]>
Commit | Line | Data |
---|---|---|
fc01f7e7 FB |
1 | /* |
2 | * QEMU System Emulator block driver | |
5fafdf24 | 3 | * |
fc01f7e7 | 4 | * Copyright (c) 2003 Fabrice Bellard |
5fafdf24 | 5 | * |
fc01f7e7 FB |
6 | * Permission is hereby granted, free of charge, to any person obtaining a copy |
7 | * of this software and associated documentation files (the "Software"), to deal | |
8 | * in the Software without restriction, including without limitation the rights | |
9 | * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell | |
10 | * copies of the Software, and to permit persons to whom the Software is | |
11 | * furnished to do so, subject to the following conditions: | |
12 | * | |
13 | * The above copyright notice and this permission notice shall be included in | |
14 | * all copies or substantial portions of the Software. | |
15 | * | |
16 | * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR | |
17 | * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, | |
18 | * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL | |
19 | * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER | |
20 | * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, | |
21 | * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN | |
22 | * THE SOFTWARE. | |
23 | */ | |
3990d09a | 24 | #include "config-host.h" |
faf07963 | 25 | #include "qemu-common.h" |
6d519a5f | 26 | #include "trace.h" |
737e150e PB |
27 | #include "block/block_int.h" |
28 | #include "block/blockjob.h" | |
1de7afc9 | 29 | #include "qemu/module.h" |
7b1b5d19 | 30 | #include "qapi/qmp/qjson.h" |
9c17d615 | 31 | #include "sysemu/sysemu.h" |
1de7afc9 | 32 | #include "qemu/notify.h" |
737e150e | 33 | #include "block/coroutine.h" |
c13163fb | 34 | #include "block/qapi.h" |
b2023818 | 35 | #include "qmp-commands.h" |
1de7afc9 | 36 | #include "qemu/timer.h" |
a5ee7bd4 | 37 | #include "qapi-event.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, | |
13af91eb | 182 | bdrv_get_aio_context(bs), |
cc0681c4 BC |
183 | QEMU_CLOCK_VIRTUAL, |
184 | bdrv_throttle_read_timer_cb, | |
185 | bdrv_throttle_write_timer_cb, | |
186 | bs); | |
187 | bs->io_limits_enabled = true; | |
188 | } | |
189 | ||
190 | /* This function makes an IO wait if needed | |
191 | * | |
192 | * @nb_sectors: the number of sectors of the IO | |
193 | * @is_write: is the IO a write | |
194 | */ | |
98f90dba | 195 | static void bdrv_io_limits_intercept(BlockDriverState *bs, |
d5103588 | 196 | unsigned int bytes, |
cc0681c4 | 197 | bool is_write) |
98f90dba | 198 | { |
cc0681c4 BC |
199 | /* does this io must wait */ |
200 | bool must_wait = throttle_schedule_timer(&bs->throttle_state, is_write); | |
98f90dba | 201 | |
cc0681c4 BC |
202 | /* if must wait or any request of this type throttled queue the IO */ |
203 | if (must_wait || | |
204 | !qemu_co_queue_empty(&bs->throttled_reqs[is_write])) { | |
205 | qemu_co_queue_wait(&bs->throttled_reqs[is_write]); | |
98f90dba ZYW |
206 | } |
207 | ||
cc0681c4 | 208 | /* the IO will be executed, do the accounting */ |
d5103588 KW |
209 | throttle_account(&bs->throttle_state, is_write, bytes); |
210 | ||
98f90dba | 211 | |
cc0681c4 BC |
212 | /* if the next request must wait -> do nothing */ |
213 | if (throttle_schedule_timer(&bs->throttle_state, is_write)) { | |
214 | return; | |
98f90dba ZYW |
215 | } |
216 | ||
cc0681c4 BC |
217 | /* else queue next request for execution */ |
218 | qemu_co_queue_next(&bs->throttled_reqs[is_write]); | |
98f90dba ZYW |
219 | } |
220 | ||
339064d5 KW |
221 | size_t bdrv_opt_mem_align(BlockDriverState *bs) |
222 | { | |
223 | if (!bs || !bs->drv) { | |
224 | /* 4k should be on the safe side */ | |
225 | return 4096; | |
226 | } | |
227 | ||
228 | return bs->bl.opt_mem_alignment; | |
229 | } | |
230 | ||
9e0b22f4 SH |
231 | /* check if the path starts with "<protocol>:" */ |
232 | static int path_has_protocol(const char *path) | |
233 | { | |
947995c0 PB |
234 | const char *p; |
235 | ||
9e0b22f4 SH |
236 | #ifdef _WIN32 |
237 | if (is_windows_drive(path) || | |
238 | is_windows_drive_prefix(path)) { | |
239 | return 0; | |
240 | } | |
947995c0 PB |
241 | p = path + strcspn(path, ":/\\"); |
242 | #else | |
243 | p = path + strcspn(path, ":/"); | |
9e0b22f4 SH |
244 | #endif |
245 | ||
947995c0 | 246 | return *p == ':'; |
9e0b22f4 SH |
247 | } |
248 | ||
83f64091 | 249 | int path_is_absolute(const char *path) |
3b0d4f61 | 250 | { |
21664424 FB |
251 | #ifdef _WIN32 |
252 | /* specific case for names like: "\\.\d:" */ | |
f53f4da9 | 253 | if (is_windows_drive(path) || is_windows_drive_prefix(path)) { |
21664424 | 254 | return 1; |
f53f4da9 PB |
255 | } |
256 | return (*path == '/' || *path == '\\'); | |
3b9f94e1 | 257 | #else |
f53f4da9 | 258 | return (*path == '/'); |
3b9f94e1 | 259 | #endif |
3b0d4f61 FB |
260 | } |
261 | ||
83f64091 FB |
262 | /* if filename is absolute, just copy it to dest. Otherwise, build a |
263 | path to it by considering it is relative to base_path. URL are | |
264 | supported. */ | |
265 | void path_combine(char *dest, int dest_size, | |
266 | const char *base_path, | |
267 | const char *filename) | |
3b0d4f61 | 268 | { |
83f64091 FB |
269 | const char *p, *p1; |
270 | int len; | |
271 | ||
272 | if (dest_size <= 0) | |
273 | return; | |
274 | if (path_is_absolute(filename)) { | |
275 | pstrcpy(dest, dest_size, filename); | |
276 | } else { | |
277 | p = strchr(base_path, ':'); | |
278 | if (p) | |
279 | p++; | |
280 | else | |
281 | p = base_path; | |
3b9f94e1 FB |
282 | p1 = strrchr(base_path, '/'); |
283 | #ifdef _WIN32 | |
284 | { | |
285 | const char *p2; | |
286 | p2 = strrchr(base_path, '\\'); | |
287 | if (!p1 || p2 > p1) | |
288 | p1 = p2; | |
289 | } | |
290 | #endif | |
83f64091 FB |
291 | if (p1) |
292 | p1++; | |
293 | else | |
294 | p1 = base_path; | |
295 | if (p1 > p) | |
296 | p = p1; | |
297 | len = p - base_path; | |
298 | if (len > dest_size - 1) | |
299 | len = dest_size - 1; | |
300 | memcpy(dest, base_path, len); | |
301 | dest[len] = '\0'; | |
302 | pstrcat(dest, dest_size, filename); | |
3b0d4f61 | 303 | } |
3b0d4f61 FB |
304 | } |
305 | ||
dc5a1371 PB |
306 | void bdrv_get_full_backing_filename(BlockDriverState *bs, char *dest, size_t sz) |
307 | { | |
308 | if (bs->backing_file[0] == '\0' || path_has_protocol(bs->backing_file)) { | |
309 | pstrcpy(dest, sz, bs->backing_file); | |
310 | } else { | |
311 | path_combine(dest, sz, bs->filename, bs->backing_file); | |
312 | } | |
313 | } | |
314 | ||
5efa9d5a | 315 | void bdrv_register(BlockDriver *bdrv) |
ea2384d3 | 316 | { |
8c5873d6 SH |
317 | /* Block drivers without coroutine functions need emulation */ |
318 | if (!bdrv->bdrv_co_readv) { | |
f9f05dc5 KW |
319 | bdrv->bdrv_co_readv = bdrv_co_readv_em; |
320 | bdrv->bdrv_co_writev = bdrv_co_writev_em; | |
321 | ||
f8c35c1d SH |
322 | /* bdrv_co_readv_em()/brdv_co_writev_em() work in terms of aio, so if |
323 | * the block driver lacks aio we need to emulate that too. | |
324 | */ | |
f9f05dc5 KW |
325 | if (!bdrv->bdrv_aio_readv) { |
326 | /* add AIO emulation layer */ | |
327 | bdrv->bdrv_aio_readv = bdrv_aio_readv_em; | |
328 | bdrv->bdrv_aio_writev = bdrv_aio_writev_em; | |
f9f05dc5 | 329 | } |
83f64091 | 330 | } |
b2e12bc6 | 331 | |
8a22f02a | 332 | QLIST_INSERT_HEAD(&bdrv_drivers, bdrv, list); |
ea2384d3 | 333 | } |
b338082b FB |
334 | |
335 | /* create a new block device (by default it is empty) */ | |
98522f63 | 336 | BlockDriverState *bdrv_new(const char *device_name, Error **errp) |
b338082b | 337 | { |
1b7bdbc1 | 338 | BlockDriverState *bs; |
fbe40ff7 | 339 | int i; |
b338082b | 340 | |
f2d953ec KW |
341 | if (bdrv_find(device_name)) { |
342 | error_setg(errp, "Device with id '%s' already exists", | |
343 | device_name); | |
344 | return NULL; | |
345 | } | |
346 | if (bdrv_find_node(device_name)) { | |
347 | error_setg(errp, "Device with node-name '%s' already exists", | |
348 | device_name); | |
349 | return NULL; | |
350 | } | |
351 | ||
7267c094 | 352 | bs = g_malloc0(sizeof(BlockDriverState)); |
e4654d2d | 353 | QLIST_INIT(&bs->dirty_bitmaps); |
b338082b | 354 | pstrcpy(bs->device_name, sizeof(bs->device_name), device_name); |
ea2384d3 | 355 | if (device_name[0] != '\0') { |
dc364f4c | 356 | QTAILQ_INSERT_TAIL(&bdrv_states, bs, device_list); |
ea2384d3 | 357 | } |
fbe40ff7 FZ |
358 | for (i = 0; i < BLOCK_OP_TYPE_MAX; i++) { |
359 | QLIST_INIT(&bs->op_blockers[i]); | |
360 | } | |
28a7282a | 361 | bdrv_iostatus_disable(bs); |
d7d512f6 | 362 | notifier_list_init(&bs->close_notifiers); |
d616b224 | 363 | notifier_with_return_list_init(&bs->before_write_notifiers); |
cc0681c4 BC |
364 | qemu_co_queue_init(&bs->throttled_reqs[0]); |
365 | qemu_co_queue_init(&bs->throttled_reqs[1]); | |
9fcb0251 | 366 | bs->refcnt = 1; |
dcd04228 | 367 | bs->aio_context = qemu_get_aio_context(); |
d7d512f6 | 368 | |
b338082b FB |
369 | return bs; |
370 | } | |
371 | ||
d7d512f6 PB |
372 | void bdrv_add_close_notifier(BlockDriverState *bs, Notifier *notify) |
373 | { | |
374 | notifier_list_add(&bs->close_notifiers, notify); | |
375 | } | |
376 | ||
ea2384d3 FB |
377 | BlockDriver *bdrv_find_format(const char *format_name) |
378 | { | |
379 | BlockDriver *drv1; | |
8a22f02a SH |
380 | QLIST_FOREACH(drv1, &bdrv_drivers, list) { |
381 | if (!strcmp(drv1->format_name, format_name)) { | |
ea2384d3 | 382 | return drv1; |
8a22f02a | 383 | } |
ea2384d3 FB |
384 | } |
385 | return NULL; | |
386 | } | |
387 | ||
b64ec4e4 | 388 | static int bdrv_is_whitelisted(BlockDriver *drv, bool read_only) |
eb852011 | 389 | { |
b64ec4e4 FZ |
390 | static const char *whitelist_rw[] = { |
391 | CONFIG_BDRV_RW_WHITELIST | |
392 | }; | |
393 | static const char *whitelist_ro[] = { | |
394 | CONFIG_BDRV_RO_WHITELIST | |
eb852011 MA |
395 | }; |
396 | const char **p; | |
397 | ||
b64ec4e4 | 398 | if (!whitelist_rw[0] && !whitelist_ro[0]) { |
eb852011 | 399 | return 1; /* no whitelist, anything goes */ |
b64ec4e4 | 400 | } |
eb852011 | 401 | |
b64ec4e4 | 402 | for (p = whitelist_rw; *p; p++) { |
eb852011 MA |
403 | if (!strcmp(drv->format_name, *p)) { |
404 | return 1; | |
405 | } | |
406 | } | |
b64ec4e4 FZ |
407 | if (read_only) { |
408 | for (p = whitelist_ro; *p; p++) { | |
409 | if (!strcmp(drv->format_name, *p)) { | |
410 | return 1; | |
411 | } | |
412 | } | |
413 | } | |
eb852011 MA |
414 | return 0; |
415 | } | |
416 | ||
b64ec4e4 FZ |
417 | BlockDriver *bdrv_find_whitelisted_format(const char *format_name, |
418 | bool read_only) | |
eb852011 MA |
419 | { |
420 | BlockDriver *drv = bdrv_find_format(format_name); | |
b64ec4e4 | 421 | return drv && bdrv_is_whitelisted(drv, read_only) ? drv : NULL; |
eb852011 MA |
422 | } |
423 | ||
5b7e1542 ZYW |
424 | typedef struct CreateCo { |
425 | BlockDriver *drv; | |
426 | char *filename; | |
83d0521a | 427 | QemuOpts *opts; |
5b7e1542 | 428 | int ret; |
cc84d90f | 429 | Error *err; |
5b7e1542 ZYW |
430 | } CreateCo; |
431 | ||
432 | static void coroutine_fn bdrv_create_co_entry(void *opaque) | |
433 | { | |
cc84d90f HR |
434 | Error *local_err = NULL; |
435 | int ret; | |
436 | ||
5b7e1542 ZYW |
437 | CreateCo *cco = opaque; |
438 | assert(cco->drv); | |
439 | ||
c282e1fd | 440 | ret = cco->drv->bdrv_create(cco->filename, cco->opts, &local_err); |
84d18f06 | 441 | if (local_err) { |
cc84d90f HR |
442 | error_propagate(&cco->err, local_err); |
443 | } | |
444 | cco->ret = ret; | |
5b7e1542 ZYW |
445 | } |
446 | ||
0e7e1989 | 447 | int bdrv_create(BlockDriver *drv, const char* filename, |
83d0521a | 448 | QemuOpts *opts, Error **errp) |
ea2384d3 | 449 | { |
5b7e1542 ZYW |
450 | int ret; |
451 | ||
452 | Coroutine *co; | |
453 | CreateCo cco = { | |
454 | .drv = drv, | |
455 | .filename = g_strdup(filename), | |
83d0521a | 456 | .opts = opts, |
5b7e1542 | 457 | .ret = NOT_DONE, |
cc84d90f | 458 | .err = NULL, |
5b7e1542 ZYW |
459 | }; |
460 | ||
c282e1fd | 461 | if (!drv->bdrv_create) { |
cc84d90f | 462 | error_setg(errp, "Driver '%s' does not support image creation", drv->format_name); |
80168bff LC |
463 | ret = -ENOTSUP; |
464 | goto out; | |
5b7e1542 ZYW |
465 | } |
466 | ||
467 | if (qemu_in_coroutine()) { | |
468 | /* Fast-path if already in coroutine context */ | |
469 | bdrv_create_co_entry(&cco); | |
470 | } else { | |
471 | co = qemu_coroutine_create(bdrv_create_co_entry); | |
472 | qemu_coroutine_enter(co, &cco); | |
473 | while (cco.ret == NOT_DONE) { | |
b47ec2c4 | 474 | aio_poll(qemu_get_aio_context(), true); |
5b7e1542 ZYW |
475 | } |
476 | } | |
477 | ||
478 | ret = cco.ret; | |
cc84d90f | 479 | if (ret < 0) { |
84d18f06 | 480 | if (cco.err) { |
cc84d90f HR |
481 | error_propagate(errp, cco.err); |
482 | } else { | |
483 | error_setg_errno(errp, -ret, "Could not create image"); | |
484 | } | |
485 | } | |
0e7e1989 | 486 | |
80168bff LC |
487 | out: |
488 | g_free(cco.filename); | |
5b7e1542 | 489 | return ret; |
ea2384d3 FB |
490 | } |
491 | ||
c282e1fd | 492 | int bdrv_create_file(const char *filename, QemuOpts *opts, Error **errp) |
84a12e66 CH |
493 | { |
494 | BlockDriver *drv; | |
cc84d90f HR |
495 | Error *local_err = NULL; |
496 | int ret; | |
84a12e66 | 497 | |
98289620 | 498 | drv = bdrv_find_protocol(filename, true); |
84a12e66 | 499 | if (drv == NULL) { |
cc84d90f | 500 | error_setg(errp, "Could not find protocol for file '%s'", filename); |
16905d71 | 501 | return -ENOENT; |
84a12e66 CH |
502 | } |
503 | ||
c282e1fd | 504 | ret = bdrv_create(drv, filename, opts, &local_err); |
84d18f06 | 505 | if (local_err) { |
cc84d90f HR |
506 | error_propagate(errp, local_err); |
507 | } | |
508 | return ret; | |
84a12e66 CH |
509 | } |
510 | ||
3baca891 | 511 | void bdrv_refresh_limits(BlockDriverState *bs, Error **errp) |
d34682cd KW |
512 | { |
513 | BlockDriver *drv = bs->drv; | |
3baca891 | 514 | Error *local_err = NULL; |
d34682cd KW |
515 | |
516 | memset(&bs->bl, 0, sizeof(bs->bl)); | |
517 | ||
466ad822 | 518 | if (!drv) { |
3baca891 | 519 | return; |
466ad822 KW |
520 | } |
521 | ||
522 | /* Take some limits from the children as a default */ | |
523 | if (bs->file) { | |
3baca891 KW |
524 | bdrv_refresh_limits(bs->file, &local_err); |
525 | if (local_err) { | |
526 | error_propagate(errp, local_err); | |
527 | return; | |
528 | } | |
466ad822 | 529 | bs->bl.opt_transfer_length = bs->file->bl.opt_transfer_length; |
339064d5 KW |
530 | bs->bl.opt_mem_alignment = bs->file->bl.opt_mem_alignment; |
531 | } else { | |
532 | bs->bl.opt_mem_alignment = 512; | |
466ad822 KW |
533 | } |
534 | ||
535 | if (bs->backing_hd) { | |
3baca891 KW |
536 | bdrv_refresh_limits(bs->backing_hd, &local_err); |
537 | if (local_err) { | |
538 | error_propagate(errp, local_err); | |
539 | return; | |
540 | } | |
466ad822 KW |
541 | bs->bl.opt_transfer_length = |
542 | MAX(bs->bl.opt_transfer_length, | |
543 | bs->backing_hd->bl.opt_transfer_length); | |
339064d5 KW |
544 | bs->bl.opt_mem_alignment = |
545 | MAX(bs->bl.opt_mem_alignment, | |
546 | bs->backing_hd->bl.opt_mem_alignment); | |
466ad822 KW |
547 | } |
548 | ||
549 | /* Then let the driver override it */ | |
550 | if (drv->bdrv_refresh_limits) { | |
3baca891 | 551 | drv->bdrv_refresh_limits(bs, errp); |
d34682cd | 552 | } |
d34682cd KW |
553 | } |
554 | ||
eba25057 JM |
555 | /* |
556 | * Create a uniquely-named empty temporary file. | |
557 | * Return 0 upon success, otherwise a negative errno value. | |
558 | */ | |
559 | int get_tmp_filename(char *filename, int size) | |
d5249393 | 560 | { |
eba25057 | 561 | #ifdef _WIN32 |
3b9f94e1 | 562 | char temp_dir[MAX_PATH]; |
eba25057 JM |
563 | /* GetTempFileName requires that its output buffer (4th param) |
564 | have length MAX_PATH or greater. */ | |
565 | assert(size >= MAX_PATH); | |
566 | return (GetTempPath(MAX_PATH, temp_dir) | |
567 | && GetTempFileName(temp_dir, "qem", 0, filename) | |
568 | ? 0 : -GetLastError()); | |
d5249393 | 569 | #else |
67b915a5 | 570 | int fd; |
7ccfb2eb | 571 | const char *tmpdir; |
0badc1ee | 572 | tmpdir = getenv("TMPDIR"); |
69bef793 AS |
573 | if (!tmpdir) { |
574 | tmpdir = "/var/tmp"; | |
575 | } | |
eba25057 JM |
576 | if (snprintf(filename, size, "%s/vl.XXXXXX", tmpdir) >= size) { |
577 | return -EOVERFLOW; | |
578 | } | |
ea2384d3 | 579 | fd = mkstemp(filename); |
fe235a06 DH |
580 | if (fd < 0) { |
581 | return -errno; | |
582 | } | |
583 | if (close(fd) != 0) { | |
584 | unlink(filename); | |
eba25057 JM |
585 | return -errno; |
586 | } | |
587 | return 0; | |
d5249393 | 588 | #endif |
eba25057 | 589 | } |
fc01f7e7 | 590 | |
84a12e66 CH |
591 | /* |
592 | * Detect host devices. By convention, /dev/cdrom[N] is always | |
593 | * recognized as a host CDROM. | |
594 | */ | |
595 | static BlockDriver *find_hdev_driver(const char *filename) | |
596 | { | |
597 | int score_max = 0, score; | |
598 | BlockDriver *drv = NULL, *d; | |
599 | ||
600 | QLIST_FOREACH(d, &bdrv_drivers, list) { | |
601 | if (d->bdrv_probe_device) { | |
602 | score = d->bdrv_probe_device(filename); | |
603 | if (score > score_max) { | |
604 | score_max = score; | |
605 | drv = d; | |
606 | } | |
607 | } | |
608 | } | |
609 | ||
610 | return drv; | |
611 | } | |
612 | ||
98289620 KW |
613 | BlockDriver *bdrv_find_protocol(const char *filename, |
614 | bool allow_protocol_prefix) | |
83f64091 FB |
615 | { |
616 | BlockDriver *drv1; | |
617 | char protocol[128]; | |
1cec71e3 | 618 | int len; |
83f64091 | 619 | const char *p; |
19cb3738 | 620 | |
66f82cee KW |
621 | /* TODO Drivers without bdrv_file_open must be specified explicitly */ |
622 | ||
39508e7a CH |
623 | /* |
624 | * XXX(hch): we really should not let host device detection | |
625 | * override an explicit protocol specification, but moving this | |
626 | * later breaks access to device names with colons in them. | |
627 | * Thanks to the brain-dead persistent naming schemes on udev- | |
628 | * based Linux systems those actually are quite common. | |
629 | */ | |
630 | drv1 = find_hdev_driver(filename); | |
631 | if (drv1) { | |
632 | return drv1; | |
633 | } | |
634 | ||
98289620 | 635 | if (!path_has_protocol(filename) || !allow_protocol_prefix) { |
39508e7a | 636 | return bdrv_find_format("file"); |
84a12e66 | 637 | } |
98289620 | 638 | |
9e0b22f4 SH |
639 | p = strchr(filename, ':'); |
640 | assert(p != NULL); | |
1cec71e3 AL |
641 | len = p - filename; |
642 | if (len > sizeof(protocol) - 1) | |
643 | len = sizeof(protocol) - 1; | |
644 | memcpy(protocol, filename, len); | |
645 | protocol[len] = '\0'; | |
8a22f02a | 646 | QLIST_FOREACH(drv1, &bdrv_drivers, list) { |
5fafdf24 | 647 | if (drv1->protocol_name && |
8a22f02a | 648 | !strcmp(drv1->protocol_name, protocol)) { |
83f64091 | 649 | return drv1; |
8a22f02a | 650 | } |
83f64091 FB |
651 | } |
652 | return NULL; | |
653 | } | |
654 | ||
f500a6d3 | 655 | static int find_image_format(BlockDriverState *bs, const char *filename, |
34b5d2c6 | 656 | BlockDriver **pdrv, Error **errp) |
f3a5d3f8 | 657 | { |
f500a6d3 | 658 | int score, score_max; |
f3a5d3f8 CH |
659 | BlockDriver *drv1, *drv; |
660 | uint8_t buf[2048]; | |
f500a6d3 | 661 | int ret = 0; |
f8ea0b00 | 662 | |
08a00559 | 663 | /* Return the raw BlockDriver * to scsi-generic devices or empty drives */ |
8e895599 | 664 | if (bs->sg || !bdrv_is_inserted(bs) || bdrv_getlength(bs) == 0) { |
c98ac35d SW |
665 | drv = bdrv_find_format("raw"); |
666 | if (!drv) { | |
34b5d2c6 | 667 | error_setg(errp, "Could not find raw image format"); |
c98ac35d SW |
668 | ret = -ENOENT; |
669 | } | |
670 | *pdrv = drv; | |
671 | return ret; | |
1a396859 | 672 | } |
f8ea0b00 | 673 | |
83f64091 | 674 | ret = bdrv_pread(bs, 0, buf, sizeof(buf)); |
83f64091 | 675 | if (ret < 0) { |
34b5d2c6 HR |
676 | error_setg_errno(errp, -ret, "Could not read image for determining its " |
677 | "format"); | |
c98ac35d SW |
678 | *pdrv = NULL; |
679 | return ret; | |
83f64091 FB |
680 | } |
681 | ||
ea2384d3 | 682 | score_max = 0; |
84a12e66 | 683 | drv = NULL; |
8a22f02a | 684 | QLIST_FOREACH(drv1, &bdrv_drivers, list) { |
83f64091 FB |
685 | if (drv1->bdrv_probe) { |
686 | score = drv1->bdrv_probe(buf, ret, filename); | |
687 | if (score > score_max) { | |
688 | score_max = score; | |
689 | drv = drv1; | |
690 | } | |
0849bf08 | 691 | } |
fc01f7e7 | 692 | } |
c98ac35d | 693 | if (!drv) { |
34b5d2c6 HR |
694 | error_setg(errp, "Could not determine image format: No compatible " |
695 | "driver found"); | |
c98ac35d SW |
696 | ret = -ENOENT; |
697 | } | |
698 | *pdrv = drv; | |
699 | return ret; | |
ea2384d3 FB |
700 | } |
701 | ||
51762288 SH |
702 | /** |
703 | * Set the current 'total_sectors' value | |
704 | */ | |
705 | static int refresh_total_sectors(BlockDriverState *bs, int64_t hint) | |
706 | { | |
707 | BlockDriver *drv = bs->drv; | |
708 | ||
396759ad NB |
709 | /* Do not attempt drv->bdrv_getlength() on scsi-generic devices */ |
710 | if (bs->sg) | |
711 | return 0; | |
712 | ||
51762288 SH |
713 | /* query actual device if possible, otherwise just trust the hint */ |
714 | if (drv->bdrv_getlength) { | |
715 | int64_t length = drv->bdrv_getlength(bs); | |
716 | if (length < 0) { | |
717 | return length; | |
718 | } | |
7e382003 | 719 | hint = DIV_ROUND_UP(length, BDRV_SECTOR_SIZE); |
51762288 SH |
720 | } |
721 | ||
722 | bs->total_sectors = hint; | |
723 | return 0; | |
724 | } | |
725 | ||
9e8f1835 PB |
726 | /** |
727 | * Set open flags for a given discard mode | |
728 | * | |
729 | * Return 0 on success, -1 if the discard mode was invalid. | |
730 | */ | |
731 | int bdrv_parse_discard_flags(const char *mode, int *flags) | |
732 | { | |
733 | *flags &= ~BDRV_O_UNMAP; | |
734 | ||
735 | if (!strcmp(mode, "off") || !strcmp(mode, "ignore")) { | |
736 | /* do nothing */ | |
737 | } else if (!strcmp(mode, "on") || !strcmp(mode, "unmap")) { | |
738 | *flags |= BDRV_O_UNMAP; | |
739 | } else { | |
740 | return -1; | |
741 | } | |
742 | ||
743 | return 0; | |
744 | } | |
745 | ||
c3993cdc SH |
746 | /** |
747 | * Set open flags for a given cache mode | |
748 | * | |
749 | * Return 0 on success, -1 if the cache mode was invalid. | |
750 | */ | |
751 | int bdrv_parse_cache_flags(const char *mode, int *flags) | |
752 | { | |
753 | *flags &= ~BDRV_O_CACHE_MASK; | |
754 | ||
755 | if (!strcmp(mode, "off") || !strcmp(mode, "none")) { | |
756 | *flags |= BDRV_O_NOCACHE | BDRV_O_CACHE_WB; | |
92196b2f SH |
757 | } else if (!strcmp(mode, "directsync")) { |
758 | *flags |= BDRV_O_NOCACHE; | |
c3993cdc SH |
759 | } else if (!strcmp(mode, "writeback")) { |
760 | *flags |= BDRV_O_CACHE_WB; | |
761 | } else if (!strcmp(mode, "unsafe")) { | |
762 | *flags |= BDRV_O_CACHE_WB; | |
763 | *flags |= BDRV_O_NO_FLUSH; | |
764 | } else if (!strcmp(mode, "writethrough")) { | |
765 | /* this is the default */ | |
766 | } else { | |
767 | return -1; | |
768 | } | |
769 | ||
770 | return 0; | |
771 | } | |
772 | ||
53fec9d3 SH |
773 | /** |
774 | * The copy-on-read flag is actually a reference count so multiple users may | |
775 | * use the feature without worrying about clobbering its previous state. | |
776 | * Copy-on-read stays enabled until all users have called to disable it. | |
777 | */ | |
778 | void bdrv_enable_copy_on_read(BlockDriverState *bs) | |
779 | { | |
780 | bs->copy_on_read++; | |
781 | } | |
782 | ||
783 | void bdrv_disable_copy_on_read(BlockDriverState *bs) | |
784 | { | |
785 | assert(bs->copy_on_read > 0); | |
786 | bs->copy_on_read--; | |
787 | } | |
788 | ||
b1e6fc08 KW |
789 | /* |
790 | * Returns the flags that a temporary snapshot should get, based on the | |
791 | * originally requested flags (the originally requested image will have flags | |
792 | * like a backing file) | |
793 | */ | |
794 | static int bdrv_temp_snapshot_flags(int flags) | |
795 | { | |
796 | return (flags & ~BDRV_O_SNAPSHOT) | BDRV_O_TEMPORARY; | |
797 | } | |
798 | ||
0b50cc88 KW |
799 | /* |
800 | * Returns the flags that bs->file should get, based on the given flags for | |
801 | * the parent BDS | |
802 | */ | |
803 | static int bdrv_inherited_flags(int flags) | |
804 | { | |
805 | /* Enable protocol handling, disable format probing for bs->file */ | |
806 | flags |= BDRV_O_PROTOCOL; | |
807 | ||
808 | /* Our block drivers take care to send flushes and respect unmap policy, | |
809 | * so we can enable both unconditionally on lower layers. */ | |
810 | flags |= BDRV_O_CACHE_WB | BDRV_O_UNMAP; | |
811 | ||
0b50cc88 | 812 | /* Clear flags that only apply to the top layer */ |
5669b44d | 813 | flags &= ~(BDRV_O_SNAPSHOT | BDRV_O_NO_BACKING | BDRV_O_COPY_ON_READ); |
0b50cc88 KW |
814 | |
815 | return flags; | |
816 | } | |
817 | ||
317fc44e KW |
818 | /* |
819 | * Returns the flags that bs->backing_hd should get, based on the given flags | |
820 | * for the parent BDS | |
821 | */ | |
822 | static int bdrv_backing_flags(int flags) | |
823 | { | |
824 | /* backing files always opened read-only */ | |
825 | flags &= ~(BDRV_O_RDWR | BDRV_O_COPY_ON_READ); | |
826 | ||
827 | /* snapshot=on is handled on the top layer */ | |
8bfea15d | 828 | flags &= ~(BDRV_O_SNAPSHOT | BDRV_O_TEMPORARY); |
317fc44e KW |
829 | |
830 | return flags; | |
831 | } | |
832 | ||
7b272452 KW |
833 | static int bdrv_open_flags(BlockDriverState *bs, int flags) |
834 | { | |
835 | int open_flags = flags | BDRV_O_CACHE_WB; | |
836 | ||
837 | /* | |
838 | * Clear flags that are internal to the block layer before opening the | |
839 | * image. | |
840 | */ | |
20cca275 | 841 | open_flags &= ~(BDRV_O_SNAPSHOT | BDRV_O_NO_BACKING | BDRV_O_PROTOCOL); |
7b272452 KW |
842 | |
843 | /* | |
844 | * Snapshots should be writable. | |
845 | */ | |
8bfea15d | 846 | if (flags & BDRV_O_TEMPORARY) { |
7b272452 KW |
847 | open_flags |= BDRV_O_RDWR; |
848 | } | |
849 | ||
850 | return open_flags; | |
851 | } | |
852 | ||
636ea370 KW |
853 | static void bdrv_assign_node_name(BlockDriverState *bs, |
854 | const char *node_name, | |
855 | Error **errp) | |
6913c0c2 BC |
856 | { |
857 | if (!node_name) { | |
636ea370 | 858 | return; |
6913c0c2 BC |
859 | } |
860 | ||
861 | /* empty string node name is invalid */ | |
862 | if (node_name[0] == '\0') { | |
863 | error_setg(errp, "Empty node name"); | |
636ea370 | 864 | return; |
6913c0c2 BC |
865 | } |
866 | ||
0c5e94ee BC |
867 | /* takes care of avoiding namespaces collisions */ |
868 | if (bdrv_find(node_name)) { | |
869 | error_setg(errp, "node-name=%s is conflicting with a device id", | |
870 | node_name); | |
636ea370 | 871 | return; |
0c5e94ee BC |
872 | } |
873 | ||
6913c0c2 BC |
874 | /* takes care of avoiding duplicates node names */ |
875 | if (bdrv_find_node(node_name)) { | |
876 | error_setg(errp, "Duplicate node name"); | |
636ea370 | 877 | return; |
6913c0c2 BC |
878 | } |
879 | ||
880 | /* copy node name into the bs and insert it into the graph list */ | |
881 | pstrcpy(bs->node_name, sizeof(bs->node_name), node_name); | |
882 | QTAILQ_INSERT_TAIL(&graph_bdrv_states, bs, node_list); | |
6913c0c2 BC |
883 | } |
884 | ||
57915332 KW |
885 | /* |
886 | * Common part for opening disk images and files | |
b6ad491a KW |
887 | * |
888 | * Removes all processed options from *options. | |
57915332 | 889 | */ |
f500a6d3 | 890 | static int bdrv_open_common(BlockDriverState *bs, BlockDriverState *file, |
34b5d2c6 | 891 | QDict *options, int flags, BlockDriver *drv, Error **errp) |
57915332 KW |
892 | { |
893 | int ret, open_flags; | |
035fccdf | 894 | const char *filename; |
6913c0c2 | 895 | const char *node_name = NULL; |
34b5d2c6 | 896 | Error *local_err = NULL; |
57915332 KW |
897 | |
898 | assert(drv != NULL); | |
6405875c | 899 | assert(bs->file == NULL); |
707ff828 | 900 | assert(options != NULL && bs->options != options); |
57915332 | 901 | |
45673671 KW |
902 | if (file != NULL) { |
903 | filename = file->filename; | |
904 | } else { | |
905 | filename = qdict_get_try_str(options, "filename"); | |
906 | } | |
907 | ||
765003db KW |
908 | if (drv->bdrv_needs_filename && !filename) { |
909 | error_setg(errp, "The '%s' block driver requires a file name", | |
910 | drv->format_name); | |
911 | return -EINVAL; | |
912 | } | |
913 | ||
45673671 | 914 | trace_bdrv_open_common(bs, filename ?: "", flags, drv->format_name); |
28dcee10 | 915 | |
6913c0c2 | 916 | node_name = qdict_get_try_str(options, "node-name"); |
636ea370 | 917 | bdrv_assign_node_name(bs, node_name, &local_err); |
0fb6395c | 918 | if (local_err) { |
636ea370 KW |
919 | error_propagate(errp, local_err); |
920 | return -EINVAL; | |
6913c0c2 BC |
921 | } |
922 | qdict_del(options, "node-name"); | |
923 | ||
5d186eb0 KW |
924 | /* bdrv_open() with directly using a protocol as drv. This layer is already |
925 | * opened, so assign it to bs (while file becomes a closed BlockDriverState) | |
926 | * and return immediately. */ | |
927 | if (file != NULL && drv->bdrv_file_open) { | |
928 | bdrv_swap(file, bs); | |
929 | return 0; | |
930 | } | |
931 | ||
57915332 | 932 | bs->open_flags = flags; |
1b7fd729 | 933 | bs->guest_block_size = 512; |
c25f53b0 | 934 | bs->request_alignment = 512; |
0d51b4de | 935 | bs->zero_beyond_eof = true; |
b64ec4e4 FZ |
936 | open_flags = bdrv_open_flags(bs, flags); |
937 | bs->read_only = !(open_flags & BDRV_O_RDWR); | |
20cca275 | 938 | bs->growable = !!(flags & BDRV_O_PROTOCOL); |
b64ec4e4 FZ |
939 | |
940 | if (use_bdrv_whitelist && !bdrv_is_whitelisted(drv, bs->read_only)) { | |
8f94a6e4 KW |
941 | error_setg(errp, |
942 | !bs->read_only && bdrv_is_whitelisted(drv, true) | |
943 | ? "Driver '%s' can only be used for read-only devices" | |
944 | : "Driver '%s' is not whitelisted", | |
945 | drv->format_name); | |
b64ec4e4 FZ |
946 | return -ENOTSUP; |
947 | } | |
57915332 | 948 | |
53fec9d3 | 949 | assert(bs->copy_on_read == 0); /* bdrv_new() and bdrv_close() make it so */ |
0ebd24e0 KW |
950 | if (flags & BDRV_O_COPY_ON_READ) { |
951 | if (!bs->read_only) { | |
952 | bdrv_enable_copy_on_read(bs); | |
953 | } else { | |
954 | error_setg(errp, "Can't use copy-on-read on read-only device"); | |
955 | return -EINVAL; | |
956 | } | |
53fec9d3 SH |
957 | } |
958 | ||
c2ad1b0c KW |
959 | if (filename != NULL) { |
960 | pstrcpy(bs->filename, sizeof(bs->filename), filename); | |
961 | } else { | |
962 | bs->filename[0] = '\0'; | |
963 | } | |
57915332 | 964 | |
57915332 | 965 | bs->drv = drv; |
7267c094 | 966 | bs->opaque = g_malloc0(drv->instance_size); |
57915332 | 967 | |
03f541bd | 968 | bs->enable_write_cache = !!(flags & BDRV_O_CACHE_WB); |
e7c63796 | 969 | |
66f82cee KW |
970 | /* Open the image, either directly or using a protocol */ |
971 | if (drv->bdrv_file_open) { | |
5d186eb0 | 972 | assert(file == NULL); |
030be321 | 973 | assert(!drv->bdrv_needs_filename || filename != NULL); |
34b5d2c6 | 974 | ret = drv->bdrv_file_open(bs, options, open_flags, &local_err); |
f500a6d3 | 975 | } else { |
2af5ef70 | 976 | if (file == NULL) { |
34b5d2c6 HR |
977 | error_setg(errp, "Can't use '%s' as a block driver for the " |
978 | "protocol level", drv->format_name); | |
2af5ef70 KW |
979 | ret = -EINVAL; |
980 | goto free_and_fail; | |
981 | } | |
f500a6d3 | 982 | bs->file = file; |
34b5d2c6 | 983 | ret = drv->bdrv_open(bs, options, open_flags, &local_err); |
66f82cee KW |
984 | } |
985 | ||
57915332 | 986 | if (ret < 0) { |
84d18f06 | 987 | if (local_err) { |
34b5d2c6 | 988 | error_propagate(errp, local_err); |
2fa9aa59 DH |
989 | } else if (bs->filename[0]) { |
990 | error_setg_errno(errp, -ret, "Could not open '%s'", bs->filename); | |
34b5d2c6 HR |
991 | } else { |
992 | error_setg_errno(errp, -ret, "Could not open image"); | |
993 | } | |
57915332 KW |
994 | goto free_and_fail; |
995 | } | |
996 | ||
51762288 SH |
997 | ret = refresh_total_sectors(bs, bs->total_sectors); |
998 | if (ret < 0) { | |
34b5d2c6 | 999 | error_setg_errno(errp, -ret, "Could not refresh total sector count"); |
51762288 | 1000 | goto free_and_fail; |
57915332 | 1001 | } |
51762288 | 1002 | |
3baca891 KW |
1003 | bdrv_refresh_limits(bs, &local_err); |
1004 | if (local_err) { | |
1005 | error_propagate(errp, local_err); | |
1006 | ret = -EINVAL; | |
1007 | goto free_and_fail; | |
1008 | } | |
1009 | ||
c25f53b0 | 1010 | assert(bdrv_opt_mem_align(bs) != 0); |
47ea2de2 | 1011 | assert((bs->request_alignment != 0) || bs->sg); |
57915332 KW |
1012 | return 0; |
1013 | ||
1014 | free_and_fail: | |
f500a6d3 | 1015 | bs->file = NULL; |
7267c094 | 1016 | g_free(bs->opaque); |
57915332 KW |
1017 | bs->opaque = NULL; |
1018 | bs->drv = NULL; | |
1019 | return ret; | |
1020 | } | |
1021 | ||
5e5c4f63 KW |
1022 | static QDict *parse_json_filename(const char *filename, Error **errp) |
1023 | { | |
1024 | QObject *options_obj; | |
1025 | QDict *options; | |
1026 | int ret; | |
1027 | ||
1028 | ret = strstart(filename, "json:", &filename); | |
1029 | assert(ret); | |
1030 | ||
1031 | options_obj = qobject_from_json(filename); | |
1032 | if (!options_obj) { | |
1033 | error_setg(errp, "Could not parse the JSON options"); | |
1034 | return NULL; | |
1035 | } | |
1036 | ||
1037 | if (qobject_type(options_obj) != QTYPE_QDICT) { | |
1038 | qobject_decref(options_obj); | |
1039 | error_setg(errp, "Invalid JSON object given"); | |
1040 | return NULL; | |
1041 | } | |
1042 | ||
1043 | options = qobject_to_qdict(options_obj); | |
1044 | qdict_flatten(options); | |
1045 | ||
1046 | return options; | |
1047 | } | |
1048 | ||
b6ce07aa | 1049 | /* |
f54120ff KW |
1050 | * Fills in default options for opening images and converts the legacy |
1051 | * filename/flags pair to option QDict entries. | |
b6ce07aa | 1052 | */ |
5e5c4f63 | 1053 | static int bdrv_fill_options(QDict **options, const char **pfilename, int flags, |
17b005f1 | 1054 | BlockDriver *drv, Error **errp) |
ea2384d3 | 1055 | { |
5e5c4f63 | 1056 | const char *filename = *pfilename; |
c2ad1b0c | 1057 | const char *drvname; |
462f5bcf | 1058 | bool protocol = flags & BDRV_O_PROTOCOL; |
e3fa4bfa | 1059 | bool parse_filename = false; |
34b5d2c6 | 1060 | Error *local_err = NULL; |
83f64091 | 1061 | |
5e5c4f63 KW |
1062 | /* Parse json: pseudo-protocol */ |
1063 | if (filename && g_str_has_prefix(filename, "json:")) { | |
1064 | QDict *json_options = parse_json_filename(filename, &local_err); | |
1065 | if (local_err) { | |
1066 | error_propagate(errp, local_err); | |
1067 | return -EINVAL; | |
1068 | } | |
1069 | ||
1070 | /* Options given in the filename have lower priority than options | |
1071 | * specified directly */ | |
1072 | qdict_join(*options, json_options, false); | |
1073 | QDECREF(json_options); | |
1074 | *pfilename = filename = NULL; | |
1075 | } | |
1076 | ||
035fccdf | 1077 | /* Fetch the file name from the options QDict if necessary */ |
17b005f1 | 1078 | if (protocol && filename) { |
f54120ff KW |
1079 | if (!qdict_haskey(*options, "filename")) { |
1080 | qdict_put(*options, "filename", qstring_from_str(filename)); | |
1081 | parse_filename = true; | |
1082 | } else { | |
1083 | error_setg(errp, "Can't specify 'file' and 'filename' options at " | |
1084 | "the same time"); | |
1085 | return -EINVAL; | |
1086 | } | |
035fccdf KW |
1087 | } |
1088 | ||
c2ad1b0c | 1089 | /* Find the right block driver */ |
f54120ff | 1090 | filename = qdict_get_try_str(*options, "filename"); |
5acd9d81 | 1091 | drvname = qdict_get_try_str(*options, "driver"); |
f54120ff | 1092 | |
17b005f1 KW |
1093 | if (drv) { |
1094 | if (drvname) { | |
1095 | error_setg(errp, "Driver specified twice"); | |
1096 | return -EINVAL; | |
1097 | } | |
1098 | drvname = drv->format_name; | |
1099 | qdict_put(*options, "driver", qstring_from_str(drvname)); | |
1100 | } else { | |
1101 | if (!drvname && protocol) { | |
1102 | if (filename) { | |
1103 | drv = bdrv_find_protocol(filename, parse_filename); | |
1104 | if (!drv) { | |
1105 | error_setg(errp, "Unknown protocol"); | |
1106 | return -EINVAL; | |
1107 | } | |
1108 | ||
1109 | drvname = drv->format_name; | |
1110 | qdict_put(*options, "driver", qstring_from_str(drvname)); | |
1111 | } else { | |
1112 | error_setg(errp, "Must specify either driver or file"); | |
f54120ff KW |
1113 | return -EINVAL; |
1114 | } | |
17b005f1 KW |
1115 | } else if (drvname) { |
1116 | drv = bdrv_find_format(drvname); | |
1117 | if (!drv) { | |
1118 | error_setg(errp, "Unknown driver '%s'", drvname); | |
1119 | return -ENOENT; | |
1120 | } | |
98289620 | 1121 | } |
c2ad1b0c KW |
1122 | } |
1123 | ||
17b005f1 | 1124 | assert(drv || !protocol); |
c2ad1b0c | 1125 | |
f54120ff | 1126 | /* Driver-specific filename parsing */ |
17b005f1 | 1127 | if (drv && drv->bdrv_parse_filename && parse_filename) { |
5acd9d81 | 1128 | drv->bdrv_parse_filename(filename, *options, &local_err); |
84d18f06 | 1129 | if (local_err) { |
34b5d2c6 | 1130 | error_propagate(errp, local_err); |
f54120ff | 1131 | return -EINVAL; |
6963a30d | 1132 | } |
cd5d031e HR |
1133 | |
1134 | if (!drv->bdrv_needs_filename) { | |
1135 | qdict_del(*options, "filename"); | |
cd5d031e | 1136 | } |
6963a30d KW |
1137 | } |
1138 | ||
f54120ff KW |
1139 | return 0; |
1140 | } | |
1141 | ||
8d24cce1 FZ |
1142 | void bdrv_set_backing_hd(BlockDriverState *bs, BlockDriverState *backing_hd) |
1143 | { | |
1144 | ||
826b6ca0 FZ |
1145 | if (bs->backing_hd) { |
1146 | assert(bs->backing_blocker); | |
1147 | bdrv_op_unblock_all(bs->backing_hd, bs->backing_blocker); | |
1148 | } else if (backing_hd) { | |
1149 | error_setg(&bs->backing_blocker, | |
1150 | "device is used as backing hd of '%s'", | |
1151 | bs->device_name); | |
1152 | } | |
1153 | ||
8d24cce1 FZ |
1154 | bs->backing_hd = backing_hd; |
1155 | if (!backing_hd) { | |
826b6ca0 FZ |
1156 | error_free(bs->backing_blocker); |
1157 | bs->backing_blocker = NULL; | |
8d24cce1 FZ |
1158 | goto out; |
1159 | } | |
1160 | bs->open_flags &= ~BDRV_O_NO_BACKING; | |
1161 | pstrcpy(bs->backing_file, sizeof(bs->backing_file), backing_hd->filename); | |
1162 | pstrcpy(bs->backing_format, sizeof(bs->backing_format), | |
1163 | backing_hd->drv ? backing_hd->drv->format_name : ""); | |
826b6ca0 FZ |
1164 | |
1165 | bdrv_op_block_all(bs->backing_hd, bs->backing_blocker); | |
1166 | /* Otherwise we won't be able to commit due to check in bdrv_commit */ | |
1167 | bdrv_op_unblock(bs->backing_hd, BLOCK_OP_TYPE_COMMIT, | |
1168 | bs->backing_blocker); | |
8d24cce1 | 1169 | out: |
3baca891 | 1170 | bdrv_refresh_limits(bs, NULL); |
8d24cce1 FZ |
1171 | } |
1172 | ||
31ca6d07 KW |
1173 | /* |
1174 | * Opens the backing file for a BlockDriverState if not yet open | |
1175 | * | |
1176 | * options is a QDict of options to pass to the block drivers, or NULL for an | |
1177 | * empty set of options. The reference to the QDict is transferred to this | |
1178 | * function (even on failure), so if the caller intends to reuse the dictionary, | |
1179 | * it needs to use QINCREF() before calling bdrv_file_open. | |
1180 | */ | |
34b5d2c6 | 1181 | int bdrv_open_backing_file(BlockDriverState *bs, QDict *options, Error **errp) |
9156df12 | 1182 | { |
1ba4b6a5 | 1183 | char *backing_filename = g_malloc0(PATH_MAX); |
317fc44e | 1184 | int ret = 0; |
9156df12 | 1185 | BlockDriver *back_drv = NULL; |
8d24cce1 | 1186 | BlockDriverState *backing_hd; |
34b5d2c6 | 1187 | Error *local_err = NULL; |
9156df12 PB |
1188 | |
1189 | if (bs->backing_hd != NULL) { | |
31ca6d07 | 1190 | QDECREF(options); |
1ba4b6a5 | 1191 | goto free_exit; |
9156df12 PB |
1192 | } |
1193 | ||
31ca6d07 KW |
1194 | /* NULL means an empty set of options */ |
1195 | if (options == NULL) { | |
1196 | options = qdict_new(); | |
1197 | } | |
1198 | ||
9156df12 | 1199 | bs->open_flags &= ~BDRV_O_NO_BACKING; |
1cb6f506 KW |
1200 | if (qdict_haskey(options, "file.filename")) { |
1201 | backing_filename[0] = '\0'; | |
1202 | } else if (bs->backing_file[0] == '\0' && qdict_size(options) == 0) { | |
31ca6d07 | 1203 | QDECREF(options); |
1ba4b6a5 | 1204 | goto free_exit; |
dbecebdd | 1205 | } else { |
1ba4b6a5 | 1206 | bdrv_get_full_backing_filename(bs, backing_filename, PATH_MAX); |
9156df12 PB |
1207 | } |
1208 | ||
8ee79e70 KW |
1209 | if (!bs->drv || !bs->drv->supports_backing) { |
1210 | ret = -EINVAL; | |
1211 | error_setg(errp, "Driver doesn't support backing files"); | |
1212 | QDECREF(options); | |
1213 | goto free_exit; | |
1214 | } | |
1215 | ||
8d24cce1 FZ |
1216 | backing_hd = bdrv_new("", errp); |
1217 | ||
9156df12 PB |
1218 | if (bs->backing_format[0] != '\0') { |
1219 | back_drv = bdrv_find_format(bs->backing_format); | |
1220 | } | |
1221 | ||
f67503e5 | 1222 | assert(bs->backing_hd == NULL); |
8d24cce1 | 1223 | ret = bdrv_open(&backing_hd, |
ddf5636d | 1224 | *backing_filename ? backing_filename : NULL, NULL, options, |
317fc44e | 1225 | bdrv_backing_flags(bs->open_flags), back_drv, &local_err); |
9156df12 | 1226 | if (ret < 0) { |
8d24cce1 FZ |
1227 | bdrv_unref(backing_hd); |
1228 | backing_hd = NULL; | |
9156df12 | 1229 | bs->open_flags |= BDRV_O_NO_BACKING; |
b04b6b6e FZ |
1230 | error_setg(errp, "Could not open backing file: %s", |
1231 | error_get_pretty(local_err)); | |
1232 | error_free(local_err); | |
1ba4b6a5 | 1233 | goto free_exit; |
9156df12 | 1234 | } |
8d24cce1 | 1235 | bdrv_set_backing_hd(bs, backing_hd); |
d80ac658 | 1236 | |
1ba4b6a5 BC |
1237 | free_exit: |
1238 | g_free(backing_filename); | |
1239 | return ret; | |
9156df12 PB |
1240 | } |
1241 | ||
da557aac HR |
1242 | /* |
1243 | * Opens a disk image whose options are given as BlockdevRef in another block | |
1244 | * device's options. | |
1245 | * | |
da557aac HR |
1246 | * If allow_none is true, no image will be opened if filename is false and no |
1247 | * BlockdevRef is given. *pbs will remain unchanged and 0 will be returned. | |
1248 | * | |
1249 | * bdrev_key specifies the key for the image's BlockdevRef in the options QDict. | |
1250 | * That QDict has to be flattened; therefore, if the BlockdevRef is a QDict | |
1251 | * itself, all options starting with "${bdref_key}." are considered part of the | |
1252 | * BlockdevRef. | |
1253 | * | |
1254 | * The BlockdevRef will be removed from the options QDict. | |
f67503e5 HR |
1255 | * |
1256 | * To conform with the behavior of bdrv_open(), *pbs has to be NULL. | |
da557aac HR |
1257 | */ |
1258 | int bdrv_open_image(BlockDriverState **pbs, const char *filename, | |
1259 | QDict *options, const char *bdref_key, int flags, | |
f7d9fd8c | 1260 | bool allow_none, Error **errp) |
da557aac HR |
1261 | { |
1262 | QDict *image_options; | |
1263 | int ret; | |
1264 | char *bdref_key_dot; | |
1265 | const char *reference; | |
1266 | ||
f67503e5 HR |
1267 | assert(pbs); |
1268 | assert(*pbs == NULL); | |
1269 | ||
da557aac HR |
1270 | bdref_key_dot = g_strdup_printf("%s.", bdref_key); |
1271 | qdict_extract_subqdict(options, &image_options, bdref_key_dot); | |
1272 | g_free(bdref_key_dot); | |
1273 | ||
1274 | reference = qdict_get_try_str(options, bdref_key); | |
1275 | if (!filename && !reference && !qdict_size(image_options)) { | |
1276 | if (allow_none) { | |
1277 | ret = 0; | |
1278 | } else { | |
1279 | error_setg(errp, "A block device must be specified for \"%s\"", | |
1280 | bdref_key); | |
1281 | ret = -EINVAL; | |
1282 | } | |
b20e61e0 | 1283 | QDECREF(image_options); |
da557aac HR |
1284 | goto done; |
1285 | } | |
1286 | ||
f7d9fd8c | 1287 | ret = bdrv_open(pbs, filename, reference, image_options, flags, NULL, errp); |
da557aac HR |
1288 | |
1289 | done: | |
1290 | qdict_del(options, bdref_key); | |
1291 | return ret; | |
1292 | } | |
1293 | ||
6b8aeca5 | 1294 | int bdrv_append_temp_snapshot(BlockDriverState *bs, int flags, Error **errp) |
b998875d KW |
1295 | { |
1296 | /* TODO: extra byte is a hack to ensure MAX_PATH space on Windows. */ | |
1ba4b6a5 | 1297 | char *tmp_filename = g_malloc0(PATH_MAX + 1); |
b998875d KW |
1298 | int64_t total_size; |
1299 | BlockDriver *bdrv_qcow2; | |
83d0521a | 1300 | QemuOpts *opts = NULL; |
b998875d KW |
1301 | QDict *snapshot_options; |
1302 | BlockDriverState *bs_snapshot; | |
1303 | Error *local_err; | |
1304 | int ret; | |
1305 | ||
1306 | /* if snapshot, we create a temporary backing file and open it | |
1307 | instead of opening 'filename' directly */ | |
1308 | ||
1309 | /* Get the required size from the image */ | |
f187743a KW |
1310 | total_size = bdrv_getlength(bs); |
1311 | if (total_size < 0) { | |
6b8aeca5 | 1312 | ret = total_size; |
f187743a | 1313 | error_setg_errno(errp, -total_size, "Could not get image size"); |
1ba4b6a5 | 1314 | goto out; |
f187743a KW |
1315 | } |
1316 | total_size &= BDRV_SECTOR_MASK; | |
b998875d KW |
1317 | |
1318 | /* Create the temporary image */ | |
1ba4b6a5 | 1319 | ret = get_tmp_filename(tmp_filename, PATH_MAX + 1); |
b998875d KW |
1320 | if (ret < 0) { |
1321 | error_setg_errno(errp, -ret, "Could not get temporary filename"); | |
1ba4b6a5 | 1322 | goto out; |
b998875d KW |
1323 | } |
1324 | ||
1325 | bdrv_qcow2 = bdrv_find_format("qcow2"); | |
c282e1fd CL |
1326 | opts = qemu_opts_create(bdrv_qcow2->create_opts, NULL, 0, |
1327 | &error_abort); | |
83d0521a | 1328 | qemu_opt_set_number(opts, BLOCK_OPT_SIZE, total_size); |
c282e1fd | 1329 | ret = bdrv_create(bdrv_qcow2, tmp_filename, opts, &local_err); |
83d0521a | 1330 | qemu_opts_del(opts); |
b998875d KW |
1331 | if (ret < 0) { |
1332 | error_setg_errno(errp, -ret, "Could not create temporary overlay " | |
1333 | "'%s': %s", tmp_filename, | |
1334 | error_get_pretty(local_err)); | |
1335 | error_free(local_err); | |
1ba4b6a5 | 1336 | goto out; |
b998875d KW |
1337 | } |
1338 | ||
1339 | /* Prepare a new options QDict for the temporary file */ | |
1340 | snapshot_options = qdict_new(); | |
1341 | qdict_put(snapshot_options, "file.driver", | |
1342 | qstring_from_str("file")); | |
1343 | qdict_put(snapshot_options, "file.filename", | |
1344 | qstring_from_str(tmp_filename)); | |
1345 | ||
98522f63 | 1346 | bs_snapshot = bdrv_new("", &error_abort); |
b998875d KW |
1347 | |
1348 | ret = bdrv_open(&bs_snapshot, NULL, NULL, snapshot_options, | |
b1e6fc08 | 1349 | flags, bdrv_qcow2, &local_err); |
b998875d KW |
1350 | if (ret < 0) { |
1351 | error_propagate(errp, local_err); | |
1ba4b6a5 | 1352 | goto out; |
b998875d KW |
1353 | } |
1354 | ||
1355 | bdrv_append(bs_snapshot, bs); | |
1ba4b6a5 BC |
1356 | |
1357 | out: | |
1358 | g_free(tmp_filename); | |
6b8aeca5 | 1359 | return ret; |
b998875d KW |
1360 | } |
1361 | ||
b6ce07aa KW |
1362 | /* |
1363 | * Opens a disk image (raw, qcow2, vmdk, ...) | |
de9c0cec KW |
1364 | * |
1365 | * options is a QDict of options to pass to the block drivers, or NULL for an | |
1366 | * empty set of options. The reference to the QDict belongs to the block layer | |
1367 | * after the call (even on failure), so if the caller intends to reuse the | |
1368 | * dictionary, it needs to use QINCREF() before calling bdrv_open. | |
f67503e5 HR |
1369 | * |
1370 | * If *pbs is NULL, a new BDS will be created with a pointer to it stored there. | |
1371 | * If it is not NULL, the referenced BDS will be reused. | |
ddf5636d HR |
1372 | * |
1373 | * The reference parameter may be used to specify an existing block device which | |
1374 | * should be opened. If specified, neither options nor a filename may be given, | |
1375 | * nor can an existing BDS be reused (that is, *pbs has to be NULL). | |
b6ce07aa | 1376 | */ |
ddf5636d HR |
1377 | int bdrv_open(BlockDriverState **pbs, const char *filename, |
1378 | const char *reference, QDict *options, int flags, | |
1379 | BlockDriver *drv, Error **errp) | |
ea2384d3 | 1380 | { |
b6ce07aa | 1381 | int ret; |
f67503e5 | 1382 | BlockDriverState *file = NULL, *bs; |
74fe54f2 | 1383 | const char *drvname; |
34b5d2c6 | 1384 | Error *local_err = NULL; |
b1e6fc08 | 1385 | int snapshot_flags = 0; |
712e7874 | 1386 | |
f67503e5 HR |
1387 | assert(pbs); |
1388 | ||
ddf5636d HR |
1389 | if (reference) { |
1390 | bool options_non_empty = options ? qdict_size(options) : false; | |
1391 | QDECREF(options); | |
1392 | ||
1393 | if (*pbs) { | |
1394 | error_setg(errp, "Cannot reuse an existing BDS when referencing " | |
1395 | "another block device"); | |
1396 | return -EINVAL; | |
1397 | } | |
1398 | ||
1399 | if (filename || options_non_empty) { | |
1400 | error_setg(errp, "Cannot reference an existing block device with " | |
1401 | "additional options or a new filename"); | |
1402 | return -EINVAL; | |
1403 | } | |
1404 | ||
1405 | bs = bdrv_lookup_bs(reference, reference, errp); | |
1406 | if (!bs) { | |
1407 | return -ENODEV; | |
1408 | } | |
1409 | bdrv_ref(bs); | |
1410 | *pbs = bs; | |
1411 | return 0; | |
1412 | } | |
1413 | ||
f67503e5 HR |
1414 | if (*pbs) { |
1415 | bs = *pbs; | |
1416 | } else { | |
98522f63 | 1417 | bs = bdrv_new("", &error_abort); |
f67503e5 HR |
1418 | } |
1419 | ||
de9c0cec KW |
1420 | /* NULL means an empty set of options */ |
1421 | if (options == NULL) { | |
1422 | options = qdict_new(); | |
1423 | } | |
1424 | ||
17b005f1 | 1425 | ret = bdrv_fill_options(&options, &filename, flags, drv, &local_err); |
462f5bcf KW |
1426 | if (local_err) { |
1427 | goto fail; | |
1428 | } | |
1429 | ||
76c591b0 KW |
1430 | /* Find the right image format driver */ |
1431 | drv = NULL; | |
1432 | drvname = qdict_get_try_str(options, "driver"); | |
1433 | if (drvname) { | |
1434 | drv = bdrv_find_format(drvname); | |
1435 | qdict_del(options, "driver"); | |
1436 | if (!drv) { | |
1437 | error_setg(errp, "Unknown driver: '%s'", drvname); | |
1438 | ret = -EINVAL; | |
1439 | goto fail; | |
1440 | } | |
1441 | } | |
1442 | ||
1443 | assert(drvname || !(flags & BDRV_O_PROTOCOL)); | |
1444 | if (drv && !drv->bdrv_file_open) { | |
1445 | /* If the user explicitly wants a format driver here, we'll need to add | |
1446 | * another layer for the protocol in bs->file */ | |
1447 | flags &= ~BDRV_O_PROTOCOL; | |
1448 | } | |
1449 | ||
de9c0cec | 1450 | bs->options = options; |
b6ad491a | 1451 | options = qdict_clone_shallow(options); |
de9c0cec | 1452 | |
f500a6d3 | 1453 | /* Open image file without format layer */ |
f4788adc KW |
1454 | if ((flags & BDRV_O_PROTOCOL) == 0) { |
1455 | if (flags & BDRV_O_RDWR) { | |
1456 | flags |= BDRV_O_ALLOW_RDWR; | |
1457 | } | |
1458 | if (flags & BDRV_O_SNAPSHOT) { | |
1459 | snapshot_flags = bdrv_temp_snapshot_flags(flags); | |
1460 | flags = bdrv_backing_flags(flags); | |
1461 | } | |
f500a6d3 | 1462 | |
f4788adc KW |
1463 | assert(file == NULL); |
1464 | ret = bdrv_open_image(&file, filename, options, "file", | |
1465 | bdrv_inherited_flags(flags), | |
1466 | true, &local_err); | |
1467 | if (ret < 0) { | |
1468 | goto fail; | |
1469 | } | |
f500a6d3 KW |
1470 | } |
1471 | ||
76c591b0 KW |
1472 | /* Image format probing */ |
1473 | if (!drv && file) { | |
17b005f1 KW |
1474 | ret = find_image_format(file, filename, &drv, &local_err); |
1475 | if (ret < 0) { | |
8bfea15d | 1476 | goto fail; |
2a05cbe4 | 1477 | } |
76c591b0 | 1478 | } else if (!drv) { |
17b005f1 KW |
1479 | error_setg(errp, "Must specify either driver or file"); |
1480 | ret = -EINVAL; | |
8bfea15d | 1481 | goto fail; |
ea2384d3 | 1482 | } |
b6ce07aa KW |
1483 | |
1484 | /* Open the image */ | |
34b5d2c6 | 1485 | ret = bdrv_open_common(bs, file, options, flags, drv, &local_err); |
b6ce07aa | 1486 | if (ret < 0) { |
8bfea15d | 1487 | goto fail; |
6987307c CH |
1488 | } |
1489 | ||
2a05cbe4 | 1490 | if (file && (bs->file != file)) { |
4f6fd349 | 1491 | bdrv_unref(file); |
f500a6d3 KW |
1492 | file = NULL; |
1493 | } | |
1494 | ||
b6ce07aa | 1495 | /* If there is a backing file, use it */ |
9156df12 | 1496 | if ((flags & BDRV_O_NO_BACKING) == 0) { |
31ca6d07 KW |
1497 | QDict *backing_options; |
1498 | ||
5726d872 | 1499 | qdict_extract_subqdict(options, &backing_options, "backing."); |
34b5d2c6 | 1500 | ret = bdrv_open_backing_file(bs, backing_options, &local_err); |
b6ce07aa | 1501 | if (ret < 0) { |
b6ad491a | 1502 | goto close_and_fail; |
b6ce07aa | 1503 | } |
b6ce07aa KW |
1504 | } |
1505 | ||
b998875d KW |
1506 | /* For snapshot=on, create a temporary qcow2 overlay. bs points to the |
1507 | * temporary snapshot afterwards. */ | |
b1e6fc08 | 1508 | if (snapshot_flags) { |
6b8aeca5 | 1509 | ret = bdrv_append_temp_snapshot(bs, snapshot_flags, &local_err); |
b998875d | 1510 | if (local_err) { |
b998875d KW |
1511 | goto close_and_fail; |
1512 | } | |
1513 | } | |
1514 | ||
b6ad491a | 1515 | /* Check if any unknown options were used */ |
5acd9d81 | 1516 | if (options && (qdict_size(options) != 0)) { |
b6ad491a | 1517 | const QDictEntry *entry = qdict_first(options); |
5acd9d81 HR |
1518 | if (flags & BDRV_O_PROTOCOL) { |
1519 | error_setg(errp, "Block protocol '%s' doesn't support the option " | |
1520 | "'%s'", drv->format_name, entry->key); | |
1521 | } else { | |
1522 | error_setg(errp, "Block format '%s' used by device '%s' doesn't " | |
1523 | "support the option '%s'", drv->format_name, | |
1524 | bs->device_name, entry->key); | |
1525 | } | |
b6ad491a KW |
1526 | |
1527 | ret = -EINVAL; | |
1528 | goto close_and_fail; | |
1529 | } | |
b6ad491a | 1530 | |
b6ce07aa | 1531 | if (!bdrv_key_required(bs)) { |
7d4b4ba5 | 1532 | bdrv_dev_change_media_cb(bs, true); |
c3adb58f MA |
1533 | } else if (!runstate_check(RUN_STATE_PRELAUNCH) |
1534 | && !runstate_check(RUN_STATE_INMIGRATE) | |
1535 | && !runstate_check(RUN_STATE_PAUSED)) { /* HACK */ | |
1536 | error_setg(errp, | |
1537 | "Guest must be stopped for opening of encrypted image"); | |
1538 | ret = -EBUSY; | |
1539 | goto close_and_fail; | |
b6ce07aa KW |
1540 | } |
1541 | ||
c3adb58f | 1542 | QDECREF(options); |
f67503e5 | 1543 | *pbs = bs; |
b6ce07aa KW |
1544 | return 0; |
1545 | ||
8bfea15d | 1546 | fail: |
f500a6d3 | 1547 | if (file != NULL) { |
4f6fd349 | 1548 | bdrv_unref(file); |
f500a6d3 | 1549 | } |
de9c0cec | 1550 | QDECREF(bs->options); |
b6ad491a | 1551 | QDECREF(options); |
de9c0cec | 1552 | bs->options = NULL; |
f67503e5 HR |
1553 | if (!*pbs) { |
1554 | /* If *pbs is NULL, a new BDS has been created in this function and | |
1555 | needs to be freed now. Otherwise, it does not need to be closed, | |
1556 | since it has not really been opened yet. */ | |
1557 | bdrv_unref(bs); | |
1558 | } | |
84d18f06 | 1559 | if (local_err) { |
34b5d2c6 HR |
1560 | error_propagate(errp, local_err); |
1561 | } | |
b6ad491a | 1562 | return ret; |
de9c0cec | 1563 | |
b6ad491a | 1564 | close_and_fail: |
f67503e5 HR |
1565 | /* See fail path, but now the BDS has to be always closed */ |
1566 | if (*pbs) { | |
1567 | bdrv_close(bs); | |
1568 | } else { | |
1569 | bdrv_unref(bs); | |
1570 | } | |
b6ad491a | 1571 | QDECREF(options); |
84d18f06 | 1572 | if (local_err) { |
34b5d2c6 HR |
1573 | error_propagate(errp, local_err); |
1574 | } | |
b6ce07aa KW |
1575 | return ret; |
1576 | } | |
1577 | ||
e971aa12 JC |
1578 | typedef struct BlockReopenQueueEntry { |
1579 | bool prepared; | |
1580 | BDRVReopenState state; | |
1581 | QSIMPLEQ_ENTRY(BlockReopenQueueEntry) entry; | |
1582 | } BlockReopenQueueEntry; | |
1583 | ||
1584 | /* | |
1585 | * Adds a BlockDriverState to a simple queue for an atomic, transactional | |
1586 | * reopen of multiple devices. | |
1587 | * | |
1588 | * bs_queue can either be an existing BlockReopenQueue that has had QSIMPLE_INIT | |
1589 | * already performed, or alternatively may be NULL a new BlockReopenQueue will | |
1590 | * be created and initialized. This newly created BlockReopenQueue should be | |
1591 | * passed back in for subsequent calls that are intended to be of the same | |
1592 | * atomic 'set'. | |
1593 | * | |
1594 | * bs is the BlockDriverState to add to the reopen queue. | |
1595 | * | |
1596 | * flags contains the open flags for the associated bs | |
1597 | * | |
1598 | * returns a pointer to bs_queue, which is either the newly allocated | |
1599 | * bs_queue, or the existing bs_queue being used. | |
1600 | * | |
1601 | */ | |
1602 | BlockReopenQueue *bdrv_reopen_queue(BlockReopenQueue *bs_queue, | |
1603 | BlockDriverState *bs, int flags) | |
1604 | { | |
1605 | assert(bs != NULL); | |
1606 | ||
1607 | BlockReopenQueueEntry *bs_entry; | |
1608 | if (bs_queue == NULL) { | |
1609 | bs_queue = g_new0(BlockReopenQueue, 1); | |
1610 | QSIMPLEQ_INIT(bs_queue); | |
1611 | } | |
1612 | ||
f1f25a2e KW |
1613 | /* bdrv_open() masks this flag out */ |
1614 | flags &= ~BDRV_O_PROTOCOL; | |
1615 | ||
e971aa12 | 1616 | if (bs->file) { |
f1f25a2e | 1617 | bdrv_reopen_queue(bs_queue, bs->file, bdrv_inherited_flags(flags)); |
e971aa12 JC |
1618 | } |
1619 | ||
1620 | bs_entry = g_new0(BlockReopenQueueEntry, 1); | |
1621 | QSIMPLEQ_INSERT_TAIL(bs_queue, bs_entry, entry); | |
1622 | ||
1623 | bs_entry->state.bs = bs; | |
1624 | bs_entry->state.flags = flags; | |
1625 | ||
1626 | return bs_queue; | |
1627 | } | |
1628 | ||
1629 | /* | |
1630 | * Reopen multiple BlockDriverStates atomically & transactionally. | |
1631 | * | |
1632 | * The queue passed in (bs_queue) must have been built up previous | |
1633 | * via bdrv_reopen_queue(). | |
1634 | * | |
1635 | * Reopens all BDS specified in the queue, with the appropriate | |
1636 | * flags. All devices are prepared for reopen, and failure of any | |
1637 | * device will cause all device changes to be abandonded, and intermediate | |
1638 | * data cleaned up. | |
1639 | * | |
1640 | * If all devices prepare successfully, then the changes are committed | |
1641 | * to all devices. | |
1642 | * | |
1643 | */ | |
1644 | int bdrv_reopen_multiple(BlockReopenQueue *bs_queue, Error **errp) | |
1645 | { | |
1646 | int ret = -1; | |
1647 | BlockReopenQueueEntry *bs_entry, *next; | |
1648 | Error *local_err = NULL; | |
1649 | ||
1650 | assert(bs_queue != NULL); | |
1651 | ||
1652 | bdrv_drain_all(); | |
1653 | ||
1654 | QSIMPLEQ_FOREACH(bs_entry, bs_queue, entry) { | |
1655 | if (bdrv_reopen_prepare(&bs_entry->state, bs_queue, &local_err)) { | |
1656 | error_propagate(errp, local_err); | |
1657 | goto cleanup; | |
1658 | } | |
1659 | bs_entry->prepared = true; | |
1660 | } | |
1661 | ||
1662 | /* If we reach this point, we have success and just need to apply the | |
1663 | * changes | |
1664 | */ | |
1665 | QSIMPLEQ_FOREACH(bs_entry, bs_queue, entry) { | |
1666 | bdrv_reopen_commit(&bs_entry->state); | |
1667 | } | |
1668 | ||
1669 | ret = 0; | |
1670 | ||
1671 | cleanup: | |
1672 | QSIMPLEQ_FOREACH_SAFE(bs_entry, bs_queue, entry, next) { | |
1673 | if (ret && bs_entry->prepared) { | |
1674 | bdrv_reopen_abort(&bs_entry->state); | |
1675 | } | |
1676 | g_free(bs_entry); | |
1677 | } | |
1678 | g_free(bs_queue); | |
1679 | return ret; | |
1680 | } | |
1681 | ||
1682 | ||
1683 | /* Reopen a single BlockDriverState with the specified flags. */ | |
1684 | int bdrv_reopen(BlockDriverState *bs, int bdrv_flags, Error **errp) | |
1685 | { | |
1686 | int ret = -1; | |
1687 | Error *local_err = NULL; | |
1688 | BlockReopenQueue *queue = bdrv_reopen_queue(NULL, bs, bdrv_flags); | |
1689 | ||
1690 | ret = bdrv_reopen_multiple(queue, &local_err); | |
1691 | if (local_err != NULL) { | |
1692 | error_propagate(errp, local_err); | |
1693 | } | |
1694 | return ret; | |
1695 | } | |
1696 | ||
1697 | ||
1698 | /* | |
1699 | * Prepares a BlockDriverState for reopen. All changes are staged in the | |
1700 | * 'opaque' field of the BDRVReopenState, which is used and allocated by | |
1701 | * the block driver layer .bdrv_reopen_prepare() | |
1702 | * | |
1703 | * bs is the BlockDriverState to reopen | |
1704 | * flags are the new open flags | |
1705 | * queue is the reopen queue | |
1706 | * | |
1707 | * Returns 0 on success, non-zero on error. On error errp will be set | |
1708 | * as well. | |
1709 | * | |
1710 | * On failure, bdrv_reopen_abort() will be called to clean up any data. | |
1711 | * It is the responsibility of the caller to then call the abort() or | |
1712 | * commit() for any other BDS that have been left in a prepare() state | |
1713 | * | |
1714 | */ | |
1715 | int bdrv_reopen_prepare(BDRVReopenState *reopen_state, BlockReopenQueue *queue, | |
1716 | Error **errp) | |
1717 | { | |
1718 | int ret = -1; | |
1719 | Error *local_err = NULL; | |
1720 | BlockDriver *drv; | |
1721 | ||
1722 | assert(reopen_state != NULL); | |
1723 | assert(reopen_state->bs->drv != NULL); | |
1724 | drv = reopen_state->bs->drv; | |
1725 | ||
1726 | /* if we are to stay read-only, do not allow permission change | |
1727 | * to r/w */ | |
1728 | if (!(reopen_state->bs->open_flags & BDRV_O_ALLOW_RDWR) && | |
1729 | reopen_state->flags & BDRV_O_RDWR) { | |
1730 | error_set(errp, QERR_DEVICE_IS_READ_ONLY, | |
1731 | reopen_state->bs->device_name); | |
1732 | goto error; | |
1733 | } | |
1734 | ||
1735 | ||
1736 | ret = bdrv_flush(reopen_state->bs); | |
1737 | if (ret) { | |
1738 | error_set(errp, ERROR_CLASS_GENERIC_ERROR, "Error (%s) flushing drive", | |
1739 | strerror(-ret)); | |
1740 | goto error; | |
1741 | } | |
1742 | ||
1743 | if (drv->bdrv_reopen_prepare) { | |
1744 | ret = drv->bdrv_reopen_prepare(reopen_state, queue, &local_err); | |
1745 | if (ret) { | |
1746 | if (local_err != NULL) { | |
1747 | error_propagate(errp, local_err); | |
1748 | } else { | |
d8b6895f LC |
1749 | error_setg(errp, "failed while preparing to reopen image '%s'", |
1750 | reopen_state->bs->filename); | |
e971aa12 JC |
1751 | } |
1752 | goto error; | |
1753 | } | |
1754 | } else { | |
1755 | /* It is currently mandatory to have a bdrv_reopen_prepare() | |
1756 | * handler for each supported drv. */ | |
1757 | error_set(errp, QERR_BLOCK_FORMAT_FEATURE_NOT_SUPPORTED, | |
1758 | drv->format_name, reopen_state->bs->device_name, | |
1759 | "reopening of file"); | |
1760 | ret = -1; | |
1761 | goto error; | |
1762 | } | |
1763 | ||
1764 | ret = 0; | |
1765 | ||
1766 | error: | |
1767 | return ret; | |
1768 | } | |
1769 | ||
1770 | /* | |
1771 | * Takes the staged changes for the reopen from bdrv_reopen_prepare(), and | |
1772 | * makes them final by swapping the staging BlockDriverState contents into | |
1773 | * the active BlockDriverState contents. | |
1774 | */ | |
1775 | void bdrv_reopen_commit(BDRVReopenState *reopen_state) | |
1776 | { | |
1777 | BlockDriver *drv; | |
1778 | ||
1779 | assert(reopen_state != NULL); | |
1780 | drv = reopen_state->bs->drv; | |
1781 | assert(drv != NULL); | |
1782 | ||
1783 | /* If there are any driver level actions to take */ | |
1784 | if (drv->bdrv_reopen_commit) { | |
1785 | drv->bdrv_reopen_commit(reopen_state); | |
1786 | } | |
1787 | ||
1788 | /* set BDS specific flags now */ | |
1789 | reopen_state->bs->open_flags = reopen_state->flags; | |
1790 | reopen_state->bs->enable_write_cache = !!(reopen_state->flags & | |
1791 | BDRV_O_CACHE_WB); | |
1792 | reopen_state->bs->read_only = !(reopen_state->flags & BDRV_O_RDWR); | |
355ef4ac | 1793 | |
3baca891 | 1794 | bdrv_refresh_limits(reopen_state->bs, NULL); |
e971aa12 JC |
1795 | } |
1796 | ||
1797 | /* | |
1798 | * Abort the reopen, and delete and free the staged changes in | |
1799 | * reopen_state | |
1800 | */ | |
1801 | void bdrv_reopen_abort(BDRVReopenState *reopen_state) | |
1802 | { | |
1803 | BlockDriver *drv; | |
1804 | ||
1805 | assert(reopen_state != NULL); | |
1806 | drv = reopen_state->bs->drv; | |
1807 | assert(drv != NULL); | |
1808 | ||
1809 | if (drv->bdrv_reopen_abort) { | |
1810 | drv->bdrv_reopen_abort(reopen_state); | |
1811 | } | |
1812 | } | |
1813 | ||
1814 | ||
fc01f7e7 FB |
1815 | void bdrv_close(BlockDriverState *bs) |
1816 | { | |
3cbc002c PB |
1817 | if (bs->job) { |
1818 | block_job_cancel_sync(bs->job); | |
1819 | } | |
58fda173 SH |
1820 | bdrv_drain_all(); /* complete I/O */ |
1821 | bdrv_flush(bs); | |
1822 | bdrv_drain_all(); /* in case flush left pending I/O */ | |
d7d512f6 | 1823 | notifier_list_notify(&bs->close_notifiers, bs); |
7094f12f | 1824 | |
3cbc002c | 1825 | if (bs->drv) { |
557df6ac | 1826 | if (bs->backing_hd) { |
826b6ca0 FZ |
1827 | BlockDriverState *backing_hd = bs->backing_hd; |
1828 | bdrv_set_backing_hd(bs, NULL); | |
1829 | bdrv_unref(backing_hd); | |
557df6ac | 1830 | } |
ea2384d3 | 1831 | bs->drv->bdrv_close(bs); |
7267c094 | 1832 | g_free(bs->opaque); |
ea2384d3 FB |
1833 | bs->opaque = NULL; |
1834 | bs->drv = NULL; | |
53fec9d3 | 1835 | bs->copy_on_read = 0; |
a275fa42 PB |
1836 | bs->backing_file[0] = '\0'; |
1837 | bs->backing_format[0] = '\0'; | |
6405875c PB |
1838 | bs->total_sectors = 0; |
1839 | bs->encrypted = 0; | |
1840 | bs->valid_key = 0; | |
1841 | bs->sg = 0; | |
1842 | bs->growable = 0; | |
0d51b4de | 1843 | bs->zero_beyond_eof = false; |
de9c0cec KW |
1844 | QDECREF(bs->options); |
1845 | bs->options = NULL; | |
b338082b | 1846 | |
66f82cee | 1847 | if (bs->file != NULL) { |
4f6fd349 | 1848 | bdrv_unref(bs->file); |
0ac9377d | 1849 | bs->file = NULL; |
66f82cee | 1850 | } |
b338082b | 1851 | } |
98f90dba | 1852 | |
9ca11154 PH |
1853 | bdrv_dev_change_media_cb(bs, false); |
1854 | ||
98f90dba ZYW |
1855 | /*throttling disk I/O limits*/ |
1856 | if (bs->io_limits_enabled) { | |
1857 | bdrv_io_limits_disable(bs); | |
1858 | } | |
b338082b FB |
1859 | } |
1860 | ||
2bc93fed MK |
1861 | void bdrv_close_all(void) |
1862 | { | |
1863 | BlockDriverState *bs; | |
1864 | ||
dc364f4c | 1865 | QTAILQ_FOREACH(bs, &bdrv_states, device_list) { |
ed78cda3 SH |
1866 | AioContext *aio_context = bdrv_get_aio_context(bs); |
1867 | ||
1868 | aio_context_acquire(aio_context); | |
2bc93fed | 1869 | bdrv_close(bs); |
ed78cda3 | 1870 | aio_context_release(aio_context); |
2bc93fed MK |
1871 | } |
1872 | } | |
1873 | ||
88266f5a SH |
1874 | /* Check if any requests are in-flight (including throttled requests) */ |
1875 | static bool bdrv_requests_pending(BlockDriverState *bs) | |
1876 | { | |
1877 | if (!QLIST_EMPTY(&bs->tracked_requests)) { | |
1878 | return true; | |
1879 | } | |
cc0681c4 BC |
1880 | if (!qemu_co_queue_empty(&bs->throttled_reqs[0])) { |
1881 | return true; | |
1882 | } | |
1883 | if (!qemu_co_queue_empty(&bs->throttled_reqs[1])) { | |
88266f5a SH |
1884 | return true; |
1885 | } | |
1886 | if (bs->file && bdrv_requests_pending(bs->file)) { | |
1887 | return true; | |
1888 | } | |
1889 | if (bs->backing_hd && bdrv_requests_pending(bs->backing_hd)) { | |
1890 | return true; | |
1891 | } | |
1892 | return false; | |
1893 | } | |
1894 | ||
922453bc SH |
1895 | /* |
1896 | * Wait for pending requests to complete across all BlockDriverStates | |
1897 | * | |
1898 | * This function does not flush data to disk, use bdrv_flush_all() for that | |
1899 | * after calling this function. | |
4c355d53 ZYW |
1900 | * |
1901 | * Note that completion of an asynchronous I/O operation can trigger any | |
1902 | * number of other I/O operations on other devices---for example a coroutine | |
1903 | * can be arbitrarily complex and a constant flow of I/O can come until the | |
1904 | * coroutine is complete. Because of this, it is not possible to have a | |
1905 | * function to drain a single device's I/O queue. | |
922453bc SH |
1906 | */ |
1907 | void bdrv_drain_all(void) | |
1908 | { | |
88266f5a SH |
1909 | /* Always run first iteration so any pending completion BHs run */ |
1910 | bool busy = true; | |
922453bc SH |
1911 | BlockDriverState *bs; |
1912 | ||
88266f5a | 1913 | while (busy) { |
9b536adc SH |
1914 | busy = false; |
1915 | ||
dc364f4c | 1916 | QTAILQ_FOREACH(bs, &bdrv_states, device_list) { |
9b536adc SH |
1917 | AioContext *aio_context = bdrv_get_aio_context(bs); |
1918 | bool bs_busy; | |
1919 | ||
1920 | aio_context_acquire(aio_context); | |
448ad91d | 1921 | bdrv_flush_io_queue(bs); |
0b06ef3b | 1922 | bdrv_start_throttled_reqs(bs); |
9b536adc SH |
1923 | bs_busy = bdrv_requests_pending(bs); |
1924 | bs_busy |= aio_poll(aio_context, bs_busy); | |
1925 | aio_context_release(aio_context); | |
922453bc | 1926 | |
9b536adc SH |
1927 | busy |= bs_busy; |
1928 | } | |
922453bc SH |
1929 | } |
1930 | } | |
1931 | ||
dc364f4c BC |
1932 | /* make a BlockDriverState anonymous by removing from bdrv_state and |
1933 | * graph_bdrv_state list. | |
d22b2f41 RH |
1934 | Also, NULL terminate the device_name to prevent double remove */ |
1935 | void bdrv_make_anon(BlockDriverState *bs) | |
1936 | { | |
1937 | if (bs->device_name[0] != '\0') { | |
dc364f4c | 1938 | QTAILQ_REMOVE(&bdrv_states, bs, device_list); |
d22b2f41 RH |
1939 | } |
1940 | bs->device_name[0] = '\0'; | |
dc364f4c BC |
1941 | if (bs->node_name[0] != '\0') { |
1942 | QTAILQ_REMOVE(&graph_bdrv_states, bs, node_list); | |
1943 | } | |
1944 | bs->node_name[0] = '\0'; | |
d22b2f41 RH |
1945 | } |
1946 | ||
e023b2e2 PB |
1947 | static void bdrv_rebind(BlockDriverState *bs) |
1948 | { | |
1949 | if (bs->drv && bs->drv->bdrv_rebind) { | |
1950 | bs->drv->bdrv_rebind(bs); | |
1951 | } | |
1952 | } | |
1953 | ||
4ddc07ca PB |
1954 | static void bdrv_move_feature_fields(BlockDriverState *bs_dest, |
1955 | BlockDriverState *bs_src) | |
8802d1fd | 1956 | { |
4ddc07ca | 1957 | /* move some fields that need to stay attached to the device */ |
8802d1fd JC |
1958 | |
1959 | /* dev info */ | |
4ddc07ca PB |
1960 | bs_dest->dev_ops = bs_src->dev_ops; |
1961 | bs_dest->dev_opaque = bs_src->dev_opaque; | |
1962 | bs_dest->dev = bs_src->dev; | |
1b7fd729 | 1963 | bs_dest->guest_block_size = bs_src->guest_block_size; |
4ddc07ca | 1964 | bs_dest->copy_on_read = bs_src->copy_on_read; |
8802d1fd | 1965 | |
4ddc07ca | 1966 | bs_dest->enable_write_cache = bs_src->enable_write_cache; |
c4a248a1 | 1967 | |
cc0681c4 BC |
1968 | /* i/o throttled req */ |
1969 | memcpy(&bs_dest->throttle_state, | |
1970 | &bs_src->throttle_state, | |
1971 | sizeof(ThrottleState)); | |
1972 | bs_dest->throttled_reqs[0] = bs_src->throttled_reqs[0]; | |
1973 | bs_dest->throttled_reqs[1] = bs_src->throttled_reqs[1]; | |
4ddc07ca | 1974 | bs_dest->io_limits_enabled = bs_src->io_limits_enabled; |
8802d1fd | 1975 | |
8802d1fd | 1976 | /* r/w error */ |
4ddc07ca PB |
1977 | bs_dest->on_read_error = bs_src->on_read_error; |
1978 | bs_dest->on_write_error = bs_src->on_write_error; | |
8802d1fd JC |
1979 | |
1980 | /* i/o status */ | |
4ddc07ca PB |
1981 | bs_dest->iostatus_enabled = bs_src->iostatus_enabled; |
1982 | bs_dest->iostatus = bs_src->iostatus; | |
8802d1fd | 1983 | |
a9fc4408 | 1984 | /* dirty bitmap */ |
e4654d2d | 1985 | bs_dest->dirty_bitmaps = bs_src->dirty_bitmaps; |
a9fc4408 | 1986 | |
9fcb0251 FZ |
1987 | /* reference count */ |
1988 | bs_dest->refcnt = bs_src->refcnt; | |
1989 | ||
a9fc4408 | 1990 | /* job */ |
4ddc07ca | 1991 | bs_dest->job = bs_src->job; |
a9fc4408 | 1992 | |
8802d1fd | 1993 | /* keep the same entry in bdrv_states */ |
4ddc07ca PB |
1994 | pstrcpy(bs_dest->device_name, sizeof(bs_dest->device_name), |
1995 | bs_src->device_name); | |
dc364f4c | 1996 | bs_dest->device_list = bs_src->device_list; |
fbe40ff7 FZ |
1997 | memcpy(bs_dest->op_blockers, bs_src->op_blockers, |
1998 | sizeof(bs_dest->op_blockers)); | |
4ddc07ca | 1999 | } |
8802d1fd | 2000 | |
4ddc07ca PB |
2001 | /* |
2002 | * Swap bs contents for two image chains while they are live, | |
2003 | * while keeping required fields on the BlockDriverState that is | |
2004 | * actually attached to a device. | |
2005 | * | |
2006 | * This will modify the BlockDriverState fields, and swap contents | |
2007 | * between bs_new and bs_old. Both bs_new and bs_old are modified. | |
2008 | * | |
2009 | * bs_new is required to be anonymous. | |
2010 | * | |
2011 | * This function does not create any image files. | |
2012 | */ | |
2013 | void bdrv_swap(BlockDriverState *bs_new, BlockDriverState *bs_old) | |
2014 | { | |
2015 | BlockDriverState tmp; | |
f6801b83 | 2016 | |
90ce8a06 BC |
2017 | /* The code needs to swap the node_name but simply swapping node_list won't |
2018 | * work so first remove the nodes from the graph list, do the swap then | |
2019 | * insert them back if needed. | |
2020 | */ | |
2021 | if (bs_new->node_name[0] != '\0') { | |
2022 | QTAILQ_REMOVE(&graph_bdrv_states, bs_new, node_list); | |
2023 | } | |
2024 | if (bs_old->node_name[0] != '\0') { | |
2025 | QTAILQ_REMOVE(&graph_bdrv_states, bs_old, node_list); | |
2026 | } | |
2027 | ||
4ddc07ca PB |
2028 | /* bs_new must be anonymous and shouldn't have anything fancy enabled */ |
2029 | assert(bs_new->device_name[0] == '\0'); | |
e4654d2d | 2030 | assert(QLIST_EMPTY(&bs_new->dirty_bitmaps)); |
4ddc07ca PB |
2031 | assert(bs_new->job == NULL); |
2032 | assert(bs_new->dev == NULL); | |
4ddc07ca | 2033 | assert(bs_new->io_limits_enabled == false); |
cc0681c4 | 2034 | assert(!throttle_have_timer(&bs_new->throttle_state)); |
8802d1fd | 2035 | |
4ddc07ca PB |
2036 | tmp = *bs_new; |
2037 | *bs_new = *bs_old; | |
2038 | *bs_old = tmp; | |
a9fc4408 | 2039 | |
4ddc07ca PB |
2040 | /* there are some fields that should not be swapped, move them back */ |
2041 | bdrv_move_feature_fields(&tmp, bs_old); | |
2042 | bdrv_move_feature_fields(bs_old, bs_new); | |
2043 | bdrv_move_feature_fields(bs_new, &tmp); | |
8802d1fd | 2044 | |
4ddc07ca PB |
2045 | /* bs_new shouldn't be in bdrv_states even after the swap! */ |
2046 | assert(bs_new->device_name[0] == '\0'); | |
2047 | ||
2048 | /* Check a few fields that should remain attached to the device */ | |
2049 | assert(bs_new->dev == NULL); | |
2050 | assert(bs_new->job == NULL); | |
4ddc07ca | 2051 | assert(bs_new->io_limits_enabled == false); |
cc0681c4 | 2052 | assert(!throttle_have_timer(&bs_new->throttle_state)); |
e023b2e2 | 2053 | |
90ce8a06 BC |
2054 | /* insert the nodes back into the graph node list if needed */ |
2055 | if (bs_new->node_name[0] != '\0') { | |
2056 | QTAILQ_INSERT_TAIL(&graph_bdrv_states, bs_new, node_list); | |
2057 | } | |
2058 | if (bs_old->node_name[0] != '\0') { | |
2059 | QTAILQ_INSERT_TAIL(&graph_bdrv_states, bs_old, node_list); | |
2060 | } | |
2061 | ||
e023b2e2 | 2062 | bdrv_rebind(bs_new); |
4ddc07ca PB |
2063 | bdrv_rebind(bs_old); |
2064 | } | |
2065 | ||
2066 | /* | |
2067 | * Add new bs contents at the top of an image chain while the chain is | |
2068 | * live, while keeping required fields on the top layer. | |
2069 | * | |
2070 | * This will modify the BlockDriverState fields, and swap contents | |
2071 | * between bs_new and bs_top. Both bs_new and bs_top are modified. | |
2072 | * | |
2073 | * bs_new is required to be anonymous. | |
2074 | * | |
2075 | * This function does not create any image files. | |
2076 | */ | |
2077 | void bdrv_append(BlockDriverState *bs_new, BlockDriverState *bs_top) | |
2078 | { | |
2079 | bdrv_swap(bs_new, bs_top); | |
2080 | ||
2081 | /* The contents of 'tmp' will become bs_top, as we are | |
2082 | * swapping bs_new and bs_top contents. */ | |
8d24cce1 | 2083 | bdrv_set_backing_hd(bs_top, bs_new); |
8802d1fd JC |
2084 | } |
2085 | ||
4f6fd349 | 2086 | static void bdrv_delete(BlockDriverState *bs) |
b338082b | 2087 | { |
fa879d62 | 2088 | assert(!bs->dev); |
3e914655 | 2089 | assert(!bs->job); |
3718d8ab | 2090 | assert(bdrv_op_blocker_is_empty(bs)); |
4f6fd349 | 2091 | assert(!bs->refcnt); |
e4654d2d | 2092 | assert(QLIST_EMPTY(&bs->dirty_bitmaps)); |
18846dee | 2093 | |
e1b5c52e SH |
2094 | bdrv_close(bs); |
2095 | ||
1b7bdbc1 | 2096 | /* remove from list, if necessary */ |
d22b2f41 | 2097 | bdrv_make_anon(bs); |
34c6f050 | 2098 | |
7267c094 | 2099 | g_free(bs); |
fc01f7e7 FB |
2100 | } |
2101 | ||
fa879d62 MA |
2102 | int bdrv_attach_dev(BlockDriverState *bs, void *dev) |
2103 | /* TODO change to DeviceState *dev when all users are qdevified */ | |
18846dee | 2104 | { |
fa879d62 | 2105 | if (bs->dev) { |
18846dee MA |
2106 | return -EBUSY; |
2107 | } | |
fa879d62 | 2108 | bs->dev = dev; |
28a7282a | 2109 | bdrv_iostatus_reset(bs); |
18846dee MA |
2110 | return 0; |
2111 | } | |
2112 | ||
fa879d62 MA |
2113 | /* TODO qdevified devices don't use this, remove when devices are qdevified */ |
2114 | void bdrv_attach_dev_nofail(BlockDriverState *bs, void *dev) | |
18846dee | 2115 | { |
fa879d62 MA |
2116 | if (bdrv_attach_dev(bs, dev) < 0) { |
2117 | abort(); | |
2118 | } | |
2119 | } | |
2120 | ||
2121 | void bdrv_detach_dev(BlockDriverState *bs, void *dev) | |
2122 | /* TODO change to DeviceState *dev when all users are qdevified */ | |
2123 | { | |
2124 | assert(bs->dev == dev); | |
2125 | bs->dev = NULL; | |
0e49de52 MA |
2126 | bs->dev_ops = NULL; |
2127 | bs->dev_opaque = NULL; | |
1b7fd729 | 2128 | bs->guest_block_size = 512; |
18846dee MA |
2129 | } |
2130 | ||
fa879d62 MA |
2131 | /* TODO change to return DeviceState * when all users are qdevified */ |
2132 | void *bdrv_get_attached_dev(BlockDriverState *bs) | |
18846dee | 2133 | { |
fa879d62 | 2134 | return bs->dev; |
18846dee MA |
2135 | } |
2136 | ||
0e49de52 MA |
2137 | void bdrv_set_dev_ops(BlockDriverState *bs, const BlockDevOps *ops, |
2138 | void *opaque) | |
2139 | { | |
2140 | bs->dev_ops = ops; | |
2141 | bs->dev_opaque = opaque; | |
2142 | } | |
2143 | ||
7d4b4ba5 | 2144 | static void bdrv_dev_change_media_cb(BlockDriverState *bs, bool load) |
0e49de52 | 2145 | { |
145feb17 | 2146 | if (bs->dev_ops && bs->dev_ops->change_media_cb) { |
6f382ed2 | 2147 | bool tray_was_closed = !bdrv_dev_is_tray_open(bs); |
7d4b4ba5 | 2148 | bs->dev_ops->change_media_cb(bs->dev_opaque, load); |
6f382ed2 LC |
2149 | if (tray_was_closed) { |
2150 | /* tray open */ | |
a5ee7bd4 WX |
2151 | qapi_event_send_device_tray_moved(bdrv_get_device_name(bs), |
2152 | true, &error_abort); | |
6f382ed2 LC |
2153 | } |
2154 | if (load) { | |
2155 | /* tray close */ | |
a5ee7bd4 WX |
2156 | qapi_event_send_device_tray_moved(bdrv_get_device_name(bs), |
2157 | false, &error_abort); | |
6f382ed2 | 2158 | } |
145feb17 MA |
2159 | } |
2160 | } | |
2161 | ||
2c6942fa MA |
2162 | bool bdrv_dev_has_removable_media(BlockDriverState *bs) |
2163 | { | |
2164 | return !bs->dev || (bs->dev_ops && bs->dev_ops->change_media_cb); | |
2165 | } | |
2166 | ||
025ccaa7 PB |
2167 | void bdrv_dev_eject_request(BlockDriverState *bs, bool force) |
2168 | { | |
2169 | if (bs->dev_ops && bs->dev_ops->eject_request_cb) { | |
2170 | bs->dev_ops->eject_request_cb(bs->dev_opaque, force); | |
2171 | } | |
2172 | } | |
2173 | ||
e4def80b MA |
2174 | bool bdrv_dev_is_tray_open(BlockDriverState *bs) |
2175 | { | |
2176 | if (bs->dev_ops && bs->dev_ops->is_tray_open) { | |
2177 | return bs->dev_ops->is_tray_open(bs->dev_opaque); | |
2178 | } | |
2179 | return false; | |
2180 | } | |
2181 | ||
145feb17 MA |
2182 | static void bdrv_dev_resize_cb(BlockDriverState *bs) |
2183 | { | |
2184 | if (bs->dev_ops && bs->dev_ops->resize_cb) { | |
2185 | bs->dev_ops->resize_cb(bs->dev_opaque); | |
0e49de52 MA |
2186 | } |
2187 | } | |
2188 | ||
f107639a MA |
2189 | bool bdrv_dev_is_medium_locked(BlockDriverState *bs) |
2190 | { | |
2191 | if (bs->dev_ops && bs->dev_ops->is_medium_locked) { | |
2192 | return bs->dev_ops->is_medium_locked(bs->dev_opaque); | |
2193 | } | |
2194 | return false; | |
2195 | } | |
2196 | ||
e97fc193 AL |
2197 | /* |
2198 | * Run consistency checks on an image | |
2199 | * | |
e076f338 | 2200 | * Returns 0 if the check could be completed (it doesn't mean that the image is |
a1c7273b | 2201 | * free of errors) or -errno when an internal error occurred. The results of the |
e076f338 | 2202 | * check are stored in res. |
e97fc193 | 2203 | */ |
4534ff54 | 2204 | int bdrv_check(BlockDriverState *bs, BdrvCheckResult *res, BdrvCheckMode fix) |
e97fc193 AL |
2205 | { |
2206 | if (bs->drv->bdrv_check == NULL) { | |
2207 | return -ENOTSUP; | |
2208 | } | |
2209 | ||
e076f338 | 2210 | memset(res, 0, sizeof(*res)); |
4534ff54 | 2211 | return bs->drv->bdrv_check(bs, res, fix); |
e97fc193 AL |
2212 | } |
2213 | ||
8a426614 KW |
2214 | #define COMMIT_BUF_SECTORS 2048 |
2215 | ||
33e3963e FB |
2216 | /* commit COW file into the raw image */ |
2217 | int bdrv_commit(BlockDriverState *bs) | |
2218 | { | |
19cb3738 | 2219 | BlockDriver *drv = bs->drv; |
72706ea4 | 2220 | int64_t sector, total_sectors, length, backing_length; |
8a426614 | 2221 | int n, ro, open_flags; |
0bce597d | 2222 | int ret = 0; |
72706ea4 | 2223 | uint8_t *buf = NULL; |
c2cba3d9 | 2224 | char filename[PATH_MAX]; |
33e3963e | 2225 | |
19cb3738 FB |
2226 | if (!drv) |
2227 | return -ENOMEDIUM; | |
4dca4b63 NS |
2228 | |
2229 | if (!bs->backing_hd) { | |
2230 | return -ENOTSUP; | |
33e3963e FB |
2231 | } |
2232 | ||
3718d8ab FZ |
2233 | if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_COMMIT, NULL) || |
2234 | bdrv_op_is_blocked(bs->backing_hd, BLOCK_OP_TYPE_COMMIT, NULL)) { | |
2d3735d3 SH |
2235 | return -EBUSY; |
2236 | } | |
2237 | ||
4dca4b63 | 2238 | ro = bs->backing_hd->read_only; |
c2cba3d9 JM |
2239 | /* Use pstrcpy (not strncpy): filename must be NUL-terminated. */ |
2240 | pstrcpy(filename, sizeof(filename), bs->backing_hd->filename); | |
4dca4b63 NS |
2241 | open_flags = bs->backing_hd->open_flags; |
2242 | ||
2243 | if (ro) { | |
0bce597d JC |
2244 | if (bdrv_reopen(bs->backing_hd, open_flags | BDRV_O_RDWR, NULL)) { |
2245 | return -EACCES; | |
4dca4b63 | 2246 | } |
ea2384d3 | 2247 | } |
33e3963e | 2248 | |
72706ea4 JC |
2249 | length = bdrv_getlength(bs); |
2250 | if (length < 0) { | |
2251 | ret = length; | |
2252 | goto ro_cleanup; | |
2253 | } | |
2254 | ||
2255 | backing_length = bdrv_getlength(bs->backing_hd); | |
2256 | if (backing_length < 0) { | |
2257 | ret = backing_length; | |
2258 | goto ro_cleanup; | |
2259 | } | |
2260 | ||
2261 | /* If our top snapshot is larger than the backing file image, | |
2262 | * grow the backing file image if possible. If not possible, | |
2263 | * we must return an error */ | |
2264 | if (length > backing_length) { | |
2265 | ret = bdrv_truncate(bs->backing_hd, length); | |
2266 | if (ret < 0) { | |
2267 | goto ro_cleanup; | |
2268 | } | |
2269 | } | |
2270 | ||
2271 | total_sectors = length >> BDRV_SECTOR_BITS; | |
7267c094 | 2272 | buf = g_malloc(COMMIT_BUF_SECTORS * BDRV_SECTOR_SIZE); |
8a426614 KW |
2273 | |
2274 | for (sector = 0; sector < total_sectors; sector += n) { | |
d663640c PB |
2275 | ret = bdrv_is_allocated(bs, sector, COMMIT_BUF_SECTORS, &n); |
2276 | if (ret < 0) { | |
2277 | goto ro_cleanup; | |
2278 | } | |
2279 | if (ret) { | |
dabfa6cc KW |
2280 | ret = bdrv_read(bs, sector, buf, n); |
2281 | if (ret < 0) { | |
8a426614 KW |
2282 | goto ro_cleanup; |
2283 | } | |
2284 | ||
dabfa6cc KW |
2285 | ret = bdrv_write(bs->backing_hd, sector, buf, n); |
2286 | if (ret < 0) { | |
8a426614 KW |
2287 | goto ro_cleanup; |
2288 | } | |
ea2384d3 | 2289 | } |
33e3963e | 2290 | } |
95389c86 | 2291 | |
1d44952f CH |
2292 | if (drv->bdrv_make_empty) { |
2293 | ret = drv->bdrv_make_empty(bs); | |
dabfa6cc KW |
2294 | if (ret < 0) { |
2295 | goto ro_cleanup; | |
2296 | } | |
1d44952f CH |
2297 | bdrv_flush(bs); |
2298 | } | |
95389c86 | 2299 | |
3f5075ae CH |
2300 | /* |
2301 | * Make sure all data we wrote to the backing device is actually | |
2302 | * stable on disk. | |
2303 | */ | |
dabfa6cc | 2304 | if (bs->backing_hd) { |
3f5075ae | 2305 | bdrv_flush(bs->backing_hd); |
dabfa6cc | 2306 | } |
4dca4b63 | 2307 | |
dabfa6cc | 2308 | ret = 0; |
4dca4b63 | 2309 | ro_cleanup: |
7267c094 | 2310 | g_free(buf); |
4dca4b63 NS |
2311 | |
2312 | if (ro) { | |
0bce597d JC |
2313 | /* ignoring error return here */ |
2314 | bdrv_reopen(bs->backing_hd, open_flags & ~BDRV_O_RDWR, NULL); | |
4dca4b63 NS |
2315 | } |
2316 | ||
1d44952f | 2317 | return ret; |
33e3963e FB |
2318 | } |
2319 | ||
e8877497 | 2320 | int bdrv_commit_all(void) |
6ab4b5ab MA |
2321 | { |
2322 | BlockDriverState *bs; | |
2323 | ||
dc364f4c | 2324 | QTAILQ_FOREACH(bs, &bdrv_states, device_list) { |
ed78cda3 SH |
2325 | AioContext *aio_context = bdrv_get_aio_context(bs); |
2326 | ||
2327 | aio_context_acquire(aio_context); | |
272d2d8e JC |
2328 | if (bs->drv && bs->backing_hd) { |
2329 | int ret = bdrv_commit(bs); | |
2330 | if (ret < 0) { | |
ed78cda3 | 2331 | aio_context_release(aio_context); |
272d2d8e JC |
2332 | return ret; |
2333 | } | |
e8877497 | 2334 | } |
ed78cda3 | 2335 | aio_context_release(aio_context); |
6ab4b5ab | 2336 | } |
e8877497 | 2337 | return 0; |
6ab4b5ab MA |
2338 | } |
2339 | ||
dbffbdcf SH |
2340 | /** |
2341 | * Remove an active request from the tracked requests list | |
2342 | * | |
2343 | * This function should be called when a tracked request is completing. | |
2344 | */ | |
2345 | static void tracked_request_end(BdrvTrackedRequest *req) | |
2346 | { | |
2dbafdc0 KW |
2347 | if (req->serialising) { |
2348 | req->bs->serialising_in_flight--; | |
2349 | } | |
2350 | ||
dbffbdcf | 2351 | QLIST_REMOVE(req, list); |
f4658285 | 2352 | qemu_co_queue_restart_all(&req->wait_queue); |
dbffbdcf SH |
2353 | } |
2354 | ||
2355 | /** | |
2356 | * Add an active request to the tracked requests list | |
2357 | */ | |
2358 | static void tracked_request_begin(BdrvTrackedRequest *req, | |
2359 | BlockDriverState *bs, | |
793ed47a KW |
2360 | int64_t offset, |
2361 | unsigned int bytes, bool is_write) | |
dbffbdcf SH |
2362 | { |
2363 | *req = (BdrvTrackedRequest){ | |
2364 | .bs = bs, | |
2dbafdc0 KW |
2365 | .offset = offset, |
2366 | .bytes = bytes, | |
2367 | .is_write = is_write, | |
2368 | .co = qemu_coroutine_self(), | |
2369 | .serialising = false, | |
7327145f KW |
2370 | .overlap_offset = offset, |
2371 | .overlap_bytes = bytes, | |
dbffbdcf SH |
2372 | }; |
2373 | ||
f4658285 SH |
2374 | qemu_co_queue_init(&req->wait_queue); |
2375 | ||
dbffbdcf SH |
2376 | QLIST_INSERT_HEAD(&bs->tracked_requests, req, list); |
2377 | } | |
2378 | ||
e96126ff | 2379 | static void mark_request_serialising(BdrvTrackedRequest *req, uint64_t align) |
2dbafdc0 | 2380 | { |
7327145f | 2381 | int64_t overlap_offset = req->offset & ~(align - 1); |
e96126ff KW |
2382 | unsigned int overlap_bytes = ROUND_UP(req->offset + req->bytes, align) |
2383 | - overlap_offset; | |
7327145f | 2384 | |
2dbafdc0 KW |
2385 | if (!req->serialising) { |
2386 | req->bs->serialising_in_flight++; | |
2387 | req->serialising = true; | |
2388 | } | |
7327145f KW |
2389 | |
2390 | req->overlap_offset = MIN(req->overlap_offset, overlap_offset); | |
2391 | req->overlap_bytes = MAX(req->overlap_bytes, overlap_bytes); | |
2dbafdc0 KW |
2392 | } |
2393 | ||
d83947ac SH |
2394 | /** |
2395 | * Round a region to cluster boundaries | |
2396 | */ | |
343bded4 PB |
2397 | void bdrv_round_to_clusters(BlockDriverState *bs, |
2398 | int64_t sector_num, int nb_sectors, | |
2399 | int64_t *cluster_sector_num, | |
2400 | int *cluster_nb_sectors) | |
d83947ac SH |
2401 | { |
2402 | BlockDriverInfo bdi; | |
2403 | ||
2404 | if (bdrv_get_info(bs, &bdi) < 0 || bdi.cluster_size == 0) { | |
2405 | *cluster_sector_num = sector_num; | |
2406 | *cluster_nb_sectors = nb_sectors; | |
2407 | } else { | |
2408 | int64_t c = bdi.cluster_size / BDRV_SECTOR_SIZE; | |
2409 | *cluster_sector_num = QEMU_ALIGN_DOWN(sector_num, c); | |
2410 | *cluster_nb_sectors = QEMU_ALIGN_UP(sector_num - *cluster_sector_num + | |
2411 | nb_sectors, c); | |
2412 | } | |
2413 | } | |
2414 | ||
7327145f | 2415 | static int bdrv_get_cluster_size(BlockDriverState *bs) |
793ed47a KW |
2416 | { |
2417 | BlockDriverInfo bdi; | |
7327145f | 2418 | int ret; |
793ed47a | 2419 | |
7327145f KW |
2420 | ret = bdrv_get_info(bs, &bdi); |
2421 | if (ret < 0 || bdi.cluster_size == 0) { | |
2422 | return bs->request_alignment; | |
793ed47a | 2423 | } else { |
7327145f | 2424 | return bdi.cluster_size; |
793ed47a KW |
2425 | } |
2426 | } | |
2427 | ||
f4658285 | 2428 | static bool tracked_request_overlaps(BdrvTrackedRequest *req, |
793ed47a KW |
2429 | int64_t offset, unsigned int bytes) |
2430 | { | |
d83947ac | 2431 | /* aaaa bbbb */ |
7327145f | 2432 | if (offset >= req->overlap_offset + req->overlap_bytes) { |
d83947ac SH |
2433 | return false; |
2434 | } | |
2435 | /* bbbb aaaa */ | |
7327145f | 2436 | if (req->overlap_offset >= offset + bytes) { |
d83947ac SH |
2437 | return false; |
2438 | } | |
2439 | return true; | |
f4658285 SH |
2440 | } |
2441 | ||
28de2dcd | 2442 | static bool coroutine_fn wait_serialising_requests(BdrvTrackedRequest *self) |
f4658285 | 2443 | { |
2dbafdc0 | 2444 | BlockDriverState *bs = self->bs; |
f4658285 SH |
2445 | BdrvTrackedRequest *req; |
2446 | bool retry; | |
28de2dcd | 2447 | bool waited = false; |
f4658285 | 2448 | |
2dbafdc0 | 2449 | if (!bs->serialising_in_flight) { |
28de2dcd | 2450 | return false; |
2dbafdc0 KW |
2451 | } |
2452 | ||
f4658285 SH |
2453 | do { |
2454 | retry = false; | |
2455 | QLIST_FOREACH(req, &bs->tracked_requests, list) { | |
2dbafdc0 | 2456 | if (req == self || (!req->serialising && !self->serialising)) { |
65afd211 KW |
2457 | continue; |
2458 | } | |
7327145f KW |
2459 | if (tracked_request_overlaps(req, self->overlap_offset, |
2460 | self->overlap_bytes)) | |
2461 | { | |
5f8b6491 SH |
2462 | /* Hitting this means there was a reentrant request, for |
2463 | * example, a block driver issuing nested requests. This must | |
2464 | * never happen since it means deadlock. | |
2465 | */ | |
2466 | assert(qemu_coroutine_self() != req->co); | |
2467 | ||
6460440f KW |
2468 | /* If the request is already (indirectly) waiting for us, or |
2469 | * will wait for us as soon as it wakes up, then just go on | |
2470 | * (instead of producing a deadlock in the former case). */ | |
2471 | if (!req->waiting_for) { | |
2472 | self->waiting_for = req; | |
2473 | qemu_co_queue_wait(&req->wait_queue); | |
2474 | self->waiting_for = NULL; | |
2475 | retry = true; | |
28de2dcd | 2476 | waited = true; |
6460440f KW |
2477 | break; |
2478 | } | |
f4658285 SH |
2479 | } |
2480 | } | |
2481 | } while (retry); | |
28de2dcd KW |
2482 | |
2483 | return waited; | |
f4658285 SH |
2484 | } |
2485 | ||
756e6736 KW |
2486 | /* |
2487 | * Return values: | |
2488 | * 0 - success | |
2489 | * -EINVAL - backing format specified, but no file | |
2490 | * -ENOSPC - can't update the backing file because no space is left in the | |
2491 | * image file header | |
2492 | * -ENOTSUP - format driver doesn't support changing the backing file | |
2493 | */ | |
2494 | int bdrv_change_backing_file(BlockDriverState *bs, | |
2495 | const char *backing_file, const char *backing_fmt) | |
2496 | { | |
2497 | BlockDriver *drv = bs->drv; | |
469ef350 | 2498 | int ret; |
756e6736 | 2499 | |
5f377794 PB |
2500 | /* Backing file format doesn't make sense without a backing file */ |
2501 | if (backing_fmt && !backing_file) { | |
2502 | return -EINVAL; | |
2503 | } | |
2504 | ||
756e6736 | 2505 | if (drv->bdrv_change_backing_file != NULL) { |
469ef350 | 2506 | ret = drv->bdrv_change_backing_file(bs, backing_file, backing_fmt); |
756e6736 | 2507 | } else { |
469ef350 | 2508 | ret = -ENOTSUP; |
756e6736 | 2509 | } |
469ef350 PB |
2510 | |
2511 | if (ret == 0) { | |
2512 | pstrcpy(bs->backing_file, sizeof(bs->backing_file), backing_file ?: ""); | |
2513 | pstrcpy(bs->backing_format, sizeof(bs->backing_format), backing_fmt ?: ""); | |
2514 | } | |
2515 | return ret; | |
756e6736 KW |
2516 | } |
2517 | ||
6ebdcee2 JC |
2518 | /* |
2519 | * Finds the image layer in the chain that has 'bs' as its backing file. | |
2520 | * | |
2521 | * active is the current topmost image. | |
2522 | * | |
2523 | * Returns NULL if bs is not found in active's image chain, | |
2524 | * or if active == bs. | |
4caf0fcd JC |
2525 | * |
2526 | * Returns the bottommost base image if bs == NULL. | |
6ebdcee2 JC |
2527 | */ |
2528 | BlockDriverState *bdrv_find_overlay(BlockDriverState *active, | |
2529 | BlockDriverState *bs) | |
2530 | { | |
4caf0fcd JC |
2531 | while (active && bs != active->backing_hd) { |
2532 | active = active->backing_hd; | |
6ebdcee2 JC |
2533 | } |
2534 | ||
4caf0fcd JC |
2535 | return active; |
2536 | } | |
6ebdcee2 | 2537 | |
4caf0fcd JC |
2538 | /* Given a BDS, searches for the base layer. */ |
2539 | BlockDriverState *bdrv_find_base(BlockDriverState *bs) | |
2540 | { | |
2541 | return bdrv_find_overlay(bs, NULL); | |
6ebdcee2 JC |
2542 | } |
2543 | ||
2544 | typedef struct BlkIntermediateStates { | |
2545 | BlockDriverState *bs; | |
2546 | QSIMPLEQ_ENTRY(BlkIntermediateStates) entry; | |
2547 | } BlkIntermediateStates; | |
2548 | ||
2549 | ||
2550 | /* | |
2551 | * Drops images above 'base' up to and including 'top', and sets the image | |
2552 | * above 'top' to have base as its backing file. | |
2553 | * | |
2554 | * Requires that the overlay to 'top' is opened r/w, so that the backing file | |
2555 | * information in 'bs' can be properly updated. | |
2556 | * | |
2557 | * E.g., this will convert the following chain: | |
2558 | * bottom <- base <- intermediate <- top <- active | |
2559 | * | |
2560 | * to | |
2561 | * | |
2562 | * bottom <- base <- active | |
2563 | * | |
2564 | * It is allowed for bottom==base, in which case it converts: | |
2565 | * | |
2566 | * base <- intermediate <- top <- active | |
2567 | * | |
2568 | * to | |
2569 | * | |
2570 | * base <- active | |
2571 | * | |
54e26900 JC |
2572 | * If backing_file_str is non-NULL, it will be used when modifying top's |
2573 | * overlay image metadata. | |
2574 | * | |
6ebdcee2 JC |
2575 | * Error conditions: |
2576 | * if active == top, that is considered an error | |
2577 | * | |
2578 | */ | |
2579 | int bdrv_drop_intermediate(BlockDriverState *active, BlockDriverState *top, | |
54e26900 | 2580 | BlockDriverState *base, const char *backing_file_str) |
6ebdcee2 JC |
2581 | { |
2582 | BlockDriverState *intermediate; | |
2583 | BlockDriverState *base_bs = NULL; | |
2584 | BlockDriverState *new_top_bs = NULL; | |
2585 | BlkIntermediateStates *intermediate_state, *next; | |
2586 | int ret = -EIO; | |
2587 | ||
2588 | QSIMPLEQ_HEAD(states_to_delete, BlkIntermediateStates) states_to_delete; | |
2589 | QSIMPLEQ_INIT(&states_to_delete); | |
2590 | ||
2591 | if (!top->drv || !base->drv) { | |
2592 | goto exit; | |
2593 | } | |
2594 | ||
2595 | new_top_bs = bdrv_find_overlay(active, top); | |
2596 | ||
2597 | if (new_top_bs == NULL) { | |
2598 | /* we could not find the image above 'top', this is an error */ | |
2599 | goto exit; | |
2600 | } | |
2601 | ||
2602 | /* special case of new_top_bs->backing_hd already pointing to base - nothing | |
2603 | * to do, no intermediate images */ | |
2604 | if (new_top_bs->backing_hd == base) { | |
2605 | ret = 0; | |
2606 | goto exit; | |
2607 | } | |
2608 | ||
2609 | intermediate = top; | |
2610 | ||
2611 | /* now we will go down through the list, and add each BDS we find | |
2612 | * into our deletion queue, until we hit the 'base' | |
2613 | */ | |
2614 | while (intermediate) { | |
2615 | intermediate_state = g_malloc0(sizeof(BlkIntermediateStates)); | |
2616 | intermediate_state->bs = intermediate; | |
2617 | QSIMPLEQ_INSERT_TAIL(&states_to_delete, intermediate_state, entry); | |
2618 | ||
2619 | if (intermediate->backing_hd == base) { | |
2620 | base_bs = intermediate->backing_hd; | |
2621 | break; | |
2622 | } | |
2623 | intermediate = intermediate->backing_hd; | |
2624 | } | |
2625 | if (base_bs == NULL) { | |
2626 | /* something went wrong, we did not end at the base. safely | |
2627 | * unravel everything, and exit with error */ | |
2628 | goto exit; | |
2629 | } | |
2630 | ||
2631 | /* success - we can delete the intermediate states, and link top->base */ | |
54e26900 JC |
2632 | backing_file_str = backing_file_str ? backing_file_str : base_bs->filename; |
2633 | ret = bdrv_change_backing_file(new_top_bs, backing_file_str, | |
6ebdcee2 JC |
2634 | base_bs->drv ? base_bs->drv->format_name : ""); |
2635 | if (ret) { | |
2636 | goto exit; | |
2637 | } | |
920beae1 | 2638 | bdrv_set_backing_hd(new_top_bs, base_bs); |
6ebdcee2 JC |
2639 | |
2640 | QSIMPLEQ_FOREACH_SAFE(intermediate_state, &states_to_delete, entry, next) { | |
2641 | /* so that bdrv_close() does not recursively close the chain */ | |
920beae1 | 2642 | bdrv_set_backing_hd(intermediate_state->bs, NULL); |
4f6fd349 | 2643 | bdrv_unref(intermediate_state->bs); |
6ebdcee2 JC |
2644 | } |
2645 | ret = 0; | |
2646 | ||
2647 | exit: | |
2648 | QSIMPLEQ_FOREACH_SAFE(intermediate_state, &states_to_delete, entry, next) { | |
2649 | g_free(intermediate_state); | |
2650 | } | |
2651 | return ret; | |
2652 | } | |
2653 | ||
2654 | ||
71d0770c AL |
2655 | static int bdrv_check_byte_request(BlockDriverState *bs, int64_t offset, |
2656 | size_t size) | |
2657 | { | |
2658 | int64_t len; | |
2659 | ||
1dd3a447 KW |
2660 | if (size > INT_MAX) { |
2661 | return -EIO; | |
2662 | } | |
2663 | ||
71d0770c AL |
2664 | if (!bdrv_is_inserted(bs)) |
2665 | return -ENOMEDIUM; | |
2666 | ||
2667 | if (bs->growable) | |
2668 | return 0; | |
2669 | ||
2670 | len = bdrv_getlength(bs); | |
2671 | ||
fbb7b4e0 KW |
2672 | if (offset < 0) |
2673 | return -EIO; | |
2674 | ||
2675 | if ((offset > len) || (len - offset < size)) | |
71d0770c AL |
2676 | return -EIO; |
2677 | ||
2678 | return 0; | |
2679 | } | |
2680 | ||
2681 | static int bdrv_check_request(BlockDriverState *bs, int64_t sector_num, | |
2682 | int nb_sectors) | |
2683 | { | |
54db38a4 | 2684 | if (nb_sectors < 0 || nb_sectors > INT_MAX / BDRV_SECTOR_SIZE) { |
8f4754ed KW |
2685 | return -EIO; |
2686 | } | |
2687 | ||
eb5a3165 JS |
2688 | return bdrv_check_byte_request(bs, sector_num * BDRV_SECTOR_SIZE, |
2689 | nb_sectors * BDRV_SECTOR_SIZE); | |
71d0770c AL |
2690 | } |
2691 | ||
1c9805a3 SH |
2692 | typedef struct RwCo { |
2693 | BlockDriverState *bs; | |
775aa8b6 | 2694 | int64_t offset; |
1c9805a3 SH |
2695 | QEMUIOVector *qiov; |
2696 | bool is_write; | |
2697 | int ret; | |
4105eaaa | 2698 | BdrvRequestFlags flags; |
1c9805a3 SH |
2699 | } RwCo; |
2700 | ||
2701 | static void coroutine_fn bdrv_rw_co_entry(void *opaque) | |
fc01f7e7 | 2702 | { |
1c9805a3 | 2703 | RwCo *rwco = opaque; |
ea2384d3 | 2704 | |
1c9805a3 | 2705 | if (!rwco->is_write) { |
775aa8b6 KW |
2706 | rwco->ret = bdrv_co_do_preadv(rwco->bs, rwco->offset, |
2707 | rwco->qiov->size, rwco->qiov, | |
4105eaaa | 2708 | rwco->flags); |
775aa8b6 KW |
2709 | } else { |
2710 | rwco->ret = bdrv_co_do_pwritev(rwco->bs, rwco->offset, | |
2711 | rwco->qiov->size, rwco->qiov, | |
2712 | rwco->flags); | |
1c9805a3 SH |
2713 | } |
2714 | } | |
e7a8a783 | 2715 | |
1c9805a3 | 2716 | /* |
8d3b1a2d | 2717 | * Process a vectored synchronous request using coroutines |
1c9805a3 | 2718 | */ |
775aa8b6 KW |
2719 | static int bdrv_prwv_co(BlockDriverState *bs, int64_t offset, |
2720 | QEMUIOVector *qiov, bool is_write, | |
2721 | BdrvRequestFlags flags) | |
1c9805a3 | 2722 | { |
1c9805a3 SH |
2723 | Coroutine *co; |
2724 | RwCo rwco = { | |
2725 | .bs = bs, | |
775aa8b6 | 2726 | .offset = offset, |
8d3b1a2d | 2727 | .qiov = qiov, |
1c9805a3 SH |
2728 | .is_write = is_write, |
2729 | .ret = NOT_DONE, | |
4105eaaa | 2730 | .flags = flags, |
1c9805a3 | 2731 | }; |
e7a8a783 | 2732 | |
498e386c ZYW |
2733 | /** |
2734 | * In sync call context, when the vcpu is blocked, this throttling timer | |
2735 | * will not fire; so the I/O throttling function has to be disabled here | |
2736 | * if it has been enabled. | |
2737 | */ | |
2738 | if (bs->io_limits_enabled) { | |
2739 | fprintf(stderr, "Disabling I/O throttling on '%s' due " | |
2740 | "to synchronous I/O.\n", bdrv_get_device_name(bs)); | |
2741 | bdrv_io_limits_disable(bs); | |
2742 | } | |
2743 | ||
1c9805a3 SH |
2744 | if (qemu_in_coroutine()) { |
2745 | /* Fast-path if already in coroutine context */ | |
2746 | bdrv_rw_co_entry(&rwco); | |
2747 | } else { | |
2572b37a SH |
2748 | AioContext *aio_context = bdrv_get_aio_context(bs); |
2749 | ||
1c9805a3 SH |
2750 | co = qemu_coroutine_create(bdrv_rw_co_entry); |
2751 | qemu_coroutine_enter(co, &rwco); | |
2752 | while (rwco.ret == NOT_DONE) { | |
2572b37a | 2753 | aio_poll(aio_context, true); |
1c9805a3 SH |
2754 | } |
2755 | } | |
2756 | return rwco.ret; | |
2757 | } | |
b338082b | 2758 | |
8d3b1a2d KW |
2759 | /* |
2760 | * Process a synchronous request using coroutines | |
2761 | */ | |
2762 | static int bdrv_rw_co(BlockDriverState *bs, int64_t sector_num, uint8_t *buf, | |
4105eaaa | 2763 | int nb_sectors, bool is_write, BdrvRequestFlags flags) |
8d3b1a2d KW |
2764 | { |
2765 | QEMUIOVector qiov; | |
2766 | struct iovec iov = { | |
2767 | .iov_base = (void *)buf, | |
2768 | .iov_len = nb_sectors * BDRV_SECTOR_SIZE, | |
2769 | }; | |
2770 | ||
da15ee51 KW |
2771 | if (nb_sectors < 0 || nb_sectors > INT_MAX / BDRV_SECTOR_SIZE) { |
2772 | return -EINVAL; | |
2773 | } | |
2774 | ||
8d3b1a2d | 2775 | qemu_iovec_init_external(&qiov, &iov, 1); |
775aa8b6 KW |
2776 | return bdrv_prwv_co(bs, sector_num << BDRV_SECTOR_BITS, |
2777 | &qiov, is_write, flags); | |
8d3b1a2d KW |
2778 | } |
2779 | ||
1c9805a3 SH |
2780 | /* return < 0 if error. See bdrv_write() for the return codes */ |
2781 | int bdrv_read(BlockDriverState *bs, int64_t sector_num, | |
2782 | uint8_t *buf, int nb_sectors) | |
2783 | { | |
4105eaaa | 2784 | return bdrv_rw_co(bs, sector_num, buf, nb_sectors, false, 0); |
fc01f7e7 FB |
2785 | } |
2786 | ||
07d27a44 MA |
2787 | /* Just like bdrv_read(), but with I/O throttling temporarily disabled */ |
2788 | int bdrv_read_unthrottled(BlockDriverState *bs, int64_t sector_num, | |
2789 | uint8_t *buf, int nb_sectors) | |
2790 | { | |
2791 | bool enabled; | |
2792 | int ret; | |
2793 | ||
2794 | enabled = bs->io_limits_enabled; | |
2795 | bs->io_limits_enabled = false; | |
4e7395e8 | 2796 | ret = bdrv_read(bs, sector_num, buf, nb_sectors); |
07d27a44 MA |
2797 | bs->io_limits_enabled = enabled; |
2798 | return ret; | |
2799 | } | |
2800 | ||
5fafdf24 | 2801 | /* Return < 0 if error. Important errors are: |
19cb3738 FB |
2802 | -EIO generic I/O error (may happen for all errors) |
2803 | -ENOMEDIUM No media inserted. | |
2804 | -EINVAL Invalid sector number or nb_sectors | |
2805 | -EACCES Trying to write a read-only device | |
2806 | */ | |
5fafdf24 | 2807 | int bdrv_write(BlockDriverState *bs, int64_t sector_num, |
fc01f7e7 FB |
2808 | const uint8_t *buf, int nb_sectors) |
2809 | { | |
4105eaaa | 2810 | return bdrv_rw_co(bs, sector_num, (uint8_t *)buf, nb_sectors, true, 0); |
83f64091 FB |
2811 | } |
2812 | ||
aa7bfbff PL |
2813 | int bdrv_write_zeroes(BlockDriverState *bs, int64_t sector_num, |
2814 | int nb_sectors, BdrvRequestFlags flags) | |
4105eaaa PL |
2815 | { |
2816 | return bdrv_rw_co(bs, sector_num, NULL, nb_sectors, true, | |
aa7bfbff | 2817 | BDRV_REQ_ZERO_WRITE | flags); |
8d3b1a2d KW |
2818 | } |
2819 | ||
d75cbb5e PL |
2820 | /* |
2821 | * Completely zero out a block device with the help of bdrv_write_zeroes. | |
2822 | * The operation is sped up by checking the block status and only writing | |
2823 | * zeroes to the device if they currently do not return zeroes. Optional | |
2824 | * flags are passed through to bdrv_write_zeroes (e.g. BDRV_REQ_MAY_UNMAP). | |
2825 | * | |
2826 | * Returns < 0 on error, 0 on success. For error codes see bdrv_write(). | |
2827 | */ | |
2828 | int bdrv_make_zero(BlockDriverState *bs, BdrvRequestFlags flags) | |
2829 | { | |
9ce10c0b | 2830 | int64_t target_size; |
d75cbb5e PL |
2831 | int64_t ret, nb_sectors, sector_num = 0; |
2832 | int n; | |
2833 | ||
9ce10c0b KW |
2834 | target_size = bdrv_getlength(bs); |
2835 | if (target_size < 0) { | |
2836 | return target_size; | |
2837 | } | |
2838 | target_size /= BDRV_SECTOR_SIZE; | |
2839 | ||
d75cbb5e PL |
2840 | for (;;) { |
2841 | nb_sectors = target_size - sector_num; | |
2842 | if (nb_sectors <= 0) { | |
2843 | return 0; | |
2844 | } | |
2845 | if (nb_sectors > INT_MAX) { | |
2846 | nb_sectors = INT_MAX; | |
2847 | } | |
2848 | ret = bdrv_get_block_status(bs, sector_num, nb_sectors, &n); | |
3d94ce60 PL |
2849 | if (ret < 0) { |
2850 | error_report("error getting block status at sector %" PRId64 ": %s", | |
2851 | sector_num, strerror(-ret)); | |
2852 | return ret; | |
2853 | } | |
d75cbb5e PL |
2854 | if (ret & BDRV_BLOCK_ZERO) { |
2855 | sector_num += n; | |
2856 | continue; | |
2857 | } | |
2858 | ret = bdrv_write_zeroes(bs, sector_num, n, flags); | |
2859 | if (ret < 0) { | |
2860 | error_report("error writing zeroes at sector %" PRId64 ": %s", | |
2861 | sector_num, strerror(-ret)); | |
2862 | return ret; | |
2863 | } | |
2864 | sector_num += n; | |
2865 | } | |
2866 | } | |
2867 | ||
a3ef6571 | 2868 | int bdrv_pread(BlockDriverState *bs, int64_t offset, void *buf, int bytes) |
83f64091 | 2869 | { |
a3ef6571 KW |
2870 | QEMUIOVector qiov; |
2871 | struct iovec iov = { | |
2872 | .iov_base = (void *)buf, | |
2873 | .iov_len = bytes, | |
2874 | }; | |
9a8c4cce | 2875 | int ret; |
83f64091 | 2876 | |
a3ef6571 KW |
2877 | if (bytes < 0) { |
2878 | return -EINVAL; | |
83f64091 FB |
2879 | } |
2880 | ||
a3ef6571 KW |
2881 | qemu_iovec_init_external(&qiov, &iov, 1); |
2882 | ret = bdrv_prwv_co(bs, offset, &qiov, false, 0); | |
2883 | if (ret < 0) { | |
2884 | return ret; | |
83f64091 | 2885 | } |
a3ef6571 KW |
2886 | |
2887 | return bytes; | |
83f64091 FB |
2888 | } |
2889 | ||
8d3b1a2d | 2890 | int bdrv_pwritev(BlockDriverState *bs, int64_t offset, QEMUIOVector *qiov) |
83f64091 | 2891 | { |
9a8c4cce | 2892 | int ret; |
83f64091 | 2893 | |
8407d5d7 KW |
2894 | ret = bdrv_prwv_co(bs, offset, qiov, true, 0); |
2895 | if (ret < 0) { | |
2896 | return ret; | |
83f64091 FB |
2897 | } |
2898 | ||
8d3b1a2d KW |
2899 | return qiov->size; |
2900 | } | |
2901 | ||
2902 | int bdrv_pwrite(BlockDriverState *bs, int64_t offset, | |
8407d5d7 | 2903 | const void *buf, int bytes) |
8d3b1a2d KW |
2904 | { |
2905 | QEMUIOVector qiov; | |
2906 | struct iovec iov = { | |
2907 | .iov_base = (void *) buf, | |
8407d5d7 | 2908 | .iov_len = bytes, |
8d3b1a2d KW |
2909 | }; |
2910 | ||
8407d5d7 KW |
2911 | if (bytes < 0) { |
2912 | return -EINVAL; | |
2913 | } | |
2914 | ||
8d3b1a2d KW |
2915 | qemu_iovec_init_external(&qiov, &iov, 1); |
2916 | return bdrv_pwritev(bs, offset, &qiov); | |
83f64091 | 2917 | } |
83f64091 | 2918 | |
f08145fe KW |
2919 | /* |
2920 | * Writes to the file and ensures that no writes are reordered across this | |
2921 | * request (acts as a barrier) | |
2922 | * | |
2923 | * Returns 0 on success, -errno in error cases. | |
2924 | */ | |
2925 | int bdrv_pwrite_sync(BlockDriverState *bs, int64_t offset, | |
2926 | const void *buf, int count) | |
2927 | { | |
2928 | int ret; | |
2929 | ||
2930 | ret = bdrv_pwrite(bs, offset, buf, count); | |
2931 | if (ret < 0) { | |
2932 | return ret; | |
2933 | } | |
2934 | ||
f05fa4ad PB |
2935 | /* No flush needed for cache modes that already do it */ |
2936 | if (bs->enable_write_cache) { | |
f08145fe KW |
2937 | bdrv_flush(bs); |
2938 | } | |
2939 | ||
2940 | return 0; | |
2941 | } | |
2942 | ||
470c0504 | 2943 | static int coroutine_fn bdrv_co_do_copy_on_readv(BlockDriverState *bs, |
ab185921 SH |
2944 | int64_t sector_num, int nb_sectors, QEMUIOVector *qiov) |
2945 | { | |
2946 | /* Perform I/O through a temporary buffer so that users who scribble over | |
2947 | * their read buffer while the operation is in progress do not end up | |
2948 | * modifying the image file. This is critical for zero-copy guest I/O | |
2949 | * where anything might happen inside guest memory. | |
2950 | */ | |
2951 | void *bounce_buffer; | |
2952 | ||
79c053bd | 2953 | BlockDriver *drv = bs->drv; |
ab185921 SH |
2954 | struct iovec iov; |
2955 | QEMUIOVector bounce_qiov; | |
2956 | int64_t cluster_sector_num; | |
2957 | int cluster_nb_sectors; | |
2958 | size_t skip_bytes; | |
2959 | int ret; | |
2960 | ||
2961 | /* Cover entire cluster so no additional backing file I/O is required when | |
2962 | * allocating cluster in the image file. | |
2963 | */ | |
343bded4 PB |
2964 | bdrv_round_to_clusters(bs, sector_num, nb_sectors, |
2965 | &cluster_sector_num, &cluster_nb_sectors); | |
ab185921 | 2966 | |
470c0504 SH |
2967 | trace_bdrv_co_do_copy_on_readv(bs, sector_num, nb_sectors, |
2968 | cluster_sector_num, cluster_nb_sectors); | |
ab185921 SH |
2969 | |
2970 | iov.iov_len = cluster_nb_sectors * BDRV_SECTOR_SIZE; | |
2971 | iov.iov_base = bounce_buffer = qemu_blockalign(bs, iov.iov_len); | |
2972 | qemu_iovec_init_external(&bounce_qiov, &iov, 1); | |
2973 | ||
79c053bd SH |
2974 | ret = drv->bdrv_co_readv(bs, cluster_sector_num, cluster_nb_sectors, |
2975 | &bounce_qiov); | |
ab185921 SH |
2976 | if (ret < 0) { |
2977 | goto err; | |
2978 | } | |
2979 | ||
79c053bd SH |
2980 | if (drv->bdrv_co_write_zeroes && |
2981 | buffer_is_zero(bounce_buffer, iov.iov_len)) { | |
621f0589 | 2982 | ret = bdrv_co_do_write_zeroes(bs, cluster_sector_num, |
aa7bfbff | 2983 | cluster_nb_sectors, 0); |
79c053bd | 2984 | } else { |
f05fa4ad PB |
2985 | /* This does not change the data on the disk, it is not necessary |
2986 | * to flush even in cache=writethrough mode. | |
2987 | */ | |
79c053bd | 2988 | ret = drv->bdrv_co_writev(bs, cluster_sector_num, cluster_nb_sectors, |
ab185921 | 2989 | &bounce_qiov); |
79c053bd SH |
2990 | } |
2991 | ||
ab185921 SH |
2992 | if (ret < 0) { |
2993 | /* It might be okay to ignore write errors for guest requests. If this | |
2994 | * is a deliberate copy-on-read then we don't want to ignore the error. | |
2995 | * Simply report it in all cases. | |
2996 | */ | |
2997 | goto err; | |
2998 | } | |
2999 | ||
3000 | skip_bytes = (sector_num - cluster_sector_num) * BDRV_SECTOR_SIZE; | |
03396148 MT |
3001 | qemu_iovec_from_buf(qiov, 0, bounce_buffer + skip_bytes, |
3002 | nb_sectors * BDRV_SECTOR_SIZE); | |
ab185921 SH |
3003 | |
3004 | err: | |
3005 | qemu_vfree(bounce_buffer); | |
3006 | return ret; | |
3007 | } | |
3008 | ||
c5fbe571 | 3009 | /* |
d0c7f642 KW |
3010 | * Forwards an already correctly aligned request to the BlockDriver. This |
3011 | * handles copy on read and zeroing after EOF; any other features must be | |
3012 | * implemented by the caller. | |
c5fbe571 | 3013 | */ |
d0c7f642 | 3014 | static int coroutine_fn bdrv_aligned_preadv(BlockDriverState *bs, |
65afd211 | 3015 | BdrvTrackedRequest *req, int64_t offset, unsigned int bytes, |
ec746e10 | 3016 | int64_t align, QEMUIOVector *qiov, int flags) |
da1fa91d KW |
3017 | { |
3018 | BlockDriver *drv = bs->drv; | |
dbffbdcf | 3019 | int ret; |
da1fa91d | 3020 | |
d0c7f642 KW |
3021 | int64_t sector_num = offset >> BDRV_SECTOR_BITS; |
3022 | unsigned int nb_sectors = bytes >> BDRV_SECTOR_BITS; | |
da1fa91d | 3023 | |
d0c7f642 KW |
3024 | assert((offset & (BDRV_SECTOR_SIZE - 1)) == 0); |
3025 | assert((bytes & (BDRV_SECTOR_SIZE - 1)) == 0); | |
8eb029c2 | 3026 | assert(!qiov || bytes == qiov->size); |
d0c7f642 KW |
3027 | |
3028 | /* Handle Copy on Read and associated serialisation */ | |
470c0504 | 3029 | if (flags & BDRV_REQ_COPY_ON_READ) { |
7327145f KW |
3030 | /* If we touch the same cluster it counts as an overlap. This |
3031 | * guarantees that allocating writes will be serialized and not race | |
3032 | * with each other for the same cluster. For example, in copy-on-read | |
3033 | * it ensures that the CoR read and write operations are atomic and | |
3034 | * guest writes cannot interleave between them. */ | |
3035 | mark_request_serialising(req, bdrv_get_cluster_size(bs)); | |
470c0504 SH |
3036 | } |
3037 | ||
2dbafdc0 | 3038 | wait_serialising_requests(req); |
f4658285 | 3039 | |
470c0504 | 3040 | if (flags & BDRV_REQ_COPY_ON_READ) { |
ab185921 SH |
3041 | int pnum; |
3042 | ||
bdad13b9 | 3043 | ret = bdrv_is_allocated(bs, sector_num, nb_sectors, &pnum); |
ab185921 SH |
3044 | if (ret < 0) { |
3045 | goto out; | |
3046 | } | |
3047 | ||
3048 | if (!ret || pnum != nb_sectors) { | |
470c0504 | 3049 | ret = bdrv_co_do_copy_on_readv(bs, sector_num, nb_sectors, qiov); |
ab185921 SH |
3050 | goto out; |
3051 | } | |
3052 | } | |
3053 | ||
d0c7f642 | 3054 | /* Forward the request to the BlockDriver */ |
893a8f62 MK |
3055 | if (!(bs->zero_beyond_eof && bs->growable)) { |
3056 | ret = drv->bdrv_co_readv(bs, sector_num, nb_sectors, qiov); | |
3057 | } else { | |
3058 | /* Read zeros after EOF of growable BDSes */ | |
3059 | int64_t len, total_sectors, max_nb_sectors; | |
3060 | ||
3061 | len = bdrv_getlength(bs); | |
3062 | if (len < 0) { | |
3063 | ret = len; | |
3064 | goto out; | |
3065 | } | |
3066 | ||
d055a1fe | 3067 | total_sectors = DIV_ROUND_UP(len, BDRV_SECTOR_SIZE); |
5f5bcd80 KW |
3068 | max_nb_sectors = ROUND_UP(MAX(0, total_sectors - sector_num), |
3069 | align >> BDRV_SECTOR_BITS); | |
893a8f62 | 3070 | if (max_nb_sectors > 0) { |
33f461e0 KW |
3071 | QEMUIOVector local_qiov; |
3072 | size_t local_sectors; | |
3073 | ||
3074 | max_nb_sectors = MIN(max_nb_sectors, SIZE_MAX / BDRV_SECTOR_BITS); | |
3075 | local_sectors = MIN(max_nb_sectors, nb_sectors); | |
3076 | ||
3077 | qemu_iovec_init(&local_qiov, qiov->niov); | |
3078 | qemu_iovec_concat(&local_qiov, qiov, 0, | |
3079 | local_sectors * BDRV_SECTOR_SIZE); | |
3080 | ||
3081 | ret = drv->bdrv_co_readv(bs, sector_num, local_sectors, | |
3082 | &local_qiov); | |
3083 | ||
3084 | qemu_iovec_destroy(&local_qiov); | |
893a8f62 MK |
3085 | } else { |
3086 | ret = 0; | |
3087 | } | |
3088 | ||
3089 | /* Reading beyond end of file is supposed to produce zeroes */ | |
3090 | if (ret == 0 && total_sectors < sector_num + nb_sectors) { | |
3091 | uint64_t offset = MAX(0, total_sectors - sector_num); | |
3092 | uint64_t bytes = (sector_num + nb_sectors - offset) * | |
3093 | BDRV_SECTOR_SIZE; | |
3094 | qemu_iovec_memset(qiov, offset * BDRV_SECTOR_SIZE, 0, bytes); | |
3095 | } | |
3096 | } | |
ab185921 SH |
3097 | |
3098 | out: | |
dbffbdcf | 3099 | return ret; |
da1fa91d KW |
3100 | } |
3101 | ||
d0c7f642 KW |
3102 | /* |
3103 | * Handle a read request in coroutine context | |
3104 | */ | |
1b0288ae KW |
3105 | static int coroutine_fn bdrv_co_do_preadv(BlockDriverState *bs, |
3106 | int64_t offset, unsigned int bytes, QEMUIOVector *qiov, | |
d0c7f642 KW |
3107 | BdrvRequestFlags flags) |
3108 | { | |
3109 | BlockDriver *drv = bs->drv; | |
65afd211 KW |
3110 | BdrvTrackedRequest req; |
3111 | ||
1b0288ae KW |
3112 | /* TODO Lift BDRV_SECTOR_SIZE restriction in BlockDriver interface */ |
3113 | uint64_t align = MAX(BDRV_SECTOR_SIZE, bs->request_alignment); | |
3114 | uint8_t *head_buf = NULL; | |
3115 | uint8_t *tail_buf = NULL; | |
3116 | QEMUIOVector local_qiov; | |
3117 | bool use_local_qiov = false; | |
d0c7f642 KW |
3118 | int ret; |
3119 | ||
3120 | if (!drv) { | |
3121 | return -ENOMEDIUM; | |
3122 | } | |
1b0288ae | 3123 | if (bdrv_check_byte_request(bs, offset, bytes)) { |
d0c7f642 KW |
3124 | return -EIO; |
3125 | } | |
3126 | ||
3127 | if (bs->copy_on_read) { | |
3128 | flags |= BDRV_REQ_COPY_ON_READ; | |
3129 | } | |
3130 | ||
3131 | /* throttling disk I/O */ | |
3132 | if (bs->io_limits_enabled) { | |
d5103588 | 3133 | bdrv_io_limits_intercept(bs, bytes, false); |
1b0288ae KW |
3134 | } |
3135 | ||
3136 | /* Align read if necessary by padding qiov */ | |
3137 | if (offset & (align - 1)) { | |
3138 | head_buf = qemu_blockalign(bs, align); | |
3139 | qemu_iovec_init(&local_qiov, qiov->niov + 2); | |
3140 | qemu_iovec_add(&local_qiov, head_buf, offset & (align - 1)); | |
3141 | qemu_iovec_concat(&local_qiov, qiov, 0, qiov->size); | |
3142 | use_local_qiov = true; | |
3143 | ||
3144 | bytes += offset & (align - 1); | |
3145 | offset = offset & ~(align - 1); | |
3146 | } | |
3147 | ||
3148 | if ((offset + bytes) & (align - 1)) { | |
3149 | if (!use_local_qiov) { | |
3150 | qemu_iovec_init(&local_qiov, qiov->niov + 1); | |
3151 | qemu_iovec_concat(&local_qiov, qiov, 0, qiov->size); | |
3152 | use_local_qiov = true; | |
3153 | } | |
3154 | tail_buf = qemu_blockalign(bs, align); | |
3155 | qemu_iovec_add(&local_qiov, tail_buf, | |
3156 | align - ((offset + bytes) & (align - 1))); | |
3157 | ||
3158 | bytes = ROUND_UP(bytes, align); | |
3159 | } | |
3160 | ||
65afd211 | 3161 | tracked_request_begin(&req, bs, offset, bytes, false); |
ec746e10 | 3162 | ret = bdrv_aligned_preadv(bs, &req, offset, bytes, align, |
1b0288ae KW |
3163 | use_local_qiov ? &local_qiov : qiov, |
3164 | flags); | |
65afd211 | 3165 | tracked_request_end(&req); |
1b0288ae KW |
3166 | |
3167 | if (use_local_qiov) { | |
3168 | qemu_iovec_destroy(&local_qiov); | |
3169 | qemu_vfree(head_buf); | |
3170 | qemu_vfree(tail_buf); | |
d0c7f642 KW |
3171 | } |
3172 | ||
d0c7f642 KW |
3173 | return ret; |
3174 | } | |
3175 | ||
1b0288ae KW |
3176 | static int coroutine_fn bdrv_co_do_readv(BlockDriverState *bs, |
3177 | int64_t sector_num, int nb_sectors, QEMUIOVector *qiov, | |
3178 | BdrvRequestFlags flags) | |
3179 | { | |
3180 | if (nb_sectors < 0 || nb_sectors > (UINT_MAX >> BDRV_SECTOR_BITS)) { | |
3181 | return -EINVAL; | |
3182 | } | |
3183 | ||
3184 | return bdrv_co_do_preadv(bs, sector_num << BDRV_SECTOR_BITS, | |
3185 | nb_sectors << BDRV_SECTOR_BITS, qiov, flags); | |
3186 | } | |
3187 | ||
c5fbe571 | 3188 | int coroutine_fn bdrv_co_readv(BlockDriverState *bs, int64_t sector_num, |
da1fa91d KW |
3189 | int nb_sectors, QEMUIOVector *qiov) |
3190 | { | |
c5fbe571 | 3191 | trace_bdrv_co_readv(bs, sector_num, nb_sectors); |
da1fa91d | 3192 | |
470c0504 SH |
3193 | return bdrv_co_do_readv(bs, sector_num, nb_sectors, qiov, 0); |
3194 | } | |
3195 | ||
3196 | int coroutine_fn bdrv_co_copy_on_readv(BlockDriverState *bs, | |
3197 | int64_t sector_num, int nb_sectors, QEMUIOVector *qiov) | |
3198 | { | |
3199 | trace_bdrv_co_copy_on_readv(bs, sector_num, nb_sectors); | |
3200 | ||
3201 | return bdrv_co_do_readv(bs, sector_num, nb_sectors, qiov, | |
3202 | BDRV_REQ_COPY_ON_READ); | |
c5fbe571 SH |
3203 | } |
3204 | ||
c31cb707 PL |
3205 | /* if no limit is specified in the BlockLimits use a default |
3206 | * of 32768 512-byte sectors (16 MiB) per request. | |
3207 | */ | |
3208 | #define MAX_WRITE_ZEROES_DEFAULT 32768 | |
3209 | ||
f08f2dda | 3210 | static int coroutine_fn bdrv_co_do_write_zeroes(BlockDriverState *bs, |
aa7bfbff | 3211 | int64_t sector_num, int nb_sectors, BdrvRequestFlags flags) |
f08f2dda SH |
3212 | { |
3213 | BlockDriver *drv = bs->drv; | |
3214 | QEMUIOVector qiov; | |
c31cb707 PL |
3215 | struct iovec iov = {0}; |
3216 | int ret = 0; | |
f08f2dda | 3217 | |
c31cb707 PL |
3218 | int max_write_zeroes = bs->bl.max_write_zeroes ? |
3219 | bs->bl.max_write_zeroes : MAX_WRITE_ZEROES_DEFAULT; | |
621f0589 | 3220 | |
c31cb707 PL |
3221 | while (nb_sectors > 0 && !ret) { |
3222 | int num = nb_sectors; | |
3223 | ||
b8d71c09 PB |
3224 | /* Align request. Block drivers can expect the "bulk" of the request |
3225 | * to be aligned. | |
3226 | */ | |
3227 | if (bs->bl.write_zeroes_alignment | |
3228 | && num > bs->bl.write_zeroes_alignment) { | |
3229 | if (sector_num % bs->bl.write_zeroes_alignment != 0) { | |
3230 | /* Make a small request up to the first aligned sector. */ | |
c31cb707 | 3231 | num = bs->bl.write_zeroes_alignment; |
b8d71c09 PB |
3232 | num -= sector_num % bs->bl.write_zeroes_alignment; |
3233 | } else if ((sector_num + num) % bs->bl.write_zeroes_alignment != 0) { | |
3234 | /* Shorten the request to the last aligned sector. num cannot | |
3235 | * underflow because num > bs->bl.write_zeroes_alignment. | |
3236 | */ | |
3237 | num -= (sector_num + num) % bs->bl.write_zeroes_alignment; | |
c31cb707 | 3238 | } |
621f0589 | 3239 | } |
f08f2dda | 3240 | |
c31cb707 PL |
3241 | /* limit request size */ |
3242 | if (num > max_write_zeroes) { | |
3243 | num = max_write_zeroes; | |
3244 | } | |
3245 | ||
3246 | ret = -ENOTSUP; | |
3247 | /* First try the efficient write zeroes operation */ | |
3248 | if (drv->bdrv_co_write_zeroes) { | |
3249 | ret = drv->bdrv_co_write_zeroes(bs, sector_num, num, flags); | |
3250 | } | |
3251 | ||
3252 | if (ret == -ENOTSUP) { | |
3253 | /* Fall back to bounce buffer if write zeroes is unsupported */ | |
3254 | iov.iov_len = num * BDRV_SECTOR_SIZE; | |
3255 | if (iov.iov_base == NULL) { | |
b8d71c09 PB |
3256 | iov.iov_base = qemu_blockalign(bs, num * BDRV_SECTOR_SIZE); |
3257 | memset(iov.iov_base, 0, num * BDRV_SECTOR_SIZE); | |
c31cb707 PL |
3258 | } |
3259 | qemu_iovec_init_external(&qiov, &iov, 1); | |
f08f2dda | 3260 | |
c31cb707 | 3261 | ret = drv->bdrv_co_writev(bs, sector_num, num, &qiov); |
b8d71c09 PB |
3262 | |
3263 | /* Keep bounce buffer around if it is big enough for all | |
3264 | * all future requests. | |
3265 | */ | |
3266 | if (num < max_write_zeroes) { | |
3267 | qemu_vfree(iov.iov_base); | |
3268 | iov.iov_base = NULL; | |
3269 | } | |
c31cb707 PL |
3270 | } |
3271 | ||
3272 | sector_num += num; | |
3273 | nb_sectors -= num; | |
3274 | } | |
f08f2dda SH |
3275 | |
3276 | qemu_vfree(iov.iov_base); | |
3277 | return ret; | |
3278 | } | |
3279 | ||
c5fbe571 | 3280 | /* |
b404f720 | 3281 | * Forwards an already correctly aligned write request to the BlockDriver. |
c5fbe571 | 3282 | */ |
b404f720 | 3283 | static int coroutine_fn bdrv_aligned_pwritev(BlockDriverState *bs, |
65afd211 KW |
3284 | BdrvTrackedRequest *req, int64_t offset, unsigned int bytes, |
3285 | QEMUIOVector *qiov, int flags) | |
c5fbe571 SH |
3286 | { |
3287 | BlockDriver *drv = bs->drv; | |
28de2dcd | 3288 | bool waited; |
6b7cb247 | 3289 | int ret; |
da1fa91d | 3290 | |
b404f720 KW |
3291 | int64_t sector_num = offset >> BDRV_SECTOR_BITS; |
3292 | unsigned int nb_sectors = bytes >> BDRV_SECTOR_BITS; | |
f4658285 | 3293 | |
b404f720 KW |
3294 | assert((offset & (BDRV_SECTOR_SIZE - 1)) == 0); |
3295 | assert((bytes & (BDRV_SECTOR_SIZE - 1)) == 0); | |
8eb029c2 | 3296 | assert(!qiov || bytes == qiov->size); |
cc0681c4 | 3297 | |
28de2dcd KW |
3298 | waited = wait_serialising_requests(req); |
3299 | assert(!waited || !req->serialising); | |
af91f9a7 KW |
3300 | assert(req->overlap_offset <= offset); |
3301 | assert(offset + bytes <= req->overlap_offset + req->overlap_bytes); | |
244eadef | 3302 | |
65afd211 | 3303 | ret = notifier_with_return_list_notify(&bs->before_write_notifiers, req); |
d616b224 | 3304 | |
465bee1d PL |
3305 | if (!ret && bs->detect_zeroes != BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF && |
3306 | !(flags & BDRV_REQ_ZERO_WRITE) && drv->bdrv_co_write_zeroes && | |
3307 | qemu_iovec_is_zero(qiov)) { | |
3308 | flags |= BDRV_REQ_ZERO_WRITE; | |
3309 | if (bs->detect_zeroes == BLOCKDEV_DETECT_ZEROES_OPTIONS_UNMAP) { | |
3310 | flags |= BDRV_REQ_MAY_UNMAP; | |
3311 | } | |
3312 | } | |
3313 | ||
d616b224 SH |
3314 | if (ret < 0) { |
3315 | /* Do nothing, write notifier decided to fail this request */ | |
3316 | } else if (flags & BDRV_REQ_ZERO_WRITE) { | |
9e1cb96d | 3317 | BLKDBG_EVENT(bs, BLKDBG_PWRITEV_ZERO); |
aa7bfbff | 3318 | ret = bdrv_co_do_write_zeroes(bs, sector_num, nb_sectors, flags); |
f08f2dda | 3319 | } else { |
9e1cb96d | 3320 | BLKDBG_EVENT(bs, BLKDBG_PWRITEV); |
f08f2dda SH |
3321 | ret = drv->bdrv_co_writev(bs, sector_num, nb_sectors, qiov); |
3322 | } | |
9e1cb96d | 3323 | BLKDBG_EVENT(bs, BLKDBG_PWRITEV_DONE); |
6b7cb247 | 3324 | |
f05fa4ad PB |
3325 | if (ret == 0 && !bs->enable_write_cache) { |
3326 | ret = bdrv_co_flush(bs); | |
3327 | } | |
3328 | ||
e4654d2d | 3329 | bdrv_set_dirty(bs, sector_num, nb_sectors); |
da1fa91d KW |
3330 | |
3331 | if (bs->wr_highest_sector < sector_num + nb_sectors - 1) { | |
3332 | bs->wr_highest_sector = sector_num + nb_sectors - 1; | |
3333 | } | |
df2a6f29 PB |
3334 | if (bs->growable && ret >= 0) { |
3335 | bs->total_sectors = MAX(bs->total_sectors, sector_num + nb_sectors); | |
3336 | } | |
da1fa91d | 3337 | |
6b7cb247 | 3338 | return ret; |
da1fa91d KW |
3339 | } |
3340 | ||
b404f720 KW |
3341 | /* |
3342 | * Handle a write request in coroutine context | |
3343 | */ | |
6601553e KW |
3344 | static int coroutine_fn bdrv_co_do_pwritev(BlockDriverState *bs, |
3345 | int64_t offset, unsigned int bytes, QEMUIOVector *qiov, | |
b404f720 KW |
3346 | BdrvRequestFlags flags) |
3347 | { | |
65afd211 | 3348 | BdrvTrackedRequest req; |
3b8242e0 KW |
3349 | /* TODO Lift BDRV_SECTOR_SIZE restriction in BlockDriver interface */ |
3350 | uint64_t align = MAX(BDRV_SECTOR_SIZE, bs->request_alignment); | |
3351 | uint8_t *head_buf = NULL; | |
3352 | uint8_t *tail_buf = NULL; | |
3353 | QEMUIOVector local_qiov; | |
3354 | bool use_local_qiov = false; | |
b404f720 KW |
3355 | int ret; |
3356 | ||
3357 | if (!bs->drv) { | |
3358 | return -ENOMEDIUM; | |
3359 | } | |
3360 | if (bs->read_only) { | |
3361 | return -EACCES; | |
3362 | } | |
6601553e | 3363 | if (bdrv_check_byte_request(bs, offset, bytes)) { |
b404f720 KW |
3364 | return -EIO; |
3365 | } | |
3366 | ||
b404f720 KW |
3367 | /* throttling disk I/O */ |
3368 | if (bs->io_limits_enabled) { | |
d5103588 | 3369 | bdrv_io_limits_intercept(bs, bytes, true); |
b404f720 KW |
3370 | } |
3371 | ||
3b8242e0 KW |
3372 | /* |
3373 | * Align write if necessary by performing a read-modify-write cycle. | |
3374 | * Pad qiov with the read parts and be sure to have a tracked request not | |
3375 | * only for bdrv_aligned_pwritev, but also for the reads of the RMW cycle. | |
3376 | */ | |
65afd211 | 3377 | tracked_request_begin(&req, bs, offset, bytes, true); |
3b8242e0 KW |
3378 | |
3379 | if (offset & (align - 1)) { | |
3380 | QEMUIOVector head_qiov; | |
3381 | struct iovec head_iov; | |
3382 | ||
3383 | mark_request_serialising(&req, align); | |
3384 | wait_serialising_requests(&req); | |
3385 | ||
3386 | head_buf = qemu_blockalign(bs, align); | |
3387 | head_iov = (struct iovec) { | |
3388 | .iov_base = head_buf, | |
3389 | .iov_len = align, | |
3390 | }; | |
3391 | qemu_iovec_init_external(&head_qiov, &head_iov, 1); | |
3392 | ||
9e1cb96d | 3393 | BLKDBG_EVENT(bs, BLKDBG_PWRITEV_RMW_HEAD); |
3b8242e0 KW |
3394 | ret = bdrv_aligned_preadv(bs, &req, offset & ~(align - 1), align, |
3395 | align, &head_qiov, 0); | |
3396 | if (ret < 0) { | |
3397 | goto fail; | |
3398 | } | |
9e1cb96d | 3399 | BLKDBG_EVENT(bs, BLKDBG_PWRITEV_RMW_AFTER_HEAD); |
3b8242e0 KW |
3400 | |
3401 | qemu_iovec_init(&local_qiov, qiov->niov + 2); | |
3402 | qemu_iovec_add(&local_qiov, head_buf, offset & (align - 1)); | |
3403 | qemu_iovec_concat(&local_qiov, qiov, 0, qiov->size); | |
3404 | use_local_qiov = true; | |
3405 | ||
3406 | bytes += offset & (align - 1); | |
3407 | offset = offset & ~(align - 1); | |
3408 | } | |
3409 | ||
3410 | if ((offset + bytes) & (align - 1)) { | |
3411 | QEMUIOVector tail_qiov; | |
3412 | struct iovec tail_iov; | |
3413 | size_t tail_bytes; | |
28de2dcd | 3414 | bool waited; |
3b8242e0 KW |
3415 | |
3416 | mark_request_serialising(&req, align); | |
28de2dcd KW |
3417 | waited = wait_serialising_requests(&req); |
3418 | assert(!waited || !use_local_qiov); | |
3b8242e0 KW |
3419 | |
3420 | tail_buf = qemu_blockalign(bs, align); | |
3421 | tail_iov = (struct iovec) { | |
3422 | .iov_base = tail_buf, | |
3423 | .iov_len = align, | |
3424 | }; | |
3425 | qemu_iovec_init_external(&tail_qiov, &tail_iov, 1); | |
3426 | ||
9e1cb96d | 3427 | BLKDBG_EVENT(bs, BLKDBG_PWRITEV_RMW_TAIL); |
3b8242e0 KW |
3428 | ret = bdrv_aligned_preadv(bs, &req, (offset + bytes) & ~(align - 1), align, |
3429 | align, &tail_qiov, 0); | |
3430 | if (ret < 0) { | |
3431 | goto fail; | |
3432 | } | |
9e1cb96d | 3433 | BLKDBG_EVENT(bs, BLKDBG_PWRITEV_RMW_AFTER_TAIL); |
3b8242e0 KW |
3434 | |
3435 | if (!use_local_qiov) { | |
3436 | qemu_iovec_init(&local_qiov, qiov->niov + 1); | |
3437 | qemu_iovec_concat(&local_qiov, qiov, 0, qiov->size); | |
3438 | use_local_qiov = true; | |
3439 | } | |
3440 | ||
3441 | tail_bytes = (offset + bytes) & (align - 1); | |
3442 | qemu_iovec_add(&local_qiov, tail_buf + tail_bytes, align - tail_bytes); | |
3443 | ||
3444 | bytes = ROUND_UP(bytes, align); | |
3445 | } | |
3446 | ||
3447 | ret = bdrv_aligned_pwritev(bs, &req, offset, bytes, | |
3448 | use_local_qiov ? &local_qiov : qiov, | |
3449 | flags); | |
3450 | ||
3451 | fail: | |
65afd211 | 3452 | tracked_request_end(&req); |
b404f720 | 3453 | |
3b8242e0 KW |
3454 | if (use_local_qiov) { |
3455 | qemu_iovec_destroy(&local_qiov); | |
3b8242e0 | 3456 | } |
99c4a85c KW |
3457 | qemu_vfree(head_buf); |
3458 | qemu_vfree(tail_buf); | |
3b8242e0 | 3459 | |
b404f720 KW |
3460 | return ret; |
3461 | } | |
3462 | ||
6601553e KW |
3463 | static int coroutine_fn bdrv_co_do_writev(BlockDriverState *bs, |
3464 | int64_t sector_num, int nb_sectors, QEMUIOVector *qiov, | |
3465 | BdrvRequestFlags flags) | |
3466 | { | |
3467 | if (nb_sectors < 0 || nb_sectors > (INT_MAX >> BDRV_SECTOR_BITS)) { | |
3468 | return -EINVAL; | |
3469 | } | |
3470 | ||
3471 | return bdrv_co_do_pwritev(bs, sector_num << BDRV_SECTOR_BITS, | |
3472 | nb_sectors << BDRV_SECTOR_BITS, qiov, flags); | |
3473 | } | |
3474 | ||
c5fbe571 SH |
3475 | int coroutine_fn bdrv_co_writev(BlockDriverState *bs, int64_t sector_num, |
3476 | int nb_sectors, QEMUIOVector *qiov) | |
3477 | { | |
3478 | trace_bdrv_co_writev(bs, sector_num, nb_sectors); | |
3479 | ||
f08f2dda SH |
3480 | return bdrv_co_do_writev(bs, sector_num, nb_sectors, qiov, 0); |
3481 | } | |
3482 | ||
3483 | int coroutine_fn bdrv_co_write_zeroes(BlockDriverState *bs, | |
aa7bfbff PL |
3484 | int64_t sector_num, int nb_sectors, |
3485 | BdrvRequestFlags flags) | |
f08f2dda | 3486 | { |
94d6ff21 | 3487 | trace_bdrv_co_write_zeroes(bs, sector_num, nb_sectors, flags); |
f08f2dda | 3488 | |
d32f35cb PL |
3489 | if (!(bs->open_flags & BDRV_O_UNMAP)) { |
3490 | flags &= ~BDRV_REQ_MAY_UNMAP; | |
3491 | } | |
3492 | ||
f08f2dda | 3493 | return bdrv_co_do_writev(bs, sector_num, nb_sectors, NULL, |
aa7bfbff | 3494 | BDRV_REQ_ZERO_WRITE | flags); |
c5fbe571 SH |
3495 | } |
3496 | ||
83f64091 FB |
3497 | /** |
3498 | * Truncate file to 'offset' bytes (needed only for file protocols) | |
3499 | */ | |
3500 | int bdrv_truncate(BlockDriverState *bs, int64_t offset) | |
3501 | { | |
3502 | BlockDriver *drv = bs->drv; | |
51762288 | 3503 | int ret; |
83f64091 | 3504 | if (!drv) |
19cb3738 | 3505 | return -ENOMEDIUM; |
83f64091 FB |
3506 | if (!drv->bdrv_truncate) |
3507 | return -ENOTSUP; | |
59f2689d NS |
3508 | if (bs->read_only) |
3509 | return -EACCES; | |
9c75e168 | 3510 | |
51762288 SH |
3511 | ret = drv->bdrv_truncate(bs, offset); |
3512 | if (ret == 0) { | |
3513 | ret = refresh_total_sectors(bs, offset >> BDRV_SECTOR_BITS); | |
145feb17 | 3514 | bdrv_dev_resize_cb(bs); |
51762288 SH |
3515 | } |
3516 | return ret; | |
83f64091 FB |
3517 | } |
3518 | ||
4a1d5e1f FZ |
3519 | /** |
3520 | * Length of a allocated file in bytes. Sparse files are counted by actual | |
3521 | * allocated space. Return < 0 if error or unknown. | |
3522 | */ | |
3523 | int64_t bdrv_get_allocated_file_size(BlockDriverState *bs) | |
3524 | { | |
3525 | BlockDriver *drv = bs->drv; | |
3526 | if (!drv) { | |
3527 | return -ENOMEDIUM; | |
3528 | } | |
3529 | if (drv->bdrv_get_allocated_file_size) { | |
3530 | return drv->bdrv_get_allocated_file_size(bs); | |
3531 | } | |
3532 | if (bs->file) { | |
3533 | return bdrv_get_allocated_file_size(bs->file); | |
3534 | } | |
3535 | return -ENOTSUP; | |
3536 | } | |
3537 | ||
83f64091 FB |
3538 | /** |
3539 | * Length of a file in bytes. Return < 0 if error or unknown. | |
3540 | */ | |
3541 | int64_t bdrv_getlength(BlockDriverState *bs) | |
3542 | { | |
3543 | BlockDriver *drv = bs->drv; | |
3544 | if (!drv) | |
19cb3738 | 3545 | return -ENOMEDIUM; |
51762288 | 3546 | |
b94a2610 KW |
3547 | if (drv->has_variable_length) { |
3548 | int ret = refresh_total_sectors(bs, bs->total_sectors); | |
3549 | if (ret < 0) { | |
3550 | return ret; | |
46a4e4e6 | 3551 | } |
83f64091 | 3552 | } |
46a4e4e6 | 3553 | return bs->total_sectors * BDRV_SECTOR_SIZE; |
fc01f7e7 FB |
3554 | } |
3555 | ||
19cb3738 | 3556 | /* return 0 as number of sectors if no device present or error */ |
96b8f136 | 3557 | void bdrv_get_geometry(BlockDriverState *bs, uint64_t *nb_sectors_ptr) |
fc01f7e7 | 3558 | { |
19cb3738 FB |
3559 | int64_t length; |
3560 | length = bdrv_getlength(bs); | |
3561 | if (length < 0) | |
3562 | length = 0; | |
3563 | else | |
6ea44308 | 3564 | length = length >> BDRV_SECTOR_BITS; |
19cb3738 | 3565 | *nb_sectors_ptr = length; |
fc01f7e7 | 3566 | } |
cf98951b | 3567 | |
ff06f5f3 PB |
3568 | void bdrv_set_on_error(BlockDriverState *bs, BlockdevOnError on_read_error, |
3569 | BlockdevOnError on_write_error) | |
abd7f68d MA |
3570 | { |
3571 | bs->on_read_error = on_read_error; | |
3572 | bs->on_write_error = on_write_error; | |
3573 | } | |
3574 | ||
1ceee0d5 | 3575 | BlockdevOnError bdrv_get_on_error(BlockDriverState *bs, bool is_read) |
abd7f68d MA |
3576 | { |
3577 | return is_read ? bs->on_read_error : bs->on_write_error; | |
3578 | } | |
3579 | ||
3e1caa5f PB |
3580 | BlockErrorAction bdrv_get_error_action(BlockDriverState *bs, bool is_read, int error) |
3581 | { | |
3582 | BlockdevOnError on_err = is_read ? bs->on_read_error : bs->on_write_error; | |
3583 | ||
3584 | switch (on_err) { | |
3585 | case BLOCKDEV_ON_ERROR_ENOSPC: | |
a589569f WX |
3586 | return (error == ENOSPC) ? |
3587 | BLOCK_ERROR_ACTION_STOP : BLOCK_ERROR_ACTION_REPORT; | |
3e1caa5f | 3588 | case BLOCKDEV_ON_ERROR_STOP: |
a589569f | 3589 | return BLOCK_ERROR_ACTION_STOP; |
3e1caa5f | 3590 | case BLOCKDEV_ON_ERROR_REPORT: |
a589569f | 3591 | return BLOCK_ERROR_ACTION_REPORT; |
3e1caa5f | 3592 | case BLOCKDEV_ON_ERROR_IGNORE: |
a589569f | 3593 | return BLOCK_ERROR_ACTION_IGNORE; |
3e1caa5f PB |
3594 | default: |
3595 | abort(); | |
3596 | } | |
3597 | } | |
3598 | ||
3599 | /* This is done by device models because, while the block layer knows | |
3600 | * about the error, it does not know whether an operation comes from | |
3601 | * the device or the block layer (from a job, for example). | |
3602 | */ | |
3603 | void bdrv_error_action(BlockDriverState *bs, BlockErrorAction action, | |
3604 | bool is_read, int error) | |
3605 | { | |
3606 | assert(error >= 0); | |
2bd3bce8 | 3607 | |
a589569f | 3608 | if (action == BLOCK_ERROR_ACTION_STOP) { |
2bd3bce8 PB |
3609 | /* First set the iostatus, so that "info block" returns an iostatus |
3610 | * that matches the events raised so far (an additional error iostatus | |
3611 | * is fine, but not a lost one). | |
3612 | */ | |
3e1caa5f | 3613 | bdrv_iostatus_set_err(bs, error); |
2bd3bce8 PB |
3614 | |
3615 | /* Then raise the request to stop the VM and the event. | |
3616 | * qemu_system_vmstop_request_prepare has two effects. First, | |
3617 | * it ensures that the STOP event always comes after the | |
3618 | * BLOCK_IO_ERROR event. Second, it ensures that even if management | |
3619 | * can observe the STOP event and do a "cont" before the STOP | |
3620 | * event is issued, the VM will not stop. In this case, vm_start() | |
3621 | * also ensures that the STOP/RESUME pair of events is emitted. | |
3622 | */ | |
3623 | qemu_system_vmstop_request_prepare(); | |
5a2d2cbd WX |
3624 | qapi_event_send_block_io_error(bdrv_get_device_name(bs), |
3625 | is_read ? IO_OPERATION_TYPE_READ : | |
3626 | IO_OPERATION_TYPE_WRITE, | |
3627 | action, &error_abort); | |
2bd3bce8 PB |
3628 | qemu_system_vmstop_request(RUN_STATE_IO_ERROR); |
3629 | } else { | |
5a2d2cbd WX |
3630 | qapi_event_send_block_io_error(bdrv_get_device_name(bs), |
3631 | is_read ? IO_OPERATION_TYPE_READ : | |
3632 | IO_OPERATION_TYPE_WRITE, | |
3633 | action, &error_abort); | |
3e1caa5f PB |
3634 | } |
3635 | } | |
3636 | ||
b338082b FB |
3637 | int bdrv_is_read_only(BlockDriverState *bs) |
3638 | { | |
3639 | return bs->read_only; | |
3640 | } | |
3641 | ||
985a03b0 TS |
3642 | int bdrv_is_sg(BlockDriverState *bs) |
3643 | { | |
3644 | return bs->sg; | |
3645 | } | |
3646 | ||
e900a7b7 CH |
3647 | int bdrv_enable_write_cache(BlockDriverState *bs) |
3648 | { | |
3649 | return bs->enable_write_cache; | |
3650 | } | |
3651 | ||
425b0148 PB |
3652 | void bdrv_set_enable_write_cache(BlockDriverState *bs, bool wce) |
3653 | { | |
3654 | bs->enable_write_cache = wce; | |
55b110f2 JC |
3655 | |
3656 | /* so a reopen() will preserve wce */ | |
3657 | if (wce) { | |
3658 | bs->open_flags |= BDRV_O_CACHE_WB; | |
3659 | } else { | |
3660 | bs->open_flags &= ~BDRV_O_CACHE_WB; | |
3661 | } | |
425b0148 PB |
3662 | } |
3663 | ||
ea2384d3 FB |
3664 | int bdrv_is_encrypted(BlockDriverState *bs) |
3665 | { | |
3666 | if (bs->backing_hd && bs->backing_hd->encrypted) | |
3667 | return 1; | |
3668 | return bs->encrypted; | |
3669 | } | |
3670 | ||
c0f4ce77 AL |
3671 | int bdrv_key_required(BlockDriverState *bs) |
3672 | { | |
3673 | BlockDriverState *backing_hd = bs->backing_hd; | |
3674 | ||
3675 | if (backing_hd && backing_hd->encrypted && !backing_hd->valid_key) | |
3676 | return 1; | |
3677 | return (bs->encrypted && !bs->valid_key); | |
3678 | } | |
3679 | ||
ea2384d3 FB |
3680 | int bdrv_set_key(BlockDriverState *bs, const char *key) |
3681 | { | |
3682 | int ret; | |
3683 | if (bs->backing_hd && bs->backing_hd->encrypted) { | |
3684 | ret = bdrv_set_key(bs->backing_hd, key); | |
3685 | if (ret < 0) | |
3686 | return ret; | |
3687 | if (!bs->encrypted) | |
3688 | return 0; | |
3689 | } | |
fd04a2ae SH |
3690 | if (!bs->encrypted) { |
3691 | return -EINVAL; | |
3692 | } else if (!bs->drv || !bs->drv->bdrv_set_key) { | |
3693 | return -ENOMEDIUM; | |
3694 | } | |
c0f4ce77 | 3695 | ret = bs->drv->bdrv_set_key(bs, key); |
bb5fc20f AL |
3696 | if (ret < 0) { |
3697 | bs->valid_key = 0; | |
3698 | } else if (!bs->valid_key) { | |
3699 | bs->valid_key = 1; | |
3700 | /* call the change callback now, we skipped it on open */ | |
7d4b4ba5 | 3701 | bdrv_dev_change_media_cb(bs, true); |
bb5fc20f | 3702 | } |
c0f4ce77 | 3703 | return ret; |
ea2384d3 FB |
3704 | } |
3705 | ||
f8d6bba1 | 3706 | const char *bdrv_get_format_name(BlockDriverState *bs) |
ea2384d3 | 3707 | { |
f8d6bba1 | 3708 | return bs->drv ? bs->drv->format_name : NULL; |
ea2384d3 FB |
3709 | } |
3710 | ||
5fafdf24 | 3711 | void bdrv_iterate_format(void (*it)(void *opaque, const char *name), |
ea2384d3 FB |
3712 | void *opaque) |
3713 | { | |
3714 | BlockDriver *drv; | |
e855e4fb JC |
3715 | int count = 0; |
3716 | const char **formats = NULL; | |
ea2384d3 | 3717 | |
8a22f02a | 3718 | QLIST_FOREACH(drv, &bdrv_drivers, list) { |
e855e4fb JC |
3719 | if (drv->format_name) { |
3720 | bool found = false; | |
3721 | int i = count; | |
3722 | while (formats && i && !found) { | |
3723 | found = !strcmp(formats[--i], drv->format_name); | |
3724 | } | |
3725 | ||
3726 | if (!found) { | |
3727 | formats = g_realloc(formats, (count + 1) * sizeof(char *)); | |
3728 | formats[count++] = drv->format_name; | |
3729 | it(opaque, drv->format_name); | |
3730 | } | |
3731 | } | |
ea2384d3 | 3732 | } |
e855e4fb | 3733 | g_free(formats); |
ea2384d3 FB |
3734 | } |
3735 | ||
dc364f4c | 3736 | /* This function is to find block backend bs */ |
b338082b FB |
3737 | BlockDriverState *bdrv_find(const char *name) |
3738 | { | |
3739 | BlockDriverState *bs; | |
3740 | ||
dc364f4c | 3741 | QTAILQ_FOREACH(bs, &bdrv_states, device_list) { |
1b7bdbc1 | 3742 | if (!strcmp(name, bs->device_name)) { |
b338082b | 3743 | return bs; |
1b7bdbc1 | 3744 | } |
b338082b FB |
3745 | } |
3746 | return NULL; | |
3747 | } | |
3748 | ||
dc364f4c BC |
3749 | /* This function is to find a node in the bs graph */ |
3750 | BlockDriverState *bdrv_find_node(const char *node_name) | |
3751 | { | |
3752 | BlockDriverState *bs; | |
3753 | ||
3754 | assert(node_name); | |
3755 | ||
3756 | QTAILQ_FOREACH(bs, &graph_bdrv_states, node_list) { | |
3757 | if (!strcmp(node_name, bs->node_name)) { | |
3758 | return bs; | |
3759 | } | |
3760 | } | |
3761 | return NULL; | |
3762 | } | |
3763 | ||
c13163fb BC |
3764 | /* Put this QMP function here so it can access the static graph_bdrv_states. */ |
3765 | BlockDeviceInfoList *bdrv_named_nodes_list(void) | |
3766 | { | |
3767 | BlockDeviceInfoList *list, *entry; | |
3768 | BlockDriverState *bs; | |
3769 | ||
3770 | list = NULL; | |
3771 | QTAILQ_FOREACH(bs, &graph_bdrv_states, node_list) { | |
3772 | entry = g_malloc0(sizeof(*entry)); | |
3773 | entry->value = bdrv_block_device_info(bs); | |
3774 | entry->next = list; | |
3775 | list = entry; | |
3776 | } | |
3777 | ||
3778 | return list; | |
3779 | } | |
3780 | ||
12d3ba82 BC |
3781 | BlockDriverState *bdrv_lookup_bs(const char *device, |
3782 | const char *node_name, | |
3783 | Error **errp) | |
3784 | { | |
3785 | BlockDriverState *bs = NULL; | |
3786 | ||
12d3ba82 BC |
3787 | if (device) { |
3788 | bs = bdrv_find(device); | |
3789 | ||
dd67fa50 BC |
3790 | if (bs) { |
3791 | return bs; | |
12d3ba82 | 3792 | } |
12d3ba82 BC |
3793 | } |
3794 | ||
dd67fa50 BC |
3795 | if (node_name) { |
3796 | bs = bdrv_find_node(node_name); | |
12d3ba82 | 3797 | |
dd67fa50 BC |
3798 | if (bs) { |
3799 | return bs; | |
3800 | } | |
12d3ba82 BC |
3801 | } |
3802 | ||
dd67fa50 BC |
3803 | error_setg(errp, "Cannot find device=%s nor node_name=%s", |
3804 | device ? device : "", | |
3805 | node_name ? node_name : ""); | |
3806 | return NULL; | |
12d3ba82 BC |
3807 | } |
3808 | ||
5a6684d2 JC |
3809 | /* If 'base' is in the same chain as 'top', return true. Otherwise, |
3810 | * return false. If either argument is NULL, return false. */ | |
3811 | bool bdrv_chain_contains(BlockDriverState *top, BlockDriverState *base) | |
3812 | { | |
3813 | while (top && top != base) { | |
3814 | top = top->backing_hd; | |
3815 | } | |
3816 | ||
3817 | return top != NULL; | |
3818 | } | |
3819 | ||
2f399b0a MA |
3820 | BlockDriverState *bdrv_next(BlockDriverState *bs) |
3821 | { | |
3822 | if (!bs) { | |
3823 | return QTAILQ_FIRST(&bdrv_states); | |
3824 | } | |
dc364f4c | 3825 | return QTAILQ_NEXT(bs, device_list); |
2f399b0a MA |
3826 | } |
3827 | ||
51de9760 | 3828 | void bdrv_iterate(void (*it)(void *opaque, BlockDriverState *bs), void *opaque) |
81d0912d FB |
3829 | { |
3830 | BlockDriverState *bs; | |
3831 | ||
dc364f4c | 3832 | QTAILQ_FOREACH(bs, &bdrv_states, device_list) { |
51de9760 | 3833 | it(opaque, bs); |
81d0912d FB |
3834 | } |
3835 | } | |
3836 | ||
ea2384d3 FB |
3837 | const char *bdrv_get_device_name(BlockDriverState *bs) |
3838 | { | |
3839 | return bs->device_name; | |
3840 | } | |
3841 | ||
c8433287 MA |
3842 | int bdrv_get_flags(BlockDriverState *bs) |
3843 | { | |
3844 | return bs->open_flags; | |
3845 | } | |
3846 | ||
f0f0fdfe | 3847 | int bdrv_flush_all(void) |
c6ca28d6 AL |
3848 | { |
3849 | BlockDriverState *bs; | |
f0f0fdfe | 3850 | int result = 0; |
c6ca28d6 | 3851 | |
dc364f4c | 3852 | QTAILQ_FOREACH(bs, &bdrv_states, device_list) { |
ed78cda3 SH |
3853 | AioContext *aio_context = bdrv_get_aio_context(bs); |
3854 | int ret; | |
3855 | ||
3856 | aio_context_acquire(aio_context); | |
3857 | ret = bdrv_flush(bs); | |
f0f0fdfe KW |
3858 | if (ret < 0 && !result) { |
3859 | result = ret; | |
3860 | } | |
ed78cda3 | 3861 | aio_context_release(aio_context); |
1b7bdbc1 | 3862 | } |
f0f0fdfe KW |
3863 | |
3864 | return result; | |
c6ca28d6 AL |
3865 | } |
3866 | ||
3ac21627 PL |
3867 | int bdrv_has_zero_init_1(BlockDriverState *bs) |
3868 | { | |
3869 | return 1; | |
3870 | } | |
3871 | ||
f2feebbd KW |
3872 | int bdrv_has_zero_init(BlockDriverState *bs) |
3873 | { | |
3874 | assert(bs->drv); | |
3875 | ||
11212d8f PB |
3876 | /* If BS is a copy on write image, it is initialized to |
3877 | the contents of the base image, which may not be zeroes. */ | |
3878 | if (bs->backing_hd) { | |
3879 | return 0; | |
3880 | } | |
336c1c12 KW |
3881 | if (bs->drv->bdrv_has_zero_init) { |
3882 | return bs->drv->bdrv_has_zero_init(bs); | |
f2feebbd KW |
3883 | } |
3884 | ||
3ac21627 PL |
3885 | /* safe default */ |
3886 | return 0; | |
f2feebbd KW |
3887 | } |
3888 | ||
4ce78691 PL |
3889 | bool bdrv_unallocated_blocks_are_zero(BlockDriverState *bs) |
3890 | { | |
3891 | BlockDriverInfo bdi; | |
3892 | ||
3893 | if (bs->backing_hd) { | |
3894 | return false; | |
3895 | } | |
3896 | ||
3897 | if (bdrv_get_info(bs, &bdi) == 0) { | |
3898 | return bdi.unallocated_blocks_are_zero; | |
3899 | } | |
3900 | ||
3901 | return false; | |
3902 | } | |
3903 | ||
3904 | bool bdrv_can_write_zeroes_with_unmap(BlockDriverState *bs) | |
3905 | { | |
3906 | BlockDriverInfo bdi; | |
3907 | ||
3908 | if (bs->backing_hd || !(bs->open_flags & BDRV_O_UNMAP)) { | |
3909 | return false; | |
3910 | } | |
3911 | ||
3912 | if (bdrv_get_info(bs, &bdi) == 0) { | |
3913 | return bdi.can_write_zeroes_with_unmap; | |
3914 | } | |
3915 | ||
3916 | return false; | |
3917 | } | |
3918 | ||
b6b8a333 | 3919 | typedef struct BdrvCoGetBlockStatusData { |
376ae3f1 | 3920 | BlockDriverState *bs; |
b35b2bba | 3921 | BlockDriverState *base; |
376ae3f1 SH |
3922 | int64_t sector_num; |
3923 | int nb_sectors; | |
3924 | int *pnum; | |
b6b8a333 | 3925 | int64_t ret; |
376ae3f1 | 3926 | bool done; |
b6b8a333 | 3927 | } BdrvCoGetBlockStatusData; |
376ae3f1 | 3928 | |
f58c7b35 TS |
3929 | /* |
3930 | * Returns true iff the specified sector is present in the disk image. Drivers | |
3931 | * not implementing the functionality are assumed to not support backing files, | |
3932 | * hence all their sectors are reported as allocated. | |
3933 | * | |
bd9533e3 SH |
3934 | * If 'sector_num' is beyond the end of the disk image the return value is 0 |
3935 | * and 'pnum' is set to 0. | |
3936 | * | |
f58c7b35 TS |
3937 | * 'pnum' is set to the number of sectors (including and immediately following |
3938 | * the specified sector) that are known to be in the same | |
3939 | * allocated/unallocated state. | |
3940 | * | |
bd9533e3 SH |
3941 | * 'nb_sectors' is the max value 'pnum' should be set to. If nb_sectors goes |
3942 | * beyond the end of the disk image it will be clamped. | |
f58c7b35 | 3943 | */ |
b6b8a333 PB |
3944 | static int64_t coroutine_fn bdrv_co_get_block_status(BlockDriverState *bs, |
3945 | int64_t sector_num, | |
3946 | int nb_sectors, int *pnum) | |
f58c7b35 | 3947 | { |
617ccb46 | 3948 | int64_t length; |
bd9533e3 | 3949 | int64_t n; |
5daa74a6 | 3950 | int64_t ret, ret2; |
bd9533e3 | 3951 | |
617ccb46 PB |
3952 | length = bdrv_getlength(bs); |
3953 | if (length < 0) { | |
3954 | return length; | |
3955 | } | |
3956 | ||
3957 | if (sector_num >= (length >> BDRV_SECTOR_BITS)) { | |
bd9533e3 SH |
3958 | *pnum = 0; |
3959 | return 0; | |
3960 | } | |
3961 | ||
3962 | n = bs->total_sectors - sector_num; | |
3963 | if (n < nb_sectors) { | |
3964 | nb_sectors = n; | |
3965 | } | |
3966 | ||
b6b8a333 | 3967 | if (!bs->drv->bdrv_co_get_block_status) { |
bd9533e3 | 3968 | *pnum = nb_sectors; |
e88ae226 | 3969 | ret = BDRV_BLOCK_DATA | BDRV_BLOCK_ALLOCATED; |
918e92d7 PB |
3970 | if (bs->drv->protocol_name) { |
3971 | ret |= BDRV_BLOCK_OFFSET_VALID | (sector_num * BDRV_SECTOR_SIZE); | |
3972 | } | |
3973 | return ret; | |
f58c7b35 | 3974 | } |
6aebab14 | 3975 | |
415b5b01 PB |
3976 | ret = bs->drv->bdrv_co_get_block_status(bs, sector_num, nb_sectors, pnum); |
3977 | if (ret < 0) { | |
3e0a233d | 3978 | *pnum = 0; |
415b5b01 PB |
3979 | return ret; |
3980 | } | |
3981 | ||
92bc50a5 PL |
3982 | if (ret & BDRV_BLOCK_RAW) { |
3983 | assert(ret & BDRV_BLOCK_OFFSET_VALID); | |
3984 | return bdrv_get_block_status(bs->file, ret >> BDRV_SECTOR_BITS, | |
3985 | *pnum, pnum); | |
3986 | } | |
3987 | ||
e88ae226 KW |
3988 | if (ret & (BDRV_BLOCK_DATA | BDRV_BLOCK_ZERO)) { |
3989 | ret |= BDRV_BLOCK_ALLOCATED; | |
3990 | } | |
3991 | ||
c3d86884 PL |
3992 | if (!(ret & BDRV_BLOCK_DATA) && !(ret & BDRV_BLOCK_ZERO)) { |
3993 | if (bdrv_unallocated_blocks_are_zero(bs)) { | |
f0ad5712 | 3994 | ret |= BDRV_BLOCK_ZERO; |
1f9db224 | 3995 | } else if (bs->backing_hd) { |
f0ad5712 PB |
3996 | BlockDriverState *bs2 = bs->backing_hd; |
3997 | int64_t length2 = bdrv_getlength(bs2); | |
3998 | if (length2 >= 0 && sector_num >= (length2 >> BDRV_SECTOR_BITS)) { | |
3999 | ret |= BDRV_BLOCK_ZERO; | |
4000 | } | |
4001 | } | |
415b5b01 | 4002 | } |
5daa74a6 PB |
4003 | |
4004 | if (bs->file && | |
4005 | (ret & BDRV_BLOCK_DATA) && !(ret & BDRV_BLOCK_ZERO) && | |
4006 | (ret & BDRV_BLOCK_OFFSET_VALID)) { | |
4007 | ret2 = bdrv_co_get_block_status(bs->file, ret >> BDRV_SECTOR_BITS, | |
4008 | *pnum, pnum); | |
4009 | if (ret2 >= 0) { | |
4010 | /* Ignore errors. This is just providing extra information, it | |
4011 | * is useful but not necessary. | |
4012 | */ | |
4013 | ret |= (ret2 & BDRV_BLOCK_ZERO); | |
4014 | } | |
4015 | } | |
4016 | ||
415b5b01 | 4017 | return ret; |
060f51c9 SH |
4018 | } |
4019 | ||
b6b8a333 PB |
4020 | /* Coroutine wrapper for bdrv_get_block_status() */ |
4021 | static void coroutine_fn bdrv_get_block_status_co_entry(void *opaque) | |
060f51c9 | 4022 | { |
b6b8a333 | 4023 | BdrvCoGetBlockStatusData *data = opaque; |
060f51c9 SH |
4024 | BlockDriverState *bs = data->bs; |
4025 | ||
b6b8a333 PB |
4026 | data->ret = bdrv_co_get_block_status(bs, data->sector_num, data->nb_sectors, |
4027 | data->pnum); | |
060f51c9 SH |
4028 | data->done = true; |
4029 | } | |
4030 | ||
4031 | /* | |
b6b8a333 | 4032 | * Synchronous wrapper around bdrv_co_get_block_status(). |
060f51c9 | 4033 | * |
b6b8a333 | 4034 | * See bdrv_co_get_block_status() for details. |
060f51c9 | 4035 | */ |
b6b8a333 PB |
4036 | int64_t bdrv_get_block_status(BlockDriverState *bs, int64_t sector_num, |
4037 | int nb_sectors, int *pnum) | |
060f51c9 | 4038 | { |
6aebab14 | 4039 | Coroutine *co; |
b6b8a333 | 4040 | BdrvCoGetBlockStatusData data = { |
6aebab14 SH |
4041 | .bs = bs, |
4042 | .sector_num = sector_num, | |
4043 | .nb_sectors = nb_sectors, | |
4044 | .pnum = pnum, | |
4045 | .done = false, | |
4046 | }; | |
4047 | ||
bdad13b9 PB |
4048 | if (qemu_in_coroutine()) { |
4049 | /* Fast-path if already in coroutine context */ | |
b6b8a333 | 4050 | bdrv_get_block_status_co_entry(&data); |
bdad13b9 | 4051 | } else { |
2572b37a SH |
4052 | AioContext *aio_context = bdrv_get_aio_context(bs); |
4053 | ||
b6b8a333 | 4054 | co = qemu_coroutine_create(bdrv_get_block_status_co_entry); |
bdad13b9 PB |
4055 | qemu_coroutine_enter(co, &data); |
4056 | while (!data.done) { | |
2572b37a | 4057 | aio_poll(aio_context, true); |
bdad13b9 | 4058 | } |
6aebab14 SH |
4059 | } |
4060 | return data.ret; | |
f58c7b35 TS |
4061 | } |
4062 | ||
b6b8a333 PB |
4063 | int coroutine_fn bdrv_is_allocated(BlockDriverState *bs, int64_t sector_num, |
4064 | int nb_sectors, int *pnum) | |
4065 | { | |
4333bb71 PB |
4066 | int64_t ret = bdrv_get_block_status(bs, sector_num, nb_sectors, pnum); |
4067 | if (ret < 0) { | |
4068 | return ret; | |
4069 | } | |
01fb2705 | 4070 | return !!(ret & BDRV_BLOCK_ALLOCATED); |
b6b8a333 PB |
4071 | } |
4072 | ||
188a7bbf PB |
4073 | /* |
4074 | * Given an image chain: ... -> [BASE] -> [INTER1] -> [INTER2] -> [TOP] | |
4075 | * | |
4076 | * Return true if the given sector is allocated in any image between | |
4077 | * BASE and TOP (inclusive). BASE can be NULL to check if the given | |
4078 | * sector is allocated in any image of the chain. Return false otherwise. | |
4079 | * | |
4080 | * 'pnum' is set to the number of sectors (including and immediately following | |
4081 | * the specified sector) that are known to be in the same | |
4082 | * allocated/unallocated state. | |
4083 | * | |
4084 | */ | |
4f578637 PB |
4085 | int bdrv_is_allocated_above(BlockDriverState *top, |
4086 | BlockDriverState *base, | |
4087 | int64_t sector_num, | |
4088 | int nb_sectors, int *pnum) | |
188a7bbf PB |
4089 | { |
4090 | BlockDriverState *intermediate; | |
4091 | int ret, n = nb_sectors; | |
4092 | ||
4093 | intermediate = top; | |
4094 | while (intermediate && intermediate != base) { | |
4095 | int pnum_inter; | |
bdad13b9 PB |
4096 | ret = bdrv_is_allocated(intermediate, sector_num, nb_sectors, |
4097 | &pnum_inter); | |
188a7bbf PB |
4098 | if (ret < 0) { |
4099 | return ret; | |
4100 | } else if (ret) { | |
4101 | *pnum = pnum_inter; | |
4102 | return 1; | |
4103 | } | |
4104 | ||
4105 | /* | |
4106 | * [sector_num, nb_sectors] is unallocated on top but intermediate | |
4107 | * might have | |
4108 | * | |
4109 | * [sector_num+x, nr_sectors] allocated. | |
4110 | */ | |
63ba17d3 VI |
4111 | if (n > pnum_inter && |
4112 | (intermediate == top || | |
4113 | sector_num + pnum_inter < intermediate->total_sectors)) { | |
188a7bbf PB |
4114 | n = pnum_inter; |
4115 | } | |
4116 | ||
4117 | intermediate = intermediate->backing_hd; | |
4118 | } | |
4119 | ||
4120 | *pnum = n; | |
4121 | return 0; | |
4122 | } | |
4123 | ||
045df330 AL |
4124 | const char *bdrv_get_encrypted_filename(BlockDriverState *bs) |
4125 | { | |
4126 | if (bs->backing_hd && bs->backing_hd->encrypted) | |
4127 | return bs->backing_file; | |
4128 | else if (bs->encrypted) | |
4129 | return bs->filename; | |
4130 | else | |
4131 | return NULL; | |
4132 | } | |
4133 | ||
5fafdf24 | 4134 | void bdrv_get_backing_filename(BlockDriverState *bs, |
83f64091 FB |
4135 | char *filename, int filename_size) |
4136 | { | |
3574c608 | 4137 | pstrcpy(filename, filename_size, bs->backing_file); |
83f64091 FB |
4138 | } |
4139 | ||
5fafdf24 | 4140 | int bdrv_write_compressed(BlockDriverState *bs, int64_t sector_num, |
faea38e7 FB |
4141 | const uint8_t *buf, int nb_sectors) |
4142 | { | |
4143 | BlockDriver *drv = bs->drv; | |
4144 | if (!drv) | |
19cb3738 | 4145 | return -ENOMEDIUM; |
faea38e7 FB |
4146 | if (!drv->bdrv_write_compressed) |
4147 | return -ENOTSUP; | |
fbb7b4e0 KW |
4148 | if (bdrv_check_request(bs, sector_num, nb_sectors)) |
4149 | return -EIO; | |
a55eb92c | 4150 | |
e4654d2d | 4151 | assert(QLIST_EMPTY(&bs->dirty_bitmaps)); |
a55eb92c | 4152 | |
faea38e7 FB |
4153 | return drv->bdrv_write_compressed(bs, sector_num, buf, nb_sectors); |
4154 | } | |
3b46e624 | 4155 | |
faea38e7 FB |
4156 | int bdrv_get_info(BlockDriverState *bs, BlockDriverInfo *bdi) |
4157 | { | |
4158 | BlockDriver *drv = bs->drv; | |
4159 | if (!drv) | |
19cb3738 | 4160 | return -ENOMEDIUM; |
faea38e7 FB |
4161 | if (!drv->bdrv_get_info) |
4162 | return -ENOTSUP; | |
4163 | memset(bdi, 0, sizeof(*bdi)); | |
4164 | return drv->bdrv_get_info(bs, bdi); | |
4165 | } | |
4166 | ||
eae041fe HR |
4167 | ImageInfoSpecific *bdrv_get_specific_info(BlockDriverState *bs) |
4168 | { | |
4169 | BlockDriver *drv = bs->drv; | |
4170 | if (drv && drv->bdrv_get_specific_info) { | |
4171 | return drv->bdrv_get_specific_info(bs); | |
4172 | } | |
4173 | return NULL; | |
4174 | } | |
4175 | ||
45566e9c CH |
4176 | int bdrv_save_vmstate(BlockDriverState *bs, const uint8_t *buf, |
4177 | int64_t pos, int size) | |
cf8074b3 KW |
4178 | { |
4179 | QEMUIOVector qiov; | |
4180 | struct iovec iov = { | |
4181 | .iov_base = (void *) buf, | |
4182 | .iov_len = size, | |
4183 | }; | |
4184 | ||
4185 | qemu_iovec_init_external(&qiov, &iov, 1); | |
4186 | return bdrv_writev_vmstate(bs, &qiov, pos); | |
4187 | } | |
4188 | ||
4189 | int bdrv_writev_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos) | |
178e08a5 AL |
4190 | { |
4191 | BlockDriver *drv = bs->drv; | |
cf8074b3 KW |
4192 | |
4193 | if (!drv) { | |
178e08a5 | 4194 | return -ENOMEDIUM; |
cf8074b3 KW |
4195 | } else if (drv->bdrv_save_vmstate) { |
4196 | return drv->bdrv_save_vmstate(bs, qiov, pos); | |
4197 | } else if (bs->file) { | |
4198 | return bdrv_writev_vmstate(bs->file, qiov, pos); | |
4199 | } | |
4200 | ||
7cdb1f6d | 4201 | return -ENOTSUP; |
178e08a5 AL |
4202 | } |
4203 | ||
45566e9c CH |
4204 | int bdrv_load_vmstate(BlockDriverState *bs, uint8_t *buf, |
4205 | int64_t pos, int size) | |
178e08a5 AL |
4206 | { |
4207 | BlockDriver *drv = bs->drv; | |
4208 | if (!drv) | |
4209 | return -ENOMEDIUM; | |
7cdb1f6d MK |
4210 | if (drv->bdrv_load_vmstate) |
4211 | return drv->bdrv_load_vmstate(bs, buf, pos, size); | |
4212 | if (bs->file) | |
4213 | return bdrv_load_vmstate(bs->file, buf, pos, size); | |
4214 | return -ENOTSUP; | |
178e08a5 AL |
4215 | } |
4216 | ||
8b9b0cc2 KW |
4217 | void bdrv_debug_event(BlockDriverState *bs, BlkDebugEvent event) |
4218 | { | |
bf736fe3 | 4219 | if (!bs || !bs->drv || !bs->drv->bdrv_debug_event) { |
8b9b0cc2 KW |
4220 | return; |
4221 | } | |
4222 | ||
bf736fe3 | 4223 | bs->drv->bdrv_debug_event(bs, event); |
41c695c7 KW |
4224 | } |
4225 | ||
4226 | int bdrv_debug_breakpoint(BlockDriverState *bs, const char *event, | |
4227 | const char *tag) | |
4228 | { | |
4229 | while (bs && bs->drv && !bs->drv->bdrv_debug_breakpoint) { | |
4230 | bs = bs->file; | |
4231 | } | |
4232 | ||
4233 | if (bs && bs->drv && bs->drv->bdrv_debug_breakpoint) { | |
4234 | return bs->drv->bdrv_debug_breakpoint(bs, event, tag); | |
4235 | } | |
4236 | ||
4237 | return -ENOTSUP; | |
4238 | } | |
4239 | ||
4cc70e93 FZ |
4240 | int bdrv_debug_remove_breakpoint(BlockDriverState *bs, const char *tag) |
4241 | { | |
4242 | while (bs && bs->drv && !bs->drv->bdrv_debug_remove_breakpoint) { | |
4243 | bs = bs->file; | |
4244 | } | |
4245 | ||
4246 | if (bs && bs->drv && bs->drv->bdrv_debug_remove_breakpoint) { | |
4247 | return bs->drv->bdrv_debug_remove_breakpoint(bs, tag); | |
4248 | } | |
4249 | ||
4250 | return -ENOTSUP; | |
4251 | } | |
4252 | ||
41c695c7 KW |
4253 | int bdrv_debug_resume(BlockDriverState *bs, const char *tag) |
4254 | { | |
938789ea | 4255 | while (bs && (!bs->drv || !bs->drv->bdrv_debug_resume)) { |
41c695c7 KW |
4256 | bs = bs->file; |
4257 | } | |
8b9b0cc2 | 4258 | |
41c695c7 KW |
4259 | if (bs && bs->drv && bs->drv->bdrv_debug_resume) { |
4260 | return bs->drv->bdrv_debug_resume(bs, tag); | |
4261 | } | |
4262 | ||
4263 | return -ENOTSUP; | |
4264 | } | |
4265 | ||
4266 | bool bdrv_debug_is_suspended(BlockDriverState *bs, const char *tag) | |
4267 | { | |
4268 | while (bs && bs->drv && !bs->drv->bdrv_debug_is_suspended) { | |
4269 | bs = bs->file; | |
4270 | } | |
4271 | ||
4272 | if (bs && bs->drv && bs->drv->bdrv_debug_is_suspended) { | |
4273 | return bs->drv->bdrv_debug_is_suspended(bs, tag); | |
4274 | } | |
4275 | ||
4276 | return false; | |
8b9b0cc2 KW |
4277 | } |
4278 | ||
199630b6 BS |
4279 | int bdrv_is_snapshot(BlockDriverState *bs) |
4280 | { | |
4281 | return !!(bs->open_flags & BDRV_O_SNAPSHOT); | |
4282 | } | |
4283 | ||
b1b1d783 JC |
4284 | /* backing_file can either be relative, or absolute, or a protocol. If it is |
4285 | * relative, it must be relative to the chain. So, passing in bs->filename | |
4286 | * from a BDS as backing_file should not be done, as that may be relative to | |
4287 | * the CWD rather than the chain. */ | |
e8a6bb9c MT |
4288 | BlockDriverState *bdrv_find_backing_image(BlockDriverState *bs, |
4289 | const char *backing_file) | |
4290 | { | |
b1b1d783 JC |
4291 | char *filename_full = NULL; |
4292 | char *backing_file_full = NULL; | |
4293 | char *filename_tmp = NULL; | |
4294 | int is_protocol = 0; | |
4295 | BlockDriverState *curr_bs = NULL; | |
4296 | BlockDriverState *retval = NULL; | |
4297 | ||
4298 | if (!bs || !bs->drv || !backing_file) { | |
e8a6bb9c MT |
4299 | return NULL; |
4300 | } | |
4301 | ||
b1b1d783 JC |
4302 | filename_full = g_malloc(PATH_MAX); |
4303 | backing_file_full = g_malloc(PATH_MAX); | |
4304 | filename_tmp = g_malloc(PATH_MAX); | |
4305 | ||
4306 | is_protocol = path_has_protocol(backing_file); | |
4307 | ||
4308 | for (curr_bs = bs; curr_bs->backing_hd; curr_bs = curr_bs->backing_hd) { | |
4309 | ||
4310 | /* If either of the filename paths is actually a protocol, then | |
4311 | * compare unmodified paths; otherwise make paths relative */ | |
4312 | if (is_protocol || path_has_protocol(curr_bs->backing_file)) { | |
4313 | if (strcmp(backing_file, curr_bs->backing_file) == 0) { | |
4314 | retval = curr_bs->backing_hd; | |
4315 | break; | |
4316 | } | |
e8a6bb9c | 4317 | } else { |
b1b1d783 JC |
4318 | /* If not an absolute filename path, make it relative to the current |
4319 | * image's filename path */ | |
4320 | path_combine(filename_tmp, PATH_MAX, curr_bs->filename, | |
4321 | backing_file); | |
4322 | ||
4323 | /* We are going to compare absolute pathnames */ | |
4324 | if (!realpath(filename_tmp, filename_full)) { | |
4325 | continue; | |
4326 | } | |
4327 | ||
4328 | /* We need to make sure the backing filename we are comparing against | |
4329 | * is relative to the current image filename (or absolute) */ | |
4330 | path_combine(filename_tmp, PATH_MAX, curr_bs->filename, | |
4331 | curr_bs->backing_file); | |
4332 | ||
4333 | if (!realpath(filename_tmp, backing_file_full)) { | |
4334 | continue; | |
4335 | } | |
4336 | ||
4337 | if (strcmp(backing_file_full, filename_full) == 0) { | |
4338 | retval = curr_bs->backing_hd; | |
4339 | break; | |
4340 | } | |
e8a6bb9c MT |
4341 | } |
4342 | } | |
4343 | ||
b1b1d783 JC |
4344 | g_free(filename_full); |
4345 | g_free(backing_file_full); | |
4346 | g_free(filename_tmp); | |
4347 | return retval; | |
e8a6bb9c MT |
4348 | } |
4349 | ||
f198fd1c BC |
4350 | int bdrv_get_backing_file_depth(BlockDriverState *bs) |
4351 | { | |
4352 | if (!bs->drv) { | |
4353 | return 0; | |
4354 | } | |
4355 | ||
4356 | if (!bs->backing_hd) { | |
4357 | return 0; | |
4358 | } | |
4359 | ||
4360 | return 1 + bdrv_get_backing_file_depth(bs->backing_hd); | |
4361 | } | |
4362 | ||
ea2384d3 | 4363 | /**************************************************************/ |
83f64091 | 4364 | /* async I/Os */ |
ea2384d3 | 4365 | |
3b69e4b9 | 4366 | BlockDriverAIOCB *bdrv_aio_readv(BlockDriverState *bs, int64_t sector_num, |
f141eafe | 4367 | QEMUIOVector *qiov, int nb_sectors, |
3b69e4b9 | 4368 | BlockDriverCompletionFunc *cb, void *opaque) |
83f64091 | 4369 | { |
bbf0a440 SH |
4370 | trace_bdrv_aio_readv(bs, sector_num, nb_sectors, opaque); |
4371 | ||
d20d9b7c | 4372 | return bdrv_co_aio_rw_vector(bs, sector_num, qiov, nb_sectors, 0, |
8c5873d6 | 4373 | cb, opaque, false); |
ea2384d3 FB |
4374 | } |
4375 | ||
f141eafe AL |
4376 | BlockDriverAIOCB *bdrv_aio_writev(BlockDriverState *bs, int64_t sector_num, |
4377 | QEMUIOVector *qiov, int nb_sectors, | |
4378 | BlockDriverCompletionFunc *cb, void *opaque) | |
ea2384d3 | 4379 | { |
bbf0a440 SH |
4380 | trace_bdrv_aio_writev(bs, sector_num, nb_sectors, opaque); |
4381 | ||
d20d9b7c | 4382 | return bdrv_co_aio_rw_vector(bs, sector_num, qiov, nb_sectors, 0, |
8c5873d6 | 4383 | cb, opaque, true); |
83f64091 FB |
4384 | } |
4385 | ||
d5ef94d4 PB |
4386 | BlockDriverAIOCB *bdrv_aio_write_zeroes(BlockDriverState *bs, |
4387 | int64_t sector_num, int nb_sectors, BdrvRequestFlags flags, | |
4388 | BlockDriverCompletionFunc *cb, void *opaque) | |
4389 | { | |
4390 | trace_bdrv_aio_write_zeroes(bs, sector_num, nb_sectors, flags, opaque); | |
4391 | ||
4392 | return bdrv_co_aio_rw_vector(bs, sector_num, NULL, nb_sectors, | |
4393 | BDRV_REQ_ZERO_WRITE | flags, | |
4394 | cb, opaque, true); | |
4395 | } | |
4396 | ||
40b4f539 KW |
4397 | |
4398 | typedef struct MultiwriteCB { | |
4399 | int error; | |
4400 | int num_requests; | |
4401 | int num_callbacks; | |
4402 | struct { | |
4403 | BlockDriverCompletionFunc *cb; | |
4404 | void *opaque; | |
4405 | QEMUIOVector *free_qiov; | |
40b4f539 KW |
4406 | } callbacks[]; |
4407 | } MultiwriteCB; | |
4408 | ||
4409 | static void multiwrite_user_cb(MultiwriteCB *mcb) | |
4410 | { | |
4411 | int i; | |
4412 | ||
4413 | for (i = 0; i < mcb->num_callbacks; i++) { | |
4414 | mcb->callbacks[i].cb(mcb->callbacks[i].opaque, mcb->error); | |
1e1ea48d SH |
4415 | if (mcb->callbacks[i].free_qiov) { |
4416 | qemu_iovec_destroy(mcb->callbacks[i].free_qiov); | |
4417 | } | |
7267c094 | 4418 | g_free(mcb->callbacks[i].free_qiov); |
40b4f539 KW |
4419 | } |
4420 | } | |
4421 | ||
4422 | static void multiwrite_cb(void *opaque, int ret) | |
4423 | { | |
4424 | MultiwriteCB *mcb = opaque; | |
4425 | ||
6d519a5f SH |
4426 | trace_multiwrite_cb(mcb, ret); |
4427 | ||
cb6d3ca0 | 4428 | if (ret < 0 && !mcb->error) { |
40b4f539 | 4429 | mcb->error = ret; |
40b4f539 KW |
4430 | } |
4431 | ||
4432 | mcb->num_requests--; | |
4433 | if (mcb->num_requests == 0) { | |
de189a1b | 4434 | multiwrite_user_cb(mcb); |
7267c094 | 4435 | g_free(mcb); |
40b4f539 KW |
4436 | } |
4437 | } | |
4438 | ||
4439 | static int multiwrite_req_compare(const void *a, const void *b) | |
4440 | { | |
77be4366 CH |
4441 | const BlockRequest *req1 = a, *req2 = b; |
4442 | ||
4443 | /* | |
4444 | * Note that we can't simply subtract req2->sector from req1->sector | |
4445 | * here as that could overflow the return value. | |
4446 | */ | |
4447 | if (req1->sector > req2->sector) { | |
4448 | return 1; | |
4449 | } else if (req1->sector < req2->sector) { | |
4450 | return -1; | |
4451 | } else { | |
4452 | return 0; | |
4453 | } | |
40b4f539 KW |
4454 | } |
4455 | ||
4456 | /* | |
4457 | * Takes a bunch of requests and tries to merge them. Returns the number of | |
4458 | * requests that remain after merging. | |
4459 | */ | |
4460 | static int multiwrite_merge(BlockDriverState *bs, BlockRequest *reqs, | |
4461 | int num_reqs, MultiwriteCB *mcb) | |
4462 | { | |
4463 | int i, outidx; | |
4464 | ||
4465 | // Sort requests by start sector | |
4466 | qsort(reqs, num_reqs, sizeof(*reqs), &multiwrite_req_compare); | |
4467 | ||
4468 | // Check if adjacent requests touch the same clusters. If so, combine them, | |
4469 | // filling up gaps with zero sectors. | |
4470 | outidx = 0; | |
4471 | for (i = 1; i < num_reqs; i++) { | |
4472 | int merge = 0; | |
4473 | int64_t oldreq_last = reqs[outidx].sector + reqs[outidx].nb_sectors; | |
4474 | ||
b6a127a1 | 4475 | // Handle exactly sequential writes and overlapping writes. |
40b4f539 KW |
4476 | if (reqs[i].sector <= oldreq_last) { |
4477 | merge = 1; | |
4478 | } | |
4479 | ||
e2a305fb CH |
4480 | if (reqs[outidx].qiov->niov + reqs[i].qiov->niov + 1 > IOV_MAX) { |
4481 | merge = 0; | |
4482 | } | |
4483 | ||
40b4f539 KW |
4484 | if (merge) { |
4485 | size_t size; | |
7267c094 | 4486 | QEMUIOVector *qiov = g_malloc0(sizeof(*qiov)); |
40b4f539 KW |
4487 | qemu_iovec_init(qiov, |
4488 | reqs[outidx].qiov->niov + reqs[i].qiov->niov + 1); | |
4489 | ||
4490 | // Add the first request to the merged one. If the requests are | |
4491 | // overlapping, drop the last sectors of the first request. | |
4492 | size = (reqs[i].sector - reqs[outidx].sector) << 9; | |
1b093c48 | 4493 | qemu_iovec_concat(qiov, reqs[outidx].qiov, 0, size); |
40b4f539 | 4494 | |
b6a127a1 PB |
4495 | // We should need to add any zeros between the two requests |
4496 | assert (reqs[i].sector <= oldreq_last); | |
40b4f539 KW |
4497 | |
4498 | // Add the second request | |
1b093c48 | 4499 | qemu_iovec_concat(qiov, reqs[i].qiov, 0, reqs[i].qiov->size); |
40b4f539 | 4500 | |
cbf1dff2 | 4501 | reqs[outidx].nb_sectors = qiov->size >> 9; |
40b4f539 KW |
4502 | reqs[outidx].qiov = qiov; |
4503 | ||
4504 | mcb->callbacks[i].free_qiov = reqs[outidx].qiov; | |
4505 | } else { | |
4506 | outidx++; | |
4507 | reqs[outidx].sector = reqs[i].sector; | |
4508 | reqs[outidx].nb_sectors = reqs[i].nb_sectors; | |
4509 | reqs[outidx].qiov = reqs[i].qiov; | |
4510 | } | |
4511 | } | |
4512 | ||
4513 | return outidx + 1; | |
4514 | } | |
4515 | ||
4516 | /* | |
4517 | * Submit multiple AIO write requests at once. | |
4518 | * | |
4519 | * On success, the function returns 0 and all requests in the reqs array have | |
4520 | * been submitted. In error case this function returns -1, and any of the | |
4521 | * requests may or may not be submitted yet. In particular, this means that the | |
4522 | * callback will be called for some of the requests, for others it won't. The | |
4523 | * caller must check the error field of the BlockRequest to wait for the right | |
4524 | * callbacks (if error != 0, no callback will be called). | |
4525 | * | |
4526 | * The implementation may modify the contents of the reqs array, e.g. to merge | |
4527 | * requests. However, the fields opaque and error are left unmodified as they | |
4528 | * are used to signal failure for a single request to the caller. | |
4529 | */ | |
4530 | int bdrv_aio_multiwrite(BlockDriverState *bs, BlockRequest *reqs, int num_reqs) | |
4531 | { | |
40b4f539 KW |
4532 | MultiwriteCB *mcb; |
4533 | int i; | |
4534 | ||
301db7c2 RH |
4535 | /* don't submit writes if we don't have a medium */ |
4536 | if (bs->drv == NULL) { | |
4537 | for (i = 0; i < num_reqs; i++) { | |
4538 | reqs[i].error = -ENOMEDIUM; | |
4539 | } | |
4540 | return -1; | |
4541 | } | |
4542 | ||
40b4f539 KW |
4543 | if (num_reqs == 0) { |
4544 | return 0; | |
4545 | } | |
4546 | ||
4547 | // Create MultiwriteCB structure | |
7267c094 | 4548 | mcb = g_malloc0(sizeof(*mcb) + num_reqs * sizeof(*mcb->callbacks)); |
40b4f539 KW |
4549 | mcb->num_requests = 0; |
4550 | mcb->num_callbacks = num_reqs; | |
4551 | ||
4552 | for (i = 0; i < num_reqs; i++) { | |
4553 | mcb->callbacks[i].cb = reqs[i].cb; | |
4554 | mcb->callbacks[i].opaque = reqs[i].opaque; | |
4555 | } | |
4556 | ||
4557 | // Check for mergable requests | |
4558 | num_reqs = multiwrite_merge(bs, reqs, num_reqs, mcb); | |
4559 | ||
6d519a5f SH |
4560 | trace_bdrv_aio_multiwrite(mcb, mcb->num_callbacks, num_reqs); |
4561 | ||
df9309fb PB |
4562 | /* Run the aio requests. */ |
4563 | mcb->num_requests = num_reqs; | |
40b4f539 | 4564 | for (i = 0; i < num_reqs; i++) { |
d20d9b7c PB |
4565 | bdrv_co_aio_rw_vector(bs, reqs[i].sector, reqs[i].qiov, |
4566 | reqs[i].nb_sectors, reqs[i].flags, | |
4567 | multiwrite_cb, mcb, | |
4568 | true); | |
40b4f539 KW |
4569 | } |
4570 | ||
4571 | return 0; | |
40b4f539 KW |
4572 | } |
4573 | ||
83f64091 | 4574 | void bdrv_aio_cancel(BlockDriverAIOCB *acb) |
83f64091 | 4575 | { |
d7331bed | 4576 | acb->aiocb_info->cancel(acb); |
83f64091 FB |
4577 | } |
4578 | ||
4579 | /**************************************************************/ | |
4580 | /* async block device emulation */ | |
4581 | ||
c16b5a2c CH |
4582 | typedef struct BlockDriverAIOCBSync { |
4583 | BlockDriverAIOCB common; | |
4584 | QEMUBH *bh; | |
4585 | int ret; | |
4586 | /* vector translation state */ | |
4587 | QEMUIOVector *qiov; | |
4588 | uint8_t *bounce; | |
4589 | int is_write; | |
4590 | } BlockDriverAIOCBSync; | |
4591 | ||
4592 | static void bdrv_aio_cancel_em(BlockDriverAIOCB *blockacb) | |
4593 | { | |
b666d239 KW |
4594 | BlockDriverAIOCBSync *acb = |
4595 | container_of(blockacb, BlockDriverAIOCBSync, common); | |
6a7ad299 | 4596 | qemu_bh_delete(acb->bh); |
36afc451 | 4597 | acb->bh = NULL; |
c16b5a2c CH |
4598 | qemu_aio_release(acb); |
4599 | } | |
4600 | ||
d7331bed | 4601 | static const AIOCBInfo bdrv_em_aiocb_info = { |
c16b5a2c CH |
4602 | .aiocb_size = sizeof(BlockDriverAIOCBSync), |
4603 | .cancel = bdrv_aio_cancel_em, | |
4604 | }; | |
4605 | ||
ce1a14dc | 4606 | static void bdrv_aio_bh_cb(void *opaque) |
83f64091 | 4607 | { |
ce1a14dc | 4608 | BlockDriverAIOCBSync *acb = opaque; |
f141eafe | 4609 | |
f141eafe | 4610 | if (!acb->is_write) |
03396148 | 4611 | qemu_iovec_from_buf(acb->qiov, 0, acb->bounce, acb->qiov->size); |
ceb42de8 | 4612 | qemu_vfree(acb->bounce); |
ce1a14dc | 4613 | acb->common.cb(acb->common.opaque, acb->ret); |
6a7ad299 | 4614 | qemu_bh_delete(acb->bh); |
36afc451 | 4615 | acb->bh = NULL; |
ce1a14dc | 4616 | qemu_aio_release(acb); |
83f64091 | 4617 | } |
beac80cd | 4618 | |
f141eafe AL |
4619 | static BlockDriverAIOCB *bdrv_aio_rw_vector(BlockDriverState *bs, |
4620 | int64_t sector_num, | |
4621 | QEMUIOVector *qiov, | |
4622 | int nb_sectors, | |
4623 | BlockDriverCompletionFunc *cb, | |
4624 | void *opaque, | |
4625 | int is_write) | |
4626 | ||
83f64091 | 4627 | { |
ce1a14dc | 4628 | BlockDriverAIOCBSync *acb; |
ce1a14dc | 4629 | |
d7331bed | 4630 | acb = qemu_aio_get(&bdrv_em_aiocb_info, bs, cb, opaque); |
f141eafe AL |
4631 | acb->is_write = is_write; |
4632 | acb->qiov = qiov; | |
e268ca52 | 4633 | acb->bounce = qemu_blockalign(bs, qiov->size); |
2572b37a | 4634 | acb->bh = aio_bh_new(bdrv_get_aio_context(bs), bdrv_aio_bh_cb, acb); |
f141eafe AL |
4635 | |
4636 | if (is_write) { | |
d5e6b161 | 4637 | qemu_iovec_to_buf(acb->qiov, 0, acb->bounce, qiov->size); |
1ed20acf | 4638 | acb->ret = bs->drv->bdrv_write(bs, sector_num, acb->bounce, nb_sectors); |
f141eafe | 4639 | } else { |
1ed20acf | 4640 | acb->ret = bs->drv->bdrv_read(bs, sector_num, acb->bounce, nb_sectors); |
f141eafe AL |
4641 | } |
4642 | ||
ce1a14dc | 4643 | qemu_bh_schedule(acb->bh); |
f141eafe | 4644 | |
ce1a14dc | 4645 | return &acb->common; |
beac80cd FB |
4646 | } |
4647 | ||
f141eafe AL |
4648 | static BlockDriverAIOCB *bdrv_aio_readv_em(BlockDriverState *bs, |
4649 | int64_t sector_num, QEMUIOVector *qiov, int nb_sectors, | |
ce1a14dc | 4650 | BlockDriverCompletionFunc *cb, void *opaque) |
beac80cd | 4651 | { |
f141eafe AL |
4652 | return bdrv_aio_rw_vector(bs, sector_num, qiov, nb_sectors, cb, opaque, 0); |
4653 | } | |
83f64091 | 4654 | |
f141eafe AL |
4655 | static BlockDriverAIOCB *bdrv_aio_writev_em(BlockDriverState *bs, |
4656 | int64_t sector_num, QEMUIOVector *qiov, int nb_sectors, | |
4657 | BlockDriverCompletionFunc *cb, void *opaque) | |
4658 | { | |
4659 | return bdrv_aio_rw_vector(bs, sector_num, qiov, nb_sectors, cb, opaque, 1); | |
beac80cd | 4660 | } |
beac80cd | 4661 | |
68485420 KW |
4662 | |
4663 | typedef struct BlockDriverAIOCBCoroutine { | |
4664 | BlockDriverAIOCB common; | |
4665 | BlockRequest req; | |
4666 | bool is_write; | |
d318aea9 | 4667 | bool *done; |
68485420 KW |
4668 | QEMUBH* bh; |
4669 | } BlockDriverAIOCBCoroutine; | |
4670 | ||
4671 | static void bdrv_aio_co_cancel_em(BlockDriverAIOCB *blockacb) | |
4672 | { | |
2572b37a | 4673 | AioContext *aio_context = bdrv_get_aio_context(blockacb->bs); |
d318aea9 KW |
4674 | BlockDriverAIOCBCoroutine *acb = |
4675 | container_of(blockacb, BlockDriverAIOCBCoroutine, common); | |
4676 | bool done = false; | |
4677 | ||
4678 | acb->done = &done; | |
4679 | while (!done) { | |
2572b37a | 4680 | aio_poll(aio_context, true); |
d318aea9 | 4681 | } |
68485420 KW |
4682 | } |
4683 | ||
d7331bed | 4684 | static const AIOCBInfo bdrv_em_co_aiocb_info = { |
68485420 KW |
4685 | .aiocb_size = sizeof(BlockDriverAIOCBCoroutine), |
4686 | .cancel = bdrv_aio_co_cancel_em, | |
4687 | }; | |
4688 | ||
35246a68 | 4689 | static void bdrv_co_em_bh(void *opaque) |
68485420 KW |
4690 | { |
4691 | BlockDriverAIOCBCoroutine *acb = opaque; | |
4692 | ||
4693 | acb->common.cb(acb->common.opaque, acb->req.error); | |
d318aea9 KW |
4694 | |
4695 | if (acb->done) { | |
4696 | *acb->done = true; | |
4697 | } | |
4698 | ||
68485420 KW |
4699 | qemu_bh_delete(acb->bh); |
4700 | qemu_aio_release(acb); | |
4701 | } | |
4702 | ||
b2a61371 SH |
4703 | /* Invoke bdrv_co_do_readv/bdrv_co_do_writev */ |
4704 | static void coroutine_fn bdrv_co_do_rw(void *opaque) | |
4705 | { | |
4706 | BlockDriverAIOCBCoroutine *acb = opaque; | |
4707 | BlockDriverState *bs = acb->common.bs; | |
4708 | ||
4709 | if (!acb->is_write) { | |
4710 | acb->req.error = bdrv_co_do_readv(bs, acb->req.sector, | |
d20d9b7c | 4711 | acb->req.nb_sectors, acb->req.qiov, acb->req.flags); |
b2a61371 SH |
4712 | } else { |
4713 | acb->req.error = bdrv_co_do_writev(bs, acb->req.sector, | |
d20d9b7c | 4714 | acb->req.nb_sectors, acb->req.qiov, acb->req.flags); |
b2a61371 SH |
4715 | } |
4716 | ||
2572b37a | 4717 | acb->bh = aio_bh_new(bdrv_get_aio_context(bs), bdrv_co_em_bh, acb); |
b2a61371 SH |
4718 | qemu_bh_schedule(acb->bh); |
4719 | } | |
4720 | ||
68485420 KW |
4721 | static BlockDriverAIOCB *bdrv_co_aio_rw_vector(BlockDriverState *bs, |
4722 | int64_t sector_num, | |
4723 | QEMUIOVector *qiov, | |
4724 | int nb_sectors, | |
d20d9b7c | 4725 | BdrvRequestFlags flags, |
68485420 KW |
4726 | BlockDriverCompletionFunc *cb, |
4727 | void *opaque, | |
8c5873d6 | 4728 | bool is_write) |
68485420 KW |
4729 | { |
4730 | Coroutine *co; | |
4731 | BlockDriverAIOCBCoroutine *acb; | |
4732 | ||
d7331bed | 4733 | acb = qemu_aio_get(&bdrv_em_co_aiocb_info, bs, cb, opaque); |
68485420 KW |
4734 | acb->req.sector = sector_num; |
4735 | acb->req.nb_sectors = nb_sectors; | |
4736 | acb->req.qiov = qiov; | |
d20d9b7c | 4737 | acb->req.flags = flags; |
68485420 | 4738 | acb->is_write = is_write; |
d318aea9 | 4739 | acb->done = NULL; |
68485420 | 4740 | |
8c5873d6 | 4741 | co = qemu_coroutine_create(bdrv_co_do_rw); |
68485420 KW |
4742 | qemu_coroutine_enter(co, acb); |
4743 | ||
4744 | return &acb->common; | |
4745 | } | |
4746 | ||
07f07615 | 4747 | static void coroutine_fn bdrv_aio_flush_co_entry(void *opaque) |
b2e12bc6 | 4748 | { |
07f07615 PB |
4749 | BlockDriverAIOCBCoroutine *acb = opaque; |
4750 | BlockDriverState *bs = acb->common.bs; | |
b2e12bc6 | 4751 | |
07f07615 | 4752 | acb->req.error = bdrv_co_flush(bs); |
2572b37a | 4753 | acb->bh = aio_bh_new(bdrv_get_aio_context(bs), bdrv_co_em_bh, acb); |
b2e12bc6 | 4754 | qemu_bh_schedule(acb->bh); |
b2e12bc6 CH |
4755 | } |
4756 | ||
07f07615 | 4757 | BlockDriverAIOCB *bdrv_aio_flush(BlockDriverState *bs, |
016f5cf6 AG |
4758 | BlockDriverCompletionFunc *cb, void *opaque) |
4759 | { | |
07f07615 | 4760 | trace_bdrv_aio_flush(bs, opaque); |
016f5cf6 | 4761 | |
07f07615 PB |
4762 | Coroutine *co; |
4763 | BlockDriverAIOCBCoroutine *acb; | |
016f5cf6 | 4764 | |
d7331bed | 4765 | acb = qemu_aio_get(&bdrv_em_co_aiocb_info, bs, cb, opaque); |
d318aea9 KW |
4766 | acb->done = NULL; |
4767 | ||
07f07615 PB |
4768 | co = qemu_coroutine_create(bdrv_aio_flush_co_entry); |
4769 | qemu_coroutine_enter(co, acb); | |
016f5cf6 | 4770 | |
016f5cf6 AG |
4771 | return &acb->common; |
4772 | } | |
4773 | ||
4265d620 PB |
4774 | static void coroutine_fn bdrv_aio_discard_co_entry(void *opaque) |
4775 | { | |
4776 | BlockDriverAIOCBCoroutine *acb = opaque; | |
4777 | BlockDriverState *bs = acb->common.bs; | |
4778 | ||
4779 | acb->req.error = bdrv_co_discard(bs, acb->req.sector, acb->req.nb_sectors); | |
2572b37a | 4780 | acb->bh = aio_bh_new(bdrv_get_aio_context(bs), bdrv_co_em_bh, acb); |
4265d620 PB |
4781 | qemu_bh_schedule(acb->bh); |
4782 | } | |
4783 | ||
4784 | BlockDriverAIOCB *bdrv_aio_discard(BlockDriverState *bs, | |
4785 | int64_t sector_num, int nb_sectors, | |
4786 | BlockDriverCompletionFunc *cb, void *opaque) | |
4787 | { | |
4788 | Coroutine *co; | |
4789 | BlockDriverAIOCBCoroutine *acb; | |
4790 | ||
4791 | trace_bdrv_aio_discard(bs, sector_num, nb_sectors, opaque); | |
4792 | ||
d7331bed | 4793 | acb = qemu_aio_get(&bdrv_em_co_aiocb_info, bs, cb, opaque); |
4265d620 PB |
4794 | acb->req.sector = sector_num; |
4795 | acb->req.nb_sectors = nb_sectors; | |
d318aea9 | 4796 | acb->done = NULL; |
4265d620 PB |
4797 | co = qemu_coroutine_create(bdrv_aio_discard_co_entry); |
4798 | qemu_coroutine_enter(co, acb); | |
4799 | ||
4800 | return &acb->common; | |
4801 | } | |
4802 | ||
ea2384d3 FB |
4803 | void bdrv_init(void) |
4804 | { | |
5efa9d5a | 4805 | module_call_init(MODULE_INIT_BLOCK); |
ea2384d3 | 4806 | } |
ce1a14dc | 4807 | |
eb852011 MA |
4808 | void bdrv_init_with_whitelist(void) |
4809 | { | |
4810 | use_bdrv_whitelist = 1; | |
4811 | bdrv_init(); | |
4812 | } | |
4813 | ||
d7331bed | 4814 | void *qemu_aio_get(const AIOCBInfo *aiocb_info, BlockDriverState *bs, |
c16b5a2c | 4815 | BlockDriverCompletionFunc *cb, void *opaque) |
ce1a14dc | 4816 | { |
ce1a14dc PB |
4817 | BlockDriverAIOCB *acb; |
4818 | ||
d7331bed SH |
4819 | acb = g_slice_alloc(aiocb_info->aiocb_size); |
4820 | acb->aiocb_info = aiocb_info; | |
ce1a14dc PB |
4821 | acb->bs = bs; |
4822 | acb->cb = cb; | |
4823 | acb->opaque = opaque; | |
4824 | return acb; | |
4825 | } | |
4826 | ||
4827 | void qemu_aio_release(void *p) | |
4828 | { | |
d37c975f | 4829 | BlockDriverAIOCB *acb = p; |
d7331bed | 4830 | g_slice_free1(acb->aiocb_info->aiocb_size, acb); |
ce1a14dc | 4831 | } |
19cb3738 | 4832 | |
f9f05dc5 KW |
4833 | /**************************************************************/ |
4834 | /* Coroutine block device emulation */ | |
4835 | ||
4836 | typedef struct CoroutineIOCompletion { | |
4837 | Coroutine *coroutine; | |
4838 | int ret; | |
4839 | } CoroutineIOCompletion; | |
4840 | ||
4841 | static void bdrv_co_io_em_complete(void *opaque, int ret) | |
4842 | { | |
4843 | CoroutineIOCompletion *co = opaque; | |
4844 | ||
4845 | co->ret = ret; | |
4846 | qemu_coroutine_enter(co->coroutine, NULL); | |
4847 | } | |
4848 | ||
4849 | static int coroutine_fn bdrv_co_io_em(BlockDriverState *bs, int64_t sector_num, | |
4850 | int nb_sectors, QEMUIOVector *iov, | |
4851 | bool is_write) | |
4852 | { | |
4853 | CoroutineIOCompletion co = { | |
4854 | .coroutine = qemu_coroutine_self(), | |
4855 | }; | |
4856 | BlockDriverAIOCB *acb; | |
4857 | ||
4858 | if (is_write) { | |
a652d160 SH |
4859 | acb = bs->drv->bdrv_aio_writev(bs, sector_num, iov, nb_sectors, |
4860 | bdrv_co_io_em_complete, &co); | |
f9f05dc5 | 4861 | } else { |
a652d160 SH |
4862 | acb = bs->drv->bdrv_aio_readv(bs, sector_num, iov, nb_sectors, |
4863 | bdrv_co_io_em_complete, &co); | |
f9f05dc5 KW |
4864 | } |
4865 | ||
59370aaa | 4866 | trace_bdrv_co_io_em(bs, sector_num, nb_sectors, is_write, acb); |
f9f05dc5 KW |
4867 | if (!acb) { |
4868 | return -EIO; | |
4869 | } | |
4870 | qemu_coroutine_yield(); | |
4871 | ||
4872 | return co.ret; | |
4873 | } | |
4874 | ||
4875 | static int coroutine_fn bdrv_co_readv_em(BlockDriverState *bs, | |
4876 | int64_t sector_num, int nb_sectors, | |
4877 | QEMUIOVector *iov) | |
4878 | { | |
4879 | return bdrv_co_io_em(bs, sector_num, nb_sectors, iov, false); | |
4880 | } | |
4881 | ||
4882 | static int coroutine_fn bdrv_co_writev_em(BlockDriverState *bs, | |
4883 | int64_t sector_num, int nb_sectors, | |
4884 | QEMUIOVector *iov) | |
4885 | { | |
4886 | return bdrv_co_io_em(bs, sector_num, nb_sectors, iov, true); | |
4887 | } | |
4888 | ||
07f07615 | 4889 | static void coroutine_fn bdrv_flush_co_entry(void *opaque) |
e7a8a783 | 4890 | { |
07f07615 PB |
4891 | RwCo *rwco = opaque; |
4892 | ||
4893 | rwco->ret = bdrv_co_flush(rwco->bs); | |
4894 | } | |
4895 | ||
4896 | int coroutine_fn bdrv_co_flush(BlockDriverState *bs) | |
4897 | { | |
eb489bb1 KW |
4898 | int ret; |
4899 | ||
29cdb251 | 4900 | if (!bs || !bdrv_is_inserted(bs) || bdrv_is_read_only(bs)) { |
07f07615 | 4901 | return 0; |
eb489bb1 KW |
4902 | } |
4903 | ||
ca716364 | 4904 | /* Write back cached data to the OS even with cache=unsafe */ |
bf736fe3 | 4905 | BLKDBG_EVENT(bs->file, BLKDBG_FLUSH_TO_OS); |
eb489bb1 KW |
4906 | if (bs->drv->bdrv_co_flush_to_os) { |
4907 | ret = bs->drv->bdrv_co_flush_to_os(bs); | |
4908 | if (ret < 0) { | |
4909 | return ret; | |
4910 | } | |
4911 | } | |
4912 | ||
ca716364 KW |
4913 | /* But don't actually force it to the disk with cache=unsafe */ |
4914 | if (bs->open_flags & BDRV_O_NO_FLUSH) { | |
d4c82329 | 4915 | goto flush_parent; |
ca716364 KW |
4916 | } |
4917 | ||
bf736fe3 | 4918 | BLKDBG_EVENT(bs->file, BLKDBG_FLUSH_TO_DISK); |
eb489bb1 | 4919 | if (bs->drv->bdrv_co_flush_to_disk) { |
29cdb251 | 4920 | ret = bs->drv->bdrv_co_flush_to_disk(bs); |
07f07615 PB |
4921 | } else if (bs->drv->bdrv_aio_flush) { |
4922 | BlockDriverAIOCB *acb; | |
4923 | CoroutineIOCompletion co = { | |
4924 | .coroutine = qemu_coroutine_self(), | |
4925 | }; | |
4926 | ||
4927 | acb = bs->drv->bdrv_aio_flush(bs, bdrv_co_io_em_complete, &co); | |
4928 | if (acb == NULL) { | |
29cdb251 | 4929 | ret = -EIO; |
07f07615 PB |
4930 | } else { |
4931 | qemu_coroutine_yield(); | |
29cdb251 | 4932 | ret = co.ret; |
07f07615 | 4933 | } |
07f07615 PB |
4934 | } else { |
4935 | /* | |
4936 | * Some block drivers always operate in either writethrough or unsafe | |
4937 | * mode and don't support bdrv_flush therefore. Usually qemu doesn't | |
4938 | * know how the server works (because the behaviour is hardcoded or | |
4939 | * depends on server-side configuration), so we can't ensure that | |
4940 | * everything is safe on disk. Returning an error doesn't work because | |
4941 | * that would break guests even if the server operates in writethrough | |
4942 | * mode. | |
4943 | * | |
4944 | * Let's hope the user knows what he's doing. | |
4945 | */ | |
29cdb251 | 4946 | ret = 0; |
07f07615 | 4947 | } |
29cdb251 PB |
4948 | if (ret < 0) { |
4949 | return ret; | |
4950 | } | |
4951 | ||
4952 | /* Now flush the underlying protocol. It will also have BDRV_O_NO_FLUSH | |
4953 | * in the case of cache=unsafe, so there are no useless flushes. | |
4954 | */ | |
d4c82329 | 4955 | flush_parent: |
29cdb251 | 4956 | return bdrv_co_flush(bs->file); |
07f07615 PB |
4957 | } |
4958 | ||
5a8a30db | 4959 | void bdrv_invalidate_cache(BlockDriverState *bs, Error **errp) |
0f15423c | 4960 | { |
5a8a30db KW |
4961 | Error *local_err = NULL; |
4962 | int ret; | |
4963 | ||
3456a8d1 KW |
4964 | if (!bs->drv) { |
4965 | return; | |
4966 | } | |
4967 | ||
4968 | if (bs->drv->bdrv_invalidate_cache) { | |
5a8a30db | 4969 | bs->drv->bdrv_invalidate_cache(bs, &local_err); |
3456a8d1 | 4970 | } else if (bs->file) { |
5a8a30db KW |
4971 | bdrv_invalidate_cache(bs->file, &local_err); |
4972 | } | |
4973 | if (local_err) { | |
4974 | error_propagate(errp, local_err); | |
4975 | return; | |
0f15423c | 4976 | } |
3456a8d1 | 4977 | |
5a8a30db KW |
4978 | ret = refresh_total_sectors(bs, bs->total_sectors); |
4979 | if (ret < 0) { | |
4980 | error_setg_errno(errp, -ret, "Could not refresh total sector count"); | |
4981 | return; | |
4982 | } | |
0f15423c AL |
4983 | } |
4984 | ||
5a8a30db | 4985 | void bdrv_invalidate_cache_all(Error **errp) |
0f15423c AL |
4986 | { |
4987 | BlockDriverState *bs; | |
5a8a30db | 4988 | Error *local_err = NULL; |
0f15423c | 4989 | |
dc364f4c | 4990 | QTAILQ_FOREACH(bs, &bdrv_states, device_list) { |
ed78cda3 SH |
4991 | AioContext *aio_context = bdrv_get_aio_context(bs); |
4992 | ||
4993 | aio_context_acquire(aio_context); | |
5a8a30db | 4994 | bdrv_invalidate_cache(bs, &local_err); |
ed78cda3 | 4995 | aio_context_release(aio_context); |
5a8a30db KW |
4996 | if (local_err) { |
4997 | error_propagate(errp, local_err); | |
4998 | return; | |
4999 | } | |
0f15423c AL |
5000 | } |
5001 | } | |
5002 | ||
07789269 BC |
5003 | void bdrv_clear_incoming_migration_all(void) |
5004 | { | |
5005 | BlockDriverState *bs; | |
5006 | ||
dc364f4c | 5007 | QTAILQ_FOREACH(bs, &bdrv_states, device_list) { |
ed78cda3 SH |
5008 | AioContext *aio_context = bdrv_get_aio_context(bs); |
5009 | ||
5010 | aio_context_acquire(aio_context); | |
07789269 | 5011 | bs->open_flags = bs->open_flags & ~(BDRV_O_INCOMING); |
ed78cda3 | 5012 | aio_context_release(aio_context); |
07789269 BC |
5013 | } |
5014 | } | |
5015 | ||
07f07615 PB |
5016 | int bdrv_flush(BlockDriverState *bs) |
5017 | { | |
5018 | Coroutine *co; | |
5019 | RwCo rwco = { | |
5020 | .bs = bs, | |
5021 | .ret = NOT_DONE, | |
e7a8a783 | 5022 | }; |
e7a8a783 | 5023 | |
07f07615 PB |
5024 | if (qemu_in_coroutine()) { |
5025 | /* Fast-path if already in coroutine context */ | |
5026 | bdrv_flush_co_entry(&rwco); | |
5027 | } else { | |
2572b37a SH |
5028 | AioContext *aio_context = bdrv_get_aio_context(bs); |
5029 | ||
07f07615 PB |
5030 | co = qemu_coroutine_create(bdrv_flush_co_entry); |
5031 | qemu_coroutine_enter(co, &rwco); | |
5032 | while (rwco.ret == NOT_DONE) { | |
2572b37a | 5033 | aio_poll(aio_context, true); |
07f07615 | 5034 | } |
e7a8a783 | 5035 | } |
07f07615 PB |
5036 | |
5037 | return rwco.ret; | |
e7a8a783 KW |
5038 | } |
5039 | ||
775aa8b6 KW |
5040 | typedef struct DiscardCo { |
5041 | BlockDriverState *bs; | |
5042 | int64_t sector_num; | |
5043 | int nb_sectors; | |
5044 | int ret; | |
5045 | } DiscardCo; | |
4265d620 PB |
5046 | static void coroutine_fn bdrv_discard_co_entry(void *opaque) |
5047 | { | |
775aa8b6 | 5048 | DiscardCo *rwco = opaque; |
4265d620 PB |
5049 | |
5050 | rwco->ret = bdrv_co_discard(rwco->bs, rwco->sector_num, rwco->nb_sectors); | |
5051 | } | |
5052 | ||
6f14da52 PL |
5053 | /* if no limit is specified in the BlockLimits use a default |
5054 | * of 32768 512-byte sectors (16 MiB) per request. | |
5055 | */ | |
5056 | #define MAX_DISCARD_DEFAULT 32768 | |
5057 | ||
4265d620 PB |
5058 | int coroutine_fn bdrv_co_discard(BlockDriverState *bs, int64_t sector_num, |
5059 | int nb_sectors) | |
5060 | { | |
d51e9fe5 PB |
5061 | int max_discard; |
5062 | ||
4265d620 PB |
5063 | if (!bs->drv) { |
5064 | return -ENOMEDIUM; | |
5065 | } else if (bdrv_check_request(bs, sector_num, nb_sectors)) { | |
5066 | return -EIO; | |
5067 | } else if (bs->read_only) { | |
5068 | return -EROFS; | |
df702c9b PB |
5069 | } |
5070 | ||
e4654d2d | 5071 | bdrv_reset_dirty(bs, sector_num, nb_sectors); |
df702c9b | 5072 | |
9e8f1835 PB |
5073 | /* Do nothing if disabled. */ |
5074 | if (!(bs->open_flags & BDRV_O_UNMAP)) { | |
5075 | return 0; | |
5076 | } | |
5077 | ||
d51e9fe5 PB |
5078 | if (!bs->drv->bdrv_co_discard && !bs->drv->bdrv_aio_discard) { |
5079 | return 0; | |
5080 | } | |
6f14da52 | 5081 | |
d51e9fe5 PB |
5082 | max_discard = bs->bl.max_discard ? bs->bl.max_discard : MAX_DISCARD_DEFAULT; |
5083 | while (nb_sectors > 0) { | |
5084 | int ret; | |
5085 | int num = nb_sectors; | |
6f14da52 | 5086 | |
d51e9fe5 PB |
5087 | /* align request */ |
5088 | if (bs->bl.discard_alignment && | |
5089 | num >= bs->bl.discard_alignment && | |
5090 | sector_num % bs->bl.discard_alignment) { | |
5091 | if (num > bs->bl.discard_alignment) { | |
5092 | num = bs->bl.discard_alignment; | |
6f14da52 | 5093 | } |
d51e9fe5 PB |
5094 | num -= sector_num % bs->bl.discard_alignment; |
5095 | } | |
6f14da52 | 5096 | |
d51e9fe5 PB |
5097 | /* limit request size */ |
5098 | if (num > max_discard) { | |
5099 | num = max_discard; | |
5100 | } | |
6f14da52 | 5101 | |
d51e9fe5 | 5102 | if (bs->drv->bdrv_co_discard) { |
6f14da52 | 5103 | ret = bs->drv->bdrv_co_discard(bs, sector_num, num); |
d51e9fe5 PB |
5104 | } else { |
5105 | BlockDriverAIOCB *acb; | |
5106 | CoroutineIOCompletion co = { | |
5107 | .coroutine = qemu_coroutine_self(), | |
5108 | }; | |
5109 | ||
5110 | acb = bs->drv->bdrv_aio_discard(bs, sector_num, nb_sectors, | |
5111 | bdrv_co_io_em_complete, &co); | |
5112 | if (acb == NULL) { | |
5113 | return -EIO; | |
5114 | } else { | |
5115 | qemu_coroutine_yield(); | |
5116 | ret = co.ret; | |
6f14da52 | 5117 | } |
6f14da52 | 5118 | } |
7ce21016 | 5119 | if (ret && ret != -ENOTSUP) { |
d51e9fe5 | 5120 | return ret; |
4265d620 | 5121 | } |
d51e9fe5 PB |
5122 | |
5123 | sector_num += num; | |
5124 | nb_sectors -= num; | |
4265d620 | 5125 | } |
d51e9fe5 | 5126 | return 0; |
4265d620 PB |
5127 | } |
5128 | ||
5129 | int bdrv_discard(BlockDriverState *bs, int64_t sector_num, int nb_sectors) | |
5130 | { | |
5131 | Coroutine *co; | |
775aa8b6 | 5132 | DiscardCo rwco = { |
4265d620 PB |
5133 | .bs = bs, |
5134 | .sector_num = sector_num, | |
5135 | .nb_sectors = nb_sectors, | |
5136 | .ret = NOT_DONE, | |
5137 | }; | |
5138 | ||
5139 | if (qemu_in_coroutine()) { | |
5140 | /* Fast-path if already in coroutine context */ | |
5141 | bdrv_discard_co_entry(&rwco); | |
5142 | } else { | |
2572b37a SH |
5143 | AioContext *aio_context = bdrv_get_aio_context(bs); |
5144 | ||
4265d620 PB |
5145 | co = qemu_coroutine_create(bdrv_discard_co_entry); |
5146 | qemu_coroutine_enter(co, &rwco); | |
5147 | while (rwco.ret == NOT_DONE) { | |
2572b37a | 5148 | aio_poll(aio_context, true); |
4265d620 PB |
5149 | } |
5150 | } | |
5151 | ||
5152 | return rwco.ret; | |
5153 | } | |
5154 | ||
19cb3738 FB |
5155 | /**************************************************************/ |
5156 | /* removable device support */ | |
5157 | ||
5158 | /** | |
5159 | * Return TRUE if the media is present | |
5160 | */ | |
5161 | int bdrv_is_inserted(BlockDriverState *bs) | |
5162 | { | |
5163 | BlockDriver *drv = bs->drv; | |
a1aff5bf | 5164 | |
19cb3738 FB |
5165 | if (!drv) |
5166 | return 0; | |
5167 | if (!drv->bdrv_is_inserted) | |
a1aff5bf MA |
5168 | return 1; |
5169 | return drv->bdrv_is_inserted(bs); | |
19cb3738 FB |
5170 | } |
5171 | ||
5172 | /** | |
8e49ca46 MA |
5173 | * Return whether the media changed since the last call to this |
5174 | * function, or -ENOTSUP if we don't know. Most drivers don't know. | |
19cb3738 FB |
5175 | */ |
5176 | int bdrv_media_changed(BlockDriverState *bs) | |
5177 | { | |
5178 | BlockDriver *drv = bs->drv; | |
19cb3738 | 5179 | |
8e49ca46 MA |
5180 | if (drv && drv->bdrv_media_changed) { |
5181 | return drv->bdrv_media_changed(bs); | |
5182 | } | |
5183 | return -ENOTSUP; | |
19cb3738 FB |
5184 | } |
5185 | ||
5186 | /** | |
5187 | * If eject_flag is TRUE, eject the media. Otherwise, close the tray | |
5188 | */ | |
f36f3949 | 5189 | void bdrv_eject(BlockDriverState *bs, bool eject_flag) |
19cb3738 FB |
5190 | { |
5191 | BlockDriver *drv = bs->drv; | |
19cb3738 | 5192 | |
822e1cd1 MA |
5193 | if (drv && drv->bdrv_eject) { |
5194 | drv->bdrv_eject(bs, eject_flag); | |
19cb3738 | 5195 | } |
6f382ed2 LC |
5196 | |
5197 | if (bs->device_name[0] != '\0') { | |
a5ee7bd4 WX |
5198 | qapi_event_send_device_tray_moved(bdrv_get_device_name(bs), |
5199 | eject_flag, &error_abort); | |
6f382ed2 | 5200 | } |
19cb3738 FB |
5201 | } |
5202 | ||
19cb3738 FB |
5203 | /** |
5204 | * Lock or unlock the media (if it is locked, the user won't be able | |
5205 | * to eject it manually). | |
5206 | */ | |
025e849a | 5207 | void bdrv_lock_medium(BlockDriverState *bs, bool locked) |
19cb3738 FB |
5208 | { |
5209 | BlockDriver *drv = bs->drv; | |
5210 | ||
025e849a | 5211 | trace_bdrv_lock_medium(bs, locked); |
b8c6d095 | 5212 | |
025e849a MA |
5213 | if (drv && drv->bdrv_lock_medium) { |
5214 | drv->bdrv_lock_medium(bs, locked); | |
19cb3738 FB |
5215 | } |
5216 | } | |
985a03b0 TS |
5217 | |
5218 | /* needed for generic scsi interface */ | |
5219 | ||
5220 | int bdrv_ioctl(BlockDriverState *bs, unsigned long int req, void *buf) | |
5221 | { | |
5222 | BlockDriver *drv = bs->drv; | |
5223 | ||
5224 | if (drv && drv->bdrv_ioctl) | |
5225 | return drv->bdrv_ioctl(bs, req, buf); | |
5226 | return -ENOTSUP; | |
5227 | } | |
7d780669 | 5228 | |
221f715d AL |
5229 | BlockDriverAIOCB *bdrv_aio_ioctl(BlockDriverState *bs, |
5230 | unsigned long int req, void *buf, | |
5231 | BlockDriverCompletionFunc *cb, void *opaque) | |
7d780669 | 5232 | { |
221f715d | 5233 | BlockDriver *drv = bs->drv; |
7d780669 | 5234 | |
221f715d AL |
5235 | if (drv && drv->bdrv_aio_ioctl) |
5236 | return drv->bdrv_aio_ioctl(bs, req, buf, cb, opaque); | |
5237 | return NULL; | |
7d780669 | 5238 | } |
e268ca52 | 5239 | |
1b7fd729 | 5240 | void bdrv_set_guest_block_size(BlockDriverState *bs, int align) |
7b6f9300 | 5241 | { |
1b7fd729 | 5242 | bs->guest_block_size = align; |
7b6f9300 | 5243 | } |
7cd1e32a | 5244 | |
e268ca52 AL |
5245 | void *qemu_blockalign(BlockDriverState *bs, size_t size) |
5246 | { | |
339064d5 | 5247 | return qemu_memalign(bdrv_opt_mem_align(bs), size); |
e268ca52 | 5248 | } |
7cd1e32a | 5249 | |
c53b1c51 SH |
5250 | /* |
5251 | * Check if all memory in this vector is sector aligned. | |
5252 | */ | |
5253 | bool bdrv_qiov_is_aligned(BlockDriverState *bs, QEMUIOVector *qiov) | |
5254 | { | |
5255 | int i; | |
339064d5 | 5256 | size_t alignment = bdrv_opt_mem_align(bs); |
c53b1c51 SH |
5257 | |
5258 | for (i = 0; i < qiov->niov; i++) { | |
339064d5 | 5259 | if ((uintptr_t) qiov->iov[i].iov_base % alignment) { |
c53b1c51 | 5260 | return false; |
1ff735bd | 5261 | } |
339064d5 | 5262 | if (qiov->iov[i].iov_len % alignment) { |
1ff735bd | 5263 | return false; |
c53b1c51 SH |
5264 | } |
5265 | } | |
5266 | ||
5267 | return true; | |
5268 | } | |
5269 | ||
b8afb520 FZ |
5270 | BdrvDirtyBitmap *bdrv_create_dirty_bitmap(BlockDriverState *bs, int granularity, |
5271 | Error **errp) | |
7cd1e32a LS |
5272 | { |
5273 | int64_t bitmap_size; | |
e4654d2d | 5274 | BdrvDirtyBitmap *bitmap; |
a55eb92c | 5275 | |
50717e94 PB |
5276 | assert((granularity & (granularity - 1)) == 0); |
5277 | ||
e4654d2d FZ |
5278 | granularity >>= BDRV_SECTOR_BITS; |
5279 | assert(granularity); | |
b8afb520 FZ |
5280 | bitmap_size = bdrv_getlength(bs); |
5281 | if (bitmap_size < 0) { | |
5282 | error_setg_errno(errp, -bitmap_size, "could not get length of device"); | |
5283 | errno = -bitmap_size; | |
5284 | return NULL; | |
5285 | } | |
5286 | bitmap_size >>= BDRV_SECTOR_BITS; | |
e4654d2d FZ |
5287 | bitmap = g_malloc0(sizeof(BdrvDirtyBitmap)); |
5288 | bitmap->bitmap = hbitmap_alloc(bitmap_size, ffs(granularity) - 1); | |
5289 | QLIST_INSERT_HEAD(&bs->dirty_bitmaps, bitmap, list); | |
5290 | return bitmap; | |
5291 | } | |
5292 | ||
5293 | void bdrv_release_dirty_bitmap(BlockDriverState *bs, BdrvDirtyBitmap *bitmap) | |
5294 | { | |
5295 | BdrvDirtyBitmap *bm, *next; | |
5296 | QLIST_FOREACH_SAFE(bm, &bs->dirty_bitmaps, list, next) { | |
5297 | if (bm == bitmap) { | |
5298 | QLIST_REMOVE(bitmap, list); | |
5299 | hbitmap_free(bitmap->bitmap); | |
5300 | g_free(bitmap); | |
5301 | return; | |
a55eb92c | 5302 | } |
7cd1e32a LS |
5303 | } |
5304 | } | |
5305 | ||
21b56835 FZ |
5306 | BlockDirtyInfoList *bdrv_query_dirty_bitmaps(BlockDriverState *bs) |
5307 | { | |
5308 | BdrvDirtyBitmap *bm; | |
5309 | BlockDirtyInfoList *list = NULL; | |
5310 | BlockDirtyInfoList **plist = &list; | |
5311 | ||
5312 | QLIST_FOREACH(bm, &bs->dirty_bitmaps, list) { | |
5313 | BlockDirtyInfo *info = g_malloc0(sizeof(BlockDirtyInfo)); | |
5314 | BlockDirtyInfoList *entry = g_malloc0(sizeof(BlockDirtyInfoList)); | |
5315 | info->count = bdrv_get_dirty_count(bs, bm); | |
5316 | info->granularity = | |
5317 | ((int64_t) BDRV_SECTOR_SIZE << hbitmap_granularity(bm->bitmap)); | |
5318 | entry->value = info; | |
5319 | *plist = entry; | |
5320 | plist = &entry->next; | |
5321 | } | |
5322 | ||
5323 | return list; | |
5324 | } | |
5325 | ||
e4654d2d | 5326 | int bdrv_get_dirty(BlockDriverState *bs, BdrvDirtyBitmap *bitmap, int64_t sector) |
7cd1e32a | 5327 | { |
e4654d2d FZ |
5328 | if (bitmap) { |
5329 | return hbitmap_get(bitmap->bitmap, sector); | |
7cd1e32a LS |
5330 | } else { |
5331 | return 0; | |
5332 | } | |
5333 | } | |
5334 | ||
e4654d2d FZ |
5335 | void bdrv_dirty_iter_init(BlockDriverState *bs, |
5336 | BdrvDirtyBitmap *bitmap, HBitmapIter *hbi) | |
1755da16 | 5337 | { |
e4654d2d | 5338 | hbitmap_iter_init(hbi, bitmap->bitmap, 0); |
1755da16 PB |
5339 | } |
5340 | ||
5341 | void bdrv_set_dirty(BlockDriverState *bs, int64_t cur_sector, | |
5342 | int nr_sectors) | |
5343 | { | |
e4654d2d FZ |
5344 | BdrvDirtyBitmap *bitmap; |
5345 | QLIST_FOREACH(bitmap, &bs->dirty_bitmaps, list) { | |
5346 | hbitmap_set(bitmap->bitmap, cur_sector, nr_sectors); | |
5347 | } | |
1755da16 PB |
5348 | } |
5349 | ||
e4654d2d | 5350 | void bdrv_reset_dirty(BlockDriverState *bs, int64_t cur_sector, int nr_sectors) |
7cd1e32a | 5351 | { |
e4654d2d FZ |
5352 | BdrvDirtyBitmap *bitmap; |
5353 | QLIST_FOREACH(bitmap, &bs->dirty_bitmaps, list) { | |
5354 | hbitmap_reset(bitmap->bitmap, cur_sector, nr_sectors); | |
5355 | } | |
7cd1e32a | 5356 | } |
aaa0eb75 | 5357 | |
e4654d2d | 5358 | int64_t bdrv_get_dirty_count(BlockDriverState *bs, BdrvDirtyBitmap *bitmap) |
aaa0eb75 | 5359 | { |
e4654d2d | 5360 | return hbitmap_count(bitmap->bitmap); |
aaa0eb75 | 5361 | } |
f88e1a42 | 5362 | |
9fcb0251 FZ |
5363 | /* Get a reference to bs */ |
5364 | void bdrv_ref(BlockDriverState *bs) | |
5365 | { | |
5366 | bs->refcnt++; | |
5367 | } | |
5368 | ||
5369 | /* Release a previously grabbed reference to bs. | |
5370 | * If after releasing, reference count is zero, the BlockDriverState is | |
5371 | * deleted. */ | |
5372 | void bdrv_unref(BlockDriverState *bs) | |
5373 | { | |
5374 | assert(bs->refcnt > 0); | |
5375 | if (--bs->refcnt == 0) { | |
5376 | bdrv_delete(bs); | |
5377 | } | |
5378 | } | |
5379 | ||
fbe40ff7 FZ |
5380 | struct BdrvOpBlocker { |
5381 | Error *reason; | |
5382 | QLIST_ENTRY(BdrvOpBlocker) list; | |
5383 | }; | |
5384 | ||
5385 | bool bdrv_op_is_blocked(BlockDriverState *bs, BlockOpType op, Error **errp) | |
5386 | { | |
5387 | BdrvOpBlocker *blocker; | |
5388 | assert((int) op >= 0 && op < BLOCK_OP_TYPE_MAX); | |
5389 | if (!QLIST_EMPTY(&bs->op_blockers[op])) { | |
5390 | blocker = QLIST_FIRST(&bs->op_blockers[op]); | |
5391 | if (errp) { | |
5392 | error_setg(errp, "Device '%s' is busy: %s", | |
5393 | bs->device_name, error_get_pretty(blocker->reason)); | |
5394 | } | |
5395 | return true; | |
5396 | } | |
5397 | return false; | |
5398 | } | |
5399 | ||
5400 | void bdrv_op_block(BlockDriverState *bs, BlockOpType op, Error *reason) | |
5401 | { | |
5402 | BdrvOpBlocker *blocker; | |
5403 | assert((int) op >= 0 && op < BLOCK_OP_TYPE_MAX); | |
5404 | ||
5405 | blocker = g_malloc0(sizeof(BdrvOpBlocker)); | |
5406 | blocker->reason = reason; | |
5407 | QLIST_INSERT_HEAD(&bs->op_blockers[op], blocker, list); | |
5408 | } | |
5409 | ||
5410 | void bdrv_op_unblock(BlockDriverState *bs, BlockOpType op, Error *reason) | |
5411 | { | |
5412 | BdrvOpBlocker *blocker, *next; | |
5413 | assert((int) op >= 0 && op < BLOCK_OP_TYPE_MAX); | |
5414 | QLIST_FOREACH_SAFE(blocker, &bs->op_blockers[op], list, next) { | |
5415 | if (blocker->reason == reason) { | |
5416 | QLIST_REMOVE(blocker, list); | |
5417 | g_free(blocker); | |
5418 | } | |
5419 | } | |
5420 | } | |
5421 | ||
5422 | void bdrv_op_block_all(BlockDriverState *bs, Error *reason) | |
5423 | { | |
5424 | int i; | |
5425 | for (i = 0; i < BLOCK_OP_TYPE_MAX; i++) { | |
5426 | bdrv_op_block(bs, i, reason); | |
5427 | } | |
5428 | } | |
5429 | ||
5430 | void bdrv_op_unblock_all(BlockDriverState *bs, Error *reason) | |
5431 | { | |
5432 | int i; | |
5433 | for (i = 0; i < BLOCK_OP_TYPE_MAX; i++) { | |
5434 | bdrv_op_unblock(bs, i, reason); | |
5435 | } | |
5436 | } | |
5437 | ||
5438 | bool bdrv_op_blocker_is_empty(BlockDriverState *bs) | |
5439 | { | |
5440 | int i; | |
5441 | ||
5442 | for (i = 0; i < BLOCK_OP_TYPE_MAX; i++) { | |
5443 | if (!QLIST_EMPTY(&bs->op_blockers[i])) { | |
5444 | return false; | |
5445 | } | |
5446 | } | |
5447 | return true; | |
5448 | } | |
5449 | ||
28a7282a LC |
5450 | void bdrv_iostatus_enable(BlockDriverState *bs) |
5451 | { | |
d6bf279e | 5452 | bs->iostatus_enabled = true; |
58e21ef5 | 5453 | bs->iostatus = BLOCK_DEVICE_IO_STATUS_OK; |
28a7282a LC |
5454 | } |
5455 | ||
5456 | /* The I/O status is only enabled if the drive explicitly | |
5457 | * enables it _and_ the VM is configured to stop on errors */ | |
5458 | bool bdrv_iostatus_is_enabled(const BlockDriverState *bs) | |
5459 | { | |
d6bf279e | 5460 | return (bs->iostatus_enabled && |
92aa5c6d PB |
5461 | (bs->on_write_error == BLOCKDEV_ON_ERROR_ENOSPC || |
5462 | bs->on_write_error == BLOCKDEV_ON_ERROR_STOP || | |
5463 | bs->on_read_error == BLOCKDEV_ON_ERROR_STOP)); | |
28a7282a LC |
5464 | } |
5465 | ||
5466 | void bdrv_iostatus_disable(BlockDriverState *bs) | |
5467 | { | |
d6bf279e | 5468 | bs->iostatus_enabled = false; |
28a7282a LC |
5469 | } |
5470 | ||
5471 | void bdrv_iostatus_reset(BlockDriverState *bs) | |
5472 | { | |
5473 | if (bdrv_iostatus_is_enabled(bs)) { | |
58e21ef5 | 5474 | bs->iostatus = BLOCK_DEVICE_IO_STATUS_OK; |
3bd293c3 PB |
5475 | if (bs->job) { |
5476 | block_job_iostatus_reset(bs->job); | |
5477 | } | |
28a7282a LC |
5478 | } |
5479 | } | |
5480 | ||
28a7282a LC |
5481 | void bdrv_iostatus_set_err(BlockDriverState *bs, int error) |
5482 | { | |
3e1caa5f PB |
5483 | assert(bdrv_iostatus_is_enabled(bs)); |
5484 | if (bs->iostatus == BLOCK_DEVICE_IO_STATUS_OK) { | |
58e21ef5 LC |
5485 | bs->iostatus = error == ENOSPC ? BLOCK_DEVICE_IO_STATUS_NOSPACE : |
5486 | BLOCK_DEVICE_IO_STATUS_FAILED; | |
28a7282a LC |
5487 | } |
5488 | } | |
5489 | ||
a597e79c CH |
5490 | void |
5491 | bdrv_acct_start(BlockDriverState *bs, BlockAcctCookie *cookie, int64_t bytes, | |
5492 | enum BlockAcctType type) | |
5493 | { | |
5494 | assert(type < BDRV_MAX_IOTYPE); | |
5495 | ||
5496 | cookie->bytes = bytes; | |
c488c7f6 | 5497 | cookie->start_time_ns = get_clock(); |
a597e79c CH |
5498 | cookie->type = type; |
5499 | } | |
5500 | ||
5501 | void | |
5502 | bdrv_acct_done(BlockDriverState *bs, BlockAcctCookie *cookie) | |
5503 | { | |
5504 | assert(cookie->type < BDRV_MAX_IOTYPE); | |
5505 | ||
5506 | bs->nr_bytes[cookie->type] += cookie->bytes; | |
5507 | bs->nr_ops[cookie->type]++; | |
c488c7f6 | 5508 | bs->total_time_ns[cookie->type] += get_clock() - cookie->start_time_ns; |
a597e79c CH |
5509 | } |
5510 | ||
d92ada22 LC |
5511 | void bdrv_img_create(const char *filename, const char *fmt, |
5512 | const char *base_filename, const char *base_fmt, | |
f382d43a MR |
5513 | char *options, uint64_t img_size, int flags, |
5514 | Error **errp, bool quiet) | |
f88e1a42 | 5515 | { |
83d0521a CL |
5516 | QemuOptsList *create_opts = NULL; |
5517 | QemuOpts *opts = NULL; | |
5518 | const char *backing_fmt, *backing_file; | |
5519 | int64_t size; | |
f88e1a42 | 5520 | BlockDriver *drv, *proto_drv; |
96df67d1 | 5521 | BlockDriver *backing_drv = NULL; |
cc84d90f | 5522 | Error *local_err = NULL; |
f88e1a42 JS |
5523 | int ret = 0; |
5524 | ||
5525 | /* Find driver and parse its options */ | |
5526 | drv = bdrv_find_format(fmt); | |
5527 | if (!drv) { | |
71c79813 | 5528 | error_setg(errp, "Unknown file format '%s'", fmt); |
d92ada22 | 5529 | return; |
f88e1a42 JS |
5530 | } |
5531 | ||
98289620 | 5532 | proto_drv = bdrv_find_protocol(filename, true); |
f88e1a42 | 5533 | if (!proto_drv) { |
71c79813 | 5534 | error_setg(errp, "Unknown protocol '%s'", filename); |
d92ada22 | 5535 | return; |
f88e1a42 JS |
5536 | } |
5537 | ||
c282e1fd CL |
5538 | create_opts = qemu_opts_append(create_opts, drv->create_opts); |
5539 | create_opts = qemu_opts_append(create_opts, proto_drv->create_opts); | |
f88e1a42 JS |
5540 | |
5541 | /* Create parameter list with default values */ | |
83d0521a CL |
5542 | opts = qemu_opts_create(create_opts, NULL, 0, &error_abort); |
5543 | qemu_opt_set_number(opts, BLOCK_OPT_SIZE, img_size); | |
f88e1a42 JS |
5544 | |
5545 | /* Parse -o options */ | |
5546 | if (options) { | |
83d0521a CL |
5547 | if (qemu_opts_do_parse(opts, options, NULL) != 0) { |
5548 | error_setg(errp, "Invalid options for file format '%s'", fmt); | |
f88e1a42 JS |
5549 | goto out; |
5550 | } | |
5551 | } | |
5552 | ||
5553 | if (base_filename) { | |
83d0521a | 5554 | if (qemu_opt_set(opts, BLOCK_OPT_BACKING_FILE, base_filename)) { |
71c79813 LC |
5555 | error_setg(errp, "Backing file not supported for file format '%s'", |
5556 | fmt); | |
f88e1a42 JS |
5557 | goto out; |
5558 | } | |
5559 | } | |
5560 | ||
5561 | if (base_fmt) { | |
83d0521a | 5562 | if (qemu_opt_set(opts, BLOCK_OPT_BACKING_FMT, base_fmt)) { |
71c79813 LC |
5563 | error_setg(errp, "Backing file format not supported for file " |
5564 | "format '%s'", fmt); | |
f88e1a42 JS |
5565 | goto out; |
5566 | } | |
5567 | } | |
5568 | ||
83d0521a CL |
5569 | backing_file = qemu_opt_get(opts, BLOCK_OPT_BACKING_FILE); |
5570 | if (backing_file) { | |
5571 | if (!strcmp(filename, backing_file)) { | |
71c79813 LC |
5572 | error_setg(errp, "Error: Trying to create an image with the " |
5573 | "same filename as the backing file"); | |
792da93a JS |
5574 | goto out; |
5575 | } | |
5576 | } | |
5577 | ||
83d0521a CL |
5578 | backing_fmt = qemu_opt_get(opts, BLOCK_OPT_BACKING_FMT); |
5579 | if (backing_fmt) { | |
5580 | backing_drv = bdrv_find_format(backing_fmt); | |
96df67d1 | 5581 | if (!backing_drv) { |
71c79813 | 5582 | error_setg(errp, "Unknown backing file format '%s'", |
83d0521a | 5583 | backing_fmt); |
f88e1a42 JS |
5584 | goto out; |
5585 | } | |
5586 | } | |
5587 | ||
5588 | // The size for the image must always be specified, with one exception: | |
5589 | // If we are using a backing file, we can obtain the size from there | |
83d0521a CL |
5590 | size = qemu_opt_get_size(opts, BLOCK_OPT_SIZE, 0); |
5591 | if (size == -1) { | |
5592 | if (backing_file) { | |
66f6b814 | 5593 | BlockDriverState *bs; |
f88e1a42 | 5594 | uint64_t size; |
63090dac PB |
5595 | int back_flags; |
5596 | ||
5597 | /* backing files always opened read-only */ | |
5598 | back_flags = | |
5599 | flags & ~(BDRV_O_RDWR | BDRV_O_SNAPSHOT | BDRV_O_NO_BACKING); | |
f88e1a42 | 5600 | |
f67503e5 | 5601 | bs = NULL; |
83d0521a | 5602 | ret = bdrv_open(&bs, backing_file, NULL, NULL, back_flags, |
cc84d90f | 5603 | backing_drv, &local_err); |
f88e1a42 | 5604 | if (ret < 0) { |
cc84d90f | 5605 | error_setg_errno(errp, -ret, "Could not open '%s': %s", |
83d0521a | 5606 | backing_file, |
cc84d90f HR |
5607 | error_get_pretty(local_err)); |
5608 | error_free(local_err); | |
5609 | local_err = NULL; | |
f88e1a42 JS |
5610 | goto out; |
5611 | } | |
5612 | bdrv_get_geometry(bs, &size); | |
5613 | size *= 512; | |
5614 | ||
83d0521a | 5615 | qemu_opt_set_number(opts, BLOCK_OPT_SIZE, size); |
66f6b814 HR |
5616 | |
5617 | bdrv_unref(bs); | |
f88e1a42 | 5618 | } else { |
71c79813 | 5619 | error_setg(errp, "Image creation needs a size parameter"); |
f88e1a42 JS |
5620 | goto out; |
5621 | } | |
5622 | } | |
5623 | ||
f382d43a MR |
5624 | if (!quiet) { |
5625 | printf("Formatting '%s', fmt=%s ", filename, fmt); | |
83d0521a | 5626 | qemu_opts_print(opts); |
f382d43a MR |
5627 | puts(""); |
5628 | } | |
83d0521a | 5629 | |
c282e1fd | 5630 | ret = bdrv_create(drv, filename, opts, &local_err); |
83d0521a | 5631 | |
cc84d90f HR |
5632 | if (ret == -EFBIG) { |
5633 | /* This is generally a better message than whatever the driver would | |
5634 | * deliver (especially because of the cluster_size_hint), since that | |
5635 | * is most probably not much different from "image too large". */ | |
5636 | const char *cluster_size_hint = ""; | |
83d0521a | 5637 | if (qemu_opt_get_size(opts, BLOCK_OPT_CLUSTER_SIZE, 0)) { |
cc84d90f | 5638 | cluster_size_hint = " (try using a larger cluster size)"; |
f88e1a42 | 5639 | } |
cc84d90f HR |
5640 | error_setg(errp, "The image size is too large for file format '%s'" |
5641 | "%s", fmt, cluster_size_hint); | |
5642 | error_free(local_err); | |
5643 | local_err = NULL; | |
f88e1a42 JS |
5644 | } |
5645 | ||
5646 | out: | |
83d0521a CL |
5647 | qemu_opts_del(opts); |
5648 | qemu_opts_free(create_opts); | |
84d18f06 | 5649 | if (local_err) { |
cc84d90f HR |
5650 | error_propagate(errp, local_err); |
5651 | } | |
f88e1a42 | 5652 | } |
85d126f3 SH |
5653 | |
5654 | AioContext *bdrv_get_aio_context(BlockDriverState *bs) | |
5655 | { | |
dcd04228 SH |
5656 | return bs->aio_context; |
5657 | } | |
5658 | ||
5659 | void bdrv_detach_aio_context(BlockDriverState *bs) | |
5660 | { | |
5661 | if (!bs->drv) { | |
5662 | return; | |
5663 | } | |
5664 | ||
13af91eb SH |
5665 | if (bs->io_limits_enabled) { |
5666 | throttle_detach_aio_context(&bs->throttle_state); | |
5667 | } | |
dcd04228 SH |
5668 | if (bs->drv->bdrv_detach_aio_context) { |
5669 | bs->drv->bdrv_detach_aio_context(bs); | |
5670 | } | |
5671 | if (bs->file) { | |
5672 | bdrv_detach_aio_context(bs->file); | |
5673 | } | |
5674 | if (bs->backing_hd) { | |
5675 | bdrv_detach_aio_context(bs->backing_hd); | |
5676 | } | |
5677 | ||
5678 | bs->aio_context = NULL; | |
5679 | } | |
5680 | ||
5681 | void bdrv_attach_aio_context(BlockDriverState *bs, | |
5682 | AioContext *new_context) | |
5683 | { | |
5684 | if (!bs->drv) { | |
5685 | return; | |
5686 | } | |
5687 | ||
5688 | bs->aio_context = new_context; | |
5689 | ||
5690 | if (bs->backing_hd) { | |
5691 | bdrv_attach_aio_context(bs->backing_hd, new_context); | |
5692 | } | |
5693 | if (bs->file) { | |
5694 | bdrv_attach_aio_context(bs->file, new_context); | |
5695 | } | |
5696 | if (bs->drv->bdrv_attach_aio_context) { | |
5697 | bs->drv->bdrv_attach_aio_context(bs, new_context); | |
5698 | } | |
13af91eb SH |
5699 | if (bs->io_limits_enabled) { |
5700 | throttle_attach_aio_context(&bs->throttle_state, new_context); | |
5701 | } | |
dcd04228 SH |
5702 | } |
5703 | ||
5704 | void bdrv_set_aio_context(BlockDriverState *bs, AioContext *new_context) | |
5705 | { | |
5706 | bdrv_drain_all(); /* ensure there are no in-flight requests */ | |
5707 | ||
5708 | bdrv_detach_aio_context(bs); | |
5709 | ||
5710 | /* This function executes in the old AioContext so acquire the new one in | |
5711 | * case it runs in a different thread. | |
5712 | */ | |
5713 | aio_context_acquire(new_context); | |
5714 | bdrv_attach_aio_context(bs, new_context); | |
5715 | aio_context_release(new_context); | |
85d126f3 | 5716 | } |
d616b224 SH |
5717 | |
5718 | void bdrv_add_before_write_notifier(BlockDriverState *bs, | |
5719 | NotifierWithReturn *notifier) | |
5720 | { | |
5721 | notifier_with_return_list_add(&bs->before_write_notifiers, notifier); | |
5722 | } | |
6f176b48 | 5723 | |
c282e1fd | 5724 | int bdrv_amend_options(BlockDriverState *bs, QemuOpts *opts) |
6f176b48 | 5725 | { |
c282e1fd | 5726 | if (!bs->drv->bdrv_amend_options) { |
6f176b48 HR |
5727 | return -ENOTSUP; |
5728 | } | |
c282e1fd | 5729 | return bs->drv->bdrv_amend_options(bs, opts); |
6f176b48 | 5730 | } |
f6186f49 | 5731 | |
b5042a36 BC |
5732 | /* This function will be called by the bdrv_recurse_is_first_non_filter method |
5733 | * of block filter and by bdrv_is_first_non_filter. | |
5734 | * It is used to test if the given bs is the candidate or recurse more in the | |
5735 | * node graph. | |
212a5a8f | 5736 | */ |
b5042a36 | 5737 | bool bdrv_recurse_is_first_non_filter(BlockDriverState *bs, |
212a5a8f | 5738 | BlockDriverState *candidate) |
f6186f49 | 5739 | { |
b5042a36 BC |
5740 | /* return false if basic checks fails */ |
5741 | if (!bs || !bs->drv) { | |
212a5a8f | 5742 | return false; |
f6186f49 BC |
5743 | } |
5744 | ||
b5042a36 BC |
5745 | /* the code reached a non block filter driver -> check if the bs is |
5746 | * the same as the candidate. It's the recursion termination condition. | |
5747 | */ | |
5748 | if (!bs->drv->is_filter) { | |
5749 | return bs == candidate; | |
212a5a8f | 5750 | } |
b5042a36 | 5751 | /* Down this path the driver is a block filter driver */ |
212a5a8f | 5752 | |
b5042a36 BC |
5753 | /* If the block filter recursion method is defined use it to recurse down |
5754 | * the node graph. | |
5755 | */ | |
5756 | if (bs->drv->bdrv_recurse_is_first_non_filter) { | |
212a5a8f | 5757 | return bs->drv->bdrv_recurse_is_first_non_filter(bs, candidate); |
f6186f49 BC |
5758 | } |
5759 | ||
b5042a36 BC |
5760 | /* the driver is a block filter but don't allow to recurse -> return false |
5761 | */ | |
5762 | return false; | |
f6186f49 BC |
5763 | } |
5764 | ||
212a5a8f BC |
5765 | /* This function checks if the candidate is the first non filter bs down it's |
5766 | * bs chain. Since we don't have pointers to parents it explore all bs chains | |
5767 | * from the top. Some filters can choose not to pass down the recursion. | |
5768 | */ | |
5769 | bool bdrv_is_first_non_filter(BlockDriverState *candidate) | |
f6186f49 | 5770 | { |
212a5a8f BC |
5771 | BlockDriverState *bs; |
5772 | ||
5773 | /* walk down the bs forest recursively */ | |
5774 | QTAILQ_FOREACH(bs, &bdrv_states, device_list) { | |
5775 | bool perm; | |
5776 | ||
b5042a36 | 5777 | /* try to recurse in this top level bs */ |
e6dc8a1f | 5778 | perm = bdrv_recurse_is_first_non_filter(bs, candidate); |
212a5a8f BC |
5779 | |
5780 | /* candidate is the first non filter */ | |
5781 | if (perm) { | |
5782 | return true; | |
5783 | } | |
5784 | } | |
5785 | ||
5786 | return false; | |
f6186f49 | 5787 | } |
09158f00 BC |
5788 | |
5789 | BlockDriverState *check_to_replace_node(const char *node_name, Error **errp) | |
5790 | { | |
5791 | BlockDriverState *to_replace_bs = bdrv_find_node(node_name); | |
5792 | if (!to_replace_bs) { | |
5793 | error_setg(errp, "Node name '%s' not found", node_name); | |
5794 | return NULL; | |
5795 | } | |
5796 | ||
5797 | if (bdrv_op_is_blocked(to_replace_bs, BLOCK_OP_TYPE_REPLACE, errp)) { | |
5798 | return NULL; | |
5799 | } | |
5800 | ||
5801 | /* We don't want arbitrary node of the BDS chain to be replaced only the top | |
5802 | * most non filter in order to prevent data corruption. | |
5803 | * Another benefit is that this tests exclude backing files which are | |
5804 | * blocked by the backing blockers. | |
5805 | */ | |
5806 | if (!bdrv_is_first_non_filter(to_replace_bs)) { | |
5807 | error_setg(errp, "Only top most non filter can be replaced"); | |
5808 | return NULL; | |
5809 | } | |
5810 | ||
5811 | return to_replace_bs; | |
5812 | } | |
448ad91d ML |
5813 | |
5814 | void bdrv_io_plug(BlockDriverState *bs) | |
5815 | { | |
5816 | BlockDriver *drv = bs->drv; | |
5817 | if (drv && drv->bdrv_io_plug) { | |
5818 | drv->bdrv_io_plug(bs); | |
5819 | } else if (bs->file) { | |
5820 | bdrv_io_plug(bs->file); | |
5821 | } | |
5822 | } | |
5823 | ||
5824 | void bdrv_io_unplug(BlockDriverState *bs) | |
5825 | { | |
5826 | BlockDriver *drv = bs->drv; | |
5827 | if (drv && drv->bdrv_io_unplug) { | |
5828 | drv->bdrv_io_unplug(bs); | |
5829 | } else if (bs->file) { | |
5830 | bdrv_io_unplug(bs->file); | |
5831 | } | |
5832 | } | |
5833 | ||
5834 | void bdrv_flush_io_queue(BlockDriverState *bs) | |
5835 | { | |
5836 | BlockDriver *drv = bs->drv; | |
5837 | if (drv && drv->bdrv_flush_io_queue) { | |
5838 | drv->bdrv_flush_io_queue(bs); | |
5839 | } else if (bs->file) { | |
5840 | bdrv_flush_io_queue(bs->file); | |
5841 | } | |
5842 | } |