]>
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 | */ | |
922a01a0 | 24 | |
80c71a24 | 25 | #include "qemu/osdep.h" |
da34e65c | 26 | #include "qapi/error.h" |
f348b6d1 | 27 | #include "qemu/cutils.h" |
d49b6836 | 28 | #include "qemu/error-report.h" |
737e150e | 29 | #include "block/block_int.h" |
1de7afc9 | 30 | #include "qemu/module.h" |
922a01a0 | 31 | #include "qemu/option.h" |
de81a169 | 32 | #include "trace.h" |
737e150e | 33 | #include "block/thread-pool.h" |
1de7afc9 | 34 | #include "qemu/iov.h" |
0187f5c9 | 35 | #include "block/raw-aio.h" |
452fcdbc | 36 | #include "qapi/qmp/qdict.h" |
d49b6836 | 37 | #include "qapi/qmp/qstring.h" |
83f64091 | 38 | |
7c9e5276 PB |
39 | #include "scsi/pr-manager.h" |
40 | #include "scsi/constants.h" | |
41 | ||
83affaa6 | 42 | #if defined(__APPLE__) && (__MACH__) |
83f64091 FB |
43 | #include <paths.h> |
44 | #include <sys/param.h> | |
45 | #include <IOKit/IOKitLib.h> | |
46 | #include <IOKit/IOBSD.h> | |
47 | #include <IOKit/storage/IOMediaBSDClient.h> | |
48 | #include <IOKit/storage/IOMedia.h> | |
49 | #include <IOKit/storage/IOCDMedia.h> | |
50 | //#include <IOKit/storage/IOCDTypes.h> | |
d0855f12 | 51 | #include <IOKit/storage/IODVDMedia.h> |
83f64091 FB |
52 | #include <CoreFoundation/CoreFoundation.h> |
53 | #endif | |
54 | ||
55 | #ifdef __sun__ | |
2e9671da | 56 | #define _POSIX_PTHREAD_SEMANTICS 1 |
83f64091 FB |
57 | #include <sys/dkio.h> |
58 | #endif | |
19cb3738 FB |
59 | #ifdef __linux__ |
60 | #include <sys/ioctl.h> | |
05acda4d | 61 | #include <sys/param.h> |
1efad060 | 62 | #include <sys/syscall.h> |
19cb3738 FB |
63 | #include <linux/cdrom.h> |
64 | #include <linux/fd.h> | |
5500316d | 65 | #include <linux/fs.h> |
1a9335e4 | 66 | #include <linux/hdreg.h> |
3307ed7b | 67 | #include <scsi/sg.h> |
1a9335e4 ET |
68 | #ifdef __s390__ |
69 | #include <asm/dasd.h> | |
70 | #endif | |
4ab15590 CL |
71 | #ifndef FS_NOCOW_FL |
72 | #define FS_NOCOW_FL 0x00800000 /* Do not cow file */ | |
73 | #endif | |
5500316d | 74 | #endif |
b953f075 | 75 | #if defined(CONFIG_FALLOCATE_PUNCH_HOLE) || defined(CONFIG_FALLOCATE_ZERO_RANGE) |
3d4fa43e KK |
76 | #include <linux/falloc.h> |
77 | #endif | |
a167ba50 | 78 | #if defined (__FreeBSD__) || defined(__FreeBSD_kernel__) |
1cb6c3fd | 79 | #include <sys/disk.h> |
9f23011a | 80 | #include <sys/cdio.h> |
1cb6c3fd | 81 | #endif |
83f64091 | 82 | |
128ab2ff BS |
83 | #ifdef __OpenBSD__ |
84 | #include <sys/ioctl.h> | |
85 | #include <sys/disklabel.h> | |
86 | #include <sys/dkio.h> | |
87 | #endif | |
88 | ||
d1f6fd8d CE |
89 | #ifdef __NetBSD__ |
90 | #include <sys/ioctl.h> | |
91 | #include <sys/disklabel.h> | |
92 | #include <sys/dkio.h> | |
93 | #include <sys/disk.h> | |
94 | #endif | |
95 | ||
c5e97233 BS |
96 | #ifdef __DragonFly__ |
97 | #include <sys/ioctl.h> | |
98 | #include <sys/diskslice.h> | |
99 | #endif | |
100 | ||
dce512de CH |
101 | #ifdef CONFIG_XFS |
102 | #include <xfs/xfs.h> | |
103 | #endif | |
104 | ||
4f7d28d7 | 105 | #include "trace.h" |
8c05dbf9 | 106 | |
f6465578 AL |
107 | /* OS X does not have O_DSYNC */ |
108 | #ifndef O_DSYNC | |
1c27a8b3 | 109 | #ifdef O_SYNC |
7ab064d2 | 110 | #define O_DSYNC O_SYNC |
1c27a8b3 JA |
111 | #elif defined(O_FSYNC) |
112 | #define O_DSYNC O_FSYNC | |
113 | #endif | |
f6465578 AL |
114 | #endif |
115 | ||
9f7965c7 AL |
116 | /* Approximate O_DIRECT with O_DSYNC if O_DIRECT isn't available */ |
117 | #ifndef O_DIRECT | |
118 | #define O_DIRECT O_DSYNC | |
119 | #endif | |
120 | ||
19cb3738 FB |
121 | #define FTYPE_FILE 0 |
122 | #define FTYPE_CD 1 | |
83f64091 | 123 | |
581b9e29 CH |
124 | #define MAX_BLOCKSIZE 4096 |
125 | ||
244a5668 FZ |
126 | /* Posix file locking bytes. Libvirt takes byte 0, we start from higher bytes, |
127 | * leaving a few more bytes for its future use. */ | |
128 | #define RAW_LOCK_PERM_BASE 100 | |
129 | #define RAW_LOCK_SHARED_BASE 200 | |
130 | ||
19cb3738 FB |
131 | typedef struct BDRVRawState { |
132 | int fd; | |
244a5668 | 133 | bool use_lock; |
19cb3738 | 134 | int type; |
0e1d8f4c | 135 | int open_flags; |
c25f53b0 PB |
136 | size_t buf_align; |
137 | ||
244a5668 FZ |
138 | /* The current permissions. */ |
139 | uint64_t perm; | |
140 | uint64_t shared_perm; | |
141 | ||
2996ffad | 142 | /* The perms bits whose corresponding bytes are already locked in |
f2e3af29 | 143 | * s->fd. */ |
2996ffad FZ |
144 | uint64_t locked_perm; |
145 | uint64_t locked_shared_perm; | |
146 | ||
dce512de | 147 | #ifdef CONFIG_XFS |
260a82e5 | 148 | bool is_xfs:1; |
dce512de | 149 | #endif |
260a82e5 | 150 | bool has_discard:1; |
97a2ae34 | 151 | bool has_write_zeroes:1; |
260a82e5 | 152 | bool discard_zeroes:1; |
0a4279d9 | 153 | bool use_linux_aio:1; |
e5bcf967 | 154 | bool page_cache_inconsistent:1; |
d50d8222 | 155 | bool has_fallocate; |
3cad8307 | 156 | bool needs_alignment; |
31be8a2a | 157 | bool check_cache_dropped; |
7c9e5276 PB |
158 | |
159 | PRManager *pr_mgr; | |
19cb3738 FB |
160 | } BDRVRawState; |
161 | ||
eeb6b45d JC |
162 | typedef struct BDRVRawReopenState { |
163 | int fd; | |
164 | int open_flags; | |
31be8a2a | 165 | bool check_cache_dropped; |
eeb6b45d JC |
166 | } BDRVRawReopenState; |
167 | ||
19cb3738 | 168 | static int fd_open(BlockDriverState *bs); |
22afa7b5 | 169 | static int64_t raw_getlength(BlockDriverState *bs); |
83f64091 | 170 | |
de81a169 PB |
171 | typedef struct RawPosixAIOData { |
172 | BlockDriverState *bs; | |
d57c44d0 | 173 | int aio_type; |
de81a169 | 174 | int aio_fildes; |
d57c44d0 | 175 | |
de81a169 | 176 | off_t aio_offset; |
d57c44d0 KW |
177 | uint64_t aio_nbytes; |
178 | ||
93f4e2ff | 179 | union { |
d57c44d0 KW |
180 | struct { |
181 | struct iovec *iov; | |
182 | int niov; | |
183 | } io; | |
184 | struct { | |
185 | uint64_t cmd; | |
186 | void *buf; | |
187 | } ioctl; | |
93f4e2ff KW |
188 | struct { |
189 | int aio_fd2; | |
190 | off_t aio_offset2; | |
d57c44d0 | 191 | } copy_range; |
93f4e2ff KW |
192 | struct { |
193 | PreallocMode prealloc; | |
194 | Error **errp; | |
d57c44d0 | 195 | } truncate; |
93f4e2ff | 196 | }; |
de81a169 PB |
197 | } RawPosixAIOData; |
198 | ||
a167ba50 | 199 | #if defined(__FreeBSD__) || defined(__FreeBSD_kernel__) |
f3a5d3f8 | 200 | static int cdrom_reopen(BlockDriverState *bs); |
9f23011a BS |
201 | #endif |
202 | ||
1de1ae0a | 203 | #if defined(__NetBSD__) |
db0754df | 204 | static int raw_normalize_devicepath(const char **filename, Error **errp) |
1de1ae0a CE |
205 | { |
206 | static char namebuf[PATH_MAX]; | |
207 | const char *dp, *fname; | |
208 | struct stat sb; | |
209 | ||
210 | fname = *filename; | |
211 | dp = strrchr(fname, '/'); | |
212 | if (lstat(fname, &sb) < 0) { | |
db0754df | 213 | error_setg_errno(errp, errno, "%s: stat failed", fname); |
1de1ae0a CE |
214 | return -errno; |
215 | } | |
216 | ||
217 | if (!S_ISBLK(sb.st_mode)) { | |
218 | return 0; | |
219 | } | |
220 | ||
221 | if (dp == NULL) { | |
222 | snprintf(namebuf, PATH_MAX, "r%s", fname); | |
223 | } else { | |
224 | snprintf(namebuf, PATH_MAX, "%.*s/r%s", | |
225 | (int)(dp - fname), fname, dp + 1); | |
226 | } | |
1de1ae0a | 227 | *filename = namebuf; |
db0754df | 228 | warn_report("%s is a block device, using %s", fname, *filename); |
1de1ae0a CE |
229 | |
230 | return 0; | |
231 | } | |
232 | #else | |
db0754df | 233 | static int raw_normalize_devicepath(const char **filename, Error **errp) |
1de1ae0a CE |
234 | { |
235 | return 0; | |
236 | } | |
237 | #endif | |
238 | ||
8a4ed0d1 ET |
239 | /* |
240 | * Get logical block size via ioctl. On success store it in @sector_size_p. | |
241 | */ | |
242 | static int probe_logical_blocksize(int fd, unsigned int *sector_size_p) | |
c25f53b0 | 243 | { |
c25f53b0 | 244 | unsigned int sector_size; |
8a4ed0d1 | 245 | bool success = false; |
700f9ce0 | 246 | int i; |
c25f53b0 | 247 | |
8a4ed0d1 | 248 | errno = ENOTSUP; |
700f9ce0 | 249 | static const unsigned long ioctl_list[] = { |
c25f53b0 | 250 | #ifdef BLKSSZGET |
700f9ce0 | 251 | BLKSSZGET, |
c25f53b0 PB |
252 | #endif |
253 | #ifdef DKIOCGETBLOCKSIZE | |
700f9ce0 | 254 | DKIOCGETBLOCKSIZE, |
c25f53b0 PB |
255 | #endif |
256 | #ifdef DIOCGSECTORSIZE | |
700f9ce0 | 257 | DIOCGSECTORSIZE, |
c25f53b0 | 258 | #endif |
700f9ce0 PM |
259 | }; |
260 | ||
261 | /* Try a few ioctls to get the right size */ | |
262 | for (i = 0; i < (int)ARRAY_SIZE(ioctl_list); i++) { | |
263 | if (ioctl(fd, ioctl_list[i], §or_size) >= 0) { | |
264 | *sector_size_p = sector_size; | |
265 | success = true; | |
266 | } | |
267 | } | |
8a4ed0d1 ET |
268 | |
269 | return success ? 0 : -errno; | |
270 | } | |
271 | ||
1a9335e4 ET |
272 | /** |
273 | * Get physical block size of @fd. | |
274 | * On success, store it in @blk_size and return 0. | |
275 | * On failure, return -errno. | |
276 | */ | |
277 | static int probe_physical_blocksize(int fd, unsigned int *blk_size) | |
278 | { | |
279 | #ifdef BLKPBSZGET | |
280 | if (ioctl(fd, BLKPBSZGET, blk_size) < 0) { | |
281 | return -errno; | |
282 | } | |
283 | return 0; | |
284 | #else | |
285 | return -ENOTSUP; | |
286 | #endif | |
287 | } | |
288 | ||
22d182e8 SH |
289 | /* Check if read is allowed with given memory buffer and length. |
290 | * | |
291 | * This function is used to check O_DIRECT memory buffer and request alignment. | |
292 | */ | |
293 | static bool raw_is_io_aligned(int fd, void *buf, size_t len) | |
294 | { | |
295 | ssize_t ret = pread(fd, buf, len, 0); | |
296 | ||
297 | if (ret >= 0) { | |
298 | return true; | |
299 | } | |
300 | ||
301 | #ifdef __linux__ | |
302 | /* The Linux kernel returns EINVAL for misaligned O_DIRECT reads. Ignore | |
303 | * other errors (e.g. real I/O error), which could happen on a failed | |
304 | * drive, since we only care about probing alignment. | |
305 | */ | |
306 | if (errno != EINVAL) { | |
307 | return true; | |
308 | } | |
309 | #endif | |
310 | ||
311 | return false; | |
312 | } | |
313 | ||
8a4ed0d1 ET |
314 | static void raw_probe_alignment(BlockDriverState *bs, int fd, Error **errp) |
315 | { | |
316 | BDRVRawState *s = bs->opaque; | |
317 | char *buf; | |
459b4e66 | 318 | size_t max_align = MAX(MAX_BLOCKSIZE, getpagesize()); |
8a4ed0d1 | 319 | |
b192af8a | 320 | /* For SCSI generic devices the alignment is not really used. |
8a4ed0d1 | 321 | With buffered I/O, we don't have any restrictions. */ |
b192af8a | 322 | if (bdrv_is_sg(bs) || !s->needs_alignment) { |
a5b8dd2c | 323 | bs->bl.request_alignment = 1; |
8a4ed0d1 ET |
324 | s->buf_align = 1; |
325 | return; | |
326 | } | |
327 | ||
a5b8dd2c | 328 | bs->bl.request_alignment = 0; |
8a4ed0d1 ET |
329 | s->buf_align = 0; |
330 | /* Let's try to use the logical blocksize for the alignment. */ | |
a5b8dd2c EB |
331 | if (probe_logical_blocksize(fd, &bs->bl.request_alignment) < 0) { |
332 | bs->bl.request_alignment = 0; | |
8a4ed0d1 | 333 | } |
c25f53b0 PB |
334 | #ifdef CONFIG_XFS |
335 | if (s->is_xfs) { | |
336 | struct dioattr da; | |
df26a350 | 337 | if (xfsctl(NULL, fd, XFS_IOC_DIOINFO, &da) >= 0) { |
a5b8dd2c | 338 | bs->bl.request_alignment = da.d_miniosz; |
c25f53b0 PB |
339 | /* The kernel returns wrong information for d_mem */ |
340 | /* s->buf_align = da.d_mem; */ | |
341 | } | |
342 | } | |
343 | #endif | |
344 | ||
345 | /* If we could not get the sizes so far, we can only guess them */ | |
346 | if (!s->buf_align) { | |
347 | size_t align; | |
459b4e66 DL |
348 | buf = qemu_memalign(max_align, 2 * max_align); |
349 | for (align = 512; align <= max_align; align <<= 1) { | |
350 | if (raw_is_io_aligned(fd, buf + align, max_align)) { | |
c25f53b0 PB |
351 | s->buf_align = align; |
352 | break; | |
353 | } | |
354 | } | |
355 | qemu_vfree(buf); | |
356 | } | |
357 | ||
a5b8dd2c | 358 | if (!bs->bl.request_alignment) { |
c25f53b0 | 359 | size_t align; |
459b4e66 DL |
360 | buf = qemu_memalign(s->buf_align, max_align); |
361 | for (align = 512; align <= max_align; align <<= 1) { | |
22d182e8 | 362 | if (raw_is_io_aligned(fd, buf, align)) { |
a5b8dd2c | 363 | bs->bl.request_alignment = align; |
c25f53b0 PB |
364 | break; |
365 | } | |
366 | } | |
367 | qemu_vfree(buf); | |
368 | } | |
df26a350 | 369 | |
a5b8dd2c | 370 | if (!s->buf_align || !bs->bl.request_alignment) { |
8cc9c6e9 EB |
371 | error_setg(errp, "Could not find working O_DIRECT alignment"); |
372 | error_append_hint(errp, "Try cache.direct=off\n"); | |
df26a350 | 373 | } |
c25f53b0 PB |
374 | } |
375 | ||
6a8dc042 JC |
376 | static void raw_parse_flags(int bdrv_flags, int *open_flags) |
377 | { | |
378 | assert(open_flags != NULL); | |
379 | ||
380 | *open_flags |= O_BINARY; | |
381 | *open_flags &= ~O_ACCMODE; | |
382 | if (bdrv_flags & BDRV_O_RDWR) { | |
383 | *open_flags |= O_RDWR; | |
384 | } else { | |
385 | *open_flags |= O_RDONLY; | |
386 | } | |
387 | ||
388 | /* Use O_DSYNC for write-through caching, no flags for write-back caching, | |
389 | * and O_DIRECT for no caching. */ | |
390 | if ((bdrv_flags & BDRV_O_NOCACHE)) { | |
391 | *open_flags |= O_DIRECT; | |
392 | } | |
6a8dc042 JC |
393 | } |
394 | ||
078896a9 HR |
395 | static void raw_parse_filename(const char *filename, QDict *options, |
396 | Error **errp) | |
397 | { | |
03c320d8 | 398 | bdrv_parse_filename_strip_prefix(filename, "file:", options); |
078896a9 HR |
399 | } |
400 | ||
c66a6157 KW |
401 | static QemuOptsList raw_runtime_opts = { |
402 | .name = "raw", | |
403 | .head = QTAILQ_HEAD_INITIALIZER(raw_runtime_opts.head), | |
404 | .desc = { | |
405 | { | |
406 | .name = "filename", | |
407 | .type = QEMU_OPT_STRING, | |
408 | .help = "File name of the image", | |
409 | }, | |
0a4279d9 KW |
410 | { |
411 | .name = "aio", | |
412 | .type = QEMU_OPT_STRING, | |
413 | .help = "host AIO implementation (threads, native)", | |
414 | }, | |
16b48d5d FZ |
415 | { |
416 | .name = "locking", | |
417 | .type = QEMU_OPT_STRING, | |
418 | .help = "file locking mode (on/off/auto, default: auto)", | |
419 | }, | |
7c9e5276 PB |
420 | { |
421 | .name = "pr-manager", | |
422 | .type = QEMU_OPT_STRING, | |
423 | .help = "id of persistent reservation manager object (default: none)", | |
424 | }, | |
31be8a2a SH |
425 | { |
426 | .name = "x-check-cache-dropped", | |
427 | .type = QEMU_OPT_BOOL, | |
428 | .help = "check that page cache was dropped on live migration (default: off)" | |
429 | }, | |
c66a6157 KW |
430 | { /* end of list */ } |
431 | }, | |
432 | }; | |
433 | ||
434 | static int raw_open_common(BlockDriverState *bs, QDict *options, | |
230ff739 JS |
435 | int bdrv_flags, int open_flags, |
436 | bool device, Error **errp) | |
83f64091 FB |
437 | { |
438 | BDRVRawState *s = bs->opaque; | |
c66a6157 KW |
439 | QemuOpts *opts; |
440 | Error *local_err = NULL; | |
8bfea15d | 441 | const char *filename = NULL; |
7c9e5276 | 442 | const char *str; |
0a4279d9 | 443 | BlockdevAioOptions aio, aio_default; |
0e1d8f4c | 444 | int fd, ret; |
260a82e5 | 445 | struct stat st; |
244a5668 | 446 | OnOffAuto locking; |
83f64091 | 447 | |
87ea75d5 | 448 | opts = qemu_opts_create(&raw_runtime_opts, NULL, 0, &error_abort); |
c66a6157 | 449 | qemu_opts_absorb_qdict(opts, options, &local_err); |
84d18f06 | 450 | if (local_err) { |
e428e439 | 451 | error_propagate(errp, local_err); |
c66a6157 KW |
452 | ret = -EINVAL; |
453 | goto fail; | |
454 | } | |
455 | ||
456 | filename = qemu_opt_get(opts, "filename"); | |
457 | ||
db0754df | 458 | ret = raw_normalize_devicepath(&filename, errp); |
1de1ae0a | 459 | if (ret != 0) { |
c66a6157 | 460 | goto fail; |
1de1ae0a CE |
461 | } |
462 | ||
0a4279d9 KW |
463 | aio_default = (bdrv_flags & BDRV_O_NATIVE_AIO) |
464 | ? BLOCKDEV_AIO_OPTIONS_NATIVE | |
465 | : BLOCKDEV_AIO_OPTIONS_THREADS; | |
f7abe0ec MAL |
466 | aio = qapi_enum_parse(&BlockdevAioOptions_lookup, |
467 | qemu_opt_get(opts, "aio"), | |
06c60b6c | 468 | aio_default, &local_err); |
0a4279d9 KW |
469 | if (local_err) { |
470 | error_propagate(errp, local_err); | |
471 | ret = -EINVAL; | |
472 | goto fail; | |
473 | } | |
474 | s->use_linux_aio = (aio == BLOCKDEV_AIO_OPTIONS_NATIVE); | |
475 | ||
f7abe0ec MAL |
476 | locking = qapi_enum_parse(&OnOffAuto_lookup, |
477 | qemu_opt_get(opts, "locking"), | |
06c60b6c | 478 | ON_OFF_AUTO_AUTO, &local_err); |
244a5668 FZ |
479 | if (local_err) { |
480 | error_propagate(errp, local_err); | |
481 | ret = -EINVAL; | |
482 | goto fail; | |
483 | } | |
484 | switch (locking) { | |
485 | case ON_OFF_AUTO_ON: | |
486 | s->use_lock = true; | |
2b218f5d | 487 | if (!qemu_has_ofd_lock()) { |
db0754df FZ |
488 | warn_report("File lock requested but OFD locking syscall is " |
489 | "unavailable, falling back to POSIX file locks"); | |
490 | error_printf("Due to the implementation, locks can be lost " | |
491 | "unexpectedly.\n"); | |
2b218f5d | 492 | } |
244a5668 FZ |
493 | break; |
494 | case ON_OFF_AUTO_OFF: | |
495 | s->use_lock = false; | |
496 | break; | |
497 | case ON_OFF_AUTO_AUTO: | |
2b218f5d | 498 | s->use_lock = qemu_has_ofd_lock(); |
244a5668 FZ |
499 | break; |
500 | default: | |
501 | abort(); | |
502 | } | |
503 | ||
7c9e5276 PB |
504 | str = qemu_opt_get(opts, "pr-manager"); |
505 | if (str) { | |
506 | s->pr_mgr = pr_manager_lookup(str, &local_err); | |
507 | if (local_err) { | |
508 | error_propagate(errp, local_err); | |
509 | ret = -EINVAL; | |
510 | goto fail; | |
511 | } | |
512 | } | |
513 | ||
31be8a2a SH |
514 | s->check_cache_dropped = qemu_opt_get_bool(opts, "x-check-cache-dropped", |
515 | false); | |
516 | ||
6a8dc042 JC |
517 | s->open_flags = open_flags; |
518 | raw_parse_flags(bdrv_flags, &s->open_flags); | |
83f64091 | 519 | |
90babde0 | 520 | s->fd = -1; |
40ff6d7e | 521 | fd = qemu_open(filename, s->open_flags, 0644); |
64107dc0 KW |
522 | ret = fd < 0 ? -errno : 0; |
523 | ||
524 | if (ret == -EACCES || ret == -EROFS) { | |
525 | /* Try to degrade to read-only, but if it doesn't work, still use the | |
526 | * normal error message. */ | |
527 | if (bdrv_apply_auto_read_only(bs, NULL, NULL) == 0) { | |
528 | bdrv_flags &= ~BDRV_O_RDWR; | |
529 | raw_parse_flags(bdrv_flags, &s->open_flags); | |
530 | assert(!(s->open_flags & O_CREAT)); | |
531 | fd = qemu_open(filename, s->open_flags); | |
532 | ret = fd < 0 ? -errno : 0; | |
533 | } | |
534 | } | |
535 | ||
536 | if (ret < 0) { | |
537 | error_setg_errno(errp, -ret, "Could not open '%s'", filename); | |
c66a6157 | 538 | if (ret == -EROFS) { |
19cb3738 | 539 | ret = -EACCES; |
c66a6157 KW |
540 | } |
541 | goto fail; | |
19cb3738 | 542 | } |
83f64091 | 543 | s->fd = fd; |
9ef91a67 | 544 | |
244a5668 FZ |
545 | s->perm = 0; |
546 | s->shared_perm = BLK_PERM_ALL; | |
547 | ||
5c6c3a6c | 548 | #ifdef CONFIG_LINUX_AIO |
0a4279d9 | 549 | /* Currently Linux does AIO only for files opened with O_DIRECT */ |
ed6e2161 NA |
550 | if (s->use_linux_aio) { |
551 | if (!(s->open_flags & O_DIRECT)) { | |
552 | error_setg(errp, "aio=native was specified, but it requires " | |
553 | "cache.direct=on, which was not specified."); | |
554 | ret = -EINVAL; | |
555 | goto fail; | |
556 | } | |
557 | if (!aio_setup_linux_aio(bdrv_get_aio_context(bs), errp)) { | |
558 | error_prepend(errp, "Unable to use native AIO: "); | |
559 | goto fail; | |
560 | } | |
96518254 | 561 | } |
1501ecc1 | 562 | #else |
0a4279d9 | 563 | if (s->use_linux_aio) { |
d657c0c2 KW |
564 | error_setg(errp, "aio=native was specified, but is not supported " |
565 | "in this build."); | |
566 | ret = -EINVAL; | |
567 | goto fail; | |
1501ecc1 SH |
568 | } |
569 | #endif /* !defined(CONFIG_LINUX_AIO) */ | |
9ef91a67 | 570 | |
7ce21016 | 571 | s->has_discard = true; |
97a2ae34 | 572 | s->has_write_zeroes = true; |
3cad8307 RPM |
573 | if ((bs->open_flags & BDRV_O_NOCACHE) != 0) { |
574 | s->needs_alignment = true; | |
575 | } | |
260a82e5 PB |
576 | |
577 | if (fstat(s->fd, &st) < 0) { | |
01212d4e | 578 | ret = -errno; |
260a82e5 PB |
579 | error_setg_errno(errp, errno, "Could not stat file"); |
580 | goto fail; | |
581 | } | |
230ff739 JS |
582 | |
583 | if (!device) { | |
584 | if (S_ISBLK(st.st_mode)) { | |
585 | warn_report("Opening a block device as a file using the '%s' " | |
586 | "driver is deprecated", bs->drv->format_name); | |
587 | } else if (S_ISCHR(st.st_mode)) { | |
588 | warn_report("Opening a character device as a file using the '%s' " | |
589 | "driver is deprecated", bs->drv->format_name); | |
590 | } else if (!S_ISREG(st.st_mode)) { | |
591 | error_setg(errp, "A regular file was expected by the '%s' driver, " | |
592 | "but something else was given", bs->drv->format_name); | |
593 | ret = -EINVAL; | |
594 | goto fail; | |
595 | } else { | |
596 | s->discard_zeroes = true; | |
597 | s->has_fallocate = true; | |
598 | } | |
599 | } else { | |
600 | if (!(S_ISCHR(st.st_mode) || S_ISBLK(st.st_mode))) { | |
601 | error_setg(errp, "'%s' driver expects either " | |
602 | "a character or block device", bs->drv->format_name); | |
603 | ret = -EINVAL; | |
604 | goto fail; | |
605 | } | |
260a82e5 | 606 | } |
230ff739 | 607 | |
d0b4503e PB |
608 | if (S_ISBLK(st.st_mode)) { |
609 | #ifdef BLKDISCARDZEROES | |
610 | unsigned int arg; | |
611 | if (ioctl(s->fd, BLKDISCARDZEROES, &arg) == 0 && arg) { | |
612 | s->discard_zeroes = true; | |
613 | } | |
614 | #endif | |
615 | #ifdef __linux__ | |
616 | /* On Linux 3.10, BLKDISCARD leaves stale data in the page cache. Do | |
617 | * not rely on the contents of discarded blocks unless using O_DIRECT. | |
97a2ae34 | 618 | * Same for BLKZEROOUT. |
d0b4503e PB |
619 | */ |
620 | if (!(bs->open_flags & BDRV_O_NOCACHE)) { | |
621 | s->discard_zeroes = false; | |
97a2ae34 | 622 | s->has_write_zeroes = false; |
d0b4503e PB |
623 | } |
624 | #endif | |
625 | } | |
3cad8307 RPM |
626 | #ifdef __FreeBSD__ |
627 | if (S_ISCHR(st.st_mode)) { | |
628 | /* | |
629 | * The file is a char device (disk), which on FreeBSD isn't behind | |
630 | * a pager, so force all requests to be aligned. This is needed | |
631 | * so QEMU makes sure all IO operations on the device are aligned | |
632 | * to sector size, or else FreeBSD will reject them with EINVAL. | |
633 | */ | |
634 | s->needs_alignment = true; | |
635 | } | |
636 | #endif | |
260a82e5 | 637 | |
dce512de CH |
638 | #ifdef CONFIG_XFS |
639 | if (platform_test_xfs_fd(s->fd)) { | |
7ce21016 | 640 | s->is_xfs = true; |
dce512de CH |
641 | } |
642 | #endif | |
643 | ||
34fa110e | 644 | bs->supported_zero_flags = BDRV_REQ_MAY_UNMAP; |
c66a6157 KW |
645 | ret = 0; |
646 | fail: | |
8bfea15d KW |
647 | if (filename && (bdrv_flags & BDRV_O_TEMPORARY)) { |
648 | unlink(filename); | |
649 | } | |
c66a6157 KW |
650 | qemu_opts_del(opts); |
651 | return ret; | |
83f64091 FB |
652 | } |
653 | ||
015a1036 HR |
654 | static int raw_open(BlockDriverState *bs, QDict *options, int flags, |
655 | Error **errp) | |
90babde0 CH |
656 | { |
657 | BDRVRawState *s = bs->opaque; | |
658 | ||
659 | s->type = FTYPE_FILE; | |
230ff739 | 660 | return raw_open_common(bs, options, flags, 0, false, errp); |
90babde0 CH |
661 | } |
662 | ||
244a5668 FZ |
663 | typedef enum { |
664 | RAW_PL_PREPARE, | |
665 | RAW_PL_COMMIT, | |
666 | RAW_PL_ABORT, | |
667 | } RawPermLockOp; | |
668 | ||
669 | #define PERM_FOREACH(i) \ | |
670 | for ((i) = 0; (1ULL << (i)) <= BLK_PERM_ALL; i++) | |
671 | ||
672 | /* Lock bytes indicated by @perm_lock_bits and @shared_perm_lock_bits in the | |
673 | * file; if @unlock == true, also unlock the unneeded bytes. | |
674 | * @shared_perm_lock_bits is the mask of all permissions that are NOT shared. | |
675 | */ | |
2996ffad | 676 | static int raw_apply_lock_bytes(BDRVRawState *s, int fd, |
244a5668 FZ |
677 | uint64_t perm_lock_bits, |
678 | uint64_t shared_perm_lock_bits, | |
679 | bool unlock, Error **errp) | |
680 | { | |
681 | int ret; | |
682 | int i; | |
2996ffad FZ |
683 | uint64_t locked_perm, locked_shared_perm; |
684 | ||
685 | if (s) { | |
686 | locked_perm = s->locked_perm; | |
687 | locked_shared_perm = s->locked_shared_perm; | |
688 | } else { | |
689 | /* | |
690 | * We don't have the previous bits, just lock/unlock for each of the | |
691 | * requested bits. | |
692 | */ | |
693 | if (unlock) { | |
694 | locked_perm = BLK_PERM_ALL; | |
695 | locked_shared_perm = BLK_PERM_ALL; | |
696 | } else { | |
697 | locked_perm = 0; | |
698 | locked_shared_perm = 0; | |
699 | } | |
700 | } | |
244a5668 FZ |
701 | |
702 | PERM_FOREACH(i) { | |
703 | int off = RAW_LOCK_PERM_BASE + i; | |
2996ffad FZ |
704 | uint64_t bit = (1ULL << i); |
705 | if ((perm_lock_bits & bit) && !(locked_perm & bit)) { | |
d0a96155 | 706 | ret = qemu_lock_fd(fd, off, 1, false); |
244a5668 FZ |
707 | if (ret) { |
708 | error_setg(errp, "Failed to lock byte %d", off); | |
709 | return ret; | |
2996ffad FZ |
710 | } else if (s) { |
711 | s->locked_perm |= bit; | |
244a5668 | 712 | } |
2996ffad | 713 | } else if (unlock && (locked_perm & bit) && !(perm_lock_bits & bit)) { |
d0a96155 | 714 | ret = qemu_unlock_fd(fd, off, 1); |
244a5668 FZ |
715 | if (ret) { |
716 | error_setg(errp, "Failed to unlock byte %d", off); | |
717 | return ret; | |
2996ffad FZ |
718 | } else if (s) { |
719 | s->locked_perm &= ~bit; | |
244a5668 FZ |
720 | } |
721 | } | |
722 | } | |
723 | PERM_FOREACH(i) { | |
724 | int off = RAW_LOCK_SHARED_BASE + i; | |
2996ffad FZ |
725 | uint64_t bit = (1ULL << i); |
726 | if ((shared_perm_lock_bits & bit) && !(locked_shared_perm & bit)) { | |
d0a96155 | 727 | ret = qemu_lock_fd(fd, off, 1, false); |
244a5668 FZ |
728 | if (ret) { |
729 | error_setg(errp, "Failed to lock byte %d", off); | |
730 | return ret; | |
2996ffad FZ |
731 | } else if (s) { |
732 | s->locked_shared_perm |= bit; | |
244a5668 | 733 | } |
2996ffad FZ |
734 | } else if (unlock && (locked_shared_perm & bit) && |
735 | !(shared_perm_lock_bits & bit)) { | |
d0a96155 | 736 | ret = qemu_unlock_fd(fd, off, 1); |
244a5668 FZ |
737 | if (ret) { |
738 | error_setg(errp, "Failed to unlock byte %d", off); | |
739 | return ret; | |
2996ffad FZ |
740 | } else if (s) { |
741 | s->locked_shared_perm &= ~bit; | |
244a5668 FZ |
742 | } |
743 | } | |
744 | } | |
745 | return 0; | |
746 | } | |
747 | ||
748 | /* Check "unshared" bytes implied by @perm and ~@shared_perm in the file. */ | |
d0a96155 | 749 | static int raw_check_lock_bytes(int fd, uint64_t perm, uint64_t shared_perm, |
244a5668 FZ |
750 | Error **errp) |
751 | { | |
752 | int ret; | |
753 | int i; | |
754 | ||
755 | PERM_FOREACH(i) { | |
756 | int off = RAW_LOCK_SHARED_BASE + i; | |
757 | uint64_t p = 1ULL << i; | |
758 | if (perm & p) { | |
d0a96155 | 759 | ret = qemu_lock_fd_test(fd, off, 1, true); |
244a5668 FZ |
760 | if (ret) { |
761 | char *perm_name = bdrv_perm_names(p); | |
762 | error_setg(errp, | |
763 | "Failed to get \"%s\" lock", | |
764 | perm_name); | |
765 | g_free(perm_name); | |
244a5668 FZ |
766 | return ret; |
767 | } | |
768 | } | |
769 | } | |
770 | PERM_FOREACH(i) { | |
771 | int off = RAW_LOCK_PERM_BASE + i; | |
772 | uint64_t p = 1ULL << i; | |
773 | if (!(shared_perm & p)) { | |
d0a96155 | 774 | ret = qemu_lock_fd_test(fd, off, 1, true); |
244a5668 FZ |
775 | if (ret) { |
776 | char *perm_name = bdrv_perm_names(p); | |
777 | error_setg(errp, | |
778 | "Failed to get shared \"%s\" lock", | |
779 | perm_name); | |
780 | g_free(perm_name); | |
244a5668 FZ |
781 | return ret; |
782 | } | |
783 | } | |
784 | } | |
785 | return 0; | |
786 | } | |
787 | ||
788 | static int raw_handle_perm_lock(BlockDriverState *bs, | |
789 | RawPermLockOp op, | |
790 | uint64_t new_perm, uint64_t new_shared, | |
791 | Error **errp) | |
792 | { | |
793 | BDRVRawState *s = bs->opaque; | |
794 | int ret = 0; | |
795 | Error *local_err = NULL; | |
796 | ||
797 | if (!s->use_lock) { | |
798 | return 0; | |
799 | } | |
800 | ||
801 | if (bdrv_get_flags(bs) & BDRV_O_INACTIVE) { | |
802 | return 0; | |
803 | } | |
804 | ||
244a5668 FZ |
805 | switch (op) { |
806 | case RAW_PL_PREPARE: | |
f2e3af29 | 807 | ret = raw_apply_lock_bytes(s, s->fd, s->perm | new_perm, |
244a5668 FZ |
808 | ~s->shared_perm | ~new_shared, |
809 | false, errp); | |
810 | if (!ret) { | |
f2e3af29 | 811 | ret = raw_check_lock_bytes(s->fd, new_perm, new_shared, errp); |
244a5668 FZ |
812 | if (!ret) { |
813 | return 0; | |
814 | } | |
b857431d FZ |
815 | error_append_hint(errp, |
816 | "Is another process using the image [%s]?\n", | |
817 | bs->filename); | |
244a5668 FZ |
818 | } |
819 | op = RAW_PL_ABORT; | |
820 | /* fall through to unlock bytes. */ | |
821 | case RAW_PL_ABORT: | |
f2e3af29 | 822 | raw_apply_lock_bytes(s, s->fd, s->perm, ~s->shared_perm, |
d0a96155 | 823 | true, &local_err); |
244a5668 FZ |
824 | if (local_err) { |
825 | /* Theoretically the above call only unlocks bytes and it cannot | |
826 | * fail. Something weird happened, report it. | |
827 | */ | |
db0754df | 828 | warn_report_err(local_err); |
244a5668 FZ |
829 | } |
830 | break; | |
831 | case RAW_PL_COMMIT: | |
f2e3af29 | 832 | raw_apply_lock_bytes(s, s->fd, new_perm, ~new_shared, |
d0a96155 | 833 | true, &local_err); |
244a5668 FZ |
834 | if (local_err) { |
835 | /* Theoretically the above call only unlocks bytes and it cannot | |
836 | * fail. Something weird happened, report it. | |
837 | */ | |
db0754df | 838 | warn_report_err(local_err); |
244a5668 FZ |
839 | } |
840 | break; | |
841 | } | |
842 | return ret; | |
843 | } | |
844 | ||
eeb6b45d JC |
845 | static int raw_reopen_prepare(BDRVReopenState *state, |
846 | BlockReopenQueue *queue, Error **errp) | |
847 | { | |
848 | BDRVRawState *s; | |
4e6d13c9 | 849 | BDRVRawReopenState *rs; |
31be8a2a | 850 | QemuOpts *opts; |
eeb6b45d | 851 | int ret = 0; |
df26a350 | 852 | Error *local_err = NULL; |
eeb6b45d JC |
853 | |
854 | assert(state != NULL); | |
855 | assert(state->bs != NULL); | |
856 | ||
857 | s = state->bs->opaque; | |
858 | ||
5839e53b | 859 | state->opaque = g_new0(BDRVRawReopenState, 1); |
4e6d13c9 | 860 | rs = state->opaque; |
31be8a2a SH |
861 | rs->fd = -1; |
862 | ||
863 | /* Handle options changes */ | |
864 | opts = qemu_opts_create(&raw_runtime_opts, NULL, 0, &error_abort); | |
865 | qemu_opts_absorb_qdict(opts, state->options, &local_err); | |
866 | if (local_err) { | |
867 | error_propagate(errp, local_err); | |
868 | ret = -EINVAL; | |
869 | goto out; | |
870 | } | |
871 | ||
8d324575 AG |
872 | rs->check_cache_dropped = |
873 | qemu_opt_get_bool_del(opts, "x-check-cache-dropped", false); | |
874 | ||
875 | /* This driver's reopen function doesn't currently allow changing | |
876 | * other options, so let's put them back in the original QDict and | |
877 | * bdrv_reopen_prepare() will detect changes and complain. */ | |
878 | qemu_opts_to_qdict(opts, state->options); | |
eeb6b45d | 879 | |
f709623b | 880 | if (s->type == FTYPE_CD) { |
4e6d13c9 | 881 | rs->open_flags |= O_NONBLOCK; |
1bc6b705 JC |
882 | } |
883 | ||
4e6d13c9 | 884 | raw_parse_flags(state->flags, &rs->open_flags); |
eeb6b45d | 885 | |
fdf263f6 | 886 | int fcntl_flags = O_APPEND | O_NONBLOCK; |
eeb6b45d JC |
887 | #ifdef O_NOATIME |
888 | fcntl_flags |= O_NOATIME; | |
889 | #endif | |
890 | ||
fdf263f6 AF |
891 | #ifdef O_ASYNC |
892 | /* Not all operating systems have O_ASYNC, and those that don't | |
4e6d13c9 | 893 | * will not let us track the state into rs->open_flags (typically |
fdf263f6 AF |
894 | * you achieve the same effect with an ioctl, for example I_SETSIG |
895 | * on Solaris). But we do not use O_ASYNC, so that's fine. | |
896 | */ | |
897 | assert((s->open_flags & O_ASYNC) == 0); | |
898 | #endif | |
899 | ||
4e6d13c9 | 900 | if ((rs->open_flags & ~fcntl_flags) == (s->open_flags & ~fcntl_flags)) { |
eeb6b45d | 901 | /* dup the original fd */ |
4e6d13c9 FZ |
902 | rs->fd = qemu_dup(s->fd); |
903 | if (rs->fd >= 0) { | |
904 | ret = fcntl_setfl(rs->fd, rs->open_flags); | |
eeb6b45d | 905 | if (ret) { |
4e6d13c9 FZ |
906 | qemu_close(rs->fd); |
907 | rs->fd = -1; | |
eeb6b45d JC |
908 | } |
909 | } | |
910 | } | |
911 | ||
912 | /* If we cannot use fcntl, or fcntl failed, fall back to qemu_open() */ | |
4e6d13c9 | 913 | if (rs->fd == -1) { |
bdd03cdf | 914 | const char *normalized_filename = state->bs->filename; |
db0754df FZ |
915 | ret = raw_normalize_devicepath(&normalized_filename, errp); |
916 | if (ret >= 0) { | |
4e6d13c9 FZ |
917 | assert(!(rs->open_flags & O_CREAT)); |
918 | rs->fd = qemu_open(normalized_filename, rs->open_flags); | |
919 | if (rs->fd == -1) { | |
bdd03cdf HR |
920 | error_setg_errno(errp, errno, "Could not reopen file"); |
921 | ret = -1; | |
922 | } | |
eeb6b45d JC |
923 | } |
924 | } | |
df26a350 KW |
925 | |
926 | /* Fail already reopen_prepare() if we can't get a working O_DIRECT | |
927 | * alignment with the new fd. */ | |
4e6d13c9 FZ |
928 | if (rs->fd != -1) { |
929 | raw_probe_alignment(state->bs, rs->fd, &local_err); | |
df26a350 | 930 | if (local_err) { |
4e6d13c9 FZ |
931 | qemu_close(rs->fd); |
932 | rs->fd = -1; | |
df26a350 KW |
933 | error_propagate(errp, local_err); |
934 | ret = -EINVAL; | |
935 | } | |
936 | } | |
937 | ||
31be8a2a SH |
938 | out: |
939 | qemu_opts_del(opts); | |
eeb6b45d JC |
940 | return ret; |
941 | } | |
942 | ||
eeb6b45d JC |
943 | static void raw_reopen_commit(BDRVReopenState *state) |
944 | { | |
4e6d13c9 | 945 | BDRVRawReopenState *rs = state->opaque; |
eeb6b45d | 946 | BDRVRawState *s = state->bs->opaque; |
f2e3af29 | 947 | Error *local_err = NULL; |
eeb6b45d | 948 | |
31be8a2a | 949 | s->check_cache_dropped = rs->check_cache_dropped; |
4e6d13c9 | 950 | s->open_flags = rs->open_flags; |
eeb6b45d | 951 | |
f2e3af29 FZ |
952 | /* Copy locks to the new fd before closing the old one. */ |
953 | raw_apply_lock_bytes(NULL, rs->fd, s->locked_perm, | |
577a1339 | 954 | s->locked_shared_perm, false, &local_err); |
f2e3af29 FZ |
955 | if (local_err) { |
956 | /* shouldn't fail in a sane host, but report it just in case. */ | |
957 | error_report_err(local_err); | |
958 | } | |
eeb6b45d | 959 | qemu_close(s->fd); |
4e6d13c9 | 960 | s->fd = rs->fd; |
eeb6b45d JC |
961 | |
962 | g_free(state->opaque); | |
963 | state->opaque = NULL; | |
964 | } | |
965 | ||
966 | ||
967 | static void raw_reopen_abort(BDRVReopenState *state) | |
968 | { | |
4e6d13c9 | 969 | BDRVRawReopenState *rs = state->opaque; |
eeb6b45d JC |
970 | |
971 | /* nothing to do if NULL, we didn't get far enough */ | |
4e6d13c9 | 972 | if (rs == NULL) { |
eeb6b45d JC |
973 | return; |
974 | } | |
975 | ||
4e6d13c9 FZ |
976 | if (rs->fd >= 0) { |
977 | qemu_close(rs->fd); | |
978 | rs->fd = -1; | |
eeb6b45d JC |
979 | } |
980 | g_free(state->opaque); | |
981 | state->opaque = NULL; | |
982 | } | |
983 | ||
48265250 | 984 | static int hdev_get_max_transfer_length(BlockDriverState *bs, int fd) |
6f607174 FZ |
985 | { |
986 | #ifdef BLKSECTGET | |
48265250 EF |
987 | int max_bytes = 0; |
988 | short max_sectors = 0; | |
989 | if (bs->sg && ioctl(fd, BLKSECTGET, &max_bytes) == 0) { | |
990 | return max_bytes; | |
991 | } else if (!bs->sg && ioctl(fd, BLKSECTGET, &max_sectors) == 0) { | |
992 | return max_sectors << BDRV_SECTOR_BITS; | |
6f607174 FZ |
993 | } else { |
994 | return -errno; | |
995 | } | |
996 | #else | |
997 | return -ENOSYS; | |
998 | #endif | |
999 | } | |
1000 | ||
9103f1ce FZ |
1001 | static int hdev_get_max_segments(const struct stat *st) |
1002 | { | |
1003 | #ifdef CONFIG_LINUX | |
1004 | char buf[32]; | |
1005 | const char *end; | |
1006 | char *sysfspath; | |
1007 | int ret; | |
1008 | int fd = -1; | |
1009 | long max_segments; | |
1010 | ||
1011 | sysfspath = g_strdup_printf("/sys/dev/block/%u:%u/queue/max_segments", | |
1012 | major(st->st_rdev), minor(st->st_rdev)); | |
1013 | fd = open(sysfspath, O_RDONLY); | |
1014 | if (fd == -1) { | |
1015 | ret = -errno; | |
1016 | goto out; | |
1017 | } | |
1018 | do { | |
69583490 | 1019 | ret = read(fd, buf, sizeof(buf) - 1); |
9103f1ce FZ |
1020 | } while (ret == -1 && errno == EINTR); |
1021 | if (ret < 0) { | |
1022 | ret = -errno; | |
1023 | goto out; | |
1024 | } else if (ret == 0) { | |
1025 | ret = -EIO; | |
1026 | goto out; | |
1027 | } | |
1028 | buf[ret] = 0; | |
1029 | /* The file is ended with '\n', pass 'end' to accept that. */ | |
1030 | ret = qemu_strtol(buf, &end, 10, &max_segments); | |
1031 | if (ret == 0 && end && *end == '\n') { | |
1032 | ret = max_segments; | |
1033 | } | |
1034 | ||
1035 | out: | |
fed414df FZ |
1036 | if (fd != -1) { |
1037 | close(fd); | |
1038 | } | |
9103f1ce FZ |
1039 | g_free(sysfspath); |
1040 | return ret; | |
1041 | #else | |
1042 | return -ENOTSUP; | |
1043 | #endif | |
1044 | } | |
1045 | ||
3baca891 | 1046 | static void raw_refresh_limits(BlockDriverState *bs, Error **errp) |
c25f53b0 PB |
1047 | { |
1048 | BDRVRawState *s = bs->opaque; | |
6f607174 FZ |
1049 | struct stat st; |
1050 | ||
1051 | if (!fstat(s->fd, &st)) { | |
c4c41a0a | 1052 | if (S_ISBLK(st.st_mode) || S_ISCHR(st.st_mode)) { |
48265250 EF |
1053 | int ret = hdev_get_max_transfer_length(bs, s->fd); |
1054 | if (ret > 0 && ret <= BDRV_REQUEST_MAX_BYTES) { | |
1055 | bs->bl.max_transfer = pow2floor(ret); | |
6f607174 | 1056 | } |
9103f1ce FZ |
1057 | ret = hdev_get_max_segments(&st); |
1058 | if (ret > 0) { | |
1059 | bs->bl.max_transfer = MIN(bs->bl.max_transfer, | |
1060 | ret * getpagesize()); | |
1061 | } | |
6f607174 FZ |
1062 | } |
1063 | } | |
eeb6b45d | 1064 | |
df26a350 | 1065 | raw_probe_alignment(bs, s->fd, errp); |
4196d2f0 | 1066 | bs->bl.min_mem_alignment = s->buf_align; |
459b4e66 | 1067 | bs->bl.opt_mem_alignment = MAX(s->buf_align, getpagesize()); |
c25f53b0 | 1068 | } |
83f64091 | 1069 | |
1a9335e4 ET |
1070 | static int check_for_dasd(int fd) |
1071 | { | |
1072 | #ifdef BIODASDINFO2 | |
1073 | struct dasd_information2_t info = {0}; | |
1074 | ||
1075 | return ioctl(fd, BIODASDINFO2, &info); | |
1076 | #else | |
1077 | return -1; | |
1078 | #endif | |
1079 | } | |
1080 | ||
1081 | /** | |
1082 | * Try to get @bs's logical and physical block size. | |
1083 | * On success, store them in @bsz and return zero. | |
1084 | * On failure, return negative errno. | |
1085 | */ | |
1086 | static int hdev_probe_blocksizes(BlockDriverState *bs, BlockSizes *bsz) | |
1087 | { | |
1088 | BDRVRawState *s = bs->opaque; | |
1089 | int ret; | |
1090 | ||
1091 | /* If DASD, get blocksizes */ | |
1092 | if (check_for_dasd(s->fd) < 0) { | |
1093 | return -ENOTSUP; | |
1094 | } | |
1095 | ret = probe_logical_blocksize(s->fd, &bsz->log); | |
1096 | if (ret < 0) { | |
1097 | return ret; | |
1098 | } | |
1099 | return probe_physical_blocksize(s->fd, &bsz->phys); | |
1100 | } | |
1101 | ||
1102 | /** | |
1103 | * Try to get @bs's geometry: cyls, heads, sectors. | |
1104 | * On success, store them in @geo and return 0. | |
1105 | * On failure return -errno. | |
1106 | * (Allows block driver to assign default geometry values that guest sees) | |
1107 | */ | |
1108 | #ifdef __linux__ | |
1109 | static int hdev_probe_geometry(BlockDriverState *bs, HDGeometry *geo) | |
1110 | { | |
1111 | BDRVRawState *s = bs->opaque; | |
1112 | struct hd_geometry ioctl_geo = {0}; | |
1a9335e4 ET |
1113 | |
1114 | /* If DASD, get its geometry */ | |
1115 | if (check_for_dasd(s->fd) < 0) { | |
1116 | return -ENOTSUP; | |
1117 | } | |
1118 | if (ioctl(s->fd, HDIO_GETGEO, &ioctl_geo) < 0) { | |
1119 | return -errno; | |
1120 | } | |
1121 | /* HDIO_GETGEO may return success even though geo contains zeros | |
1122 | (e.g. certain multipath setups) */ | |
1123 | if (!ioctl_geo.heads || !ioctl_geo.sectors || !ioctl_geo.cylinders) { | |
1124 | return -ENOTSUP; | |
1125 | } | |
1126 | /* Do not return a geometry for partition */ | |
1127 | if (ioctl_geo.start != 0) { | |
1128 | return -ENOTSUP; | |
1129 | } | |
1130 | geo->heads = ioctl_geo.heads; | |
1131 | geo->sectors = ioctl_geo.sectors; | |
1a9335e4 ET |
1132 | geo->cylinders = ioctl_geo.cylinders; |
1133 | ||
1134 | return 0; | |
1135 | } | |
1136 | #else /* __linux__ */ | |
1137 | static int hdev_probe_geometry(BlockDriverState *bs, HDGeometry *geo) | |
1138 | { | |
1139 | return -ENOTSUP; | |
1140 | } | |
1141 | #endif | |
1142 | ||
03425671 KW |
1143 | #if defined(__linux__) |
1144 | static int handle_aiocb_ioctl(void *opaque) | |
de81a169 | 1145 | { |
03425671 | 1146 | RawPosixAIOData *aiocb = opaque; |
de81a169 PB |
1147 | int ret; |
1148 | ||
d57c44d0 | 1149 | ret = ioctl(aiocb->aio_fildes, aiocb->ioctl.cmd, aiocb->ioctl.buf); |
de81a169 PB |
1150 | if (ret == -1) { |
1151 | return -errno; | |
1152 | } | |
1153 | ||
b608c8dc | 1154 | return 0; |
de81a169 | 1155 | } |
03425671 | 1156 | #endif /* linux */ |
de81a169 | 1157 | |
06dc9bd5 | 1158 | static int handle_aiocb_flush(void *opaque) |
de81a169 | 1159 | { |
06dc9bd5 | 1160 | RawPosixAIOData *aiocb = opaque; |
e5bcf967 | 1161 | BDRVRawState *s = aiocb->bs->opaque; |
de81a169 PB |
1162 | int ret; |
1163 | ||
e5bcf967 KW |
1164 | if (s->page_cache_inconsistent) { |
1165 | return -EIO; | |
1166 | } | |
1167 | ||
de81a169 PB |
1168 | ret = qemu_fdatasync(aiocb->aio_fildes); |
1169 | if (ret == -1) { | |
e5bcf967 KW |
1170 | /* There is no clear definition of the semantics of a failing fsync(), |
1171 | * so we may have to assume the worst. The sad truth is that this | |
1172 | * assumption is correct for Linux. Some pages are now probably marked | |
1173 | * clean in the page cache even though they are inconsistent with the | |
1174 | * on-disk contents. The next fdatasync() call would succeed, but no | |
1175 | * further writeback attempt will be made. We can't get back to a state | |
1176 | * in which we know what is on disk (we would have to rewrite | |
1177 | * everything that was touched since the last fdatasync() at least), so | |
1178 | * make bdrv_flush() fail permanently. Given that the behaviour isn't | |
1179 | * really defined, I have little hope that other OSes are doing better. | |
1180 | * | |
1181 | * Obviously, this doesn't affect O_DIRECT, which bypasses the page | |
1182 | * cache. */ | |
1183 | if ((s->open_flags & O_DIRECT) == 0) { | |
1184 | s->page_cache_inconsistent = true; | |
1185 | } | |
de81a169 PB |
1186 | return -errno; |
1187 | } | |
1188 | return 0; | |
1189 | } | |
1190 | ||
1191 | #ifdef CONFIG_PREADV | |
1192 | ||
1193 | static bool preadv_present = true; | |
1194 | ||
1195 | static ssize_t | |
1196 | qemu_preadv(int fd, const struct iovec *iov, int nr_iov, off_t offset) | |
1197 | { | |
1198 | return preadv(fd, iov, nr_iov, offset); | |
1199 | } | |
1200 | ||
1201 | static ssize_t | |
1202 | qemu_pwritev(int fd, const struct iovec *iov, int nr_iov, off_t offset) | |
1203 | { | |
1204 | return pwritev(fd, iov, nr_iov, offset); | |
1205 | } | |
1206 | ||
1207 | #else | |
1208 | ||
1209 | static bool preadv_present = false; | |
1210 | ||
1211 | static ssize_t | |
1212 | qemu_preadv(int fd, const struct iovec *iov, int nr_iov, off_t offset) | |
1213 | { | |
1214 | return -ENOSYS; | |
1215 | } | |
1216 | ||
1217 | static ssize_t | |
1218 | qemu_pwritev(int fd, const struct iovec *iov, int nr_iov, off_t offset) | |
1219 | { | |
1220 | return -ENOSYS; | |
1221 | } | |
1222 | ||
1223 | #endif | |
1224 | ||
1225 | static ssize_t handle_aiocb_rw_vector(RawPosixAIOData *aiocb) | |
1226 | { | |
1227 | ssize_t len; | |
1228 | ||
1229 | do { | |
1230 | if (aiocb->aio_type & QEMU_AIO_WRITE) | |
1231 | len = qemu_pwritev(aiocb->aio_fildes, | |
d57c44d0 KW |
1232 | aiocb->io.iov, |
1233 | aiocb->io.niov, | |
de81a169 PB |
1234 | aiocb->aio_offset); |
1235 | else | |
1236 | len = qemu_preadv(aiocb->aio_fildes, | |
d57c44d0 KW |
1237 | aiocb->io.iov, |
1238 | aiocb->io.niov, | |
de81a169 PB |
1239 | aiocb->aio_offset); |
1240 | } while (len == -1 && errno == EINTR); | |
1241 | ||
1242 | if (len == -1) { | |
1243 | return -errno; | |
1244 | } | |
1245 | return len; | |
1246 | } | |
1247 | ||
1248 | /* | |
1249 | * Read/writes the data to/from a given linear buffer. | |
1250 | * | |
1251 | * Returns the number of bytes handles or -errno in case of an error. Short | |
1252 | * reads are only returned if the end of the file is reached. | |
1253 | */ | |
1254 | static ssize_t handle_aiocb_rw_linear(RawPosixAIOData *aiocb, char *buf) | |
1255 | { | |
1256 | ssize_t offset = 0; | |
1257 | ssize_t len; | |
1258 | ||
1259 | while (offset < aiocb->aio_nbytes) { | |
1260 | if (aiocb->aio_type & QEMU_AIO_WRITE) { | |
1261 | len = pwrite(aiocb->aio_fildes, | |
1262 | (const char *)buf + offset, | |
1263 | aiocb->aio_nbytes - offset, | |
1264 | aiocb->aio_offset + offset); | |
1265 | } else { | |
1266 | len = pread(aiocb->aio_fildes, | |
1267 | buf + offset, | |
1268 | aiocb->aio_nbytes - offset, | |
1269 | aiocb->aio_offset + offset); | |
1270 | } | |
1271 | if (len == -1 && errno == EINTR) { | |
1272 | continue; | |
61ed73cf SH |
1273 | } else if (len == -1 && errno == EINVAL && |
1274 | (aiocb->bs->open_flags & BDRV_O_NOCACHE) && | |
1275 | !(aiocb->aio_type & QEMU_AIO_WRITE) && | |
1276 | offset > 0) { | |
1277 | /* O_DIRECT pread() may fail with EINVAL when offset is unaligned | |
1278 | * after a short read. Assume that O_DIRECT short reads only occur | |
1279 | * at EOF. Therefore this is a short read, not an I/O error. | |
1280 | */ | |
1281 | break; | |
de81a169 PB |
1282 | } else if (len == -1) { |
1283 | offset = -errno; | |
1284 | break; | |
1285 | } else if (len == 0) { | |
1286 | break; | |
1287 | } | |
1288 | offset += len; | |
1289 | } | |
1290 | ||
1291 | return offset; | |
1292 | } | |
1293 | ||
999e6b69 | 1294 | static int handle_aiocb_rw(void *opaque) |
de81a169 | 1295 | { |
999e6b69 | 1296 | RawPosixAIOData *aiocb = opaque; |
de81a169 PB |
1297 | ssize_t nbytes; |
1298 | char *buf; | |
1299 | ||
1300 | if (!(aiocb->aio_type & QEMU_AIO_MISALIGNED)) { | |
1301 | /* | |
1302 | * If there is just a single buffer, and it is properly aligned | |
1303 | * we can just use plain pread/pwrite without any problems. | |
1304 | */ | |
d57c44d0 | 1305 | if (aiocb->io.niov == 1) { |
54c7ca1b KW |
1306 | nbytes = handle_aiocb_rw_linear(aiocb, aiocb->io.iov->iov_base); |
1307 | goto out; | |
de81a169 PB |
1308 | } |
1309 | /* | |
1310 | * We have more than one iovec, and all are properly aligned. | |
1311 | * | |
1312 | * Try preadv/pwritev first and fall back to linearizing the | |
1313 | * buffer if it's not supported. | |
1314 | */ | |
1315 | if (preadv_present) { | |
1316 | nbytes = handle_aiocb_rw_vector(aiocb); | |
1317 | if (nbytes == aiocb->aio_nbytes || | |
1318 | (nbytes < 0 && nbytes != -ENOSYS)) { | |
54c7ca1b | 1319 | goto out; |
de81a169 PB |
1320 | } |
1321 | preadv_present = false; | |
1322 | } | |
1323 | ||
1324 | /* | |
1325 | * XXX(hch): short read/write. no easy way to handle the reminder | |
1326 | * using these interfaces. For now retry using plain | |
1327 | * pread/pwrite? | |
1328 | */ | |
1329 | } | |
1330 | ||
1331 | /* | |
1332 | * Ok, we have to do it the hard way, copy all segments into | |
1333 | * a single aligned buffer. | |
1334 | */ | |
50d4a858 KW |
1335 | buf = qemu_try_blockalign(aiocb->bs, aiocb->aio_nbytes); |
1336 | if (buf == NULL) { | |
54c7ca1b KW |
1337 | nbytes = -ENOMEM; |
1338 | goto out; | |
50d4a858 KW |
1339 | } |
1340 | ||
de81a169 PB |
1341 | if (aiocb->aio_type & QEMU_AIO_WRITE) { |
1342 | char *p = buf; | |
1343 | int i; | |
1344 | ||
d57c44d0 KW |
1345 | for (i = 0; i < aiocb->io.niov; ++i) { |
1346 | memcpy(p, aiocb->io.iov[i].iov_base, aiocb->io.iov[i].iov_len); | |
1347 | p += aiocb->io.iov[i].iov_len; | |
de81a169 | 1348 | } |
8eb029c2 | 1349 | assert(p - buf == aiocb->aio_nbytes); |
de81a169 PB |
1350 | } |
1351 | ||
1352 | nbytes = handle_aiocb_rw_linear(aiocb, buf); | |
1353 | if (!(aiocb->aio_type & QEMU_AIO_WRITE)) { | |
1354 | char *p = buf; | |
1355 | size_t count = aiocb->aio_nbytes, copy; | |
1356 | int i; | |
1357 | ||
d57c44d0 | 1358 | for (i = 0; i < aiocb->io.niov && count; ++i) { |
de81a169 | 1359 | copy = count; |
d57c44d0 KW |
1360 | if (copy > aiocb->io.iov[i].iov_len) { |
1361 | copy = aiocb->io.iov[i].iov_len; | |
de81a169 | 1362 | } |
d57c44d0 | 1363 | memcpy(aiocb->io.iov[i].iov_base, p, copy); |
8eb029c2 | 1364 | assert(count >= copy); |
de81a169 PB |
1365 | p += copy; |
1366 | count -= copy; | |
1367 | } | |
8eb029c2 | 1368 | assert(count == 0); |
de81a169 PB |
1369 | } |
1370 | qemu_vfree(buf); | |
1371 | ||
54c7ca1b KW |
1372 | out: |
1373 | if (nbytes == aiocb->aio_nbytes) { | |
1374 | return 0; | |
1375 | } else if (nbytes >= 0 && nbytes < aiocb->aio_nbytes) { | |
1376 | if (aiocb->aio_type & QEMU_AIO_WRITE) { | |
1377 | return -EINVAL; | |
1378 | } else { | |
1379 | iov_memset(aiocb->io.iov, aiocb->io.niov, nbytes, | |
1380 | 0, aiocb->aio_nbytes - nbytes); | |
1381 | return 0; | |
1382 | } | |
1383 | } else { | |
1384 | assert(nbytes < 0); | |
1385 | return nbytes; | |
1386 | } | |
de81a169 PB |
1387 | } |
1388 | ||
8238010b | 1389 | #ifdef CONFIG_XFS |
97a2ae34 PB |
1390 | static int xfs_write_zeroes(BDRVRawState *s, int64_t offset, uint64_t bytes) |
1391 | { | |
1392 | struct xfs_flock64 fl; | |
bcb22555 | 1393 | int err; |
97a2ae34 PB |
1394 | |
1395 | memset(&fl, 0, sizeof(fl)); | |
1396 | fl.l_whence = SEEK_SET; | |
1397 | fl.l_start = offset; | |
1398 | fl.l_len = bytes; | |
1399 | ||
1400 | if (xfsctl(NULL, s->fd, XFS_IOC_ZERO_RANGE, &fl) < 0) { | |
bcb22555 | 1401 | err = errno; |
4f7d28d7 | 1402 | trace_file_xfs_write_zeroes(strerror(errno)); |
bcb22555 | 1403 | return -err; |
97a2ae34 PB |
1404 | } |
1405 | ||
1406 | return 0; | |
1407 | } | |
1408 | ||
8238010b PB |
1409 | static int xfs_discard(BDRVRawState *s, int64_t offset, uint64_t bytes) |
1410 | { | |
1411 | struct xfs_flock64 fl; | |
bcb22555 | 1412 | int err; |
8238010b PB |
1413 | |
1414 | memset(&fl, 0, sizeof(fl)); | |
1415 | fl.l_whence = SEEK_SET; | |
1416 | fl.l_start = offset; | |
1417 | fl.l_len = bytes; | |
1418 | ||
1419 | if (xfsctl(NULL, s->fd, XFS_IOC_UNRESVSP64, &fl) < 0) { | |
bcb22555 | 1420 | err = errno; |
4f7d28d7 | 1421 | trace_file_xfs_discard(strerror(errno)); |
bcb22555 | 1422 | return -err; |
8238010b PB |
1423 | } |
1424 | ||
1425 | return 0; | |
1426 | } | |
1427 | #endif | |
1428 | ||
1486df0e DL |
1429 | static int translate_err(int err) |
1430 | { | |
1431 | if (err == -ENODEV || err == -ENOSYS || err == -EOPNOTSUPP || | |
1432 | err == -ENOTTY) { | |
1433 | err = -ENOTSUP; | |
1434 | } | |
1435 | return err; | |
1436 | } | |
1437 | ||
d50d8222 | 1438 | #ifdef CONFIG_FALLOCATE |
0b991712 DL |
1439 | static int do_fallocate(int fd, int mode, off_t offset, off_t len) |
1440 | { | |
1441 | do { | |
1442 | if (fallocate(fd, mode, offset, len) == 0) { | |
1443 | return 0; | |
1444 | } | |
1445 | } while (errno == EINTR); | |
1446 | return translate_err(-errno); | |
1447 | } | |
1448 | #endif | |
1449 | ||
37cc9f7f | 1450 | static ssize_t handle_aiocb_write_zeroes_block(RawPosixAIOData *aiocb) |
97a2ae34 | 1451 | { |
37cc9f7f | 1452 | int ret = -ENOTSUP; |
97a2ae34 PB |
1453 | BDRVRawState *s = aiocb->bs->opaque; |
1454 | ||
37cc9f7f | 1455 | if (!s->has_write_zeroes) { |
97a2ae34 PB |
1456 | return -ENOTSUP; |
1457 | } | |
1458 | ||
97a2ae34 | 1459 | #ifdef BLKZEROOUT |
37cc9f7f DL |
1460 | do { |
1461 | uint64_t range[2] = { aiocb->aio_offset, aiocb->aio_nbytes }; | |
1462 | if (ioctl(aiocb->aio_fildes, BLKZEROOUT, range) == 0) { | |
1463 | return 0; | |
97a2ae34 | 1464 | } |
37cc9f7f DL |
1465 | } while (errno == EINTR); |
1466 | ||
1467 | ret = translate_err(-errno); | |
97a2ae34 | 1468 | #endif |
97a2ae34 | 1469 | |
1486df0e | 1470 | if (ret == -ENOTSUP) { |
97a2ae34 | 1471 | s->has_write_zeroes = false; |
97a2ae34 PB |
1472 | } |
1473 | return ret; | |
1474 | } | |
1475 | ||
7154d8ae | 1476 | static int handle_aiocb_write_zeroes(void *opaque) |
37cc9f7f | 1477 | { |
7154d8ae | 1478 | RawPosixAIOData *aiocb = opaque; |
a6dcf097 | 1479 | #if defined(CONFIG_FALLOCATE) || defined(CONFIG_XFS) |
37cc9f7f | 1480 | BDRVRawState *s = aiocb->bs->opaque; |
a6dcf097 | 1481 | #endif |
70d9110b DL |
1482 | #ifdef CONFIG_FALLOCATE |
1483 | int64_t len; | |
1484 | #endif | |
37cc9f7f DL |
1485 | |
1486 | if (aiocb->aio_type & QEMU_AIO_BLKDEV) { | |
1487 | return handle_aiocb_write_zeroes_block(aiocb); | |
1488 | } | |
1489 | ||
1490 | #ifdef CONFIG_XFS | |
1491 | if (s->is_xfs) { | |
1492 | return xfs_write_zeroes(s, aiocb->aio_offset, aiocb->aio_nbytes); | |
1493 | } | |
1494 | #endif | |
1495 | ||
b953f075 DL |
1496 | #ifdef CONFIG_FALLOCATE_ZERO_RANGE |
1497 | if (s->has_write_zeroes) { | |
1498 | int ret = do_fallocate(s->fd, FALLOC_FL_ZERO_RANGE, | |
1499 | aiocb->aio_offset, aiocb->aio_nbytes); | |
1500 | if (ret == 0 || ret != -ENOTSUP) { | |
1501 | return ret; | |
1502 | } | |
1503 | s->has_write_zeroes = false; | |
1504 | } | |
1505 | #endif | |
1506 | ||
1cdc3239 DL |
1507 | #ifdef CONFIG_FALLOCATE_PUNCH_HOLE |
1508 | if (s->has_discard && s->has_fallocate) { | |
1509 | int ret = do_fallocate(s->fd, | |
1510 | FALLOC_FL_PUNCH_HOLE | FALLOC_FL_KEEP_SIZE, | |
1511 | aiocb->aio_offset, aiocb->aio_nbytes); | |
1512 | if (ret == 0) { | |
1513 | ret = do_fallocate(s->fd, 0, aiocb->aio_offset, aiocb->aio_nbytes); | |
1514 | if (ret == 0 || ret != -ENOTSUP) { | |
1515 | return ret; | |
1516 | } | |
1517 | s->has_fallocate = false; | |
1518 | } else if (ret != -ENOTSUP) { | |
1519 | return ret; | |
1520 | } else { | |
1521 | s->has_discard = false; | |
1522 | } | |
1523 | } | |
1524 | #endif | |
1525 | ||
d50d8222 | 1526 | #ifdef CONFIG_FALLOCATE |
70d9110b DL |
1527 | /* Last resort: we are trying to extend the file with zeroed data. This |
1528 | * can be done via fallocate(fd, 0) */ | |
1529 | len = bdrv_getlength(aiocb->bs); | |
1530 | if (s->has_fallocate && len >= 0 && aiocb->aio_offset >= len) { | |
d50d8222 DL |
1531 | int ret = do_fallocate(s->fd, 0, aiocb->aio_offset, aiocb->aio_nbytes); |
1532 | if (ret == 0 || ret != -ENOTSUP) { | |
1533 | return ret; | |
1534 | } | |
1535 | s->has_fallocate = false; | |
1536 | } | |
1537 | #endif | |
1538 | ||
37cc9f7f DL |
1539 | return -ENOTSUP; |
1540 | } | |
1541 | ||
7154d8ae | 1542 | static int handle_aiocb_write_zeroes_unmap(void *opaque) |
34fa110e | 1543 | { |
7154d8ae | 1544 | RawPosixAIOData *aiocb = opaque; |
34fa110e KW |
1545 | BDRVRawState *s G_GNUC_UNUSED = aiocb->bs->opaque; |
1546 | int ret; | |
1547 | ||
1548 | /* First try to write zeros and unmap at the same time */ | |
1549 | ||
1550 | #ifdef CONFIG_FALLOCATE_PUNCH_HOLE | |
1551 | ret = do_fallocate(s->fd, FALLOC_FL_PUNCH_HOLE | FALLOC_FL_KEEP_SIZE, | |
1552 | aiocb->aio_offset, aiocb->aio_nbytes); | |
1553 | if (ret != -ENOTSUP) { | |
1554 | return ret; | |
1555 | } | |
1556 | #endif | |
1557 | ||
1558 | #ifdef CONFIG_XFS | |
1559 | if (s->is_xfs) { | |
1560 | /* xfs_discard() guarantees that the discarded area reads as all-zero | |
1561 | * afterwards, so we can use it here. */ | |
1562 | return xfs_discard(s, aiocb->aio_offset, aiocb->aio_nbytes); | |
1563 | } | |
1564 | #endif | |
1565 | ||
1566 | /* If we couldn't manage to unmap while guaranteed that the area reads as | |
1567 | * all-zero afterwards, just write zeroes without unmapping */ | |
1568 | ret = handle_aiocb_write_zeroes(aiocb); | |
1569 | return ret; | |
1570 | } | |
1571 | ||
1efad060 FZ |
1572 | #ifndef HAVE_COPY_FILE_RANGE |
1573 | static off_t copy_file_range(int in_fd, off_t *in_off, int out_fd, | |
1574 | off_t *out_off, size_t len, unsigned int flags) | |
1575 | { | |
1576 | #ifdef __NR_copy_file_range | |
1577 | return syscall(__NR_copy_file_range, in_fd, in_off, out_fd, | |
1578 | out_off, len, flags); | |
1579 | #else | |
1580 | errno = ENOSYS; | |
1581 | return -1; | |
1582 | #endif | |
1583 | } | |
1584 | #endif | |
1585 | ||
58a209c4 | 1586 | static int handle_aiocb_copy_range(void *opaque) |
1efad060 | 1587 | { |
58a209c4 | 1588 | RawPosixAIOData *aiocb = opaque; |
1efad060 FZ |
1589 | uint64_t bytes = aiocb->aio_nbytes; |
1590 | off_t in_off = aiocb->aio_offset; | |
d57c44d0 | 1591 | off_t out_off = aiocb->copy_range.aio_offset2; |
1efad060 FZ |
1592 | |
1593 | while (bytes) { | |
1594 | ssize_t ret = copy_file_range(aiocb->aio_fildes, &in_off, | |
d57c44d0 | 1595 | aiocb->copy_range.aio_fd2, &out_off, |
1efad060 | 1596 | bytes, 0); |
ecc983a5 | 1597 | trace_file_copy_file_range(aiocb->bs, aiocb->aio_fildes, in_off, |
d57c44d0 KW |
1598 | aiocb->copy_range.aio_fd2, out_off, bytes, |
1599 | 0, ret); | |
c436e3d0 FZ |
1600 | if (ret == 0) { |
1601 | /* No progress (e.g. when beyond EOF), let the caller fall back to | |
1602 | * buffer I/O. */ | |
1603 | return -ENOSPC; | |
1efad060 FZ |
1604 | } |
1605 | if (ret < 0) { | |
c436e3d0 FZ |
1606 | switch (errno) { |
1607 | case ENOSYS: | |
1efad060 | 1608 | return -ENOTSUP; |
c436e3d0 FZ |
1609 | case EINTR: |
1610 | continue; | |
1611 | default: | |
1efad060 FZ |
1612 | return -errno; |
1613 | } | |
1614 | } | |
1efad060 FZ |
1615 | bytes -= ret; |
1616 | } | |
1617 | return 0; | |
1618 | } | |
1619 | ||
46ee0f46 | 1620 | static int handle_aiocb_discard(void *opaque) |
8238010b | 1621 | { |
46ee0f46 | 1622 | RawPosixAIOData *aiocb = opaque; |
8238010b PB |
1623 | int ret = -EOPNOTSUPP; |
1624 | BDRVRawState *s = aiocb->bs->opaque; | |
1625 | ||
7ce21016 PB |
1626 | if (!s->has_discard) { |
1627 | return -ENOTSUP; | |
8238010b PB |
1628 | } |
1629 | ||
1630 | if (aiocb->aio_type & QEMU_AIO_BLKDEV) { | |
1631 | #ifdef BLKDISCARD | |
1632 | do { | |
1633 | uint64_t range[2] = { aiocb->aio_offset, aiocb->aio_nbytes }; | |
1634 | if (ioctl(aiocb->aio_fildes, BLKDISCARD, range) == 0) { | |
1635 | return 0; | |
1636 | } | |
1637 | } while (errno == EINTR); | |
1638 | ||
1639 | ret = -errno; | |
1640 | #endif | |
1641 | } else { | |
1642 | #ifdef CONFIG_XFS | |
1643 | if (s->is_xfs) { | |
1644 | return xfs_discard(s, aiocb->aio_offset, aiocb->aio_nbytes); | |
1645 | } | |
1646 | #endif | |
1647 | ||
1648 | #ifdef CONFIG_FALLOCATE_PUNCH_HOLE | |
0b991712 DL |
1649 | ret = do_fallocate(s->fd, FALLOC_FL_PUNCH_HOLE | FALLOC_FL_KEEP_SIZE, |
1650 | aiocb->aio_offset, aiocb->aio_nbytes); | |
8238010b PB |
1651 | #endif |
1652 | } | |
1653 | ||
1486df0e DL |
1654 | ret = translate_err(ret); |
1655 | if (ret == -ENOTSUP) { | |
7ce21016 | 1656 | s->has_discard = false; |
8238010b PB |
1657 | } |
1658 | return ret; | |
1659 | } | |
1660 | ||
29cb4c01 | 1661 | static int handle_aiocb_truncate(void *opaque) |
93f4e2ff | 1662 | { |
29cb4c01 | 1663 | RawPosixAIOData *aiocb = opaque; |
93f4e2ff KW |
1664 | int result = 0; |
1665 | int64_t current_length = 0; | |
1666 | char *buf = NULL; | |
1667 | struct stat st; | |
1668 | int fd = aiocb->aio_fildes; | |
1669 | int64_t offset = aiocb->aio_offset; | |
d57c44d0 KW |
1670 | PreallocMode prealloc = aiocb->truncate.prealloc; |
1671 | Error **errp = aiocb->truncate.errp; | |
93f4e2ff KW |
1672 | |
1673 | if (fstat(fd, &st) < 0) { | |
1674 | result = -errno; | |
1675 | error_setg_errno(errp, -result, "Could not stat file"); | |
1676 | return result; | |
1677 | } | |
1678 | ||
1679 | current_length = st.st_size; | |
d57c44d0 | 1680 | if (current_length > offset && prealloc != PREALLOC_MODE_OFF) { |
93f4e2ff KW |
1681 | error_setg(errp, "Cannot use preallocation for shrinking files"); |
1682 | return -ENOTSUP; | |
1683 | } | |
1684 | ||
d57c44d0 | 1685 | switch (prealloc) { |
93f4e2ff KW |
1686 | #ifdef CONFIG_POSIX_FALLOCATE |
1687 | case PREALLOC_MODE_FALLOC: | |
1688 | /* | |
1689 | * Truncating before posix_fallocate() makes it about twice slower on | |
1690 | * file systems that do not support fallocate(), trying to check if a | |
1691 | * block is allocated before allocating it, so don't do that here. | |
1692 | */ | |
1693 | if (offset != current_length) { | |
1694 | result = -posix_fallocate(fd, current_length, | |
1695 | offset - current_length); | |
1696 | if (result != 0) { | |
1697 | /* posix_fallocate() doesn't set errno. */ | |
1698 | error_setg_errno(errp, -result, | |
1699 | "Could not preallocate new data"); | |
1700 | } | |
1701 | } else { | |
1702 | result = 0; | |
1703 | } | |
1704 | goto out; | |
1705 | #endif | |
1706 | case PREALLOC_MODE_FULL: | |
1707 | { | |
1708 | int64_t num = 0, left = offset - current_length; | |
1709 | off_t seek_result; | |
1710 | ||
1711 | /* | |
1712 | * Knowing the final size from the beginning could allow the file | |
1713 | * system driver to do less allocations and possibly avoid | |
1714 | * fragmentation of the file. | |
1715 | */ | |
1716 | if (ftruncate(fd, offset) != 0) { | |
1717 | result = -errno; | |
1718 | error_setg_errno(errp, -result, "Could not resize file"); | |
1719 | goto out; | |
1720 | } | |
1721 | ||
1722 | buf = g_malloc0(65536); | |
1723 | ||
1724 | seek_result = lseek(fd, current_length, SEEK_SET); | |
1725 | if (seek_result < 0) { | |
1726 | result = -errno; | |
1727 | error_setg_errno(errp, -result, | |
1728 | "Failed to seek to the old end of file"); | |
1729 | goto out; | |
1730 | } | |
1731 | ||
1732 | while (left > 0) { | |
1733 | num = MIN(left, 65536); | |
1734 | result = write(fd, buf, num); | |
1735 | if (result < 0) { | |
a1c81f4f FZ |
1736 | if (errno == EINTR) { |
1737 | continue; | |
1738 | } | |
93f4e2ff KW |
1739 | result = -errno; |
1740 | error_setg_errno(errp, -result, | |
1741 | "Could not write zeros for preallocation"); | |
1742 | goto out; | |
1743 | } | |
1744 | left -= result; | |
1745 | } | |
1746 | if (result >= 0) { | |
1747 | result = fsync(fd); | |
1748 | if (result < 0) { | |
1749 | result = -errno; | |
1750 | error_setg_errno(errp, -result, | |
1751 | "Could not flush file to disk"); | |
1752 | goto out; | |
1753 | } | |
1754 | } | |
1755 | goto out; | |
1756 | } | |
1757 | case PREALLOC_MODE_OFF: | |
1758 | if (ftruncate(fd, offset) != 0) { | |
1759 | result = -errno; | |
1760 | error_setg_errno(errp, -result, "Could not resize file"); | |
1761 | } | |
1762 | return result; | |
1763 | default: | |
1764 | result = -ENOTSUP; | |
1765 | error_setg(errp, "Unsupported preallocation mode: %s", | |
d57c44d0 | 1766 | PreallocMode_str(prealloc)); |
93f4e2ff KW |
1767 | return result; |
1768 | } | |
1769 | ||
1770 | out: | |
1771 | if (result < 0) { | |
1772 | if (ftruncate(fd, current_length) < 0) { | |
1773 | error_report("Failed to restore old file length: %s", | |
1774 | strerror(errno)); | |
1775 | } | |
1776 | } | |
1777 | ||
1778 | g_free(buf); | |
1779 | return result; | |
1780 | } | |
1781 | ||
5d5de250 KW |
1782 | static int coroutine_fn raw_thread_pool_submit(BlockDriverState *bs, |
1783 | ThreadPoolFunc func, void *arg) | |
1784 | { | |
1785 | /* @bs can be NULL, bdrv_get_aio_context() returns the main context then */ | |
1786 | ThreadPool *pool = aio_get_thread_pool(bdrv_get_aio_context(bs)); | |
1787 | return thread_pool_submit_co(pool, func, arg); | |
1788 | } | |
1789 | ||
9d52aa3c KW |
1790 | static int coroutine_fn raw_co_prw(BlockDriverState *bs, uint64_t offset, |
1791 | uint64_t bytes, QEMUIOVector *qiov, int type) | |
83f64091 | 1792 | { |
ce1a14dc | 1793 | BDRVRawState *s = bs->opaque; |
999e6b69 | 1794 | RawPosixAIOData acb; |
ce1a14dc | 1795 | |
19cb3738 | 1796 | if (fd_open(bs) < 0) |
2174f12b | 1797 | return -EIO; |
19cb3738 | 1798 | |
f141eafe | 1799 | /* |
3cad8307 RPM |
1800 | * Check if the underlying device requires requests to be aligned, |
1801 | * and if the request we are trying to submit is aligned or not. | |
1802 | * If this is the case tell the low-level driver that it needs | |
1803 | * to copy the buffer. | |
f141eafe | 1804 | */ |
3cad8307 | 1805 | if (s->needs_alignment) { |
c53b1c51 | 1806 | if (!bdrv_qiov_is_aligned(bs, qiov)) { |
5c6c3a6c | 1807 | type |= QEMU_AIO_MISALIGNED; |
e44bd6fc | 1808 | #ifdef CONFIG_LINUX_AIO |
0a4279d9 | 1809 | } else if (s->use_linux_aio) { |
0187f5c9 | 1810 | LinuxAioState *aio = aio_get_linux_aio(bdrv_get_aio_context(bs)); |
9d52aa3c | 1811 | assert(qiov->size == bytes); |
0187f5c9 | 1812 | return laio_co_submit(bs, aio, s->fd, offset, qiov, type); |
e44bd6fc | 1813 | #endif |
5c6c3a6c | 1814 | } |
9ef91a67 | 1815 | } |
f141eafe | 1816 | |
999e6b69 KW |
1817 | acb = (RawPosixAIOData) { |
1818 | .bs = bs, | |
1819 | .aio_fildes = s->fd, | |
1820 | .aio_type = type, | |
1821 | .aio_offset = offset, | |
1822 | .aio_nbytes = bytes, | |
1823 | .io = { | |
1824 | .iov = qiov->iov, | |
1825 | .niov = qiov->niov, | |
1826 | }, | |
1827 | }; | |
1828 | ||
1829 | assert(qiov->size == bytes); | |
1830 | return raw_thread_pool_submit(bs, handle_aiocb_rw, &acb); | |
2174f12b KW |
1831 | } |
1832 | ||
9d52aa3c KW |
1833 | static int coroutine_fn raw_co_preadv(BlockDriverState *bs, uint64_t offset, |
1834 | uint64_t bytes, QEMUIOVector *qiov, | |
1835 | int flags) | |
2174f12b | 1836 | { |
9d52aa3c | 1837 | return raw_co_prw(bs, offset, bytes, qiov, QEMU_AIO_READ); |
2174f12b KW |
1838 | } |
1839 | ||
9d52aa3c KW |
1840 | static int coroutine_fn raw_co_pwritev(BlockDriverState *bs, uint64_t offset, |
1841 | uint64_t bytes, QEMUIOVector *qiov, | |
1842 | int flags) | |
2174f12b | 1843 | { |
9d52aa3c KW |
1844 | assert(flags == 0); |
1845 | return raw_co_prw(bs, offset, bytes, qiov, QEMU_AIO_WRITE); | |
83f64091 FB |
1846 | } |
1847 | ||
1b3abdcc ML |
1848 | static void raw_aio_plug(BlockDriverState *bs) |
1849 | { | |
1850 | #ifdef CONFIG_LINUX_AIO | |
0a4279d9 KW |
1851 | BDRVRawState *s = bs->opaque; |
1852 | if (s->use_linux_aio) { | |
0187f5c9 PB |
1853 | LinuxAioState *aio = aio_get_linux_aio(bdrv_get_aio_context(bs)); |
1854 | laio_io_plug(bs, aio); | |
1b3abdcc ML |
1855 | } |
1856 | #endif | |
1857 | } | |
1858 | ||
1859 | static void raw_aio_unplug(BlockDriverState *bs) | |
1860 | { | |
1861 | #ifdef CONFIG_LINUX_AIO | |
0a4279d9 KW |
1862 | BDRVRawState *s = bs->opaque; |
1863 | if (s->use_linux_aio) { | |
0187f5c9 PB |
1864 | LinuxAioState *aio = aio_get_linux_aio(bdrv_get_aio_context(bs)); |
1865 | laio_io_unplug(bs, aio); | |
1b3abdcc ML |
1866 | } |
1867 | #endif | |
1868 | } | |
1869 | ||
33d70fb6 | 1870 | static int raw_co_flush_to_disk(BlockDriverState *bs) |
b2e12bc6 CH |
1871 | { |
1872 | BDRVRawState *s = bs->opaque; | |
06dc9bd5 | 1873 | RawPosixAIOData acb; |
33d70fb6 | 1874 | int ret; |
b2e12bc6 | 1875 | |
33d70fb6 KW |
1876 | ret = fd_open(bs); |
1877 | if (ret < 0) { | |
1878 | return ret; | |
1879 | } | |
b2e12bc6 | 1880 | |
06dc9bd5 KW |
1881 | acb = (RawPosixAIOData) { |
1882 | .bs = bs, | |
1883 | .aio_fildes = s->fd, | |
1884 | .aio_type = QEMU_AIO_FLUSH, | |
1885 | }; | |
1886 | ||
1887 | return raw_thread_pool_submit(bs, handle_aiocb_flush, &acb); | |
b2e12bc6 CH |
1888 | } |
1889 | ||
ed6e2161 NA |
1890 | static void raw_aio_attach_aio_context(BlockDriverState *bs, |
1891 | AioContext *new_context) | |
1892 | { | |
1893 | #ifdef CONFIG_LINUX_AIO | |
1894 | BDRVRawState *s = bs->opaque; | |
1895 | if (s->use_linux_aio) { | |
1896 | Error *local_err; | |
1897 | if (!aio_setup_linux_aio(new_context, &local_err)) { | |
1898 | error_reportf_err(local_err, "Unable to use native AIO, " | |
1899 | "falling back to thread pool: "); | |
1900 | s->use_linux_aio = false; | |
1901 | } | |
1902 | } | |
1903 | #endif | |
1904 | } | |
1905 | ||
83f64091 FB |
1906 | static void raw_close(BlockDriverState *bs) |
1907 | { | |
1908 | BDRVRawState *s = bs->opaque; | |
c2f3426c | 1909 | |
19cb3738 | 1910 | if (s->fd >= 0) { |
2e1e79da | 1911 | qemu_close(s->fd); |
19cb3738 FB |
1912 | s->fd = -1; |
1913 | } | |
83f64091 FB |
1914 | } |
1915 | ||
d0bc9e5d HR |
1916 | /** |
1917 | * Truncates the given regular file @fd to @offset and, when growing, fills the | |
1918 | * new space according to @prealloc. | |
1919 | * | |
1920 | * Returns: 0 on success, -errno on failure. | |
1921 | */ | |
93f4e2ff KW |
1922 | static int coroutine_fn |
1923 | raw_regular_truncate(BlockDriverState *bs, int fd, int64_t offset, | |
1924 | PreallocMode prealloc, Error **errp) | |
9f63b07e | 1925 | { |
29cb4c01 | 1926 | RawPosixAIOData acb; |
d0bc9e5d | 1927 | |
29cb4c01 | 1928 | acb = (RawPosixAIOData) { |
93f4e2ff KW |
1929 | .bs = bs, |
1930 | .aio_fildes = fd, | |
1931 | .aio_type = QEMU_AIO_TRUNCATE, | |
1932 | .aio_offset = offset, | |
d57c44d0 KW |
1933 | .truncate = { |
1934 | .prealloc = prealloc, | |
1935 | .errp = errp, | |
1936 | }, | |
93f4e2ff | 1937 | }; |
d0bc9e5d | 1938 | |
29cb4c01 | 1939 | return raw_thread_pool_submit(bs, handle_aiocb_truncate, &acb); |
9f63b07e HR |
1940 | } |
1941 | ||
061ca8a3 KW |
1942 | static int coroutine_fn raw_co_truncate(BlockDriverState *bs, int64_t offset, |
1943 | PreallocMode prealloc, Error **errp) | |
83f64091 FB |
1944 | { |
1945 | BDRVRawState *s = bs->opaque; | |
55b949c8 | 1946 | struct stat st; |
f59adb32 | 1947 | int ret; |
55b949c8 CH |
1948 | |
1949 | if (fstat(s->fd, &st)) { | |
f59adb32 HR |
1950 | ret = -errno; |
1951 | error_setg_errno(errp, -ret, "Failed to fstat() the file"); | |
1952 | return ret; | |
55b949c8 CH |
1953 | } |
1954 | ||
1955 | if (S_ISREG(st.st_mode)) { | |
93f4e2ff | 1956 | return raw_regular_truncate(bs, s->fd, offset, prealloc, errp); |
35d72602 HR |
1957 | } |
1958 | ||
1959 | if (prealloc != PREALLOC_MODE_OFF) { | |
1960 | error_setg(errp, "Preallocation mode '%s' unsupported for this " | |
977c736f | 1961 | "non-regular file", PreallocMode_str(prealloc)); |
35d72602 HR |
1962 | return -ENOTSUP; |
1963 | } | |
1964 | ||
1965 | if (S_ISCHR(st.st_mode) || S_ISBLK(st.st_mode)) { | |
f59adb32 HR |
1966 | if (offset > raw_getlength(bs)) { |
1967 | error_setg(errp, "Cannot grow device files"); | |
1968 | return -EINVAL; | |
1969 | } | |
55b949c8 | 1970 | } else { |
f59adb32 | 1971 | error_setg(errp, "Resizing this file is not supported"); |
55b949c8 CH |
1972 | return -ENOTSUP; |
1973 | } | |
1974 | ||
83f64091 FB |
1975 | return 0; |
1976 | } | |
1977 | ||
128ab2ff BS |
1978 | #ifdef __OpenBSD__ |
1979 | static int64_t raw_getlength(BlockDriverState *bs) | |
1980 | { | |
1981 | BDRVRawState *s = bs->opaque; | |
1982 | int fd = s->fd; | |
1983 | struct stat st; | |
1984 | ||
1985 | if (fstat(fd, &st)) | |
aa729704 | 1986 | return -errno; |
128ab2ff BS |
1987 | if (S_ISCHR(st.st_mode) || S_ISBLK(st.st_mode)) { |
1988 | struct disklabel dl; | |
1989 | ||
1990 | if (ioctl(fd, DIOCGDINFO, &dl)) | |
aa729704 | 1991 | return -errno; |
128ab2ff BS |
1992 | return (uint64_t)dl.d_secsize * |
1993 | dl.d_partitions[DISKPART(st.st_rdev)].p_size; | |
1994 | } else | |
1995 | return st.st_size; | |
1996 | } | |
d1f6fd8d CE |
1997 | #elif defined(__NetBSD__) |
1998 | static int64_t raw_getlength(BlockDriverState *bs) | |
1999 | { | |
2000 | BDRVRawState *s = bs->opaque; | |
2001 | int fd = s->fd; | |
2002 | struct stat st; | |
2003 | ||
2004 | if (fstat(fd, &st)) | |
aa729704 | 2005 | return -errno; |
d1f6fd8d CE |
2006 | if (S_ISCHR(st.st_mode) || S_ISBLK(st.st_mode)) { |
2007 | struct dkwedge_info dkw; | |
2008 | ||
2009 | if (ioctl(fd, DIOCGWEDGEINFO, &dkw) != -1) { | |
2010 | return dkw.dkw_size * 512; | |
2011 | } else { | |
2012 | struct disklabel dl; | |
2013 | ||
2014 | if (ioctl(fd, DIOCGDINFO, &dl)) | |
aa729704 | 2015 | return -errno; |
d1f6fd8d CE |
2016 | return (uint64_t)dl.d_secsize * |
2017 | dl.d_partitions[DISKPART(st.st_rdev)].p_size; | |
2018 | } | |
2019 | } else | |
2020 | return st.st_size; | |
2021 | } | |
50779cc2 CH |
2022 | #elif defined(__sun__) |
2023 | static int64_t raw_getlength(BlockDriverState *bs) | |
2024 | { | |
2025 | BDRVRawState *s = bs->opaque; | |
2026 | struct dk_minfo minfo; | |
2027 | int ret; | |
aa729704 | 2028 | int64_t size; |
50779cc2 CH |
2029 | |
2030 | ret = fd_open(bs); | |
2031 | if (ret < 0) { | |
2032 | return ret; | |
2033 | } | |
2034 | ||
2035 | /* | |
2036 | * Use the DKIOCGMEDIAINFO ioctl to read the size. | |
2037 | */ | |
2038 | ret = ioctl(s->fd, DKIOCGMEDIAINFO, &minfo); | |
2039 | if (ret != -1) { | |
2040 | return minfo.dki_lbsize * minfo.dki_capacity; | |
2041 | } | |
2042 | ||
2043 | /* | |
2044 | * There are reports that lseek on some devices fails, but | |
2045 | * irc discussion said that contingency on contingency was overkill. | |
2046 | */ | |
aa729704 MA |
2047 | size = lseek(s->fd, 0, SEEK_END); |
2048 | if (size < 0) { | |
2049 | return -errno; | |
2050 | } | |
2051 | return size; | |
50779cc2 CH |
2052 | } |
2053 | #elif defined(CONFIG_BSD) | |
2054 | static int64_t raw_getlength(BlockDriverState *bs) | |
83f64091 FB |
2055 | { |
2056 | BDRVRawState *s = bs->opaque; | |
2057 | int fd = s->fd; | |
2058 | int64_t size; | |
83f64091 | 2059 | struct stat sb; |
a167ba50 | 2060 | #if defined (__FreeBSD__) || defined(__FreeBSD_kernel__) |
9f23011a | 2061 | int reopened = 0; |
83f64091 | 2062 | #endif |
19cb3738 FB |
2063 | int ret; |
2064 | ||
2065 | ret = fd_open(bs); | |
2066 | if (ret < 0) | |
2067 | return ret; | |
83f64091 | 2068 | |
a167ba50 | 2069 | #if defined (__FreeBSD__) || defined(__FreeBSD_kernel__) |
9f23011a BS |
2070 | again: |
2071 | #endif | |
83f64091 FB |
2072 | if (!fstat(fd, &sb) && (S_IFCHR & sb.st_mode)) { |
2073 | #ifdef DIOCGMEDIASIZE | |
7d37435b | 2074 | if (ioctl(fd, DIOCGMEDIASIZE, (off_t *)&size)) |
c5e97233 BS |
2075 | #elif defined(DIOCGPART) |
2076 | { | |
2077 | struct partinfo pi; | |
2078 | if (ioctl(fd, DIOCGPART, &pi) == 0) | |
2079 | size = pi.media_size; | |
2080 | else | |
2081 | size = 0; | |
2082 | } | |
2083 | if (size == 0) | |
83f64091 | 2084 | #endif |
83affaa6 | 2085 | #if defined(__APPLE__) && defined(__MACH__) |
728dacbd JA |
2086 | { |
2087 | uint64_t sectors = 0; | |
2088 | uint32_t sector_size = 0; | |
2089 | ||
2090 | if (ioctl(fd, DKIOCGETBLOCKCOUNT, §ors) == 0 | |
2091 | && ioctl(fd, DKIOCGETBLOCKSIZE, §or_size) == 0) { | |
2092 | size = sectors * sector_size; | |
2093 | } else { | |
2094 | size = lseek(fd, 0LL, SEEK_END); | |
2095 | if (size < 0) { | |
2096 | return -errno; | |
2097 | } | |
2098 | } | |
2099 | } | |
83f64091 FB |
2100 | #else |
2101 | size = lseek(fd, 0LL, SEEK_END); | |
aa729704 MA |
2102 | if (size < 0) { |
2103 | return -errno; | |
2104 | } | |
9f23011a | 2105 | #endif |
a167ba50 | 2106 | #if defined(__FreeBSD__) || defined(__FreeBSD_kernel__) |
9f23011a BS |
2107 | switch(s->type) { |
2108 | case FTYPE_CD: | |
2109 | /* XXX FreeBSD acd returns UINT_MAX sectors for an empty drive */ | |
2110 | if (size == 2048LL * (unsigned)-1) | |
2111 | size = 0; | |
2112 | /* XXX no disc? maybe we need to reopen... */ | |
f3a5d3f8 | 2113 | if (size <= 0 && !reopened && cdrom_reopen(bs) >= 0) { |
9f23011a BS |
2114 | reopened = 1; |
2115 | goto again; | |
2116 | } | |
2117 | } | |
83f64091 | 2118 | #endif |
50779cc2 | 2119 | } else { |
83f64091 | 2120 | size = lseek(fd, 0, SEEK_END); |
aa729704 MA |
2121 | if (size < 0) { |
2122 | return -errno; | |
2123 | } | |
83f64091 | 2124 | } |
83f64091 FB |
2125 | return size; |
2126 | } | |
50779cc2 CH |
2127 | #else |
2128 | static int64_t raw_getlength(BlockDriverState *bs) | |
2129 | { | |
2130 | BDRVRawState *s = bs->opaque; | |
2131 | int ret; | |
aa729704 | 2132 | int64_t size; |
50779cc2 CH |
2133 | |
2134 | ret = fd_open(bs); | |
2135 | if (ret < 0) { | |
2136 | return ret; | |
2137 | } | |
2138 | ||
aa729704 MA |
2139 | size = lseek(s->fd, 0, SEEK_END); |
2140 | if (size < 0) { | |
2141 | return -errno; | |
2142 | } | |
2143 | return size; | |
50779cc2 | 2144 | } |
128ab2ff | 2145 | #endif |
83f64091 | 2146 | |
4a1d5e1f FZ |
2147 | static int64_t raw_get_allocated_file_size(BlockDriverState *bs) |
2148 | { | |
2149 | struct stat st; | |
2150 | BDRVRawState *s = bs->opaque; | |
2151 | ||
2152 | if (fstat(s->fd, &st) < 0) { | |
2153 | return -errno; | |
2154 | } | |
2155 | return (int64_t)st.st_blocks * 512; | |
2156 | } | |
2157 | ||
93f4e2ff KW |
2158 | static int coroutine_fn |
2159 | raw_co_create(BlockdevCreateOptions *options, Error **errp) | |
83f64091 | 2160 | { |
927f11e1 | 2161 | BlockdevCreateOptionsFile *file_opts; |
7c20c808 | 2162 | Error *local_err = NULL; |
83f64091 | 2163 | int fd; |
d815efca | 2164 | uint64_t perm, shared; |
1e37d059 | 2165 | int result = 0; |
83f64091 | 2166 | |
927f11e1 KW |
2167 | /* Validate options and set default values */ |
2168 | assert(options->driver == BLOCKDEV_DRIVER_FILE); | |
2169 | file_opts = &options->u.file; | |
464d9f64 | 2170 | |
927f11e1 KW |
2171 | if (!file_opts->has_nocow) { |
2172 | file_opts->nocow = false; | |
2173 | } | |
2174 | if (!file_opts->has_preallocation) { | |
2175 | file_opts->preallocation = PREALLOC_MODE_OFF; | |
06247428 | 2176 | } |
83f64091 | 2177 | |
927f11e1 | 2178 | /* Create file */ |
b8cf1913 | 2179 | fd = qemu_open(file_opts->filename, O_RDWR | O_CREAT | O_BINARY, 0644); |
1e37d059 SW |
2180 | if (fd < 0) { |
2181 | result = -errno; | |
e428e439 | 2182 | error_setg_errno(errp, -result, "Could not create file"); |
06247428 HT |
2183 | goto out; |
2184 | } | |
2185 | ||
b8cf1913 HR |
2186 | /* Take permissions: We want to discard everything, so we need |
2187 | * BLK_PERM_WRITE; and truncation to the desired size requires | |
2188 | * BLK_PERM_RESIZE. | |
2189 | * On the other hand, we cannot share the RESIZE permission | |
2190 | * because we promise that after this function, the file has the | |
2191 | * size given in the options. If someone else were to resize it | |
2192 | * concurrently, we could not guarantee that. | |
2193 | * Note that after this function, we can no longer guarantee that | |
2194 | * the file is not touched by a third party, so it may be resized | |
2195 | * then. */ | |
2196 | perm = BLK_PERM_WRITE | BLK_PERM_RESIZE; | |
2197 | shared = BLK_PERM_ALL & ~BLK_PERM_RESIZE; | |
2198 | ||
2199 | /* Step one: Take locks */ | |
2996ffad | 2200 | result = raw_apply_lock_bytes(NULL, fd, perm, ~shared, false, errp); |
b8cf1913 HR |
2201 | if (result < 0) { |
2202 | goto out_close; | |
2203 | } | |
2204 | ||
2205 | /* Step two: Check that nobody else has taken conflicting locks */ | |
2206 | result = raw_check_lock_bytes(fd, perm, shared, errp); | |
2207 | if (result < 0) { | |
b857431d FZ |
2208 | error_append_hint(errp, |
2209 | "Is another process using the image [%s]?\n", | |
2210 | file_opts->filename); | |
7c20c808 | 2211 | goto out_unlock; |
b8cf1913 HR |
2212 | } |
2213 | ||
2214 | /* Clear the file by truncating it to 0 */ | |
93f4e2ff | 2215 | result = raw_regular_truncate(NULL, fd, 0, PREALLOC_MODE_OFF, errp); |
b8cf1913 | 2216 | if (result < 0) { |
7c20c808 | 2217 | goto out_unlock; |
b8cf1913 HR |
2218 | } |
2219 | ||
927f11e1 | 2220 | if (file_opts->nocow) { |
4ab15590 | 2221 | #ifdef __linux__ |
06247428 HT |
2222 | /* Set NOCOW flag to solve performance issue on fs like btrfs. |
2223 | * This is an optimisation. The FS_IOC_SETFLAGS ioctl return value | |
2224 | * will be ignored since any failure of this operation should not | |
2225 | * block the left work. | |
2226 | */ | |
2227 | int attr; | |
2228 | if (ioctl(fd, FS_IOC_GETFLAGS, &attr) == 0) { | |
2229 | attr |= FS_NOCOW_FL; | |
2230 | ioctl(fd, FS_IOC_SETFLAGS, &attr); | |
4ab15590 | 2231 | } |
06247428 HT |
2232 | #endif |
2233 | } | |
2234 | ||
b8cf1913 HR |
2235 | /* Resize and potentially preallocate the file to the desired |
2236 | * final size */ | |
93f4e2ff KW |
2237 | result = raw_regular_truncate(NULL, fd, file_opts->size, |
2238 | file_opts->preallocation, errp); | |
9f63b07e | 2239 | if (result < 0) { |
7c20c808 HR |
2240 | goto out_unlock; |
2241 | } | |
2242 | ||
2243 | out_unlock: | |
2996ffad | 2244 | raw_apply_lock_bytes(NULL, fd, 0, 0, true, &local_err); |
7c20c808 HR |
2245 | if (local_err) { |
2246 | /* The above call should not fail, and if it does, that does | |
2247 | * not mean the whole creation operation has failed. So | |
2248 | * report it the user for their convenience, but do not report | |
2249 | * it to the caller. */ | |
db0754df | 2250 | warn_report_err(local_err); |
1e37d059 | 2251 | } |
06247428 | 2252 | |
5a1dad9d | 2253 | out_close: |
06247428 HT |
2254 | if (qemu_close(fd) != 0 && result == 0) { |
2255 | result = -errno; | |
2256 | error_setg_errno(errp, -result, "Could not close the new file"); | |
2257 | } | |
2258 | out: | |
1e37d059 | 2259 | return result; |
83f64091 FB |
2260 | } |
2261 | ||
927f11e1 KW |
2262 | static int coroutine_fn raw_co_create_opts(const char *filename, QemuOpts *opts, |
2263 | Error **errp) | |
2264 | { | |
2265 | BlockdevCreateOptions options; | |
2266 | int64_t total_size = 0; | |
2267 | bool nocow = false; | |
2268 | PreallocMode prealloc; | |
2269 | char *buf = NULL; | |
2270 | Error *local_err = NULL; | |
2271 | ||
2272 | /* Skip file: protocol prefix */ | |
2273 | strstart(filename, "file:", &filename); | |
2274 | ||
2275 | /* Read out options */ | |
2276 | total_size = ROUND_UP(qemu_opt_get_size_del(opts, BLOCK_OPT_SIZE, 0), | |
2277 | BDRV_SECTOR_SIZE); | |
2278 | nocow = qemu_opt_get_bool(opts, BLOCK_OPT_NOCOW, false); | |
2279 | buf = qemu_opt_get_del(opts, BLOCK_OPT_PREALLOC); | |
2280 | prealloc = qapi_enum_parse(&PreallocMode_lookup, buf, | |
2281 | PREALLOC_MODE_OFF, &local_err); | |
2282 | g_free(buf); | |
2283 | if (local_err) { | |
2284 | error_propagate(errp, local_err); | |
2285 | return -EINVAL; | |
2286 | } | |
2287 | ||
2288 | options = (BlockdevCreateOptions) { | |
2289 | .driver = BLOCKDEV_DRIVER_FILE, | |
2290 | .u.file = { | |
2291 | .filename = (char *) filename, | |
2292 | .size = total_size, | |
2293 | .has_preallocation = true, | |
2294 | .preallocation = prealloc, | |
2295 | .has_nocow = true, | |
2296 | .nocow = nocow, | |
2297 | }, | |
2298 | }; | |
2299 | return raw_co_create(&options, errp); | |
2300 | } | |
2301 | ||
d1f06fe6 MA |
2302 | /* |
2303 | * Find allocation range in @bs around offset @start. | |
2304 | * May change underlying file descriptor's file offset. | |
2305 | * If @start is not in a hole, store @start in @data, and the | |
2306 | * beginning of the next hole in @hole, and return 0. | |
2307 | * If @start is in a non-trailing hole, store @start in @hole and the | |
2308 | * beginning of the next non-hole in @data, and return 0. | |
2309 | * If @start is in a trailing hole or beyond EOF, return -ENXIO. | |
2310 | * If we can't find out, return a negative errno other than -ENXIO. | |
2311 | */ | |
2312 | static int find_allocation(BlockDriverState *bs, off_t start, | |
2313 | off_t *data, off_t *hole) | |
4f11aa8a HR |
2314 | { |
2315 | #if defined SEEK_HOLE && defined SEEK_DATA | |
94282e71 | 2316 | BDRVRawState *s = bs->opaque; |
d1f06fe6 | 2317 | off_t offs; |
94282e71 | 2318 | |
d1f06fe6 MA |
2319 | /* |
2320 | * SEEK_DATA cases: | |
2321 | * D1. offs == start: start is in data | |
2322 | * D2. offs > start: start is in a hole, next data at offs | |
2323 | * D3. offs < 0, errno = ENXIO: either start is in a trailing hole | |
2324 | * or start is beyond EOF | |
2325 | * If the latter happens, the file has been truncated behind | |
2326 | * our back since we opened it. All bets are off then. | |
2327 | * Treating like a trailing hole is simplest. | |
2328 | * D4. offs < 0, errno != ENXIO: we learned nothing | |
2329 | */ | |
2330 | offs = lseek(s->fd, start, SEEK_DATA); | |
2331 | if (offs < 0) { | |
2332 | return -errno; /* D3 or D4 */ | |
2333 | } | |
a03083a0 JC |
2334 | |
2335 | if (offs < start) { | |
2336 | /* This is not a valid return by lseek(). We are safe to just return | |
2337 | * -EIO in this case, and we'll treat it like D4. */ | |
2338 | return -EIO; | |
2339 | } | |
d1f06fe6 MA |
2340 | |
2341 | if (offs > start) { | |
2342 | /* D2: in hole, next data at offs */ | |
2343 | *hole = start; | |
2344 | *data = offs; | |
2345 | return 0; | |
5500316d PB |
2346 | } |
2347 | ||
d1f06fe6 MA |
2348 | /* D1: in data, end not yet known */ |
2349 | ||
2350 | /* | |
2351 | * SEEK_HOLE cases: | |
2352 | * H1. offs == start: start is in a hole | |
2353 | * If this happens here, a hole has been dug behind our back | |
2354 | * since the previous lseek(). | |
2355 | * H2. offs > start: either start is in data, next hole at offs, | |
2356 | * or start is in trailing hole, EOF at offs | |
2357 | * Linux treats trailing holes like any other hole: offs == | |
2358 | * start. Solaris seeks to EOF instead: offs > start (blech). | |
2359 | * If that happens here, a hole has been dug behind our back | |
2360 | * since the previous lseek(). | |
2361 | * H3. offs < 0, errno = ENXIO: start is beyond EOF | |
2362 | * If this happens, the file has been truncated behind our | |
2363 | * back since we opened it. Treat it like a trailing hole. | |
2364 | * H4. offs < 0, errno != ENXIO: we learned nothing | |
2365 | * Pretend we know nothing at all, i.e. "forget" about D1. | |
2366 | */ | |
2367 | offs = lseek(s->fd, start, SEEK_HOLE); | |
2368 | if (offs < 0) { | |
2369 | return -errno; /* D1 and (H3 or H4) */ | |
2370 | } | |
a03083a0 JC |
2371 | |
2372 | if (offs < start) { | |
2373 | /* This is not a valid return by lseek(). We are safe to just return | |
2374 | * -EIO in this case, and we'll treat it like H4. */ | |
2375 | return -EIO; | |
2376 | } | |
d1f06fe6 MA |
2377 | |
2378 | if (offs > start) { | |
2379 | /* | |
2380 | * D1 and H2: either in data, next hole at offs, or it was in | |
2381 | * data but is now in a trailing hole. In the latter case, | |
2382 | * all bets are off. Treating it as if it there was data all | |
2383 | * the way to EOF is safe, so simply do that. | |
2384 | */ | |
4f11aa8a | 2385 | *data = start; |
d1f06fe6 MA |
2386 | *hole = offs; |
2387 | return 0; | |
5500316d | 2388 | } |
4f11aa8a | 2389 | |
d1f06fe6 MA |
2390 | /* D1 and H1 */ |
2391 | return -EBUSY; | |
5500316d | 2392 | #else |
4f11aa8a | 2393 | return -ENOTSUP; |
5500316d | 2394 | #endif |
4f11aa8a HR |
2395 | } |
2396 | ||
2397 | /* | |
a290f085 | 2398 | * Returns the allocation status of the specified offset. |
4f11aa8a | 2399 | * |
a290f085 | 2400 | * The block layer guarantees 'offset' and 'bytes' are within bounds. |
4f11aa8a | 2401 | * |
a290f085 EB |
2402 | * 'pnum' is set to the number of bytes (including and immediately following |
2403 | * the specified offset) that are known to be in the same | |
4f11aa8a HR |
2404 | * allocated/unallocated state. |
2405 | * | |
a290f085 | 2406 | * 'bytes' is the max value 'pnum' should be set to. |
4f11aa8a | 2407 | */ |
a290f085 EB |
2408 | static int coroutine_fn raw_co_block_status(BlockDriverState *bs, |
2409 | bool want_zero, | |
2410 | int64_t offset, | |
2411 | int64_t bytes, int64_t *pnum, | |
2412 | int64_t *map, | |
2413 | BlockDriverState **file) | |
2414 | { | |
2415 | off_t data = 0, hole = 0; | |
d7f62751 | 2416 | int ret; |
4f11aa8a HR |
2417 | |
2418 | ret = fd_open(bs); | |
2419 | if (ret < 0) { | |
2420 | return ret; | |
2421 | } | |
2422 | ||
a290f085 EB |
2423 | if (!want_zero) { |
2424 | *pnum = bytes; | |
2425 | *map = offset; | |
2426 | *file = bs; | |
2427 | return BDRV_BLOCK_DATA | BDRV_BLOCK_OFFSET_VALID; | |
e6d7ec32 | 2428 | } |
4f11aa8a | 2429 | |
a290f085 | 2430 | ret = find_allocation(bs, offset, &data, &hole); |
d1f06fe6 MA |
2431 | if (ret == -ENXIO) { |
2432 | /* Trailing hole */ | |
a290f085 | 2433 | *pnum = bytes; |
d1f06fe6 MA |
2434 | ret = BDRV_BLOCK_ZERO; |
2435 | } else if (ret < 0) { | |
2436 | /* No info available, so pretend there are no holes */ | |
a290f085 | 2437 | *pnum = bytes; |
d1f06fe6 | 2438 | ret = BDRV_BLOCK_DATA; |
a290f085 EB |
2439 | } else if (data == offset) { |
2440 | /* On a data extent, compute bytes to the end of the extent, | |
f4a769ab | 2441 | * possibly including a partial sector at EOF. */ |
a290f085 | 2442 | *pnum = MIN(bytes, hole - offset); |
d1f06fe6 | 2443 | ret = BDRV_BLOCK_DATA; |
5500316d | 2444 | } else { |
a290f085 EB |
2445 | /* On a hole, compute bytes to the beginning of the next extent. */ |
2446 | assert(hole == offset); | |
2447 | *pnum = MIN(bytes, data - offset); | |
d1f06fe6 | 2448 | ret = BDRV_BLOCK_ZERO; |
5500316d | 2449 | } |
a290f085 | 2450 | *map = offset; |
02650acb | 2451 | *file = bs; |
a290f085 | 2452 | return ret | BDRV_BLOCK_OFFSET_VALID; |
5500316d PB |
2453 | } |
2454 | ||
31be8a2a SH |
2455 | #if defined(__linux__) |
2456 | /* Verify that the file is not in the page cache */ | |
2457 | static void check_cache_dropped(BlockDriverState *bs, Error **errp) | |
2458 | { | |
2459 | const size_t window_size = 128 * 1024 * 1024; | |
2460 | BDRVRawState *s = bs->opaque; | |
2461 | void *window = NULL; | |
2462 | size_t length = 0; | |
2463 | unsigned char *vec; | |
2464 | size_t page_size; | |
2465 | off_t offset; | |
2466 | off_t end; | |
2467 | ||
2468 | /* mincore(2) page status information requires 1 byte per page */ | |
2469 | page_size = sysconf(_SC_PAGESIZE); | |
2470 | vec = g_malloc(DIV_ROUND_UP(window_size, page_size)); | |
2471 | ||
2472 | end = raw_getlength(bs); | |
2473 | ||
2474 | for (offset = 0; offset < end; offset += window_size) { | |
2475 | void *new_window; | |
2476 | size_t new_length; | |
2477 | size_t vec_end; | |
2478 | size_t i; | |
2479 | int ret; | |
2480 | ||
2481 | /* Unmap previous window if size has changed */ | |
2482 | new_length = MIN(end - offset, window_size); | |
2483 | if (new_length != length) { | |
2484 | munmap(window, length); | |
2485 | window = NULL; | |
2486 | length = 0; | |
2487 | } | |
2488 | ||
2489 | new_window = mmap(window, new_length, PROT_NONE, MAP_PRIVATE, | |
2490 | s->fd, offset); | |
2491 | if (new_window == MAP_FAILED) { | |
2492 | error_setg_errno(errp, errno, "mmap failed"); | |
2493 | break; | |
2494 | } | |
2495 | ||
2496 | window = new_window; | |
2497 | length = new_length; | |
2498 | ||
2499 | ret = mincore(window, length, vec); | |
2500 | if (ret < 0) { | |
2501 | error_setg_errno(errp, errno, "mincore failed"); | |
2502 | break; | |
2503 | } | |
2504 | ||
2505 | vec_end = DIV_ROUND_UP(length, page_size); | |
2506 | for (i = 0; i < vec_end; i++) { | |
2507 | if (vec[i] & 0x1) { | |
2508 | error_setg(errp, "page cache still in use!"); | |
2509 | break; | |
2510 | } | |
2511 | } | |
2512 | } | |
2513 | ||
2514 | if (window) { | |
2515 | munmap(window, length); | |
2516 | } | |
2517 | ||
2518 | g_free(vec); | |
2519 | } | |
2520 | #endif /* __linux__ */ | |
2521 | ||
dd577a26 SH |
2522 | static void coroutine_fn raw_co_invalidate_cache(BlockDriverState *bs, |
2523 | Error **errp) | |
2524 | { | |
2525 | BDRVRawState *s = bs->opaque; | |
2526 | int ret; | |
2527 | ||
2528 | ret = fd_open(bs); | |
2529 | if (ret < 0) { | |
2530 | error_setg_errno(errp, -ret, "The file descriptor is not open"); | |
2531 | return; | |
2532 | } | |
2533 | ||
2534 | if (s->open_flags & O_DIRECT) { | |
2535 | return; /* No host kernel page cache */ | |
2536 | } | |
2537 | ||
2538 | #if defined(__linux__) | |
2539 | /* This sets the scene for the next syscall... */ | |
2540 | ret = bdrv_co_flush(bs); | |
2541 | if (ret < 0) { | |
2542 | error_setg_errno(errp, -ret, "flush failed"); | |
2543 | return; | |
2544 | } | |
2545 | ||
2546 | /* Linux does not invalidate pages that are dirty, locked, or mmapped by a | |
2547 | * process. These limitations are okay because we just fsynced the file, | |
2548 | * we don't use mmap, and the file should not be in use by other processes. | |
2549 | */ | |
2550 | ret = posix_fadvise(s->fd, 0, 0, POSIX_FADV_DONTNEED); | |
2551 | if (ret != 0) { /* the return value is a positive errno */ | |
2552 | error_setg_errno(errp, ret, "fadvise failed"); | |
2553 | return; | |
2554 | } | |
31be8a2a SH |
2555 | |
2556 | if (s->check_cache_dropped) { | |
2557 | check_cache_dropped(bs, errp); | |
2558 | } | |
dd577a26 SH |
2559 | #else /* __linux__ */ |
2560 | /* Do nothing. Live migration to a remote host with cache.direct=off is | |
2561 | * unsupported on other host operating systems. Cache consistency issues | |
2562 | * may occur but no error is reported here, partly because that's the | |
2563 | * historical behavior and partly because it's hard to differentiate valid | |
2564 | * configurations that should not cause errors. | |
2565 | */ | |
2566 | #endif /* !__linux__ */ | |
2567 | } | |
2568 | ||
33d70fb6 | 2569 | static coroutine_fn int |
46ee0f46 | 2570 | raw_do_pdiscard(BlockDriverState *bs, int64_t offset, int bytes, bool blkdev) |
dce512de | 2571 | { |
dce512de | 2572 | BDRVRawState *s = bs->opaque; |
46ee0f46 KW |
2573 | RawPosixAIOData acb; |
2574 | ||
2575 | acb = (RawPosixAIOData) { | |
2576 | .bs = bs, | |
2577 | .aio_fildes = s->fd, | |
2578 | .aio_type = QEMU_AIO_DISCARD, | |
2579 | .aio_offset = offset, | |
2580 | .aio_nbytes = bytes, | |
2581 | }; | |
dce512de | 2582 | |
46ee0f46 KW |
2583 | if (blkdev) { |
2584 | acb.aio_type |= QEMU_AIO_BLKDEV; | |
2585 | } | |
2586 | ||
2587 | return raw_thread_pool_submit(bs, handle_aiocb_discard, &acb); | |
2588 | } | |
2589 | ||
2590 | static coroutine_fn int | |
2591 | raw_co_pdiscard(BlockDriverState *bs, int64_t offset, int bytes) | |
2592 | { | |
2593 | return raw_do_pdiscard(bs, offset, bytes, false); | |
dce512de | 2594 | } |
0e7e1989 | 2595 | |
7154d8ae KW |
2596 | static int coroutine_fn |
2597 | raw_do_pwrite_zeroes(BlockDriverState *bs, int64_t offset, int bytes, | |
2598 | BdrvRequestFlags flags, bool blkdev) | |
260a82e5 PB |
2599 | { |
2600 | BDRVRawState *s = bs->opaque; | |
7154d8ae KW |
2601 | RawPosixAIOData acb; |
2602 | ThreadPoolFunc *handler; | |
2603 | ||
2604 | acb = (RawPosixAIOData) { | |
2605 | .bs = bs, | |
2606 | .aio_fildes = s->fd, | |
2607 | .aio_type = QEMU_AIO_WRITE_ZEROES, | |
2608 | .aio_offset = offset, | |
2609 | .aio_nbytes = bytes, | |
2610 | }; | |
2611 | ||
2612 | if (blkdev) { | |
2613 | acb.aio_type |= QEMU_AIO_BLKDEV; | |
2614 | } | |
260a82e5 | 2615 | |
34fa110e | 2616 | if (flags & BDRV_REQ_MAY_UNMAP) { |
7154d8ae KW |
2617 | acb.aio_type |= QEMU_AIO_DISCARD; |
2618 | handler = handle_aiocb_write_zeroes_unmap; | |
2619 | } else { | |
2620 | handler = handle_aiocb_write_zeroes; | |
260a82e5 | 2621 | } |
34fa110e | 2622 | |
7154d8ae KW |
2623 | return raw_thread_pool_submit(bs, handler, &acb); |
2624 | } | |
2625 | ||
2626 | static int coroutine_fn raw_co_pwrite_zeroes( | |
2627 | BlockDriverState *bs, int64_t offset, | |
2628 | int bytes, BdrvRequestFlags flags) | |
2629 | { | |
2630 | return raw_do_pwrite_zeroes(bs, offset, bytes, flags, false); | |
260a82e5 PB |
2631 | } |
2632 | ||
2633 | static int raw_get_info(BlockDriverState *bs, BlockDriverInfo *bdi) | |
2634 | { | |
2635 | BDRVRawState *s = bs->opaque; | |
2636 | ||
2637 | bdi->unallocated_blocks_are_zero = s->discard_zeroes; | |
260a82e5 PB |
2638 | return 0; |
2639 | } | |
2640 | ||
6f482f74 CL |
2641 | static QemuOptsList raw_create_opts = { |
2642 | .name = "raw-create-opts", | |
2643 | .head = QTAILQ_HEAD_INITIALIZER(raw_create_opts.head), | |
2644 | .desc = { | |
2645 | { | |
2646 | .name = BLOCK_OPT_SIZE, | |
2647 | .type = QEMU_OPT_SIZE, | |
2648 | .help = "Virtual disk size" | |
2649 | }, | |
4ab15590 CL |
2650 | { |
2651 | .name = BLOCK_OPT_NOCOW, | |
2652 | .type = QEMU_OPT_BOOL, | |
2653 | .help = "Turn off copy-on-write (valid only on btrfs)" | |
2654 | }, | |
06247428 HT |
2655 | { |
2656 | .name = BLOCK_OPT_PREALLOC, | |
2657 | .type = QEMU_OPT_STRING, | |
2658 | .help = "Preallocation mode (allowed values: off, falloc, full)" | |
2659 | }, | |
6f482f74 CL |
2660 | { /* end of list */ } |
2661 | } | |
0e7e1989 KW |
2662 | }; |
2663 | ||
244a5668 FZ |
2664 | static int raw_check_perm(BlockDriverState *bs, uint64_t perm, uint64_t shared, |
2665 | Error **errp) | |
2666 | { | |
2667 | return raw_handle_perm_lock(bs, RAW_PL_PREPARE, perm, shared, errp); | |
2668 | } | |
2669 | ||
2670 | static void raw_set_perm(BlockDriverState *bs, uint64_t perm, uint64_t shared) | |
2671 | { | |
2672 | BDRVRawState *s = bs->opaque; | |
2673 | raw_handle_perm_lock(bs, RAW_PL_COMMIT, perm, shared, NULL); | |
2674 | s->perm = perm; | |
2675 | s->shared_perm = shared; | |
2676 | } | |
2677 | ||
2678 | static void raw_abort_perm_update(BlockDriverState *bs) | |
2679 | { | |
2680 | raw_handle_perm_lock(bs, RAW_PL_ABORT, 0, 0, NULL); | |
2681 | } | |
2682 | ||
67b51fb9 VSO |
2683 | static int coroutine_fn raw_co_copy_range_from( |
2684 | BlockDriverState *bs, BdrvChild *src, uint64_t src_offset, | |
2685 | BdrvChild *dst, uint64_t dst_offset, uint64_t bytes, | |
2686 | BdrvRequestFlags read_flags, BdrvRequestFlags write_flags) | |
1efad060 | 2687 | { |
67b51fb9 VSO |
2688 | return bdrv_co_copy_range_to(src, src_offset, dst, dst_offset, bytes, |
2689 | read_flags, write_flags); | |
1efad060 FZ |
2690 | } |
2691 | ||
2692 | static int coroutine_fn raw_co_copy_range_to(BlockDriverState *bs, | |
67b51fb9 VSO |
2693 | BdrvChild *src, |
2694 | uint64_t src_offset, | |
2695 | BdrvChild *dst, | |
2696 | uint64_t dst_offset, | |
2697 | uint64_t bytes, | |
2698 | BdrvRequestFlags read_flags, | |
2699 | BdrvRequestFlags write_flags) | |
1efad060 | 2700 | { |
58a209c4 | 2701 | RawPosixAIOData acb; |
1efad060 FZ |
2702 | BDRVRawState *s = bs->opaque; |
2703 | BDRVRawState *src_s; | |
2704 | ||
2705 | assert(dst->bs == bs); | |
2706 | if (src->bs->drv->bdrv_co_copy_range_to != raw_co_copy_range_to) { | |
2707 | return -ENOTSUP; | |
2708 | } | |
2709 | ||
2710 | src_s = src->bs->opaque; | |
9f850f67 | 2711 | if (fd_open(src->bs) < 0 || fd_open(dst->bs) < 0) { |
1efad060 FZ |
2712 | return -EIO; |
2713 | } | |
58a209c4 KW |
2714 | |
2715 | acb = (RawPosixAIOData) { | |
2716 | .bs = bs, | |
2717 | .aio_type = QEMU_AIO_COPY_RANGE, | |
2718 | .aio_fildes = src_s->fd, | |
2719 | .aio_offset = src_offset, | |
2720 | .aio_nbytes = bytes, | |
2721 | .copy_range = { | |
2722 | .aio_fd2 = s->fd, | |
2723 | .aio_offset2 = dst_offset, | |
2724 | }, | |
2725 | }; | |
2726 | ||
2727 | return raw_thread_pool_submit(bs, handle_aiocb_copy_range, &acb); | |
1efad060 FZ |
2728 | } |
2729 | ||
5f535a94 | 2730 | BlockDriver bdrv_file = { |
84a12e66 CH |
2731 | .format_name = "file", |
2732 | .protocol_name = "file", | |
856ae5c3 | 2733 | .instance_size = sizeof(BDRVRawState), |
030be321 | 2734 | .bdrv_needs_filename = true, |
856ae5c3 | 2735 | .bdrv_probe = NULL, /* no probe for protocols */ |
078896a9 | 2736 | .bdrv_parse_filename = raw_parse_filename, |
66f82cee | 2737 | .bdrv_file_open = raw_open, |
eeb6b45d JC |
2738 | .bdrv_reopen_prepare = raw_reopen_prepare, |
2739 | .bdrv_reopen_commit = raw_reopen_commit, | |
2740 | .bdrv_reopen_abort = raw_reopen_abort, | |
856ae5c3 | 2741 | .bdrv_close = raw_close, |
927f11e1 | 2742 | .bdrv_co_create = raw_co_create, |
efc75e2a | 2743 | .bdrv_co_create_opts = raw_co_create_opts, |
3ac21627 | 2744 | .bdrv_has_zero_init = bdrv_has_zero_init_1, |
a290f085 | 2745 | .bdrv_co_block_status = raw_co_block_status, |
dd577a26 | 2746 | .bdrv_co_invalidate_cache = raw_co_invalidate_cache, |
2ffa76c2 | 2747 | .bdrv_co_pwrite_zeroes = raw_co_pwrite_zeroes, |
3b46e624 | 2748 | |
9d52aa3c KW |
2749 | .bdrv_co_preadv = raw_co_preadv, |
2750 | .bdrv_co_pwritev = raw_co_pwritev, | |
33d70fb6 KW |
2751 | .bdrv_co_flush_to_disk = raw_co_flush_to_disk, |
2752 | .bdrv_co_pdiscard = raw_co_pdiscard, | |
1efad060 FZ |
2753 | .bdrv_co_copy_range_from = raw_co_copy_range_from, |
2754 | .bdrv_co_copy_range_to = raw_co_copy_range_to, | |
c25f53b0 | 2755 | .bdrv_refresh_limits = raw_refresh_limits, |
1b3abdcc ML |
2756 | .bdrv_io_plug = raw_aio_plug, |
2757 | .bdrv_io_unplug = raw_aio_unplug, | |
ed6e2161 | 2758 | .bdrv_attach_aio_context = raw_aio_attach_aio_context, |
3c529d93 | 2759 | |
061ca8a3 | 2760 | .bdrv_co_truncate = raw_co_truncate, |
83f64091 | 2761 | .bdrv_getlength = raw_getlength, |
260a82e5 | 2762 | .bdrv_get_info = raw_get_info, |
4a1d5e1f FZ |
2763 | .bdrv_get_allocated_file_size |
2764 | = raw_get_allocated_file_size, | |
244a5668 FZ |
2765 | .bdrv_check_perm = raw_check_perm, |
2766 | .bdrv_set_perm = raw_set_perm, | |
2767 | .bdrv_abort_perm_update = raw_abort_perm_update, | |
6f482f74 | 2768 | .create_opts = &raw_create_opts, |
83f64091 FB |
2769 | }; |
2770 | ||
19cb3738 FB |
2771 | /***********************************************/ |
2772 | /* host device */ | |
2773 | ||
83affaa6 | 2774 | #if defined(__APPLE__) && defined(__MACH__) |
98caa5bc JA |
2775 | static kern_return_t GetBSDPath(io_iterator_t mediaIterator, char *bsdPath, |
2776 | CFIndex maxPathSize, int flags); | |
d0855f12 | 2777 | static char *FindEjectableOpticalMedia(io_iterator_t *mediaIterator) |
19cb3738 | 2778 | { |
d0855f12 | 2779 | kern_return_t kernResult = KERN_FAILURE; |
19cb3738 FB |
2780 | mach_port_t masterPort; |
2781 | CFMutableDictionaryRef classesToMatch; | |
d0855f12 JA |
2782 | const char *matching_array[] = {kIODVDMediaClass, kIOCDMediaClass}; |
2783 | char *mediaType = NULL; | |
19cb3738 FB |
2784 | |
2785 | kernResult = IOMasterPort( MACH_PORT_NULL, &masterPort ); | |
2786 | if ( KERN_SUCCESS != kernResult ) { | |
2787 | printf( "IOMasterPort returned %d\n", kernResult ); | |
2788 | } | |
3b46e624 | 2789 | |
d0855f12 JA |
2790 | int index; |
2791 | for (index = 0; index < ARRAY_SIZE(matching_array); index++) { | |
2792 | classesToMatch = IOServiceMatching(matching_array[index]); | |
2793 | if (classesToMatch == NULL) { | |
2794 | error_report("IOServiceMatching returned NULL for %s", | |
2795 | matching_array[index]); | |
2796 | continue; | |
2797 | } | |
2798 | CFDictionarySetValue(classesToMatch, CFSTR(kIOMediaEjectableKey), | |
2799 | kCFBooleanTrue); | |
2800 | kernResult = IOServiceGetMatchingServices(masterPort, classesToMatch, | |
2801 | mediaIterator); | |
2802 | if (kernResult != KERN_SUCCESS) { | |
2803 | error_report("Note: IOServiceGetMatchingServices returned %d", | |
2804 | kernResult); | |
2805 | continue; | |
2806 | } | |
3b46e624 | 2807 | |
d0855f12 JA |
2808 | /* If a match was found, leave the loop */ |
2809 | if (*mediaIterator != 0) { | |
4f7d28d7 | 2810 | trace_file_FindEjectableOpticalMedia(matching_array[index]); |
d0855f12 JA |
2811 | mediaType = g_strdup(matching_array[index]); |
2812 | break; | |
2813 | } | |
2814 | } | |
2815 | return mediaType; | |
19cb3738 FB |
2816 | } |
2817 | ||
98caa5bc JA |
2818 | kern_return_t GetBSDPath(io_iterator_t mediaIterator, char *bsdPath, |
2819 | CFIndex maxPathSize, int flags) | |
19cb3738 FB |
2820 | { |
2821 | io_object_t nextMedia; | |
2822 | kern_return_t kernResult = KERN_FAILURE; | |
2823 | *bsdPath = '\0'; | |
2824 | nextMedia = IOIteratorNext( mediaIterator ); | |
2825 | if ( nextMedia ) | |
2826 | { | |
2827 | CFTypeRef bsdPathAsCFString; | |
2828 | bsdPathAsCFString = IORegistryEntryCreateCFProperty( nextMedia, CFSTR( kIOBSDNameKey ), kCFAllocatorDefault, 0 ); | |
2829 | if ( bsdPathAsCFString ) { | |
2830 | size_t devPathLength; | |
2831 | strcpy( bsdPath, _PATH_DEV ); | |
98caa5bc JA |
2832 | if (flags & BDRV_O_NOCACHE) { |
2833 | strcat(bsdPath, "r"); | |
2834 | } | |
19cb3738 FB |
2835 | devPathLength = strlen( bsdPath ); |
2836 | if ( CFStringGetCString( bsdPathAsCFString, bsdPath + devPathLength, maxPathSize - devPathLength, kCFStringEncodingASCII ) ) { | |
2837 | kernResult = KERN_SUCCESS; | |
2838 | } | |
2839 | CFRelease( bsdPathAsCFString ); | |
2840 | } | |
2841 | IOObjectRelease( nextMedia ); | |
2842 | } | |
3b46e624 | 2843 | |
19cb3738 FB |
2844 | return kernResult; |
2845 | } | |
2846 | ||
d0855f12 JA |
2847 | /* Sets up a real cdrom for use in QEMU */ |
2848 | static bool setup_cdrom(char *bsd_path, Error **errp) | |
2849 | { | |
2850 | int index, num_of_test_partitions = 2, fd; | |
2851 | char test_partition[MAXPATHLEN]; | |
2852 | bool partition_found = false; | |
2853 | ||
2854 | /* look for a working partition */ | |
2855 | for (index = 0; index < num_of_test_partitions; index++) { | |
2856 | snprintf(test_partition, sizeof(test_partition), "%ss%d", bsd_path, | |
2857 | index); | |
2858 | fd = qemu_open(test_partition, O_RDONLY | O_BINARY | O_LARGEFILE); | |
2859 | if (fd >= 0) { | |
2860 | partition_found = true; | |
2861 | qemu_close(fd); | |
2862 | break; | |
2863 | } | |
2864 | } | |
2865 | ||
2866 | /* if a working partition on the device was not found */ | |
2867 | if (partition_found == false) { | |
2868 | error_setg(errp, "Failed to find a working partition on disc"); | |
2869 | } else { | |
4f7d28d7 | 2870 | trace_file_setup_cdrom(test_partition); |
d0855f12 JA |
2871 | pstrcpy(bsd_path, MAXPATHLEN, test_partition); |
2872 | } | |
2873 | return partition_found; | |
2874 | } | |
2875 | ||
2876 | /* Prints directions on mounting and unmounting a device */ | |
2877 | static void print_unmounting_directions(const char *file_name) | |
2878 | { | |
2879 | error_report("If device %s is mounted on the desktop, unmount" | |
2880 | " it first before using it in QEMU", file_name); | |
2881 | error_report("Command to unmount device: diskutil unmountDisk %s", | |
2882 | file_name); | |
2883 | error_report("Command to mount device: diskutil mountDisk %s", file_name); | |
2884 | } | |
2885 | ||
2886 | #endif /* defined(__APPLE__) && defined(__MACH__) */ | |
19cb3738 | 2887 | |
508c7cb3 CH |
2888 | static int hdev_probe_device(const char *filename) |
2889 | { | |
2890 | struct stat st; | |
2891 | ||
2892 | /* allow a dedicated CD-ROM driver to match with a higher priority */ | |
2893 | if (strstart(filename, "/dev/cdrom", NULL)) | |
2894 | return 50; | |
2895 | ||
2896 | if (stat(filename, &st) >= 0 && | |
2897 | (S_ISCHR(st.st_mode) || S_ISBLK(st.st_mode))) { | |
2898 | return 100; | |
2899 | } | |
2900 | ||
2901 | return 0; | |
2902 | } | |
2903 | ||
da888d37 SH |
2904 | static int check_hdev_writable(BDRVRawState *s) |
2905 | { | |
2906 | #if defined(BLKROGET) | |
2907 | /* Linux block devices can be configured "read-only" using blockdev(8). | |
2908 | * This is independent of device node permissions and therefore open(2) | |
2909 | * with O_RDWR succeeds. Actual writes fail with EPERM. | |
2910 | * | |
2911 | * bdrv_open() is supposed to fail if the disk is read-only. Explicitly | |
2912 | * check for read-only block devices so that Linux block devices behave | |
2913 | * properly. | |
2914 | */ | |
2915 | struct stat st; | |
2916 | int readonly = 0; | |
2917 | ||
2918 | if (fstat(s->fd, &st)) { | |
2919 | return -errno; | |
2920 | } | |
2921 | ||
2922 | if (!S_ISBLK(st.st_mode)) { | |
2923 | return 0; | |
2924 | } | |
2925 | ||
2926 | if (ioctl(s->fd, BLKROGET, &readonly) < 0) { | |
2927 | return -errno; | |
2928 | } | |
2929 | ||
2930 | if (readonly) { | |
2931 | return -EACCES; | |
2932 | } | |
2933 | #endif /* defined(BLKROGET) */ | |
2934 | return 0; | |
2935 | } | |
2936 | ||
7af803d4 HR |
2937 | static void hdev_parse_filename(const char *filename, QDict *options, |
2938 | Error **errp) | |
2939 | { | |
03c320d8 | 2940 | bdrv_parse_filename_strip_prefix(filename, "host_device:", options); |
7af803d4 HR |
2941 | } |
2942 | ||
3307ed7b DA |
2943 | static bool hdev_is_sg(BlockDriverState *bs) |
2944 | { | |
2945 | ||
2946 | #if defined(__linux__) | |
2947 | ||
0d4377b3 | 2948 | BDRVRawState *s = bs->opaque; |
3307ed7b DA |
2949 | struct stat st; |
2950 | struct sg_scsi_id scsiid; | |
2951 | int sg_version; | |
0d4377b3 KW |
2952 | int ret; |
2953 | ||
2954 | if (stat(bs->filename, &st) < 0 || !S_ISCHR(st.st_mode)) { | |
2955 | return false; | |
2956 | } | |
3307ed7b | 2957 | |
0d4377b3 KW |
2958 | ret = ioctl(s->fd, SG_GET_VERSION_NUM, &sg_version); |
2959 | if (ret < 0) { | |
2960 | return false; | |
2961 | } | |
2962 | ||
2963 | ret = ioctl(s->fd, SG_GET_SCSI_ID, &scsiid); | |
2964 | if (ret >= 0) { | |
4f7d28d7 | 2965 | trace_file_hdev_is_sg(scsiid.scsi_type, sg_version); |
3307ed7b DA |
2966 | return true; |
2967 | } | |
2968 | ||
2969 | #endif | |
2970 | ||
2971 | return false; | |
2972 | } | |
2973 | ||
015a1036 HR |
2974 | static int hdev_open(BlockDriverState *bs, QDict *options, int flags, |
2975 | Error **errp) | |
19cb3738 FB |
2976 | { |
2977 | BDRVRawState *s = bs->opaque; | |
e428e439 | 2978 | Error *local_err = NULL; |
da888d37 | 2979 | int ret; |
a76bab49 | 2980 | |
83affaa6 | 2981 | #if defined(__APPLE__) && defined(__MACH__) |
129c7d1c MA |
2982 | /* |
2983 | * Caution: while qdict_get_str() is fine, getting non-string types | |
2984 | * would require more care. When @options come from -blockdev or | |
2985 | * blockdev_add, its members are typed according to the QAPI | |
2986 | * schema, but when they come from -drive, they're all QString. | |
2987 | */ | |
3307ed7b | 2988 | const char *filename = qdict_get_str(options, "filename"); |
d0855f12 JA |
2989 | char bsd_path[MAXPATHLEN] = ""; |
2990 | bool error_occurred = false; | |
2991 | ||
2992 | /* If using a real cdrom */ | |
2993 | if (strcmp(filename, "/dev/cdrom") == 0) { | |
2994 | char *mediaType = NULL; | |
2995 | kern_return_t ret_val; | |
2996 | io_iterator_t mediaIterator = 0; | |
2997 | ||
2998 | mediaType = FindEjectableOpticalMedia(&mediaIterator); | |
2999 | if (mediaType == NULL) { | |
3000 | error_setg(errp, "Please make sure your CD/DVD is in the optical" | |
3001 | " drive"); | |
3002 | error_occurred = true; | |
3003 | goto hdev_open_Mac_error; | |
3004 | } | |
3307ed7b | 3005 | |
d0855f12 JA |
3006 | ret_val = GetBSDPath(mediaIterator, bsd_path, sizeof(bsd_path), flags); |
3007 | if (ret_val != KERN_SUCCESS) { | |
3008 | error_setg(errp, "Could not get BSD path for optical drive"); | |
3009 | error_occurred = true; | |
3010 | goto hdev_open_Mac_error; | |
3011 | } | |
3012 | ||
3013 | /* If a real optical drive was not found */ | |
3014 | if (bsd_path[0] == '\0') { | |
3015 | error_setg(errp, "Failed to obtain bsd path for optical drive"); | |
3016 | error_occurred = true; | |
3017 | goto hdev_open_Mac_error; | |
3018 | } | |
3019 | ||
3020 | /* If using a cdrom disc and finding a partition on the disc failed */ | |
3021 | if (strncmp(mediaType, kIOCDMediaClass, 9) == 0 && | |
3022 | setup_cdrom(bsd_path, errp) == false) { | |
3023 | print_unmounting_directions(bsd_path); | |
3024 | error_occurred = true; | |
3025 | goto hdev_open_Mac_error; | |
19cb3738 | 3026 | } |
3b46e624 | 3027 | |
46f5ac20 | 3028 | qdict_put_str(options, "filename", bsd_path); |
d0855f12 JA |
3029 | |
3030 | hdev_open_Mac_error: | |
3031 | g_free(mediaType); | |
3032 | if (mediaIterator) { | |
3033 | IOObjectRelease(mediaIterator); | |
3034 | } | |
3035 | if (error_occurred) { | |
3036 | return -ENOENT; | |
3037 | } | |
19cb3738 | 3038 | } |
d0855f12 | 3039 | #endif /* defined(__APPLE__) && defined(__MACH__) */ |
19cb3738 FB |
3040 | |
3041 | s->type = FTYPE_FILE; | |
90babde0 | 3042 | |
230ff739 | 3043 | ret = raw_open_common(bs, options, flags, 0, true, &local_err); |
da888d37 | 3044 | if (ret < 0) { |
621ff94d | 3045 | error_propagate(errp, local_err); |
d0855f12 JA |
3046 | #if defined(__APPLE__) && defined(__MACH__) |
3047 | if (*bsd_path) { | |
3048 | filename = bsd_path; | |
3049 | } | |
3050 | /* if a physical device experienced an error while being opened */ | |
3051 | if (strncmp(filename, "/dev/", 5) == 0) { | |
3052 | print_unmounting_directions(filename); | |
3053 | } | |
3054 | #endif /* defined(__APPLE__) && defined(__MACH__) */ | |
da888d37 SH |
3055 | return ret; |
3056 | } | |
3057 | ||
3307ed7b DA |
3058 | /* Since this does ioctl the device must be already opened */ |
3059 | bs->sg = hdev_is_sg(bs); | |
3060 | ||
da888d37 SH |
3061 | if (flags & BDRV_O_RDWR) { |
3062 | ret = check_hdev_writable(s); | |
3063 | if (ret < 0) { | |
3064 | raw_close(bs); | |
e428e439 | 3065 | error_setg_errno(errp, -ret, "The device is not writable"); |
da888d37 SH |
3066 | return ret; |
3067 | } | |
3068 | } | |
3069 | ||
3070 | return ret; | |
19cb3738 FB |
3071 | } |
3072 | ||
03ff3ca3 | 3073 | #if defined(__linux__) |
2f3a7ab3 KW |
3074 | static int coroutine_fn |
3075 | hdev_co_ioctl(BlockDriverState *bs, unsigned long int req, void *buf) | |
221f715d | 3076 | { |
f141eafe | 3077 | BDRVRawState *s = bs->opaque; |
03425671 | 3078 | RawPosixAIOData acb; |
2f3a7ab3 | 3079 | int ret; |
221f715d | 3080 | |
2f3a7ab3 KW |
3081 | ret = fd_open(bs); |
3082 | if (ret < 0) { | |
3083 | return ret; | |
3084 | } | |
c208e8c2 | 3085 | |
7c9e5276 PB |
3086 | if (req == SG_IO && s->pr_mgr) { |
3087 | struct sg_io_hdr *io_hdr = buf; | |
3088 | if (io_hdr->cmdp[0] == PERSISTENT_RESERVE_OUT || | |
3089 | io_hdr->cmdp[0] == PERSISTENT_RESERVE_IN) { | |
3090 | return pr_manager_execute(s->pr_mgr, bdrv_get_aio_context(bs), | |
2f3a7ab3 | 3091 | s->fd, io_hdr); |
7c9e5276 PB |
3092 | } |
3093 | } | |
3094 | ||
03425671 KW |
3095 | acb = (RawPosixAIOData) { |
3096 | .bs = bs, | |
3097 | .aio_type = QEMU_AIO_IOCTL, | |
3098 | .aio_fildes = s->fd, | |
3099 | .aio_offset = 0, | |
3100 | .ioctl = { | |
3101 | .buf = buf, | |
3102 | .cmd = req, | |
3103 | }, | |
3104 | }; | |
3105 | ||
3106 | return raw_thread_pool_submit(bs, handle_aiocb_ioctl, &acb); | |
221f715d | 3107 | } |
f709623b | 3108 | #endif /* linux */ |
221f715d | 3109 | |
9f23011a BS |
3110 | static int fd_open(BlockDriverState *bs) |
3111 | { | |
3112 | BDRVRawState *s = bs->opaque; | |
3113 | ||
3114 | /* this is just to ensure s->fd is sane (its called by io ops) */ | |
3115 | if (s->fd >= 0) | |
3116 | return 0; | |
3117 | return -EIO; | |
3118 | } | |
04eeb8b6 | 3119 | |
33d70fb6 KW |
3120 | static coroutine_fn int |
3121 | hdev_co_pdiscard(BlockDriverState *bs, int64_t offset, int bytes) | |
c36dd8a0 | 3122 | { |
33d70fb6 | 3123 | int ret; |
c36dd8a0 | 3124 | |
33d70fb6 KW |
3125 | ret = fd_open(bs); |
3126 | if (ret < 0) { | |
3127 | return ret; | |
c36dd8a0 | 3128 | } |
46ee0f46 | 3129 | return raw_do_pdiscard(bs, offset, bytes, true); |
c36dd8a0 AF |
3130 | } |
3131 | ||
2ffa76c2 | 3132 | static coroutine_fn int hdev_co_pwrite_zeroes(BlockDriverState *bs, |
f5a5ca79 | 3133 | int64_t offset, int bytes, BdrvRequestFlags flags) |
d0b4503e | 3134 | { |
d0b4503e PB |
3135 | int rc; |
3136 | ||
3137 | rc = fd_open(bs); | |
3138 | if (rc < 0) { | |
3139 | return rc; | |
3140 | } | |
34fa110e | 3141 | |
7154d8ae | 3142 | return raw_do_pwrite_zeroes(bs, offset, bytes, flags, true); |
d0b4503e PB |
3143 | } |
3144 | ||
efc75e2a SH |
3145 | static int coroutine_fn hdev_co_create_opts(const char *filename, QemuOpts *opts, |
3146 | Error **errp) | |
93c65b47 AL |
3147 | { |
3148 | int fd; | |
3149 | int ret = 0; | |
3150 | struct stat stat_buf; | |
0e7e1989 | 3151 | int64_t total_size = 0; |
cc28c6aa HR |
3152 | bool has_prefix; |
3153 | ||
f709623b HR |
3154 | /* This function is used by both protocol block drivers and therefore either |
3155 | * of these prefixes may be given. | |
cc28c6aa HR |
3156 | * The return value has to be stored somewhere, otherwise this is an error |
3157 | * due to -Werror=unused-value. */ | |
3158 | has_prefix = | |
3159 | strstart(filename, "host_device:", &filename) || | |
f709623b | 3160 | strstart(filename, "host_cdrom:" , &filename); |
cc28c6aa HR |
3161 | |
3162 | (void)has_prefix; | |
93c65b47 | 3163 | |
db0754df | 3164 | ret = raw_normalize_devicepath(&filename, errp); |
bdd03cdf | 3165 | if (ret < 0) { |
bdd03cdf HR |
3166 | return ret; |
3167 | } | |
3168 | ||
0e7e1989 | 3169 | /* Read out options */ |
180e9526 HT |
3170 | total_size = ROUND_UP(qemu_opt_get_size_del(opts, BLOCK_OPT_SIZE, 0), |
3171 | BDRV_SECTOR_SIZE); | |
93c65b47 | 3172 | |
6165f4d8 | 3173 | fd = qemu_open(filename, O_WRONLY | O_BINARY); |
e428e439 HR |
3174 | if (fd < 0) { |
3175 | ret = -errno; | |
3176 | error_setg_errno(errp, -ret, "Could not open device"); | |
3177 | return ret; | |
3178 | } | |
93c65b47 | 3179 | |
e428e439 | 3180 | if (fstat(fd, &stat_buf) < 0) { |
57e69b7d | 3181 | ret = -errno; |
e428e439 HR |
3182 | error_setg_errno(errp, -ret, "Could not stat device"); |
3183 | } else if (!S_ISBLK(stat_buf.st_mode) && !S_ISCHR(stat_buf.st_mode)) { | |
3184 | error_setg(errp, | |
3185 | "The given file is neither a block nor a character device"); | |
57e69b7d | 3186 | ret = -ENODEV; |
180e9526 | 3187 | } else if (lseek(fd, 0, SEEK_END) < total_size) { |
e428e439 | 3188 | error_setg(errp, "Device is too small"); |
93c65b47 | 3189 | ret = -ENOSPC; |
e428e439 | 3190 | } |
93c65b47 | 3191 | |
97ec9117 FZ |
3192 | if (!ret && total_size) { |
3193 | uint8_t buf[BDRV_SECTOR_SIZE] = { 0 }; | |
3194 | int64_t zero_size = MIN(BDRV_SECTOR_SIZE, total_size); | |
3195 | if (lseek(fd, 0, SEEK_SET) == -1) { | |
3196 | ret = -errno; | |
3197 | } else { | |
3198 | ret = qemu_write_full(fd, buf, zero_size); | |
3199 | ret = ret == zero_size ? 0 : -errno; | |
3200 | } | |
3201 | } | |
2e1e79da | 3202 | qemu_close(fd); |
93c65b47 AL |
3203 | return ret; |
3204 | } | |
3205 | ||
5efa9d5a | 3206 | static BlockDriver bdrv_host_device = { |
0b4ce02e | 3207 | .format_name = "host_device", |
84a12e66 | 3208 | .protocol_name = "host_device", |
0b4ce02e | 3209 | .instance_size = sizeof(BDRVRawState), |
030be321 | 3210 | .bdrv_needs_filename = true, |
0b4ce02e | 3211 | .bdrv_probe_device = hdev_probe_device, |
7af803d4 | 3212 | .bdrv_parse_filename = hdev_parse_filename, |
66f82cee | 3213 | .bdrv_file_open = hdev_open, |
0b4ce02e | 3214 | .bdrv_close = raw_close, |
1bc6b705 JC |
3215 | .bdrv_reopen_prepare = raw_reopen_prepare, |
3216 | .bdrv_reopen_commit = raw_reopen_commit, | |
3217 | .bdrv_reopen_abort = raw_reopen_abort, | |
efc75e2a | 3218 | .bdrv_co_create_opts = hdev_co_create_opts, |
6f482f74 | 3219 | .create_opts = &raw_create_opts, |
dd577a26 | 3220 | .bdrv_co_invalidate_cache = raw_co_invalidate_cache, |
2ffa76c2 | 3221 | .bdrv_co_pwrite_zeroes = hdev_co_pwrite_zeroes, |
3b46e624 | 3222 | |
9d52aa3c KW |
3223 | .bdrv_co_preadv = raw_co_preadv, |
3224 | .bdrv_co_pwritev = raw_co_pwritev, | |
33d70fb6 KW |
3225 | .bdrv_co_flush_to_disk = raw_co_flush_to_disk, |
3226 | .bdrv_co_pdiscard = hdev_co_pdiscard, | |
1efad060 FZ |
3227 | .bdrv_co_copy_range_from = raw_co_copy_range_from, |
3228 | .bdrv_co_copy_range_to = raw_co_copy_range_to, | |
c25f53b0 | 3229 | .bdrv_refresh_limits = raw_refresh_limits, |
1b3abdcc ML |
3230 | .bdrv_io_plug = raw_aio_plug, |
3231 | .bdrv_io_unplug = raw_aio_unplug, | |
042b757c | 3232 | .bdrv_attach_aio_context = raw_aio_attach_aio_context, |
3c529d93 | 3233 | |
061ca8a3 | 3234 | .bdrv_co_truncate = raw_co_truncate, |
e60f469c | 3235 | .bdrv_getlength = raw_getlength, |
260a82e5 | 3236 | .bdrv_get_info = raw_get_info, |
4a1d5e1f FZ |
3237 | .bdrv_get_allocated_file_size |
3238 | = raw_get_allocated_file_size, | |
244a5668 FZ |
3239 | .bdrv_check_perm = raw_check_perm, |
3240 | .bdrv_set_perm = raw_set_perm, | |
3241 | .bdrv_abort_perm_update = raw_abort_perm_update, | |
1a9335e4 ET |
3242 | .bdrv_probe_blocksizes = hdev_probe_blocksizes, |
3243 | .bdrv_probe_geometry = hdev_probe_geometry, | |
19cb3738 | 3244 | |
f3a5d3f8 | 3245 | /* generic scsi device */ |
63ec93db | 3246 | #ifdef __linux__ |
2f3a7ab3 | 3247 | .bdrv_co_ioctl = hdev_co_ioctl, |
63ec93db | 3248 | #endif |
f3a5d3f8 CH |
3249 | }; |
3250 | ||
18fa1c42 HR |
3251 | #if defined(__linux__) || defined(__FreeBSD__) || defined(__FreeBSD_kernel__) |
3252 | static void cdrom_parse_filename(const char *filename, QDict *options, | |
3253 | Error **errp) | |
3254 | { | |
03c320d8 | 3255 | bdrv_parse_filename_strip_prefix(filename, "host_cdrom:", options); |
18fa1c42 HR |
3256 | } |
3257 | #endif | |
3258 | ||
3259 | #ifdef __linux__ | |
015a1036 HR |
3260 | static int cdrom_open(BlockDriverState *bs, QDict *options, int flags, |
3261 | Error **errp) | |
f3a5d3f8 CH |
3262 | { |
3263 | BDRVRawState *s = bs->opaque; | |
3264 | ||
f3a5d3f8 CH |
3265 | s->type = FTYPE_CD; |
3266 | ||
19a3da7f | 3267 | /* open will not fail even if no CD is inserted, so add O_NONBLOCK */ |
230ff739 | 3268 | return raw_open_common(bs, options, flags, O_NONBLOCK, true, errp); |
f3a5d3f8 CH |
3269 | } |
3270 | ||
508c7cb3 CH |
3271 | static int cdrom_probe_device(const char *filename) |
3272 | { | |
3baf720e CR |
3273 | int fd, ret; |
3274 | int prio = 0; | |
343f8568 | 3275 | struct stat st; |
3baf720e | 3276 | |
6165f4d8 | 3277 | fd = qemu_open(filename, O_RDONLY | O_NONBLOCK); |
3baf720e CR |
3278 | if (fd < 0) { |
3279 | goto out; | |
3280 | } | |
343f8568 JS |
3281 | ret = fstat(fd, &st); |
3282 | if (ret == -1 || !S_ISBLK(st.st_mode)) { | |
3283 | goto outc; | |
3284 | } | |
3baf720e CR |
3285 | |
3286 | /* Attempt to detect via a CDROM specific ioctl */ | |
3287 | ret = ioctl(fd, CDROM_DRIVE_STATUS, CDSL_CURRENT); | |
3288 | if (ret >= 0) | |
3289 | prio = 100; | |
3290 | ||
343f8568 | 3291 | outc: |
2e1e79da | 3292 | qemu_close(fd); |
3baf720e CR |
3293 | out: |
3294 | return prio; | |
508c7cb3 CH |
3295 | } |
3296 | ||
e031f750 | 3297 | static bool cdrom_is_inserted(BlockDriverState *bs) |
f3a5d3f8 CH |
3298 | { |
3299 | BDRVRawState *s = bs->opaque; | |
3300 | int ret; | |
3301 | ||
3302 | ret = ioctl(s->fd, CDROM_DRIVE_STATUS, CDSL_CURRENT); | |
e031f750 | 3303 | return ret == CDS_DISC_OK; |
f3a5d3f8 CH |
3304 | } |
3305 | ||
f36f3949 | 3306 | static void cdrom_eject(BlockDriverState *bs, bool eject_flag) |
f3a5d3f8 CH |
3307 | { |
3308 | BDRVRawState *s = bs->opaque; | |
3309 | ||
3310 | if (eject_flag) { | |
3311 | if (ioctl(s->fd, CDROMEJECT, NULL) < 0) | |
3312 | perror("CDROMEJECT"); | |
3313 | } else { | |
3314 | if (ioctl(s->fd, CDROMCLOSETRAY, NULL) < 0) | |
3315 | perror("CDROMEJECT"); | |
3316 | } | |
f3a5d3f8 CH |
3317 | } |
3318 | ||
025e849a | 3319 | static void cdrom_lock_medium(BlockDriverState *bs, bool locked) |
f3a5d3f8 CH |
3320 | { |
3321 | BDRVRawState *s = bs->opaque; | |
3322 | ||
3323 | if (ioctl(s->fd, CDROM_LOCKDOOR, locked) < 0) { | |
3324 | /* | |
3325 | * Note: an error can happen if the distribution automatically | |
3326 | * mounts the CD-ROM | |
3327 | */ | |
3328 | /* perror("CDROM_LOCKDOOR"); */ | |
3329 | } | |
f3a5d3f8 CH |
3330 | } |
3331 | ||
3332 | static BlockDriver bdrv_host_cdrom = { | |
3333 | .format_name = "host_cdrom", | |
84a12e66 | 3334 | .protocol_name = "host_cdrom", |
f3a5d3f8 | 3335 | .instance_size = sizeof(BDRVRawState), |
030be321 | 3336 | .bdrv_needs_filename = true, |
508c7cb3 | 3337 | .bdrv_probe_device = cdrom_probe_device, |
18fa1c42 | 3338 | .bdrv_parse_filename = cdrom_parse_filename, |
66f82cee | 3339 | .bdrv_file_open = cdrom_open, |
f3a5d3f8 | 3340 | .bdrv_close = raw_close, |
1bc6b705 JC |
3341 | .bdrv_reopen_prepare = raw_reopen_prepare, |
3342 | .bdrv_reopen_commit = raw_reopen_commit, | |
3343 | .bdrv_reopen_abort = raw_reopen_abort, | |
efc75e2a | 3344 | .bdrv_co_create_opts = hdev_co_create_opts, |
6f482f74 | 3345 | .create_opts = &raw_create_opts, |
dd577a26 | 3346 | .bdrv_co_invalidate_cache = raw_co_invalidate_cache, |
f3a5d3f8 | 3347 | |
9d52aa3c KW |
3348 | |
3349 | .bdrv_co_preadv = raw_co_preadv, | |
3350 | .bdrv_co_pwritev = raw_co_pwritev, | |
33d70fb6 | 3351 | .bdrv_co_flush_to_disk = raw_co_flush_to_disk, |
c25f53b0 | 3352 | .bdrv_refresh_limits = raw_refresh_limits, |
1b3abdcc ML |
3353 | .bdrv_io_plug = raw_aio_plug, |
3354 | .bdrv_io_unplug = raw_aio_unplug, | |
042b757c | 3355 | .bdrv_attach_aio_context = raw_aio_attach_aio_context, |
f3a5d3f8 | 3356 | |
061ca8a3 | 3357 | .bdrv_co_truncate = raw_co_truncate, |
b94a2610 KW |
3358 | .bdrv_getlength = raw_getlength, |
3359 | .has_variable_length = true, | |
4a1d5e1f FZ |
3360 | .bdrv_get_allocated_file_size |
3361 | = raw_get_allocated_file_size, | |
f3a5d3f8 CH |
3362 | |
3363 | /* removable device support */ | |
3364 | .bdrv_is_inserted = cdrom_is_inserted, | |
3365 | .bdrv_eject = cdrom_eject, | |
025e849a | 3366 | .bdrv_lock_medium = cdrom_lock_medium, |
f3a5d3f8 CH |
3367 | |
3368 | /* generic scsi device */ | |
2f3a7ab3 | 3369 | .bdrv_co_ioctl = hdev_co_ioctl, |
f3a5d3f8 CH |
3370 | }; |
3371 | #endif /* __linux__ */ | |
3372 | ||
a167ba50 | 3373 | #if defined (__FreeBSD__) || defined(__FreeBSD_kernel__) |
511018e4 AT |
3374 | static int cdrom_open(BlockDriverState *bs, QDict *options, int flags, |
3375 | Error **errp) | |
f3a5d3f8 CH |
3376 | { |
3377 | BDRVRawState *s = bs->opaque; | |
e428e439 | 3378 | Error *local_err = NULL; |
f3a5d3f8 CH |
3379 | int ret; |
3380 | ||
3381 | s->type = FTYPE_CD; | |
3382 | ||
230ff739 | 3383 | ret = raw_open_common(bs, options, flags, 0, true, &local_err); |
e428e439 | 3384 | if (ret) { |
621ff94d | 3385 | error_propagate(errp, local_err); |
f3a5d3f8 | 3386 | return ret; |
e428e439 | 3387 | } |
f3a5d3f8 | 3388 | |
9b2260cb | 3389 | /* make sure the door isn't locked at this time */ |
f3a5d3f8 CH |
3390 | ioctl(s->fd, CDIOCALLOW); |
3391 | return 0; | |
3392 | } | |
3393 | ||
508c7cb3 CH |
3394 | static int cdrom_probe_device(const char *filename) |
3395 | { | |
3396 | if (strstart(filename, "/dev/cd", NULL) || | |
3397 | strstart(filename, "/dev/acd", NULL)) | |
3398 | return 100; | |
3399 | return 0; | |
3400 | } | |
3401 | ||
f3a5d3f8 CH |
3402 | static int cdrom_reopen(BlockDriverState *bs) |
3403 | { | |
3404 | BDRVRawState *s = bs->opaque; | |
3405 | int fd; | |
3406 | ||
3407 | /* | |
3408 | * Force reread of possibly changed/newly loaded disc, | |
3409 | * FreeBSD seems to not notice sometimes... | |
3410 | */ | |
3411 | if (s->fd >= 0) | |
2e1e79da | 3412 | qemu_close(s->fd); |
6165f4d8 | 3413 | fd = qemu_open(bs->filename, s->open_flags, 0644); |
f3a5d3f8 CH |
3414 | if (fd < 0) { |
3415 | s->fd = -1; | |
3416 | return -EIO; | |
3417 | } | |
3418 | s->fd = fd; | |
3419 | ||
9b2260cb | 3420 | /* make sure the door isn't locked at this time */ |
f3a5d3f8 CH |
3421 | ioctl(s->fd, CDIOCALLOW); |
3422 | return 0; | |
3423 | } | |
3424 | ||
e031f750 | 3425 | static bool cdrom_is_inserted(BlockDriverState *bs) |
f3a5d3f8 CH |
3426 | { |
3427 | return raw_getlength(bs) > 0; | |
3428 | } | |
3429 | ||
f36f3949 | 3430 | static void cdrom_eject(BlockDriverState *bs, bool eject_flag) |
f3a5d3f8 CH |
3431 | { |
3432 | BDRVRawState *s = bs->opaque; | |
3433 | ||
3434 | if (s->fd < 0) | |
822e1cd1 | 3435 | return; |
f3a5d3f8 CH |
3436 | |
3437 | (void) ioctl(s->fd, CDIOCALLOW); | |
3438 | ||
3439 | if (eject_flag) { | |
3440 | if (ioctl(s->fd, CDIOCEJECT) < 0) | |
3441 | perror("CDIOCEJECT"); | |
3442 | } else { | |
3443 | if (ioctl(s->fd, CDIOCCLOSE) < 0) | |
3444 | perror("CDIOCCLOSE"); | |
3445 | } | |
3446 | ||
822e1cd1 | 3447 | cdrom_reopen(bs); |
f3a5d3f8 CH |
3448 | } |
3449 | ||
025e849a | 3450 | static void cdrom_lock_medium(BlockDriverState *bs, bool locked) |
f3a5d3f8 CH |
3451 | { |
3452 | BDRVRawState *s = bs->opaque; | |
3453 | ||
3454 | if (s->fd < 0) | |
7bf37fed | 3455 | return; |
f3a5d3f8 CH |
3456 | if (ioctl(s->fd, (locked ? CDIOCPREVENT : CDIOCALLOW)) < 0) { |
3457 | /* | |
3458 | * Note: an error can happen if the distribution automatically | |
3459 | * mounts the CD-ROM | |
3460 | */ | |
3461 | /* perror("CDROM_LOCKDOOR"); */ | |
3462 | } | |
f3a5d3f8 CH |
3463 | } |
3464 | ||
3465 | static BlockDriver bdrv_host_cdrom = { | |
3466 | .format_name = "host_cdrom", | |
84a12e66 | 3467 | .protocol_name = "host_cdrom", |
f3a5d3f8 | 3468 | .instance_size = sizeof(BDRVRawState), |
030be321 | 3469 | .bdrv_needs_filename = true, |
508c7cb3 | 3470 | .bdrv_probe_device = cdrom_probe_device, |
18fa1c42 | 3471 | .bdrv_parse_filename = cdrom_parse_filename, |
66f82cee | 3472 | .bdrv_file_open = cdrom_open, |
f3a5d3f8 | 3473 | .bdrv_close = raw_close, |
1bc6b705 JC |
3474 | .bdrv_reopen_prepare = raw_reopen_prepare, |
3475 | .bdrv_reopen_commit = raw_reopen_commit, | |
3476 | .bdrv_reopen_abort = raw_reopen_abort, | |
efc75e2a | 3477 | .bdrv_co_create_opts = hdev_co_create_opts, |
6f482f74 | 3478 | .create_opts = &raw_create_opts, |
f3a5d3f8 | 3479 | |
9d52aa3c KW |
3480 | .bdrv_co_preadv = raw_co_preadv, |
3481 | .bdrv_co_pwritev = raw_co_pwritev, | |
33d70fb6 | 3482 | .bdrv_co_flush_to_disk = raw_co_flush_to_disk, |
c25f53b0 | 3483 | .bdrv_refresh_limits = raw_refresh_limits, |
1b3abdcc ML |
3484 | .bdrv_io_plug = raw_aio_plug, |
3485 | .bdrv_io_unplug = raw_aio_unplug, | |
042b757c | 3486 | .bdrv_attach_aio_context = raw_aio_attach_aio_context, |
f3a5d3f8 | 3487 | |
061ca8a3 | 3488 | .bdrv_co_truncate = raw_co_truncate, |
b94a2610 KW |
3489 | .bdrv_getlength = raw_getlength, |
3490 | .has_variable_length = true, | |
4a1d5e1f FZ |
3491 | .bdrv_get_allocated_file_size |
3492 | = raw_get_allocated_file_size, | |
f3a5d3f8 | 3493 | |
19cb3738 | 3494 | /* removable device support */ |
f3a5d3f8 CH |
3495 | .bdrv_is_inserted = cdrom_is_inserted, |
3496 | .bdrv_eject = cdrom_eject, | |
025e849a | 3497 | .bdrv_lock_medium = cdrom_lock_medium, |
19cb3738 | 3498 | }; |
f3a5d3f8 | 3499 | #endif /* __FreeBSD__ */ |
5efa9d5a | 3500 | |
84a12e66 | 3501 | static void bdrv_file_init(void) |
5efa9d5a | 3502 | { |
508c7cb3 CH |
3503 | /* |
3504 | * Register all the drivers. Note that order is important, the driver | |
3505 | * registered last will get probed first. | |
3506 | */ | |
84a12e66 | 3507 | bdrv_register(&bdrv_file); |
5efa9d5a | 3508 | bdrv_register(&bdrv_host_device); |
f3a5d3f8 | 3509 | #ifdef __linux__ |
f3a5d3f8 CH |
3510 | bdrv_register(&bdrv_host_cdrom); |
3511 | #endif | |
a167ba50 | 3512 | #if defined(__FreeBSD__) || defined(__FreeBSD_kernel__) |
f3a5d3f8 CH |
3513 | bdrv_register(&bdrv_host_cdrom); |
3514 | #endif | |
5efa9d5a AL |
3515 | } |
3516 | ||
84a12e66 | 3517 | block_init(bdrv_file_init); |