]>
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) */ | |
98522f63 | 335 | BlockDriverState *bdrv_new(const char *device_name, Error **errp) |
b338082b | 336 | { |
1b7bdbc1 | 337 | BlockDriverState *bs; |
b338082b | 338 | |
f2d953ec KW |
339 | if (bdrv_find(device_name)) { |
340 | error_setg(errp, "Device with id '%s' already exists", | |
341 | device_name); | |
342 | return NULL; | |
343 | } | |
344 | if (bdrv_find_node(device_name)) { | |
345 | error_setg(errp, "Device with node-name '%s' already exists", | |
346 | device_name); | |
347 | return NULL; | |
348 | } | |
349 | ||
7267c094 | 350 | bs = g_malloc0(sizeof(BlockDriverState)); |
e4654d2d | 351 | QLIST_INIT(&bs->dirty_bitmaps); |
b338082b | 352 | pstrcpy(bs->device_name, sizeof(bs->device_name), device_name); |
ea2384d3 | 353 | if (device_name[0] != '\0') { |
dc364f4c | 354 | QTAILQ_INSERT_TAIL(&bdrv_states, bs, device_list); |
ea2384d3 | 355 | } |
28a7282a | 356 | bdrv_iostatus_disable(bs); |
d7d512f6 | 357 | notifier_list_init(&bs->close_notifiers); |
d616b224 | 358 | notifier_with_return_list_init(&bs->before_write_notifiers); |
cc0681c4 BC |
359 | qemu_co_queue_init(&bs->throttled_reqs[0]); |
360 | qemu_co_queue_init(&bs->throttled_reqs[1]); | |
9fcb0251 | 361 | bs->refcnt = 1; |
d7d512f6 | 362 | |
b338082b FB |
363 | return bs; |
364 | } | |
365 | ||
d7d512f6 PB |
366 | void bdrv_add_close_notifier(BlockDriverState *bs, Notifier *notify) |
367 | { | |
368 | notifier_list_add(&bs->close_notifiers, notify); | |
369 | } | |
370 | ||
ea2384d3 FB |
371 | BlockDriver *bdrv_find_format(const char *format_name) |
372 | { | |
373 | BlockDriver *drv1; | |
8a22f02a SH |
374 | QLIST_FOREACH(drv1, &bdrv_drivers, list) { |
375 | if (!strcmp(drv1->format_name, format_name)) { | |
ea2384d3 | 376 | return drv1; |
8a22f02a | 377 | } |
ea2384d3 FB |
378 | } |
379 | return NULL; | |
380 | } | |
381 | ||
b64ec4e4 | 382 | static int bdrv_is_whitelisted(BlockDriver *drv, bool read_only) |
eb852011 | 383 | { |
b64ec4e4 FZ |
384 | static const char *whitelist_rw[] = { |
385 | CONFIG_BDRV_RW_WHITELIST | |
386 | }; | |
387 | static const char *whitelist_ro[] = { | |
388 | CONFIG_BDRV_RO_WHITELIST | |
eb852011 MA |
389 | }; |
390 | const char **p; | |
391 | ||
b64ec4e4 | 392 | if (!whitelist_rw[0] && !whitelist_ro[0]) { |
eb852011 | 393 | return 1; /* no whitelist, anything goes */ |
b64ec4e4 | 394 | } |
eb852011 | 395 | |
b64ec4e4 | 396 | for (p = whitelist_rw; *p; p++) { |
eb852011 MA |
397 | if (!strcmp(drv->format_name, *p)) { |
398 | return 1; | |
399 | } | |
400 | } | |
b64ec4e4 FZ |
401 | if (read_only) { |
402 | for (p = whitelist_ro; *p; p++) { | |
403 | if (!strcmp(drv->format_name, *p)) { | |
404 | return 1; | |
405 | } | |
406 | } | |
407 | } | |
eb852011 MA |
408 | return 0; |
409 | } | |
410 | ||
b64ec4e4 FZ |
411 | BlockDriver *bdrv_find_whitelisted_format(const char *format_name, |
412 | bool read_only) | |
eb852011 MA |
413 | { |
414 | BlockDriver *drv = bdrv_find_format(format_name); | |
b64ec4e4 | 415 | return drv && bdrv_is_whitelisted(drv, read_only) ? drv : NULL; |
eb852011 MA |
416 | } |
417 | ||
5b7e1542 ZYW |
418 | typedef struct CreateCo { |
419 | BlockDriver *drv; | |
420 | char *filename; | |
421 | QEMUOptionParameter *options; | |
422 | int ret; | |
cc84d90f | 423 | Error *err; |
5b7e1542 ZYW |
424 | } CreateCo; |
425 | ||
426 | static void coroutine_fn bdrv_create_co_entry(void *opaque) | |
427 | { | |
cc84d90f HR |
428 | Error *local_err = NULL; |
429 | int ret; | |
430 | ||
5b7e1542 ZYW |
431 | CreateCo *cco = opaque; |
432 | assert(cco->drv); | |
433 | ||
cc84d90f | 434 | ret = cco->drv->bdrv_create(cco->filename, cco->options, &local_err); |
84d18f06 | 435 | if (local_err) { |
cc84d90f HR |
436 | error_propagate(&cco->err, local_err); |
437 | } | |
438 | cco->ret = ret; | |
5b7e1542 ZYW |
439 | } |
440 | ||
0e7e1989 | 441 | int bdrv_create(BlockDriver *drv, const char* filename, |
cc84d90f | 442 | QEMUOptionParameter *options, Error **errp) |
ea2384d3 | 443 | { |
5b7e1542 ZYW |
444 | int ret; |
445 | ||
446 | Coroutine *co; | |
447 | CreateCo cco = { | |
448 | .drv = drv, | |
449 | .filename = g_strdup(filename), | |
450 | .options = options, | |
451 | .ret = NOT_DONE, | |
cc84d90f | 452 | .err = NULL, |
5b7e1542 ZYW |
453 | }; |
454 | ||
455 | if (!drv->bdrv_create) { | |
cc84d90f | 456 | error_setg(errp, "Driver '%s' does not support image creation", drv->format_name); |
80168bff LC |
457 | ret = -ENOTSUP; |
458 | goto out; | |
5b7e1542 ZYW |
459 | } |
460 | ||
461 | if (qemu_in_coroutine()) { | |
462 | /* Fast-path if already in coroutine context */ | |
463 | bdrv_create_co_entry(&cco); | |
464 | } else { | |
465 | co = qemu_coroutine_create(bdrv_create_co_entry); | |
466 | qemu_coroutine_enter(co, &cco); | |
467 | while (cco.ret == NOT_DONE) { | |
468 | qemu_aio_wait(); | |
469 | } | |
470 | } | |
471 | ||
472 | ret = cco.ret; | |
cc84d90f | 473 | if (ret < 0) { |
84d18f06 | 474 | if (cco.err) { |
cc84d90f HR |
475 | error_propagate(errp, cco.err); |
476 | } else { | |
477 | error_setg_errno(errp, -ret, "Could not create image"); | |
478 | } | |
479 | } | |
0e7e1989 | 480 | |
80168bff LC |
481 | out: |
482 | g_free(cco.filename); | |
5b7e1542 | 483 | return ret; |
ea2384d3 FB |
484 | } |
485 | ||
cc84d90f HR |
486 | int bdrv_create_file(const char* filename, QEMUOptionParameter *options, |
487 | Error **errp) | |
84a12e66 CH |
488 | { |
489 | BlockDriver *drv; | |
cc84d90f HR |
490 | Error *local_err = NULL; |
491 | int ret; | |
84a12e66 | 492 | |
98289620 | 493 | drv = bdrv_find_protocol(filename, true); |
84a12e66 | 494 | if (drv == NULL) { |
cc84d90f | 495 | error_setg(errp, "Could not find protocol for file '%s'", filename); |
16905d71 | 496 | return -ENOENT; |
84a12e66 CH |
497 | } |
498 | ||
cc84d90f | 499 | ret = bdrv_create(drv, filename, options, &local_err); |
84d18f06 | 500 | if (local_err) { |
cc84d90f HR |
501 | error_propagate(errp, local_err); |
502 | } | |
503 | return ret; | |
84a12e66 CH |
504 | } |
505 | ||
355ef4ac | 506 | int bdrv_refresh_limits(BlockDriverState *bs) |
d34682cd KW |
507 | { |
508 | BlockDriver *drv = bs->drv; | |
509 | ||
510 | memset(&bs->bl, 0, sizeof(bs->bl)); | |
511 | ||
466ad822 KW |
512 | if (!drv) { |
513 | return 0; | |
514 | } | |
515 | ||
516 | /* Take some limits from the children as a default */ | |
517 | if (bs->file) { | |
518 | bdrv_refresh_limits(bs->file); | |
519 | bs->bl.opt_transfer_length = bs->file->bl.opt_transfer_length; | |
339064d5 KW |
520 | bs->bl.opt_mem_alignment = bs->file->bl.opt_mem_alignment; |
521 | } else { | |
522 | bs->bl.opt_mem_alignment = 512; | |
466ad822 KW |
523 | } |
524 | ||
525 | if (bs->backing_hd) { | |
526 | bdrv_refresh_limits(bs->backing_hd); | |
527 | bs->bl.opt_transfer_length = | |
528 | MAX(bs->bl.opt_transfer_length, | |
529 | bs->backing_hd->bl.opt_transfer_length); | |
339064d5 KW |
530 | bs->bl.opt_mem_alignment = |
531 | MAX(bs->bl.opt_mem_alignment, | |
532 | bs->backing_hd->bl.opt_mem_alignment); | |
466ad822 KW |
533 | } |
534 | ||
535 | /* Then let the driver override it */ | |
536 | if (drv->bdrv_refresh_limits) { | |
d34682cd KW |
537 | return drv->bdrv_refresh_limits(bs); |
538 | } | |
539 | ||
540 | return 0; | |
541 | } | |
542 | ||
eba25057 JM |
543 | /* |
544 | * Create a uniquely-named empty temporary file. | |
545 | * Return 0 upon success, otherwise a negative errno value. | |
546 | */ | |
547 | int get_tmp_filename(char *filename, int size) | |
d5249393 | 548 | { |
eba25057 | 549 | #ifdef _WIN32 |
3b9f94e1 | 550 | char temp_dir[MAX_PATH]; |
eba25057 JM |
551 | /* GetTempFileName requires that its output buffer (4th param) |
552 | have length MAX_PATH or greater. */ | |
553 | assert(size >= MAX_PATH); | |
554 | return (GetTempPath(MAX_PATH, temp_dir) | |
555 | && GetTempFileName(temp_dir, "qem", 0, filename) | |
556 | ? 0 : -GetLastError()); | |
d5249393 | 557 | #else |
67b915a5 | 558 | int fd; |
7ccfb2eb | 559 | const char *tmpdir; |
0badc1ee | 560 | tmpdir = getenv("TMPDIR"); |
69bef793 AS |
561 | if (!tmpdir) { |
562 | tmpdir = "/var/tmp"; | |
563 | } | |
eba25057 JM |
564 | if (snprintf(filename, size, "%s/vl.XXXXXX", tmpdir) >= size) { |
565 | return -EOVERFLOW; | |
566 | } | |
ea2384d3 | 567 | fd = mkstemp(filename); |
fe235a06 DH |
568 | if (fd < 0) { |
569 | return -errno; | |
570 | } | |
571 | if (close(fd) != 0) { | |
572 | unlink(filename); | |
eba25057 JM |
573 | return -errno; |
574 | } | |
575 | return 0; | |
d5249393 | 576 | #endif |
eba25057 | 577 | } |
fc01f7e7 | 578 | |
84a12e66 CH |
579 | /* |
580 | * Detect host devices. By convention, /dev/cdrom[N] is always | |
581 | * recognized as a host CDROM. | |
582 | */ | |
583 | static BlockDriver *find_hdev_driver(const char *filename) | |
584 | { | |
585 | int score_max = 0, score; | |
586 | BlockDriver *drv = NULL, *d; | |
587 | ||
588 | QLIST_FOREACH(d, &bdrv_drivers, list) { | |
589 | if (d->bdrv_probe_device) { | |
590 | score = d->bdrv_probe_device(filename); | |
591 | if (score > score_max) { | |
592 | score_max = score; | |
593 | drv = d; | |
594 | } | |
595 | } | |
596 | } | |
597 | ||
598 | return drv; | |
599 | } | |
600 | ||
98289620 KW |
601 | BlockDriver *bdrv_find_protocol(const char *filename, |
602 | bool allow_protocol_prefix) | |
83f64091 FB |
603 | { |
604 | BlockDriver *drv1; | |
605 | char protocol[128]; | |
1cec71e3 | 606 | int len; |
83f64091 | 607 | const char *p; |
19cb3738 | 608 | |
66f82cee KW |
609 | /* TODO Drivers without bdrv_file_open must be specified explicitly */ |
610 | ||
39508e7a CH |
611 | /* |
612 | * XXX(hch): we really should not let host device detection | |
613 | * override an explicit protocol specification, but moving this | |
614 | * later breaks access to device names with colons in them. | |
615 | * Thanks to the brain-dead persistent naming schemes on udev- | |
616 | * based Linux systems those actually are quite common. | |
617 | */ | |
618 | drv1 = find_hdev_driver(filename); | |
619 | if (drv1) { | |
620 | return drv1; | |
621 | } | |
622 | ||
98289620 | 623 | if (!path_has_protocol(filename) || !allow_protocol_prefix) { |
39508e7a | 624 | return bdrv_find_format("file"); |
84a12e66 | 625 | } |
98289620 | 626 | |
9e0b22f4 SH |
627 | p = strchr(filename, ':'); |
628 | assert(p != NULL); | |
1cec71e3 AL |
629 | len = p - filename; |
630 | if (len > sizeof(protocol) - 1) | |
631 | len = sizeof(protocol) - 1; | |
632 | memcpy(protocol, filename, len); | |
633 | protocol[len] = '\0'; | |
8a22f02a | 634 | QLIST_FOREACH(drv1, &bdrv_drivers, list) { |
5fafdf24 | 635 | if (drv1->protocol_name && |
8a22f02a | 636 | !strcmp(drv1->protocol_name, protocol)) { |
83f64091 | 637 | return drv1; |
8a22f02a | 638 | } |
83f64091 FB |
639 | } |
640 | return NULL; | |
641 | } | |
642 | ||
f500a6d3 | 643 | static int find_image_format(BlockDriverState *bs, const char *filename, |
34b5d2c6 | 644 | BlockDriver **pdrv, Error **errp) |
f3a5d3f8 | 645 | { |
f500a6d3 | 646 | int score, score_max; |
f3a5d3f8 CH |
647 | BlockDriver *drv1, *drv; |
648 | uint8_t buf[2048]; | |
f500a6d3 | 649 | int ret = 0; |
f8ea0b00 | 650 | |
08a00559 | 651 | /* Return the raw BlockDriver * to scsi-generic devices or empty drives */ |
8e895599 | 652 | if (bs->sg || !bdrv_is_inserted(bs) || bdrv_getlength(bs) == 0) { |
c98ac35d SW |
653 | drv = bdrv_find_format("raw"); |
654 | if (!drv) { | |
34b5d2c6 | 655 | error_setg(errp, "Could not find raw image format"); |
c98ac35d SW |
656 | ret = -ENOENT; |
657 | } | |
658 | *pdrv = drv; | |
659 | return ret; | |
1a396859 | 660 | } |
f8ea0b00 | 661 | |
83f64091 | 662 | ret = bdrv_pread(bs, 0, buf, sizeof(buf)); |
83f64091 | 663 | if (ret < 0) { |
34b5d2c6 HR |
664 | error_setg_errno(errp, -ret, "Could not read image for determining its " |
665 | "format"); | |
c98ac35d SW |
666 | *pdrv = NULL; |
667 | return ret; | |
83f64091 FB |
668 | } |
669 | ||
ea2384d3 | 670 | score_max = 0; |
84a12e66 | 671 | drv = NULL; |
8a22f02a | 672 | QLIST_FOREACH(drv1, &bdrv_drivers, list) { |
83f64091 FB |
673 | if (drv1->bdrv_probe) { |
674 | score = drv1->bdrv_probe(buf, ret, filename); | |
675 | if (score > score_max) { | |
676 | score_max = score; | |
677 | drv = drv1; | |
678 | } | |
0849bf08 | 679 | } |
fc01f7e7 | 680 | } |
c98ac35d | 681 | if (!drv) { |
34b5d2c6 HR |
682 | error_setg(errp, "Could not determine image format: No compatible " |
683 | "driver found"); | |
c98ac35d SW |
684 | ret = -ENOENT; |
685 | } | |
686 | *pdrv = drv; | |
687 | return ret; | |
ea2384d3 FB |
688 | } |
689 | ||
51762288 SH |
690 | /** |
691 | * Set the current 'total_sectors' value | |
692 | */ | |
693 | static int refresh_total_sectors(BlockDriverState *bs, int64_t hint) | |
694 | { | |
695 | BlockDriver *drv = bs->drv; | |
696 | ||
396759ad NB |
697 | /* Do not attempt drv->bdrv_getlength() on scsi-generic devices */ |
698 | if (bs->sg) | |
699 | return 0; | |
700 | ||
51762288 SH |
701 | /* query actual device if possible, otherwise just trust the hint */ |
702 | if (drv->bdrv_getlength) { | |
703 | int64_t length = drv->bdrv_getlength(bs); | |
704 | if (length < 0) { | |
705 | return length; | |
706 | } | |
7e382003 | 707 | hint = DIV_ROUND_UP(length, BDRV_SECTOR_SIZE); |
51762288 SH |
708 | } |
709 | ||
710 | bs->total_sectors = hint; | |
711 | return 0; | |
712 | } | |
713 | ||
9e8f1835 PB |
714 | /** |
715 | * Set open flags for a given discard mode | |
716 | * | |
717 | * Return 0 on success, -1 if the discard mode was invalid. | |
718 | */ | |
719 | int bdrv_parse_discard_flags(const char *mode, int *flags) | |
720 | { | |
721 | *flags &= ~BDRV_O_UNMAP; | |
722 | ||
723 | if (!strcmp(mode, "off") || !strcmp(mode, "ignore")) { | |
724 | /* do nothing */ | |
725 | } else if (!strcmp(mode, "on") || !strcmp(mode, "unmap")) { | |
726 | *flags |= BDRV_O_UNMAP; | |
727 | } else { | |
728 | return -1; | |
729 | } | |
730 | ||
731 | return 0; | |
732 | } | |
733 | ||
c3993cdc SH |
734 | /** |
735 | * Set open flags for a given cache mode | |
736 | * | |
737 | * Return 0 on success, -1 if the cache mode was invalid. | |
738 | */ | |
739 | int bdrv_parse_cache_flags(const char *mode, int *flags) | |
740 | { | |
741 | *flags &= ~BDRV_O_CACHE_MASK; | |
742 | ||
743 | if (!strcmp(mode, "off") || !strcmp(mode, "none")) { | |
744 | *flags |= BDRV_O_NOCACHE | BDRV_O_CACHE_WB; | |
92196b2f SH |
745 | } else if (!strcmp(mode, "directsync")) { |
746 | *flags |= BDRV_O_NOCACHE; | |
c3993cdc SH |
747 | } else if (!strcmp(mode, "writeback")) { |
748 | *flags |= BDRV_O_CACHE_WB; | |
749 | } else if (!strcmp(mode, "unsafe")) { | |
750 | *flags |= BDRV_O_CACHE_WB; | |
751 | *flags |= BDRV_O_NO_FLUSH; | |
752 | } else if (!strcmp(mode, "writethrough")) { | |
753 | /* this is the default */ | |
754 | } else { | |
755 | return -1; | |
756 | } | |
757 | ||
758 | return 0; | |
759 | } | |
760 | ||
53fec9d3 SH |
761 | /** |
762 | * The copy-on-read flag is actually a reference count so multiple users may | |
763 | * use the feature without worrying about clobbering its previous state. | |
764 | * Copy-on-read stays enabled until all users have called to disable it. | |
765 | */ | |
766 | void bdrv_enable_copy_on_read(BlockDriverState *bs) | |
767 | { | |
768 | bs->copy_on_read++; | |
769 | } | |
770 | ||
771 | void bdrv_disable_copy_on_read(BlockDriverState *bs) | |
772 | { | |
773 | assert(bs->copy_on_read > 0); | |
774 | bs->copy_on_read--; | |
775 | } | |
776 | ||
b1e6fc08 KW |
777 | /* |
778 | * Returns the flags that a temporary snapshot should get, based on the | |
779 | * originally requested flags (the originally requested image will have flags | |
780 | * like a backing file) | |
781 | */ | |
782 | static int bdrv_temp_snapshot_flags(int flags) | |
783 | { | |
784 | return (flags & ~BDRV_O_SNAPSHOT) | BDRV_O_TEMPORARY; | |
785 | } | |
786 | ||
0b50cc88 KW |
787 | /* |
788 | * Returns the flags that bs->file should get, based on the given flags for | |
789 | * the parent BDS | |
790 | */ | |
791 | static int bdrv_inherited_flags(int flags) | |
792 | { | |
793 | /* Enable protocol handling, disable format probing for bs->file */ | |
794 | flags |= BDRV_O_PROTOCOL; | |
795 | ||
796 | /* Our block drivers take care to send flushes and respect unmap policy, | |
797 | * so we can enable both unconditionally on lower layers. */ | |
798 | flags |= BDRV_O_CACHE_WB | BDRV_O_UNMAP; | |
799 | ||
0b50cc88 | 800 | /* Clear flags that only apply to the top layer */ |
5669b44d | 801 | flags &= ~(BDRV_O_SNAPSHOT | BDRV_O_NO_BACKING | BDRV_O_COPY_ON_READ); |
0b50cc88 KW |
802 | |
803 | return flags; | |
804 | } | |
805 | ||
317fc44e KW |
806 | /* |
807 | * Returns the flags that bs->backing_hd should get, based on the given flags | |
808 | * for the parent BDS | |
809 | */ | |
810 | static int bdrv_backing_flags(int flags) | |
811 | { | |
812 | /* backing files always opened read-only */ | |
813 | flags &= ~(BDRV_O_RDWR | BDRV_O_COPY_ON_READ); | |
814 | ||
815 | /* snapshot=on is handled on the top layer */ | |
8bfea15d | 816 | flags &= ~(BDRV_O_SNAPSHOT | BDRV_O_TEMPORARY); |
317fc44e KW |
817 | |
818 | return flags; | |
819 | } | |
820 | ||
7b272452 KW |
821 | static int bdrv_open_flags(BlockDriverState *bs, int flags) |
822 | { | |
823 | int open_flags = flags | BDRV_O_CACHE_WB; | |
824 | ||
825 | /* | |
826 | * Clear flags that are internal to the block layer before opening the | |
827 | * image. | |
828 | */ | |
829 | open_flags &= ~(BDRV_O_SNAPSHOT | BDRV_O_NO_BACKING); | |
830 | ||
831 | /* | |
832 | * Snapshots should be writable. | |
833 | */ | |
8bfea15d | 834 | if (flags & BDRV_O_TEMPORARY) { |
7b272452 KW |
835 | open_flags |= BDRV_O_RDWR; |
836 | } | |
837 | ||
838 | return open_flags; | |
839 | } | |
840 | ||
636ea370 KW |
841 | static void bdrv_assign_node_name(BlockDriverState *bs, |
842 | const char *node_name, | |
843 | Error **errp) | |
6913c0c2 BC |
844 | { |
845 | if (!node_name) { | |
636ea370 | 846 | return; |
6913c0c2 BC |
847 | } |
848 | ||
849 | /* empty string node name is invalid */ | |
850 | if (node_name[0] == '\0') { | |
851 | error_setg(errp, "Empty node name"); | |
636ea370 | 852 | return; |
6913c0c2 BC |
853 | } |
854 | ||
0c5e94ee BC |
855 | /* takes care of avoiding namespaces collisions */ |
856 | if (bdrv_find(node_name)) { | |
857 | error_setg(errp, "node-name=%s is conflicting with a device id", | |
858 | node_name); | |
636ea370 | 859 | return; |
0c5e94ee BC |
860 | } |
861 | ||
6913c0c2 BC |
862 | /* takes care of avoiding duplicates node names */ |
863 | if (bdrv_find_node(node_name)) { | |
864 | error_setg(errp, "Duplicate node name"); | |
636ea370 | 865 | return; |
6913c0c2 BC |
866 | } |
867 | ||
868 | /* copy node name into the bs and insert it into the graph list */ | |
869 | pstrcpy(bs->node_name, sizeof(bs->node_name), node_name); | |
870 | QTAILQ_INSERT_TAIL(&graph_bdrv_states, bs, node_list); | |
6913c0c2 BC |
871 | } |
872 | ||
57915332 KW |
873 | /* |
874 | * Common part for opening disk images and files | |
b6ad491a KW |
875 | * |
876 | * Removes all processed options from *options. | |
57915332 | 877 | */ |
f500a6d3 | 878 | static int bdrv_open_common(BlockDriverState *bs, BlockDriverState *file, |
34b5d2c6 | 879 | QDict *options, int flags, BlockDriver *drv, Error **errp) |
57915332 KW |
880 | { |
881 | int ret, open_flags; | |
035fccdf | 882 | const char *filename; |
6913c0c2 | 883 | const char *node_name = NULL; |
34b5d2c6 | 884 | Error *local_err = NULL; |
57915332 KW |
885 | |
886 | assert(drv != NULL); | |
6405875c | 887 | assert(bs->file == NULL); |
707ff828 | 888 | assert(options != NULL && bs->options != options); |
57915332 | 889 | |
45673671 KW |
890 | if (file != NULL) { |
891 | filename = file->filename; | |
892 | } else { | |
893 | filename = qdict_get_try_str(options, "filename"); | |
894 | } | |
895 | ||
765003db KW |
896 | if (drv->bdrv_needs_filename && !filename) { |
897 | error_setg(errp, "The '%s' block driver requires a file name", | |
898 | drv->format_name); | |
899 | return -EINVAL; | |
900 | } | |
901 | ||
45673671 | 902 | trace_bdrv_open_common(bs, filename ?: "", flags, drv->format_name); |
28dcee10 | 903 | |
6913c0c2 | 904 | node_name = qdict_get_try_str(options, "node-name"); |
636ea370 | 905 | bdrv_assign_node_name(bs, node_name, &local_err); |
0fb6395c | 906 | if (local_err) { |
636ea370 KW |
907 | error_propagate(errp, local_err); |
908 | return -EINVAL; | |
6913c0c2 BC |
909 | } |
910 | qdict_del(options, "node-name"); | |
911 | ||
5d186eb0 KW |
912 | /* bdrv_open() with directly using a protocol as drv. This layer is already |
913 | * opened, so assign it to bs (while file becomes a closed BlockDriverState) | |
914 | * and return immediately. */ | |
915 | if (file != NULL && drv->bdrv_file_open) { | |
916 | bdrv_swap(file, bs); | |
917 | return 0; | |
918 | } | |
919 | ||
57915332 | 920 | bs->open_flags = flags; |
1b7fd729 | 921 | bs->guest_block_size = 512; |
c25f53b0 | 922 | bs->request_alignment = 512; |
0d51b4de | 923 | bs->zero_beyond_eof = true; |
b64ec4e4 FZ |
924 | open_flags = bdrv_open_flags(bs, flags); |
925 | bs->read_only = !(open_flags & BDRV_O_RDWR); | |
926 | ||
927 | if (use_bdrv_whitelist && !bdrv_is_whitelisted(drv, bs->read_only)) { | |
8f94a6e4 KW |
928 | error_setg(errp, |
929 | !bs->read_only && bdrv_is_whitelisted(drv, true) | |
930 | ? "Driver '%s' can only be used for read-only devices" | |
931 | : "Driver '%s' is not whitelisted", | |
932 | drv->format_name); | |
b64ec4e4 FZ |
933 | return -ENOTSUP; |
934 | } | |
57915332 | 935 | |
53fec9d3 | 936 | assert(bs->copy_on_read == 0); /* bdrv_new() and bdrv_close() make it so */ |
0ebd24e0 KW |
937 | if (flags & BDRV_O_COPY_ON_READ) { |
938 | if (!bs->read_only) { | |
939 | bdrv_enable_copy_on_read(bs); | |
940 | } else { | |
941 | error_setg(errp, "Can't use copy-on-read on read-only device"); | |
942 | return -EINVAL; | |
943 | } | |
53fec9d3 SH |
944 | } |
945 | ||
c2ad1b0c KW |
946 | if (filename != NULL) { |
947 | pstrcpy(bs->filename, sizeof(bs->filename), filename); | |
948 | } else { | |
949 | bs->filename[0] = '\0'; | |
950 | } | |
57915332 | 951 | |
57915332 | 952 | bs->drv = drv; |
7267c094 | 953 | bs->opaque = g_malloc0(drv->instance_size); |
57915332 | 954 | |
03f541bd | 955 | bs->enable_write_cache = !!(flags & BDRV_O_CACHE_WB); |
e7c63796 | 956 | |
66f82cee KW |
957 | /* Open the image, either directly or using a protocol */ |
958 | if (drv->bdrv_file_open) { | |
5d186eb0 | 959 | assert(file == NULL); |
030be321 | 960 | assert(!drv->bdrv_needs_filename || filename != NULL); |
34b5d2c6 | 961 | ret = drv->bdrv_file_open(bs, options, open_flags, &local_err); |
f500a6d3 | 962 | } else { |
2af5ef70 | 963 | if (file == NULL) { |
34b5d2c6 HR |
964 | error_setg(errp, "Can't use '%s' as a block driver for the " |
965 | "protocol level", drv->format_name); | |
2af5ef70 KW |
966 | ret = -EINVAL; |
967 | goto free_and_fail; | |
968 | } | |
f500a6d3 | 969 | bs->file = file; |
34b5d2c6 | 970 | ret = drv->bdrv_open(bs, options, open_flags, &local_err); |
66f82cee KW |
971 | } |
972 | ||
57915332 | 973 | if (ret < 0) { |
84d18f06 | 974 | if (local_err) { |
34b5d2c6 | 975 | error_propagate(errp, local_err); |
2fa9aa59 DH |
976 | } else if (bs->filename[0]) { |
977 | error_setg_errno(errp, -ret, "Could not open '%s'", bs->filename); | |
34b5d2c6 HR |
978 | } else { |
979 | error_setg_errno(errp, -ret, "Could not open image"); | |
980 | } | |
57915332 KW |
981 | goto free_and_fail; |
982 | } | |
983 | ||
51762288 SH |
984 | ret = refresh_total_sectors(bs, bs->total_sectors); |
985 | if (ret < 0) { | |
34b5d2c6 | 986 | error_setg_errno(errp, -ret, "Could not refresh total sector count"); |
51762288 | 987 | goto free_and_fail; |
57915332 | 988 | } |
51762288 | 989 | |
d34682cd | 990 | bdrv_refresh_limits(bs); |
c25f53b0 | 991 | assert(bdrv_opt_mem_align(bs) != 0); |
47ea2de2 | 992 | assert((bs->request_alignment != 0) || bs->sg); |
57915332 KW |
993 | return 0; |
994 | ||
995 | free_and_fail: | |
f500a6d3 | 996 | bs->file = NULL; |
7267c094 | 997 | g_free(bs->opaque); |
57915332 KW |
998 | bs->opaque = NULL; |
999 | bs->drv = NULL; | |
1000 | return ret; | |
1001 | } | |
1002 | ||
b6ce07aa KW |
1003 | /* |
1004 | * Opens a file using a protocol (file, host_device, nbd, ...) | |
787e4a85 | 1005 | * |
5acd9d81 HR |
1006 | * options is an indirect pointer to a QDict of options to pass to the block |
1007 | * drivers, or pointer to NULL for an empty set of options. If this function | |
1008 | * takes ownership of the QDict reference, it will set *options to NULL; | |
1009 | * otherwise, it will contain unused/unrecognized options after this function | |
1010 | * returns. Then, the caller is responsible for freeing it. If it intends to | |
1011 | * reuse the QDict, QINCREF() should be called beforehand. | |
b6ce07aa | 1012 | */ |
d4446eae | 1013 | static int bdrv_file_open(BlockDriverState *bs, const char *filename, |
5acd9d81 | 1014 | QDict **options, int flags, Error **errp) |
ea2384d3 | 1015 | { |
6db95603 | 1016 | BlockDriver *drv; |
c2ad1b0c | 1017 | const char *drvname; |
e3fa4bfa | 1018 | bool parse_filename = false; |
34b5d2c6 | 1019 | Error *local_err = NULL; |
83f64091 FB |
1020 | int ret; |
1021 | ||
035fccdf KW |
1022 | /* Fetch the file name from the options QDict if necessary */ |
1023 | if (!filename) { | |
5acd9d81 HR |
1024 | filename = qdict_get_try_str(*options, "filename"); |
1025 | } else if (filename && !qdict_haskey(*options, "filename")) { | |
1026 | qdict_put(*options, "filename", qstring_from_str(filename)); | |
e3fa4bfa | 1027 | parse_filename = true; |
035fccdf | 1028 | } else { |
34b5d2c6 HR |
1029 | error_setg(errp, "Can't specify 'file' and 'filename' options at the " |
1030 | "same time"); | |
035fccdf KW |
1031 | ret = -EINVAL; |
1032 | goto fail; | |
1033 | } | |
1034 | ||
c2ad1b0c | 1035 | /* Find the right block driver */ |
5acd9d81 | 1036 | drvname = qdict_get_try_str(*options, "driver"); |
c2ad1b0c | 1037 | if (drvname) { |
8f94a6e4 | 1038 | drv = bdrv_find_format(drvname); |
34b5d2c6 HR |
1039 | if (!drv) { |
1040 | error_setg(errp, "Unknown driver '%s'", drvname); | |
1041 | } | |
5acd9d81 | 1042 | qdict_del(*options, "driver"); |
c2ad1b0c | 1043 | } else if (filename) { |
e3fa4bfa | 1044 | drv = bdrv_find_protocol(filename, parse_filename); |
98289620 | 1045 | if (!drv) { |
34b5d2c6 | 1046 | error_setg(errp, "Unknown protocol"); |
98289620 | 1047 | } |
c2ad1b0c | 1048 | } else { |
34b5d2c6 | 1049 | error_setg(errp, "Must specify either driver or file"); |
c2ad1b0c KW |
1050 | drv = NULL; |
1051 | } | |
1052 | ||
1053 | if (!drv) { | |
34b5d2c6 | 1054 | /* errp has been set already */ |
c2ad1b0c KW |
1055 | ret = -ENOENT; |
1056 | goto fail; | |
1057 | } | |
1058 | ||
1059 | /* Parse the filename and open it */ | |
e3fa4bfa | 1060 | if (drv->bdrv_parse_filename && parse_filename) { |
5acd9d81 | 1061 | drv->bdrv_parse_filename(filename, *options, &local_err); |
84d18f06 | 1062 | if (local_err) { |
34b5d2c6 | 1063 | error_propagate(errp, local_err); |
6963a30d KW |
1064 | ret = -EINVAL; |
1065 | goto fail; | |
1066 | } | |
cd5d031e HR |
1067 | |
1068 | if (!drv->bdrv_needs_filename) { | |
1069 | qdict_del(*options, "filename"); | |
1070 | } else { | |
1071 | filename = qdict_get_str(*options, "filename"); | |
1072 | } | |
6963a30d KW |
1073 | } |
1074 | ||
505d7583 | 1075 | if (!drv->bdrv_file_open) { |
5acd9d81 HR |
1076 | ret = bdrv_open(&bs, filename, NULL, *options, flags, drv, &local_err); |
1077 | *options = NULL; | |
505d7583 | 1078 | } else { |
5acd9d81 | 1079 | ret = bdrv_open_common(bs, NULL, *options, flags, drv, &local_err); |
505d7583 | 1080 | } |
83f64091 | 1081 | if (ret < 0) { |
34b5d2c6 | 1082 | error_propagate(errp, local_err); |
707ff828 KW |
1083 | goto fail; |
1084 | } | |
1085 | ||
71d0770c | 1086 | bs->growable = 1; |
83f64091 | 1087 | return 0; |
707ff828 KW |
1088 | |
1089 | fail: | |
707ff828 | 1090 | return ret; |
83f64091 FB |
1091 | } |
1092 | ||
31ca6d07 KW |
1093 | /* |
1094 | * Opens the backing file for a BlockDriverState if not yet open | |
1095 | * | |
1096 | * options is a QDict of options to pass to the block drivers, or NULL for an | |
1097 | * empty set of options. The reference to the QDict is transferred to this | |
1098 | * function (even on failure), so if the caller intends to reuse the dictionary, | |
1099 | * it needs to use QINCREF() before calling bdrv_file_open. | |
1100 | */ | |
34b5d2c6 | 1101 | int bdrv_open_backing_file(BlockDriverState *bs, QDict *options, Error **errp) |
9156df12 | 1102 | { |
1ba4b6a5 | 1103 | char *backing_filename = g_malloc0(PATH_MAX); |
317fc44e | 1104 | int ret = 0; |
9156df12 | 1105 | BlockDriver *back_drv = NULL; |
34b5d2c6 | 1106 | Error *local_err = NULL; |
9156df12 PB |
1107 | |
1108 | if (bs->backing_hd != NULL) { | |
31ca6d07 | 1109 | QDECREF(options); |
1ba4b6a5 | 1110 | goto free_exit; |
9156df12 PB |
1111 | } |
1112 | ||
31ca6d07 KW |
1113 | /* NULL means an empty set of options */ |
1114 | if (options == NULL) { | |
1115 | options = qdict_new(); | |
1116 | } | |
1117 | ||
9156df12 | 1118 | bs->open_flags &= ~BDRV_O_NO_BACKING; |
1cb6f506 KW |
1119 | if (qdict_haskey(options, "file.filename")) { |
1120 | backing_filename[0] = '\0'; | |
1121 | } else if (bs->backing_file[0] == '\0' && qdict_size(options) == 0) { | |
31ca6d07 | 1122 | QDECREF(options); |
1ba4b6a5 | 1123 | goto free_exit; |
dbecebdd | 1124 | } else { |
1ba4b6a5 | 1125 | bdrv_get_full_backing_filename(bs, backing_filename, PATH_MAX); |
9156df12 PB |
1126 | } |
1127 | ||
9156df12 PB |
1128 | if (bs->backing_format[0] != '\0') { |
1129 | back_drv = bdrv_find_format(bs->backing_format); | |
1130 | } | |
1131 | ||
f67503e5 HR |
1132 | assert(bs->backing_hd == NULL); |
1133 | ret = bdrv_open(&bs->backing_hd, | |
ddf5636d | 1134 | *backing_filename ? backing_filename : NULL, NULL, options, |
317fc44e | 1135 | bdrv_backing_flags(bs->open_flags), back_drv, &local_err); |
9156df12 | 1136 | if (ret < 0) { |
9156df12 PB |
1137 | bs->backing_hd = NULL; |
1138 | bs->open_flags |= BDRV_O_NO_BACKING; | |
b04b6b6e FZ |
1139 | error_setg(errp, "Could not open backing file: %s", |
1140 | error_get_pretty(local_err)); | |
1141 | error_free(local_err); | |
1ba4b6a5 | 1142 | goto free_exit; |
9156df12 | 1143 | } |
d80ac658 PF |
1144 | |
1145 | if (bs->backing_hd->file) { | |
1146 | pstrcpy(bs->backing_file, sizeof(bs->backing_file), | |
1147 | bs->backing_hd->file->filename); | |
1148 | } | |
1149 | ||
d34682cd KW |
1150 | /* Recalculate the BlockLimits with the backing file */ |
1151 | bdrv_refresh_limits(bs); | |
1152 | ||
1ba4b6a5 BC |
1153 | free_exit: |
1154 | g_free(backing_filename); | |
1155 | return ret; | |
9156df12 PB |
1156 | } |
1157 | ||
da557aac HR |
1158 | /* |
1159 | * Opens a disk image whose options are given as BlockdevRef in another block | |
1160 | * device's options. | |
1161 | * | |
da557aac HR |
1162 | * If allow_none is true, no image will be opened if filename is false and no |
1163 | * BlockdevRef is given. *pbs will remain unchanged and 0 will be returned. | |
1164 | * | |
1165 | * bdrev_key specifies the key for the image's BlockdevRef in the options QDict. | |
1166 | * That QDict has to be flattened; therefore, if the BlockdevRef is a QDict | |
1167 | * itself, all options starting with "${bdref_key}." are considered part of the | |
1168 | * BlockdevRef. | |
1169 | * | |
1170 | * The BlockdevRef will be removed from the options QDict. | |
f67503e5 HR |
1171 | * |
1172 | * To conform with the behavior of bdrv_open(), *pbs has to be NULL. | |
da557aac HR |
1173 | */ |
1174 | int bdrv_open_image(BlockDriverState **pbs, const char *filename, | |
1175 | QDict *options, const char *bdref_key, int flags, | |
f7d9fd8c | 1176 | bool allow_none, Error **errp) |
da557aac HR |
1177 | { |
1178 | QDict *image_options; | |
1179 | int ret; | |
1180 | char *bdref_key_dot; | |
1181 | const char *reference; | |
1182 | ||
f67503e5 HR |
1183 | assert(pbs); |
1184 | assert(*pbs == NULL); | |
1185 | ||
da557aac HR |
1186 | bdref_key_dot = g_strdup_printf("%s.", bdref_key); |
1187 | qdict_extract_subqdict(options, &image_options, bdref_key_dot); | |
1188 | g_free(bdref_key_dot); | |
1189 | ||
1190 | reference = qdict_get_try_str(options, bdref_key); | |
1191 | if (!filename && !reference && !qdict_size(image_options)) { | |
1192 | if (allow_none) { | |
1193 | ret = 0; | |
1194 | } else { | |
1195 | error_setg(errp, "A block device must be specified for \"%s\"", | |
1196 | bdref_key); | |
1197 | ret = -EINVAL; | |
1198 | } | |
1199 | goto done; | |
1200 | } | |
1201 | ||
f7d9fd8c | 1202 | ret = bdrv_open(pbs, filename, reference, image_options, flags, NULL, errp); |
da557aac HR |
1203 | |
1204 | done: | |
1205 | qdict_del(options, bdref_key); | |
1206 | return ret; | |
1207 | } | |
1208 | ||
b1e6fc08 | 1209 | void bdrv_append_temp_snapshot(BlockDriverState *bs, int flags, Error **errp) |
b998875d KW |
1210 | { |
1211 | /* TODO: extra byte is a hack to ensure MAX_PATH space on Windows. */ | |
1ba4b6a5 | 1212 | char *tmp_filename = g_malloc0(PATH_MAX + 1); |
b998875d KW |
1213 | int64_t total_size; |
1214 | BlockDriver *bdrv_qcow2; | |
1215 | QEMUOptionParameter *create_options; | |
1216 | QDict *snapshot_options; | |
1217 | BlockDriverState *bs_snapshot; | |
1218 | Error *local_err; | |
1219 | int ret; | |
1220 | ||
1221 | /* if snapshot, we create a temporary backing file and open it | |
1222 | instead of opening 'filename' directly */ | |
1223 | ||
1224 | /* Get the required size from the image */ | |
f187743a KW |
1225 | total_size = bdrv_getlength(bs); |
1226 | if (total_size < 0) { | |
1227 | error_setg_errno(errp, -total_size, "Could not get image size"); | |
1ba4b6a5 | 1228 | goto out; |
f187743a KW |
1229 | } |
1230 | total_size &= BDRV_SECTOR_MASK; | |
b998875d KW |
1231 | |
1232 | /* Create the temporary image */ | |
1ba4b6a5 | 1233 | ret = get_tmp_filename(tmp_filename, PATH_MAX + 1); |
b998875d KW |
1234 | if (ret < 0) { |
1235 | error_setg_errno(errp, -ret, "Could not get temporary filename"); | |
1ba4b6a5 | 1236 | goto out; |
b998875d KW |
1237 | } |
1238 | ||
1239 | bdrv_qcow2 = bdrv_find_format("qcow2"); | |
1240 | create_options = parse_option_parameters("", bdrv_qcow2->create_options, | |
1241 | NULL); | |
1242 | ||
1243 | set_option_parameter_int(create_options, BLOCK_OPT_SIZE, total_size); | |
1244 | ||
1245 | ret = bdrv_create(bdrv_qcow2, tmp_filename, create_options, &local_err); | |
1246 | free_option_parameters(create_options); | |
1247 | if (ret < 0) { | |
1248 | error_setg_errno(errp, -ret, "Could not create temporary overlay " | |
1249 | "'%s': %s", tmp_filename, | |
1250 | error_get_pretty(local_err)); | |
1251 | error_free(local_err); | |
1ba4b6a5 | 1252 | goto out; |
b998875d KW |
1253 | } |
1254 | ||
1255 | /* Prepare a new options QDict for the temporary file */ | |
1256 | snapshot_options = qdict_new(); | |
1257 | qdict_put(snapshot_options, "file.driver", | |
1258 | qstring_from_str("file")); | |
1259 | qdict_put(snapshot_options, "file.filename", | |
1260 | qstring_from_str(tmp_filename)); | |
1261 | ||
98522f63 | 1262 | bs_snapshot = bdrv_new("", &error_abort); |
b998875d KW |
1263 | |
1264 | ret = bdrv_open(&bs_snapshot, NULL, NULL, snapshot_options, | |
b1e6fc08 | 1265 | flags, bdrv_qcow2, &local_err); |
b998875d KW |
1266 | if (ret < 0) { |
1267 | error_propagate(errp, local_err); | |
1ba4b6a5 | 1268 | goto out; |
b998875d KW |
1269 | } |
1270 | ||
1271 | bdrv_append(bs_snapshot, bs); | |
1ba4b6a5 BC |
1272 | |
1273 | out: | |
1274 | g_free(tmp_filename); | |
b998875d KW |
1275 | } |
1276 | ||
b6ce07aa KW |
1277 | /* |
1278 | * Opens a disk image (raw, qcow2, vmdk, ...) | |
de9c0cec KW |
1279 | * |
1280 | * options is a QDict of options to pass to the block drivers, or NULL for an | |
1281 | * empty set of options. The reference to the QDict belongs to the block layer | |
1282 | * after the call (even on failure), so if the caller intends to reuse the | |
1283 | * dictionary, it needs to use QINCREF() before calling bdrv_open. | |
f67503e5 HR |
1284 | * |
1285 | * If *pbs is NULL, a new BDS will be created with a pointer to it stored there. | |
1286 | * If it is not NULL, the referenced BDS will be reused. | |
ddf5636d HR |
1287 | * |
1288 | * The reference parameter may be used to specify an existing block device which | |
1289 | * should be opened. If specified, neither options nor a filename may be given, | |
1290 | * nor can an existing BDS be reused (that is, *pbs has to be NULL). | |
b6ce07aa | 1291 | */ |
ddf5636d HR |
1292 | int bdrv_open(BlockDriverState **pbs, const char *filename, |
1293 | const char *reference, QDict *options, int flags, | |
1294 | BlockDriver *drv, Error **errp) | |
ea2384d3 | 1295 | { |
b6ce07aa | 1296 | int ret; |
f67503e5 | 1297 | BlockDriverState *file = NULL, *bs; |
74fe54f2 | 1298 | const char *drvname; |
34b5d2c6 | 1299 | Error *local_err = NULL; |
b1e6fc08 | 1300 | int snapshot_flags = 0; |
712e7874 | 1301 | |
f67503e5 HR |
1302 | assert(pbs); |
1303 | ||
ddf5636d HR |
1304 | if (reference) { |
1305 | bool options_non_empty = options ? qdict_size(options) : false; | |
1306 | QDECREF(options); | |
1307 | ||
1308 | if (*pbs) { | |
1309 | error_setg(errp, "Cannot reuse an existing BDS when referencing " | |
1310 | "another block device"); | |
1311 | return -EINVAL; | |
1312 | } | |
1313 | ||
1314 | if (filename || options_non_empty) { | |
1315 | error_setg(errp, "Cannot reference an existing block device with " | |
1316 | "additional options or a new filename"); | |
1317 | return -EINVAL; | |
1318 | } | |
1319 | ||
1320 | bs = bdrv_lookup_bs(reference, reference, errp); | |
1321 | if (!bs) { | |
1322 | return -ENODEV; | |
1323 | } | |
1324 | bdrv_ref(bs); | |
1325 | *pbs = bs; | |
1326 | return 0; | |
1327 | } | |
1328 | ||
f67503e5 HR |
1329 | if (*pbs) { |
1330 | bs = *pbs; | |
1331 | } else { | |
98522f63 | 1332 | bs = bdrv_new("", &error_abort); |
f67503e5 HR |
1333 | } |
1334 | ||
de9c0cec KW |
1335 | /* NULL means an empty set of options */ |
1336 | if (options == NULL) { | |
1337 | options = qdict_new(); | |
1338 | } | |
1339 | ||
1340 | bs->options = options; | |
b6ad491a | 1341 | options = qdict_clone_shallow(options); |
de9c0cec | 1342 | |
5469a2a6 HR |
1343 | if (flags & BDRV_O_PROTOCOL) { |
1344 | assert(!drv); | |
5acd9d81 | 1345 | ret = bdrv_file_open(bs, filename, &options, flags & ~BDRV_O_PROTOCOL, |
5469a2a6 | 1346 | &local_err); |
5469a2a6 | 1347 | if (!ret) { |
eb909c7f | 1348 | drv = bs->drv; |
5acd9d81 | 1349 | goto done; |
5469a2a6 HR |
1350 | } else if (bs->drv) { |
1351 | goto close_and_fail; | |
1352 | } else { | |
1353 | goto fail; | |
1354 | } | |
1355 | } | |
1356 | ||
f500a6d3 KW |
1357 | /* Open image file without format layer */ |
1358 | if (flags & BDRV_O_RDWR) { | |
1359 | flags |= BDRV_O_ALLOW_RDWR; | |
1360 | } | |
b1e6fc08 KW |
1361 | if (flags & BDRV_O_SNAPSHOT) { |
1362 | snapshot_flags = bdrv_temp_snapshot_flags(flags); | |
1363 | flags = bdrv_backing_flags(flags); | |
1364 | } | |
f500a6d3 | 1365 | |
f67503e5 | 1366 | assert(file == NULL); |
054963f8 | 1367 | ret = bdrv_open_image(&file, filename, options, "file", |
0b50cc88 KW |
1368 | bdrv_inherited_flags(flags), |
1369 | true, &local_err); | |
054963f8 | 1370 | if (ret < 0) { |
8bfea15d | 1371 | goto fail; |
f500a6d3 KW |
1372 | } |
1373 | ||
b6ce07aa | 1374 | /* Find the right image format driver */ |
74fe54f2 KW |
1375 | drvname = qdict_get_try_str(options, "driver"); |
1376 | if (drvname) { | |
8f94a6e4 | 1377 | drv = bdrv_find_format(drvname); |
74fe54f2 | 1378 | qdict_del(options, "driver"); |
06d22aa3 KW |
1379 | if (!drv) { |
1380 | error_setg(errp, "Invalid driver: '%s'", drvname); | |
1381 | ret = -EINVAL; | |
8bfea15d | 1382 | goto fail; |
06d22aa3 | 1383 | } |
74fe54f2 KW |
1384 | } |
1385 | ||
6db95603 | 1386 | if (!drv) { |
2a05cbe4 HR |
1387 | if (file) { |
1388 | ret = find_image_format(file, filename, &drv, &local_err); | |
1389 | } else { | |
1390 | error_setg(errp, "Must specify either driver or file"); | |
1391 | ret = -EINVAL; | |
8bfea15d | 1392 | goto fail; |
2a05cbe4 | 1393 | } |
51d7c00c | 1394 | } |
6987307c | 1395 | |
51d7c00c | 1396 | if (!drv) { |
8bfea15d | 1397 | goto fail; |
ea2384d3 | 1398 | } |
b6ce07aa KW |
1399 | |
1400 | /* Open the image */ | |
34b5d2c6 | 1401 | ret = bdrv_open_common(bs, file, options, flags, drv, &local_err); |
b6ce07aa | 1402 | if (ret < 0) { |
8bfea15d | 1403 | goto fail; |
6987307c CH |
1404 | } |
1405 | ||
2a05cbe4 | 1406 | if (file && (bs->file != file)) { |
4f6fd349 | 1407 | bdrv_unref(file); |
f500a6d3 KW |
1408 | file = NULL; |
1409 | } | |
1410 | ||
b6ce07aa | 1411 | /* If there is a backing file, use it */ |
9156df12 | 1412 | if ((flags & BDRV_O_NO_BACKING) == 0) { |
31ca6d07 KW |
1413 | QDict *backing_options; |
1414 | ||
5726d872 | 1415 | qdict_extract_subqdict(options, &backing_options, "backing."); |
34b5d2c6 | 1416 | ret = bdrv_open_backing_file(bs, backing_options, &local_err); |
b6ce07aa | 1417 | if (ret < 0) { |
b6ad491a | 1418 | goto close_and_fail; |
b6ce07aa | 1419 | } |
b6ce07aa KW |
1420 | } |
1421 | ||
b998875d KW |
1422 | /* For snapshot=on, create a temporary qcow2 overlay. bs points to the |
1423 | * temporary snapshot afterwards. */ | |
b1e6fc08 KW |
1424 | if (snapshot_flags) { |
1425 | bdrv_append_temp_snapshot(bs, snapshot_flags, &local_err); | |
b998875d KW |
1426 | if (local_err) { |
1427 | error_propagate(errp, local_err); | |
1428 | goto close_and_fail; | |
1429 | } | |
1430 | } | |
1431 | ||
1432 | ||
5acd9d81 | 1433 | done: |
b6ad491a | 1434 | /* Check if any unknown options were used */ |
5acd9d81 | 1435 | if (options && (qdict_size(options) != 0)) { |
b6ad491a | 1436 | const QDictEntry *entry = qdict_first(options); |
5acd9d81 HR |
1437 | if (flags & BDRV_O_PROTOCOL) { |
1438 | error_setg(errp, "Block protocol '%s' doesn't support the option " | |
1439 | "'%s'", drv->format_name, entry->key); | |
1440 | } else { | |
1441 | error_setg(errp, "Block format '%s' used by device '%s' doesn't " | |
1442 | "support the option '%s'", drv->format_name, | |
1443 | bs->device_name, entry->key); | |
1444 | } | |
b6ad491a KW |
1445 | |
1446 | ret = -EINVAL; | |
1447 | goto close_and_fail; | |
1448 | } | |
b6ad491a | 1449 | |
b6ce07aa | 1450 | if (!bdrv_key_required(bs)) { |
7d4b4ba5 | 1451 | bdrv_dev_change_media_cb(bs, true); |
c3adb58f MA |
1452 | } else if (!runstate_check(RUN_STATE_PRELAUNCH) |
1453 | && !runstate_check(RUN_STATE_INMIGRATE) | |
1454 | && !runstate_check(RUN_STATE_PAUSED)) { /* HACK */ | |
1455 | error_setg(errp, | |
1456 | "Guest must be stopped for opening of encrypted image"); | |
1457 | ret = -EBUSY; | |
1458 | goto close_and_fail; | |
b6ce07aa KW |
1459 | } |
1460 | ||
c3adb58f | 1461 | QDECREF(options); |
f67503e5 | 1462 | *pbs = bs; |
b6ce07aa KW |
1463 | return 0; |
1464 | ||
8bfea15d | 1465 | fail: |
f500a6d3 | 1466 | if (file != NULL) { |
4f6fd349 | 1467 | bdrv_unref(file); |
f500a6d3 | 1468 | } |
de9c0cec | 1469 | QDECREF(bs->options); |
b6ad491a | 1470 | QDECREF(options); |
de9c0cec | 1471 | bs->options = NULL; |
f67503e5 HR |
1472 | if (!*pbs) { |
1473 | /* If *pbs is NULL, a new BDS has been created in this function and | |
1474 | needs to be freed now. Otherwise, it does not need to be closed, | |
1475 | since it has not really been opened yet. */ | |
1476 | bdrv_unref(bs); | |
1477 | } | |
84d18f06 | 1478 | if (local_err) { |
34b5d2c6 HR |
1479 | error_propagate(errp, local_err); |
1480 | } | |
b6ad491a | 1481 | return ret; |
de9c0cec | 1482 | |
b6ad491a | 1483 | close_and_fail: |
f67503e5 HR |
1484 | /* See fail path, but now the BDS has to be always closed */ |
1485 | if (*pbs) { | |
1486 | bdrv_close(bs); | |
1487 | } else { | |
1488 | bdrv_unref(bs); | |
1489 | } | |
b6ad491a | 1490 | QDECREF(options); |
84d18f06 | 1491 | if (local_err) { |
34b5d2c6 HR |
1492 | error_propagate(errp, local_err); |
1493 | } | |
b6ce07aa KW |
1494 | return ret; |
1495 | } | |
1496 | ||
e971aa12 JC |
1497 | typedef struct BlockReopenQueueEntry { |
1498 | bool prepared; | |
1499 | BDRVReopenState state; | |
1500 | QSIMPLEQ_ENTRY(BlockReopenQueueEntry) entry; | |
1501 | } BlockReopenQueueEntry; | |
1502 | ||
1503 | /* | |
1504 | * Adds a BlockDriverState to a simple queue for an atomic, transactional | |
1505 | * reopen of multiple devices. | |
1506 | * | |
1507 | * bs_queue can either be an existing BlockReopenQueue that has had QSIMPLE_INIT | |
1508 | * already performed, or alternatively may be NULL a new BlockReopenQueue will | |
1509 | * be created and initialized. This newly created BlockReopenQueue should be | |
1510 | * passed back in for subsequent calls that are intended to be of the same | |
1511 | * atomic 'set'. | |
1512 | * | |
1513 | * bs is the BlockDriverState to add to the reopen queue. | |
1514 | * | |
1515 | * flags contains the open flags for the associated bs | |
1516 | * | |
1517 | * returns a pointer to bs_queue, which is either the newly allocated | |
1518 | * bs_queue, or the existing bs_queue being used. | |
1519 | * | |
1520 | */ | |
1521 | BlockReopenQueue *bdrv_reopen_queue(BlockReopenQueue *bs_queue, | |
1522 | BlockDriverState *bs, int flags) | |
1523 | { | |
1524 | assert(bs != NULL); | |
1525 | ||
1526 | BlockReopenQueueEntry *bs_entry; | |
1527 | if (bs_queue == NULL) { | |
1528 | bs_queue = g_new0(BlockReopenQueue, 1); | |
1529 | QSIMPLEQ_INIT(bs_queue); | |
1530 | } | |
1531 | ||
f1f25a2e KW |
1532 | /* bdrv_open() masks this flag out */ |
1533 | flags &= ~BDRV_O_PROTOCOL; | |
1534 | ||
e971aa12 | 1535 | if (bs->file) { |
f1f25a2e | 1536 | bdrv_reopen_queue(bs_queue, bs->file, bdrv_inherited_flags(flags)); |
e971aa12 JC |
1537 | } |
1538 | ||
1539 | bs_entry = g_new0(BlockReopenQueueEntry, 1); | |
1540 | QSIMPLEQ_INSERT_TAIL(bs_queue, bs_entry, entry); | |
1541 | ||
1542 | bs_entry->state.bs = bs; | |
1543 | bs_entry->state.flags = flags; | |
1544 | ||
1545 | return bs_queue; | |
1546 | } | |
1547 | ||
1548 | /* | |
1549 | * Reopen multiple BlockDriverStates atomically & transactionally. | |
1550 | * | |
1551 | * The queue passed in (bs_queue) must have been built up previous | |
1552 | * via bdrv_reopen_queue(). | |
1553 | * | |
1554 | * Reopens all BDS specified in the queue, with the appropriate | |
1555 | * flags. All devices are prepared for reopen, and failure of any | |
1556 | * device will cause all device changes to be abandonded, and intermediate | |
1557 | * data cleaned up. | |
1558 | * | |
1559 | * If all devices prepare successfully, then the changes are committed | |
1560 | * to all devices. | |
1561 | * | |
1562 | */ | |
1563 | int bdrv_reopen_multiple(BlockReopenQueue *bs_queue, Error **errp) | |
1564 | { | |
1565 | int ret = -1; | |
1566 | BlockReopenQueueEntry *bs_entry, *next; | |
1567 | Error *local_err = NULL; | |
1568 | ||
1569 | assert(bs_queue != NULL); | |
1570 | ||
1571 | bdrv_drain_all(); | |
1572 | ||
1573 | QSIMPLEQ_FOREACH(bs_entry, bs_queue, entry) { | |
1574 | if (bdrv_reopen_prepare(&bs_entry->state, bs_queue, &local_err)) { | |
1575 | error_propagate(errp, local_err); | |
1576 | goto cleanup; | |
1577 | } | |
1578 | bs_entry->prepared = true; | |
1579 | } | |
1580 | ||
1581 | /* If we reach this point, we have success and just need to apply the | |
1582 | * changes | |
1583 | */ | |
1584 | QSIMPLEQ_FOREACH(bs_entry, bs_queue, entry) { | |
1585 | bdrv_reopen_commit(&bs_entry->state); | |
1586 | } | |
1587 | ||
1588 | ret = 0; | |
1589 | ||
1590 | cleanup: | |
1591 | QSIMPLEQ_FOREACH_SAFE(bs_entry, bs_queue, entry, next) { | |
1592 | if (ret && bs_entry->prepared) { | |
1593 | bdrv_reopen_abort(&bs_entry->state); | |
1594 | } | |
1595 | g_free(bs_entry); | |
1596 | } | |
1597 | g_free(bs_queue); | |
1598 | return ret; | |
1599 | } | |
1600 | ||
1601 | ||
1602 | /* Reopen a single BlockDriverState with the specified flags. */ | |
1603 | int bdrv_reopen(BlockDriverState *bs, int bdrv_flags, Error **errp) | |
1604 | { | |
1605 | int ret = -1; | |
1606 | Error *local_err = NULL; | |
1607 | BlockReopenQueue *queue = bdrv_reopen_queue(NULL, bs, bdrv_flags); | |
1608 | ||
1609 | ret = bdrv_reopen_multiple(queue, &local_err); | |
1610 | if (local_err != NULL) { | |
1611 | error_propagate(errp, local_err); | |
1612 | } | |
1613 | return ret; | |
1614 | } | |
1615 | ||
1616 | ||
1617 | /* | |
1618 | * Prepares a BlockDriverState for reopen. All changes are staged in the | |
1619 | * 'opaque' field of the BDRVReopenState, which is used and allocated by | |
1620 | * the block driver layer .bdrv_reopen_prepare() | |
1621 | * | |
1622 | * bs is the BlockDriverState to reopen | |
1623 | * flags are the new open flags | |
1624 | * queue is the reopen queue | |
1625 | * | |
1626 | * Returns 0 on success, non-zero on error. On error errp will be set | |
1627 | * as well. | |
1628 | * | |
1629 | * On failure, bdrv_reopen_abort() will be called to clean up any data. | |
1630 | * It is the responsibility of the caller to then call the abort() or | |
1631 | * commit() for any other BDS that have been left in a prepare() state | |
1632 | * | |
1633 | */ | |
1634 | int bdrv_reopen_prepare(BDRVReopenState *reopen_state, BlockReopenQueue *queue, | |
1635 | Error **errp) | |
1636 | { | |
1637 | int ret = -1; | |
1638 | Error *local_err = NULL; | |
1639 | BlockDriver *drv; | |
1640 | ||
1641 | assert(reopen_state != NULL); | |
1642 | assert(reopen_state->bs->drv != NULL); | |
1643 | drv = reopen_state->bs->drv; | |
1644 | ||
1645 | /* if we are to stay read-only, do not allow permission change | |
1646 | * to r/w */ | |
1647 | if (!(reopen_state->bs->open_flags & BDRV_O_ALLOW_RDWR) && | |
1648 | reopen_state->flags & BDRV_O_RDWR) { | |
1649 | error_set(errp, QERR_DEVICE_IS_READ_ONLY, | |
1650 | reopen_state->bs->device_name); | |
1651 | goto error; | |
1652 | } | |
1653 | ||
1654 | ||
1655 | ret = bdrv_flush(reopen_state->bs); | |
1656 | if (ret) { | |
1657 | error_set(errp, ERROR_CLASS_GENERIC_ERROR, "Error (%s) flushing drive", | |
1658 | strerror(-ret)); | |
1659 | goto error; | |
1660 | } | |
1661 | ||
1662 | if (drv->bdrv_reopen_prepare) { | |
1663 | ret = drv->bdrv_reopen_prepare(reopen_state, queue, &local_err); | |
1664 | if (ret) { | |
1665 | if (local_err != NULL) { | |
1666 | error_propagate(errp, local_err); | |
1667 | } else { | |
d8b6895f LC |
1668 | error_setg(errp, "failed while preparing to reopen image '%s'", |
1669 | reopen_state->bs->filename); | |
e971aa12 JC |
1670 | } |
1671 | goto error; | |
1672 | } | |
1673 | } else { | |
1674 | /* It is currently mandatory to have a bdrv_reopen_prepare() | |
1675 | * handler for each supported drv. */ | |
1676 | error_set(errp, QERR_BLOCK_FORMAT_FEATURE_NOT_SUPPORTED, | |
1677 | drv->format_name, reopen_state->bs->device_name, | |
1678 | "reopening of file"); | |
1679 | ret = -1; | |
1680 | goto error; | |
1681 | } | |
1682 | ||
1683 | ret = 0; | |
1684 | ||
1685 | error: | |
1686 | return ret; | |
1687 | } | |
1688 | ||
1689 | /* | |
1690 | * Takes the staged changes for the reopen from bdrv_reopen_prepare(), and | |
1691 | * makes them final by swapping the staging BlockDriverState contents into | |
1692 | * the active BlockDriverState contents. | |
1693 | */ | |
1694 | void bdrv_reopen_commit(BDRVReopenState *reopen_state) | |
1695 | { | |
1696 | BlockDriver *drv; | |
1697 | ||
1698 | assert(reopen_state != NULL); | |
1699 | drv = reopen_state->bs->drv; | |
1700 | assert(drv != NULL); | |
1701 | ||
1702 | /* If there are any driver level actions to take */ | |
1703 | if (drv->bdrv_reopen_commit) { | |
1704 | drv->bdrv_reopen_commit(reopen_state); | |
1705 | } | |
1706 | ||
1707 | /* set BDS specific flags now */ | |
1708 | reopen_state->bs->open_flags = reopen_state->flags; | |
1709 | reopen_state->bs->enable_write_cache = !!(reopen_state->flags & | |
1710 | BDRV_O_CACHE_WB); | |
1711 | reopen_state->bs->read_only = !(reopen_state->flags & BDRV_O_RDWR); | |
355ef4ac KW |
1712 | |
1713 | bdrv_refresh_limits(reopen_state->bs); | |
e971aa12 JC |
1714 | } |
1715 | ||
1716 | /* | |
1717 | * Abort the reopen, and delete and free the staged changes in | |
1718 | * reopen_state | |
1719 | */ | |
1720 | void bdrv_reopen_abort(BDRVReopenState *reopen_state) | |
1721 | { | |
1722 | BlockDriver *drv; | |
1723 | ||
1724 | assert(reopen_state != NULL); | |
1725 | drv = reopen_state->bs->drv; | |
1726 | assert(drv != NULL); | |
1727 | ||
1728 | if (drv->bdrv_reopen_abort) { | |
1729 | drv->bdrv_reopen_abort(reopen_state); | |
1730 | } | |
1731 | } | |
1732 | ||
1733 | ||
fc01f7e7 FB |
1734 | void bdrv_close(BlockDriverState *bs) |
1735 | { | |
3cbc002c PB |
1736 | if (bs->job) { |
1737 | block_job_cancel_sync(bs->job); | |
1738 | } | |
58fda173 SH |
1739 | bdrv_drain_all(); /* complete I/O */ |
1740 | bdrv_flush(bs); | |
1741 | bdrv_drain_all(); /* in case flush left pending I/O */ | |
d7d512f6 | 1742 | notifier_list_notify(&bs->close_notifiers, bs); |
7094f12f | 1743 | |
3cbc002c | 1744 | if (bs->drv) { |
557df6ac | 1745 | if (bs->backing_hd) { |
4f6fd349 | 1746 | bdrv_unref(bs->backing_hd); |
557df6ac SH |
1747 | bs->backing_hd = NULL; |
1748 | } | |
ea2384d3 | 1749 | bs->drv->bdrv_close(bs); |
7267c094 | 1750 | g_free(bs->opaque); |
ea2384d3 FB |
1751 | bs->opaque = NULL; |
1752 | bs->drv = NULL; | |
53fec9d3 | 1753 | bs->copy_on_read = 0; |
a275fa42 PB |
1754 | bs->backing_file[0] = '\0'; |
1755 | bs->backing_format[0] = '\0'; | |
6405875c PB |
1756 | bs->total_sectors = 0; |
1757 | bs->encrypted = 0; | |
1758 | bs->valid_key = 0; | |
1759 | bs->sg = 0; | |
1760 | bs->growable = 0; | |
0d51b4de | 1761 | bs->zero_beyond_eof = false; |
de9c0cec KW |
1762 | QDECREF(bs->options); |
1763 | bs->options = NULL; | |
b338082b | 1764 | |
66f82cee | 1765 | if (bs->file != NULL) { |
4f6fd349 | 1766 | bdrv_unref(bs->file); |
0ac9377d | 1767 | bs->file = NULL; |
66f82cee | 1768 | } |
b338082b | 1769 | } |
98f90dba | 1770 | |
9ca11154 PH |
1771 | bdrv_dev_change_media_cb(bs, false); |
1772 | ||
98f90dba ZYW |
1773 | /*throttling disk I/O limits*/ |
1774 | if (bs->io_limits_enabled) { | |
1775 | bdrv_io_limits_disable(bs); | |
1776 | } | |
b338082b FB |
1777 | } |
1778 | ||
2bc93fed MK |
1779 | void bdrv_close_all(void) |
1780 | { | |
1781 | BlockDriverState *bs; | |
1782 | ||
dc364f4c | 1783 | QTAILQ_FOREACH(bs, &bdrv_states, device_list) { |
2bc93fed MK |
1784 | bdrv_close(bs); |
1785 | } | |
1786 | } | |
1787 | ||
88266f5a SH |
1788 | /* Check if any requests are in-flight (including throttled requests) */ |
1789 | static bool bdrv_requests_pending(BlockDriverState *bs) | |
1790 | { | |
1791 | if (!QLIST_EMPTY(&bs->tracked_requests)) { | |
1792 | return true; | |
1793 | } | |
cc0681c4 BC |
1794 | if (!qemu_co_queue_empty(&bs->throttled_reqs[0])) { |
1795 | return true; | |
1796 | } | |
1797 | if (!qemu_co_queue_empty(&bs->throttled_reqs[1])) { | |
88266f5a SH |
1798 | return true; |
1799 | } | |
1800 | if (bs->file && bdrv_requests_pending(bs->file)) { | |
1801 | return true; | |
1802 | } | |
1803 | if (bs->backing_hd && bdrv_requests_pending(bs->backing_hd)) { | |
1804 | return true; | |
1805 | } | |
1806 | return false; | |
1807 | } | |
1808 | ||
1809 | static bool bdrv_requests_pending_all(void) | |
1810 | { | |
1811 | BlockDriverState *bs; | |
dc364f4c | 1812 | QTAILQ_FOREACH(bs, &bdrv_states, device_list) { |
88266f5a SH |
1813 | if (bdrv_requests_pending(bs)) { |
1814 | return true; | |
1815 | } | |
1816 | } | |
1817 | return false; | |
1818 | } | |
1819 | ||
922453bc SH |
1820 | /* |
1821 | * Wait for pending requests to complete across all BlockDriverStates | |
1822 | * | |
1823 | * This function does not flush data to disk, use bdrv_flush_all() for that | |
1824 | * after calling this function. | |
4c355d53 ZYW |
1825 | * |
1826 | * Note that completion of an asynchronous I/O operation can trigger any | |
1827 | * number of other I/O operations on other devices---for example a coroutine | |
1828 | * can be arbitrarily complex and a constant flow of I/O can come until the | |
1829 | * coroutine is complete. Because of this, it is not possible to have a | |
1830 | * function to drain a single device's I/O queue. | |
922453bc SH |
1831 | */ |
1832 | void bdrv_drain_all(void) | |
1833 | { | |
88266f5a SH |
1834 | /* Always run first iteration so any pending completion BHs run */ |
1835 | bool busy = true; | |
922453bc SH |
1836 | BlockDriverState *bs; |
1837 | ||
88266f5a | 1838 | while (busy) { |
dc364f4c | 1839 | QTAILQ_FOREACH(bs, &bdrv_states, device_list) { |
0b06ef3b | 1840 | bdrv_start_throttled_reqs(bs); |
4c355d53 | 1841 | } |
922453bc | 1842 | |
88266f5a SH |
1843 | busy = bdrv_requests_pending_all(); |
1844 | busy |= aio_poll(qemu_get_aio_context(), busy); | |
922453bc SH |
1845 | } |
1846 | } | |
1847 | ||
dc364f4c BC |
1848 | /* make a BlockDriverState anonymous by removing from bdrv_state and |
1849 | * graph_bdrv_state list. | |
d22b2f41 RH |
1850 | Also, NULL terminate the device_name to prevent double remove */ |
1851 | void bdrv_make_anon(BlockDriverState *bs) | |
1852 | { | |
1853 | if (bs->device_name[0] != '\0') { | |
dc364f4c | 1854 | QTAILQ_REMOVE(&bdrv_states, bs, device_list); |
d22b2f41 RH |
1855 | } |
1856 | bs->device_name[0] = '\0'; | |
dc364f4c BC |
1857 | if (bs->node_name[0] != '\0') { |
1858 | QTAILQ_REMOVE(&graph_bdrv_states, bs, node_list); | |
1859 | } | |
1860 | bs->node_name[0] = '\0'; | |
d22b2f41 RH |
1861 | } |
1862 | ||
e023b2e2 PB |
1863 | static void bdrv_rebind(BlockDriverState *bs) |
1864 | { | |
1865 | if (bs->drv && bs->drv->bdrv_rebind) { | |
1866 | bs->drv->bdrv_rebind(bs); | |
1867 | } | |
1868 | } | |
1869 | ||
4ddc07ca PB |
1870 | static void bdrv_move_feature_fields(BlockDriverState *bs_dest, |
1871 | BlockDriverState *bs_src) | |
8802d1fd | 1872 | { |
4ddc07ca | 1873 | /* move some fields that need to stay attached to the device */ |
8802d1fd JC |
1874 | |
1875 | /* dev info */ | |
4ddc07ca PB |
1876 | bs_dest->dev_ops = bs_src->dev_ops; |
1877 | bs_dest->dev_opaque = bs_src->dev_opaque; | |
1878 | bs_dest->dev = bs_src->dev; | |
1b7fd729 | 1879 | bs_dest->guest_block_size = bs_src->guest_block_size; |
4ddc07ca | 1880 | bs_dest->copy_on_read = bs_src->copy_on_read; |
8802d1fd | 1881 | |
4ddc07ca | 1882 | bs_dest->enable_write_cache = bs_src->enable_write_cache; |
c4a248a1 | 1883 | |
cc0681c4 BC |
1884 | /* i/o throttled req */ |
1885 | memcpy(&bs_dest->throttle_state, | |
1886 | &bs_src->throttle_state, | |
1887 | sizeof(ThrottleState)); | |
1888 | bs_dest->throttled_reqs[0] = bs_src->throttled_reqs[0]; | |
1889 | bs_dest->throttled_reqs[1] = bs_src->throttled_reqs[1]; | |
4ddc07ca | 1890 | bs_dest->io_limits_enabled = bs_src->io_limits_enabled; |
8802d1fd | 1891 | |
8802d1fd | 1892 | /* r/w error */ |
4ddc07ca PB |
1893 | bs_dest->on_read_error = bs_src->on_read_error; |
1894 | bs_dest->on_write_error = bs_src->on_write_error; | |
8802d1fd JC |
1895 | |
1896 | /* i/o status */ | |
4ddc07ca PB |
1897 | bs_dest->iostatus_enabled = bs_src->iostatus_enabled; |
1898 | bs_dest->iostatus = bs_src->iostatus; | |
8802d1fd | 1899 | |
a9fc4408 | 1900 | /* dirty bitmap */ |
e4654d2d | 1901 | bs_dest->dirty_bitmaps = bs_src->dirty_bitmaps; |
a9fc4408 | 1902 | |
9fcb0251 FZ |
1903 | /* reference count */ |
1904 | bs_dest->refcnt = bs_src->refcnt; | |
1905 | ||
a9fc4408 | 1906 | /* job */ |
4ddc07ca PB |
1907 | bs_dest->in_use = bs_src->in_use; |
1908 | bs_dest->job = bs_src->job; | |
a9fc4408 | 1909 | |
8802d1fd | 1910 | /* keep the same entry in bdrv_states */ |
4ddc07ca PB |
1911 | pstrcpy(bs_dest->device_name, sizeof(bs_dest->device_name), |
1912 | bs_src->device_name); | |
dc364f4c | 1913 | bs_dest->device_list = bs_src->device_list; |
4ddc07ca | 1914 | } |
8802d1fd | 1915 | |
4ddc07ca PB |
1916 | /* |
1917 | * Swap bs contents for two image chains while they are live, | |
1918 | * while keeping required fields on the BlockDriverState that is | |
1919 | * actually attached to a device. | |
1920 | * | |
1921 | * This will modify the BlockDriverState fields, and swap contents | |
1922 | * between bs_new and bs_old. Both bs_new and bs_old are modified. | |
1923 | * | |
1924 | * bs_new is required to be anonymous. | |
1925 | * | |
1926 | * This function does not create any image files. | |
1927 | */ | |
1928 | void bdrv_swap(BlockDriverState *bs_new, BlockDriverState *bs_old) | |
1929 | { | |
1930 | BlockDriverState tmp; | |
f6801b83 | 1931 | |
90ce8a06 BC |
1932 | /* The code needs to swap the node_name but simply swapping node_list won't |
1933 | * work so first remove the nodes from the graph list, do the swap then | |
1934 | * insert them back if needed. | |
1935 | */ | |
1936 | if (bs_new->node_name[0] != '\0') { | |
1937 | QTAILQ_REMOVE(&graph_bdrv_states, bs_new, node_list); | |
1938 | } | |
1939 | if (bs_old->node_name[0] != '\0') { | |
1940 | QTAILQ_REMOVE(&graph_bdrv_states, bs_old, node_list); | |
1941 | } | |
1942 | ||
4ddc07ca PB |
1943 | /* bs_new must be anonymous and shouldn't have anything fancy enabled */ |
1944 | assert(bs_new->device_name[0] == '\0'); | |
e4654d2d | 1945 | assert(QLIST_EMPTY(&bs_new->dirty_bitmaps)); |
4ddc07ca PB |
1946 | assert(bs_new->job == NULL); |
1947 | assert(bs_new->dev == NULL); | |
1948 | assert(bs_new->in_use == 0); | |
1949 | assert(bs_new->io_limits_enabled == false); | |
cc0681c4 | 1950 | assert(!throttle_have_timer(&bs_new->throttle_state)); |
8802d1fd | 1951 | |
4ddc07ca PB |
1952 | tmp = *bs_new; |
1953 | *bs_new = *bs_old; | |
1954 | *bs_old = tmp; | |
a9fc4408 | 1955 | |
4ddc07ca PB |
1956 | /* there are some fields that should not be swapped, move them back */ |
1957 | bdrv_move_feature_fields(&tmp, bs_old); | |
1958 | bdrv_move_feature_fields(bs_old, bs_new); | |
1959 | bdrv_move_feature_fields(bs_new, &tmp); | |
8802d1fd | 1960 | |
4ddc07ca PB |
1961 | /* bs_new shouldn't be in bdrv_states even after the swap! */ |
1962 | assert(bs_new->device_name[0] == '\0'); | |
1963 | ||
1964 | /* Check a few fields that should remain attached to the device */ | |
1965 | assert(bs_new->dev == NULL); | |
1966 | assert(bs_new->job == NULL); | |
1967 | assert(bs_new->in_use == 0); | |
1968 | assert(bs_new->io_limits_enabled == false); | |
cc0681c4 | 1969 | assert(!throttle_have_timer(&bs_new->throttle_state)); |
e023b2e2 | 1970 | |
90ce8a06 BC |
1971 | /* insert the nodes back into the graph node list if needed */ |
1972 | if (bs_new->node_name[0] != '\0') { | |
1973 | QTAILQ_INSERT_TAIL(&graph_bdrv_states, bs_new, node_list); | |
1974 | } | |
1975 | if (bs_old->node_name[0] != '\0') { | |
1976 | QTAILQ_INSERT_TAIL(&graph_bdrv_states, bs_old, node_list); | |
1977 | } | |
1978 | ||
e023b2e2 | 1979 | bdrv_rebind(bs_new); |
4ddc07ca PB |
1980 | bdrv_rebind(bs_old); |
1981 | } | |
1982 | ||
1983 | /* | |
1984 | * Add new bs contents at the top of an image chain while the chain is | |
1985 | * live, while keeping required fields on the top layer. | |
1986 | * | |
1987 | * This will modify the BlockDriverState fields, and swap contents | |
1988 | * between bs_new and bs_top. Both bs_new and bs_top are modified. | |
1989 | * | |
1990 | * bs_new is required to be anonymous. | |
1991 | * | |
1992 | * This function does not create any image files. | |
1993 | */ | |
1994 | void bdrv_append(BlockDriverState *bs_new, BlockDriverState *bs_top) | |
1995 | { | |
1996 | bdrv_swap(bs_new, bs_top); | |
1997 | ||
1998 | /* The contents of 'tmp' will become bs_top, as we are | |
1999 | * swapping bs_new and bs_top contents. */ | |
2000 | bs_top->backing_hd = bs_new; | |
2001 | bs_top->open_flags &= ~BDRV_O_NO_BACKING; | |
2002 | pstrcpy(bs_top->backing_file, sizeof(bs_top->backing_file), | |
2003 | bs_new->filename); | |
2004 | pstrcpy(bs_top->backing_format, sizeof(bs_top->backing_format), | |
2005 | bs_new->drv ? bs_new->drv->format_name : ""); | |
8802d1fd JC |
2006 | } |
2007 | ||
4f6fd349 | 2008 | static void bdrv_delete(BlockDriverState *bs) |
b338082b | 2009 | { |
fa879d62 | 2010 | assert(!bs->dev); |
3e914655 PB |
2011 | assert(!bs->job); |
2012 | assert(!bs->in_use); | |
4f6fd349 | 2013 | assert(!bs->refcnt); |
e4654d2d | 2014 | assert(QLIST_EMPTY(&bs->dirty_bitmaps)); |
18846dee | 2015 | |
e1b5c52e SH |
2016 | bdrv_close(bs); |
2017 | ||
1b7bdbc1 | 2018 | /* remove from list, if necessary */ |
d22b2f41 | 2019 | bdrv_make_anon(bs); |
34c6f050 | 2020 | |
7267c094 | 2021 | g_free(bs); |
fc01f7e7 FB |
2022 | } |
2023 | ||
fa879d62 MA |
2024 | int bdrv_attach_dev(BlockDriverState *bs, void *dev) |
2025 | /* TODO change to DeviceState *dev when all users are qdevified */ | |
18846dee | 2026 | { |
fa879d62 | 2027 | if (bs->dev) { |
18846dee MA |
2028 | return -EBUSY; |
2029 | } | |
fa879d62 | 2030 | bs->dev = dev; |
28a7282a | 2031 | bdrv_iostatus_reset(bs); |
18846dee MA |
2032 | return 0; |
2033 | } | |
2034 | ||
fa879d62 MA |
2035 | /* TODO qdevified devices don't use this, remove when devices are qdevified */ |
2036 | void bdrv_attach_dev_nofail(BlockDriverState *bs, void *dev) | |
18846dee | 2037 | { |
fa879d62 MA |
2038 | if (bdrv_attach_dev(bs, dev) < 0) { |
2039 | abort(); | |
2040 | } | |
2041 | } | |
2042 | ||
2043 | void bdrv_detach_dev(BlockDriverState *bs, void *dev) | |
2044 | /* TODO change to DeviceState *dev when all users are qdevified */ | |
2045 | { | |
2046 | assert(bs->dev == dev); | |
2047 | bs->dev = NULL; | |
0e49de52 MA |
2048 | bs->dev_ops = NULL; |
2049 | bs->dev_opaque = NULL; | |
1b7fd729 | 2050 | bs->guest_block_size = 512; |
18846dee MA |
2051 | } |
2052 | ||
fa879d62 MA |
2053 | /* TODO change to return DeviceState * when all users are qdevified */ |
2054 | void *bdrv_get_attached_dev(BlockDriverState *bs) | |
18846dee | 2055 | { |
fa879d62 | 2056 | return bs->dev; |
18846dee MA |
2057 | } |
2058 | ||
0e49de52 MA |
2059 | void bdrv_set_dev_ops(BlockDriverState *bs, const BlockDevOps *ops, |
2060 | void *opaque) | |
2061 | { | |
2062 | bs->dev_ops = ops; | |
2063 | bs->dev_opaque = opaque; | |
2064 | } | |
2065 | ||
32c81a4a PB |
2066 | void bdrv_emit_qmp_error_event(const BlockDriverState *bdrv, |
2067 | enum MonitorEvent ev, | |
2068 | BlockErrorAction action, bool is_read) | |
329c0a48 LC |
2069 | { |
2070 | QObject *data; | |
2071 | const char *action_str; | |
2072 | ||
2073 | switch (action) { | |
2074 | case BDRV_ACTION_REPORT: | |
2075 | action_str = "report"; | |
2076 | break; | |
2077 | case BDRV_ACTION_IGNORE: | |
2078 | action_str = "ignore"; | |
2079 | break; | |
2080 | case BDRV_ACTION_STOP: | |
2081 | action_str = "stop"; | |
2082 | break; | |
2083 | default: | |
2084 | abort(); | |
2085 | } | |
2086 | ||
2087 | data = qobject_from_jsonf("{ 'device': %s, 'action': %s, 'operation': %s }", | |
2088 | bdrv->device_name, | |
2089 | action_str, | |
2090 | is_read ? "read" : "write"); | |
32c81a4a | 2091 | monitor_protocol_event(ev, data); |
329c0a48 LC |
2092 | |
2093 | qobject_decref(data); | |
2094 | } | |
2095 | ||
6f382ed2 LC |
2096 | static void bdrv_emit_qmp_eject_event(BlockDriverState *bs, bool ejected) |
2097 | { | |
2098 | QObject *data; | |
2099 | ||
2100 | data = qobject_from_jsonf("{ 'device': %s, 'tray-open': %i }", | |
2101 | bdrv_get_device_name(bs), ejected); | |
2102 | monitor_protocol_event(QEVENT_DEVICE_TRAY_MOVED, data); | |
2103 | ||
2104 | qobject_decref(data); | |
2105 | } | |
2106 | ||
7d4b4ba5 | 2107 | static void bdrv_dev_change_media_cb(BlockDriverState *bs, bool load) |
0e49de52 | 2108 | { |
145feb17 | 2109 | if (bs->dev_ops && bs->dev_ops->change_media_cb) { |
6f382ed2 | 2110 | bool tray_was_closed = !bdrv_dev_is_tray_open(bs); |
7d4b4ba5 | 2111 | bs->dev_ops->change_media_cb(bs->dev_opaque, load); |
6f382ed2 LC |
2112 | if (tray_was_closed) { |
2113 | /* tray open */ | |
2114 | bdrv_emit_qmp_eject_event(bs, true); | |
2115 | } | |
2116 | if (load) { | |
2117 | /* tray close */ | |
2118 | bdrv_emit_qmp_eject_event(bs, false); | |
2119 | } | |
145feb17 MA |
2120 | } |
2121 | } | |
2122 | ||
2c6942fa MA |
2123 | bool bdrv_dev_has_removable_media(BlockDriverState *bs) |
2124 | { | |
2125 | return !bs->dev || (bs->dev_ops && bs->dev_ops->change_media_cb); | |
2126 | } | |
2127 | ||
025ccaa7 PB |
2128 | void bdrv_dev_eject_request(BlockDriverState *bs, bool force) |
2129 | { | |
2130 | if (bs->dev_ops && bs->dev_ops->eject_request_cb) { | |
2131 | bs->dev_ops->eject_request_cb(bs->dev_opaque, force); | |
2132 | } | |
2133 | } | |
2134 | ||
e4def80b MA |
2135 | bool bdrv_dev_is_tray_open(BlockDriverState *bs) |
2136 | { | |
2137 | if (bs->dev_ops && bs->dev_ops->is_tray_open) { | |
2138 | return bs->dev_ops->is_tray_open(bs->dev_opaque); | |
2139 | } | |
2140 | return false; | |
2141 | } | |
2142 | ||
145feb17 MA |
2143 | static void bdrv_dev_resize_cb(BlockDriverState *bs) |
2144 | { | |
2145 | if (bs->dev_ops && bs->dev_ops->resize_cb) { | |
2146 | bs->dev_ops->resize_cb(bs->dev_opaque); | |
0e49de52 MA |
2147 | } |
2148 | } | |
2149 | ||
f107639a MA |
2150 | bool bdrv_dev_is_medium_locked(BlockDriverState *bs) |
2151 | { | |
2152 | if (bs->dev_ops && bs->dev_ops->is_medium_locked) { | |
2153 | return bs->dev_ops->is_medium_locked(bs->dev_opaque); | |
2154 | } | |
2155 | return false; | |
2156 | } | |
2157 | ||
e97fc193 AL |
2158 | /* |
2159 | * Run consistency checks on an image | |
2160 | * | |
e076f338 | 2161 | * Returns 0 if the check could be completed (it doesn't mean that the image is |
a1c7273b | 2162 | * free of errors) or -errno when an internal error occurred. The results of the |
e076f338 | 2163 | * check are stored in res. |
e97fc193 | 2164 | */ |
4534ff54 | 2165 | int bdrv_check(BlockDriverState *bs, BdrvCheckResult *res, BdrvCheckMode fix) |
e97fc193 AL |
2166 | { |
2167 | if (bs->drv->bdrv_check == NULL) { | |
2168 | return -ENOTSUP; | |
2169 | } | |
2170 | ||
e076f338 | 2171 | memset(res, 0, sizeof(*res)); |
4534ff54 | 2172 | return bs->drv->bdrv_check(bs, res, fix); |
e97fc193 AL |
2173 | } |
2174 | ||
8a426614 KW |
2175 | #define COMMIT_BUF_SECTORS 2048 |
2176 | ||
33e3963e FB |
2177 | /* commit COW file into the raw image */ |
2178 | int bdrv_commit(BlockDriverState *bs) | |
2179 | { | |
19cb3738 | 2180 | BlockDriver *drv = bs->drv; |
72706ea4 | 2181 | int64_t sector, total_sectors, length, backing_length; |
8a426614 | 2182 | int n, ro, open_flags; |
0bce597d | 2183 | int ret = 0; |
72706ea4 | 2184 | uint8_t *buf = NULL; |
c2cba3d9 | 2185 | char filename[PATH_MAX]; |
33e3963e | 2186 | |
19cb3738 FB |
2187 | if (!drv) |
2188 | return -ENOMEDIUM; | |
4dca4b63 NS |
2189 | |
2190 | if (!bs->backing_hd) { | |
2191 | return -ENOTSUP; | |
33e3963e FB |
2192 | } |
2193 | ||
2d3735d3 SH |
2194 | if (bdrv_in_use(bs) || bdrv_in_use(bs->backing_hd)) { |
2195 | return -EBUSY; | |
2196 | } | |
2197 | ||
4dca4b63 | 2198 | ro = bs->backing_hd->read_only; |
c2cba3d9 JM |
2199 | /* Use pstrcpy (not strncpy): filename must be NUL-terminated. */ |
2200 | pstrcpy(filename, sizeof(filename), bs->backing_hd->filename); | |
4dca4b63 NS |
2201 | open_flags = bs->backing_hd->open_flags; |
2202 | ||
2203 | if (ro) { | |
0bce597d JC |
2204 | if (bdrv_reopen(bs->backing_hd, open_flags | BDRV_O_RDWR, NULL)) { |
2205 | return -EACCES; | |
4dca4b63 | 2206 | } |
ea2384d3 | 2207 | } |
33e3963e | 2208 | |
72706ea4 JC |
2209 | length = bdrv_getlength(bs); |
2210 | if (length < 0) { | |
2211 | ret = length; | |
2212 | goto ro_cleanup; | |
2213 | } | |
2214 | ||
2215 | backing_length = bdrv_getlength(bs->backing_hd); | |
2216 | if (backing_length < 0) { | |
2217 | ret = backing_length; | |
2218 | goto ro_cleanup; | |
2219 | } | |
2220 | ||
2221 | /* If our top snapshot is larger than the backing file image, | |
2222 | * grow the backing file image if possible. If not possible, | |
2223 | * we must return an error */ | |
2224 | if (length > backing_length) { | |
2225 | ret = bdrv_truncate(bs->backing_hd, length); | |
2226 | if (ret < 0) { | |
2227 | goto ro_cleanup; | |
2228 | } | |
2229 | } | |
2230 | ||
2231 | total_sectors = length >> BDRV_SECTOR_BITS; | |
7267c094 | 2232 | buf = g_malloc(COMMIT_BUF_SECTORS * BDRV_SECTOR_SIZE); |
8a426614 KW |
2233 | |
2234 | for (sector = 0; sector < total_sectors; sector += n) { | |
d663640c PB |
2235 | ret = bdrv_is_allocated(bs, sector, COMMIT_BUF_SECTORS, &n); |
2236 | if (ret < 0) { | |
2237 | goto ro_cleanup; | |
2238 | } | |
2239 | if (ret) { | |
dabfa6cc KW |
2240 | ret = bdrv_read(bs, sector, buf, n); |
2241 | if (ret < 0) { | |
8a426614 KW |
2242 | goto ro_cleanup; |
2243 | } | |
2244 | ||
dabfa6cc KW |
2245 | ret = bdrv_write(bs->backing_hd, sector, buf, n); |
2246 | if (ret < 0) { | |
8a426614 KW |
2247 | goto ro_cleanup; |
2248 | } | |
ea2384d3 | 2249 | } |
33e3963e | 2250 | } |
95389c86 | 2251 | |
1d44952f CH |
2252 | if (drv->bdrv_make_empty) { |
2253 | ret = drv->bdrv_make_empty(bs); | |
dabfa6cc KW |
2254 | if (ret < 0) { |
2255 | goto ro_cleanup; | |
2256 | } | |
1d44952f CH |
2257 | bdrv_flush(bs); |
2258 | } | |
95389c86 | 2259 | |
3f5075ae CH |
2260 | /* |
2261 | * Make sure all data we wrote to the backing device is actually | |
2262 | * stable on disk. | |
2263 | */ | |
dabfa6cc | 2264 | if (bs->backing_hd) { |
3f5075ae | 2265 | bdrv_flush(bs->backing_hd); |
dabfa6cc | 2266 | } |
4dca4b63 | 2267 | |
dabfa6cc | 2268 | ret = 0; |
4dca4b63 | 2269 | ro_cleanup: |
7267c094 | 2270 | g_free(buf); |
4dca4b63 NS |
2271 | |
2272 | if (ro) { | |
0bce597d JC |
2273 | /* ignoring error return here */ |
2274 | bdrv_reopen(bs->backing_hd, open_flags & ~BDRV_O_RDWR, NULL); | |
4dca4b63 NS |
2275 | } |
2276 | ||
1d44952f | 2277 | return ret; |
33e3963e FB |
2278 | } |
2279 | ||
e8877497 | 2280 | int bdrv_commit_all(void) |
6ab4b5ab MA |
2281 | { |
2282 | BlockDriverState *bs; | |
2283 | ||
dc364f4c | 2284 | QTAILQ_FOREACH(bs, &bdrv_states, device_list) { |
272d2d8e JC |
2285 | if (bs->drv && bs->backing_hd) { |
2286 | int ret = bdrv_commit(bs); | |
2287 | if (ret < 0) { | |
2288 | return ret; | |
2289 | } | |
e8877497 | 2290 | } |
6ab4b5ab | 2291 | } |
e8877497 | 2292 | return 0; |
6ab4b5ab MA |
2293 | } |
2294 | ||
dbffbdcf SH |
2295 | /** |
2296 | * Remove an active request from the tracked requests list | |
2297 | * | |
2298 | * This function should be called when a tracked request is completing. | |
2299 | */ | |
2300 | static void tracked_request_end(BdrvTrackedRequest *req) | |
2301 | { | |
2dbafdc0 KW |
2302 | if (req->serialising) { |
2303 | req->bs->serialising_in_flight--; | |
2304 | } | |
2305 | ||
dbffbdcf | 2306 | QLIST_REMOVE(req, list); |
f4658285 | 2307 | qemu_co_queue_restart_all(&req->wait_queue); |
dbffbdcf SH |
2308 | } |
2309 | ||
2310 | /** | |
2311 | * Add an active request to the tracked requests list | |
2312 | */ | |
2313 | static void tracked_request_begin(BdrvTrackedRequest *req, | |
2314 | BlockDriverState *bs, | |
793ed47a KW |
2315 | int64_t offset, |
2316 | unsigned int bytes, bool is_write) | |
dbffbdcf SH |
2317 | { |
2318 | *req = (BdrvTrackedRequest){ | |
2319 | .bs = bs, | |
2dbafdc0 KW |
2320 | .offset = offset, |
2321 | .bytes = bytes, | |
2322 | .is_write = is_write, | |
2323 | .co = qemu_coroutine_self(), | |
2324 | .serialising = false, | |
7327145f KW |
2325 | .overlap_offset = offset, |
2326 | .overlap_bytes = bytes, | |
dbffbdcf SH |
2327 | }; |
2328 | ||
f4658285 SH |
2329 | qemu_co_queue_init(&req->wait_queue); |
2330 | ||
dbffbdcf SH |
2331 | QLIST_INSERT_HEAD(&bs->tracked_requests, req, list); |
2332 | } | |
2333 | ||
e96126ff | 2334 | static void mark_request_serialising(BdrvTrackedRequest *req, uint64_t align) |
2dbafdc0 | 2335 | { |
7327145f | 2336 | int64_t overlap_offset = req->offset & ~(align - 1); |
e96126ff KW |
2337 | unsigned int overlap_bytes = ROUND_UP(req->offset + req->bytes, align) |
2338 | - overlap_offset; | |
7327145f | 2339 | |
2dbafdc0 KW |
2340 | if (!req->serialising) { |
2341 | req->bs->serialising_in_flight++; | |
2342 | req->serialising = true; | |
2343 | } | |
7327145f KW |
2344 | |
2345 | req->overlap_offset = MIN(req->overlap_offset, overlap_offset); | |
2346 | req->overlap_bytes = MAX(req->overlap_bytes, overlap_bytes); | |
2dbafdc0 KW |
2347 | } |
2348 | ||
d83947ac SH |
2349 | /** |
2350 | * Round a region to cluster boundaries | |
2351 | */ | |
343bded4 PB |
2352 | void bdrv_round_to_clusters(BlockDriverState *bs, |
2353 | int64_t sector_num, int nb_sectors, | |
2354 | int64_t *cluster_sector_num, | |
2355 | int *cluster_nb_sectors) | |
d83947ac SH |
2356 | { |
2357 | BlockDriverInfo bdi; | |
2358 | ||
2359 | if (bdrv_get_info(bs, &bdi) < 0 || bdi.cluster_size == 0) { | |
2360 | *cluster_sector_num = sector_num; | |
2361 | *cluster_nb_sectors = nb_sectors; | |
2362 | } else { | |
2363 | int64_t c = bdi.cluster_size / BDRV_SECTOR_SIZE; | |
2364 | *cluster_sector_num = QEMU_ALIGN_DOWN(sector_num, c); | |
2365 | *cluster_nb_sectors = QEMU_ALIGN_UP(sector_num - *cluster_sector_num + | |
2366 | nb_sectors, c); | |
2367 | } | |
2368 | } | |
2369 | ||
7327145f | 2370 | static int bdrv_get_cluster_size(BlockDriverState *bs) |
793ed47a KW |
2371 | { |
2372 | BlockDriverInfo bdi; | |
7327145f | 2373 | int ret; |
793ed47a | 2374 | |
7327145f KW |
2375 | ret = bdrv_get_info(bs, &bdi); |
2376 | if (ret < 0 || bdi.cluster_size == 0) { | |
2377 | return bs->request_alignment; | |
793ed47a | 2378 | } else { |
7327145f | 2379 | return bdi.cluster_size; |
793ed47a KW |
2380 | } |
2381 | } | |
2382 | ||
f4658285 | 2383 | static bool tracked_request_overlaps(BdrvTrackedRequest *req, |
793ed47a KW |
2384 | int64_t offset, unsigned int bytes) |
2385 | { | |
d83947ac | 2386 | /* aaaa bbbb */ |
7327145f | 2387 | if (offset >= req->overlap_offset + req->overlap_bytes) { |
d83947ac SH |
2388 | return false; |
2389 | } | |
2390 | /* bbbb aaaa */ | |
7327145f | 2391 | if (req->overlap_offset >= offset + bytes) { |
d83947ac SH |
2392 | return false; |
2393 | } | |
2394 | return true; | |
f4658285 SH |
2395 | } |
2396 | ||
28de2dcd | 2397 | static bool coroutine_fn wait_serialising_requests(BdrvTrackedRequest *self) |
f4658285 | 2398 | { |
2dbafdc0 | 2399 | BlockDriverState *bs = self->bs; |
f4658285 SH |
2400 | BdrvTrackedRequest *req; |
2401 | bool retry; | |
28de2dcd | 2402 | bool waited = false; |
f4658285 | 2403 | |
2dbafdc0 | 2404 | if (!bs->serialising_in_flight) { |
28de2dcd | 2405 | return false; |
2dbafdc0 KW |
2406 | } |
2407 | ||
f4658285 SH |
2408 | do { |
2409 | retry = false; | |
2410 | QLIST_FOREACH(req, &bs->tracked_requests, list) { | |
2dbafdc0 | 2411 | if (req == self || (!req->serialising && !self->serialising)) { |
65afd211 KW |
2412 | continue; |
2413 | } | |
7327145f KW |
2414 | if (tracked_request_overlaps(req, self->overlap_offset, |
2415 | self->overlap_bytes)) | |
2416 | { | |
5f8b6491 SH |
2417 | /* Hitting this means there was a reentrant request, for |
2418 | * example, a block driver issuing nested requests. This must | |
2419 | * never happen since it means deadlock. | |
2420 | */ | |
2421 | assert(qemu_coroutine_self() != req->co); | |
2422 | ||
6460440f KW |
2423 | /* If the request is already (indirectly) waiting for us, or |
2424 | * will wait for us as soon as it wakes up, then just go on | |
2425 | * (instead of producing a deadlock in the former case). */ | |
2426 | if (!req->waiting_for) { | |
2427 | self->waiting_for = req; | |
2428 | qemu_co_queue_wait(&req->wait_queue); | |
2429 | self->waiting_for = NULL; | |
2430 | retry = true; | |
28de2dcd | 2431 | waited = true; |
6460440f KW |
2432 | break; |
2433 | } | |
f4658285 SH |
2434 | } |
2435 | } | |
2436 | } while (retry); | |
28de2dcd KW |
2437 | |
2438 | return waited; | |
f4658285 SH |
2439 | } |
2440 | ||
756e6736 KW |
2441 | /* |
2442 | * Return values: | |
2443 | * 0 - success | |
2444 | * -EINVAL - backing format specified, but no file | |
2445 | * -ENOSPC - can't update the backing file because no space is left in the | |
2446 | * image file header | |
2447 | * -ENOTSUP - format driver doesn't support changing the backing file | |
2448 | */ | |
2449 | int bdrv_change_backing_file(BlockDriverState *bs, | |
2450 | const char *backing_file, const char *backing_fmt) | |
2451 | { | |
2452 | BlockDriver *drv = bs->drv; | |
469ef350 | 2453 | int ret; |
756e6736 | 2454 | |
5f377794 PB |
2455 | /* Backing file format doesn't make sense without a backing file */ |
2456 | if (backing_fmt && !backing_file) { | |
2457 | return -EINVAL; | |
2458 | } | |
2459 | ||
756e6736 | 2460 | if (drv->bdrv_change_backing_file != NULL) { |
469ef350 | 2461 | ret = drv->bdrv_change_backing_file(bs, backing_file, backing_fmt); |
756e6736 | 2462 | } else { |
469ef350 | 2463 | ret = -ENOTSUP; |
756e6736 | 2464 | } |
469ef350 PB |
2465 | |
2466 | if (ret == 0) { | |
2467 | pstrcpy(bs->backing_file, sizeof(bs->backing_file), backing_file ?: ""); | |
2468 | pstrcpy(bs->backing_format, sizeof(bs->backing_format), backing_fmt ?: ""); | |
2469 | } | |
2470 | return ret; | |
756e6736 KW |
2471 | } |
2472 | ||
6ebdcee2 JC |
2473 | /* |
2474 | * Finds the image layer in the chain that has 'bs' as its backing file. | |
2475 | * | |
2476 | * active is the current topmost image. | |
2477 | * | |
2478 | * Returns NULL if bs is not found in active's image chain, | |
2479 | * or if active == bs. | |
2480 | */ | |
2481 | BlockDriverState *bdrv_find_overlay(BlockDriverState *active, | |
2482 | BlockDriverState *bs) | |
2483 | { | |
2484 | BlockDriverState *overlay = NULL; | |
2485 | BlockDriverState *intermediate; | |
2486 | ||
2487 | assert(active != NULL); | |
2488 | assert(bs != NULL); | |
2489 | ||
2490 | /* if bs is the same as active, then by definition it has no overlay | |
2491 | */ | |
2492 | if (active == bs) { | |
2493 | return NULL; | |
2494 | } | |
2495 | ||
2496 | intermediate = active; | |
2497 | while (intermediate->backing_hd) { | |
2498 | if (intermediate->backing_hd == bs) { | |
2499 | overlay = intermediate; | |
2500 | break; | |
2501 | } | |
2502 | intermediate = intermediate->backing_hd; | |
2503 | } | |
2504 | ||
2505 | return overlay; | |
2506 | } | |
2507 | ||
2508 | typedef struct BlkIntermediateStates { | |
2509 | BlockDriverState *bs; | |
2510 | QSIMPLEQ_ENTRY(BlkIntermediateStates) entry; | |
2511 | } BlkIntermediateStates; | |
2512 | ||
2513 | ||
2514 | /* | |
2515 | * Drops images above 'base' up to and including 'top', and sets the image | |
2516 | * above 'top' to have base as its backing file. | |
2517 | * | |
2518 | * Requires that the overlay to 'top' is opened r/w, so that the backing file | |
2519 | * information in 'bs' can be properly updated. | |
2520 | * | |
2521 | * E.g., this will convert the following chain: | |
2522 | * bottom <- base <- intermediate <- top <- active | |
2523 | * | |
2524 | * to | |
2525 | * | |
2526 | * bottom <- base <- active | |
2527 | * | |
2528 | * It is allowed for bottom==base, in which case it converts: | |
2529 | * | |
2530 | * base <- intermediate <- top <- active | |
2531 | * | |
2532 | * to | |
2533 | * | |
2534 | * base <- active | |
2535 | * | |
2536 | * Error conditions: | |
2537 | * if active == top, that is considered an error | |
2538 | * | |
2539 | */ | |
2540 | int bdrv_drop_intermediate(BlockDriverState *active, BlockDriverState *top, | |
2541 | BlockDriverState *base) | |
2542 | { | |
2543 | BlockDriverState *intermediate; | |
2544 | BlockDriverState *base_bs = NULL; | |
2545 | BlockDriverState *new_top_bs = NULL; | |
2546 | BlkIntermediateStates *intermediate_state, *next; | |
2547 | int ret = -EIO; | |
2548 | ||
2549 | QSIMPLEQ_HEAD(states_to_delete, BlkIntermediateStates) states_to_delete; | |
2550 | QSIMPLEQ_INIT(&states_to_delete); | |
2551 | ||
2552 | if (!top->drv || !base->drv) { | |
2553 | goto exit; | |
2554 | } | |
2555 | ||
2556 | new_top_bs = bdrv_find_overlay(active, top); | |
2557 | ||
2558 | if (new_top_bs == NULL) { | |
2559 | /* we could not find the image above 'top', this is an error */ | |
2560 | goto exit; | |
2561 | } | |
2562 | ||
2563 | /* special case of new_top_bs->backing_hd already pointing to base - nothing | |
2564 | * to do, no intermediate images */ | |
2565 | if (new_top_bs->backing_hd == base) { | |
2566 | ret = 0; | |
2567 | goto exit; | |
2568 | } | |
2569 | ||
2570 | intermediate = top; | |
2571 | ||
2572 | /* now we will go down through the list, and add each BDS we find | |
2573 | * into our deletion queue, until we hit the 'base' | |
2574 | */ | |
2575 | while (intermediate) { | |
2576 | intermediate_state = g_malloc0(sizeof(BlkIntermediateStates)); | |
2577 | intermediate_state->bs = intermediate; | |
2578 | QSIMPLEQ_INSERT_TAIL(&states_to_delete, intermediate_state, entry); | |
2579 | ||
2580 | if (intermediate->backing_hd == base) { | |
2581 | base_bs = intermediate->backing_hd; | |
2582 | break; | |
2583 | } | |
2584 | intermediate = intermediate->backing_hd; | |
2585 | } | |
2586 | if (base_bs == NULL) { | |
2587 | /* something went wrong, we did not end at the base. safely | |
2588 | * unravel everything, and exit with error */ | |
2589 | goto exit; | |
2590 | } | |
2591 | ||
2592 | /* success - we can delete the intermediate states, and link top->base */ | |
2593 | ret = bdrv_change_backing_file(new_top_bs, base_bs->filename, | |
2594 | base_bs->drv ? base_bs->drv->format_name : ""); | |
2595 | if (ret) { | |
2596 | goto exit; | |
2597 | } | |
2598 | new_top_bs->backing_hd = base_bs; | |
2599 | ||
355ef4ac | 2600 | bdrv_refresh_limits(new_top_bs); |
6ebdcee2 JC |
2601 | |
2602 | QSIMPLEQ_FOREACH_SAFE(intermediate_state, &states_to_delete, entry, next) { | |
2603 | /* so that bdrv_close() does not recursively close the chain */ | |
2604 | intermediate_state->bs->backing_hd = NULL; | |
4f6fd349 | 2605 | bdrv_unref(intermediate_state->bs); |
6ebdcee2 JC |
2606 | } |
2607 | ret = 0; | |
2608 | ||
2609 | exit: | |
2610 | QSIMPLEQ_FOREACH_SAFE(intermediate_state, &states_to_delete, entry, next) { | |
2611 | g_free(intermediate_state); | |
2612 | } | |
2613 | return ret; | |
2614 | } | |
2615 | ||
2616 | ||
71d0770c AL |
2617 | static int bdrv_check_byte_request(BlockDriverState *bs, int64_t offset, |
2618 | size_t size) | |
2619 | { | |
2620 | int64_t len; | |
2621 | ||
1dd3a447 KW |
2622 | if (size > INT_MAX) { |
2623 | return -EIO; | |
2624 | } | |
2625 | ||
71d0770c AL |
2626 | if (!bdrv_is_inserted(bs)) |
2627 | return -ENOMEDIUM; | |
2628 | ||
2629 | if (bs->growable) | |
2630 | return 0; | |
2631 | ||
2632 | len = bdrv_getlength(bs); | |
2633 | ||
fbb7b4e0 KW |
2634 | if (offset < 0) |
2635 | return -EIO; | |
2636 | ||
2637 | if ((offset > len) || (len - offset < size)) | |
71d0770c AL |
2638 | return -EIO; |
2639 | ||
2640 | return 0; | |
2641 | } | |
2642 | ||
2643 | static int bdrv_check_request(BlockDriverState *bs, int64_t sector_num, | |
2644 | int nb_sectors) | |
2645 | { | |
54db38a4 | 2646 | if (nb_sectors < 0 || nb_sectors > INT_MAX / BDRV_SECTOR_SIZE) { |
8f4754ed KW |
2647 | return -EIO; |
2648 | } | |
2649 | ||
eb5a3165 JS |
2650 | return bdrv_check_byte_request(bs, sector_num * BDRV_SECTOR_SIZE, |
2651 | nb_sectors * BDRV_SECTOR_SIZE); | |
71d0770c AL |
2652 | } |
2653 | ||
1c9805a3 SH |
2654 | typedef struct RwCo { |
2655 | BlockDriverState *bs; | |
775aa8b6 | 2656 | int64_t offset; |
1c9805a3 SH |
2657 | QEMUIOVector *qiov; |
2658 | bool is_write; | |
2659 | int ret; | |
4105eaaa | 2660 | BdrvRequestFlags flags; |
1c9805a3 SH |
2661 | } RwCo; |
2662 | ||
2663 | static void coroutine_fn bdrv_rw_co_entry(void *opaque) | |
fc01f7e7 | 2664 | { |
1c9805a3 | 2665 | RwCo *rwco = opaque; |
ea2384d3 | 2666 | |
1c9805a3 | 2667 | if (!rwco->is_write) { |
775aa8b6 KW |
2668 | rwco->ret = bdrv_co_do_preadv(rwco->bs, rwco->offset, |
2669 | rwco->qiov->size, rwco->qiov, | |
4105eaaa | 2670 | rwco->flags); |
775aa8b6 KW |
2671 | } else { |
2672 | rwco->ret = bdrv_co_do_pwritev(rwco->bs, rwco->offset, | |
2673 | rwco->qiov->size, rwco->qiov, | |
2674 | rwco->flags); | |
1c9805a3 SH |
2675 | } |
2676 | } | |
e7a8a783 | 2677 | |
1c9805a3 | 2678 | /* |
8d3b1a2d | 2679 | * Process a vectored synchronous request using coroutines |
1c9805a3 | 2680 | */ |
775aa8b6 KW |
2681 | static int bdrv_prwv_co(BlockDriverState *bs, int64_t offset, |
2682 | QEMUIOVector *qiov, bool is_write, | |
2683 | BdrvRequestFlags flags) | |
1c9805a3 | 2684 | { |
1c9805a3 SH |
2685 | Coroutine *co; |
2686 | RwCo rwco = { | |
2687 | .bs = bs, | |
775aa8b6 | 2688 | .offset = offset, |
8d3b1a2d | 2689 | .qiov = qiov, |
1c9805a3 SH |
2690 | .is_write = is_write, |
2691 | .ret = NOT_DONE, | |
4105eaaa | 2692 | .flags = flags, |
1c9805a3 | 2693 | }; |
e7a8a783 | 2694 | |
498e386c ZYW |
2695 | /** |
2696 | * In sync call context, when the vcpu is blocked, this throttling timer | |
2697 | * will not fire; so the I/O throttling function has to be disabled here | |
2698 | * if it has been enabled. | |
2699 | */ | |
2700 | if (bs->io_limits_enabled) { | |
2701 | fprintf(stderr, "Disabling I/O throttling on '%s' due " | |
2702 | "to synchronous I/O.\n", bdrv_get_device_name(bs)); | |
2703 | bdrv_io_limits_disable(bs); | |
2704 | } | |
2705 | ||
1c9805a3 SH |
2706 | if (qemu_in_coroutine()) { |
2707 | /* Fast-path if already in coroutine context */ | |
2708 | bdrv_rw_co_entry(&rwco); | |
2709 | } else { | |
2710 | co = qemu_coroutine_create(bdrv_rw_co_entry); | |
2711 | qemu_coroutine_enter(co, &rwco); | |
2712 | while (rwco.ret == NOT_DONE) { | |
2713 | qemu_aio_wait(); | |
2714 | } | |
2715 | } | |
2716 | return rwco.ret; | |
2717 | } | |
b338082b | 2718 | |
8d3b1a2d KW |
2719 | /* |
2720 | * Process a synchronous request using coroutines | |
2721 | */ | |
2722 | static int bdrv_rw_co(BlockDriverState *bs, int64_t sector_num, uint8_t *buf, | |
4105eaaa | 2723 | int nb_sectors, bool is_write, BdrvRequestFlags flags) |
8d3b1a2d KW |
2724 | { |
2725 | QEMUIOVector qiov; | |
2726 | struct iovec iov = { | |
2727 | .iov_base = (void *)buf, | |
2728 | .iov_len = nb_sectors * BDRV_SECTOR_SIZE, | |
2729 | }; | |
2730 | ||
da15ee51 KW |
2731 | if (nb_sectors < 0 || nb_sectors > INT_MAX / BDRV_SECTOR_SIZE) { |
2732 | return -EINVAL; | |
2733 | } | |
2734 | ||
8d3b1a2d | 2735 | qemu_iovec_init_external(&qiov, &iov, 1); |
775aa8b6 KW |
2736 | return bdrv_prwv_co(bs, sector_num << BDRV_SECTOR_BITS, |
2737 | &qiov, is_write, flags); | |
8d3b1a2d KW |
2738 | } |
2739 | ||
1c9805a3 SH |
2740 | /* return < 0 if error. See bdrv_write() for the return codes */ |
2741 | int bdrv_read(BlockDriverState *bs, int64_t sector_num, | |
2742 | uint8_t *buf, int nb_sectors) | |
2743 | { | |
4105eaaa | 2744 | return bdrv_rw_co(bs, sector_num, buf, nb_sectors, false, 0); |
fc01f7e7 FB |
2745 | } |
2746 | ||
07d27a44 MA |
2747 | /* Just like bdrv_read(), but with I/O throttling temporarily disabled */ |
2748 | int bdrv_read_unthrottled(BlockDriverState *bs, int64_t sector_num, | |
2749 | uint8_t *buf, int nb_sectors) | |
2750 | { | |
2751 | bool enabled; | |
2752 | int ret; | |
2753 | ||
2754 | enabled = bs->io_limits_enabled; | |
2755 | bs->io_limits_enabled = false; | |
4e7395e8 | 2756 | ret = bdrv_read(bs, sector_num, buf, nb_sectors); |
07d27a44 MA |
2757 | bs->io_limits_enabled = enabled; |
2758 | return ret; | |
2759 | } | |
2760 | ||
5fafdf24 | 2761 | /* Return < 0 if error. Important errors are: |
19cb3738 FB |
2762 | -EIO generic I/O error (may happen for all errors) |
2763 | -ENOMEDIUM No media inserted. | |
2764 | -EINVAL Invalid sector number or nb_sectors | |
2765 | -EACCES Trying to write a read-only device | |
2766 | */ | |
5fafdf24 | 2767 | int bdrv_write(BlockDriverState *bs, int64_t sector_num, |
fc01f7e7 FB |
2768 | const uint8_t *buf, int nb_sectors) |
2769 | { | |
4105eaaa | 2770 | return bdrv_rw_co(bs, sector_num, (uint8_t *)buf, nb_sectors, true, 0); |
83f64091 FB |
2771 | } |
2772 | ||
aa7bfbff PL |
2773 | int bdrv_write_zeroes(BlockDriverState *bs, int64_t sector_num, |
2774 | int nb_sectors, BdrvRequestFlags flags) | |
4105eaaa PL |
2775 | { |
2776 | return bdrv_rw_co(bs, sector_num, NULL, nb_sectors, true, | |
aa7bfbff | 2777 | BDRV_REQ_ZERO_WRITE | flags); |
8d3b1a2d KW |
2778 | } |
2779 | ||
d75cbb5e PL |
2780 | /* |
2781 | * Completely zero out a block device with the help of bdrv_write_zeroes. | |
2782 | * The operation is sped up by checking the block status and only writing | |
2783 | * zeroes to the device if they currently do not return zeroes. Optional | |
2784 | * flags are passed through to bdrv_write_zeroes (e.g. BDRV_REQ_MAY_UNMAP). | |
2785 | * | |
2786 | * Returns < 0 on error, 0 on success. For error codes see bdrv_write(). | |
2787 | */ | |
2788 | int bdrv_make_zero(BlockDriverState *bs, BdrvRequestFlags flags) | |
2789 | { | |
9ce10c0b | 2790 | int64_t target_size; |
d75cbb5e PL |
2791 | int64_t ret, nb_sectors, sector_num = 0; |
2792 | int n; | |
2793 | ||
9ce10c0b KW |
2794 | target_size = bdrv_getlength(bs); |
2795 | if (target_size < 0) { | |
2796 | return target_size; | |
2797 | } | |
2798 | target_size /= BDRV_SECTOR_SIZE; | |
2799 | ||
d75cbb5e PL |
2800 | for (;;) { |
2801 | nb_sectors = target_size - sector_num; | |
2802 | if (nb_sectors <= 0) { | |
2803 | return 0; | |
2804 | } | |
2805 | if (nb_sectors > INT_MAX) { | |
2806 | nb_sectors = INT_MAX; | |
2807 | } | |
2808 | ret = bdrv_get_block_status(bs, sector_num, nb_sectors, &n); | |
3d94ce60 PL |
2809 | if (ret < 0) { |
2810 | error_report("error getting block status at sector %" PRId64 ": %s", | |
2811 | sector_num, strerror(-ret)); | |
2812 | return ret; | |
2813 | } | |
d75cbb5e PL |
2814 | if (ret & BDRV_BLOCK_ZERO) { |
2815 | sector_num += n; | |
2816 | continue; | |
2817 | } | |
2818 | ret = bdrv_write_zeroes(bs, sector_num, n, flags); | |
2819 | if (ret < 0) { | |
2820 | error_report("error writing zeroes at sector %" PRId64 ": %s", | |
2821 | sector_num, strerror(-ret)); | |
2822 | return ret; | |
2823 | } | |
2824 | sector_num += n; | |
2825 | } | |
2826 | } | |
2827 | ||
a3ef6571 | 2828 | int bdrv_pread(BlockDriverState *bs, int64_t offset, void *buf, int bytes) |
83f64091 | 2829 | { |
a3ef6571 KW |
2830 | QEMUIOVector qiov; |
2831 | struct iovec iov = { | |
2832 | .iov_base = (void *)buf, | |
2833 | .iov_len = bytes, | |
2834 | }; | |
9a8c4cce | 2835 | int ret; |
83f64091 | 2836 | |
a3ef6571 KW |
2837 | if (bytes < 0) { |
2838 | return -EINVAL; | |
83f64091 FB |
2839 | } |
2840 | ||
a3ef6571 KW |
2841 | qemu_iovec_init_external(&qiov, &iov, 1); |
2842 | ret = bdrv_prwv_co(bs, offset, &qiov, false, 0); | |
2843 | if (ret < 0) { | |
2844 | return ret; | |
83f64091 | 2845 | } |
a3ef6571 KW |
2846 | |
2847 | return bytes; | |
83f64091 FB |
2848 | } |
2849 | ||
8d3b1a2d | 2850 | int bdrv_pwritev(BlockDriverState *bs, int64_t offset, QEMUIOVector *qiov) |
83f64091 | 2851 | { |
9a8c4cce | 2852 | int ret; |
83f64091 | 2853 | |
8407d5d7 KW |
2854 | ret = bdrv_prwv_co(bs, offset, qiov, true, 0); |
2855 | if (ret < 0) { | |
2856 | return ret; | |
83f64091 FB |
2857 | } |
2858 | ||
8d3b1a2d KW |
2859 | return qiov->size; |
2860 | } | |
2861 | ||
2862 | int bdrv_pwrite(BlockDriverState *bs, int64_t offset, | |
8407d5d7 | 2863 | const void *buf, int bytes) |
8d3b1a2d KW |
2864 | { |
2865 | QEMUIOVector qiov; | |
2866 | struct iovec iov = { | |
2867 | .iov_base = (void *) buf, | |
8407d5d7 | 2868 | .iov_len = bytes, |
8d3b1a2d KW |
2869 | }; |
2870 | ||
8407d5d7 KW |
2871 | if (bytes < 0) { |
2872 | return -EINVAL; | |
2873 | } | |
2874 | ||
8d3b1a2d KW |
2875 | qemu_iovec_init_external(&qiov, &iov, 1); |
2876 | return bdrv_pwritev(bs, offset, &qiov); | |
83f64091 | 2877 | } |
83f64091 | 2878 | |
f08145fe KW |
2879 | /* |
2880 | * Writes to the file and ensures that no writes are reordered across this | |
2881 | * request (acts as a barrier) | |
2882 | * | |
2883 | * Returns 0 on success, -errno in error cases. | |
2884 | */ | |
2885 | int bdrv_pwrite_sync(BlockDriverState *bs, int64_t offset, | |
2886 | const void *buf, int count) | |
2887 | { | |
2888 | int ret; | |
2889 | ||
2890 | ret = bdrv_pwrite(bs, offset, buf, count); | |
2891 | if (ret < 0) { | |
2892 | return ret; | |
2893 | } | |
2894 | ||
f05fa4ad PB |
2895 | /* No flush needed for cache modes that already do it */ |
2896 | if (bs->enable_write_cache) { | |
f08145fe KW |
2897 | bdrv_flush(bs); |
2898 | } | |
2899 | ||
2900 | return 0; | |
2901 | } | |
2902 | ||
470c0504 | 2903 | static int coroutine_fn bdrv_co_do_copy_on_readv(BlockDriverState *bs, |
ab185921 SH |
2904 | int64_t sector_num, int nb_sectors, QEMUIOVector *qiov) |
2905 | { | |
2906 | /* Perform I/O through a temporary buffer so that users who scribble over | |
2907 | * their read buffer while the operation is in progress do not end up | |
2908 | * modifying the image file. This is critical for zero-copy guest I/O | |
2909 | * where anything might happen inside guest memory. | |
2910 | */ | |
2911 | void *bounce_buffer; | |
2912 | ||
79c053bd | 2913 | BlockDriver *drv = bs->drv; |
ab185921 SH |
2914 | struct iovec iov; |
2915 | QEMUIOVector bounce_qiov; | |
2916 | int64_t cluster_sector_num; | |
2917 | int cluster_nb_sectors; | |
2918 | size_t skip_bytes; | |
2919 | int ret; | |
2920 | ||
2921 | /* Cover entire cluster so no additional backing file I/O is required when | |
2922 | * allocating cluster in the image file. | |
2923 | */ | |
343bded4 PB |
2924 | bdrv_round_to_clusters(bs, sector_num, nb_sectors, |
2925 | &cluster_sector_num, &cluster_nb_sectors); | |
ab185921 | 2926 | |
470c0504 SH |
2927 | trace_bdrv_co_do_copy_on_readv(bs, sector_num, nb_sectors, |
2928 | cluster_sector_num, cluster_nb_sectors); | |
ab185921 SH |
2929 | |
2930 | iov.iov_len = cluster_nb_sectors * BDRV_SECTOR_SIZE; | |
2931 | iov.iov_base = bounce_buffer = qemu_blockalign(bs, iov.iov_len); | |
2932 | qemu_iovec_init_external(&bounce_qiov, &iov, 1); | |
2933 | ||
79c053bd SH |
2934 | ret = drv->bdrv_co_readv(bs, cluster_sector_num, cluster_nb_sectors, |
2935 | &bounce_qiov); | |
ab185921 SH |
2936 | if (ret < 0) { |
2937 | goto err; | |
2938 | } | |
2939 | ||
79c053bd SH |
2940 | if (drv->bdrv_co_write_zeroes && |
2941 | buffer_is_zero(bounce_buffer, iov.iov_len)) { | |
621f0589 | 2942 | ret = bdrv_co_do_write_zeroes(bs, cluster_sector_num, |
aa7bfbff | 2943 | cluster_nb_sectors, 0); |
79c053bd | 2944 | } else { |
f05fa4ad PB |
2945 | /* This does not change the data on the disk, it is not necessary |
2946 | * to flush even in cache=writethrough mode. | |
2947 | */ | |
79c053bd | 2948 | ret = drv->bdrv_co_writev(bs, cluster_sector_num, cluster_nb_sectors, |
ab185921 | 2949 | &bounce_qiov); |
79c053bd SH |
2950 | } |
2951 | ||
ab185921 SH |
2952 | if (ret < 0) { |
2953 | /* It might be okay to ignore write errors for guest requests. If this | |
2954 | * is a deliberate copy-on-read then we don't want to ignore the error. | |
2955 | * Simply report it in all cases. | |
2956 | */ | |
2957 | goto err; | |
2958 | } | |
2959 | ||
2960 | skip_bytes = (sector_num - cluster_sector_num) * BDRV_SECTOR_SIZE; | |
03396148 MT |
2961 | qemu_iovec_from_buf(qiov, 0, bounce_buffer + skip_bytes, |
2962 | nb_sectors * BDRV_SECTOR_SIZE); | |
ab185921 SH |
2963 | |
2964 | err: | |
2965 | qemu_vfree(bounce_buffer); | |
2966 | return ret; | |
2967 | } | |
2968 | ||
c5fbe571 | 2969 | /* |
d0c7f642 KW |
2970 | * Forwards an already correctly aligned request to the BlockDriver. This |
2971 | * handles copy on read and zeroing after EOF; any other features must be | |
2972 | * implemented by the caller. | |
c5fbe571 | 2973 | */ |
d0c7f642 | 2974 | static int coroutine_fn bdrv_aligned_preadv(BlockDriverState *bs, |
65afd211 | 2975 | BdrvTrackedRequest *req, int64_t offset, unsigned int bytes, |
ec746e10 | 2976 | int64_t align, QEMUIOVector *qiov, int flags) |
da1fa91d KW |
2977 | { |
2978 | BlockDriver *drv = bs->drv; | |
dbffbdcf | 2979 | int ret; |
da1fa91d | 2980 | |
d0c7f642 KW |
2981 | int64_t sector_num = offset >> BDRV_SECTOR_BITS; |
2982 | unsigned int nb_sectors = bytes >> BDRV_SECTOR_BITS; | |
da1fa91d | 2983 | |
d0c7f642 KW |
2984 | assert((offset & (BDRV_SECTOR_SIZE - 1)) == 0); |
2985 | assert((bytes & (BDRV_SECTOR_SIZE - 1)) == 0); | |
2986 | ||
2987 | /* Handle Copy on Read and associated serialisation */ | |
470c0504 | 2988 | if (flags & BDRV_REQ_COPY_ON_READ) { |
7327145f KW |
2989 | /* If we touch the same cluster it counts as an overlap. This |
2990 | * guarantees that allocating writes will be serialized and not race | |
2991 | * with each other for the same cluster. For example, in copy-on-read | |
2992 | * it ensures that the CoR read and write operations are atomic and | |
2993 | * guest writes cannot interleave between them. */ | |
2994 | mark_request_serialising(req, bdrv_get_cluster_size(bs)); | |
470c0504 SH |
2995 | } |
2996 | ||
2dbafdc0 | 2997 | wait_serialising_requests(req); |
f4658285 | 2998 | |
470c0504 | 2999 | if (flags & BDRV_REQ_COPY_ON_READ) { |
ab185921 SH |
3000 | int pnum; |
3001 | ||
bdad13b9 | 3002 | ret = bdrv_is_allocated(bs, sector_num, nb_sectors, &pnum); |
ab185921 SH |
3003 | if (ret < 0) { |
3004 | goto out; | |
3005 | } | |
3006 | ||
3007 | if (!ret || pnum != nb_sectors) { | |
470c0504 | 3008 | ret = bdrv_co_do_copy_on_readv(bs, sector_num, nb_sectors, qiov); |
ab185921 SH |
3009 | goto out; |
3010 | } | |
3011 | } | |
3012 | ||
d0c7f642 | 3013 | /* Forward the request to the BlockDriver */ |
893a8f62 MK |
3014 | if (!(bs->zero_beyond_eof && bs->growable)) { |
3015 | ret = drv->bdrv_co_readv(bs, sector_num, nb_sectors, qiov); | |
3016 | } else { | |
3017 | /* Read zeros after EOF of growable BDSes */ | |
3018 | int64_t len, total_sectors, max_nb_sectors; | |
3019 | ||
3020 | len = bdrv_getlength(bs); | |
3021 | if (len < 0) { | |
3022 | ret = len; | |
3023 | goto out; | |
3024 | } | |
3025 | ||
d055a1fe | 3026 | total_sectors = DIV_ROUND_UP(len, BDRV_SECTOR_SIZE); |
5f5bcd80 KW |
3027 | max_nb_sectors = ROUND_UP(MAX(0, total_sectors - sector_num), |
3028 | align >> BDRV_SECTOR_BITS); | |
893a8f62 MK |
3029 | if (max_nb_sectors > 0) { |
3030 | ret = drv->bdrv_co_readv(bs, sector_num, | |
3031 | MIN(nb_sectors, max_nb_sectors), qiov); | |
3032 | } else { | |
3033 | ret = 0; | |
3034 | } | |
3035 | ||
3036 | /* Reading beyond end of file is supposed to produce zeroes */ | |
3037 | if (ret == 0 && total_sectors < sector_num + nb_sectors) { | |
3038 | uint64_t offset = MAX(0, total_sectors - sector_num); | |
3039 | uint64_t bytes = (sector_num + nb_sectors - offset) * | |
3040 | BDRV_SECTOR_SIZE; | |
3041 | qemu_iovec_memset(qiov, offset * BDRV_SECTOR_SIZE, 0, bytes); | |
3042 | } | |
3043 | } | |
ab185921 SH |
3044 | |
3045 | out: | |
dbffbdcf | 3046 | return ret; |
da1fa91d KW |
3047 | } |
3048 | ||
d0c7f642 KW |
3049 | /* |
3050 | * Handle a read request in coroutine context | |
3051 | */ | |
1b0288ae KW |
3052 | static int coroutine_fn bdrv_co_do_preadv(BlockDriverState *bs, |
3053 | int64_t offset, unsigned int bytes, QEMUIOVector *qiov, | |
d0c7f642 KW |
3054 | BdrvRequestFlags flags) |
3055 | { | |
3056 | BlockDriver *drv = bs->drv; | |
65afd211 KW |
3057 | BdrvTrackedRequest req; |
3058 | ||
1b0288ae KW |
3059 | /* TODO Lift BDRV_SECTOR_SIZE restriction in BlockDriver interface */ |
3060 | uint64_t align = MAX(BDRV_SECTOR_SIZE, bs->request_alignment); | |
3061 | uint8_t *head_buf = NULL; | |
3062 | uint8_t *tail_buf = NULL; | |
3063 | QEMUIOVector local_qiov; | |
3064 | bool use_local_qiov = false; | |
d0c7f642 KW |
3065 | int ret; |
3066 | ||
3067 | if (!drv) { | |
3068 | return -ENOMEDIUM; | |
3069 | } | |
1b0288ae | 3070 | if (bdrv_check_byte_request(bs, offset, bytes)) { |
d0c7f642 KW |
3071 | return -EIO; |
3072 | } | |
3073 | ||
3074 | if (bs->copy_on_read) { | |
3075 | flags |= BDRV_REQ_COPY_ON_READ; | |
3076 | } | |
3077 | ||
3078 | /* throttling disk I/O */ | |
3079 | if (bs->io_limits_enabled) { | |
d5103588 | 3080 | bdrv_io_limits_intercept(bs, bytes, false); |
1b0288ae KW |
3081 | } |
3082 | ||
3083 | /* Align read if necessary by padding qiov */ | |
3084 | if (offset & (align - 1)) { | |
3085 | head_buf = qemu_blockalign(bs, align); | |
3086 | qemu_iovec_init(&local_qiov, qiov->niov + 2); | |
3087 | qemu_iovec_add(&local_qiov, head_buf, offset & (align - 1)); | |
3088 | qemu_iovec_concat(&local_qiov, qiov, 0, qiov->size); | |
3089 | use_local_qiov = true; | |
3090 | ||
3091 | bytes += offset & (align - 1); | |
3092 | offset = offset & ~(align - 1); | |
3093 | } | |
3094 | ||
3095 | if ((offset + bytes) & (align - 1)) { | |
3096 | if (!use_local_qiov) { | |
3097 | qemu_iovec_init(&local_qiov, qiov->niov + 1); | |
3098 | qemu_iovec_concat(&local_qiov, qiov, 0, qiov->size); | |
3099 | use_local_qiov = true; | |
3100 | } | |
3101 | tail_buf = qemu_blockalign(bs, align); | |
3102 | qemu_iovec_add(&local_qiov, tail_buf, | |
3103 | align - ((offset + bytes) & (align - 1))); | |
3104 | ||
3105 | bytes = ROUND_UP(bytes, align); | |
3106 | } | |
3107 | ||
65afd211 | 3108 | tracked_request_begin(&req, bs, offset, bytes, false); |
ec746e10 | 3109 | ret = bdrv_aligned_preadv(bs, &req, offset, bytes, align, |
1b0288ae KW |
3110 | use_local_qiov ? &local_qiov : qiov, |
3111 | flags); | |
65afd211 | 3112 | tracked_request_end(&req); |
1b0288ae KW |
3113 | |
3114 | if (use_local_qiov) { | |
3115 | qemu_iovec_destroy(&local_qiov); | |
3116 | qemu_vfree(head_buf); | |
3117 | qemu_vfree(tail_buf); | |
d0c7f642 KW |
3118 | } |
3119 | ||
d0c7f642 KW |
3120 | return ret; |
3121 | } | |
3122 | ||
1b0288ae KW |
3123 | static int coroutine_fn bdrv_co_do_readv(BlockDriverState *bs, |
3124 | int64_t sector_num, int nb_sectors, QEMUIOVector *qiov, | |
3125 | BdrvRequestFlags flags) | |
3126 | { | |
3127 | if (nb_sectors < 0 || nb_sectors > (UINT_MAX >> BDRV_SECTOR_BITS)) { | |
3128 | return -EINVAL; | |
3129 | } | |
3130 | ||
3131 | return bdrv_co_do_preadv(bs, sector_num << BDRV_SECTOR_BITS, | |
3132 | nb_sectors << BDRV_SECTOR_BITS, qiov, flags); | |
3133 | } | |
3134 | ||
c5fbe571 | 3135 | int coroutine_fn bdrv_co_readv(BlockDriverState *bs, int64_t sector_num, |
da1fa91d KW |
3136 | int nb_sectors, QEMUIOVector *qiov) |
3137 | { | |
c5fbe571 | 3138 | trace_bdrv_co_readv(bs, sector_num, nb_sectors); |
da1fa91d | 3139 | |
470c0504 SH |
3140 | return bdrv_co_do_readv(bs, sector_num, nb_sectors, qiov, 0); |
3141 | } | |
3142 | ||
3143 | int coroutine_fn bdrv_co_copy_on_readv(BlockDriverState *bs, | |
3144 | int64_t sector_num, int nb_sectors, QEMUIOVector *qiov) | |
3145 | { | |
3146 | trace_bdrv_co_copy_on_readv(bs, sector_num, nb_sectors); | |
3147 | ||
3148 | return bdrv_co_do_readv(bs, sector_num, nb_sectors, qiov, | |
3149 | BDRV_REQ_COPY_ON_READ); | |
c5fbe571 SH |
3150 | } |
3151 | ||
c31cb707 PL |
3152 | /* if no limit is specified in the BlockLimits use a default |
3153 | * of 32768 512-byte sectors (16 MiB) per request. | |
3154 | */ | |
3155 | #define MAX_WRITE_ZEROES_DEFAULT 32768 | |
3156 | ||
f08f2dda | 3157 | static int coroutine_fn bdrv_co_do_write_zeroes(BlockDriverState *bs, |
aa7bfbff | 3158 | int64_t sector_num, int nb_sectors, BdrvRequestFlags flags) |
f08f2dda SH |
3159 | { |
3160 | BlockDriver *drv = bs->drv; | |
3161 | QEMUIOVector qiov; | |
c31cb707 PL |
3162 | struct iovec iov = {0}; |
3163 | int ret = 0; | |
f08f2dda | 3164 | |
c31cb707 PL |
3165 | int max_write_zeroes = bs->bl.max_write_zeroes ? |
3166 | bs->bl.max_write_zeroes : MAX_WRITE_ZEROES_DEFAULT; | |
621f0589 | 3167 | |
c31cb707 PL |
3168 | while (nb_sectors > 0 && !ret) { |
3169 | int num = nb_sectors; | |
3170 | ||
b8d71c09 PB |
3171 | /* Align request. Block drivers can expect the "bulk" of the request |
3172 | * to be aligned. | |
3173 | */ | |
3174 | if (bs->bl.write_zeroes_alignment | |
3175 | && num > bs->bl.write_zeroes_alignment) { | |
3176 | if (sector_num % bs->bl.write_zeroes_alignment != 0) { | |
3177 | /* Make a small request up to the first aligned sector. */ | |
c31cb707 | 3178 | num = bs->bl.write_zeroes_alignment; |
b8d71c09 PB |
3179 | num -= sector_num % bs->bl.write_zeroes_alignment; |
3180 | } else if ((sector_num + num) % bs->bl.write_zeroes_alignment != 0) { | |
3181 | /* Shorten the request to the last aligned sector. num cannot | |
3182 | * underflow because num > bs->bl.write_zeroes_alignment. | |
3183 | */ | |
3184 | num -= (sector_num + num) % bs->bl.write_zeroes_alignment; | |
c31cb707 | 3185 | } |
621f0589 | 3186 | } |
f08f2dda | 3187 | |
c31cb707 PL |
3188 | /* limit request size */ |
3189 | if (num > max_write_zeroes) { | |
3190 | num = max_write_zeroes; | |
3191 | } | |
3192 | ||
3193 | ret = -ENOTSUP; | |
3194 | /* First try the efficient write zeroes operation */ | |
3195 | if (drv->bdrv_co_write_zeroes) { | |
3196 | ret = drv->bdrv_co_write_zeroes(bs, sector_num, num, flags); | |
3197 | } | |
3198 | ||
3199 | if (ret == -ENOTSUP) { | |
3200 | /* Fall back to bounce buffer if write zeroes is unsupported */ | |
3201 | iov.iov_len = num * BDRV_SECTOR_SIZE; | |
3202 | if (iov.iov_base == NULL) { | |
b8d71c09 PB |
3203 | iov.iov_base = qemu_blockalign(bs, num * BDRV_SECTOR_SIZE); |
3204 | memset(iov.iov_base, 0, num * BDRV_SECTOR_SIZE); | |
c31cb707 PL |
3205 | } |
3206 | qemu_iovec_init_external(&qiov, &iov, 1); | |
f08f2dda | 3207 | |
c31cb707 | 3208 | ret = drv->bdrv_co_writev(bs, sector_num, num, &qiov); |
b8d71c09 PB |
3209 | |
3210 | /* Keep bounce buffer around if it is big enough for all | |
3211 | * all future requests. | |
3212 | */ | |
3213 | if (num < max_write_zeroes) { | |
3214 | qemu_vfree(iov.iov_base); | |
3215 | iov.iov_base = NULL; | |
3216 | } | |
c31cb707 PL |
3217 | } |
3218 | ||
3219 | sector_num += num; | |
3220 | nb_sectors -= num; | |
3221 | } | |
f08f2dda SH |
3222 | |
3223 | qemu_vfree(iov.iov_base); | |
3224 | return ret; | |
3225 | } | |
3226 | ||
c5fbe571 | 3227 | /* |
b404f720 | 3228 | * Forwards an already correctly aligned write request to the BlockDriver. |
c5fbe571 | 3229 | */ |
b404f720 | 3230 | static int coroutine_fn bdrv_aligned_pwritev(BlockDriverState *bs, |
65afd211 KW |
3231 | BdrvTrackedRequest *req, int64_t offset, unsigned int bytes, |
3232 | QEMUIOVector *qiov, int flags) | |
c5fbe571 SH |
3233 | { |
3234 | BlockDriver *drv = bs->drv; | |
28de2dcd | 3235 | bool waited; |
6b7cb247 | 3236 | int ret; |
da1fa91d | 3237 | |
b404f720 KW |
3238 | int64_t sector_num = offset >> BDRV_SECTOR_BITS; |
3239 | unsigned int nb_sectors = bytes >> BDRV_SECTOR_BITS; | |
f4658285 | 3240 | |
b404f720 KW |
3241 | assert((offset & (BDRV_SECTOR_SIZE - 1)) == 0); |
3242 | assert((bytes & (BDRV_SECTOR_SIZE - 1)) == 0); | |
cc0681c4 | 3243 | |
28de2dcd KW |
3244 | waited = wait_serialising_requests(req); |
3245 | assert(!waited || !req->serialising); | |
af91f9a7 KW |
3246 | assert(req->overlap_offset <= offset); |
3247 | assert(offset + bytes <= req->overlap_offset + req->overlap_bytes); | |
244eadef | 3248 | |
65afd211 | 3249 | ret = notifier_with_return_list_notify(&bs->before_write_notifiers, req); |
d616b224 SH |
3250 | |
3251 | if (ret < 0) { | |
3252 | /* Do nothing, write notifier decided to fail this request */ | |
3253 | } else if (flags & BDRV_REQ_ZERO_WRITE) { | |
9e1cb96d | 3254 | BLKDBG_EVENT(bs, BLKDBG_PWRITEV_ZERO); |
aa7bfbff | 3255 | ret = bdrv_co_do_write_zeroes(bs, sector_num, nb_sectors, flags); |
f08f2dda | 3256 | } else { |
9e1cb96d | 3257 | BLKDBG_EVENT(bs, BLKDBG_PWRITEV); |
f08f2dda SH |
3258 | ret = drv->bdrv_co_writev(bs, sector_num, nb_sectors, qiov); |
3259 | } | |
9e1cb96d | 3260 | BLKDBG_EVENT(bs, BLKDBG_PWRITEV_DONE); |
6b7cb247 | 3261 | |
f05fa4ad PB |
3262 | if (ret == 0 && !bs->enable_write_cache) { |
3263 | ret = bdrv_co_flush(bs); | |
3264 | } | |
3265 | ||
e4654d2d | 3266 | bdrv_set_dirty(bs, sector_num, nb_sectors); |
da1fa91d KW |
3267 | |
3268 | if (bs->wr_highest_sector < sector_num + nb_sectors - 1) { | |
3269 | bs->wr_highest_sector = sector_num + nb_sectors - 1; | |
3270 | } | |
df2a6f29 PB |
3271 | if (bs->growable && ret >= 0) { |
3272 | bs->total_sectors = MAX(bs->total_sectors, sector_num + nb_sectors); | |
3273 | } | |
da1fa91d | 3274 | |
6b7cb247 | 3275 | return ret; |
da1fa91d KW |
3276 | } |
3277 | ||
b404f720 KW |
3278 | /* |
3279 | * Handle a write request in coroutine context | |
3280 | */ | |
6601553e KW |
3281 | static int coroutine_fn bdrv_co_do_pwritev(BlockDriverState *bs, |
3282 | int64_t offset, unsigned int bytes, QEMUIOVector *qiov, | |
b404f720 KW |
3283 | BdrvRequestFlags flags) |
3284 | { | |
65afd211 | 3285 | BdrvTrackedRequest req; |
3b8242e0 KW |
3286 | /* TODO Lift BDRV_SECTOR_SIZE restriction in BlockDriver interface */ |
3287 | uint64_t align = MAX(BDRV_SECTOR_SIZE, bs->request_alignment); | |
3288 | uint8_t *head_buf = NULL; | |
3289 | uint8_t *tail_buf = NULL; | |
3290 | QEMUIOVector local_qiov; | |
3291 | bool use_local_qiov = false; | |
b404f720 KW |
3292 | int ret; |
3293 | ||
3294 | if (!bs->drv) { | |
3295 | return -ENOMEDIUM; | |
3296 | } | |
3297 | if (bs->read_only) { | |
3298 | return -EACCES; | |
3299 | } | |
6601553e | 3300 | if (bdrv_check_byte_request(bs, offset, bytes)) { |
b404f720 KW |
3301 | return -EIO; |
3302 | } | |
3303 | ||
b404f720 KW |
3304 | /* throttling disk I/O */ |
3305 | if (bs->io_limits_enabled) { | |
d5103588 | 3306 | bdrv_io_limits_intercept(bs, bytes, true); |
b404f720 KW |
3307 | } |
3308 | ||
3b8242e0 KW |
3309 | /* |
3310 | * Align write if necessary by performing a read-modify-write cycle. | |
3311 | * Pad qiov with the read parts and be sure to have a tracked request not | |
3312 | * only for bdrv_aligned_pwritev, but also for the reads of the RMW cycle. | |
3313 | */ | |
65afd211 | 3314 | tracked_request_begin(&req, bs, offset, bytes, true); |
3b8242e0 KW |
3315 | |
3316 | if (offset & (align - 1)) { | |
3317 | QEMUIOVector head_qiov; | |
3318 | struct iovec head_iov; | |
3319 | ||
3320 | mark_request_serialising(&req, align); | |
3321 | wait_serialising_requests(&req); | |
3322 | ||
3323 | head_buf = qemu_blockalign(bs, align); | |
3324 | head_iov = (struct iovec) { | |
3325 | .iov_base = head_buf, | |
3326 | .iov_len = align, | |
3327 | }; | |
3328 | qemu_iovec_init_external(&head_qiov, &head_iov, 1); | |
3329 | ||
9e1cb96d | 3330 | BLKDBG_EVENT(bs, BLKDBG_PWRITEV_RMW_HEAD); |
3b8242e0 KW |
3331 | ret = bdrv_aligned_preadv(bs, &req, offset & ~(align - 1), align, |
3332 | align, &head_qiov, 0); | |
3333 | if (ret < 0) { | |
3334 | goto fail; | |
3335 | } | |
9e1cb96d | 3336 | BLKDBG_EVENT(bs, BLKDBG_PWRITEV_RMW_AFTER_HEAD); |
3b8242e0 KW |
3337 | |
3338 | qemu_iovec_init(&local_qiov, qiov->niov + 2); | |
3339 | qemu_iovec_add(&local_qiov, head_buf, offset & (align - 1)); | |
3340 | qemu_iovec_concat(&local_qiov, qiov, 0, qiov->size); | |
3341 | use_local_qiov = true; | |
3342 | ||
3343 | bytes += offset & (align - 1); | |
3344 | offset = offset & ~(align - 1); | |
3345 | } | |
3346 | ||
3347 | if ((offset + bytes) & (align - 1)) { | |
3348 | QEMUIOVector tail_qiov; | |
3349 | struct iovec tail_iov; | |
3350 | size_t tail_bytes; | |
28de2dcd | 3351 | bool waited; |
3b8242e0 KW |
3352 | |
3353 | mark_request_serialising(&req, align); | |
28de2dcd KW |
3354 | waited = wait_serialising_requests(&req); |
3355 | assert(!waited || !use_local_qiov); | |
3b8242e0 KW |
3356 | |
3357 | tail_buf = qemu_blockalign(bs, align); | |
3358 | tail_iov = (struct iovec) { | |
3359 | .iov_base = tail_buf, | |
3360 | .iov_len = align, | |
3361 | }; | |
3362 | qemu_iovec_init_external(&tail_qiov, &tail_iov, 1); | |
3363 | ||
9e1cb96d | 3364 | BLKDBG_EVENT(bs, BLKDBG_PWRITEV_RMW_TAIL); |
3b8242e0 KW |
3365 | ret = bdrv_aligned_preadv(bs, &req, (offset + bytes) & ~(align - 1), align, |
3366 | align, &tail_qiov, 0); | |
3367 | if (ret < 0) { | |
3368 | goto fail; | |
3369 | } | |
9e1cb96d | 3370 | BLKDBG_EVENT(bs, BLKDBG_PWRITEV_RMW_AFTER_TAIL); |
3b8242e0 KW |
3371 | |
3372 | if (!use_local_qiov) { | |
3373 | qemu_iovec_init(&local_qiov, qiov->niov + 1); | |
3374 | qemu_iovec_concat(&local_qiov, qiov, 0, qiov->size); | |
3375 | use_local_qiov = true; | |
3376 | } | |
3377 | ||
3378 | tail_bytes = (offset + bytes) & (align - 1); | |
3379 | qemu_iovec_add(&local_qiov, tail_buf + tail_bytes, align - tail_bytes); | |
3380 | ||
3381 | bytes = ROUND_UP(bytes, align); | |
3382 | } | |
3383 | ||
3384 | ret = bdrv_aligned_pwritev(bs, &req, offset, bytes, | |
3385 | use_local_qiov ? &local_qiov : qiov, | |
3386 | flags); | |
3387 | ||
3388 | fail: | |
65afd211 | 3389 | tracked_request_end(&req); |
b404f720 | 3390 | |
3b8242e0 KW |
3391 | if (use_local_qiov) { |
3392 | qemu_iovec_destroy(&local_qiov); | |
3b8242e0 | 3393 | } |
99c4a85c KW |
3394 | qemu_vfree(head_buf); |
3395 | qemu_vfree(tail_buf); | |
3b8242e0 | 3396 | |
b404f720 KW |
3397 | return ret; |
3398 | } | |
3399 | ||
6601553e KW |
3400 | static int coroutine_fn bdrv_co_do_writev(BlockDriverState *bs, |
3401 | int64_t sector_num, int nb_sectors, QEMUIOVector *qiov, | |
3402 | BdrvRequestFlags flags) | |
3403 | { | |
3404 | if (nb_sectors < 0 || nb_sectors > (INT_MAX >> BDRV_SECTOR_BITS)) { | |
3405 | return -EINVAL; | |
3406 | } | |
3407 | ||
3408 | return bdrv_co_do_pwritev(bs, sector_num << BDRV_SECTOR_BITS, | |
3409 | nb_sectors << BDRV_SECTOR_BITS, qiov, flags); | |
3410 | } | |
3411 | ||
c5fbe571 SH |
3412 | int coroutine_fn bdrv_co_writev(BlockDriverState *bs, int64_t sector_num, |
3413 | int nb_sectors, QEMUIOVector *qiov) | |
3414 | { | |
3415 | trace_bdrv_co_writev(bs, sector_num, nb_sectors); | |
3416 | ||
f08f2dda SH |
3417 | return bdrv_co_do_writev(bs, sector_num, nb_sectors, qiov, 0); |
3418 | } | |
3419 | ||
3420 | int coroutine_fn bdrv_co_write_zeroes(BlockDriverState *bs, | |
aa7bfbff PL |
3421 | int64_t sector_num, int nb_sectors, |
3422 | BdrvRequestFlags flags) | |
f08f2dda | 3423 | { |
94d6ff21 | 3424 | trace_bdrv_co_write_zeroes(bs, sector_num, nb_sectors, flags); |
f08f2dda | 3425 | |
d32f35cb PL |
3426 | if (!(bs->open_flags & BDRV_O_UNMAP)) { |
3427 | flags &= ~BDRV_REQ_MAY_UNMAP; | |
3428 | } | |
3429 | ||
f08f2dda | 3430 | return bdrv_co_do_writev(bs, sector_num, nb_sectors, NULL, |
aa7bfbff | 3431 | BDRV_REQ_ZERO_WRITE | flags); |
c5fbe571 SH |
3432 | } |
3433 | ||
83f64091 FB |
3434 | /** |
3435 | * Truncate file to 'offset' bytes (needed only for file protocols) | |
3436 | */ | |
3437 | int bdrv_truncate(BlockDriverState *bs, int64_t offset) | |
3438 | { | |
3439 | BlockDriver *drv = bs->drv; | |
51762288 | 3440 | int ret; |
83f64091 | 3441 | if (!drv) |
19cb3738 | 3442 | return -ENOMEDIUM; |
83f64091 FB |
3443 | if (!drv->bdrv_truncate) |
3444 | return -ENOTSUP; | |
59f2689d NS |
3445 | if (bs->read_only) |
3446 | return -EACCES; | |
8591675f MT |
3447 | if (bdrv_in_use(bs)) |
3448 | return -EBUSY; | |
51762288 SH |
3449 | ret = drv->bdrv_truncate(bs, offset); |
3450 | if (ret == 0) { | |
3451 | ret = refresh_total_sectors(bs, offset >> BDRV_SECTOR_BITS); | |
145feb17 | 3452 | bdrv_dev_resize_cb(bs); |
51762288 SH |
3453 | } |
3454 | return ret; | |
83f64091 FB |
3455 | } |
3456 | ||
4a1d5e1f FZ |
3457 | /** |
3458 | * Length of a allocated file in bytes. Sparse files are counted by actual | |
3459 | * allocated space. Return < 0 if error or unknown. | |
3460 | */ | |
3461 | int64_t bdrv_get_allocated_file_size(BlockDriverState *bs) | |
3462 | { | |
3463 | BlockDriver *drv = bs->drv; | |
3464 | if (!drv) { | |
3465 | return -ENOMEDIUM; | |
3466 | } | |
3467 | if (drv->bdrv_get_allocated_file_size) { | |
3468 | return drv->bdrv_get_allocated_file_size(bs); | |
3469 | } | |
3470 | if (bs->file) { | |
3471 | return bdrv_get_allocated_file_size(bs->file); | |
3472 | } | |
3473 | return -ENOTSUP; | |
3474 | } | |
3475 | ||
83f64091 FB |
3476 | /** |
3477 | * Length of a file in bytes. Return < 0 if error or unknown. | |
3478 | */ | |
3479 | int64_t bdrv_getlength(BlockDriverState *bs) | |
3480 | { | |
3481 | BlockDriver *drv = bs->drv; | |
3482 | if (!drv) | |
19cb3738 | 3483 | return -ENOMEDIUM; |
51762288 | 3484 | |
b94a2610 KW |
3485 | if (drv->has_variable_length) { |
3486 | int ret = refresh_total_sectors(bs, bs->total_sectors); | |
3487 | if (ret < 0) { | |
3488 | return ret; | |
46a4e4e6 | 3489 | } |
83f64091 | 3490 | } |
46a4e4e6 | 3491 | return bs->total_sectors * BDRV_SECTOR_SIZE; |
fc01f7e7 FB |
3492 | } |
3493 | ||
19cb3738 | 3494 | /* return 0 as number of sectors if no device present or error */ |
96b8f136 | 3495 | void bdrv_get_geometry(BlockDriverState *bs, uint64_t *nb_sectors_ptr) |
fc01f7e7 | 3496 | { |
19cb3738 FB |
3497 | int64_t length; |
3498 | length = bdrv_getlength(bs); | |
3499 | if (length < 0) | |
3500 | length = 0; | |
3501 | else | |
6ea44308 | 3502 | length = length >> BDRV_SECTOR_BITS; |
19cb3738 | 3503 | *nb_sectors_ptr = length; |
fc01f7e7 | 3504 | } |
cf98951b | 3505 | |
ff06f5f3 PB |
3506 | void bdrv_set_on_error(BlockDriverState *bs, BlockdevOnError on_read_error, |
3507 | BlockdevOnError on_write_error) | |
abd7f68d MA |
3508 | { |
3509 | bs->on_read_error = on_read_error; | |
3510 | bs->on_write_error = on_write_error; | |
3511 | } | |
3512 | ||
1ceee0d5 | 3513 | BlockdevOnError bdrv_get_on_error(BlockDriverState *bs, bool is_read) |
abd7f68d MA |
3514 | { |
3515 | return is_read ? bs->on_read_error : bs->on_write_error; | |
3516 | } | |
3517 | ||
3e1caa5f PB |
3518 | BlockErrorAction bdrv_get_error_action(BlockDriverState *bs, bool is_read, int error) |
3519 | { | |
3520 | BlockdevOnError on_err = is_read ? bs->on_read_error : bs->on_write_error; | |
3521 | ||
3522 | switch (on_err) { | |
3523 | case BLOCKDEV_ON_ERROR_ENOSPC: | |
3524 | return (error == ENOSPC) ? BDRV_ACTION_STOP : BDRV_ACTION_REPORT; | |
3525 | case BLOCKDEV_ON_ERROR_STOP: | |
3526 | return BDRV_ACTION_STOP; | |
3527 | case BLOCKDEV_ON_ERROR_REPORT: | |
3528 | return BDRV_ACTION_REPORT; | |
3529 | case BLOCKDEV_ON_ERROR_IGNORE: | |
3530 | return BDRV_ACTION_IGNORE; | |
3531 | default: | |
3532 | abort(); | |
3533 | } | |
3534 | } | |
3535 | ||
3536 | /* This is done by device models because, while the block layer knows | |
3537 | * about the error, it does not know whether an operation comes from | |
3538 | * the device or the block layer (from a job, for example). | |
3539 | */ | |
3540 | void bdrv_error_action(BlockDriverState *bs, BlockErrorAction action, | |
3541 | bool is_read, int error) | |
3542 | { | |
3543 | assert(error >= 0); | |
32c81a4a | 3544 | bdrv_emit_qmp_error_event(bs, QEVENT_BLOCK_IO_ERROR, action, is_read); |
3e1caa5f PB |
3545 | if (action == BDRV_ACTION_STOP) { |
3546 | vm_stop(RUN_STATE_IO_ERROR); | |
3547 | bdrv_iostatus_set_err(bs, error); | |
3548 | } | |
3549 | } | |
3550 | ||
b338082b FB |
3551 | int bdrv_is_read_only(BlockDriverState *bs) |
3552 | { | |
3553 | return bs->read_only; | |
3554 | } | |
3555 | ||
985a03b0 TS |
3556 | int bdrv_is_sg(BlockDriverState *bs) |
3557 | { | |
3558 | return bs->sg; | |
3559 | } | |
3560 | ||
e900a7b7 CH |
3561 | int bdrv_enable_write_cache(BlockDriverState *bs) |
3562 | { | |
3563 | return bs->enable_write_cache; | |
3564 | } | |
3565 | ||
425b0148 PB |
3566 | void bdrv_set_enable_write_cache(BlockDriverState *bs, bool wce) |
3567 | { | |
3568 | bs->enable_write_cache = wce; | |
55b110f2 JC |
3569 | |
3570 | /* so a reopen() will preserve wce */ | |
3571 | if (wce) { | |
3572 | bs->open_flags |= BDRV_O_CACHE_WB; | |
3573 | } else { | |
3574 | bs->open_flags &= ~BDRV_O_CACHE_WB; | |
3575 | } | |
425b0148 PB |
3576 | } |
3577 | ||
ea2384d3 FB |
3578 | int bdrv_is_encrypted(BlockDriverState *bs) |
3579 | { | |
3580 | if (bs->backing_hd && bs->backing_hd->encrypted) | |
3581 | return 1; | |
3582 | return bs->encrypted; | |
3583 | } | |
3584 | ||
c0f4ce77 AL |
3585 | int bdrv_key_required(BlockDriverState *bs) |
3586 | { | |
3587 | BlockDriverState *backing_hd = bs->backing_hd; | |
3588 | ||
3589 | if (backing_hd && backing_hd->encrypted && !backing_hd->valid_key) | |
3590 | return 1; | |
3591 | return (bs->encrypted && !bs->valid_key); | |
3592 | } | |
3593 | ||
ea2384d3 FB |
3594 | int bdrv_set_key(BlockDriverState *bs, const char *key) |
3595 | { | |
3596 | int ret; | |
3597 | if (bs->backing_hd && bs->backing_hd->encrypted) { | |
3598 | ret = bdrv_set_key(bs->backing_hd, key); | |
3599 | if (ret < 0) | |
3600 | return ret; | |
3601 | if (!bs->encrypted) | |
3602 | return 0; | |
3603 | } | |
fd04a2ae SH |
3604 | if (!bs->encrypted) { |
3605 | return -EINVAL; | |
3606 | } else if (!bs->drv || !bs->drv->bdrv_set_key) { | |
3607 | return -ENOMEDIUM; | |
3608 | } | |
c0f4ce77 | 3609 | ret = bs->drv->bdrv_set_key(bs, key); |
bb5fc20f AL |
3610 | if (ret < 0) { |
3611 | bs->valid_key = 0; | |
3612 | } else if (!bs->valid_key) { | |
3613 | bs->valid_key = 1; | |
3614 | /* call the change callback now, we skipped it on open */ | |
7d4b4ba5 | 3615 | bdrv_dev_change_media_cb(bs, true); |
bb5fc20f | 3616 | } |
c0f4ce77 | 3617 | return ret; |
ea2384d3 FB |
3618 | } |
3619 | ||
f8d6bba1 | 3620 | const char *bdrv_get_format_name(BlockDriverState *bs) |
ea2384d3 | 3621 | { |
f8d6bba1 | 3622 | return bs->drv ? bs->drv->format_name : NULL; |
ea2384d3 FB |
3623 | } |
3624 | ||
5fafdf24 | 3625 | void bdrv_iterate_format(void (*it)(void *opaque, const char *name), |
ea2384d3 FB |
3626 | void *opaque) |
3627 | { | |
3628 | BlockDriver *drv; | |
e855e4fb JC |
3629 | int count = 0; |
3630 | const char **formats = NULL; | |
ea2384d3 | 3631 | |
8a22f02a | 3632 | QLIST_FOREACH(drv, &bdrv_drivers, list) { |
e855e4fb JC |
3633 | if (drv->format_name) { |
3634 | bool found = false; | |
3635 | int i = count; | |
3636 | while (formats && i && !found) { | |
3637 | found = !strcmp(formats[--i], drv->format_name); | |
3638 | } | |
3639 | ||
3640 | if (!found) { | |
3641 | formats = g_realloc(formats, (count + 1) * sizeof(char *)); | |
3642 | formats[count++] = drv->format_name; | |
3643 | it(opaque, drv->format_name); | |
3644 | } | |
3645 | } | |
ea2384d3 | 3646 | } |
e855e4fb | 3647 | g_free(formats); |
ea2384d3 FB |
3648 | } |
3649 | ||
dc364f4c | 3650 | /* This function is to find block backend bs */ |
b338082b FB |
3651 | BlockDriverState *bdrv_find(const char *name) |
3652 | { | |
3653 | BlockDriverState *bs; | |
3654 | ||
dc364f4c | 3655 | QTAILQ_FOREACH(bs, &bdrv_states, device_list) { |
1b7bdbc1 | 3656 | if (!strcmp(name, bs->device_name)) { |
b338082b | 3657 | return bs; |
1b7bdbc1 | 3658 | } |
b338082b FB |
3659 | } |
3660 | return NULL; | |
3661 | } | |
3662 | ||
dc364f4c BC |
3663 | /* This function is to find a node in the bs graph */ |
3664 | BlockDriverState *bdrv_find_node(const char *node_name) | |
3665 | { | |
3666 | BlockDriverState *bs; | |
3667 | ||
3668 | assert(node_name); | |
3669 | ||
3670 | QTAILQ_FOREACH(bs, &graph_bdrv_states, node_list) { | |
3671 | if (!strcmp(node_name, bs->node_name)) { | |
3672 | return bs; | |
3673 | } | |
3674 | } | |
3675 | return NULL; | |
3676 | } | |
3677 | ||
c13163fb BC |
3678 | /* Put this QMP function here so it can access the static graph_bdrv_states. */ |
3679 | BlockDeviceInfoList *bdrv_named_nodes_list(void) | |
3680 | { | |
3681 | BlockDeviceInfoList *list, *entry; | |
3682 | BlockDriverState *bs; | |
3683 | ||
3684 | list = NULL; | |
3685 | QTAILQ_FOREACH(bs, &graph_bdrv_states, node_list) { | |
3686 | entry = g_malloc0(sizeof(*entry)); | |
3687 | entry->value = bdrv_block_device_info(bs); | |
3688 | entry->next = list; | |
3689 | list = entry; | |
3690 | } | |
3691 | ||
3692 | return list; | |
3693 | } | |
3694 | ||
12d3ba82 BC |
3695 | BlockDriverState *bdrv_lookup_bs(const char *device, |
3696 | const char *node_name, | |
3697 | Error **errp) | |
3698 | { | |
3699 | BlockDriverState *bs = NULL; | |
3700 | ||
12d3ba82 BC |
3701 | if (device) { |
3702 | bs = bdrv_find(device); | |
3703 | ||
dd67fa50 BC |
3704 | if (bs) { |
3705 | return bs; | |
12d3ba82 | 3706 | } |
12d3ba82 BC |
3707 | } |
3708 | ||
dd67fa50 BC |
3709 | if (node_name) { |
3710 | bs = bdrv_find_node(node_name); | |
12d3ba82 | 3711 | |
dd67fa50 BC |
3712 | if (bs) { |
3713 | return bs; | |
3714 | } | |
12d3ba82 BC |
3715 | } |
3716 | ||
dd67fa50 BC |
3717 | error_setg(errp, "Cannot find device=%s nor node_name=%s", |
3718 | device ? device : "", | |
3719 | node_name ? node_name : ""); | |
3720 | return NULL; | |
12d3ba82 BC |
3721 | } |
3722 | ||
2f399b0a MA |
3723 | BlockDriverState *bdrv_next(BlockDriverState *bs) |
3724 | { | |
3725 | if (!bs) { | |
3726 | return QTAILQ_FIRST(&bdrv_states); | |
3727 | } | |
dc364f4c | 3728 | return QTAILQ_NEXT(bs, device_list); |
2f399b0a MA |
3729 | } |
3730 | ||
51de9760 | 3731 | void bdrv_iterate(void (*it)(void *opaque, BlockDriverState *bs), void *opaque) |
81d0912d FB |
3732 | { |
3733 | BlockDriverState *bs; | |
3734 | ||
dc364f4c | 3735 | QTAILQ_FOREACH(bs, &bdrv_states, device_list) { |
51de9760 | 3736 | it(opaque, bs); |
81d0912d FB |
3737 | } |
3738 | } | |
3739 | ||
ea2384d3 FB |
3740 | const char *bdrv_get_device_name(BlockDriverState *bs) |
3741 | { | |
3742 | return bs->device_name; | |
3743 | } | |
3744 | ||
c8433287 MA |
3745 | int bdrv_get_flags(BlockDriverState *bs) |
3746 | { | |
3747 | return bs->open_flags; | |
3748 | } | |
3749 | ||
f0f0fdfe | 3750 | int bdrv_flush_all(void) |
c6ca28d6 AL |
3751 | { |
3752 | BlockDriverState *bs; | |
f0f0fdfe | 3753 | int result = 0; |
c6ca28d6 | 3754 | |
dc364f4c | 3755 | QTAILQ_FOREACH(bs, &bdrv_states, device_list) { |
f0f0fdfe KW |
3756 | int ret = bdrv_flush(bs); |
3757 | if (ret < 0 && !result) { | |
3758 | result = ret; | |
3759 | } | |
1b7bdbc1 | 3760 | } |
f0f0fdfe KW |
3761 | |
3762 | return result; | |
c6ca28d6 AL |
3763 | } |
3764 | ||
3ac21627 PL |
3765 | int bdrv_has_zero_init_1(BlockDriverState *bs) |
3766 | { | |
3767 | return 1; | |
3768 | } | |
3769 | ||
f2feebbd KW |
3770 | int bdrv_has_zero_init(BlockDriverState *bs) |
3771 | { | |
3772 | assert(bs->drv); | |
3773 | ||
11212d8f PB |
3774 | /* If BS is a copy on write image, it is initialized to |
3775 | the contents of the base image, which may not be zeroes. */ | |
3776 | if (bs->backing_hd) { | |
3777 | return 0; | |
3778 | } | |
336c1c12 KW |
3779 | if (bs->drv->bdrv_has_zero_init) { |
3780 | return bs->drv->bdrv_has_zero_init(bs); | |
f2feebbd KW |
3781 | } |
3782 | ||
3ac21627 PL |
3783 | /* safe default */ |
3784 | return 0; | |
f2feebbd KW |
3785 | } |
3786 | ||
4ce78691 PL |
3787 | bool bdrv_unallocated_blocks_are_zero(BlockDriverState *bs) |
3788 | { | |
3789 | BlockDriverInfo bdi; | |
3790 | ||
3791 | if (bs->backing_hd) { | |
3792 | return false; | |
3793 | } | |
3794 | ||
3795 | if (bdrv_get_info(bs, &bdi) == 0) { | |
3796 | return bdi.unallocated_blocks_are_zero; | |
3797 | } | |
3798 | ||
3799 | return false; | |
3800 | } | |
3801 | ||
3802 | bool bdrv_can_write_zeroes_with_unmap(BlockDriverState *bs) | |
3803 | { | |
3804 | BlockDriverInfo bdi; | |
3805 | ||
3806 | if (bs->backing_hd || !(bs->open_flags & BDRV_O_UNMAP)) { | |
3807 | return false; | |
3808 | } | |
3809 | ||
3810 | if (bdrv_get_info(bs, &bdi) == 0) { | |
3811 | return bdi.can_write_zeroes_with_unmap; | |
3812 | } | |
3813 | ||
3814 | return false; | |
3815 | } | |
3816 | ||
b6b8a333 | 3817 | typedef struct BdrvCoGetBlockStatusData { |
376ae3f1 | 3818 | BlockDriverState *bs; |
b35b2bba | 3819 | BlockDriverState *base; |
376ae3f1 SH |
3820 | int64_t sector_num; |
3821 | int nb_sectors; | |
3822 | int *pnum; | |
b6b8a333 | 3823 | int64_t ret; |
376ae3f1 | 3824 | bool done; |
b6b8a333 | 3825 | } BdrvCoGetBlockStatusData; |
376ae3f1 | 3826 | |
f58c7b35 TS |
3827 | /* |
3828 | * Returns true iff the specified sector is present in the disk image. Drivers | |
3829 | * not implementing the functionality are assumed to not support backing files, | |
3830 | * hence all their sectors are reported as allocated. | |
3831 | * | |
bd9533e3 SH |
3832 | * If 'sector_num' is beyond the end of the disk image the return value is 0 |
3833 | * and 'pnum' is set to 0. | |
3834 | * | |
f58c7b35 TS |
3835 | * 'pnum' is set to the number of sectors (including and immediately following |
3836 | * the specified sector) that are known to be in the same | |
3837 | * allocated/unallocated state. | |
3838 | * | |
bd9533e3 SH |
3839 | * 'nb_sectors' is the max value 'pnum' should be set to. If nb_sectors goes |
3840 | * beyond the end of the disk image it will be clamped. | |
f58c7b35 | 3841 | */ |
b6b8a333 PB |
3842 | static int64_t coroutine_fn bdrv_co_get_block_status(BlockDriverState *bs, |
3843 | int64_t sector_num, | |
3844 | int nb_sectors, int *pnum) | |
f58c7b35 | 3845 | { |
617ccb46 | 3846 | int64_t length; |
bd9533e3 | 3847 | int64_t n; |
5daa74a6 | 3848 | int64_t ret, ret2; |
bd9533e3 | 3849 | |
617ccb46 PB |
3850 | length = bdrv_getlength(bs); |
3851 | if (length < 0) { | |
3852 | return length; | |
3853 | } | |
3854 | ||
3855 | if (sector_num >= (length >> BDRV_SECTOR_BITS)) { | |
bd9533e3 SH |
3856 | *pnum = 0; |
3857 | return 0; | |
3858 | } | |
3859 | ||
3860 | n = bs->total_sectors - sector_num; | |
3861 | if (n < nb_sectors) { | |
3862 | nb_sectors = n; | |
3863 | } | |
3864 | ||
b6b8a333 | 3865 | if (!bs->drv->bdrv_co_get_block_status) { |
bd9533e3 | 3866 | *pnum = nb_sectors; |
918e92d7 PB |
3867 | ret = BDRV_BLOCK_DATA; |
3868 | if (bs->drv->protocol_name) { | |
3869 | ret |= BDRV_BLOCK_OFFSET_VALID | (sector_num * BDRV_SECTOR_SIZE); | |
3870 | } | |
3871 | return ret; | |
f58c7b35 | 3872 | } |
6aebab14 | 3873 | |
415b5b01 PB |
3874 | ret = bs->drv->bdrv_co_get_block_status(bs, sector_num, nb_sectors, pnum); |
3875 | if (ret < 0) { | |
3e0a233d | 3876 | *pnum = 0; |
415b5b01 PB |
3877 | return ret; |
3878 | } | |
3879 | ||
92bc50a5 PL |
3880 | if (ret & BDRV_BLOCK_RAW) { |
3881 | assert(ret & BDRV_BLOCK_OFFSET_VALID); | |
3882 | return bdrv_get_block_status(bs->file, ret >> BDRV_SECTOR_BITS, | |
3883 | *pnum, pnum); | |
3884 | } | |
3885 | ||
c3d86884 PL |
3886 | if (!(ret & BDRV_BLOCK_DATA) && !(ret & BDRV_BLOCK_ZERO)) { |
3887 | if (bdrv_unallocated_blocks_are_zero(bs)) { | |
f0ad5712 | 3888 | ret |= BDRV_BLOCK_ZERO; |
1f9db224 | 3889 | } else if (bs->backing_hd) { |
f0ad5712 PB |
3890 | BlockDriverState *bs2 = bs->backing_hd; |
3891 | int64_t length2 = bdrv_getlength(bs2); | |
3892 | if (length2 >= 0 && sector_num >= (length2 >> BDRV_SECTOR_BITS)) { | |
3893 | ret |= BDRV_BLOCK_ZERO; | |
3894 | } | |
3895 | } | |
415b5b01 | 3896 | } |
5daa74a6 PB |
3897 | |
3898 | if (bs->file && | |
3899 | (ret & BDRV_BLOCK_DATA) && !(ret & BDRV_BLOCK_ZERO) && | |
3900 | (ret & BDRV_BLOCK_OFFSET_VALID)) { | |
3901 | ret2 = bdrv_co_get_block_status(bs->file, ret >> BDRV_SECTOR_BITS, | |
3902 | *pnum, pnum); | |
3903 | if (ret2 >= 0) { | |
3904 | /* Ignore errors. This is just providing extra information, it | |
3905 | * is useful but not necessary. | |
3906 | */ | |
3907 | ret |= (ret2 & BDRV_BLOCK_ZERO); | |
3908 | } | |
3909 | } | |
3910 | ||
415b5b01 | 3911 | return ret; |
060f51c9 SH |
3912 | } |
3913 | ||
b6b8a333 PB |
3914 | /* Coroutine wrapper for bdrv_get_block_status() */ |
3915 | static void coroutine_fn bdrv_get_block_status_co_entry(void *opaque) | |
060f51c9 | 3916 | { |
b6b8a333 | 3917 | BdrvCoGetBlockStatusData *data = opaque; |
060f51c9 SH |
3918 | BlockDriverState *bs = data->bs; |
3919 | ||
b6b8a333 PB |
3920 | data->ret = bdrv_co_get_block_status(bs, data->sector_num, data->nb_sectors, |
3921 | data->pnum); | |
060f51c9 SH |
3922 | data->done = true; |
3923 | } | |
3924 | ||
3925 | /* | |
b6b8a333 | 3926 | * Synchronous wrapper around bdrv_co_get_block_status(). |
060f51c9 | 3927 | * |
b6b8a333 | 3928 | * See bdrv_co_get_block_status() for details. |
060f51c9 | 3929 | */ |
b6b8a333 PB |
3930 | int64_t bdrv_get_block_status(BlockDriverState *bs, int64_t sector_num, |
3931 | int nb_sectors, int *pnum) | |
060f51c9 | 3932 | { |
6aebab14 | 3933 | Coroutine *co; |
b6b8a333 | 3934 | BdrvCoGetBlockStatusData data = { |
6aebab14 SH |
3935 | .bs = bs, |
3936 | .sector_num = sector_num, | |
3937 | .nb_sectors = nb_sectors, | |
3938 | .pnum = pnum, | |
3939 | .done = false, | |
3940 | }; | |
3941 | ||
bdad13b9 PB |
3942 | if (qemu_in_coroutine()) { |
3943 | /* Fast-path if already in coroutine context */ | |
b6b8a333 | 3944 | bdrv_get_block_status_co_entry(&data); |
bdad13b9 | 3945 | } else { |
b6b8a333 | 3946 | co = qemu_coroutine_create(bdrv_get_block_status_co_entry); |
bdad13b9 PB |
3947 | qemu_coroutine_enter(co, &data); |
3948 | while (!data.done) { | |
3949 | qemu_aio_wait(); | |
3950 | } | |
6aebab14 SH |
3951 | } |
3952 | return data.ret; | |
f58c7b35 TS |
3953 | } |
3954 | ||
b6b8a333 PB |
3955 | int coroutine_fn bdrv_is_allocated(BlockDriverState *bs, int64_t sector_num, |
3956 | int nb_sectors, int *pnum) | |
3957 | { | |
4333bb71 PB |
3958 | int64_t ret = bdrv_get_block_status(bs, sector_num, nb_sectors, pnum); |
3959 | if (ret < 0) { | |
3960 | return ret; | |
3961 | } | |
3962 | return | |
3963 | (ret & BDRV_BLOCK_DATA) || | |
3964 | ((ret & BDRV_BLOCK_ZERO) && !bdrv_has_zero_init(bs)); | |
b6b8a333 PB |
3965 | } |
3966 | ||
188a7bbf PB |
3967 | /* |
3968 | * Given an image chain: ... -> [BASE] -> [INTER1] -> [INTER2] -> [TOP] | |
3969 | * | |
3970 | * Return true if the given sector is allocated in any image between | |
3971 | * BASE and TOP (inclusive). BASE can be NULL to check if the given | |
3972 | * sector is allocated in any image of the chain. Return false otherwise. | |
3973 | * | |
3974 | * 'pnum' is set to the number of sectors (including and immediately following | |
3975 | * the specified sector) that are known to be in the same | |
3976 | * allocated/unallocated state. | |
3977 | * | |
3978 | */ | |
4f578637 PB |
3979 | int bdrv_is_allocated_above(BlockDriverState *top, |
3980 | BlockDriverState *base, | |
3981 | int64_t sector_num, | |
3982 | int nb_sectors, int *pnum) | |
188a7bbf PB |
3983 | { |
3984 | BlockDriverState *intermediate; | |
3985 | int ret, n = nb_sectors; | |
3986 | ||
3987 | intermediate = top; | |
3988 | while (intermediate && intermediate != base) { | |
3989 | int pnum_inter; | |
bdad13b9 PB |
3990 | ret = bdrv_is_allocated(intermediate, sector_num, nb_sectors, |
3991 | &pnum_inter); | |
188a7bbf PB |
3992 | if (ret < 0) { |
3993 | return ret; | |
3994 | } else if (ret) { | |
3995 | *pnum = pnum_inter; | |
3996 | return 1; | |
3997 | } | |
3998 | ||
3999 | /* | |
4000 | * [sector_num, nb_sectors] is unallocated on top but intermediate | |
4001 | * might have | |
4002 | * | |
4003 | * [sector_num+x, nr_sectors] allocated. | |
4004 | */ | |
63ba17d3 VI |
4005 | if (n > pnum_inter && |
4006 | (intermediate == top || | |
4007 | sector_num + pnum_inter < intermediate->total_sectors)) { | |
188a7bbf PB |
4008 | n = pnum_inter; |
4009 | } | |
4010 | ||
4011 | intermediate = intermediate->backing_hd; | |
4012 | } | |
4013 | ||
4014 | *pnum = n; | |
4015 | return 0; | |
4016 | } | |
4017 | ||
045df330 AL |
4018 | const char *bdrv_get_encrypted_filename(BlockDriverState *bs) |
4019 | { | |
4020 | if (bs->backing_hd && bs->backing_hd->encrypted) | |
4021 | return bs->backing_file; | |
4022 | else if (bs->encrypted) | |
4023 | return bs->filename; | |
4024 | else | |
4025 | return NULL; | |
4026 | } | |
4027 | ||
5fafdf24 | 4028 | void bdrv_get_backing_filename(BlockDriverState *bs, |
83f64091 FB |
4029 | char *filename, int filename_size) |
4030 | { | |
3574c608 | 4031 | pstrcpy(filename, filename_size, bs->backing_file); |
83f64091 FB |
4032 | } |
4033 | ||
5fafdf24 | 4034 | int bdrv_write_compressed(BlockDriverState *bs, int64_t sector_num, |
faea38e7 FB |
4035 | const uint8_t *buf, int nb_sectors) |
4036 | { | |
4037 | BlockDriver *drv = bs->drv; | |
4038 | if (!drv) | |
19cb3738 | 4039 | return -ENOMEDIUM; |
faea38e7 FB |
4040 | if (!drv->bdrv_write_compressed) |
4041 | return -ENOTSUP; | |
fbb7b4e0 KW |
4042 | if (bdrv_check_request(bs, sector_num, nb_sectors)) |
4043 | return -EIO; | |
a55eb92c | 4044 | |
e4654d2d | 4045 | assert(QLIST_EMPTY(&bs->dirty_bitmaps)); |
a55eb92c | 4046 | |
faea38e7 FB |
4047 | return drv->bdrv_write_compressed(bs, sector_num, buf, nb_sectors); |
4048 | } | |
3b46e624 | 4049 | |
faea38e7 FB |
4050 | int bdrv_get_info(BlockDriverState *bs, BlockDriverInfo *bdi) |
4051 | { | |
4052 | BlockDriver *drv = bs->drv; | |
4053 | if (!drv) | |
19cb3738 | 4054 | return -ENOMEDIUM; |
faea38e7 FB |
4055 | if (!drv->bdrv_get_info) |
4056 | return -ENOTSUP; | |
4057 | memset(bdi, 0, sizeof(*bdi)); | |
4058 | return drv->bdrv_get_info(bs, bdi); | |
4059 | } | |
4060 | ||
eae041fe HR |
4061 | ImageInfoSpecific *bdrv_get_specific_info(BlockDriverState *bs) |
4062 | { | |
4063 | BlockDriver *drv = bs->drv; | |
4064 | if (drv && drv->bdrv_get_specific_info) { | |
4065 | return drv->bdrv_get_specific_info(bs); | |
4066 | } | |
4067 | return NULL; | |
4068 | } | |
4069 | ||
45566e9c CH |
4070 | int bdrv_save_vmstate(BlockDriverState *bs, const uint8_t *buf, |
4071 | int64_t pos, int size) | |
cf8074b3 KW |
4072 | { |
4073 | QEMUIOVector qiov; | |
4074 | struct iovec iov = { | |
4075 | .iov_base = (void *) buf, | |
4076 | .iov_len = size, | |
4077 | }; | |
4078 | ||
4079 | qemu_iovec_init_external(&qiov, &iov, 1); | |
4080 | return bdrv_writev_vmstate(bs, &qiov, pos); | |
4081 | } | |
4082 | ||
4083 | int bdrv_writev_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos) | |
178e08a5 AL |
4084 | { |
4085 | BlockDriver *drv = bs->drv; | |
cf8074b3 KW |
4086 | |
4087 | if (!drv) { | |
178e08a5 | 4088 | return -ENOMEDIUM; |
cf8074b3 KW |
4089 | } else if (drv->bdrv_save_vmstate) { |
4090 | return drv->bdrv_save_vmstate(bs, qiov, pos); | |
4091 | } else if (bs->file) { | |
4092 | return bdrv_writev_vmstate(bs->file, qiov, pos); | |
4093 | } | |
4094 | ||
7cdb1f6d | 4095 | return -ENOTSUP; |
178e08a5 AL |
4096 | } |
4097 | ||
45566e9c CH |
4098 | int bdrv_load_vmstate(BlockDriverState *bs, uint8_t *buf, |
4099 | int64_t pos, int size) | |
178e08a5 AL |
4100 | { |
4101 | BlockDriver *drv = bs->drv; | |
4102 | if (!drv) | |
4103 | return -ENOMEDIUM; | |
7cdb1f6d MK |
4104 | if (drv->bdrv_load_vmstate) |
4105 | return drv->bdrv_load_vmstate(bs, buf, pos, size); | |
4106 | if (bs->file) | |
4107 | return bdrv_load_vmstate(bs->file, buf, pos, size); | |
4108 | return -ENOTSUP; | |
178e08a5 AL |
4109 | } |
4110 | ||
8b9b0cc2 KW |
4111 | void bdrv_debug_event(BlockDriverState *bs, BlkDebugEvent event) |
4112 | { | |
bf736fe3 | 4113 | if (!bs || !bs->drv || !bs->drv->bdrv_debug_event) { |
8b9b0cc2 KW |
4114 | return; |
4115 | } | |
4116 | ||
bf736fe3 | 4117 | bs->drv->bdrv_debug_event(bs, event); |
41c695c7 KW |
4118 | } |
4119 | ||
4120 | int bdrv_debug_breakpoint(BlockDriverState *bs, const char *event, | |
4121 | const char *tag) | |
4122 | { | |
4123 | while (bs && bs->drv && !bs->drv->bdrv_debug_breakpoint) { | |
4124 | bs = bs->file; | |
4125 | } | |
4126 | ||
4127 | if (bs && bs->drv && bs->drv->bdrv_debug_breakpoint) { | |
4128 | return bs->drv->bdrv_debug_breakpoint(bs, event, tag); | |
4129 | } | |
4130 | ||
4131 | return -ENOTSUP; | |
4132 | } | |
4133 | ||
4cc70e93 FZ |
4134 | int bdrv_debug_remove_breakpoint(BlockDriverState *bs, const char *tag) |
4135 | { | |
4136 | while (bs && bs->drv && !bs->drv->bdrv_debug_remove_breakpoint) { | |
4137 | bs = bs->file; | |
4138 | } | |
4139 | ||
4140 | if (bs && bs->drv && bs->drv->bdrv_debug_remove_breakpoint) { | |
4141 | return bs->drv->bdrv_debug_remove_breakpoint(bs, tag); | |
4142 | } | |
4143 | ||
4144 | return -ENOTSUP; | |
4145 | } | |
4146 | ||
41c695c7 KW |
4147 | int bdrv_debug_resume(BlockDriverState *bs, const char *tag) |
4148 | { | |
938789ea | 4149 | while (bs && (!bs->drv || !bs->drv->bdrv_debug_resume)) { |
41c695c7 KW |
4150 | bs = bs->file; |
4151 | } | |
8b9b0cc2 | 4152 | |
41c695c7 KW |
4153 | if (bs && bs->drv && bs->drv->bdrv_debug_resume) { |
4154 | return bs->drv->bdrv_debug_resume(bs, tag); | |
4155 | } | |
4156 | ||
4157 | return -ENOTSUP; | |
4158 | } | |
4159 | ||
4160 | bool bdrv_debug_is_suspended(BlockDriverState *bs, const char *tag) | |
4161 | { | |
4162 | while (bs && bs->drv && !bs->drv->bdrv_debug_is_suspended) { | |
4163 | bs = bs->file; | |
4164 | } | |
4165 | ||
4166 | if (bs && bs->drv && bs->drv->bdrv_debug_is_suspended) { | |
4167 | return bs->drv->bdrv_debug_is_suspended(bs, tag); | |
4168 | } | |
4169 | ||
4170 | return false; | |
8b9b0cc2 KW |
4171 | } |
4172 | ||
199630b6 BS |
4173 | int bdrv_is_snapshot(BlockDriverState *bs) |
4174 | { | |
4175 | return !!(bs->open_flags & BDRV_O_SNAPSHOT); | |
4176 | } | |
4177 | ||
b1b1d783 JC |
4178 | /* backing_file can either be relative, or absolute, or a protocol. If it is |
4179 | * relative, it must be relative to the chain. So, passing in bs->filename | |
4180 | * from a BDS as backing_file should not be done, as that may be relative to | |
4181 | * the CWD rather than the chain. */ | |
e8a6bb9c MT |
4182 | BlockDriverState *bdrv_find_backing_image(BlockDriverState *bs, |
4183 | const char *backing_file) | |
4184 | { | |
b1b1d783 JC |
4185 | char *filename_full = NULL; |
4186 | char *backing_file_full = NULL; | |
4187 | char *filename_tmp = NULL; | |
4188 | int is_protocol = 0; | |
4189 | BlockDriverState *curr_bs = NULL; | |
4190 | BlockDriverState *retval = NULL; | |
4191 | ||
4192 | if (!bs || !bs->drv || !backing_file) { | |
e8a6bb9c MT |
4193 | return NULL; |
4194 | } | |
4195 | ||
b1b1d783 JC |
4196 | filename_full = g_malloc(PATH_MAX); |
4197 | backing_file_full = g_malloc(PATH_MAX); | |
4198 | filename_tmp = g_malloc(PATH_MAX); | |
4199 | ||
4200 | is_protocol = path_has_protocol(backing_file); | |
4201 | ||
4202 | for (curr_bs = bs; curr_bs->backing_hd; curr_bs = curr_bs->backing_hd) { | |
4203 | ||
4204 | /* If either of the filename paths is actually a protocol, then | |
4205 | * compare unmodified paths; otherwise make paths relative */ | |
4206 | if (is_protocol || path_has_protocol(curr_bs->backing_file)) { | |
4207 | if (strcmp(backing_file, curr_bs->backing_file) == 0) { | |
4208 | retval = curr_bs->backing_hd; | |
4209 | break; | |
4210 | } | |
e8a6bb9c | 4211 | } else { |
b1b1d783 JC |
4212 | /* If not an absolute filename path, make it relative to the current |
4213 | * image's filename path */ | |
4214 | path_combine(filename_tmp, PATH_MAX, curr_bs->filename, | |
4215 | backing_file); | |
4216 | ||
4217 | /* We are going to compare absolute pathnames */ | |
4218 | if (!realpath(filename_tmp, filename_full)) { | |
4219 | continue; | |
4220 | } | |
4221 | ||
4222 | /* We need to make sure the backing filename we are comparing against | |
4223 | * is relative to the current image filename (or absolute) */ | |
4224 | path_combine(filename_tmp, PATH_MAX, curr_bs->filename, | |
4225 | curr_bs->backing_file); | |
4226 | ||
4227 | if (!realpath(filename_tmp, backing_file_full)) { | |
4228 | continue; | |
4229 | } | |
4230 | ||
4231 | if (strcmp(backing_file_full, filename_full) == 0) { | |
4232 | retval = curr_bs->backing_hd; | |
4233 | break; | |
4234 | } | |
e8a6bb9c MT |
4235 | } |
4236 | } | |
4237 | ||
b1b1d783 JC |
4238 | g_free(filename_full); |
4239 | g_free(backing_file_full); | |
4240 | g_free(filename_tmp); | |
4241 | return retval; | |
e8a6bb9c MT |
4242 | } |
4243 | ||
f198fd1c BC |
4244 | int bdrv_get_backing_file_depth(BlockDriverState *bs) |
4245 | { | |
4246 | if (!bs->drv) { | |
4247 | return 0; | |
4248 | } | |
4249 | ||
4250 | if (!bs->backing_hd) { | |
4251 | return 0; | |
4252 | } | |
4253 | ||
4254 | return 1 + bdrv_get_backing_file_depth(bs->backing_hd); | |
4255 | } | |
4256 | ||
79fac568 JC |
4257 | BlockDriverState *bdrv_find_base(BlockDriverState *bs) |
4258 | { | |
4259 | BlockDriverState *curr_bs = NULL; | |
4260 | ||
4261 | if (!bs) { | |
4262 | return NULL; | |
4263 | } | |
4264 | ||
4265 | curr_bs = bs; | |
4266 | ||
4267 | while (curr_bs->backing_hd) { | |
4268 | curr_bs = curr_bs->backing_hd; | |
4269 | } | |
4270 | return curr_bs; | |
4271 | } | |
4272 | ||
ea2384d3 | 4273 | /**************************************************************/ |
83f64091 | 4274 | /* async I/Os */ |
ea2384d3 | 4275 | |
3b69e4b9 | 4276 | BlockDriverAIOCB *bdrv_aio_readv(BlockDriverState *bs, int64_t sector_num, |
f141eafe | 4277 | QEMUIOVector *qiov, int nb_sectors, |
3b69e4b9 | 4278 | BlockDriverCompletionFunc *cb, void *opaque) |
83f64091 | 4279 | { |
bbf0a440 SH |
4280 | trace_bdrv_aio_readv(bs, sector_num, nb_sectors, opaque); |
4281 | ||
d20d9b7c | 4282 | return bdrv_co_aio_rw_vector(bs, sector_num, qiov, nb_sectors, 0, |
8c5873d6 | 4283 | cb, opaque, false); |
ea2384d3 FB |
4284 | } |
4285 | ||
f141eafe AL |
4286 | BlockDriverAIOCB *bdrv_aio_writev(BlockDriverState *bs, int64_t sector_num, |
4287 | QEMUIOVector *qiov, int nb_sectors, | |
4288 | BlockDriverCompletionFunc *cb, void *opaque) | |
ea2384d3 | 4289 | { |
bbf0a440 SH |
4290 | trace_bdrv_aio_writev(bs, sector_num, nb_sectors, opaque); |
4291 | ||
d20d9b7c | 4292 | return bdrv_co_aio_rw_vector(bs, sector_num, qiov, nb_sectors, 0, |
8c5873d6 | 4293 | cb, opaque, true); |
83f64091 FB |
4294 | } |
4295 | ||
d5ef94d4 PB |
4296 | BlockDriverAIOCB *bdrv_aio_write_zeroes(BlockDriverState *bs, |
4297 | int64_t sector_num, int nb_sectors, BdrvRequestFlags flags, | |
4298 | BlockDriverCompletionFunc *cb, void *opaque) | |
4299 | { | |
4300 | trace_bdrv_aio_write_zeroes(bs, sector_num, nb_sectors, flags, opaque); | |
4301 | ||
4302 | return bdrv_co_aio_rw_vector(bs, sector_num, NULL, nb_sectors, | |
4303 | BDRV_REQ_ZERO_WRITE | flags, | |
4304 | cb, opaque, true); | |
4305 | } | |
4306 | ||
40b4f539 KW |
4307 | |
4308 | typedef struct MultiwriteCB { | |
4309 | int error; | |
4310 | int num_requests; | |
4311 | int num_callbacks; | |
4312 | struct { | |
4313 | BlockDriverCompletionFunc *cb; | |
4314 | void *opaque; | |
4315 | QEMUIOVector *free_qiov; | |
40b4f539 KW |
4316 | } callbacks[]; |
4317 | } MultiwriteCB; | |
4318 | ||
4319 | static void multiwrite_user_cb(MultiwriteCB *mcb) | |
4320 | { | |
4321 | int i; | |
4322 | ||
4323 | for (i = 0; i < mcb->num_callbacks; i++) { | |
4324 | mcb->callbacks[i].cb(mcb->callbacks[i].opaque, mcb->error); | |
1e1ea48d SH |
4325 | if (mcb->callbacks[i].free_qiov) { |
4326 | qemu_iovec_destroy(mcb->callbacks[i].free_qiov); | |
4327 | } | |
7267c094 | 4328 | g_free(mcb->callbacks[i].free_qiov); |
40b4f539 KW |
4329 | } |
4330 | } | |
4331 | ||
4332 | static void multiwrite_cb(void *opaque, int ret) | |
4333 | { | |
4334 | MultiwriteCB *mcb = opaque; | |
4335 | ||
6d519a5f SH |
4336 | trace_multiwrite_cb(mcb, ret); |
4337 | ||
cb6d3ca0 | 4338 | if (ret < 0 && !mcb->error) { |
40b4f539 | 4339 | mcb->error = ret; |
40b4f539 KW |
4340 | } |
4341 | ||
4342 | mcb->num_requests--; | |
4343 | if (mcb->num_requests == 0) { | |
de189a1b | 4344 | multiwrite_user_cb(mcb); |
7267c094 | 4345 | g_free(mcb); |
40b4f539 KW |
4346 | } |
4347 | } | |
4348 | ||
4349 | static int multiwrite_req_compare(const void *a, const void *b) | |
4350 | { | |
77be4366 CH |
4351 | const BlockRequest *req1 = a, *req2 = b; |
4352 | ||
4353 | /* | |
4354 | * Note that we can't simply subtract req2->sector from req1->sector | |
4355 | * here as that could overflow the return value. | |
4356 | */ | |
4357 | if (req1->sector > req2->sector) { | |
4358 | return 1; | |
4359 | } else if (req1->sector < req2->sector) { | |
4360 | return -1; | |
4361 | } else { | |
4362 | return 0; | |
4363 | } | |
40b4f539 KW |
4364 | } |
4365 | ||
4366 | /* | |
4367 | * Takes a bunch of requests and tries to merge them. Returns the number of | |
4368 | * requests that remain after merging. | |
4369 | */ | |
4370 | static int multiwrite_merge(BlockDriverState *bs, BlockRequest *reqs, | |
4371 | int num_reqs, MultiwriteCB *mcb) | |
4372 | { | |
4373 | int i, outidx; | |
4374 | ||
4375 | // Sort requests by start sector | |
4376 | qsort(reqs, num_reqs, sizeof(*reqs), &multiwrite_req_compare); | |
4377 | ||
4378 | // Check if adjacent requests touch the same clusters. If so, combine them, | |
4379 | // filling up gaps with zero sectors. | |
4380 | outidx = 0; | |
4381 | for (i = 1; i < num_reqs; i++) { | |
4382 | int merge = 0; | |
4383 | int64_t oldreq_last = reqs[outidx].sector + reqs[outidx].nb_sectors; | |
4384 | ||
b6a127a1 | 4385 | // Handle exactly sequential writes and overlapping writes. |
40b4f539 KW |
4386 | if (reqs[i].sector <= oldreq_last) { |
4387 | merge = 1; | |
4388 | } | |
4389 | ||
e2a305fb CH |
4390 | if (reqs[outidx].qiov->niov + reqs[i].qiov->niov + 1 > IOV_MAX) { |
4391 | merge = 0; | |
4392 | } | |
4393 | ||
40b4f539 KW |
4394 | if (merge) { |
4395 | size_t size; | |
7267c094 | 4396 | QEMUIOVector *qiov = g_malloc0(sizeof(*qiov)); |
40b4f539 KW |
4397 | qemu_iovec_init(qiov, |
4398 | reqs[outidx].qiov->niov + reqs[i].qiov->niov + 1); | |
4399 | ||
4400 | // Add the first request to the merged one. If the requests are | |
4401 | // overlapping, drop the last sectors of the first request. | |
4402 | size = (reqs[i].sector - reqs[outidx].sector) << 9; | |
1b093c48 | 4403 | qemu_iovec_concat(qiov, reqs[outidx].qiov, 0, size); |
40b4f539 | 4404 | |
b6a127a1 PB |
4405 | // We should need to add any zeros between the two requests |
4406 | assert (reqs[i].sector <= oldreq_last); | |
40b4f539 KW |
4407 | |
4408 | // Add the second request | |
1b093c48 | 4409 | qemu_iovec_concat(qiov, reqs[i].qiov, 0, reqs[i].qiov->size); |
40b4f539 | 4410 | |
cbf1dff2 | 4411 | reqs[outidx].nb_sectors = qiov->size >> 9; |
40b4f539 KW |
4412 | reqs[outidx].qiov = qiov; |
4413 | ||
4414 | mcb->callbacks[i].free_qiov = reqs[outidx].qiov; | |
4415 | } else { | |
4416 | outidx++; | |
4417 | reqs[outidx].sector = reqs[i].sector; | |
4418 | reqs[outidx].nb_sectors = reqs[i].nb_sectors; | |
4419 | reqs[outidx].qiov = reqs[i].qiov; | |
4420 | } | |
4421 | } | |
4422 | ||
4423 | return outidx + 1; | |
4424 | } | |
4425 | ||
4426 | /* | |
4427 | * Submit multiple AIO write requests at once. | |
4428 | * | |
4429 | * On success, the function returns 0 and all requests in the reqs array have | |
4430 | * been submitted. In error case this function returns -1, and any of the | |
4431 | * requests may or may not be submitted yet. In particular, this means that the | |
4432 | * callback will be called for some of the requests, for others it won't. The | |
4433 | * caller must check the error field of the BlockRequest to wait for the right | |
4434 | * callbacks (if error != 0, no callback will be called). | |
4435 | * | |
4436 | * The implementation may modify the contents of the reqs array, e.g. to merge | |
4437 | * requests. However, the fields opaque and error are left unmodified as they | |
4438 | * are used to signal failure for a single request to the caller. | |
4439 | */ | |
4440 | int bdrv_aio_multiwrite(BlockDriverState *bs, BlockRequest *reqs, int num_reqs) | |
4441 | { | |
40b4f539 KW |
4442 | MultiwriteCB *mcb; |
4443 | int i; | |
4444 | ||
301db7c2 RH |
4445 | /* don't submit writes if we don't have a medium */ |
4446 | if (bs->drv == NULL) { | |
4447 | for (i = 0; i < num_reqs; i++) { | |
4448 | reqs[i].error = -ENOMEDIUM; | |
4449 | } | |
4450 | return -1; | |
4451 | } | |
4452 | ||
40b4f539 KW |
4453 | if (num_reqs == 0) { |
4454 | return 0; | |
4455 | } | |
4456 | ||
4457 | // Create MultiwriteCB structure | |
7267c094 | 4458 | mcb = g_malloc0(sizeof(*mcb) + num_reqs * sizeof(*mcb->callbacks)); |
40b4f539 KW |
4459 | mcb->num_requests = 0; |
4460 | mcb->num_callbacks = num_reqs; | |
4461 | ||
4462 | for (i = 0; i < num_reqs; i++) { | |
4463 | mcb->callbacks[i].cb = reqs[i].cb; | |
4464 | mcb->callbacks[i].opaque = reqs[i].opaque; | |
4465 | } | |
4466 | ||
4467 | // Check for mergable requests | |
4468 | num_reqs = multiwrite_merge(bs, reqs, num_reqs, mcb); | |
4469 | ||
6d519a5f SH |
4470 | trace_bdrv_aio_multiwrite(mcb, mcb->num_callbacks, num_reqs); |
4471 | ||
df9309fb PB |
4472 | /* Run the aio requests. */ |
4473 | mcb->num_requests = num_reqs; | |
40b4f539 | 4474 | for (i = 0; i < num_reqs; i++) { |
d20d9b7c PB |
4475 | bdrv_co_aio_rw_vector(bs, reqs[i].sector, reqs[i].qiov, |
4476 | reqs[i].nb_sectors, reqs[i].flags, | |
4477 | multiwrite_cb, mcb, | |
4478 | true); | |
40b4f539 KW |
4479 | } |
4480 | ||
4481 | return 0; | |
40b4f539 KW |
4482 | } |
4483 | ||
83f64091 | 4484 | void bdrv_aio_cancel(BlockDriverAIOCB *acb) |
83f64091 | 4485 | { |
d7331bed | 4486 | acb->aiocb_info->cancel(acb); |
83f64091 FB |
4487 | } |
4488 | ||
4489 | /**************************************************************/ | |
4490 | /* async block device emulation */ | |
4491 | ||
c16b5a2c CH |
4492 | typedef struct BlockDriverAIOCBSync { |
4493 | BlockDriverAIOCB common; | |
4494 | QEMUBH *bh; | |
4495 | int ret; | |
4496 | /* vector translation state */ | |
4497 | QEMUIOVector *qiov; | |
4498 | uint8_t *bounce; | |
4499 | int is_write; | |
4500 | } BlockDriverAIOCBSync; | |
4501 | ||
4502 | static void bdrv_aio_cancel_em(BlockDriverAIOCB *blockacb) | |
4503 | { | |
b666d239 KW |
4504 | BlockDriverAIOCBSync *acb = |
4505 | container_of(blockacb, BlockDriverAIOCBSync, common); | |
6a7ad299 | 4506 | qemu_bh_delete(acb->bh); |
36afc451 | 4507 | acb->bh = NULL; |
c16b5a2c CH |
4508 | qemu_aio_release(acb); |
4509 | } | |
4510 | ||
d7331bed | 4511 | static const AIOCBInfo bdrv_em_aiocb_info = { |
c16b5a2c CH |
4512 | .aiocb_size = sizeof(BlockDriverAIOCBSync), |
4513 | .cancel = bdrv_aio_cancel_em, | |
4514 | }; | |
4515 | ||
ce1a14dc | 4516 | static void bdrv_aio_bh_cb(void *opaque) |
83f64091 | 4517 | { |
ce1a14dc | 4518 | BlockDriverAIOCBSync *acb = opaque; |
f141eafe | 4519 | |
f141eafe | 4520 | if (!acb->is_write) |
03396148 | 4521 | qemu_iovec_from_buf(acb->qiov, 0, acb->bounce, acb->qiov->size); |
ceb42de8 | 4522 | qemu_vfree(acb->bounce); |
ce1a14dc | 4523 | acb->common.cb(acb->common.opaque, acb->ret); |
6a7ad299 | 4524 | qemu_bh_delete(acb->bh); |
36afc451 | 4525 | acb->bh = NULL; |
ce1a14dc | 4526 | qemu_aio_release(acb); |
83f64091 | 4527 | } |
beac80cd | 4528 | |
f141eafe AL |
4529 | static BlockDriverAIOCB *bdrv_aio_rw_vector(BlockDriverState *bs, |
4530 | int64_t sector_num, | |
4531 | QEMUIOVector *qiov, | |
4532 | int nb_sectors, | |
4533 | BlockDriverCompletionFunc *cb, | |
4534 | void *opaque, | |
4535 | int is_write) | |
4536 | ||
83f64091 | 4537 | { |
ce1a14dc | 4538 | BlockDriverAIOCBSync *acb; |
ce1a14dc | 4539 | |
d7331bed | 4540 | acb = qemu_aio_get(&bdrv_em_aiocb_info, bs, cb, opaque); |
f141eafe AL |
4541 | acb->is_write = is_write; |
4542 | acb->qiov = qiov; | |
e268ca52 | 4543 | acb->bounce = qemu_blockalign(bs, qiov->size); |
3f3aace8 | 4544 | acb->bh = qemu_bh_new(bdrv_aio_bh_cb, acb); |
f141eafe AL |
4545 | |
4546 | if (is_write) { | |
d5e6b161 | 4547 | qemu_iovec_to_buf(acb->qiov, 0, acb->bounce, qiov->size); |
1ed20acf | 4548 | acb->ret = bs->drv->bdrv_write(bs, sector_num, acb->bounce, nb_sectors); |
f141eafe | 4549 | } else { |
1ed20acf | 4550 | acb->ret = bs->drv->bdrv_read(bs, sector_num, acb->bounce, nb_sectors); |
f141eafe AL |
4551 | } |
4552 | ||
ce1a14dc | 4553 | qemu_bh_schedule(acb->bh); |
f141eafe | 4554 | |
ce1a14dc | 4555 | return &acb->common; |
beac80cd FB |
4556 | } |
4557 | ||
f141eafe AL |
4558 | static BlockDriverAIOCB *bdrv_aio_readv_em(BlockDriverState *bs, |
4559 | int64_t sector_num, QEMUIOVector *qiov, int nb_sectors, | |
ce1a14dc | 4560 | BlockDriverCompletionFunc *cb, void *opaque) |
beac80cd | 4561 | { |
f141eafe AL |
4562 | return bdrv_aio_rw_vector(bs, sector_num, qiov, nb_sectors, cb, opaque, 0); |
4563 | } | |
83f64091 | 4564 | |
f141eafe AL |
4565 | static BlockDriverAIOCB *bdrv_aio_writev_em(BlockDriverState *bs, |
4566 | int64_t sector_num, QEMUIOVector *qiov, int nb_sectors, | |
4567 | BlockDriverCompletionFunc *cb, void *opaque) | |
4568 | { | |
4569 | return bdrv_aio_rw_vector(bs, sector_num, qiov, nb_sectors, cb, opaque, 1); | |
beac80cd | 4570 | } |
beac80cd | 4571 | |
68485420 KW |
4572 | |
4573 | typedef struct BlockDriverAIOCBCoroutine { | |
4574 | BlockDriverAIOCB common; | |
4575 | BlockRequest req; | |
4576 | bool is_write; | |
d318aea9 | 4577 | bool *done; |
68485420 KW |
4578 | QEMUBH* bh; |
4579 | } BlockDriverAIOCBCoroutine; | |
4580 | ||
4581 | static void bdrv_aio_co_cancel_em(BlockDriverAIOCB *blockacb) | |
4582 | { | |
d318aea9 KW |
4583 | BlockDriverAIOCBCoroutine *acb = |
4584 | container_of(blockacb, BlockDriverAIOCBCoroutine, common); | |
4585 | bool done = false; | |
4586 | ||
4587 | acb->done = &done; | |
4588 | while (!done) { | |
4589 | qemu_aio_wait(); | |
4590 | } | |
68485420 KW |
4591 | } |
4592 | ||
d7331bed | 4593 | static const AIOCBInfo bdrv_em_co_aiocb_info = { |
68485420 KW |
4594 | .aiocb_size = sizeof(BlockDriverAIOCBCoroutine), |
4595 | .cancel = bdrv_aio_co_cancel_em, | |
4596 | }; | |
4597 | ||
35246a68 | 4598 | static void bdrv_co_em_bh(void *opaque) |
68485420 KW |
4599 | { |
4600 | BlockDriverAIOCBCoroutine *acb = opaque; | |
4601 | ||
4602 | acb->common.cb(acb->common.opaque, acb->req.error); | |
d318aea9 KW |
4603 | |
4604 | if (acb->done) { | |
4605 | *acb->done = true; | |
4606 | } | |
4607 | ||
68485420 KW |
4608 | qemu_bh_delete(acb->bh); |
4609 | qemu_aio_release(acb); | |
4610 | } | |
4611 | ||
b2a61371 SH |
4612 | /* Invoke bdrv_co_do_readv/bdrv_co_do_writev */ |
4613 | static void coroutine_fn bdrv_co_do_rw(void *opaque) | |
4614 | { | |
4615 | BlockDriverAIOCBCoroutine *acb = opaque; | |
4616 | BlockDriverState *bs = acb->common.bs; | |
4617 | ||
4618 | if (!acb->is_write) { | |
4619 | acb->req.error = bdrv_co_do_readv(bs, acb->req.sector, | |
d20d9b7c | 4620 | acb->req.nb_sectors, acb->req.qiov, acb->req.flags); |
b2a61371 SH |
4621 | } else { |
4622 | acb->req.error = bdrv_co_do_writev(bs, acb->req.sector, | |
d20d9b7c | 4623 | acb->req.nb_sectors, acb->req.qiov, acb->req.flags); |
b2a61371 SH |
4624 | } |
4625 | ||
35246a68 | 4626 | acb->bh = qemu_bh_new(bdrv_co_em_bh, acb); |
b2a61371 SH |
4627 | qemu_bh_schedule(acb->bh); |
4628 | } | |
4629 | ||
68485420 KW |
4630 | static BlockDriverAIOCB *bdrv_co_aio_rw_vector(BlockDriverState *bs, |
4631 | int64_t sector_num, | |
4632 | QEMUIOVector *qiov, | |
4633 | int nb_sectors, | |
d20d9b7c | 4634 | BdrvRequestFlags flags, |
68485420 KW |
4635 | BlockDriverCompletionFunc *cb, |
4636 | void *opaque, | |
8c5873d6 | 4637 | bool is_write) |
68485420 KW |
4638 | { |
4639 | Coroutine *co; | |
4640 | BlockDriverAIOCBCoroutine *acb; | |
4641 | ||
d7331bed | 4642 | acb = qemu_aio_get(&bdrv_em_co_aiocb_info, bs, cb, opaque); |
68485420 KW |
4643 | acb->req.sector = sector_num; |
4644 | acb->req.nb_sectors = nb_sectors; | |
4645 | acb->req.qiov = qiov; | |
d20d9b7c | 4646 | acb->req.flags = flags; |
68485420 | 4647 | acb->is_write = is_write; |
d318aea9 | 4648 | acb->done = NULL; |
68485420 | 4649 | |
8c5873d6 | 4650 | co = qemu_coroutine_create(bdrv_co_do_rw); |
68485420 KW |
4651 | qemu_coroutine_enter(co, acb); |
4652 | ||
4653 | return &acb->common; | |
4654 | } | |
4655 | ||
07f07615 | 4656 | static void coroutine_fn bdrv_aio_flush_co_entry(void *opaque) |
b2e12bc6 | 4657 | { |
07f07615 PB |
4658 | BlockDriverAIOCBCoroutine *acb = opaque; |
4659 | BlockDriverState *bs = acb->common.bs; | |
b2e12bc6 | 4660 | |
07f07615 PB |
4661 | acb->req.error = bdrv_co_flush(bs); |
4662 | acb->bh = qemu_bh_new(bdrv_co_em_bh, acb); | |
b2e12bc6 | 4663 | qemu_bh_schedule(acb->bh); |
b2e12bc6 CH |
4664 | } |
4665 | ||
07f07615 | 4666 | BlockDriverAIOCB *bdrv_aio_flush(BlockDriverState *bs, |
016f5cf6 AG |
4667 | BlockDriverCompletionFunc *cb, void *opaque) |
4668 | { | |
07f07615 | 4669 | trace_bdrv_aio_flush(bs, opaque); |
016f5cf6 | 4670 | |
07f07615 PB |
4671 | Coroutine *co; |
4672 | BlockDriverAIOCBCoroutine *acb; | |
016f5cf6 | 4673 | |
d7331bed | 4674 | acb = qemu_aio_get(&bdrv_em_co_aiocb_info, bs, cb, opaque); |
d318aea9 KW |
4675 | acb->done = NULL; |
4676 | ||
07f07615 PB |
4677 | co = qemu_coroutine_create(bdrv_aio_flush_co_entry); |
4678 | qemu_coroutine_enter(co, acb); | |
016f5cf6 | 4679 | |
016f5cf6 AG |
4680 | return &acb->common; |
4681 | } | |
4682 | ||
4265d620 PB |
4683 | static void coroutine_fn bdrv_aio_discard_co_entry(void *opaque) |
4684 | { | |
4685 | BlockDriverAIOCBCoroutine *acb = opaque; | |
4686 | BlockDriverState *bs = acb->common.bs; | |
4687 | ||
4688 | acb->req.error = bdrv_co_discard(bs, acb->req.sector, acb->req.nb_sectors); | |
4689 | acb->bh = qemu_bh_new(bdrv_co_em_bh, acb); | |
4690 | qemu_bh_schedule(acb->bh); | |
4691 | } | |
4692 | ||
4693 | BlockDriverAIOCB *bdrv_aio_discard(BlockDriverState *bs, | |
4694 | int64_t sector_num, int nb_sectors, | |
4695 | BlockDriverCompletionFunc *cb, void *opaque) | |
4696 | { | |
4697 | Coroutine *co; | |
4698 | BlockDriverAIOCBCoroutine *acb; | |
4699 | ||
4700 | trace_bdrv_aio_discard(bs, sector_num, nb_sectors, opaque); | |
4701 | ||
d7331bed | 4702 | acb = qemu_aio_get(&bdrv_em_co_aiocb_info, bs, cb, opaque); |
4265d620 PB |
4703 | acb->req.sector = sector_num; |
4704 | acb->req.nb_sectors = nb_sectors; | |
d318aea9 | 4705 | acb->done = NULL; |
4265d620 PB |
4706 | co = qemu_coroutine_create(bdrv_aio_discard_co_entry); |
4707 | qemu_coroutine_enter(co, acb); | |
4708 | ||
4709 | return &acb->common; | |
4710 | } | |
4711 | ||
ea2384d3 FB |
4712 | void bdrv_init(void) |
4713 | { | |
5efa9d5a | 4714 | module_call_init(MODULE_INIT_BLOCK); |
ea2384d3 | 4715 | } |
ce1a14dc | 4716 | |
eb852011 MA |
4717 | void bdrv_init_with_whitelist(void) |
4718 | { | |
4719 | use_bdrv_whitelist = 1; | |
4720 | bdrv_init(); | |
4721 | } | |
4722 | ||
d7331bed | 4723 | void *qemu_aio_get(const AIOCBInfo *aiocb_info, BlockDriverState *bs, |
c16b5a2c | 4724 | BlockDriverCompletionFunc *cb, void *opaque) |
ce1a14dc | 4725 | { |
ce1a14dc PB |
4726 | BlockDriverAIOCB *acb; |
4727 | ||
d7331bed SH |
4728 | acb = g_slice_alloc(aiocb_info->aiocb_size); |
4729 | acb->aiocb_info = aiocb_info; | |
ce1a14dc PB |
4730 | acb->bs = bs; |
4731 | acb->cb = cb; | |
4732 | acb->opaque = opaque; | |
4733 | return acb; | |
4734 | } | |
4735 | ||
4736 | void qemu_aio_release(void *p) | |
4737 | { | |
d37c975f | 4738 | BlockDriverAIOCB *acb = p; |
d7331bed | 4739 | g_slice_free1(acb->aiocb_info->aiocb_size, acb); |
ce1a14dc | 4740 | } |
19cb3738 | 4741 | |
f9f05dc5 KW |
4742 | /**************************************************************/ |
4743 | /* Coroutine block device emulation */ | |
4744 | ||
4745 | typedef struct CoroutineIOCompletion { | |
4746 | Coroutine *coroutine; | |
4747 | int ret; | |
4748 | } CoroutineIOCompletion; | |
4749 | ||
4750 | static void bdrv_co_io_em_complete(void *opaque, int ret) | |
4751 | { | |
4752 | CoroutineIOCompletion *co = opaque; | |
4753 | ||
4754 | co->ret = ret; | |
4755 | qemu_coroutine_enter(co->coroutine, NULL); | |
4756 | } | |
4757 | ||
4758 | static int coroutine_fn bdrv_co_io_em(BlockDriverState *bs, int64_t sector_num, | |
4759 | int nb_sectors, QEMUIOVector *iov, | |
4760 | bool is_write) | |
4761 | { | |
4762 | CoroutineIOCompletion co = { | |
4763 | .coroutine = qemu_coroutine_self(), | |
4764 | }; | |
4765 | BlockDriverAIOCB *acb; | |
4766 | ||
4767 | if (is_write) { | |
a652d160 SH |
4768 | acb = bs->drv->bdrv_aio_writev(bs, sector_num, iov, nb_sectors, |
4769 | bdrv_co_io_em_complete, &co); | |
f9f05dc5 | 4770 | } else { |
a652d160 SH |
4771 | acb = bs->drv->bdrv_aio_readv(bs, sector_num, iov, nb_sectors, |
4772 | bdrv_co_io_em_complete, &co); | |
f9f05dc5 KW |
4773 | } |
4774 | ||
59370aaa | 4775 | trace_bdrv_co_io_em(bs, sector_num, nb_sectors, is_write, acb); |
f9f05dc5 KW |
4776 | if (!acb) { |
4777 | return -EIO; | |
4778 | } | |
4779 | qemu_coroutine_yield(); | |
4780 | ||
4781 | return co.ret; | |
4782 | } | |
4783 | ||
4784 | static int coroutine_fn bdrv_co_readv_em(BlockDriverState *bs, | |
4785 | int64_t sector_num, int nb_sectors, | |
4786 | QEMUIOVector *iov) | |
4787 | { | |
4788 | return bdrv_co_io_em(bs, sector_num, nb_sectors, iov, false); | |
4789 | } | |
4790 | ||
4791 | static int coroutine_fn bdrv_co_writev_em(BlockDriverState *bs, | |
4792 | int64_t sector_num, int nb_sectors, | |
4793 | QEMUIOVector *iov) | |
4794 | { | |
4795 | return bdrv_co_io_em(bs, sector_num, nb_sectors, iov, true); | |
4796 | } | |
4797 | ||
07f07615 | 4798 | static void coroutine_fn bdrv_flush_co_entry(void *opaque) |
e7a8a783 | 4799 | { |
07f07615 PB |
4800 | RwCo *rwco = opaque; |
4801 | ||
4802 | rwco->ret = bdrv_co_flush(rwco->bs); | |
4803 | } | |
4804 | ||
4805 | int coroutine_fn bdrv_co_flush(BlockDriverState *bs) | |
4806 | { | |
eb489bb1 KW |
4807 | int ret; |
4808 | ||
29cdb251 | 4809 | if (!bs || !bdrv_is_inserted(bs) || bdrv_is_read_only(bs)) { |
07f07615 | 4810 | return 0; |
eb489bb1 KW |
4811 | } |
4812 | ||
ca716364 | 4813 | /* Write back cached data to the OS even with cache=unsafe */ |
bf736fe3 | 4814 | BLKDBG_EVENT(bs->file, BLKDBG_FLUSH_TO_OS); |
eb489bb1 KW |
4815 | if (bs->drv->bdrv_co_flush_to_os) { |
4816 | ret = bs->drv->bdrv_co_flush_to_os(bs); | |
4817 | if (ret < 0) { | |
4818 | return ret; | |
4819 | } | |
4820 | } | |
4821 | ||
ca716364 KW |
4822 | /* But don't actually force it to the disk with cache=unsafe */ |
4823 | if (bs->open_flags & BDRV_O_NO_FLUSH) { | |
d4c82329 | 4824 | goto flush_parent; |
ca716364 KW |
4825 | } |
4826 | ||
bf736fe3 | 4827 | BLKDBG_EVENT(bs->file, BLKDBG_FLUSH_TO_DISK); |
eb489bb1 | 4828 | if (bs->drv->bdrv_co_flush_to_disk) { |
29cdb251 | 4829 | ret = bs->drv->bdrv_co_flush_to_disk(bs); |
07f07615 PB |
4830 | } else if (bs->drv->bdrv_aio_flush) { |
4831 | BlockDriverAIOCB *acb; | |
4832 | CoroutineIOCompletion co = { | |
4833 | .coroutine = qemu_coroutine_self(), | |
4834 | }; | |
4835 | ||
4836 | acb = bs->drv->bdrv_aio_flush(bs, bdrv_co_io_em_complete, &co); | |
4837 | if (acb == NULL) { | |
29cdb251 | 4838 | ret = -EIO; |
07f07615 PB |
4839 | } else { |
4840 | qemu_coroutine_yield(); | |
29cdb251 | 4841 | ret = co.ret; |
07f07615 | 4842 | } |
07f07615 PB |
4843 | } else { |
4844 | /* | |
4845 | * Some block drivers always operate in either writethrough or unsafe | |
4846 | * mode and don't support bdrv_flush therefore. Usually qemu doesn't | |
4847 | * know how the server works (because the behaviour is hardcoded or | |
4848 | * depends on server-side configuration), so we can't ensure that | |
4849 | * everything is safe on disk. Returning an error doesn't work because | |
4850 | * that would break guests even if the server operates in writethrough | |
4851 | * mode. | |
4852 | * | |
4853 | * Let's hope the user knows what he's doing. | |
4854 | */ | |
29cdb251 | 4855 | ret = 0; |
07f07615 | 4856 | } |
29cdb251 PB |
4857 | if (ret < 0) { |
4858 | return ret; | |
4859 | } | |
4860 | ||
4861 | /* Now flush the underlying protocol. It will also have BDRV_O_NO_FLUSH | |
4862 | * in the case of cache=unsafe, so there are no useless flushes. | |
4863 | */ | |
d4c82329 | 4864 | flush_parent: |
29cdb251 | 4865 | return bdrv_co_flush(bs->file); |
07f07615 PB |
4866 | } |
4867 | ||
5a8a30db | 4868 | void bdrv_invalidate_cache(BlockDriverState *bs, Error **errp) |
0f15423c | 4869 | { |
5a8a30db KW |
4870 | Error *local_err = NULL; |
4871 | int ret; | |
4872 | ||
3456a8d1 KW |
4873 | if (!bs->drv) { |
4874 | return; | |
4875 | } | |
4876 | ||
4877 | if (bs->drv->bdrv_invalidate_cache) { | |
5a8a30db | 4878 | bs->drv->bdrv_invalidate_cache(bs, &local_err); |
3456a8d1 | 4879 | } else if (bs->file) { |
5a8a30db KW |
4880 | bdrv_invalidate_cache(bs->file, &local_err); |
4881 | } | |
4882 | if (local_err) { | |
4883 | error_propagate(errp, local_err); | |
4884 | return; | |
0f15423c | 4885 | } |
3456a8d1 | 4886 | |
5a8a30db KW |
4887 | ret = refresh_total_sectors(bs, bs->total_sectors); |
4888 | if (ret < 0) { | |
4889 | error_setg_errno(errp, -ret, "Could not refresh total sector count"); | |
4890 | return; | |
4891 | } | |
0f15423c AL |
4892 | } |
4893 | ||
5a8a30db | 4894 | void bdrv_invalidate_cache_all(Error **errp) |
0f15423c AL |
4895 | { |
4896 | BlockDriverState *bs; | |
5a8a30db | 4897 | Error *local_err = NULL; |
0f15423c | 4898 | |
dc364f4c | 4899 | QTAILQ_FOREACH(bs, &bdrv_states, device_list) { |
5a8a30db KW |
4900 | bdrv_invalidate_cache(bs, &local_err); |
4901 | if (local_err) { | |
4902 | error_propagate(errp, local_err); | |
4903 | return; | |
4904 | } | |
0f15423c AL |
4905 | } |
4906 | } | |
4907 | ||
07789269 BC |
4908 | void bdrv_clear_incoming_migration_all(void) |
4909 | { | |
4910 | BlockDriverState *bs; | |
4911 | ||
dc364f4c | 4912 | QTAILQ_FOREACH(bs, &bdrv_states, device_list) { |
07789269 BC |
4913 | bs->open_flags = bs->open_flags & ~(BDRV_O_INCOMING); |
4914 | } | |
4915 | } | |
4916 | ||
07f07615 PB |
4917 | int bdrv_flush(BlockDriverState *bs) |
4918 | { | |
4919 | Coroutine *co; | |
4920 | RwCo rwco = { | |
4921 | .bs = bs, | |
4922 | .ret = NOT_DONE, | |
e7a8a783 | 4923 | }; |
e7a8a783 | 4924 | |
07f07615 PB |
4925 | if (qemu_in_coroutine()) { |
4926 | /* Fast-path if already in coroutine context */ | |
4927 | bdrv_flush_co_entry(&rwco); | |
4928 | } else { | |
4929 | co = qemu_coroutine_create(bdrv_flush_co_entry); | |
4930 | qemu_coroutine_enter(co, &rwco); | |
4931 | while (rwco.ret == NOT_DONE) { | |
4932 | qemu_aio_wait(); | |
4933 | } | |
e7a8a783 | 4934 | } |
07f07615 PB |
4935 | |
4936 | return rwco.ret; | |
e7a8a783 KW |
4937 | } |
4938 | ||
775aa8b6 KW |
4939 | typedef struct DiscardCo { |
4940 | BlockDriverState *bs; | |
4941 | int64_t sector_num; | |
4942 | int nb_sectors; | |
4943 | int ret; | |
4944 | } DiscardCo; | |
4265d620 PB |
4945 | static void coroutine_fn bdrv_discard_co_entry(void *opaque) |
4946 | { | |
775aa8b6 | 4947 | DiscardCo *rwco = opaque; |
4265d620 PB |
4948 | |
4949 | rwco->ret = bdrv_co_discard(rwco->bs, rwco->sector_num, rwco->nb_sectors); | |
4950 | } | |
4951 | ||
6f14da52 PL |
4952 | /* if no limit is specified in the BlockLimits use a default |
4953 | * of 32768 512-byte sectors (16 MiB) per request. | |
4954 | */ | |
4955 | #define MAX_DISCARD_DEFAULT 32768 | |
4956 | ||
4265d620 PB |
4957 | int coroutine_fn bdrv_co_discard(BlockDriverState *bs, int64_t sector_num, |
4958 | int nb_sectors) | |
4959 | { | |
d51e9fe5 PB |
4960 | int max_discard; |
4961 | ||
4265d620 PB |
4962 | if (!bs->drv) { |
4963 | return -ENOMEDIUM; | |
4964 | } else if (bdrv_check_request(bs, sector_num, nb_sectors)) { | |
4965 | return -EIO; | |
4966 | } else if (bs->read_only) { | |
4967 | return -EROFS; | |
df702c9b PB |
4968 | } |
4969 | ||
e4654d2d | 4970 | bdrv_reset_dirty(bs, sector_num, nb_sectors); |
df702c9b | 4971 | |
9e8f1835 PB |
4972 | /* Do nothing if disabled. */ |
4973 | if (!(bs->open_flags & BDRV_O_UNMAP)) { | |
4974 | return 0; | |
4975 | } | |
4976 | ||
d51e9fe5 PB |
4977 | if (!bs->drv->bdrv_co_discard && !bs->drv->bdrv_aio_discard) { |
4978 | return 0; | |
4979 | } | |
6f14da52 | 4980 | |
d51e9fe5 PB |
4981 | max_discard = bs->bl.max_discard ? bs->bl.max_discard : MAX_DISCARD_DEFAULT; |
4982 | while (nb_sectors > 0) { | |
4983 | int ret; | |
4984 | int num = nb_sectors; | |
6f14da52 | 4985 | |
d51e9fe5 PB |
4986 | /* align request */ |
4987 | if (bs->bl.discard_alignment && | |
4988 | num >= bs->bl.discard_alignment && | |
4989 | sector_num % bs->bl.discard_alignment) { | |
4990 | if (num > bs->bl.discard_alignment) { | |
4991 | num = bs->bl.discard_alignment; | |
6f14da52 | 4992 | } |
d51e9fe5 PB |
4993 | num -= sector_num % bs->bl.discard_alignment; |
4994 | } | |
6f14da52 | 4995 | |
d51e9fe5 PB |
4996 | /* limit request size */ |
4997 | if (num > max_discard) { | |
4998 | num = max_discard; | |
4999 | } | |
6f14da52 | 5000 | |
d51e9fe5 | 5001 | if (bs->drv->bdrv_co_discard) { |
6f14da52 | 5002 | ret = bs->drv->bdrv_co_discard(bs, sector_num, num); |
d51e9fe5 PB |
5003 | } else { |
5004 | BlockDriverAIOCB *acb; | |
5005 | CoroutineIOCompletion co = { | |
5006 | .coroutine = qemu_coroutine_self(), | |
5007 | }; | |
5008 | ||
5009 | acb = bs->drv->bdrv_aio_discard(bs, sector_num, nb_sectors, | |
5010 | bdrv_co_io_em_complete, &co); | |
5011 | if (acb == NULL) { | |
5012 | return -EIO; | |
5013 | } else { | |
5014 | qemu_coroutine_yield(); | |
5015 | ret = co.ret; | |
6f14da52 | 5016 | } |
6f14da52 | 5017 | } |
7ce21016 | 5018 | if (ret && ret != -ENOTSUP) { |
d51e9fe5 | 5019 | return ret; |
4265d620 | 5020 | } |
d51e9fe5 PB |
5021 | |
5022 | sector_num += num; | |
5023 | nb_sectors -= num; | |
4265d620 | 5024 | } |
d51e9fe5 | 5025 | return 0; |
4265d620 PB |
5026 | } |
5027 | ||
5028 | int bdrv_discard(BlockDriverState *bs, int64_t sector_num, int nb_sectors) | |
5029 | { | |
5030 | Coroutine *co; | |
775aa8b6 | 5031 | DiscardCo rwco = { |
4265d620 PB |
5032 | .bs = bs, |
5033 | .sector_num = sector_num, | |
5034 | .nb_sectors = nb_sectors, | |
5035 | .ret = NOT_DONE, | |
5036 | }; | |
5037 | ||
5038 | if (qemu_in_coroutine()) { | |
5039 | /* Fast-path if already in coroutine context */ | |
5040 | bdrv_discard_co_entry(&rwco); | |
5041 | } else { | |
5042 | co = qemu_coroutine_create(bdrv_discard_co_entry); | |
5043 | qemu_coroutine_enter(co, &rwco); | |
5044 | while (rwco.ret == NOT_DONE) { | |
5045 | qemu_aio_wait(); | |
5046 | } | |
5047 | } | |
5048 | ||
5049 | return rwco.ret; | |
5050 | } | |
5051 | ||
19cb3738 FB |
5052 | /**************************************************************/ |
5053 | /* removable device support */ | |
5054 | ||
5055 | /** | |
5056 | * Return TRUE if the media is present | |
5057 | */ | |
5058 | int bdrv_is_inserted(BlockDriverState *bs) | |
5059 | { | |
5060 | BlockDriver *drv = bs->drv; | |
a1aff5bf | 5061 | |
19cb3738 FB |
5062 | if (!drv) |
5063 | return 0; | |
5064 | if (!drv->bdrv_is_inserted) | |
a1aff5bf MA |
5065 | return 1; |
5066 | return drv->bdrv_is_inserted(bs); | |
19cb3738 FB |
5067 | } |
5068 | ||
5069 | /** | |
8e49ca46 MA |
5070 | * Return whether the media changed since the last call to this |
5071 | * function, or -ENOTSUP if we don't know. Most drivers don't know. | |
19cb3738 FB |
5072 | */ |
5073 | int bdrv_media_changed(BlockDriverState *bs) | |
5074 | { | |
5075 | BlockDriver *drv = bs->drv; | |
19cb3738 | 5076 | |
8e49ca46 MA |
5077 | if (drv && drv->bdrv_media_changed) { |
5078 | return drv->bdrv_media_changed(bs); | |
5079 | } | |
5080 | return -ENOTSUP; | |
19cb3738 FB |
5081 | } |
5082 | ||
5083 | /** | |
5084 | * If eject_flag is TRUE, eject the media. Otherwise, close the tray | |
5085 | */ | |
f36f3949 | 5086 | void bdrv_eject(BlockDriverState *bs, bool eject_flag) |
19cb3738 FB |
5087 | { |
5088 | BlockDriver *drv = bs->drv; | |
19cb3738 | 5089 | |
822e1cd1 MA |
5090 | if (drv && drv->bdrv_eject) { |
5091 | drv->bdrv_eject(bs, eject_flag); | |
19cb3738 | 5092 | } |
6f382ed2 LC |
5093 | |
5094 | if (bs->device_name[0] != '\0') { | |
5095 | bdrv_emit_qmp_eject_event(bs, eject_flag); | |
5096 | } | |
19cb3738 FB |
5097 | } |
5098 | ||
19cb3738 FB |
5099 | /** |
5100 | * Lock or unlock the media (if it is locked, the user won't be able | |
5101 | * to eject it manually). | |
5102 | */ | |
025e849a | 5103 | void bdrv_lock_medium(BlockDriverState *bs, bool locked) |
19cb3738 FB |
5104 | { |
5105 | BlockDriver *drv = bs->drv; | |
5106 | ||
025e849a | 5107 | trace_bdrv_lock_medium(bs, locked); |
b8c6d095 | 5108 | |
025e849a MA |
5109 | if (drv && drv->bdrv_lock_medium) { |
5110 | drv->bdrv_lock_medium(bs, locked); | |
19cb3738 FB |
5111 | } |
5112 | } | |
985a03b0 TS |
5113 | |
5114 | /* needed for generic scsi interface */ | |
5115 | ||
5116 | int bdrv_ioctl(BlockDriverState *bs, unsigned long int req, void *buf) | |
5117 | { | |
5118 | BlockDriver *drv = bs->drv; | |
5119 | ||
5120 | if (drv && drv->bdrv_ioctl) | |
5121 | return drv->bdrv_ioctl(bs, req, buf); | |
5122 | return -ENOTSUP; | |
5123 | } | |
7d780669 | 5124 | |
221f715d AL |
5125 | BlockDriverAIOCB *bdrv_aio_ioctl(BlockDriverState *bs, |
5126 | unsigned long int req, void *buf, | |
5127 | BlockDriverCompletionFunc *cb, void *opaque) | |
7d780669 | 5128 | { |
221f715d | 5129 | BlockDriver *drv = bs->drv; |
7d780669 | 5130 | |
221f715d AL |
5131 | if (drv && drv->bdrv_aio_ioctl) |
5132 | return drv->bdrv_aio_ioctl(bs, req, buf, cb, opaque); | |
5133 | return NULL; | |
7d780669 | 5134 | } |
e268ca52 | 5135 | |
1b7fd729 | 5136 | void bdrv_set_guest_block_size(BlockDriverState *bs, int align) |
7b6f9300 | 5137 | { |
1b7fd729 | 5138 | bs->guest_block_size = align; |
7b6f9300 | 5139 | } |
7cd1e32a | 5140 | |
e268ca52 AL |
5141 | void *qemu_blockalign(BlockDriverState *bs, size_t size) |
5142 | { | |
339064d5 | 5143 | return qemu_memalign(bdrv_opt_mem_align(bs), size); |
e268ca52 | 5144 | } |
7cd1e32a | 5145 | |
c53b1c51 SH |
5146 | /* |
5147 | * Check if all memory in this vector is sector aligned. | |
5148 | */ | |
5149 | bool bdrv_qiov_is_aligned(BlockDriverState *bs, QEMUIOVector *qiov) | |
5150 | { | |
5151 | int i; | |
339064d5 | 5152 | size_t alignment = bdrv_opt_mem_align(bs); |
c53b1c51 SH |
5153 | |
5154 | for (i = 0; i < qiov->niov; i++) { | |
339064d5 | 5155 | if ((uintptr_t) qiov->iov[i].iov_base % alignment) { |
c53b1c51 | 5156 | return false; |
1ff735bd | 5157 | } |
339064d5 | 5158 | if (qiov->iov[i].iov_len % alignment) { |
1ff735bd | 5159 | return false; |
c53b1c51 SH |
5160 | } |
5161 | } | |
5162 | ||
5163 | return true; | |
5164 | } | |
5165 | ||
b8afb520 FZ |
5166 | BdrvDirtyBitmap *bdrv_create_dirty_bitmap(BlockDriverState *bs, int granularity, |
5167 | Error **errp) | |
7cd1e32a LS |
5168 | { |
5169 | int64_t bitmap_size; | |
e4654d2d | 5170 | BdrvDirtyBitmap *bitmap; |
a55eb92c | 5171 | |
50717e94 PB |
5172 | assert((granularity & (granularity - 1)) == 0); |
5173 | ||
e4654d2d FZ |
5174 | granularity >>= BDRV_SECTOR_BITS; |
5175 | assert(granularity); | |
b8afb520 FZ |
5176 | bitmap_size = bdrv_getlength(bs); |
5177 | if (bitmap_size < 0) { | |
5178 | error_setg_errno(errp, -bitmap_size, "could not get length of device"); | |
5179 | errno = -bitmap_size; | |
5180 | return NULL; | |
5181 | } | |
5182 | bitmap_size >>= BDRV_SECTOR_BITS; | |
e4654d2d FZ |
5183 | bitmap = g_malloc0(sizeof(BdrvDirtyBitmap)); |
5184 | bitmap->bitmap = hbitmap_alloc(bitmap_size, ffs(granularity) - 1); | |
5185 | QLIST_INSERT_HEAD(&bs->dirty_bitmaps, bitmap, list); | |
5186 | return bitmap; | |
5187 | } | |
5188 | ||
5189 | void bdrv_release_dirty_bitmap(BlockDriverState *bs, BdrvDirtyBitmap *bitmap) | |
5190 | { | |
5191 | BdrvDirtyBitmap *bm, *next; | |
5192 | QLIST_FOREACH_SAFE(bm, &bs->dirty_bitmaps, list, next) { | |
5193 | if (bm == bitmap) { | |
5194 | QLIST_REMOVE(bitmap, list); | |
5195 | hbitmap_free(bitmap->bitmap); | |
5196 | g_free(bitmap); | |
5197 | return; | |
a55eb92c | 5198 | } |
7cd1e32a LS |
5199 | } |
5200 | } | |
5201 | ||
21b56835 FZ |
5202 | BlockDirtyInfoList *bdrv_query_dirty_bitmaps(BlockDriverState *bs) |
5203 | { | |
5204 | BdrvDirtyBitmap *bm; | |
5205 | BlockDirtyInfoList *list = NULL; | |
5206 | BlockDirtyInfoList **plist = &list; | |
5207 | ||
5208 | QLIST_FOREACH(bm, &bs->dirty_bitmaps, list) { | |
5209 | BlockDirtyInfo *info = g_malloc0(sizeof(BlockDirtyInfo)); | |
5210 | BlockDirtyInfoList *entry = g_malloc0(sizeof(BlockDirtyInfoList)); | |
5211 | info->count = bdrv_get_dirty_count(bs, bm); | |
5212 | info->granularity = | |
5213 | ((int64_t) BDRV_SECTOR_SIZE << hbitmap_granularity(bm->bitmap)); | |
5214 | entry->value = info; | |
5215 | *plist = entry; | |
5216 | plist = &entry->next; | |
5217 | } | |
5218 | ||
5219 | return list; | |
5220 | } | |
5221 | ||
e4654d2d | 5222 | int bdrv_get_dirty(BlockDriverState *bs, BdrvDirtyBitmap *bitmap, int64_t sector) |
7cd1e32a | 5223 | { |
e4654d2d FZ |
5224 | if (bitmap) { |
5225 | return hbitmap_get(bitmap->bitmap, sector); | |
7cd1e32a LS |
5226 | } else { |
5227 | return 0; | |
5228 | } | |
5229 | } | |
5230 | ||
e4654d2d FZ |
5231 | void bdrv_dirty_iter_init(BlockDriverState *bs, |
5232 | BdrvDirtyBitmap *bitmap, HBitmapIter *hbi) | |
1755da16 | 5233 | { |
e4654d2d | 5234 | hbitmap_iter_init(hbi, bitmap->bitmap, 0); |
1755da16 PB |
5235 | } |
5236 | ||
5237 | void bdrv_set_dirty(BlockDriverState *bs, int64_t cur_sector, | |
5238 | int nr_sectors) | |
5239 | { | |
e4654d2d FZ |
5240 | BdrvDirtyBitmap *bitmap; |
5241 | QLIST_FOREACH(bitmap, &bs->dirty_bitmaps, list) { | |
5242 | hbitmap_set(bitmap->bitmap, cur_sector, nr_sectors); | |
5243 | } | |
1755da16 PB |
5244 | } |
5245 | ||
e4654d2d | 5246 | void bdrv_reset_dirty(BlockDriverState *bs, int64_t cur_sector, int nr_sectors) |
7cd1e32a | 5247 | { |
e4654d2d FZ |
5248 | BdrvDirtyBitmap *bitmap; |
5249 | QLIST_FOREACH(bitmap, &bs->dirty_bitmaps, list) { | |
5250 | hbitmap_reset(bitmap->bitmap, cur_sector, nr_sectors); | |
5251 | } | |
7cd1e32a | 5252 | } |
aaa0eb75 | 5253 | |
e4654d2d | 5254 | int64_t bdrv_get_dirty_count(BlockDriverState *bs, BdrvDirtyBitmap *bitmap) |
aaa0eb75 | 5255 | { |
e4654d2d | 5256 | return hbitmap_count(bitmap->bitmap); |
aaa0eb75 | 5257 | } |
f88e1a42 | 5258 | |
9fcb0251 FZ |
5259 | /* Get a reference to bs */ |
5260 | void bdrv_ref(BlockDriverState *bs) | |
5261 | { | |
5262 | bs->refcnt++; | |
5263 | } | |
5264 | ||
5265 | /* Release a previously grabbed reference to bs. | |
5266 | * If after releasing, reference count is zero, the BlockDriverState is | |
5267 | * deleted. */ | |
5268 | void bdrv_unref(BlockDriverState *bs) | |
5269 | { | |
5270 | assert(bs->refcnt > 0); | |
5271 | if (--bs->refcnt == 0) { | |
5272 | bdrv_delete(bs); | |
5273 | } | |
5274 | } | |
5275 | ||
db593f25 MT |
5276 | void bdrv_set_in_use(BlockDriverState *bs, int in_use) |
5277 | { | |
5278 | assert(bs->in_use != in_use); | |
5279 | bs->in_use = in_use; | |
5280 | } | |
5281 | ||
5282 | int bdrv_in_use(BlockDriverState *bs) | |
5283 | { | |
5284 | return bs->in_use; | |
5285 | } | |
5286 | ||
28a7282a LC |
5287 | void bdrv_iostatus_enable(BlockDriverState *bs) |
5288 | { | |
d6bf279e | 5289 | bs->iostatus_enabled = true; |
58e21ef5 | 5290 | bs->iostatus = BLOCK_DEVICE_IO_STATUS_OK; |
28a7282a LC |
5291 | } |
5292 | ||
5293 | /* The I/O status is only enabled if the drive explicitly | |
5294 | * enables it _and_ the VM is configured to stop on errors */ | |
5295 | bool bdrv_iostatus_is_enabled(const BlockDriverState *bs) | |
5296 | { | |
d6bf279e | 5297 | return (bs->iostatus_enabled && |
92aa5c6d PB |
5298 | (bs->on_write_error == BLOCKDEV_ON_ERROR_ENOSPC || |
5299 | bs->on_write_error == BLOCKDEV_ON_ERROR_STOP || | |
5300 | bs->on_read_error == BLOCKDEV_ON_ERROR_STOP)); | |
28a7282a LC |
5301 | } |
5302 | ||
5303 | void bdrv_iostatus_disable(BlockDriverState *bs) | |
5304 | { | |
d6bf279e | 5305 | bs->iostatus_enabled = false; |
28a7282a LC |
5306 | } |
5307 | ||
5308 | void bdrv_iostatus_reset(BlockDriverState *bs) | |
5309 | { | |
5310 | if (bdrv_iostatus_is_enabled(bs)) { | |
58e21ef5 | 5311 | bs->iostatus = BLOCK_DEVICE_IO_STATUS_OK; |
3bd293c3 PB |
5312 | if (bs->job) { |
5313 | block_job_iostatus_reset(bs->job); | |
5314 | } | |
28a7282a LC |
5315 | } |
5316 | } | |
5317 | ||
28a7282a LC |
5318 | void bdrv_iostatus_set_err(BlockDriverState *bs, int error) |
5319 | { | |
3e1caa5f PB |
5320 | assert(bdrv_iostatus_is_enabled(bs)); |
5321 | if (bs->iostatus == BLOCK_DEVICE_IO_STATUS_OK) { | |
58e21ef5 LC |
5322 | bs->iostatus = error == ENOSPC ? BLOCK_DEVICE_IO_STATUS_NOSPACE : |
5323 | BLOCK_DEVICE_IO_STATUS_FAILED; | |
28a7282a LC |
5324 | } |
5325 | } | |
5326 | ||
a597e79c CH |
5327 | void |
5328 | bdrv_acct_start(BlockDriverState *bs, BlockAcctCookie *cookie, int64_t bytes, | |
5329 | enum BlockAcctType type) | |
5330 | { | |
5331 | assert(type < BDRV_MAX_IOTYPE); | |
5332 | ||
5333 | cookie->bytes = bytes; | |
c488c7f6 | 5334 | cookie->start_time_ns = get_clock(); |
a597e79c CH |
5335 | cookie->type = type; |
5336 | } | |
5337 | ||
5338 | void | |
5339 | bdrv_acct_done(BlockDriverState *bs, BlockAcctCookie *cookie) | |
5340 | { | |
5341 | assert(cookie->type < BDRV_MAX_IOTYPE); | |
5342 | ||
5343 | bs->nr_bytes[cookie->type] += cookie->bytes; | |
5344 | bs->nr_ops[cookie->type]++; | |
c488c7f6 | 5345 | bs->total_time_ns[cookie->type] += get_clock() - cookie->start_time_ns; |
a597e79c CH |
5346 | } |
5347 | ||
d92ada22 LC |
5348 | void bdrv_img_create(const char *filename, const char *fmt, |
5349 | const char *base_filename, const char *base_fmt, | |
f382d43a MR |
5350 | char *options, uint64_t img_size, int flags, |
5351 | Error **errp, bool quiet) | |
f88e1a42 JS |
5352 | { |
5353 | QEMUOptionParameter *param = NULL, *create_options = NULL; | |
d220894e | 5354 | QEMUOptionParameter *backing_fmt, *backing_file, *size; |
f88e1a42 | 5355 | BlockDriver *drv, *proto_drv; |
96df67d1 | 5356 | BlockDriver *backing_drv = NULL; |
cc84d90f | 5357 | Error *local_err = NULL; |
f88e1a42 JS |
5358 | int ret = 0; |
5359 | ||
5360 | /* Find driver and parse its options */ | |
5361 | drv = bdrv_find_format(fmt); | |
5362 | if (!drv) { | |
71c79813 | 5363 | error_setg(errp, "Unknown file format '%s'", fmt); |
d92ada22 | 5364 | return; |
f88e1a42 JS |
5365 | } |
5366 | ||
98289620 | 5367 | proto_drv = bdrv_find_protocol(filename, true); |
f88e1a42 | 5368 | if (!proto_drv) { |
71c79813 | 5369 | error_setg(errp, "Unknown protocol '%s'", filename); |
d92ada22 | 5370 | return; |
f88e1a42 JS |
5371 | } |
5372 | ||
5373 | create_options = append_option_parameters(create_options, | |
5374 | drv->create_options); | |
5375 | create_options = append_option_parameters(create_options, | |
5376 | proto_drv->create_options); | |
5377 | ||
5378 | /* Create parameter list with default values */ | |
5379 | param = parse_option_parameters("", create_options, param); | |
5380 | ||
5381 | set_option_parameter_int(param, BLOCK_OPT_SIZE, img_size); | |
5382 | ||
5383 | /* Parse -o options */ | |
5384 | if (options) { | |
5385 | param = parse_option_parameters(options, create_options, param); | |
5386 | if (param == NULL) { | |
71c79813 | 5387 | error_setg(errp, "Invalid options for file format '%s'.", fmt); |
f88e1a42 JS |
5388 | goto out; |
5389 | } | |
5390 | } | |
5391 | ||
5392 | if (base_filename) { | |
5393 | if (set_option_parameter(param, BLOCK_OPT_BACKING_FILE, | |
5394 | base_filename)) { | |
71c79813 LC |
5395 | error_setg(errp, "Backing file not supported for file format '%s'", |
5396 | fmt); | |
f88e1a42 JS |
5397 | goto out; |
5398 | } | |
5399 | } | |
5400 | ||
5401 | if (base_fmt) { | |
5402 | if (set_option_parameter(param, BLOCK_OPT_BACKING_FMT, base_fmt)) { | |
71c79813 LC |
5403 | error_setg(errp, "Backing file format not supported for file " |
5404 | "format '%s'", fmt); | |
f88e1a42 JS |
5405 | goto out; |
5406 | } | |
5407 | } | |
5408 | ||
792da93a JS |
5409 | backing_file = get_option_parameter(param, BLOCK_OPT_BACKING_FILE); |
5410 | if (backing_file && backing_file->value.s) { | |
5411 | if (!strcmp(filename, backing_file->value.s)) { | |
71c79813 LC |
5412 | error_setg(errp, "Error: Trying to create an image with the " |
5413 | "same filename as the backing file"); | |
792da93a JS |
5414 | goto out; |
5415 | } | |
5416 | } | |
5417 | ||
f88e1a42 JS |
5418 | backing_fmt = get_option_parameter(param, BLOCK_OPT_BACKING_FMT); |
5419 | if (backing_fmt && backing_fmt->value.s) { | |
96df67d1 SH |
5420 | backing_drv = bdrv_find_format(backing_fmt->value.s); |
5421 | if (!backing_drv) { | |
71c79813 LC |
5422 | error_setg(errp, "Unknown backing file format '%s'", |
5423 | backing_fmt->value.s); | |
f88e1a42 JS |
5424 | goto out; |
5425 | } | |
5426 | } | |
5427 | ||
5428 | // The size for the image must always be specified, with one exception: | |
5429 | // If we are using a backing file, we can obtain the size from there | |
d220894e KW |
5430 | size = get_option_parameter(param, BLOCK_OPT_SIZE); |
5431 | if (size && size->value.n == -1) { | |
f88e1a42 | 5432 | if (backing_file && backing_file->value.s) { |
66f6b814 | 5433 | BlockDriverState *bs; |
f88e1a42 | 5434 | uint64_t size; |
f88e1a42 | 5435 | char buf[32]; |
63090dac PB |
5436 | int back_flags; |
5437 | ||
5438 | /* backing files always opened read-only */ | |
5439 | back_flags = | |
5440 | flags & ~(BDRV_O_RDWR | BDRV_O_SNAPSHOT | BDRV_O_NO_BACKING); | |
f88e1a42 | 5441 | |
f67503e5 | 5442 | bs = NULL; |
ddf5636d | 5443 | ret = bdrv_open(&bs, backing_file->value.s, NULL, NULL, back_flags, |
cc84d90f | 5444 | backing_drv, &local_err); |
f88e1a42 | 5445 | if (ret < 0) { |
cc84d90f HR |
5446 | error_setg_errno(errp, -ret, "Could not open '%s': %s", |
5447 | backing_file->value.s, | |
5448 | error_get_pretty(local_err)); | |
5449 | error_free(local_err); | |
5450 | local_err = NULL; | |
f88e1a42 JS |
5451 | goto out; |
5452 | } | |
5453 | bdrv_get_geometry(bs, &size); | |
5454 | size *= 512; | |
5455 | ||
5456 | snprintf(buf, sizeof(buf), "%" PRId64, size); | |
5457 | set_option_parameter(param, BLOCK_OPT_SIZE, buf); | |
66f6b814 HR |
5458 | |
5459 | bdrv_unref(bs); | |
f88e1a42 | 5460 | } else { |
71c79813 | 5461 | error_setg(errp, "Image creation needs a size parameter"); |
f88e1a42 JS |
5462 | goto out; |
5463 | } | |
5464 | } | |
5465 | ||
f382d43a MR |
5466 | if (!quiet) { |
5467 | printf("Formatting '%s', fmt=%s ", filename, fmt); | |
5468 | print_option_parameters(param); | |
5469 | puts(""); | |
5470 | } | |
cc84d90f HR |
5471 | ret = bdrv_create(drv, filename, param, &local_err); |
5472 | if (ret == -EFBIG) { | |
5473 | /* This is generally a better message than whatever the driver would | |
5474 | * deliver (especially because of the cluster_size_hint), since that | |
5475 | * is most probably not much different from "image too large". */ | |
5476 | const char *cluster_size_hint = ""; | |
5477 | if (get_option_parameter(create_options, BLOCK_OPT_CLUSTER_SIZE)) { | |
5478 | cluster_size_hint = " (try using a larger cluster size)"; | |
f88e1a42 | 5479 | } |
cc84d90f HR |
5480 | error_setg(errp, "The image size is too large for file format '%s'" |
5481 | "%s", fmt, cluster_size_hint); | |
5482 | error_free(local_err); | |
5483 | local_err = NULL; | |
f88e1a42 JS |
5484 | } |
5485 | ||
5486 | out: | |
5487 | free_option_parameters(create_options); | |
5488 | free_option_parameters(param); | |
5489 | ||
84d18f06 | 5490 | if (local_err) { |
cc84d90f HR |
5491 | error_propagate(errp, local_err); |
5492 | } | |
f88e1a42 | 5493 | } |
85d126f3 SH |
5494 | |
5495 | AioContext *bdrv_get_aio_context(BlockDriverState *bs) | |
5496 | { | |
5497 | /* Currently BlockDriverState always uses the main loop AioContext */ | |
5498 | return qemu_get_aio_context(); | |
5499 | } | |
d616b224 SH |
5500 | |
5501 | void bdrv_add_before_write_notifier(BlockDriverState *bs, | |
5502 | NotifierWithReturn *notifier) | |
5503 | { | |
5504 | notifier_with_return_list_add(&bs->before_write_notifiers, notifier); | |
5505 | } | |
6f176b48 HR |
5506 | |
5507 | int bdrv_amend_options(BlockDriverState *bs, QEMUOptionParameter *options) | |
5508 | { | |
5509 | if (bs->drv->bdrv_amend_options == NULL) { | |
5510 | return -ENOTSUP; | |
5511 | } | |
5512 | return bs->drv->bdrv_amend_options(bs, options); | |
5513 | } | |
f6186f49 | 5514 | |
b5042a36 BC |
5515 | /* This function will be called by the bdrv_recurse_is_first_non_filter method |
5516 | * of block filter and by bdrv_is_first_non_filter. | |
5517 | * It is used to test if the given bs is the candidate or recurse more in the | |
5518 | * node graph. | |
212a5a8f | 5519 | */ |
b5042a36 | 5520 | bool bdrv_recurse_is_first_non_filter(BlockDriverState *bs, |
212a5a8f | 5521 | BlockDriverState *candidate) |
f6186f49 | 5522 | { |
b5042a36 BC |
5523 | /* return false if basic checks fails */ |
5524 | if (!bs || !bs->drv) { | |
212a5a8f | 5525 | return false; |
f6186f49 BC |
5526 | } |
5527 | ||
b5042a36 BC |
5528 | /* the code reached a non block filter driver -> check if the bs is |
5529 | * the same as the candidate. It's the recursion termination condition. | |
5530 | */ | |
5531 | if (!bs->drv->is_filter) { | |
5532 | return bs == candidate; | |
212a5a8f | 5533 | } |
b5042a36 | 5534 | /* Down this path the driver is a block filter driver */ |
212a5a8f | 5535 | |
b5042a36 BC |
5536 | /* If the block filter recursion method is defined use it to recurse down |
5537 | * the node graph. | |
5538 | */ | |
5539 | if (bs->drv->bdrv_recurse_is_first_non_filter) { | |
212a5a8f | 5540 | return bs->drv->bdrv_recurse_is_first_non_filter(bs, candidate); |
f6186f49 BC |
5541 | } |
5542 | ||
b5042a36 BC |
5543 | /* the driver is a block filter but don't allow to recurse -> return false |
5544 | */ | |
5545 | return false; | |
f6186f49 BC |
5546 | } |
5547 | ||
212a5a8f BC |
5548 | /* This function checks if the candidate is the first non filter bs down it's |
5549 | * bs chain. Since we don't have pointers to parents it explore all bs chains | |
5550 | * from the top. Some filters can choose not to pass down the recursion. | |
5551 | */ | |
5552 | bool bdrv_is_first_non_filter(BlockDriverState *candidate) | |
f6186f49 | 5553 | { |
212a5a8f BC |
5554 | BlockDriverState *bs; |
5555 | ||
5556 | /* walk down the bs forest recursively */ | |
5557 | QTAILQ_FOREACH(bs, &bdrv_states, device_list) { | |
5558 | bool perm; | |
5559 | ||
b5042a36 | 5560 | /* try to recurse in this top level bs */ |
e6dc8a1f | 5561 | perm = bdrv_recurse_is_first_non_filter(bs, candidate); |
212a5a8f BC |
5562 | |
5563 | /* candidate is the first non filter */ | |
5564 | if (perm) { | |
5565 | return true; | |
5566 | } | |
5567 | } | |
5568 | ||
5569 | return false; | |
f6186f49 | 5570 | } |