]>
Commit | Line | Data |
---|---|---|
fc01f7e7 FB |
1 | /* |
2 | * QEMU System Emulator block driver | |
5fafdf24 | 3 | * |
fc01f7e7 | 4 | * Copyright (c) 2003 Fabrice Bellard |
5fafdf24 | 5 | * |
fc01f7e7 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 | */ | |
3990d09a | 24 | #include "config-host.h" |
faf07963 | 25 | #include "qemu-common.h" |
6d519a5f | 26 | #include "trace.h" |
737e150e PB |
27 | #include "block/block_int.h" |
28 | #include "block/blockjob.h" | |
1de7afc9 | 29 | #include "qemu/module.h" |
7b1b5d19 | 30 | #include "qapi/qmp/qjson.h" |
bfb197e0 | 31 | #include "sysemu/block-backend.h" |
9c17d615 | 32 | #include "sysemu/sysemu.h" |
1de7afc9 | 33 | #include "qemu/notify.h" |
737e150e | 34 | #include "block/coroutine.h" |
c13163fb | 35 | #include "block/qapi.h" |
b2023818 | 36 | #include "qmp-commands.h" |
1de7afc9 | 37 | #include "qemu/timer.h" |
a5ee7bd4 | 38 | #include "qapi-event.h" |
fc01f7e7 | 39 | |
71e72a19 | 40 | #ifdef CONFIG_BSD |
7674e7bf FB |
41 | #include <sys/types.h> |
42 | #include <sys/stat.h> | |
43 | #include <sys/ioctl.h> | |
72cf2d4f | 44 | #include <sys/queue.h> |
c5e97233 | 45 | #ifndef __DragonFly__ |
7674e7bf FB |
46 | #include <sys/disk.h> |
47 | #endif | |
c5e97233 | 48 | #endif |
7674e7bf | 49 | |
49dc768d AL |
50 | #ifdef _WIN32 |
51 | #include <windows.h> | |
52 | #endif | |
53 | ||
e4654d2d FZ |
54 | struct BdrvDirtyBitmap { |
55 | HBitmap *bitmap; | |
56 | QLIST_ENTRY(BdrvDirtyBitmap) list; | |
57 | }; | |
58 | ||
1c9805a3 SH |
59 | #define NOT_DONE 0x7fffffff /* used while emulated sync operation in progress */ |
60 | ||
7c84b1b8 | 61 | static BlockAIOCB *bdrv_aio_readv_em(BlockDriverState *bs, |
f141eafe | 62 | int64_t sector_num, QEMUIOVector *qiov, int nb_sectors, |
097310b5 | 63 | BlockCompletionFunc *cb, void *opaque); |
7c84b1b8 | 64 | static BlockAIOCB *bdrv_aio_writev_em(BlockDriverState *bs, |
f141eafe | 65 | int64_t sector_num, QEMUIOVector *qiov, int nb_sectors, |
097310b5 | 66 | BlockCompletionFunc *cb, void *opaque); |
f9f05dc5 KW |
67 | static int coroutine_fn bdrv_co_readv_em(BlockDriverState *bs, |
68 | int64_t sector_num, int nb_sectors, | |
69 | QEMUIOVector *iov); | |
70 | static int coroutine_fn bdrv_co_writev_em(BlockDriverState *bs, | |
71 | int64_t sector_num, int nb_sectors, | |
72 | QEMUIOVector *iov); | |
775aa8b6 KW |
73 | static int coroutine_fn bdrv_co_do_preadv(BlockDriverState *bs, |
74 | int64_t offset, unsigned int bytes, QEMUIOVector *qiov, | |
470c0504 | 75 | BdrvRequestFlags flags); |
775aa8b6 KW |
76 | static int coroutine_fn bdrv_co_do_pwritev(BlockDriverState *bs, |
77 | int64_t offset, unsigned int bytes, QEMUIOVector *qiov, | |
f08f2dda | 78 | BdrvRequestFlags flags); |
7c84b1b8 MA |
79 | static BlockAIOCB *bdrv_co_aio_rw_vector(BlockDriverState *bs, |
80 | int64_t sector_num, | |
81 | QEMUIOVector *qiov, | |
82 | int nb_sectors, | |
83 | BdrvRequestFlags flags, | |
097310b5 | 84 | BlockCompletionFunc *cb, |
7c84b1b8 MA |
85 | void *opaque, |
86 | bool is_write); | |
b2a61371 | 87 | static void coroutine_fn bdrv_co_do_rw(void *opaque); |
621f0589 | 88 | static int coroutine_fn bdrv_co_do_write_zeroes(BlockDriverState *bs, |
aa7bfbff | 89 | int64_t sector_num, int nb_sectors, BdrvRequestFlags flags); |
ec530c81 | 90 | |
1b7bdbc1 SH |
91 | static QTAILQ_HEAD(, BlockDriverState) bdrv_states = |
92 | QTAILQ_HEAD_INITIALIZER(bdrv_states); | |
7ee930d0 | 93 | |
dc364f4c BC |
94 | static QTAILQ_HEAD(, BlockDriverState) graph_bdrv_states = |
95 | QTAILQ_HEAD_INITIALIZER(graph_bdrv_states); | |
96 | ||
8a22f02a SH |
97 | static QLIST_HEAD(, BlockDriver) bdrv_drivers = |
98 | QLIST_HEAD_INITIALIZER(bdrv_drivers); | |
ea2384d3 | 99 | |
eb852011 MA |
100 | /* If non-zero, use only whitelisted block drivers */ |
101 | static int use_bdrv_whitelist; | |
102 | ||
9e0b22f4 SH |
103 | #ifdef _WIN32 |
104 | static int is_windows_drive_prefix(const char *filename) | |
105 | { | |
106 | return (((filename[0] >= 'a' && filename[0] <= 'z') || | |
107 | (filename[0] >= 'A' && filename[0] <= 'Z')) && | |
108 | filename[1] == ':'); | |
109 | } | |
110 | ||
111 | int is_windows_drive(const char *filename) | |
112 | { | |
113 | if (is_windows_drive_prefix(filename) && | |
114 | filename[2] == '\0') | |
115 | return 1; | |
116 | if (strstart(filename, "\\\\.\\", NULL) || | |
117 | strstart(filename, "//./", NULL)) | |
118 | return 1; | |
119 | return 0; | |
120 | } | |
121 | #endif | |
122 | ||
0563e191 | 123 | /* throttling disk I/O limits */ |
cc0681c4 BC |
124 | void bdrv_set_io_limits(BlockDriverState *bs, |
125 | ThrottleConfig *cfg) | |
98f90dba | 126 | { |
cc0681c4 | 127 | int i; |
98f90dba | 128 | |
cc0681c4 | 129 | throttle_config(&bs->throttle_state, cfg); |
98f90dba | 130 | |
cc0681c4 BC |
131 | for (i = 0; i < 2; i++) { |
132 | qemu_co_enter_next(&bs->throttled_reqs[i]); | |
98f90dba | 133 | } |
cc0681c4 BC |
134 | } |
135 | ||
136 | /* this function drain all the throttled IOs */ | |
137 | static bool bdrv_start_throttled_reqs(BlockDriverState *bs) | |
138 | { | |
139 | bool drained = false; | |
140 | bool enabled = bs->io_limits_enabled; | |
141 | int i; | |
142 | ||
143 | bs->io_limits_enabled = false; | |
144 | ||
145 | for (i = 0; i < 2; i++) { | |
146 | while (qemu_co_enter_next(&bs->throttled_reqs[i])) { | |
147 | drained = true; | |
148 | } | |
149 | } | |
150 | ||
151 | bs->io_limits_enabled = enabled; | |
98f90dba | 152 | |
cc0681c4 | 153 | return drained; |
98f90dba ZYW |
154 | } |
155 | ||
cc0681c4 | 156 | void bdrv_io_limits_disable(BlockDriverState *bs) |
0563e191 | 157 | { |
cc0681c4 | 158 | bs->io_limits_enabled = false; |
0563e191 | 159 | |
cc0681c4 BC |
160 | bdrv_start_throttled_reqs(bs); |
161 | ||
162 | throttle_destroy(&bs->throttle_state); | |
0563e191 ZYW |
163 | } |
164 | ||
cc0681c4 | 165 | static void bdrv_throttle_read_timer_cb(void *opaque) |
0563e191 | 166 | { |
cc0681c4 BC |
167 | BlockDriverState *bs = opaque; |
168 | qemu_co_enter_next(&bs->throttled_reqs[0]); | |
0563e191 ZYW |
169 | } |
170 | ||
cc0681c4 | 171 | static void bdrv_throttle_write_timer_cb(void *opaque) |
0563e191 | 172 | { |
cc0681c4 BC |
173 | BlockDriverState *bs = opaque; |
174 | qemu_co_enter_next(&bs->throttled_reqs[1]); | |
0563e191 ZYW |
175 | } |
176 | ||
cc0681c4 BC |
177 | /* should be called before bdrv_set_io_limits if a limit is set */ |
178 | void bdrv_io_limits_enable(BlockDriverState *bs) | |
179 | { | |
180 | assert(!bs->io_limits_enabled); | |
181 | throttle_init(&bs->throttle_state, | |
13af91eb | 182 | bdrv_get_aio_context(bs), |
cc0681c4 BC |
183 | QEMU_CLOCK_VIRTUAL, |
184 | bdrv_throttle_read_timer_cb, | |
185 | bdrv_throttle_write_timer_cb, | |
186 | bs); | |
187 | bs->io_limits_enabled = true; | |
188 | } | |
189 | ||
190 | /* This function makes an IO wait if needed | |
191 | * | |
192 | * @nb_sectors: the number of sectors of the IO | |
193 | * @is_write: is the IO a write | |
194 | */ | |
98f90dba | 195 | static void bdrv_io_limits_intercept(BlockDriverState *bs, |
d5103588 | 196 | unsigned int bytes, |
cc0681c4 | 197 | bool is_write) |
98f90dba | 198 | { |
cc0681c4 BC |
199 | /* does this io must wait */ |
200 | bool must_wait = throttle_schedule_timer(&bs->throttle_state, is_write); | |
98f90dba | 201 | |
cc0681c4 BC |
202 | /* if must wait or any request of this type throttled queue the IO */ |
203 | if (must_wait || | |
204 | !qemu_co_queue_empty(&bs->throttled_reqs[is_write])) { | |
205 | qemu_co_queue_wait(&bs->throttled_reqs[is_write]); | |
98f90dba ZYW |
206 | } |
207 | ||
cc0681c4 | 208 | /* the IO will be executed, do the accounting */ |
d5103588 KW |
209 | throttle_account(&bs->throttle_state, is_write, bytes); |
210 | ||
98f90dba | 211 | |
cc0681c4 BC |
212 | /* if the next request must wait -> do nothing */ |
213 | if (throttle_schedule_timer(&bs->throttle_state, is_write)) { | |
214 | return; | |
98f90dba ZYW |
215 | } |
216 | ||
cc0681c4 BC |
217 | /* else queue next request for execution */ |
218 | qemu_co_queue_next(&bs->throttled_reqs[is_write]); | |
98f90dba ZYW |
219 | } |
220 | ||
339064d5 KW |
221 | size_t bdrv_opt_mem_align(BlockDriverState *bs) |
222 | { | |
223 | if (!bs || !bs->drv) { | |
224 | /* 4k should be on the safe side */ | |
225 | return 4096; | |
226 | } | |
227 | ||
228 | return bs->bl.opt_mem_alignment; | |
229 | } | |
230 | ||
9e0b22f4 | 231 | /* check if the path starts with "<protocol>:" */ |
5c98415b | 232 | int path_has_protocol(const char *path) |
9e0b22f4 | 233 | { |
947995c0 PB |
234 | const char *p; |
235 | ||
9e0b22f4 SH |
236 | #ifdef _WIN32 |
237 | if (is_windows_drive(path) || | |
238 | is_windows_drive_prefix(path)) { | |
239 | return 0; | |
240 | } | |
947995c0 PB |
241 | p = path + strcspn(path, ":/\\"); |
242 | #else | |
243 | p = path + strcspn(path, ":/"); | |
9e0b22f4 SH |
244 | #endif |
245 | ||
947995c0 | 246 | return *p == ':'; |
9e0b22f4 SH |
247 | } |
248 | ||
83f64091 | 249 | int path_is_absolute(const char *path) |
3b0d4f61 | 250 | { |
21664424 FB |
251 | #ifdef _WIN32 |
252 | /* specific case for names like: "\\.\d:" */ | |
f53f4da9 | 253 | if (is_windows_drive(path) || is_windows_drive_prefix(path)) { |
21664424 | 254 | return 1; |
f53f4da9 PB |
255 | } |
256 | return (*path == '/' || *path == '\\'); | |
3b9f94e1 | 257 | #else |
f53f4da9 | 258 | return (*path == '/'); |
3b9f94e1 | 259 | #endif |
3b0d4f61 FB |
260 | } |
261 | ||
83f64091 FB |
262 | /* if filename is absolute, just copy it to dest. Otherwise, build a |
263 | path to it by considering it is relative to base_path. URL are | |
264 | supported. */ | |
265 | void path_combine(char *dest, int dest_size, | |
266 | const char *base_path, | |
267 | const char *filename) | |
3b0d4f61 | 268 | { |
83f64091 FB |
269 | const char *p, *p1; |
270 | int len; | |
271 | ||
272 | if (dest_size <= 0) | |
273 | return; | |
274 | if (path_is_absolute(filename)) { | |
275 | pstrcpy(dest, dest_size, filename); | |
276 | } else { | |
277 | p = strchr(base_path, ':'); | |
278 | if (p) | |
279 | p++; | |
280 | else | |
281 | p = base_path; | |
3b9f94e1 FB |
282 | p1 = strrchr(base_path, '/'); |
283 | #ifdef _WIN32 | |
284 | { | |
285 | const char *p2; | |
286 | p2 = strrchr(base_path, '\\'); | |
287 | if (!p1 || p2 > p1) | |
288 | p1 = p2; | |
289 | } | |
290 | #endif | |
83f64091 FB |
291 | if (p1) |
292 | p1++; | |
293 | else | |
294 | p1 = base_path; | |
295 | if (p1 > p) | |
296 | p = p1; | |
297 | len = p - base_path; | |
298 | if (len > dest_size - 1) | |
299 | len = dest_size - 1; | |
300 | memcpy(dest, base_path, len); | |
301 | dest[len] = '\0'; | |
302 | pstrcat(dest, dest_size, filename); | |
3b0d4f61 | 303 | } |
3b0d4f61 FB |
304 | } |
305 | ||
0a82855a HR |
306 | void bdrv_get_full_backing_filename_from_filename(const char *backed, |
307 | const char *backing, | |
9f07429e HR |
308 | char *dest, size_t sz, |
309 | Error **errp) | |
dc5a1371 | 310 | { |
9f07429e HR |
311 | if (backing[0] == '\0' || path_has_protocol(backing) || |
312 | path_is_absolute(backing)) | |
313 | { | |
0a82855a | 314 | pstrcpy(dest, sz, backing); |
9f07429e HR |
315 | } else if (backed[0] == '\0' || strstart(backed, "json:", NULL)) { |
316 | error_setg(errp, "Cannot use relative backing file names for '%s'", | |
317 | backed); | |
dc5a1371 | 318 | } else { |
0a82855a | 319 | path_combine(dest, sz, backed, backing); |
dc5a1371 PB |
320 | } |
321 | } | |
322 | ||
9f07429e HR |
323 | void bdrv_get_full_backing_filename(BlockDriverState *bs, char *dest, size_t sz, |
324 | Error **errp) | |
0a82855a | 325 | { |
9f07429e HR |
326 | char *backed = bs->exact_filename[0] ? bs->exact_filename : bs->filename; |
327 | ||
328 | bdrv_get_full_backing_filename_from_filename(backed, bs->backing_file, | |
329 | dest, sz, errp); | |
0a82855a HR |
330 | } |
331 | ||
5efa9d5a | 332 | void bdrv_register(BlockDriver *bdrv) |
ea2384d3 | 333 | { |
8c5873d6 SH |
334 | /* Block drivers without coroutine functions need emulation */ |
335 | if (!bdrv->bdrv_co_readv) { | |
f9f05dc5 KW |
336 | bdrv->bdrv_co_readv = bdrv_co_readv_em; |
337 | bdrv->bdrv_co_writev = bdrv_co_writev_em; | |
338 | ||
f8c35c1d SH |
339 | /* bdrv_co_readv_em()/brdv_co_writev_em() work in terms of aio, so if |
340 | * the block driver lacks aio we need to emulate that too. | |
341 | */ | |
f9f05dc5 KW |
342 | if (!bdrv->bdrv_aio_readv) { |
343 | /* add AIO emulation layer */ | |
344 | bdrv->bdrv_aio_readv = bdrv_aio_readv_em; | |
345 | bdrv->bdrv_aio_writev = bdrv_aio_writev_em; | |
f9f05dc5 | 346 | } |
83f64091 | 347 | } |
b2e12bc6 | 348 | |
8a22f02a | 349 | QLIST_INSERT_HEAD(&bdrv_drivers, bdrv, list); |
ea2384d3 | 350 | } |
b338082b | 351 | |
7f06d47e | 352 | BlockDriverState *bdrv_new_root(void) |
b338082b | 353 | { |
7f06d47e | 354 | BlockDriverState *bs = bdrv_new(); |
e4e9986b | 355 | |
e4e9986b | 356 | QTAILQ_INSERT_TAIL(&bdrv_states, bs, device_list); |
e4e9986b MA |
357 | return bs; |
358 | } | |
359 | ||
360 | BlockDriverState *bdrv_new(void) | |
361 | { | |
362 | BlockDriverState *bs; | |
363 | int i; | |
364 | ||
5839e53b | 365 | bs = g_new0(BlockDriverState, 1); |
e4654d2d | 366 | QLIST_INIT(&bs->dirty_bitmaps); |
fbe40ff7 FZ |
367 | for (i = 0; i < BLOCK_OP_TYPE_MAX; i++) { |
368 | QLIST_INIT(&bs->op_blockers[i]); | |
369 | } | |
28a7282a | 370 | bdrv_iostatus_disable(bs); |
d7d512f6 | 371 | notifier_list_init(&bs->close_notifiers); |
d616b224 | 372 | notifier_with_return_list_init(&bs->before_write_notifiers); |
cc0681c4 BC |
373 | qemu_co_queue_init(&bs->throttled_reqs[0]); |
374 | qemu_co_queue_init(&bs->throttled_reqs[1]); | |
9fcb0251 | 375 | bs->refcnt = 1; |
dcd04228 | 376 | bs->aio_context = qemu_get_aio_context(); |
d7d512f6 | 377 | |
b338082b FB |
378 | return bs; |
379 | } | |
380 | ||
d7d512f6 PB |
381 | void bdrv_add_close_notifier(BlockDriverState *bs, Notifier *notify) |
382 | { | |
383 | notifier_list_add(&bs->close_notifiers, notify); | |
384 | } | |
385 | ||
ea2384d3 FB |
386 | BlockDriver *bdrv_find_format(const char *format_name) |
387 | { | |
388 | BlockDriver *drv1; | |
8a22f02a SH |
389 | QLIST_FOREACH(drv1, &bdrv_drivers, list) { |
390 | if (!strcmp(drv1->format_name, format_name)) { | |
ea2384d3 | 391 | return drv1; |
8a22f02a | 392 | } |
ea2384d3 FB |
393 | } |
394 | return NULL; | |
395 | } | |
396 | ||
b64ec4e4 | 397 | static int bdrv_is_whitelisted(BlockDriver *drv, bool read_only) |
eb852011 | 398 | { |
b64ec4e4 FZ |
399 | static const char *whitelist_rw[] = { |
400 | CONFIG_BDRV_RW_WHITELIST | |
401 | }; | |
402 | static const char *whitelist_ro[] = { | |
403 | CONFIG_BDRV_RO_WHITELIST | |
eb852011 MA |
404 | }; |
405 | const char **p; | |
406 | ||
b64ec4e4 | 407 | if (!whitelist_rw[0] && !whitelist_ro[0]) { |
eb852011 | 408 | return 1; /* no whitelist, anything goes */ |
b64ec4e4 | 409 | } |
eb852011 | 410 | |
b64ec4e4 | 411 | for (p = whitelist_rw; *p; p++) { |
eb852011 MA |
412 | if (!strcmp(drv->format_name, *p)) { |
413 | return 1; | |
414 | } | |
415 | } | |
b64ec4e4 FZ |
416 | if (read_only) { |
417 | for (p = whitelist_ro; *p; p++) { | |
418 | if (!strcmp(drv->format_name, *p)) { | |
419 | return 1; | |
420 | } | |
421 | } | |
422 | } | |
eb852011 MA |
423 | return 0; |
424 | } | |
425 | ||
b64ec4e4 FZ |
426 | BlockDriver *bdrv_find_whitelisted_format(const char *format_name, |
427 | bool read_only) | |
eb852011 MA |
428 | { |
429 | BlockDriver *drv = bdrv_find_format(format_name); | |
b64ec4e4 | 430 | return drv && bdrv_is_whitelisted(drv, read_only) ? drv : NULL; |
eb852011 MA |
431 | } |
432 | ||
5b7e1542 ZYW |
433 | typedef struct CreateCo { |
434 | BlockDriver *drv; | |
435 | char *filename; | |
83d0521a | 436 | QemuOpts *opts; |
5b7e1542 | 437 | int ret; |
cc84d90f | 438 | Error *err; |
5b7e1542 ZYW |
439 | } CreateCo; |
440 | ||
441 | static void coroutine_fn bdrv_create_co_entry(void *opaque) | |
442 | { | |
cc84d90f HR |
443 | Error *local_err = NULL; |
444 | int ret; | |
445 | ||
5b7e1542 ZYW |
446 | CreateCo *cco = opaque; |
447 | assert(cco->drv); | |
448 | ||
c282e1fd | 449 | ret = cco->drv->bdrv_create(cco->filename, cco->opts, &local_err); |
84d18f06 | 450 | if (local_err) { |
cc84d90f HR |
451 | error_propagate(&cco->err, local_err); |
452 | } | |
453 | cco->ret = ret; | |
5b7e1542 ZYW |
454 | } |
455 | ||
0e7e1989 | 456 | int bdrv_create(BlockDriver *drv, const char* filename, |
83d0521a | 457 | QemuOpts *opts, Error **errp) |
ea2384d3 | 458 | { |
5b7e1542 ZYW |
459 | int ret; |
460 | ||
461 | Coroutine *co; | |
462 | CreateCo cco = { | |
463 | .drv = drv, | |
464 | .filename = g_strdup(filename), | |
83d0521a | 465 | .opts = opts, |
5b7e1542 | 466 | .ret = NOT_DONE, |
cc84d90f | 467 | .err = NULL, |
5b7e1542 ZYW |
468 | }; |
469 | ||
c282e1fd | 470 | if (!drv->bdrv_create) { |
cc84d90f | 471 | error_setg(errp, "Driver '%s' does not support image creation", drv->format_name); |
80168bff LC |
472 | ret = -ENOTSUP; |
473 | goto out; | |
5b7e1542 ZYW |
474 | } |
475 | ||
476 | if (qemu_in_coroutine()) { | |
477 | /* Fast-path if already in coroutine context */ | |
478 | bdrv_create_co_entry(&cco); | |
479 | } else { | |
480 | co = qemu_coroutine_create(bdrv_create_co_entry); | |
481 | qemu_coroutine_enter(co, &cco); | |
482 | while (cco.ret == NOT_DONE) { | |
b47ec2c4 | 483 | aio_poll(qemu_get_aio_context(), true); |
5b7e1542 ZYW |
484 | } |
485 | } | |
486 | ||
487 | ret = cco.ret; | |
cc84d90f | 488 | if (ret < 0) { |
84d18f06 | 489 | if (cco.err) { |
cc84d90f HR |
490 | error_propagate(errp, cco.err); |
491 | } else { | |
492 | error_setg_errno(errp, -ret, "Could not create image"); | |
493 | } | |
494 | } | |
0e7e1989 | 495 | |
80168bff LC |
496 | out: |
497 | g_free(cco.filename); | |
5b7e1542 | 498 | return ret; |
ea2384d3 FB |
499 | } |
500 | ||
c282e1fd | 501 | int bdrv_create_file(const char *filename, QemuOpts *opts, Error **errp) |
84a12e66 CH |
502 | { |
503 | BlockDriver *drv; | |
cc84d90f HR |
504 | Error *local_err = NULL; |
505 | int ret; | |
84a12e66 | 506 | |
98289620 | 507 | drv = bdrv_find_protocol(filename, true); |
84a12e66 | 508 | if (drv == NULL) { |
cc84d90f | 509 | error_setg(errp, "Could not find protocol for file '%s'", filename); |
16905d71 | 510 | return -ENOENT; |
84a12e66 CH |
511 | } |
512 | ||
c282e1fd | 513 | ret = bdrv_create(drv, filename, opts, &local_err); |
84d18f06 | 514 | if (local_err) { |
cc84d90f HR |
515 | error_propagate(errp, local_err); |
516 | } | |
517 | return ret; | |
84a12e66 CH |
518 | } |
519 | ||
3baca891 | 520 | void bdrv_refresh_limits(BlockDriverState *bs, Error **errp) |
d34682cd KW |
521 | { |
522 | BlockDriver *drv = bs->drv; | |
3baca891 | 523 | Error *local_err = NULL; |
d34682cd KW |
524 | |
525 | memset(&bs->bl, 0, sizeof(bs->bl)); | |
526 | ||
466ad822 | 527 | if (!drv) { |
3baca891 | 528 | return; |
466ad822 KW |
529 | } |
530 | ||
531 | /* Take some limits from the children as a default */ | |
532 | if (bs->file) { | |
3baca891 KW |
533 | bdrv_refresh_limits(bs->file, &local_err); |
534 | if (local_err) { | |
535 | error_propagate(errp, local_err); | |
536 | return; | |
537 | } | |
466ad822 | 538 | bs->bl.opt_transfer_length = bs->file->bl.opt_transfer_length; |
2647fab5 | 539 | bs->bl.max_transfer_length = bs->file->bl.max_transfer_length; |
339064d5 KW |
540 | bs->bl.opt_mem_alignment = bs->file->bl.opt_mem_alignment; |
541 | } else { | |
542 | bs->bl.opt_mem_alignment = 512; | |
466ad822 KW |
543 | } |
544 | ||
545 | if (bs->backing_hd) { | |
3baca891 KW |
546 | bdrv_refresh_limits(bs->backing_hd, &local_err); |
547 | if (local_err) { | |
548 | error_propagate(errp, local_err); | |
549 | return; | |
550 | } | |
466ad822 KW |
551 | bs->bl.opt_transfer_length = |
552 | MAX(bs->bl.opt_transfer_length, | |
553 | bs->backing_hd->bl.opt_transfer_length); | |
2647fab5 PL |
554 | bs->bl.max_transfer_length = |
555 | MIN_NON_ZERO(bs->bl.max_transfer_length, | |
556 | bs->backing_hd->bl.max_transfer_length); | |
339064d5 KW |
557 | bs->bl.opt_mem_alignment = |
558 | MAX(bs->bl.opt_mem_alignment, | |
559 | bs->backing_hd->bl.opt_mem_alignment); | |
466ad822 KW |
560 | } |
561 | ||
562 | /* Then let the driver override it */ | |
563 | if (drv->bdrv_refresh_limits) { | |
3baca891 | 564 | drv->bdrv_refresh_limits(bs, errp); |
d34682cd | 565 | } |
d34682cd KW |
566 | } |
567 | ||
eba25057 JM |
568 | /* |
569 | * Create a uniquely-named empty temporary file. | |
570 | * Return 0 upon success, otherwise a negative errno value. | |
571 | */ | |
572 | int get_tmp_filename(char *filename, int size) | |
d5249393 | 573 | { |
eba25057 | 574 | #ifdef _WIN32 |
3b9f94e1 | 575 | char temp_dir[MAX_PATH]; |
eba25057 JM |
576 | /* GetTempFileName requires that its output buffer (4th param) |
577 | have length MAX_PATH or greater. */ | |
578 | assert(size >= MAX_PATH); | |
579 | return (GetTempPath(MAX_PATH, temp_dir) | |
580 | && GetTempFileName(temp_dir, "qem", 0, filename) | |
581 | ? 0 : -GetLastError()); | |
d5249393 | 582 | #else |
67b915a5 | 583 | int fd; |
7ccfb2eb | 584 | const char *tmpdir; |
0badc1ee | 585 | tmpdir = getenv("TMPDIR"); |
69bef793 AS |
586 | if (!tmpdir) { |
587 | tmpdir = "/var/tmp"; | |
588 | } | |
eba25057 JM |
589 | if (snprintf(filename, size, "%s/vl.XXXXXX", tmpdir) >= size) { |
590 | return -EOVERFLOW; | |
591 | } | |
ea2384d3 | 592 | fd = mkstemp(filename); |
fe235a06 DH |
593 | if (fd < 0) { |
594 | return -errno; | |
595 | } | |
596 | if (close(fd) != 0) { | |
597 | unlink(filename); | |
eba25057 JM |
598 | return -errno; |
599 | } | |
600 | return 0; | |
d5249393 | 601 | #endif |
eba25057 | 602 | } |
fc01f7e7 | 603 | |
84a12e66 CH |
604 | /* |
605 | * Detect host devices. By convention, /dev/cdrom[N] is always | |
606 | * recognized as a host CDROM. | |
607 | */ | |
608 | static BlockDriver *find_hdev_driver(const char *filename) | |
609 | { | |
610 | int score_max = 0, score; | |
611 | BlockDriver *drv = NULL, *d; | |
612 | ||
613 | QLIST_FOREACH(d, &bdrv_drivers, list) { | |
614 | if (d->bdrv_probe_device) { | |
615 | score = d->bdrv_probe_device(filename); | |
616 | if (score > score_max) { | |
617 | score_max = score; | |
618 | drv = d; | |
619 | } | |
620 | } | |
621 | } | |
622 | ||
623 | return drv; | |
624 | } | |
625 | ||
98289620 KW |
626 | BlockDriver *bdrv_find_protocol(const char *filename, |
627 | bool allow_protocol_prefix) | |
83f64091 FB |
628 | { |
629 | BlockDriver *drv1; | |
630 | char protocol[128]; | |
1cec71e3 | 631 | int len; |
83f64091 | 632 | const char *p; |
19cb3738 | 633 | |
66f82cee KW |
634 | /* TODO Drivers without bdrv_file_open must be specified explicitly */ |
635 | ||
39508e7a CH |
636 | /* |
637 | * XXX(hch): we really should not let host device detection | |
638 | * override an explicit protocol specification, but moving this | |
639 | * later breaks access to device names with colons in them. | |
640 | * Thanks to the brain-dead persistent naming schemes on udev- | |
641 | * based Linux systems those actually are quite common. | |
642 | */ | |
643 | drv1 = find_hdev_driver(filename); | |
644 | if (drv1) { | |
645 | return drv1; | |
646 | } | |
647 | ||
98289620 | 648 | if (!path_has_protocol(filename) || !allow_protocol_prefix) { |
ef810437 | 649 | return &bdrv_file; |
84a12e66 | 650 | } |
98289620 | 651 | |
9e0b22f4 SH |
652 | p = strchr(filename, ':'); |
653 | assert(p != NULL); | |
1cec71e3 AL |
654 | len = p - filename; |
655 | if (len > sizeof(protocol) - 1) | |
656 | len = sizeof(protocol) - 1; | |
657 | memcpy(protocol, filename, len); | |
658 | protocol[len] = '\0'; | |
8a22f02a | 659 | QLIST_FOREACH(drv1, &bdrv_drivers, list) { |
5fafdf24 | 660 | if (drv1->protocol_name && |
8a22f02a | 661 | !strcmp(drv1->protocol_name, protocol)) { |
83f64091 | 662 | return drv1; |
8a22f02a | 663 | } |
83f64091 FB |
664 | } |
665 | return NULL; | |
666 | } | |
667 | ||
c6684249 MA |
668 | /* |
669 | * Guess image format by probing its contents. | |
670 | * This is not a good idea when your image is raw (CVE-2008-2004), but | |
671 | * we do it anyway for backward compatibility. | |
672 | * | |
673 | * @buf contains the image's first @buf_size bytes. | |
7cddd372 KW |
674 | * @buf_size is the buffer size in bytes (generally BLOCK_PROBE_BUF_SIZE, |
675 | * but can be smaller if the image file is smaller) | |
c6684249 MA |
676 | * @filename is its filename. |
677 | * | |
678 | * For all block drivers, call the bdrv_probe() method to get its | |
679 | * probing score. | |
680 | * Return the first block driver with the highest probing score. | |
681 | */ | |
38f3ef57 KW |
682 | BlockDriver *bdrv_probe_all(const uint8_t *buf, int buf_size, |
683 | const char *filename) | |
c6684249 MA |
684 | { |
685 | int score_max = 0, score; | |
686 | BlockDriver *drv = NULL, *d; | |
687 | ||
688 | QLIST_FOREACH(d, &bdrv_drivers, list) { | |
689 | if (d->bdrv_probe) { | |
690 | score = d->bdrv_probe(buf, buf_size, filename); | |
691 | if (score > score_max) { | |
692 | score_max = score; | |
693 | drv = d; | |
694 | } | |
695 | } | |
696 | } | |
697 | ||
698 | return drv; | |
699 | } | |
700 | ||
f500a6d3 | 701 | static int find_image_format(BlockDriverState *bs, const char *filename, |
34b5d2c6 | 702 | BlockDriver **pdrv, Error **errp) |
f3a5d3f8 | 703 | { |
c6684249 | 704 | BlockDriver *drv; |
7cddd372 | 705 | uint8_t buf[BLOCK_PROBE_BUF_SIZE]; |
f500a6d3 | 706 | int ret = 0; |
f8ea0b00 | 707 | |
08a00559 | 708 | /* Return the raw BlockDriver * to scsi-generic devices or empty drives */ |
8e895599 | 709 | if (bs->sg || !bdrv_is_inserted(bs) || bdrv_getlength(bs) == 0) { |
ef810437 | 710 | *pdrv = &bdrv_raw; |
c98ac35d | 711 | return ret; |
1a396859 | 712 | } |
f8ea0b00 | 713 | |
83f64091 | 714 | ret = bdrv_pread(bs, 0, buf, sizeof(buf)); |
83f64091 | 715 | if (ret < 0) { |
34b5d2c6 HR |
716 | error_setg_errno(errp, -ret, "Could not read image for determining its " |
717 | "format"); | |
c98ac35d SW |
718 | *pdrv = NULL; |
719 | return ret; | |
83f64091 FB |
720 | } |
721 | ||
c6684249 | 722 | drv = bdrv_probe_all(buf, ret, filename); |
c98ac35d | 723 | if (!drv) { |
34b5d2c6 HR |
724 | error_setg(errp, "Could not determine image format: No compatible " |
725 | "driver found"); | |
c98ac35d SW |
726 | ret = -ENOENT; |
727 | } | |
728 | *pdrv = drv; | |
729 | return ret; | |
ea2384d3 FB |
730 | } |
731 | ||
51762288 SH |
732 | /** |
733 | * Set the current 'total_sectors' value | |
65a9bb25 | 734 | * Return 0 on success, -errno on error. |
51762288 SH |
735 | */ |
736 | static int refresh_total_sectors(BlockDriverState *bs, int64_t hint) | |
737 | { | |
738 | BlockDriver *drv = bs->drv; | |
739 | ||
396759ad NB |
740 | /* Do not attempt drv->bdrv_getlength() on scsi-generic devices */ |
741 | if (bs->sg) | |
742 | return 0; | |
743 | ||
51762288 SH |
744 | /* query actual device if possible, otherwise just trust the hint */ |
745 | if (drv->bdrv_getlength) { | |
746 | int64_t length = drv->bdrv_getlength(bs); | |
747 | if (length < 0) { | |
748 | return length; | |
749 | } | |
7e382003 | 750 | hint = DIV_ROUND_UP(length, BDRV_SECTOR_SIZE); |
51762288 SH |
751 | } |
752 | ||
753 | bs->total_sectors = hint; | |
754 | return 0; | |
755 | } | |
756 | ||
9e8f1835 PB |
757 | /** |
758 | * Set open flags for a given discard mode | |
759 | * | |
760 | * Return 0 on success, -1 if the discard mode was invalid. | |
761 | */ | |
762 | int bdrv_parse_discard_flags(const char *mode, int *flags) | |
763 | { | |
764 | *flags &= ~BDRV_O_UNMAP; | |
765 | ||
766 | if (!strcmp(mode, "off") || !strcmp(mode, "ignore")) { | |
767 | /* do nothing */ | |
768 | } else if (!strcmp(mode, "on") || !strcmp(mode, "unmap")) { | |
769 | *flags |= BDRV_O_UNMAP; | |
770 | } else { | |
771 | return -1; | |
772 | } | |
773 | ||
774 | return 0; | |
775 | } | |
776 | ||
c3993cdc SH |
777 | /** |
778 | * Set open flags for a given cache mode | |
779 | * | |
780 | * Return 0 on success, -1 if the cache mode was invalid. | |
781 | */ | |
782 | int bdrv_parse_cache_flags(const char *mode, int *flags) | |
783 | { | |
784 | *flags &= ~BDRV_O_CACHE_MASK; | |
785 | ||
786 | if (!strcmp(mode, "off") || !strcmp(mode, "none")) { | |
787 | *flags |= BDRV_O_NOCACHE | BDRV_O_CACHE_WB; | |
92196b2f SH |
788 | } else if (!strcmp(mode, "directsync")) { |
789 | *flags |= BDRV_O_NOCACHE; | |
c3993cdc SH |
790 | } else if (!strcmp(mode, "writeback")) { |
791 | *flags |= BDRV_O_CACHE_WB; | |
792 | } else if (!strcmp(mode, "unsafe")) { | |
793 | *flags |= BDRV_O_CACHE_WB; | |
794 | *flags |= BDRV_O_NO_FLUSH; | |
795 | } else if (!strcmp(mode, "writethrough")) { | |
796 | /* this is the default */ | |
797 | } else { | |
798 | return -1; | |
799 | } | |
800 | ||
801 | return 0; | |
802 | } | |
803 | ||
53fec9d3 SH |
804 | /** |
805 | * The copy-on-read flag is actually a reference count so multiple users may | |
806 | * use the feature without worrying about clobbering its previous state. | |
807 | * Copy-on-read stays enabled until all users have called to disable it. | |
808 | */ | |
809 | void bdrv_enable_copy_on_read(BlockDriverState *bs) | |
810 | { | |
811 | bs->copy_on_read++; | |
812 | } | |
813 | ||
814 | void bdrv_disable_copy_on_read(BlockDriverState *bs) | |
815 | { | |
816 | assert(bs->copy_on_read > 0); | |
817 | bs->copy_on_read--; | |
818 | } | |
819 | ||
b1e6fc08 KW |
820 | /* |
821 | * Returns the flags that a temporary snapshot should get, based on the | |
822 | * originally requested flags (the originally requested image will have flags | |
823 | * like a backing file) | |
824 | */ | |
825 | static int bdrv_temp_snapshot_flags(int flags) | |
826 | { | |
827 | return (flags & ~BDRV_O_SNAPSHOT) | BDRV_O_TEMPORARY; | |
828 | } | |
829 | ||
0b50cc88 KW |
830 | /* |
831 | * Returns the flags that bs->file should get, based on the given flags for | |
832 | * the parent BDS | |
833 | */ | |
834 | static int bdrv_inherited_flags(int flags) | |
835 | { | |
836 | /* Enable protocol handling, disable format probing for bs->file */ | |
837 | flags |= BDRV_O_PROTOCOL; | |
838 | ||
839 | /* Our block drivers take care to send flushes and respect unmap policy, | |
840 | * so we can enable both unconditionally on lower layers. */ | |
841 | flags |= BDRV_O_CACHE_WB | BDRV_O_UNMAP; | |
842 | ||
0b50cc88 | 843 | /* Clear flags that only apply to the top layer */ |
5669b44d | 844 | flags &= ~(BDRV_O_SNAPSHOT | BDRV_O_NO_BACKING | BDRV_O_COPY_ON_READ); |
0b50cc88 KW |
845 | |
846 | return flags; | |
847 | } | |
848 | ||
317fc44e KW |
849 | /* |
850 | * Returns the flags that bs->backing_hd should get, based on the given flags | |
851 | * for the parent BDS | |
852 | */ | |
853 | static int bdrv_backing_flags(int flags) | |
854 | { | |
855 | /* backing files always opened read-only */ | |
856 | flags &= ~(BDRV_O_RDWR | BDRV_O_COPY_ON_READ); | |
857 | ||
858 | /* snapshot=on is handled on the top layer */ | |
8bfea15d | 859 | flags &= ~(BDRV_O_SNAPSHOT | BDRV_O_TEMPORARY); |
317fc44e KW |
860 | |
861 | return flags; | |
862 | } | |
863 | ||
7b272452 KW |
864 | static int bdrv_open_flags(BlockDriverState *bs, int flags) |
865 | { | |
866 | int open_flags = flags | BDRV_O_CACHE_WB; | |
867 | ||
868 | /* | |
869 | * Clear flags that are internal to the block layer before opening the | |
870 | * image. | |
871 | */ | |
20cca275 | 872 | open_flags &= ~(BDRV_O_SNAPSHOT | BDRV_O_NO_BACKING | BDRV_O_PROTOCOL); |
7b272452 KW |
873 | |
874 | /* | |
875 | * Snapshots should be writable. | |
876 | */ | |
8bfea15d | 877 | if (flags & BDRV_O_TEMPORARY) { |
7b272452 KW |
878 | open_flags |= BDRV_O_RDWR; |
879 | } | |
880 | ||
881 | return open_flags; | |
882 | } | |
883 | ||
636ea370 KW |
884 | static void bdrv_assign_node_name(BlockDriverState *bs, |
885 | const char *node_name, | |
886 | Error **errp) | |
6913c0c2 BC |
887 | { |
888 | if (!node_name) { | |
636ea370 | 889 | return; |
6913c0c2 BC |
890 | } |
891 | ||
9aebf3b8 | 892 | /* Check for empty string or invalid characters */ |
f5bebbbb | 893 | if (!id_wellformed(node_name)) { |
9aebf3b8 | 894 | error_setg(errp, "Invalid node name"); |
636ea370 | 895 | return; |
6913c0c2 BC |
896 | } |
897 | ||
0c5e94ee | 898 | /* takes care of avoiding namespaces collisions */ |
7f06d47e | 899 | if (blk_by_name(node_name)) { |
0c5e94ee BC |
900 | error_setg(errp, "node-name=%s is conflicting with a device id", |
901 | node_name); | |
636ea370 | 902 | return; |
0c5e94ee BC |
903 | } |
904 | ||
6913c0c2 BC |
905 | /* takes care of avoiding duplicates node names */ |
906 | if (bdrv_find_node(node_name)) { | |
907 | error_setg(errp, "Duplicate node name"); | |
636ea370 | 908 | return; |
6913c0c2 BC |
909 | } |
910 | ||
911 | /* copy node name into the bs and insert it into the graph list */ | |
912 | pstrcpy(bs->node_name, sizeof(bs->node_name), node_name); | |
913 | QTAILQ_INSERT_TAIL(&graph_bdrv_states, bs, node_list); | |
6913c0c2 BC |
914 | } |
915 | ||
57915332 KW |
916 | /* |
917 | * Common part for opening disk images and files | |
b6ad491a KW |
918 | * |
919 | * Removes all processed options from *options. | |
57915332 | 920 | */ |
f500a6d3 | 921 | static int bdrv_open_common(BlockDriverState *bs, BlockDriverState *file, |
34b5d2c6 | 922 | QDict *options, int flags, BlockDriver *drv, Error **errp) |
57915332 KW |
923 | { |
924 | int ret, open_flags; | |
035fccdf | 925 | const char *filename; |
6913c0c2 | 926 | const char *node_name = NULL; |
34b5d2c6 | 927 | Error *local_err = NULL; |
57915332 KW |
928 | |
929 | assert(drv != NULL); | |
6405875c | 930 | assert(bs->file == NULL); |
707ff828 | 931 | assert(options != NULL && bs->options != options); |
57915332 | 932 | |
45673671 KW |
933 | if (file != NULL) { |
934 | filename = file->filename; | |
935 | } else { | |
936 | filename = qdict_get_try_str(options, "filename"); | |
937 | } | |
938 | ||
765003db KW |
939 | if (drv->bdrv_needs_filename && !filename) { |
940 | error_setg(errp, "The '%s' block driver requires a file name", | |
941 | drv->format_name); | |
942 | return -EINVAL; | |
943 | } | |
944 | ||
45673671 | 945 | trace_bdrv_open_common(bs, filename ?: "", flags, drv->format_name); |
28dcee10 | 946 | |
6913c0c2 | 947 | node_name = qdict_get_try_str(options, "node-name"); |
636ea370 | 948 | bdrv_assign_node_name(bs, node_name, &local_err); |
0fb6395c | 949 | if (local_err) { |
636ea370 KW |
950 | error_propagate(errp, local_err); |
951 | return -EINVAL; | |
6913c0c2 BC |
952 | } |
953 | qdict_del(options, "node-name"); | |
954 | ||
5d186eb0 KW |
955 | /* bdrv_open() with directly using a protocol as drv. This layer is already |
956 | * opened, so assign it to bs (while file becomes a closed BlockDriverState) | |
957 | * and return immediately. */ | |
958 | if (file != NULL && drv->bdrv_file_open) { | |
959 | bdrv_swap(file, bs); | |
960 | return 0; | |
961 | } | |
962 | ||
57915332 | 963 | bs->open_flags = flags; |
1b7fd729 | 964 | bs->guest_block_size = 512; |
c25f53b0 | 965 | bs->request_alignment = 512; |
0d51b4de | 966 | bs->zero_beyond_eof = true; |
b64ec4e4 FZ |
967 | open_flags = bdrv_open_flags(bs, flags); |
968 | bs->read_only = !(open_flags & BDRV_O_RDWR); | |
20cca275 | 969 | bs->growable = !!(flags & BDRV_O_PROTOCOL); |
b64ec4e4 FZ |
970 | |
971 | if (use_bdrv_whitelist && !bdrv_is_whitelisted(drv, bs->read_only)) { | |
8f94a6e4 KW |
972 | error_setg(errp, |
973 | !bs->read_only && bdrv_is_whitelisted(drv, true) | |
974 | ? "Driver '%s' can only be used for read-only devices" | |
975 | : "Driver '%s' is not whitelisted", | |
976 | drv->format_name); | |
b64ec4e4 FZ |
977 | return -ENOTSUP; |
978 | } | |
57915332 | 979 | |
53fec9d3 | 980 | assert(bs->copy_on_read == 0); /* bdrv_new() and bdrv_close() make it so */ |
0ebd24e0 KW |
981 | if (flags & BDRV_O_COPY_ON_READ) { |
982 | if (!bs->read_only) { | |
983 | bdrv_enable_copy_on_read(bs); | |
984 | } else { | |
985 | error_setg(errp, "Can't use copy-on-read on read-only device"); | |
986 | return -EINVAL; | |
987 | } | |
53fec9d3 SH |
988 | } |
989 | ||
c2ad1b0c KW |
990 | if (filename != NULL) { |
991 | pstrcpy(bs->filename, sizeof(bs->filename), filename); | |
992 | } else { | |
993 | bs->filename[0] = '\0'; | |
994 | } | |
91af7014 | 995 | pstrcpy(bs->exact_filename, sizeof(bs->exact_filename), bs->filename); |
57915332 | 996 | |
57915332 | 997 | bs->drv = drv; |
7267c094 | 998 | bs->opaque = g_malloc0(drv->instance_size); |
57915332 | 999 | |
03f541bd | 1000 | bs->enable_write_cache = !!(flags & BDRV_O_CACHE_WB); |
e7c63796 | 1001 | |
66f82cee KW |
1002 | /* Open the image, either directly or using a protocol */ |
1003 | if (drv->bdrv_file_open) { | |
5d186eb0 | 1004 | assert(file == NULL); |
030be321 | 1005 | assert(!drv->bdrv_needs_filename || filename != NULL); |
34b5d2c6 | 1006 | ret = drv->bdrv_file_open(bs, options, open_flags, &local_err); |
f500a6d3 | 1007 | } else { |
2af5ef70 | 1008 | if (file == NULL) { |
34b5d2c6 HR |
1009 | error_setg(errp, "Can't use '%s' as a block driver for the " |
1010 | "protocol level", drv->format_name); | |
2af5ef70 KW |
1011 | ret = -EINVAL; |
1012 | goto free_and_fail; | |
1013 | } | |
f500a6d3 | 1014 | bs->file = file; |
34b5d2c6 | 1015 | ret = drv->bdrv_open(bs, options, open_flags, &local_err); |
66f82cee KW |
1016 | } |
1017 | ||
57915332 | 1018 | if (ret < 0) { |
84d18f06 | 1019 | if (local_err) { |
34b5d2c6 | 1020 | error_propagate(errp, local_err); |
2fa9aa59 DH |
1021 | } else if (bs->filename[0]) { |
1022 | error_setg_errno(errp, -ret, "Could not open '%s'", bs->filename); | |
34b5d2c6 HR |
1023 | } else { |
1024 | error_setg_errno(errp, -ret, "Could not open image"); | |
1025 | } | |
57915332 KW |
1026 | goto free_and_fail; |
1027 | } | |
1028 | ||
51762288 SH |
1029 | ret = refresh_total_sectors(bs, bs->total_sectors); |
1030 | if (ret < 0) { | |
34b5d2c6 | 1031 | error_setg_errno(errp, -ret, "Could not refresh total sector count"); |
51762288 | 1032 | goto free_and_fail; |
57915332 | 1033 | } |
51762288 | 1034 | |
3baca891 KW |
1035 | bdrv_refresh_limits(bs, &local_err); |
1036 | if (local_err) { | |
1037 | error_propagate(errp, local_err); | |
1038 | ret = -EINVAL; | |
1039 | goto free_and_fail; | |
1040 | } | |
1041 | ||
c25f53b0 | 1042 | assert(bdrv_opt_mem_align(bs) != 0); |
47ea2de2 | 1043 | assert((bs->request_alignment != 0) || bs->sg); |
57915332 KW |
1044 | return 0; |
1045 | ||
1046 | free_and_fail: | |
f500a6d3 | 1047 | bs->file = NULL; |
7267c094 | 1048 | g_free(bs->opaque); |
57915332 KW |
1049 | bs->opaque = NULL; |
1050 | bs->drv = NULL; | |
1051 | return ret; | |
1052 | } | |
1053 | ||
5e5c4f63 KW |
1054 | static QDict *parse_json_filename(const char *filename, Error **errp) |
1055 | { | |
1056 | QObject *options_obj; | |
1057 | QDict *options; | |
1058 | int ret; | |
1059 | ||
1060 | ret = strstart(filename, "json:", &filename); | |
1061 | assert(ret); | |
1062 | ||
1063 | options_obj = qobject_from_json(filename); | |
1064 | if (!options_obj) { | |
1065 | error_setg(errp, "Could not parse the JSON options"); | |
1066 | return NULL; | |
1067 | } | |
1068 | ||
1069 | if (qobject_type(options_obj) != QTYPE_QDICT) { | |
1070 | qobject_decref(options_obj); | |
1071 | error_setg(errp, "Invalid JSON object given"); | |
1072 | return NULL; | |
1073 | } | |
1074 | ||
1075 | options = qobject_to_qdict(options_obj); | |
1076 | qdict_flatten(options); | |
1077 | ||
1078 | return options; | |
1079 | } | |
1080 | ||
b6ce07aa | 1081 | /* |
f54120ff KW |
1082 | * Fills in default options for opening images and converts the legacy |
1083 | * filename/flags pair to option QDict entries. | |
b6ce07aa | 1084 | */ |
5e5c4f63 | 1085 | static int bdrv_fill_options(QDict **options, const char **pfilename, int flags, |
17b005f1 | 1086 | BlockDriver *drv, Error **errp) |
ea2384d3 | 1087 | { |
5e5c4f63 | 1088 | const char *filename = *pfilename; |
c2ad1b0c | 1089 | const char *drvname; |
462f5bcf | 1090 | bool protocol = flags & BDRV_O_PROTOCOL; |
e3fa4bfa | 1091 | bool parse_filename = false; |
34b5d2c6 | 1092 | Error *local_err = NULL; |
83f64091 | 1093 | |
5e5c4f63 KW |
1094 | /* Parse json: pseudo-protocol */ |
1095 | if (filename && g_str_has_prefix(filename, "json:")) { | |
1096 | QDict *json_options = parse_json_filename(filename, &local_err); | |
1097 | if (local_err) { | |
1098 | error_propagate(errp, local_err); | |
1099 | return -EINVAL; | |
1100 | } | |
1101 | ||
1102 | /* Options given in the filename have lower priority than options | |
1103 | * specified directly */ | |
1104 | qdict_join(*options, json_options, false); | |
1105 | QDECREF(json_options); | |
1106 | *pfilename = filename = NULL; | |
1107 | } | |
1108 | ||
035fccdf | 1109 | /* Fetch the file name from the options QDict if necessary */ |
17b005f1 | 1110 | if (protocol && filename) { |
f54120ff KW |
1111 | if (!qdict_haskey(*options, "filename")) { |
1112 | qdict_put(*options, "filename", qstring_from_str(filename)); | |
1113 | parse_filename = true; | |
1114 | } else { | |
1115 | error_setg(errp, "Can't specify 'file' and 'filename' options at " | |
1116 | "the same time"); | |
1117 | return -EINVAL; | |
1118 | } | |
035fccdf KW |
1119 | } |
1120 | ||
c2ad1b0c | 1121 | /* Find the right block driver */ |
f54120ff | 1122 | filename = qdict_get_try_str(*options, "filename"); |
5acd9d81 | 1123 | drvname = qdict_get_try_str(*options, "driver"); |
f54120ff | 1124 | |
17b005f1 KW |
1125 | if (drv) { |
1126 | if (drvname) { | |
1127 | error_setg(errp, "Driver specified twice"); | |
1128 | return -EINVAL; | |
1129 | } | |
1130 | drvname = drv->format_name; | |
1131 | qdict_put(*options, "driver", qstring_from_str(drvname)); | |
1132 | } else { | |
1133 | if (!drvname && protocol) { | |
1134 | if (filename) { | |
1135 | drv = bdrv_find_protocol(filename, parse_filename); | |
1136 | if (!drv) { | |
1137 | error_setg(errp, "Unknown protocol"); | |
1138 | return -EINVAL; | |
1139 | } | |
1140 | ||
1141 | drvname = drv->format_name; | |
1142 | qdict_put(*options, "driver", qstring_from_str(drvname)); | |
1143 | } else { | |
1144 | error_setg(errp, "Must specify either driver or file"); | |
f54120ff KW |
1145 | return -EINVAL; |
1146 | } | |
17b005f1 KW |
1147 | } else if (drvname) { |
1148 | drv = bdrv_find_format(drvname); | |
1149 | if (!drv) { | |
1150 | error_setg(errp, "Unknown driver '%s'", drvname); | |
1151 | return -ENOENT; | |
1152 | } | |
98289620 | 1153 | } |
c2ad1b0c KW |
1154 | } |
1155 | ||
17b005f1 | 1156 | assert(drv || !protocol); |
c2ad1b0c | 1157 | |
f54120ff | 1158 | /* Driver-specific filename parsing */ |
17b005f1 | 1159 | if (drv && drv->bdrv_parse_filename && parse_filename) { |
5acd9d81 | 1160 | drv->bdrv_parse_filename(filename, *options, &local_err); |
84d18f06 | 1161 | if (local_err) { |
34b5d2c6 | 1162 | error_propagate(errp, local_err); |
f54120ff | 1163 | return -EINVAL; |
6963a30d | 1164 | } |
cd5d031e HR |
1165 | |
1166 | if (!drv->bdrv_needs_filename) { | |
1167 | qdict_del(*options, "filename"); | |
cd5d031e | 1168 | } |
6963a30d KW |
1169 | } |
1170 | ||
f54120ff KW |
1171 | return 0; |
1172 | } | |
1173 | ||
8d24cce1 FZ |
1174 | void bdrv_set_backing_hd(BlockDriverState *bs, BlockDriverState *backing_hd) |
1175 | { | |
1176 | ||
826b6ca0 FZ |
1177 | if (bs->backing_hd) { |
1178 | assert(bs->backing_blocker); | |
1179 | bdrv_op_unblock_all(bs->backing_hd, bs->backing_blocker); | |
1180 | } else if (backing_hd) { | |
1181 | error_setg(&bs->backing_blocker, | |
1182 | "device is used as backing hd of '%s'", | |
bfb197e0 | 1183 | bdrv_get_device_name(bs)); |
826b6ca0 FZ |
1184 | } |
1185 | ||
8d24cce1 FZ |
1186 | bs->backing_hd = backing_hd; |
1187 | if (!backing_hd) { | |
826b6ca0 FZ |
1188 | error_free(bs->backing_blocker); |
1189 | bs->backing_blocker = NULL; | |
8d24cce1 FZ |
1190 | goto out; |
1191 | } | |
1192 | bs->open_flags &= ~BDRV_O_NO_BACKING; | |
1193 | pstrcpy(bs->backing_file, sizeof(bs->backing_file), backing_hd->filename); | |
1194 | pstrcpy(bs->backing_format, sizeof(bs->backing_format), | |
1195 | backing_hd->drv ? backing_hd->drv->format_name : ""); | |
826b6ca0 FZ |
1196 | |
1197 | bdrv_op_block_all(bs->backing_hd, bs->backing_blocker); | |
1198 | /* Otherwise we won't be able to commit due to check in bdrv_commit */ | |
1199 | bdrv_op_unblock(bs->backing_hd, BLOCK_OP_TYPE_COMMIT, | |
1200 | bs->backing_blocker); | |
8d24cce1 | 1201 | out: |
3baca891 | 1202 | bdrv_refresh_limits(bs, NULL); |
8d24cce1 FZ |
1203 | } |
1204 | ||
31ca6d07 KW |
1205 | /* |
1206 | * Opens the backing file for a BlockDriverState if not yet open | |
1207 | * | |
1208 | * options is a QDict of options to pass to the block drivers, or NULL for an | |
1209 | * empty set of options. The reference to the QDict is transferred to this | |
1210 | * function (even on failure), so if the caller intends to reuse the dictionary, | |
1211 | * it needs to use QINCREF() before calling bdrv_file_open. | |
1212 | */ | |
34b5d2c6 | 1213 | int bdrv_open_backing_file(BlockDriverState *bs, QDict *options, Error **errp) |
9156df12 | 1214 | { |
1ba4b6a5 | 1215 | char *backing_filename = g_malloc0(PATH_MAX); |
317fc44e | 1216 | int ret = 0; |
8d24cce1 | 1217 | BlockDriverState *backing_hd; |
34b5d2c6 | 1218 | Error *local_err = NULL; |
9156df12 PB |
1219 | |
1220 | if (bs->backing_hd != NULL) { | |
31ca6d07 | 1221 | QDECREF(options); |
1ba4b6a5 | 1222 | goto free_exit; |
9156df12 PB |
1223 | } |
1224 | ||
31ca6d07 KW |
1225 | /* NULL means an empty set of options */ |
1226 | if (options == NULL) { | |
1227 | options = qdict_new(); | |
1228 | } | |
1229 | ||
9156df12 | 1230 | bs->open_flags &= ~BDRV_O_NO_BACKING; |
1cb6f506 KW |
1231 | if (qdict_haskey(options, "file.filename")) { |
1232 | backing_filename[0] = '\0'; | |
1233 | } else if (bs->backing_file[0] == '\0' && qdict_size(options) == 0) { | |
31ca6d07 | 1234 | QDECREF(options); |
1ba4b6a5 | 1235 | goto free_exit; |
dbecebdd | 1236 | } else { |
9f07429e HR |
1237 | bdrv_get_full_backing_filename(bs, backing_filename, PATH_MAX, |
1238 | &local_err); | |
1239 | if (local_err) { | |
1240 | ret = -EINVAL; | |
1241 | error_propagate(errp, local_err); | |
1242 | QDECREF(options); | |
1243 | goto free_exit; | |
1244 | } | |
9156df12 PB |
1245 | } |
1246 | ||
8ee79e70 KW |
1247 | if (!bs->drv || !bs->drv->supports_backing) { |
1248 | ret = -EINVAL; | |
1249 | error_setg(errp, "Driver doesn't support backing files"); | |
1250 | QDECREF(options); | |
1251 | goto free_exit; | |
1252 | } | |
1253 | ||
e4e9986b | 1254 | backing_hd = bdrv_new(); |
8d24cce1 | 1255 | |
c5f6e493 KW |
1256 | if (bs->backing_format[0] != '\0' && !qdict_haskey(options, "driver")) { |
1257 | qdict_put(options, "driver", qstring_from_str(bs->backing_format)); | |
9156df12 PB |
1258 | } |
1259 | ||
f67503e5 | 1260 | assert(bs->backing_hd == NULL); |
8d24cce1 | 1261 | ret = bdrv_open(&backing_hd, |
ddf5636d | 1262 | *backing_filename ? backing_filename : NULL, NULL, options, |
c5f6e493 | 1263 | bdrv_backing_flags(bs->open_flags), NULL, &local_err); |
9156df12 | 1264 | if (ret < 0) { |
8d24cce1 FZ |
1265 | bdrv_unref(backing_hd); |
1266 | backing_hd = NULL; | |
9156df12 | 1267 | bs->open_flags |= BDRV_O_NO_BACKING; |
b04b6b6e FZ |
1268 | error_setg(errp, "Could not open backing file: %s", |
1269 | error_get_pretty(local_err)); | |
1270 | error_free(local_err); | |
1ba4b6a5 | 1271 | goto free_exit; |
9156df12 | 1272 | } |
8d24cce1 | 1273 | bdrv_set_backing_hd(bs, backing_hd); |
d80ac658 | 1274 | |
1ba4b6a5 BC |
1275 | free_exit: |
1276 | g_free(backing_filename); | |
1277 | return ret; | |
9156df12 PB |
1278 | } |
1279 | ||
da557aac HR |
1280 | /* |
1281 | * Opens a disk image whose options are given as BlockdevRef in another block | |
1282 | * device's options. | |
1283 | * | |
da557aac HR |
1284 | * If allow_none is true, no image will be opened if filename is false and no |
1285 | * BlockdevRef is given. *pbs will remain unchanged and 0 will be returned. | |
1286 | * | |
1287 | * bdrev_key specifies the key for the image's BlockdevRef in the options QDict. | |
1288 | * That QDict has to be flattened; therefore, if the BlockdevRef is a QDict | |
1289 | * itself, all options starting with "${bdref_key}." are considered part of the | |
1290 | * BlockdevRef. | |
1291 | * | |
1292 | * The BlockdevRef will be removed from the options QDict. | |
f67503e5 HR |
1293 | * |
1294 | * To conform with the behavior of bdrv_open(), *pbs has to be NULL. | |
da557aac HR |
1295 | */ |
1296 | int bdrv_open_image(BlockDriverState **pbs, const char *filename, | |
1297 | QDict *options, const char *bdref_key, int flags, | |
f7d9fd8c | 1298 | bool allow_none, Error **errp) |
da557aac HR |
1299 | { |
1300 | QDict *image_options; | |
1301 | int ret; | |
1302 | char *bdref_key_dot; | |
1303 | const char *reference; | |
1304 | ||
f67503e5 HR |
1305 | assert(pbs); |
1306 | assert(*pbs == NULL); | |
1307 | ||
da557aac HR |
1308 | bdref_key_dot = g_strdup_printf("%s.", bdref_key); |
1309 | qdict_extract_subqdict(options, &image_options, bdref_key_dot); | |
1310 | g_free(bdref_key_dot); | |
1311 | ||
1312 | reference = qdict_get_try_str(options, bdref_key); | |
1313 | if (!filename && !reference && !qdict_size(image_options)) { | |
1314 | if (allow_none) { | |
1315 | ret = 0; | |
1316 | } else { | |
1317 | error_setg(errp, "A block device must be specified for \"%s\"", | |
1318 | bdref_key); | |
1319 | ret = -EINVAL; | |
1320 | } | |
b20e61e0 | 1321 | QDECREF(image_options); |
da557aac HR |
1322 | goto done; |
1323 | } | |
1324 | ||
f7d9fd8c | 1325 | ret = bdrv_open(pbs, filename, reference, image_options, flags, NULL, errp); |
da557aac HR |
1326 | |
1327 | done: | |
1328 | qdict_del(options, bdref_key); | |
1329 | return ret; | |
1330 | } | |
1331 | ||
6b8aeca5 | 1332 | int bdrv_append_temp_snapshot(BlockDriverState *bs, int flags, Error **errp) |
b998875d KW |
1333 | { |
1334 | /* TODO: extra byte is a hack to ensure MAX_PATH space on Windows. */ | |
1ba4b6a5 | 1335 | char *tmp_filename = g_malloc0(PATH_MAX + 1); |
b998875d | 1336 | int64_t total_size; |
83d0521a | 1337 | QemuOpts *opts = NULL; |
b998875d KW |
1338 | QDict *snapshot_options; |
1339 | BlockDriverState *bs_snapshot; | |
1340 | Error *local_err; | |
1341 | int ret; | |
1342 | ||
1343 | /* if snapshot, we create a temporary backing file and open it | |
1344 | instead of opening 'filename' directly */ | |
1345 | ||
1346 | /* Get the required size from the image */ | |
f187743a KW |
1347 | total_size = bdrv_getlength(bs); |
1348 | if (total_size < 0) { | |
6b8aeca5 | 1349 | ret = total_size; |
f187743a | 1350 | error_setg_errno(errp, -total_size, "Could not get image size"); |
1ba4b6a5 | 1351 | goto out; |
f187743a | 1352 | } |
b998875d KW |
1353 | |
1354 | /* Create the temporary image */ | |
1ba4b6a5 | 1355 | ret = get_tmp_filename(tmp_filename, PATH_MAX + 1); |
b998875d KW |
1356 | if (ret < 0) { |
1357 | error_setg_errno(errp, -ret, "Could not get temporary filename"); | |
1ba4b6a5 | 1358 | goto out; |
b998875d KW |
1359 | } |
1360 | ||
ef810437 | 1361 | opts = qemu_opts_create(bdrv_qcow2.create_opts, NULL, 0, |
c282e1fd | 1362 | &error_abort); |
83d0521a | 1363 | qemu_opt_set_number(opts, BLOCK_OPT_SIZE, total_size); |
ef810437 | 1364 | ret = bdrv_create(&bdrv_qcow2, tmp_filename, opts, &local_err); |
83d0521a | 1365 | qemu_opts_del(opts); |
b998875d KW |
1366 | if (ret < 0) { |
1367 | error_setg_errno(errp, -ret, "Could not create temporary overlay " | |
1368 | "'%s': %s", tmp_filename, | |
1369 | error_get_pretty(local_err)); | |
1370 | error_free(local_err); | |
1ba4b6a5 | 1371 | goto out; |
b998875d KW |
1372 | } |
1373 | ||
1374 | /* Prepare a new options QDict for the temporary file */ | |
1375 | snapshot_options = qdict_new(); | |
1376 | qdict_put(snapshot_options, "file.driver", | |
1377 | qstring_from_str("file")); | |
1378 | qdict_put(snapshot_options, "file.filename", | |
1379 | qstring_from_str(tmp_filename)); | |
1380 | ||
e4e9986b | 1381 | bs_snapshot = bdrv_new(); |
b998875d KW |
1382 | |
1383 | ret = bdrv_open(&bs_snapshot, NULL, NULL, snapshot_options, | |
ef810437 | 1384 | flags, &bdrv_qcow2, &local_err); |
b998875d KW |
1385 | if (ret < 0) { |
1386 | error_propagate(errp, local_err); | |
1ba4b6a5 | 1387 | goto out; |
b998875d KW |
1388 | } |
1389 | ||
1390 | bdrv_append(bs_snapshot, bs); | |
1ba4b6a5 BC |
1391 | |
1392 | out: | |
1393 | g_free(tmp_filename); | |
6b8aeca5 | 1394 | return ret; |
b998875d KW |
1395 | } |
1396 | ||
b6ce07aa KW |
1397 | /* |
1398 | * Opens a disk image (raw, qcow2, vmdk, ...) | |
de9c0cec KW |
1399 | * |
1400 | * options is a QDict of options to pass to the block drivers, or NULL for an | |
1401 | * empty set of options. The reference to the QDict belongs to the block layer | |
1402 | * after the call (even on failure), so if the caller intends to reuse the | |
1403 | * dictionary, it needs to use QINCREF() before calling bdrv_open. | |
f67503e5 HR |
1404 | * |
1405 | * If *pbs is NULL, a new BDS will be created with a pointer to it stored there. | |
1406 | * If it is not NULL, the referenced BDS will be reused. | |
ddf5636d HR |
1407 | * |
1408 | * The reference parameter may be used to specify an existing block device which | |
1409 | * should be opened. If specified, neither options nor a filename may be given, | |
1410 | * nor can an existing BDS be reused (that is, *pbs has to be NULL). | |
b6ce07aa | 1411 | */ |
ddf5636d HR |
1412 | int bdrv_open(BlockDriverState **pbs, const char *filename, |
1413 | const char *reference, QDict *options, int flags, | |
1414 | BlockDriver *drv, Error **errp) | |
ea2384d3 | 1415 | { |
b6ce07aa | 1416 | int ret; |
f67503e5 | 1417 | BlockDriverState *file = NULL, *bs; |
74fe54f2 | 1418 | const char *drvname; |
34b5d2c6 | 1419 | Error *local_err = NULL; |
b1e6fc08 | 1420 | int snapshot_flags = 0; |
712e7874 | 1421 | |
f67503e5 HR |
1422 | assert(pbs); |
1423 | ||
ddf5636d HR |
1424 | if (reference) { |
1425 | bool options_non_empty = options ? qdict_size(options) : false; | |
1426 | QDECREF(options); | |
1427 | ||
1428 | if (*pbs) { | |
1429 | error_setg(errp, "Cannot reuse an existing BDS when referencing " | |
1430 | "another block device"); | |
1431 | return -EINVAL; | |
1432 | } | |
1433 | ||
1434 | if (filename || options_non_empty) { | |
1435 | error_setg(errp, "Cannot reference an existing block device with " | |
1436 | "additional options or a new filename"); | |
1437 | return -EINVAL; | |
1438 | } | |
1439 | ||
1440 | bs = bdrv_lookup_bs(reference, reference, errp); | |
1441 | if (!bs) { | |
1442 | return -ENODEV; | |
1443 | } | |
1444 | bdrv_ref(bs); | |
1445 | *pbs = bs; | |
1446 | return 0; | |
1447 | } | |
1448 | ||
f67503e5 HR |
1449 | if (*pbs) { |
1450 | bs = *pbs; | |
1451 | } else { | |
e4e9986b | 1452 | bs = bdrv_new(); |
f67503e5 HR |
1453 | } |
1454 | ||
de9c0cec KW |
1455 | /* NULL means an empty set of options */ |
1456 | if (options == NULL) { | |
1457 | options = qdict_new(); | |
1458 | } | |
1459 | ||
17b005f1 | 1460 | ret = bdrv_fill_options(&options, &filename, flags, drv, &local_err); |
462f5bcf KW |
1461 | if (local_err) { |
1462 | goto fail; | |
1463 | } | |
1464 | ||
76c591b0 KW |
1465 | /* Find the right image format driver */ |
1466 | drv = NULL; | |
1467 | drvname = qdict_get_try_str(options, "driver"); | |
1468 | if (drvname) { | |
1469 | drv = bdrv_find_format(drvname); | |
1470 | qdict_del(options, "driver"); | |
1471 | if (!drv) { | |
1472 | error_setg(errp, "Unknown driver: '%s'", drvname); | |
1473 | ret = -EINVAL; | |
1474 | goto fail; | |
1475 | } | |
1476 | } | |
1477 | ||
1478 | assert(drvname || !(flags & BDRV_O_PROTOCOL)); | |
1479 | if (drv && !drv->bdrv_file_open) { | |
1480 | /* If the user explicitly wants a format driver here, we'll need to add | |
1481 | * another layer for the protocol in bs->file */ | |
1482 | flags &= ~BDRV_O_PROTOCOL; | |
1483 | } | |
1484 | ||
de9c0cec | 1485 | bs->options = options; |
b6ad491a | 1486 | options = qdict_clone_shallow(options); |
de9c0cec | 1487 | |
f500a6d3 | 1488 | /* Open image file without format layer */ |
f4788adc KW |
1489 | if ((flags & BDRV_O_PROTOCOL) == 0) { |
1490 | if (flags & BDRV_O_RDWR) { | |
1491 | flags |= BDRV_O_ALLOW_RDWR; | |
1492 | } | |
1493 | if (flags & BDRV_O_SNAPSHOT) { | |
1494 | snapshot_flags = bdrv_temp_snapshot_flags(flags); | |
1495 | flags = bdrv_backing_flags(flags); | |
1496 | } | |
f500a6d3 | 1497 | |
f4788adc KW |
1498 | assert(file == NULL); |
1499 | ret = bdrv_open_image(&file, filename, options, "file", | |
1500 | bdrv_inherited_flags(flags), | |
1501 | true, &local_err); | |
1502 | if (ret < 0) { | |
1503 | goto fail; | |
1504 | } | |
f500a6d3 KW |
1505 | } |
1506 | ||
76c591b0 | 1507 | /* Image format probing */ |
38f3ef57 | 1508 | bs->probed = !drv; |
76c591b0 | 1509 | if (!drv && file) { |
17b005f1 KW |
1510 | ret = find_image_format(file, filename, &drv, &local_err); |
1511 | if (ret < 0) { | |
8bfea15d | 1512 | goto fail; |
2a05cbe4 | 1513 | } |
76c591b0 | 1514 | } else if (!drv) { |
17b005f1 KW |
1515 | error_setg(errp, "Must specify either driver or file"); |
1516 | ret = -EINVAL; | |
8bfea15d | 1517 | goto fail; |
ea2384d3 | 1518 | } |
b6ce07aa KW |
1519 | |
1520 | /* Open the image */ | |
34b5d2c6 | 1521 | ret = bdrv_open_common(bs, file, options, flags, drv, &local_err); |
b6ce07aa | 1522 | if (ret < 0) { |
8bfea15d | 1523 | goto fail; |
6987307c CH |
1524 | } |
1525 | ||
2a05cbe4 | 1526 | if (file && (bs->file != file)) { |
4f6fd349 | 1527 | bdrv_unref(file); |
f500a6d3 KW |
1528 | file = NULL; |
1529 | } | |
1530 | ||
b6ce07aa | 1531 | /* If there is a backing file, use it */ |
9156df12 | 1532 | if ((flags & BDRV_O_NO_BACKING) == 0) { |
31ca6d07 KW |
1533 | QDict *backing_options; |
1534 | ||
5726d872 | 1535 | qdict_extract_subqdict(options, &backing_options, "backing."); |
34b5d2c6 | 1536 | ret = bdrv_open_backing_file(bs, backing_options, &local_err); |
b6ce07aa | 1537 | if (ret < 0) { |
b6ad491a | 1538 | goto close_and_fail; |
b6ce07aa | 1539 | } |
b6ce07aa KW |
1540 | } |
1541 | ||
91af7014 HR |
1542 | bdrv_refresh_filename(bs); |
1543 | ||
b998875d KW |
1544 | /* For snapshot=on, create a temporary qcow2 overlay. bs points to the |
1545 | * temporary snapshot afterwards. */ | |
b1e6fc08 | 1546 | if (snapshot_flags) { |
6b8aeca5 | 1547 | ret = bdrv_append_temp_snapshot(bs, snapshot_flags, &local_err); |
b998875d | 1548 | if (local_err) { |
b998875d KW |
1549 | goto close_and_fail; |
1550 | } | |
1551 | } | |
1552 | ||
b6ad491a | 1553 | /* Check if any unknown options were used */ |
5acd9d81 | 1554 | if (options && (qdict_size(options) != 0)) { |
b6ad491a | 1555 | const QDictEntry *entry = qdict_first(options); |
5acd9d81 HR |
1556 | if (flags & BDRV_O_PROTOCOL) { |
1557 | error_setg(errp, "Block protocol '%s' doesn't support the option " | |
1558 | "'%s'", drv->format_name, entry->key); | |
1559 | } else { | |
1560 | error_setg(errp, "Block format '%s' used by device '%s' doesn't " | |
1561 | "support the option '%s'", drv->format_name, | |
bfb197e0 | 1562 | bdrv_get_device_name(bs), entry->key); |
5acd9d81 | 1563 | } |
b6ad491a KW |
1564 | |
1565 | ret = -EINVAL; | |
1566 | goto close_and_fail; | |
1567 | } | |
b6ad491a | 1568 | |
b6ce07aa | 1569 | if (!bdrv_key_required(bs)) { |
a7f53e26 MA |
1570 | if (bs->blk) { |
1571 | blk_dev_change_media_cb(bs->blk, true); | |
1572 | } | |
c3adb58f MA |
1573 | } else if (!runstate_check(RUN_STATE_PRELAUNCH) |
1574 | && !runstate_check(RUN_STATE_INMIGRATE) | |
1575 | && !runstate_check(RUN_STATE_PAUSED)) { /* HACK */ | |
1576 | error_setg(errp, | |
1577 | "Guest must be stopped for opening of encrypted image"); | |
1578 | ret = -EBUSY; | |
1579 | goto close_and_fail; | |
b6ce07aa KW |
1580 | } |
1581 | ||
c3adb58f | 1582 | QDECREF(options); |
f67503e5 | 1583 | *pbs = bs; |
b6ce07aa KW |
1584 | return 0; |
1585 | ||
8bfea15d | 1586 | fail: |
f500a6d3 | 1587 | if (file != NULL) { |
4f6fd349 | 1588 | bdrv_unref(file); |
f500a6d3 | 1589 | } |
de9c0cec | 1590 | QDECREF(bs->options); |
b6ad491a | 1591 | QDECREF(options); |
de9c0cec | 1592 | bs->options = NULL; |
f67503e5 HR |
1593 | if (!*pbs) { |
1594 | /* If *pbs is NULL, a new BDS has been created in this function and | |
1595 | needs to be freed now. Otherwise, it does not need to be closed, | |
1596 | since it has not really been opened yet. */ | |
1597 | bdrv_unref(bs); | |
1598 | } | |
84d18f06 | 1599 | if (local_err) { |
34b5d2c6 HR |
1600 | error_propagate(errp, local_err); |
1601 | } | |
b6ad491a | 1602 | return ret; |
de9c0cec | 1603 | |
b6ad491a | 1604 | close_and_fail: |
f67503e5 HR |
1605 | /* See fail path, but now the BDS has to be always closed */ |
1606 | if (*pbs) { | |
1607 | bdrv_close(bs); | |
1608 | } else { | |
1609 | bdrv_unref(bs); | |
1610 | } | |
b6ad491a | 1611 | QDECREF(options); |
84d18f06 | 1612 | if (local_err) { |
34b5d2c6 HR |
1613 | error_propagate(errp, local_err); |
1614 | } | |
b6ce07aa KW |
1615 | return ret; |
1616 | } | |
1617 | ||
e971aa12 JC |
1618 | typedef struct BlockReopenQueueEntry { |
1619 | bool prepared; | |
1620 | BDRVReopenState state; | |
1621 | QSIMPLEQ_ENTRY(BlockReopenQueueEntry) entry; | |
1622 | } BlockReopenQueueEntry; | |
1623 | ||
1624 | /* | |
1625 | * Adds a BlockDriverState to a simple queue for an atomic, transactional | |
1626 | * reopen of multiple devices. | |
1627 | * | |
1628 | * bs_queue can either be an existing BlockReopenQueue that has had QSIMPLE_INIT | |
1629 | * already performed, or alternatively may be NULL a new BlockReopenQueue will | |
1630 | * be created and initialized. This newly created BlockReopenQueue should be | |
1631 | * passed back in for subsequent calls that are intended to be of the same | |
1632 | * atomic 'set'. | |
1633 | * | |
1634 | * bs is the BlockDriverState to add to the reopen queue. | |
1635 | * | |
1636 | * flags contains the open flags for the associated bs | |
1637 | * | |
1638 | * returns a pointer to bs_queue, which is either the newly allocated | |
1639 | * bs_queue, or the existing bs_queue being used. | |
1640 | * | |
1641 | */ | |
1642 | BlockReopenQueue *bdrv_reopen_queue(BlockReopenQueue *bs_queue, | |
1643 | BlockDriverState *bs, int flags) | |
1644 | { | |
1645 | assert(bs != NULL); | |
1646 | ||
1647 | BlockReopenQueueEntry *bs_entry; | |
1648 | if (bs_queue == NULL) { | |
1649 | bs_queue = g_new0(BlockReopenQueue, 1); | |
1650 | QSIMPLEQ_INIT(bs_queue); | |
1651 | } | |
1652 | ||
f1f25a2e KW |
1653 | /* bdrv_open() masks this flag out */ |
1654 | flags &= ~BDRV_O_PROTOCOL; | |
1655 | ||
e971aa12 | 1656 | if (bs->file) { |
f1f25a2e | 1657 | bdrv_reopen_queue(bs_queue, bs->file, bdrv_inherited_flags(flags)); |
e971aa12 JC |
1658 | } |
1659 | ||
1660 | bs_entry = g_new0(BlockReopenQueueEntry, 1); | |
1661 | QSIMPLEQ_INSERT_TAIL(bs_queue, bs_entry, entry); | |
1662 | ||
1663 | bs_entry->state.bs = bs; | |
1664 | bs_entry->state.flags = flags; | |
1665 | ||
1666 | return bs_queue; | |
1667 | } | |
1668 | ||
1669 | /* | |
1670 | * Reopen multiple BlockDriverStates atomically & transactionally. | |
1671 | * | |
1672 | * The queue passed in (bs_queue) must have been built up previous | |
1673 | * via bdrv_reopen_queue(). | |
1674 | * | |
1675 | * Reopens all BDS specified in the queue, with the appropriate | |
1676 | * flags. All devices are prepared for reopen, and failure of any | |
1677 | * device will cause all device changes to be abandonded, and intermediate | |
1678 | * data cleaned up. | |
1679 | * | |
1680 | * If all devices prepare successfully, then the changes are committed | |
1681 | * to all devices. | |
1682 | * | |
1683 | */ | |
1684 | int bdrv_reopen_multiple(BlockReopenQueue *bs_queue, Error **errp) | |
1685 | { | |
1686 | int ret = -1; | |
1687 | BlockReopenQueueEntry *bs_entry, *next; | |
1688 | Error *local_err = NULL; | |
1689 | ||
1690 | assert(bs_queue != NULL); | |
1691 | ||
1692 | bdrv_drain_all(); | |
1693 | ||
1694 | QSIMPLEQ_FOREACH(bs_entry, bs_queue, entry) { | |
1695 | if (bdrv_reopen_prepare(&bs_entry->state, bs_queue, &local_err)) { | |
1696 | error_propagate(errp, local_err); | |
1697 | goto cleanup; | |
1698 | } | |
1699 | bs_entry->prepared = true; | |
1700 | } | |
1701 | ||
1702 | /* If we reach this point, we have success and just need to apply the | |
1703 | * changes | |
1704 | */ | |
1705 | QSIMPLEQ_FOREACH(bs_entry, bs_queue, entry) { | |
1706 | bdrv_reopen_commit(&bs_entry->state); | |
1707 | } | |
1708 | ||
1709 | ret = 0; | |
1710 | ||
1711 | cleanup: | |
1712 | QSIMPLEQ_FOREACH_SAFE(bs_entry, bs_queue, entry, next) { | |
1713 | if (ret && bs_entry->prepared) { | |
1714 | bdrv_reopen_abort(&bs_entry->state); | |
1715 | } | |
1716 | g_free(bs_entry); | |
1717 | } | |
1718 | g_free(bs_queue); | |
1719 | return ret; | |
1720 | } | |
1721 | ||
1722 | ||
1723 | /* Reopen a single BlockDriverState with the specified flags. */ | |
1724 | int bdrv_reopen(BlockDriverState *bs, int bdrv_flags, Error **errp) | |
1725 | { | |
1726 | int ret = -1; | |
1727 | Error *local_err = NULL; | |
1728 | BlockReopenQueue *queue = bdrv_reopen_queue(NULL, bs, bdrv_flags); | |
1729 | ||
1730 | ret = bdrv_reopen_multiple(queue, &local_err); | |
1731 | if (local_err != NULL) { | |
1732 | error_propagate(errp, local_err); | |
1733 | } | |
1734 | return ret; | |
1735 | } | |
1736 | ||
1737 | ||
1738 | /* | |
1739 | * Prepares a BlockDriverState for reopen. All changes are staged in the | |
1740 | * 'opaque' field of the BDRVReopenState, which is used and allocated by | |
1741 | * the block driver layer .bdrv_reopen_prepare() | |
1742 | * | |
1743 | * bs is the BlockDriverState to reopen | |
1744 | * flags are the new open flags | |
1745 | * queue is the reopen queue | |
1746 | * | |
1747 | * Returns 0 on success, non-zero on error. On error errp will be set | |
1748 | * as well. | |
1749 | * | |
1750 | * On failure, bdrv_reopen_abort() will be called to clean up any data. | |
1751 | * It is the responsibility of the caller to then call the abort() or | |
1752 | * commit() for any other BDS that have been left in a prepare() state | |
1753 | * | |
1754 | */ | |
1755 | int bdrv_reopen_prepare(BDRVReopenState *reopen_state, BlockReopenQueue *queue, | |
1756 | Error **errp) | |
1757 | { | |
1758 | int ret = -1; | |
1759 | Error *local_err = NULL; | |
1760 | BlockDriver *drv; | |
1761 | ||
1762 | assert(reopen_state != NULL); | |
1763 | assert(reopen_state->bs->drv != NULL); | |
1764 | drv = reopen_state->bs->drv; | |
1765 | ||
1766 | /* if we are to stay read-only, do not allow permission change | |
1767 | * to r/w */ | |
1768 | if (!(reopen_state->bs->open_flags & BDRV_O_ALLOW_RDWR) && | |
1769 | reopen_state->flags & BDRV_O_RDWR) { | |
1770 | error_set(errp, QERR_DEVICE_IS_READ_ONLY, | |
bfb197e0 | 1771 | bdrv_get_device_name(reopen_state->bs)); |
e971aa12 JC |
1772 | goto error; |
1773 | } | |
1774 | ||
1775 | ||
1776 | ret = bdrv_flush(reopen_state->bs); | |
1777 | if (ret) { | |
1778 | error_set(errp, ERROR_CLASS_GENERIC_ERROR, "Error (%s) flushing drive", | |
1779 | strerror(-ret)); | |
1780 | goto error; | |
1781 | } | |
1782 | ||
1783 | if (drv->bdrv_reopen_prepare) { | |
1784 | ret = drv->bdrv_reopen_prepare(reopen_state, queue, &local_err); | |
1785 | if (ret) { | |
1786 | if (local_err != NULL) { | |
1787 | error_propagate(errp, local_err); | |
1788 | } else { | |
d8b6895f LC |
1789 | error_setg(errp, "failed while preparing to reopen image '%s'", |
1790 | reopen_state->bs->filename); | |
e971aa12 JC |
1791 | } |
1792 | goto error; | |
1793 | } | |
1794 | } else { | |
1795 | /* It is currently mandatory to have a bdrv_reopen_prepare() | |
1796 | * handler for each supported drv. */ | |
1797 | error_set(errp, QERR_BLOCK_FORMAT_FEATURE_NOT_SUPPORTED, | |
bfb197e0 | 1798 | drv->format_name, bdrv_get_device_name(reopen_state->bs), |
e971aa12 JC |
1799 | "reopening of file"); |
1800 | ret = -1; | |
1801 | goto error; | |
1802 | } | |
1803 | ||
1804 | ret = 0; | |
1805 | ||
1806 | error: | |
1807 | return ret; | |
1808 | } | |
1809 | ||
1810 | /* | |
1811 | * Takes the staged changes for the reopen from bdrv_reopen_prepare(), and | |
1812 | * makes them final by swapping the staging BlockDriverState contents into | |
1813 | * the active BlockDriverState contents. | |
1814 | */ | |
1815 | void bdrv_reopen_commit(BDRVReopenState *reopen_state) | |
1816 | { | |
1817 | BlockDriver *drv; | |
1818 | ||
1819 | assert(reopen_state != NULL); | |
1820 | drv = reopen_state->bs->drv; | |
1821 | assert(drv != NULL); | |
1822 | ||
1823 | /* If there are any driver level actions to take */ | |
1824 | if (drv->bdrv_reopen_commit) { | |
1825 | drv->bdrv_reopen_commit(reopen_state); | |
1826 | } | |
1827 | ||
1828 | /* set BDS specific flags now */ | |
1829 | reopen_state->bs->open_flags = reopen_state->flags; | |
1830 | reopen_state->bs->enable_write_cache = !!(reopen_state->flags & | |
1831 | BDRV_O_CACHE_WB); | |
1832 | reopen_state->bs->read_only = !(reopen_state->flags & BDRV_O_RDWR); | |
355ef4ac | 1833 | |
3baca891 | 1834 | bdrv_refresh_limits(reopen_state->bs, NULL); |
e971aa12 JC |
1835 | } |
1836 | ||
1837 | /* | |
1838 | * Abort the reopen, and delete and free the staged changes in | |
1839 | * reopen_state | |
1840 | */ | |
1841 | void bdrv_reopen_abort(BDRVReopenState *reopen_state) | |
1842 | { | |
1843 | BlockDriver *drv; | |
1844 | ||
1845 | assert(reopen_state != NULL); | |
1846 | drv = reopen_state->bs->drv; | |
1847 | assert(drv != NULL); | |
1848 | ||
1849 | if (drv->bdrv_reopen_abort) { | |
1850 | drv->bdrv_reopen_abort(reopen_state); | |
1851 | } | |
1852 | } | |
1853 | ||
1854 | ||
fc01f7e7 FB |
1855 | void bdrv_close(BlockDriverState *bs) |
1856 | { | |
33384421 HR |
1857 | BdrvAioNotifier *ban, *ban_next; |
1858 | ||
3cbc002c PB |
1859 | if (bs->job) { |
1860 | block_job_cancel_sync(bs->job); | |
1861 | } | |
58fda173 SH |
1862 | bdrv_drain_all(); /* complete I/O */ |
1863 | bdrv_flush(bs); | |
1864 | bdrv_drain_all(); /* in case flush left pending I/O */ | |
d7d512f6 | 1865 | notifier_list_notify(&bs->close_notifiers, bs); |
7094f12f | 1866 | |
3cbc002c | 1867 | if (bs->drv) { |
557df6ac | 1868 | if (bs->backing_hd) { |
826b6ca0 FZ |
1869 | BlockDriverState *backing_hd = bs->backing_hd; |
1870 | bdrv_set_backing_hd(bs, NULL); | |
1871 | bdrv_unref(backing_hd); | |
557df6ac | 1872 | } |
ea2384d3 | 1873 | bs->drv->bdrv_close(bs); |
7267c094 | 1874 | g_free(bs->opaque); |
ea2384d3 FB |
1875 | bs->opaque = NULL; |
1876 | bs->drv = NULL; | |
53fec9d3 | 1877 | bs->copy_on_read = 0; |
a275fa42 PB |
1878 | bs->backing_file[0] = '\0'; |
1879 | bs->backing_format[0] = '\0'; | |
6405875c PB |
1880 | bs->total_sectors = 0; |
1881 | bs->encrypted = 0; | |
1882 | bs->valid_key = 0; | |
1883 | bs->sg = 0; | |
1884 | bs->growable = 0; | |
0d51b4de | 1885 | bs->zero_beyond_eof = false; |
de9c0cec KW |
1886 | QDECREF(bs->options); |
1887 | bs->options = NULL; | |
91af7014 HR |
1888 | QDECREF(bs->full_open_options); |
1889 | bs->full_open_options = NULL; | |
b338082b | 1890 | |
66f82cee | 1891 | if (bs->file != NULL) { |
4f6fd349 | 1892 | bdrv_unref(bs->file); |
0ac9377d | 1893 | bs->file = NULL; |
66f82cee | 1894 | } |
b338082b | 1895 | } |
98f90dba | 1896 | |
a7f53e26 MA |
1897 | if (bs->blk) { |
1898 | blk_dev_change_media_cb(bs->blk, false); | |
1899 | } | |
9ca11154 | 1900 | |
98f90dba ZYW |
1901 | /*throttling disk I/O limits*/ |
1902 | if (bs->io_limits_enabled) { | |
1903 | bdrv_io_limits_disable(bs); | |
1904 | } | |
33384421 HR |
1905 | |
1906 | QLIST_FOREACH_SAFE(ban, &bs->aio_notifiers, list, ban_next) { | |
1907 | g_free(ban); | |
1908 | } | |
1909 | QLIST_INIT(&bs->aio_notifiers); | |
b338082b FB |
1910 | } |
1911 | ||
2bc93fed MK |
1912 | void bdrv_close_all(void) |
1913 | { | |
1914 | BlockDriverState *bs; | |
1915 | ||
dc364f4c | 1916 | QTAILQ_FOREACH(bs, &bdrv_states, device_list) { |
ed78cda3 SH |
1917 | AioContext *aio_context = bdrv_get_aio_context(bs); |
1918 | ||
1919 | aio_context_acquire(aio_context); | |
2bc93fed | 1920 | bdrv_close(bs); |
ed78cda3 | 1921 | aio_context_release(aio_context); |
2bc93fed MK |
1922 | } |
1923 | } | |
1924 | ||
88266f5a SH |
1925 | /* Check if any requests are in-flight (including throttled requests) */ |
1926 | static bool bdrv_requests_pending(BlockDriverState *bs) | |
1927 | { | |
1928 | if (!QLIST_EMPTY(&bs->tracked_requests)) { | |
1929 | return true; | |
1930 | } | |
cc0681c4 BC |
1931 | if (!qemu_co_queue_empty(&bs->throttled_reqs[0])) { |
1932 | return true; | |
1933 | } | |
1934 | if (!qemu_co_queue_empty(&bs->throttled_reqs[1])) { | |
88266f5a SH |
1935 | return true; |
1936 | } | |
1937 | if (bs->file && bdrv_requests_pending(bs->file)) { | |
1938 | return true; | |
1939 | } | |
1940 | if (bs->backing_hd && bdrv_requests_pending(bs->backing_hd)) { | |
1941 | return true; | |
1942 | } | |
1943 | return false; | |
1944 | } | |
1945 | ||
5b98db0a SH |
1946 | static bool bdrv_drain_one(BlockDriverState *bs) |
1947 | { | |
1948 | bool bs_busy; | |
1949 | ||
1950 | bdrv_flush_io_queue(bs); | |
1951 | bdrv_start_throttled_reqs(bs); | |
1952 | bs_busy = bdrv_requests_pending(bs); | |
1953 | bs_busy |= aio_poll(bdrv_get_aio_context(bs), bs_busy); | |
1954 | return bs_busy; | |
1955 | } | |
1956 | ||
1957 | /* | |
1958 | * Wait for pending requests to complete on a single BlockDriverState subtree | |
1959 | * | |
1960 | * See the warning in bdrv_drain_all(). This function can only be called if | |
1961 | * you are sure nothing can generate I/O because you have op blockers | |
1962 | * installed. | |
1963 | * | |
1964 | * Note that unlike bdrv_drain_all(), the caller must hold the BlockDriverState | |
1965 | * AioContext. | |
1966 | */ | |
1967 | void bdrv_drain(BlockDriverState *bs) | |
1968 | { | |
1969 | while (bdrv_drain_one(bs)) { | |
1970 | /* Keep iterating */ | |
1971 | } | |
1972 | } | |
1973 | ||
922453bc SH |
1974 | /* |
1975 | * Wait for pending requests to complete across all BlockDriverStates | |
1976 | * | |
1977 | * This function does not flush data to disk, use bdrv_flush_all() for that | |
1978 | * after calling this function. | |
4c355d53 ZYW |
1979 | * |
1980 | * Note that completion of an asynchronous I/O operation can trigger any | |
1981 | * number of other I/O operations on other devices---for example a coroutine | |
1982 | * can be arbitrarily complex and a constant flow of I/O can come until the | |
1983 | * coroutine is complete. Because of this, it is not possible to have a | |
1984 | * function to drain a single device's I/O queue. | |
922453bc SH |
1985 | */ |
1986 | void bdrv_drain_all(void) | |
1987 | { | |
88266f5a SH |
1988 | /* Always run first iteration so any pending completion BHs run */ |
1989 | bool busy = true; | |
922453bc SH |
1990 | BlockDriverState *bs; |
1991 | ||
88266f5a | 1992 | while (busy) { |
9b536adc SH |
1993 | busy = false; |
1994 | ||
dc364f4c | 1995 | QTAILQ_FOREACH(bs, &bdrv_states, device_list) { |
9b536adc | 1996 | AioContext *aio_context = bdrv_get_aio_context(bs); |
9b536adc SH |
1997 | |
1998 | aio_context_acquire(aio_context); | |
5b98db0a | 1999 | busy |= bdrv_drain_one(bs); |
9b536adc | 2000 | aio_context_release(aio_context); |
9b536adc | 2001 | } |
922453bc SH |
2002 | } |
2003 | } | |
2004 | ||
dc364f4c BC |
2005 | /* make a BlockDriverState anonymous by removing from bdrv_state and |
2006 | * graph_bdrv_state list. | |
d22b2f41 RH |
2007 | Also, NULL terminate the device_name to prevent double remove */ |
2008 | void bdrv_make_anon(BlockDriverState *bs) | |
2009 | { | |
bfb197e0 MA |
2010 | /* |
2011 | * Take care to remove bs from bdrv_states only when it's actually | |
2012 | * in it. Note that bs->device_list.tqe_prev is initially null, | |
2013 | * and gets set to non-null by QTAILQ_INSERT_TAIL(). Establish | |
2014 | * the useful invariant "bs in bdrv_states iff bs->tqe_prev" by | |
2015 | * resetting it to null on remove. | |
2016 | */ | |
2017 | if (bs->device_list.tqe_prev) { | |
dc364f4c | 2018 | QTAILQ_REMOVE(&bdrv_states, bs, device_list); |
bfb197e0 | 2019 | bs->device_list.tqe_prev = NULL; |
d22b2f41 | 2020 | } |
dc364f4c BC |
2021 | if (bs->node_name[0] != '\0') { |
2022 | QTAILQ_REMOVE(&graph_bdrv_states, bs, node_list); | |
2023 | } | |
2024 | bs->node_name[0] = '\0'; | |
d22b2f41 RH |
2025 | } |
2026 | ||
e023b2e2 PB |
2027 | static void bdrv_rebind(BlockDriverState *bs) |
2028 | { | |
2029 | if (bs->drv && bs->drv->bdrv_rebind) { | |
2030 | bs->drv->bdrv_rebind(bs); | |
2031 | } | |
2032 | } | |
2033 | ||
4ddc07ca PB |
2034 | static void bdrv_move_feature_fields(BlockDriverState *bs_dest, |
2035 | BlockDriverState *bs_src) | |
8802d1fd | 2036 | { |
4ddc07ca | 2037 | /* move some fields that need to stay attached to the device */ |
8802d1fd JC |
2038 | |
2039 | /* dev info */ | |
1b7fd729 | 2040 | bs_dest->guest_block_size = bs_src->guest_block_size; |
4ddc07ca | 2041 | bs_dest->copy_on_read = bs_src->copy_on_read; |
8802d1fd | 2042 | |
4ddc07ca | 2043 | bs_dest->enable_write_cache = bs_src->enable_write_cache; |
c4a248a1 | 2044 | |
cc0681c4 BC |
2045 | /* i/o throttled req */ |
2046 | memcpy(&bs_dest->throttle_state, | |
2047 | &bs_src->throttle_state, | |
2048 | sizeof(ThrottleState)); | |
2049 | bs_dest->throttled_reqs[0] = bs_src->throttled_reqs[0]; | |
2050 | bs_dest->throttled_reqs[1] = bs_src->throttled_reqs[1]; | |
4ddc07ca | 2051 | bs_dest->io_limits_enabled = bs_src->io_limits_enabled; |
8802d1fd | 2052 | |
8802d1fd | 2053 | /* r/w error */ |
4ddc07ca PB |
2054 | bs_dest->on_read_error = bs_src->on_read_error; |
2055 | bs_dest->on_write_error = bs_src->on_write_error; | |
8802d1fd JC |
2056 | |
2057 | /* i/o status */ | |
4ddc07ca PB |
2058 | bs_dest->iostatus_enabled = bs_src->iostatus_enabled; |
2059 | bs_dest->iostatus = bs_src->iostatus; | |
8802d1fd | 2060 | |
a9fc4408 | 2061 | /* dirty bitmap */ |
e4654d2d | 2062 | bs_dest->dirty_bitmaps = bs_src->dirty_bitmaps; |
a9fc4408 | 2063 | |
9fcb0251 FZ |
2064 | /* reference count */ |
2065 | bs_dest->refcnt = bs_src->refcnt; | |
2066 | ||
a9fc4408 | 2067 | /* job */ |
4ddc07ca | 2068 | bs_dest->job = bs_src->job; |
a9fc4408 | 2069 | |
8802d1fd | 2070 | /* keep the same entry in bdrv_states */ |
dc364f4c | 2071 | bs_dest->device_list = bs_src->device_list; |
7e7d56d9 MA |
2072 | bs_dest->blk = bs_src->blk; |
2073 | ||
fbe40ff7 FZ |
2074 | memcpy(bs_dest->op_blockers, bs_src->op_blockers, |
2075 | sizeof(bs_dest->op_blockers)); | |
4ddc07ca | 2076 | } |
8802d1fd | 2077 | |
4ddc07ca PB |
2078 | /* |
2079 | * Swap bs contents for two image chains while they are live, | |
2080 | * while keeping required fields on the BlockDriverState that is | |
2081 | * actually attached to a device. | |
2082 | * | |
2083 | * This will modify the BlockDriverState fields, and swap contents | |
2084 | * between bs_new and bs_old. Both bs_new and bs_old are modified. | |
2085 | * | |
bfb197e0 | 2086 | * bs_new must not be attached to a BlockBackend. |
4ddc07ca PB |
2087 | * |
2088 | * This function does not create any image files. | |
2089 | */ | |
2090 | void bdrv_swap(BlockDriverState *bs_new, BlockDriverState *bs_old) | |
2091 | { | |
2092 | BlockDriverState tmp; | |
f6801b83 | 2093 | |
90ce8a06 BC |
2094 | /* The code needs to swap the node_name but simply swapping node_list won't |
2095 | * work so first remove the nodes from the graph list, do the swap then | |
2096 | * insert them back if needed. | |
2097 | */ | |
2098 | if (bs_new->node_name[0] != '\0') { | |
2099 | QTAILQ_REMOVE(&graph_bdrv_states, bs_new, node_list); | |
2100 | } | |
2101 | if (bs_old->node_name[0] != '\0') { | |
2102 | QTAILQ_REMOVE(&graph_bdrv_states, bs_old, node_list); | |
2103 | } | |
2104 | ||
bfb197e0 | 2105 | /* bs_new must be unattached and shouldn't have anything fancy enabled */ |
7e7d56d9 | 2106 | assert(!bs_new->blk); |
e4654d2d | 2107 | assert(QLIST_EMPTY(&bs_new->dirty_bitmaps)); |
4ddc07ca | 2108 | assert(bs_new->job == NULL); |
4ddc07ca | 2109 | assert(bs_new->io_limits_enabled == false); |
cc0681c4 | 2110 | assert(!throttle_have_timer(&bs_new->throttle_state)); |
8802d1fd | 2111 | |
4ddc07ca PB |
2112 | tmp = *bs_new; |
2113 | *bs_new = *bs_old; | |
2114 | *bs_old = tmp; | |
a9fc4408 | 2115 | |
4ddc07ca PB |
2116 | /* there are some fields that should not be swapped, move them back */ |
2117 | bdrv_move_feature_fields(&tmp, bs_old); | |
2118 | bdrv_move_feature_fields(bs_old, bs_new); | |
2119 | bdrv_move_feature_fields(bs_new, &tmp); | |
8802d1fd | 2120 | |
bfb197e0 | 2121 | /* bs_new must remain unattached */ |
7e7d56d9 | 2122 | assert(!bs_new->blk); |
4ddc07ca PB |
2123 | |
2124 | /* Check a few fields that should remain attached to the device */ | |
4ddc07ca | 2125 | assert(bs_new->job == NULL); |
4ddc07ca | 2126 | assert(bs_new->io_limits_enabled == false); |
cc0681c4 | 2127 | assert(!throttle_have_timer(&bs_new->throttle_state)); |
e023b2e2 | 2128 | |
90ce8a06 BC |
2129 | /* insert the nodes back into the graph node list if needed */ |
2130 | if (bs_new->node_name[0] != '\0') { | |
2131 | QTAILQ_INSERT_TAIL(&graph_bdrv_states, bs_new, node_list); | |
2132 | } | |
2133 | if (bs_old->node_name[0] != '\0') { | |
2134 | QTAILQ_INSERT_TAIL(&graph_bdrv_states, bs_old, node_list); | |
2135 | } | |
2136 | ||
e023b2e2 | 2137 | bdrv_rebind(bs_new); |
4ddc07ca PB |
2138 | bdrv_rebind(bs_old); |
2139 | } | |
2140 | ||
2141 | /* | |
2142 | * Add new bs contents at the top of an image chain while the chain is | |
2143 | * live, while keeping required fields on the top layer. | |
2144 | * | |
2145 | * This will modify the BlockDriverState fields, and swap contents | |
2146 | * between bs_new and bs_top. Both bs_new and bs_top are modified. | |
2147 | * | |
bfb197e0 | 2148 | * bs_new must not be attached to a BlockBackend. |
4ddc07ca PB |
2149 | * |
2150 | * This function does not create any image files. | |
2151 | */ | |
2152 | void bdrv_append(BlockDriverState *bs_new, BlockDriverState *bs_top) | |
2153 | { | |
2154 | bdrv_swap(bs_new, bs_top); | |
2155 | ||
2156 | /* The contents of 'tmp' will become bs_top, as we are | |
2157 | * swapping bs_new and bs_top contents. */ | |
8d24cce1 | 2158 | bdrv_set_backing_hd(bs_top, bs_new); |
8802d1fd JC |
2159 | } |
2160 | ||
4f6fd349 | 2161 | static void bdrv_delete(BlockDriverState *bs) |
b338082b | 2162 | { |
3e914655 | 2163 | assert(!bs->job); |
3718d8ab | 2164 | assert(bdrv_op_blocker_is_empty(bs)); |
4f6fd349 | 2165 | assert(!bs->refcnt); |
e4654d2d | 2166 | assert(QLIST_EMPTY(&bs->dirty_bitmaps)); |
18846dee | 2167 | |
e1b5c52e SH |
2168 | bdrv_close(bs); |
2169 | ||
1b7bdbc1 | 2170 | /* remove from list, if necessary */ |
d22b2f41 | 2171 | bdrv_make_anon(bs); |
34c6f050 | 2172 | |
7267c094 | 2173 | g_free(bs); |
fc01f7e7 FB |
2174 | } |
2175 | ||
e97fc193 AL |
2176 | /* |
2177 | * Run consistency checks on an image | |
2178 | * | |
e076f338 | 2179 | * Returns 0 if the check could be completed (it doesn't mean that the image is |
a1c7273b | 2180 | * free of errors) or -errno when an internal error occurred. The results of the |
e076f338 | 2181 | * check are stored in res. |
e97fc193 | 2182 | */ |
4534ff54 | 2183 | int bdrv_check(BlockDriverState *bs, BdrvCheckResult *res, BdrvCheckMode fix) |
e97fc193 | 2184 | { |
908bcd54 HR |
2185 | if (bs->drv == NULL) { |
2186 | return -ENOMEDIUM; | |
2187 | } | |
e97fc193 AL |
2188 | if (bs->drv->bdrv_check == NULL) { |
2189 | return -ENOTSUP; | |
2190 | } | |
2191 | ||
e076f338 | 2192 | memset(res, 0, sizeof(*res)); |
4534ff54 | 2193 | return bs->drv->bdrv_check(bs, res, fix); |
e97fc193 AL |
2194 | } |
2195 | ||
8a426614 KW |
2196 | #define COMMIT_BUF_SECTORS 2048 |
2197 | ||
33e3963e FB |
2198 | /* commit COW file into the raw image */ |
2199 | int bdrv_commit(BlockDriverState *bs) | |
2200 | { | |
19cb3738 | 2201 | BlockDriver *drv = bs->drv; |
72706ea4 | 2202 | int64_t sector, total_sectors, length, backing_length; |
8a426614 | 2203 | int n, ro, open_flags; |
0bce597d | 2204 | int ret = 0; |
72706ea4 | 2205 | uint8_t *buf = NULL; |
c2cba3d9 | 2206 | char filename[PATH_MAX]; |
33e3963e | 2207 | |
19cb3738 FB |
2208 | if (!drv) |
2209 | return -ENOMEDIUM; | |
6bb45158 | 2210 | |
4dca4b63 NS |
2211 | if (!bs->backing_hd) { |
2212 | return -ENOTSUP; | |
33e3963e FB |
2213 | } |
2214 | ||
3718d8ab FZ |
2215 | if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_COMMIT, NULL) || |
2216 | bdrv_op_is_blocked(bs->backing_hd, BLOCK_OP_TYPE_COMMIT, NULL)) { | |
2d3735d3 SH |
2217 | return -EBUSY; |
2218 | } | |
2219 | ||
4dca4b63 | 2220 | ro = bs->backing_hd->read_only; |
c2cba3d9 JM |
2221 | /* Use pstrcpy (not strncpy): filename must be NUL-terminated. */ |
2222 | pstrcpy(filename, sizeof(filename), bs->backing_hd->filename); | |
4dca4b63 NS |
2223 | open_flags = bs->backing_hd->open_flags; |
2224 | ||
2225 | if (ro) { | |
0bce597d JC |
2226 | if (bdrv_reopen(bs->backing_hd, open_flags | BDRV_O_RDWR, NULL)) { |
2227 | return -EACCES; | |
4dca4b63 | 2228 | } |
ea2384d3 | 2229 | } |
33e3963e | 2230 | |
72706ea4 JC |
2231 | length = bdrv_getlength(bs); |
2232 | if (length < 0) { | |
2233 | ret = length; | |
2234 | goto ro_cleanup; | |
2235 | } | |
2236 | ||
2237 | backing_length = bdrv_getlength(bs->backing_hd); | |
2238 | if (backing_length < 0) { | |
2239 | ret = backing_length; | |
2240 | goto ro_cleanup; | |
2241 | } | |
2242 | ||
2243 | /* If our top snapshot is larger than the backing file image, | |
2244 | * grow the backing file image if possible. If not possible, | |
2245 | * we must return an error */ | |
2246 | if (length > backing_length) { | |
2247 | ret = bdrv_truncate(bs->backing_hd, length); | |
2248 | if (ret < 0) { | |
2249 | goto ro_cleanup; | |
2250 | } | |
2251 | } | |
2252 | ||
2253 | total_sectors = length >> BDRV_SECTOR_BITS; | |
857d4f46 KW |
2254 | |
2255 | /* qemu_try_blockalign() for bs will choose an alignment that works for | |
2256 | * bs->backing_hd as well, so no need to compare the alignment manually. */ | |
2257 | buf = qemu_try_blockalign(bs, COMMIT_BUF_SECTORS * BDRV_SECTOR_SIZE); | |
2258 | if (buf == NULL) { | |
2259 | ret = -ENOMEM; | |
2260 | goto ro_cleanup; | |
2261 | } | |
8a426614 KW |
2262 | |
2263 | for (sector = 0; sector < total_sectors; sector += n) { | |
d663640c PB |
2264 | ret = bdrv_is_allocated(bs, sector, COMMIT_BUF_SECTORS, &n); |
2265 | if (ret < 0) { | |
2266 | goto ro_cleanup; | |
2267 | } | |
2268 | if (ret) { | |
dabfa6cc KW |
2269 | ret = bdrv_read(bs, sector, buf, n); |
2270 | if (ret < 0) { | |
8a426614 KW |
2271 | goto ro_cleanup; |
2272 | } | |
2273 | ||
dabfa6cc KW |
2274 | ret = bdrv_write(bs->backing_hd, sector, buf, n); |
2275 | if (ret < 0) { | |
8a426614 KW |
2276 | goto ro_cleanup; |
2277 | } | |
ea2384d3 | 2278 | } |
33e3963e | 2279 | } |
95389c86 | 2280 | |
1d44952f CH |
2281 | if (drv->bdrv_make_empty) { |
2282 | ret = drv->bdrv_make_empty(bs); | |
dabfa6cc KW |
2283 | if (ret < 0) { |
2284 | goto ro_cleanup; | |
2285 | } | |
1d44952f CH |
2286 | bdrv_flush(bs); |
2287 | } | |
95389c86 | 2288 | |
3f5075ae CH |
2289 | /* |
2290 | * Make sure all data we wrote to the backing device is actually | |
2291 | * stable on disk. | |
2292 | */ | |
dabfa6cc | 2293 | if (bs->backing_hd) { |
3f5075ae | 2294 | bdrv_flush(bs->backing_hd); |
dabfa6cc | 2295 | } |
4dca4b63 | 2296 | |
dabfa6cc | 2297 | ret = 0; |
4dca4b63 | 2298 | ro_cleanup: |
857d4f46 | 2299 | qemu_vfree(buf); |
4dca4b63 NS |
2300 | |
2301 | if (ro) { | |
0bce597d JC |
2302 | /* ignoring error return here */ |
2303 | bdrv_reopen(bs->backing_hd, open_flags & ~BDRV_O_RDWR, NULL); | |
4dca4b63 NS |
2304 | } |
2305 | ||
1d44952f | 2306 | return ret; |
33e3963e FB |
2307 | } |
2308 | ||
e8877497 | 2309 | int bdrv_commit_all(void) |
6ab4b5ab MA |
2310 | { |
2311 | BlockDriverState *bs; | |
2312 | ||
dc364f4c | 2313 | QTAILQ_FOREACH(bs, &bdrv_states, device_list) { |
ed78cda3 SH |
2314 | AioContext *aio_context = bdrv_get_aio_context(bs); |
2315 | ||
2316 | aio_context_acquire(aio_context); | |
272d2d8e JC |
2317 | if (bs->drv && bs->backing_hd) { |
2318 | int ret = bdrv_commit(bs); | |
2319 | if (ret < 0) { | |
ed78cda3 | 2320 | aio_context_release(aio_context); |
272d2d8e JC |
2321 | return ret; |
2322 | } | |
e8877497 | 2323 | } |
ed78cda3 | 2324 | aio_context_release(aio_context); |
6ab4b5ab | 2325 | } |
e8877497 | 2326 | return 0; |
6ab4b5ab MA |
2327 | } |
2328 | ||
dbffbdcf SH |
2329 | /** |
2330 | * Remove an active request from the tracked requests list | |
2331 | * | |
2332 | * This function should be called when a tracked request is completing. | |
2333 | */ | |
2334 | static void tracked_request_end(BdrvTrackedRequest *req) | |
2335 | { | |
2dbafdc0 KW |
2336 | if (req->serialising) { |
2337 | req->bs->serialising_in_flight--; | |
2338 | } | |
2339 | ||
dbffbdcf | 2340 | QLIST_REMOVE(req, list); |
f4658285 | 2341 | qemu_co_queue_restart_all(&req->wait_queue); |
dbffbdcf SH |
2342 | } |
2343 | ||
2344 | /** | |
2345 | * Add an active request to the tracked requests list | |
2346 | */ | |
2347 | static void tracked_request_begin(BdrvTrackedRequest *req, | |
2348 | BlockDriverState *bs, | |
793ed47a KW |
2349 | int64_t offset, |
2350 | unsigned int bytes, bool is_write) | |
dbffbdcf SH |
2351 | { |
2352 | *req = (BdrvTrackedRequest){ | |
2353 | .bs = bs, | |
2dbafdc0 KW |
2354 | .offset = offset, |
2355 | .bytes = bytes, | |
2356 | .is_write = is_write, | |
2357 | .co = qemu_coroutine_self(), | |
2358 | .serialising = false, | |
7327145f KW |
2359 | .overlap_offset = offset, |
2360 | .overlap_bytes = bytes, | |
dbffbdcf SH |
2361 | }; |
2362 | ||
f4658285 SH |
2363 | qemu_co_queue_init(&req->wait_queue); |
2364 | ||
dbffbdcf SH |
2365 | QLIST_INSERT_HEAD(&bs->tracked_requests, req, list); |
2366 | } | |
2367 | ||
e96126ff | 2368 | static void mark_request_serialising(BdrvTrackedRequest *req, uint64_t align) |
2dbafdc0 | 2369 | { |
7327145f | 2370 | int64_t overlap_offset = req->offset & ~(align - 1); |
e96126ff KW |
2371 | unsigned int overlap_bytes = ROUND_UP(req->offset + req->bytes, align) |
2372 | - overlap_offset; | |
7327145f | 2373 | |
2dbafdc0 KW |
2374 | if (!req->serialising) { |
2375 | req->bs->serialising_in_flight++; | |
2376 | req->serialising = true; | |
2377 | } | |
7327145f KW |
2378 | |
2379 | req->overlap_offset = MIN(req->overlap_offset, overlap_offset); | |
2380 | req->overlap_bytes = MAX(req->overlap_bytes, overlap_bytes); | |
2dbafdc0 KW |
2381 | } |
2382 | ||
d83947ac SH |
2383 | /** |
2384 | * Round a region to cluster boundaries | |
2385 | */ | |
343bded4 PB |
2386 | void bdrv_round_to_clusters(BlockDriverState *bs, |
2387 | int64_t sector_num, int nb_sectors, | |
2388 | int64_t *cluster_sector_num, | |
2389 | int *cluster_nb_sectors) | |
d83947ac SH |
2390 | { |
2391 | BlockDriverInfo bdi; | |
2392 | ||
2393 | if (bdrv_get_info(bs, &bdi) < 0 || bdi.cluster_size == 0) { | |
2394 | *cluster_sector_num = sector_num; | |
2395 | *cluster_nb_sectors = nb_sectors; | |
2396 | } else { | |
2397 | int64_t c = bdi.cluster_size / BDRV_SECTOR_SIZE; | |
2398 | *cluster_sector_num = QEMU_ALIGN_DOWN(sector_num, c); | |
2399 | *cluster_nb_sectors = QEMU_ALIGN_UP(sector_num - *cluster_sector_num + | |
2400 | nb_sectors, c); | |
2401 | } | |
2402 | } | |
2403 | ||
7327145f | 2404 | static int bdrv_get_cluster_size(BlockDriverState *bs) |
793ed47a KW |
2405 | { |
2406 | BlockDriverInfo bdi; | |
7327145f | 2407 | int ret; |
793ed47a | 2408 | |
7327145f KW |
2409 | ret = bdrv_get_info(bs, &bdi); |
2410 | if (ret < 0 || bdi.cluster_size == 0) { | |
2411 | return bs->request_alignment; | |
793ed47a | 2412 | } else { |
7327145f | 2413 | return bdi.cluster_size; |
793ed47a KW |
2414 | } |
2415 | } | |
2416 | ||
f4658285 | 2417 | static bool tracked_request_overlaps(BdrvTrackedRequest *req, |
793ed47a KW |
2418 | int64_t offset, unsigned int bytes) |
2419 | { | |
d83947ac | 2420 | /* aaaa bbbb */ |
7327145f | 2421 | if (offset >= req->overlap_offset + req->overlap_bytes) { |
d83947ac SH |
2422 | return false; |
2423 | } | |
2424 | /* bbbb aaaa */ | |
7327145f | 2425 | if (req->overlap_offset >= offset + bytes) { |
d83947ac SH |
2426 | return false; |
2427 | } | |
2428 | return true; | |
f4658285 SH |
2429 | } |
2430 | ||
28de2dcd | 2431 | static bool coroutine_fn wait_serialising_requests(BdrvTrackedRequest *self) |
f4658285 | 2432 | { |
2dbafdc0 | 2433 | BlockDriverState *bs = self->bs; |
f4658285 SH |
2434 | BdrvTrackedRequest *req; |
2435 | bool retry; | |
28de2dcd | 2436 | bool waited = false; |
f4658285 | 2437 | |
2dbafdc0 | 2438 | if (!bs->serialising_in_flight) { |
28de2dcd | 2439 | return false; |
2dbafdc0 KW |
2440 | } |
2441 | ||
f4658285 SH |
2442 | do { |
2443 | retry = false; | |
2444 | QLIST_FOREACH(req, &bs->tracked_requests, list) { | |
2dbafdc0 | 2445 | if (req == self || (!req->serialising && !self->serialising)) { |
65afd211 KW |
2446 | continue; |
2447 | } | |
7327145f KW |
2448 | if (tracked_request_overlaps(req, self->overlap_offset, |
2449 | self->overlap_bytes)) | |
2450 | { | |
5f8b6491 SH |
2451 | /* Hitting this means there was a reentrant request, for |
2452 | * example, a block driver issuing nested requests. This must | |
2453 | * never happen since it means deadlock. | |
2454 | */ | |
2455 | assert(qemu_coroutine_self() != req->co); | |
2456 | ||
6460440f KW |
2457 | /* If the request is already (indirectly) waiting for us, or |
2458 | * will wait for us as soon as it wakes up, then just go on | |
2459 | * (instead of producing a deadlock in the former case). */ | |
2460 | if (!req->waiting_for) { | |
2461 | self->waiting_for = req; | |
2462 | qemu_co_queue_wait(&req->wait_queue); | |
2463 | self->waiting_for = NULL; | |
2464 | retry = true; | |
28de2dcd | 2465 | waited = true; |
6460440f KW |
2466 | break; |
2467 | } | |
f4658285 SH |
2468 | } |
2469 | } | |
2470 | } while (retry); | |
28de2dcd KW |
2471 | |
2472 | return waited; | |
f4658285 SH |
2473 | } |
2474 | ||
756e6736 KW |
2475 | /* |
2476 | * Return values: | |
2477 | * 0 - success | |
2478 | * -EINVAL - backing format specified, but no file | |
2479 | * -ENOSPC - can't update the backing file because no space is left in the | |
2480 | * image file header | |
2481 | * -ENOTSUP - format driver doesn't support changing the backing file | |
2482 | */ | |
2483 | int bdrv_change_backing_file(BlockDriverState *bs, | |
2484 | const char *backing_file, const char *backing_fmt) | |
2485 | { | |
2486 | BlockDriver *drv = bs->drv; | |
469ef350 | 2487 | int ret; |
756e6736 | 2488 | |
5f377794 PB |
2489 | /* Backing file format doesn't make sense without a backing file */ |
2490 | if (backing_fmt && !backing_file) { | |
2491 | return -EINVAL; | |
2492 | } | |
2493 | ||
756e6736 | 2494 | if (drv->bdrv_change_backing_file != NULL) { |
469ef350 | 2495 | ret = drv->bdrv_change_backing_file(bs, backing_file, backing_fmt); |
756e6736 | 2496 | } else { |
469ef350 | 2497 | ret = -ENOTSUP; |
756e6736 | 2498 | } |
469ef350 PB |
2499 | |
2500 | if (ret == 0) { | |
2501 | pstrcpy(bs->backing_file, sizeof(bs->backing_file), backing_file ?: ""); | |
2502 | pstrcpy(bs->backing_format, sizeof(bs->backing_format), backing_fmt ?: ""); | |
2503 | } | |
2504 | return ret; | |
756e6736 KW |
2505 | } |
2506 | ||
6ebdcee2 JC |
2507 | /* |
2508 | * Finds the image layer in the chain that has 'bs' as its backing file. | |
2509 | * | |
2510 | * active is the current topmost image. | |
2511 | * | |
2512 | * Returns NULL if bs is not found in active's image chain, | |
2513 | * or if active == bs. | |
4caf0fcd JC |
2514 | * |
2515 | * Returns the bottommost base image if bs == NULL. | |
6ebdcee2 JC |
2516 | */ |
2517 | BlockDriverState *bdrv_find_overlay(BlockDriverState *active, | |
2518 | BlockDriverState *bs) | |
2519 | { | |
4caf0fcd JC |
2520 | while (active && bs != active->backing_hd) { |
2521 | active = active->backing_hd; | |
6ebdcee2 JC |
2522 | } |
2523 | ||
4caf0fcd JC |
2524 | return active; |
2525 | } | |
6ebdcee2 | 2526 | |
4caf0fcd JC |
2527 | /* Given a BDS, searches for the base layer. */ |
2528 | BlockDriverState *bdrv_find_base(BlockDriverState *bs) | |
2529 | { | |
2530 | return bdrv_find_overlay(bs, NULL); | |
6ebdcee2 JC |
2531 | } |
2532 | ||
2533 | typedef struct BlkIntermediateStates { | |
2534 | BlockDriverState *bs; | |
2535 | QSIMPLEQ_ENTRY(BlkIntermediateStates) entry; | |
2536 | } BlkIntermediateStates; | |
2537 | ||
2538 | ||
2539 | /* | |
2540 | * Drops images above 'base' up to and including 'top', and sets the image | |
2541 | * above 'top' to have base as its backing file. | |
2542 | * | |
2543 | * Requires that the overlay to 'top' is opened r/w, so that the backing file | |
2544 | * information in 'bs' can be properly updated. | |
2545 | * | |
2546 | * E.g., this will convert the following chain: | |
2547 | * bottom <- base <- intermediate <- top <- active | |
2548 | * | |
2549 | * to | |
2550 | * | |
2551 | * bottom <- base <- active | |
2552 | * | |
2553 | * It is allowed for bottom==base, in which case it converts: | |
2554 | * | |
2555 | * base <- intermediate <- top <- active | |
2556 | * | |
2557 | * to | |
2558 | * | |
2559 | * base <- active | |
2560 | * | |
54e26900 JC |
2561 | * If backing_file_str is non-NULL, it will be used when modifying top's |
2562 | * overlay image metadata. | |
2563 | * | |
6ebdcee2 JC |
2564 | * Error conditions: |
2565 | * if active == top, that is considered an error | |
2566 | * | |
2567 | */ | |
2568 | int bdrv_drop_intermediate(BlockDriverState *active, BlockDriverState *top, | |
54e26900 | 2569 | BlockDriverState *base, const char *backing_file_str) |
6ebdcee2 JC |
2570 | { |
2571 | BlockDriverState *intermediate; | |
2572 | BlockDriverState *base_bs = NULL; | |
2573 | BlockDriverState *new_top_bs = NULL; | |
2574 | BlkIntermediateStates *intermediate_state, *next; | |
2575 | int ret = -EIO; | |
2576 | ||
2577 | QSIMPLEQ_HEAD(states_to_delete, BlkIntermediateStates) states_to_delete; | |
2578 | QSIMPLEQ_INIT(&states_to_delete); | |
2579 | ||
2580 | if (!top->drv || !base->drv) { | |
2581 | goto exit; | |
2582 | } | |
2583 | ||
2584 | new_top_bs = bdrv_find_overlay(active, top); | |
2585 | ||
2586 | if (new_top_bs == NULL) { | |
2587 | /* we could not find the image above 'top', this is an error */ | |
2588 | goto exit; | |
2589 | } | |
2590 | ||
2591 | /* special case of new_top_bs->backing_hd already pointing to base - nothing | |
2592 | * to do, no intermediate images */ | |
2593 | if (new_top_bs->backing_hd == base) { | |
2594 | ret = 0; | |
2595 | goto exit; | |
2596 | } | |
2597 | ||
2598 | intermediate = top; | |
2599 | ||
2600 | /* now we will go down through the list, and add each BDS we find | |
2601 | * into our deletion queue, until we hit the 'base' | |
2602 | */ | |
2603 | while (intermediate) { | |
5839e53b | 2604 | intermediate_state = g_new0(BlkIntermediateStates, 1); |
6ebdcee2 JC |
2605 | intermediate_state->bs = intermediate; |
2606 | QSIMPLEQ_INSERT_TAIL(&states_to_delete, intermediate_state, entry); | |
2607 | ||
2608 | if (intermediate->backing_hd == base) { | |
2609 | base_bs = intermediate->backing_hd; | |
2610 | break; | |
2611 | } | |
2612 | intermediate = intermediate->backing_hd; | |
2613 | } | |
2614 | if (base_bs == NULL) { | |
2615 | /* something went wrong, we did not end at the base. safely | |
2616 | * unravel everything, and exit with error */ | |
2617 | goto exit; | |
2618 | } | |
2619 | ||
2620 | /* success - we can delete the intermediate states, and link top->base */ | |
54e26900 JC |
2621 | backing_file_str = backing_file_str ? backing_file_str : base_bs->filename; |
2622 | ret = bdrv_change_backing_file(new_top_bs, backing_file_str, | |
6ebdcee2 JC |
2623 | base_bs->drv ? base_bs->drv->format_name : ""); |
2624 | if (ret) { | |
2625 | goto exit; | |
2626 | } | |
920beae1 | 2627 | bdrv_set_backing_hd(new_top_bs, base_bs); |
6ebdcee2 JC |
2628 | |
2629 | QSIMPLEQ_FOREACH_SAFE(intermediate_state, &states_to_delete, entry, next) { | |
2630 | /* so that bdrv_close() does not recursively close the chain */ | |
920beae1 | 2631 | bdrv_set_backing_hd(intermediate_state->bs, NULL); |
4f6fd349 | 2632 | bdrv_unref(intermediate_state->bs); |
6ebdcee2 JC |
2633 | } |
2634 | ret = 0; | |
2635 | ||
2636 | exit: | |
2637 | QSIMPLEQ_FOREACH_SAFE(intermediate_state, &states_to_delete, entry, next) { | |
2638 | g_free(intermediate_state); | |
2639 | } | |
2640 | return ret; | |
2641 | } | |
2642 | ||
2643 | ||
71d0770c AL |
2644 | static int bdrv_check_byte_request(BlockDriverState *bs, int64_t offset, |
2645 | size_t size) | |
2646 | { | |
2647 | int64_t len; | |
2648 | ||
1dd3a447 KW |
2649 | if (size > INT_MAX) { |
2650 | return -EIO; | |
2651 | } | |
2652 | ||
71d0770c AL |
2653 | if (!bdrv_is_inserted(bs)) |
2654 | return -ENOMEDIUM; | |
2655 | ||
2656 | if (bs->growable) | |
2657 | return 0; | |
2658 | ||
2659 | len = bdrv_getlength(bs); | |
2660 | ||
fbb7b4e0 KW |
2661 | if (offset < 0) |
2662 | return -EIO; | |
2663 | ||
2664 | if ((offset > len) || (len - offset < size)) | |
71d0770c AL |
2665 | return -EIO; |
2666 | ||
2667 | return 0; | |
2668 | } | |
2669 | ||
2670 | static int bdrv_check_request(BlockDriverState *bs, int64_t sector_num, | |
2671 | int nb_sectors) | |
2672 | { | |
54db38a4 | 2673 | if (nb_sectors < 0 || nb_sectors > INT_MAX / BDRV_SECTOR_SIZE) { |
8f4754ed KW |
2674 | return -EIO; |
2675 | } | |
2676 | ||
eb5a3165 JS |
2677 | return bdrv_check_byte_request(bs, sector_num * BDRV_SECTOR_SIZE, |
2678 | nb_sectors * BDRV_SECTOR_SIZE); | |
71d0770c AL |
2679 | } |
2680 | ||
1c9805a3 SH |
2681 | typedef struct RwCo { |
2682 | BlockDriverState *bs; | |
775aa8b6 | 2683 | int64_t offset; |
1c9805a3 SH |
2684 | QEMUIOVector *qiov; |
2685 | bool is_write; | |
2686 | int ret; | |
4105eaaa | 2687 | BdrvRequestFlags flags; |
1c9805a3 SH |
2688 | } RwCo; |
2689 | ||
2690 | static void coroutine_fn bdrv_rw_co_entry(void *opaque) | |
fc01f7e7 | 2691 | { |
1c9805a3 | 2692 | RwCo *rwco = opaque; |
ea2384d3 | 2693 | |
1c9805a3 | 2694 | if (!rwco->is_write) { |
775aa8b6 KW |
2695 | rwco->ret = bdrv_co_do_preadv(rwco->bs, rwco->offset, |
2696 | rwco->qiov->size, rwco->qiov, | |
4105eaaa | 2697 | rwco->flags); |
775aa8b6 KW |
2698 | } else { |
2699 | rwco->ret = bdrv_co_do_pwritev(rwco->bs, rwco->offset, | |
2700 | rwco->qiov->size, rwco->qiov, | |
2701 | rwco->flags); | |
1c9805a3 SH |
2702 | } |
2703 | } | |
e7a8a783 | 2704 | |
1c9805a3 | 2705 | /* |
8d3b1a2d | 2706 | * Process a vectored synchronous request using coroutines |
1c9805a3 | 2707 | */ |
775aa8b6 KW |
2708 | static int bdrv_prwv_co(BlockDriverState *bs, int64_t offset, |
2709 | QEMUIOVector *qiov, bool is_write, | |
2710 | BdrvRequestFlags flags) | |
1c9805a3 | 2711 | { |
1c9805a3 SH |
2712 | Coroutine *co; |
2713 | RwCo rwco = { | |
2714 | .bs = bs, | |
775aa8b6 | 2715 | .offset = offset, |
8d3b1a2d | 2716 | .qiov = qiov, |
1c9805a3 SH |
2717 | .is_write = is_write, |
2718 | .ret = NOT_DONE, | |
4105eaaa | 2719 | .flags = flags, |
1c9805a3 | 2720 | }; |
e7a8a783 | 2721 | |
498e386c ZYW |
2722 | /** |
2723 | * In sync call context, when the vcpu is blocked, this throttling timer | |
2724 | * will not fire; so the I/O throttling function has to be disabled here | |
2725 | * if it has been enabled. | |
2726 | */ | |
2727 | if (bs->io_limits_enabled) { | |
2728 | fprintf(stderr, "Disabling I/O throttling on '%s' due " | |
2729 | "to synchronous I/O.\n", bdrv_get_device_name(bs)); | |
2730 | bdrv_io_limits_disable(bs); | |
2731 | } | |
2732 | ||
1c9805a3 SH |
2733 | if (qemu_in_coroutine()) { |
2734 | /* Fast-path if already in coroutine context */ | |
2735 | bdrv_rw_co_entry(&rwco); | |
2736 | } else { | |
2572b37a SH |
2737 | AioContext *aio_context = bdrv_get_aio_context(bs); |
2738 | ||
1c9805a3 SH |
2739 | co = qemu_coroutine_create(bdrv_rw_co_entry); |
2740 | qemu_coroutine_enter(co, &rwco); | |
2741 | while (rwco.ret == NOT_DONE) { | |
2572b37a | 2742 | aio_poll(aio_context, true); |
1c9805a3 SH |
2743 | } |
2744 | } | |
2745 | return rwco.ret; | |
2746 | } | |
b338082b | 2747 | |
8d3b1a2d KW |
2748 | /* |
2749 | * Process a synchronous request using coroutines | |
2750 | */ | |
2751 | static int bdrv_rw_co(BlockDriverState *bs, int64_t sector_num, uint8_t *buf, | |
4105eaaa | 2752 | int nb_sectors, bool is_write, BdrvRequestFlags flags) |
8d3b1a2d KW |
2753 | { |
2754 | QEMUIOVector qiov; | |
2755 | struct iovec iov = { | |
2756 | .iov_base = (void *)buf, | |
2757 | .iov_len = nb_sectors * BDRV_SECTOR_SIZE, | |
2758 | }; | |
2759 | ||
da15ee51 KW |
2760 | if (nb_sectors < 0 || nb_sectors > INT_MAX / BDRV_SECTOR_SIZE) { |
2761 | return -EINVAL; | |
2762 | } | |
2763 | ||
8d3b1a2d | 2764 | qemu_iovec_init_external(&qiov, &iov, 1); |
775aa8b6 KW |
2765 | return bdrv_prwv_co(bs, sector_num << BDRV_SECTOR_BITS, |
2766 | &qiov, is_write, flags); | |
8d3b1a2d KW |
2767 | } |
2768 | ||
1c9805a3 SH |
2769 | /* return < 0 if error. See bdrv_write() for the return codes */ |
2770 | int bdrv_read(BlockDriverState *bs, int64_t sector_num, | |
2771 | uint8_t *buf, int nb_sectors) | |
2772 | { | |
4105eaaa | 2773 | return bdrv_rw_co(bs, sector_num, buf, nb_sectors, false, 0); |
fc01f7e7 FB |
2774 | } |
2775 | ||
07d27a44 MA |
2776 | /* Just like bdrv_read(), but with I/O throttling temporarily disabled */ |
2777 | int bdrv_read_unthrottled(BlockDriverState *bs, int64_t sector_num, | |
2778 | uint8_t *buf, int nb_sectors) | |
2779 | { | |
2780 | bool enabled; | |
2781 | int ret; | |
2782 | ||
2783 | enabled = bs->io_limits_enabled; | |
2784 | bs->io_limits_enabled = false; | |
4e7395e8 | 2785 | ret = bdrv_read(bs, sector_num, buf, nb_sectors); |
07d27a44 MA |
2786 | bs->io_limits_enabled = enabled; |
2787 | return ret; | |
2788 | } | |
2789 | ||
5fafdf24 | 2790 | /* Return < 0 if error. Important errors are: |
19cb3738 FB |
2791 | -EIO generic I/O error (may happen for all errors) |
2792 | -ENOMEDIUM No media inserted. | |
2793 | -EINVAL Invalid sector number or nb_sectors | |
2794 | -EACCES Trying to write a read-only device | |
2795 | */ | |
5fafdf24 | 2796 | int bdrv_write(BlockDriverState *bs, int64_t sector_num, |
fc01f7e7 FB |
2797 | const uint8_t *buf, int nb_sectors) |
2798 | { | |
4105eaaa | 2799 | return bdrv_rw_co(bs, sector_num, (uint8_t *)buf, nb_sectors, true, 0); |
83f64091 FB |
2800 | } |
2801 | ||
aa7bfbff PL |
2802 | int bdrv_write_zeroes(BlockDriverState *bs, int64_t sector_num, |
2803 | int nb_sectors, BdrvRequestFlags flags) | |
4105eaaa PL |
2804 | { |
2805 | return bdrv_rw_co(bs, sector_num, NULL, nb_sectors, true, | |
aa7bfbff | 2806 | BDRV_REQ_ZERO_WRITE | flags); |
8d3b1a2d KW |
2807 | } |
2808 | ||
d75cbb5e PL |
2809 | /* |
2810 | * Completely zero out a block device with the help of bdrv_write_zeroes. | |
2811 | * The operation is sped up by checking the block status and only writing | |
2812 | * zeroes to the device if they currently do not return zeroes. Optional | |
2813 | * flags are passed through to bdrv_write_zeroes (e.g. BDRV_REQ_MAY_UNMAP). | |
2814 | * | |
2815 | * Returns < 0 on error, 0 on success. For error codes see bdrv_write(). | |
2816 | */ | |
2817 | int bdrv_make_zero(BlockDriverState *bs, BdrvRequestFlags flags) | |
2818 | { | |
d32f7c10 | 2819 | int64_t target_sectors, ret, nb_sectors, sector_num = 0; |
d75cbb5e PL |
2820 | int n; |
2821 | ||
d32f7c10 MA |
2822 | target_sectors = bdrv_nb_sectors(bs); |
2823 | if (target_sectors < 0) { | |
2824 | return target_sectors; | |
9ce10c0b | 2825 | } |
9ce10c0b | 2826 | |
d75cbb5e | 2827 | for (;;) { |
d32f7c10 | 2828 | nb_sectors = target_sectors - sector_num; |
d75cbb5e PL |
2829 | if (nb_sectors <= 0) { |
2830 | return 0; | |
2831 | } | |
f3a9cfdd FZ |
2832 | if (nb_sectors > INT_MAX / BDRV_SECTOR_SIZE) { |
2833 | nb_sectors = INT_MAX / BDRV_SECTOR_SIZE; | |
d75cbb5e PL |
2834 | } |
2835 | ret = bdrv_get_block_status(bs, sector_num, nb_sectors, &n); | |
3d94ce60 PL |
2836 | if (ret < 0) { |
2837 | error_report("error getting block status at sector %" PRId64 ": %s", | |
2838 | sector_num, strerror(-ret)); | |
2839 | return ret; | |
2840 | } | |
d75cbb5e PL |
2841 | if (ret & BDRV_BLOCK_ZERO) { |
2842 | sector_num += n; | |
2843 | continue; | |
2844 | } | |
2845 | ret = bdrv_write_zeroes(bs, sector_num, n, flags); | |
2846 | if (ret < 0) { | |
2847 | error_report("error writing zeroes at sector %" PRId64 ": %s", | |
2848 | sector_num, strerror(-ret)); | |
2849 | return ret; | |
2850 | } | |
2851 | sector_num += n; | |
2852 | } | |
2853 | } | |
2854 | ||
a3ef6571 | 2855 | int bdrv_pread(BlockDriverState *bs, int64_t offset, void *buf, int bytes) |
83f64091 | 2856 | { |
a3ef6571 KW |
2857 | QEMUIOVector qiov; |
2858 | struct iovec iov = { | |
2859 | .iov_base = (void *)buf, | |
2860 | .iov_len = bytes, | |
2861 | }; | |
9a8c4cce | 2862 | int ret; |
83f64091 | 2863 | |
a3ef6571 KW |
2864 | if (bytes < 0) { |
2865 | return -EINVAL; | |
83f64091 FB |
2866 | } |
2867 | ||
a3ef6571 KW |
2868 | qemu_iovec_init_external(&qiov, &iov, 1); |
2869 | ret = bdrv_prwv_co(bs, offset, &qiov, false, 0); | |
2870 | if (ret < 0) { | |
2871 | return ret; | |
83f64091 | 2872 | } |
a3ef6571 KW |
2873 | |
2874 | return bytes; | |
83f64091 FB |
2875 | } |
2876 | ||
8d3b1a2d | 2877 | int bdrv_pwritev(BlockDriverState *bs, int64_t offset, QEMUIOVector *qiov) |
83f64091 | 2878 | { |
9a8c4cce | 2879 | int ret; |
83f64091 | 2880 | |
8407d5d7 KW |
2881 | ret = bdrv_prwv_co(bs, offset, qiov, true, 0); |
2882 | if (ret < 0) { | |
2883 | return ret; | |
83f64091 FB |
2884 | } |
2885 | ||
8d3b1a2d KW |
2886 | return qiov->size; |
2887 | } | |
2888 | ||
2889 | int bdrv_pwrite(BlockDriverState *bs, int64_t offset, | |
8407d5d7 | 2890 | const void *buf, int bytes) |
8d3b1a2d KW |
2891 | { |
2892 | QEMUIOVector qiov; | |
2893 | struct iovec iov = { | |
2894 | .iov_base = (void *) buf, | |
8407d5d7 | 2895 | .iov_len = bytes, |
8d3b1a2d KW |
2896 | }; |
2897 | ||
8407d5d7 KW |
2898 | if (bytes < 0) { |
2899 | return -EINVAL; | |
2900 | } | |
2901 | ||
8d3b1a2d KW |
2902 | qemu_iovec_init_external(&qiov, &iov, 1); |
2903 | return bdrv_pwritev(bs, offset, &qiov); | |
83f64091 | 2904 | } |
83f64091 | 2905 | |
f08145fe KW |
2906 | /* |
2907 | * Writes to the file and ensures that no writes are reordered across this | |
2908 | * request (acts as a barrier) | |
2909 | * | |
2910 | * Returns 0 on success, -errno in error cases. | |
2911 | */ | |
2912 | int bdrv_pwrite_sync(BlockDriverState *bs, int64_t offset, | |
2913 | const void *buf, int count) | |
2914 | { | |
2915 | int ret; | |
2916 | ||
2917 | ret = bdrv_pwrite(bs, offset, buf, count); | |
2918 | if (ret < 0) { | |
2919 | return ret; | |
2920 | } | |
2921 | ||
f05fa4ad PB |
2922 | /* No flush needed for cache modes that already do it */ |
2923 | if (bs->enable_write_cache) { | |
f08145fe KW |
2924 | bdrv_flush(bs); |
2925 | } | |
2926 | ||
2927 | return 0; | |
2928 | } | |
2929 | ||
470c0504 | 2930 | static int coroutine_fn bdrv_co_do_copy_on_readv(BlockDriverState *bs, |
ab185921 SH |
2931 | int64_t sector_num, int nb_sectors, QEMUIOVector *qiov) |
2932 | { | |
2933 | /* Perform I/O through a temporary buffer so that users who scribble over | |
2934 | * their read buffer while the operation is in progress do not end up | |
2935 | * modifying the image file. This is critical for zero-copy guest I/O | |
2936 | * where anything might happen inside guest memory. | |
2937 | */ | |
2938 | void *bounce_buffer; | |
2939 | ||
79c053bd | 2940 | BlockDriver *drv = bs->drv; |
ab185921 SH |
2941 | struct iovec iov; |
2942 | QEMUIOVector bounce_qiov; | |
2943 | int64_t cluster_sector_num; | |
2944 | int cluster_nb_sectors; | |
2945 | size_t skip_bytes; | |
2946 | int ret; | |
2947 | ||
2948 | /* Cover entire cluster so no additional backing file I/O is required when | |
2949 | * allocating cluster in the image file. | |
2950 | */ | |
343bded4 PB |
2951 | bdrv_round_to_clusters(bs, sector_num, nb_sectors, |
2952 | &cluster_sector_num, &cluster_nb_sectors); | |
ab185921 | 2953 | |
470c0504 SH |
2954 | trace_bdrv_co_do_copy_on_readv(bs, sector_num, nb_sectors, |
2955 | cluster_sector_num, cluster_nb_sectors); | |
ab185921 SH |
2956 | |
2957 | iov.iov_len = cluster_nb_sectors * BDRV_SECTOR_SIZE; | |
857d4f46 KW |
2958 | iov.iov_base = bounce_buffer = qemu_try_blockalign(bs, iov.iov_len); |
2959 | if (bounce_buffer == NULL) { | |
2960 | ret = -ENOMEM; | |
2961 | goto err; | |
2962 | } | |
2963 | ||
ab185921 SH |
2964 | qemu_iovec_init_external(&bounce_qiov, &iov, 1); |
2965 | ||
79c053bd SH |
2966 | ret = drv->bdrv_co_readv(bs, cluster_sector_num, cluster_nb_sectors, |
2967 | &bounce_qiov); | |
ab185921 SH |
2968 | if (ret < 0) { |
2969 | goto err; | |
2970 | } | |
2971 | ||
79c053bd SH |
2972 | if (drv->bdrv_co_write_zeroes && |
2973 | buffer_is_zero(bounce_buffer, iov.iov_len)) { | |
621f0589 | 2974 | ret = bdrv_co_do_write_zeroes(bs, cluster_sector_num, |
aa7bfbff | 2975 | cluster_nb_sectors, 0); |
79c053bd | 2976 | } else { |
f05fa4ad PB |
2977 | /* This does not change the data on the disk, it is not necessary |
2978 | * to flush even in cache=writethrough mode. | |
2979 | */ | |
79c053bd | 2980 | ret = drv->bdrv_co_writev(bs, cluster_sector_num, cluster_nb_sectors, |
ab185921 | 2981 | &bounce_qiov); |
79c053bd SH |
2982 | } |
2983 | ||
ab185921 SH |
2984 | if (ret < 0) { |
2985 | /* It might be okay to ignore write errors for guest requests. If this | |
2986 | * is a deliberate copy-on-read then we don't want to ignore the error. | |
2987 | * Simply report it in all cases. | |
2988 | */ | |
2989 | goto err; | |
2990 | } | |
2991 | ||
2992 | skip_bytes = (sector_num - cluster_sector_num) * BDRV_SECTOR_SIZE; | |
03396148 MT |
2993 | qemu_iovec_from_buf(qiov, 0, bounce_buffer + skip_bytes, |
2994 | nb_sectors * BDRV_SECTOR_SIZE); | |
ab185921 SH |
2995 | |
2996 | err: | |
2997 | qemu_vfree(bounce_buffer); | |
2998 | return ret; | |
2999 | } | |
3000 | ||
c5fbe571 | 3001 | /* |
d0c7f642 KW |
3002 | * Forwards an already correctly aligned request to the BlockDriver. This |
3003 | * handles copy on read and zeroing after EOF; any other features must be | |
3004 | * implemented by the caller. | |
c5fbe571 | 3005 | */ |
d0c7f642 | 3006 | static int coroutine_fn bdrv_aligned_preadv(BlockDriverState *bs, |
65afd211 | 3007 | BdrvTrackedRequest *req, int64_t offset, unsigned int bytes, |
ec746e10 | 3008 | int64_t align, QEMUIOVector *qiov, int flags) |
da1fa91d KW |
3009 | { |
3010 | BlockDriver *drv = bs->drv; | |
dbffbdcf | 3011 | int ret; |
da1fa91d | 3012 | |
d0c7f642 KW |
3013 | int64_t sector_num = offset >> BDRV_SECTOR_BITS; |
3014 | unsigned int nb_sectors = bytes >> BDRV_SECTOR_BITS; | |
da1fa91d | 3015 | |
d0c7f642 KW |
3016 | assert((offset & (BDRV_SECTOR_SIZE - 1)) == 0); |
3017 | assert((bytes & (BDRV_SECTOR_SIZE - 1)) == 0); | |
8eb029c2 | 3018 | assert(!qiov || bytes == qiov->size); |
d0c7f642 KW |
3019 | |
3020 | /* Handle Copy on Read and associated serialisation */ | |
470c0504 | 3021 | if (flags & BDRV_REQ_COPY_ON_READ) { |
7327145f KW |
3022 | /* If we touch the same cluster it counts as an overlap. This |
3023 | * guarantees that allocating writes will be serialized and not race | |
3024 | * with each other for the same cluster. For example, in copy-on-read | |
3025 | * it ensures that the CoR read and write operations are atomic and | |
3026 | * guest writes cannot interleave between them. */ | |
3027 | mark_request_serialising(req, bdrv_get_cluster_size(bs)); | |
470c0504 SH |
3028 | } |
3029 | ||
2dbafdc0 | 3030 | wait_serialising_requests(req); |
f4658285 | 3031 | |
470c0504 | 3032 | if (flags & BDRV_REQ_COPY_ON_READ) { |
ab185921 SH |
3033 | int pnum; |
3034 | ||
bdad13b9 | 3035 | ret = bdrv_is_allocated(bs, sector_num, nb_sectors, &pnum); |
ab185921 SH |
3036 | if (ret < 0) { |
3037 | goto out; | |
3038 | } | |
3039 | ||
3040 | if (!ret || pnum != nb_sectors) { | |
470c0504 | 3041 | ret = bdrv_co_do_copy_on_readv(bs, sector_num, nb_sectors, qiov); |
ab185921 SH |
3042 | goto out; |
3043 | } | |
3044 | } | |
3045 | ||
d0c7f642 | 3046 | /* Forward the request to the BlockDriver */ |
893a8f62 MK |
3047 | if (!(bs->zero_beyond_eof && bs->growable)) { |
3048 | ret = drv->bdrv_co_readv(bs, sector_num, nb_sectors, qiov); | |
3049 | } else { | |
3050 | /* Read zeros after EOF of growable BDSes */ | |
4049082c | 3051 | int64_t total_sectors, max_nb_sectors; |
893a8f62 | 3052 | |
4049082c MA |
3053 | total_sectors = bdrv_nb_sectors(bs); |
3054 | if (total_sectors < 0) { | |
3055 | ret = total_sectors; | |
893a8f62 MK |
3056 | goto out; |
3057 | } | |
3058 | ||
5f5bcd80 KW |
3059 | max_nb_sectors = ROUND_UP(MAX(0, total_sectors - sector_num), |
3060 | align >> BDRV_SECTOR_BITS); | |
e012b78c PB |
3061 | if (nb_sectors < max_nb_sectors) { |
3062 | ret = drv->bdrv_co_readv(bs, sector_num, nb_sectors, qiov); | |
3063 | } else if (max_nb_sectors > 0) { | |
33f461e0 | 3064 | QEMUIOVector local_qiov; |
33f461e0 KW |
3065 | |
3066 | qemu_iovec_init(&local_qiov, qiov->niov); | |
3067 | qemu_iovec_concat(&local_qiov, qiov, 0, | |
e012b78c | 3068 | max_nb_sectors * BDRV_SECTOR_SIZE); |
33f461e0 | 3069 | |
e012b78c | 3070 | ret = drv->bdrv_co_readv(bs, sector_num, max_nb_sectors, |
33f461e0 KW |
3071 | &local_qiov); |
3072 | ||
3073 | qemu_iovec_destroy(&local_qiov); | |
893a8f62 MK |
3074 | } else { |
3075 | ret = 0; | |
3076 | } | |
3077 | ||
3078 | /* Reading beyond end of file is supposed to produce zeroes */ | |
3079 | if (ret == 0 && total_sectors < sector_num + nb_sectors) { | |
3080 | uint64_t offset = MAX(0, total_sectors - sector_num); | |
3081 | uint64_t bytes = (sector_num + nb_sectors - offset) * | |
3082 | BDRV_SECTOR_SIZE; | |
3083 | qemu_iovec_memset(qiov, offset * BDRV_SECTOR_SIZE, 0, bytes); | |
3084 | } | |
3085 | } | |
ab185921 SH |
3086 | |
3087 | out: | |
dbffbdcf | 3088 | return ret; |
da1fa91d KW |
3089 | } |
3090 | ||
d0c7f642 KW |
3091 | /* |
3092 | * Handle a read request in coroutine context | |
3093 | */ | |
1b0288ae KW |
3094 | static int coroutine_fn bdrv_co_do_preadv(BlockDriverState *bs, |
3095 | int64_t offset, unsigned int bytes, QEMUIOVector *qiov, | |
d0c7f642 KW |
3096 | BdrvRequestFlags flags) |
3097 | { | |
3098 | BlockDriver *drv = bs->drv; | |
65afd211 KW |
3099 | BdrvTrackedRequest req; |
3100 | ||
1b0288ae KW |
3101 | /* TODO Lift BDRV_SECTOR_SIZE restriction in BlockDriver interface */ |
3102 | uint64_t align = MAX(BDRV_SECTOR_SIZE, bs->request_alignment); | |
3103 | uint8_t *head_buf = NULL; | |
3104 | uint8_t *tail_buf = NULL; | |
3105 | QEMUIOVector local_qiov; | |
3106 | bool use_local_qiov = false; | |
d0c7f642 KW |
3107 | int ret; |
3108 | ||
3109 | if (!drv) { | |
3110 | return -ENOMEDIUM; | |
3111 | } | |
1b0288ae | 3112 | if (bdrv_check_byte_request(bs, offset, bytes)) { |
d0c7f642 KW |
3113 | return -EIO; |
3114 | } | |
3115 | ||
3116 | if (bs->copy_on_read) { | |
3117 | flags |= BDRV_REQ_COPY_ON_READ; | |
3118 | } | |
3119 | ||
3120 | /* throttling disk I/O */ | |
3121 | if (bs->io_limits_enabled) { | |
d5103588 | 3122 | bdrv_io_limits_intercept(bs, bytes, false); |
1b0288ae KW |
3123 | } |
3124 | ||
3125 | /* Align read if necessary by padding qiov */ | |
3126 | if (offset & (align - 1)) { | |
3127 | head_buf = qemu_blockalign(bs, align); | |
3128 | qemu_iovec_init(&local_qiov, qiov->niov + 2); | |
3129 | qemu_iovec_add(&local_qiov, head_buf, offset & (align - 1)); | |
3130 | qemu_iovec_concat(&local_qiov, qiov, 0, qiov->size); | |
3131 | use_local_qiov = true; | |
3132 | ||
3133 | bytes += offset & (align - 1); | |
3134 | offset = offset & ~(align - 1); | |
3135 | } | |
3136 | ||
3137 | if ((offset + bytes) & (align - 1)) { | |
3138 | if (!use_local_qiov) { | |
3139 | qemu_iovec_init(&local_qiov, qiov->niov + 1); | |
3140 | qemu_iovec_concat(&local_qiov, qiov, 0, qiov->size); | |
3141 | use_local_qiov = true; | |
3142 | } | |
3143 | tail_buf = qemu_blockalign(bs, align); | |
3144 | qemu_iovec_add(&local_qiov, tail_buf, | |
3145 | align - ((offset + bytes) & (align - 1))); | |
3146 | ||
3147 | bytes = ROUND_UP(bytes, align); | |
3148 | } | |
3149 | ||
65afd211 | 3150 | tracked_request_begin(&req, bs, offset, bytes, false); |
ec746e10 | 3151 | ret = bdrv_aligned_preadv(bs, &req, offset, bytes, align, |
1b0288ae KW |
3152 | use_local_qiov ? &local_qiov : qiov, |
3153 | flags); | |
65afd211 | 3154 | tracked_request_end(&req); |
1b0288ae KW |
3155 | |
3156 | if (use_local_qiov) { | |
3157 | qemu_iovec_destroy(&local_qiov); | |
3158 | qemu_vfree(head_buf); | |
3159 | qemu_vfree(tail_buf); | |
d0c7f642 KW |
3160 | } |
3161 | ||
d0c7f642 KW |
3162 | return ret; |
3163 | } | |
3164 | ||
1b0288ae KW |
3165 | static int coroutine_fn bdrv_co_do_readv(BlockDriverState *bs, |
3166 | int64_t sector_num, int nb_sectors, QEMUIOVector *qiov, | |
3167 | BdrvRequestFlags flags) | |
3168 | { | |
3169 | if (nb_sectors < 0 || nb_sectors > (UINT_MAX >> BDRV_SECTOR_BITS)) { | |
3170 | return -EINVAL; | |
3171 | } | |
3172 | ||
3173 | return bdrv_co_do_preadv(bs, sector_num << BDRV_SECTOR_BITS, | |
3174 | nb_sectors << BDRV_SECTOR_BITS, qiov, flags); | |
3175 | } | |
3176 | ||
c5fbe571 | 3177 | int coroutine_fn bdrv_co_readv(BlockDriverState *bs, int64_t sector_num, |
da1fa91d KW |
3178 | int nb_sectors, QEMUIOVector *qiov) |
3179 | { | |
c5fbe571 | 3180 | trace_bdrv_co_readv(bs, sector_num, nb_sectors); |
da1fa91d | 3181 | |
470c0504 SH |
3182 | return bdrv_co_do_readv(bs, sector_num, nb_sectors, qiov, 0); |
3183 | } | |
3184 | ||
3185 | int coroutine_fn bdrv_co_copy_on_readv(BlockDriverState *bs, | |
3186 | int64_t sector_num, int nb_sectors, QEMUIOVector *qiov) | |
3187 | { | |
3188 | trace_bdrv_co_copy_on_readv(bs, sector_num, nb_sectors); | |
3189 | ||
3190 | return bdrv_co_do_readv(bs, sector_num, nb_sectors, qiov, | |
3191 | BDRV_REQ_COPY_ON_READ); | |
c5fbe571 SH |
3192 | } |
3193 | ||
c31cb707 PL |
3194 | /* if no limit is specified in the BlockLimits use a default |
3195 | * of 32768 512-byte sectors (16 MiB) per request. | |
3196 | */ | |
3197 | #define MAX_WRITE_ZEROES_DEFAULT 32768 | |
3198 | ||
f08f2dda | 3199 | static int coroutine_fn bdrv_co_do_write_zeroes(BlockDriverState *bs, |
aa7bfbff | 3200 | int64_t sector_num, int nb_sectors, BdrvRequestFlags flags) |
f08f2dda SH |
3201 | { |
3202 | BlockDriver *drv = bs->drv; | |
3203 | QEMUIOVector qiov; | |
c31cb707 PL |
3204 | struct iovec iov = {0}; |
3205 | int ret = 0; | |
f08f2dda | 3206 | |
c31cb707 PL |
3207 | int max_write_zeroes = bs->bl.max_write_zeroes ? |
3208 | bs->bl.max_write_zeroes : MAX_WRITE_ZEROES_DEFAULT; | |
621f0589 | 3209 | |
c31cb707 PL |
3210 | while (nb_sectors > 0 && !ret) { |
3211 | int num = nb_sectors; | |
3212 | ||
b8d71c09 PB |
3213 | /* Align request. Block drivers can expect the "bulk" of the request |
3214 | * to be aligned. | |
3215 | */ | |
3216 | if (bs->bl.write_zeroes_alignment | |
3217 | && num > bs->bl.write_zeroes_alignment) { | |
3218 | if (sector_num % bs->bl.write_zeroes_alignment != 0) { | |
3219 | /* Make a small request up to the first aligned sector. */ | |
c31cb707 | 3220 | num = bs->bl.write_zeroes_alignment; |
b8d71c09 PB |
3221 | num -= sector_num % bs->bl.write_zeroes_alignment; |
3222 | } else if ((sector_num + num) % bs->bl.write_zeroes_alignment != 0) { | |
3223 | /* Shorten the request to the last aligned sector. num cannot | |
3224 | * underflow because num > bs->bl.write_zeroes_alignment. | |
3225 | */ | |
3226 | num -= (sector_num + num) % bs->bl.write_zeroes_alignment; | |
c31cb707 | 3227 | } |
621f0589 | 3228 | } |
f08f2dda | 3229 | |
c31cb707 PL |
3230 | /* limit request size */ |
3231 | if (num > max_write_zeroes) { | |
3232 | num = max_write_zeroes; | |
3233 | } | |
3234 | ||
3235 | ret = -ENOTSUP; | |
3236 | /* First try the efficient write zeroes operation */ | |
3237 | if (drv->bdrv_co_write_zeroes) { | |
3238 | ret = drv->bdrv_co_write_zeroes(bs, sector_num, num, flags); | |
3239 | } | |
3240 | ||
3241 | if (ret == -ENOTSUP) { | |
3242 | /* Fall back to bounce buffer if write zeroes is unsupported */ | |
3243 | iov.iov_len = num * BDRV_SECTOR_SIZE; | |
3244 | if (iov.iov_base == NULL) { | |
857d4f46 KW |
3245 | iov.iov_base = qemu_try_blockalign(bs, num * BDRV_SECTOR_SIZE); |
3246 | if (iov.iov_base == NULL) { | |
3247 | ret = -ENOMEM; | |
3248 | goto fail; | |
3249 | } | |
b8d71c09 | 3250 | memset(iov.iov_base, 0, num * BDRV_SECTOR_SIZE); |
c31cb707 PL |
3251 | } |
3252 | qemu_iovec_init_external(&qiov, &iov, 1); | |
f08f2dda | 3253 | |
c31cb707 | 3254 | ret = drv->bdrv_co_writev(bs, sector_num, num, &qiov); |
b8d71c09 PB |
3255 | |
3256 | /* Keep bounce buffer around if it is big enough for all | |
3257 | * all future requests. | |
3258 | */ | |
3259 | if (num < max_write_zeroes) { | |
3260 | qemu_vfree(iov.iov_base); | |
3261 | iov.iov_base = NULL; | |
3262 | } | |
c31cb707 PL |
3263 | } |
3264 | ||
3265 | sector_num += num; | |
3266 | nb_sectors -= num; | |
3267 | } | |
f08f2dda | 3268 | |
857d4f46 | 3269 | fail: |
f08f2dda SH |
3270 | qemu_vfree(iov.iov_base); |
3271 | return ret; | |
3272 | } | |
3273 | ||
c5fbe571 | 3274 | /* |
b404f720 | 3275 | * Forwards an already correctly aligned write request to the BlockDriver. |
c5fbe571 | 3276 | */ |
b404f720 | 3277 | static int coroutine_fn bdrv_aligned_pwritev(BlockDriverState *bs, |
65afd211 KW |
3278 | BdrvTrackedRequest *req, int64_t offset, unsigned int bytes, |
3279 | QEMUIOVector *qiov, int flags) | |
c5fbe571 SH |
3280 | { |
3281 | BlockDriver *drv = bs->drv; | |
28de2dcd | 3282 | bool waited; |
6b7cb247 | 3283 | int ret; |
da1fa91d | 3284 | |
b404f720 KW |
3285 | int64_t sector_num = offset >> BDRV_SECTOR_BITS; |
3286 | unsigned int nb_sectors = bytes >> BDRV_SECTOR_BITS; | |
f4658285 | 3287 | |
b404f720 KW |
3288 | assert((offset & (BDRV_SECTOR_SIZE - 1)) == 0); |
3289 | assert((bytes & (BDRV_SECTOR_SIZE - 1)) == 0); | |
8eb029c2 | 3290 | assert(!qiov || bytes == qiov->size); |
cc0681c4 | 3291 | |
28de2dcd KW |
3292 | waited = wait_serialising_requests(req); |
3293 | assert(!waited || !req->serialising); | |
af91f9a7 KW |
3294 | assert(req->overlap_offset <= offset); |
3295 | assert(offset + bytes <= req->overlap_offset + req->overlap_bytes); | |
244eadef | 3296 | |
65afd211 | 3297 | ret = notifier_with_return_list_notify(&bs->before_write_notifiers, req); |
d616b224 | 3298 | |
465bee1d PL |
3299 | if (!ret && bs->detect_zeroes != BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF && |
3300 | !(flags & BDRV_REQ_ZERO_WRITE) && drv->bdrv_co_write_zeroes && | |
3301 | qemu_iovec_is_zero(qiov)) { | |
3302 | flags |= BDRV_REQ_ZERO_WRITE; | |
3303 | if (bs->detect_zeroes == BLOCKDEV_DETECT_ZEROES_OPTIONS_UNMAP) { | |
3304 | flags |= BDRV_REQ_MAY_UNMAP; | |
3305 | } | |
3306 | } | |
3307 | ||
d616b224 SH |
3308 | if (ret < 0) { |
3309 | /* Do nothing, write notifier decided to fail this request */ | |
3310 | } else if (flags & BDRV_REQ_ZERO_WRITE) { | |
9e1cb96d | 3311 | BLKDBG_EVENT(bs, BLKDBG_PWRITEV_ZERO); |
aa7bfbff | 3312 | ret = bdrv_co_do_write_zeroes(bs, sector_num, nb_sectors, flags); |
f08f2dda | 3313 | } else { |
9e1cb96d | 3314 | BLKDBG_EVENT(bs, BLKDBG_PWRITEV); |
f08f2dda SH |
3315 | ret = drv->bdrv_co_writev(bs, sector_num, nb_sectors, qiov); |
3316 | } | |
9e1cb96d | 3317 | BLKDBG_EVENT(bs, BLKDBG_PWRITEV_DONE); |
6b7cb247 | 3318 | |
f05fa4ad PB |
3319 | if (ret == 0 && !bs->enable_write_cache) { |
3320 | ret = bdrv_co_flush(bs); | |
3321 | } | |
3322 | ||
e4654d2d | 3323 | bdrv_set_dirty(bs, sector_num, nb_sectors); |
da1fa91d | 3324 | |
5366d0c8 | 3325 | block_acct_highest_sector(&bs->stats, sector_num, nb_sectors); |
5e5a94b6 | 3326 | |
df2a6f29 PB |
3327 | if (bs->growable && ret >= 0) { |
3328 | bs->total_sectors = MAX(bs->total_sectors, sector_num + nb_sectors); | |
3329 | } | |
da1fa91d | 3330 | |
6b7cb247 | 3331 | return ret; |
da1fa91d KW |
3332 | } |
3333 | ||
b404f720 KW |
3334 | /* |
3335 | * Handle a write request in coroutine context | |
3336 | */ | |
6601553e KW |
3337 | static int coroutine_fn bdrv_co_do_pwritev(BlockDriverState *bs, |
3338 | int64_t offset, unsigned int bytes, QEMUIOVector *qiov, | |
b404f720 KW |
3339 | BdrvRequestFlags flags) |
3340 | { | |
65afd211 | 3341 | BdrvTrackedRequest req; |
3b8242e0 KW |
3342 | /* TODO Lift BDRV_SECTOR_SIZE restriction in BlockDriver interface */ |
3343 | uint64_t align = MAX(BDRV_SECTOR_SIZE, bs->request_alignment); | |
3344 | uint8_t *head_buf = NULL; | |
3345 | uint8_t *tail_buf = NULL; | |
3346 | QEMUIOVector local_qiov; | |
3347 | bool use_local_qiov = false; | |
b404f720 KW |
3348 | int ret; |
3349 | ||
3350 | if (!bs->drv) { | |
3351 | return -ENOMEDIUM; | |
3352 | } | |
3353 | if (bs->read_only) { | |
3354 | return -EACCES; | |
3355 | } | |
6601553e | 3356 | if (bdrv_check_byte_request(bs, offset, bytes)) { |
b404f720 KW |
3357 | return -EIO; |
3358 | } | |
3359 | ||
b404f720 KW |
3360 | /* throttling disk I/O */ |
3361 | if (bs->io_limits_enabled) { | |
d5103588 | 3362 | bdrv_io_limits_intercept(bs, bytes, true); |
b404f720 KW |
3363 | } |
3364 | ||
3b8242e0 KW |
3365 | /* |
3366 | * Align write if necessary by performing a read-modify-write cycle. | |
3367 | * Pad qiov with the read parts and be sure to have a tracked request not | |
3368 | * only for bdrv_aligned_pwritev, but also for the reads of the RMW cycle. | |
3369 | */ | |
65afd211 | 3370 | tracked_request_begin(&req, bs, offset, bytes, true); |
3b8242e0 KW |
3371 | |
3372 | if (offset & (align - 1)) { | |
3373 | QEMUIOVector head_qiov; | |
3374 | struct iovec head_iov; | |
3375 | ||
3376 | mark_request_serialising(&req, align); | |
3377 | wait_serialising_requests(&req); | |
3378 | ||
3379 | head_buf = qemu_blockalign(bs, align); | |
3380 | head_iov = (struct iovec) { | |
3381 | .iov_base = head_buf, | |
3382 | .iov_len = align, | |
3383 | }; | |
3384 | qemu_iovec_init_external(&head_qiov, &head_iov, 1); | |
3385 | ||
9e1cb96d | 3386 | BLKDBG_EVENT(bs, BLKDBG_PWRITEV_RMW_HEAD); |
3b8242e0 KW |
3387 | ret = bdrv_aligned_preadv(bs, &req, offset & ~(align - 1), align, |
3388 | align, &head_qiov, 0); | |
3389 | if (ret < 0) { | |
3390 | goto fail; | |
3391 | } | |
9e1cb96d | 3392 | BLKDBG_EVENT(bs, BLKDBG_PWRITEV_RMW_AFTER_HEAD); |
3b8242e0 KW |
3393 | |
3394 | qemu_iovec_init(&local_qiov, qiov->niov + 2); | |
3395 | qemu_iovec_add(&local_qiov, head_buf, offset & (align - 1)); | |
3396 | qemu_iovec_concat(&local_qiov, qiov, 0, qiov->size); | |
3397 | use_local_qiov = true; | |
3398 | ||
3399 | bytes += offset & (align - 1); | |
3400 | offset = offset & ~(align - 1); | |
3401 | } | |
3402 | ||
3403 | if ((offset + bytes) & (align - 1)) { | |
3404 | QEMUIOVector tail_qiov; | |
3405 | struct iovec tail_iov; | |
3406 | size_t tail_bytes; | |
28de2dcd | 3407 | bool waited; |
3b8242e0 KW |
3408 | |
3409 | mark_request_serialising(&req, align); | |
28de2dcd KW |
3410 | waited = wait_serialising_requests(&req); |
3411 | assert(!waited || !use_local_qiov); | |
3b8242e0 KW |
3412 | |
3413 | tail_buf = qemu_blockalign(bs, align); | |
3414 | tail_iov = (struct iovec) { | |
3415 | .iov_base = tail_buf, | |
3416 | .iov_len = align, | |
3417 | }; | |
3418 | qemu_iovec_init_external(&tail_qiov, &tail_iov, 1); | |
3419 | ||
9e1cb96d | 3420 | BLKDBG_EVENT(bs, BLKDBG_PWRITEV_RMW_TAIL); |
3b8242e0 KW |
3421 | ret = bdrv_aligned_preadv(bs, &req, (offset + bytes) & ~(align - 1), align, |
3422 | align, &tail_qiov, 0); | |
3423 | if (ret < 0) { | |
3424 | goto fail; | |
3425 | } | |
9e1cb96d | 3426 | BLKDBG_EVENT(bs, BLKDBG_PWRITEV_RMW_AFTER_TAIL); |
3b8242e0 KW |
3427 | |
3428 | if (!use_local_qiov) { | |
3429 | qemu_iovec_init(&local_qiov, qiov->niov + 1); | |
3430 | qemu_iovec_concat(&local_qiov, qiov, 0, qiov->size); | |
3431 | use_local_qiov = true; | |
3432 | } | |
3433 | ||
3434 | tail_bytes = (offset + bytes) & (align - 1); | |
3435 | qemu_iovec_add(&local_qiov, tail_buf + tail_bytes, align - tail_bytes); | |
3436 | ||
3437 | bytes = ROUND_UP(bytes, align); | |
3438 | } | |
3439 | ||
3440 | ret = bdrv_aligned_pwritev(bs, &req, offset, bytes, | |
3441 | use_local_qiov ? &local_qiov : qiov, | |
3442 | flags); | |
3443 | ||
3444 | fail: | |
65afd211 | 3445 | tracked_request_end(&req); |
b404f720 | 3446 | |
3b8242e0 KW |
3447 | if (use_local_qiov) { |
3448 | qemu_iovec_destroy(&local_qiov); | |
3b8242e0 | 3449 | } |
99c4a85c KW |
3450 | qemu_vfree(head_buf); |
3451 | qemu_vfree(tail_buf); | |
3b8242e0 | 3452 | |
b404f720 KW |
3453 | return ret; |
3454 | } | |
3455 | ||
6601553e KW |
3456 | static int coroutine_fn bdrv_co_do_writev(BlockDriverState *bs, |
3457 | int64_t sector_num, int nb_sectors, QEMUIOVector *qiov, | |
3458 | BdrvRequestFlags flags) | |
3459 | { | |
3460 | if (nb_sectors < 0 || nb_sectors > (INT_MAX >> BDRV_SECTOR_BITS)) { | |
3461 | return -EINVAL; | |
3462 | } | |
3463 | ||
3464 | return bdrv_co_do_pwritev(bs, sector_num << BDRV_SECTOR_BITS, | |
3465 | nb_sectors << BDRV_SECTOR_BITS, qiov, flags); | |
3466 | } | |
3467 | ||
c5fbe571 SH |
3468 | int coroutine_fn bdrv_co_writev(BlockDriverState *bs, int64_t sector_num, |
3469 | int nb_sectors, QEMUIOVector *qiov) | |
3470 | { | |
3471 | trace_bdrv_co_writev(bs, sector_num, nb_sectors); | |
3472 | ||
f08f2dda SH |
3473 | return bdrv_co_do_writev(bs, sector_num, nb_sectors, qiov, 0); |
3474 | } | |
3475 | ||
3476 | int coroutine_fn bdrv_co_write_zeroes(BlockDriverState *bs, | |
aa7bfbff PL |
3477 | int64_t sector_num, int nb_sectors, |
3478 | BdrvRequestFlags flags) | |
f08f2dda | 3479 | { |
94d6ff21 | 3480 | trace_bdrv_co_write_zeroes(bs, sector_num, nb_sectors, flags); |
f08f2dda | 3481 | |
d32f35cb PL |
3482 | if (!(bs->open_flags & BDRV_O_UNMAP)) { |
3483 | flags &= ~BDRV_REQ_MAY_UNMAP; | |
3484 | } | |
3485 | ||
f08f2dda | 3486 | return bdrv_co_do_writev(bs, sector_num, nb_sectors, NULL, |
aa7bfbff | 3487 | BDRV_REQ_ZERO_WRITE | flags); |
c5fbe571 SH |
3488 | } |
3489 | ||
83f64091 FB |
3490 | /** |
3491 | * Truncate file to 'offset' bytes (needed only for file protocols) | |
3492 | */ | |
3493 | int bdrv_truncate(BlockDriverState *bs, int64_t offset) | |
3494 | { | |
3495 | BlockDriver *drv = bs->drv; | |
51762288 | 3496 | int ret; |
83f64091 | 3497 | if (!drv) |
19cb3738 | 3498 | return -ENOMEDIUM; |
83f64091 FB |
3499 | if (!drv->bdrv_truncate) |
3500 | return -ENOTSUP; | |
59f2689d NS |
3501 | if (bs->read_only) |
3502 | return -EACCES; | |
9c75e168 | 3503 | |
51762288 SH |
3504 | ret = drv->bdrv_truncate(bs, offset); |
3505 | if (ret == 0) { | |
3506 | ret = refresh_total_sectors(bs, offset >> BDRV_SECTOR_BITS); | |
a7f53e26 MA |
3507 | if (bs->blk) { |
3508 | blk_dev_resize_cb(bs->blk); | |
3509 | } | |
51762288 SH |
3510 | } |
3511 | return ret; | |
83f64091 FB |
3512 | } |
3513 | ||
4a1d5e1f FZ |
3514 | /** |
3515 | * Length of a allocated file in bytes. Sparse files are counted by actual | |
3516 | * allocated space. Return < 0 if error or unknown. | |
3517 | */ | |
3518 | int64_t bdrv_get_allocated_file_size(BlockDriverState *bs) | |
3519 | { | |
3520 | BlockDriver *drv = bs->drv; | |
3521 | if (!drv) { | |
3522 | return -ENOMEDIUM; | |
3523 | } | |
3524 | if (drv->bdrv_get_allocated_file_size) { | |
3525 | return drv->bdrv_get_allocated_file_size(bs); | |
3526 | } | |
3527 | if (bs->file) { | |
3528 | return bdrv_get_allocated_file_size(bs->file); | |
3529 | } | |
3530 | return -ENOTSUP; | |
3531 | } | |
3532 | ||
83f64091 | 3533 | /** |
65a9bb25 | 3534 | * Return number of sectors on success, -errno on error. |
83f64091 | 3535 | */ |
65a9bb25 | 3536 | int64_t bdrv_nb_sectors(BlockDriverState *bs) |
83f64091 FB |
3537 | { |
3538 | BlockDriver *drv = bs->drv; | |
65a9bb25 | 3539 | |
83f64091 | 3540 | if (!drv) |
19cb3738 | 3541 | return -ENOMEDIUM; |
51762288 | 3542 | |
b94a2610 KW |
3543 | if (drv->has_variable_length) { |
3544 | int ret = refresh_total_sectors(bs, bs->total_sectors); | |
3545 | if (ret < 0) { | |
3546 | return ret; | |
46a4e4e6 | 3547 | } |
83f64091 | 3548 | } |
65a9bb25 MA |
3549 | return bs->total_sectors; |
3550 | } | |
3551 | ||
3552 | /** | |
3553 | * Return length in bytes on success, -errno on error. | |
3554 | * The length is always a multiple of BDRV_SECTOR_SIZE. | |
3555 | */ | |
3556 | int64_t bdrv_getlength(BlockDriverState *bs) | |
3557 | { | |
3558 | int64_t ret = bdrv_nb_sectors(bs); | |
3559 | ||
3560 | return ret < 0 ? ret : ret * BDRV_SECTOR_SIZE; | |
fc01f7e7 FB |
3561 | } |
3562 | ||
19cb3738 | 3563 | /* return 0 as number of sectors if no device present or error */ |
96b8f136 | 3564 | void bdrv_get_geometry(BlockDriverState *bs, uint64_t *nb_sectors_ptr) |
fc01f7e7 | 3565 | { |
65a9bb25 MA |
3566 | int64_t nb_sectors = bdrv_nb_sectors(bs); |
3567 | ||
3568 | *nb_sectors_ptr = nb_sectors < 0 ? 0 : nb_sectors; | |
fc01f7e7 | 3569 | } |
cf98951b | 3570 | |
ff06f5f3 PB |
3571 | void bdrv_set_on_error(BlockDriverState *bs, BlockdevOnError on_read_error, |
3572 | BlockdevOnError on_write_error) | |
abd7f68d MA |
3573 | { |
3574 | bs->on_read_error = on_read_error; | |
3575 | bs->on_write_error = on_write_error; | |
3576 | } | |
3577 | ||
1ceee0d5 | 3578 | BlockdevOnError bdrv_get_on_error(BlockDriverState *bs, bool is_read) |
abd7f68d MA |
3579 | { |
3580 | return is_read ? bs->on_read_error : bs->on_write_error; | |
3581 | } | |
3582 | ||
3e1caa5f PB |
3583 | BlockErrorAction bdrv_get_error_action(BlockDriverState *bs, bool is_read, int error) |
3584 | { | |
3585 | BlockdevOnError on_err = is_read ? bs->on_read_error : bs->on_write_error; | |
3586 | ||
3587 | switch (on_err) { | |
3588 | case BLOCKDEV_ON_ERROR_ENOSPC: | |
a589569f WX |
3589 | return (error == ENOSPC) ? |
3590 | BLOCK_ERROR_ACTION_STOP : BLOCK_ERROR_ACTION_REPORT; | |
3e1caa5f | 3591 | case BLOCKDEV_ON_ERROR_STOP: |
a589569f | 3592 | return BLOCK_ERROR_ACTION_STOP; |
3e1caa5f | 3593 | case BLOCKDEV_ON_ERROR_REPORT: |
a589569f | 3594 | return BLOCK_ERROR_ACTION_REPORT; |
3e1caa5f | 3595 | case BLOCKDEV_ON_ERROR_IGNORE: |
a589569f | 3596 | return BLOCK_ERROR_ACTION_IGNORE; |
3e1caa5f PB |
3597 | default: |
3598 | abort(); | |
3599 | } | |
3600 | } | |
3601 | ||
c7c2ff0c LC |
3602 | static void send_qmp_error_event(BlockDriverState *bs, |
3603 | BlockErrorAction action, | |
3604 | bool is_read, int error) | |
3605 | { | |
573742a5 | 3606 | IoOperationType optype; |
c7c2ff0c | 3607 | |
573742a5 PM |
3608 | optype = is_read ? IO_OPERATION_TYPE_READ : IO_OPERATION_TYPE_WRITE; |
3609 | qapi_event_send_block_io_error(bdrv_get_device_name(bs), optype, action, | |
c7c2ff0c | 3610 | bdrv_iostatus_is_enabled(bs), |
624ff573 LC |
3611 | error == ENOSPC, strerror(error), |
3612 | &error_abort); | |
c7c2ff0c LC |
3613 | } |
3614 | ||
3e1caa5f PB |
3615 | /* This is done by device models because, while the block layer knows |
3616 | * about the error, it does not know whether an operation comes from | |
3617 | * the device or the block layer (from a job, for example). | |
3618 | */ | |
3619 | void bdrv_error_action(BlockDriverState *bs, BlockErrorAction action, | |
3620 | bool is_read, int error) | |
3621 | { | |
3622 | assert(error >= 0); | |
2bd3bce8 | 3623 | |
a589569f | 3624 | if (action == BLOCK_ERROR_ACTION_STOP) { |
2bd3bce8 PB |
3625 | /* First set the iostatus, so that "info block" returns an iostatus |
3626 | * that matches the events raised so far (an additional error iostatus | |
3627 | * is fine, but not a lost one). | |
3628 | */ | |
3e1caa5f | 3629 | bdrv_iostatus_set_err(bs, error); |
2bd3bce8 PB |
3630 | |
3631 | /* Then raise the request to stop the VM and the event. | |
3632 | * qemu_system_vmstop_request_prepare has two effects. First, | |
3633 | * it ensures that the STOP event always comes after the | |
3634 | * BLOCK_IO_ERROR event. Second, it ensures that even if management | |
3635 | * can observe the STOP event and do a "cont" before the STOP | |
3636 | * event is issued, the VM will not stop. In this case, vm_start() | |
3637 | * also ensures that the STOP/RESUME pair of events is emitted. | |
3638 | */ | |
3639 | qemu_system_vmstop_request_prepare(); | |
c7c2ff0c | 3640 | send_qmp_error_event(bs, action, is_read, error); |
2bd3bce8 PB |
3641 | qemu_system_vmstop_request(RUN_STATE_IO_ERROR); |
3642 | } else { | |
c7c2ff0c | 3643 | send_qmp_error_event(bs, action, is_read, error); |
3e1caa5f PB |
3644 | } |
3645 | } | |
3646 | ||
b338082b FB |
3647 | int bdrv_is_read_only(BlockDriverState *bs) |
3648 | { | |
3649 | return bs->read_only; | |
3650 | } | |
3651 | ||
985a03b0 TS |
3652 | int bdrv_is_sg(BlockDriverState *bs) |
3653 | { | |
3654 | return bs->sg; | |
3655 | } | |
3656 | ||
e900a7b7 CH |
3657 | int bdrv_enable_write_cache(BlockDriverState *bs) |
3658 | { | |
3659 | return bs->enable_write_cache; | |
3660 | } | |
3661 | ||
425b0148 PB |
3662 | void bdrv_set_enable_write_cache(BlockDriverState *bs, bool wce) |
3663 | { | |
3664 | bs->enable_write_cache = wce; | |
55b110f2 JC |
3665 | |
3666 | /* so a reopen() will preserve wce */ | |
3667 | if (wce) { | |
3668 | bs->open_flags |= BDRV_O_CACHE_WB; | |
3669 | } else { | |
3670 | bs->open_flags &= ~BDRV_O_CACHE_WB; | |
3671 | } | |
425b0148 PB |
3672 | } |
3673 | ||
ea2384d3 FB |
3674 | int bdrv_is_encrypted(BlockDriverState *bs) |
3675 | { | |
3676 | if (bs->backing_hd && bs->backing_hd->encrypted) | |
3677 | return 1; | |
3678 | return bs->encrypted; | |
3679 | } | |
3680 | ||
c0f4ce77 AL |
3681 | int bdrv_key_required(BlockDriverState *bs) |
3682 | { | |
3683 | BlockDriverState *backing_hd = bs->backing_hd; | |
3684 | ||
3685 | if (backing_hd && backing_hd->encrypted && !backing_hd->valid_key) | |
3686 | return 1; | |
3687 | return (bs->encrypted && !bs->valid_key); | |
3688 | } | |
3689 | ||
ea2384d3 FB |
3690 | int bdrv_set_key(BlockDriverState *bs, const char *key) |
3691 | { | |
3692 | int ret; | |
3693 | if (bs->backing_hd && bs->backing_hd->encrypted) { | |
3694 | ret = bdrv_set_key(bs->backing_hd, key); | |
3695 | if (ret < 0) | |
3696 | return ret; | |
3697 | if (!bs->encrypted) | |
3698 | return 0; | |
3699 | } | |
fd04a2ae SH |
3700 | if (!bs->encrypted) { |
3701 | return -EINVAL; | |
3702 | } else if (!bs->drv || !bs->drv->bdrv_set_key) { | |
3703 | return -ENOMEDIUM; | |
3704 | } | |
c0f4ce77 | 3705 | ret = bs->drv->bdrv_set_key(bs, key); |
bb5fc20f AL |
3706 | if (ret < 0) { |
3707 | bs->valid_key = 0; | |
3708 | } else if (!bs->valid_key) { | |
3709 | bs->valid_key = 1; | |
a7f53e26 MA |
3710 | if (bs->blk) { |
3711 | /* call the change callback now, we skipped it on open */ | |
3712 | blk_dev_change_media_cb(bs->blk, true); | |
3713 | } | |
bb5fc20f | 3714 | } |
c0f4ce77 | 3715 | return ret; |
ea2384d3 FB |
3716 | } |
3717 | ||
f8d6bba1 | 3718 | const char *bdrv_get_format_name(BlockDriverState *bs) |
ea2384d3 | 3719 | { |
f8d6bba1 | 3720 | return bs->drv ? bs->drv->format_name : NULL; |
ea2384d3 FB |
3721 | } |
3722 | ||
ada42401 SH |
3723 | static int qsort_strcmp(const void *a, const void *b) |
3724 | { | |
3725 | return strcmp(a, b); | |
3726 | } | |
3727 | ||
5fafdf24 | 3728 | void bdrv_iterate_format(void (*it)(void *opaque, const char *name), |
ea2384d3 FB |
3729 | void *opaque) |
3730 | { | |
3731 | BlockDriver *drv; | |
e855e4fb | 3732 | int count = 0; |
ada42401 | 3733 | int i; |
e855e4fb | 3734 | const char **formats = NULL; |
ea2384d3 | 3735 | |
8a22f02a | 3736 | QLIST_FOREACH(drv, &bdrv_drivers, list) { |
e855e4fb JC |
3737 | if (drv->format_name) { |
3738 | bool found = false; | |
3739 | int i = count; | |
3740 | while (formats && i && !found) { | |
3741 | found = !strcmp(formats[--i], drv->format_name); | |
3742 | } | |
3743 | ||
3744 | if (!found) { | |
5839e53b | 3745 | formats = g_renew(const char *, formats, count + 1); |
e855e4fb | 3746 | formats[count++] = drv->format_name; |
e855e4fb JC |
3747 | } |
3748 | } | |
ea2384d3 | 3749 | } |
ada42401 SH |
3750 | |
3751 | qsort(formats, count, sizeof(formats[0]), qsort_strcmp); | |
3752 | ||
3753 | for (i = 0; i < count; i++) { | |
3754 | it(opaque, formats[i]); | |
3755 | } | |
3756 | ||
e855e4fb | 3757 | g_free(formats); |
ea2384d3 FB |
3758 | } |
3759 | ||
dc364f4c | 3760 | /* This function is to find block backend bs */ |
7f06d47e | 3761 | /* TODO convert callers to blk_by_name(), then remove */ |
b338082b FB |
3762 | BlockDriverState *bdrv_find(const char *name) |
3763 | { | |
7f06d47e | 3764 | BlockBackend *blk = blk_by_name(name); |
b338082b | 3765 | |
7f06d47e | 3766 | return blk ? blk_bs(blk) : NULL; |
b338082b FB |
3767 | } |
3768 | ||
dc364f4c BC |
3769 | /* This function is to find a node in the bs graph */ |
3770 | BlockDriverState *bdrv_find_node(const char *node_name) | |
3771 | { | |
3772 | BlockDriverState *bs; | |
3773 | ||
3774 | assert(node_name); | |
3775 | ||
3776 | QTAILQ_FOREACH(bs, &graph_bdrv_states, node_list) { | |
3777 | if (!strcmp(node_name, bs->node_name)) { | |
3778 | return bs; | |
3779 | } | |
3780 | } | |
3781 | return NULL; | |
3782 | } | |
3783 | ||
c13163fb BC |
3784 | /* Put this QMP function here so it can access the static graph_bdrv_states. */ |
3785 | BlockDeviceInfoList *bdrv_named_nodes_list(void) | |
3786 | { | |
3787 | BlockDeviceInfoList *list, *entry; | |
3788 | BlockDriverState *bs; | |
3789 | ||
3790 | list = NULL; | |
3791 | QTAILQ_FOREACH(bs, &graph_bdrv_states, node_list) { | |
3792 | entry = g_malloc0(sizeof(*entry)); | |
3793 | entry->value = bdrv_block_device_info(bs); | |
3794 | entry->next = list; | |
3795 | list = entry; | |
3796 | } | |
3797 | ||
3798 | return list; | |
3799 | } | |
3800 | ||
12d3ba82 BC |
3801 | BlockDriverState *bdrv_lookup_bs(const char *device, |
3802 | const char *node_name, | |
3803 | Error **errp) | |
3804 | { | |
7f06d47e MA |
3805 | BlockBackend *blk; |
3806 | BlockDriverState *bs; | |
12d3ba82 | 3807 | |
12d3ba82 | 3808 | if (device) { |
7f06d47e | 3809 | blk = blk_by_name(device); |
12d3ba82 | 3810 | |
7f06d47e MA |
3811 | if (blk) { |
3812 | return blk_bs(blk); | |
12d3ba82 | 3813 | } |
12d3ba82 BC |
3814 | } |
3815 | ||
dd67fa50 BC |
3816 | if (node_name) { |
3817 | bs = bdrv_find_node(node_name); | |
12d3ba82 | 3818 | |
dd67fa50 BC |
3819 | if (bs) { |
3820 | return bs; | |
3821 | } | |
12d3ba82 BC |
3822 | } |
3823 | ||
dd67fa50 BC |
3824 | error_setg(errp, "Cannot find device=%s nor node_name=%s", |
3825 | device ? device : "", | |
3826 | node_name ? node_name : ""); | |
3827 | return NULL; | |
12d3ba82 BC |
3828 | } |
3829 | ||
5a6684d2 JC |
3830 | /* If 'base' is in the same chain as 'top', return true. Otherwise, |
3831 | * return false. If either argument is NULL, return false. */ | |
3832 | bool bdrv_chain_contains(BlockDriverState *top, BlockDriverState *base) | |
3833 | { | |
3834 | while (top && top != base) { | |
3835 | top = top->backing_hd; | |
3836 | } | |
3837 | ||
3838 | return top != NULL; | |
3839 | } | |
3840 | ||
04df765a FZ |
3841 | BlockDriverState *bdrv_next_node(BlockDriverState *bs) |
3842 | { | |
3843 | if (!bs) { | |
3844 | return QTAILQ_FIRST(&graph_bdrv_states); | |
3845 | } | |
3846 | return QTAILQ_NEXT(bs, node_list); | |
3847 | } | |
3848 | ||
2f399b0a MA |
3849 | BlockDriverState *bdrv_next(BlockDriverState *bs) |
3850 | { | |
3851 | if (!bs) { | |
3852 | return QTAILQ_FIRST(&bdrv_states); | |
3853 | } | |
dc364f4c | 3854 | return QTAILQ_NEXT(bs, device_list); |
2f399b0a MA |
3855 | } |
3856 | ||
20a9e77d FZ |
3857 | const char *bdrv_get_node_name(const BlockDriverState *bs) |
3858 | { | |
3859 | return bs->node_name; | |
3860 | } | |
3861 | ||
7f06d47e | 3862 | /* TODO check what callers really want: bs->node_name or blk_name() */ |
bfb197e0 | 3863 | const char *bdrv_get_device_name(const BlockDriverState *bs) |
ea2384d3 | 3864 | { |
bfb197e0 | 3865 | return bs->blk ? blk_name(bs->blk) : ""; |
ea2384d3 FB |
3866 | } |
3867 | ||
c8433287 MA |
3868 | int bdrv_get_flags(BlockDriverState *bs) |
3869 | { | |
3870 | return bs->open_flags; | |
3871 | } | |
3872 | ||
f0f0fdfe | 3873 | int bdrv_flush_all(void) |
c6ca28d6 AL |
3874 | { |
3875 | BlockDriverState *bs; | |
f0f0fdfe | 3876 | int result = 0; |
c6ca28d6 | 3877 | |
dc364f4c | 3878 | QTAILQ_FOREACH(bs, &bdrv_states, device_list) { |
ed78cda3 SH |
3879 | AioContext *aio_context = bdrv_get_aio_context(bs); |
3880 | int ret; | |
3881 | ||
3882 | aio_context_acquire(aio_context); | |
3883 | ret = bdrv_flush(bs); | |
f0f0fdfe KW |
3884 | if (ret < 0 && !result) { |
3885 | result = ret; | |
3886 | } | |
ed78cda3 | 3887 | aio_context_release(aio_context); |
1b7bdbc1 | 3888 | } |
f0f0fdfe KW |
3889 | |
3890 | return result; | |
c6ca28d6 AL |
3891 | } |
3892 | ||
3ac21627 PL |
3893 | int bdrv_has_zero_init_1(BlockDriverState *bs) |
3894 | { | |
3895 | return 1; | |
3896 | } | |
3897 | ||
f2feebbd KW |
3898 | int bdrv_has_zero_init(BlockDriverState *bs) |
3899 | { | |
3900 | assert(bs->drv); | |
3901 | ||
11212d8f PB |
3902 | /* If BS is a copy on write image, it is initialized to |
3903 | the contents of the base image, which may not be zeroes. */ | |
3904 | if (bs->backing_hd) { | |
3905 | return 0; | |
3906 | } | |
336c1c12 KW |
3907 | if (bs->drv->bdrv_has_zero_init) { |
3908 | return bs->drv->bdrv_has_zero_init(bs); | |
f2feebbd KW |
3909 | } |
3910 | ||
3ac21627 PL |
3911 | /* safe default */ |
3912 | return 0; | |
f2feebbd KW |
3913 | } |
3914 | ||
4ce78691 PL |
3915 | bool bdrv_unallocated_blocks_are_zero(BlockDriverState *bs) |
3916 | { | |
3917 | BlockDriverInfo bdi; | |
3918 | ||
3919 | if (bs->backing_hd) { | |
3920 | return false; | |
3921 | } | |
3922 | ||
3923 | if (bdrv_get_info(bs, &bdi) == 0) { | |
3924 | return bdi.unallocated_blocks_are_zero; | |
3925 | } | |
3926 | ||
3927 | return false; | |
3928 | } | |
3929 | ||
3930 | bool bdrv_can_write_zeroes_with_unmap(BlockDriverState *bs) | |
3931 | { | |
3932 | BlockDriverInfo bdi; | |
3933 | ||
3934 | if (bs->backing_hd || !(bs->open_flags & BDRV_O_UNMAP)) { | |
3935 | return false; | |
3936 | } | |
3937 | ||
3938 | if (bdrv_get_info(bs, &bdi) == 0) { | |
3939 | return bdi.can_write_zeroes_with_unmap; | |
3940 | } | |
3941 | ||
3942 | return false; | |
3943 | } | |
3944 | ||
b6b8a333 | 3945 | typedef struct BdrvCoGetBlockStatusData { |
376ae3f1 | 3946 | BlockDriverState *bs; |
b35b2bba | 3947 | BlockDriverState *base; |
376ae3f1 SH |
3948 | int64_t sector_num; |
3949 | int nb_sectors; | |
3950 | int *pnum; | |
b6b8a333 | 3951 | int64_t ret; |
376ae3f1 | 3952 | bool done; |
b6b8a333 | 3953 | } BdrvCoGetBlockStatusData; |
376ae3f1 | 3954 | |
f58c7b35 | 3955 | /* |
705be728 FZ |
3956 | * Returns the allocation status of the specified sectors. |
3957 | * Drivers not implementing the functionality are assumed to not support | |
3958 | * backing files, hence all their sectors are reported as allocated. | |
f58c7b35 | 3959 | * |
bd9533e3 SH |
3960 | * If 'sector_num' is beyond the end of the disk image the return value is 0 |
3961 | * and 'pnum' is set to 0. | |
3962 | * | |
f58c7b35 TS |
3963 | * 'pnum' is set to the number of sectors (including and immediately following |
3964 | * the specified sector) that are known to be in the same | |
3965 | * allocated/unallocated state. | |
3966 | * | |
bd9533e3 SH |
3967 | * 'nb_sectors' is the max value 'pnum' should be set to. If nb_sectors goes |
3968 | * beyond the end of the disk image it will be clamped. | |
f58c7b35 | 3969 | */ |
b6b8a333 PB |
3970 | static int64_t coroutine_fn bdrv_co_get_block_status(BlockDriverState *bs, |
3971 | int64_t sector_num, | |
3972 | int nb_sectors, int *pnum) | |
f58c7b35 | 3973 | { |
30a7f2fc | 3974 | int64_t total_sectors; |
bd9533e3 | 3975 | int64_t n; |
5daa74a6 | 3976 | int64_t ret, ret2; |
bd9533e3 | 3977 | |
30a7f2fc MA |
3978 | total_sectors = bdrv_nb_sectors(bs); |
3979 | if (total_sectors < 0) { | |
3980 | return total_sectors; | |
617ccb46 PB |
3981 | } |
3982 | ||
30a7f2fc | 3983 | if (sector_num >= total_sectors) { |
bd9533e3 SH |
3984 | *pnum = 0; |
3985 | return 0; | |
3986 | } | |
3987 | ||
30a7f2fc | 3988 | n = total_sectors - sector_num; |
bd9533e3 SH |
3989 | if (n < nb_sectors) { |
3990 | nb_sectors = n; | |
3991 | } | |
3992 | ||
b6b8a333 | 3993 | if (!bs->drv->bdrv_co_get_block_status) { |
bd9533e3 | 3994 | *pnum = nb_sectors; |
e88ae226 | 3995 | ret = BDRV_BLOCK_DATA | BDRV_BLOCK_ALLOCATED; |
918e92d7 PB |
3996 | if (bs->drv->protocol_name) { |
3997 | ret |= BDRV_BLOCK_OFFSET_VALID | (sector_num * BDRV_SECTOR_SIZE); | |
3998 | } | |
3999 | return ret; | |
f58c7b35 | 4000 | } |
6aebab14 | 4001 | |
415b5b01 PB |
4002 | ret = bs->drv->bdrv_co_get_block_status(bs, sector_num, nb_sectors, pnum); |
4003 | if (ret < 0) { | |
3e0a233d | 4004 | *pnum = 0; |
415b5b01 PB |
4005 | return ret; |
4006 | } | |
4007 | ||
92bc50a5 PL |
4008 | if (ret & BDRV_BLOCK_RAW) { |
4009 | assert(ret & BDRV_BLOCK_OFFSET_VALID); | |
4010 | return bdrv_get_block_status(bs->file, ret >> BDRV_SECTOR_BITS, | |
4011 | *pnum, pnum); | |
4012 | } | |
4013 | ||
e88ae226 KW |
4014 | if (ret & (BDRV_BLOCK_DATA | BDRV_BLOCK_ZERO)) { |
4015 | ret |= BDRV_BLOCK_ALLOCATED; | |
4016 | } | |
4017 | ||
c3d86884 PL |
4018 | if (!(ret & BDRV_BLOCK_DATA) && !(ret & BDRV_BLOCK_ZERO)) { |
4019 | if (bdrv_unallocated_blocks_are_zero(bs)) { | |
f0ad5712 | 4020 | ret |= BDRV_BLOCK_ZERO; |
1f9db224 | 4021 | } else if (bs->backing_hd) { |
f0ad5712 | 4022 | BlockDriverState *bs2 = bs->backing_hd; |
30a7f2fc MA |
4023 | int64_t nb_sectors2 = bdrv_nb_sectors(bs2); |
4024 | if (nb_sectors2 >= 0 && sector_num >= nb_sectors2) { | |
f0ad5712 PB |
4025 | ret |= BDRV_BLOCK_ZERO; |
4026 | } | |
4027 | } | |
415b5b01 | 4028 | } |
5daa74a6 PB |
4029 | |
4030 | if (bs->file && | |
4031 | (ret & BDRV_BLOCK_DATA) && !(ret & BDRV_BLOCK_ZERO) && | |
4032 | (ret & BDRV_BLOCK_OFFSET_VALID)) { | |
59c9a95f HR |
4033 | int file_pnum; |
4034 | ||
5daa74a6 | 4035 | ret2 = bdrv_co_get_block_status(bs->file, ret >> BDRV_SECTOR_BITS, |
59c9a95f | 4036 | *pnum, &file_pnum); |
5daa74a6 PB |
4037 | if (ret2 >= 0) { |
4038 | /* Ignore errors. This is just providing extra information, it | |
4039 | * is useful but not necessary. | |
4040 | */ | |
59c9a95f HR |
4041 | if (!file_pnum) { |
4042 | /* !file_pnum indicates an offset at or beyond the EOF; it is | |
4043 | * perfectly valid for the format block driver to point to such | |
4044 | * offsets, so catch it and mark everything as zero */ | |
4045 | ret |= BDRV_BLOCK_ZERO; | |
4046 | } else { | |
4047 | /* Limit request to the range reported by the protocol driver */ | |
4048 | *pnum = file_pnum; | |
4049 | ret |= (ret2 & BDRV_BLOCK_ZERO); | |
4050 | } | |
5daa74a6 PB |
4051 | } |
4052 | } | |
4053 | ||
415b5b01 | 4054 | return ret; |
060f51c9 SH |
4055 | } |
4056 | ||
b6b8a333 PB |
4057 | /* Coroutine wrapper for bdrv_get_block_status() */ |
4058 | static void coroutine_fn bdrv_get_block_status_co_entry(void *opaque) | |
060f51c9 | 4059 | { |
b6b8a333 | 4060 | BdrvCoGetBlockStatusData *data = opaque; |
060f51c9 SH |
4061 | BlockDriverState *bs = data->bs; |
4062 | ||
b6b8a333 PB |
4063 | data->ret = bdrv_co_get_block_status(bs, data->sector_num, data->nb_sectors, |
4064 | data->pnum); | |
060f51c9 SH |
4065 | data->done = true; |
4066 | } | |
4067 | ||
4068 | /* | |
b6b8a333 | 4069 | * Synchronous wrapper around bdrv_co_get_block_status(). |
060f51c9 | 4070 | * |
b6b8a333 | 4071 | * See bdrv_co_get_block_status() for details. |
060f51c9 | 4072 | */ |
b6b8a333 PB |
4073 | int64_t bdrv_get_block_status(BlockDriverState *bs, int64_t sector_num, |
4074 | int nb_sectors, int *pnum) | |
060f51c9 | 4075 | { |
6aebab14 | 4076 | Coroutine *co; |
b6b8a333 | 4077 | BdrvCoGetBlockStatusData data = { |
6aebab14 SH |
4078 | .bs = bs, |
4079 | .sector_num = sector_num, | |
4080 | .nb_sectors = nb_sectors, | |
4081 | .pnum = pnum, | |
4082 | .done = false, | |
4083 | }; | |
4084 | ||
bdad13b9 PB |
4085 | if (qemu_in_coroutine()) { |
4086 | /* Fast-path if already in coroutine context */ | |
b6b8a333 | 4087 | bdrv_get_block_status_co_entry(&data); |
bdad13b9 | 4088 | } else { |
2572b37a SH |
4089 | AioContext *aio_context = bdrv_get_aio_context(bs); |
4090 | ||
b6b8a333 | 4091 | co = qemu_coroutine_create(bdrv_get_block_status_co_entry); |
bdad13b9 PB |
4092 | qemu_coroutine_enter(co, &data); |
4093 | while (!data.done) { | |
2572b37a | 4094 | aio_poll(aio_context, true); |
bdad13b9 | 4095 | } |
6aebab14 SH |
4096 | } |
4097 | return data.ret; | |
f58c7b35 TS |
4098 | } |
4099 | ||
b6b8a333 PB |
4100 | int coroutine_fn bdrv_is_allocated(BlockDriverState *bs, int64_t sector_num, |
4101 | int nb_sectors, int *pnum) | |
4102 | { | |
4333bb71 PB |
4103 | int64_t ret = bdrv_get_block_status(bs, sector_num, nb_sectors, pnum); |
4104 | if (ret < 0) { | |
4105 | return ret; | |
4106 | } | |
01fb2705 | 4107 | return !!(ret & BDRV_BLOCK_ALLOCATED); |
b6b8a333 PB |
4108 | } |
4109 | ||
188a7bbf PB |
4110 | /* |
4111 | * Given an image chain: ... -> [BASE] -> [INTER1] -> [INTER2] -> [TOP] | |
4112 | * | |
4113 | * Return true if the given sector is allocated in any image between | |
4114 | * BASE and TOP (inclusive). BASE can be NULL to check if the given | |
4115 | * sector is allocated in any image of the chain. Return false otherwise. | |
4116 | * | |
4117 | * 'pnum' is set to the number of sectors (including and immediately following | |
4118 | * the specified sector) that are known to be in the same | |
4119 | * allocated/unallocated state. | |
4120 | * | |
4121 | */ | |
4f578637 PB |
4122 | int bdrv_is_allocated_above(BlockDriverState *top, |
4123 | BlockDriverState *base, | |
4124 | int64_t sector_num, | |
4125 | int nb_sectors, int *pnum) | |
188a7bbf PB |
4126 | { |
4127 | BlockDriverState *intermediate; | |
4128 | int ret, n = nb_sectors; | |
4129 | ||
4130 | intermediate = top; | |
4131 | while (intermediate && intermediate != base) { | |
4132 | int pnum_inter; | |
bdad13b9 PB |
4133 | ret = bdrv_is_allocated(intermediate, sector_num, nb_sectors, |
4134 | &pnum_inter); | |
188a7bbf PB |
4135 | if (ret < 0) { |
4136 | return ret; | |
4137 | } else if (ret) { | |
4138 | *pnum = pnum_inter; | |
4139 | return 1; | |
4140 | } | |
4141 | ||
4142 | /* | |
4143 | * [sector_num, nb_sectors] is unallocated on top but intermediate | |
4144 | * might have | |
4145 | * | |
4146 | * [sector_num+x, nr_sectors] allocated. | |
4147 | */ | |
63ba17d3 VI |
4148 | if (n > pnum_inter && |
4149 | (intermediate == top || | |
4150 | sector_num + pnum_inter < intermediate->total_sectors)) { | |
188a7bbf PB |
4151 | n = pnum_inter; |
4152 | } | |
4153 | ||
4154 | intermediate = intermediate->backing_hd; | |
4155 | } | |
4156 | ||
4157 | *pnum = n; | |
4158 | return 0; | |
4159 | } | |
4160 | ||
045df330 AL |
4161 | const char *bdrv_get_encrypted_filename(BlockDriverState *bs) |
4162 | { | |
4163 | if (bs->backing_hd && bs->backing_hd->encrypted) | |
4164 | return bs->backing_file; | |
4165 | else if (bs->encrypted) | |
4166 | return bs->filename; | |
4167 | else | |
4168 | return NULL; | |
4169 | } | |
4170 | ||
5fafdf24 | 4171 | void bdrv_get_backing_filename(BlockDriverState *bs, |
83f64091 FB |
4172 | char *filename, int filename_size) |
4173 | { | |
3574c608 | 4174 | pstrcpy(filename, filename_size, bs->backing_file); |
83f64091 FB |
4175 | } |
4176 | ||
5fafdf24 | 4177 | int bdrv_write_compressed(BlockDriverState *bs, int64_t sector_num, |
faea38e7 FB |
4178 | const uint8_t *buf, int nb_sectors) |
4179 | { | |
4180 | BlockDriver *drv = bs->drv; | |
4181 | if (!drv) | |
19cb3738 | 4182 | return -ENOMEDIUM; |
faea38e7 FB |
4183 | if (!drv->bdrv_write_compressed) |
4184 | return -ENOTSUP; | |
fbb7b4e0 KW |
4185 | if (bdrv_check_request(bs, sector_num, nb_sectors)) |
4186 | return -EIO; | |
a55eb92c | 4187 | |
e4654d2d | 4188 | assert(QLIST_EMPTY(&bs->dirty_bitmaps)); |
a55eb92c | 4189 | |
faea38e7 FB |
4190 | return drv->bdrv_write_compressed(bs, sector_num, buf, nb_sectors); |
4191 | } | |
3b46e624 | 4192 | |
faea38e7 FB |
4193 | int bdrv_get_info(BlockDriverState *bs, BlockDriverInfo *bdi) |
4194 | { | |
4195 | BlockDriver *drv = bs->drv; | |
4196 | if (!drv) | |
19cb3738 | 4197 | return -ENOMEDIUM; |
faea38e7 FB |
4198 | if (!drv->bdrv_get_info) |
4199 | return -ENOTSUP; | |
4200 | memset(bdi, 0, sizeof(*bdi)); | |
4201 | return drv->bdrv_get_info(bs, bdi); | |
4202 | } | |
4203 | ||
eae041fe HR |
4204 | ImageInfoSpecific *bdrv_get_specific_info(BlockDriverState *bs) |
4205 | { | |
4206 | BlockDriver *drv = bs->drv; | |
4207 | if (drv && drv->bdrv_get_specific_info) { | |
4208 | return drv->bdrv_get_specific_info(bs); | |
4209 | } | |
4210 | return NULL; | |
4211 | } | |
4212 | ||
45566e9c CH |
4213 | int bdrv_save_vmstate(BlockDriverState *bs, const uint8_t *buf, |
4214 | int64_t pos, int size) | |
cf8074b3 KW |
4215 | { |
4216 | QEMUIOVector qiov; | |
4217 | struct iovec iov = { | |
4218 | .iov_base = (void *) buf, | |
4219 | .iov_len = size, | |
4220 | }; | |
4221 | ||
4222 | qemu_iovec_init_external(&qiov, &iov, 1); | |
4223 | return bdrv_writev_vmstate(bs, &qiov, pos); | |
4224 | } | |
4225 | ||
4226 | int bdrv_writev_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos) | |
178e08a5 AL |
4227 | { |
4228 | BlockDriver *drv = bs->drv; | |
cf8074b3 KW |
4229 | |
4230 | if (!drv) { | |
178e08a5 | 4231 | return -ENOMEDIUM; |
cf8074b3 KW |
4232 | } else if (drv->bdrv_save_vmstate) { |
4233 | return drv->bdrv_save_vmstate(bs, qiov, pos); | |
4234 | } else if (bs->file) { | |
4235 | return bdrv_writev_vmstate(bs->file, qiov, pos); | |
4236 | } | |
4237 | ||
7cdb1f6d | 4238 | return -ENOTSUP; |
178e08a5 AL |
4239 | } |
4240 | ||
45566e9c CH |
4241 | int bdrv_load_vmstate(BlockDriverState *bs, uint8_t *buf, |
4242 | int64_t pos, int size) | |
178e08a5 AL |
4243 | { |
4244 | BlockDriver *drv = bs->drv; | |
4245 | if (!drv) | |
4246 | return -ENOMEDIUM; | |
7cdb1f6d MK |
4247 | if (drv->bdrv_load_vmstate) |
4248 | return drv->bdrv_load_vmstate(bs, buf, pos, size); | |
4249 | if (bs->file) | |
4250 | return bdrv_load_vmstate(bs->file, buf, pos, size); | |
4251 | return -ENOTSUP; | |
178e08a5 AL |
4252 | } |
4253 | ||
8b9b0cc2 KW |
4254 | void bdrv_debug_event(BlockDriverState *bs, BlkDebugEvent event) |
4255 | { | |
bf736fe3 | 4256 | if (!bs || !bs->drv || !bs->drv->bdrv_debug_event) { |
8b9b0cc2 KW |
4257 | return; |
4258 | } | |
4259 | ||
bf736fe3 | 4260 | bs->drv->bdrv_debug_event(bs, event); |
41c695c7 KW |
4261 | } |
4262 | ||
4263 | int bdrv_debug_breakpoint(BlockDriverState *bs, const char *event, | |
4264 | const char *tag) | |
4265 | { | |
4266 | while (bs && bs->drv && !bs->drv->bdrv_debug_breakpoint) { | |
4267 | bs = bs->file; | |
4268 | } | |
4269 | ||
4270 | if (bs && bs->drv && bs->drv->bdrv_debug_breakpoint) { | |
4271 | return bs->drv->bdrv_debug_breakpoint(bs, event, tag); | |
4272 | } | |
4273 | ||
4274 | return -ENOTSUP; | |
4275 | } | |
4276 | ||
4cc70e93 FZ |
4277 | int bdrv_debug_remove_breakpoint(BlockDriverState *bs, const char *tag) |
4278 | { | |
4279 | while (bs && bs->drv && !bs->drv->bdrv_debug_remove_breakpoint) { | |
4280 | bs = bs->file; | |
4281 | } | |
4282 | ||
4283 | if (bs && bs->drv && bs->drv->bdrv_debug_remove_breakpoint) { | |
4284 | return bs->drv->bdrv_debug_remove_breakpoint(bs, tag); | |
4285 | } | |
4286 | ||
4287 | return -ENOTSUP; | |
4288 | } | |
4289 | ||
41c695c7 KW |
4290 | int bdrv_debug_resume(BlockDriverState *bs, const char *tag) |
4291 | { | |
938789ea | 4292 | while (bs && (!bs->drv || !bs->drv->bdrv_debug_resume)) { |
41c695c7 KW |
4293 | bs = bs->file; |
4294 | } | |
8b9b0cc2 | 4295 | |
41c695c7 KW |
4296 | if (bs && bs->drv && bs->drv->bdrv_debug_resume) { |
4297 | return bs->drv->bdrv_debug_resume(bs, tag); | |
4298 | } | |
4299 | ||
4300 | return -ENOTSUP; | |
4301 | } | |
4302 | ||
4303 | bool bdrv_debug_is_suspended(BlockDriverState *bs, const char *tag) | |
4304 | { | |
4305 | while (bs && bs->drv && !bs->drv->bdrv_debug_is_suspended) { | |
4306 | bs = bs->file; | |
4307 | } | |
4308 | ||
4309 | if (bs && bs->drv && bs->drv->bdrv_debug_is_suspended) { | |
4310 | return bs->drv->bdrv_debug_is_suspended(bs, tag); | |
4311 | } | |
4312 | ||
4313 | return false; | |
8b9b0cc2 KW |
4314 | } |
4315 | ||
199630b6 BS |
4316 | int bdrv_is_snapshot(BlockDriverState *bs) |
4317 | { | |
4318 | return !!(bs->open_flags & BDRV_O_SNAPSHOT); | |
4319 | } | |
4320 | ||
b1b1d783 JC |
4321 | /* backing_file can either be relative, or absolute, or a protocol. If it is |
4322 | * relative, it must be relative to the chain. So, passing in bs->filename | |
4323 | * from a BDS as backing_file should not be done, as that may be relative to | |
4324 | * the CWD rather than the chain. */ | |
e8a6bb9c MT |
4325 | BlockDriverState *bdrv_find_backing_image(BlockDriverState *bs, |
4326 | const char *backing_file) | |
4327 | { | |
b1b1d783 JC |
4328 | char *filename_full = NULL; |
4329 | char *backing_file_full = NULL; | |
4330 | char *filename_tmp = NULL; | |
4331 | int is_protocol = 0; | |
4332 | BlockDriverState *curr_bs = NULL; | |
4333 | BlockDriverState *retval = NULL; | |
4334 | ||
4335 | if (!bs || !bs->drv || !backing_file) { | |
e8a6bb9c MT |
4336 | return NULL; |
4337 | } | |
4338 | ||
b1b1d783 JC |
4339 | filename_full = g_malloc(PATH_MAX); |
4340 | backing_file_full = g_malloc(PATH_MAX); | |
4341 | filename_tmp = g_malloc(PATH_MAX); | |
4342 | ||
4343 | is_protocol = path_has_protocol(backing_file); | |
4344 | ||
4345 | for (curr_bs = bs; curr_bs->backing_hd; curr_bs = curr_bs->backing_hd) { | |
4346 | ||
4347 | /* If either of the filename paths is actually a protocol, then | |
4348 | * compare unmodified paths; otherwise make paths relative */ | |
4349 | if (is_protocol || path_has_protocol(curr_bs->backing_file)) { | |
4350 | if (strcmp(backing_file, curr_bs->backing_file) == 0) { | |
4351 | retval = curr_bs->backing_hd; | |
4352 | break; | |
4353 | } | |
e8a6bb9c | 4354 | } else { |
b1b1d783 JC |
4355 | /* If not an absolute filename path, make it relative to the current |
4356 | * image's filename path */ | |
4357 | path_combine(filename_tmp, PATH_MAX, curr_bs->filename, | |
4358 | backing_file); | |
4359 | ||
4360 | /* We are going to compare absolute pathnames */ | |
4361 | if (!realpath(filename_tmp, filename_full)) { | |
4362 | continue; | |
4363 | } | |
4364 | ||
4365 | /* We need to make sure the backing filename we are comparing against | |
4366 | * is relative to the current image filename (or absolute) */ | |
4367 | path_combine(filename_tmp, PATH_MAX, curr_bs->filename, | |
4368 | curr_bs->backing_file); | |
4369 | ||
4370 | if (!realpath(filename_tmp, backing_file_full)) { | |
4371 | continue; | |
4372 | } | |
4373 | ||
4374 | if (strcmp(backing_file_full, filename_full) == 0) { | |
4375 | retval = curr_bs->backing_hd; | |
4376 | break; | |
4377 | } | |
e8a6bb9c MT |
4378 | } |
4379 | } | |
4380 | ||
b1b1d783 JC |
4381 | g_free(filename_full); |
4382 | g_free(backing_file_full); | |
4383 | g_free(filename_tmp); | |
4384 | return retval; | |
e8a6bb9c MT |
4385 | } |
4386 | ||
f198fd1c BC |
4387 | int bdrv_get_backing_file_depth(BlockDriverState *bs) |
4388 | { | |
4389 | if (!bs->drv) { | |
4390 | return 0; | |
4391 | } | |
4392 | ||
4393 | if (!bs->backing_hd) { | |
4394 | return 0; | |
4395 | } | |
4396 | ||
4397 | return 1 + bdrv_get_backing_file_depth(bs->backing_hd); | |
4398 | } | |
4399 | ||
ea2384d3 | 4400 | /**************************************************************/ |
83f64091 | 4401 | /* async I/Os */ |
ea2384d3 | 4402 | |
7c84b1b8 MA |
4403 | BlockAIOCB *bdrv_aio_readv(BlockDriverState *bs, int64_t sector_num, |
4404 | QEMUIOVector *qiov, int nb_sectors, | |
097310b5 | 4405 | BlockCompletionFunc *cb, void *opaque) |
83f64091 | 4406 | { |
bbf0a440 SH |
4407 | trace_bdrv_aio_readv(bs, sector_num, nb_sectors, opaque); |
4408 | ||
d20d9b7c | 4409 | return bdrv_co_aio_rw_vector(bs, sector_num, qiov, nb_sectors, 0, |
8c5873d6 | 4410 | cb, opaque, false); |
ea2384d3 FB |
4411 | } |
4412 | ||
7c84b1b8 MA |
4413 | BlockAIOCB *bdrv_aio_writev(BlockDriverState *bs, int64_t sector_num, |
4414 | QEMUIOVector *qiov, int nb_sectors, | |
097310b5 | 4415 | BlockCompletionFunc *cb, void *opaque) |
ea2384d3 | 4416 | { |
bbf0a440 SH |
4417 | trace_bdrv_aio_writev(bs, sector_num, nb_sectors, opaque); |
4418 | ||
d20d9b7c | 4419 | return bdrv_co_aio_rw_vector(bs, sector_num, qiov, nb_sectors, 0, |
8c5873d6 | 4420 | cb, opaque, true); |
83f64091 FB |
4421 | } |
4422 | ||
7c84b1b8 | 4423 | BlockAIOCB *bdrv_aio_write_zeroes(BlockDriverState *bs, |
d5ef94d4 | 4424 | int64_t sector_num, int nb_sectors, BdrvRequestFlags flags, |
097310b5 | 4425 | BlockCompletionFunc *cb, void *opaque) |
d5ef94d4 PB |
4426 | { |
4427 | trace_bdrv_aio_write_zeroes(bs, sector_num, nb_sectors, flags, opaque); | |
4428 | ||
4429 | return bdrv_co_aio_rw_vector(bs, sector_num, NULL, nb_sectors, | |
4430 | BDRV_REQ_ZERO_WRITE | flags, | |
4431 | cb, opaque, true); | |
4432 | } | |
4433 | ||
40b4f539 KW |
4434 | |
4435 | typedef struct MultiwriteCB { | |
4436 | int error; | |
4437 | int num_requests; | |
4438 | int num_callbacks; | |
4439 | struct { | |
097310b5 | 4440 | BlockCompletionFunc *cb; |
40b4f539 KW |
4441 | void *opaque; |
4442 | QEMUIOVector *free_qiov; | |
40b4f539 KW |
4443 | } callbacks[]; |
4444 | } MultiwriteCB; | |
4445 | ||
4446 | static void multiwrite_user_cb(MultiwriteCB *mcb) | |
4447 | { | |
4448 | int i; | |
4449 | ||
4450 | for (i = 0; i < mcb->num_callbacks; i++) { | |
4451 | mcb->callbacks[i].cb(mcb->callbacks[i].opaque, mcb->error); | |
1e1ea48d SH |
4452 | if (mcb->callbacks[i].free_qiov) { |
4453 | qemu_iovec_destroy(mcb->callbacks[i].free_qiov); | |
4454 | } | |
7267c094 | 4455 | g_free(mcb->callbacks[i].free_qiov); |
40b4f539 KW |
4456 | } |
4457 | } | |
4458 | ||
4459 | static void multiwrite_cb(void *opaque, int ret) | |
4460 | { | |
4461 | MultiwriteCB *mcb = opaque; | |
4462 | ||
6d519a5f SH |
4463 | trace_multiwrite_cb(mcb, ret); |
4464 | ||
cb6d3ca0 | 4465 | if (ret < 0 && !mcb->error) { |
40b4f539 | 4466 | mcb->error = ret; |
40b4f539 KW |
4467 | } |
4468 | ||
4469 | mcb->num_requests--; | |
4470 | if (mcb->num_requests == 0) { | |
de189a1b | 4471 | multiwrite_user_cb(mcb); |
7267c094 | 4472 | g_free(mcb); |
40b4f539 KW |
4473 | } |
4474 | } | |
4475 | ||
4476 | static int multiwrite_req_compare(const void *a, const void *b) | |
4477 | { | |
77be4366 CH |
4478 | const BlockRequest *req1 = a, *req2 = b; |
4479 | ||
4480 | /* | |
4481 | * Note that we can't simply subtract req2->sector from req1->sector | |
4482 | * here as that could overflow the return value. | |
4483 | */ | |
4484 | if (req1->sector > req2->sector) { | |
4485 | return 1; | |
4486 | } else if (req1->sector < req2->sector) { | |
4487 | return -1; | |
4488 | } else { | |
4489 | return 0; | |
4490 | } | |
40b4f539 KW |
4491 | } |
4492 | ||
4493 | /* | |
4494 | * Takes a bunch of requests and tries to merge them. Returns the number of | |
4495 | * requests that remain after merging. | |
4496 | */ | |
4497 | static int multiwrite_merge(BlockDriverState *bs, BlockRequest *reqs, | |
4498 | int num_reqs, MultiwriteCB *mcb) | |
4499 | { | |
4500 | int i, outidx; | |
4501 | ||
4502 | // Sort requests by start sector | |
4503 | qsort(reqs, num_reqs, sizeof(*reqs), &multiwrite_req_compare); | |
4504 | ||
4505 | // Check if adjacent requests touch the same clusters. If so, combine them, | |
4506 | // filling up gaps with zero sectors. | |
4507 | outidx = 0; | |
4508 | for (i = 1; i < num_reqs; i++) { | |
4509 | int merge = 0; | |
4510 | int64_t oldreq_last = reqs[outidx].sector + reqs[outidx].nb_sectors; | |
4511 | ||
b6a127a1 | 4512 | // Handle exactly sequential writes and overlapping writes. |
40b4f539 KW |
4513 | if (reqs[i].sector <= oldreq_last) { |
4514 | merge = 1; | |
4515 | } | |
4516 | ||
e2a305fb CH |
4517 | if (reqs[outidx].qiov->niov + reqs[i].qiov->niov + 1 > IOV_MAX) { |
4518 | merge = 0; | |
4519 | } | |
4520 | ||
6c5a42ac PL |
4521 | if (bs->bl.max_transfer_length && reqs[outidx].nb_sectors + |
4522 | reqs[i].nb_sectors > bs->bl.max_transfer_length) { | |
4523 | merge = 0; | |
4524 | } | |
4525 | ||
40b4f539 KW |
4526 | if (merge) { |
4527 | size_t size; | |
7267c094 | 4528 | QEMUIOVector *qiov = g_malloc0(sizeof(*qiov)); |
40b4f539 KW |
4529 | qemu_iovec_init(qiov, |
4530 | reqs[outidx].qiov->niov + reqs[i].qiov->niov + 1); | |
4531 | ||
4532 | // Add the first request to the merged one. If the requests are | |
4533 | // overlapping, drop the last sectors of the first request. | |
4534 | size = (reqs[i].sector - reqs[outidx].sector) << 9; | |
1b093c48 | 4535 | qemu_iovec_concat(qiov, reqs[outidx].qiov, 0, size); |
40b4f539 | 4536 | |
b6a127a1 PB |
4537 | // We should need to add any zeros between the two requests |
4538 | assert (reqs[i].sector <= oldreq_last); | |
40b4f539 KW |
4539 | |
4540 | // Add the second request | |
1b093c48 | 4541 | qemu_iovec_concat(qiov, reqs[i].qiov, 0, reqs[i].qiov->size); |
40b4f539 | 4542 | |
391827eb SH |
4543 | // Add tail of first request, if necessary |
4544 | if (qiov->size < reqs[outidx].qiov->size) { | |
4545 | qemu_iovec_concat(qiov, reqs[outidx].qiov, qiov->size, | |
4546 | reqs[outidx].qiov->size - qiov->size); | |
4547 | } | |
4548 | ||
cbf1dff2 | 4549 | reqs[outidx].nb_sectors = qiov->size >> 9; |
40b4f539 KW |
4550 | reqs[outidx].qiov = qiov; |
4551 | ||
4552 | mcb->callbacks[i].free_qiov = reqs[outidx].qiov; | |
4553 | } else { | |
4554 | outidx++; | |
4555 | reqs[outidx].sector = reqs[i].sector; | |
4556 | reqs[outidx].nb_sectors = reqs[i].nb_sectors; | |
4557 | reqs[outidx].qiov = reqs[i].qiov; | |
4558 | } | |
4559 | } | |
4560 | ||
4561 | return outidx + 1; | |
4562 | } | |
4563 | ||
4564 | /* | |
4565 | * Submit multiple AIO write requests at once. | |
4566 | * | |
4567 | * On success, the function returns 0 and all requests in the reqs array have | |
4568 | * been submitted. In error case this function returns -1, and any of the | |
4569 | * requests may or may not be submitted yet. In particular, this means that the | |
4570 | * callback will be called for some of the requests, for others it won't. The | |
4571 | * caller must check the error field of the BlockRequest to wait for the right | |
4572 | * callbacks (if error != 0, no callback will be called). | |
4573 | * | |
4574 | * The implementation may modify the contents of the reqs array, e.g. to merge | |
4575 | * requests. However, the fields opaque and error are left unmodified as they | |
4576 | * are used to signal failure for a single request to the caller. | |
4577 | */ | |
4578 | int bdrv_aio_multiwrite(BlockDriverState *bs, BlockRequest *reqs, int num_reqs) | |
4579 | { | |
40b4f539 KW |
4580 | MultiwriteCB *mcb; |
4581 | int i; | |
4582 | ||
301db7c2 RH |
4583 | /* don't submit writes if we don't have a medium */ |
4584 | if (bs->drv == NULL) { | |
4585 | for (i = 0; i < num_reqs; i++) { | |
4586 | reqs[i].error = -ENOMEDIUM; | |
4587 | } | |
4588 | return -1; | |
4589 | } | |
4590 | ||
40b4f539 KW |
4591 | if (num_reqs == 0) { |
4592 | return 0; | |
4593 | } | |
4594 | ||
4595 | // Create MultiwriteCB structure | |
7267c094 | 4596 | mcb = g_malloc0(sizeof(*mcb) + num_reqs * sizeof(*mcb->callbacks)); |
40b4f539 KW |
4597 | mcb->num_requests = 0; |
4598 | mcb->num_callbacks = num_reqs; | |
4599 | ||
4600 | for (i = 0; i < num_reqs; i++) { | |
4601 | mcb->callbacks[i].cb = reqs[i].cb; | |
4602 | mcb->callbacks[i].opaque = reqs[i].opaque; | |
4603 | } | |
4604 | ||
4605 | // Check for mergable requests | |
4606 | num_reqs = multiwrite_merge(bs, reqs, num_reqs, mcb); | |
4607 | ||
6d519a5f SH |
4608 | trace_bdrv_aio_multiwrite(mcb, mcb->num_callbacks, num_reqs); |
4609 | ||
df9309fb PB |
4610 | /* Run the aio requests. */ |
4611 | mcb->num_requests = num_reqs; | |
40b4f539 | 4612 | for (i = 0; i < num_reqs; i++) { |
d20d9b7c PB |
4613 | bdrv_co_aio_rw_vector(bs, reqs[i].sector, reqs[i].qiov, |
4614 | reqs[i].nb_sectors, reqs[i].flags, | |
4615 | multiwrite_cb, mcb, | |
4616 | true); | |
40b4f539 KW |
4617 | } |
4618 | ||
4619 | return 0; | |
40b4f539 KW |
4620 | } |
4621 | ||
7c84b1b8 | 4622 | void bdrv_aio_cancel(BlockAIOCB *acb) |
83f64091 | 4623 | { |
ca5fd113 FZ |
4624 | qemu_aio_ref(acb); |
4625 | bdrv_aio_cancel_async(acb); | |
4626 | while (acb->refcnt > 1) { | |
4627 | if (acb->aiocb_info->get_aio_context) { | |
4628 | aio_poll(acb->aiocb_info->get_aio_context(acb), true); | |
4629 | } else if (acb->bs) { | |
4630 | aio_poll(bdrv_get_aio_context(acb->bs), true); | |
4631 | } else { | |
4632 | abort(); | |
02c50efe | 4633 | } |
02c50efe | 4634 | } |
8007429a | 4635 | qemu_aio_unref(acb); |
02c50efe FZ |
4636 | } |
4637 | ||
4638 | /* Async version of aio cancel. The caller is not blocked if the acb implements | |
4639 | * cancel_async, otherwise we do nothing and let the request normally complete. | |
4640 | * In either case the completion callback must be called. */ | |
7c84b1b8 | 4641 | void bdrv_aio_cancel_async(BlockAIOCB *acb) |
02c50efe FZ |
4642 | { |
4643 | if (acb->aiocb_info->cancel_async) { | |
4644 | acb->aiocb_info->cancel_async(acb); | |
4645 | } | |
83f64091 FB |
4646 | } |
4647 | ||
4648 | /**************************************************************/ | |
4649 | /* async block device emulation */ | |
4650 | ||
7c84b1b8 MA |
4651 | typedef struct BlockAIOCBSync { |
4652 | BlockAIOCB common; | |
c16b5a2c CH |
4653 | QEMUBH *bh; |
4654 | int ret; | |
4655 | /* vector translation state */ | |
4656 | QEMUIOVector *qiov; | |
4657 | uint8_t *bounce; | |
4658 | int is_write; | |
7c84b1b8 | 4659 | } BlockAIOCBSync; |
c16b5a2c | 4660 | |
d7331bed | 4661 | static const AIOCBInfo bdrv_em_aiocb_info = { |
7c84b1b8 | 4662 | .aiocb_size = sizeof(BlockAIOCBSync), |
c16b5a2c CH |
4663 | }; |
4664 | ||
ce1a14dc | 4665 | static void bdrv_aio_bh_cb(void *opaque) |
83f64091 | 4666 | { |
7c84b1b8 | 4667 | BlockAIOCBSync *acb = opaque; |
f141eafe | 4668 | |
857d4f46 | 4669 | if (!acb->is_write && acb->ret >= 0) { |
03396148 | 4670 | qemu_iovec_from_buf(acb->qiov, 0, acb->bounce, acb->qiov->size); |
857d4f46 | 4671 | } |
ceb42de8 | 4672 | qemu_vfree(acb->bounce); |
ce1a14dc | 4673 | acb->common.cb(acb->common.opaque, acb->ret); |
6a7ad299 | 4674 | qemu_bh_delete(acb->bh); |
36afc451 | 4675 | acb->bh = NULL; |
8007429a | 4676 | qemu_aio_unref(acb); |
83f64091 | 4677 | } |
beac80cd | 4678 | |
7c84b1b8 MA |
4679 | static BlockAIOCB *bdrv_aio_rw_vector(BlockDriverState *bs, |
4680 | int64_t sector_num, | |
4681 | QEMUIOVector *qiov, | |
4682 | int nb_sectors, | |
097310b5 | 4683 | BlockCompletionFunc *cb, |
7c84b1b8 MA |
4684 | void *opaque, |
4685 | int is_write) | |
f141eafe | 4686 | |
83f64091 | 4687 | { |
7c84b1b8 | 4688 | BlockAIOCBSync *acb; |
ce1a14dc | 4689 | |
d7331bed | 4690 | acb = qemu_aio_get(&bdrv_em_aiocb_info, bs, cb, opaque); |
f141eafe AL |
4691 | acb->is_write = is_write; |
4692 | acb->qiov = qiov; | |
857d4f46 | 4693 | acb->bounce = qemu_try_blockalign(bs, qiov->size); |
2572b37a | 4694 | acb->bh = aio_bh_new(bdrv_get_aio_context(bs), bdrv_aio_bh_cb, acb); |
f141eafe | 4695 | |
857d4f46 KW |
4696 | if (acb->bounce == NULL) { |
4697 | acb->ret = -ENOMEM; | |
4698 | } else if (is_write) { | |
d5e6b161 | 4699 | qemu_iovec_to_buf(acb->qiov, 0, acb->bounce, qiov->size); |
1ed20acf | 4700 | acb->ret = bs->drv->bdrv_write(bs, sector_num, acb->bounce, nb_sectors); |
f141eafe | 4701 | } else { |
1ed20acf | 4702 | acb->ret = bs->drv->bdrv_read(bs, sector_num, acb->bounce, nb_sectors); |
f141eafe AL |
4703 | } |
4704 | ||
ce1a14dc | 4705 | qemu_bh_schedule(acb->bh); |
f141eafe | 4706 | |
ce1a14dc | 4707 | return &acb->common; |
beac80cd FB |
4708 | } |
4709 | ||
7c84b1b8 | 4710 | static BlockAIOCB *bdrv_aio_readv_em(BlockDriverState *bs, |
f141eafe | 4711 | int64_t sector_num, QEMUIOVector *qiov, int nb_sectors, |
097310b5 | 4712 | BlockCompletionFunc *cb, void *opaque) |
beac80cd | 4713 | { |
f141eafe AL |
4714 | return bdrv_aio_rw_vector(bs, sector_num, qiov, nb_sectors, cb, opaque, 0); |
4715 | } | |
83f64091 | 4716 | |
7c84b1b8 | 4717 | static BlockAIOCB *bdrv_aio_writev_em(BlockDriverState *bs, |
f141eafe | 4718 | int64_t sector_num, QEMUIOVector *qiov, int nb_sectors, |
097310b5 | 4719 | BlockCompletionFunc *cb, void *opaque) |
f141eafe AL |
4720 | { |
4721 | return bdrv_aio_rw_vector(bs, sector_num, qiov, nb_sectors, cb, opaque, 1); | |
beac80cd | 4722 | } |
beac80cd | 4723 | |
68485420 | 4724 | |
7c84b1b8 MA |
4725 | typedef struct BlockAIOCBCoroutine { |
4726 | BlockAIOCB common; | |
68485420 KW |
4727 | BlockRequest req; |
4728 | bool is_write; | |
d318aea9 | 4729 | bool *done; |
68485420 | 4730 | QEMUBH* bh; |
7c84b1b8 | 4731 | } BlockAIOCBCoroutine; |
68485420 | 4732 | |
d7331bed | 4733 | static const AIOCBInfo bdrv_em_co_aiocb_info = { |
7c84b1b8 | 4734 | .aiocb_size = sizeof(BlockAIOCBCoroutine), |
68485420 KW |
4735 | }; |
4736 | ||
35246a68 | 4737 | static void bdrv_co_em_bh(void *opaque) |
68485420 | 4738 | { |
7c84b1b8 | 4739 | BlockAIOCBCoroutine *acb = opaque; |
68485420 KW |
4740 | |
4741 | acb->common.cb(acb->common.opaque, acb->req.error); | |
d318aea9 | 4742 | |
68485420 | 4743 | qemu_bh_delete(acb->bh); |
8007429a | 4744 | qemu_aio_unref(acb); |
68485420 KW |
4745 | } |
4746 | ||
b2a61371 SH |
4747 | /* Invoke bdrv_co_do_readv/bdrv_co_do_writev */ |
4748 | static void coroutine_fn bdrv_co_do_rw(void *opaque) | |
4749 | { | |
7c84b1b8 | 4750 | BlockAIOCBCoroutine *acb = opaque; |
b2a61371 SH |
4751 | BlockDriverState *bs = acb->common.bs; |
4752 | ||
4753 | if (!acb->is_write) { | |
4754 | acb->req.error = bdrv_co_do_readv(bs, acb->req.sector, | |
d20d9b7c | 4755 | acb->req.nb_sectors, acb->req.qiov, acb->req.flags); |
b2a61371 SH |
4756 | } else { |
4757 | acb->req.error = bdrv_co_do_writev(bs, acb->req.sector, | |
d20d9b7c | 4758 | acb->req.nb_sectors, acb->req.qiov, acb->req.flags); |
b2a61371 SH |
4759 | } |
4760 | ||
2572b37a | 4761 | acb->bh = aio_bh_new(bdrv_get_aio_context(bs), bdrv_co_em_bh, acb); |
b2a61371 SH |
4762 | qemu_bh_schedule(acb->bh); |
4763 | } | |
4764 | ||
7c84b1b8 MA |
4765 | static BlockAIOCB *bdrv_co_aio_rw_vector(BlockDriverState *bs, |
4766 | int64_t sector_num, | |
4767 | QEMUIOVector *qiov, | |
4768 | int nb_sectors, | |
4769 | BdrvRequestFlags flags, | |
097310b5 | 4770 | BlockCompletionFunc *cb, |
7c84b1b8 MA |
4771 | void *opaque, |
4772 | bool is_write) | |
68485420 KW |
4773 | { |
4774 | Coroutine *co; | |
7c84b1b8 | 4775 | BlockAIOCBCoroutine *acb; |
68485420 | 4776 | |
d7331bed | 4777 | acb = qemu_aio_get(&bdrv_em_co_aiocb_info, bs, cb, opaque); |
68485420 KW |
4778 | acb->req.sector = sector_num; |
4779 | acb->req.nb_sectors = nb_sectors; | |
4780 | acb->req.qiov = qiov; | |
d20d9b7c | 4781 | acb->req.flags = flags; |
68485420 KW |
4782 | acb->is_write = is_write; |
4783 | ||
8c5873d6 | 4784 | co = qemu_coroutine_create(bdrv_co_do_rw); |
68485420 KW |
4785 | qemu_coroutine_enter(co, acb); |
4786 | ||
4787 | return &acb->common; | |
4788 | } | |
4789 | ||
07f07615 | 4790 | static void coroutine_fn bdrv_aio_flush_co_entry(void *opaque) |
b2e12bc6 | 4791 | { |
7c84b1b8 | 4792 | BlockAIOCBCoroutine *acb = opaque; |
07f07615 | 4793 | BlockDriverState *bs = acb->common.bs; |
b2e12bc6 | 4794 | |
07f07615 | 4795 | acb->req.error = bdrv_co_flush(bs); |
2572b37a | 4796 | acb->bh = aio_bh_new(bdrv_get_aio_context(bs), bdrv_co_em_bh, acb); |
b2e12bc6 | 4797 | qemu_bh_schedule(acb->bh); |
b2e12bc6 CH |
4798 | } |
4799 | ||
7c84b1b8 | 4800 | BlockAIOCB *bdrv_aio_flush(BlockDriverState *bs, |
097310b5 | 4801 | BlockCompletionFunc *cb, void *opaque) |
016f5cf6 | 4802 | { |
07f07615 | 4803 | trace_bdrv_aio_flush(bs, opaque); |
016f5cf6 | 4804 | |
07f07615 | 4805 | Coroutine *co; |
7c84b1b8 | 4806 | BlockAIOCBCoroutine *acb; |
016f5cf6 | 4807 | |
d7331bed | 4808 | acb = qemu_aio_get(&bdrv_em_co_aiocb_info, bs, cb, opaque); |
d318aea9 | 4809 | |
07f07615 PB |
4810 | co = qemu_coroutine_create(bdrv_aio_flush_co_entry); |
4811 | qemu_coroutine_enter(co, acb); | |
016f5cf6 | 4812 | |
016f5cf6 AG |
4813 | return &acb->common; |
4814 | } | |
4815 | ||
4265d620 PB |
4816 | static void coroutine_fn bdrv_aio_discard_co_entry(void *opaque) |
4817 | { | |
7c84b1b8 | 4818 | BlockAIOCBCoroutine *acb = opaque; |
4265d620 PB |
4819 | BlockDriverState *bs = acb->common.bs; |
4820 | ||
4821 | acb->req.error = bdrv_co_discard(bs, acb->req.sector, acb->req.nb_sectors); | |
2572b37a | 4822 | acb->bh = aio_bh_new(bdrv_get_aio_context(bs), bdrv_co_em_bh, acb); |
4265d620 PB |
4823 | qemu_bh_schedule(acb->bh); |
4824 | } | |
4825 | ||
7c84b1b8 | 4826 | BlockAIOCB *bdrv_aio_discard(BlockDriverState *bs, |
4265d620 | 4827 | int64_t sector_num, int nb_sectors, |
097310b5 | 4828 | BlockCompletionFunc *cb, void *opaque) |
4265d620 PB |
4829 | { |
4830 | Coroutine *co; | |
7c84b1b8 | 4831 | BlockAIOCBCoroutine *acb; |
4265d620 PB |
4832 | |
4833 | trace_bdrv_aio_discard(bs, sector_num, nb_sectors, opaque); | |
4834 | ||
d7331bed | 4835 | acb = qemu_aio_get(&bdrv_em_co_aiocb_info, bs, cb, opaque); |
4265d620 PB |
4836 | acb->req.sector = sector_num; |
4837 | acb->req.nb_sectors = nb_sectors; | |
4838 | co = qemu_coroutine_create(bdrv_aio_discard_co_entry); | |
4839 | qemu_coroutine_enter(co, acb); | |
4840 | ||
4841 | return &acb->common; | |
4842 | } | |
4843 | ||
ea2384d3 FB |
4844 | void bdrv_init(void) |
4845 | { | |
5efa9d5a | 4846 | module_call_init(MODULE_INIT_BLOCK); |
ea2384d3 | 4847 | } |
ce1a14dc | 4848 | |
eb852011 MA |
4849 | void bdrv_init_with_whitelist(void) |
4850 | { | |
4851 | use_bdrv_whitelist = 1; | |
4852 | bdrv_init(); | |
4853 | } | |
4854 | ||
d7331bed | 4855 | void *qemu_aio_get(const AIOCBInfo *aiocb_info, BlockDriverState *bs, |
097310b5 | 4856 | BlockCompletionFunc *cb, void *opaque) |
ce1a14dc | 4857 | { |
7c84b1b8 | 4858 | BlockAIOCB *acb; |
ce1a14dc | 4859 | |
d7331bed SH |
4860 | acb = g_slice_alloc(aiocb_info->aiocb_size); |
4861 | acb->aiocb_info = aiocb_info; | |
ce1a14dc PB |
4862 | acb->bs = bs; |
4863 | acb->cb = cb; | |
4864 | acb->opaque = opaque; | |
f197fe2b | 4865 | acb->refcnt = 1; |
ce1a14dc PB |
4866 | return acb; |
4867 | } | |
4868 | ||
f197fe2b FZ |
4869 | void qemu_aio_ref(void *p) |
4870 | { | |
7c84b1b8 | 4871 | BlockAIOCB *acb = p; |
f197fe2b FZ |
4872 | acb->refcnt++; |
4873 | } | |
4874 | ||
8007429a | 4875 | void qemu_aio_unref(void *p) |
ce1a14dc | 4876 | { |
7c84b1b8 | 4877 | BlockAIOCB *acb = p; |
f197fe2b FZ |
4878 | assert(acb->refcnt > 0); |
4879 | if (--acb->refcnt == 0) { | |
4880 | g_slice_free1(acb->aiocb_info->aiocb_size, acb); | |
4881 | } | |
ce1a14dc | 4882 | } |
19cb3738 | 4883 | |
f9f05dc5 KW |
4884 | /**************************************************************/ |
4885 | /* Coroutine block device emulation */ | |
4886 | ||
4887 | typedef struct CoroutineIOCompletion { | |
4888 | Coroutine *coroutine; | |
4889 | int ret; | |
4890 | } CoroutineIOCompletion; | |
4891 | ||
4892 | static void bdrv_co_io_em_complete(void *opaque, int ret) | |
4893 | { | |
4894 | CoroutineIOCompletion *co = opaque; | |
4895 | ||
4896 | co->ret = ret; | |
4897 | qemu_coroutine_enter(co->coroutine, NULL); | |
4898 | } | |
4899 | ||
4900 | static int coroutine_fn bdrv_co_io_em(BlockDriverState *bs, int64_t sector_num, | |
4901 | int nb_sectors, QEMUIOVector *iov, | |
4902 | bool is_write) | |
4903 | { | |
4904 | CoroutineIOCompletion co = { | |
4905 | .coroutine = qemu_coroutine_self(), | |
4906 | }; | |
7c84b1b8 | 4907 | BlockAIOCB *acb; |
f9f05dc5 KW |
4908 | |
4909 | if (is_write) { | |
a652d160 SH |
4910 | acb = bs->drv->bdrv_aio_writev(bs, sector_num, iov, nb_sectors, |
4911 | bdrv_co_io_em_complete, &co); | |
f9f05dc5 | 4912 | } else { |
a652d160 SH |
4913 | acb = bs->drv->bdrv_aio_readv(bs, sector_num, iov, nb_sectors, |
4914 | bdrv_co_io_em_complete, &co); | |
f9f05dc5 KW |
4915 | } |
4916 | ||
59370aaa | 4917 | trace_bdrv_co_io_em(bs, sector_num, nb_sectors, is_write, acb); |
f9f05dc5 KW |
4918 | if (!acb) { |
4919 | return -EIO; | |
4920 | } | |
4921 | qemu_coroutine_yield(); | |
4922 | ||
4923 | return co.ret; | |
4924 | } | |
4925 | ||
4926 | static int coroutine_fn bdrv_co_readv_em(BlockDriverState *bs, | |
4927 | int64_t sector_num, int nb_sectors, | |
4928 | QEMUIOVector *iov) | |
4929 | { | |
4930 | return bdrv_co_io_em(bs, sector_num, nb_sectors, iov, false); | |
4931 | } | |
4932 | ||
4933 | static int coroutine_fn bdrv_co_writev_em(BlockDriverState *bs, | |
4934 | int64_t sector_num, int nb_sectors, | |
4935 | QEMUIOVector *iov) | |
4936 | { | |
4937 | return bdrv_co_io_em(bs, sector_num, nb_sectors, iov, true); | |
4938 | } | |
4939 | ||
07f07615 | 4940 | static void coroutine_fn bdrv_flush_co_entry(void *opaque) |
e7a8a783 | 4941 | { |
07f07615 PB |
4942 | RwCo *rwco = opaque; |
4943 | ||
4944 | rwco->ret = bdrv_co_flush(rwco->bs); | |
4945 | } | |
4946 | ||
4947 | int coroutine_fn bdrv_co_flush(BlockDriverState *bs) | |
4948 | { | |
eb489bb1 KW |
4949 | int ret; |
4950 | ||
29cdb251 | 4951 | if (!bs || !bdrv_is_inserted(bs) || bdrv_is_read_only(bs)) { |
07f07615 | 4952 | return 0; |
eb489bb1 KW |
4953 | } |
4954 | ||
ca716364 | 4955 | /* Write back cached data to the OS even with cache=unsafe */ |
bf736fe3 | 4956 | BLKDBG_EVENT(bs->file, BLKDBG_FLUSH_TO_OS); |
eb489bb1 KW |
4957 | if (bs->drv->bdrv_co_flush_to_os) { |
4958 | ret = bs->drv->bdrv_co_flush_to_os(bs); | |
4959 | if (ret < 0) { | |
4960 | return ret; | |
4961 | } | |
4962 | } | |
4963 | ||
ca716364 KW |
4964 | /* But don't actually force it to the disk with cache=unsafe */ |
4965 | if (bs->open_flags & BDRV_O_NO_FLUSH) { | |
d4c82329 | 4966 | goto flush_parent; |
ca716364 KW |
4967 | } |
4968 | ||
bf736fe3 | 4969 | BLKDBG_EVENT(bs->file, BLKDBG_FLUSH_TO_DISK); |
eb489bb1 | 4970 | if (bs->drv->bdrv_co_flush_to_disk) { |
29cdb251 | 4971 | ret = bs->drv->bdrv_co_flush_to_disk(bs); |
07f07615 | 4972 | } else if (bs->drv->bdrv_aio_flush) { |
7c84b1b8 | 4973 | BlockAIOCB *acb; |
07f07615 PB |
4974 | CoroutineIOCompletion co = { |
4975 | .coroutine = qemu_coroutine_self(), | |
4976 | }; | |
4977 | ||
4978 | acb = bs->drv->bdrv_aio_flush(bs, bdrv_co_io_em_complete, &co); | |
4979 | if (acb == NULL) { | |
29cdb251 | 4980 | ret = -EIO; |
07f07615 PB |
4981 | } else { |
4982 | qemu_coroutine_yield(); | |
29cdb251 | 4983 | ret = co.ret; |
07f07615 | 4984 | } |
07f07615 PB |
4985 | } else { |
4986 | /* | |
4987 | * Some block drivers always operate in either writethrough or unsafe | |
4988 | * mode and don't support bdrv_flush therefore. Usually qemu doesn't | |
4989 | * know how the server works (because the behaviour is hardcoded or | |
4990 | * depends on server-side configuration), so we can't ensure that | |
4991 | * everything is safe on disk. Returning an error doesn't work because | |
4992 | * that would break guests even if the server operates in writethrough | |
4993 | * mode. | |
4994 | * | |
4995 | * Let's hope the user knows what he's doing. | |
4996 | */ | |
29cdb251 | 4997 | ret = 0; |
07f07615 | 4998 | } |
29cdb251 PB |
4999 | if (ret < 0) { |
5000 | return ret; | |
5001 | } | |
5002 | ||
5003 | /* Now flush the underlying protocol. It will also have BDRV_O_NO_FLUSH | |
5004 | * in the case of cache=unsafe, so there are no useless flushes. | |
5005 | */ | |
d4c82329 | 5006 | flush_parent: |
29cdb251 | 5007 | return bdrv_co_flush(bs->file); |
07f07615 PB |
5008 | } |
5009 | ||
5a8a30db | 5010 | void bdrv_invalidate_cache(BlockDriverState *bs, Error **errp) |
0f15423c | 5011 | { |
5a8a30db KW |
5012 | Error *local_err = NULL; |
5013 | int ret; | |
5014 | ||
3456a8d1 KW |
5015 | if (!bs->drv) { |
5016 | return; | |
5017 | } | |
5018 | ||
7ea2d269 AK |
5019 | if (!(bs->open_flags & BDRV_O_INCOMING)) { |
5020 | return; | |
5021 | } | |
5022 | bs->open_flags &= ~BDRV_O_INCOMING; | |
5023 | ||
3456a8d1 | 5024 | if (bs->drv->bdrv_invalidate_cache) { |
5a8a30db | 5025 | bs->drv->bdrv_invalidate_cache(bs, &local_err); |
3456a8d1 | 5026 | } else if (bs->file) { |
5a8a30db KW |
5027 | bdrv_invalidate_cache(bs->file, &local_err); |
5028 | } | |
5029 | if (local_err) { | |
5030 | error_propagate(errp, local_err); | |
5031 | return; | |
0f15423c | 5032 | } |
3456a8d1 | 5033 | |
5a8a30db KW |
5034 | ret = refresh_total_sectors(bs, bs->total_sectors); |
5035 | if (ret < 0) { | |
5036 | error_setg_errno(errp, -ret, "Could not refresh total sector count"); | |
5037 | return; | |
5038 | } | |
0f15423c AL |
5039 | } |
5040 | ||
5a8a30db | 5041 | void bdrv_invalidate_cache_all(Error **errp) |
0f15423c AL |
5042 | { |
5043 | BlockDriverState *bs; | |
5a8a30db | 5044 | Error *local_err = NULL; |
0f15423c | 5045 | |
dc364f4c | 5046 | QTAILQ_FOREACH(bs, &bdrv_states, device_list) { |
ed78cda3 SH |
5047 | AioContext *aio_context = bdrv_get_aio_context(bs); |
5048 | ||
5049 | aio_context_acquire(aio_context); | |
5a8a30db | 5050 | bdrv_invalidate_cache(bs, &local_err); |
ed78cda3 | 5051 | aio_context_release(aio_context); |
5a8a30db KW |
5052 | if (local_err) { |
5053 | error_propagate(errp, local_err); | |
5054 | return; | |
5055 | } | |
0f15423c AL |
5056 | } |
5057 | } | |
5058 | ||
07f07615 PB |
5059 | int bdrv_flush(BlockDriverState *bs) |
5060 | { | |
5061 | Coroutine *co; | |
5062 | RwCo rwco = { | |
5063 | .bs = bs, | |
5064 | .ret = NOT_DONE, | |
e7a8a783 | 5065 | }; |
e7a8a783 | 5066 | |
07f07615 PB |
5067 | if (qemu_in_coroutine()) { |
5068 | /* Fast-path if already in coroutine context */ | |
5069 | bdrv_flush_co_entry(&rwco); | |
5070 | } else { | |
2572b37a SH |
5071 | AioContext *aio_context = bdrv_get_aio_context(bs); |
5072 | ||
07f07615 PB |
5073 | co = qemu_coroutine_create(bdrv_flush_co_entry); |
5074 | qemu_coroutine_enter(co, &rwco); | |
5075 | while (rwco.ret == NOT_DONE) { | |
2572b37a | 5076 | aio_poll(aio_context, true); |
07f07615 | 5077 | } |
e7a8a783 | 5078 | } |
07f07615 PB |
5079 | |
5080 | return rwco.ret; | |
e7a8a783 KW |
5081 | } |
5082 | ||
775aa8b6 KW |
5083 | typedef struct DiscardCo { |
5084 | BlockDriverState *bs; | |
5085 | int64_t sector_num; | |
5086 | int nb_sectors; | |
5087 | int ret; | |
5088 | } DiscardCo; | |
4265d620 PB |
5089 | static void coroutine_fn bdrv_discard_co_entry(void *opaque) |
5090 | { | |
775aa8b6 | 5091 | DiscardCo *rwco = opaque; |
4265d620 PB |
5092 | |
5093 | rwco->ret = bdrv_co_discard(rwco->bs, rwco->sector_num, rwco->nb_sectors); | |
5094 | } | |
5095 | ||
6f14da52 PL |
5096 | /* if no limit is specified in the BlockLimits use a default |
5097 | * of 32768 512-byte sectors (16 MiB) per request. | |
5098 | */ | |
5099 | #define MAX_DISCARD_DEFAULT 32768 | |
5100 | ||
4265d620 PB |
5101 | int coroutine_fn bdrv_co_discard(BlockDriverState *bs, int64_t sector_num, |
5102 | int nb_sectors) | |
5103 | { | |
d51e9fe5 PB |
5104 | int max_discard; |
5105 | ||
4265d620 PB |
5106 | if (!bs->drv) { |
5107 | return -ENOMEDIUM; | |
5108 | } else if (bdrv_check_request(bs, sector_num, nb_sectors)) { | |
5109 | return -EIO; | |
5110 | } else if (bs->read_only) { | |
5111 | return -EROFS; | |
df702c9b PB |
5112 | } |
5113 | ||
e4654d2d | 5114 | bdrv_reset_dirty(bs, sector_num, nb_sectors); |
df702c9b | 5115 | |
9e8f1835 PB |
5116 | /* Do nothing if disabled. */ |
5117 | if (!(bs->open_flags & BDRV_O_UNMAP)) { | |
5118 | return 0; | |
5119 | } | |
5120 | ||
d51e9fe5 PB |
5121 | if (!bs->drv->bdrv_co_discard && !bs->drv->bdrv_aio_discard) { |
5122 | return 0; | |
5123 | } | |
6f14da52 | 5124 | |
d51e9fe5 PB |
5125 | max_discard = bs->bl.max_discard ? bs->bl.max_discard : MAX_DISCARD_DEFAULT; |
5126 | while (nb_sectors > 0) { | |
5127 | int ret; | |
5128 | int num = nb_sectors; | |
6f14da52 | 5129 | |
d51e9fe5 PB |
5130 | /* align request */ |
5131 | if (bs->bl.discard_alignment && | |
5132 | num >= bs->bl.discard_alignment && | |
5133 | sector_num % bs->bl.discard_alignment) { | |
5134 | if (num > bs->bl.discard_alignment) { | |
5135 | num = bs->bl.discard_alignment; | |
6f14da52 | 5136 | } |
d51e9fe5 PB |
5137 | num -= sector_num % bs->bl.discard_alignment; |
5138 | } | |
6f14da52 | 5139 | |
d51e9fe5 PB |
5140 | /* limit request size */ |
5141 | if (num > max_discard) { | |
5142 | num = max_discard; | |
5143 | } | |
6f14da52 | 5144 | |
d51e9fe5 | 5145 | if (bs->drv->bdrv_co_discard) { |
6f14da52 | 5146 | ret = bs->drv->bdrv_co_discard(bs, sector_num, num); |
d51e9fe5 | 5147 | } else { |
7c84b1b8 | 5148 | BlockAIOCB *acb; |
d51e9fe5 PB |
5149 | CoroutineIOCompletion co = { |
5150 | .coroutine = qemu_coroutine_self(), | |
5151 | }; | |
5152 | ||
5153 | acb = bs->drv->bdrv_aio_discard(bs, sector_num, nb_sectors, | |
5154 | bdrv_co_io_em_complete, &co); | |
5155 | if (acb == NULL) { | |
5156 | return -EIO; | |
5157 | } else { | |
5158 | qemu_coroutine_yield(); | |
5159 | ret = co.ret; | |
6f14da52 | 5160 | } |
6f14da52 | 5161 | } |
7ce21016 | 5162 | if (ret && ret != -ENOTSUP) { |
d51e9fe5 | 5163 | return ret; |
4265d620 | 5164 | } |
d51e9fe5 PB |
5165 | |
5166 | sector_num += num; | |
5167 | nb_sectors -= num; | |
4265d620 | 5168 | } |
d51e9fe5 | 5169 | return 0; |
4265d620 PB |
5170 | } |
5171 | ||
5172 | int bdrv_discard(BlockDriverState *bs, int64_t sector_num, int nb_sectors) | |
5173 | { | |
5174 | Coroutine *co; | |
775aa8b6 | 5175 | DiscardCo rwco = { |
4265d620 PB |
5176 | .bs = bs, |
5177 | .sector_num = sector_num, | |
5178 | .nb_sectors = nb_sectors, | |
5179 | .ret = NOT_DONE, | |
5180 | }; | |
5181 | ||
5182 | if (qemu_in_coroutine()) { | |
5183 | /* Fast-path if already in coroutine context */ | |
5184 | bdrv_discard_co_entry(&rwco); | |
5185 | } else { | |
2572b37a SH |
5186 | AioContext *aio_context = bdrv_get_aio_context(bs); |
5187 | ||
4265d620 PB |
5188 | co = qemu_coroutine_create(bdrv_discard_co_entry); |
5189 | qemu_coroutine_enter(co, &rwco); | |
5190 | while (rwco.ret == NOT_DONE) { | |
2572b37a | 5191 | aio_poll(aio_context, true); |
4265d620 PB |
5192 | } |
5193 | } | |
5194 | ||
5195 | return rwco.ret; | |
5196 | } | |
5197 | ||
19cb3738 FB |
5198 | /**************************************************************/ |
5199 | /* removable device support */ | |
5200 | ||
5201 | /** | |
5202 | * Return TRUE if the media is present | |
5203 | */ | |
5204 | int bdrv_is_inserted(BlockDriverState *bs) | |
5205 | { | |
5206 | BlockDriver *drv = bs->drv; | |
a1aff5bf | 5207 | |
19cb3738 FB |
5208 | if (!drv) |
5209 | return 0; | |
5210 | if (!drv->bdrv_is_inserted) | |
a1aff5bf MA |
5211 | return 1; |
5212 | return drv->bdrv_is_inserted(bs); | |
19cb3738 FB |
5213 | } |
5214 | ||
5215 | /** | |
8e49ca46 MA |
5216 | * Return whether the media changed since the last call to this |
5217 | * function, or -ENOTSUP if we don't know. Most drivers don't know. | |
19cb3738 FB |
5218 | */ |
5219 | int bdrv_media_changed(BlockDriverState *bs) | |
5220 | { | |
5221 | BlockDriver *drv = bs->drv; | |
19cb3738 | 5222 | |
8e49ca46 MA |
5223 | if (drv && drv->bdrv_media_changed) { |
5224 | return drv->bdrv_media_changed(bs); | |
5225 | } | |
5226 | return -ENOTSUP; | |
19cb3738 FB |
5227 | } |
5228 | ||
5229 | /** | |
5230 | * If eject_flag is TRUE, eject the media. Otherwise, close the tray | |
5231 | */ | |
f36f3949 | 5232 | void bdrv_eject(BlockDriverState *bs, bool eject_flag) |
19cb3738 FB |
5233 | { |
5234 | BlockDriver *drv = bs->drv; | |
bfb197e0 | 5235 | const char *device_name; |
19cb3738 | 5236 | |
822e1cd1 MA |
5237 | if (drv && drv->bdrv_eject) { |
5238 | drv->bdrv_eject(bs, eject_flag); | |
19cb3738 | 5239 | } |
6f382ed2 | 5240 | |
bfb197e0 MA |
5241 | device_name = bdrv_get_device_name(bs); |
5242 | if (device_name[0] != '\0') { | |
5243 | qapi_event_send_device_tray_moved(device_name, | |
a5ee7bd4 | 5244 | eject_flag, &error_abort); |
6f382ed2 | 5245 | } |
19cb3738 FB |
5246 | } |
5247 | ||
19cb3738 FB |
5248 | /** |
5249 | * Lock or unlock the media (if it is locked, the user won't be able | |
5250 | * to eject it manually). | |
5251 | */ | |
025e849a | 5252 | void bdrv_lock_medium(BlockDriverState *bs, bool locked) |
19cb3738 FB |
5253 | { |
5254 | BlockDriver *drv = bs->drv; | |
5255 | ||
025e849a | 5256 | trace_bdrv_lock_medium(bs, locked); |
b8c6d095 | 5257 | |
025e849a MA |
5258 | if (drv && drv->bdrv_lock_medium) { |
5259 | drv->bdrv_lock_medium(bs, locked); | |
19cb3738 FB |
5260 | } |
5261 | } | |
985a03b0 TS |
5262 | |
5263 | /* needed for generic scsi interface */ | |
5264 | ||
5265 | int bdrv_ioctl(BlockDriverState *bs, unsigned long int req, void *buf) | |
5266 | { | |
5267 | BlockDriver *drv = bs->drv; | |
5268 | ||
5269 | if (drv && drv->bdrv_ioctl) | |
5270 | return drv->bdrv_ioctl(bs, req, buf); | |
5271 | return -ENOTSUP; | |
5272 | } | |
7d780669 | 5273 | |
7c84b1b8 | 5274 | BlockAIOCB *bdrv_aio_ioctl(BlockDriverState *bs, |
221f715d | 5275 | unsigned long int req, void *buf, |
097310b5 | 5276 | BlockCompletionFunc *cb, void *opaque) |
7d780669 | 5277 | { |
221f715d | 5278 | BlockDriver *drv = bs->drv; |
7d780669 | 5279 | |
221f715d AL |
5280 | if (drv && drv->bdrv_aio_ioctl) |
5281 | return drv->bdrv_aio_ioctl(bs, req, buf, cb, opaque); | |
5282 | return NULL; | |
7d780669 | 5283 | } |
e268ca52 | 5284 | |
1b7fd729 | 5285 | void bdrv_set_guest_block_size(BlockDriverState *bs, int align) |
7b6f9300 | 5286 | { |
1b7fd729 | 5287 | bs->guest_block_size = align; |
7b6f9300 | 5288 | } |
7cd1e32a | 5289 | |
e268ca52 AL |
5290 | void *qemu_blockalign(BlockDriverState *bs, size_t size) |
5291 | { | |
339064d5 | 5292 | return qemu_memalign(bdrv_opt_mem_align(bs), size); |
e268ca52 | 5293 | } |
7cd1e32a | 5294 | |
9ebd8448 HR |
5295 | void *qemu_blockalign0(BlockDriverState *bs, size_t size) |
5296 | { | |
5297 | return memset(qemu_blockalign(bs, size), 0, size); | |
5298 | } | |
5299 | ||
7d2a35cc KW |
5300 | void *qemu_try_blockalign(BlockDriverState *bs, size_t size) |
5301 | { | |
5302 | size_t align = bdrv_opt_mem_align(bs); | |
5303 | ||
5304 | /* Ensure that NULL is never returned on success */ | |
5305 | assert(align > 0); | |
5306 | if (size == 0) { | |
5307 | size = align; | |
5308 | } | |
5309 | ||
5310 | return qemu_try_memalign(align, size); | |
5311 | } | |
5312 | ||
9ebd8448 HR |
5313 | void *qemu_try_blockalign0(BlockDriverState *bs, size_t size) |
5314 | { | |
5315 | void *mem = qemu_try_blockalign(bs, size); | |
5316 | ||
5317 | if (mem) { | |
5318 | memset(mem, 0, size); | |
5319 | } | |
5320 | ||
5321 | return mem; | |
5322 | } | |
5323 | ||
c53b1c51 SH |
5324 | /* |
5325 | * Check if all memory in this vector is sector aligned. | |
5326 | */ | |
5327 | bool bdrv_qiov_is_aligned(BlockDriverState *bs, QEMUIOVector *qiov) | |
5328 | { | |
5329 | int i; | |
339064d5 | 5330 | size_t alignment = bdrv_opt_mem_align(bs); |
c53b1c51 SH |
5331 | |
5332 | for (i = 0; i < qiov->niov; i++) { | |
339064d5 | 5333 | if ((uintptr_t) qiov->iov[i].iov_base % alignment) { |
c53b1c51 | 5334 | return false; |
1ff735bd | 5335 | } |
339064d5 | 5336 | if (qiov->iov[i].iov_len % alignment) { |
1ff735bd | 5337 | return false; |
c53b1c51 SH |
5338 | } |
5339 | } | |
5340 | ||
5341 | return true; | |
5342 | } | |
5343 | ||
b8afb520 FZ |
5344 | BdrvDirtyBitmap *bdrv_create_dirty_bitmap(BlockDriverState *bs, int granularity, |
5345 | Error **errp) | |
7cd1e32a LS |
5346 | { |
5347 | int64_t bitmap_size; | |
e4654d2d | 5348 | BdrvDirtyBitmap *bitmap; |
a55eb92c | 5349 | |
50717e94 PB |
5350 | assert((granularity & (granularity - 1)) == 0); |
5351 | ||
e4654d2d FZ |
5352 | granularity >>= BDRV_SECTOR_BITS; |
5353 | assert(granularity); | |
57322b78 | 5354 | bitmap_size = bdrv_nb_sectors(bs); |
b8afb520 FZ |
5355 | if (bitmap_size < 0) { |
5356 | error_setg_errno(errp, -bitmap_size, "could not get length of device"); | |
5357 | errno = -bitmap_size; | |
5358 | return NULL; | |
5359 | } | |
5839e53b | 5360 | bitmap = g_new0(BdrvDirtyBitmap, 1); |
e4654d2d FZ |
5361 | bitmap->bitmap = hbitmap_alloc(bitmap_size, ffs(granularity) - 1); |
5362 | QLIST_INSERT_HEAD(&bs->dirty_bitmaps, bitmap, list); | |
5363 | return bitmap; | |
5364 | } | |
5365 | ||
5366 | void bdrv_release_dirty_bitmap(BlockDriverState *bs, BdrvDirtyBitmap *bitmap) | |
5367 | { | |
5368 | BdrvDirtyBitmap *bm, *next; | |
5369 | QLIST_FOREACH_SAFE(bm, &bs->dirty_bitmaps, list, next) { | |
5370 | if (bm == bitmap) { | |
5371 | QLIST_REMOVE(bitmap, list); | |
5372 | hbitmap_free(bitmap->bitmap); | |
5373 | g_free(bitmap); | |
5374 | return; | |
a55eb92c | 5375 | } |
7cd1e32a LS |
5376 | } |
5377 | } | |
5378 | ||
21b56835 FZ |
5379 | BlockDirtyInfoList *bdrv_query_dirty_bitmaps(BlockDriverState *bs) |
5380 | { | |
5381 | BdrvDirtyBitmap *bm; | |
5382 | BlockDirtyInfoList *list = NULL; | |
5383 | BlockDirtyInfoList **plist = &list; | |
5384 | ||
5385 | QLIST_FOREACH(bm, &bs->dirty_bitmaps, list) { | |
5839e53b MA |
5386 | BlockDirtyInfo *info = g_new0(BlockDirtyInfo, 1); |
5387 | BlockDirtyInfoList *entry = g_new0(BlockDirtyInfoList, 1); | |
21b56835 FZ |
5388 | info->count = bdrv_get_dirty_count(bs, bm); |
5389 | info->granularity = | |
5390 | ((int64_t) BDRV_SECTOR_SIZE << hbitmap_granularity(bm->bitmap)); | |
5391 | entry->value = info; | |
5392 | *plist = entry; | |
5393 | plist = &entry->next; | |
5394 | } | |
5395 | ||
5396 | return list; | |
5397 | } | |
5398 | ||
e4654d2d | 5399 | int bdrv_get_dirty(BlockDriverState *bs, BdrvDirtyBitmap *bitmap, int64_t sector) |
7cd1e32a | 5400 | { |
e4654d2d FZ |
5401 | if (bitmap) { |
5402 | return hbitmap_get(bitmap->bitmap, sector); | |
7cd1e32a LS |
5403 | } else { |
5404 | return 0; | |
5405 | } | |
5406 | } | |
5407 | ||
e4654d2d FZ |
5408 | void bdrv_dirty_iter_init(BlockDriverState *bs, |
5409 | BdrvDirtyBitmap *bitmap, HBitmapIter *hbi) | |
1755da16 | 5410 | { |
e4654d2d | 5411 | hbitmap_iter_init(hbi, bitmap->bitmap, 0); |
1755da16 PB |
5412 | } |
5413 | ||
5414 | void bdrv_set_dirty(BlockDriverState *bs, int64_t cur_sector, | |
5415 | int nr_sectors) | |
5416 | { | |
e4654d2d FZ |
5417 | BdrvDirtyBitmap *bitmap; |
5418 | QLIST_FOREACH(bitmap, &bs->dirty_bitmaps, list) { | |
5419 | hbitmap_set(bitmap->bitmap, cur_sector, nr_sectors); | |
5420 | } | |
1755da16 PB |
5421 | } |
5422 | ||
e4654d2d | 5423 | void bdrv_reset_dirty(BlockDriverState *bs, int64_t cur_sector, int nr_sectors) |
7cd1e32a | 5424 | { |
e4654d2d FZ |
5425 | BdrvDirtyBitmap *bitmap; |
5426 | QLIST_FOREACH(bitmap, &bs->dirty_bitmaps, list) { | |
5427 | hbitmap_reset(bitmap->bitmap, cur_sector, nr_sectors); | |
5428 | } | |
7cd1e32a | 5429 | } |
aaa0eb75 | 5430 | |
e4654d2d | 5431 | int64_t bdrv_get_dirty_count(BlockDriverState *bs, BdrvDirtyBitmap *bitmap) |
aaa0eb75 | 5432 | { |
e4654d2d | 5433 | return hbitmap_count(bitmap->bitmap); |
aaa0eb75 | 5434 | } |
f88e1a42 | 5435 | |
9fcb0251 FZ |
5436 | /* Get a reference to bs */ |
5437 | void bdrv_ref(BlockDriverState *bs) | |
5438 | { | |
5439 | bs->refcnt++; | |
5440 | } | |
5441 | ||
5442 | /* Release a previously grabbed reference to bs. | |
5443 | * If after releasing, reference count is zero, the BlockDriverState is | |
5444 | * deleted. */ | |
5445 | void bdrv_unref(BlockDriverState *bs) | |
5446 | { | |
9a4d5ca6 JC |
5447 | if (!bs) { |
5448 | return; | |
5449 | } | |
9fcb0251 FZ |
5450 | assert(bs->refcnt > 0); |
5451 | if (--bs->refcnt == 0) { | |
5452 | bdrv_delete(bs); | |
5453 | } | |
5454 | } | |
5455 | ||
fbe40ff7 FZ |
5456 | struct BdrvOpBlocker { |
5457 | Error *reason; | |
5458 | QLIST_ENTRY(BdrvOpBlocker) list; | |
5459 | }; | |
5460 | ||
5461 | bool bdrv_op_is_blocked(BlockDriverState *bs, BlockOpType op, Error **errp) | |
5462 | { | |
5463 | BdrvOpBlocker *blocker; | |
5464 | assert((int) op >= 0 && op < BLOCK_OP_TYPE_MAX); | |
5465 | if (!QLIST_EMPTY(&bs->op_blockers[op])) { | |
5466 | blocker = QLIST_FIRST(&bs->op_blockers[op]); | |
5467 | if (errp) { | |
5468 | error_setg(errp, "Device '%s' is busy: %s", | |
bfb197e0 MA |
5469 | bdrv_get_device_name(bs), |
5470 | error_get_pretty(blocker->reason)); | |
fbe40ff7 FZ |
5471 | } |
5472 | return true; | |
5473 | } | |
5474 | return false; | |
5475 | } | |
5476 | ||
5477 | void bdrv_op_block(BlockDriverState *bs, BlockOpType op, Error *reason) | |
5478 | { | |
5479 | BdrvOpBlocker *blocker; | |
5480 | assert((int) op >= 0 && op < BLOCK_OP_TYPE_MAX); | |
5481 | ||
5839e53b | 5482 | blocker = g_new0(BdrvOpBlocker, 1); |
fbe40ff7 FZ |
5483 | blocker->reason = reason; |
5484 | QLIST_INSERT_HEAD(&bs->op_blockers[op], blocker, list); | |
5485 | } | |
5486 | ||
5487 | void bdrv_op_unblock(BlockDriverState *bs, BlockOpType op, Error *reason) | |
5488 | { | |
5489 | BdrvOpBlocker *blocker, *next; | |
5490 | assert((int) op >= 0 && op < BLOCK_OP_TYPE_MAX); | |
5491 | QLIST_FOREACH_SAFE(blocker, &bs->op_blockers[op], list, next) { | |
5492 | if (blocker->reason == reason) { | |
5493 | QLIST_REMOVE(blocker, list); | |
5494 | g_free(blocker); | |
5495 | } | |
5496 | } | |
5497 | } | |
5498 | ||
5499 | void bdrv_op_block_all(BlockDriverState *bs, Error *reason) | |
5500 | { | |
5501 | int i; | |
5502 | for (i = 0; i < BLOCK_OP_TYPE_MAX; i++) { | |
5503 | bdrv_op_block(bs, i, reason); | |
5504 | } | |
5505 | } | |
5506 | ||
5507 | void bdrv_op_unblock_all(BlockDriverState *bs, Error *reason) | |
5508 | { | |
5509 | int i; | |
5510 | for (i = 0; i < BLOCK_OP_TYPE_MAX; i++) { | |
5511 | bdrv_op_unblock(bs, i, reason); | |
5512 | } | |
5513 | } | |
5514 | ||
5515 | bool bdrv_op_blocker_is_empty(BlockDriverState *bs) | |
5516 | { | |
5517 | int i; | |
5518 | ||
5519 | for (i = 0; i < BLOCK_OP_TYPE_MAX; i++) { | |
5520 | if (!QLIST_EMPTY(&bs->op_blockers[i])) { | |
5521 | return false; | |
5522 | } | |
5523 | } | |
5524 | return true; | |
5525 | } | |
5526 | ||
28a7282a LC |
5527 | void bdrv_iostatus_enable(BlockDriverState *bs) |
5528 | { | |
d6bf279e | 5529 | bs->iostatus_enabled = true; |
58e21ef5 | 5530 | bs->iostatus = BLOCK_DEVICE_IO_STATUS_OK; |
28a7282a LC |
5531 | } |
5532 | ||
5533 | /* The I/O status is only enabled if the drive explicitly | |
5534 | * enables it _and_ the VM is configured to stop on errors */ | |
5535 | bool bdrv_iostatus_is_enabled(const BlockDriverState *bs) | |
5536 | { | |
d6bf279e | 5537 | return (bs->iostatus_enabled && |
92aa5c6d PB |
5538 | (bs->on_write_error == BLOCKDEV_ON_ERROR_ENOSPC || |
5539 | bs->on_write_error == BLOCKDEV_ON_ERROR_STOP || | |
5540 | bs->on_read_error == BLOCKDEV_ON_ERROR_STOP)); | |
28a7282a LC |
5541 | } |
5542 | ||
5543 | void bdrv_iostatus_disable(BlockDriverState *bs) | |
5544 | { | |
d6bf279e | 5545 | bs->iostatus_enabled = false; |
28a7282a LC |
5546 | } |
5547 | ||
5548 | void bdrv_iostatus_reset(BlockDriverState *bs) | |
5549 | { | |
5550 | if (bdrv_iostatus_is_enabled(bs)) { | |
58e21ef5 | 5551 | bs->iostatus = BLOCK_DEVICE_IO_STATUS_OK; |
3bd293c3 PB |
5552 | if (bs->job) { |
5553 | block_job_iostatus_reset(bs->job); | |
5554 | } | |
28a7282a LC |
5555 | } |
5556 | } | |
5557 | ||
28a7282a LC |
5558 | void bdrv_iostatus_set_err(BlockDriverState *bs, int error) |
5559 | { | |
3e1caa5f PB |
5560 | assert(bdrv_iostatus_is_enabled(bs)); |
5561 | if (bs->iostatus == BLOCK_DEVICE_IO_STATUS_OK) { | |
58e21ef5 LC |
5562 | bs->iostatus = error == ENOSPC ? BLOCK_DEVICE_IO_STATUS_NOSPACE : |
5563 | BLOCK_DEVICE_IO_STATUS_FAILED; | |
28a7282a LC |
5564 | } |
5565 | } | |
5566 | ||
d92ada22 LC |
5567 | void bdrv_img_create(const char *filename, const char *fmt, |
5568 | const char *base_filename, const char *base_fmt, | |
f382d43a MR |
5569 | char *options, uint64_t img_size, int flags, |
5570 | Error **errp, bool quiet) | |
f88e1a42 | 5571 | { |
83d0521a CL |
5572 | QemuOptsList *create_opts = NULL; |
5573 | QemuOpts *opts = NULL; | |
5574 | const char *backing_fmt, *backing_file; | |
5575 | int64_t size; | |
f88e1a42 | 5576 | BlockDriver *drv, *proto_drv; |
96df67d1 | 5577 | BlockDriver *backing_drv = NULL; |
cc84d90f | 5578 | Error *local_err = NULL; |
f88e1a42 JS |
5579 | int ret = 0; |
5580 | ||
5581 | /* Find driver and parse its options */ | |
5582 | drv = bdrv_find_format(fmt); | |
5583 | if (!drv) { | |
71c79813 | 5584 | error_setg(errp, "Unknown file format '%s'", fmt); |
d92ada22 | 5585 | return; |
f88e1a42 JS |
5586 | } |
5587 | ||
98289620 | 5588 | proto_drv = bdrv_find_protocol(filename, true); |
f88e1a42 | 5589 | if (!proto_drv) { |
71c79813 | 5590 | error_setg(errp, "Unknown protocol '%s'", filename); |
d92ada22 | 5591 | return; |
f88e1a42 JS |
5592 | } |
5593 | ||
c6149724 HR |
5594 | if (!drv->create_opts) { |
5595 | error_setg(errp, "Format driver '%s' does not support image creation", | |
5596 | drv->format_name); | |
5597 | return; | |
5598 | } | |
5599 | ||
5600 | if (!proto_drv->create_opts) { | |
5601 | error_setg(errp, "Protocol driver '%s' does not support image creation", | |
5602 | proto_drv->format_name); | |
5603 | return; | |
5604 | } | |
5605 | ||
c282e1fd CL |
5606 | create_opts = qemu_opts_append(create_opts, drv->create_opts); |
5607 | create_opts = qemu_opts_append(create_opts, proto_drv->create_opts); | |
f88e1a42 JS |
5608 | |
5609 | /* Create parameter list with default values */ | |
83d0521a CL |
5610 | opts = qemu_opts_create(create_opts, NULL, 0, &error_abort); |
5611 | qemu_opt_set_number(opts, BLOCK_OPT_SIZE, img_size); | |
f88e1a42 JS |
5612 | |
5613 | /* Parse -o options */ | |
5614 | if (options) { | |
83d0521a CL |
5615 | if (qemu_opts_do_parse(opts, options, NULL) != 0) { |
5616 | error_setg(errp, "Invalid options for file format '%s'", fmt); | |
f88e1a42 JS |
5617 | goto out; |
5618 | } | |
5619 | } | |
5620 | ||
5621 | if (base_filename) { | |
83d0521a | 5622 | if (qemu_opt_set(opts, BLOCK_OPT_BACKING_FILE, base_filename)) { |
71c79813 LC |
5623 | error_setg(errp, "Backing file not supported for file format '%s'", |
5624 | fmt); | |
f88e1a42 JS |
5625 | goto out; |
5626 | } | |
5627 | } | |
5628 | ||
5629 | if (base_fmt) { | |
83d0521a | 5630 | if (qemu_opt_set(opts, BLOCK_OPT_BACKING_FMT, base_fmt)) { |
71c79813 LC |
5631 | error_setg(errp, "Backing file format not supported for file " |
5632 | "format '%s'", fmt); | |
f88e1a42 JS |
5633 | goto out; |
5634 | } | |
5635 | } | |
5636 | ||
83d0521a CL |
5637 | backing_file = qemu_opt_get(opts, BLOCK_OPT_BACKING_FILE); |
5638 | if (backing_file) { | |
5639 | if (!strcmp(filename, backing_file)) { | |
71c79813 LC |
5640 | error_setg(errp, "Error: Trying to create an image with the " |
5641 | "same filename as the backing file"); | |
792da93a JS |
5642 | goto out; |
5643 | } | |
5644 | } | |
5645 | ||
83d0521a CL |
5646 | backing_fmt = qemu_opt_get(opts, BLOCK_OPT_BACKING_FMT); |
5647 | if (backing_fmt) { | |
5648 | backing_drv = bdrv_find_format(backing_fmt); | |
96df67d1 | 5649 | if (!backing_drv) { |
71c79813 | 5650 | error_setg(errp, "Unknown backing file format '%s'", |
83d0521a | 5651 | backing_fmt); |
f88e1a42 JS |
5652 | goto out; |
5653 | } | |
5654 | } | |
5655 | ||
5656 | // The size for the image must always be specified, with one exception: | |
5657 | // If we are using a backing file, we can obtain the size from there | |
83d0521a CL |
5658 | size = qemu_opt_get_size(opts, BLOCK_OPT_SIZE, 0); |
5659 | if (size == -1) { | |
5660 | if (backing_file) { | |
66f6b814 | 5661 | BlockDriverState *bs; |
52bf1e72 | 5662 | int64_t size; |
63090dac PB |
5663 | int back_flags; |
5664 | ||
5665 | /* backing files always opened read-only */ | |
5666 | back_flags = | |
5667 | flags & ~(BDRV_O_RDWR | BDRV_O_SNAPSHOT | BDRV_O_NO_BACKING); | |
f88e1a42 | 5668 | |
f67503e5 | 5669 | bs = NULL; |
83d0521a | 5670 | ret = bdrv_open(&bs, backing_file, NULL, NULL, back_flags, |
cc84d90f | 5671 | backing_drv, &local_err); |
f88e1a42 | 5672 | if (ret < 0) { |
f88e1a42 JS |
5673 | goto out; |
5674 | } | |
52bf1e72 MA |
5675 | size = bdrv_getlength(bs); |
5676 | if (size < 0) { | |
5677 | error_setg_errno(errp, -size, "Could not get size of '%s'", | |
5678 | backing_file); | |
5679 | bdrv_unref(bs); | |
5680 | goto out; | |
5681 | } | |
f88e1a42 | 5682 | |
83d0521a | 5683 | qemu_opt_set_number(opts, BLOCK_OPT_SIZE, size); |
66f6b814 HR |
5684 | |
5685 | bdrv_unref(bs); | |
f88e1a42 | 5686 | } else { |
71c79813 | 5687 | error_setg(errp, "Image creation needs a size parameter"); |
f88e1a42 JS |
5688 | goto out; |
5689 | } | |
5690 | } | |
5691 | ||
f382d43a | 5692 | if (!quiet) { |
43c5d8f8 FZ |
5693 | printf("Formatting '%s', fmt=%s", filename, fmt); |
5694 | qemu_opts_print(opts, " "); | |
f382d43a MR |
5695 | puts(""); |
5696 | } | |
83d0521a | 5697 | |
c282e1fd | 5698 | ret = bdrv_create(drv, filename, opts, &local_err); |
83d0521a | 5699 | |
cc84d90f HR |
5700 | if (ret == -EFBIG) { |
5701 | /* This is generally a better message than whatever the driver would | |
5702 | * deliver (especially because of the cluster_size_hint), since that | |
5703 | * is most probably not much different from "image too large". */ | |
5704 | const char *cluster_size_hint = ""; | |
83d0521a | 5705 | if (qemu_opt_get_size(opts, BLOCK_OPT_CLUSTER_SIZE, 0)) { |
cc84d90f | 5706 | cluster_size_hint = " (try using a larger cluster size)"; |
f88e1a42 | 5707 | } |
cc84d90f HR |
5708 | error_setg(errp, "The image size is too large for file format '%s'" |
5709 | "%s", fmt, cluster_size_hint); | |
5710 | error_free(local_err); | |
5711 | local_err = NULL; | |
f88e1a42 JS |
5712 | } |
5713 | ||
5714 | out: | |
83d0521a CL |
5715 | qemu_opts_del(opts); |
5716 | qemu_opts_free(create_opts); | |
84d18f06 | 5717 | if (local_err) { |
cc84d90f HR |
5718 | error_propagate(errp, local_err); |
5719 | } | |
f88e1a42 | 5720 | } |
85d126f3 SH |
5721 | |
5722 | AioContext *bdrv_get_aio_context(BlockDriverState *bs) | |
5723 | { | |
dcd04228 SH |
5724 | return bs->aio_context; |
5725 | } | |
5726 | ||
5727 | void bdrv_detach_aio_context(BlockDriverState *bs) | |
5728 | { | |
33384421 HR |
5729 | BdrvAioNotifier *baf; |
5730 | ||
dcd04228 SH |
5731 | if (!bs->drv) { |
5732 | return; | |
5733 | } | |
5734 | ||
33384421 HR |
5735 | QLIST_FOREACH(baf, &bs->aio_notifiers, list) { |
5736 | baf->detach_aio_context(baf->opaque); | |
5737 | } | |
5738 | ||
13af91eb SH |
5739 | if (bs->io_limits_enabled) { |
5740 | throttle_detach_aio_context(&bs->throttle_state); | |
5741 | } | |
dcd04228 SH |
5742 | if (bs->drv->bdrv_detach_aio_context) { |
5743 | bs->drv->bdrv_detach_aio_context(bs); | |
5744 | } | |
5745 | if (bs->file) { | |
5746 | bdrv_detach_aio_context(bs->file); | |
5747 | } | |
5748 | if (bs->backing_hd) { | |
5749 | bdrv_detach_aio_context(bs->backing_hd); | |
5750 | } | |
5751 | ||
5752 | bs->aio_context = NULL; | |
5753 | } | |
5754 | ||
5755 | void bdrv_attach_aio_context(BlockDriverState *bs, | |
5756 | AioContext *new_context) | |
5757 | { | |
33384421 HR |
5758 | BdrvAioNotifier *ban; |
5759 | ||
dcd04228 SH |
5760 | if (!bs->drv) { |
5761 | return; | |
5762 | } | |
5763 | ||
5764 | bs->aio_context = new_context; | |
5765 | ||
5766 | if (bs->backing_hd) { | |
5767 | bdrv_attach_aio_context(bs->backing_hd, new_context); | |
5768 | } | |
5769 | if (bs->file) { | |
5770 | bdrv_attach_aio_context(bs->file, new_context); | |
5771 | } | |
5772 | if (bs->drv->bdrv_attach_aio_context) { | |
5773 | bs->drv->bdrv_attach_aio_context(bs, new_context); | |
5774 | } | |
13af91eb SH |
5775 | if (bs->io_limits_enabled) { |
5776 | throttle_attach_aio_context(&bs->throttle_state, new_context); | |
5777 | } | |
33384421 HR |
5778 | |
5779 | QLIST_FOREACH(ban, &bs->aio_notifiers, list) { | |
5780 | ban->attached_aio_context(new_context, ban->opaque); | |
5781 | } | |
dcd04228 SH |
5782 | } |
5783 | ||
5784 | void bdrv_set_aio_context(BlockDriverState *bs, AioContext *new_context) | |
5785 | { | |
5786 | bdrv_drain_all(); /* ensure there are no in-flight requests */ | |
5787 | ||
5788 | bdrv_detach_aio_context(bs); | |
5789 | ||
5790 | /* This function executes in the old AioContext so acquire the new one in | |
5791 | * case it runs in a different thread. | |
5792 | */ | |
5793 | aio_context_acquire(new_context); | |
5794 | bdrv_attach_aio_context(bs, new_context); | |
5795 | aio_context_release(new_context); | |
85d126f3 | 5796 | } |
d616b224 | 5797 | |
33384421 HR |
5798 | void bdrv_add_aio_context_notifier(BlockDriverState *bs, |
5799 | void (*attached_aio_context)(AioContext *new_context, void *opaque), | |
5800 | void (*detach_aio_context)(void *opaque), void *opaque) | |
5801 | { | |
5802 | BdrvAioNotifier *ban = g_new(BdrvAioNotifier, 1); | |
5803 | *ban = (BdrvAioNotifier){ | |
5804 | .attached_aio_context = attached_aio_context, | |
5805 | .detach_aio_context = detach_aio_context, | |
5806 | .opaque = opaque | |
5807 | }; | |
5808 | ||
5809 | QLIST_INSERT_HEAD(&bs->aio_notifiers, ban, list); | |
5810 | } | |
5811 | ||
5812 | void bdrv_remove_aio_context_notifier(BlockDriverState *bs, | |
5813 | void (*attached_aio_context)(AioContext *, | |
5814 | void *), | |
5815 | void (*detach_aio_context)(void *), | |
5816 | void *opaque) | |
5817 | { | |
5818 | BdrvAioNotifier *ban, *ban_next; | |
5819 | ||
5820 | QLIST_FOREACH_SAFE(ban, &bs->aio_notifiers, list, ban_next) { | |
5821 | if (ban->attached_aio_context == attached_aio_context && | |
5822 | ban->detach_aio_context == detach_aio_context && | |
5823 | ban->opaque == opaque) | |
5824 | { | |
5825 | QLIST_REMOVE(ban, list); | |
5826 | g_free(ban); | |
5827 | ||
5828 | return; | |
5829 | } | |
5830 | } | |
5831 | ||
5832 | abort(); | |
5833 | } | |
5834 | ||
d616b224 SH |
5835 | void bdrv_add_before_write_notifier(BlockDriverState *bs, |
5836 | NotifierWithReturn *notifier) | |
5837 | { | |
5838 | notifier_with_return_list_add(&bs->before_write_notifiers, notifier); | |
5839 | } | |
6f176b48 | 5840 | |
77485434 HR |
5841 | int bdrv_amend_options(BlockDriverState *bs, QemuOpts *opts, |
5842 | BlockDriverAmendStatusCB *status_cb) | |
6f176b48 | 5843 | { |
c282e1fd | 5844 | if (!bs->drv->bdrv_amend_options) { |
6f176b48 HR |
5845 | return -ENOTSUP; |
5846 | } | |
77485434 | 5847 | return bs->drv->bdrv_amend_options(bs, opts, status_cb); |
6f176b48 | 5848 | } |
f6186f49 | 5849 | |
b5042a36 BC |
5850 | /* This function will be called by the bdrv_recurse_is_first_non_filter method |
5851 | * of block filter and by bdrv_is_first_non_filter. | |
5852 | * It is used to test if the given bs is the candidate or recurse more in the | |
5853 | * node graph. | |
212a5a8f | 5854 | */ |
b5042a36 | 5855 | bool bdrv_recurse_is_first_non_filter(BlockDriverState *bs, |
212a5a8f | 5856 | BlockDriverState *candidate) |
f6186f49 | 5857 | { |
b5042a36 BC |
5858 | /* return false if basic checks fails */ |
5859 | if (!bs || !bs->drv) { | |
212a5a8f | 5860 | return false; |
f6186f49 BC |
5861 | } |
5862 | ||
b5042a36 BC |
5863 | /* the code reached a non block filter driver -> check if the bs is |
5864 | * the same as the candidate. It's the recursion termination condition. | |
5865 | */ | |
5866 | if (!bs->drv->is_filter) { | |
5867 | return bs == candidate; | |
212a5a8f | 5868 | } |
b5042a36 | 5869 | /* Down this path the driver is a block filter driver */ |
212a5a8f | 5870 | |
b5042a36 BC |
5871 | /* If the block filter recursion method is defined use it to recurse down |
5872 | * the node graph. | |
5873 | */ | |
5874 | if (bs->drv->bdrv_recurse_is_first_non_filter) { | |
212a5a8f | 5875 | return bs->drv->bdrv_recurse_is_first_non_filter(bs, candidate); |
f6186f49 BC |
5876 | } |
5877 | ||
b5042a36 BC |
5878 | /* the driver is a block filter but don't allow to recurse -> return false |
5879 | */ | |
5880 | return false; | |
f6186f49 BC |
5881 | } |
5882 | ||
212a5a8f BC |
5883 | /* This function checks if the candidate is the first non filter bs down it's |
5884 | * bs chain. Since we don't have pointers to parents it explore all bs chains | |
5885 | * from the top. Some filters can choose not to pass down the recursion. | |
5886 | */ | |
5887 | bool bdrv_is_first_non_filter(BlockDriverState *candidate) | |
f6186f49 | 5888 | { |
212a5a8f BC |
5889 | BlockDriverState *bs; |
5890 | ||
5891 | /* walk down the bs forest recursively */ | |
5892 | QTAILQ_FOREACH(bs, &bdrv_states, device_list) { | |
5893 | bool perm; | |
5894 | ||
b5042a36 | 5895 | /* try to recurse in this top level bs */ |
e6dc8a1f | 5896 | perm = bdrv_recurse_is_first_non_filter(bs, candidate); |
212a5a8f BC |
5897 | |
5898 | /* candidate is the first non filter */ | |
5899 | if (perm) { | |
5900 | return true; | |
5901 | } | |
5902 | } | |
5903 | ||
5904 | return false; | |
f6186f49 | 5905 | } |
09158f00 BC |
5906 | |
5907 | BlockDriverState *check_to_replace_node(const char *node_name, Error **errp) | |
5908 | { | |
5909 | BlockDriverState *to_replace_bs = bdrv_find_node(node_name); | |
5a7e7a0b SH |
5910 | AioContext *aio_context; |
5911 | ||
09158f00 BC |
5912 | if (!to_replace_bs) { |
5913 | error_setg(errp, "Node name '%s' not found", node_name); | |
5914 | return NULL; | |
5915 | } | |
5916 | ||
5a7e7a0b SH |
5917 | aio_context = bdrv_get_aio_context(to_replace_bs); |
5918 | aio_context_acquire(aio_context); | |
5919 | ||
09158f00 | 5920 | if (bdrv_op_is_blocked(to_replace_bs, BLOCK_OP_TYPE_REPLACE, errp)) { |
5a7e7a0b SH |
5921 | to_replace_bs = NULL; |
5922 | goto out; | |
09158f00 BC |
5923 | } |
5924 | ||
5925 | /* We don't want arbitrary node of the BDS chain to be replaced only the top | |
5926 | * most non filter in order to prevent data corruption. | |
5927 | * Another benefit is that this tests exclude backing files which are | |
5928 | * blocked by the backing blockers. | |
5929 | */ | |
5930 | if (!bdrv_is_first_non_filter(to_replace_bs)) { | |
5931 | error_setg(errp, "Only top most non filter can be replaced"); | |
5a7e7a0b SH |
5932 | to_replace_bs = NULL; |
5933 | goto out; | |
09158f00 BC |
5934 | } |
5935 | ||
5a7e7a0b SH |
5936 | out: |
5937 | aio_context_release(aio_context); | |
09158f00 BC |
5938 | return to_replace_bs; |
5939 | } | |
448ad91d ML |
5940 | |
5941 | void bdrv_io_plug(BlockDriverState *bs) | |
5942 | { | |
5943 | BlockDriver *drv = bs->drv; | |
5944 | if (drv && drv->bdrv_io_plug) { | |
5945 | drv->bdrv_io_plug(bs); | |
5946 | } else if (bs->file) { | |
5947 | bdrv_io_plug(bs->file); | |
5948 | } | |
5949 | } | |
5950 | ||
5951 | void bdrv_io_unplug(BlockDriverState *bs) | |
5952 | { | |
5953 | BlockDriver *drv = bs->drv; | |
5954 | if (drv && drv->bdrv_io_unplug) { | |
5955 | drv->bdrv_io_unplug(bs); | |
5956 | } else if (bs->file) { | |
5957 | bdrv_io_unplug(bs->file); | |
5958 | } | |
5959 | } | |
5960 | ||
5961 | void bdrv_flush_io_queue(BlockDriverState *bs) | |
5962 | { | |
5963 | BlockDriver *drv = bs->drv; | |
5964 | if (drv && drv->bdrv_flush_io_queue) { | |
5965 | drv->bdrv_flush_io_queue(bs); | |
5966 | } else if (bs->file) { | |
5967 | bdrv_flush_io_queue(bs->file); | |
5968 | } | |
5969 | } | |
91af7014 HR |
5970 | |
5971 | static bool append_open_options(QDict *d, BlockDriverState *bs) | |
5972 | { | |
5973 | const QDictEntry *entry; | |
5974 | bool found_any = false; | |
5975 | ||
5976 | for (entry = qdict_first(bs->options); entry; | |
5977 | entry = qdict_next(bs->options, entry)) | |
5978 | { | |
5979 | /* Only take options for this level and exclude all non-driver-specific | |
5980 | * options */ | |
5981 | if (!strchr(qdict_entry_key(entry), '.') && | |
5982 | strcmp(qdict_entry_key(entry), "node-name")) | |
5983 | { | |
5984 | qobject_incref(qdict_entry_value(entry)); | |
5985 | qdict_put_obj(d, qdict_entry_key(entry), qdict_entry_value(entry)); | |
5986 | found_any = true; | |
5987 | } | |
5988 | } | |
5989 | ||
5990 | return found_any; | |
5991 | } | |
5992 | ||
5993 | /* Updates the following BDS fields: | |
5994 | * - exact_filename: A filename which may be used for opening a block device | |
5995 | * which (mostly) equals the given BDS (even without any | |
5996 | * other options; so reading and writing must return the same | |
5997 | * results, but caching etc. may be different) | |
5998 | * - full_open_options: Options which, when given when opening a block device | |
5999 | * (without a filename), result in a BDS (mostly) | |
6000 | * equalling the given one | |
6001 | * - filename: If exact_filename is set, it is copied here. Otherwise, | |
6002 | * full_open_options is converted to a JSON object, prefixed with | |
6003 | * "json:" (for use through the JSON pseudo protocol) and put here. | |
6004 | */ | |
6005 | void bdrv_refresh_filename(BlockDriverState *bs) | |
6006 | { | |
6007 | BlockDriver *drv = bs->drv; | |
6008 | QDict *opts; | |
6009 | ||
6010 | if (!drv) { | |
6011 | return; | |
6012 | } | |
6013 | ||
6014 | /* This BDS's file name will most probably depend on its file's name, so | |
6015 | * refresh that first */ | |
6016 | if (bs->file) { | |
6017 | bdrv_refresh_filename(bs->file); | |
6018 | } | |
6019 | ||
6020 | if (drv->bdrv_refresh_filename) { | |
6021 | /* Obsolete information is of no use here, so drop the old file name | |
6022 | * information before refreshing it */ | |
6023 | bs->exact_filename[0] = '\0'; | |
6024 | if (bs->full_open_options) { | |
6025 | QDECREF(bs->full_open_options); | |
6026 | bs->full_open_options = NULL; | |
6027 | } | |
6028 | ||
6029 | drv->bdrv_refresh_filename(bs); | |
6030 | } else if (bs->file) { | |
6031 | /* Try to reconstruct valid information from the underlying file */ | |
6032 | bool has_open_options; | |
6033 | ||
6034 | bs->exact_filename[0] = '\0'; | |
6035 | if (bs->full_open_options) { | |
6036 | QDECREF(bs->full_open_options); | |
6037 | bs->full_open_options = NULL; | |
6038 | } | |
6039 | ||
6040 | opts = qdict_new(); | |
6041 | has_open_options = append_open_options(opts, bs); | |
6042 | ||
6043 | /* If no specific options have been given for this BDS, the filename of | |
6044 | * the underlying file should suffice for this one as well */ | |
6045 | if (bs->file->exact_filename[0] && !has_open_options) { | |
6046 | strcpy(bs->exact_filename, bs->file->exact_filename); | |
6047 | } | |
6048 | /* Reconstructing the full options QDict is simple for most format block | |
6049 | * drivers, as long as the full options are known for the underlying | |
6050 | * file BDS. The full options QDict of that file BDS should somehow | |
6051 | * contain a representation of the filename, therefore the following | |
6052 | * suffices without querying the (exact_)filename of this BDS. */ | |
6053 | if (bs->file->full_open_options) { | |
6054 | qdict_put_obj(opts, "driver", | |
6055 | QOBJECT(qstring_from_str(drv->format_name))); | |
6056 | QINCREF(bs->file->full_open_options); | |
6057 | qdict_put_obj(opts, "file", QOBJECT(bs->file->full_open_options)); | |
6058 | ||
6059 | bs->full_open_options = opts; | |
6060 | } else { | |
6061 | QDECREF(opts); | |
6062 | } | |
6063 | } else if (!bs->full_open_options && qdict_size(bs->options)) { | |
6064 | /* There is no underlying file BDS (at least referenced by BDS.file), | |
6065 | * so the full options QDict should be equal to the options given | |
6066 | * specifically for this block device when it was opened (plus the | |
6067 | * driver specification). | |
6068 | * Because those options don't change, there is no need to update | |
6069 | * full_open_options when it's already set. */ | |
6070 | ||
6071 | opts = qdict_new(); | |
6072 | append_open_options(opts, bs); | |
6073 | qdict_put_obj(opts, "driver", | |
6074 | QOBJECT(qstring_from_str(drv->format_name))); | |
6075 | ||
6076 | if (bs->exact_filename[0]) { | |
6077 | /* This may not work for all block protocol drivers (some may | |
6078 | * require this filename to be parsed), but we have to find some | |
6079 | * default solution here, so just include it. If some block driver | |
6080 | * does not support pure options without any filename at all or | |
6081 | * needs some special format of the options QDict, it needs to | |
6082 | * implement the driver-specific bdrv_refresh_filename() function. | |
6083 | */ | |
6084 | qdict_put_obj(opts, "filename", | |
6085 | QOBJECT(qstring_from_str(bs->exact_filename))); | |
6086 | } | |
6087 | ||
6088 | bs->full_open_options = opts; | |
6089 | } | |
6090 | ||
6091 | if (bs->exact_filename[0]) { | |
6092 | pstrcpy(bs->filename, sizeof(bs->filename), bs->exact_filename); | |
6093 | } else if (bs->full_open_options) { | |
6094 | QString *json = qobject_to_json(QOBJECT(bs->full_open_options)); | |
6095 | snprintf(bs->filename, sizeof(bs->filename), "json:%s", | |
6096 | qstring_get_str(json)); | |
6097 | QDECREF(json); | |
6098 | } | |
6099 | } | |
5366d0c8 BC |
6100 | |
6101 | /* This accessor function purpose is to allow the device models to access the | |
6102 | * BlockAcctStats structure embedded inside a BlockDriverState without being | |
6103 | * aware of the BlockDriverState structure layout. | |
6104 | * It will go away when the BlockAcctStats structure will be moved inside | |
6105 | * the device models. | |
6106 | */ | |
6107 | BlockAcctStats *bdrv_get_stats(BlockDriverState *bs) | |
6108 | { | |
6109 | return &bs->stats; | |
6110 | } |