]>
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" |
83c9089e | 27 | #include "monitor/monitor.h" |
737e150e PB |
28 | #include "block/block_int.h" |
29 | #include "block/blockjob.h" | |
1de7afc9 | 30 | #include "qemu/module.h" |
7b1b5d19 | 31 | #include "qapi/qmp/qjson.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" |
fc01f7e7 | 38 | |
71e72a19 | 39 | #ifdef CONFIG_BSD |
7674e7bf FB |
40 | #include <sys/types.h> |
41 | #include <sys/stat.h> | |
42 | #include <sys/ioctl.h> | |
72cf2d4f | 43 | #include <sys/queue.h> |
c5e97233 | 44 | #ifndef __DragonFly__ |
7674e7bf FB |
45 | #include <sys/disk.h> |
46 | #endif | |
c5e97233 | 47 | #endif |
7674e7bf | 48 | |
49dc768d AL |
49 | #ifdef _WIN32 |
50 | #include <windows.h> | |
51 | #endif | |
52 | ||
e4654d2d FZ |
53 | struct BdrvDirtyBitmap { |
54 | HBitmap *bitmap; | |
55 | QLIST_ENTRY(BdrvDirtyBitmap) list; | |
56 | }; | |
57 | ||
1c9805a3 SH |
58 | #define NOT_DONE 0x7fffffff /* used while emulated sync operation in progress */ |
59 | ||
7d4b4ba5 | 60 | static void bdrv_dev_change_media_cb(BlockDriverState *bs, bool load); |
f141eafe AL |
61 | static BlockDriverAIOCB *bdrv_aio_readv_em(BlockDriverState *bs, |
62 | int64_t sector_num, QEMUIOVector *qiov, int nb_sectors, | |
c87c0672 | 63 | BlockDriverCompletionFunc *cb, void *opaque); |
f141eafe AL |
64 | static BlockDriverAIOCB *bdrv_aio_writev_em(BlockDriverState *bs, |
65 | int64_t sector_num, QEMUIOVector *qiov, int nb_sectors, | |
ce1a14dc | 66 | BlockDriverCompletionFunc *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); |
b2a61371 SH |
79 | static BlockDriverAIOCB *bdrv_co_aio_rw_vector(BlockDriverState *bs, |
80 | int64_t sector_num, | |
81 | QEMUIOVector *qiov, | |
82 | int nb_sectors, | |
d20d9b7c | 83 | BdrvRequestFlags flags, |
b2a61371 SH |
84 | BlockDriverCompletionFunc *cb, |
85 | void *opaque, | |
8c5873d6 | 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, | |
182 | QEMU_CLOCK_VIRTUAL, | |
183 | bdrv_throttle_read_timer_cb, | |
184 | bdrv_throttle_write_timer_cb, | |
185 | bs); | |
186 | bs->io_limits_enabled = true; | |
187 | } | |
188 | ||
189 | /* This function makes an IO wait if needed | |
190 | * | |
191 | * @nb_sectors: the number of sectors of the IO | |
192 | * @is_write: is the IO a write | |
193 | */ | |
98f90dba | 194 | static void bdrv_io_limits_intercept(BlockDriverState *bs, |
d5103588 | 195 | unsigned int bytes, |
cc0681c4 | 196 | bool is_write) |
98f90dba | 197 | { |
cc0681c4 BC |
198 | /* does this io must wait */ |
199 | bool must_wait = throttle_schedule_timer(&bs->throttle_state, is_write); | |
98f90dba | 200 | |
cc0681c4 BC |
201 | /* if must wait or any request of this type throttled queue the IO */ |
202 | if (must_wait || | |
203 | !qemu_co_queue_empty(&bs->throttled_reqs[is_write])) { | |
204 | qemu_co_queue_wait(&bs->throttled_reqs[is_write]); | |
98f90dba ZYW |
205 | } |
206 | ||
cc0681c4 | 207 | /* the IO will be executed, do the accounting */ |
d5103588 KW |
208 | throttle_account(&bs->throttle_state, is_write, bytes); |
209 | ||
98f90dba | 210 | |
cc0681c4 BC |
211 | /* if the next request must wait -> do nothing */ |
212 | if (throttle_schedule_timer(&bs->throttle_state, is_write)) { | |
213 | return; | |
98f90dba ZYW |
214 | } |
215 | ||
cc0681c4 BC |
216 | /* else queue next request for execution */ |
217 | qemu_co_queue_next(&bs->throttled_reqs[is_write]); | |
98f90dba ZYW |
218 | } |
219 | ||
339064d5 KW |
220 | size_t bdrv_opt_mem_align(BlockDriverState *bs) |
221 | { | |
222 | if (!bs || !bs->drv) { | |
223 | /* 4k should be on the safe side */ | |
224 | return 4096; | |
225 | } | |
226 | ||
227 | return bs->bl.opt_mem_alignment; | |
228 | } | |
229 | ||
9e0b22f4 SH |
230 | /* check if the path starts with "<protocol>:" */ |
231 | static int path_has_protocol(const char *path) | |
232 | { | |
947995c0 PB |
233 | const char *p; |
234 | ||
9e0b22f4 SH |
235 | #ifdef _WIN32 |
236 | if (is_windows_drive(path) || | |
237 | is_windows_drive_prefix(path)) { | |
238 | return 0; | |
239 | } | |
947995c0 PB |
240 | p = path + strcspn(path, ":/\\"); |
241 | #else | |
242 | p = path + strcspn(path, ":/"); | |
9e0b22f4 SH |
243 | #endif |
244 | ||
947995c0 | 245 | return *p == ':'; |
9e0b22f4 SH |
246 | } |
247 | ||
83f64091 | 248 | int path_is_absolute(const char *path) |
3b0d4f61 | 249 | { |
21664424 FB |
250 | #ifdef _WIN32 |
251 | /* specific case for names like: "\\.\d:" */ | |
f53f4da9 | 252 | if (is_windows_drive(path) || is_windows_drive_prefix(path)) { |
21664424 | 253 | return 1; |
f53f4da9 PB |
254 | } |
255 | return (*path == '/' || *path == '\\'); | |
3b9f94e1 | 256 | #else |
f53f4da9 | 257 | return (*path == '/'); |
3b9f94e1 | 258 | #endif |
3b0d4f61 FB |
259 | } |
260 | ||
83f64091 FB |
261 | /* if filename is absolute, just copy it to dest. Otherwise, build a |
262 | path to it by considering it is relative to base_path. URL are | |
263 | supported. */ | |
264 | void path_combine(char *dest, int dest_size, | |
265 | const char *base_path, | |
266 | const char *filename) | |
3b0d4f61 | 267 | { |
83f64091 FB |
268 | const char *p, *p1; |
269 | int len; | |
270 | ||
271 | if (dest_size <= 0) | |
272 | return; | |
273 | if (path_is_absolute(filename)) { | |
274 | pstrcpy(dest, dest_size, filename); | |
275 | } else { | |
276 | p = strchr(base_path, ':'); | |
277 | if (p) | |
278 | p++; | |
279 | else | |
280 | p = base_path; | |
3b9f94e1 FB |
281 | p1 = strrchr(base_path, '/'); |
282 | #ifdef _WIN32 | |
283 | { | |
284 | const char *p2; | |
285 | p2 = strrchr(base_path, '\\'); | |
286 | if (!p1 || p2 > p1) | |
287 | p1 = p2; | |
288 | } | |
289 | #endif | |
83f64091 FB |
290 | if (p1) |
291 | p1++; | |
292 | else | |
293 | p1 = base_path; | |
294 | if (p1 > p) | |
295 | p = p1; | |
296 | len = p - base_path; | |
297 | if (len > dest_size - 1) | |
298 | len = dest_size - 1; | |
299 | memcpy(dest, base_path, len); | |
300 | dest[len] = '\0'; | |
301 | pstrcat(dest, dest_size, filename); | |
3b0d4f61 | 302 | } |
3b0d4f61 FB |
303 | } |
304 | ||
dc5a1371 PB |
305 | void bdrv_get_full_backing_filename(BlockDriverState *bs, char *dest, size_t sz) |
306 | { | |
307 | if (bs->backing_file[0] == '\0' || path_has_protocol(bs->backing_file)) { | |
308 | pstrcpy(dest, sz, bs->backing_file); | |
309 | } else { | |
310 | path_combine(dest, sz, bs->filename, bs->backing_file); | |
311 | } | |
312 | } | |
313 | ||
5efa9d5a | 314 | void bdrv_register(BlockDriver *bdrv) |
ea2384d3 | 315 | { |
8c5873d6 SH |
316 | /* Block drivers without coroutine functions need emulation */ |
317 | if (!bdrv->bdrv_co_readv) { | |
f9f05dc5 KW |
318 | bdrv->bdrv_co_readv = bdrv_co_readv_em; |
319 | bdrv->bdrv_co_writev = bdrv_co_writev_em; | |
320 | ||
f8c35c1d SH |
321 | /* bdrv_co_readv_em()/brdv_co_writev_em() work in terms of aio, so if |
322 | * the block driver lacks aio we need to emulate that too. | |
323 | */ | |
f9f05dc5 KW |
324 | if (!bdrv->bdrv_aio_readv) { |
325 | /* add AIO emulation layer */ | |
326 | bdrv->bdrv_aio_readv = bdrv_aio_readv_em; | |
327 | bdrv->bdrv_aio_writev = bdrv_aio_writev_em; | |
f9f05dc5 | 328 | } |
83f64091 | 329 | } |
b2e12bc6 | 330 | |
8a22f02a | 331 | QLIST_INSERT_HEAD(&bdrv_drivers, bdrv, list); |
ea2384d3 | 332 | } |
b338082b FB |
333 | |
334 | /* create a new block device (by default it is empty) */ | |
335 | BlockDriverState *bdrv_new(const char *device_name) | |
336 | { | |
1b7bdbc1 | 337 | BlockDriverState *bs; |
b338082b | 338 | |
7267c094 | 339 | bs = g_malloc0(sizeof(BlockDriverState)); |
e4654d2d | 340 | QLIST_INIT(&bs->dirty_bitmaps); |
b338082b | 341 | pstrcpy(bs->device_name, sizeof(bs->device_name), device_name); |
ea2384d3 | 342 | if (device_name[0] != '\0') { |
dc364f4c | 343 | QTAILQ_INSERT_TAIL(&bdrv_states, bs, device_list); |
ea2384d3 | 344 | } |
28a7282a | 345 | bdrv_iostatus_disable(bs); |
d7d512f6 | 346 | notifier_list_init(&bs->close_notifiers); |
d616b224 | 347 | notifier_with_return_list_init(&bs->before_write_notifiers); |
cc0681c4 BC |
348 | qemu_co_queue_init(&bs->throttled_reqs[0]); |
349 | qemu_co_queue_init(&bs->throttled_reqs[1]); | |
9fcb0251 | 350 | bs->refcnt = 1; |
d7d512f6 | 351 | |
b338082b FB |
352 | return bs; |
353 | } | |
354 | ||
d7d512f6 PB |
355 | void bdrv_add_close_notifier(BlockDriverState *bs, Notifier *notify) |
356 | { | |
357 | notifier_list_add(&bs->close_notifiers, notify); | |
358 | } | |
359 | ||
ea2384d3 FB |
360 | BlockDriver *bdrv_find_format(const char *format_name) |
361 | { | |
362 | BlockDriver *drv1; | |
8a22f02a SH |
363 | QLIST_FOREACH(drv1, &bdrv_drivers, list) { |
364 | if (!strcmp(drv1->format_name, format_name)) { | |
ea2384d3 | 365 | return drv1; |
8a22f02a | 366 | } |
ea2384d3 FB |
367 | } |
368 | return NULL; | |
369 | } | |
370 | ||
b64ec4e4 | 371 | static int bdrv_is_whitelisted(BlockDriver *drv, bool read_only) |
eb852011 | 372 | { |
b64ec4e4 FZ |
373 | static const char *whitelist_rw[] = { |
374 | CONFIG_BDRV_RW_WHITELIST | |
375 | }; | |
376 | static const char *whitelist_ro[] = { | |
377 | CONFIG_BDRV_RO_WHITELIST | |
eb852011 MA |
378 | }; |
379 | const char **p; | |
380 | ||
b64ec4e4 | 381 | if (!whitelist_rw[0] && !whitelist_ro[0]) { |
eb852011 | 382 | return 1; /* no whitelist, anything goes */ |
b64ec4e4 | 383 | } |
eb852011 | 384 | |
b64ec4e4 | 385 | for (p = whitelist_rw; *p; p++) { |
eb852011 MA |
386 | if (!strcmp(drv->format_name, *p)) { |
387 | return 1; | |
388 | } | |
389 | } | |
b64ec4e4 FZ |
390 | if (read_only) { |
391 | for (p = whitelist_ro; *p; p++) { | |
392 | if (!strcmp(drv->format_name, *p)) { | |
393 | return 1; | |
394 | } | |
395 | } | |
396 | } | |
eb852011 MA |
397 | return 0; |
398 | } | |
399 | ||
b64ec4e4 FZ |
400 | BlockDriver *bdrv_find_whitelisted_format(const char *format_name, |
401 | bool read_only) | |
eb852011 MA |
402 | { |
403 | BlockDriver *drv = bdrv_find_format(format_name); | |
b64ec4e4 | 404 | return drv && bdrv_is_whitelisted(drv, read_only) ? drv : NULL; |
eb852011 MA |
405 | } |
406 | ||
5b7e1542 ZYW |
407 | typedef struct CreateCo { |
408 | BlockDriver *drv; | |
409 | char *filename; | |
410 | QEMUOptionParameter *options; | |
411 | int ret; | |
cc84d90f | 412 | Error *err; |
5b7e1542 ZYW |
413 | } CreateCo; |
414 | ||
415 | static void coroutine_fn bdrv_create_co_entry(void *opaque) | |
416 | { | |
cc84d90f HR |
417 | Error *local_err = NULL; |
418 | int ret; | |
419 | ||
5b7e1542 ZYW |
420 | CreateCo *cco = opaque; |
421 | assert(cco->drv); | |
422 | ||
cc84d90f | 423 | ret = cco->drv->bdrv_create(cco->filename, cco->options, &local_err); |
84d18f06 | 424 | if (local_err) { |
cc84d90f HR |
425 | error_propagate(&cco->err, local_err); |
426 | } | |
427 | cco->ret = ret; | |
5b7e1542 ZYW |
428 | } |
429 | ||
0e7e1989 | 430 | int bdrv_create(BlockDriver *drv, const char* filename, |
cc84d90f | 431 | QEMUOptionParameter *options, Error **errp) |
ea2384d3 | 432 | { |
5b7e1542 ZYW |
433 | int ret; |
434 | ||
435 | Coroutine *co; | |
436 | CreateCo cco = { | |
437 | .drv = drv, | |
438 | .filename = g_strdup(filename), | |
439 | .options = options, | |
440 | .ret = NOT_DONE, | |
cc84d90f | 441 | .err = NULL, |
5b7e1542 ZYW |
442 | }; |
443 | ||
444 | if (!drv->bdrv_create) { | |
cc84d90f | 445 | error_setg(errp, "Driver '%s' does not support image creation", drv->format_name); |
80168bff LC |
446 | ret = -ENOTSUP; |
447 | goto out; | |
5b7e1542 ZYW |
448 | } |
449 | ||
450 | if (qemu_in_coroutine()) { | |
451 | /* Fast-path if already in coroutine context */ | |
452 | bdrv_create_co_entry(&cco); | |
453 | } else { | |
454 | co = qemu_coroutine_create(bdrv_create_co_entry); | |
455 | qemu_coroutine_enter(co, &cco); | |
456 | while (cco.ret == NOT_DONE) { | |
457 | qemu_aio_wait(); | |
458 | } | |
459 | } | |
460 | ||
461 | ret = cco.ret; | |
cc84d90f | 462 | if (ret < 0) { |
84d18f06 | 463 | if (cco.err) { |
cc84d90f HR |
464 | error_propagate(errp, cco.err); |
465 | } else { | |
466 | error_setg_errno(errp, -ret, "Could not create image"); | |
467 | } | |
468 | } | |
0e7e1989 | 469 | |
80168bff LC |
470 | out: |
471 | g_free(cco.filename); | |
5b7e1542 | 472 | return ret; |
ea2384d3 FB |
473 | } |
474 | ||
cc84d90f HR |
475 | int bdrv_create_file(const char* filename, QEMUOptionParameter *options, |
476 | Error **errp) | |
84a12e66 CH |
477 | { |
478 | BlockDriver *drv; | |
cc84d90f HR |
479 | Error *local_err = NULL; |
480 | int ret; | |
84a12e66 | 481 | |
98289620 | 482 | drv = bdrv_find_protocol(filename, true); |
84a12e66 | 483 | if (drv == NULL) { |
cc84d90f | 484 | error_setg(errp, "Could not find protocol for file '%s'", filename); |
16905d71 | 485 | return -ENOENT; |
84a12e66 CH |
486 | } |
487 | ||
cc84d90f | 488 | ret = bdrv_create(drv, filename, options, &local_err); |
84d18f06 | 489 | if (local_err) { |
cc84d90f HR |
490 | error_propagate(errp, local_err); |
491 | } | |
492 | return ret; | |
84a12e66 CH |
493 | } |
494 | ||
355ef4ac | 495 | int bdrv_refresh_limits(BlockDriverState *bs) |
d34682cd KW |
496 | { |
497 | BlockDriver *drv = bs->drv; | |
498 | ||
499 | memset(&bs->bl, 0, sizeof(bs->bl)); | |
500 | ||
466ad822 KW |
501 | if (!drv) { |
502 | return 0; | |
503 | } | |
504 | ||
505 | /* Take some limits from the children as a default */ | |
506 | if (bs->file) { | |
507 | bdrv_refresh_limits(bs->file); | |
508 | bs->bl.opt_transfer_length = bs->file->bl.opt_transfer_length; | |
339064d5 KW |
509 | bs->bl.opt_mem_alignment = bs->file->bl.opt_mem_alignment; |
510 | } else { | |
511 | bs->bl.opt_mem_alignment = 512; | |
466ad822 KW |
512 | } |
513 | ||
514 | if (bs->backing_hd) { | |
515 | bdrv_refresh_limits(bs->backing_hd); | |
516 | bs->bl.opt_transfer_length = | |
517 | MAX(bs->bl.opt_transfer_length, | |
518 | bs->backing_hd->bl.opt_transfer_length); | |
339064d5 KW |
519 | bs->bl.opt_mem_alignment = |
520 | MAX(bs->bl.opt_mem_alignment, | |
521 | bs->backing_hd->bl.opt_mem_alignment); | |
466ad822 KW |
522 | } |
523 | ||
524 | /* Then let the driver override it */ | |
525 | if (drv->bdrv_refresh_limits) { | |
d34682cd KW |
526 | return drv->bdrv_refresh_limits(bs); |
527 | } | |
528 | ||
529 | return 0; | |
530 | } | |
531 | ||
eba25057 JM |
532 | /* |
533 | * Create a uniquely-named empty temporary file. | |
534 | * Return 0 upon success, otherwise a negative errno value. | |
535 | */ | |
536 | int get_tmp_filename(char *filename, int size) | |
d5249393 | 537 | { |
eba25057 | 538 | #ifdef _WIN32 |
3b9f94e1 | 539 | char temp_dir[MAX_PATH]; |
eba25057 JM |
540 | /* GetTempFileName requires that its output buffer (4th param) |
541 | have length MAX_PATH or greater. */ | |
542 | assert(size >= MAX_PATH); | |
543 | return (GetTempPath(MAX_PATH, temp_dir) | |
544 | && GetTempFileName(temp_dir, "qem", 0, filename) | |
545 | ? 0 : -GetLastError()); | |
d5249393 | 546 | #else |
67b915a5 | 547 | int fd; |
7ccfb2eb | 548 | const char *tmpdir; |
0badc1ee | 549 | tmpdir = getenv("TMPDIR"); |
69bef793 AS |
550 | if (!tmpdir) { |
551 | tmpdir = "/var/tmp"; | |
552 | } | |
eba25057 JM |
553 | if (snprintf(filename, size, "%s/vl.XXXXXX", tmpdir) >= size) { |
554 | return -EOVERFLOW; | |
555 | } | |
ea2384d3 | 556 | fd = mkstemp(filename); |
fe235a06 DH |
557 | if (fd < 0) { |
558 | return -errno; | |
559 | } | |
560 | if (close(fd) != 0) { | |
561 | unlink(filename); | |
eba25057 JM |
562 | return -errno; |
563 | } | |
564 | return 0; | |
d5249393 | 565 | #endif |
eba25057 | 566 | } |
fc01f7e7 | 567 | |
84a12e66 CH |
568 | /* |
569 | * Detect host devices. By convention, /dev/cdrom[N] is always | |
570 | * recognized as a host CDROM. | |
571 | */ | |
572 | static BlockDriver *find_hdev_driver(const char *filename) | |
573 | { | |
574 | int score_max = 0, score; | |
575 | BlockDriver *drv = NULL, *d; | |
576 | ||
577 | QLIST_FOREACH(d, &bdrv_drivers, list) { | |
578 | if (d->bdrv_probe_device) { | |
579 | score = d->bdrv_probe_device(filename); | |
580 | if (score > score_max) { | |
581 | score_max = score; | |
582 | drv = d; | |
583 | } | |
584 | } | |
585 | } | |
586 | ||
587 | return drv; | |
588 | } | |
589 | ||
98289620 KW |
590 | BlockDriver *bdrv_find_protocol(const char *filename, |
591 | bool allow_protocol_prefix) | |
83f64091 FB |
592 | { |
593 | BlockDriver *drv1; | |
594 | char protocol[128]; | |
1cec71e3 | 595 | int len; |
83f64091 | 596 | const char *p; |
19cb3738 | 597 | |
66f82cee KW |
598 | /* TODO Drivers without bdrv_file_open must be specified explicitly */ |
599 | ||
39508e7a CH |
600 | /* |
601 | * XXX(hch): we really should not let host device detection | |
602 | * override an explicit protocol specification, but moving this | |
603 | * later breaks access to device names with colons in them. | |
604 | * Thanks to the brain-dead persistent naming schemes on udev- | |
605 | * based Linux systems those actually are quite common. | |
606 | */ | |
607 | drv1 = find_hdev_driver(filename); | |
608 | if (drv1) { | |
609 | return drv1; | |
610 | } | |
611 | ||
98289620 | 612 | if (!path_has_protocol(filename) || !allow_protocol_prefix) { |
39508e7a | 613 | return bdrv_find_format("file"); |
84a12e66 | 614 | } |
98289620 | 615 | |
9e0b22f4 SH |
616 | p = strchr(filename, ':'); |
617 | assert(p != NULL); | |
1cec71e3 AL |
618 | len = p - filename; |
619 | if (len > sizeof(protocol) - 1) | |
620 | len = sizeof(protocol) - 1; | |
621 | memcpy(protocol, filename, len); | |
622 | protocol[len] = '\0'; | |
8a22f02a | 623 | QLIST_FOREACH(drv1, &bdrv_drivers, list) { |
5fafdf24 | 624 | if (drv1->protocol_name && |
8a22f02a | 625 | !strcmp(drv1->protocol_name, protocol)) { |
83f64091 | 626 | return drv1; |
8a22f02a | 627 | } |
83f64091 FB |
628 | } |
629 | return NULL; | |
630 | } | |
631 | ||
f500a6d3 | 632 | static int find_image_format(BlockDriverState *bs, const char *filename, |
34b5d2c6 | 633 | BlockDriver **pdrv, Error **errp) |
f3a5d3f8 | 634 | { |
f500a6d3 | 635 | int score, score_max; |
f3a5d3f8 CH |
636 | BlockDriver *drv1, *drv; |
637 | uint8_t buf[2048]; | |
f500a6d3 | 638 | int ret = 0; |
f8ea0b00 | 639 | |
08a00559 | 640 | /* Return the raw BlockDriver * to scsi-generic devices or empty drives */ |
8e895599 | 641 | if (bs->sg || !bdrv_is_inserted(bs) || bdrv_getlength(bs) == 0) { |
c98ac35d SW |
642 | drv = bdrv_find_format("raw"); |
643 | if (!drv) { | |
34b5d2c6 | 644 | error_setg(errp, "Could not find raw image format"); |
c98ac35d SW |
645 | ret = -ENOENT; |
646 | } | |
647 | *pdrv = drv; | |
648 | return ret; | |
1a396859 | 649 | } |
f8ea0b00 | 650 | |
83f64091 | 651 | ret = bdrv_pread(bs, 0, buf, sizeof(buf)); |
83f64091 | 652 | if (ret < 0) { |
34b5d2c6 HR |
653 | error_setg_errno(errp, -ret, "Could not read image for determining its " |
654 | "format"); | |
c98ac35d SW |
655 | *pdrv = NULL; |
656 | return ret; | |
83f64091 FB |
657 | } |
658 | ||
ea2384d3 | 659 | score_max = 0; |
84a12e66 | 660 | drv = NULL; |
8a22f02a | 661 | QLIST_FOREACH(drv1, &bdrv_drivers, list) { |
83f64091 FB |
662 | if (drv1->bdrv_probe) { |
663 | score = drv1->bdrv_probe(buf, ret, filename); | |
664 | if (score > score_max) { | |
665 | score_max = score; | |
666 | drv = drv1; | |
667 | } | |
0849bf08 | 668 | } |
fc01f7e7 | 669 | } |
c98ac35d | 670 | if (!drv) { |
34b5d2c6 HR |
671 | error_setg(errp, "Could not determine image format: No compatible " |
672 | "driver found"); | |
c98ac35d SW |
673 | ret = -ENOENT; |
674 | } | |
675 | *pdrv = drv; | |
676 | return ret; | |
ea2384d3 FB |
677 | } |
678 | ||
51762288 SH |
679 | /** |
680 | * Set the current 'total_sectors' value | |
681 | */ | |
682 | static int refresh_total_sectors(BlockDriverState *bs, int64_t hint) | |
683 | { | |
684 | BlockDriver *drv = bs->drv; | |
685 | ||
396759ad NB |
686 | /* Do not attempt drv->bdrv_getlength() on scsi-generic devices */ |
687 | if (bs->sg) | |
688 | return 0; | |
689 | ||
51762288 SH |
690 | /* query actual device if possible, otherwise just trust the hint */ |
691 | if (drv->bdrv_getlength) { | |
692 | int64_t length = drv->bdrv_getlength(bs); | |
693 | if (length < 0) { | |
694 | return length; | |
695 | } | |
7e382003 | 696 | hint = DIV_ROUND_UP(length, BDRV_SECTOR_SIZE); |
51762288 SH |
697 | } |
698 | ||
699 | bs->total_sectors = hint; | |
700 | return 0; | |
701 | } | |
702 | ||
9e8f1835 PB |
703 | /** |
704 | * Set open flags for a given discard mode | |
705 | * | |
706 | * Return 0 on success, -1 if the discard mode was invalid. | |
707 | */ | |
708 | int bdrv_parse_discard_flags(const char *mode, int *flags) | |
709 | { | |
710 | *flags &= ~BDRV_O_UNMAP; | |
711 | ||
712 | if (!strcmp(mode, "off") || !strcmp(mode, "ignore")) { | |
713 | /* do nothing */ | |
714 | } else if (!strcmp(mode, "on") || !strcmp(mode, "unmap")) { | |
715 | *flags |= BDRV_O_UNMAP; | |
716 | } else { | |
717 | return -1; | |
718 | } | |
719 | ||
720 | return 0; | |
721 | } | |
722 | ||
c3993cdc SH |
723 | /** |
724 | * Set open flags for a given cache mode | |
725 | * | |
726 | * Return 0 on success, -1 if the cache mode was invalid. | |
727 | */ | |
728 | int bdrv_parse_cache_flags(const char *mode, int *flags) | |
729 | { | |
730 | *flags &= ~BDRV_O_CACHE_MASK; | |
731 | ||
732 | if (!strcmp(mode, "off") || !strcmp(mode, "none")) { | |
733 | *flags |= BDRV_O_NOCACHE | BDRV_O_CACHE_WB; | |
92196b2f SH |
734 | } else if (!strcmp(mode, "directsync")) { |
735 | *flags |= BDRV_O_NOCACHE; | |
c3993cdc SH |
736 | } else if (!strcmp(mode, "writeback")) { |
737 | *flags |= BDRV_O_CACHE_WB; | |
738 | } else if (!strcmp(mode, "unsafe")) { | |
739 | *flags |= BDRV_O_CACHE_WB; | |
740 | *flags |= BDRV_O_NO_FLUSH; | |
741 | } else if (!strcmp(mode, "writethrough")) { | |
742 | /* this is the default */ | |
743 | } else { | |
744 | return -1; | |
745 | } | |
746 | ||
747 | return 0; | |
748 | } | |
749 | ||
53fec9d3 SH |
750 | /** |
751 | * The copy-on-read flag is actually a reference count so multiple users may | |
752 | * use the feature without worrying about clobbering its previous state. | |
753 | * Copy-on-read stays enabled until all users have called to disable it. | |
754 | */ | |
755 | void bdrv_enable_copy_on_read(BlockDriverState *bs) | |
756 | { | |
757 | bs->copy_on_read++; | |
758 | } | |
759 | ||
760 | void bdrv_disable_copy_on_read(BlockDriverState *bs) | |
761 | { | |
762 | assert(bs->copy_on_read > 0); | |
763 | bs->copy_on_read--; | |
764 | } | |
765 | ||
7b272452 KW |
766 | static int bdrv_open_flags(BlockDriverState *bs, int flags) |
767 | { | |
768 | int open_flags = flags | BDRV_O_CACHE_WB; | |
769 | ||
b998875d KW |
770 | /* The backing file of a temporary snapshot is read-only */ |
771 | if (flags & BDRV_O_SNAPSHOT) { | |
772 | open_flags &= ~BDRV_O_RDWR; | |
773 | } | |
774 | ||
7b272452 KW |
775 | /* |
776 | * Clear flags that are internal to the block layer before opening the | |
777 | * image. | |
778 | */ | |
779 | open_flags &= ~(BDRV_O_SNAPSHOT | BDRV_O_NO_BACKING); | |
780 | ||
781 | /* | |
782 | * Snapshots should be writable. | |
783 | */ | |
784 | if (bs->is_temporary) { | |
785 | open_flags |= BDRV_O_RDWR; | |
786 | } | |
787 | ||
788 | return open_flags; | |
789 | } | |
790 | ||
6913c0c2 BC |
791 | static int bdrv_assign_node_name(BlockDriverState *bs, |
792 | const char *node_name, | |
793 | Error **errp) | |
794 | { | |
795 | if (!node_name) { | |
796 | return 0; | |
797 | } | |
798 | ||
799 | /* empty string node name is invalid */ | |
800 | if (node_name[0] == '\0') { | |
801 | error_setg(errp, "Empty node name"); | |
802 | return -EINVAL; | |
803 | } | |
804 | ||
0c5e94ee BC |
805 | /* takes care of avoiding namespaces collisions */ |
806 | if (bdrv_find(node_name)) { | |
807 | error_setg(errp, "node-name=%s is conflicting with a device id", | |
808 | node_name); | |
809 | return -EINVAL; | |
810 | } | |
811 | ||
6913c0c2 BC |
812 | /* takes care of avoiding duplicates node names */ |
813 | if (bdrv_find_node(node_name)) { | |
814 | error_setg(errp, "Duplicate node name"); | |
815 | return -EINVAL; | |
816 | } | |
817 | ||
818 | /* copy node name into the bs and insert it into the graph list */ | |
819 | pstrcpy(bs->node_name, sizeof(bs->node_name), node_name); | |
820 | QTAILQ_INSERT_TAIL(&graph_bdrv_states, bs, node_list); | |
821 | ||
822 | return 0; | |
823 | } | |
824 | ||
57915332 KW |
825 | /* |
826 | * Common part for opening disk images and files | |
b6ad491a KW |
827 | * |
828 | * Removes all processed options from *options. | |
57915332 | 829 | */ |
f500a6d3 | 830 | static int bdrv_open_common(BlockDriverState *bs, BlockDriverState *file, |
34b5d2c6 | 831 | QDict *options, int flags, BlockDriver *drv, Error **errp) |
57915332 KW |
832 | { |
833 | int ret, open_flags; | |
035fccdf | 834 | const char *filename; |
6913c0c2 | 835 | const char *node_name = NULL; |
34b5d2c6 | 836 | Error *local_err = NULL; |
57915332 KW |
837 | |
838 | assert(drv != NULL); | |
6405875c | 839 | assert(bs->file == NULL); |
707ff828 | 840 | assert(options != NULL && bs->options != options); |
57915332 | 841 | |
45673671 KW |
842 | if (file != NULL) { |
843 | filename = file->filename; | |
844 | } else { | |
845 | filename = qdict_get_try_str(options, "filename"); | |
846 | } | |
847 | ||
765003db KW |
848 | if (drv->bdrv_needs_filename && !filename) { |
849 | error_setg(errp, "The '%s' block driver requires a file name", | |
850 | drv->format_name); | |
851 | return -EINVAL; | |
852 | } | |
853 | ||
45673671 | 854 | trace_bdrv_open_common(bs, filename ?: "", flags, drv->format_name); |
28dcee10 | 855 | |
6913c0c2 BC |
856 | node_name = qdict_get_try_str(options, "node-name"); |
857 | ret = bdrv_assign_node_name(bs, node_name, errp); | |
858 | if (ret < 0) { | |
859 | return ret; | |
860 | } | |
861 | qdict_del(options, "node-name"); | |
862 | ||
5d186eb0 KW |
863 | /* bdrv_open() with directly using a protocol as drv. This layer is already |
864 | * opened, so assign it to bs (while file becomes a closed BlockDriverState) | |
865 | * and return immediately. */ | |
866 | if (file != NULL && drv->bdrv_file_open) { | |
867 | bdrv_swap(file, bs); | |
868 | return 0; | |
869 | } | |
870 | ||
57915332 | 871 | bs->open_flags = flags; |
1b7fd729 | 872 | bs->guest_block_size = 512; |
c25f53b0 | 873 | bs->request_alignment = 512; |
0d51b4de | 874 | bs->zero_beyond_eof = true; |
b64ec4e4 FZ |
875 | open_flags = bdrv_open_flags(bs, flags); |
876 | bs->read_only = !(open_flags & BDRV_O_RDWR); | |
877 | ||
878 | if (use_bdrv_whitelist && !bdrv_is_whitelisted(drv, bs->read_only)) { | |
8f94a6e4 KW |
879 | error_setg(errp, |
880 | !bs->read_only && bdrv_is_whitelisted(drv, true) | |
881 | ? "Driver '%s' can only be used for read-only devices" | |
882 | : "Driver '%s' is not whitelisted", | |
883 | drv->format_name); | |
b64ec4e4 FZ |
884 | return -ENOTSUP; |
885 | } | |
57915332 | 886 | |
53fec9d3 | 887 | assert(bs->copy_on_read == 0); /* bdrv_new() and bdrv_close() make it so */ |
0ebd24e0 KW |
888 | if (flags & BDRV_O_COPY_ON_READ) { |
889 | if (!bs->read_only) { | |
890 | bdrv_enable_copy_on_read(bs); | |
891 | } else { | |
892 | error_setg(errp, "Can't use copy-on-read on read-only device"); | |
893 | return -EINVAL; | |
894 | } | |
53fec9d3 SH |
895 | } |
896 | ||
c2ad1b0c KW |
897 | if (filename != NULL) { |
898 | pstrcpy(bs->filename, sizeof(bs->filename), filename); | |
899 | } else { | |
900 | bs->filename[0] = '\0'; | |
901 | } | |
57915332 | 902 | |
57915332 | 903 | bs->drv = drv; |
7267c094 | 904 | bs->opaque = g_malloc0(drv->instance_size); |
57915332 | 905 | |
03f541bd | 906 | bs->enable_write_cache = !!(flags & BDRV_O_CACHE_WB); |
e7c63796 | 907 | |
66f82cee KW |
908 | /* Open the image, either directly or using a protocol */ |
909 | if (drv->bdrv_file_open) { | |
5d186eb0 | 910 | assert(file == NULL); |
030be321 | 911 | assert(!drv->bdrv_needs_filename || filename != NULL); |
34b5d2c6 | 912 | ret = drv->bdrv_file_open(bs, options, open_flags, &local_err); |
f500a6d3 | 913 | } else { |
2af5ef70 | 914 | if (file == NULL) { |
34b5d2c6 HR |
915 | error_setg(errp, "Can't use '%s' as a block driver for the " |
916 | "protocol level", drv->format_name); | |
2af5ef70 KW |
917 | ret = -EINVAL; |
918 | goto free_and_fail; | |
919 | } | |
f500a6d3 | 920 | bs->file = file; |
34b5d2c6 | 921 | ret = drv->bdrv_open(bs, options, open_flags, &local_err); |
66f82cee KW |
922 | } |
923 | ||
57915332 | 924 | if (ret < 0) { |
84d18f06 | 925 | if (local_err) { |
34b5d2c6 | 926 | error_propagate(errp, local_err); |
2fa9aa59 DH |
927 | } else if (bs->filename[0]) { |
928 | error_setg_errno(errp, -ret, "Could not open '%s'", bs->filename); | |
34b5d2c6 HR |
929 | } else { |
930 | error_setg_errno(errp, -ret, "Could not open image"); | |
931 | } | |
57915332 KW |
932 | goto free_and_fail; |
933 | } | |
934 | ||
51762288 SH |
935 | ret = refresh_total_sectors(bs, bs->total_sectors); |
936 | if (ret < 0) { | |
34b5d2c6 | 937 | error_setg_errno(errp, -ret, "Could not refresh total sector count"); |
51762288 | 938 | goto free_and_fail; |
57915332 | 939 | } |
51762288 | 940 | |
d34682cd | 941 | bdrv_refresh_limits(bs); |
c25f53b0 | 942 | assert(bdrv_opt_mem_align(bs) != 0); |
47ea2de2 | 943 | assert((bs->request_alignment != 0) || bs->sg); |
d34682cd | 944 | |
57915332 KW |
945 | #ifndef _WIN32 |
946 | if (bs->is_temporary) { | |
d4cea8df DH |
947 | assert(bs->filename[0] != '\0'); |
948 | unlink(bs->filename); | |
57915332 KW |
949 | } |
950 | #endif | |
951 | return 0; | |
952 | ||
953 | free_and_fail: | |
f500a6d3 | 954 | bs->file = NULL; |
7267c094 | 955 | g_free(bs->opaque); |
57915332 KW |
956 | bs->opaque = NULL; |
957 | bs->drv = NULL; | |
958 | return ret; | |
959 | } | |
960 | ||
b6ce07aa KW |
961 | /* |
962 | * Opens a file using a protocol (file, host_device, nbd, ...) | |
787e4a85 | 963 | * |
5acd9d81 HR |
964 | * options is an indirect pointer to a QDict of options to pass to the block |
965 | * drivers, or pointer to NULL for an empty set of options. If this function | |
966 | * takes ownership of the QDict reference, it will set *options to NULL; | |
967 | * otherwise, it will contain unused/unrecognized options after this function | |
968 | * returns. Then, the caller is responsible for freeing it. If it intends to | |
969 | * reuse the QDict, QINCREF() should be called beforehand. | |
b6ce07aa | 970 | */ |
d4446eae | 971 | static int bdrv_file_open(BlockDriverState *bs, const char *filename, |
5acd9d81 | 972 | QDict **options, int flags, Error **errp) |
ea2384d3 | 973 | { |
6db95603 | 974 | BlockDriver *drv; |
c2ad1b0c | 975 | const char *drvname; |
e3fa4bfa | 976 | bool parse_filename = false; |
34b5d2c6 | 977 | Error *local_err = NULL; |
83f64091 FB |
978 | int ret; |
979 | ||
035fccdf KW |
980 | /* Fetch the file name from the options QDict if necessary */ |
981 | if (!filename) { | |
5acd9d81 HR |
982 | filename = qdict_get_try_str(*options, "filename"); |
983 | } else if (filename && !qdict_haskey(*options, "filename")) { | |
984 | qdict_put(*options, "filename", qstring_from_str(filename)); | |
e3fa4bfa | 985 | parse_filename = true; |
035fccdf | 986 | } else { |
34b5d2c6 HR |
987 | error_setg(errp, "Can't specify 'file' and 'filename' options at the " |
988 | "same time"); | |
035fccdf KW |
989 | ret = -EINVAL; |
990 | goto fail; | |
991 | } | |
992 | ||
c2ad1b0c | 993 | /* Find the right block driver */ |
5acd9d81 | 994 | drvname = qdict_get_try_str(*options, "driver"); |
c2ad1b0c | 995 | if (drvname) { |
8f94a6e4 | 996 | drv = bdrv_find_format(drvname); |
34b5d2c6 HR |
997 | if (!drv) { |
998 | error_setg(errp, "Unknown driver '%s'", drvname); | |
999 | } | |
5acd9d81 | 1000 | qdict_del(*options, "driver"); |
c2ad1b0c | 1001 | } else if (filename) { |
e3fa4bfa | 1002 | drv = bdrv_find_protocol(filename, parse_filename); |
98289620 | 1003 | if (!drv) { |
34b5d2c6 | 1004 | error_setg(errp, "Unknown protocol"); |
98289620 | 1005 | } |
c2ad1b0c | 1006 | } else { |
34b5d2c6 | 1007 | error_setg(errp, "Must specify either driver or file"); |
c2ad1b0c KW |
1008 | drv = NULL; |
1009 | } | |
1010 | ||
1011 | if (!drv) { | |
34b5d2c6 | 1012 | /* errp has been set already */ |
c2ad1b0c KW |
1013 | ret = -ENOENT; |
1014 | goto fail; | |
1015 | } | |
1016 | ||
1017 | /* Parse the filename and open it */ | |
e3fa4bfa | 1018 | if (drv->bdrv_parse_filename && parse_filename) { |
5acd9d81 | 1019 | drv->bdrv_parse_filename(filename, *options, &local_err); |
84d18f06 | 1020 | if (local_err) { |
34b5d2c6 | 1021 | error_propagate(errp, local_err); |
6963a30d KW |
1022 | ret = -EINVAL; |
1023 | goto fail; | |
1024 | } | |
cd5d031e HR |
1025 | |
1026 | if (!drv->bdrv_needs_filename) { | |
1027 | qdict_del(*options, "filename"); | |
1028 | } else { | |
1029 | filename = qdict_get_str(*options, "filename"); | |
1030 | } | |
6963a30d KW |
1031 | } |
1032 | ||
505d7583 | 1033 | if (!drv->bdrv_file_open) { |
5acd9d81 HR |
1034 | ret = bdrv_open(&bs, filename, NULL, *options, flags, drv, &local_err); |
1035 | *options = NULL; | |
505d7583 | 1036 | } else { |
5acd9d81 | 1037 | ret = bdrv_open_common(bs, NULL, *options, flags, drv, &local_err); |
505d7583 | 1038 | } |
83f64091 | 1039 | if (ret < 0) { |
34b5d2c6 | 1040 | error_propagate(errp, local_err); |
707ff828 KW |
1041 | goto fail; |
1042 | } | |
1043 | ||
71d0770c | 1044 | bs->growable = 1; |
83f64091 | 1045 | return 0; |
707ff828 KW |
1046 | |
1047 | fail: | |
707ff828 | 1048 | return ret; |
83f64091 FB |
1049 | } |
1050 | ||
31ca6d07 KW |
1051 | /* |
1052 | * Opens the backing file for a BlockDriverState if not yet open | |
1053 | * | |
1054 | * options is a QDict of options to pass to the block drivers, or NULL for an | |
1055 | * empty set of options. The reference to the QDict is transferred to this | |
1056 | * function (even on failure), so if the caller intends to reuse the dictionary, | |
1057 | * it needs to use QINCREF() before calling bdrv_file_open. | |
1058 | */ | |
34b5d2c6 | 1059 | int bdrv_open_backing_file(BlockDriverState *bs, QDict *options, Error **errp) |
9156df12 PB |
1060 | { |
1061 | char backing_filename[PATH_MAX]; | |
1062 | int back_flags, ret; | |
1063 | BlockDriver *back_drv = NULL; | |
34b5d2c6 | 1064 | Error *local_err = NULL; |
9156df12 PB |
1065 | |
1066 | if (bs->backing_hd != NULL) { | |
31ca6d07 | 1067 | QDECREF(options); |
9156df12 PB |
1068 | return 0; |
1069 | } | |
1070 | ||
31ca6d07 KW |
1071 | /* NULL means an empty set of options */ |
1072 | if (options == NULL) { | |
1073 | options = qdict_new(); | |
1074 | } | |
1075 | ||
9156df12 | 1076 | bs->open_flags &= ~BDRV_O_NO_BACKING; |
1cb6f506 KW |
1077 | if (qdict_haskey(options, "file.filename")) { |
1078 | backing_filename[0] = '\0'; | |
1079 | } else if (bs->backing_file[0] == '\0' && qdict_size(options) == 0) { | |
31ca6d07 | 1080 | QDECREF(options); |
9156df12 | 1081 | return 0; |
dbecebdd FZ |
1082 | } else { |
1083 | bdrv_get_full_backing_filename(bs, backing_filename, | |
1084 | sizeof(backing_filename)); | |
9156df12 PB |
1085 | } |
1086 | ||
9156df12 PB |
1087 | if (bs->backing_format[0] != '\0') { |
1088 | back_drv = bdrv_find_format(bs->backing_format); | |
1089 | } | |
1090 | ||
1091 | /* backing files always opened read-only */ | |
87a5debd TL |
1092 | back_flags = bs->open_flags & ~(BDRV_O_RDWR | BDRV_O_SNAPSHOT | |
1093 | BDRV_O_COPY_ON_READ); | |
9156df12 | 1094 | |
f67503e5 HR |
1095 | assert(bs->backing_hd == NULL); |
1096 | ret = bdrv_open(&bs->backing_hd, | |
ddf5636d | 1097 | *backing_filename ? backing_filename : NULL, NULL, options, |
34b5d2c6 | 1098 | back_flags, back_drv, &local_err); |
9156df12 | 1099 | if (ret < 0) { |
9156df12 PB |
1100 | bs->backing_hd = NULL; |
1101 | bs->open_flags |= BDRV_O_NO_BACKING; | |
b04b6b6e FZ |
1102 | error_setg(errp, "Could not open backing file: %s", |
1103 | error_get_pretty(local_err)); | |
1104 | error_free(local_err); | |
9156df12 PB |
1105 | return ret; |
1106 | } | |
d80ac658 PF |
1107 | |
1108 | if (bs->backing_hd->file) { | |
1109 | pstrcpy(bs->backing_file, sizeof(bs->backing_file), | |
1110 | bs->backing_hd->file->filename); | |
1111 | } | |
1112 | ||
d34682cd KW |
1113 | /* Recalculate the BlockLimits with the backing file */ |
1114 | bdrv_refresh_limits(bs); | |
1115 | ||
9156df12 PB |
1116 | return 0; |
1117 | } | |
1118 | ||
da557aac HR |
1119 | /* |
1120 | * Opens a disk image whose options are given as BlockdevRef in another block | |
1121 | * device's options. | |
1122 | * | |
da557aac HR |
1123 | * If allow_none is true, no image will be opened if filename is false and no |
1124 | * BlockdevRef is given. *pbs will remain unchanged and 0 will be returned. | |
1125 | * | |
1126 | * bdrev_key specifies the key for the image's BlockdevRef in the options QDict. | |
1127 | * That QDict has to be flattened; therefore, if the BlockdevRef is a QDict | |
1128 | * itself, all options starting with "${bdref_key}." are considered part of the | |
1129 | * BlockdevRef. | |
1130 | * | |
1131 | * The BlockdevRef will be removed from the options QDict. | |
f67503e5 HR |
1132 | * |
1133 | * To conform with the behavior of bdrv_open(), *pbs has to be NULL. | |
da557aac HR |
1134 | */ |
1135 | int bdrv_open_image(BlockDriverState **pbs, const char *filename, | |
1136 | QDict *options, const char *bdref_key, int flags, | |
f7d9fd8c | 1137 | bool allow_none, Error **errp) |
da557aac HR |
1138 | { |
1139 | QDict *image_options; | |
1140 | int ret; | |
1141 | char *bdref_key_dot; | |
1142 | const char *reference; | |
1143 | ||
f67503e5 HR |
1144 | assert(pbs); |
1145 | assert(*pbs == NULL); | |
1146 | ||
da557aac HR |
1147 | bdref_key_dot = g_strdup_printf("%s.", bdref_key); |
1148 | qdict_extract_subqdict(options, &image_options, bdref_key_dot); | |
1149 | g_free(bdref_key_dot); | |
1150 | ||
1151 | reference = qdict_get_try_str(options, bdref_key); | |
1152 | if (!filename && !reference && !qdict_size(image_options)) { | |
1153 | if (allow_none) { | |
1154 | ret = 0; | |
1155 | } else { | |
1156 | error_setg(errp, "A block device must be specified for \"%s\"", | |
1157 | bdref_key); | |
1158 | ret = -EINVAL; | |
1159 | } | |
1160 | goto done; | |
1161 | } | |
1162 | ||
f7d9fd8c | 1163 | ret = bdrv_open(pbs, filename, reference, image_options, flags, NULL, errp); |
da557aac HR |
1164 | |
1165 | done: | |
1166 | qdict_del(options, bdref_key); | |
1167 | return ret; | |
1168 | } | |
1169 | ||
b998875d KW |
1170 | void bdrv_append_temp_snapshot(BlockDriverState *bs, Error **errp) |
1171 | { | |
1172 | /* TODO: extra byte is a hack to ensure MAX_PATH space on Windows. */ | |
1173 | char tmp_filename[PATH_MAX + 1]; | |
1174 | ||
1175 | int64_t total_size; | |
1176 | BlockDriver *bdrv_qcow2; | |
1177 | QEMUOptionParameter *create_options; | |
1178 | QDict *snapshot_options; | |
1179 | BlockDriverState *bs_snapshot; | |
1180 | Error *local_err; | |
1181 | int ret; | |
1182 | ||
1183 | /* if snapshot, we create a temporary backing file and open it | |
1184 | instead of opening 'filename' directly */ | |
1185 | ||
1186 | /* Get the required size from the image */ | |
f187743a KW |
1187 | total_size = bdrv_getlength(bs); |
1188 | if (total_size < 0) { | |
1189 | error_setg_errno(errp, -total_size, "Could not get image size"); | |
1190 | return; | |
1191 | } | |
1192 | total_size &= BDRV_SECTOR_MASK; | |
b998875d KW |
1193 | |
1194 | /* Create the temporary image */ | |
1195 | ret = get_tmp_filename(tmp_filename, sizeof(tmp_filename)); | |
1196 | if (ret < 0) { | |
1197 | error_setg_errno(errp, -ret, "Could not get temporary filename"); | |
1198 | return; | |
1199 | } | |
1200 | ||
1201 | bdrv_qcow2 = bdrv_find_format("qcow2"); | |
1202 | create_options = parse_option_parameters("", bdrv_qcow2->create_options, | |
1203 | NULL); | |
1204 | ||
1205 | set_option_parameter_int(create_options, BLOCK_OPT_SIZE, total_size); | |
1206 | ||
1207 | ret = bdrv_create(bdrv_qcow2, tmp_filename, create_options, &local_err); | |
1208 | free_option_parameters(create_options); | |
1209 | if (ret < 0) { | |
1210 | error_setg_errno(errp, -ret, "Could not create temporary overlay " | |
1211 | "'%s': %s", tmp_filename, | |
1212 | error_get_pretty(local_err)); | |
1213 | error_free(local_err); | |
1214 | return; | |
1215 | } | |
1216 | ||
1217 | /* Prepare a new options QDict for the temporary file */ | |
1218 | snapshot_options = qdict_new(); | |
1219 | qdict_put(snapshot_options, "file.driver", | |
1220 | qstring_from_str("file")); | |
1221 | qdict_put(snapshot_options, "file.filename", | |
1222 | qstring_from_str(tmp_filename)); | |
1223 | ||
1224 | bs_snapshot = bdrv_new(""); | |
1225 | bs_snapshot->is_temporary = 1; | |
1226 | ||
1227 | ret = bdrv_open(&bs_snapshot, NULL, NULL, snapshot_options, | |
1228 | bs->open_flags & ~BDRV_O_SNAPSHOT, bdrv_qcow2, &local_err); | |
1229 | if (ret < 0) { | |
1230 | error_propagate(errp, local_err); | |
1231 | return; | |
1232 | } | |
1233 | ||
1234 | bdrv_append(bs_snapshot, bs); | |
1235 | } | |
1236 | ||
b6ce07aa KW |
1237 | /* |
1238 | * Opens a disk image (raw, qcow2, vmdk, ...) | |
de9c0cec KW |
1239 | * |
1240 | * options is a QDict of options to pass to the block drivers, or NULL for an | |
1241 | * empty set of options. The reference to the QDict belongs to the block layer | |
1242 | * after the call (even on failure), so if the caller intends to reuse the | |
1243 | * dictionary, it needs to use QINCREF() before calling bdrv_open. | |
f67503e5 HR |
1244 | * |
1245 | * If *pbs is NULL, a new BDS will be created with a pointer to it stored there. | |
1246 | * If it is not NULL, the referenced BDS will be reused. | |
ddf5636d HR |
1247 | * |
1248 | * The reference parameter may be used to specify an existing block device which | |
1249 | * should be opened. If specified, neither options nor a filename may be given, | |
1250 | * nor can an existing BDS be reused (that is, *pbs has to be NULL). | |
b6ce07aa | 1251 | */ |
ddf5636d HR |
1252 | int bdrv_open(BlockDriverState **pbs, const char *filename, |
1253 | const char *reference, QDict *options, int flags, | |
1254 | BlockDriver *drv, Error **errp) | |
ea2384d3 | 1255 | { |
b6ce07aa | 1256 | int ret; |
f67503e5 | 1257 | BlockDriverState *file = NULL, *bs; |
74fe54f2 | 1258 | const char *drvname; |
34b5d2c6 | 1259 | Error *local_err = NULL; |
712e7874 | 1260 | |
f67503e5 HR |
1261 | assert(pbs); |
1262 | ||
ddf5636d HR |
1263 | if (reference) { |
1264 | bool options_non_empty = options ? qdict_size(options) : false; | |
1265 | QDECREF(options); | |
1266 | ||
1267 | if (*pbs) { | |
1268 | error_setg(errp, "Cannot reuse an existing BDS when referencing " | |
1269 | "another block device"); | |
1270 | return -EINVAL; | |
1271 | } | |
1272 | ||
1273 | if (filename || options_non_empty) { | |
1274 | error_setg(errp, "Cannot reference an existing block device with " | |
1275 | "additional options or a new filename"); | |
1276 | return -EINVAL; | |
1277 | } | |
1278 | ||
1279 | bs = bdrv_lookup_bs(reference, reference, errp); | |
1280 | if (!bs) { | |
1281 | return -ENODEV; | |
1282 | } | |
1283 | bdrv_ref(bs); | |
1284 | *pbs = bs; | |
1285 | return 0; | |
1286 | } | |
1287 | ||
f67503e5 HR |
1288 | if (*pbs) { |
1289 | bs = *pbs; | |
1290 | } else { | |
1291 | bs = bdrv_new(""); | |
1292 | } | |
1293 | ||
de9c0cec KW |
1294 | /* NULL means an empty set of options */ |
1295 | if (options == NULL) { | |
1296 | options = qdict_new(); | |
1297 | } | |
1298 | ||
1299 | bs->options = options; | |
b6ad491a | 1300 | options = qdict_clone_shallow(options); |
de9c0cec | 1301 | |
5469a2a6 HR |
1302 | if (flags & BDRV_O_PROTOCOL) { |
1303 | assert(!drv); | |
5acd9d81 | 1304 | ret = bdrv_file_open(bs, filename, &options, flags & ~BDRV_O_PROTOCOL, |
5469a2a6 | 1305 | &local_err); |
5469a2a6 | 1306 | if (!ret) { |
eb909c7f | 1307 | drv = bs->drv; |
5acd9d81 | 1308 | goto done; |
5469a2a6 HR |
1309 | } else if (bs->drv) { |
1310 | goto close_and_fail; | |
1311 | } else { | |
1312 | goto fail; | |
1313 | } | |
1314 | } | |
1315 | ||
f500a6d3 KW |
1316 | /* Open image file without format layer */ |
1317 | if (flags & BDRV_O_RDWR) { | |
1318 | flags |= BDRV_O_ALLOW_RDWR; | |
1319 | } | |
1320 | ||
f67503e5 | 1321 | assert(file == NULL); |
054963f8 | 1322 | ret = bdrv_open_image(&file, filename, options, "file", |
f7d9fd8c HR |
1323 | bdrv_open_flags(bs, flags | BDRV_O_UNMAP) | |
1324 | BDRV_O_PROTOCOL, true, &local_err); | |
054963f8 | 1325 | if (ret < 0) { |
9562f69c | 1326 | goto unlink_and_fail; |
f500a6d3 KW |
1327 | } |
1328 | ||
b6ce07aa | 1329 | /* Find the right image format driver */ |
74fe54f2 KW |
1330 | drvname = qdict_get_try_str(options, "driver"); |
1331 | if (drvname) { | |
8f94a6e4 | 1332 | drv = bdrv_find_format(drvname); |
74fe54f2 | 1333 | qdict_del(options, "driver"); |
06d22aa3 KW |
1334 | if (!drv) { |
1335 | error_setg(errp, "Invalid driver: '%s'", drvname); | |
1336 | ret = -EINVAL; | |
1337 | goto unlink_and_fail; | |
1338 | } | |
74fe54f2 KW |
1339 | } |
1340 | ||
6db95603 | 1341 | if (!drv) { |
2a05cbe4 HR |
1342 | if (file) { |
1343 | ret = find_image_format(file, filename, &drv, &local_err); | |
1344 | } else { | |
1345 | error_setg(errp, "Must specify either driver or file"); | |
1346 | ret = -EINVAL; | |
1347 | goto unlink_and_fail; | |
1348 | } | |
51d7c00c | 1349 | } |
6987307c | 1350 | |
51d7c00c | 1351 | if (!drv) { |
51d7c00c | 1352 | goto unlink_and_fail; |
ea2384d3 | 1353 | } |
b6ce07aa KW |
1354 | |
1355 | /* Open the image */ | |
34b5d2c6 | 1356 | ret = bdrv_open_common(bs, file, options, flags, drv, &local_err); |
b6ce07aa | 1357 | if (ret < 0) { |
6987307c CH |
1358 | goto unlink_and_fail; |
1359 | } | |
1360 | ||
2a05cbe4 | 1361 | if (file && (bs->file != file)) { |
4f6fd349 | 1362 | bdrv_unref(file); |
f500a6d3 KW |
1363 | file = NULL; |
1364 | } | |
1365 | ||
b6ce07aa | 1366 | /* If there is a backing file, use it */ |
9156df12 | 1367 | if ((flags & BDRV_O_NO_BACKING) == 0) { |
31ca6d07 KW |
1368 | QDict *backing_options; |
1369 | ||
5726d872 | 1370 | qdict_extract_subqdict(options, &backing_options, "backing."); |
34b5d2c6 | 1371 | ret = bdrv_open_backing_file(bs, backing_options, &local_err); |
b6ce07aa | 1372 | if (ret < 0) { |
b6ad491a | 1373 | goto close_and_fail; |
b6ce07aa | 1374 | } |
b6ce07aa KW |
1375 | } |
1376 | ||
b998875d KW |
1377 | /* For snapshot=on, create a temporary qcow2 overlay. bs points to the |
1378 | * temporary snapshot afterwards. */ | |
1379 | if (flags & BDRV_O_SNAPSHOT) { | |
1380 | bdrv_append_temp_snapshot(bs, &local_err); | |
1381 | if (local_err) { | |
1382 | error_propagate(errp, local_err); | |
1383 | goto close_and_fail; | |
1384 | } | |
1385 | } | |
1386 | ||
1387 | ||
5acd9d81 | 1388 | done: |
b6ad491a | 1389 | /* Check if any unknown options were used */ |
5acd9d81 | 1390 | if (options && (qdict_size(options) != 0)) { |
b6ad491a | 1391 | const QDictEntry *entry = qdict_first(options); |
5acd9d81 HR |
1392 | if (flags & BDRV_O_PROTOCOL) { |
1393 | error_setg(errp, "Block protocol '%s' doesn't support the option " | |
1394 | "'%s'", drv->format_name, entry->key); | |
1395 | } else { | |
1396 | error_setg(errp, "Block format '%s' used by device '%s' doesn't " | |
1397 | "support the option '%s'", drv->format_name, | |
1398 | bs->device_name, entry->key); | |
1399 | } | |
b6ad491a KW |
1400 | |
1401 | ret = -EINVAL; | |
1402 | goto close_and_fail; | |
1403 | } | |
b6ad491a | 1404 | |
b6ce07aa | 1405 | if (!bdrv_key_required(bs)) { |
7d4b4ba5 | 1406 | bdrv_dev_change_media_cb(bs, true); |
c3adb58f MA |
1407 | } else if (!runstate_check(RUN_STATE_PRELAUNCH) |
1408 | && !runstate_check(RUN_STATE_INMIGRATE) | |
1409 | && !runstate_check(RUN_STATE_PAUSED)) { /* HACK */ | |
1410 | error_setg(errp, | |
1411 | "Guest must be stopped for opening of encrypted image"); | |
1412 | ret = -EBUSY; | |
1413 | goto close_and_fail; | |
b6ce07aa KW |
1414 | } |
1415 | ||
c3adb58f | 1416 | QDECREF(options); |
f67503e5 | 1417 | *pbs = bs; |
b6ce07aa KW |
1418 | return 0; |
1419 | ||
1420 | unlink_and_fail: | |
f500a6d3 | 1421 | if (file != NULL) { |
4f6fd349 | 1422 | bdrv_unref(file); |
f500a6d3 | 1423 | } |
b6ce07aa KW |
1424 | if (bs->is_temporary) { |
1425 | unlink(filename); | |
1426 | } | |
de9c0cec KW |
1427 | fail: |
1428 | QDECREF(bs->options); | |
b6ad491a | 1429 | QDECREF(options); |
de9c0cec | 1430 | bs->options = NULL; |
f67503e5 HR |
1431 | if (!*pbs) { |
1432 | /* If *pbs is NULL, a new BDS has been created in this function and | |
1433 | needs to be freed now. Otherwise, it does not need to be closed, | |
1434 | since it has not really been opened yet. */ | |
1435 | bdrv_unref(bs); | |
1436 | } | |
84d18f06 | 1437 | if (local_err) { |
34b5d2c6 HR |
1438 | error_propagate(errp, local_err); |
1439 | } | |
b6ad491a | 1440 | return ret; |
de9c0cec | 1441 | |
b6ad491a | 1442 | close_and_fail: |
f67503e5 HR |
1443 | /* See fail path, but now the BDS has to be always closed */ |
1444 | if (*pbs) { | |
1445 | bdrv_close(bs); | |
1446 | } else { | |
1447 | bdrv_unref(bs); | |
1448 | } | |
b6ad491a | 1449 | QDECREF(options); |
84d18f06 | 1450 | if (local_err) { |
34b5d2c6 HR |
1451 | error_propagate(errp, local_err); |
1452 | } | |
b6ce07aa KW |
1453 | return ret; |
1454 | } | |
1455 | ||
e971aa12 JC |
1456 | typedef struct BlockReopenQueueEntry { |
1457 | bool prepared; | |
1458 | BDRVReopenState state; | |
1459 | QSIMPLEQ_ENTRY(BlockReopenQueueEntry) entry; | |
1460 | } BlockReopenQueueEntry; | |
1461 | ||
1462 | /* | |
1463 | * Adds a BlockDriverState to a simple queue for an atomic, transactional | |
1464 | * reopen of multiple devices. | |
1465 | * | |
1466 | * bs_queue can either be an existing BlockReopenQueue that has had QSIMPLE_INIT | |
1467 | * already performed, or alternatively may be NULL a new BlockReopenQueue will | |
1468 | * be created and initialized. This newly created BlockReopenQueue should be | |
1469 | * passed back in for subsequent calls that are intended to be of the same | |
1470 | * atomic 'set'. | |
1471 | * | |
1472 | * bs is the BlockDriverState to add to the reopen queue. | |
1473 | * | |
1474 | * flags contains the open flags for the associated bs | |
1475 | * | |
1476 | * returns a pointer to bs_queue, which is either the newly allocated | |
1477 | * bs_queue, or the existing bs_queue being used. | |
1478 | * | |
1479 | */ | |
1480 | BlockReopenQueue *bdrv_reopen_queue(BlockReopenQueue *bs_queue, | |
1481 | BlockDriverState *bs, int flags) | |
1482 | { | |
1483 | assert(bs != NULL); | |
1484 | ||
1485 | BlockReopenQueueEntry *bs_entry; | |
1486 | if (bs_queue == NULL) { | |
1487 | bs_queue = g_new0(BlockReopenQueue, 1); | |
1488 | QSIMPLEQ_INIT(bs_queue); | |
1489 | } | |
1490 | ||
1491 | if (bs->file) { | |
1492 | bdrv_reopen_queue(bs_queue, bs->file, flags); | |
1493 | } | |
1494 | ||
1495 | bs_entry = g_new0(BlockReopenQueueEntry, 1); | |
1496 | QSIMPLEQ_INSERT_TAIL(bs_queue, bs_entry, entry); | |
1497 | ||
1498 | bs_entry->state.bs = bs; | |
1499 | bs_entry->state.flags = flags; | |
1500 | ||
1501 | return bs_queue; | |
1502 | } | |
1503 | ||
1504 | /* | |
1505 | * Reopen multiple BlockDriverStates atomically & transactionally. | |
1506 | * | |
1507 | * The queue passed in (bs_queue) must have been built up previous | |
1508 | * via bdrv_reopen_queue(). | |
1509 | * | |
1510 | * Reopens all BDS specified in the queue, with the appropriate | |
1511 | * flags. All devices are prepared for reopen, and failure of any | |
1512 | * device will cause all device changes to be abandonded, and intermediate | |
1513 | * data cleaned up. | |
1514 | * | |
1515 | * If all devices prepare successfully, then the changes are committed | |
1516 | * to all devices. | |
1517 | * | |
1518 | */ | |
1519 | int bdrv_reopen_multiple(BlockReopenQueue *bs_queue, Error **errp) | |
1520 | { | |
1521 | int ret = -1; | |
1522 | BlockReopenQueueEntry *bs_entry, *next; | |
1523 | Error *local_err = NULL; | |
1524 | ||
1525 | assert(bs_queue != NULL); | |
1526 | ||
1527 | bdrv_drain_all(); | |
1528 | ||
1529 | QSIMPLEQ_FOREACH(bs_entry, bs_queue, entry) { | |
1530 | if (bdrv_reopen_prepare(&bs_entry->state, bs_queue, &local_err)) { | |
1531 | error_propagate(errp, local_err); | |
1532 | goto cleanup; | |
1533 | } | |
1534 | bs_entry->prepared = true; | |
1535 | } | |
1536 | ||
1537 | /* If we reach this point, we have success and just need to apply the | |
1538 | * changes | |
1539 | */ | |
1540 | QSIMPLEQ_FOREACH(bs_entry, bs_queue, entry) { | |
1541 | bdrv_reopen_commit(&bs_entry->state); | |
1542 | } | |
1543 | ||
1544 | ret = 0; | |
1545 | ||
1546 | cleanup: | |
1547 | QSIMPLEQ_FOREACH_SAFE(bs_entry, bs_queue, entry, next) { | |
1548 | if (ret && bs_entry->prepared) { | |
1549 | bdrv_reopen_abort(&bs_entry->state); | |
1550 | } | |
1551 | g_free(bs_entry); | |
1552 | } | |
1553 | g_free(bs_queue); | |
1554 | return ret; | |
1555 | } | |
1556 | ||
1557 | ||
1558 | /* Reopen a single BlockDriverState with the specified flags. */ | |
1559 | int bdrv_reopen(BlockDriverState *bs, int bdrv_flags, Error **errp) | |
1560 | { | |
1561 | int ret = -1; | |
1562 | Error *local_err = NULL; | |
1563 | BlockReopenQueue *queue = bdrv_reopen_queue(NULL, bs, bdrv_flags); | |
1564 | ||
1565 | ret = bdrv_reopen_multiple(queue, &local_err); | |
1566 | if (local_err != NULL) { | |
1567 | error_propagate(errp, local_err); | |
1568 | } | |
1569 | return ret; | |
1570 | } | |
1571 | ||
1572 | ||
1573 | /* | |
1574 | * Prepares a BlockDriverState for reopen. All changes are staged in the | |
1575 | * 'opaque' field of the BDRVReopenState, which is used and allocated by | |
1576 | * the block driver layer .bdrv_reopen_prepare() | |
1577 | * | |
1578 | * bs is the BlockDriverState to reopen | |
1579 | * flags are the new open flags | |
1580 | * queue is the reopen queue | |
1581 | * | |
1582 | * Returns 0 on success, non-zero on error. On error errp will be set | |
1583 | * as well. | |
1584 | * | |
1585 | * On failure, bdrv_reopen_abort() will be called to clean up any data. | |
1586 | * It is the responsibility of the caller to then call the abort() or | |
1587 | * commit() for any other BDS that have been left in a prepare() state | |
1588 | * | |
1589 | */ | |
1590 | int bdrv_reopen_prepare(BDRVReopenState *reopen_state, BlockReopenQueue *queue, | |
1591 | Error **errp) | |
1592 | { | |
1593 | int ret = -1; | |
1594 | Error *local_err = NULL; | |
1595 | BlockDriver *drv; | |
1596 | ||
1597 | assert(reopen_state != NULL); | |
1598 | assert(reopen_state->bs->drv != NULL); | |
1599 | drv = reopen_state->bs->drv; | |
1600 | ||
1601 | /* if we are to stay read-only, do not allow permission change | |
1602 | * to r/w */ | |
1603 | if (!(reopen_state->bs->open_flags & BDRV_O_ALLOW_RDWR) && | |
1604 | reopen_state->flags & BDRV_O_RDWR) { | |
1605 | error_set(errp, QERR_DEVICE_IS_READ_ONLY, | |
1606 | reopen_state->bs->device_name); | |
1607 | goto error; | |
1608 | } | |
1609 | ||
1610 | ||
1611 | ret = bdrv_flush(reopen_state->bs); | |
1612 | if (ret) { | |
1613 | error_set(errp, ERROR_CLASS_GENERIC_ERROR, "Error (%s) flushing drive", | |
1614 | strerror(-ret)); | |
1615 | goto error; | |
1616 | } | |
1617 | ||
1618 | if (drv->bdrv_reopen_prepare) { | |
1619 | ret = drv->bdrv_reopen_prepare(reopen_state, queue, &local_err); | |
1620 | if (ret) { | |
1621 | if (local_err != NULL) { | |
1622 | error_propagate(errp, local_err); | |
1623 | } else { | |
d8b6895f LC |
1624 | error_setg(errp, "failed while preparing to reopen image '%s'", |
1625 | reopen_state->bs->filename); | |
e971aa12 JC |
1626 | } |
1627 | goto error; | |
1628 | } | |
1629 | } else { | |
1630 | /* It is currently mandatory to have a bdrv_reopen_prepare() | |
1631 | * handler for each supported drv. */ | |
1632 | error_set(errp, QERR_BLOCK_FORMAT_FEATURE_NOT_SUPPORTED, | |
1633 | drv->format_name, reopen_state->bs->device_name, | |
1634 | "reopening of file"); | |
1635 | ret = -1; | |
1636 | goto error; | |
1637 | } | |
1638 | ||
1639 | ret = 0; | |
1640 | ||
1641 | error: | |
1642 | return ret; | |
1643 | } | |
1644 | ||
1645 | /* | |
1646 | * Takes the staged changes for the reopen from bdrv_reopen_prepare(), and | |
1647 | * makes them final by swapping the staging BlockDriverState contents into | |
1648 | * the active BlockDriverState contents. | |
1649 | */ | |
1650 | void bdrv_reopen_commit(BDRVReopenState *reopen_state) | |
1651 | { | |
1652 | BlockDriver *drv; | |
1653 | ||
1654 | assert(reopen_state != NULL); | |
1655 | drv = reopen_state->bs->drv; | |
1656 | assert(drv != NULL); | |
1657 | ||
1658 | /* If there are any driver level actions to take */ | |
1659 | if (drv->bdrv_reopen_commit) { | |
1660 | drv->bdrv_reopen_commit(reopen_state); | |
1661 | } | |
1662 | ||
1663 | /* set BDS specific flags now */ | |
1664 | reopen_state->bs->open_flags = reopen_state->flags; | |
1665 | reopen_state->bs->enable_write_cache = !!(reopen_state->flags & | |
1666 | BDRV_O_CACHE_WB); | |
1667 | reopen_state->bs->read_only = !(reopen_state->flags & BDRV_O_RDWR); | |
355ef4ac KW |
1668 | |
1669 | bdrv_refresh_limits(reopen_state->bs); | |
e971aa12 JC |
1670 | } |
1671 | ||
1672 | /* | |
1673 | * Abort the reopen, and delete and free the staged changes in | |
1674 | * reopen_state | |
1675 | */ | |
1676 | void bdrv_reopen_abort(BDRVReopenState *reopen_state) | |
1677 | { | |
1678 | BlockDriver *drv; | |
1679 | ||
1680 | assert(reopen_state != NULL); | |
1681 | drv = reopen_state->bs->drv; | |
1682 | assert(drv != NULL); | |
1683 | ||
1684 | if (drv->bdrv_reopen_abort) { | |
1685 | drv->bdrv_reopen_abort(reopen_state); | |
1686 | } | |
1687 | } | |
1688 | ||
1689 | ||
fc01f7e7 FB |
1690 | void bdrv_close(BlockDriverState *bs) |
1691 | { | |
3cbc002c PB |
1692 | if (bs->job) { |
1693 | block_job_cancel_sync(bs->job); | |
1694 | } | |
58fda173 SH |
1695 | bdrv_drain_all(); /* complete I/O */ |
1696 | bdrv_flush(bs); | |
1697 | bdrv_drain_all(); /* in case flush left pending I/O */ | |
d7d512f6 | 1698 | notifier_list_notify(&bs->close_notifiers, bs); |
7094f12f | 1699 | |
3cbc002c | 1700 | if (bs->drv) { |
557df6ac | 1701 | if (bs->backing_hd) { |
4f6fd349 | 1702 | bdrv_unref(bs->backing_hd); |
557df6ac SH |
1703 | bs->backing_hd = NULL; |
1704 | } | |
ea2384d3 | 1705 | bs->drv->bdrv_close(bs); |
7267c094 | 1706 | g_free(bs->opaque); |
ea2384d3 FB |
1707 | #ifdef _WIN32 |
1708 | if (bs->is_temporary) { | |
1709 | unlink(bs->filename); | |
1710 | } | |
67b915a5 | 1711 | #endif |
ea2384d3 FB |
1712 | bs->opaque = NULL; |
1713 | bs->drv = NULL; | |
53fec9d3 | 1714 | bs->copy_on_read = 0; |
a275fa42 PB |
1715 | bs->backing_file[0] = '\0'; |
1716 | bs->backing_format[0] = '\0'; | |
6405875c PB |
1717 | bs->total_sectors = 0; |
1718 | bs->encrypted = 0; | |
1719 | bs->valid_key = 0; | |
1720 | bs->sg = 0; | |
1721 | bs->growable = 0; | |
0d51b4de | 1722 | bs->zero_beyond_eof = false; |
de9c0cec KW |
1723 | QDECREF(bs->options); |
1724 | bs->options = NULL; | |
b338082b | 1725 | |
66f82cee | 1726 | if (bs->file != NULL) { |
4f6fd349 | 1727 | bdrv_unref(bs->file); |
0ac9377d | 1728 | bs->file = NULL; |
66f82cee | 1729 | } |
b338082b | 1730 | } |
98f90dba | 1731 | |
9ca11154 PH |
1732 | bdrv_dev_change_media_cb(bs, false); |
1733 | ||
98f90dba ZYW |
1734 | /*throttling disk I/O limits*/ |
1735 | if (bs->io_limits_enabled) { | |
1736 | bdrv_io_limits_disable(bs); | |
1737 | } | |
b338082b FB |
1738 | } |
1739 | ||
2bc93fed MK |
1740 | void bdrv_close_all(void) |
1741 | { | |
1742 | BlockDriverState *bs; | |
1743 | ||
dc364f4c | 1744 | QTAILQ_FOREACH(bs, &bdrv_states, device_list) { |
2bc93fed MK |
1745 | bdrv_close(bs); |
1746 | } | |
1747 | } | |
1748 | ||
88266f5a SH |
1749 | /* Check if any requests are in-flight (including throttled requests) */ |
1750 | static bool bdrv_requests_pending(BlockDriverState *bs) | |
1751 | { | |
1752 | if (!QLIST_EMPTY(&bs->tracked_requests)) { | |
1753 | return true; | |
1754 | } | |
cc0681c4 BC |
1755 | if (!qemu_co_queue_empty(&bs->throttled_reqs[0])) { |
1756 | return true; | |
1757 | } | |
1758 | if (!qemu_co_queue_empty(&bs->throttled_reqs[1])) { | |
88266f5a SH |
1759 | return true; |
1760 | } | |
1761 | if (bs->file && bdrv_requests_pending(bs->file)) { | |
1762 | return true; | |
1763 | } | |
1764 | if (bs->backing_hd && bdrv_requests_pending(bs->backing_hd)) { | |
1765 | return true; | |
1766 | } | |
1767 | return false; | |
1768 | } | |
1769 | ||
1770 | static bool bdrv_requests_pending_all(void) | |
1771 | { | |
1772 | BlockDriverState *bs; | |
dc364f4c | 1773 | QTAILQ_FOREACH(bs, &bdrv_states, device_list) { |
88266f5a SH |
1774 | if (bdrv_requests_pending(bs)) { |
1775 | return true; | |
1776 | } | |
1777 | } | |
1778 | return false; | |
1779 | } | |
1780 | ||
922453bc SH |
1781 | /* |
1782 | * Wait for pending requests to complete across all BlockDriverStates | |
1783 | * | |
1784 | * This function does not flush data to disk, use bdrv_flush_all() for that | |
1785 | * after calling this function. | |
4c355d53 ZYW |
1786 | * |
1787 | * Note that completion of an asynchronous I/O operation can trigger any | |
1788 | * number of other I/O operations on other devices---for example a coroutine | |
1789 | * can be arbitrarily complex and a constant flow of I/O can come until the | |
1790 | * coroutine is complete. Because of this, it is not possible to have a | |
1791 | * function to drain a single device's I/O queue. | |
922453bc SH |
1792 | */ |
1793 | void bdrv_drain_all(void) | |
1794 | { | |
88266f5a SH |
1795 | /* Always run first iteration so any pending completion BHs run */ |
1796 | bool busy = true; | |
922453bc SH |
1797 | BlockDriverState *bs; |
1798 | ||
88266f5a | 1799 | while (busy) { |
dc364f4c | 1800 | QTAILQ_FOREACH(bs, &bdrv_states, device_list) { |
0b06ef3b | 1801 | bdrv_start_throttled_reqs(bs); |
4c355d53 | 1802 | } |
922453bc | 1803 | |
88266f5a SH |
1804 | busy = bdrv_requests_pending_all(); |
1805 | busy |= aio_poll(qemu_get_aio_context(), busy); | |
922453bc SH |
1806 | } |
1807 | } | |
1808 | ||
dc364f4c BC |
1809 | /* make a BlockDriverState anonymous by removing from bdrv_state and |
1810 | * graph_bdrv_state list. | |
d22b2f41 RH |
1811 | Also, NULL terminate the device_name to prevent double remove */ |
1812 | void bdrv_make_anon(BlockDriverState *bs) | |
1813 | { | |
1814 | if (bs->device_name[0] != '\0') { | |
dc364f4c | 1815 | QTAILQ_REMOVE(&bdrv_states, bs, device_list); |
d22b2f41 RH |
1816 | } |
1817 | bs->device_name[0] = '\0'; | |
dc364f4c BC |
1818 | if (bs->node_name[0] != '\0') { |
1819 | QTAILQ_REMOVE(&graph_bdrv_states, bs, node_list); | |
1820 | } | |
1821 | bs->node_name[0] = '\0'; | |
d22b2f41 RH |
1822 | } |
1823 | ||
e023b2e2 PB |
1824 | static void bdrv_rebind(BlockDriverState *bs) |
1825 | { | |
1826 | if (bs->drv && bs->drv->bdrv_rebind) { | |
1827 | bs->drv->bdrv_rebind(bs); | |
1828 | } | |
1829 | } | |
1830 | ||
4ddc07ca PB |
1831 | static void bdrv_move_feature_fields(BlockDriverState *bs_dest, |
1832 | BlockDriverState *bs_src) | |
8802d1fd | 1833 | { |
4ddc07ca PB |
1834 | /* move some fields that need to stay attached to the device */ |
1835 | bs_dest->open_flags = bs_src->open_flags; | |
8802d1fd JC |
1836 | |
1837 | /* dev info */ | |
4ddc07ca PB |
1838 | bs_dest->dev_ops = bs_src->dev_ops; |
1839 | bs_dest->dev_opaque = bs_src->dev_opaque; | |
1840 | bs_dest->dev = bs_src->dev; | |
1b7fd729 | 1841 | bs_dest->guest_block_size = bs_src->guest_block_size; |
4ddc07ca | 1842 | bs_dest->copy_on_read = bs_src->copy_on_read; |
8802d1fd | 1843 | |
4ddc07ca | 1844 | bs_dest->enable_write_cache = bs_src->enable_write_cache; |
c4a248a1 | 1845 | |
cc0681c4 BC |
1846 | /* i/o throttled req */ |
1847 | memcpy(&bs_dest->throttle_state, | |
1848 | &bs_src->throttle_state, | |
1849 | sizeof(ThrottleState)); | |
1850 | bs_dest->throttled_reqs[0] = bs_src->throttled_reqs[0]; | |
1851 | bs_dest->throttled_reqs[1] = bs_src->throttled_reqs[1]; | |
4ddc07ca | 1852 | bs_dest->io_limits_enabled = bs_src->io_limits_enabled; |
8802d1fd | 1853 | |
8802d1fd | 1854 | /* r/w error */ |
4ddc07ca PB |
1855 | bs_dest->on_read_error = bs_src->on_read_error; |
1856 | bs_dest->on_write_error = bs_src->on_write_error; | |
8802d1fd JC |
1857 | |
1858 | /* i/o status */ | |
4ddc07ca PB |
1859 | bs_dest->iostatus_enabled = bs_src->iostatus_enabled; |
1860 | bs_dest->iostatus = bs_src->iostatus; | |
8802d1fd | 1861 | |
a9fc4408 | 1862 | /* dirty bitmap */ |
e4654d2d | 1863 | bs_dest->dirty_bitmaps = bs_src->dirty_bitmaps; |
a9fc4408 | 1864 | |
9fcb0251 FZ |
1865 | /* reference count */ |
1866 | bs_dest->refcnt = bs_src->refcnt; | |
1867 | ||
a9fc4408 | 1868 | /* job */ |
4ddc07ca PB |
1869 | bs_dest->in_use = bs_src->in_use; |
1870 | bs_dest->job = bs_src->job; | |
a9fc4408 | 1871 | |
8802d1fd | 1872 | /* keep the same entry in bdrv_states */ |
4ddc07ca PB |
1873 | pstrcpy(bs_dest->device_name, sizeof(bs_dest->device_name), |
1874 | bs_src->device_name); | |
dc364f4c | 1875 | bs_dest->device_list = bs_src->device_list; |
4ddc07ca | 1876 | } |
8802d1fd | 1877 | |
4ddc07ca PB |
1878 | /* |
1879 | * Swap bs contents for two image chains while they are live, | |
1880 | * while keeping required fields on the BlockDriverState that is | |
1881 | * actually attached to a device. | |
1882 | * | |
1883 | * This will modify the BlockDriverState fields, and swap contents | |
1884 | * between bs_new and bs_old. Both bs_new and bs_old are modified. | |
1885 | * | |
1886 | * bs_new is required to be anonymous. | |
1887 | * | |
1888 | * This function does not create any image files. | |
1889 | */ | |
1890 | void bdrv_swap(BlockDriverState *bs_new, BlockDriverState *bs_old) | |
1891 | { | |
1892 | BlockDriverState tmp; | |
f6801b83 | 1893 | |
90ce8a06 BC |
1894 | /* The code needs to swap the node_name but simply swapping node_list won't |
1895 | * work so first remove the nodes from the graph list, do the swap then | |
1896 | * insert them back if needed. | |
1897 | */ | |
1898 | if (bs_new->node_name[0] != '\0') { | |
1899 | QTAILQ_REMOVE(&graph_bdrv_states, bs_new, node_list); | |
1900 | } | |
1901 | if (bs_old->node_name[0] != '\0') { | |
1902 | QTAILQ_REMOVE(&graph_bdrv_states, bs_old, node_list); | |
1903 | } | |
1904 | ||
4ddc07ca PB |
1905 | /* bs_new must be anonymous and shouldn't have anything fancy enabled */ |
1906 | assert(bs_new->device_name[0] == '\0'); | |
e4654d2d | 1907 | assert(QLIST_EMPTY(&bs_new->dirty_bitmaps)); |
4ddc07ca PB |
1908 | assert(bs_new->job == NULL); |
1909 | assert(bs_new->dev == NULL); | |
1910 | assert(bs_new->in_use == 0); | |
1911 | assert(bs_new->io_limits_enabled == false); | |
cc0681c4 | 1912 | assert(!throttle_have_timer(&bs_new->throttle_state)); |
8802d1fd | 1913 | |
4ddc07ca PB |
1914 | tmp = *bs_new; |
1915 | *bs_new = *bs_old; | |
1916 | *bs_old = tmp; | |
a9fc4408 | 1917 | |
4ddc07ca PB |
1918 | /* there are some fields that should not be swapped, move them back */ |
1919 | bdrv_move_feature_fields(&tmp, bs_old); | |
1920 | bdrv_move_feature_fields(bs_old, bs_new); | |
1921 | bdrv_move_feature_fields(bs_new, &tmp); | |
8802d1fd | 1922 | |
4ddc07ca PB |
1923 | /* bs_new shouldn't be in bdrv_states even after the swap! */ |
1924 | assert(bs_new->device_name[0] == '\0'); | |
1925 | ||
1926 | /* Check a few fields that should remain attached to the device */ | |
1927 | assert(bs_new->dev == NULL); | |
1928 | assert(bs_new->job == NULL); | |
1929 | assert(bs_new->in_use == 0); | |
1930 | assert(bs_new->io_limits_enabled == false); | |
cc0681c4 | 1931 | assert(!throttle_have_timer(&bs_new->throttle_state)); |
e023b2e2 | 1932 | |
90ce8a06 BC |
1933 | /* insert the nodes back into the graph node list if needed */ |
1934 | if (bs_new->node_name[0] != '\0') { | |
1935 | QTAILQ_INSERT_TAIL(&graph_bdrv_states, bs_new, node_list); | |
1936 | } | |
1937 | if (bs_old->node_name[0] != '\0') { | |
1938 | QTAILQ_INSERT_TAIL(&graph_bdrv_states, bs_old, node_list); | |
1939 | } | |
1940 | ||
e023b2e2 | 1941 | bdrv_rebind(bs_new); |
4ddc07ca PB |
1942 | bdrv_rebind(bs_old); |
1943 | } | |
1944 | ||
1945 | /* | |
1946 | * Add new bs contents at the top of an image chain while the chain is | |
1947 | * live, while keeping required fields on the top layer. | |
1948 | * | |
1949 | * This will modify the BlockDriverState fields, and swap contents | |
1950 | * between bs_new and bs_top. Both bs_new and bs_top are modified. | |
1951 | * | |
1952 | * bs_new is required to be anonymous. | |
1953 | * | |
1954 | * This function does not create any image files. | |
1955 | */ | |
1956 | void bdrv_append(BlockDriverState *bs_new, BlockDriverState *bs_top) | |
1957 | { | |
1958 | bdrv_swap(bs_new, bs_top); | |
1959 | ||
1960 | /* The contents of 'tmp' will become bs_top, as we are | |
1961 | * swapping bs_new and bs_top contents. */ | |
1962 | bs_top->backing_hd = bs_new; | |
1963 | bs_top->open_flags &= ~BDRV_O_NO_BACKING; | |
1964 | pstrcpy(bs_top->backing_file, sizeof(bs_top->backing_file), | |
1965 | bs_new->filename); | |
1966 | pstrcpy(bs_top->backing_format, sizeof(bs_top->backing_format), | |
1967 | bs_new->drv ? bs_new->drv->format_name : ""); | |
8802d1fd JC |
1968 | } |
1969 | ||
4f6fd349 | 1970 | static void bdrv_delete(BlockDriverState *bs) |
b338082b | 1971 | { |
fa879d62 | 1972 | assert(!bs->dev); |
3e914655 PB |
1973 | assert(!bs->job); |
1974 | assert(!bs->in_use); | |
4f6fd349 | 1975 | assert(!bs->refcnt); |
e4654d2d | 1976 | assert(QLIST_EMPTY(&bs->dirty_bitmaps)); |
18846dee | 1977 | |
e1b5c52e SH |
1978 | bdrv_close(bs); |
1979 | ||
1b7bdbc1 | 1980 | /* remove from list, if necessary */ |
d22b2f41 | 1981 | bdrv_make_anon(bs); |
34c6f050 | 1982 | |
7267c094 | 1983 | g_free(bs); |
fc01f7e7 FB |
1984 | } |
1985 | ||
fa879d62 MA |
1986 | int bdrv_attach_dev(BlockDriverState *bs, void *dev) |
1987 | /* TODO change to DeviceState *dev when all users are qdevified */ | |
18846dee | 1988 | { |
fa879d62 | 1989 | if (bs->dev) { |
18846dee MA |
1990 | return -EBUSY; |
1991 | } | |
fa879d62 | 1992 | bs->dev = dev; |
28a7282a | 1993 | bdrv_iostatus_reset(bs); |
18846dee MA |
1994 | return 0; |
1995 | } | |
1996 | ||
fa879d62 MA |
1997 | /* TODO qdevified devices don't use this, remove when devices are qdevified */ |
1998 | void bdrv_attach_dev_nofail(BlockDriverState *bs, void *dev) | |
18846dee | 1999 | { |
fa879d62 MA |
2000 | if (bdrv_attach_dev(bs, dev) < 0) { |
2001 | abort(); | |
2002 | } | |
2003 | } | |
2004 | ||
2005 | void bdrv_detach_dev(BlockDriverState *bs, void *dev) | |
2006 | /* TODO change to DeviceState *dev when all users are qdevified */ | |
2007 | { | |
2008 | assert(bs->dev == dev); | |
2009 | bs->dev = NULL; | |
0e49de52 MA |
2010 | bs->dev_ops = NULL; |
2011 | bs->dev_opaque = NULL; | |
1b7fd729 | 2012 | bs->guest_block_size = 512; |
18846dee MA |
2013 | } |
2014 | ||
fa879d62 MA |
2015 | /* TODO change to return DeviceState * when all users are qdevified */ |
2016 | void *bdrv_get_attached_dev(BlockDriverState *bs) | |
18846dee | 2017 | { |
fa879d62 | 2018 | return bs->dev; |
18846dee MA |
2019 | } |
2020 | ||
0e49de52 MA |
2021 | void bdrv_set_dev_ops(BlockDriverState *bs, const BlockDevOps *ops, |
2022 | void *opaque) | |
2023 | { | |
2024 | bs->dev_ops = ops; | |
2025 | bs->dev_opaque = opaque; | |
2026 | } | |
2027 | ||
32c81a4a PB |
2028 | void bdrv_emit_qmp_error_event(const BlockDriverState *bdrv, |
2029 | enum MonitorEvent ev, | |
2030 | BlockErrorAction action, bool is_read) | |
329c0a48 LC |
2031 | { |
2032 | QObject *data; | |
2033 | const char *action_str; | |
2034 | ||
2035 | switch (action) { | |
2036 | case BDRV_ACTION_REPORT: | |
2037 | action_str = "report"; | |
2038 | break; | |
2039 | case BDRV_ACTION_IGNORE: | |
2040 | action_str = "ignore"; | |
2041 | break; | |
2042 | case BDRV_ACTION_STOP: | |
2043 | action_str = "stop"; | |
2044 | break; | |
2045 | default: | |
2046 | abort(); | |
2047 | } | |
2048 | ||
2049 | data = qobject_from_jsonf("{ 'device': %s, 'action': %s, 'operation': %s }", | |
2050 | bdrv->device_name, | |
2051 | action_str, | |
2052 | is_read ? "read" : "write"); | |
32c81a4a | 2053 | monitor_protocol_event(ev, data); |
329c0a48 LC |
2054 | |
2055 | qobject_decref(data); | |
2056 | } | |
2057 | ||
6f382ed2 LC |
2058 | static void bdrv_emit_qmp_eject_event(BlockDriverState *bs, bool ejected) |
2059 | { | |
2060 | QObject *data; | |
2061 | ||
2062 | data = qobject_from_jsonf("{ 'device': %s, 'tray-open': %i }", | |
2063 | bdrv_get_device_name(bs), ejected); | |
2064 | monitor_protocol_event(QEVENT_DEVICE_TRAY_MOVED, data); | |
2065 | ||
2066 | qobject_decref(data); | |
2067 | } | |
2068 | ||
7d4b4ba5 | 2069 | static void bdrv_dev_change_media_cb(BlockDriverState *bs, bool load) |
0e49de52 | 2070 | { |
145feb17 | 2071 | if (bs->dev_ops && bs->dev_ops->change_media_cb) { |
6f382ed2 | 2072 | bool tray_was_closed = !bdrv_dev_is_tray_open(bs); |
7d4b4ba5 | 2073 | bs->dev_ops->change_media_cb(bs->dev_opaque, load); |
6f382ed2 LC |
2074 | if (tray_was_closed) { |
2075 | /* tray open */ | |
2076 | bdrv_emit_qmp_eject_event(bs, true); | |
2077 | } | |
2078 | if (load) { | |
2079 | /* tray close */ | |
2080 | bdrv_emit_qmp_eject_event(bs, false); | |
2081 | } | |
145feb17 MA |
2082 | } |
2083 | } | |
2084 | ||
2c6942fa MA |
2085 | bool bdrv_dev_has_removable_media(BlockDriverState *bs) |
2086 | { | |
2087 | return !bs->dev || (bs->dev_ops && bs->dev_ops->change_media_cb); | |
2088 | } | |
2089 | ||
025ccaa7 PB |
2090 | void bdrv_dev_eject_request(BlockDriverState *bs, bool force) |
2091 | { | |
2092 | if (bs->dev_ops && bs->dev_ops->eject_request_cb) { | |
2093 | bs->dev_ops->eject_request_cb(bs->dev_opaque, force); | |
2094 | } | |
2095 | } | |
2096 | ||
e4def80b MA |
2097 | bool bdrv_dev_is_tray_open(BlockDriverState *bs) |
2098 | { | |
2099 | if (bs->dev_ops && bs->dev_ops->is_tray_open) { | |
2100 | return bs->dev_ops->is_tray_open(bs->dev_opaque); | |
2101 | } | |
2102 | return false; | |
2103 | } | |
2104 | ||
145feb17 MA |
2105 | static void bdrv_dev_resize_cb(BlockDriverState *bs) |
2106 | { | |
2107 | if (bs->dev_ops && bs->dev_ops->resize_cb) { | |
2108 | bs->dev_ops->resize_cb(bs->dev_opaque); | |
0e49de52 MA |
2109 | } |
2110 | } | |
2111 | ||
f107639a MA |
2112 | bool bdrv_dev_is_medium_locked(BlockDriverState *bs) |
2113 | { | |
2114 | if (bs->dev_ops && bs->dev_ops->is_medium_locked) { | |
2115 | return bs->dev_ops->is_medium_locked(bs->dev_opaque); | |
2116 | } | |
2117 | return false; | |
2118 | } | |
2119 | ||
e97fc193 AL |
2120 | /* |
2121 | * Run consistency checks on an image | |
2122 | * | |
e076f338 | 2123 | * Returns 0 if the check could be completed (it doesn't mean that the image is |
a1c7273b | 2124 | * free of errors) or -errno when an internal error occurred. The results of the |
e076f338 | 2125 | * check are stored in res. |
e97fc193 | 2126 | */ |
4534ff54 | 2127 | int bdrv_check(BlockDriverState *bs, BdrvCheckResult *res, BdrvCheckMode fix) |
e97fc193 AL |
2128 | { |
2129 | if (bs->drv->bdrv_check == NULL) { | |
2130 | return -ENOTSUP; | |
2131 | } | |
2132 | ||
e076f338 | 2133 | memset(res, 0, sizeof(*res)); |
4534ff54 | 2134 | return bs->drv->bdrv_check(bs, res, fix); |
e97fc193 AL |
2135 | } |
2136 | ||
8a426614 KW |
2137 | #define COMMIT_BUF_SECTORS 2048 |
2138 | ||
33e3963e FB |
2139 | /* commit COW file into the raw image */ |
2140 | int bdrv_commit(BlockDriverState *bs) | |
2141 | { | |
19cb3738 | 2142 | BlockDriver *drv = bs->drv; |
72706ea4 | 2143 | int64_t sector, total_sectors, length, backing_length; |
8a426614 | 2144 | int n, ro, open_flags; |
0bce597d | 2145 | int ret = 0; |
72706ea4 | 2146 | uint8_t *buf = NULL; |
c2cba3d9 | 2147 | char filename[PATH_MAX]; |
33e3963e | 2148 | |
19cb3738 FB |
2149 | if (!drv) |
2150 | return -ENOMEDIUM; | |
4dca4b63 NS |
2151 | |
2152 | if (!bs->backing_hd) { | |
2153 | return -ENOTSUP; | |
33e3963e FB |
2154 | } |
2155 | ||
2d3735d3 SH |
2156 | if (bdrv_in_use(bs) || bdrv_in_use(bs->backing_hd)) { |
2157 | return -EBUSY; | |
2158 | } | |
2159 | ||
4dca4b63 | 2160 | ro = bs->backing_hd->read_only; |
c2cba3d9 JM |
2161 | /* Use pstrcpy (not strncpy): filename must be NUL-terminated. */ |
2162 | pstrcpy(filename, sizeof(filename), bs->backing_hd->filename); | |
4dca4b63 NS |
2163 | open_flags = bs->backing_hd->open_flags; |
2164 | ||
2165 | if (ro) { | |
0bce597d JC |
2166 | if (bdrv_reopen(bs->backing_hd, open_flags | BDRV_O_RDWR, NULL)) { |
2167 | return -EACCES; | |
4dca4b63 | 2168 | } |
ea2384d3 | 2169 | } |
33e3963e | 2170 | |
72706ea4 JC |
2171 | length = bdrv_getlength(bs); |
2172 | if (length < 0) { | |
2173 | ret = length; | |
2174 | goto ro_cleanup; | |
2175 | } | |
2176 | ||
2177 | backing_length = bdrv_getlength(bs->backing_hd); | |
2178 | if (backing_length < 0) { | |
2179 | ret = backing_length; | |
2180 | goto ro_cleanup; | |
2181 | } | |
2182 | ||
2183 | /* If our top snapshot is larger than the backing file image, | |
2184 | * grow the backing file image if possible. If not possible, | |
2185 | * we must return an error */ | |
2186 | if (length > backing_length) { | |
2187 | ret = bdrv_truncate(bs->backing_hd, length); | |
2188 | if (ret < 0) { | |
2189 | goto ro_cleanup; | |
2190 | } | |
2191 | } | |
2192 | ||
2193 | total_sectors = length >> BDRV_SECTOR_BITS; | |
7267c094 | 2194 | buf = g_malloc(COMMIT_BUF_SECTORS * BDRV_SECTOR_SIZE); |
8a426614 KW |
2195 | |
2196 | for (sector = 0; sector < total_sectors; sector += n) { | |
d663640c PB |
2197 | ret = bdrv_is_allocated(bs, sector, COMMIT_BUF_SECTORS, &n); |
2198 | if (ret < 0) { | |
2199 | goto ro_cleanup; | |
2200 | } | |
2201 | if (ret) { | |
dabfa6cc KW |
2202 | ret = bdrv_read(bs, sector, buf, n); |
2203 | if (ret < 0) { | |
8a426614 KW |
2204 | goto ro_cleanup; |
2205 | } | |
2206 | ||
dabfa6cc KW |
2207 | ret = bdrv_write(bs->backing_hd, sector, buf, n); |
2208 | if (ret < 0) { | |
8a426614 KW |
2209 | goto ro_cleanup; |
2210 | } | |
ea2384d3 | 2211 | } |
33e3963e | 2212 | } |
95389c86 | 2213 | |
1d44952f CH |
2214 | if (drv->bdrv_make_empty) { |
2215 | ret = drv->bdrv_make_empty(bs); | |
dabfa6cc KW |
2216 | if (ret < 0) { |
2217 | goto ro_cleanup; | |
2218 | } | |
1d44952f CH |
2219 | bdrv_flush(bs); |
2220 | } | |
95389c86 | 2221 | |
3f5075ae CH |
2222 | /* |
2223 | * Make sure all data we wrote to the backing device is actually | |
2224 | * stable on disk. | |
2225 | */ | |
dabfa6cc | 2226 | if (bs->backing_hd) { |
3f5075ae | 2227 | bdrv_flush(bs->backing_hd); |
dabfa6cc | 2228 | } |
4dca4b63 | 2229 | |
dabfa6cc | 2230 | ret = 0; |
4dca4b63 | 2231 | ro_cleanup: |
7267c094 | 2232 | g_free(buf); |
4dca4b63 NS |
2233 | |
2234 | if (ro) { | |
0bce597d JC |
2235 | /* ignoring error return here */ |
2236 | bdrv_reopen(bs->backing_hd, open_flags & ~BDRV_O_RDWR, NULL); | |
4dca4b63 NS |
2237 | } |
2238 | ||
1d44952f | 2239 | return ret; |
33e3963e FB |
2240 | } |
2241 | ||
e8877497 | 2242 | int bdrv_commit_all(void) |
6ab4b5ab MA |
2243 | { |
2244 | BlockDriverState *bs; | |
2245 | ||
dc364f4c | 2246 | QTAILQ_FOREACH(bs, &bdrv_states, device_list) { |
272d2d8e JC |
2247 | if (bs->drv && bs->backing_hd) { |
2248 | int ret = bdrv_commit(bs); | |
2249 | if (ret < 0) { | |
2250 | return ret; | |
2251 | } | |
e8877497 | 2252 | } |
6ab4b5ab | 2253 | } |
e8877497 | 2254 | return 0; |
6ab4b5ab MA |
2255 | } |
2256 | ||
dbffbdcf SH |
2257 | /** |
2258 | * Remove an active request from the tracked requests list | |
2259 | * | |
2260 | * This function should be called when a tracked request is completing. | |
2261 | */ | |
2262 | static void tracked_request_end(BdrvTrackedRequest *req) | |
2263 | { | |
2dbafdc0 KW |
2264 | if (req->serialising) { |
2265 | req->bs->serialising_in_flight--; | |
2266 | } | |
2267 | ||
dbffbdcf | 2268 | QLIST_REMOVE(req, list); |
f4658285 | 2269 | qemu_co_queue_restart_all(&req->wait_queue); |
dbffbdcf SH |
2270 | } |
2271 | ||
2272 | /** | |
2273 | * Add an active request to the tracked requests list | |
2274 | */ | |
2275 | static void tracked_request_begin(BdrvTrackedRequest *req, | |
2276 | BlockDriverState *bs, | |
793ed47a KW |
2277 | int64_t offset, |
2278 | unsigned int bytes, bool is_write) | |
dbffbdcf SH |
2279 | { |
2280 | *req = (BdrvTrackedRequest){ | |
2281 | .bs = bs, | |
2dbafdc0 KW |
2282 | .offset = offset, |
2283 | .bytes = bytes, | |
2284 | .is_write = is_write, | |
2285 | .co = qemu_coroutine_self(), | |
2286 | .serialising = false, | |
7327145f KW |
2287 | .overlap_offset = offset, |
2288 | .overlap_bytes = bytes, | |
dbffbdcf SH |
2289 | }; |
2290 | ||
f4658285 SH |
2291 | qemu_co_queue_init(&req->wait_queue); |
2292 | ||
dbffbdcf SH |
2293 | QLIST_INSERT_HEAD(&bs->tracked_requests, req, list); |
2294 | } | |
2295 | ||
e96126ff | 2296 | static void mark_request_serialising(BdrvTrackedRequest *req, uint64_t align) |
2dbafdc0 | 2297 | { |
7327145f | 2298 | int64_t overlap_offset = req->offset & ~(align - 1); |
e96126ff KW |
2299 | unsigned int overlap_bytes = ROUND_UP(req->offset + req->bytes, align) |
2300 | - overlap_offset; | |
7327145f | 2301 | |
2dbafdc0 KW |
2302 | if (!req->serialising) { |
2303 | req->bs->serialising_in_flight++; | |
2304 | req->serialising = true; | |
2305 | } | |
7327145f KW |
2306 | |
2307 | req->overlap_offset = MIN(req->overlap_offset, overlap_offset); | |
2308 | req->overlap_bytes = MAX(req->overlap_bytes, overlap_bytes); | |
2dbafdc0 KW |
2309 | } |
2310 | ||
d83947ac SH |
2311 | /** |
2312 | * Round a region to cluster boundaries | |
2313 | */ | |
343bded4 PB |
2314 | void bdrv_round_to_clusters(BlockDriverState *bs, |
2315 | int64_t sector_num, int nb_sectors, | |
2316 | int64_t *cluster_sector_num, | |
2317 | int *cluster_nb_sectors) | |
d83947ac SH |
2318 | { |
2319 | BlockDriverInfo bdi; | |
2320 | ||
2321 | if (bdrv_get_info(bs, &bdi) < 0 || bdi.cluster_size == 0) { | |
2322 | *cluster_sector_num = sector_num; | |
2323 | *cluster_nb_sectors = nb_sectors; | |
2324 | } else { | |
2325 | int64_t c = bdi.cluster_size / BDRV_SECTOR_SIZE; | |
2326 | *cluster_sector_num = QEMU_ALIGN_DOWN(sector_num, c); | |
2327 | *cluster_nb_sectors = QEMU_ALIGN_UP(sector_num - *cluster_sector_num + | |
2328 | nb_sectors, c); | |
2329 | } | |
2330 | } | |
2331 | ||
7327145f | 2332 | static int bdrv_get_cluster_size(BlockDriverState *bs) |
793ed47a KW |
2333 | { |
2334 | BlockDriverInfo bdi; | |
7327145f | 2335 | int ret; |
793ed47a | 2336 | |
7327145f KW |
2337 | ret = bdrv_get_info(bs, &bdi); |
2338 | if (ret < 0 || bdi.cluster_size == 0) { | |
2339 | return bs->request_alignment; | |
793ed47a | 2340 | } else { |
7327145f | 2341 | return bdi.cluster_size; |
793ed47a KW |
2342 | } |
2343 | } | |
2344 | ||
f4658285 | 2345 | static bool tracked_request_overlaps(BdrvTrackedRequest *req, |
793ed47a KW |
2346 | int64_t offset, unsigned int bytes) |
2347 | { | |
d83947ac | 2348 | /* aaaa bbbb */ |
7327145f | 2349 | if (offset >= req->overlap_offset + req->overlap_bytes) { |
d83947ac SH |
2350 | return false; |
2351 | } | |
2352 | /* bbbb aaaa */ | |
7327145f | 2353 | if (req->overlap_offset >= offset + bytes) { |
d83947ac SH |
2354 | return false; |
2355 | } | |
2356 | return true; | |
f4658285 SH |
2357 | } |
2358 | ||
28de2dcd | 2359 | static bool coroutine_fn wait_serialising_requests(BdrvTrackedRequest *self) |
f4658285 | 2360 | { |
2dbafdc0 | 2361 | BlockDriverState *bs = self->bs; |
f4658285 SH |
2362 | BdrvTrackedRequest *req; |
2363 | bool retry; | |
28de2dcd | 2364 | bool waited = false; |
f4658285 | 2365 | |
2dbafdc0 | 2366 | if (!bs->serialising_in_flight) { |
28de2dcd | 2367 | return false; |
2dbafdc0 KW |
2368 | } |
2369 | ||
f4658285 SH |
2370 | do { |
2371 | retry = false; | |
2372 | QLIST_FOREACH(req, &bs->tracked_requests, list) { | |
2dbafdc0 | 2373 | if (req == self || (!req->serialising && !self->serialising)) { |
65afd211 KW |
2374 | continue; |
2375 | } | |
7327145f KW |
2376 | if (tracked_request_overlaps(req, self->overlap_offset, |
2377 | self->overlap_bytes)) | |
2378 | { | |
5f8b6491 SH |
2379 | /* Hitting this means there was a reentrant request, for |
2380 | * example, a block driver issuing nested requests. This must | |
2381 | * never happen since it means deadlock. | |
2382 | */ | |
2383 | assert(qemu_coroutine_self() != req->co); | |
2384 | ||
6460440f KW |
2385 | /* If the request is already (indirectly) waiting for us, or |
2386 | * will wait for us as soon as it wakes up, then just go on | |
2387 | * (instead of producing a deadlock in the former case). */ | |
2388 | if (!req->waiting_for) { | |
2389 | self->waiting_for = req; | |
2390 | qemu_co_queue_wait(&req->wait_queue); | |
2391 | self->waiting_for = NULL; | |
2392 | retry = true; | |
28de2dcd | 2393 | waited = true; |
6460440f KW |
2394 | break; |
2395 | } | |
f4658285 SH |
2396 | } |
2397 | } | |
2398 | } while (retry); | |
28de2dcd KW |
2399 | |
2400 | return waited; | |
f4658285 SH |
2401 | } |
2402 | ||
756e6736 KW |
2403 | /* |
2404 | * Return values: | |
2405 | * 0 - success | |
2406 | * -EINVAL - backing format specified, but no file | |
2407 | * -ENOSPC - can't update the backing file because no space is left in the | |
2408 | * image file header | |
2409 | * -ENOTSUP - format driver doesn't support changing the backing file | |
2410 | */ | |
2411 | int bdrv_change_backing_file(BlockDriverState *bs, | |
2412 | const char *backing_file, const char *backing_fmt) | |
2413 | { | |
2414 | BlockDriver *drv = bs->drv; | |
469ef350 | 2415 | int ret; |
756e6736 | 2416 | |
5f377794 PB |
2417 | /* Backing file format doesn't make sense without a backing file */ |
2418 | if (backing_fmt && !backing_file) { | |
2419 | return -EINVAL; | |
2420 | } | |
2421 | ||
756e6736 | 2422 | if (drv->bdrv_change_backing_file != NULL) { |
469ef350 | 2423 | ret = drv->bdrv_change_backing_file(bs, backing_file, backing_fmt); |
756e6736 | 2424 | } else { |
469ef350 | 2425 | ret = -ENOTSUP; |
756e6736 | 2426 | } |
469ef350 PB |
2427 | |
2428 | if (ret == 0) { | |
2429 | pstrcpy(bs->backing_file, sizeof(bs->backing_file), backing_file ?: ""); | |
2430 | pstrcpy(bs->backing_format, sizeof(bs->backing_format), backing_fmt ?: ""); | |
2431 | } | |
2432 | return ret; | |
756e6736 KW |
2433 | } |
2434 | ||
6ebdcee2 JC |
2435 | /* |
2436 | * Finds the image layer in the chain that has 'bs' as its backing file. | |
2437 | * | |
2438 | * active is the current topmost image. | |
2439 | * | |
2440 | * Returns NULL if bs is not found in active's image chain, | |
2441 | * or if active == bs. | |
2442 | */ | |
2443 | BlockDriverState *bdrv_find_overlay(BlockDriverState *active, | |
2444 | BlockDriverState *bs) | |
2445 | { | |
2446 | BlockDriverState *overlay = NULL; | |
2447 | BlockDriverState *intermediate; | |
2448 | ||
2449 | assert(active != NULL); | |
2450 | assert(bs != NULL); | |
2451 | ||
2452 | /* if bs is the same as active, then by definition it has no overlay | |
2453 | */ | |
2454 | if (active == bs) { | |
2455 | return NULL; | |
2456 | } | |
2457 | ||
2458 | intermediate = active; | |
2459 | while (intermediate->backing_hd) { | |
2460 | if (intermediate->backing_hd == bs) { | |
2461 | overlay = intermediate; | |
2462 | break; | |
2463 | } | |
2464 | intermediate = intermediate->backing_hd; | |
2465 | } | |
2466 | ||
2467 | return overlay; | |
2468 | } | |
2469 | ||
2470 | typedef struct BlkIntermediateStates { | |
2471 | BlockDriverState *bs; | |
2472 | QSIMPLEQ_ENTRY(BlkIntermediateStates) entry; | |
2473 | } BlkIntermediateStates; | |
2474 | ||
2475 | ||
2476 | /* | |
2477 | * Drops images above 'base' up to and including 'top', and sets the image | |
2478 | * above 'top' to have base as its backing file. | |
2479 | * | |
2480 | * Requires that the overlay to 'top' is opened r/w, so that the backing file | |
2481 | * information in 'bs' can be properly updated. | |
2482 | * | |
2483 | * E.g., this will convert the following chain: | |
2484 | * bottom <- base <- intermediate <- top <- active | |
2485 | * | |
2486 | * to | |
2487 | * | |
2488 | * bottom <- base <- active | |
2489 | * | |
2490 | * It is allowed for bottom==base, in which case it converts: | |
2491 | * | |
2492 | * base <- intermediate <- top <- active | |
2493 | * | |
2494 | * to | |
2495 | * | |
2496 | * base <- active | |
2497 | * | |
2498 | * Error conditions: | |
2499 | * if active == top, that is considered an error | |
2500 | * | |
2501 | */ | |
2502 | int bdrv_drop_intermediate(BlockDriverState *active, BlockDriverState *top, | |
2503 | BlockDriverState *base) | |
2504 | { | |
2505 | BlockDriverState *intermediate; | |
2506 | BlockDriverState *base_bs = NULL; | |
2507 | BlockDriverState *new_top_bs = NULL; | |
2508 | BlkIntermediateStates *intermediate_state, *next; | |
2509 | int ret = -EIO; | |
2510 | ||
2511 | QSIMPLEQ_HEAD(states_to_delete, BlkIntermediateStates) states_to_delete; | |
2512 | QSIMPLEQ_INIT(&states_to_delete); | |
2513 | ||
2514 | if (!top->drv || !base->drv) { | |
2515 | goto exit; | |
2516 | } | |
2517 | ||
2518 | new_top_bs = bdrv_find_overlay(active, top); | |
2519 | ||
2520 | if (new_top_bs == NULL) { | |
2521 | /* we could not find the image above 'top', this is an error */ | |
2522 | goto exit; | |
2523 | } | |
2524 | ||
2525 | /* special case of new_top_bs->backing_hd already pointing to base - nothing | |
2526 | * to do, no intermediate images */ | |
2527 | if (new_top_bs->backing_hd == base) { | |
2528 | ret = 0; | |
2529 | goto exit; | |
2530 | } | |
2531 | ||
2532 | intermediate = top; | |
2533 | ||
2534 | /* now we will go down through the list, and add each BDS we find | |
2535 | * into our deletion queue, until we hit the 'base' | |
2536 | */ | |
2537 | while (intermediate) { | |
2538 | intermediate_state = g_malloc0(sizeof(BlkIntermediateStates)); | |
2539 | intermediate_state->bs = intermediate; | |
2540 | QSIMPLEQ_INSERT_TAIL(&states_to_delete, intermediate_state, entry); | |
2541 | ||
2542 | if (intermediate->backing_hd == base) { | |
2543 | base_bs = intermediate->backing_hd; | |
2544 | break; | |
2545 | } | |
2546 | intermediate = intermediate->backing_hd; | |
2547 | } | |
2548 | if (base_bs == NULL) { | |
2549 | /* something went wrong, we did not end at the base. safely | |
2550 | * unravel everything, and exit with error */ | |
2551 | goto exit; | |
2552 | } | |
2553 | ||
2554 | /* success - we can delete the intermediate states, and link top->base */ | |
2555 | ret = bdrv_change_backing_file(new_top_bs, base_bs->filename, | |
2556 | base_bs->drv ? base_bs->drv->format_name : ""); | |
2557 | if (ret) { | |
2558 | goto exit; | |
2559 | } | |
2560 | new_top_bs->backing_hd = base_bs; | |
2561 | ||
355ef4ac | 2562 | bdrv_refresh_limits(new_top_bs); |
6ebdcee2 JC |
2563 | |
2564 | QSIMPLEQ_FOREACH_SAFE(intermediate_state, &states_to_delete, entry, next) { | |
2565 | /* so that bdrv_close() does not recursively close the chain */ | |
2566 | intermediate_state->bs->backing_hd = NULL; | |
4f6fd349 | 2567 | bdrv_unref(intermediate_state->bs); |
6ebdcee2 JC |
2568 | } |
2569 | ret = 0; | |
2570 | ||
2571 | exit: | |
2572 | QSIMPLEQ_FOREACH_SAFE(intermediate_state, &states_to_delete, entry, next) { | |
2573 | g_free(intermediate_state); | |
2574 | } | |
2575 | return ret; | |
2576 | } | |
2577 | ||
2578 | ||
71d0770c AL |
2579 | static int bdrv_check_byte_request(BlockDriverState *bs, int64_t offset, |
2580 | size_t size) | |
2581 | { | |
2582 | int64_t len; | |
2583 | ||
1dd3a447 KW |
2584 | if (size > INT_MAX) { |
2585 | return -EIO; | |
2586 | } | |
2587 | ||
71d0770c AL |
2588 | if (!bdrv_is_inserted(bs)) |
2589 | return -ENOMEDIUM; | |
2590 | ||
2591 | if (bs->growable) | |
2592 | return 0; | |
2593 | ||
2594 | len = bdrv_getlength(bs); | |
2595 | ||
fbb7b4e0 KW |
2596 | if (offset < 0) |
2597 | return -EIO; | |
2598 | ||
2599 | if ((offset > len) || (len - offset < size)) | |
71d0770c AL |
2600 | return -EIO; |
2601 | ||
2602 | return 0; | |
2603 | } | |
2604 | ||
2605 | static int bdrv_check_request(BlockDriverState *bs, int64_t sector_num, | |
2606 | int nb_sectors) | |
2607 | { | |
54db38a4 | 2608 | if (nb_sectors < 0 || nb_sectors > INT_MAX / BDRV_SECTOR_SIZE) { |
8f4754ed KW |
2609 | return -EIO; |
2610 | } | |
2611 | ||
eb5a3165 JS |
2612 | return bdrv_check_byte_request(bs, sector_num * BDRV_SECTOR_SIZE, |
2613 | nb_sectors * BDRV_SECTOR_SIZE); | |
71d0770c AL |
2614 | } |
2615 | ||
1c9805a3 SH |
2616 | typedef struct RwCo { |
2617 | BlockDriverState *bs; | |
775aa8b6 | 2618 | int64_t offset; |
1c9805a3 SH |
2619 | QEMUIOVector *qiov; |
2620 | bool is_write; | |
2621 | int ret; | |
4105eaaa | 2622 | BdrvRequestFlags flags; |
1c9805a3 SH |
2623 | } RwCo; |
2624 | ||
2625 | static void coroutine_fn bdrv_rw_co_entry(void *opaque) | |
fc01f7e7 | 2626 | { |
1c9805a3 | 2627 | RwCo *rwco = opaque; |
ea2384d3 | 2628 | |
1c9805a3 | 2629 | if (!rwco->is_write) { |
775aa8b6 KW |
2630 | rwco->ret = bdrv_co_do_preadv(rwco->bs, rwco->offset, |
2631 | rwco->qiov->size, rwco->qiov, | |
4105eaaa | 2632 | rwco->flags); |
775aa8b6 KW |
2633 | } else { |
2634 | rwco->ret = bdrv_co_do_pwritev(rwco->bs, rwco->offset, | |
2635 | rwco->qiov->size, rwco->qiov, | |
2636 | rwco->flags); | |
1c9805a3 SH |
2637 | } |
2638 | } | |
e7a8a783 | 2639 | |
1c9805a3 | 2640 | /* |
8d3b1a2d | 2641 | * Process a vectored synchronous request using coroutines |
1c9805a3 | 2642 | */ |
775aa8b6 KW |
2643 | static int bdrv_prwv_co(BlockDriverState *bs, int64_t offset, |
2644 | QEMUIOVector *qiov, bool is_write, | |
2645 | BdrvRequestFlags flags) | |
1c9805a3 | 2646 | { |
1c9805a3 SH |
2647 | Coroutine *co; |
2648 | RwCo rwco = { | |
2649 | .bs = bs, | |
775aa8b6 | 2650 | .offset = offset, |
8d3b1a2d | 2651 | .qiov = qiov, |
1c9805a3 SH |
2652 | .is_write = is_write, |
2653 | .ret = NOT_DONE, | |
4105eaaa | 2654 | .flags = flags, |
1c9805a3 | 2655 | }; |
e7a8a783 | 2656 | |
498e386c ZYW |
2657 | /** |
2658 | * In sync call context, when the vcpu is blocked, this throttling timer | |
2659 | * will not fire; so the I/O throttling function has to be disabled here | |
2660 | * if it has been enabled. | |
2661 | */ | |
2662 | if (bs->io_limits_enabled) { | |
2663 | fprintf(stderr, "Disabling I/O throttling on '%s' due " | |
2664 | "to synchronous I/O.\n", bdrv_get_device_name(bs)); | |
2665 | bdrv_io_limits_disable(bs); | |
2666 | } | |
2667 | ||
1c9805a3 SH |
2668 | if (qemu_in_coroutine()) { |
2669 | /* Fast-path if already in coroutine context */ | |
2670 | bdrv_rw_co_entry(&rwco); | |
2671 | } else { | |
2672 | co = qemu_coroutine_create(bdrv_rw_co_entry); | |
2673 | qemu_coroutine_enter(co, &rwco); | |
2674 | while (rwco.ret == NOT_DONE) { | |
2675 | qemu_aio_wait(); | |
2676 | } | |
2677 | } | |
2678 | return rwco.ret; | |
2679 | } | |
b338082b | 2680 | |
8d3b1a2d KW |
2681 | /* |
2682 | * Process a synchronous request using coroutines | |
2683 | */ | |
2684 | static int bdrv_rw_co(BlockDriverState *bs, int64_t sector_num, uint8_t *buf, | |
4105eaaa | 2685 | int nb_sectors, bool is_write, BdrvRequestFlags flags) |
8d3b1a2d KW |
2686 | { |
2687 | QEMUIOVector qiov; | |
2688 | struct iovec iov = { | |
2689 | .iov_base = (void *)buf, | |
2690 | .iov_len = nb_sectors * BDRV_SECTOR_SIZE, | |
2691 | }; | |
2692 | ||
da15ee51 KW |
2693 | if (nb_sectors < 0 || nb_sectors > INT_MAX / BDRV_SECTOR_SIZE) { |
2694 | return -EINVAL; | |
2695 | } | |
2696 | ||
8d3b1a2d | 2697 | qemu_iovec_init_external(&qiov, &iov, 1); |
775aa8b6 KW |
2698 | return bdrv_prwv_co(bs, sector_num << BDRV_SECTOR_BITS, |
2699 | &qiov, is_write, flags); | |
8d3b1a2d KW |
2700 | } |
2701 | ||
1c9805a3 SH |
2702 | /* return < 0 if error. See bdrv_write() for the return codes */ |
2703 | int bdrv_read(BlockDriverState *bs, int64_t sector_num, | |
2704 | uint8_t *buf, int nb_sectors) | |
2705 | { | |
4105eaaa | 2706 | return bdrv_rw_co(bs, sector_num, buf, nb_sectors, false, 0); |
fc01f7e7 FB |
2707 | } |
2708 | ||
07d27a44 MA |
2709 | /* Just like bdrv_read(), but with I/O throttling temporarily disabled */ |
2710 | int bdrv_read_unthrottled(BlockDriverState *bs, int64_t sector_num, | |
2711 | uint8_t *buf, int nb_sectors) | |
2712 | { | |
2713 | bool enabled; | |
2714 | int ret; | |
2715 | ||
2716 | enabled = bs->io_limits_enabled; | |
2717 | bs->io_limits_enabled = false; | |
4e7395e8 | 2718 | ret = bdrv_read(bs, sector_num, buf, nb_sectors); |
07d27a44 MA |
2719 | bs->io_limits_enabled = enabled; |
2720 | return ret; | |
2721 | } | |
2722 | ||
5fafdf24 | 2723 | /* Return < 0 if error. Important errors are: |
19cb3738 FB |
2724 | -EIO generic I/O error (may happen for all errors) |
2725 | -ENOMEDIUM No media inserted. | |
2726 | -EINVAL Invalid sector number or nb_sectors | |
2727 | -EACCES Trying to write a read-only device | |
2728 | */ | |
5fafdf24 | 2729 | int bdrv_write(BlockDriverState *bs, int64_t sector_num, |
fc01f7e7 FB |
2730 | const uint8_t *buf, int nb_sectors) |
2731 | { | |
4105eaaa | 2732 | return bdrv_rw_co(bs, sector_num, (uint8_t *)buf, nb_sectors, true, 0); |
83f64091 FB |
2733 | } |
2734 | ||
aa7bfbff PL |
2735 | int bdrv_write_zeroes(BlockDriverState *bs, int64_t sector_num, |
2736 | int nb_sectors, BdrvRequestFlags flags) | |
4105eaaa PL |
2737 | { |
2738 | return bdrv_rw_co(bs, sector_num, NULL, nb_sectors, true, | |
aa7bfbff | 2739 | BDRV_REQ_ZERO_WRITE | flags); |
8d3b1a2d KW |
2740 | } |
2741 | ||
d75cbb5e PL |
2742 | /* |
2743 | * Completely zero out a block device with the help of bdrv_write_zeroes. | |
2744 | * The operation is sped up by checking the block status and only writing | |
2745 | * zeroes to the device if they currently do not return zeroes. Optional | |
2746 | * flags are passed through to bdrv_write_zeroes (e.g. BDRV_REQ_MAY_UNMAP). | |
2747 | * | |
2748 | * Returns < 0 on error, 0 on success. For error codes see bdrv_write(). | |
2749 | */ | |
2750 | int bdrv_make_zero(BlockDriverState *bs, BdrvRequestFlags flags) | |
2751 | { | |
9ce10c0b | 2752 | int64_t target_size; |
d75cbb5e PL |
2753 | int64_t ret, nb_sectors, sector_num = 0; |
2754 | int n; | |
2755 | ||
9ce10c0b KW |
2756 | target_size = bdrv_getlength(bs); |
2757 | if (target_size < 0) { | |
2758 | return target_size; | |
2759 | } | |
2760 | target_size /= BDRV_SECTOR_SIZE; | |
2761 | ||
d75cbb5e PL |
2762 | for (;;) { |
2763 | nb_sectors = target_size - sector_num; | |
2764 | if (nb_sectors <= 0) { | |
2765 | return 0; | |
2766 | } | |
2767 | if (nb_sectors > INT_MAX) { | |
2768 | nb_sectors = INT_MAX; | |
2769 | } | |
2770 | ret = bdrv_get_block_status(bs, sector_num, nb_sectors, &n); | |
3d94ce60 PL |
2771 | if (ret < 0) { |
2772 | error_report("error getting block status at sector %" PRId64 ": %s", | |
2773 | sector_num, strerror(-ret)); | |
2774 | return ret; | |
2775 | } | |
d75cbb5e PL |
2776 | if (ret & BDRV_BLOCK_ZERO) { |
2777 | sector_num += n; | |
2778 | continue; | |
2779 | } | |
2780 | ret = bdrv_write_zeroes(bs, sector_num, n, flags); | |
2781 | if (ret < 0) { | |
2782 | error_report("error writing zeroes at sector %" PRId64 ": %s", | |
2783 | sector_num, strerror(-ret)); | |
2784 | return ret; | |
2785 | } | |
2786 | sector_num += n; | |
2787 | } | |
2788 | } | |
2789 | ||
a3ef6571 | 2790 | int bdrv_pread(BlockDriverState *bs, int64_t offset, void *buf, int bytes) |
83f64091 | 2791 | { |
a3ef6571 KW |
2792 | QEMUIOVector qiov; |
2793 | struct iovec iov = { | |
2794 | .iov_base = (void *)buf, | |
2795 | .iov_len = bytes, | |
2796 | }; | |
9a8c4cce | 2797 | int ret; |
83f64091 | 2798 | |
a3ef6571 KW |
2799 | if (bytes < 0) { |
2800 | return -EINVAL; | |
83f64091 FB |
2801 | } |
2802 | ||
a3ef6571 KW |
2803 | qemu_iovec_init_external(&qiov, &iov, 1); |
2804 | ret = bdrv_prwv_co(bs, offset, &qiov, false, 0); | |
2805 | if (ret < 0) { | |
2806 | return ret; | |
83f64091 | 2807 | } |
a3ef6571 KW |
2808 | |
2809 | return bytes; | |
83f64091 FB |
2810 | } |
2811 | ||
8d3b1a2d | 2812 | int bdrv_pwritev(BlockDriverState *bs, int64_t offset, QEMUIOVector *qiov) |
83f64091 | 2813 | { |
9a8c4cce | 2814 | int ret; |
83f64091 | 2815 | |
8407d5d7 KW |
2816 | ret = bdrv_prwv_co(bs, offset, qiov, true, 0); |
2817 | if (ret < 0) { | |
2818 | return ret; | |
83f64091 FB |
2819 | } |
2820 | ||
8d3b1a2d KW |
2821 | return qiov->size; |
2822 | } | |
2823 | ||
2824 | int bdrv_pwrite(BlockDriverState *bs, int64_t offset, | |
8407d5d7 | 2825 | const void *buf, int bytes) |
8d3b1a2d KW |
2826 | { |
2827 | QEMUIOVector qiov; | |
2828 | struct iovec iov = { | |
2829 | .iov_base = (void *) buf, | |
8407d5d7 | 2830 | .iov_len = bytes, |
8d3b1a2d KW |
2831 | }; |
2832 | ||
8407d5d7 KW |
2833 | if (bytes < 0) { |
2834 | return -EINVAL; | |
2835 | } | |
2836 | ||
8d3b1a2d KW |
2837 | qemu_iovec_init_external(&qiov, &iov, 1); |
2838 | return bdrv_pwritev(bs, offset, &qiov); | |
83f64091 | 2839 | } |
83f64091 | 2840 | |
f08145fe KW |
2841 | /* |
2842 | * Writes to the file and ensures that no writes are reordered across this | |
2843 | * request (acts as a barrier) | |
2844 | * | |
2845 | * Returns 0 on success, -errno in error cases. | |
2846 | */ | |
2847 | int bdrv_pwrite_sync(BlockDriverState *bs, int64_t offset, | |
2848 | const void *buf, int count) | |
2849 | { | |
2850 | int ret; | |
2851 | ||
2852 | ret = bdrv_pwrite(bs, offset, buf, count); | |
2853 | if (ret < 0) { | |
2854 | return ret; | |
2855 | } | |
2856 | ||
f05fa4ad PB |
2857 | /* No flush needed for cache modes that already do it */ |
2858 | if (bs->enable_write_cache) { | |
f08145fe KW |
2859 | bdrv_flush(bs); |
2860 | } | |
2861 | ||
2862 | return 0; | |
2863 | } | |
2864 | ||
470c0504 | 2865 | static int coroutine_fn bdrv_co_do_copy_on_readv(BlockDriverState *bs, |
ab185921 SH |
2866 | int64_t sector_num, int nb_sectors, QEMUIOVector *qiov) |
2867 | { | |
2868 | /* Perform I/O through a temporary buffer so that users who scribble over | |
2869 | * their read buffer while the operation is in progress do not end up | |
2870 | * modifying the image file. This is critical for zero-copy guest I/O | |
2871 | * where anything might happen inside guest memory. | |
2872 | */ | |
2873 | void *bounce_buffer; | |
2874 | ||
79c053bd | 2875 | BlockDriver *drv = bs->drv; |
ab185921 SH |
2876 | struct iovec iov; |
2877 | QEMUIOVector bounce_qiov; | |
2878 | int64_t cluster_sector_num; | |
2879 | int cluster_nb_sectors; | |
2880 | size_t skip_bytes; | |
2881 | int ret; | |
2882 | ||
2883 | /* Cover entire cluster so no additional backing file I/O is required when | |
2884 | * allocating cluster in the image file. | |
2885 | */ | |
343bded4 PB |
2886 | bdrv_round_to_clusters(bs, sector_num, nb_sectors, |
2887 | &cluster_sector_num, &cluster_nb_sectors); | |
ab185921 | 2888 | |
470c0504 SH |
2889 | trace_bdrv_co_do_copy_on_readv(bs, sector_num, nb_sectors, |
2890 | cluster_sector_num, cluster_nb_sectors); | |
ab185921 SH |
2891 | |
2892 | iov.iov_len = cluster_nb_sectors * BDRV_SECTOR_SIZE; | |
2893 | iov.iov_base = bounce_buffer = qemu_blockalign(bs, iov.iov_len); | |
2894 | qemu_iovec_init_external(&bounce_qiov, &iov, 1); | |
2895 | ||
79c053bd SH |
2896 | ret = drv->bdrv_co_readv(bs, cluster_sector_num, cluster_nb_sectors, |
2897 | &bounce_qiov); | |
ab185921 SH |
2898 | if (ret < 0) { |
2899 | goto err; | |
2900 | } | |
2901 | ||
79c053bd SH |
2902 | if (drv->bdrv_co_write_zeroes && |
2903 | buffer_is_zero(bounce_buffer, iov.iov_len)) { | |
621f0589 | 2904 | ret = bdrv_co_do_write_zeroes(bs, cluster_sector_num, |
aa7bfbff | 2905 | cluster_nb_sectors, 0); |
79c053bd | 2906 | } else { |
f05fa4ad PB |
2907 | /* This does not change the data on the disk, it is not necessary |
2908 | * to flush even in cache=writethrough mode. | |
2909 | */ | |
79c053bd | 2910 | ret = drv->bdrv_co_writev(bs, cluster_sector_num, cluster_nb_sectors, |
ab185921 | 2911 | &bounce_qiov); |
79c053bd SH |
2912 | } |
2913 | ||
ab185921 SH |
2914 | if (ret < 0) { |
2915 | /* It might be okay to ignore write errors for guest requests. If this | |
2916 | * is a deliberate copy-on-read then we don't want to ignore the error. | |
2917 | * Simply report it in all cases. | |
2918 | */ | |
2919 | goto err; | |
2920 | } | |
2921 | ||
2922 | skip_bytes = (sector_num - cluster_sector_num) * BDRV_SECTOR_SIZE; | |
03396148 MT |
2923 | qemu_iovec_from_buf(qiov, 0, bounce_buffer + skip_bytes, |
2924 | nb_sectors * BDRV_SECTOR_SIZE); | |
ab185921 SH |
2925 | |
2926 | err: | |
2927 | qemu_vfree(bounce_buffer); | |
2928 | return ret; | |
2929 | } | |
2930 | ||
c5fbe571 | 2931 | /* |
d0c7f642 KW |
2932 | * Forwards an already correctly aligned request to the BlockDriver. This |
2933 | * handles copy on read and zeroing after EOF; any other features must be | |
2934 | * implemented by the caller. | |
c5fbe571 | 2935 | */ |
d0c7f642 | 2936 | static int coroutine_fn bdrv_aligned_preadv(BlockDriverState *bs, |
65afd211 | 2937 | BdrvTrackedRequest *req, int64_t offset, unsigned int bytes, |
ec746e10 | 2938 | int64_t align, QEMUIOVector *qiov, int flags) |
da1fa91d KW |
2939 | { |
2940 | BlockDriver *drv = bs->drv; | |
dbffbdcf | 2941 | int ret; |
da1fa91d | 2942 | |
d0c7f642 KW |
2943 | int64_t sector_num = offset >> BDRV_SECTOR_BITS; |
2944 | unsigned int nb_sectors = bytes >> BDRV_SECTOR_BITS; | |
da1fa91d | 2945 | |
d0c7f642 KW |
2946 | assert((offset & (BDRV_SECTOR_SIZE - 1)) == 0); |
2947 | assert((bytes & (BDRV_SECTOR_SIZE - 1)) == 0); | |
2948 | ||
2949 | /* Handle Copy on Read and associated serialisation */ | |
470c0504 | 2950 | if (flags & BDRV_REQ_COPY_ON_READ) { |
7327145f KW |
2951 | /* If we touch the same cluster it counts as an overlap. This |
2952 | * guarantees that allocating writes will be serialized and not race | |
2953 | * with each other for the same cluster. For example, in copy-on-read | |
2954 | * it ensures that the CoR read and write operations are atomic and | |
2955 | * guest writes cannot interleave between them. */ | |
2956 | mark_request_serialising(req, bdrv_get_cluster_size(bs)); | |
470c0504 SH |
2957 | } |
2958 | ||
2dbafdc0 | 2959 | wait_serialising_requests(req); |
f4658285 | 2960 | |
470c0504 | 2961 | if (flags & BDRV_REQ_COPY_ON_READ) { |
ab185921 SH |
2962 | int pnum; |
2963 | ||
bdad13b9 | 2964 | ret = bdrv_is_allocated(bs, sector_num, nb_sectors, &pnum); |
ab185921 SH |
2965 | if (ret < 0) { |
2966 | goto out; | |
2967 | } | |
2968 | ||
2969 | if (!ret || pnum != nb_sectors) { | |
470c0504 | 2970 | ret = bdrv_co_do_copy_on_readv(bs, sector_num, nb_sectors, qiov); |
ab185921 SH |
2971 | goto out; |
2972 | } | |
2973 | } | |
2974 | ||
d0c7f642 | 2975 | /* Forward the request to the BlockDriver */ |
893a8f62 MK |
2976 | if (!(bs->zero_beyond_eof && bs->growable)) { |
2977 | ret = drv->bdrv_co_readv(bs, sector_num, nb_sectors, qiov); | |
2978 | } else { | |
2979 | /* Read zeros after EOF of growable BDSes */ | |
2980 | int64_t len, total_sectors, max_nb_sectors; | |
2981 | ||
2982 | len = bdrv_getlength(bs); | |
2983 | if (len < 0) { | |
2984 | ret = len; | |
2985 | goto out; | |
2986 | } | |
2987 | ||
d055a1fe | 2988 | total_sectors = DIV_ROUND_UP(len, BDRV_SECTOR_SIZE); |
5f5bcd80 KW |
2989 | max_nb_sectors = ROUND_UP(MAX(0, total_sectors - sector_num), |
2990 | align >> BDRV_SECTOR_BITS); | |
893a8f62 MK |
2991 | if (max_nb_sectors > 0) { |
2992 | ret = drv->bdrv_co_readv(bs, sector_num, | |
2993 | MIN(nb_sectors, max_nb_sectors), qiov); | |
2994 | } else { | |
2995 | ret = 0; | |
2996 | } | |
2997 | ||
2998 | /* Reading beyond end of file is supposed to produce zeroes */ | |
2999 | if (ret == 0 && total_sectors < sector_num + nb_sectors) { | |
3000 | uint64_t offset = MAX(0, total_sectors - sector_num); | |
3001 | uint64_t bytes = (sector_num + nb_sectors - offset) * | |
3002 | BDRV_SECTOR_SIZE; | |
3003 | qemu_iovec_memset(qiov, offset * BDRV_SECTOR_SIZE, 0, bytes); | |
3004 | } | |
3005 | } | |
ab185921 SH |
3006 | |
3007 | out: | |
dbffbdcf | 3008 | return ret; |
da1fa91d KW |
3009 | } |
3010 | ||
d0c7f642 KW |
3011 | /* |
3012 | * Handle a read request in coroutine context | |
3013 | */ | |
1b0288ae KW |
3014 | static int coroutine_fn bdrv_co_do_preadv(BlockDriverState *bs, |
3015 | int64_t offset, unsigned int bytes, QEMUIOVector *qiov, | |
d0c7f642 KW |
3016 | BdrvRequestFlags flags) |
3017 | { | |
3018 | BlockDriver *drv = bs->drv; | |
65afd211 KW |
3019 | BdrvTrackedRequest req; |
3020 | ||
1b0288ae KW |
3021 | /* TODO Lift BDRV_SECTOR_SIZE restriction in BlockDriver interface */ |
3022 | uint64_t align = MAX(BDRV_SECTOR_SIZE, bs->request_alignment); | |
3023 | uint8_t *head_buf = NULL; | |
3024 | uint8_t *tail_buf = NULL; | |
3025 | QEMUIOVector local_qiov; | |
3026 | bool use_local_qiov = false; | |
d0c7f642 KW |
3027 | int ret; |
3028 | ||
3029 | if (!drv) { | |
3030 | return -ENOMEDIUM; | |
3031 | } | |
1b0288ae | 3032 | if (bdrv_check_byte_request(bs, offset, bytes)) { |
d0c7f642 KW |
3033 | return -EIO; |
3034 | } | |
3035 | ||
3036 | if (bs->copy_on_read) { | |
3037 | flags |= BDRV_REQ_COPY_ON_READ; | |
3038 | } | |
3039 | ||
3040 | /* throttling disk I/O */ | |
3041 | if (bs->io_limits_enabled) { | |
d5103588 | 3042 | bdrv_io_limits_intercept(bs, bytes, false); |
1b0288ae KW |
3043 | } |
3044 | ||
3045 | /* Align read if necessary by padding qiov */ | |
3046 | if (offset & (align - 1)) { | |
3047 | head_buf = qemu_blockalign(bs, align); | |
3048 | qemu_iovec_init(&local_qiov, qiov->niov + 2); | |
3049 | qemu_iovec_add(&local_qiov, head_buf, offset & (align - 1)); | |
3050 | qemu_iovec_concat(&local_qiov, qiov, 0, qiov->size); | |
3051 | use_local_qiov = true; | |
3052 | ||
3053 | bytes += offset & (align - 1); | |
3054 | offset = offset & ~(align - 1); | |
3055 | } | |
3056 | ||
3057 | if ((offset + bytes) & (align - 1)) { | |
3058 | if (!use_local_qiov) { | |
3059 | qemu_iovec_init(&local_qiov, qiov->niov + 1); | |
3060 | qemu_iovec_concat(&local_qiov, qiov, 0, qiov->size); | |
3061 | use_local_qiov = true; | |
3062 | } | |
3063 | tail_buf = qemu_blockalign(bs, align); | |
3064 | qemu_iovec_add(&local_qiov, tail_buf, | |
3065 | align - ((offset + bytes) & (align - 1))); | |
3066 | ||
3067 | bytes = ROUND_UP(bytes, align); | |
3068 | } | |
3069 | ||
65afd211 | 3070 | tracked_request_begin(&req, bs, offset, bytes, false); |
ec746e10 | 3071 | ret = bdrv_aligned_preadv(bs, &req, offset, bytes, align, |
1b0288ae KW |
3072 | use_local_qiov ? &local_qiov : qiov, |
3073 | flags); | |
65afd211 | 3074 | tracked_request_end(&req); |
1b0288ae KW |
3075 | |
3076 | if (use_local_qiov) { | |
3077 | qemu_iovec_destroy(&local_qiov); | |
3078 | qemu_vfree(head_buf); | |
3079 | qemu_vfree(tail_buf); | |
d0c7f642 KW |
3080 | } |
3081 | ||
d0c7f642 KW |
3082 | return ret; |
3083 | } | |
3084 | ||
1b0288ae KW |
3085 | static int coroutine_fn bdrv_co_do_readv(BlockDriverState *bs, |
3086 | int64_t sector_num, int nb_sectors, QEMUIOVector *qiov, | |
3087 | BdrvRequestFlags flags) | |
3088 | { | |
3089 | if (nb_sectors < 0 || nb_sectors > (UINT_MAX >> BDRV_SECTOR_BITS)) { | |
3090 | return -EINVAL; | |
3091 | } | |
3092 | ||
3093 | return bdrv_co_do_preadv(bs, sector_num << BDRV_SECTOR_BITS, | |
3094 | nb_sectors << BDRV_SECTOR_BITS, qiov, flags); | |
3095 | } | |
3096 | ||
c5fbe571 | 3097 | int coroutine_fn bdrv_co_readv(BlockDriverState *bs, int64_t sector_num, |
da1fa91d KW |
3098 | int nb_sectors, QEMUIOVector *qiov) |
3099 | { | |
c5fbe571 | 3100 | trace_bdrv_co_readv(bs, sector_num, nb_sectors); |
da1fa91d | 3101 | |
470c0504 SH |
3102 | return bdrv_co_do_readv(bs, sector_num, nb_sectors, qiov, 0); |
3103 | } | |
3104 | ||
3105 | int coroutine_fn bdrv_co_copy_on_readv(BlockDriverState *bs, | |
3106 | int64_t sector_num, int nb_sectors, QEMUIOVector *qiov) | |
3107 | { | |
3108 | trace_bdrv_co_copy_on_readv(bs, sector_num, nb_sectors); | |
3109 | ||
3110 | return bdrv_co_do_readv(bs, sector_num, nb_sectors, qiov, | |
3111 | BDRV_REQ_COPY_ON_READ); | |
c5fbe571 SH |
3112 | } |
3113 | ||
c31cb707 PL |
3114 | /* if no limit is specified in the BlockLimits use a default |
3115 | * of 32768 512-byte sectors (16 MiB) per request. | |
3116 | */ | |
3117 | #define MAX_WRITE_ZEROES_DEFAULT 32768 | |
3118 | ||
f08f2dda | 3119 | static int coroutine_fn bdrv_co_do_write_zeroes(BlockDriverState *bs, |
aa7bfbff | 3120 | int64_t sector_num, int nb_sectors, BdrvRequestFlags flags) |
f08f2dda SH |
3121 | { |
3122 | BlockDriver *drv = bs->drv; | |
3123 | QEMUIOVector qiov; | |
c31cb707 PL |
3124 | struct iovec iov = {0}; |
3125 | int ret = 0; | |
f08f2dda | 3126 | |
c31cb707 PL |
3127 | int max_write_zeroes = bs->bl.max_write_zeroes ? |
3128 | bs->bl.max_write_zeroes : MAX_WRITE_ZEROES_DEFAULT; | |
621f0589 | 3129 | |
c31cb707 PL |
3130 | while (nb_sectors > 0 && !ret) { |
3131 | int num = nb_sectors; | |
3132 | ||
b8d71c09 PB |
3133 | /* Align request. Block drivers can expect the "bulk" of the request |
3134 | * to be aligned. | |
3135 | */ | |
3136 | if (bs->bl.write_zeroes_alignment | |
3137 | && num > bs->bl.write_zeroes_alignment) { | |
3138 | if (sector_num % bs->bl.write_zeroes_alignment != 0) { | |
3139 | /* Make a small request up to the first aligned sector. */ | |
c31cb707 | 3140 | num = bs->bl.write_zeroes_alignment; |
b8d71c09 PB |
3141 | num -= sector_num % bs->bl.write_zeroes_alignment; |
3142 | } else if ((sector_num + num) % bs->bl.write_zeroes_alignment != 0) { | |
3143 | /* Shorten the request to the last aligned sector. num cannot | |
3144 | * underflow because num > bs->bl.write_zeroes_alignment. | |
3145 | */ | |
3146 | num -= (sector_num + num) % bs->bl.write_zeroes_alignment; | |
c31cb707 | 3147 | } |
621f0589 | 3148 | } |
f08f2dda | 3149 | |
c31cb707 PL |
3150 | /* limit request size */ |
3151 | if (num > max_write_zeroes) { | |
3152 | num = max_write_zeroes; | |
3153 | } | |
3154 | ||
3155 | ret = -ENOTSUP; | |
3156 | /* First try the efficient write zeroes operation */ | |
3157 | if (drv->bdrv_co_write_zeroes) { | |
3158 | ret = drv->bdrv_co_write_zeroes(bs, sector_num, num, flags); | |
3159 | } | |
3160 | ||
3161 | if (ret == -ENOTSUP) { | |
3162 | /* Fall back to bounce buffer if write zeroes is unsupported */ | |
3163 | iov.iov_len = num * BDRV_SECTOR_SIZE; | |
3164 | if (iov.iov_base == NULL) { | |
b8d71c09 PB |
3165 | iov.iov_base = qemu_blockalign(bs, num * BDRV_SECTOR_SIZE); |
3166 | memset(iov.iov_base, 0, num * BDRV_SECTOR_SIZE); | |
c31cb707 PL |
3167 | } |
3168 | qemu_iovec_init_external(&qiov, &iov, 1); | |
f08f2dda | 3169 | |
c31cb707 | 3170 | ret = drv->bdrv_co_writev(bs, sector_num, num, &qiov); |
b8d71c09 PB |
3171 | |
3172 | /* Keep bounce buffer around if it is big enough for all | |
3173 | * all future requests. | |
3174 | */ | |
3175 | if (num < max_write_zeroes) { | |
3176 | qemu_vfree(iov.iov_base); | |
3177 | iov.iov_base = NULL; | |
3178 | } | |
c31cb707 PL |
3179 | } |
3180 | ||
3181 | sector_num += num; | |
3182 | nb_sectors -= num; | |
3183 | } | |
f08f2dda SH |
3184 | |
3185 | qemu_vfree(iov.iov_base); | |
3186 | return ret; | |
3187 | } | |
3188 | ||
c5fbe571 | 3189 | /* |
b404f720 | 3190 | * Forwards an already correctly aligned write request to the BlockDriver. |
c5fbe571 | 3191 | */ |
b404f720 | 3192 | static int coroutine_fn bdrv_aligned_pwritev(BlockDriverState *bs, |
65afd211 KW |
3193 | BdrvTrackedRequest *req, int64_t offset, unsigned int bytes, |
3194 | QEMUIOVector *qiov, int flags) | |
c5fbe571 SH |
3195 | { |
3196 | BlockDriver *drv = bs->drv; | |
28de2dcd | 3197 | bool waited; |
6b7cb247 | 3198 | int ret; |
da1fa91d | 3199 | |
b404f720 KW |
3200 | int64_t sector_num = offset >> BDRV_SECTOR_BITS; |
3201 | unsigned int nb_sectors = bytes >> BDRV_SECTOR_BITS; | |
f4658285 | 3202 | |
b404f720 KW |
3203 | assert((offset & (BDRV_SECTOR_SIZE - 1)) == 0); |
3204 | assert((bytes & (BDRV_SECTOR_SIZE - 1)) == 0); | |
cc0681c4 | 3205 | |
28de2dcd KW |
3206 | waited = wait_serialising_requests(req); |
3207 | assert(!waited || !req->serialising); | |
af91f9a7 KW |
3208 | assert(req->overlap_offset <= offset); |
3209 | assert(offset + bytes <= req->overlap_offset + req->overlap_bytes); | |
244eadef | 3210 | |
65afd211 | 3211 | ret = notifier_with_return_list_notify(&bs->before_write_notifiers, req); |
d616b224 SH |
3212 | |
3213 | if (ret < 0) { | |
3214 | /* Do nothing, write notifier decided to fail this request */ | |
3215 | } else if (flags & BDRV_REQ_ZERO_WRITE) { | |
9e1cb96d | 3216 | BLKDBG_EVENT(bs, BLKDBG_PWRITEV_ZERO); |
aa7bfbff | 3217 | ret = bdrv_co_do_write_zeroes(bs, sector_num, nb_sectors, flags); |
f08f2dda | 3218 | } else { |
9e1cb96d | 3219 | BLKDBG_EVENT(bs, BLKDBG_PWRITEV); |
f08f2dda SH |
3220 | ret = drv->bdrv_co_writev(bs, sector_num, nb_sectors, qiov); |
3221 | } | |
9e1cb96d | 3222 | BLKDBG_EVENT(bs, BLKDBG_PWRITEV_DONE); |
6b7cb247 | 3223 | |
f05fa4ad PB |
3224 | if (ret == 0 && !bs->enable_write_cache) { |
3225 | ret = bdrv_co_flush(bs); | |
3226 | } | |
3227 | ||
e4654d2d | 3228 | bdrv_set_dirty(bs, sector_num, nb_sectors); |
da1fa91d KW |
3229 | |
3230 | if (bs->wr_highest_sector < sector_num + nb_sectors - 1) { | |
3231 | bs->wr_highest_sector = sector_num + nb_sectors - 1; | |
3232 | } | |
df2a6f29 PB |
3233 | if (bs->growable && ret >= 0) { |
3234 | bs->total_sectors = MAX(bs->total_sectors, sector_num + nb_sectors); | |
3235 | } | |
da1fa91d | 3236 | |
6b7cb247 | 3237 | return ret; |
da1fa91d KW |
3238 | } |
3239 | ||
b404f720 KW |
3240 | /* |
3241 | * Handle a write request in coroutine context | |
3242 | */ | |
6601553e KW |
3243 | static int coroutine_fn bdrv_co_do_pwritev(BlockDriverState *bs, |
3244 | int64_t offset, unsigned int bytes, QEMUIOVector *qiov, | |
b404f720 KW |
3245 | BdrvRequestFlags flags) |
3246 | { | |
65afd211 | 3247 | BdrvTrackedRequest req; |
3b8242e0 KW |
3248 | /* TODO Lift BDRV_SECTOR_SIZE restriction in BlockDriver interface */ |
3249 | uint64_t align = MAX(BDRV_SECTOR_SIZE, bs->request_alignment); | |
3250 | uint8_t *head_buf = NULL; | |
3251 | uint8_t *tail_buf = NULL; | |
3252 | QEMUIOVector local_qiov; | |
3253 | bool use_local_qiov = false; | |
b404f720 KW |
3254 | int ret; |
3255 | ||
3256 | if (!bs->drv) { | |
3257 | return -ENOMEDIUM; | |
3258 | } | |
3259 | if (bs->read_only) { | |
3260 | return -EACCES; | |
3261 | } | |
6601553e | 3262 | if (bdrv_check_byte_request(bs, offset, bytes)) { |
b404f720 KW |
3263 | return -EIO; |
3264 | } | |
3265 | ||
b404f720 KW |
3266 | /* throttling disk I/O */ |
3267 | if (bs->io_limits_enabled) { | |
d5103588 | 3268 | bdrv_io_limits_intercept(bs, bytes, true); |
b404f720 KW |
3269 | } |
3270 | ||
3b8242e0 KW |
3271 | /* |
3272 | * Align write if necessary by performing a read-modify-write cycle. | |
3273 | * Pad qiov with the read parts and be sure to have a tracked request not | |
3274 | * only for bdrv_aligned_pwritev, but also for the reads of the RMW cycle. | |
3275 | */ | |
65afd211 | 3276 | tracked_request_begin(&req, bs, offset, bytes, true); |
3b8242e0 KW |
3277 | |
3278 | if (offset & (align - 1)) { | |
3279 | QEMUIOVector head_qiov; | |
3280 | struct iovec head_iov; | |
3281 | ||
3282 | mark_request_serialising(&req, align); | |
3283 | wait_serialising_requests(&req); | |
3284 | ||
3285 | head_buf = qemu_blockalign(bs, align); | |
3286 | head_iov = (struct iovec) { | |
3287 | .iov_base = head_buf, | |
3288 | .iov_len = align, | |
3289 | }; | |
3290 | qemu_iovec_init_external(&head_qiov, &head_iov, 1); | |
3291 | ||
9e1cb96d | 3292 | BLKDBG_EVENT(bs, BLKDBG_PWRITEV_RMW_HEAD); |
3b8242e0 KW |
3293 | ret = bdrv_aligned_preadv(bs, &req, offset & ~(align - 1), align, |
3294 | align, &head_qiov, 0); | |
3295 | if (ret < 0) { | |
3296 | goto fail; | |
3297 | } | |
9e1cb96d | 3298 | BLKDBG_EVENT(bs, BLKDBG_PWRITEV_RMW_AFTER_HEAD); |
3b8242e0 KW |
3299 | |
3300 | qemu_iovec_init(&local_qiov, qiov->niov + 2); | |
3301 | qemu_iovec_add(&local_qiov, head_buf, offset & (align - 1)); | |
3302 | qemu_iovec_concat(&local_qiov, qiov, 0, qiov->size); | |
3303 | use_local_qiov = true; | |
3304 | ||
3305 | bytes += offset & (align - 1); | |
3306 | offset = offset & ~(align - 1); | |
3307 | } | |
3308 | ||
3309 | if ((offset + bytes) & (align - 1)) { | |
3310 | QEMUIOVector tail_qiov; | |
3311 | struct iovec tail_iov; | |
3312 | size_t tail_bytes; | |
28de2dcd | 3313 | bool waited; |
3b8242e0 KW |
3314 | |
3315 | mark_request_serialising(&req, align); | |
28de2dcd KW |
3316 | waited = wait_serialising_requests(&req); |
3317 | assert(!waited || !use_local_qiov); | |
3b8242e0 KW |
3318 | |
3319 | tail_buf = qemu_blockalign(bs, align); | |
3320 | tail_iov = (struct iovec) { | |
3321 | .iov_base = tail_buf, | |
3322 | .iov_len = align, | |
3323 | }; | |
3324 | qemu_iovec_init_external(&tail_qiov, &tail_iov, 1); | |
3325 | ||
9e1cb96d | 3326 | BLKDBG_EVENT(bs, BLKDBG_PWRITEV_RMW_TAIL); |
3b8242e0 KW |
3327 | ret = bdrv_aligned_preadv(bs, &req, (offset + bytes) & ~(align - 1), align, |
3328 | align, &tail_qiov, 0); | |
3329 | if (ret < 0) { | |
3330 | goto fail; | |
3331 | } | |
9e1cb96d | 3332 | BLKDBG_EVENT(bs, BLKDBG_PWRITEV_RMW_AFTER_TAIL); |
3b8242e0 KW |
3333 | |
3334 | if (!use_local_qiov) { | |
3335 | qemu_iovec_init(&local_qiov, qiov->niov + 1); | |
3336 | qemu_iovec_concat(&local_qiov, qiov, 0, qiov->size); | |
3337 | use_local_qiov = true; | |
3338 | } | |
3339 | ||
3340 | tail_bytes = (offset + bytes) & (align - 1); | |
3341 | qemu_iovec_add(&local_qiov, tail_buf + tail_bytes, align - tail_bytes); | |
3342 | ||
3343 | bytes = ROUND_UP(bytes, align); | |
3344 | } | |
3345 | ||
3346 | ret = bdrv_aligned_pwritev(bs, &req, offset, bytes, | |
3347 | use_local_qiov ? &local_qiov : qiov, | |
3348 | flags); | |
3349 | ||
3350 | fail: | |
65afd211 | 3351 | tracked_request_end(&req); |
b404f720 | 3352 | |
3b8242e0 KW |
3353 | if (use_local_qiov) { |
3354 | qemu_iovec_destroy(&local_qiov); | |
3b8242e0 | 3355 | } |
99c4a85c KW |
3356 | qemu_vfree(head_buf); |
3357 | qemu_vfree(tail_buf); | |
3b8242e0 | 3358 | |
b404f720 KW |
3359 | return ret; |
3360 | } | |
3361 | ||
6601553e KW |
3362 | static int coroutine_fn bdrv_co_do_writev(BlockDriverState *bs, |
3363 | int64_t sector_num, int nb_sectors, QEMUIOVector *qiov, | |
3364 | BdrvRequestFlags flags) | |
3365 | { | |
3366 | if (nb_sectors < 0 || nb_sectors > (INT_MAX >> BDRV_SECTOR_BITS)) { | |
3367 | return -EINVAL; | |
3368 | } | |
3369 | ||
3370 | return bdrv_co_do_pwritev(bs, sector_num << BDRV_SECTOR_BITS, | |
3371 | nb_sectors << BDRV_SECTOR_BITS, qiov, flags); | |
3372 | } | |
3373 | ||
c5fbe571 SH |
3374 | int coroutine_fn bdrv_co_writev(BlockDriverState *bs, int64_t sector_num, |
3375 | int nb_sectors, QEMUIOVector *qiov) | |
3376 | { | |
3377 | trace_bdrv_co_writev(bs, sector_num, nb_sectors); | |
3378 | ||
f08f2dda SH |
3379 | return bdrv_co_do_writev(bs, sector_num, nb_sectors, qiov, 0); |
3380 | } | |
3381 | ||
3382 | int coroutine_fn bdrv_co_write_zeroes(BlockDriverState *bs, | |
aa7bfbff PL |
3383 | int64_t sector_num, int nb_sectors, |
3384 | BdrvRequestFlags flags) | |
f08f2dda | 3385 | { |
94d6ff21 | 3386 | trace_bdrv_co_write_zeroes(bs, sector_num, nb_sectors, flags); |
f08f2dda | 3387 | |
d32f35cb PL |
3388 | if (!(bs->open_flags & BDRV_O_UNMAP)) { |
3389 | flags &= ~BDRV_REQ_MAY_UNMAP; | |
3390 | } | |
3391 | ||
f08f2dda | 3392 | return bdrv_co_do_writev(bs, sector_num, nb_sectors, NULL, |
aa7bfbff | 3393 | BDRV_REQ_ZERO_WRITE | flags); |
c5fbe571 SH |
3394 | } |
3395 | ||
83f64091 FB |
3396 | /** |
3397 | * Truncate file to 'offset' bytes (needed only for file protocols) | |
3398 | */ | |
3399 | int bdrv_truncate(BlockDriverState *bs, int64_t offset) | |
3400 | { | |
3401 | BlockDriver *drv = bs->drv; | |
51762288 | 3402 | int ret; |
83f64091 | 3403 | if (!drv) |
19cb3738 | 3404 | return -ENOMEDIUM; |
83f64091 FB |
3405 | if (!drv->bdrv_truncate) |
3406 | return -ENOTSUP; | |
59f2689d NS |
3407 | if (bs->read_only) |
3408 | return -EACCES; | |
8591675f MT |
3409 | if (bdrv_in_use(bs)) |
3410 | return -EBUSY; | |
51762288 SH |
3411 | ret = drv->bdrv_truncate(bs, offset); |
3412 | if (ret == 0) { | |
3413 | ret = refresh_total_sectors(bs, offset >> BDRV_SECTOR_BITS); | |
145feb17 | 3414 | bdrv_dev_resize_cb(bs); |
51762288 SH |
3415 | } |
3416 | return ret; | |
83f64091 FB |
3417 | } |
3418 | ||
4a1d5e1f FZ |
3419 | /** |
3420 | * Length of a allocated file in bytes. Sparse files are counted by actual | |
3421 | * allocated space. Return < 0 if error or unknown. | |
3422 | */ | |
3423 | int64_t bdrv_get_allocated_file_size(BlockDriverState *bs) | |
3424 | { | |
3425 | BlockDriver *drv = bs->drv; | |
3426 | if (!drv) { | |
3427 | return -ENOMEDIUM; | |
3428 | } | |
3429 | if (drv->bdrv_get_allocated_file_size) { | |
3430 | return drv->bdrv_get_allocated_file_size(bs); | |
3431 | } | |
3432 | if (bs->file) { | |
3433 | return bdrv_get_allocated_file_size(bs->file); | |
3434 | } | |
3435 | return -ENOTSUP; | |
3436 | } | |
3437 | ||
83f64091 FB |
3438 | /** |
3439 | * Length of a file in bytes. Return < 0 if error or unknown. | |
3440 | */ | |
3441 | int64_t bdrv_getlength(BlockDriverState *bs) | |
3442 | { | |
3443 | BlockDriver *drv = bs->drv; | |
3444 | if (!drv) | |
19cb3738 | 3445 | return -ENOMEDIUM; |
51762288 | 3446 | |
b94a2610 KW |
3447 | if (drv->has_variable_length) { |
3448 | int ret = refresh_total_sectors(bs, bs->total_sectors); | |
3449 | if (ret < 0) { | |
3450 | return ret; | |
46a4e4e6 | 3451 | } |
83f64091 | 3452 | } |
46a4e4e6 | 3453 | return bs->total_sectors * BDRV_SECTOR_SIZE; |
fc01f7e7 FB |
3454 | } |
3455 | ||
19cb3738 | 3456 | /* return 0 as number of sectors if no device present or error */ |
96b8f136 | 3457 | void bdrv_get_geometry(BlockDriverState *bs, uint64_t *nb_sectors_ptr) |
fc01f7e7 | 3458 | { |
19cb3738 FB |
3459 | int64_t length; |
3460 | length = bdrv_getlength(bs); | |
3461 | if (length < 0) | |
3462 | length = 0; | |
3463 | else | |
6ea44308 | 3464 | length = length >> BDRV_SECTOR_BITS; |
19cb3738 | 3465 | *nb_sectors_ptr = length; |
fc01f7e7 | 3466 | } |
cf98951b | 3467 | |
ff06f5f3 PB |
3468 | void bdrv_set_on_error(BlockDriverState *bs, BlockdevOnError on_read_error, |
3469 | BlockdevOnError on_write_error) | |
abd7f68d MA |
3470 | { |
3471 | bs->on_read_error = on_read_error; | |
3472 | bs->on_write_error = on_write_error; | |
3473 | } | |
3474 | ||
1ceee0d5 | 3475 | BlockdevOnError bdrv_get_on_error(BlockDriverState *bs, bool is_read) |
abd7f68d MA |
3476 | { |
3477 | return is_read ? bs->on_read_error : bs->on_write_error; | |
3478 | } | |
3479 | ||
3e1caa5f PB |
3480 | BlockErrorAction bdrv_get_error_action(BlockDriverState *bs, bool is_read, int error) |
3481 | { | |
3482 | BlockdevOnError on_err = is_read ? bs->on_read_error : bs->on_write_error; | |
3483 | ||
3484 | switch (on_err) { | |
3485 | case BLOCKDEV_ON_ERROR_ENOSPC: | |
3486 | return (error == ENOSPC) ? BDRV_ACTION_STOP : BDRV_ACTION_REPORT; | |
3487 | case BLOCKDEV_ON_ERROR_STOP: | |
3488 | return BDRV_ACTION_STOP; | |
3489 | case BLOCKDEV_ON_ERROR_REPORT: | |
3490 | return BDRV_ACTION_REPORT; | |
3491 | case BLOCKDEV_ON_ERROR_IGNORE: | |
3492 | return BDRV_ACTION_IGNORE; | |
3493 | default: | |
3494 | abort(); | |
3495 | } | |
3496 | } | |
3497 | ||
3498 | /* This is done by device models because, while the block layer knows | |
3499 | * about the error, it does not know whether an operation comes from | |
3500 | * the device or the block layer (from a job, for example). | |
3501 | */ | |
3502 | void bdrv_error_action(BlockDriverState *bs, BlockErrorAction action, | |
3503 | bool is_read, int error) | |
3504 | { | |
3505 | assert(error >= 0); | |
32c81a4a | 3506 | bdrv_emit_qmp_error_event(bs, QEVENT_BLOCK_IO_ERROR, action, is_read); |
3e1caa5f PB |
3507 | if (action == BDRV_ACTION_STOP) { |
3508 | vm_stop(RUN_STATE_IO_ERROR); | |
3509 | bdrv_iostatus_set_err(bs, error); | |
3510 | } | |
3511 | } | |
3512 | ||
b338082b FB |
3513 | int bdrv_is_read_only(BlockDriverState *bs) |
3514 | { | |
3515 | return bs->read_only; | |
3516 | } | |
3517 | ||
985a03b0 TS |
3518 | int bdrv_is_sg(BlockDriverState *bs) |
3519 | { | |
3520 | return bs->sg; | |
3521 | } | |
3522 | ||
e900a7b7 CH |
3523 | int bdrv_enable_write_cache(BlockDriverState *bs) |
3524 | { | |
3525 | return bs->enable_write_cache; | |
3526 | } | |
3527 | ||
425b0148 PB |
3528 | void bdrv_set_enable_write_cache(BlockDriverState *bs, bool wce) |
3529 | { | |
3530 | bs->enable_write_cache = wce; | |
55b110f2 JC |
3531 | |
3532 | /* so a reopen() will preserve wce */ | |
3533 | if (wce) { | |
3534 | bs->open_flags |= BDRV_O_CACHE_WB; | |
3535 | } else { | |
3536 | bs->open_flags &= ~BDRV_O_CACHE_WB; | |
3537 | } | |
425b0148 PB |
3538 | } |
3539 | ||
ea2384d3 FB |
3540 | int bdrv_is_encrypted(BlockDriverState *bs) |
3541 | { | |
3542 | if (bs->backing_hd && bs->backing_hd->encrypted) | |
3543 | return 1; | |
3544 | return bs->encrypted; | |
3545 | } | |
3546 | ||
c0f4ce77 AL |
3547 | int bdrv_key_required(BlockDriverState *bs) |
3548 | { | |
3549 | BlockDriverState *backing_hd = bs->backing_hd; | |
3550 | ||
3551 | if (backing_hd && backing_hd->encrypted && !backing_hd->valid_key) | |
3552 | return 1; | |
3553 | return (bs->encrypted && !bs->valid_key); | |
3554 | } | |
3555 | ||
ea2384d3 FB |
3556 | int bdrv_set_key(BlockDriverState *bs, const char *key) |
3557 | { | |
3558 | int ret; | |
3559 | if (bs->backing_hd && bs->backing_hd->encrypted) { | |
3560 | ret = bdrv_set_key(bs->backing_hd, key); | |
3561 | if (ret < 0) | |
3562 | return ret; | |
3563 | if (!bs->encrypted) | |
3564 | return 0; | |
3565 | } | |
fd04a2ae SH |
3566 | if (!bs->encrypted) { |
3567 | return -EINVAL; | |
3568 | } else if (!bs->drv || !bs->drv->bdrv_set_key) { | |
3569 | return -ENOMEDIUM; | |
3570 | } | |
c0f4ce77 | 3571 | ret = bs->drv->bdrv_set_key(bs, key); |
bb5fc20f AL |
3572 | if (ret < 0) { |
3573 | bs->valid_key = 0; | |
3574 | } else if (!bs->valid_key) { | |
3575 | bs->valid_key = 1; | |
3576 | /* call the change callback now, we skipped it on open */ | |
7d4b4ba5 | 3577 | bdrv_dev_change_media_cb(bs, true); |
bb5fc20f | 3578 | } |
c0f4ce77 | 3579 | return ret; |
ea2384d3 FB |
3580 | } |
3581 | ||
f8d6bba1 | 3582 | const char *bdrv_get_format_name(BlockDriverState *bs) |
ea2384d3 | 3583 | { |
f8d6bba1 | 3584 | return bs->drv ? bs->drv->format_name : NULL; |
ea2384d3 FB |
3585 | } |
3586 | ||
5fafdf24 | 3587 | void bdrv_iterate_format(void (*it)(void *opaque, const char *name), |
ea2384d3 FB |
3588 | void *opaque) |
3589 | { | |
3590 | BlockDriver *drv; | |
3591 | ||
8a22f02a | 3592 | QLIST_FOREACH(drv, &bdrv_drivers, list) { |
ea2384d3 FB |
3593 | it(opaque, drv->format_name); |
3594 | } | |
3595 | } | |
3596 | ||
dc364f4c | 3597 | /* This function is to find block backend bs */ |
b338082b FB |
3598 | BlockDriverState *bdrv_find(const char *name) |
3599 | { | |
3600 | BlockDriverState *bs; | |
3601 | ||
dc364f4c | 3602 | QTAILQ_FOREACH(bs, &bdrv_states, device_list) { |
1b7bdbc1 | 3603 | if (!strcmp(name, bs->device_name)) { |
b338082b | 3604 | return bs; |
1b7bdbc1 | 3605 | } |
b338082b FB |
3606 | } |
3607 | return NULL; | |
3608 | } | |
3609 | ||
dc364f4c BC |
3610 | /* This function is to find a node in the bs graph */ |
3611 | BlockDriverState *bdrv_find_node(const char *node_name) | |
3612 | { | |
3613 | BlockDriverState *bs; | |
3614 | ||
3615 | assert(node_name); | |
3616 | ||
3617 | QTAILQ_FOREACH(bs, &graph_bdrv_states, node_list) { | |
3618 | if (!strcmp(node_name, bs->node_name)) { | |
3619 | return bs; | |
3620 | } | |
3621 | } | |
3622 | return NULL; | |
3623 | } | |
3624 | ||
c13163fb BC |
3625 | /* Put this QMP function here so it can access the static graph_bdrv_states. */ |
3626 | BlockDeviceInfoList *bdrv_named_nodes_list(void) | |
3627 | { | |
3628 | BlockDeviceInfoList *list, *entry; | |
3629 | BlockDriverState *bs; | |
3630 | ||
3631 | list = NULL; | |
3632 | QTAILQ_FOREACH(bs, &graph_bdrv_states, node_list) { | |
3633 | entry = g_malloc0(sizeof(*entry)); | |
3634 | entry->value = bdrv_block_device_info(bs); | |
3635 | entry->next = list; | |
3636 | list = entry; | |
3637 | } | |
3638 | ||
3639 | return list; | |
3640 | } | |
3641 | ||
12d3ba82 BC |
3642 | BlockDriverState *bdrv_lookup_bs(const char *device, |
3643 | const char *node_name, | |
3644 | Error **errp) | |
3645 | { | |
3646 | BlockDriverState *bs = NULL; | |
3647 | ||
12d3ba82 BC |
3648 | if (device) { |
3649 | bs = bdrv_find(device); | |
3650 | ||
dd67fa50 BC |
3651 | if (bs) { |
3652 | return bs; | |
12d3ba82 | 3653 | } |
12d3ba82 BC |
3654 | } |
3655 | ||
dd67fa50 BC |
3656 | if (node_name) { |
3657 | bs = bdrv_find_node(node_name); | |
12d3ba82 | 3658 | |
dd67fa50 BC |
3659 | if (bs) { |
3660 | return bs; | |
3661 | } | |
12d3ba82 BC |
3662 | } |
3663 | ||
dd67fa50 BC |
3664 | error_setg(errp, "Cannot find device=%s nor node_name=%s", |
3665 | device ? device : "", | |
3666 | node_name ? node_name : ""); | |
3667 | return NULL; | |
12d3ba82 BC |
3668 | } |
3669 | ||
2f399b0a MA |
3670 | BlockDriverState *bdrv_next(BlockDriverState *bs) |
3671 | { | |
3672 | if (!bs) { | |
3673 | return QTAILQ_FIRST(&bdrv_states); | |
3674 | } | |
dc364f4c | 3675 | return QTAILQ_NEXT(bs, device_list); |
2f399b0a MA |
3676 | } |
3677 | ||
51de9760 | 3678 | void bdrv_iterate(void (*it)(void *opaque, BlockDriverState *bs), void *opaque) |
81d0912d FB |
3679 | { |
3680 | BlockDriverState *bs; | |
3681 | ||
dc364f4c | 3682 | QTAILQ_FOREACH(bs, &bdrv_states, device_list) { |
51de9760 | 3683 | it(opaque, bs); |
81d0912d FB |
3684 | } |
3685 | } | |
3686 | ||
ea2384d3 FB |
3687 | const char *bdrv_get_device_name(BlockDriverState *bs) |
3688 | { | |
3689 | return bs->device_name; | |
3690 | } | |
3691 | ||
c8433287 MA |
3692 | int bdrv_get_flags(BlockDriverState *bs) |
3693 | { | |
3694 | return bs->open_flags; | |
3695 | } | |
3696 | ||
f0f0fdfe | 3697 | int bdrv_flush_all(void) |
c6ca28d6 AL |
3698 | { |
3699 | BlockDriverState *bs; | |
f0f0fdfe | 3700 | int result = 0; |
c6ca28d6 | 3701 | |
dc364f4c | 3702 | QTAILQ_FOREACH(bs, &bdrv_states, device_list) { |
f0f0fdfe KW |
3703 | int ret = bdrv_flush(bs); |
3704 | if (ret < 0 && !result) { | |
3705 | result = ret; | |
3706 | } | |
1b7bdbc1 | 3707 | } |
f0f0fdfe KW |
3708 | |
3709 | return result; | |
c6ca28d6 AL |
3710 | } |
3711 | ||
3ac21627 PL |
3712 | int bdrv_has_zero_init_1(BlockDriverState *bs) |
3713 | { | |
3714 | return 1; | |
3715 | } | |
3716 | ||
f2feebbd KW |
3717 | int bdrv_has_zero_init(BlockDriverState *bs) |
3718 | { | |
3719 | assert(bs->drv); | |
3720 | ||
11212d8f PB |
3721 | /* If BS is a copy on write image, it is initialized to |
3722 | the contents of the base image, which may not be zeroes. */ | |
3723 | if (bs->backing_hd) { | |
3724 | return 0; | |
3725 | } | |
336c1c12 KW |
3726 | if (bs->drv->bdrv_has_zero_init) { |
3727 | return bs->drv->bdrv_has_zero_init(bs); | |
f2feebbd KW |
3728 | } |
3729 | ||
3ac21627 PL |
3730 | /* safe default */ |
3731 | return 0; | |
f2feebbd KW |
3732 | } |
3733 | ||
4ce78691 PL |
3734 | bool bdrv_unallocated_blocks_are_zero(BlockDriverState *bs) |
3735 | { | |
3736 | BlockDriverInfo bdi; | |
3737 | ||
3738 | if (bs->backing_hd) { | |
3739 | return false; | |
3740 | } | |
3741 | ||
3742 | if (bdrv_get_info(bs, &bdi) == 0) { | |
3743 | return bdi.unallocated_blocks_are_zero; | |
3744 | } | |
3745 | ||
3746 | return false; | |
3747 | } | |
3748 | ||
3749 | bool bdrv_can_write_zeroes_with_unmap(BlockDriverState *bs) | |
3750 | { | |
3751 | BlockDriverInfo bdi; | |
3752 | ||
3753 | if (bs->backing_hd || !(bs->open_flags & BDRV_O_UNMAP)) { | |
3754 | return false; | |
3755 | } | |
3756 | ||
3757 | if (bdrv_get_info(bs, &bdi) == 0) { | |
3758 | return bdi.can_write_zeroes_with_unmap; | |
3759 | } | |
3760 | ||
3761 | return false; | |
3762 | } | |
3763 | ||
b6b8a333 | 3764 | typedef struct BdrvCoGetBlockStatusData { |
376ae3f1 | 3765 | BlockDriverState *bs; |
b35b2bba | 3766 | BlockDriverState *base; |
376ae3f1 SH |
3767 | int64_t sector_num; |
3768 | int nb_sectors; | |
3769 | int *pnum; | |
b6b8a333 | 3770 | int64_t ret; |
376ae3f1 | 3771 | bool done; |
b6b8a333 | 3772 | } BdrvCoGetBlockStatusData; |
376ae3f1 | 3773 | |
f58c7b35 TS |
3774 | /* |
3775 | * Returns true iff the specified sector is present in the disk image. Drivers | |
3776 | * not implementing the functionality are assumed to not support backing files, | |
3777 | * hence all their sectors are reported as allocated. | |
3778 | * | |
bd9533e3 SH |
3779 | * If 'sector_num' is beyond the end of the disk image the return value is 0 |
3780 | * and 'pnum' is set to 0. | |
3781 | * | |
f58c7b35 TS |
3782 | * 'pnum' is set to the number of sectors (including and immediately following |
3783 | * the specified sector) that are known to be in the same | |
3784 | * allocated/unallocated state. | |
3785 | * | |
bd9533e3 SH |
3786 | * 'nb_sectors' is the max value 'pnum' should be set to. If nb_sectors goes |
3787 | * beyond the end of the disk image it will be clamped. | |
f58c7b35 | 3788 | */ |
b6b8a333 PB |
3789 | static int64_t coroutine_fn bdrv_co_get_block_status(BlockDriverState *bs, |
3790 | int64_t sector_num, | |
3791 | int nb_sectors, int *pnum) | |
f58c7b35 | 3792 | { |
617ccb46 | 3793 | int64_t length; |
bd9533e3 | 3794 | int64_t n; |
5daa74a6 | 3795 | int64_t ret, ret2; |
bd9533e3 | 3796 | |
617ccb46 PB |
3797 | length = bdrv_getlength(bs); |
3798 | if (length < 0) { | |
3799 | return length; | |
3800 | } | |
3801 | ||
3802 | if (sector_num >= (length >> BDRV_SECTOR_BITS)) { | |
bd9533e3 SH |
3803 | *pnum = 0; |
3804 | return 0; | |
3805 | } | |
3806 | ||
3807 | n = bs->total_sectors - sector_num; | |
3808 | if (n < nb_sectors) { | |
3809 | nb_sectors = n; | |
3810 | } | |
3811 | ||
b6b8a333 | 3812 | if (!bs->drv->bdrv_co_get_block_status) { |
bd9533e3 | 3813 | *pnum = nb_sectors; |
918e92d7 PB |
3814 | ret = BDRV_BLOCK_DATA; |
3815 | if (bs->drv->protocol_name) { | |
3816 | ret |= BDRV_BLOCK_OFFSET_VALID | (sector_num * BDRV_SECTOR_SIZE); | |
3817 | } | |
3818 | return ret; | |
f58c7b35 | 3819 | } |
6aebab14 | 3820 | |
415b5b01 PB |
3821 | ret = bs->drv->bdrv_co_get_block_status(bs, sector_num, nb_sectors, pnum); |
3822 | if (ret < 0) { | |
3e0a233d | 3823 | *pnum = 0; |
415b5b01 PB |
3824 | return ret; |
3825 | } | |
3826 | ||
92bc50a5 PL |
3827 | if (ret & BDRV_BLOCK_RAW) { |
3828 | assert(ret & BDRV_BLOCK_OFFSET_VALID); | |
3829 | return bdrv_get_block_status(bs->file, ret >> BDRV_SECTOR_BITS, | |
3830 | *pnum, pnum); | |
3831 | } | |
3832 | ||
c3d86884 PL |
3833 | if (!(ret & BDRV_BLOCK_DATA) && !(ret & BDRV_BLOCK_ZERO)) { |
3834 | if (bdrv_unallocated_blocks_are_zero(bs)) { | |
f0ad5712 | 3835 | ret |= BDRV_BLOCK_ZERO; |
1f9db224 | 3836 | } else if (bs->backing_hd) { |
f0ad5712 PB |
3837 | BlockDriverState *bs2 = bs->backing_hd; |
3838 | int64_t length2 = bdrv_getlength(bs2); | |
3839 | if (length2 >= 0 && sector_num >= (length2 >> BDRV_SECTOR_BITS)) { | |
3840 | ret |= BDRV_BLOCK_ZERO; | |
3841 | } | |
3842 | } | |
415b5b01 | 3843 | } |
5daa74a6 PB |
3844 | |
3845 | if (bs->file && | |
3846 | (ret & BDRV_BLOCK_DATA) && !(ret & BDRV_BLOCK_ZERO) && | |
3847 | (ret & BDRV_BLOCK_OFFSET_VALID)) { | |
3848 | ret2 = bdrv_co_get_block_status(bs->file, ret >> BDRV_SECTOR_BITS, | |
3849 | *pnum, pnum); | |
3850 | if (ret2 >= 0) { | |
3851 | /* Ignore errors. This is just providing extra information, it | |
3852 | * is useful but not necessary. | |
3853 | */ | |
3854 | ret |= (ret2 & BDRV_BLOCK_ZERO); | |
3855 | } | |
3856 | } | |
3857 | ||
415b5b01 | 3858 | return ret; |
060f51c9 SH |
3859 | } |
3860 | ||
b6b8a333 PB |
3861 | /* Coroutine wrapper for bdrv_get_block_status() */ |
3862 | static void coroutine_fn bdrv_get_block_status_co_entry(void *opaque) | |
060f51c9 | 3863 | { |
b6b8a333 | 3864 | BdrvCoGetBlockStatusData *data = opaque; |
060f51c9 SH |
3865 | BlockDriverState *bs = data->bs; |
3866 | ||
b6b8a333 PB |
3867 | data->ret = bdrv_co_get_block_status(bs, data->sector_num, data->nb_sectors, |
3868 | data->pnum); | |
060f51c9 SH |
3869 | data->done = true; |
3870 | } | |
3871 | ||
3872 | /* | |
b6b8a333 | 3873 | * Synchronous wrapper around bdrv_co_get_block_status(). |
060f51c9 | 3874 | * |
b6b8a333 | 3875 | * See bdrv_co_get_block_status() for details. |
060f51c9 | 3876 | */ |
b6b8a333 PB |
3877 | int64_t bdrv_get_block_status(BlockDriverState *bs, int64_t sector_num, |
3878 | int nb_sectors, int *pnum) | |
060f51c9 | 3879 | { |
6aebab14 | 3880 | Coroutine *co; |
b6b8a333 | 3881 | BdrvCoGetBlockStatusData data = { |
6aebab14 SH |
3882 | .bs = bs, |
3883 | .sector_num = sector_num, | |
3884 | .nb_sectors = nb_sectors, | |
3885 | .pnum = pnum, | |
3886 | .done = false, | |
3887 | }; | |
3888 | ||
bdad13b9 PB |
3889 | if (qemu_in_coroutine()) { |
3890 | /* Fast-path if already in coroutine context */ | |
b6b8a333 | 3891 | bdrv_get_block_status_co_entry(&data); |
bdad13b9 | 3892 | } else { |
b6b8a333 | 3893 | co = qemu_coroutine_create(bdrv_get_block_status_co_entry); |
bdad13b9 PB |
3894 | qemu_coroutine_enter(co, &data); |
3895 | while (!data.done) { | |
3896 | qemu_aio_wait(); | |
3897 | } | |
6aebab14 SH |
3898 | } |
3899 | return data.ret; | |
f58c7b35 TS |
3900 | } |
3901 | ||
b6b8a333 PB |
3902 | int coroutine_fn bdrv_is_allocated(BlockDriverState *bs, int64_t sector_num, |
3903 | int nb_sectors, int *pnum) | |
3904 | { | |
4333bb71 PB |
3905 | int64_t ret = bdrv_get_block_status(bs, sector_num, nb_sectors, pnum); |
3906 | if (ret < 0) { | |
3907 | return ret; | |
3908 | } | |
3909 | return | |
3910 | (ret & BDRV_BLOCK_DATA) || | |
3911 | ((ret & BDRV_BLOCK_ZERO) && !bdrv_has_zero_init(bs)); | |
b6b8a333 PB |
3912 | } |
3913 | ||
188a7bbf PB |
3914 | /* |
3915 | * Given an image chain: ... -> [BASE] -> [INTER1] -> [INTER2] -> [TOP] | |
3916 | * | |
3917 | * Return true if the given sector is allocated in any image between | |
3918 | * BASE and TOP (inclusive). BASE can be NULL to check if the given | |
3919 | * sector is allocated in any image of the chain. Return false otherwise. | |
3920 | * | |
3921 | * 'pnum' is set to the number of sectors (including and immediately following | |
3922 | * the specified sector) that are known to be in the same | |
3923 | * allocated/unallocated state. | |
3924 | * | |
3925 | */ | |
4f578637 PB |
3926 | int bdrv_is_allocated_above(BlockDriverState *top, |
3927 | BlockDriverState *base, | |
3928 | int64_t sector_num, | |
3929 | int nb_sectors, int *pnum) | |
188a7bbf PB |
3930 | { |
3931 | BlockDriverState *intermediate; | |
3932 | int ret, n = nb_sectors; | |
3933 | ||
3934 | intermediate = top; | |
3935 | while (intermediate && intermediate != base) { | |
3936 | int pnum_inter; | |
bdad13b9 PB |
3937 | ret = bdrv_is_allocated(intermediate, sector_num, nb_sectors, |
3938 | &pnum_inter); | |
188a7bbf PB |
3939 | if (ret < 0) { |
3940 | return ret; | |
3941 | } else if (ret) { | |
3942 | *pnum = pnum_inter; | |
3943 | return 1; | |
3944 | } | |
3945 | ||
3946 | /* | |
3947 | * [sector_num, nb_sectors] is unallocated on top but intermediate | |
3948 | * might have | |
3949 | * | |
3950 | * [sector_num+x, nr_sectors] allocated. | |
3951 | */ | |
63ba17d3 VI |
3952 | if (n > pnum_inter && |
3953 | (intermediate == top || | |
3954 | sector_num + pnum_inter < intermediate->total_sectors)) { | |
188a7bbf PB |
3955 | n = pnum_inter; |
3956 | } | |
3957 | ||
3958 | intermediate = intermediate->backing_hd; | |
3959 | } | |
3960 | ||
3961 | *pnum = n; | |
3962 | return 0; | |
3963 | } | |
3964 | ||
045df330 AL |
3965 | const char *bdrv_get_encrypted_filename(BlockDriverState *bs) |
3966 | { | |
3967 | if (bs->backing_hd && bs->backing_hd->encrypted) | |
3968 | return bs->backing_file; | |
3969 | else if (bs->encrypted) | |
3970 | return bs->filename; | |
3971 | else | |
3972 | return NULL; | |
3973 | } | |
3974 | ||
5fafdf24 | 3975 | void bdrv_get_backing_filename(BlockDriverState *bs, |
83f64091 FB |
3976 | char *filename, int filename_size) |
3977 | { | |
3574c608 | 3978 | pstrcpy(filename, filename_size, bs->backing_file); |
83f64091 FB |
3979 | } |
3980 | ||
5fafdf24 | 3981 | int bdrv_write_compressed(BlockDriverState *bs, int64_t sector_num, |
faea38e7 FB |
3982 | const uint8_t *buf, int nb_sectors) |
3983 | { | |
3984 | BlockDriver *drv = bs->drv; | |
3985 | if (!drv) | |
19cb3738 | 3986 | return -ENOMEDIUM; |
faea38e7 FB |
3987 | if (!drv->bdrv_write_compressed) |
3988 | return -ENOTSUP; | |
fbb7b4e0 KW |
3989 | if (bdrv_check_request(bs, sector_num, nb_sectors)) |
3990 | return -EIO; | |
a55eb92c | 3991 | |
e4654d2d | 3992 | assert(QLIST_EMPTY(&bs->dirty_bitmaps)); |
a55eb92c | 3993 | |
faea38e7 FB |
3994 | return drv->bdrv_write_compressed(bs, sector_num, buf, nb_sectors); |
3995 | } | |
3b46e624 | 3996 | |
faea38e7 FB |
3997 | int bdrv_get_info(BlockDriverState *bs, BlockDriverInfo *bdi) |
3998 | { | |
3999 | BlockDriver *drv = bs->drv; | |
4000 | if (!drv) | |
19cb3738 | 4001 | return -ENOMEDIUM; |
faea38e7 FB |
4002 | if (!drv->bdrv_get_info) |
4003 | return -ENOTSUP; | |
4004 | memset(bdi, 0, sizeof(*bdi)); | |
4005 | return drv->bdrv_get_info(bs, bdi); | |
4006 | } | |
4007 | ||
eae041fe HR |
4008 | ImageInfoSpecific *bdrv_get_specific_info(BlockDriverState *bs) |
4009 | { | |
4010 | BlockDriver *drv = bs->drv; | |
4011 | if (drv && drv->bdrv_get_specific_info) { | |
4012 | return drv->bdrv_get_specific_info(bs); | |
4013 | } | |
4014 | return NULL; | |
4015 | } | |
4016 | ||
45566e9c CH |
4017 | int bdrv_save_vmstate(BlockDriverState *bs, const uint8_t *buf, |
4018 | int64_t pos, int size) | |
cf8074b3 KW |
4019 | { |
4020 | QEMUIOVector qiov; | |
4021 | struct iovec iov = { | |
4022 | .iov_base = (void *) buf, | |
4023 | .iov_len = size, | |
4024 | }; | |
4025 | ||
4026 | qemu_iovec_init_external(&qiov, &iov, 1); | |
4027 | return bdrv_writev_vmstate(bs, &qiov, pos); | |
4028 | } | |
4029 | ||
4030 | int bdrv_writev_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos) | |
178e08a5 AL |
4031 | { |
4032 | BlockDriver *drv = bs->drv; | |
cf8074b3 KW |
4033 | |
4034 | if (!drv) { | |
178e08a5 | 4035 | return -ENOMEDIUM; |
cf8074b3 KW |
4036 | } else if (drv->bdrv_save_vmstate) { |
4037 | return drv->bdrv_save_vmstate(bs, qiov, pos); | |
4038 | } else if (bs->file) { | |
4039 | return bdrv_writev_vmstate(bs->file, qiov, pos); | |
4040 | } | |
4041 | ||
7cdb1f6d | 4042 | return -ENOTSUP; |
178e08a5 AL |
4043 | } |
4044 | ||
45566e9c CH |
4045 | int bdrv_load_vmstate(BlockDriverState *bs, uint8_t *buf, |
4046 | int64_t pos, int size) | |
178e08a5 AL |
4047 | { |
4048 | BlockDriver *drv = bs->drv; | |
4049 | if (!drv) | |
4050 | return -ENOMEDIUM; | |
7cdb1f6d MK |
4051 | if (drv->bdrv_load_vmstate) |
4052 | return drv->bdrv_load_vmstate(bs, buf, pos, size); | |
4053 | if (bs->file) | |
4054 | return bdrv_load_vmstate(bs->file, buf, pos, size); | |
4055 | return -ENOTSUP; | |
178e08a5 AL |
4056 | } |
4057 | ||
8b9b0cc2 KW |
4058 | void bdrv_debug_event(BlockDriverState *bs, BlkDebugEvent event) |
4059 | { | |
bf736fe3 | 4060 | if (!bs || !bs->drv || !bs->drv->bdrv_debug_event) { |
8b9b0cc2 KW |
4061 | return; |
4062 | } | |
4063 | ||
bf736fe3 | 4064 | bs->drv->bdrv_debug_event(bs, event); |
41c695c7 KW |
4065 | } |
4066 | ||
4067 | int bdrv_debug_breakpoint(BlockDriverState *bs, const char *event, | |
4068 | const char *tag) | |
4069 | { | |
4070 | while (bs && bs->drv && !bs->drv->bdrv_debug_breakpoint) { | |
4071 | bs = bs->file; | |
4072 | } | |
4073 | ||
4074 | if (bs && bs->drv && bs->drv->bdrv_debug_breakpoint) { | |
4075 | return bs->drv->bdrv_debug_breakpoint(bs, event, tag); | |
4076 | } | |
4077 | ||
4078 | return -ENOTSUP; | |
4079 | } | |
4080 | ||
4cc70e93 FZ |
4081 | int bdrv_debug_remove_breakpoint(BlockDriverState *bs, const char *tag) |
4082 | { | |
4083 | while (bs && bs->drv && !bs->drv->bdrv_debug_remove_breakpoint) { | |
4084 | bs = bs->file; | |
4085 | } | |
4086 | ||
4087 | if (bs && bs->drv && bs->drv->bdrv_debug_remove_breakpoint) { | |
4088 | return bs->drv->bdrv_debug_remove_breakpoint(bs, tag); | |
4089 | } | |
4090 | ||
4091 | return -ENOTSUP; | |
4092 | } | |
4093 | ||
41c695c7 KW |
4094 | int bdrv_debug_resume(BlockDriverState *bs, const char *tag) |
4095 | { | |
938789ea | 4096 | while (bs && (!bs->drv || !bs->drv->bdrv_debug_resume)) { |
41c695c7 KW |
4097 | bs = bs->file; |
4098 | } | |
8b9b0cc2 | 4099 | |
41c695c7 KW |
4100 | if (bs && bs->drv && bs->drv->bdrv_debug_resume) { |
4101 | return bs->drv->bdrv_debug_resume(bs, tag); | |
4102 | } | |
4103 | ||
4104 | return -ENOTSUP; | |
4105 | } | |
4106 | ||
4107 | bool bdrv_debug_is_suspended(BlockDriverState *bs, const char *tag) | |
4108 | { | |
4109 | while (bs && bs->drv && !bs->drv->bdrv_debug_is_suspended) { | |
4110 | bs = bs->file; | |
4111 | } | |
4112 | ||
4113 | if (bs && bs->drv && bs->drv->bdrv_debug_is_suspended) { | |
4114 | return bs->drv->bdrv_debug_is_suspended(bs, tag); | |
4115 | } | |
4116 | ||
4117 | return false; | |
8b9b0cc2 KW |
4118 | } |
4119 | ||
199630b6 BS |
4120 | int bdrv_is_snapshot(BlockDriverState *bs) |
4121 | { | |
4122 | return !!(bs->open_flags & BDRV_O_SNAPSHOT); | |
4123 | } | |
4124 | ||
b1b1d783 JC |
4125 | /* backing_file can either be relative, or absolute, or a protocol. If it is |
4126 | * relative, it must be relative to the chain. So, passing in bs->filename | |
4127 | * from a BDS as backing_file should not be done, as that may be relative to | |
4128 | * the CWD rather than the chain. */ | |
e8a6bb9c MT |
4129 | BlockDriverState *bdrv_find_backing_image(BlockDriverState *bs, |
4130 | const char *backing_file) | |
4131 | { | |
b1b1d783 JC |
4132 | char *filename_full = NULL; |
4133 | char *backing_file_full = NULL; | |
4134 | char *filename_tmp = NULL; | |
4135 | int is_protocol = 0; | |
4136 | BlockDriverState *curr_bs = NULL; | |
4137 | BlockDriverState *retval = NULL; | |
4138 | ||
4139 | if (!bs || !bs->drv || !backing_file) { | |
e8a6bb9c MT |
4140 | return NULL; |
4141 | } | |
4142 | ||
b1b1d783 JC |
4143 | filename_full = g_malloc(PATH_MAX); |
4144 | backing_file_full = g_malloc(PATH_MAX); | |
4145 | filename_tmp = g_malloc(PATH_MAX); | |
4146 | ||
4147 | is_protocol = path_has_protocol(backing_file); | |
4148 | ||
4149 | for (curr_bs = bs; curr_bs->backing_hd; curr_bs = curr_bs->backing_hd) { | |
4150 | ||
4151 | /* If either of the filename paths is actually a protocol, then | |
4152 | * compare unmodified paths; otherwise make paths relative */ | |
4153 | if (is_protocol || path_has_protocol(curr_bs->backing_file)) { | |
4154 | if (strcmp(backing_file, curr_bs->backing_file) == 0) { | |
4155 | retval = curr_bs->backing_hd; | |
4156 | break; | |
4157 | } | |
e8a6bb9c | 4158 | } else { |
b1b1d783 JC |
4159 | /* If not an absolute filename path, make it relative to the current |
4160 | * image's filename path */ | |
4161 | path_combine(filename_tmp, PATH_MAX, curr_bs->filename, | |
4162 | backing_file); | |
4163 | ||
4164 | /* We are going to compare absolute pathnames */ | |
4165 | if (!realpath(filename_tmp, filename_full)) { | |
4166 | continue; | |
4167 | } | |
4168 | ||
4169 | /* We need to make sure the backing filename we are comparing against | |
4170 | * is relative to the current image filename (or absolute) */ | |
4171 | path_combine(filename_tmp, PATH_MAX, curr_bs->filename, | |
4172 | curr_bs->backing_file); | |
4173 | ||
4174 | if (!realpath(filename_tmp, backing_file_full)) { | |
4175 | continue; | |
4176 | } | |
4177 | ||
4178 | if (strcmp(backing_file_full, filename_full) == 0) { | |
4179 | retval = curr_bs->backing_hd; | |
4180 | break; | |
4181 | } | |
e8a6bb9c MT |
4182 | } |
4183 | } | |
4184 | ||
b1b1d783 JC |
4185 | g_free(filename_full); |
4186 | g_free(backing_file_full); | |
4187 | g_free(filename_tmp); | |
4188 | return retval; | |
e8a6bb9c MT |
4189 | } |
4190 | ||
f198fd1c BC |
4191 | int bdrv_get_backing_file_depth(BlockDriverState *bs) |
4192 | { | |
4193 | if (!bs->drv) { | |
4194 | return 0; | |
4195 | } | |
4196 | ||
4197 | if (!bs->backing_hd) { | |
4198 | return 0; | |
4199 | } | |
4200 | ||
4201 | return 1 + bdrv_get_backing_file_depth(bs->backing_hd); | |
4202 | } | |
4203 | ||
79fac568 JC |
4204 | BlockDriverState *bdrv_find_base(BlockDriverState *bs) |
4205 | { | |
4206 | BlockDriverState *curr_bs = NULL; | |
4207 | ||
4208 | if (!bs) { | |
4209 | return NULL; | |
4210 | } | |
4211 | ||
4212 | curr_bs = bs; | |
4213 | ||
4214 | while (curr_bs->backing_hd) { | |
4215 | curr_bs = curr_bs->backing_hd; | |
4216 | } | |
4217 | return curr_bs; | |
4218 | } | |
4219 | ||
ea2384d3 | 4220 | /**************************************************************/ |
83f64091 | 4221 | /* async I/Os */ |
ea2384d3 | 4222 | |
3b69e4b9 | 4223 | BlockDriverAIOCB *bdrv_aio_readv(BlockDriverState *bs, int64_t sector_num, |
f141eafe | 4224 | QEMUIOVector *qiov, int nb_sectors, |
3b69e4b9 | 4225 | BlockDriverCompletionFunc *cb, void *opaque) |
83f64091 | 4226 | { |
bbf0a440 SH |
4227 | trace_bdrv_aio_readv(bs, sector_num, nb_sectors, opaque); |
4228 | ||
d20d9b7c | 4229 | return bdrv_co_aio_rw_vector(bs, sector_num, qiov, nb_sectors, 0, |
8c5873d6 | 4230 | cb, opaque, false); |
ea2384d3 FB |
4231 | } |
4232 | ||
f141eafe AL |
4233 | BlockDriverAIOCB *bdrv_aio_writev(BlockDriverState *bs, int64_t sector_num, |
4234 | QEMUIOVector *qiov, int nb_sectors, | |
4235 | BlockDriverCompletionFunc *cb, void *opaque) | |
ea2384d3 | 4236 | { |
bbf0a440 SH |
4237 | trace_bdrv_aio_writev(bs, sector_num, nb_sectors, opaque); |
4238 | ||
d20d9b7c | 4239 | return bdrv_co_aio_rw_vector(bs, sector_num, qiov, nb_sectors, 0, |
8c5873d6 | 4240 | cb, opaque, true); |
83f64091 FB |
4241 | } |
4242 | ||
d5ef94d4 PB |
4243 | BlockDriverAIOCB *bdrv_aio_write_zeroes(BlockDriverState *bs, |
4244 | int64_t sector_num, int nb_sectors, BdrvRequestFlags flags, | |
4245 | BlockDriverCompletionFunc *cb, void *opaque) | |
4246 | { | |
4247 | trace_bdrv_aio_write_zeroes(bs, sector_num, nb_sectors, flags, opaque); | |
4248 | ||
4249 | return bdrv_co_aio_rw_vector(bs, sector_num, NULL, nb_sectors, | |
4250 | BDRV_REQ_ZERO_WRITE | flags, | |
4251 | cb, opaque, true); | |
4252 | } | |
4253 | ||
40b4f539 KW |
4254 | |
4255 | typedef struct MultiwriteCB { | |
4256 | int error; | |
4257 | int num_requests; | |
4258 | int num_callbacks; | |
4259 | struct { | |
4260 | BlockDriverCompletionFunc *cb; | |
4261 | void *opaque; | |
4262 | QEMUIOVector *free_qiov; | |
40b4f539 KW |
4263 | } callbacks[]; |
4264 | } MultiwriteCB; | |
4265 | ||
4266 | static void multiwrite_user_cb(MultiwriteCB *mcb) | |
4267 | { | |
4268 | int i; | |
4269 | ||
4270 | for (i = 0; i < mcb->num_callbacks; i++) { | |
4271 | mcb->callbacks[i].cb(mcb->callbacks[i].opaque, mcb->error); | |
1e1ea48d SH |
4272 | if (mcb->callbacks[i].free_qiov) { |
4273 | qemu_iovec_destroy(mcb->callbacks[i].free_qiov); | |
4274 | } | |
7267c094 | 4275 | g_free(mcb->callbacks[i].free_qiov); |
40b4f539 KW |
4276 | } |
4277 | } | |
4278 | ||
4279 | static void multiwrite_cb(void *opaque, int ret) | |
4280 | { | |
4281 | MultiwriteCB *mcb = opaque; | |
4282 | ||
6d519a5f SH |
4283 | trace_multiwrite_cb(mcb, ret); |
4284 | ||
cb6d3ca0 | 4285 | if (ret < 0 && !mcb->error) { |
40b4f539 | 4286 | mcb->error = ret; |
40b4f539 KW |
4287 | } |
4288 | ||
4289 | mcb->num_requests--; | |
4290 | if (mcb->num_requests == 0) { | |
de189a1b | 4291 | multiwrite_user_cb(mcb); |
7267c094 | 4292 | g_free(mcb); |
40b4f539 KW |
4293 | } |
4294 | } | |
4295 | ||
4296 | static int multiwrite_req_compare(const void *a, const void *b) | |
4297 | { | |
77be4366 CH |
4298 | const BlockRequest *req1 = a, *req2 = b; |
4299 | ||
4300 | /* | |
4301 | * Note that we can't simply subtract req2->sector from req1->sector | |
4302 | * here as that could overflow the return value. | |
4303 | */ | |
4304 | if (req1->sector > req2->sector) { | |
4305 | return 1; | |
4306 | } else if (req1->sector < req2->sector) { | |
4307 | return -1; | |
4308 | } else { | |
4309 | return 0; | |
4310 | } | |
40b4f539 KW |
4311 | } |
4312 | ||
4313 | /* | |
4314 | * Takes a bunch of requests and tries to merge them. Returns the number of | |
4315 | * requests that remain after merging. | |
4316 | */ | |
4317 | static int multiwrite_merge(BlockDriverState *bs, BlockRequest *reqs, | |
4318 | int num_reqs, MultiwriteCB *mcb) | |
4319 | { | |
4320 | int i, outidx; | |
4321 | ||
4322 | // Sort requests by start sector | |
4323 | qsort(reqs, num_reqs, sizeof(*reqs), &multiwrite_req_compare); | |
4324 | ||
4325 | // Check if adjacent requests touch the same clusters. If so, combine them, | |
4326 | // filling up gaps with zero sectors. | |
4327 | outidx = 0; | |
4328 | for (i = 1; i < num_reqs; i++) { | |
4329 | int merge = 0; | |
4330 | int64_t oldreq_last = reqs[outidx].sector + reqs[outidx].nb_sectors; | |
4331 | ||
b6a127a1 | 4332 | // Handle exactly sequential writes and overlapping writes. |
40b4f539 KW |
4333 | if (reqs[i].sector <= oldreq_last) { |
4334 | merge = 1; | |
4335 | } | |
4336 | ||
e2a305fb CH |
4337 | if (reqs[outidx].qiov->niov + reqs[i].qiov->niov + 1 > IOV_MAX) { |
4338 | merge = 0; | |
4339 | } | |
4340 | ||
40b4f539 KW |
4341 | if (merge) { |
4342 | size_t size; | |
7267c094 | 4343 | QEMUIOVector *qiov = g_malloc0(sizeof(*qiov)); |
40b4f539 KW |
4344 | qemu_iovec_init(qiov, |
4345 | reqs[outidx].qiov->niov + reqs[i].qiov->niov + 1); | |
4346 | ||
4347 | // Add the first request to the merged one. If the requests are | |
4348 | // overlapping, drop the last sectors of the first request. | |
4349 | size = (reqs[i].sector - reqs[outidx].sector) << 9; | |
1b093c48 | 4350 | qemu_iovec_concat(qiov, reqs[outidx].qiov, 0, size); |
40b4f539 | 4351 | |
b6a127a1 PB |
4352 | // We should need to add any zeros between the two requests |
4353 | assert (reqs[i].sector <= oldreq_last); | |
40b4f539 KW |
4354 | |
4355 | // Add the second request | |
1b093c48 | 4356 | qemu_iovec_concat(qiov, reqs[i].qiov, 0, reqs[i].qiov->size); |
40b4f539 | 4357 | |
cbf1dff2 | 4358 | reqs[outidx].nb_sectors = qiov->size >> 9; |
40b4f539 KW |
4359 | reqs[outidx].qiov = qiov; |
4360 | ||
4361 | mcb->callbacks[i].free_qiov = reqs[outidx].qiov; | |
4362 | } else { | |
4363 | outidx++; | |
4364 | reqs[outidx].sector = reqs[i].sector; | |
4365 | reqs[outidx].nb_sectors = reqs[i].nb_sectors; | |
4366 | reqs[outidx].qiov = reqs[i].qiov; | |
4367 | } | |
4368 | } | |
4369 | ||
4370 | return outidx + 1; | |
4371 | } | |
4372 | ||
4373 | /* | |
4374 | * Submit multiple AIO write requests at once. | |
4375 | * | |
4376 | * On success, the function returns 0 and all requests in the reqs array have | |
4377 | * been submitted. In error case this function returns -1, and any of the | |
4378 | * requests may or may not be submitted yet. In particular, this means that the | |
4379 | * callback will be called for some of the requests, for others it won't. The | |
4380 | * caller must check the error field of the BlockRequest to wait for the right | |
4381 | * callbacks (if error != 0, no callback will be called). | |
4382 | * | |
4383 | * The implementation may modify the contents of the reqs array, e.g. to merge | |
4384 | * requests. However, the fields opaque and error are left unmodified as they | |
4385 | * are used to signal failure for a single request to the caller. | |
4386 | */ | |
4387 | int bdrv_aio_multiwrite(BlockDriverState *bs, BlockRequest *reqs, int num_reqs) | |
4388 | { | |
40b4f539 KW |
4389 | MultiwriteCB *mcb; |
4390 | int i; | |
4391 | ||
301db7c2 RH |
4392 | /* don't submit writes if we don't have a medium */ |
4393 | if (bs->drv == NULL) { | |
4394 | for (i = 0; i < num_reqs; i++) { | |
4395 | reqs[i].error = -ENOMEDIUM; | |
4396 | } | |
4397 | return -1; | |
4398 | } | |
4399 | ||
40b4f539 KW |
4400 | if (num_reqs == 0) { |
4401 | return 0; | |
4402 | } | |
4403 | ||
4404 | // Create MultiwriteCB structure | |
7267c094 | 4405 | mcb = g_malloc0(sizeof(*mcb) + num_reqs * sizeof(*mcb->callbacks)); |
40b4f539 KW |
4406 | mcb->num_requests = 0; |
4407 | mcb->num_callbacks = num_reqs; | |
4408 | ||
4409 | for (i = 0; i < num_reqs; i++) { | |
4410 | mcb->callbacks[i].cb = reqs[i].cb; | |
4411 | mcb->callbacks[i].opaque = reqs[i].opaque; | |
4412 | } | |
4413 | ||
4414 | // Check for mergable requests | |
4415 | num_reqs = multiwrite_merge(bs, reqs, num_reqs, mcb); | |
4416 | ||
6d519a5f SH |
4417 | trace_bdrv_aio_multiwrite(mcb, mcb->num_callbacks, num_reqs); |
4418 | ||
df9309fb PB |
4419 | /* Run the aio requests. */ |
4420 | mcb->num_requests = num_reqs; | |
40b4f539 | 4421 | for (i = 0; i < num_reqs; i++) { |
d20d9b7c PB |
4422 | bdrv_co_aio_rw_vector(bs, reqs[i].sector, reqs[i].qiov, |
4423 | reqs[i].nb_sectors, reqs[i].flags, | |
4424 | multiwrite_cb, mcb, | |
4425 | true); | |
40b4f539 KW |
4426 | } |
4427 | ||
4428 | return 0; | |
40b4f539 KW |
4429 | } |
4430 | ||
83f64091 | 4431 | void bdrv_aio_cancel(BlockDriverAIOCB *acb) |
83f64091 | 4432 | { |
d7331bed | 4433 | acb->aiocb_info->cancel(acb); |
83f64091 FB |
4434 | } |
4435 | ||
4436 | /**************************************************************/ | |
4437 | /* async block device emulation */ | |
4438 | ||
c16b5a2c CH |
4439 | typedef struct BlockDriverAIOCBSync { |
4440 | BlockDriverAIOCB common; | |
4441 | QEMUBH *bh; | |
4442 | int ret; | |
4443 | /* vector translation state */ | |
4444 | QEMUIOVector *qiov; | |
4445 | uint8_t *bounce; | |
4446 | int is_write; | |
4447 | } BlockDriverAIOCBSync; | |
4448 | ||
4449 | static void bdrv_aio_cancel_em(BlockDriverAIOCB *blockacb) | |
4450 | { | |
b666d239 KW |
4451 | BlockDriverAIOCBSync *acb = |
4452 | container_of(blockacb, BlockDriverAIOCBSync, common); | |
6a7ad299 | 4453 | qemu_bh_delete(acb->bh); |
36afc451 | 4454 | acb->bh = NULL; |
c16b5a2c CH |
4455 | qemu_aio_release(acb); |
4456 | } | |
4457 | ||
d7331bed | 4458 | static const AIOCBInfo bdrv_em_aiocb_info = { |
c16b5a2c CH |
4459 | .aiocb_size = sizeof(BlockDriverAIOCBSync), |
4460 | .cancel = bdrv_aio_cancel_em, | |
4461 | }; | |
4462 | ||
ce1a14dc | 4463 | static void bdrv_aio_bh_cb(void *opaque) |
83f64091 | 4464 | { |
ce1a14dc | 4465 | BlockDriverAIOCBSync *acb = opaque; |
f141eafe | 4466 | |
f141eafe | 4467 | if (!acb->is_write) |
03396148 | 4468 | qemu_iovec_from_buf(acb->qiov, 0, acb->bounce, acb->qiov->size); |
ceb42de8 | 4469 | qemu_vfree(acb->bounce); |
ce1a14dc | 4470 | acb->common.cb(acb->common.opaque, acb->ret); |
6a7ad299 | 4471 | qemu_bh_delete(acb->bh); |
36afc451 | 4472 | acb->bh = NULL; |
ce1a14dc | 4473 | qemu_aio_release(acb); |
83f64091 | 4474 | } |
beac80cd | 4475 | |
f141eafe AL |
4476 | static BlockDriverAIOCB *bdrv_aio_rw_vector(BlockDriverState *bs, |
4477 | int64_t sector_num, | |
4478 | QEMUIOVector *qiov, | |
4479 | int nb_sectors, | |
4480 | BlockDriverCompletionFunc *cb, | |
4481 | void *opaque, | |
4482 | int is_write) | |
4483 | ||
83f64091 | 4484 | { |
ce1a14dc | 4485 | BlockDriverAIOCBSync *acb; |
ce1a14dc | 4486 | |
d7331bed | 4487 | acb = qemu_aio_get(&bdrv_em_aiocb_info, bs, cb, opaque); |
f141eafe AL |
4488 | acb->is_write = is_write; |
4489 | acb->qiov = qiov; | |
e268ca52 | 4490 | acb->bounce = qemu_blockalign(bs, qiov->size); |
3f3aace8 | 4491 | acb->bh = qemu_bh_new(bdrv_aio_bh_cb, acb); |
f141eafe AL |
4492 | |
4493 | if (is_write) { | |
d5e6b161 | 4494 | qemu_iovec_to_buf(acb->qiov, 0, acb->bounce, qiov->size); |
1ed20acf | 4495 | acb->ret = bs->drv->bdrv_write(bs, sector_num, acb->bounce, nb_sectors); |
f141eafe | 4496 | } else { |
1ed20acf | 4497 | acb->ret = bs->drv->bdrv_read(bs, sector_num, acb->bounce, nb_sectors); |
f141eafe AL |
4498 | } |
4499 | ||
ce1a14dc | 4500 | qemu_bh_schedule(acb->bh); |
f141eafe | 4501 | |
ce1a14dc | 4502 | return &acb->common; |
beac80cd FB |
4503 | } |
4504 | ||
f141eafe AL |
4505 | static BlockDriverAIOCB *bdrv_aio_readv_em(BlockDriverState *bs, |
4506 | int64_t sector_num, QEMUIOVector *qiov, int nb_sectors, | |
ce1a14dc | 4507 | BlockDriverCompletionFunc *cb, void *opaque) |
beac80cd | 4508 | { |
f141eafe AL |
4509 | return bdrv_aio_rw_vector(bs, sector_num, qiov, nb_sectors, cb, opaque, 0); |
4510 | } | |
83f64091 | 4511 | |
f141eafe AL |
4512 | static BlockDriverAIOCB *bdrv_aio_writev_em(BlockDriverState *bs, |
4513 | int64_t sector_num, QEMUIOVector *qiov, int nb_sectors, | |
4514 | BlockDriverCompletionFunc *cb, void *opaque) | |
4515 | { | |
4516 | return bdrv_aio_rw_vector(bs, sector_num, qiov, nb_sectors, cb, opaque, 1); | |
beac80cd | 4517 | } |
beac80cd | 4518 | |
68485420 KW |
4519 | |
4520 | typedef struct BlockDriverAIOCBCoroutine { | |
4521 | BlockDriverAIOCB common; | |
4522 | BlockRequest req; | |
4523 | bool is_write; | |
d318aea9 | 4524 | bool *done; |
68485420 KW |
4525 | QEMUBH* bh; |
4526 | } BlockDriverAIOCBCoroutine; | |
4527 | ||
4528 | static void bdrv_aio_co_cancel_em(BlockDriverAIOCB *blockacb) | |
4529 | { | |
d318aea9 KW |
4530 | BlockDriverAIOCBCoroutine *acb = |
4531 | container_of(blockacb, BlockDriverAIOCBCoroutine, common); | |
4532 | bool done = false; | |
4533 | ||
4534 | acb->done = &done; | |
4535 | while (!done) { | |
4536 | qemu_aio_wait(); | |
4537 | } | |
68485420 KW |
4538 | } |
4539 | ||
d7331bed | 4540 | static const AIOCBInfo bdrv_em_co_aiocb_info = { |
68485420 KW |
4541 | .aiocb_size = sizeof(BlockDriverAIOCBCoroutine), |
4542 | .cancel = bdrv_aio_co_cancel_em, | |
4543 | }; | |
4544 | ||
35246a68 | 4545 | static void bdrv_co_em_bh(void *opaque) |
68485420 KW |
4546 | { |
4547 | BlockDriverAIOCBCoroutine *acb = opaque; | |
4548 | ||
4549 | acb->common.cb(acb->common.opaque, acb->req.error); | |
d318aea9 KW |
4550 | |
4551 | if (acb->done) { | |
4552 | *acb->done = true; | |
4553 | } | |
4554 | ||
68485420 KW |
4555 | qemu_bh_delete(acb->bh); |
4556 | qemu_aio_release(acb); | |
4557 | } | |
4558 | ||
b2a61371 SH |
4559 | /* Invoke bdrv_co_do_readv/bdrv_co_do_writev */ |
4560 | static void coroutine_fn bdrv_co_do_rw(void *opaque) | |
4561 | { | |
4562 | BlockDriverAIOCBCoroutine *acb = opaque; | |
4563 | BlockDriverState *bs = acb->common.bs; | |
4564 | ||
4565 | if (!acb->is_write) { | |
4566 | acb->req.error = bdrv_co_do_readv(bs, acb->req.sector, | |
d20d9b7c | 4567 | acb->req.nb_sectors, acb->req.qiov, acb->req.flags); |
b2a61371 SH |
4568 | } else { |
4569 | acb->req.error = bdrv_co_do_writev(bs, acb->req.sector, | |
d20d9b7c | 4570 | acb->req.nb_sectors, acb->req.qiov, acb->req.flags); |
b2a61371 SH |
4571 | } |
4572 | ||
35246a68 | 4573 | acb->bh = qemu_bh_new(bdrv_co_em_bh, acb); |
b2a61371 SH |
4574 | qemu_bh_schedule(acb->bh); |
4575 | } | |
4576 | ||
68485420 KW |
4577 | static BlockDriverAIOCB *bdrv_co_aio_rw_vector(BlockDriverState *bs, |
4578 | int64_t sector_num, | |
4579 | QEMUIOVector *qiov, | |
4580 | int nb_sectors, | |
d20d9b7c | 4581 | BdrvRequestFlags flags, |
68485420 KW |
4582 | BlockDriverCompletionFunc *cb, |
4583 | void *opaque, | |
8c5873d6 | 4584 | bool is_write) |
68485420 KW |
4585 | { |
4586 | Coroutine *co; | |
4587 | BlockDriverAIOCBCoroutine *acb; | |
4588 | ||
d7331bed | 4589 | acb = qemu_aio_get(&bdrv_em_co_aiocb_info, bs, cb, opaque); |
68485420 KW |
4590 | acb->req.sector = sector_num; |
4591 | acb->req.nb_sectors = nb_sectors; | |
4592 | acb->req.qiov = qiov; | |
d20d9b7c | 4593 | acb->req.flags = flags; |
68485420 | 4594 | acb->is_write = is_write; |
d318aea9 | 4595 | acb->done = NULL; |
68485420 | 4596 | |
8c5873d6 | 4597 | co = qemu_coroutine_create(bdrv_co_do_rw); |
68485420 KW |
4598 | qemu_coroutine_enter(co, acb); |
4599 | ||
4600 | return &acb->common; | |
4601 | } | |
4602 | ||
07f07615 | 4603 | static void coroutine_fn bdrv_aio_flush_co_entry(void *opaque) |
b2e12bc6 | 4604 | { |
07f07615 PB |
4605 | BlockDriverAIOCBCoroutine *acb = opaque; |
4606 | BlockDriverState *bs = acb->common.bs; | |
b2e12bc6 | 4607 | |
07f07615 PB |
4608 | acb->req.error = bdrv_co_flush(bs); |
4609 | acb->bh = qemu_bh_new(bdrv_co_em_bh, acb); | |
b2e12bc6 | 4610 | qemu_bh_schedule(acb->bh); |
b2e12bc6 CH |
4611 | } |
4612 | ||
07f07615 | 4613 | BlockDriverAIOCB *bdrv_aio_flush(BlockDriverState *bs, |
016f5cf6 AG |
4614 | BlockDriverCompletionFunc *cb, void *opaque) |
4615 | { | |
07f07615 | 4616 | trace_bdrv_aio_flush(bs, opaque); |
016f5cf6 | 4617 | |
07f07615 PB |
4618 | Coroutine *co; |
4619 | BlockDriverAIOCBCoroutine *acb; | |
016f5cf6 | 4620 | |
d7331bed | 4621 | acb = qemu_aio_get(&bdrv_em_co_aiocb_info, bs, cb, opaque); |
d318aea9 KW |
4622 | acb->done = NULL; |
4623 | ||
07f07615 PB |
4624 | co = qemu_coroutine_create(bdrv_aio_flush_co_entry); |
4625 | qemu_coroutine_enter(co, acb); | |
016f5cf6 | 4626 | |
016f5cf6 AG |
4627 | return &acb->common; |
4628 | } | |
4629 | ||
4265d620 PB |
4630 | static void coroutine_fn bdrv_aio_discard_co_entry(void *opaque) |
4631 | { | |
4632 | BlockDriverAIOCBCoroutine *acb = opaque; | |
4633 | BlockDriverState *bs = acb->common.bs; | |
4634 | ||
4635 | acb->req.error = bdrv_co_discard(bs, acb->req.sector, acb->req.nb_sectors); | |
4636 | acb->bh = qemu_bh_new(bdrv_co_em_bh, acb); | |
4637 | qemu_bh_schedule(acb->bh); | |
4638 | } | |
4639 | ||
4640 | BlockDriverAIOCB *bdrv_aio_discard(BlockDriverState *bs, | |
4641 | int64_t sector_num, int nb_sectors, | |
4642 | BlockDriverCompletionFunc *cb, void *opaque) | |
4643 | { | |
4644 | Coroutine *co; | |
4645 | BlockDriverAIOCBCoroutine *acb; | |
4646 | ||
4647 | trace_bdrv_aio_discard(bs, sector_num, nb_sectors, opaque); | |
4648 | ||
d7331bed | 4649 | acb = qemu_aio_get(&bdrv_em_co_aiocb_info, bs, cb, opaque); |
4265d620 PB |
4650 | acb->req.sector = sector_num; |
4651 | acb->req.nb_sectors = nb_sectors; | |
d318aea9 | 4652 | acb->done = NULL; |
4265d620 PB |
4653 | co = qemu_coroutine_create(bdrv_aio_discard_co_entry); |
4654 | qemu_coroutine_enter(co, acb); | |
4655 | ||
4656 | return &acb->common; | |
4657 | } | |
4658 | ||
ea2384d3 FB |
4659 | void bdrv_init(void) |
4660 | { | |
5efa9d5a | 4661 | module_call_init(MODULE_INIT_BLOCK); |
ea2384d3 | 4662 | } |
ce1a14dc | 4663 | |
eb852011 MA |
4664 | void bdrv_init_with_whitelist(void) |
4665 | { | |
4666 | use_bdrv_whitelist = 1; | |
4667 | bdrv_init(); | |
4668 | } | |
4669 | ||
d7331bed | 4670 | void *qemu_aio_get(const AIOCBInfo *aiocb_info, BlockDriverState *bs, |
c16b5a2c | 4671 | BlockDriverCompletionFunc *cb, void *opaque) |
ce1a14dc | 4672 | { |
ce1a14dc PB |
4673 | BlockDriverAIOCB *acb; |
4674 | ||
d7331bed SH |
4675 | acb = g_slice_alloc(aiocb_info->aiocb_size); |
4676 | acb->aiocb_info = aiocb_info; | |
ce1a14dc PB |
4677 | acb->bs = bs; |
4678 | acb->cb = cb; | |
4679 | acb->opaque = opaque; | |
4680 | return acb; | |
4681 | } | |
4682 | ||
4683 | void qemu_aio_release(void *p) | |
4684 | { | |
d37c975f | 4685 | BlockDriverAIOCB *acb = p; |
d7331bed | 4686 | g_slice_free1(acb->aiocb_info->aiocb_size, acb); |
ce1a14dc | 4687 | } |
19cb3738 | 4688 | |
f9f05dc5 KW |
4689 | /**************************************************************/ |
4690 | /* Coroutine block device emulation */ | |
4691 | ||
4692 | typedef struct CoroutineIOCompletion { | |
4693 | Coroutine *coroutine; | |
4694 | int ret; | |
4695 | } CoroutineIOCompletion; | |
4696 | ||
4697 | static void bdrv_co_io_em_complete(void *opaque, int ret) | |
4698 | { | |
4699 | CoroutineIOCompletion *co = opaque; | |
4700 | ||
4701 | co->ret = ret; | |
4702 | qemu_coroutine_enter(co->coroutine, NULL); | |
4703 | } | |
4704 | ||
4705 | static int coroutine_fn bdrv_co_io_em(BlockDriverState *bs, int64_t sector_num, | |
4706 | int nb_sectors, QEMUIOVector *iov, | |
4707 | bool is_write) | |
4708 | { | |
4709 | CoroutineIOCompletion co = { | |
4710 | .coroutine = qemu_coroutine_self(), | |
4711 | }; | |
4712 | BlockDriverAIOCB *acb; | |
4713 | ||
4714 | if (is_write) { | |
a652d160 SH |
4715 | acb = bs->drv->bdrv_aio_writev(bs, sector_num, iov, nb_sectors, |
4716 | bdrv_co_io_em_complete, &co); | |
f9f05dc5 | 4717 | } else { |
a652d160 SH |
4718 | acb = bs->drv->bdrv_aio_readv(bs, sector_num, iov, nb_sectors, |
4719 | bdrv_co_io_em_complete, &co); | |
f9f05dc5 KW |
4720 | } |
4721 | ||
59370aaa | 4722 | trace_bdrv_co_io_em(bs, sector_num, nb_sectors, is_write, acb); |
f9f05dc5 KW |
4723 | if (!acb) { |
4724 | return -EIO; | |
4725 | } | |
4726 | qemu_coroutine_yield(); | |
4727 | ||
4728 | return co.ret; | |
4729 | } | |
4730 | ||
4731 | static int coroutine_fn bdrv_co_readv_em(BlockDriverState *bs, | |
4732 | int64_t sector_num, int nb_sectors, | |
4733 | QEMUIOVector *iov) | |
4734 | { | |
4735 | return bdrv_co_io_em(bs, sector_num, nb_sectors, iov, false); | |
4736 | } | |
4737 | ||
4738 | static int coroutine_fn bdrv_co_writev_em(BlockDriverState *bs, | |
4739 | int64_t sector_num, int nb_sectors, | |
4740 | QEMUIOVector *iov) | |
4741 | { | |
4742 | return bdrv_co_io_em(bs, sector_num, nb_sectors, iov, true); | |
4743 | } | |
4744 | ||
07f07615 | 4745 | static void coroutine_fn bdrv_flush_co_entry(void *opaque) |
e7a8a783 | 4746 | { |
07f07615 PB |
4747 | RwCo *rwco = opaque; |
4748 | ||
4749 | rwco->ret = bdrv_co_flush(rwco->bs); | |
4750 | } | |
4751 | ||
4752 | int coroutine_fn bdrv_co_flush(BlockDriverState *bs) | |
4753 | { | |
eb489bb1 KW |
4754 | int ret; |
4755 | ||
29cdb251 | 4756 | if (!bs || !bdrv_is_inserted(bs) || bdrv_is_read_only(bs)) { |
07f07615 | 4757 | return 0; |
eb489bb1 KW |
4758 | } |
4759 | ||
ca716364 | 4760 | /* Write back cached data to the OS even with cache=unsafe */ |
bf736fe3 | 4761 | BLKDBG_EVENT(bs->file, BLKDBG_FLUSH_TO_OS); |
eb489bb1 KW |
4762 | if (bs->drv->bdrv_co_flush_to_os) { |
4763 | ret = bs->drv->bdrv_co_flush_to_os(bs); | |
4764 | if (ret < 0) { | |
4765 | return ret; | |
4766 | } | |
4767 | } | |
4768 | ||
ca716364 KW |
4769 | /* But don't actually force it to the disk with cache=unsafe */ |
4770 | if (bs->open_flags & BDRV_O_NO_FLUSH) { | |
d4c82329 | 4771 | goto flush_parent; |
ca716364 KW |
4772 | } |
4773 | ||
bf736fe3 | 4774 | BLKDBG_EVENT(bs->file, BLKDBG_FLUSH_TO_DISK); |
eb489bb1 | 4775 | if (bs->drv->bdrv_co_flush_to_disk) { |
29cdb251 | 4776 | ret = bs->drv->bdrv_co_flush_to_disk(bs); |
07f07615 PB |
4777 | } else if (bs->drv->bdrv_aio_flush) { |
4778 | BlockDriverAIOCB *acb; | |
4779 | CoroutineIOCompletion co = { | |
4780 | .coroutine = qemu_coroutine_self(), | |
4781 | }; | |
4782 | ||
4783 | acb = bs->drv->bdrv_aio_flush(bs, bdrv_co_io_em_complete, &co); | |
4784 | if (acb == NULL) { | |
29cdb251 | 4785 | ret = -EIO; |
07f07615 PB |
4786 | } else { |
4787 | qemu_coroutine_yield(); | |
29cdb251 | 4788 | ret = co.ret; |
07f07615 | 4789 | } |
07f07615 PB |
4790 | } else { |
4791 | /* | |
4792 | * Some block drivers always operate in either writethrough or unsafe | |
4793 | * mode and don't support bdrv_flush therefore. Usually qemu doesn't | |
4794 | * know how the server works (because the behaviour is hardcoded or | |
4795 | * depends on server-side configuration), so we can't ensure that | |
4796 | * everything is safe on disk. Returning an error doesn't work because | |
4797 | * that would break guests even if the server operates in writethrough | |
4798 | * mode. | |
4799 | * | |
4800 | * Let's hope the user knows what he's doing. | |
4801 | */ | |
29cdb251 | 4802 | ret = 0; |
07f07615 | 4803 | } |
29cdb251 PB |
4804 | if (ret < 0) { |
4805 | return ret; | |
4806 | } | |
4807 | ||
4808 | /* Now flush the underlying protocol. It will also have BDRV_O_NO_FLUSH | |
4809 | * in the case of cache=unsafe, so there are no useless flushes. | |
4810 | */ | |
d4c82329 | 4811 | flush_parent: |
29cdb251 | 4812 | return bdrv_co_flush(bs->file); |
07f07615 PB |
4813 | } |
4814 | ||
5a8a30db | 4815 | void bdrv_invalidate_cache(BlockDriverState *bs, Error **errp) |
0f15423c | 4816 | { |
5a8a30db KW |
4817 | Error *local_err = NULL; |
4818 | int ret; | |
4819 | ||
3456a8d1 KW |
4820 | if (!bs->drv) { |
4821 | return; | |
4822 | } | |
4823 | ||
4824 | if (bs->drv->bdrv_invalidate_cache) { | |
5a8a30db | 4825 | bs->drv->bdrv_invalidate_cache(bs, &local_err); |
3456a8d1 | 4826 | } else if (bs->file) { |
5a8a30db KW |
4827 | bdrv_invalidate_cache(bs->file, &local_err); |
4828 | } | |
4829 | if (local_err) { | |
4830 | error_propagate(errp, local_err); | |
4831 | return; | |
0f15423c | 4832 | } |
3456a8d1 | 4833 | |
5a8a30db KW |
4834 | ret = refresh_total_sectors(bs, bs->total_sectors); |
4835 | if (ret < 0) { | |
4836 | error_setg_errno(errp, -ret, "Could not refresh total sector count"); | |
4837 | return; | |
4838 | } | |
0f15423c AL |
4839 | } |
4840 | ||
5a8a30db | 4841 | void bdrv_invalidate_cache_all(Error **errp) |
0f15423c AL |
4842 | { |
4843 | BlockDriverState *bs; | |
5a8a30db | 4844 | Error *local_err = NULL; |
0f15423c | 4845 | |
dc364f4c | 4846 | QTAILQ_FOREACH(bs, &bdrv_states, device_list) { |
5a8a30db KW |
4847 | bdrv_invalidate_cache(bs, &local_err); |
4848 | if (local_err) { | |
4849 | error_propagate(errp, local_err); | |
4850 | return; | |
4851 | } | |
0f15423c AL |
4852 | } |
4853 | } | |
4854 | ||
07789269 BC |
4855 | void bdrv_clear_incoming_migration_all(void) |
4856 | { | |
4857 | BlockDriverState *bs; | |
4858 | ||
dc364f4c | 4859 | QTAILQ_FOREACH(bs, &bdrv_states, device_list) { |
07789269 BC |
4860 | bs->open_flags = bs->open_flags & ~(BDRV_O_INCOMING); |
4861 | } | |
4862 | } | |
4863 | ||
07f07615 PB |
4864 | int bdrv_flush(BlockDriverState *bs) |
4865 | { | |
4866 | Coroutine *co; | |
4867 | RwCo rwco = { | |
4868 | .bs = bs, | |
4869 | .ret = NOT_DONE, | |
e7a8a783 | 4870 | }; |
e7a8a783 | 4871 | |
07f07615 PB |
4872 | if (qemu_in_coroutine()) { |
4873 | /* Fast-path if already in coroutine context */ | |
4874 | bdrv_flush_co_entry(&rwco); | |
4875 | } else { | |
4876 | co = qemu_coroutine_create(bdrv_flush_co_entry); | |
4877 | qemu_coroutine_enter(co, &rwco); | |
4878 | while (rwco.ret == NOT_DONE) { | |
4879 | qemu_aio_wait(); | |
4880 | } | |
e7a8a783 | 4881 | } |
07f07615 PB |
4882 | |
4883 | return rwco.ret; | |
e7a8a783 KW |
4884 | } |
4885 | ||
775aa8b6 KW |
4886 | typedef struct DiscardCo { |
4887 | BlockDriverState *bs; | |
4888 | int64_t sector_num; | |
4889 | int nb_sectors; | |
4890 | int ret; | |
4891 | } DiscardCo; | |
4265d620 PB |
4892 | static void coroutine_fn bdrv_discard_co_entry(void *opaque) |
4893 | { | |
775aa8b6 | 4894 | DiscardCo *rwco = opaque; |
4265d620 PB |
4895 | |
4896 | rwco->ret = bdrv_co_discard(rwco->bs, rwco->sector_num, rwco->nb_sectors); | |
4897 | } | |
4898 | ||
6f14da52 PL |
4899 | /* if no limit is specified in the BlockLimits use a default |
4900 | * of 32768 512-byte sectors (16 MiB) per request. | |
4901 | */ | |
4902 | #define MAX_DISCARD_DEFAULT 32768 | |
4903 | ||
4265d620 PB |
4904 | int coroutine_fn bdrv_co_discard(BlockDriverState *bs, int64_t sector_num, |
4905 | int nb_sectors) | |
4906 | { | |
d51e9fe5 PB |
4907 | int max_discard; |
4908 | ||
4265d620 PB |
4909 | if (!bs->drv) { |
4910 | return -ENOMEDIUM; | |
4911 | } else if (bdrv_check_request(bs, sector_num, nb_sectors)) { | |
4912 | return -EIO; | |
4913 | } else if (bs->read_only) { | |
4914 | return -EROFS; | |
df702c9b PB |
4915 | } |
4916 | ||
e4654d2d | 4917 | bdrv_reset_dirty(bs, sector_num, nb_sectors); |
df702c9b | 4918 | |
9e8f1835 PB |
4919 | /* Do nothing if disabled. */ |
4920 | if (!(bs->open_flags & BDRV_O_UNMAP)) { | |
4921 | return 0; | |
4922 | } | |
4923 | ||
d51e9fe5 PB |
4924 | if (!bs->drv->bdrv_co_discard && !bs->drv->bdrv_aio_discard) { |
4925 | return 0; | |
4926 | } | |
6f14da52 | 4927 | |
d51e9fe5 PB |
4928 | max_discard = bs->bl.max_discard ? bs->bl.max_discard : MAX_DISCARD_DEFAULT; |
4929 | while (nb_sectors > 0) { | |
4930 | int ret; | |
4931 | int num = nb_sectors; | |
6f14da52 | 4932 | |
d51e9fe5 PB |
4933 | /* align request */ |
4934 | if (bs->bl.discard_alignment && | |
4935 | num >= bs->bl.discard_alignment && | |
4936 | sector_num % bs->bl.discard_alignment) { | |
4937 | if (num > bs->bl.discard_alignment) { | |
4938 | num = bs->bl.discard_alignment; | |
6f14da52 | 4939 | } |
d51e9fe5 PB |
4940 | num -= sector_num % bs->bl.discard_alignment; |
4941 | } | |
6f14da52 | 4942 | |
d51e9fe5 PB |
4943 | /* limit request size */ |
4944 | if (num > max_discard) { | |
4945 | num = max_discard; | |
4946 | } | |
6f14da52 | 4947 | |
d51e9fe5 | 4948 | if (bs->drv->bdrv_co_discard) { |
6f14da52 | 4949 | ret = bs->drv->bdrv_co_discard(bs, sector_num, num); |
d51e9fe5 PB |
4950 | } else { |
4951 | BlockDriverAIOCB *acb; | |
4952 | CoroutineIOCompletion co = { | |
4953 | .coroutine = qemu_coroutine_self(), | |
4954 | }; | |
4955 | ||
4956 | acb = bs->drv->bdrv_aio_discard(bs, sector_num, nb_sectors, | |
4957 | bdrv_co_io_em_complete, &co); | |
4958 | if (acb == NULL) { | |
4959 | return -EIO; | |
4960 | } else { | |
4961 | qemu_coroutine_yield(); | |
4962 | ret = co.ret; | |
6f14da52 | 4963 | } |
6f14da52 | 4964 | } |
7ce21016 | 4965 | if (ret && ret != -ENOTSUP) { |
d51e9fe5 | 4966 | return ret; |
4265d620 | 4967 | } |
d51e9fe5 PB |
4968 | |
4969 | sector_num += num; | |
4970 | nb_sectors -= num; | |
4265d620 | 4971 | } |
d51e9fe5 | 4972 | return 0; |
4265d620 PB |
4973 | } |
4974 | ||
4975 | int bdrv_discard(BlockDriverState *bs, int64_t sector_num, int nb_sectors) | |
4976 | { | |
4977 | Coroutine *co; | |
775aa8b6 | 4978 | DiscardCo rwco = { |
4265d620 PB |
4979 | .bs = bs, |
4980 | .sector_num = sector_num, | |
4981 | .nb_sectors = nb_sectors, | |
4982 | .ret = NOT_DONE, | |
4983 | }; | |
4984 | ||
4985 | if (qemu_in_coroutine()) { | |
4986 | /* Fast-path if already in coroutine context */ | |
4987 | bdrv_discard_co_entry(&rwco); | |
4988 | } else { | |
4989 | co = qemu_coroutine_create(bdrv_discard_co_entry); | |
4990 | qemu_coroutine_enter(co, &rwco); | |
4991 | while (rwco.ret == NOT_DONE) { | |
4992 | qemu_aio_wait(); | |
4993 | } | |
4994 | } | |
4995 | ||
4996 | return rwco.ret; | |
4997 | } | |
4998 | ||
19cb3738 FB |
4999 | /**************************************************************/ |
5000 | /* removable device support */ | |
5001 | ||
5002 | /** | |
5003 | * Return TRUE if the media is present | |
5004 | */ | |
5005 | int bdrv_is_inserted(BlockDriverState *bs) | |
5006 | { | |
5007 | BlockDriver *drv = bs->drv; | |
a1aff5bf | 5008 | |
19cb3738 FB |
5009 | if (!drv) |
5010 | return 0; | |
5011 | if (!drv->bdrv_is_inserted) | |
a1aff5bf MA |
5012 | return 1; |
5013 | return drv->bdrv_is_inserted(bs); | |
19cb3738 FB |
5014 | } |
5015 | ||
5016 | /** | |
8e49ca46 MA |
5017 | * Return whether the media changed since the last call to this |
5018 | * function, or -ENOTSUP if we don't know. Most drivers don't know. | |
19cb3738 FB |
5019 | */ |
5020 | int bdrv_media_changed(BlockDriverState *bs) | |
5021 | { | |
5022 | BlockDriver *drv = bs->drv; | |
19cb3738 | 5023 | |
8e49ca46 MA |
5024 | if (drv && drv->bdrv_media_changed) { |
5025 | return drv->bdrv_media_changed(bs); | |
5026 | } | |
5027 | return -ENOTSUP; | |
19cb3738 FB |
5028 | } |
5029 | ||
5030 | /** | |
5031 | * If eject_flag is TRUE, eject the media. Otherwise, close the tray | |
5032 | */ | |
f36f3949 | 5033 | void bdrv_eject(BlockDriverState *bs, bool eject_flag) |
19cb3738 FB |
5034 | { |
5035 | BlockDriver *drv = bs->drv; | |
19cb3738 | 5036 | |
822e1cd1 MA |
5037 | if (drv && drv->bdrv_eject) { |
5038 | drv->bdrv_eject(bs, eject_flag); | |
19cb3738 | 5039 | } |
6f382ed2 LC |
5040 | |
5041 | if (bs->device_name[0] != '\0') { | |
5042 | bdrv_emit_qmp_eject_event(bs, eject_flag); | |
5043 | } | |
19cb3738 FB |
5044 | } |
5045 | ||
19cb3738 FB |
5046 | /** |
5047 | * Lock or unlock the media (if it is locked, the user won't be able | |
5048 | * to eject it manually). | |
5049 | */ | |
025e849a | 5050 | void bdrv_lock_medium(BlockDriverState *bs, bool locked) |
19cb3738 FB |
5051 | { |
5052 | BlockDriver *drv = bs->drv; | |
5053 | ||
025e849a | 5054 | trace_bdrv_lock_medium(bs, locked); |
b8c6d095 | 5055 | |
025e849a MA |
5056 | if (drv && drv->bdrv_lock_medium) { |
5057 | drv->bdrv_lock_medium(bs, locked); | |
19cb3738 FB |
5058 | } |
5059 | } | |
985a03b0 TS |
5060 | |
5061 | /* needed for generic scsi interface */ | |
5062 | ||
5063 | int bdrv_ioctl(BlockDriverState *bs, unsigned long int req, void *buf) | |
5064 | { | |
5065 | BlockDriver *drv = bs->drv; | |
5066 | ||
5067 | if (drv && drv->bdrv_ioctl) | |
5068 | return drv->bdrv_ioctl(bs, req, buf); | |
5069 | return -ENOTSUP; | |
5070 | } | |
7d780669 | 5071 | |
221f715d AL |
5072 | BlockDriverAIOCB *bdrv_aio_ioctl(BlockDriverState *bs, |
5073 | unsigned long int req, void *buf, | |
5074 | BlockDriverCompletionFunc *cb, void *opaque) | |
7d780669 | 5075 | { |
221f715d | 5076 | BlockDriver *drv = bs->drv; |
7d780669 | 5077 | |
221f715d AL |
5078 | if (drv && drv->bdrv_aio_ioctl) |
5079 | return drv->bdrv_aio_ioctl(bs, req, buf, cb, opaque); | |
5080 | return NULL; | |
7d780669 | 5081 | } |
e268ca52 | 5082 | |
1b7fd729 | 5083 | void bdrv_set_guest_block_size(BlockDriverState *bs, int align) |
7b6f9300 | 5084 | { |
1b7fd729 | 5085 | bs->guest_block_size = align; |
7b6f9300 | 5086 | } |
7cd1e32a | 5087 | |
e268ca52 AL |
5088 | void *qemu_blockalign(BlockDriverState *bs, size_t size) |
5089 | { | |
339064d5 | 5090 | return qemu_memalign(bdrv_opt_mem_align(bs), size); |
e268ca52 | 5091 | } |
7cd1e32a | 5092 | |
c53b1c51 SH |
5093 | /* |
5094 | * Check if all memory in this vector is sector aligned. | |
5095 | */ | |
5096 | bool bdrv_qiov_is_aligned(BlockDriverState *bs, QEMUIOVector *qiov) | |
5097 | { | |
5098 | int i; | |
339064d5 | 5099 | size_t alignment = bdrv_opt_mem_align(bs); |
c53b1c51 SH |
5100 | |
5101 | for (i = 0; i < qiov->niov; i++) { | |
339064d5 | 5102 | if ((uintptr_t) qiov->iov[i].iov_base % alignment) { |
c53b1c51 | 5103 | return false; |
1ff735bd | 5104 | } |
339064d5 | 5105 | if (qiov->iov[i].iov_len % alignment) { |
1ff735bd | 5106 | return false; |
c53b1c51 SH |
5107 | } |
5108 | } | |
5109 | ||
5110 | return true; | |
5111 | } | |
5112 | ||
e4654d2d | 5113 | BdrvDirtyBitmap *bdrv_create_dirty_bitmap(BlockDriverState *bs, int granularity) |
7cd1e32a LS |
5114 | { |
5115 | int64_t bitmap_size; | |
e4654d2d | 5116 | BdrvDirtyBitmap *bitmap; |
a55eb92c | 5117 | |
50717e94 PB |
5118 | assert((granularity & (granularity - 1)) == 0); |
5119 | ||
e4654d2d FZ |
5120 | granularity >>= BDRV_SECTOR_BITS; |
5121 | assert(granularity); | |
5122 | bitmap_size = (bdrv_getlength(bs) >> BDRV_SECTOR_BITS); | |
5123 | bitmap = g_malloc0(sizeof(BdrvDirtyBitmap)); | |
5124 | bitmap->bitmap = hbitmap_alloc(bitmap_size, ffs(granularity) - 1); | |
5125 | QLIST_INSERT_HEAD(&bs->dirty_bitmaps, bitmap, list); | |
5126 | return bitmap; | |
5127 | } | |
5128 | ||
5129 | void bdrv_release_dirty_bitmap(BlockDriverState *bs, BdrvDirtyBitmap *bitmap) | |
5130 | { | |
5131 | BdrvDirtyBitmap *bm, *next; | |
5132 | QLIST_FOREACH_SAFE(bm, &bs->dirty_bitmaps, list, next) { | |
5133 | if (bm == bitmap) { | |
5134 | QLIST_REMOVE(bitmap, list); | |
5135 | hbitmap_free(bitmap->bitmap); | |
5136 | g_free(bitmap); | |
5137 | return; | |
a55eb92c | 5138 | } |
7cd1e32a LS |
5139 | } |
5140 | } | |
5141 | ||
21b56835 FZ |
5142 | BlockDirtyInfoList *bdrv_query_dirty_bitmaps(BlockDriverState *bs) |
5143 | { | |
5144 | BdrvDirtyBitmap *bm; | |
5145 | BlockDirtyInfoList *list = NULL; | |
5146 | BlockDirtyInfoList **plist = &list; | |
5147 | ||
5148 | QLIST_FOREACH(bm, &bs->dirty_bitmaps, list) { | |
5149 | BlockDirtyInfo *info = g_malloc0(sizeof(BlockDirtyInfo)); | |
5150 | BlockDirtyInfoList *entry = g_malloc0(sizeof(BlockDirtyInfoList)); | |
5151 | info->count = bdrv_get_dirty_count(bs, bm); | |
5152 | info->granularity = | |
5153 | ((int64_t) BDRV_SECTOR_SIZE << hbitmap_granularity(bm->bitmap)); | |
5154 | entry->value = info; | |
5155 | *plist = entry; | |
5156 | plist = &entry->next; | |
5157 | } | |
5158 | ||
5159 | return list; | |
5160 | } | |
5161 | ||
e4654d2d | 5162 | int bdrv_get_dirty(BlockDriverState *bs, BdrvDirtyBitmap *bitmap, int64_t sector) |
7cd1e32a | 5163 | { |
e4654d2d FZ |
5164 | if (bitmap) { |
5165 | return hbitmap_get(bitmap->bitmap, sector); | |
7cd1e32a LS |
5166 | } else { |
5167 | return 0; | |
5168 | } | |
5169 | } | |
5170 | ||
e4654d2d FZ |
5171 | void bdrv_dirty_iter_init(BlockDriverState *bs, |
5172 | BdrvDirtyBitmap *bitmap, HBitmapIter *hbi) | |
1755da16 | 5173 | { |
e4654d2d | 5174 | hbitmap_iter_init(hbi, bitmap->bitmap, 0); |
1755da16 PB |
5175 | } |
5176 | ||
5177 | void bdrv_set_dirty(BlockDriverState *bs, int64_t cur_sector, | |
5178 | int nr_sectors) | |
5179 | { | |
e4654d2d FZ |
5180 | BdrvDirtyBitmap *bitmap; |
5181 | QLIST_FOREACH(bitmap, &bs->dirty_bitmaps, list) { | |
5182 | hbitmap_set(bitmap->bitmap, cur_sector, nr_sectors); | |
5183 | } | |
1755da16 PB |
5184 | } |
5185 | ||
e4654d2d | 5186 | void bdrv_reset_dirty(BlockDriverState *bs, int64_t cur_sector, int nr_sectors) |
7cd1e32a | 5187 | { |
e4654d2d FZ |
5188 | BdrvDirtyBitmap *bitmap; |
5189 | QLIST_FOREACH(bitmap, &bs->dirty_bitmaps, list) { | |
5190 | hbitmap_reset(bitmap->bitmap, cur_sector, nr_sectors); | |
5191 | } | |
7cd1e32a | 5192 | } |
aaa0eb75 | 5193 | |
e4654d2d | 5194 | int64_t bdrv_get_dirty_count(BlockDriverState *bs, BdrvDirtyBitmap *bitmap) |
aaa0eb75 | 5195 | { |
e4654d2d | 5196 | return hbitmap_count(bitmap->bitmap); |
aaa0eb75 | 5197 | } |
f88e1a42 | 5198 | |
9fcb0251 FZ |
5199 | /* Get a reference to bs */ |
5200 | void bdrv_ref(BlockDriverState *bs) | |
5201 | { | |
5202 | bs->refcnt++; | |
5203 | } | |
5204 | ||
5205 | /* Release a previously grabbed reference to bs. | |
5206 | * If after releasing, reference count is zero, the BlockDriverState is | |
5207 | * deleted. */ | |
5208 | void bdrv_unref(BlockDriverState *bs) | |
5209 | { | |
5210 | assert(bs->refcnt > 0); | |
5211 | if (--bs->refcnt == 0) { | |
5212 | bdrv_delete(bs); | |
5213 | } | |
5214 | } | |
5215 | ||
db593f25 MT |
5216 | void bdrv_set_in_use(BlockDriverState *bs, int in_use) |
5217 | { | |
5218 | assert(bs->in_use != in_use); | |
5219 | bs->in_use = in_use; | |
5220 | } | |
5221 | ||
5222 | int bdrv_in_use(BlockDriverState *bs) | |
5223 | { | |
5224 | return bs->in_use; | |
5225 | } | |
5226 | ||
28a7282a LC |
5227 | void bdrv_iostatus_enable(BlockDriverState *bs) |
5228 | { | |
d6bf279e | 5229 | bs->iostatus_enabled = true; |
58e21ef5 | 5230 | bs->iostatus = BLOCK_DEVICE_IO_STATUS_OK; |
28a7282a LC |
5231 | } |
5232 | ||
5233 | /* The I/O status is only enabled if the drive explicitly | |
5234 | * enables it _and_ the VM is configured to stop on errors */ | |
5235 | bool bdrv_iostatus_is_enabled(const BlockDriverState *bs) | |
5236 | { | |
d6bf279e | 5237 | return (bs->iostatus_enabled && |
92aa5c6d PB |
5238 | (bs->on_write_error == BLOCKDEV_ON_ERROR_ENOSPC || |
5239 | bs->on_write_error == BLOCKDEV_ON_ERROR_STOP || | |
5240 | bs->on_read_error == BLOCKDEV_ON_ERROR_STOP)); | |
28a7282a LC |
5241 | } |
5242 | ||
5243 | void bdrv_iostatus_disable(BlockDriverState *bs) | |
5244 | { | |
d6bf279e | 5245 | bs->iostatus_enabled = false; |
28a7282a LC |
5246 | } |
5247 | ||
5248 | void bdrv_iostatus_reset(BlockDriverState *bs) | |
5249 | { | |
5250 | if (bdrv_iostatus_is_enabled(bs)) { | |
58e21ef5 | 5251 | bs->iostatus = BLOCK_DEVICE_IO_STATUS_OK; |
3bd293c3 PB |
5252 | if (bs->job) { |
5253 | block_job_iostatus_reset(bs->job); | |
5254 | } | |
28a7282a LC |
5255 | } |
5256 | } | |
5257 | ||
28a7282a LC |
5258 | void bdrv_iostatus_set_err(BlockDriverState *bs, int error) |
5259 | { | |
3e1caa5f PB |
5260 | assert(bdrv_iostatus_is_enabled(bs)); |
5261 | if (bs->iostatus == BLOCK_DEVICE_IO_STATUS_OK) { | |
58e21ef5 LC |
5262 | bs->iostatus = error == ENOSPC ? BLOCK_DEVICE_IO_STATUS_NOSPACE : |
5263 | BLOCK_DEVICE_IO_STATUS_FAILED; | |
28a7282a LC |
5264 | } |
5265 | } | |
5266 | ||
a597e79c CH |
5267 | void |
5268 | bdrv_acct_start(BlockDriverState *bs, BlockAcctCookie *cookie, int64_t bytes, | |
5269 | enum BlockAcctType type) | |
5270 | { | |
5271 | assert(type < BDRV_MAX_IOTYPE); | |
5272 | ||
5273 | cookie->bytes = bytes; | |
c488c7f6 | 5274 | cookie->start_time_ns = get_clock(); |
a597e79c CH |
5275 | cookie->type = type; |
5276 | } | |
5277 | ||
5278 | void | |
5279 | bdrv_acct_done(BlockDriverState *bs, BlockAcctCookie *cookie) | |
5280 | { | |
5281 | assert(cookie->type < BDRV_MAX_IOTYPE); | |
5282 | ||
5283 | bs->nr_bytes[cookie->type] += cookie->bytes; | |
5284 | bs->nr_ops[cookie->type]++; | |
c488c7f6 | 5285 | bs->total_time_ns[cookie->type] += get_clock() - cookie->start_time_ns; |
a597e79c CH |
5286 | } |
5287 | ||
d92ada22 LC |
5288 | void bdrv_img_create(const char *filename, const char *fmt, |
5289 | const char *base_filename, const char *base_fmt, | |
f382d43a MR |
5290 | char *options, uint64_t img_size, int flags, |
5291 | Error **errp, bool quiet) | |
f88e1a42 JS |
5292 | { |
5293 | QEMUOptionParameter *param = NULL, *create_options = NULL; | |
d220894e | 5294 | QEMUOptionParameter *backing_fmt, *backing_file, *size; |
f88e1a42 | 5295 | BlockDriver *drv, *proto_drv; |
96df67d1 | 5296 | BlockDriver *backing_drv = NULL; |
cc84d90f | 5297 | Error *local_err = NULL; |
f88e1a42 JS |
5298 | int ret = 0; |
5299 | ||
5300 | /* Find driver and parse its options */ | |
5301 | drv = bdrv_find_format(fmt); | |
5302 | if (!drv) { | |
71c79813 | 5303 | error_setg(errp, "Unknown file format '%s'", fmt); |
d92ada22 | 5304 | return; |
f88e1a42 JS |
5305 | } |
5306 | ||
98289620 | 5307 | proto_drv = bdrv_find_protocol(filename, true); |
f88e1a42 | 5308 | if (!proto_drv) { |
71c79813 | 5309 | error_setg(errp, "Unknown protocol '%s'", filename); |
d92ada22 | 5310 | return; |
f88e1a42 JS |
5311 | } |
5312 | ||
5313 | create_options = append_option_parameters(create_options, | |
5314 | drv->create_options); | |
5315 | create_options = append_option_parameters(create_options, | |
5316 | proto_drv->create_options); | |
5317 | ||
5318 | /* Create parameter list with default values */ | |
5319 | param = parse_option_parameters("", create_options, param); | |
5320 | ||
5321 | set_option_parameter_int(param, BLOCK_OPT_SIZE, img_size); | |
5322 | ||
5323 | /* Parse -o options */ | |
5324 | if (options) { | |
5325 | param = parse_option_parameters(options, create_options, param); | |
5326 | if (param == NULL) { | |
71c79813 | 5327 | error_setg(errp, "Invalid options for file format '%s'.", fmt); |
f88e1a42 JS |
5328 | goto out; |
5329 | } | |
5330 | } | |
5331 | ||
5332 | if (base_filename) { | |
5333 | if (set_option_parameter(param, BLOCK_OPT_BACKING_FILE, | |
5334 | base_filename)) { | |
71c79813 LC |
5335 | error_setg(errp, "Backing file not supported for file format '%s'", |
5336 | fmt); | |
f88e1a42 JS |
5337 | goto out; |
5338 | } | |
5339 | } | |
5340 | ||
5341 | if (base_fmt) { | |
5342 | if (set_option_parameter(param, BLOCK_OPT_BACKING_FMT, base_fmt)) { | |
71c79813 LC |
5343 | error_setg(errp, "Backing file format not supported for file " |
5344 | "format '%s'", fmt); | |
f88e1a42 JS |
5345 | goto out; |
5346 | } | |
5347 | } | |
5348 | ||
792da93a JS |
5349 | backing_file = get_option_parameter(param, BLOCK_OPT_BACKING_FILE); |
5350 | if (backing_file && backing_file->value.s) { | |
5351 | if (!strcmp(filename, backing_file->value.s)) { | |
71c79813 LC |
5352 | error_setg(errp, "Error: Trying to create an image with the " |
5353 | "same filename as the backing file"); | |
792da93a JS |
5354 | goto out; |
5355 | } | |
5356 | } | |
5357 | ||
f88e1a42 JS |
5358 | backing_fmt = get_option_parameter(param, BLOCK_OPT_BACKING_FMT); |
5359 | if (backing_fmt && backing_fmt->value.s) { | |
96df67d1 SH |
5360 | backing_drv = bdrv_find_format(backing_fmt->value.s); |
5361 | if (!backing_drv) { | |
71c79813 LC |
5362 | error_setg(errp, "Unknown backing file format '%s'", |
5363 | backing_fmt->value.s); | |
f88e1a42 JS |
5364 | goto out; |
5365 | } | |
5366 | } | |
5367 | ||
5368 | // The size for the image must always be specified, with one exception: | |
5369 | // If we are using a backing file, we can obtain the size from there | |
d220894e KW |
5370 | size = get_option_parameter(param, BLOCK_OPT_SIZE); |
5371 | if (size && size->value.n == -1) { | |
f88e1a42 | 5372 | if (backing_file && backing_file->value.s) { |
66f6b814 | 5373 | BlockDriverState *bs; |
f88e1a42 | 5374 | uint64_t size; |
f88e1a42 | 5375 | char buf[32]; |
63090dac PB |
5376 | int back_flags; |
5377 | ||
5378 | /* backing files always opened read-only */ | |
5379 | back_flags = | |
5380 | flags & ~(BDRV_O_RDWR | BDRV_O_SNAPSHOT | BDRV_O_NO_BACKING); | |
f88e1a42 | 5381 | |
f67503e5 | 5382 | bs = NULL; |
ddf5636d | 5383 | ret = bdrv_open(&bs, backing_file->value.s, NULL, NULL, back_flags, |
cc84d90f | 5384 | backing_drv, &local_err); |
f88e1a42 | 5385 | if (ret < 0) { |
cc84d90f HR |
5386 | error_setg_errno(errp, -ret, "Could not open '%s': %s", |
5387 | backing_file->value.s, | |
5388 | error_get_pretty(local_err)); | |
5389 | error_free(local_err); | |
5390 | local_err = NULL; | |
f88e1a42 JS |
5391 | goto out; |
5392 | } | |
5393 | bdrv_get_geometry(bs, &size); | |
5394 | size *= 512; | |
5395 | ||
5396 | snprintf(buf, sizeof(buf), "%" PRId64, size); | |
5397 | set_option_parameter(param, BLOCK_OPT_SIZE, buf); | |
66f6b814 HR |
5398 | |
5399 | bdrv_unref(bs); | |
f88e1a42 | 5400 | } else { |
71c79813 | 5401 | error_setg(errp, "Image creation needs a size parameter"); |
f88e1a42 JS |
5402 | goto out; |
5403 | } | |
5404 | } | |
5405 | ||
f382d43a MR |
5406 | if (!quiet) { |
5407 | printf("Formatting '%s', fmt=%s ", filename, fmt); | |
5408 | print_option_parameters(param); | |
5409 | puts(""); | |
5410 | } | |
cc84d90f HR |
5411 | ret = bdrv_create(drv, filename, param, &local_err); |
5412 | if (ret == -EFBIG) { | |
5413 | /* This is generally a better message than whatever the driver would | |
5414 | * deliver (especially because of the cluster_size_hint), since that | |
5415 | * is most probably not much different from "image too large". */ | |
5416 | const char *cluster_size_hint = ""; | |
5417 | if (get_option_parameter(create_options, BLOCK_OPT_CLUSTER_SIZE)) { | |
5418 | cluster_size_hint = " (try using a larger cluster size)"; | |
f88e1a42 | 5419 | } |
cc84d90f HR |
5420 | error_setg(errp, "The image size is too large for file format '%s'" |
5421 | "%s", fmt, cluster_size_hint); | |
5422 | error_free(local_err); | |
5423 | local_err = NULL; | |
f88e1a42 JS |
5424 | } |
5425 | ||
5426 | out: | |
5427 | free_option_parameters(create_options); | |
5428 | free_option_parameters(param); | |
5429 | ||
84d18f06 | 5430 | if (local_err) { |
cc84d90f HR |
5431 | error_propagate(errp, local_err); |
5432 | } | |
f88e1a42 | 5433 | } |
85d126f3 SH |
5434 | |
5435 | AioContext *bdrv_get_aio_context(BlockDriverState *bs) | |
5436 | { | |
5437 | /* Currently BlockDriverState always uses the main loop AioContext */ | |
5438 | return qemu_get_aio_context(); | |
5439 | } | |
d616b224 SH |
5440 | |
5441 | void bdrv_add_before_write_notifier(BlockDriverState *bs, | |
5442 | NotifierWithReturn *notifier) | |
5443 | { | |
5444 | notifier_with_return_list_add(&bs->before_write_notifiers, notifier); | |
5445 | } | |
6f176b48 HR |
5446 | |
5447 | int bdrv_amend_options(BlockDriverState *bs, QEMUOptionParameter *options) | |
5448 | { | |
5449 | if (bs->drv->bdrv_amend_options == NULL) { | |
5450 | return -ENOTSUP; | |
5451 | } | |
5452 | return bs->drv->bdrv_amend_options(bs, options); | |
5453 | } | |
f6186f49 | 5454 | |
b5042a36 BC |
5455 | /* This function will be called by the bdrv_recurse_is_first_non_filter method |
5456 | * of block filter and by bdrv_is_first_non_filter. | |
5457 | * It is used to test if the given bs is the candidate or recurse more in the | |
5458 | * node graph. | |
212a5a8f | 5459 | */ |
b5042a36 | 5460 | bool bdrv_recurse_is_first_non_filter(BlockDriverState *bs, |
212a5a8f | 5461 | BlockDriverState *candidate) |
f6186f49 | 5462 | { |
b5042a36 BC |
5463 | /* return false if basic checks fails */ |
5464 | if (!bs || !bs->drv) { | |
212a5a8f | 5465 | return false; |
f6186f49 BC |
5466 | } |
5467 | ||
b5042a36 BC |
5468 | /* the code reached a non block filter driver -> check if the bs is |
5469 | * the same as the candidate. It's the recursion termination condition. | |
5470 | */ | |
5471 | if (!bs->drv->is_filter) { | |
5472 | return bs == candidate; | |
212a5a8f | 5473 | } |
b5042a36 | 5474 | /* Down this path the driver is a block filter driver */ |
212a5a8f | 5475 | |
b5042a36 BC |
5476 | /* If the block filter recursion method is defined use it to recurse down |
5477 | * the node graph. | |
5478 | */ | |
5479 | if (bs->drv->bdrv_recurse_is_first_non_filter) { | |
212a5a8f | 5480 | return bs->drv->bdrv_recurse_is_first_non_filter(bs, candidate); |
f6186f49 BC |
5481 | } |
5482 | ||
b5042a36 BC |
5483 | /* the driver is a block filter but don't allow to recurse -> return false |
5484 | */ | |
5485 | return false; | |
f6186f49 BC |
5486 | } |
5487 | ||
212a5a8f BC |
5488 | /* This function checks if the candidate is the first non filter bs down it's |
5489 | * bs chain. Since we don't have pointers to parents it explore all bs chains | |
5490 | * from the top. Some filters can choose not to pass down the recursion. | |
5491 | */ | |
5492 | bool bdrv_is_first_non_filter(BlockDriverState *candidate) | |
f6186f49 | 5493 | { |
212a5a8f BC |
5494 | BlockDriverState *bs; |
5495 | ||
5496 | /* walk down the bs forest recursively */ | |
5497 | QTAILQ_FOREACH(bs, &bdrv_states, device_list) { | |
5498 | bool perm; | |
5499 | ||
b5042a36 | 5500 | /* try to recurse in this top level bs */ |
e6dc8a1f | 5501 | perm = bdrv_recurse_is_first_non_filter(bs, candidate); |
212a5a8f BC |
5502 | |
5503 | /* candidate is the first non filter */ | |
5504 | if (perm) { | |
5505 | return true; | |
5506 | } | |
5507 | } | |
5508 | ||
5509 | return false; | |
f6186f49 | 5510 | } |