRITE_ZEROES: |
1247 | ret = handle_aiocb_write_zeroes(aiocb); | |
1248 | break; | |
de81a169 PB |
1249 | default: |
1250 | fprintf(stderr, "invalid aio request (0x%x)\n", aiocb->aio_type); | |
1251 | ret = -EINVAL; | |
1252 | break; | |
1253 | } | |
1254 | ||
1255 | g_slice_free(RawPosixAIOData, aiocb); | |
1256 | return ret; | |
1257 | } | |
1258 | ||
260a82e5 PB |
1259 | static int paio_submit_co(BlockDriverState *bs, int fd, |
1260 | int64_t sector_num, QEMUIOVector *qiov, int nb_sectors, | |
1261 | int type) | |
1262 | { | |
1263 | RawPosixAIOData *acb = g_slice_new(RawPosixAIOData); | |
1264 | ThreadPool *pool; | |
1265 | ||
1266 | acb->bs = bs; | |
1267 | acb->aio_type = type; | |
1268 | acb->aio_fildes = fd; | |
1269 | ||
8eb029c2 KW |
1270 | acb->aio_nbytes = nb_sectors * BDRV_SECTOR_SIZE; |
1271 | acb->aio_offset = sector_num * BDRV_SECTOR_SIZE; | |
1272 | ||
260a82e5 PB |
1273 | if (qiov) { |
1274 | acb->aio_iov = qiov->iov; | |
1275 | acb->aio_niov = qiov->niov; | |
8eb029c2 | 1276 | assert(qiov->size == acb->aio_nbytes); |
260a82e5 | 1277 | } |
260a82e5 PB |
1278 | |
1279 | trace_paio_submit_co(sector_num, nb_sectors, type); | |
1280 | pool = aio_get_thread_pool(bdrv_get_aio_context(bs)); | |
1281 | return thread_pool_submit_co(pool, aio_worker, acb); | |
1282 | } | |
1283 | ||
7c84b1b8 | 1284 | static BlockAIOCB *paio_submit(BlockDriverState *bs, int fd, |
de81a169 | 1285 | int64_t sector_num, QEMUIOVector *qiov, int nb_sectors, |
097310b5 | 1286 | BlockCompletionFunc *cb, void *opaque, int type) |
de81a169 PB |
1287 | { |
1288 | RawPosixAIOData *acb = g_slice_new(RawPosixAIOData); | |
c4d9d196 | 1289 | ThreadPool *pool; |
de81a169 PB |
1290 | |
1291 | acb->bs = bs; | |
1292 | acb->aio_type = type; | |
1293 | acb->aio_fildes = fd; | |
1294 | ||
8eb029c2 KW |
1295 | acb->aio_nbytes = nb_sectors * BDRV_SECTOR_SIZE; |
1296 | acb->aio_offset = sector_num * BDRV_SECTOR_SIZE; | |
1297 | ||
de81a169 PB |
1298 | if (qiov) { |
1299 | acb->aio_iov = qiov->iov; | |
1300 | acb->aio_niov = qiov->niov; | |
8eb029c2 | 1301 | assert(qiov->size == acb->aio_nbytes); |
de81a169 | 1302 | } |
de81a169 PB |
1303 | |
1304 | trace_paio_submit(acb, opaque, sector_num, nb_sectors, type); | |
c4d9d196 SH |
1305 | pool = aio_get_thread_pool(bdrv_get_aio_context(bs)); |
1306 | return thread_pool_submit_aio(pool, aio_worker, acb, cb, opaque); | |
de81a169 PB |
1307 | } |
1308 | ||
7c84b1b8 | 1309 | static BlockAIOCB *raw_aio_submit(BlockDriverState *bs, |
9ef91a67 | 1310 | int64_t sector_num, QEMUIOVector *qiov, int nb_sectors, |
097310b5 | 1311 | BlockCompletionFunc *cb, void *opaque, int type) |
83f64091 | 1312 | { |
ce1a14dc | 1313 | BDRVRawState *s = bs->opaque; |
ce1a14dc | 1314 | |
19cb3738 FB |
1315 | if (fd_open(bs) < 0) |
1316 | return NULL; | |
1317 | ||
f141eafe | 1318 | /* |
3cad8307 RPM |
1319 | * Check if the underlying device requires requests to be aligned, |
1320 | * and if the request we are trying to submit is aligned or not. | |
1321 | * If this is the case tell the low-level driver that it needs | |
1322 | * to copy the buffer. | |
f141eafe | 1323 | */ |
3cad8307 | 1324 | if (s->needs_alignment) { |
c53b1c51 | 1325 | if (!bdrv_qiov_is_aligned(bs, qiov)) { |
5c6c3a6c | 1326 | type |= QEMU_AIO_MISALIGNED; |
e44bd6fc | 1327 | #ifdef CONFIG_LINUX_AIO |
5c6c3a6c CH |
1328 | } else if (s->use_aio) { |
1329 | return laio_submit(bs, s->aio_ctx, s->fd, sector_num, qiov, | |
e44bd6fc SW |
1330 | nb_sectors, cb, opaque, type); |
1331 | #endif | |
5c6c3a6c | 1332 | } |
9ef91a67 | 1333 | } |
f141eafe | 1334 | |
1e5b9d2f | 1335 | return paio_submit(bs, s->fd, sector_num, qiov, nb_sectors, |
9ef91a67 | 1336 | cb, opaque, type); |
83f64091 FB |
1337 | } |
1338 | ||
1b3abdcc ML |
1339 | static void raw_aio_plug(BlockDriverState *bs) |
1340 | { | |
1341 | #ifdef CONFIG_LINUX_AIO | |
1342 | BDRVRawState *s = bs->opaque; | |
1343 | if (s->use_aio) { | |
1344 | laio_io_plug(bs, s->aio_ctx); | |
1345 | } | |
1346 | #endif | |
1347 | } | |
1348 | ||
1349 | static void raw_aio_unplug(BlockDriverState *bs) | |
1350 | { | |
1351 | #ifdef CONFIG_LINUX_AIO | |
1352 | BDRVRawState *s = bs->opaque; | |
1353 | if (s->use_aio) { | |
1354 | laio_io_unplug(bs, s->aio_ctx, true); | |
1355 | } | |
1356 | #endif | |
1357 | } | |
1358 | ||
1359 | static void raw_aio_flush_io_queue(BlockDriverState *bs) | |
1360 | { | |
1361 | #ifdef CONFIG_LINUX_AIO | |
1362 | BDRVRawState *s = bs->opaque; | |
1363 | if (s->use_aio) { | |
1364 | laio_io_unplug(bs, s->aio_ctx, false); | |
1365 | } | |
1366 | #endif | |
1367 | } | |
1368 | ||
7c84b1b8 | 1369 | static BlockAIOCB *raw_aio_readv(BlockDriverState *bs, |
f141eafe | 1370 | int64_t sector_num, QEMUIOVector *qiov, int nb_sectors, |
097310b5 | 1371 | BlockCompletionFunc *cb, void *opaque) |
83f64091 | 1372 | { |
9ef91a67 CH |
1373 | return raw_aio_submit(bs, sector_num, qiov, nb_sectors, |
1374 | cb, opaque, QEMU_AIO_READ); | |
83f64091 FB |
1375 | } |
1376 | ||
7c84b1b8 | 1377 | static BlockAIOCB *raw_aio_writev(BlockDriverState *bs, |
f141eafe | 1378 | int64_t sector_num, QEMUIOVector *qiov, int nb_sectors, |
097310b5 | 1379 | BlockCompletionFunc *cb, void *opaque) |
83f64091 | 1380 | { |
9ef91a67 CH |
1381 | return raw_aio_submit(bs, sector_num, qiov, nb_sectors, |
1382 | cb, opaque, QEMU_AIO_WRITE); | |
83f64091 | 1383 | } |
53538725 | 1384 | |
7c84b1b8 | 1385 | static BlockAIOCB *raw_aio_flush(BlockDriverState *bs, |
097310b5 | 1386 | BlockCompletionFunc *cb, void *opaque) |
b2e12bc6 CH |
1387 | { |
1388 | BDRVRawState *s = bs->opaque; | |
1389 | ||
1390 | if (fd_open(bs) < 0) | |
1391 | return NULL; | |
1392 | ||
1e5b9d2f | 1393 | return paio_submit(bs, s->fd, 0, NULL, 0, cb, opaque, QEMU_AIO_FLUSH); |
b2e12bc6 CH |
1394 | } |
1395 | ||
83f64091 FB |
1396 | static void raw_close(BlockDriverState *bs) |
1397 | { | |
1398 | BDRVRawState *s = bs->opaque; | |
c2f3426c SH |
1399 | |
1400 | raw_detach_aio_context(bs); | |
1401 | ||
abd269b7 SH |
1402 | #ifdef CONFIG_LINUX_AIO |
1403 | if (s->use_aio) { | |
1404 | laio_cleanup(s->aio_ctx); | |
1405 | } | |
1406 | #endif | |
19cb3738 | 1407 | if (s->fd >= 0) { |
2e1e79da | 1408 | qemu_close(s->fd); |
19cb3738 FB |
1409 | s->fd = -1; |
1410 | } | |
83f64091 FB |
1411 | } |
1412 | ||
1413 | static int raw_truncate(BlockDriverState *bs, int64_t offset) | |
1414 | { | |
1415 | BDRVRawState *s = bs->opaque; | |
55b949c8 CH |
1416 | struct stat st; |
1417 | ||
1418 | if (fstat(s->fd, &st)) { | |
83f64091 | 1419 | return -errno; |
55b949c8 CH |
1420 | } |
1421 | ||
1422 | if (S_ISREG(st.st_mode)) { | |
1423 | if (ftruncate(s->fd, offset) < 0) { | |
1424 | return -errno; | |
1425 | } | |
1426 | } else if (S_ISCHR(st.st_mode) || S_ISBLK(st.st_mode)) { | |
1427 | if (offset > raw_getlength(bs)) { | |
1428 | return -EINVAL; | |
1429 | } | |
1430 | } else { | |
1431 | return -ENOTSUP; | |
1432 | } | |
1433 | ||
83f64091 FB |
1434 | return 0; |
1435 | } | |
1436 | ||
128ab2ff BS |
1437 | #ifdef __OpenBSD__ |
1438 | static int64_t raw_getlength(BlockDriverState *bs) | |
1439 | { | |
1440 | BDRVRawState *s = bs->opaque; | |
1441 | int fd = s->fd; | |
1442 | struct stat st; | |
1443 | ||
1444 | if (fstat(fd, &st)) | |
aa729704 | 1445 | return -errno; |
128ab2ff BS |
1446 | if (S_ISCHR(st.st_mode) || S_ISBLK(st.st_mode)) { |
1447 | struct disklabel dl; | |
1448 | ||
1449 | if (ioctl(fd, DIOCGDINFO, &dl)) | |
aa729704 | 1450 | return -errno; |
128ab2ff BS |
1451 | return (uint64_t)dl.d_secsize * |
1452 | dl.d_partitions[DISKPART(st.st_rdev)].p_size; | |
1453 | } else | |
1454 | return st.st_size; | |
1455 | } | |
d1f6fd8d CE |
1456 | #elif defined(__NetBSD__) |
1457 | static int64_t raw_getlength(BlockDriverState *bs) | |
1458 | { | |
1459 | BDRVRawState *s = bs->opaque; | |
1460 | int fd = s->fd; | |
1461 | struct stat st; | |
1462 | ||
1463 | if (fstat(fd, &st)) | |
aa729704 | 1464 | return -errno; |
d1f6fd8d CE |
1465 | if (S_ISCHR(st.st_mode) || S_ISBLK(st.st_mode)) { |
1466 | struct dkwedge_info dkw; | |
1467 | ||
1468 | if (ioctl(fd, DIOCGWEDGEINFO, &dkw) != -1) { | |
1469 | return dkw.dkw_size * 512; | |
1470 | } else { | |
1471 | struct disklabel dl; | |
1472 | ||
1473 | if (ioctl(fd, DIOCGDINFO, &dl)) | |
aa729704 | 1474 | return -errno; |
d1f6fd8d CE |
1475 | return (uint64_t)dl.d_secsize * |
1476 | dl.d_partitions[DISKPART(st.st_rdev)].p_size; | |
1477 | } | |
1478 | } else | |
1479 | return st.st_size; | |
1480 | } | |
50779cc2 CH |
1481 | #elif defined(__sun__) |
1482 | static int64_t raw_getlength(BlockDriverState *bs) | |
1483 | { | |
1484 | BDRVRawState *s = bs->opaque; | |
1485 | struct dk_minfo minfo; | |
1486 | int ret; | |
aa729704 | 1487 | int64_t size; |
50779cc2 CH |
1488 | |
1489 | ret = fd_open(bs); | |
1490 | if (ret < 0) { | |
1491 | return ret; | |
1492 | } | |
1493 | ||
1494 | /* | |
1495 | * Use the DKIOCGMEDIAINFO ioctl to read the size. | |
1496 | */ | |
1497 | ret = ioctl(s->fd, DKIOCGMEDIAINFO, &minfo); | |
1498 | if (ret != -1) { | |
1499 | return minfo.dki_lbsize * minfo.dki_capacity; | |
1500 | } | |
1501 | ||
1502 | /* | |
1503 | * There are reports that lseek on some devices fails, but | |
1504 | * irc discussion said that contingency on contingency was overkill. | |
1505 | */ | |
aa729704 MA |
1506 | size = lseek(s->fd, 0, SEEK_END); |
1507 | if (size < 0) { | |
1508 | return -errno; | |
1509 | } | |
1510 | return size; | |
50779cc2 CH |
1511 | } |
1512 | #elif defined(CONFIG_BSD) | |
1513 | static int64_t raw_getlength(BlockDriverState *bs) | |
83f64091 FB |
1514 | { |
1515 | BDRVRawState *s = bs->opaque; | |
1516 | int fd = s->fd; | |
1517 | int64_t size; | |
83f64091 | 1518 | struct stat sb; |
a167ba50 | 1519 | #if defined (__FreeBSD__) || defined(__FreeBSD_kernel__) |
9f23011a | 1520 | int reopened = 0; |
83f64091 | 1521 | #endif |
19cb3738 FB |
1522 | int ret; |
1523 | ||
1524 | ret = fd_open(bs); | |
1525 | if (ret < 0) | |
1526 | return ret; | |
83f64091 | 1527 | |
a167ba50 | 1528 | #if defined (__FreeBSD__) || defined(__FreeBSD_kernel__) |
9f23011a BS |
1529 | again: |
1530 | #endif | |
83f64091 FB |
1531 | if (!fstat(fd, &sb) && (S_IFCHR & sb.st_mode)) { |
1532 | #ifdef DIOCGMEDIASIZE | |
1533 | if (ioctl(fd, DIOCGMEDIASIZE, (off_t *)&size)) | |
c5e97233 BS |
1534 | #elif defined(DIOCGPART) |
1535 | { | |
1536 | struct partinfo pi; | |
1537 | if (ioctl(fd, DIOCGPART, &pi) == 0) | |
1538 | size = pi.media_size; | |
1539 | else | |
1540 | size = 0; | |
1541 | } | |
1542 | if (size == 0) | |
83f64091 | 1543 | #endif |
83affaa6 | 1544 | #if defined(__APPLE__) && defined(__MACH__) |
728dacbd JA |
1545 | { |
1546 | uint64_t sectors = 0; | |
1547 | uint32_t sector_size = 0; | |
1548 | ||
1549 | if (ioctl(fd, DKIOCGETBLOCKCOUNT, §ors) == 0 | |
1550 | && ioctl(fd, DKIOCGETBLOCKSIZE, §or_size) == 0) { | |
1551 | size = sectors * sector_size; | |
1552 | } else { | |
1553 | size = lseek(fd, 0LL, SEEK_END); | |
1554 | if (size < 0) { | |
1555 | return -errno; | |
1556 | } | |
1557 | } | |
1558 | } | |
83f64091 FB |
1559 | #else |
1560 | size = lseek(fd, 0LL, SEEK_END); | |
aa729704 MA |
1561 | if (size < 0) { |
1562 | return -errno; | |
1563 | } | |
9f23011a | 1564 | #endif |
a167ba50 | 1565 | #if defined(__FreeBSD__) || defined(__FreeBSD_kernel__) |
9f23011a BS |
1566 | switch(s->type) { |
1567 | case FTYPE_CD: | |
1568 | /* XXX FreeBSD acd returns UINT_MAX sectors for an empty drive */ | |
1569 | if (size == 2048LL * (unsigned)-1) | |
1570 | size = 0; | |
1571 | /* XXX no disc? maybe we need to reopen... */ | |
f3a5d3f8 | 1572 | if (size <= 0 && !reopened && cdrom_reopen(bs) >= 0) { |
9f23011a BS |
1573 | reopened = 1; |
1574 | goto again; | |
1575 | } | |
1576 | } | |
83f64091 | 1577 | #endif |
50779cc2 | 1578 | } else { |
83f64091 | 1579 | size = lseek(fd, 0, SEEK_END); |
aa729704 MA |
1580 | if (size < 0) { |
1581 | return -errno; | |
1582 | } | |
83f64091 | 1583 | } |
83f64091 FB |
1584 | return size; |
1585 | } | |
50779cc2 CH |
1586 | #else |
1587 | static int64_t raw_getlength(BlockDriverState *bs) | |
1588 | { | |
1589 | BDRVRawState *s = bs->opaque; | |
1590 | int ret; | |
aa729704 | 1591 | int64_t size; |
50779cc2 CH |
1592 | |
1593 | ret = fd_open(bs); | |
1594 | if (ret < 0) { | |
1595 | return ret; | |
1596 | } | |
1597 | ||
aa729704 MA |
1598 | size = lseek(s->fd, 0, SEEK_END); |
1599 | if (size < 0) { | |
1600 | return -errno; | |
1601 | } | |
1602 | return size; | |
50779cc2 | 1603 | } |
128ab2ff | 1604 | #endif |
83f64091 | 1605 | |
4a1d5e1f FZ |
1606 | static int64_t raw_get_allocated_file_size(BlockDriverState *bs) |
1607 | { | |
1608 | struct stat st; | |
1609 | BDRVRawState *s = bs->opaque; | |
1610 | ||
1611 | if (fstat(s->fd, &st) < 0) { | |
1612 | return -errno; | |
1613 | } | |
1614 | return (int64_t)st.st_blocks * 512; | |
1615 | } | |
1616 | ||
6f482f74 | 1617 | static int raw_create(const char *filename, QemuOpts *opts, Error **errp) |
83f64091 FB |
1618 | { |
1619 | int fd; | |
1e37d059 | 1620 | int result = 0; |
0e7e1989 | 1621 | int64_t total_size = 0; |
4ab15590 | 1622 | bool nocow = false; |
06247428 HT |
1623 | PreallocMode prealloc; |
1624 | char *buf = NULL; | |
1625 | Error *local_err = NULL; | |
83f64091 | 1626 | |
464d9f64 HR |
1627 | strstart(filename, "file:", &filename); |
1628 | ||
0e7e1989 | 1629 | /* Read out options */ |
180e9526 HT |
1630 | total_size = ROUND_UP(qemu_opt_get_size_del(opts, BLOCK_OPT_SIZE, 0), |
1631 | BDRV_SECTOR_SIZE); | |
4ab15590 | 1632 | nocow = qemu_opt_get_bool(opts, BLOCK_OPT_NOCOW, false); |
06247428 HT |
1633 | buf = qemu_opt_get_del(opts, BLOCK_OPT_PREALLOC); |
1634 | prealloc = qapi_enum_parse(PreallocMode_lookup, buf, | |
1635 | PREALLOC_MODE_MAX, PREALLOC_MODE_OFF, | |
1636 | &local_err); | |
1637 | g_free(buf); | |
1638 | if (local_err) { | |
1639 | error_propagate(errp, local_err); | |
1640 | result = -EINVAL; | |
1641 | goto out; | |
1642 | } | |
83f64091 | 1643 | |
6165f4d8 CB |
1644 | fd = qemu_open(filename, O_WRONLY | O_CREAT | O_TRUNC | O_BINARY, |
1645 | 0644); | |
1e37d059 SW |
1646 | if (fd < 0) { |
1647 | result = -errno; | |
e428e439 | 1648 | error_setg_errno(errp, -result, "Could not create file"); |
06247428 HT |
1649 | goto out; |
1650 | } | |
1651 | ||
1652 | if (nocow) { | |
4ab15590 | 1653 | #ifdef __linux__ |
06247428 HT |
1654 | /* Set NOCOW flag to solve performance issue on fs like btrfs. |
1655 | * This is an optimisation. The FS_IOC_SETFLAGS ioctl return value | |
1656 | * will be ignored since any failure of this operation should not | |
1657 | * block the left work. | |
1658 | */ | |
1659 | int attr; | |
1660 | if (ioctl(fd, FS_IOC_GETFLAGS, &attr) == 0) { | |
1661 | attr |= FS_NOCOW_FL; | |
1662 | ioctl(fd, FS_IOC_SETFLAGS, &attr); | |
4ab15590 | 1663 | } |
06247428 HT |
1664 | #endif |
1665 | } | |
1666 | ||
1667 | if (ftruncate(fd, total_size) != 0) { | |
1668 | result = -errno; | |
1669 | error_setg_errno(errp, -result, "Could not resize file"); | |
1670 | goto out_close; | |
1671 | } | |
4ab15590 | 1672 | |
ed911435 KW |
1673 | switch (prealloc) { |
1674 | #ifdef CONFIG_POSIX_FALLOCATE | |
1675 | case PREALLOC_MODE_FALLOC: | |
06247428 HT |
1676 | /* posix_fallocate() doesn't set errno. */ |
1677 | result = -posix_fallocate(fd, 0, total_size); | |
1678 | if (result != 0) { | |
1679 | error_setg_errno(errp, -result, | |
1680 | "Could not preallocate data for the new file"); | |
1e37d059 | 1681 | } |
ed911435 KW |
1682 | break; |
1683 | #endif | |
1684 | case PREALLOC_MODE_FULL: | |
1685 | { | |
06247428 | 1686 | int64_t num = 0, left = total_size; |
ed911435 | 1687 | buf = g_malloc0(65536); |
06247428 HT |
1688 | |
1689 | while (left > 0) { | |
1690 | num = MIN(left, 65536); | |
1691 | result = write(fd, buf, num); | |
1692 | if (result < 0) { | |
1693 | result = -errno; | |
1694 | error_setg_errno(errp, -result, | |
1695 | "Could not write to the new file"); | |
1696 | break; | |
1697 | } | |
39411cf3 | 1698 | left -= result; |
1e37d059 | 1699 | } |
731de380 | 1700 | if (result >= 0) { |
098ffa66 HR |
1701 | result = fsync(fd); |
1702 | if (result < 0) { | |
1703 | result = -errno; | |
1704 | error_setg_errno(errp, -result, | |
1705 | "Could not flush new file to disk"); | |
1706 | } | |
731de380 | 1707 | } |
06247428 | 1708 | g_free(buf); |
ed911435 KW |
1709 | break; |
1710 | } | |
1711 | case PREALLOC_MODE_OFF: | |
1712 | break; | |
1713 | default: | |
06247428 HT |
1714 | result = -EINVAL; |
1715 | error_setg(errp, "Unsupported preallocation mode: %s", | |
1716 | PreallocMode_lookup[prealloc]); | |
ed911435 | 1717 | break; |
1e37d059 | 1718 | } |
06247428 HT |
1719 | |
1720 | out_close: | |
1721 | if (qemu_close(fd) != 0 && result == 0) { | |
1722 | result = -errno; | |
1723 | error_setg_errno(errp, -result, "Could not close the new file"); | |
1724 | } | |
1725 | out: | |
1e37d059 | 1726 | return result; |
83f64091 FB |
1727 | } |
1728 | ||
d1f06fe6 MA |
1729 | /* |
1730 | * Find allocation range in @bs around offset @start. | |
1731 | * May change underlying file descriptor's file offset. | |
1732 | * If @start is not in a hole, store @start in @data, and the | |
1733 | * beginning of the next hole in @hole, and return 0. | |
1734 | * If @start is in a non-trailing hole, store @start in @hole and the | |
1735 | * beginning of the next non-hole in @data, and return 0. | |
1736 | * If @start is in a trailing hole or beyond EOF, return -ENXIO. | |
1737 | * If we can't find out, return a negative errno other than -ENXIO. | |
1738 | */ | |
1739 | static int find_allocation(BlockDriverState *bs, off_t start, | |
1740 | off_t *data, off_t *hole) | |
4f11aa8a HR |
1741 | { |
1742 | #if defined SEEK_HOLE && defined SEEK_DATA | |
94282e71 | 1743 | BDRVRawState *s = bs->opaque; |
d1f06fe6 | 1744 | off_t offs; |
94282e71 | 1745 | |
d1f06fe6 MA |
1746 | /* |
1747 | * SEEK_DATA cases: | |
1748 | * D1. offs == start: start is in data | |
1749 | * D2. offs > start: start is in a hole, next data at offs | |
1750 | * D3. offs < 0, errno = ENXIO: either start is in a trailing hole | |
1751 | * or start is beyond EOF | |
1752 | * If the latter happens, the file has been truncated behind | |
1753 | * our back since we opened it. All bets are off then. | |
1754 | * Treating like a trailing hole is simplest. | |
1755 | * D4. offs < 0, errno != ENXIO: we learned nothing | |
1756 | */ | |
1757 | offs = lseek(s->fd, start, SEEK_DATA); | |
1758 | if (offs < 0) { | |
1759 | return -errno; /* D3 or D4 */ | |
1760 | } | |
1761 | assert(offs >= start); | |
1762 | ||
1763 | if (offs > start) { | |
1764 | /* D2: in hole, next data at offs */ | |
1765 | *hole = start; | |
1766 | *data = offs; | |
1767 | return 0; | |
5500316d PB |
1768 | } |
1769 | ||
d1f06fe6 MA |
1770 | /* D1: in data, end not yet known */ |
1771 | ||
1772 | /* | |
1773 | * SEEK_HOLE cases: | |
1774 | * H1. offs == start: start is in a hole | |
1775 | * If this happens here, a hole has been dug behind our back | |
1776 | * since the previous lseek(). | |
1777 | * H2. offs > start: either start is in data, next hole at offs, | |
1778 | * or start is in trailing hole, EOF at offs | |
1779 | * Linux treats trailing holes like any other hole: offs == | |
1780 | * start. Solaris seeks to EOF instead: offs > start (blech). | |
1781 | * If that happens here, a hole has been dug behind our back | |
1782 | * since the previous lseek(). | |
1783 | * H3. offs < 0, errno = ENXIO: start is beyond EOF | |
1784 | * If this happens, the file has been truncated behind our | |
1785 | * back since we opened it. Treat it like a trailing hole. | |
1786 | * H4. offs < 0, errno != ENXIO: we learned nothing | |
1787 | * Pretend we know nothing at all, i.e. "forget" about D1. | |
1788 | */ | |
1789 | offs = lseek(s->fd, start, SEEK_HOLE); | |
1790 | if (offs < 0) { | |
1791 | return -errno; /* D1 and (H3 or H4) */ | |
1792 | } | |
1793 | assert(offs >= start); | |
1794 | ||
1795 | if (offs > start) { | |
1796 | /* | |
1797 | * D1 and H2: either in data, next hole at offs, or it was in | |
1798 | * data but is now in a trailing hole. In the latter case, | |
1799 | * all bets are off. Treating it as if it there was data all | |
1800 | * the way to EOF is safe, so simply do that. | |
1801 | */ | |
4f11aa8a | 1802 | *data = start; |
d1f06fe6 MA |
1803 | *hole = offs; |
1804 | return 0; | |
5500316d | 1805 | } |
4f11aa8a | 1806 | |
d1f06fe6 MA |
1807 | /* D1 and H1 */ |
1808 | return -EBUSY; | |
5500316d | 1809 | #else |
4f11aa8a | 1810 | return -ENOTSUP; |
5500316d | 1811 | #endif |
4f11aa8a HR |
1812 | } |
1813 | ||
1814 | /* | |
be2ebc6d | 1815 | * Returns the allocation status of the specified sectors. |
4f11aa8a HR |
1816 | * |
1817 | * If 'sector_num' is beyond the end of the disk image the return value is 0 | |
1818 | * and 'pnum' is set to 0. | |
1819 | * | |
1820 | * 'pnum' is set to the number of sectors (including and immediately following | |
1821 | * the specified sector) that are known to be in the same | |
1822 | * allocated/unallocated state. | |
1823 | * | |
1824 | * 'nb_sectors' is the max value 'pnum' should be set to. If nb_sectors goes | |
1825 | * beyond the end of the disk image it will be clamped. | |
1826 | */ | |
1827 | static int64_t coroutine_fn raw_co_get_block_status(BlockDriverState *bs, | |
1828 | int64_t sector_num, | |
1829 | int nb_sectors, int *pnum) | |
1830 | { | |
1831 | off_t start, data = 0, hole = 0; | |
e6d7ec32 | 1832 | int64_t total_size; |
d7f62751 | 1833 | int ret; |
4f11aa8a HR |
1834 | |
1835 | ret = fd_open(bs); | |
1836 | if (ret < 0) { | |
1837 | return ret; | |
1838 | } | |
1839 | ||
1840 | start = sector_num * BDRV_SECTOR_SIZE; | |
e6d7ec32 HR |
1841 | total_size = bdrv_getlength(bs); |
1842 | if (total_size < 0) { | |
1843 | return total_size; | |
1844 | } else if (start >= total_size) { | |
1845 | *pnum = 0; | |
1846 | return 0; | |
1847 | } else if (start + nb_sectors * BDRV_SECTOR_SIZE > total_size) { | |
1848 | nb_sectors = DIV_ROUND_UP(total_size - start, BDRV_SECTOR_SIZE); | |
1849 | } | |
4f11aa8a | 1850 | |
d1f06fe6 MA |
1851 | ret = find_allocation(bs, start, &data, &hole); |
1852 | if (ret == -ENXIO) { | |
1853 | /* Trailing hole */ | |
1854 | *pnum = nb_sectors; | |
1855 | ret = BDRV_BLOCK_ZERO; | |
1856 | } else if (ret < 0) { | |
1857 | /* No info available, so pretend there are no holes */ | |
1858 | *pnum = nb_sectors; | |
1859 | ret = BDRV_BLOCK_DATA; | |
1860 | } else if (data == start) { | |
f4a769ab KW |
1861 | /* On a data extent, compute sectors to the end of the extent, |
1862 | * possibly including a partial sector at EOF. */ | |
1863 | *pnum = MIN(nb_sectors, DIV_ROUND_UP(hole - start, BDRV_SECTOR_SIZE)); | |
d1f06fe6 | 1864 | ret = BDRV_BLOCK_DATA; |
5500316d PB |
1865 | } else { |
1866 | /* On a hole, compute sectors to the beginning of the next extent. */ | |
d1f06fe6 | 1867 | assert(hole == start); |
5500316d | 1868 | *pnum = MIN(nb_sectors, (data - start) / BDRV_SECTOR_SIZE); |
d1f06fe6 | 1869 | ret = BDRV_BLOCK_ZERO; |
5500316d | 1870 | } |
d1f06fe6 | 1871 | return ret | BDRV_BLOCK_OFFSET_VALID | start; |
5500316d PB |
1872 | } |
1873 | ||
7c84b1b8 | 1874 | static coroutine_fn BlockAIOCB *raw_aio_discard(BlockDriverState *bs, |
8238010b | 1875 | int64_t sector_num, int nb_sectors, |
097310b5 | 1876 | BlockCompletionFunc *cb, void *opaque) |
dce512de | 1877 | { |
dce512de CH |
1878 | BDRVRawState *s = bs->opaque; |
1879 | ||
8238010b PB |
1880 | return paio_submit(bs, s->fd, sector_num, NULL, nb_sectors, |
1881 | cb, opaque, QEMU_AIO_DISCARD); | |
dce512de | 1882 | } |
0e7e1989 | 1883 | |
260a82e5 PB |
1884 | static int coroutine_fn raw_co_write_zeroes( |
1885 | BlockDriverState *bs, int64_t sector_num, | |
1886 | int nb_sectors, BdrvRequestFlags flags) | |
1887 | { | |
1888 | BDRVRawState *s = bs->opaque; | |
1889 | ||
1890 | if (!(flags & BDRV_REQ_MAY_UNMAP)) { | |
97a2ae34 PB |
1891 | return paio_submit_co(bs, s->fd, sector_num, NULL, nb_sectors, |
1892 | QEMU_AIO_WRITE_ZEROES); | |
1893 | } else if (s->discard_zeroes) { | |
1894 | return paio_submit_co(bs, s->fd, sector_num, NULL, nb_sectors, | |
1895 | QEMU_AIO_DISCARD); | |
260a82e5 | 1896 | } |
97a2ae34 | 1897 | return -ENOTSUP; |
260a82e5 PB |
1898 | } |
1899 | ||
1900 | static int raw_get_info(BlockDriverState *bs, BlockDriverInfo *bdi) | |
1901 | { | |
1902 | BDRVRawState *s = bs->opaque; | |
1903 | ||
1904 | bdi->unallocated_blocks_are_zero = s->discard_zeroes; | |
1905 | bdi->can_write_zeroes_with_unmap = s->discard_zeroes; | |
1906 | return 0; | |
1907 | } | |
1908 | ||
6f482f74 CL |
1909 | static QemuOptsList raw_create_opts = { |
1910 | .name = "raw-create-opts", | |
1911 | .head = QTAILQ_HEAD_INITIALIZER(raw_create_opts.head), | |
1912 | .desc = { | |
1913 | { | |
1914 | .name = BLOCK_OPT_SIZE, | |
1915 | .type = QEMU_OPT_SIZE, | |
1916 | .help = "Virtual disk size" | |
1917 | }, | |
4ab15590 CL |
1918 | { |
1919 | .name = BLOCK_OPT_NOCOW, | |
1920 | .type = QEMU_OPT_BOOL, | |
1921 | .help = "Turn off copy-on-write (valid only on btrfs)" | |
1922 | }, | |
06247428 HT |
1923 | { |
1924 | .name = BLOCK_OPT_PREALLOC, | |
1925 | .type = QEMU_OPT_STRING, | |
1926 | .help = "Preallocation mode (allowed values: off, falloc, full)" | |
1927 | }, | |
6f482f74 CL |
1928 | { /* end of list */ } |
1929 | } | |
0e7e1989 KW |
1930 | }; |
1931 | ||
5f535a94 | 1932 | BlockDriver bdrv_file = { |
84a12e66 CH |
1933 | .format_name = "file", |
1934 | .protocol_name = "file", | |
856ae5c3 | 1935 | .instance_size = sizeof(BDRVRawState), |
030be321 | 1936 | .bdrv_needs_filename = true, |
856ae5c3 | 1937 | .bdrv_probe = NULL, /* no probe for protocols */ |
078896a9 | 1938 | .bdrv_parse_filename = raw_parse_filename, |
66f82cee | 1939 | .bdrv_file_open = raw_open, |
eeb6b45d JC |
1940 | .bdrv_reopen_prepare = raw_reopen_prepare, |
1941 | .bdrv_reopen_commit = raw_reopen_commit, | |
1942 | .bdrv_reopen_abort = raw_reopen_abort, | |
856ae5c3 | 1943 | .bdrv_close = raw_close, |
c282e1fd | 1944 | .bdrv_create = raw_create, |
3ac21627 | 1945 | .bdrv_has_zero_init = bdrv_has_zero_init_1, |
b6b8a333 | 1946 | .bdrv_co_get_block_status = raw_co_get_block_status, |
260a82e5 | 1947 | .bdrv_co_write_zeroes = raw_co_write_zeroes, |
3b46e624 | 1948 | |
f141eafe AL |
1949 | .bdrv_aio_readv = raw_aio_readv, |
1950 | .bdrv_aio_writev = raw_aio_writev, | |
b2e12bc6 | 1951 | .bdrv_aio_flush = raw_aio_flush, |
8238010b | 1952 | .bdrv_aio_discard = raw_aio_discard, |
c25f53b0 | 1953 | .bdrv_refresh_limits = raw_refresh_limits, |
1b3abdcc ML |
1954 | .bdrv_io_plug = raw_aio_plug, |
1955 | .bdrv_io_unplug = raw_aio_unplug, | |
1956 | .bdrv_flush_io_queue = raw_aio_flush_io_queue, | |
3c529d93 | 1957 | |
83f64091 FB |
1958 | .bdrv_truncate = raw_truncate, |
1959 | .bdrv_getlength = raw_getlength, | |
260a82e5 | 1960 | .bdrv_get_info = raw_get_info, |
4a1d5e1f FZ |
1961 | .bdrv_get_allocated_file_size |
1962 | = raw_get_allocated_file_size, | |
0e7e1989 | 1963 | |
c2f3426c SH |
1964 | .bdrv_detach_aio_context = raw_detach_aio_context, |
1965 | .bdrv_attach_aio_context = raw_attach_aio_context, | |
1966 | ||
6f482f74 | 1967 | .create_opts = &raw_create_opts, |
83f64091 FB |
1968 | }; |
1969 | ||
19cb3738 FB |
1970 | /***********************************************/ |
1971 | /* host device */ | |
1972 | ||
83affaa6 | 1973 | #if defined(__APPLE__) && defined(__MACH__) |
19cb3738 FB |
1974 | static kern_return_t FindEjectableCDMedia( io_iterator_t *mediaIterator ); |
1975 | static kern_return_t GetBSDPath( io_iterator_t mediaIterator, char *bsdPath, CFIndex maxPathSize ); | |
1976 | ||
1977 | kern_return_t FindEjectableCDMedia( io_iterator_t *mediaIterator ) | |
1978 | { | |
5fafdf24 | 1979 | kern_return_t kernResult; |
19cb3738 FB |
1980 | mach_port_t masterPort; |
1981 | CFMutableDictionaryRef classesToMatch; | |
1982 | ||
1983 | kernResult = IOMasterPort( MACH_PORT_NULL, &masterPort ); | |
1984 | if ( KERN_SUCCESS != kernResult ) { | |
1985 | printf( "IOMasterPort returned %d\n", kernResult ); | |
1986 | } | |
3b46e624 | 1987 | |
5fafdf24 | 1988 | classesToMatch = IOServiceMatching( kIOCDMediaClass ); |
19cb3738 FB |
1989 | if ( classesToMatch == NULL ) { |
1990 | printf( "IOServiceMatching returned a NULL dictionary.\n" ); | |
1991 | } else { | |
1992 | CFDictionarySetValue( classesToMatch, CFSTR( kIOMediaEjectableKey ), kCFBooleanTrue ); | |
1993 | } | |
1994 | kernResult = IOServiceGetMatchingServices( masterPort, classesToMatch, mediaIterator ); | |
1995 | if ( KERN_SUCCESS != kernResult ) | |
1996 | { | |
1997 | printf( "IOServiceGetMatchingServices returned %d\n", kernResult ); | |
1998 | } | |
3b46e624 | 1999 | |
19cb3738 FB |
2000 | return kernResult; |
2001 | } | |
2002 | ||
2003 | kern_return_t GetBSDPath( io_iterator_t mediaIterator, char *bsdPath, CFIndex maxPathSize ) | |
2004 | { | |
2005 | io_object_t nextMedia; | |
2006 | kern_return_t kernResult = KERN_FAILURE; | |
2007 | *bsdPath = '\0'; | |
2008 | nextMedia = IOIteratorNext( mediaIterator ); | |
2009 | if ( nextMedia ) | |
2010 | { | |
2011 | CFTypeRef bsdPathAsCFString; | |
2012 | bsdPathAsCFString = IORegistryEntryCreateCFProperty( nextMedia, CFSTR( kIOBSDNameKey ), kCFAllocatorDefault, 0 ); | |
2013 | if ( bsdPathAsCFString ) { | |
2014 | size_t devPathLength; | |
2015 | strcpy( bsdPath, _PATH_DEV ); | |
2016 | strcat( bsdPath, "r" ); | |
2017 | devPathLength = strlen( bsdPath ); | |
2018 | if ( CFStringGetCString( bsdPathAsCFString, bsdPath + devPathLength, maxPathSize - devPathLength, kCFStringEncodingASCII ) ) { | |
2019 | kernResult = KERN_SUCCESS; | |
2020 | } | |
2021 | CFRelease( bsdPathAsCFString ); | |
2022 | } | |
2023 | IOObjectRelease( nextMedia ); | |
2024 | } | |
3b46e624 | 2025 | |
19cb3738 FB |
2026 | return kernResult; |
2027 | } | |
2028 | ||
2029 | #endif | |
2030 | ||
508c7cb3 CH |
2031 | static int hdev_probe_device(const char *filename) |
2032 | { | |
2033 | struct stat st; | |
2034 | ||
2035 | /* allow a dedicated CD-ROM driver to match with a higher priority */ | |
2036 | if (strstart(filename, "/dev/cdrom", NULL)) | |
2037 | return 50; | |
2038 | ||
2039 | if (stat(filename, &st) >= 0 && | |
2040 | (S_ISCHR(st.st_mode) || S_ISBLK(st.st_mode))) { | |
2041 | return 100; | |
2042 | } | |
2043 | ||
2044 | return 0; | |
2045 | } | |
2046 | ||
da888d37 SH |
2047 | static int check_hdev_writable(BDRVRawState *s) |
2048 | { | |
2049 | #if defined(BLKROGET) | |
2050 | /* Linux block devices can be configured "read-only" using blockdev(8). | |
2051 | * This is independent of device node permissions and therefore open(2) | |
2052 | * with O_RDWR succeeds. Actual writes fail with EPERM. | |
2053 | * | |
2054 | * bdrv_open() is supposed to fail if the disk is read-only. Explicitly | |
2055 | * check for read-only block devices so that Linux block devices behave | |
2056 | * properly. | |
2057 | */ | |
2058 | struct stat st; | |
2059 | int readonly = 0; | |
2060 | ||
2061 | if (fstat(s->fd, &st)) { | |
2062 | return -errno; | |
2063 | } | |
2064 | ||
2065 | if (!S_ISBLK(st.st_mode)) { | |
2066 | return 0; | |
2067 | } | |
2068 | ||
2069 | if (ioctl(s->fd, BLKROGET, &readonly) < 0) { | |
2070 | return -errno; | |
2071 | } | |
2072 | ||
2073 | if (readonly) { | |
2074 | return -EACCES; | |
2075 | } | |
2076 | #endif /* defined(BLKROGET) */ | |
2077 | return 0; | |
2078 | } | |
2079 | ||
7af803d4 HR |
2080 | static void hdev_parse_filename(const char *filename, QDict *options, |
2081 | Error **errp) | |
2082 | { | |
2083 | /* The prefix is optional, just as for "file". */ | |
2084 | strstart(filename, "host_device:", &filename); | |
2085 | ||
2086 | qdict_put_obj(options, "filename", QOBJECT(qstring_from_str(filename))); | |
2087 | } | |
2088 | ||
015a1036 HR |
2089 | static int hdev_open(BlockDriverState *bs, QDict *options, int flags, |
2090 | Error **errp) | |
19cb3738 FB |
2091 | { |
2092 | BDRVRawState *s = bs->opaque; | |
e428e439 | 2093 | Error *local_err = NULL; |
da888d37 | 2094 | int ret; |
c66a6157 | 2095 | const char *filename = qdict_get_str(options, "filename"); |
a76bab49 | 2096 | |
83affaa6 | 2097 | #if defined(__APPLE__) && defined(__MACH__) |
19cb3738 FB |
2098 | if (strstart(filename, "/dev/cdrom", NULL)) { |
2099 | kern_return_t kernResult; | |
2100 | io_iterator_t mediaIterator; | |
2101 | char bsdPath[ MAXPATHLEN ]; | |
2102 | int fd; | |
5fafdf24 | 2103 | |
19cb3738 FB |
2104 | kernResult = FindEjectableCDMedia( &mediaIterator ); |
2105 | kernResult = GetBSDPath( mediaIterator, bsdPath, sizeof( bsdPath ) ); | |
3b46e624 | 2106 | |
19cb3738 FB |
2107 | if ( bsdPath[ 0 ] != '\0' ) { |
2108 | strcat(bsdPath,"s0"); | |
2109 | /* some CDs don't have a partition 0 */ | |
6165f4d8 | 2110 | fd = qemu_open(bsdPath, O_RDONLY | O_BINARY | O_LARGEFILE); |
19cb3738 FB |
2111 | if (fd < 0) { |
2112 | bsdPath[strlen(bsdPath)-1] = '1'; | |
2113 | } else { | |
2e1e79da | 2114 | qemu_close(fd); |
19cb3738 FB |
2115 | } |
2116 | filename = bsdPath; | |
a5c5ea3f | 2117 | qdict_put(options, "filename", qstring_from_str(filename)); |
19cb3738 | 2118 | } |
3b46e624 | 2119 | |
19cb3738 FB |
2120 | if ( mediaIterator ) |
2121 | IOObjectRelease( mediaIterator ); | |
2122 | } | |
2123 | #endif | |
19cb3738 FB |
2124 | |
2125 | s->type = FTYPE_FILE; | |
4dd75c70 | 2126 | #if defined(__linux__) |
05acda4d BK |
2127 | { |
2128 | char resolved_path[ MAXPATHLEN ], *temp; | |
2129 | ||
2130 | temp = realpath(filename, resolved_path); | |
2131 | if (temp && strstart(temp, "/dev/sg", NULL)) { | |
2132 | bs->sg = 1; | |
2133 | } | |
19cb3738 FB |
2134 | } |
2135 | #endif | |
90babde0 | 2136 | |
e428e439 | 2137 | ret = raw_open_common(bs, options, flags, 0, &local_err); |
da888d37 | 2138 | if (ret < 0) { |
84d18f06 | 2139 | if (local_err) { |
e428e439 HR |
2140 | error_propagate(errp, local_err); |
2141 | } | |
da888d37 SH |
2142 | return ret; |
2143 | } | |
2144 | ||
2145 | if (flags & BDRV_O_RDWR) { | |
2146 | ret = check_hdev_writable(s); | |
2147 | if (ret < 0) { | |
2148 | raw_close(bs); | |
e428e439 | 2149 | error_setg_errno(errp, -ret, "The device is not writable"); |
da888d37 SH |
2150 | return ret; |
2151 | } | |
2152 | } | |
2153 | ||
2154 | return ret; | |
19cb3738 FB |
2155 | } |
2156 | ||
03ff3ca3 | 2157 | #if defined(__linux__) |
19cb3738 FB |
2158 | /* Note: we do not have a reliable method to detect if the floppy is |
2159 | present. The current method is to try to open the floppy at every | |
2160 | I/O and to keep it opened during a few hundreds of ms. */ | |
2161 | static int fd_open(BlockDriverState *bs) | |
2162 | { | |
2163 | BDRVRawState *s = bs->opaque; | |
2164 | int last_media_present; | |
2165 | ||
2166 | if (s->type != FTYPE_FD) | |
2167 | return 0; | |
2168 | last_media_present = (s->fd >= 0); | |
5fafdf24 | 2169 | if (s->fd >= 0 && |
a56ebc6b | 2170 | (qemu_clock_get_ns(QEMU_CLOCK_REALTIME) - s->fd_open_time) >= FD_OPEN_TIMEOUT) { |
2e1e79da | 2171 | qemu_close(s->fd); |
19cb3738 | 2172 | s->fd = -1; |
a93a3982 | 2173 | DPRINTF("Floppy closed\n"); |
19cb3738 FB |
2174 | } |
2175 | if (s->fd < 0) { | |
5fafdf24 | 2176 | if (s->fd_got_error && |
a56ebc6b | 2177 | (qemu_clock_get_ns(QEMU_CLOCK_REALTIME) - s->fd_error_time) < FD_OPEN_TIMEOUT) { |
a93a3982 | 2178 | DPRINTF("No floppy (open delayed)\n"); |
19cb3738 FB |
2179 | return -EIO; |
2180 | } | |
6165f4d8 | 2181 | s->fd = qemu_open(bs->filename, s->open_flags & ~O_NONBLOCK); |
19cb3738 | 2182 | if (s->fd < 0) { |
a56ebc6b | 2183 | s->fd_error_time = qemu_clock_get_ns(QEMU_CLOCK_REALTIME); |
19cb3738 FB |
2184 | s->fd_got_error = 1; |
2185 | if (last_media_present) | |
2186 | s->fd_media_changed = 1; | |
a93a3982 | 2187 | DPRINTF("No floppy\n"); |
19cb3738 FB |
2188 | return -EIO; |
2189 | } | |
a93a3982 | 2190 | DPRINTF("Floppy opened\n"); |
19cb3738 FB |
2191 | } |
2192 | if (!last_media_present) | |
2193 | s->fd_media_changed = 1; | |
a56ebc6b | 2194 | s->fd_open_time = qemu_clock_get_ns(QEMU_CLOCK_REALTIME); |
19cb3738 FB |
2195 | s->fd_got_error = 0; |
2196 | return 0; | |
2197 | } | |
19cb3738 | 2198 | |
63ec93db | 2199 | static int hdev_ioctl(BlockDriverState *bs, unsigned long int req, void *buf) |
985a03b0 TS |
2200 | { |
2201 | BDRVRawState *s = bs->opaque; | |
2202 | ||
2203 | return ioctl(s->fd, req, buf); | |
2204 | } | |
221f715d | 2205 | |
7c84b1b8 | 2206 | static BlockAIOCB *hdev_aio_ioctl(BlockDriverState *bs, |
221f715d | 2207 | unsigned long int req, void *buf, |
097310b5 | 2208 | BlockCompletionFunc *cb, void *opaque) |
221f715d | 2209 | { |
f141eafe | 2210 | BDRVRawState *s = bs->opaque; |
c208e8c2 | 2211 | RawPosixAIOData *acb; |
c4d9d196 | 2212 | ThreadPool *pool; |
221f715d | 2213 | |
f141eafe AL |
2214 | if (fd_open(bs) < 0) |
2215 | return NULL; | |
c208e8c2 PB |
2216 | |
2217 | acb = g_slice_new(RawPosixAIOData); | |
2218 | acb->bs = bs; | |
2219 | acb->aio_type = QEMU_AIO_IOCTL; | |
2220 | acb->aio_fildes = s->fd; | |
2221 | acb->aio_offset = 0; | |
2222 | acb->aio_ioctl_buf = buf; | |
2223 | acb->aio_ioctl_cmd = req; | |
c4d9d196 SH |
2224 | pool = aio_get_thread_pool(bdrv_get_aio_context(bs)); |
2225 | return thread_pool_submit_aio(pool, aio_worker, acb, cb, opaque); | |
221f715d AL |
2226 | } |
2227 | ||
a167ba50 | 2228 | #elif defined(__FreeBSD__) || defined(__FreeBSD_kernel__) |
9f23011a BS |
2229 | static int fd_open(BlockDriverState *bs) |
2230 | { | |
2231 | BDRVRawState *s = bs->opaque; | |
2232 | ||
2233 | /* this is just to ensure s->fd is sane (its called by io ops) */ | |
2234 | if (s->fd >= 0) | |
2235 | return 0; | |
2236 | return -EIO; | |
2237 | } | |
9f23011a | 2238 | #else /* !linux && !FreeBSD */ |
19cb3738 | 2239 | |
08af02e2 AL |
2240 | static int fd_open(BlockDriverState *bs) |
2241 | { | |
2242 | return 0; | |
2243 | } | |
2244 | ||
221f715d | 2245 | #endif /* !linux && !FreeBSD */ |
04eeb8b6 | 2246 | |
7c84b1b8 | 2247 | static coroutine_fn BlockAIOCB *hdev_aio_discard(BlockDriverState *bs, |
c36dd8a0 | 2248 | int64_t sector_num, int nb_sectors, |
097310b5 | 2249 | BlockCompletionFunc *cb, void *opaque) |
c36dd8a0 AF |
2250 | { |
2251 | BDRVRawState *s = bs->opaque; | |
2252 | ||
2253 | if (fd_open(bs) < 0) { | |
2254 | return NULL; | |
2255 | } | |
2256 | return paio_submit(bs, s->fd, sector_num, NULL, nb_sectors, | |
2257 | cb, opaque, QEMU_AIO_DISCARD|QEMU_AIO_BLKDEV); | |
2258 | } | |
2259 | ||
d0b4503e PB |
2260 | static coroutine_fn int hdev_co_write_zeroes(BlockDriverState *bs, |
2261 | int64_t sector_num, int nb_sectors, BdrvRequestFlags flags) | |
2262 | { | |
2263 | BDRVRawState *s = bs->opaque; | |
2264 | int rc; | |
2265 | ||
2266 | rc = fd_open(bs); | |
2267 | if (rc < 0) { | |
2268 | return rc; | |
2269 | } | |
2270 | if (!(flags & BDRV_REQ_MAY_UNMAP)) { | |
97a2ae34 PB |
2271 | return paio_submit_co(bs, s->fd, sector_num, NULL, nb_sectors, |
2272 | QEMU_AIO_WRITE_ZEROES|QEMU_AIO_BLKDEV); | |
2273 | } else if (s->discard_zeroes) { | |
2274 | return paio_submit_co(bs, s->fd, sector_num, NULL, nb_sectors, | |
2275 | QEMU_AIO_DISCARD|QEMU_AIO_BLKDEV); | |
d0b4503e | 2276 | } |
97a2ae34 | 2277 | return -ENOTSUP; |
d0b4503e PB |
2278 | } |
2279 | ||
6f482f74 | 2280 | static int hdev_create(const char *filename, QemuOpts *opts, |
d5124c00 | 2281 | Error **errp) |
93c65b47 AL |
2282 | { |
2283 | int fd; | |
2284 | int ret = 0; | |
2285 | struct stat stat_buf; | |
0e7e1989 | 2286 | int64_t total_size = 0; |
cc28c6aa HR |
2287 | bool has_prefix; |
2288 | ||
2289 | /* This function is used by all three protocol block drivers and therefore | |
2290 | * any of these three prefixes may be given. | |
2291 | * The return value has to be stored somewhere, otherwise this is an error | |
2292 | * due to -Werror=unused-value. */ | |
2293 | has_prefix = | |
2294 | strstart(filename, "host_device:", &filename) || | |
2295 | strstart(filename, "host_cdrom:" , &filename) || | |
2296 | strstart(filename, "host_floppy:", &filename); | |
2297 | ||
2298 | (void)has_prefix; | |
93c65b47 | 2299 | |
0e7e1989 | 2300 | /* Read out options */ |
180e9526 HT |
2301 | total_size = ROUND_UP(qemu_opt_get_size_del(opts, BLOCK_OPT_SIZE, 0), |
2302 | BDRV_SECTOR_SIZE); | |
93c65b47 | 2303 | |
6165f4d8 | 2304 | fd = qemu_open(filename, O_WRONLY | O_BINARY); |
e428e439 HR |
2305 | if (fd < 0) { |
2306 | ret = -errno; | |
2307 | error_setg_errno(errp, -ret, "Could not open device"); | |
2308 | return ret; | |
2309 | } | |
93c65b47 | 2310 | |
e428e439 | 2311 | if (fstat(fd, &stat_buf) < 0) { |
57e69b7d | 2312 | ret = -errno; |
e428e439 HR |
2313 | error_setg_errno(errp, -ret, "Could not stat device"); |
2314 | } else if (!S_ISBLK(stat_buf.st_mode) && !S_ISCHR(stat_buf.st_mode)) { | |
2315 | error_setg(errp, | |
2316 | "The given file is neither a block nor a character device"); | |
57e69b7d | 2317 | ret = -ENODEV; |
180e9526 | 2318 | } else if (lseek(fd, 0, SEEK_END) < total_size) { |
e428e439 | 2319 | error_setg(errp, "Device is too small"); |
93c65b47 | 2320 | ret = -ENOSPC; |
e428e439 | 2321 | } |
93c65b47 | 2322 | |
2e1e79da | 2323 | qemu_close(fd); |
93c65b47 AL |
2324 | return ret; |
2325 | } | |
2326 | ||
5efa9d5a | 2327 | static BlockDriver bdrv_host_device = { |
0b4ce02e | 2328 | .format_name = "host_device", |
84a12e66 | 2329 | .protocol_name = "host_device", |
0b4ce02e | 2330 | .instance_size = sizeof(BDRVRawState), |
030be321 | 2331 | .bdrv_needs_filename = true, |
0b4ce02e | 2332 | .bdrv_probe_device = hdev_probe_device, |
7af803d4 | 2333 | .bdrv_parse_filename = hdev_parse_filename, |
66f82cee | 2334 | .bdrv_file_open = hdev_open, |
0b4ce02e | 2335 | .bdrv_close = raw_close, |
1bc6b705 JC |
2336 | .bdrv_reopen_prepare = raw_reopen_prepare, |
2337 | .bdrv_reopen_commit = raw_reopen_commit, | |
2338 | .bdrv_reopen_abort = raw_reopen_abort, | |
c282e1fd | 2339 | .bdrv_create = hdev_create, |
6f482f74 | 2340 | .create_opts = &raw_create_opts, |
d0b4503e | 2341 | .bdrv_co_write_zeroes = hdev_co_write_zeroes, |
3b46e624 | 2342 | |
f141eafe AL |
2343 | .bdrv_aio_readv = raw_aio_readv, |
2344 | .bdrv_aio_writev = raw_aio_writev, | |
b2e12bc6 | 2345 | .bdrv_aio_flush = raw_aio_flush, |
8238010b | 2346 | .bdrv_aio_discard = hdev_aio_discard, |
c25f53b0 | 2347 | .bdrv_refresh_limits = raw_refresh_limits, |
1b3abdcc ML |
2348 | .bdrv_io_plug = raw_aio_plug, |
2349 | .bdrv_io_unplug = raw_aio_unplug, | |
2350 | .bdrv_flush_io_queue = raw_aio_flush_io_queue, | |
3c529d93 | 2351 | |
55b949c8 | 2352 | .bdrv_truncate = raw_truncate, |
e60f469c | 2353 | .bdrv_getlength = raw_getlength, |
260a82e5 | 2354 | .bdrv_get_info = raw_get_info, |
4a1d5e1f FZ |
2355 | .bdrv_get_allocated_file_size |
2356 | = raw_get_allocated_file_size, | |
1a9335e4 ET |
2357 | .bdrv_probe_blocksizes = hdev_probe_blocksizes, |
2358 | .bdrv_probe_geometry = hdev_probe_geometry, | |
19cb3738 | 2359 | |
c2f3426c SH |
2360 | .bdrv_detach_aio_context = raw_detach_aio_context, |
2361 | .bdrv_attach_aio_context = raw_attach_aio_context, | |
2362 | ||
f3a5d3f8 | 2363 | /* generic scsi device */ |
63ec93db CH |
2364 | #ifdef __linux__ |
2365 | .bdrv_ioctl = hdev_ioctl, | |
63ec93db CH |
2366 | .bdrv_aio_ioctl = hdev_aio_ioctl, |
2367 | #endif | |
f3a5d3f8 CH |
2368 | }; |
2369 | ||
2370 | #ifdef __linux__ | |
d3f49845 HR |
2371 | static void floppy_parse_filename(const char *filename, QDict *options, |
2372 | Error **errp) | |
2373 | { | |
2374 | /* The prefix is optional, just as for "file". */ | |
2375 | strstart(filename, "host_floppy:", &filename); | |
2376 | ||
2377 | qdict_put_obj(options, "filename", QOBJECT(qstring_from_str(filename))); | |
2378 | } | |
2379 | ||
015a1036 HR |
2380 | static int floppy_open(BlockDriverState *bs, QDict *options, int flags, |
2381 | Error **errp) | |
f3a5d3f8 CH |
2382 | { |
2383 | BDRVRawState *s = bs->opaque; | |
e428e439 | 2384 | Error *local_err = NULL; |
f3a5d3f8 CH |
2385 | int ret; |
2386 | ||
f3a5d3f8 | 2387 | s->type = FTYPE_FD; |
f3a5d3f8 | 2388 | |
19a3da7f | 2389 | /* open will not fail even if no floppy is inserted, so add O_NONBLOCK */ |
e428e439 HR |
2390 | ret = raw_open_common(bs, options, flags, O_NONBLOCK, &local_err); |
2391 | if (ret) { | |
84d18f06 | 2392 | if (local_err) { |
e428e439 HR |
2393 | error_propagate(errp, local_err); |
2394 | } | |
f3a5d3f8 | 2395 | return ret; |
e428e439 | 2396 | } |
f3a5d3f8 CH |
2397 | |
2398 | /* close fd so that we can reopen it as needed */ | |
2e1e79da | 2399 | qemu_close(s->fd); |
f3a5d3f8 CH |
2400 | s->fd = -1; |
2401 | s->fd_media_changed = 1; | |
2402 | ||
92a539d2 MA |
2403 | error_report("Host floppy pass-through is deprecated"); |
2404 | error_printf("Support for it will be removed in a future release.\n"); | |
f3a5d3f8 CH |
2405 | return 0; |
2406 | } | |
2407 | ||
508c7cb3 CH |
2408 | static int floppy_probe_device(const char *filename) |
2409 | { | |
2ebf7c4b CR |
2410 | int fd, ret; |
2411 | int prio = 0; | |
2412 | struct floppy_struct fdparam; | |
343f8568 | 2413 | struct stat st; |
2ebf7c4b | 2414 | |
e1740828 CB |
2415 | if (strstart(filename, "/dev/fd", NULL) && |
2416 | !strstart(filename, "/dev/fdset/", NULL)) { | |
2ebf7c4b | 2417 | prio = 50; |
e1740828 | 2418 | } |
2ebf7c4b | 2419 | |
6165f4d8 | 2420 | fd = qemu_open(filename, O_RDONLY | O_NONBLOCK); |
2ebf7c4b CR |
2421 | if (fd < 0) { |
2422 | goto out; | |
2423 | } | |
343f8568 JS |
2424 | ret = fstat(fd, &st); |
2425 | if (ret == -1 || !S_ISBLK(st.st_mode)) { | |
2426 | goto outc; | |
2427 | } | |
2ebf7c4b CR |
2428 | |
2429 | /* Attempt to detect via a floppy specific ioctl */ | |
2430 | ret = ioctl(fd, FDGETPRM, &fdparam); | |
2431 | if (ret >= 0) | |
2432 | prio = 100; | |
2433 | ||
343f8568 | 2434 | outc: |
2e1e79da | 2435 | qemu_close(fd); |
2ebf7c4b CR |
2436 | out: |
2437 | return prio; | |
508c7cb3 CH |
2438 | } |
2439 | ||
2440 | ||
f3a5d3f8 CH |
2441 | static int floppy_is_inserted(BlockDriverState *bs) |
2442 | { | |
2443 | return fd_open(bs) >= 0; | |
2444 | } | |
2445 | ||
2446 | static int floppy_media_changed(BlockDriverState *bs) | |
2447 | { | |
2448 | BDRVRawState *s = bs->opaque; | |
2449 | int ret; | |
2450 | ||
2451 | /* | |
2452 | * XXX: we do not have a true media changed indication. | |
2453 | * It does not work if the floppy is changed without trying to read it. | |
2454 | */ | |
2455 | fd_open(bs); | |
2456 | ret = s->fd_media_changed; | |
2457 | s->fd_media_changed = 0; | |
a93a3982 | 2458 | DPRINTF("Floppy changed=%d\n", ret); |
f3a5d3f8 CH |
2459 | return ret; |
2460 | } | |
2461 | ||
f36f3949 | 2462 | static void floppy_eject(BlockDriverState *bs, bool eject_flag) |
f3a5d3f8 CH |
2463 | { |
2464 | BDRVRawState *s = bs->opaque; | |
2465 | int fd; | |
2466 | ||
2467 | if (s->fd >= 0) { | |
2e1e79da | 2468 | qemu_close(s->fd); |
f3a5d3f8 CH |
2469 | s->fd = -1; |
2470 | } | |
6165f4d8 | 2471 | fd = qemu_open(bs->filename, s->open_flags | O_NONBLOCK); |
f3a5d3f8 CH |
2472 | if (fd >= 0) { |
2473 | if (ioctl(fd, FDEJECT, 0) < 0) | |
2474 | perror("FDEJECT"); | |
2e1e79da | 2475 | qemu_close(fd); |
f3a5d3f8 | 2476 | } |
f3a5d3f8 CH |
2477 | } |
2478 | ||
2479 | static BlockDriver bdrv_host_floppy = { | |
2480 | .format_name = "host_floppy", | |
84a12e66 | 2481 | .protocol_name = "host_floppy", |
f3a5d3f8 | 2482 | .instance_size = sizeof(BDRVRawState), |
030be321 | 2483 | .bdrv_needs_filename = true, |
508c7cb3 | 2484 | .bdrv_probe_device = floppy_probe_device, |
d3f49845 | 2485 | .bdrv_parse_filename = floppy_parse_filename, |
66f82cee | 2486 | .bdrv_file_open = floppy_open, |
f3a5d3f8 | 2487 | .bdrv_close = raw_close, |
1bc6b705 JC |
2488 | .bdrv_reopen_prepare = raw_reopen_prepare, |
2489 | .bdrv_reopen_commit = raw_reopen_commit, | |
2490 | .bdrv_reopen_abort = raw_reopen_abort, | |
c282e1fd | 2491 | .bdrv_create = hdev_create, |
6f482f74 | 2492 | .create_opts = &raw_create_opts, |
f3a5d3f8 | 2493 | |
f3a5d3f8 CH |
2494 | .bdrv_aio_readv = raw_aio_readv, |
2495 | .bdrv_aio_writev = raw_aio_writev, | |
b2e12bc6 | 2496 | .bdrv_aio_flush = raw_aio_flush, |
c25f53b0 | 2497 | .bdrv_refresh_limits = raw_refresh_limits, |
1b3abdcc ML |
2498 | .bdrv_io_plug = raw_aio_plug, |
2499 | .bdrv_io_unplug = raw_aio_unplug, | |
2500 | .bdrv_flush_io_queue = raw_aio_flush_io_queue, | |
f3a5d3f8 | 2501 | |
55b949c8 | 2502 | .bdrv_truncate = raw_truncate, |
b94a2610 KW |
2503 | .bdrv_getlength = raw_getlength, |
2504 | .has_variable_length = true, | |
4a1d5e1f FZ |
2505 | .bdrv_get_allocated_file_size |
2506 | = raw_get_allocated_file_size, | |
f3a5d3f8 | 2507 | |
c2f3426c SH |
2508 | .bdrv_detach_aio_context = raw_detach_aio_context, |
2509 | .bdrv_attach_aio_context = raw_attach_aio_context, | |
2510 | ||
f3a5d3f8 CH |
2511 | /* removable device support */ |
2512 | .bdrv_is_inserted = floppy_is_inserted, | |
2513 | .bdrv_media_changed = floppy_media_changed, | |
2514 | .bdrv_eject = floppy_eject, | |
f3a5d3f8 | 2515 | }; |
18fa1c42 | 2516 | #endif |
f3a5d3f8 | 2517 | |
18fa1c42 HR |
2518 | #if defined(__linux__) || defined(__FreeBSD__) || defined(__FreeBSD_kernel__) |
2519 | static void cdrom_parse_filename(const char *filename, QDict *options, | |
2520 | Error **errp) | |
2521 | { | |
2522 | /* The prefix is optional, just as for "file". */ | |
2523 | strstart(filename, "host_cdrom:", &filename); | |
2524 | ||
2525 | qdict_put_obj(options, "filename", QOBJECT(qstring_from_str(filename))); | |
2526 | } | |
2527 | #endif | |
2528 | ||
2529 | #ifdef __linux__ | |
015a1036 HR |
2530 | static int cdrom_open(BlockDriverState *bs, QDict *options, int flags, |
2531 | Error **errp) | |
f3a5d3f8 CH |
2532 | { |
2533 | BDRVRawState *s = bs->opaque; | |
e428e439 HR |
2534 | Error *local_err = NULL; |
2535 | int ret; | |
f3a5d3f8 | 2536 | |
f3a5d3f8 CH |
2537 | s->type = FTYPE_CD; |
2538 | ||
19a3da7f | 2539 | /* open will not fail even if no CD is inserted, so add O_NONBLOCK */ |
e428e439 | 2540 | ret = raw_open_common(bs, options, flags, O_NONBLOCK, &local_err); |
84d18f06 | 2541 | if (local_err) { |
e428e439 HR |
2542 | error_propagate(errp, local_err); |
2543 | } | |
2544 | return ret; | |
f3a5d3f8 CH |
2545 | } |
2546 | ||
508c7cb3 CH |
2547 | static int cdrom_probe_device(const char *filename) |
2548 | { | |
3baf720e CR |
2549 | int fd, ret; |
2550 | int prio = 0; | |
343f8568 | 2551 | struct stat st; |
3baf720e | 2552 | |
6165f4d8 | 2553 | fd = qemu_open(filename, O_RDONLY | O_NONBLOCK); |
3baf720e CR |
2554 | if (fd < 0) { |
2555 | goto out; | |
2556 | } | |
343f8568 JS |
2557 | ret = fstat(fd, &st); |
2558 | if (ret == -1 || !S_ISBLK(st.st_mode)) { | |
2559 | goto outc; | |
2560 | } | |
3baf720e CR |
2561 | |
2562 | /* Attempt to detect via a CDROM specific ioctl */ | |
2563 | ret = ioctl(fd, CDROM_DRIVE_STATUS, CDSL_CURRENT); | |
2564 | if (ret >= 0) | |
2565 | prio = 100; | |
2566 | ||
343f8568 | 2567 | outc: |
2e1e79da | 2568 | qemu_close(fd); |
3baf720e CR |
2569 | out: |
2570 | return prio; | |
508c7cb3 CH |
2571 | } |
2572 | ||
f3a5d3f8 CH |
2573 | static int cdrom_is_inserted(BlockDriverState *bs) |
2574 | { | |
2575 | BDRVRawState *s = bs->opaque; | |
2576 | int ret; | |
2577 | ||
2578 | ret = ioctl(s->fd, CDROM_DRIVE_STATUS, CDSL_CURRENT); | |
2579 | if (ret == CDS_DISC_OK) | |
2580 | return 1; | |
2581 | return 0; | |
2582 | } | |
2583 | ||
f36f3949 | 2584 | static void cdrom_eject(BlockDriverState *bs, bool eject_flag) |
f3a5d3f8 CH |
2585 | { |
2586 | BDRVRawState *s = bs->opaque; | |
2587 | ||
2588 | if (eject_flag) { | |
2589 | if (ioctl(s->fd, CDROMEJECT, NULL) < 0) | |
2590 | perror("CDROMEJECT"); | |
2591 | } else { | |
2592 | if (ioctl(s->fd, CDROMCLOSETRAY, NULL) < 0) | |
2593 | perror("CDROMEJECT"); | |
2594 | } | |
f3a5d3f8 CH |
2595 | } |
2596 | ||
025e849a | 2597 | static void cdrom_lock_medium(BlockDriverState *bs, bool locked) |
f3a5d3f8 CH |
2598 | { |
2599 | BDRVRawState *s = bs->opaque; | |
2600 | ||
2601 | if (ioctl(s->fd, CDROM_LOCKDOOR, locked) < 0) { | |
2602 | /* | |
2603 | * Note: an error can happen if the distribution automatically | |
2604 | * mounts the CD-ROM | |
2605 | */ | |
2606 | /* perror("CDROM_LOCKDOOR"); */ | |
2607 | } | |
f3a5d3f8 CH |
2608 | } |
2609 | ||
2610 | static BlockDriver bdrv_host_cdrom = { | |
2611 | .format_name = "host_cdrom", | |
84a12e66 | 2612 | .protocol_name = "host_cdrom", |
f3a5d3f8 | 2613 | .instance_size = sizeof(BDRVRawState), |
030be321 | 2614 | .bdrv_needs_filename = true, |
508c7cb3 | 2615 | .bdrv_probe_device = cdrom_probe_device, |
18fa1c42 | 2616 | .bdrv_parse_filename = cdrom_parse_filename, |
66f82cee | 2617 | .bdrv_file_open = cdrom_open, |
f3a5d3f8 | 2618 | .bdrv_close = raw_close, |
1bc6b705 JC |
2619 | .bdrv_reopen_prepare = raw_reopen_prepare, |
2620 | .bdrv_reopen_commit = raw_reopen_commit, | |
2621 | .bdrv_reopen_abort = raw_reopen_abort, | |
c282e1fd | 2622 | .bdrv_create = hdev_create, |
6f482f74 | 2623 | .create_opts = &raw_create_opts, |
f3a5d3f8 | 2624 | |
f3a5d3f8 CH |
2625 | .bdrv_aio_readv = raw_aio_readv, |
2626 | .bdrv_aio_writev = raw_aio_writev, | |
b2e12bc6 | 2627 | .bdrv_aio_flush = raw_aio_flush, |
c25f53b0 | 2628 | .bdrv_refresh_limits = raw_refresh_limits, |
1b3abdcc ML |
2629 | .bdrv_io_plug = raw_aio_plug, |
2630 | .bdrv_io_unplug = raw_aio_unplug, | |
2631 | .bdrv_flush_io_queue = raw_aio_flush_io_queue, | |
f3a5d3f8 | 2632 | |
55b949c8 | 2633 | .bdrv_truncate = raw_truncate, |
b94a2610 KW |
2634 | .bdrv_getlength = raw_getlength, |
2635 | .has_variable_length = true, | |
4a1d5e1f FZ |
2636 | .bdrv_get_allocated_file_size |
2637 | = raw_get_allocated_file_size, | |
f3a5d3f8 | 2638 | |
c2f3426c SH |
2639 | .bdrv_detach_aio_context = raw_detach_aio_context, |
2640 | .bdrv_attach_aio_context = raw_attach_aio_context, | |
2641 | ||
f3a5d3f8 CH |
2642 | /* removable device support */ |
2643 | .bdrv_is_inserted = cdrom_is_inserted, | |
2644 | .bdrv_eject = cdrom_eject, | |
025e849a | 2645 | .bdrv_lock_medium = cdrom_lock_medium, |
f3a5d3f8 CH |
2646 | |
2647 | /* generic scsi device */ | |
63ec93db | 2648 | .bdrv_ioctl = hdev_ioctl, |
63ec93db | 2649 | .bdrv_aio_ioctl = hdev_aio_ioctl, |
f3a5d3f8 CH |
2650 | }; |
2651 | #endif /* __linux__ */ | |
2652 | ||
a167ba50 | 2653 | #if defined (__FreeBSD__) || defined(__FreeBSD_kernel__) |
511018e4 AT |
2654 | static int cdrom_open(BlockDriverState *bs, QDict *options, int flags, |
2655 | Error **errp) | |
f3a5d3f8 CH |
2656 | { |
2657 | BDRVRawState *s = bs->opaque; | |
e428e439 | 2658 | Error *local_err = NULL; |
f3a5d3f8 CH |
2659 | int ret; |
2660 | ||
2661 | s->type = FTYPE_CD; | |
2662 | ||
e428e439 HR |
2663 | ret = raw_open_common(bs, options, flags, 0, &local_err); |
2664 | if (ret) { | |
84d18f06 | 2665 | if (local_err) { |
e428e439 HR |
2666 | error_propagate(errp, local_err); |
2667 | } | |
f3a5d3f8 | 2668 | return ret; |
e428e439 | 2669 | } |
f3a5d3f8 | 2670 | |
9b2260cb | 2671 | /* make sure the door isn't locked at this time */ |
f3a5d3f8 CH |
2672 | ioctl(s->fd, CDIOCALLOW); |
2673 | return 0; | |
2674 | } | |
2675 | ||
508c7cb3 CH |
2676 | static int cdrom_probe_device(const char *filename) |
2677 | { | |
2678 | if (strstart(filename, "/dev/cd", NULL) || | |
2679 | strstart(filename, "/dev/acd", NULL)) | |
2680 | return 100; | |
2681 | return 0; | |
2682 | } | |
2683 | ||
f3a5d3f8 CH |
2684 | static int cdrom_reopen(BlockDriverState *bs) |
2685 | { | |
2686 | BDRVRawState *s = bs->opaque; | |
2687 | int fd; | |
2688 | ||
2689 | /* | |
2690 | * Force reread of possibly changed/newly loaded disc, | |
2691 | * FreeBSD seems to not notice sometimes... | |
2692 | */ | |
2693 | if (s->fd >= 0) | |
2e1e79da | 2694 | qemu_close(s->fd); |
6165f4d8 | 2695 | fd = qemu_open(bs->filename, s->open_flags, 0644); |
f3a5d3f8 CH |
2696 | if (fd < 0) { |
2697 | s->fd = -1; | |
2698 | return -EIO; | |
2699 | } | |
2700 | s->fd = fd; | |
2701 | ||
9b2260cb | 2702 | /* make sure the door isn't locked at this time */ |
f3a5d3f8 CH |
2703 | ioctl(s->fd, CDIOCALLOW); |
2704 | return 0; | |
2705 | } | |
2706 | ||
2707 | static int cdrom_is_inserted(BlockDriverState *bs) | |
2708 | { | |
2709 | return raw_getlength(bs) > 0; | |
2710 | } | |
2711 | ||
f36f3949 | 2712 | static void cdrom_eject(BlockDriverState *bs, bool eject_flag) |
f3a5d3f8 CH |
2713 | { |
2714 | BDRVRawState *s = bs->opaque; | |
2715 | ||
2716 | if (s->fd < 0) | |
822e1cd1 | 2717 | return; |
f3a5d3f8 CH |
2718 | |
2719 | (void) ioctl(s->fd, CDIOCALLOW); | |
2720 | ||
2721 | if (eject_flag) { | |
2722 | if (ioctl(s->fd, CDIOCEJECT) < 0) | |
2723 | perror("CDIOCEJECT"); | |
2724 | } else { | |
2725 | if (ioctl(s->fd, CDIOCCLOSE) < 0) | |
2726 | perror("CDIOCCLOSE"); | |
2727 | } | |
2728 | ||
822e1cd1 | 2729 | cdrom_reopen(bs); |
f3a5d3f8 CH |
2730 | } |
2731 | ||
025e849a | 2732 | static void cdrom_lock_medium(BlockDriverState *bs, bool locked) |
f3a5d3f8 CH |
2733 | { |
2734 | BDRVRawState *s = bs->opaque; | |
2735 | ||
2736 | if (s->fd < 0) | |
7bf37fed | 2737 | return; |
f3a5d3f8 CH |
2738 | if (ioctl(s->fd, (locked ? CDIOCPREVENT : CDIOCALLOW)) < 0) { |
2739 | /* | |
2740 | * Note: an error can happen if the distribution automatically | |
2741 | * mounts the CD-ROM | |
2742 | */ | |
2743 | /* perror("CDROM_LOCKDOOR"); */ | |
2744 | } | |
f3a5d3f8 CH |
2745 | } |
2746 | ||
2747 | static BlockDriver bdrv_host_cdrom = { | |
2748 | .format_name = "host_cdrom", | |
84a12e66 | 2749 | .protocol_name = "host_cdrom", |
f3a5d3f8 | 2750 | .instance_size = sizeof(BDRVRawState), |
030be321 | 2751 | .bdrv_needs_filename = true, |
508c7cb3 | 2752 | .bdrv_probe_device = cdrom_probe_device, |
18fa1c42 | 2753 | .bdrv_parse_filename = cdrom_parse_filename, |
66f82cee | 2754 | .bdrv_file_open = cdrom_open, |
f3a5d3f8 | 2755 | .bdrv_close = raw_close, |
1bc6b705 JC |
2756 | .bdrv_reopen_prepare = raw_reopen_prepare, |
2757 | .bdrv_reopen_commit = raw_reopen_commit, | |
2758 | .bdrv_reopen_abort = raw_reopen_abort, | |
c282e1fd | 2759 | .bdrv_create = hdev_create, |
6f482f74 | 2760 | .create_opts = &raw_create_opts, |
f3a5d3f8 | 2761 | |
f3a5d3f8 CH |
2762 | .bdrv_aio_readv = raw_aio_readv, |
2763 | .bdrv_aio_writev = raw_aio_writev, | |
b2e12bc6 | 2764 | .bdrv_aio_flush = raw_aio_flush, |
c25f53b0 | 2765 | .bdrv_refresh_limits = raw_refresh_limits, |
1b3abdcc ML |
2766 | .bdrv_io_plug = raw_aio_plug, |
2767 | .bdrv_io_unplug = raw_aio_unplug, | |
2768 | .bdrv_flush_io_queue = raw_aio_flush_io_queue, | |
f3a5d3f8 | 2769 | |
55b949c8 | 2770 | .bdrv_truncate = raw_truncate, |
b94a2610 KW |
2771 | .bdrv_getlength = raw_getlength, |
2772 | .has_variable_length = true, | |
4a1d5e1f FZ |
2773 | .bdrv_get_allocated_file_size |
2774 | = raw_get_allocated_file_size, | |
f3a5d3f8 | 2775 | |
c2f3426c SH |
2776 | .bdrv_detach_aio_context = raw_detach_aio_context, |
2777 | .bdrv_attach_aio_context = raw_attach_aio_context, | |
2778 | ||
19cb3738 | 2779 | /* removable device support */ |
f3a5d3f8 CH |
2780 | .bdrv_is_inserted = cdrom_is_inserted, |
2781 | .bdrv_eject = cdrom_eject, | |
025e849a | 2782 | .bdrv_lock_medium = cdrom_lock_medium, |
19cb3738 | 2783 | }; |
f3a5d3f8 | 2784 | #endif /* __FreeBSD__ */ |
5efa9d5a | 2785 | |
84a12e66 | 2786 | static void bdrv_file_init(void) |
5efa9d5a | 2787 | { |
508c7cb3 CH |
2788 | /* |
2789 | * Register all the drivers. Note that order is important, the driver | |
2790 | * registered last will get probed first. | |
2791 | */ | |
84a12e66 | 2792 | bdrv_register(&bdrv_file); |
5efa9d5a | 2793 | bdrv_register(&bdrv_host_device); |
f3a5d3f8 CH |
2794 | #ifdef __linux__ |
2795 | bdrv_register(&bdrv_host_floppy); | |
2796 | bdrv_register(&bdrv_host_cdrom); | |
2797 | #endif | |
a167ba50 | 2798 | #if defined(__FreeBSD__) || defined(__FreeBSD_kernel__) |
f3a5d3f8 CH |
2799 | bdrv_register(&bdrv_host_cdrom); |
2800 | #endif | |
5efa9d5a AL |
2801 | } |
2802 | ||
84a12e66 | 2803 | block_init(bdrv_file_init); |