]>
Commit | Line | Data |
---|---|---|
83f64091 | 1 | /* |
223d4670 | 2 | * Block driver for RAW files (posix) |
5fafdf24 | 3 | * |
83f64091 | 4 | * Copyright (c) 2006 Fabrice Bellard |
5fafdf24 | 5 | * |
83f64091 FB |
6 | * Permission is hereby granted, free of charge, to any person obtaining a copy |
7 | * of this software and associated documentation files (the "Software"), to deal | |
8 | * in the Software without restriction, including without limitation the rights | |
9 | * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell | |
10 | * copies of the Software, and to permit persons to whom the Software is | |
11 | * furnished to do so, subject to the following conditions: | |
12 | * | |
13 | * The above copyright notice and this permission notice shall be included in | |
14 | * all copies or substantial portions of the Software. | |
15 | * | |
16 | * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR | |
17 | * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, | |
18 | * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL | |
19 | * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER | |
20 | * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, | |
21 | * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN | |
22 | * THE SOFTWARE. | |
23 | */ | |
faf07963 | 24 | #include "qemu-common.h" |
1de7afc9 PB |
25 | #include "qemu/timer.h" |
26 | #include "qemu/log.h" | |
737e150e | 27 | #include "block/block_int.h" |
1de7afc9 | 28 | #include "qemu/module.h" |
de81a169 | 29 | #include "trace.h" |
737e150e | 30 | #include "block/thread-pool.h" |
1de7afc9 | 31 | #include "qemu/iov.h" |
9f8540ec | 32 | #include "raw-aio.h" |
06247428 | 33 | #include "qapi/util.h" |
83f64091 | 34 | |
83affaa6 | 35 | #if defined(__APPLE__) && (__MACH__) |
83f64091 FB |
36 | #include <paths.h> |
37 | #include <sys/param.h> | |
38 | #include <IOKit/IOKitLib.h> | |
39 | #include <IOKit/IOBSD.h> | |
40 | #include <IOKit/storage/IOMediaBSDClient.h> | |
41 | #include <IOKit/storage/IOMedia.h> | |
42 | #include <IOKit/storage/IOCDMedia.h> | |
43 | //#include <IOKit/storage/IOCDTypes.h> | |
44 | #include <CoreFoundation/CoreFoundation.h> | |
45 | #endif | |
46 | ||
47 | #ifdef __sun__ | |
2e9671da | 48 | #define _POSIX_PTHREAD_SEMANTICS 1 |
83f64091 FB |
49 | #include <sys/dkio.h> |
50 | #endif | |
19cb3738 | 51 | #ifdef __linux__ |
343f8568 JS |
52 | #include <sys/types.h> |
53 | #include <sys/stat.h> | |
19cb3738 | 54 | #include <sys/ioctl.h> |
05acda4d | 55 | #include <sys/param.h> |
19cb3738 FB |
56 | #include <linux/cdrom.h> |
57 | #include <linux/fd.h> | |
5500316d | 58 | #include <linux/fs.h> |
1a9335e4 ET |
59 | #include <linux/hdreg.h> |
60 | #ifdef __s390__ | |
61 | #include <asm/dasd.h> | |
62 | #endif | |
4ab15590 CL |
63 | #ifndef FS_NOCOW_FL |
64 | #define FS_NOCOW_FL 0x00800000 /* Do not cow file */ | |
65 | #endif | |
5500316d | 66 | #endif |
b953f075 | 67 | #if defined(CONFIG_FALLOCATE_PUNCH_HOLE) || defined(CONFIG_FALLOCATE_ZERO_RANGE) |
3d4fa43e KK |
68 | #include <linux/falloc.h> |
69 | #endif | |
a167ba50 | 70 | #if defined (__FreeBSD__) || defined(__FreeBSD_kernel__) |
1cb6c3fd | 71 | #include <sys/disk.h> |
9f23011a | 72 | #include <sys/cdio.h> |
1cb6c3fd | 73 | #endif |
83f64091 | 74 | |
128ab2ff BS |
75 | #ifdef __OpenBSD__ |
76 | #include <sys/ioctl.h> | |
77 | #include <sys/disklabel.h> | |
78 | #include <sys/dkio.h> | |
79 | #endif | |
80 | ||
d1f6fd8d CE |
81 | #ifdef __NetBSD__ |
82 | #include <sys/ioctl.h> | |
83 | #include <sys/disklabel.h> | |
84 | #include <sys/dkio.h> | |
85 | #include <sys/disk.h> | |
86 | #endif | |
87 | ||
c5e97233 BS |
88 | #ifdef __DragonFly__ |
89 | #include <sys/ioctl.h> | |
90 | #include <sys/diskslice.h> | |
91 | #endif | |
92 | ||
dce512de CH |
93 | #ifdef CONFIG_XFS |
94 | #include <xfs/xfs.h> | |
95 | #endif | |
96 | ||
19cb3738 | 97 | //#define DEBUG_FLOPPY |
83f64091 | 98 | |
faf07963 | 99 | //#define DEBUG_BLOCK |
03ff3ca3 | 100 | #if defined(DEBUG_BLOCK) |
001faf32 BS |
101 | #define DEBUG_BLOCK_PRINT(formatCstr, ...) do { if (qemu_log_enabled()) \ |
102 | { qemu_log(formatCstr, ## __VA_ARGS__); qemu_log_flush(); } } while (0) | |
8c05dbf9 | 103 | #else |
001faf32 | 104 | #define DEBUG_BLOCK_PRINT(formatCstr, ...) |
8c05dbf9 TS |
105 | #endif |
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 | |
123 | #define FTYPE_FD 2 | |
83f64091 | 124 | |
c57c846a | 125 | /* if the FD is not accessed during that time (in ns), we try to |
19cb3738 | 126 | reopen it to see if the disk has been changed */ |
c57c846a | 127 | #define FD_OPEN_TIMEOUT (1000000000) |
83f64091 | 128 | |
581b9e29 CH |
129 | #define MAX_BLOCKSIZE 4096 |
130 | ||
19cb3738 FB |
131 | typedef struct BDRVRawState { |
132 | int fd; | |
133 | int type; | |
0e1d8f4c | 134 | int open_flags; |
c25f53b0 PB |
135 | size_t buf_align; |
136 | ||
19cb3738 FB |
137 | #if defined(__linux__) |
138 | /* linux floppy specific */ | |
19cb3738 FB |
139 | int64_t fd_open_time; |
140 | int64_t fd_error_time; | |
141 | int fd_got_error; | |
142 | int fd_media_changed; | |
83f64091 | 143 | #endif |
e44bd6fc | 144 | #ifdef CONFIG_LINUX_AIO |
5c6c3a6c | 145 | int use_aio; |
1e5b9d2f | 146 | void *aio_ctx; |
e44bd6fc | 147 | #endif |
dce512de | 148 | #ifdef CONFIG_XFS |
260a82e5 | 149 | bool is_xfs:1; |
dce512de | 150 | #endif |
260a82e5 | 151 | bool has_discard:1; |
97a2ae34 | 152 | bool has_write_zeroes:1; |
260a82e5 | 153 | bool discard_zeroes:1; |
d50d8222 | 154 | bool has_fallocate; |
3cad8307 | 155 | bool needs_alignment; |
19cb3738 FB |
156 | } BDRVRawState; |
157 | ||
eeb6b45d JC |
158 | typedef struct BDRVRawReopenState { |
159 | int fd; | |
160 | int open_flags; | |
161 | #ifdef CONFIG_LINUX_AIO | |
162 | int use_aio; | |
163 | #endif | |
164 | } BDRVRawReopenState; | |
165 | ||
19cb3738 | 166 | static int fd_open(BlockDriverState *bs); |
22afa7b5 | 167 | static int64_t raw_getlength(BlockDriverState *bs); |
83f64091 | 168 | |
de81a169 PB |
169 | typedef struct RawPosixAIOData { |
170 | BlockDriverState *bs; | |
171 | int aio_fildes; | |
172 | union { | |
173 | struct iovec *aio_iov; | |
174 | void *aio_ioctl_buf; | |
175 | }; | |
176 | int aio_niov; | |
8238010b | 177 | uint64_t aio_nbytes; |
de81a169 PB |
178 | #define aio_ioctl_cmd aio_nbytes /* for QEMU_AIO_IOCTL */ |
179 | off_t aio_offset; | |
180 | int aio_type; | |
181 | } RawPosixAIOData; | |
182 | ||
a167ba50 | 183 | #if defined(__FreeBSD__) || defined(__FreeBSD_kernel__) |
f3a5d3f8 | 184 | static int cdrom_reopen(BlockDriverState *bs); |
9f23011a BS |
185 | #endif |
186 | ||
1de1ae0a CE |
187 | #if defined(__NetBSD__) |
188 | static int raw_normalize_devicepath(const char **filename) | |
189 | { | |
190 | static char namebuf[PATH_MAX]; | |
191 | const char *dp, *fname; | |
192 | struct stat sb; | |
193 | ||
194 | fname = *filename; | |
195 | dp = strrchr(fname, '/'); | |
196 | if (lstat(fname, &sb) < 0) { | |
197 | fprintf(stderr, "%s: stat failed: %s\n", | |
198 | fname, strerror(errno)); | |
199 | return -errno; | |
200 | } | |
201 | ||
202 | if (!S_ISBLK(sb.st_mode)) { | |
203 | return 0; | |
204 | } | |
205 | ||
206 | if (dp == NULL) { | |
207 | snprintf(namebuf, PATH_MAX, "r%s", fname); | |
208 | } else { | |
209 | snprintf(namebuf, PATH_MAX, "%.*s/r%s", | |
210 | (int)(dp - fname), fname, dp + 1); | |
211 | } | |
212 | fprintf(stderr, "%s is a block device", fname); | |
213 | *filename = namebuf; | |
214 | fprintf(stderr, ", using %s\n", *filename); | |
215 | ||
216 | return 0; | |
217 | } | |
218 | #else | |
219 | static int raw_normalize_devicepath(const char **filename) | |
220 | { | |
221 | return 0; | |
222 | } | |
223 | #endif | |
224 | ||
8a4ed0d1 ET |
225 | /* |
226 | * Get logical block size via ioctl. On success store it in @sector_size_p. | |
227 | */ | |
228 | static int probe_logical_blocksize(int fd, unsigned int *sector_size_p) | |
c25f53b0 | 229 | { |
c25f53b0 | 230 | unsigned int sector_size; |
8a4ed0d1 | 231 | bool success = false; |
c25f53b0 | 232 | |
8a4ed0d1 | 233 | errno = ENOTSUP; |
c25f53b0 PB |
234 | |
235 | /* Try a few ioctls to get the right size */ | |
c25f53b0 | 236 | #ifdef BLKSSZGET |
df26a350 | 237 | if (ioctl(fd, BLKSSZGET, §or_size) >= 0) { |
8a4ed0d1 ET |
238 | *sector_size_p = sector_size; |
239 | success = true; | |
c25f53b0 PB |
240 | } |
241 | #endif | |
242 | #ifdef DKIOCGETBLOCKSIZE | |
df26a350 | 243 | if (ioctl(fd, DKIOCGETBLOCKSIZE, §or_size) >= 0) { |
8a4ed0d1 ET |
244 | *sector_size_p = sector_size; |
245 | success = true; | |
c25f53b0 PB |
246 | } |
247 | #endif | |
248 | #ifdef DIOCGSECTORSIZE | |
df26a350 | 249 | if (ioctl(fd, DIOCGSECTORSIZE, §or_size) >= 0) { |
8a4ed0d1 ET |
250 | *sector_size_p = sector_size; |
251 | success = true; | |
c25f53b0 PB |
252 | } |
253 | #endif | |
8a4ed0d1 ET |
254 | |
255 | return success ? 0 : -errno; | |
256 | } | |
257 | ||
1a9335e4 ET |
258 | /** |
259 | * Get physical block size of @fd. | |
260 | * On success, store it in @blk_size and return 0. | |
261 | * On failure, return -errno. | |
262 | */ | |
263 | static int probe_physical_blocksize(int fd, unsigned int *blk_size) | |
264 | { | |
265 | #ifdef BLKPBSZGET | |
266 | if (ioctl(fd, BLKPBSZGET, blk_size) < 0) { | |
267 | return -errno; | |
268 | } | |
269 | return 0; | |
270 | #else | |
271 | return -ENOTSUP; | |
272 | #endif | |
273 | } | |
274 | ||
22d182e8 SH |
275 | /* Check if read is allowed with given memory buffer and length. |
276 | * | |
277 | * This function is used to check O_DIRECT memory buffer and request alignment. | |
278 | */ | |
279 | static bool raw_is_io_aligned(int fd, void *buf, size_t len) | |
280 | { | |
281 | ssize_t ret = pread(fd, buf, len, 0); | |
282 | ||
283 | if (ret >= 0) { | |
284 | return true; | |
285 | } | |
286 | ||
287 | #ifdef __linux__ | |
288 | /* The Linux kernel returns EINVAL for misaligned O_DIRECT reads. Ignore | |
289 | * other errors (e.g. real I/O error), which could happen on a failed | |
290 | * drive, since we only care about probing alignment. | |
291 | */ | |
292 | if (errno != EINVAL) { | |
293 | return true; | |
294 | } | |
295 | #endif | |
296 | ||
297 | return false; | |
298 | } | |
299 | ||
8a4ed0d1 ET |
300 | static void raw_probe_alignment(BlockDriverState *bs, int fd, Error **errp) |
301 | { | |
302 | BDRVRawState *s = bs->opaque; | |
303 | char *buf; | |
459b4e66 | 304 | size_t max_align = MAX(MAX_BLOCKSIZE, getpagesize()); |
8a4ed0d1 ET |
305 | |
306 | /* For /dev/sg devices the alignment is not really used. | |
307 | With buffered I/O, we don't have any restrictions. */ | |
308 | if (bs->sg || !s->needs_alignment) { | |
309 | bs->request_alignment = 1; | |
310 | s->buf_align = 1; | |
311 | return; | |
312 | } | |
313 | ||
314 | bs->request_alignment = 0; | |
315 | s->buf_align = 0; | |
316 | /* Let's try to use the logical blocksize for the alignment. */ | |
317 | if (probe_logical_blocksize(fd, &bs->request_alignment) < 0) { | |
318 | bs->request_alignment = 0; | |
319 | } | |
c25f53b0 PB |
320 | #ifdef CONFIG_XFS |
321 | if (s->is_xfs) { | |
322 | struct dioattr da; | |
df26a350 | 323 | if (xfsctl(NULL, fd, XFS_IOC_DIOINFO, &da) >= 0) { |
c25f53b0 PB |
324 | bs->request_alignment = da.d_miniosz; |
325 | /* The kernel returns wrong information for d_mem */ | |
326 | /* s->buf_align = da.d_mem; */ | |
327 | } | |
328 | } | |
329 | #endif | |
330 | ||
331 | /* If we could not get the sizes so far, we can only guess them */ | |
332 | if (!s->buf_align) { | |
333 | size_t align; | |
459b4e66 DL |
334 | buf = qemu_memalign(max_align, 2 * max_align); |
335 | for (align = 512; align <= max_align; align <<= 1) { | |
336 | if (raw_is_io_aligned(fd, buf + align, max_align)) { | |
c25f53b0 PB |
337 | s->buf_align = align; |
338 | break; | |
339 | } | |
340 | } | |
341 | qemu_vfree(buf); | |
342 | } | |
343 | ||
344 | if (!bs->request_alignment) { | |
345 | size_t align; | |
459b4e66 DL |
346 | buf = qemu_memalign(s->buf_align, max_align); |
347 | for (align = 512; align <= max_align; align <<= 1) { | |
22d182e8 | 348 | if (raw_is_io_aligned(fd, buf, align)) { |
c25f53b0 PB |
349 | bs->request_alignment = align; |
350 | break; | |
351 | } | |
352 | } | |
353 | qemu_vfree(buf); | |
354 | } | |
df26a350 KW |
355 | |
356 | if (!s->buf_align || !bs->request_alignment) { | |
357 | error_setg(errp, "Could not find working O_DIRECT alignment. " | |
358 | "Try cache.direct=off."); | |
359 | } | |
c25f53b0 PB |
360 | } |
361 | ||
6a8dc042 JC |
362 | static void raw_parse_flags(int bdrv_flags, int *open_flags) |
363 | { | |
364 | assert(open_flags != NULL); | |
365 | ||
366 | *open_flags |= O_BINARY; | |
367 | *open_flags &= ~O_ACCMODE; | |
368 | if (bdrv_flags & BDRV_O_RDWR) { | |
369 | *open_flags |= O_RDWR; | |
370 | } else { | |
371 | *open_flags |= O_RDONLY; | |
372 | } | |
373 | ||
374 | /* Use O_DSYNC for write-through caching, no flags for write-back caching, | |
375 | * and O_DIRECT for no caching. */ | |
376 | if ((bdrv_flags & BDRV_O_NOCACHE)) { | |
377 | *open_flags |= O_DIRECT; | |
378 | } | |
6a8dc042 JC |
379 | } |
380 | ||
c2f3426c SH |
381 | static void raw_detach_aio_context(BlockDriverState *bs) |
382 | { | |
383 | #ifdef CONFIG_LINUX_AIO | |
384 | BDRVRawState *s = bs->opaque; | |
385 | ||
386 | if (s->use_aio) { | |
387 | laio_detach_aio_context(s->aio_ctx, bdrv_get_aio_context(bs)); | |
388 | } | |
389 | #endif | |
390 | } | |
391 | ||
392 | static void raw_attach_aio_context(BlockDriverState *bs, | |
393 | AioContext *new_context) | |
394 | { | |
395 | #ifdef CONFIG_LINUX_AIO | |
396 | BDRVRawState *s = bs->opaque; | |
397 | ||
398 | if (s->use_aio) { | |
399 | laio_attach_aio_context(s->aio_ctx, new_context); | |
400 | } | |
401 | #endif | |
402 | } | |
403 | ||
fc32a72d JC |
404 | #ifdef CONFIG_LINUX_AIO |
405 | static int raw_set_aio(void **aio_ctx, int *use_aio, int bdrv_flags) | |
406 | { | |
407 | int ret = -1; | |
408 | assert(aio_ctx != NULL); | |
409 | assert(use_aio != NULL); | |
410 | /* | |
411 | * Currently Linux do AIO only for files opened with O_DIRECT | |
412 | * specified so check NOCACHE flag too | |
413 | */ | |
414 | if ((bdrv_flags & (BDRV_O_NOCACHE|BDRV_O_NATIVE_AIO)) == | |
415 | (BDRV_O_NOCACHE|BDRV_O_NATIVE_AIO)) { | |
416 | ||
417 | /* if non-NULL, laio_init() has already been run */ | |
418 | if (*aio_ctx == NULL) { | |
419 | *aio_ctx = laio_init(); | |
420 | if (!*aio_ctx) { | |
421 | goto error; | |
422 | } | |
423 | } | |
424 | *use_aio = 1; | |
425 | } else { | |
426 | *use_aio = 0; | |
427 | } | |
428 | ||
429 | ret = 0; | |
430 | ||
431 | error: | |
432 | return ret; | |
433 | } | |
434 | #endif | |
435 | ||
078896a9 HR |
436 | static void raw_parse_filename(const char *filename, QDict *options, |
437 | Error **errp) | |
438 | { | |
439 | /* The filename does not have to be prefixed by the protocol name, since | |
440 | * "file" is the default protocol; therefore, the return value of this | |
441 | * function call can be ignored. */ | |
442 | strstart(filename, "file:", &filename); | |
443 | ||
444 | qdict_put_obj(options, "filename", QOBJECT(qstring_from_str(filename))); | |
445 | } | |
446 | ||
c66a6157 KW |
447 | static QemuOptsList raw_runtime_opts = { |
448 | .name = "raw", | |
449 | .head = QTAILQ_HEAD_INITIALIZER(raw_runtime_opts.head), | |
450 | .desc = { | |
451 | { | |
452 | .name = "filename", | |
453 | .type = QEMU_OPT_STRING, | |
454 | .help = "File name of the image", | |
455 | }, | |
456 | { /* end of list */ } | |
457 | }, | |
458 | }; | |
459 | ||
460 | static int raw_open_common(BlockDriverState *bs, QDict *options, | |
e428e439 | 461 | int bdrv_flags, int open_flags, Error **errp) |
83f64091 FB |
462 | { |
463 | BDRVRawState *s = bs->opaque; | |
c66a6157 KW |
464 | QemuOpts *opts; |
465 | Error *local_err = NULL; | |
8bfea15d | 466 | const char *filename = NULL; |
0e1d8f4c | 467 | int fd, ret; |
260a82e5 | 468 | struct stat st; |
83f64091 | 469 | |
87ea75d5 | 470 | opts = qemu_opts_create(&raw_runtime_opts, NULL, 0, &error_abort); |
c66a6157 | 471 | qemu_opts_absorb_qdict(opts, options, &local_err); |
84d18f06 | 472 | if (local_err) { |
e428e439 | 473 | error_propagate(errp, local_err); |
c66a6157 KW |
474 | ret = -EINVAL; |
475 | goto fail; | |
476 | } | |
477 | ||
478 | filename = qemu_opt_get(opts, "filename"); | |
479 | ||
1de1ae0a CE |
480 | ret = raw_normalize_devicepath(&filename); |
481 | if (ret != 0) { | |
e428e439 | 482 | error_setg_errno(errp, -ret, "Could not normalize device path"); |
c66a6157 | 483 | goto fail; |
1de1ae0a CE |
484 | } |
485 | ||
6a8dc042 JC |
486 | s->open_flags = open_flags; |
487 | raw_parse_flags(bdrv_flags, &s->open_flags); | |
83f64091 | 488 | |
90babde0 | 489 | s->fd = -1; |
40ff6d7e | 490 | fd = qemu_open(filename, s->open_flags, 0644); |
19cb3738 FB |
491 | if (fd < 0) { |
492 | ret = -errno; | |
c66a6157 | 493 | if (ret == -EROFS) { |
19cb3738 | 494 | ret = -EACCES; |
c66a6157 KW |
495 | } |
496 | goto fail; | |
19cb3738 | 497 | } |
83f64091 | 498 | s->fd = fd; |
9ef91a67 | 499 | |
5c6c3a6c | 500 | #ifdef CONFIG_LINUX_AIO |
fc32a72d | 501 | if (raw_set_aio(&s->aio_ctx, &s->use_aio, bdrv_flags)) { |
47e6b251 | 502 | qemu_close(fd); |
c66a6157 | 503 | ret = -errno; |
e428e439 | 504 | error_setg_errno(errp, -ret, "Could not set AIO state"); |
c66a6157 | 505 | goto fail; |
9ef91a67 | 506 | } |
96518254 KW |
507 | if (!s->use_aio && (bdrv_flags & BDRV_O_NATIVE_AIO)) { |
508 | error_printf("WARNING: aio=native was specified for '%s', but " | |
509 | "it requires cache.direct=on, which was not " | |
510 | "specified. Falling back to aio=threads.\n" | |
511 | " This will become an error condition in " | |
512 | "future QEMU versions.\n", | |
513 | bs->filename); | |
514 | } | |
fc32a72d | 515 | #endif |
9ef91a67 | 516 | |
7ce21016 | 517 | s->has_discard = true; |
97a2ae34 | 518 | s->has_write_zeroes = true; |
3cad8307 RPM |
519 | if ((bs->open_flags & BDRV_O_NOCACHE) != 0) { |
520 | s->needs_alignment = true; | |
521 | } | |
260a82e5 PB |
522 | |
523 | if (fstat(s->fd, &st) < 0) { | |
01212d4e | 524 | ret = -errno; |
260a82e5 PB |
525 | error_setg_errno(errp, errno, "Could not stat file"); |
526 | goto fail; | |
527 | } | |
528 | if (S_ISREG(st.st_mode)) { | |
529 | s->discard_zeroes = true; | |
d50d8222 | 530 | s->has_fallocate = true; |
260a82e5 | 531 | } |
d0b4503e PB |
532 | if (S_ISBLK(st.st_mode)) { |
533 | #ifdef BLKDISCARDZEROES | |
534 | unsigned int arg; | |
535 | if (ioctl(s->fd, BLKDISCARDZEROES, &arg) == 0 && arg) { | |
536 | s->discard_zeroes = true; | |
537 | } | |
538 | #endif | |
539 | #ifdef __linux__ | |
540 | /* On Linux 3.10, BLKDISCARD leaves stale data in the page cache. Do | |
541 | * not rely on the contents of discarded blocks unless using O_DIRECT. | |
97a2ae34 | 542 | * Same for BLKZEROOUT. |
d0b4503e PB |
543 | */ |
544 | if (!(bs->open_flags & BDRV_O_NOCACHE)) { | |
545 | s->discard_zeroes = false; | |
97a2ae34 | 546 | s->has_write_zeroes = false; |
d0b4503e PB |
547 | } |
548 | #endif | |
549 | } | |
3cad8307 RPM |
550 | #ifdef __FreeBSD__ |
551 | if (S_ISCHR(st.st_mode)) { | |
552 | /* | |
553 | * The file is a char device (disk), which on FreeBSD isn't behind | |
554 | * a pager, so force all requests to be aligned. This is needed | |
555 | * so QEMU makes sure all IO operations on the device are aligned | |
556 | * to sector size, or else FreeBSD will reject them with EINVAL. | |
557 | */ | |
558 | s->needs_alignment = true; | |
559 | } | |
560 | #endif | |
260a82e5 | 561 | |
dce512de CH |
562 | #ifdef CONFIG_XFS |
563 | if (platform_test_xfs_fd(s->fd)) { | |
7ce21016 | 564 | s->is_xfs = true; |
dce512de CH |
565 | } |
566 | #endif | |
567 | ||
c2f3426c SH |
568 | raw_attach_aio_context(bs, bdrv_get_aio_context(bs)); |
569 | ||
c66a6157 KW |
570 | ret = 0; |
571 | fail: | |
8bfea15d KW |
572 | if (filename && (bdrv_flags & BDRV_O_TEMPORARY)) { |
573 | unlink(filename); | |
574 | } | |
c66a6157 KW |
575 | qemu_opts_del(opts); |
576 | return ret; | |
83f64091 FB |
577 | } |
578 | ||
015a1036 HR |
579 | static int raw_open(BlockDriverState *bs, QDict *options, int flags, |
580 | Error **errp) | |
90babde0 CH |
581 | { |
582 | BDRVRawState *s = bs->opaque; | |
e428e439 HR |
583 | Error *local_err = NULL; |
584 | int ret; | |
90babde0 CH |
585 | |
586 | s->type = FTYPE_FILE; | |
e428e439 | 587 | ret = raw_open_common(bs, options, flags, 0, &local_err); |
84d18f06 | 588 | if (local_err) { |
e428e439 HR |
589 | error_propagate(errp, local_err); |
590 | } | |
591 | return ret; | |
90babde0 CH |
592 | } |
593 | ||
eeb6b45d JC |
594 | static int raw_reopen_prepare(BDRVReopenState *state, |
595 | BlockReopenQueue *queue, Error **errp) | |
596 | { | |
597 | BDRVRawState *s; | |
598 | BDRVRawReopenState *raw_s; | |
599 | int ret = 0; | |
df26a350 | 600 | Error *local_err = NULL; |
eeb6b45d JC |
601 | |
602 | assert(state != NULL); | |
603 | assert(state->bs != NULL); | |
604 | ||
605 | s = state->bs->opaque; | |
606 | ||
5839e53b | 607 | state->opaque = g_new0(BDRVRawReopenState, 1); |
eeb6b45d JC |
608 | raw_s = state->opaque; |
609 | ||
610 | #ifdef CONFIG_LINUX_AIO | |
611 | raw_s->use_aio = s->use_aio; | |
612 | ||
613 | /* we can use s->aio_ctx instead of a copy, because the use_aio flag is | |
614 | * valid in the 'false' condition even if aio_ctx is set, and raw_set_aio() | |
615 | * won't override aio_ctx if aio_ctx is non-NULL */ | |
616 | if (raw_set_aio(&s->aio_ctx, &raw_s->use_aio, state->flags)) { | |
e428e439 | 617 | error_setg(errp, "Could not set AIO state"); |
eeb6b45d JC |
618 | return -1; |
619 | } | |
620 | #endif | |
621 | ||
1bc6b705 JC |
622 | if (s->type == FTYPE_FD || s->type == FTYPE_CD) { |
623 | raw_s->open_flags |= O_NONBLOCK; | |
624 | } | |
625 | ||
eeb6b45d JC |
626 | raw_parse_flags(state->flags, &raw_s->open_flags); |
627 | ||
628 | raw_s->fd = -1; | |
629 | ||
fdf263f6 | 630 | int fcntl_flags = O_APPEND | O_NONBLOCK; |
eeb6b45d JC |
631 | #ifdef O_NOATIME |
632 | fcntl_flags |= O_NOATIME; | |
633 | #endif | |
634 | ||
fdf263f6 AF |
635 | #ifdef O_ASYNC |
636 | /* Not all operating systems have O_ASYNC, and those that don't | |
637 | * will not let us track the state into raw_s->open_flags (typically | |
638 | * you achieve the same effect with an ioctl, for example I_SETSIG | |
639 | * on Solaris). But we do not use O_ASYNC, so that's fine. | |
640 | */ | |
641 | assert((s->open_flags & O_ASYNC) == 0); | |
642 | #endif | |
643 | ||
eeb6b45d JC |
644 | if ((raw_s->open_flags & ~fcntl_flags) == (s->open_flags & ~fcntl_flags)) { |
645 | /* dup the original fd */ | |
646 | /* TODO: use qemu fcntl wrapper */ | |
647 | #ifdef F_DUPFD_CLOEXEC | |
648 | raw_s->fd = fcntl(s->fd, F_DUPFD_CLOEXEC, 0); | |
649 | #else | |
650 | raw_s->fd = dup(s->fd); | |
651 | if (raw_s->fd != -1) { | |
652 | qemu_set_cloexec(raw_s->fd); | |
653 | } | |
654 | #endif | |
655 | if (raw_s->fd >= 0) { | |
656 | ret = fcntl_setfl(raw_s->fd, raw_s->open_flags); | |
657 | if (ret) { | |
658 | qemu_close(raw_s->fd); | |
659 | raw_s->fd = -1; | |
660 | } | |
661 | } | |
662 | } | |
663 | ||
664 | /* If we cannot use fcntl, or fcntl failed, fall back to qemu_open() */ | |
665 | if (raw_s->fd == -1) { | |
666 | assert(!(raw_s->open_flags & O_CREAT)); | |
667 | raw_s->fd = qemu_open(state->bs->filename, raw_s->open_flags); | |
668 | if (raw_s->fd == -1) { | |
e428e439 | 669 | error_setg_errno(errp, errno, "Could not reopen file"); |
eeb6b45d JC |
670 | ret = -1; |
671 | } | |
672 | } | |
df26a350 KW |
673 | |
674 | /* Fail already reopen_prepare() if we can't get a working O_DIRECT | |
675 | * alignment with the new fd. */ | |
676 | if (raw_s->fd != -1) { | |
677 | raw_probe_alignment(state->bs, raw_s->fd, &local_err); | |
678 | if (local_err) { | |
679 | qemu_close(raw_s->fd); | |
680 | raw_s->fd = -1; | |
681 | error_propagate(errp, local_err); | |
682 | ret = -EINVAL; | |
683 | } | |
684 | } | |
685 | ||
eeb6b45d JC |
686 | return ret; |
687 | } | |
688 | ||
eeb6b45d JC |
689 | static void raw_reopen_commit(BDRVReopenState *state) |
690 | { | |
691 | BDRVRawReopenState *raw_s = state->opaque; | |
692 | BDRVRawState *s = state->bs->opaque; | |
693 | ||
694 | s->open_flags = raw_s->open_flags; | |
695 | ||
696 | qemu_close(s->fd); | |
697 | s->fd = raw_s->fd; | |
698 | #ifdef CONFIG_LINUX_AIO | |
699 | s->use_aio = raw_s->use_aio; | |
700 | #endif | |
701 | ||
702 | g_free(state->opaque); | |
703 | state->opaque = NULL; | |
704 | } | |
705 | ||
706 | ||
707 | static void raw_reopen_abort(BDRVReopenState *state) | |
708 | { | |
709 | BDRVRawReopenState *raw_s = state->opaque; | |
710 | ||
711 | /* nothing to do if NULL, we didn't get far enough */ | |
712 | if (raw_s == NULL) { | |
713 | return; | |
714 | } | |
715 | ||
716 | if (raw_s->fd >= 0) { | |
717 | qemu_close(raw_s->fd); | |
718 | raw_s->fd = -1; | |
719 | } | |
720 | g_free(state->opaque); | |
721 | state->opaque = NULL; | |
722 | } | |
723 | ||
3baca891 | 724 | static void raw_refresh_limits(BlockDriverState *bs, Error **errp) |
c25f53b0 PB |
725 | { |
726 | BDRVRawState *s = bs->opaque; | |
eeb6b45d | 727 | |
df26a350 | 728 | raw_probe_alignment(bs, s->fd, errp); |
4196d2f0 | 729 | bs->bl.min_mem_alignment = s->buf_align; |
459b4e66 | 730 | bs->bl.opt_mem_alignment = MAX(s->buf_align, getpagesize()); |
c25f53b0 | 731 | } |
83f64091 | 732 | |
1a9335e4 ET |
733 | static int check_for_dasd(int fd) |
734 | { | |
735 | #ifdef BIODASDINFO2 | |
736 | struct dasd_information2_t info = {0}; | |
737 | ||
738 | return ioctl(fd, BIODASDINFO2, &info); | |
739 | #else | |
740 | return -1; | |
741 | #endif | |
742 | } | |
743 | ||
744 | /** | |
745 | * Try to get @bs's logical and physical block size. | |
746 | * On success, store them in @bsz and return zero. | |
747 | * On failure, return negative errno. | |
748 | */ | |
749 | static int hdev_probe_blocksizes(BlockDriverState *bs, BlockSizes *bsz) | |
750 | { | |
751 | BDRVRawState *s = bs->opaque; | |
752 | int ret; | |
753 | ||
754 | /* If DASD, get blocksizes */ | |
755 | if (check_for_dasd(s->fd) < 0) { | |
756 | return -ENOTSUP; | |
757 | } | |
758 | ret = probe_logical_blocksize(s->fd, &bsz->log); | |
759 | if (ret < 0) { | |
760 | return ret; | |
761 | } | |
762 | return probe_physical_blocksize(s->fd, &bsz->phys); | |
763 | } | |
764 | ||
765 | /** | |
766 | * Try to get @bs's geometry: cyls, heads, sectors. | |
767 | * On success, store them in @geo and return 0. | |
768 | * On failure return -errno. | |
769 | * (Allows block driver to assign default geometry values that guest sees) | |
770 | */ | |
771 | #ifdef __linux__ | |
772 | static int hdev_probe_geometry(BlockDriverState *bs, HDGeometry *geo) | |
773 | { | |
774 | BDRVRawState *s = bs->opaque; | |
775 | struct hd_geometry ioctl_geo = {0}; | |
776 | uint32_t blksize; | |
777 | ||
778 | /* If DASD, get its geometry */ | |
779 | if (check_for_dasd(s->fd) < 0) { | |
780 | return -ENOTSUP; | |
781 | } | |
782 | if (ioctl(s->fd, HDIO_GETGEO, &ioctl_geo) < 0) { | |
783 | return -errno; | |
784 | } | |
785 | /* HDIO_GETGEO may return success even though geo contains zeros | |
786 | (e.g. certain multipath setups) */ | |
787 | if (!ioctl_geo.heads || !ioctl_geo.sectors || !ioctl_geo.cylinders) { | |
788 | return -ENOTSUP; | |
789 | } | |
790 | /* Do not return a geometry for partition */ | |
791 | if (ioctl_geo.start != 0) { | |
792 | return -ENOTSUP; | |
793 | } | |
794 | geo->heads = ioctl_geo.heads; | |
795 | geo->sectors = ioctl_geo.sectors; | |
796 | if (!probe_physical_blocksize(s->fd, &blksize)) { | |
797 | /* overwrite cyls: HDIO_GETGEO result is incorrect for big drives */ | |
798 | geo->cylinders = bdrv_nb_sectors(bs) / (blksize / BDRV_SECTOR_SIZE) | |
799 | / (geo->heads * geo->sectors); | |
800 | return 0; | |
801 | } | |
802 | geo->cylinders = ioctl_geo.cylinders; | |
803 | ||
804 | return 0; | |
805 | } | |
806 | #else /* __linux__ */ | |
807 | static int hdev_probe_geometry(BlockDriverState *bs, HDGeometry *geo) | |
808 | { | |
809 | return -ENOTSUP; | |
810 | } | |
811 | #endif | |
812 | ||
de81a169 PB |
813 | static ssize_t handle_aiocb_ioctl(RawPosixAIOData *aiocb) |
814 | { | |
815 | int ret; | |
816 | ||
817 | ret = ioctl(aiocb->aio_fildes, aiocb->aio_ioctl_cmd, aiocb->aio_ioctl_buf); | |
818 | if (ret == -1) { | |
819 | return -errno; | |
820 | } | |
821 | ||
b608c8dc | 822 | return 0; |
de81a169 PB |
823 | } |
824 | ||
825 | static ssize_t handle_aiocb_flush(RawPosixAIOData *aiocb) | |
826 | { | |
827 | int ret; | |
828 | ||
829 | ret = qemu_fdatasync(aiocb->aio_fildes); | |
830 | if (ret == -1) { | |
831 | return -errno; | |
832 | } | |
833 | return 0; | |
834 | } | |
835 | ||
836 | #ifdef CONFIG_PREADV | |
837 | ||
838 | static bool preadv_present = true; | |
839 | ||
840 | static ssize_t | |
841 | qemu_preadv(int fd, const struct iovec *iov, int nr_iov, off_t offset) | |
842 | { | |
843 | return preadv(fd, iov, nr_iov, offset); | |
844 | } | |
845 | ||
846 | static ssize_t | |
847 | qemu_pwritev(int fd, const struct iovec *iov, int nr_iov, off_t offset) | |
848 | { | |
849 | return pwritev(fd, iov, nr_iov, offset); | |
850 | } | |
851 | ||
852 | #else | |
853 | ||
854 | static bool preadv_present = false; | |
855 | ||
856 | static ssize_t | |
857 | qemu_preadv(int fd, const struct iovec *iov, int nr_iov, off_t offset) | |
858 | { | |
859 | return -ENOSYS; | |
860 | } | |
861 | ||
862 | static ssize_t | |
863 | qemu_pwritev(int fd, const struct iovec *iov, int nr_iov, off_t offset) | |
864 | { | |
865 | return -ENOSYS; | |
866 | } | |
867 | ||
868 | #endif | |
869 | ||
870 | static ssize_t handle_aiocb_rw_vector(RawPosixAIOData *aiocb) | |
871 | { | |
872 | ssize_t len; | |
873 | ||
874 | do { | |
875 | if (aiocb->aio_type & QEMU_AIO_WRITE) | |
876 | len = qemu_pwritev(aiocb->aio_fildes, | |
877 | aiocb->aio_iov, | |
878 | aiocb->aio_niov, | |
879 | aiocb->aio_offset); | |
880 | else | |
881 | len = qemu_preadv(aiocb->aio_fildes, | |
882 | aiocb->aio_iov, | |
883 | aiocb->aio_niov, | |
884 | aiocb->aio_offset); | |
885 | } while (len == -1 && errno == EINTR); | |
886 | ||
887 | if (len == -1) { | |
888 | return -errno; | |
889 | } | |
890 | return len; | |
891 | } | |
892 | ||
893 | /* | |
894 | * Read/writes the data to/from a given linear buffer. | |
895 | * | |
896 | * Returns the number of bytes handles or -errno in case of an error. Short | |
897 | * reads are only returned if the end of the file is reached. | |
898 | */ | |
899 | static ssize_t handle_aiocb_rw_linear(RawPosixAIOData *aiocb, char *buf) | |
900 | { | |
901 | ssize_t offset = 0; | |
902 | ssize_t len; | |
903 | ||
904 | while (offset < aiocb->aio_nbytes) { | |
905 | if (aiocb->aio_type & QEMU_AIO_WRITE) { | |
906 | len = pwrite(aiocb->aio_fildes, | |
907 | (const char *)buf + offset, | |
908 | aiocb->aio_nbytes - offset, | |
909 | aiocb->aio_offset + offset); | |
910 | } else { | |
911 | len = pread(aiocb->aio_fildes, | |
912 | buf + offset, | |
913 | aiocb->aio_nbytes - offset, | |
914 | aiocb->aio_offset + offset); | |
915 | } | |
916 | if (len == -1 && errno == EINTR) { | |
917 | continue; | |
61ed73cf SH |
918 | } else if (len == -1 && errno == EINVAL && |
919 | (aiocb->bs->open_flags & BDRV_O_NOCACHE) && | |
920 | !(aiocb->aio_type & QEMU_AIO_WRITE) && | |
921 | offset > 0) { | |
922 | /* O_DIRECT pread() may fail with EINVAL when offset is unaligned | |
923 | * after a short read. Assume that O_DIRECT short reads only occur | |
924 | * at EOF. Therefore this is a short read, not an I/O error. | |
925 | */ | |
926 | break; | |
de81a169 PB |
927 | } else if (len == -1) { |
928 | offset = -errno; | |
929 | break; | |
930 | } else if (len == 0) { | |
931 | break; | |
932 | } | |
933 | offset += len; | |
934 | } | |
935 | ||
936 | return offset; | |
937 | } | |
938 | ||
939 | static ssize_t handle_aiocb_rw(RawPosixAIOData *aiocb) | |
940 | { | |
941 | ssize_t nbytes; | |
942 | char *buf; | |
943 | ||
944 | if (!(aiocb->aio_type & QEMU_AIO_MISALIGNED)) { | |
945 | /* | |
946 | * If there is just a single buffer, and it is properly aligned | |
947 | * we can just use plain pread/pwrite without any problems. | |
948 | */ | |
949 | if (aiocb->aio_niov == 1) { | |
950 | return handle_aiocb_rw_linear(aiocb, aiocb->aio_iov->iov_base); | |
951 | } | |
952 | /* | |
953 | * We have more than one iovec, and all are properly aligned. | |
954 | * | |
955 | * Try preadv/pwritev first and fall back to linearizing the | |
956 | * buffer if it's not supported. | |
957 | */ | |
958 | if (preadv_present) { | |
959 | nbytes = handle_aiocb_rw_vector(aiocb); | |
960 | if (nbytes == aiocb->aio_nbytes || | |
961 | (nbytes < 0 && nbytes != -ENOSYS)) { | |
962 | return nbytes; | |
963 | } | |
964 | preadv_present = false; | |
965 | } | |
966 | ||
967 | /* | |
968 | * XXX(hch): short read/write. no easy way to handle the reminder | |
969 | * using these interfaces. For now retry using plain | |
970 | * pread/pwrite? | |
971 | */ | |
972 | } | |
973 | ||
974 | /* | |
975 | * Ok, we have to do it the hard way, copy all segments into | |
976 | * a single aligned buffer. | |
977 | */ | |
50d4a858 KW |
978 | buf = qemu_try_blockalign(aiocb->bs, aiocb->aio_nbytes); |
979 | if (buf == NULL) { | |
980 | return -ENOMEM; | |
981 | } | |
982 | ||
de81a169 PB |
983 | if (aiocb->aio_type & QEMU_AIO_WRITE) { |
984 | char *p = buf; | |
985 | int i; | |
986 | ||
987 | for (i = 0; i < aiocb->aio_niov; ++i) { | |
988 | memcpy(p, aiocb->aio_iov[i].iov_base, aiocb->aio_iov[i].iov_len); | |
989 | p += aiocb->aio_iov[i].iov_len; | |
990 | } | |
8eb029c2 | 991 | assert(p - buf == aiocb->aio_nbytes); |
de81a169 PB |
992 | } |
993 | ||
994 | nbytes = handle_aiocb_rw_linear(aiocb, buf); | |
995 | if (!(aiocb->aio_type & QEMU_AIO_WRITE)) { | |
996 | char *p = buf; | |
997 | size_t count = aiocb->aio_nbytes, copy; | |
998 | int i; | |
999 | ||
1000 | for (i = 0; i < aiocb->aio_niov && count; ++i) { | |
1001 | copy = count; | |
1002 | if (copy > aiocb->aio_iov[i].iov_len) { | |
1003 | copy = aiocb->aio_iov[i].iov_len; | |
1004 | } | |
1005 | memcpy(aiocb->aio_iov[i].iov_base, p, copy); | |
8eb029c2 | 1006 | assert(count >= copy); |
de81a169 PB |
1007 | p += copy; |
1008 | count -= copy; | |
1009 | } | |
8eb029c2 | 1010 | assert(count == 0); |
de81a169 PB |
1011 | } |
1012 | qemu_vfree(buf); | |
1013 | ||
1014 | return nbytes; | |
1015 | } | |
1016 | ||
8238010b | 1017 | #ifdef CONFIG_XFS |
97a2ae34 PB |
1018 | static int xfs_write_zeroes(BDRVRawState *s, int64_t offset, uint64_t bytes) |
1019 | { | |
1020 | struct xfs_flock64 fl; | |
1021 | ||
1022 | memset(&fl, 0, sizeof(fl)); | |
1023 | fl.l_whence = SEEK_SET; | |
1024 | fl.l_start = offset; | |
1025 | fl.l_len = bytes; | |
1026 | ||
1027 | if (xfsctl(NULL, s->fd, XFS_IOC_ZERO_RANGE, &fl) < 0) { | |
1028 | DEBUG_BLOCK_PRINT("cannot write zero range (%s)\n", strerror(errno)); | |
1029 | return -errno; | |
1030 | } | |
1031 | ||
1032 | return 0; | |
1033 | } | |
1034 | ||
8238010b PB |
1035 | static int xfs_discard(BDRVRawState *s, int64_t offset, uint64_t bytes) |
1036 | { | |
1037 | struct xfs_flock64 fl; | |
1038 | ||
1039 | memset(&fl, 0, sizeof(fl)); | |
1040 | fl.l_whence = SEEK_SET; | |
1041 | fl.l_start = offset; | |
1042 | fl.l_len = bytes; | |
1043 | ||
1044 | if (xfsctl(NULL, s->fd, XFS_IOC_UNRESVSP64, &fl) < 0) { | |
1045 | DEBUG_BLOCK_PRINT("cannot punch hole (%s)\n", strerror(errno)); | |
1046 | return -errno; | |
1047 | } | |
1048 | ||
1049 | return 0; | |
1050 | } | |
1051 | #endif | |
1052 | ||
1486df0e DL |
1053 | static int translate_err(int err) |
1054 | { | |
1055 | if (err == -ENODEV || err == -ENOSYS || err == -EOPNOTSUPP || | |
1056 | err == -ENOTTY) { | |
1057 | err = -ENOTSUP; | |
1058 | } | |
1059 | return err; | |
1060 | } | |
1061 | ||
d50d8222 | 1062 | #ifdef CONFIG_FALLOCATE |
0b991712 DL |
1063 | static int do_fallocate(int fd, int mode, off_t offset, off_t len) |
1064 | { | |
1065 | do { | |
1066 | if (fallocate(fd, mode, offset, len) == 0) { | |
1067 | return 0; | |
1068 | } | |
1069 | } while (errno == EINTR); | |
1070 | return translate_err(-errno); | |
1071 | } | |
1072 | #endif | |
1073 | ||
37cc9f7f | 1074 | static ssize_t handle_aiocb_write_zeroes_block(RawPosixAIOData *aiocb) |
97a2ae34 | 1075 | { |
37cc9f7f | 1076 | int ret = -ENOTSUP; |
97a2ae34 PB |
1077 | BDRVRawState *s = aiocb->bs->opaque; |
1078 | ||
37cc9f7f | 1079 | if (!s->has_write_zeroes) { |
97a2ae34 PB |
1080 | return -ENOTSUP; |
1081 | } | |
1082 | ||
97a2ae34 | 1083 | #ifdef BLKZEROOUT |
37cc9f7f DL |
1084 | do { |
1085 | uint64_t range[2] = { aiocb->aio_offset, aiocb->aio_nbytes }; | |
1086 | if (ioctl(aiocb->aio_fildes, BLKZEROOUT, range) == 0) { | |
1087 | return 0; | |
97a2ae34 | 1088 | } |
37cc9f7f DL |
1089 | } while (errno == EINTR); |
1090 | ||
1091 | ret = translate_err(-errno); | |
97a2ae34 | 1092 | #endif |
97a2ae34 | 1093 | |
1486df0e | 1094 | if (ret == -ENOTSUP) { |
97a2ae34 | 1095 | s->has_write_zeroes = false; |
97a2ae34 PB |
1096 | } |
1097 | return ret; | |
1098 | } | |
1099 | ||
37cc9f7f DL |
1100 | static ssize_t handle_aiocb_write_zeroes(RawPosixAIOData *aiocb) |
1101 | { | |
a6dcf097 | 1102 | #if defined(CONFIG_FALLOCATE) || defined(CONFIG_XFS) |
37cc9f7f | 1103 | BDRVRawState *s = aiocb->bs->opaque; |
a6dcf097 | 1104 | #endif |
37cc9f7f DL |
1105 | |
1106 | if (aiocb->aio_type & QEMU_AIO_BLKDEV) { | |
1107 | return handle_aiocb_write_zeroes_block(aiocb); | |
1108 | } | |
1109 | ||
1110 | #ifdef CONFIG_XFS | |
1111 | if (s->is_xfs) { | |
1112 | return xfs_write_zeroes(s, aiocb->aio_offset, aiocb->aio_nbytes); | |
1113 | } | |
1114 | #endif | |
1115 | ||
b953f075 DL |
1116 | #ifdef CONFIG_FALLOCATE_ZERO_RANGE |
1117 | if (s->has_write_zeroes) { | |
1118 | int ret = do_fallocate(s->fd, FALLOC_FL_ZERO_RANGE, | |
1119 | aiocb->aio_offset, aiocb->aio_nbytes); | |
1120 | if (ret == 0 || ret != -ENOTSUP) { | |
1121 | return ret; | |
1122 | } | |
1123 | s->has_write_zeroes = false; | |
1124 | } | |
1125 | #endif | |
1126 | ||
1cdc3239 DL |
1127 | #ifdef CONFIG_FALLOCATE_PUNCH_HOLE |
1128 | if (s->has_discard && s->has_fallocate) { | |
1129 | int ret = do_fallocate(s->fd, | |
1130 | FALLOC_FL_PUNCH_HOLE | FALLOC_FL_KEEP_SIZE, | |
1131 | aiocb->aio_offset, aiocb->aio_nbytes); | |
1132 | if (ret == 0) { | |
1133 | ret = do_fallocate(s->fd, 0, aiocb->aio_offset, aiocb->aio_nbytes); | |
1134 | if (ret == 0 || ret != -ENOTSUP) { | |
1135 | return ret; | |
1136 | } | |
1137 | s->has_fallocate = false; | |
1138 | } else if (ret != -ENOTSUP) { | |
1139 | return ret; | |
1140 | } else { | |
1141 | s->has_discard = false; | |
1142 | } | |
1143 | } | |
1144 | #endif | |
1145 | ||
d50d8222 DL |
1146 | #ifdef CONFIG_FALLOCATE |
1147 | if (s->has_fallocate && aiocb->aio_offset >= bdrv_getlength(aiocb->bs)) { | |
1148 | int ret = do_fallocate(s->fd, 0, aiocb->aio_offset, aiocb->aio_nbytes); | |
1149 | if (ret == 0 || ret != -ENOTSUP) { | |
1150 | return ret; | |
1151 | } | |
1152 | s->has_fallocate = false; | |
1153 | } | |
1154 | #endif | |
1155 | ||
37cc9f7f DL |
1156 | return -ENOTSUP; |
1157 | } | |
1158 | ||
8238010b PB |
1159 | static ssize_t handle_aiocb_discard(RawPosixAIOData *aiocb) |
1160 | { | |
1161 | int ret = -EOPNOTSUPP; | |
1162 | BDRVRawState *s = aiocb->bs->opaque; | |
1163 | ||
7ce21016 PB |
1164 | if (!s->has_discard) { |
1165 | return -ENOTSUP; | |
8238010b PB |
1166 | } |
1167 | ||
1168 | if (aiocb->aio_type & QEMU_AIO_BLKDEV) { | |
1169 | #ifdef BLKDISCARD | |
1170 | do { | |
1171 | uint64_t range[2] = { aiocb->aio_offset, aiocb->aio_nbytes }; | |
1172 | if (ioctl(aiocb->aio_fildes, BLKDISCARD, range) == 0) { | |
1173 | return 0; | |
1174 | } | |
1175 | } while (errno == EINTR); | |
1176 | ||
1177 | ret = -errno; | |
1178 | #endif | |
1179 | } else { | |
1180 | #ifdef CONFIG_XFS | |
1181 | if (s->is_xfs) { | |
1182 | return xfs_discard(s, aiocb->aio_offset, aiocb->aio_nbytes); | |
1183 | } | |
1184 | #endif | |
1185 | ||
1186 | #ifdef CONFIG_FALLOCATE_PUNCH_HOLE | |
0b991712 DL |
1187 | ret = do_fallocate(s->fd, FALLOC_FL_PUNCH_HOLE | FALLOC_FL_KEEP_SIZE, |
1188 | aiocb->aio_offset, aiocb->aio_nbytes); | |
8238010b PB |
1189 | #endif |
1190 | } | |
1191 | ||
1486df0e DL |
1192 | ret = translate_err(ret); |
1193 | if (ret == -ENOTSUP) { | |
7ce21016 | 1194 | s->has_discard = false; |
8238010b PB |
1195 | } |
1196 | return ret; | |
1197 | } | |
1198 | ||
de81a169 PB |
1199 | static int aio_worker(void *arg) |
1200 | { | |
1201 | RawPosixAIOData *aiocb = arg; | |
1202 | ssize_t ret = 0; | |
1203 | ||
1204 | switch (aiocb->aio_type & QEMU_AIO_TYPE_MASK) { | |
1205 | case QEMU_AIO_READ: | |
1206 | ret = handle_aiocb_rw(aiocb); | |
c0191e76 | 1207 | if (ret >= 0 && ret < aiocb->aio_nbytes) { |
de81a169 PB |
1208 | iov_memset(aiocb->aio_iov, aiocb->aio_niov, ret, |
1209 | 0, aiocb->aio_nbytes - ret); | |
1210 | ||
1211 | ret = aiocb->aio_nbytes; | |
1212 | } | |
1213 | if (ret == aiocb->aio_nbytes) { | |
1214 | ret = 0; | |
1215 | } else if (ret >= 0 && ret < aiocb->aio_nbytes) { | |
1216 | ret = -EINVAL; | |
1217 | } | |
1218 | break; | |
1219 | case QEMU_AIO_WRITE: | |
1220 | ret = handle_aiocb_rw(aiocb); | |
1221 | if (ret == aiocb->aio_nbytes) { | |
1222 | ret = 0; | |
1223 | } else if (ret >= 0 && ret < aiocb->aio_nbytes) { | |
1224 | ret = -EINVAL; | |
1225 | } | |
1226 | break; | |
1227 | case QEMU_AIO_FLUSH: | |
1228 | ret = handle_aiocb_flush(aiocb); | |
1229 | break; | |
1230 | case QEMU_AIO_IOCTL: | |
1231 | ret = handle_aiocb_ioctl(aiocb); | |
1232 | break; | |
8238010b PB |
1233 | case QEMU_AIO_DISCARD: |
1234 | ret = handle_aiocb_discard(aiocb); | |
1235 | break; | |
97a2ae34 PB |
1236 | case QEMU_AIO_WRITE_ZEROES: |
1237 | ret = handle_aiocb_write_zeroes(aiocb); | |
1238 | break; | |
de81a169 PB |
1239 | default: |
1240 | fprintf(stderr, "invalid aio request (0x%x)\n", aiocb->aio_type); | |
1241 | ret = -EINVAL; | |
1242 | break; | |
1243 | } | |
1244 | ||
1245 | g_slice_free(RawPosixAIOData, aiocb); | |
1246 | return ret; | |
1247 | } | |
1248 | ||
260a82e5 PB |
1249 | static int paio_submit_co(BlockDriverState *bs, int fd, |
1250 | int64_t sector_num, QEMUIOVector *qiov, int nb_sectors, | |
1251 | int type) | |
1252 | { | |
1253 | RawPosixAIOData *acb = g_slice_new(RawPosixAIOData); | |
1254 | ThreadPool *pool; | |
1255 | ||
1256 | acb->bs = bs; | |
1257 | acb->aio_type = type; | |
1258 | acb->aio_fildes = fd; | |
1259 | ||
8eb029c2 KW |
1260 | acb->aio_nbytes = nb_sectors * BDRV_SECTOR_SIZE; |
1261 | acb->aio_offset = sector_num * BDRV_SECTOR_SIZE; | |
1262 | ||
260a82e5 PB |
1263 | if (qiov) { |
1264 | acb->aio_iov = qiov->iov; | |
1265 | acb->aio_niov = qiov->niov; | |
8eb029c2 | 1266 | assert(qiov->size == acb->aio_nbytes); |
260a82e5 | 1267 | } |
260a82e5 PB |
1268 | |
1269 | trace_paio_submit_co(sector_num, nb_sectors, type); | |
1270 | pool = aio_get_thread_pool(bdrv_get_aio_context(bs)); | |
1271 | return thread_pool_submit_co(pool, aio_worker, acb); | |
1272 | } | |
1273 | ||
7c84b1b8 | 1274 | static BlockAIOCB *paio_submit(BlockDriverState *bs, int fd, |
de81a169 | 1275 | int64_t sector_num, QEMUIOVector *qiov, int nb_sectors, |
097310b5 | 1276 | BlockCompletionFunc *cb, void *opaque, int type) |
de81a169 PB |
1277 | { |
1278 | RawPosixAIOData *acb = g_slice_new(RawPosixAIOData); | |
c4d9d196 | 1279 | ThreadPool *pool; |
de81a169 PB |
1280 | |
1281 | acb->bs = bs; | |
1282 | acb->aio_type = type; | |
1283 | acb->aio_fildes = fd; | |
1284 | ||
8eb029c2 KW |
1285 | acb->aio_nbytes = nb_sectors * BDRV_SECTOR_SIZE; |
1286 | acb->aio_offset = sector_num * BDRV_SECTOR_SIZE; | |
1287 | ||
de81a169 PB |
1288 | if (qiov) { |
1289 | acb->aio_iov = qiov->iov; | |
1290 | acb->aio_niov = qiov->niov; | |
8eb029c2 | 1291 | assert(qiov->size == acb->aio_nbytes); |
de81a169 | 1292 | } |
de81a169 PB |
1293 | |
1294 | trace_paio_submit(acb, opaque, sector_num, nb_sectors, type); | |
c4d9d196 SH |
1295 | pool = aio_get_thread_pool(bdrv_get_aio_context(bs)); |
1296 | return thread_pool_submit_aio(pool, aio_worker, acb, cb, opaque); | |
de81a169 PB |
1297 | } |
1298 | ||
7c84b1b8 | 1299 | static BlockAIOCB *raw_aio_submit(BlockDriverState *bs, |
9ef91a67 | 1300 | int64_t sector_num, QEMUIOVector *qiov, int nb_sectors, |
097310b5 | 1301 | BlockCompletionFunc *cb, void *opaque, int type) |
83f64091 | 1302 | { |
ce1a14dc | 1303 | BDRVRawState *s = bs->opaque; |
ce1a14dc | 1304 | |
19cb3738 FB |
1305 | if (fd_open(bs) < 0) |
1306 | return NULL; | |
1307 | ||
f141eafe | 1308 | /* |
3cad8307 RPM |
1309 | * Check if the underlying device requires requests to be aligned, |
1310 | * and if the request we are trying to submit is aligned or not. | |
1311 | * If this is the case tell the low-level driver that it needs | |
1312 | * to copy the buffer. | |
f141eafe | 1313 | */ |
3cad8307 | 1314 | if (s->needs_alignment) { |
c53b1c51 | 1315 | if (!bdrv_qiov_is_aligned(bs, qiov)) { |
5c6c3a6c | 1316 | type |= QEMU_AIO_MISALIGNED; |
e44bd6fc | 1317 | #ifdef CONFIG_LINUX_AIO |
5c6c3a6c CH |
1318 | } else if (s->use_aio) { |
1319 | return laio_submit(bs, s->aio_ctx, s->fd, sector_num, qiov, | |
e44bd6fc SW |
1320 | nb_sectors, cb, opaque, type); |
1321 | #endif | |
5c6c3a6c | 1322 | } |
9ef91a67 | 1323 | } |
f141eafe | 1324 | |
1e5b9d2f | 1325 | return paio_submit(bs, s->fd, sector_num, qiov, nb_sectors, |
9ef91a67 | 1326 | cb, opaque, type); |
83f64091 FB |
1327 | } |
1328 | ||
1b3abdcc ML |
1329 | static void raw_aio_plug(BlockDriverState *bs) |
1330 | { | |
1331 | #ifdef CONFIG_LINUX_AIO | |
1332 | BDRVRawState *s = bs->opaque; | |
1333 | if (s->use_aio) { | |
1334 | laio_io_plug(bs, s->aio_ctx); | |
1335 | } | |
1336 | #endif | |
1337 | } | |
1338 | ||
1339 | static void raw_aio_unplug(BlockDriverState *bs) | |
1340 | { | |
1341 | #ifdef CONFIG_LINUX_AIO | |
1342 | BDRVRawState *s = bs->opaque; | |
1343 | if (s->use_aio) { | |
1344 | laio_io_unplug(bs, s->aio_ctx, true); | |
1345 | } | |
1346 | #endif | |
1347 | } | |
1348 | ||
1349 | static void raw_aio_flush_io_queue(BlockDriverState *bs) | |
1350 | { | |
1351 | #ifdef CONFIG_LINUX_AIO | |
1352 | BDRVRawState *s = bs->opaque; | |
1353 | if (s->use_aio) { | |
1354 | laio_io_unplug(bs, s->aio_ctx, false); | |
1355 | } | |
1356 | #endif | |
1357 | } | |
1358 | ||
7c84b1b8 | 1359 | static BlockAIOCB *raw_aio_readv(BlockDriverState *bs, |
f141eafe | 1360 | int64_t sector_num, QEMUIOVector *qiov, int nb_sectors, |
097310b5 | 1361 | BlockCompletionFunc *cb, void *opaque) |
83f64091 | 1362 | { |
9ef91a67 CH |
1363 | return raw_aio_submit(bs, sector_num, qiov, nb_sectors, |
1364 | cb, opaque, QEMU_AIO_READ); | |
83f64091 FB |
1365 | } |
1366 | ||
7c84b1b8 | 1367 | static BlockAIOCB *raw_aio_writev(BlockDriverState *bs, |
f141eafe | 1368 | int64_t sector_num, QEMUIOVector *qiov, int nb_sectors, |
097310b5 | 1369 | BlockCompletionFunc *cb, void *opaque) |
83f64091 | 1370 | { |
9ef91a67 CH |
1371 | return raw_aio_submit(bs, sector_num, qiov, nb_sectors, |
1372 | cb, opaque, QEMU_AIO_WRITE); | |
83f64091 | 1373 | } |
53538725 | 1374 | |
7c84b1b8 | 1375 | static BlockAIOCB *raw_aio_flush(BlockDriverState *bs, |
097310b5 | 1376 | BlockCompletionFunc *cb, void *opaque) |
b2e12bc6 CH |
1377 | { |
1378 | BDRVRawState *s = bs->opaque; | |
1379 | ||
1380 | if (fd_open(bs) < 0) | |
1381 | return NULL; | |
1382 | ||
1e5b9d2f | 1383 | return paio_submit(bs, s->fd, 0, NULL, 0, cb, opaque, QEMU_AIO_FLUSH); |
b2e12bc6 CH |
1384 | } |
1385 | ||
83f64091 FB |
1386 | static void raw_close(BlockDriverState *bs) |
1387 | { | |
1388 | BDRVRawState *s = bs->opaque; | |
c2f3426c SH |
1389 | |
1390 | raw_detach_aio_context(bs); | |
1391 | ||
abd269b7 SH |
1392 | #ifdef CONFIG_LINUX_AIO |
1393 | if (s->use_aio) { | |
1394 | laio_cleanup(s->aio_ctx); | |
1395 | } | |
1396 | #endif | |
19cb3738 | 1397 | if (s->fd >= 0) { |
2e1e79da | 1398 | qemu_close(s->fd); |
19cb3738 FB |
1399 | s->fd = -1; |
1400 | } | |
83f64091 FB |
1401 | } |
1402 | ||
1403 | static int raw_truncate(BlockDriverState *bs, int64_t offset) | |
1404 | { | |
1405 | BDRVRawState *s = bs->opaque; | |
55b949c8 CH |
1406 | struct stat st; |
1407 | ||
1408 | if (fstat(s->fd, &st)) { | |
83f64091 | 1409 | return -errno; |
55b949c8 CH |
1410 | } |
1411 | ||
1412 | if (S_ISREG(st.st_mode)) { | |
1413 | if (ftruncate(s->fd, offset) < 0) { | |
1414 | return -errno; | |
1415 | } | |
1416 | } else if (S_ISCHR(st.st_mode) || S_ISBLK(st.st_mode)) { | |
1417 | if (offset > raw_getlength(bs)) { | |
1418 | return -EINVAL; | |
1419 | } | |
1420 | } else { | |
1421 | return -ENOTSUP; | |
1422 | } | |
1423 | ||
83f64091 FB |
1424 | return 0; |
1425 | } | |
1426 | ||
128ab2ff BS |
1427 | #ifdef __OpenBSD__ |
1428 | static int64_t raw_getlength(BlockDriverState *bs) | |
1429 | { | |
1430 | BDRVRawState *s = bs->opaque; | |
1431 | int fd = s->fd; | |
1432 | struct stat st; | |
1433 | ||
1434 | if (fstat(fd, &st)) | |
aa729704 | 1435 | return -errno; |
128ab2ff BS |
1436 | if (S_ISCHR(st.st_mode) || S_ISBLK(st.st_mode)) { |
1437 | struct disklabel dl; | |
1438 | ||
1439 | if (ioctl(fd, DIOCGDINFO, &dl)) | |
aa729704 | 1440 | return -errno; |
128ab2ff BS |
1441 | return (uint64_t)dl.d_secsize * |
1442 | dl.d_partitions[DISKPART(st.st_rdev)].p_size; | |
1443 | } else | |
1444 | return st.st_size; | |
1445 | } | |
d1f6fd8d CE |
1446 | #elif defined(__NetBSD__) |
1447 | static int64_t raw_getlength(BlockDriverState *bs) | |
1448 | { | |
1449 | BDRVRawState *s = bs->opaque; | |
1450 | int fd = s->fd; | |
1451 | struct stat st; | |
1452 | ||
1453 | if (fstat(fd, &st)) | |
aa729704 | 1454 | return -errno; |
d1f6fd8d CE |
1455 | if (S_ISCHR(st.st_mode) || S_ISBLK(st.st_mode)) { |
1456 | struct dkwedge_info dkw; | |
1457 | ||
1458 | if (ioctl(fd, DIOCGWEDGEINFO, &dkw) != -1) { | |
1459 | return dkw.dkw_size * 512; | |
1460 | } else { | |
1461 | struct disklabel dl; | |
1462 | ||
1463 | if (ioctl(fd, DIOCGDINFO, &dl)) | |
aa729704 | 1464 | return -errno; |
d1f6fd8d CE |
1465 | return (uint64_t)dl.d_secsize * |
1466 | dl.d_partitions[DISKPART(st.st_rdev)].p_size; | |
1467 | } | |
1468 | } else | |
1469 | return st.st_size; | |
1470 | } | |
50779cc2 CH |
1471 | #elif defined(__sun__) |
1472 | static int64_t raw_getlength(BlockDriverState *bs) | |
1473 | { | |
1474 | BDRVRawState *s = bs->opaque; | |
1475 | struct dk_minfo minfo; | |
1476 | int ret; | |
aa729704 | 1477 | int64_t size; |
50779cc2 CH |
1478 | |
1479 | ret = fd_open(bs); | |
1480 | if (ret < 0) { | |
1481 | return ret; | |
1482 | } | |
1483 | ||
1484 | /* | |
1485 | * Use the DKIOCGMEDIAINFO ioctl to read the size. | |
1486 | */ | |
1487 | ret = ioctl(s->fd, DKIOCGMEDIAINFO, &minfo); | |
1488 | if (ret != -1) { | |
1489 | return minfo.dki_lbsize * minfo.dki_capacity; | |
1490 | } | |
1491 | ||
1492 | /* | |
1493 | * There are reports that lseek on some devices fails, but | |
1494 | * irc discussion said that contingency on contingency was overkill. | |
1495 | */ | |
aa729704 MA |
1496 | size = lseek(s->fd, 0, SEEK_END); |
1497 | if (size < 0) { | |
1498 | return -errno; | |
1499 | } | |
1500 | return size; | |
50779cc2 CH |
1501 | } |
1502 | #elif defined(CONFIG_BSD) | |
1503 | static int64_t raw_getlength(BlockDriverState *bs) | |
83f64091 FB |
1504 | { |
1505 | BDRVRawState *s = bs->opaque; | |
1506 | int fd = s->fd; | |
1507 | int64_t size; | |
83f64091 | 1508 | struct stat sb; |
a167ba50 | 1509 | #if defined (__FreeBSD__) || defined(__FreeBSD_kernel__) |
9f23011a | 1510 | int reopened = 0; |
83f64091 | 1511 | #endif |
19cb3738 FB |
1512 | int ret; |
1513 | ||
1514 | ret = fd_open(bs); | |
1515 | if (ret < 0) | |
1516 | return ret; | |
83f64091 | 1517 | |
a167ba50 | 1518 | #if defined (__FreeBSD__) || defined(__FreeBSD_kernel__) |
9f23011a BS |
1519 | again: |
1520 | #endif | |
83f64091 FB |
1521 | if (!fstat(fd, &sb) && (S_IFCHR & sb.st_mode)) { |
1522 | #ifdef DIOCGMEDIASIZE | |
1523 | if (ioctl(fd, DIOCGMEDIASIZE, (off_t *)&size)) | |
c5e97233 BS |
1524 | #elif defined(DIOCGPART) |
1525 | { | |
1526 | struct partinfo pi; | |
1527 | if (ioctl(fd, DIOCGPART, &pi) == 0) | |
1528 | size = pi.media_size; | |
1529 | else | |
1530 | size = 0; | |
1531 | } | |
1532 | if (size == 0) | |
83f64091 | 1533 | #endif |
83affaa6 | 1534 | #if defined(__APPLE__) && defined(__MACH__) |
728dacbd JA |
1535 | { |
1536 | uint64_t sectors = 0; | |
1537 | uint32_t sector_size = 0; | |
1538 | ||
1539 | if (ioctl(fd, DKIOCGETBLOCKCOUNT, §ors) == 0 | |
1540 | && ioctl(fd, DKIOCGETBLOCKSIZE, §or_size) == 0) { | |
1541 | size = sectors * sector_size; | |
1542 | } else { | |
1543 | size = lseek(fd, 0LL, SEEK_END); | |
1544 | if (size < 0) { | |
1545 | return -errno; | |
1546 | } | |
1547 | } | |
1548 | } | |
83f64091 FB |
1549 | #else |
1550 | size = lseek(fd, 0LL, SEEK_END); | |
aa729704 MA |
1551 | if (size < 0) { |
1552 | return -errno; | |
1553 | } | |
9f23011a | 1554 | #endif |
a167ba50 | 1555 | #if defined(__FreeBSD__) || defined(__FreeBSD_kernel__) |
9f23011a BS |
1556 | switch(s->type) { |
1557 | case FTYPE_CD: | |
1558 | /* XXX FreeBSD acd returns UINT_MAX sectors for an empty drive */ | |
1559 | if (size == 2048LL * (unsigned)-1) | |
1560 | size = 0; | |
1561 | /* XXX no disc? maybe we need to reopen... */ | |
f3a5d3f8 | 1562 | if (size <= 0 && !reopened && cdrom_reopen(bs) >= 0) { |
9f23011a BS |
1563 | reopened = 1; |
1564 | goto again; | |
1565 | } | |
1566 | } | |
83f64091 | 1567 | #endif |
50779cc2 | 1568 | } else { |
83f64091 | 1569 | size = lseek(fd, 0, SEEK_END); |
aa729704 MA |
1570 | if (size < 0) { |
1571 | return -errno; | |
1572 | } | |
83f64091 | 1573 | } |
83f64091 FB |
1574 | return size; |
1575 | } | |
50779cc2 CH |
1576 | #else |
1577 | static int64_t raw_getlength(BlockDriverState *bs) | |
1578 | { | |
1579 | BDRVRawState *s = bs->opaque; | |
1580 | int ret; | |
aa729704 | 1581 | int64_t size; |
50779cc2 CH |
1582 | |
1583 | ret = fd_open(bs); | |
1584 | if (ret < 0) { | |
1585 | return ret; | |
1586 | } | |
1587 | ||
aa729704 MA |
1588 | size = lseek(s->fd, 0, SEEK_END); |
1589 | if (size < 0) { | |
1590 | return -errno; | |
1591 | } | |
1592 | return size; | |
50779cc2 | 1593 | } |
128ab2ff | 1594 | #endif |
83f64091 | 1595 | |
4a1d5e1f FZ |
1596 | static int64_t raw_get_allocated_file_size(BlockDriverState *bs) |
1597 | { | |
1598 | struct stat st; | |
1599 | BDRVRawState *s = bs->opaque; | |
1600 | ||
1601 | if (fstat(s->fd, &st) < 0) { | |
1602 | return -errno; | |
1603 | } | |
1604 | return (int64_t)st.st_blocks * 512; | |
1605 | } | |
1606 | ||
6f482f74 | 1607 | static int raw_create(const char *filename, QemuOpts *opts, Error **errp) |
83f64091 FB |
1608 | { |
1609 | int fd; | |
1e37d059 | 1610 | int result = 0; |
0e7e1989 | 1611 | int64_t total_size = 0; |
4ab15590 | 1612 | bool nocow = false; |
06247428 HT |
1613 | PreallocMode prealloc; |
1614 | char *buf = NULL; | |
1615 | Error *local_err = NULL; | |
83f64091 | 1616 | |
464d9f64 HR |
1617 | strstart(filename, "file:", &filename); |
1618 | ||
0e7e1989 | 1619 | /* Read out options */ |
180e9526 HT |
1620 | total_size = ROUND_UP(qemu_opt_get_size_del(opts, BLOCK_OPT_SIZE, 0), |
1621 | BDRV_SECTOR_SIZE); | |
4ab15590 | 1622 | nocow = qemu_opt_get_bool(opts, BLOCK_OPT_NOCOW, false); |
06247428 HT |
1623 | buf = qemu_opt_get_del(opts, BLOCK_OPT_PREALLOC); |
1624 | prealloc = qapi_enum_parse(PreallocMode_lookup, buf, | |
1625 | PREALLOC_MODE_MAX, PREALLOC_MODE_OFF, | |
1626 | &local_err); | |
1627 | g_free(buf); | |
1628 | if (local_err) { | |
1629 | error_propagate(errp, local_err); | |
1630 | result = -EINVAL; | |
1631 | goto out; | |
1632 | } | |
83f64091 | 1633 | |
6165f4d8 CB |
1634 | fd = qemu_open(filename, O_WRONLY | O_CREAT | O_TRUNC | O_BINARY, |
1635 | 0644); | |
1e37d059 SW |
1636 | if (fd < 0) { |
1637 | result = -errno; | |
e428e439 | 1638 | error_setg_errno(errp, -result, "Could not create file"); |
06247428 HT |
1639 | goto out; |
1640 | } | |
1641 | ||
1642 | if (nocow) { | |
4ab15590 | 1643 | #ifdef __linux__ |
06247428 HT |
1644 | /* Set NOCOW flag to solve performance issue on fs like btrfs. |
1645 | * This is an optimisation. The FS_IOC_SETFLAGS ioctl return value | |
1646 | * will be ignored since any failure of this operation should not | |
1647 | * block the left work. | |
1648 | */ | |
1649 | int attr; | |
1650 | if (ioctl(fd, FS_IOC_GETFLAGS, &attr) == 0) { | |
1651 | attr |= FS_NOCOW_FL; | |
1652 | ioctl(fd, FS_IOC_SETFLAGS, &attr); | |
4ab15590 | 1653 | } |
06247428 HT |
1654 | #endif |
1655 | } | |
1656 | ||
1657 | if (ftruncate(fd, total_size) != 0) { | |
1658 | result = -errno; | |
1659 | error_setg_errno(errp, -result, "Could not resize file"); | |
1660 | goto out_close; | |
1661 | } | |
4ab15590 | 1662 | |
ed911435 KW |
1663 | switch (prealloc) { |
1664 | #ifdef CONFIG_POSIX_FALLOCATE | |
1665 | case PREALLOC_MODE_FALLOC: | |
06247428 HT |
1666 | /* posix_fallocate() doesn't set errno. */ |
1667 | result = -posix_fallocate(fd, 0, total_size); | |
1668 | if (result != 0) { | |
1669 | error_setg_errno(errp, -result, | |
1670 | "Could not preallocate data for the new file"); | |
1e37d059 | 1671 | } |
ed911435 KW |
1672 | break; |
1673 | #endif | |
1674 | case PREALLOC_MODE_FULL: | |
1675 | { | |
06247428 | 1676 | int64_t num = 0, left = total_size; |
ed911435 | 1677 | buf = g_malloc0(65536); |
06247428 HT |
1678 | |
1679 | while (left > 0) { | |
1680 | num = MIN(left, 65536); | |
1681 | result = write(fd, buf, num); | |
1682 | if (result < 0) { | |
1683 | result = -errno; | |
1684 | error_setg_errno(errp, -result, | |
1685 | "Could not write to the new file"); | |
1686 | break; | |
1687 | } | |
39411cf3 | 1688 | left -= result; |
1e37d059 | 1689 | } |
731de380 | 1690 | if (result >= 0) { |
098ffa66 HR |
1691 | result = fsync(fd); |
1692 | if (result < 0) { | |
1693 | result = -errno; | |
1694 | error_setg_errno(errp, -result, | |
1695 | "Could not flush new file to disk"); | |
1696 | } | |
731de380 | 1697 | } |
06247428 | 1698 | g_free(buf); |
ed911435 KW |
1699 | break; |
1700 | } | |
1701 | case PREALLOC_MODE_OFF: | |
1702 | break; | |
1703 | default: | |
06247428 HT |
1704 | result = -EINVAL; |
1705 | error_setg(errp, "Unsupported preallocation mode: %s", | |
1706 | PreallocMode_lookup[prealloc]); | |
ed911435 | 1707 | break; |
1e37d059 | 1708 | } |
06247428 HT |
1709 | |
1710 | out_close: | |
1711 | if (qemu_close(fd) != 0 && result == 0) { | |
1712 | result = -errno; | |
1713 | error_setg_errno(errp, -result, "Could not close the new file"); | |
1714 | } | |
1715 | out: | |
1e37d059 | 1716 | return result; |
83f64091 FB |
1717 | } |
1718 | ||
d1f06fe6 MA |
1719 | /* |
1720 | * Find allocation range in @bs around offset @start. | |
1721 | * May change underlying file descriptor's file offset. | |
1722 | * If @start is not in a hole, store @start in @data, and the | |
1723 | * beginning of the next hole in @hole, and return 0. | |
1724 | * If @start is in a non-trailing hole, store @start in @hole and the | |
1725 | * beginning of the next non-hole in @data, and return 0. | |
1726 | * If @start is in a trailing hole or beyond EOF, return -ENXIO. | |
1727 | * If we can't find out, return a negative errno other than -ENXIO. | |
1728 | */ | |
1729 | static int find_allocation(BlockDriverState *bs, off_t start, | |
1730 | off_t *data, off_t *hole) | |
4f11aa8a HR |
1731 | { |
1732 | #if defined SEEK_HOLE && defined SEEK_DATA | |
94282e71 | 1733 | BDRVRawState *s = bs->opaque; |
d1f06fe6 | 1734 | off_t offs; |
94282e71 | 1735 | |
d1f06fe6 MA |
1736 | /* |
1737 | * SEEK_DATA cases: | |
1738 | * D1. offs == start: start is in data | |
1739 | * D2. offs > start: start is in a hole, next data at offs | |
1740 | * D3. offs < 0, errno = ENXIO: either start is in a trailing hole | |
1741 | * or start is beyond EOF | |
1742 | * If the latter happens, the file has been truncated behind | |
1743 | * our back since we opened it. All bets are off then. | |
1744 | * Treating like a trailing hole is simplest. | |
1745 | * D4. offs < 0, errno != ENXIO: we learned nothing | |
1746 | */ | |
1747 | offs = lseek(s->fd, start, SEEK_DATA); | |
1748 | if (offs < 0) { | |
1749 | return -errno; /* D3 or D4 */ | |
1750 | } | |
1751 | assert(offs >= start); | |
1752 | ||
1753 | if (offs > start) { | |
1754 | /* D2: in hole, next data at offs */ | |
1755 | *hole = start; | |
1756 | *data = offs; | |
1757 | return 0; | |
5500316d PB |
1758 | } |
1759 | ||
d1f06fe6 MA |
1760 | /* D1: in data, end not yet known */ |
1761 | ||
1762 | /* | |
1763 | * SEEK_HOLE cases: | |
1764 | * H1. offs == start: start is in a hole | |
1765 | * If this happens here, a hole has been dug behind our back | |
1766 | * since the previous lseek(). | |
1767 | * H2. offs > start: either start is in data, next hole at offs, | |
1768 | * or start is in trailing hole, EOF at offs | |
1769 | * Linux treats trailing holes like any other hole: offs == | |
1770 | * start. Solaris seeks to EOF instead: offs > start (blech). | |
1771 | * If that happens here, a hole has been dug behind our back | |
1772 | * since the previous lseek(). | |
1773 | * H3. offs < 0, errno = ENXIO: start is beyond EOF | |
1774 | * If this happens, the file has been truncated behind our | |
1775 | * back since we opened it. Treat it like a trailing hole. | |
1776 | * H4. offs < 0, errno != ENXIO: we learned nothing | |
1777 | * Pretend we know nothing at all, i.e. "forget" about D1. | |
1778 | */ | |
1779 | offs = lseek(s->fd, start, SEEK_HOLE); | |
1780 | if (offs < 0) { | |
1781 | return -errno; /* D1 and (H3 or H4) */ | |
1782 | } | |
1783 | assert(offs >= start); | |
1784 | ||
1785 | if (offs > start) { | |
1786 | /* | |
1787 | * D1 and H2: either in data, next hole at offs, or it was in | |
1788 | * data but is now in a trailing hole. In the latter case, | |
1789 | * all bets are off. Treating it as if it there was data all | |
1790 | * the way to EOF is safe, so simply do that. | |
1791 | */ | |
4f11aa8a | 1792 | *data = start; |
d1f06fe6 MA |
1793 | *hole = offs; |
1794 | return 0; | |
5500316d | 1795 | } |
4f11aa8a | 1796 | |
d1f06fe6 MA |
1797 | /* D1 and H1 */ |
1798 | return -EBUSY; | |
5500316d | 1799 | #else |
4f11aa8a | 1800 | return -ENOTSUP; |
5500316d | 1801 | #endif |
4f11aa8a HR |
1802 | } |
1803 | ||
1804 | /* | |
be2ebc6d | 1805 | * Returns the allocation status of the specified sectors. |
4f11aa8a HR |
1806 | * |
1807 | * If 'sector_num' is beyond the end of the disk image the return value is 0 | |
1808 | * and 'pnum' is set to 0. | |
1809 | * | |
1810 | * 'pnum' is set to the number of sectors (including and immediately following | |
1811 | * the specified sector) that are known to be in the same | |
1812 | * allocated/unallocated state. | |
1813 | * | |
1814 | * 'nb_sectors' is the max value 'pnum' should be set to. If nb_sectors goes | |
1815 | * beyond the end of the disk image it will be clamped. | |
1816 | */ | |
1817 | static int64_t coroutine_fn raw_co_get_block_status(BlockDriverState *bs, | |
1818 | int64_t sector_num, | |
1819 | int nb_sectors, int *pnum) | |
1820 | { | |
1821 | off_t start, data = 0, hole = 0; | |
e6d7ec32 | 1822 | int64_t total_size; |
d7f62751 | 1823 | int ret; |
4f11aa8a HR |
1824 | |
1825 | ret = fd_open(bs); | |
1826 | if (ret < 0) { | |
1827 | return ret; | |
1828 | } | |
1829 | ||
1830 | start = sector_num * BDRV_SECTOR_SIZE; | |
e6d7ec32 HR |
1831 | total_size = bdrv_getlength(bs); |
1832 | if (total_size < 0) { | |
1833 | return total_size; | |
1834 | } else if (start >= total_size) { | |
1835 | *pnum = 0; | |
1836 | return 0; | |
1837 | } else if (start + nb_sectors * BDRV_SECTOR_SIZE > total_size) { | |
1838 | nb_sectors = DIV_ROUND_UP(total_size - start, BDRV_SECTOR_SIZE); | |
1839 | } | |
4f11aa8a | 1840 | |
d1f06fe6 MA |
1841 | ret = find_allocation(bs, start, &data, &hole); |
1842 | if (ret == -ENXIO) { | |
1843 | /* Trailing hole */ | |
1844 | *pnum = nb_sectors; | |
1845 | ret = BDRV_BLOCK_ZERO; | |
1846 | } else if (ret < 0) { | |
1847 | /* No info available, so pretend there are no holes */ | |
1848 | *pnum = nb_sectors; | |
1849 | ret = BDRV_BLOCK_DATA; | |
1850 | } else if (data == start) { | |
f4a769ab KW |
1851 | /* On a data extent, compute sectors to the end of the extent, |
1852 | * possibly including a partial sector at EOF. */ | |
1853 | *pnum = MIN(nb_sectors, DIV_ROUND_UP(hole - start, BDRV_SECTOR_SIZE)); | |
d1f06fe6 | 1854 | ret = BDRV_BLOCK_DATA; |
5500316d PB |
1855 | } else { |
1856 | /* On a hole, compute sectors to the beginning of the next extent. */ | |
d1f06fe6 | 1857 | assert(hole == start); |
5500316d | 1858 | *pnum = MIN(nb_sectors, (data - start) / BDRV_SECTOR_SIZE); |
d1f06fe6 | 1859 | ret = BDRV_BLOCK_ZERO; |
5500316d | 1860 | } |
d1f06fe6 | 1861 | return ret | BDRV_BLOCK_OFFSET_VALID | start; |
5500316d PB |
1862 | } |
1863 | ||
7c84b1b8 | 1864 | static coroutine_fn BlockAIOCB *raw_aio_discard(BlockDriverState *bs, |
8238010b | 1865 | int64_t sector_num, int nb_sectors, |
097310b5 | 1866 | BlockCompletionFunc *cb, void *opaque) |
dce512de | 1867 | { |
dce512de CH |
1868 | BDRVRawState *s = bs->opaque; |
1869 | ||
8238010b PB |
1870 | return paio_submit(bs, s->fd, sector_num, NULL, nb_sectors, |
1871 | cb, opaque, QEMU_AIO_DISCARD); | |
dce512de | 1872 | } |
0e7e1989 | 1873 | |
260a82e5 PB |
1874 | static int coroutine_fn raw_co_write_zeroes( |
1875 | BlockDriverState *bs, int64_t sector_num, | |
1876 | int nb_sectors, BdrvRequestFlags flags) | |
1877 | { | |
1878 | BDRVRawState *s = bs->opaque; | |
1879 | ||
1880 | if (!(flags & BDRV_REQ_MAY_UNMAP)) { | |
97a2ae34 PB |
1881 | return paio_submit_co(bs, s->fd, sector_num, NULL, nb_sectors, |
1882 | QEMU_AIO_WRITE_ZEROES); | |
1883 | } else if (s->discard_zeroes) { | |
1884 | return paio_submit_co(bs, s->fd, sector_num, NULL, nb_sectors, | |
1885 | QEMU_AIO_DISCARD); | |
260a82e5 | 1886 | } |
97a2ae34 | 1887 | return -ENOTSUP; |
260a82e5 PB |
1888 | } |
1889 | ||
1890 | static int raw_get_info(BlockDriverState *bs, BlockDriverInfo *bdi) | |
1891 | { | |
1892 | BDRVRawState *s = bs->opaque; | |
1893 | ||
1894 | bdi->unallocated_blocks_are_zero = s->discard_zeroes; | |
1895 | bdi->can_write_zeroes_with_unmap = s->discard_zeroes; | |
1896 | return 0; | |
1897 | } | |
1898 | ||
6f482f74 CL |
1899 | static QemuOptsList raw_create_opts = { |
1900 | .name = "raw-create-opts", | |
1901 | .head = QTAILQ_HEAD_INITIALIZER(raw_create_opts.head), | |
1902 | .desc = { | |
1903 | { | |
1904 | .name = BLOCK_OPT_SIZE, | |
1905 | .type = QEMU_OPT_SIZE, | |
1906 | .help = "Virtual disk size" | |
1907 | }, | |
4ab15590 CL |
1908 | { |
1909 | .name = BLOCK_OPT_NOCOW, | |
1910 | .type = QEMU_OPT_BOOL, | |
1911 | .help = "Turn off copy-on-write (valid only on btrfs)" | |
1912 | }, | |
06247428 HT |
1913 | { |
1914 | .name = BLOCK_OPT_PREALLOC, | |
1915 | .type = QEMU_OPT_STRING, | |
1916 | .help = "Preallocation mode (allowed values: off, falloc, full)" | |
1917 | }, | |
6f482f74 CL |
1918 | { /* end of list */ } |
1919 | } | |
0e7e1989 KW |
1920 | }; |
1921 | ||
5f535a94 | 1922 | BlockDriver bdrv_file = { |
84a12e66 CH |
1923 | .format_name = "file", |
1924 | .protocol_name = "file", | |
856ae5c3 | 1925 | .instance_size = sizeof(BDRVRawState), |
030be321 | 1926 | .bdrv_needs_filename = true, |
856ae5c3 | 1927 | .bdrv_probe = NULL, /* no probe for protocols */ |
078896a9 | 1928 | .bdrv_parse_filename = raw_parse_filename, |
66f82cee | 1929 | .bdrv_file_open = raw_open, |
eeb6b45d JC |
1930 | .bdrv_reopen_prepare = raw_reopen_prepare, |
1931 | .bdrv_reopen_commit = raw_reopen_commit, | |
1932 | .bdrv_reopen_abort = raw_reopen_abort, | |
856ae5c3 | 1933 | .bdrv_close = raw_close, |
c282e1fd | 1934 | .bdrv_create = raw_create, |
3ac21627 | 1935 | .bdrv_has_zero_init = bdrv_has_zero_init_1, |
b6b8a333 | 1936 | .bdrv_co_get_block_status = raw_co_get_block_status, |
260a82e5 | 1937 | .bdrv_co_write_zeroes = raw_co_write_zeroes, |
3b46e624 | 1938 | |
f141eafe AL |
1939 | .bdrv_aio_readv = raw_aio_readv, |
1940 | .bdrv_aio_writev = raw_aio_writev, | |
b2e12bc6 | 1941 | .bdrv_aio_flush = raw_aio_flush, |
8238010b | 1942 | .bdrv_aio_discard = raw_aio_discard, |
c25f53b0 | 1943 | .bdrv_refresh_limits = raw_refresh_limits, |
1b3abdcc ML |
1944 | .bdrv_io_plug = raw_aio_plug, |
1945 | .bdrv_io_unplug = raw_aio_unplug, | |
1946 | .bdrv_flush_io_queue = raw_aio_flush_io_queue, | |
3c529d93 | 1947 | |
83f64091 FB |
1948 | .bdrv_truncate = raw_truncate, |
1949 | .bdrv_getlength = raw_getlength, | |
260a82e5 | 1950 | .bdrv_get_info = raw_get_info, |
4a1d5e1f FZ |
1951 | .bdrv_get_allocated_file_size |
1952 | = raw_get_allocated_file_size, | |
0e7e1989 | 1953 | |
c2f3426c SH |
1954 | .bdrv_detach_aio_context = raw_detach_aio_context, |
1955 | .bdrv_attach_aio_context = raw_attach_aio_context, | |
1956 | ||
6f482f74 | 1957 | .create_opts = &raw_create_opts, |
83f64091 FB |
1958 | }; |
1959 | ||
19cb3738 FB |
1960 | /***********************************************/ |
1961 | /* host device */ | |
1962 | ||
83affaa6 | 1963 | #if defined(__APPLE__) && defined(__MACH__) |
19cb3738 FB |
1964 | static kern_return_t FindEjectableCDMedia( io_iterator_t *mediaIterator ); |
1965 | static kern_return_t GetBSDPath( io_iterator_t mediaIterator, char *bsdPath, CFIndex maxPathSize ); | |
1966 | ||
1967 | kern_return_t FindEjectableCDMedia( io_iterator_t *mediaIterator ) | |
1968 | { | |
5fafdf24 | 1969 | kern_return_t kernResult; |
19cb3738 FB |
1970 | mach_port_t masterPort; |
1971 | CFMutableDictionaryRef classesToMatch; | |
1972 | ||
1973 | kernResult = IOMasterPort( MACH_PORT_NULL, &masterPort ); | |
1974 | if ( KERN_SUCCESS != kernResult ) { | |
1975 | printf( "IOMasterPort returned %d\n", kernResult ); | |
1976 | } | |
3b46e624 | 1977 | |
5fafdf24 | 1978 | classesToMatch = IOServiceMatching( kIOCDMediaClass ); |
19cb3738 FB |
1979 | if ( classesToMatch == NULL ) { |
1980 | printf( "IOServiceMatching returned a NULL dictionary.\n" ); | |
1981 | } else { | |
1982 | CFDictionarySetValue( classesToMatch, CFSTR( kIOMediaEjectableKey ), kCFBooleanTrue ); | |
1983 | } | |
1984 | kernResult = IOServiceGetMatchingServices( masterPort, classesToMatch, mediaIterator ); | |
1985 | if ( KERN_SUCCESS != kernResult ) | |
1986 | { | |
1987 | printf( "IOServiceGetMatchingServices returned %d\n", kernResult ); | |
1988 | } | |
3b46e624 | 1989 | |
19cb3738 FB |
1990 | return kernResult; |
1991 | } | |
1992 | ||
1993 | kern_return_t GetBSDPath( io_iterator_t mediaIterator, char *bsdPath, CFIndex maxPathSize ) | |
1994 | { | |
1995 | io_object_t nextMedia; | |
1996 | kern_return_t kernResult = KERN_FAILURE; | |
1997 | *bsdPath = '\0'; | |
1998 | nextMedia = IOIteratorNext( mediaIterator ); | |
1999 | if ( nextMedia ) | |
2000 | { | |
2001 | CFTypeRef bsdPathAsCFString; | |
2002 | bsdPathAsCFString = IORegistryEntryCreateCFProperty( nextMedia, CFSTR( kIOBSDNameKey ), kCFAllocatorDefault, 0 ); | |
2003 | if ( bsdPathAsCFString ) { | |
2004 | size_t devPathLength; | |
2005 | strcpy( bsdPath, _PATH_DEV ); | |
2006 | strcat( bsdPath, "r" ); | |
2007 | devPathLength = strlen( bsdPath ); | |
2008 | if ( CFStringGetCString( bsdPathAsCFString, bsdPath + devPathLength, maxPathSize - devPathLength, kCFStringEncodingASCII ) ) { | |
2009 | kernResult = KERN_SUCCESS; | |
2010 | } | |
2011 | CFRelease( bsdPathAsCFString ); | |
2012 | } | |
2013 | IOObjectRelease( nextMedia ); | |
2014 | } | |
3b46e624 | 2015 | |
19cb3738 FB |
2016 | return kernResult; |
2017 | } | |
2018 | ||
2019 | #endif | |
2020 | ||
508c7cb3 CH |
2021 | static int hdev_probe_device(const char *filename) |
2022 | { | |
2023 | struct stat st; | |
2024 | ||
2025 | /* allow a dedicated CD-ROM driver to match with a higher priority */ | |
2026 | if (strstart(filename, "/dev/cdrom", NULL)) | |
2027 | return 50; | |
2028 | ||
2029 | if (stat(filename, &st) >= 0 && | |
2030 | (S_ISCHR(st.st_mode) || S_ISBLK(st.st_mode))) { | |
2031 | return 100; | |
2032 | } | |
2033 | ||
2034 | return 0; | |
2035 | } | |
2036 | ||
da888d37 SH |
2037 | static int check_hdev_writable(BDRVRawState *s) |
2038 | { | |
2039 | #if defined(BLKROGET) | |
2040 | /* Linux block devices can be configured "read-only" using blockdev(8). | |
2041 | * This is independent of device node permissions and therefore open(2) | |
2042 | * with O_RDWR succeeds. Actual writes fail with EPERM. | |
2043 | * | |
2044 | * bdrv_open() is supposed to fail if the disk is read-only. Explicitly | |
2045 | * check for read-only block devices so that Linux block devices behave | |
2046 | * properly. | |
2047 | */ | |
2048 | struct stat st; | |
2049 | int readonly = 0; | |
2050 | ||
2051 | if (fstat(s->fd, &st)) { | |
2052 | return -errno; | |
2053 | } | |
2054 | ||
2055 | if (!S_ISBLK(st.st_mode)) { | |
2056 | return 0; | |
2057 | } | |
2058 | ||
2059 | if (ioctl(s->fd, BLKROGET, &readonly) < 0) { | |
2060 | return -errno; | |
2061 | } | |
2062 | ||
2063 | if (readonly) { | |
2064 | return -EACCES; | |
2065 | } | |
2066 | #endif /* defined(BLKROGET) */ | |
2067 | return 0; | |
2068 | } | |
2069 | ||
7af803d4 HR |
2070 | static void hdev_parse_filename(const char *filename, QDict *options, |
2071 | Error **errp) | |
2072 | { | |
2073 | /* The prefix is optional, just as for "file". */ | |
2074 | strstart(filename, "host_device:", &filename); | |
2075 | ||
2076 | qdict_put_obj(options, "filename", QOBJECT(qstring_from_str(filename))); | |
2077 | } | |
2078 | ||
015a1036 HR |
2079 | static int hdev_open(BlockDriverState *bs, QDict *options, int flags, |
2080 | Error **errp) | |
19cb3738 FB |
2081 | { |
2082 | BDRVRawState *s = bs->opaque; | |
e428e439 | 2083 | Error *local_err = NULL; |
da888d37 | 2084 | int ret; |
c66a6157 | 2085 | const char *filename = qdict_get_str(options, "filename"); |
a76bab49 | 2086 | |
83affaa6 | 2087 | #if defined(__APPLE__) && defined(__MACH__) |
19cb3738 FB |
2088 | if (strstart(filename, "/dev/cdrom", NULL)) { |
2089 | kern_return_t kernResult; | |
2090 | io_iterator_t mediaIterator; | |
2091 | char bsdPath[ MAXPATHLEN ]; | |
2092 | int fd; | |
5fafdf24 | 2093 | |
19cb3738 FB |
2094 | kernResult = FindEjectableCDMedia( &mediaIterator ); |
2095 | kernResult = GetBSDPath( mediaIterator, bsdPath, sizeof( bsdPath ) ); | |
3b46e624 | 2096 | |
19cb3738 FB |
2097 | if ( bsdPath[ 0 ] != '\0' ) { |
2098 | strcat(bsdPath,"s0"); | |
2099 | /* some CDs don't have a partition 0 */ | |
6165f4d8 | 2100 | fd = qemu_open(bsdPath, O_RDONLY | O_BINARY | O_LARGEFILE); |
19cb3738 FB |
2101 | if (fd < 0) { |
2102 | bsdPath[strlen(bsdPath)-1] = '1'; | |
2103 | } else { | |
2e1e79da | 2104 | qemu_close(fd); |
19cb3738 FB |
2105 | } |
2106 | filename = bsdPath; | |
a5c5ea3f | 2107 | qdict_put(options, "filename", qstring_from_str(filename)); |
19cb3738 | 2108 | } |
3b46e624 | 2109 | |
19cb3738 FB |
2110 | if ( mediaIterator ) |
2111 | IOObjectRelease( mediaIterator ); | |
2112 | } | |
2113 | #endif | |
19cb3738 FB |
2114 | |
2115 | s->type = FTYPE_FILE; | |
4dd75c70 | 2116 | #if defined(__linux__) |
05acda4d BK |
2117 | { |
2118 | char resolved_path[ MAXPATHLEN ], *temp; | |
2119 | ||
2120 | temp = realpath(filename, resolved_path); | |
2121 | if (temp && strstart(temp, "/dev/sg", NULL)) { | |
2122 | bs->sg = 1; | |
2123 | } | |
19cb3738 FB |
2124 | } |
2125 | #endif | |
90babde0 | 2126 | |
e428e439 | 2127 | ret = raw_open_common(bs, options, flags, 0, &local_err); |
da888d37 | 2128 | if (ret < 0) { |
84d18f06 | 2129 | if (local_err) { |
e428e439 HR |
2130 | error_propagate(errp, local_err); |
2131 | } | |
da888d37 SH |
2132 | return ret; |
2133 | } | |
2134 | ||
2135 | if (flags & BDRV_O_RDWR) { | |
2136 | ret = check_hdev_writable(s); | |
2137 | if (ret < 0) { | |
2138 | raw_close(bs); | |
e428e439 | 2139 | error_setg_errno(errp, -ret, "The device is not writable"); |
da888d37 SH |
2140 | return ret; |
2141 | } | |
2142 | } | |
2143 | ||
2144 | return ret; | |
19cb3738 FB |
2145 | } |
2146 | ||
03ff3ca3 | 2147 | #if defined(__linux__) |
19cb3738 FB |
2148 | /* Note: we do not have a reliable method to detect if the floppy is |
2149 | present. The current method is to try to open the floppy at every | |
2150 | I/O and to keep it opened during a few hundreds of ms. */ | |
2151 | static int fd_open(BlockDriverState *bs) | |
2152 | { | |
2153 | BDRVRawState *s = bs->opaque; | |
2154 | int last_media_present; | |
2155 | ||
2156 | if (s->type != FTYPE_FD) | |
2157 | return 0; | |
2158 | last_media_present = (s->fd >= 0); | |
5fafdf24 | 2159 | if (s->fd >= 0 && |
a56ebc6b | 2160 | (qemu_clock_get_ns(QEMU_CLOCK_REALTIME) - s->fd_open_time) >= FD_OPEN_TIMEOUT) { |
2e1e79da | 2161 | qemu_close(s->fd); |
19cb3738 FB |
2162 | s->fd = -1; |
2163 | #ifdef DEBUG_FLOPPY | |
2164 | printf("Floppy closed\n"); | |
2165 | #endif | |
2166 | } | |
2167 | if (s->fd < 0) { | |
5fafdf24 | 2168 | if (s->fd_got_error && |
a56ebc6b | 2169 | (qemu_clock_get_ns(QEMU_CLOCK_REALTIME) - s->fd_error_time) < FD_OPEN_TIMEOUT) { |
19cb3738 FB |
2170 | #ifdef DEBUG_FLOPPY |
2171 | printf("No floppy (open delayed)\n"); | |
2172 | #endif | |
2173 | return -EIO; | |
2174 | } | |
6165f4d8 | 2175 | s->fd = qemu_open(bs->filename, s->open_flags & ~O_NONBLOCK); |
19cb3738 | 2176 | if (s->fd < 0) { |
a56ebc6b | 2177 | s->fd_error_time = qemu_clock_get_ns(QEMU_CLOCK_REALTIME); |
19cb3738 FB |
2178 | s->fd_got_error = 1; |
2179 | if (last_media_present) | |
2180 | s->fd_media_changed = 1; | |
2181 | #ifdef DEBUG_FLOPPY | |
2182 | printf("No floppy\n"); | |
2183 | #endif | |
2184 | return -EIO; | |
2185 | } | |
2186 | #ifdef DEBUG_FLOPPY | |
2187 | printf("Floppy opened\n"); | |
2188 | #endif | |
2189 | } | |
2190 | if (!last_media_present) | |
2191 | s->fd_media_changed = 1; | |
a56ebc6b | 2192 | s->fd_open_time = qemu_clock_get_ns(QEMU_CLOCK_REALTIME); |
19cb3738 FB |
2193 | s->fd_got_error = 0; |
2194 | return 0; | |
2195 | } | |
19cb3738 | 2196 | |
63ec93db | 2197 | static int hdev_ioctl(BlockDriverState *bs, unsigned long int req, void *buf) |
985a03b0 TS |
2198 | { |
2199 | BDRVRawState *s = bs->opaque; | |
2200 | ||
2201 | return ioctl(s->fd, req, buf); | |
2202 | } | |
221f715d | 2203 | |
7c84b1b8 | 2204 | static BlockAIOCB *hdev_aio_ioctl(BlockDriverState *bs, |
221f715d | 2205 | unsigned long int req, void *buf, |
097310b5 | 2206 | BlockCompletionFunc *cb, void *opaque) |
221f715d | 2207 | { |
f141eafe | 2208 | BDRVRawState *s = bs->opaque; |
c208e8c2 | 2209 | RawPosixAIOData *acb; |
c4d9d196 | 2210 | ThreadPool *pool; |
221f715d | 2211 | |
f141eafe AL |
2212 | if (fd_open(bs) < 0) |
2213 | return NULL; | |
c208e8c2 PB |
2214 | |
2215 | acb = g_slice_new(RawPosixAIOData); | |
2216 | acb->bs = bs; | |
2217 | acb->aio_type = QEMU_AIO_IOCTL; | |
2218 | acb->aio_fildes = s->fd; | |
2219 | acb->aio_offset = 0; | |
2220 | acb->aio_ioctl_buf = buf; | |
2221 | acb->aio_ioctl_cmd = req; | |
c4d9d196 SH |
2222 | pool = aio_get_thread_pool(bdrv_get_aio_context(bs)); |
2223 | return thread_pool_submit_aio(pool, aio_worker, acb, cb, opaque); | |
221f715d AL |
2224 | } |
2225 | ||
a167ba50 | 2226 | #elif defined(__FreeBSD__) || defined(__FreeBSD_kernel__) |
9f23011a BS |
2227 | static int fd_open(BlockDriverState *bs) |
2228 | { | |
2229 | BDRVRawState *s = bs->opaque; | |
2230 | ||
2231 | /* this is just to ensure s->fd is sane (its called by io ops) */ | |
2232 | if (s->fd >= 0) | |
2233 | return 0; | |
2234 | return -EIO; | |
2235 | } | |
9f23011a | 2236 | #else /* !linux && !FreeBSD */ |
19cb3738 | 2237 | |
08af02e2 AL |
2238 | static int fd_open(BlockDriverState *bs) |
2239 | { | |
2240 | return 0; | |
2241 | } | |
2242 | ||
221f715d | 2243 | #endif /* !linux && !FreeBSD */ |
04eeb8b6 | 2244 | |
7c84b1b8 | 2245 | static coroutine_fn BlockAIOCB *hdev_aio_discard(BlockDriverState *bs, |
c36dd8a0 | 2246 | int64_t sector_num, int nb_sectors, |
097310b5 | 2247 | BlockCompletionFunc *cb, void *opaque) |
c36dd8a0 AF |
2248 | { |
2249 | BDRVRawState *s = bs->opaque; | |
2250 | ||
2251 | if (fd_open(bs) < 0) { | |
2252 | return NULL; | |
2253 | } | |
2254 | return paio_submit(bs, s->fd, sector_num, NULL, nb_sectors, | |
2255 | cb, opaque, QEMU_AIO_DISCARD|QEMU_AIO_BLKDEV); | |
2256 | } | |
2257 | ||
d0b4503e PB |
2258 | static coroutine_fn int hdev_co_write_zeroes(BlockDriverState *bs, |
2259 | int64_t sector_num, int nb_sectors, BdrvRequestFlags flags) | |
2260 | { | |
2261 | BDRVRawState *s = bs->opaque; | |
2262 | int rc; | |
2263 | ||
2264 | rc = fd_open(bs); | |
2265 | if (rc < 0) { | |
2266 | return rc; | |
2267 | } | |
2268 | if (!(flags & BDRV_REQ_MAY_UNMAP)) { | |
97a2ae34 PB |
2269 | return paio_submit_co(bs, s->fd, sector_num, NULL, nb_sectors, |
2270 | QEMU_AIO_WRITE_ZEROES|QEMU_AIO_BLKDEV); | |
2271 | } else if (s->discard_zeroes) { | |
2272 | return paio_submit_co(bs, s->fd, sector_num, NULL, nb_sectors, | |
2273 | QEMU_AIO_DISCARD|QEMU_AIO_BLKDEV); | |
d0b4503e | 2274 | } |
97a2ae34 | 2275 | return -ENOTSUP; |
d0b4503e PB |
2276 | } |
2277 | ||
6f482f74 | 2278 | static int hdev_create(const char *filename, QemuOpts *opts, |
d5124c00 | 2279 | Error **errp) |
93c65b47 AL |
2280 | { |
2281 | int fd; | |
2282 | int ret = 0; | |
2283 | struct stat stat_buf; | |
0e7e1989 | 2284 | int64_t total_size = 0; |
cc28c6aa HR |
2285 | bool has_prefix; |
2286 | ||
2287 | /* This function is used by all three protocol block drivers and therefore | |
2288 | * any of these three prefixes may be given. | |
2289 | * The return value has to be stored somewhere, otherwise this is an error | |
2290 | * due to -Werror=unused-value. */ | |
2291 | has_prefix = | |
2292 | strstart(filename, "host_device:", &filename) || | |
2293 | strstart(filename, "host_cdrom:" , &filename) || | |
2294 | strstart(filename, "host_floppy:", &filename); | |
2295 | ||
2296 | (void)has_prefix; | |
93c65b47 | 2297 | |
0e7e1989 | 2298 | /* Read out options */ |
180e9526 HT |
2299 | total_size = ROUND_UP(qemu_opt_get_size_del(opts, BLOCK_OPT_SIZE, 0), |
2300 | BDRV_SECTOR_SIZE); | |
93c65b47 | 2301 | |
6165f4d8 | 2302 | fd = qemu_open(filename, O_WRONLY | O_BINARY); |
e428e439 HR |
2303 | if (fd < 0) { |
2304 | ret = -errno; | |
2305 | error_setg_errno(errp, -ret, "Could not open device"); | |
2306 | return ret; | |
2307 | } | |
93c65b47 | 2308 | |
e428e439 | 2309 | if (fstat(fd, &stat_buf) < 0) { |
57e69b7d | 2310 | ret = -errno; |
e428e439 HR |
2311 | error_setg_errno(errp, -ret, "Could not stat device"); |
2312 | } else if (!S_ISBLK(stat_buf.st_mode) && !S_ISCHR(stat_buf.st_mode)) { | |
2313 | error_setg(errp, | |
2314 | "The given file is neither a block nor a character device"); | |
57e69b7d | 2315 | ret = -ENODEV; |
180e9526 | 2316 | } else if (lseek(fd, 0, SEEK_END) < total_size) { |
e428e439 | 2317 | error_setg(errp, "Device is too small"); |
93c65b47 | 2318 | ret = -ENOSPC; |
e428e439 | 2319 | } |
93c65b47 | 2320 | |
2e1e79da | 2321 | qemu_close(fd); |
93c65b47 AL |
2322 | return ret; |
2323 | } | |
2324 | ||
5efa9d5a | 2325 | static BlockDriver bdrv_host_device = { |
0b4ce02e | 2326 | .format_name = "host_device", |
84a12e66 | 2327 | .protocol_name = "host_device", |
0b4ce02e | 2328 | .instance_size = sizeof(BDRVRawState), |
030be321 | 2329 | .bdrv_needs_filename = true, |
0b4ce02e | 2330 | .bdrv_probe_device = hdev_probe_device, |
7af803d4 | 2331 | .bdrv_parse_filename = hdev_parse_filename, |
66f82cee | 2332 | .bdrv_file_open = hdev_open, |
0b4ce02e | 2333 | .bdrv_close = raw_close, |
1bc6b705 JC |
2334 | .bdrv_reopen_prepare = raw_reopen_prepare, |
2335 | .bdrv_reopen_commit = raw_reopen_commit, | |
2336 | .bdrv_reopen_abort = raw_reopen_abort, | |
c282e1fd | 2337 | .bdrv_create = hdev_create, |
6f482f74 | 2338 | .create_opts = &raw_create_opts, |
d0b4503e | 2339 | .bdrv_co_write_zeroes = hdev_co_write_zeroes, |
3b46e624 | 2340 | |
f141eafe AL |
2341 | .bdrv_aio_readv = raw_aio_readv, |
2342 | .bdrv_aio_writev = raw_aio_writev, | |
b2e12bc6 | 2343 | .bdrv_aio_flush = raw_aio_flush, |
8238010b | 2344 | .bdrv_aio_discard = hdev_aio_discard, |
c25f53b0 | 2345 | .bdrv_refresh_limits = raw_refresh_limits, |
1b3abdcc ML |
2346 | .bdrv_io_plug = raw_aio_plug, |
2347 | .bdrv_io_unplug = raw_aio_unplug, | |
2348 | .bdrv_flush_io_queue = raw_aio_flush_io_queue, | |
3c529d93 | 2349 | |
55b949c8 | 2350 | .bdrv_truncate = raw_truncate, |
e60f469c | 2351 | .bdrv_getlength = raw_getlength, |
260a82e5 | 2352 | .bdrv_get_info = raw_get_info, |
4a1d5e1f FZ |
2353 | .bdrv_get_allocated_file_size |
2354 | = raw_get_allocated_file_size, | |
1a9335e4 ET |
2355 | .bdrv_probe_blocksizes = hdev_probe_blocksizes, |
2356 | .bdrv_probe_geometry = hdev_probe_geometry, | |
19cb3738 | 2357 | |
c2f3426c SH |
2358 | .bdrv_detach_aio_context = raw_detach_aio_context, |
2359 | .bdrv_attach_aio_context = raw_attach_aio_context, | |
2360 | ||
f3a5d3f8 | 2361 | /* generic scsi device */ |
63ec93db CH |
2362 | #ifdef __linux__ |
2363 | .bdrv_ioctl = hdev_ioctl, | |
63ec93db CH |
2364 | .bdrv_aio_ioctl = hdev_aio_ioctl, |
2365 | #endif | |
f3a5d3f8 CH |
2366 | }; |
2367 | ||
2368 | #ifdef __linux__ | |
d3f49845 HR |
2369 | static void floppy_parse_filename(const char *filename, QDict *options, |
2370 | Error **errp) | |
2371 | { | |
2372 | /* The prefix is optional, just as for "file". */ | |
2373 | strstart(filename, "host_floppy:", &filename); | |
2374 | ||
2375 | qdict_put_obj(options, "filename", QOBJECT(qstring_from_str(filename))); | |
2376 | } | |
2377 | ||
015a1036 HR |
2378 | static int floppy_open(BlockDriverState *bs, QDict *options, int flags, |
2379 | Error **errp) | |
f3a5d3f8 CH |
2380 | { |
2381 | BDRVRawState *s = bs->opaque; | |
e428e439 | 2382 | Error *local_err = NULL; |
f3a5d3f8 CH |
2383 | int ret; |
2384 | ||
f3a5d3f8 | 2385 | s->type = FTYPE_FD; |
f3a5d3f8 | 2386 | |
19a3da7f | 2387 | /* open will not fail even if no floppy is inserted, so add O_NONBLOCK */ |
e428e439 HR |
2388 | ret = raw_open_common(bs, options, flags, O_NONBLOCK, &local_err); |
2389 | if (ret) { | |
84d18f06 | 2390 | if (local_err) { |
e428e439 HR |
2391 | error_propagate(errp, local_err); |
2392 | } | |
f3a5d3f8 | 2393 | return ret; |
e428e439 | 2394 | } |
f3a5d3f8 CH |
2395 | |
2396 | /* close fd so that we can reopen it as needed */ | |
2e1e79da | 2397 | qemu_close(s->fd); |
f3a5d3f8 CH |
2398 | s->fd = -1; |
2399 | s->fd_media_changed = 1; | |
2400 | ||
92a539d2 MA |
2401 | error_report("Host floppy pass-through is deprecated"); |
2402 | error_printf("Support for it will be removed in a future release.\n"); | |
f3a5d3f8 CH |
2403 | return 0; |
2404 | } | |
2405 | ||
508c7cb3 CH |
2406 | static int floppy_probe_device(const char *filename) |
2407 | { | |
2ebf7c4b CR |
2408 | int fd, ret; |
2409 | int prio = 0; | |
2410 | struct floppy_struct fdparam; | |
343f8568 | 2411 | struct stat st; |
2ebf7c4b | 2412 | |
e1740828 CB |
2413 | if (strstart(filename, "/dev/fd", NULL) && |
2414 | !strstart(filename, "/dev/fdset/", NULL)) { | |
2ebf7c4b | 2415 | prio = 50; |
e1740828 | 2416 | } |
2ebf7c4b | 2417 | |
6165f4d8 | 2418 | fd = qemu_open(filename, O_RDONLY | O_NONBLOCK); |
2ebf7c4b CR |
2419 | if (fd < 0) { |
2420 | goto out; | |
2421 | } | |
343f8568 JS |
2422 | ret = fstat(fd, &st); |
2423 | if (ret == -1 || !S_ISBLK(st.st_mode)) { | |
2424 | goto outc; | |
2425 | } | |
2ebf7c4b CR |
2426 | |
2427 | /* Attempt to detect via a floppy specific ioctl */ | |
2428 | ret = ioctl(fd, FDGETPRM, &fdparam); | |
2429 | if (ret >= 0) | |
2430 | prio = 100; | |
2431 | ||
343f8568 | 2432 | outc: |
2e1e79da | 2433 | qemu_close(fd); |
2ebf7c4b CR |
2434 | out: |
2435 | return prio; | |
508c7cb3 CH |
2436 | } |
2437 | ||
2438 | ||
f3a5d3f8 CH |
2439 | static int floppy_is_inserted(BlockDriverState *bs) |
2440 | { | |
2441 | return fd_open(bs) >= 0; | |
2442 | } | |
2443 | ||
2444 | static int floppy_media_changed(BlockDriverState *bs) | |
2445 | { | |
2446 | BDRVRawState *s = bs->opaque; | |
2447 | int ret; | |
2448 | ||
2449 | /* | |
2450 | * XXX: we do not have a true media changed indication. | |
2451 | * It does not work if the floppy is changed without trying to read it. | |
2452 | */ | |
2453 | fd_open(bs); | |
2454 | ret = s->fd_media_changed; | |
2455 | s->fd_media_changed = 0; | |
2456 | #ifdef DEBUG_FLOPPY | |
2457 | printf("Floppy changed=%d\n", ret); | |
2458 | #endif | |
2459 | return ret; | |
2460 | } | |
2461 | ||
f36f3949 | 2462 | static void floppy_eject(BlockDriverState *bs, bool eject_flag) |
f3a5d3f8 CH |
2463 | { |
2464 | BDRVRawState *s = bs->opaque; | |
2465 | int fd; | |
2466 | ||
2467 | if (s->fd >= 0) { | |
2e1e79da | 2468 | qemu_close(s->fd); |
f3a5d3f8 CH |
2469 | s->fd = -1; |
2470 | } | |
6165f4d8 | 2471 | fd = qemu_open(bs->filename, s->open_flags | O_NONBLOCK); |
f3a5d3f8 CH |
2472 | if (fd >= 0) { |
2473 | if (ioctl(fd, FDEJECT, 0) < 0) | |
2474 | perror("FDEJECT"); | |
2e1e79da | 2475 | qemu_close(fd); |
f3a5d3f8 | 2476 | } |
f3a5d3f8 CH |
2477 | } |
2478 | ||
2479 | static BlockDriver bdrv_host_floppy = { | |
2480 | .format_name = "host_floppy", | |
84a12e66 | 2481 | .protocol_name = "host_floppy", |
f3a5d3f8 | 2482 | .instance_size = sizeof(BDRVRawState), |
030be321 | 2483 | .bdrv_needs_filename = true, |
508c7cb3 | 2484 | .bdrv_probe_device = floppy_probe_device, |
d3f49845 | 2485 | .bdrv_parse_filename = floppy_parse_filename, |
66f82cee | 2486 | .bdrv_file_open = floppy_open, |
f3a5d3f8 | 2487 | .bdrv_close = raw_close, |
1bc6b705 JC |
2488 | .bdrv_reopen_prepare = raw_reopen_prepare, |
2489 | .bdrv_reopen_commit = raw_reopen_commit, | |
2490 | .bdrv_reopen_abort = raw_reopen_abort, | |
c282e1fd | 2491 | .bdrv_create = hdev_create, |
6f482f74 | 2492 | .create_opts = &raw_create_opts, |
f3a5d3f8 | 2493 | |
f3a5d3f8 CH |
2494 | .bdrv_aio_readv = raw_aio_readv, |
2495 | .bdrv_aio_writev = raw_aio_writev, | |
b2e12bc6 | 2496 | .bdrv_aio_flush = raw_aio_flush, |
c25f53b0 | 2497 | .bdrv_refresh_limits = raw_refresh_limits, |
1b3abdcc ML |
2498 | .bdrv_io_plug = raw_aio_plug, |
2499 | .bdrv_io_unplug = raw_aio_unplug, | |
2500 | .bdrv_flush_io_queue = raw_aio_flush_io_queue, | |
f3a5d3f8 | 2501 | |
55b949c8 | 2502 | .bdrv_truncate = raw_truncate, |
b94a2610 KW |
2503 | .bdrv_getlength = raw_getlength, |
2504 | .has_variable_length = true, | |
4a1d5e1f FZ |
2505 | .bdrv_get_allocated_file_size |
2506 | = raw_get_allocated_file_size, | |
f3a5d3f8 | 2507 | |
c2f3426c SH |
2508 | .bdrv_detach_aio_context = raw_detach_aio_context, |
2509 | .bdrv_attach_aio_context = raw_attach_aio_context, | |
2510 | ||
f3a5d3f8 CH |
2511 | /* removable device support */ |
2512 | .bdrv_is_inserted = floppy_is_inserted, | |
2513 | .bdrv_media_changed = floppy_media_changed, | |
2514 | .bdrv_eject = floppy_eject, | |
f3a5d3f8 | 2515 | }; |
18fa1c42 | 2516 | #endif |
f3a5d3f8 | 2517 | |
18fa1c42 HR |
2518 | #if defined(__linux__) || defined(__FreeBSD__) || defined(__FreeBSD_kernel__) |
2519 | static void cdrom_parse_filename(const char *filename, QDict *options, | |
2520 | Error **errp) | |
2521 | { | |
2522 | /* The prefix is optional, just as for "file". */ | |
2523 | strstart(filename, "host_cdrom:", &filename); | |
2524 | ||
2525 | qdict_put_obj(options, "filename", QOBJECT(qstring_from_str(filename))); | |
2526 | } | |
2527 | #endif | |
2528 | ||
2529 | #ifdef __linux__ | |
015a1036 HR |
2530 | static int cdrom_open(BlockDriverState *bs, QDict *options, int flags, |
2531 | Error **errp) | |
f3a5d3f8 CH |
2532 | { |
2533 | BDRVRawState *s = bs->opaque; | |
e428e439 HR |
2534 | Error *local_err = NULL; |
2535 | int ret; | |
f3a5d3f8 | 2536 | |
f3a5d3f8 CH |
2537 | s->type = FTYPE_CD; |
2538 | ||
19a3da7f | 2539 | /* open will not fail even if no CD is inserted, so add O_NONBLOCK */ |
e428e439 | 2540 | ret = raw_open_common(bs, options, flags, O_NONBLOCK, &local_err); |
84d18f06 | 2541 | if (local_err) { |
e428e439 HR |
2542 | error_propagate(errp, local_err); |
2543 | } | |
2544 | return ret; | |
f3a5d3f8 CH |
2545 | } |
2546 | ||
508c7cb3 CH |
2547 | static int cdrom_probe_device(const char *filename) |
2548 | { | |
3baf720e CR |
2549 | int fd, ret; |
2550 | int prio = 0; | |
343f8568 | 2551 | struct stat st; |
3baf720e | 2552 | |
6165f4d8 | 2553 | fd = qemu_open(filename, O_RDONLY | O_NONBLOCK); |
3baf720e CR |
2554 | if (fd < 0) { |
2555 | goto out; | |
2556 | } | |
343f8568 JS |
2557 | ret = fstat(fd, &st); |
2558 | if (ret == -1 || !S_ISBLK(st.st_mode)) { | |
2559 | goto outc; | |
2560 | } | |
3baf720e CR |
2561 | |
2562 | /* Attempt to detect via a CDROM specific ioctl */ | |
2563 | ret = ioctl(fd, CDROM_DRIVE_STATUS, CDSL_CURRENT); | |
2564 | if (ret >= 0) | |
2565 | prio = 100; | |
2566 | ||
343f8568 | 2567 | outc: |
2e1e79da | 2568 | qemu_close(fd); |
3baf720e CR |
2569 | out: |
2570 | return prio; | |
508c7cb3 CH |
2571 | } |
2572 | ||
f3a5d3f8 CH |
2573 | static int cdrom_is_inserted(BlockDriverState *bs) |
2574 | { | |
2575 | BDRVRawState *s = bs->opaque; | |
2576 | int ret; | |
2577 | ||
2578 | ret = ioctl(s->fd, CDROM_DRIVE_STATUS, CDSL_CURRENT); | |
2579 | if (ret == CDS_DISC_OK) | |
2580 | return 1; | |
2581 | return 0; | |
2582 | } | |
2583 | ||
f36f3949 | 2584 | static void cdrom_eject(BlockDriverState *bs, bool eject_flag) |
f3a5d3f8 CH |
2585 | { |
2586 | BDRVRawState *s = bs->opaque; | |
2587 | ||
2588 | if (eject_flag) { | |
2589 | if (ioctl(s->fd, CDROMEJECT, NULL) < 0) | |
2590 | perror("CDROMEJECT"); | |
2591 | } else { | |
2592 | if (ioctl(s->fd, CDROMCLOSETRAY, NULL) < 0) | |
2593 | perror("CDROMEJECT"); | |
2594 | } | |
f3a5d3f8 CH |
2595 | } |
2596 | ||
025e849a | 2597 | static void cdrom_lock_medium(BlockDriverState *bs, bool locked) |
f3a5d3f8 CH |
2598 | { |
2599 | BDRVRawState *s = bs->opaque; | |
2600 | ||
2601 | if (ioctl(s->fd, CDROM_LOCKDOOR, locked) < 0) { | |
2602 | /* | |
2603 | * Note: an error can happen if the distribution automatically | |
2604 | * mounts the CD-ROM | |
2605 | */ | |
2606 | /* perror("CDROM_LOCKDOOR"); */ | |
2607 | } | |
f3a5d3f8 CH |
2608 | } |
2609 | ||
2610 | static BlockDriver bdrv_host_cdrom = { | |
2611 | .format_name = "host_cdrom", | |
84a12e66 | 2612 | .protocol_name = "host_cdrom", |
f3a5d3f8 | 2613 | .instance_size = sizeof(BDRVRawState), |
030be321 | 2614 | .bdrv_needs_filename = true, |
508c7cb3 | 2615 | .bdrv_probe_device = cdrom_probe_device, |
18fa1c42 | 2616 | .bdrv_parse_filename = cdrom_parse_filename, |
66f82cee | 2617 | .bdrv_file_open = cdrom_open, |
f3a5d3f8 | 2618 | .bdrv_close = raw_close, |
1bc6b705 JC |
2619 | .bdrv_reopen_prepare = raw_reopen_prepare, |
2620 | .bdrv_reopen_commit = raw_reopen_commit, | |
2621 | .bdrv_reopen_abort = raw_reopen_abort, | |
c282e1fd | 2622 | .bdrv_create = hdev_create, |
6f482f74 | 2623 | .create_opts = &raw_create_opts, |
f3a5d3f8 | 2624 | |
f3a5d3f8 CH |
2625 | .bdrv_aio_readv = raw_aio_readv, |
2626 | .bdrv_aio_writev = raw_aio_writev, | |
b2e12bc6 | 2627 | .bdrv_aio_flush = raw_aio_flush, |
c25f53b0 | 2628 | .bdrv_refresh_limits = raw_refresh_limits, |
1b3abdcc ML |
2629 | .bdrv_io_plug = raw_aio_plug, |
2630 | .bdrv_io_unplug = raw_aio_unplug, | |
2631 | .bdrv_flush_io_queue = raw_aio_flush_io_queue, | |
f3a5d3f8 | 2632 | |
55b949c8 | 2633 | .bdrv_truncate = raw_truncate, |
b94a2610 KW |
2634 | .bdrv_getlength = raw_getlength, |
2635 | .has_variable_length = true, | |
4a1d5e1f FZ |
2636 | .bdrv_get_allocated_file_size |
2637 | = raw_get_allocated_file_size, | |
f3a5d3f8 | 2638 | |
c2f3426c SH |
2639 | .bdrv_detach_aio_context = raw_detach_aio_context, |
2640 | .bdrv_attach_aio_context = raw_attach_aio_context, | |
2641 | ||
f3a5d3f8 CH |
2642 | /* removable device support */ |
2643 | .bdrv_is_inserted = cdrom_is_inserted, | |
2644 | .bdrv_eject = cdrom_eject, | |
025e849a | 2645 | .bdrv_lock_medium = cdrom_lock_medium, |
f3a5d3f8 CH |
2646 | |
2647 | /* generic scsi device */ | |
63ec93db | 2648 | .bdrv_ioctl = hdev_ioctl, |
63ec93db | 2649 | .bdrv_aio_ioctl = hdev_aio_ioctl, |
f3a5d3f8 CH |
2650 | }; |
2651 | #endif /* __linux__ */ | |
2652 | ||
a167ba50 | 2653 | #if defined (__FreeBSD__) || defined(__FreeBSD_kernel__) |
511018e4 AT |
2654 | static int cdrom_open(BlockDriverState *bs, QDict *options, int flags, |
2655 | Error **errp) | |
f3a5d3f8 CH |
2656 | { |
2657 | BDRVRawState *s = bs->opaque; | |
e428e439 | 2658 | Error *local_err = NULL; |
f3a5d3f8 CH |
2659 | int ret; |
2660 | ||
2661 | s->type = FTYPE_CD; | |
2662 | ||
e428e439 HR |
2663 | ret = raw_open_common(bs, options, flags, 0, &local_err); |
2664 | if (ret) { | |
84d18f06 | 2665 | if (local_err) { |
e428e439 HR |
2666 | error_propagate(errp, local_err); |
2667 | } | |
f3a5d3f8 | 2668 | return ret; |
e428e439 | 2669 | } |
f3a5d3f8 | 2670 | |
9b2260cb | 2671 | /* make sure the door isn't locked at this time */ |
f3a5d3f8 CH |
2672 | ioctl(s->fd, CDIOCALLOW); |
2673 | return 0; | |
2674 | } | |
2675 | ||
508c7cb3 CH |
2676 | static int cdrom_probe_device(const char *filename) |
2677 | { | |
2678 | if (strstart(filename, "/dev/cd", NULL) || | |
2679 | strstart(filename, "/dev/acd", NULL)) | |
2680 | return 100; | |
2681 | return 0; | |
2682 | } | |
2683 | ||
f3a5d3f8 CH |
2684 | static int cdrom_reopen(BlockDriverState *bs) |
2685 | { | |
2686 | BDRVRawState *s = bs->opaque; | |
2687 | int fd; | |
2688 | ||
2689 | /* | |
2690 | * Force reread of possibly changed/newly loaded disc, | |
2691 | * FreeBSD seems to not notice sometimes... | |
2692 | */ | |
2693 | if (s->fd >= 0) | |
2e1e79da | 2694 | qemu_close(s->fd); |
6165f4d8 | 2695 | fd = qemu_open(bs->filename, s->open_flags, 0644); |
f3a5d3f8 CH |
2696 | if (fd < 0) { |
2697 | s->fd = -1; | |
2698 | return -EIO; | |
2699 | } | |
2700 | s->fd = fd; | |
2701 | ||
9b2260cb | 2702 | /* make sure the door isn't locked at this time */ |
f3a5d3f8 CH |
2703 | ioctl(s->fd, CDIOCALLOW); |
2704 | return 0; | |
2705 | } | |
2706 | ||
2707 | static int cdrom_is_inserted(BlockDriverState *bs) | |
2708 | { | |
2709 | return raw_getlength(bs) > 0; | |
2710 | } | |
2711 | ||
f36f3949 | 2712 | static void cdrom_eject(BlockDriverState *bs, bool eject_flag) |
f3a5d3f8 CH |
2713 | { |
2714 | BDRVRawState *s = bs->opaque; | |
2715 | ||
2716 | if (s->fd < 0) | |
822e1cd1 | 2717 | return; |
f3a5d3f8 CH |
2718 | |
2719 | (void) ioctl(s->fd, CDIOCALLOW); | |
2720 | ||
2721 | if (eject_flag) { | |
2722 | if (ioctl(s->fd, CDIOCEJECT) < 0) | |
2723 | perror("CDIOCEJECT"); | |
2724 | } else { | |
2725 | if (ioctl(s->fd, CDIOCCLOSE) < 0) | |
2726 | perror("CDIOCCLOSE"); | |
2727 | } | |
2728 | ||
822e1cd1 | 2729 | cdrom_reopen(bs); |
f3a5d3f8 CH |
2730 | } |
2731 | ||
025e849a | 2732 | static void cdrom_lock_medium(BlockDriverState *bs, bool locked) |
f3a5d3f8 CH |
2733 | { |
2734 | BDRVRawState *s = bs->opaque; | |
2735 | ||
2736 | if (s->fd < 0) | |
7bf37fed | 2737 | return; |
f3a5d3f8 CH |
2738 | if (ioctl(s->fd, (locked ? CDIOCPREVENT : CDIOCALLOW)) < 0) { |
2739 | /* | |
2740 | * Note: an error can happen if the distribution automatically | |
2741 | * mounts the CD-ROM | |
2742 | */ | |
2743 | /* perror("CDROM_LOCKDOOR"); */ | |
2744 | } | |
f3a5d3f8 CH |
2745 | } |
2746 | ||
2747 | static BlockDriver bdrv_host_cdrom = { | |
2748 | .format_name = "host_cdrom", | |
84a12e66 | 2749 | .protocol_name = "host_cdrom", |
f3a5d3f8 | 2750 | .instance_size = sizeof(BDRVRawState), |
030be321 | 2751 | .bdrv_needs_filename = true, |
508c7cb3 | 2752 | .bdrv_probe_device = cdrom_probe_device, |
18fa1c42 | 2753 | .bdrv_parse_filename = cdrom_parse_filename, |
66f82cee | 2754 | .bdrv_file_open = cdrom_open, |
f3a5d3f8 | 2755 | .bdrv_close = raw_close, |
1bc6b705 JC |
2756 | .bdrv_reopen_prepare = raw_reopen_prepare, |
2757 | .bdrv_reopen_commit = raw_reopen_commit, | |
2758 | .bdrv_reopen_abort = raw_reopen_abort, | |
c282e1fd | 2759 | .bdrv_create = hdev_create, |
6f482f74 | 2760 | .create_opts = &raw_create_opts, |
f3a5d3f8 | 2761 | |
f3a5d3f8 CH |
2762 | .bdrv_aio_readv = raw_aio_readv, |
2763 | .bdrv_aio_writev = raw_aio_writev, | |
b2e12bc6 | 2764 | .bdrv_aio_flush = raw_aio_flush, |
c25f53b0 | 2765 | .bdrv_refresh_limits = raw_refresh_limits, |
1b3abdcc ML |
2766 | .bdrv_io_plug = raw_aio_plug, |
2767 | .bdrv_io_unplug = raw_aio_unplug, | |
2768 | .bdrv_flush_io_queue = raw_aio_flush_io_queue, | |
f3a5d3f8 | 2769 | |
55b949c8 | 2770 | .bdrv_truncate = raw_truncate, |
b94a2610 KW |
2771 | .bdrv_getlength = raw_getlength, |
2772 | .has_variable_length = true, | |
4a1d5e1f FZ |
2773 | .bdrv_get_allocated_file_size |
2774 | = raw_get_allocated_file_size, | |
f3a5d3f8 | 2775 | |
c2f3426c SH |
2776 | .bdrv_detach_aio_context = raw_detach_aio_context, |
2777 | .bdrv_attach_aio_context = raw_attach_aio_context, | |
2778 | ||
19cb3738 | 2779 | /* removable device support */ |
f3a5d3f8 CH |
2780 | .bdrv_is_inserted = cdrom_is_inserted, |
2781 | .bdrv_eject = cdrom_eject, | |
025e849a | 2782 | .bdrv_lock_medium = cdrom_lock_medium, |
19cb3738 | 2783 | }; |
f3a5d3f8 | 2784 | #endif /* __FreeBSD__ */ |
5efa9d5a | 2785 | |
84a12e66 | 2786 | static void bdrv_file_init(void) |
5efa9d5a | 2787 | { |
508c7cb3 CH |
2788 | /* |
2789 | * Register all the drivers. Note that order is important, the driver | |
2790 | * registered last will get probed first. | |
2791 | */ | |
84a12e66 | 2792 | bdrv_register(&bdrv_file); |
5efa9d5a | 2793 | bdrv_register(&bdrv_host_device); |
f3a5d3f8 CH |
2794 | #ifdef __linux__ |
2795 | bdrv_register(&bdrv_host_floppy); | |
2796 | bdrv_register(&bdrv_host_cdrom); | |
2797 | #endif | |
a167ba50 | 2798 | #if defined(__FreeBSD__) || defined(__FreeBSD_kernel__) |
f3a5d3f8 CH |
2799 | bdrv_register(&bdrv_host_cdrom); |
2800 | #endif | |
5efa9d5a AL |
2801 | } |
2802 | ||
84a12e66 | 2803 | block_init(bdrv_file_init); |