]>
Commit | Line | Data |
---|---|---|
fc01f7e7 FB |
1 | /* |
2 | * QEMU System Emulator block driver | |
5fafdf24 | 3 | * |
fc01f7e7 | 4 | * Copyright (c) 2003 Fabrice Bellard |
5fafdf24 | 5 | * |
fc01f7e7 FB |
6 | * Permission is hereby granted, free of charge, to any person obtaining a copy |
7 | * of this software and associated documentation files (the "Software"), to deal | |
8 | * in the Software without restriction, including without limitation the rights | |
9 | * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell | |
10 | * copies of the Software, and to permit persons to whom the Software is | |
11 | * furnished to do so, subject to the following conditions: | |
12 | * | |
13 | * The above copyright notice and this permission notice shall be included in | |
14 | * all copies or substantial portions of the Software. | |
15 | * | |
16 | * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR | |
17 | * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, | |
18 | * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL | |
19 | * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER | |
20 | * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, | |
21 | * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN | |
22 | * THE SOFTWARE. | |
23 | */ | |
3990d09a | 24 | #include "config-host.h" |
faf07963 | 25 | #include "qemu-common.h" |
6d519a5f | 26 | #include "trace.h" |
83c9089e | 27 | #include "monitor/monitor.h" |
737e150e PB |
28 | #include "block/block_int.h" |
29 | #include "block/blockjob.h" | |
1de7afc9 | 30 | #include "qemu/module.h" |
7b1b5d19 | 31 | #include "qapi/qmp/qjson.h" |
9c17d615 | 32 | #include "sysemu/sysemu.h" |
1de7afc9 | 33 | #include "qemu/notify.h" |
737e150e | 34 | #include "block/coroutine.h" |
b2023818 | 35 | #include "qmp-commands.h" |
1de7afc9 | 36 | #include "qemu/timer.h" |
fc01f7e7 | 37 | |
71e72a19 | 38 | #ifdef CONFIG_BSD |
7674e7bf FB |
39 | #include <sys/types.h> |
40 | #include <sys/stat.h> | |
41 | #include <sys/ioctl.h> | |
72cf2d4f | 42 | #include <sys/queue.h> |
c5e97233 | 43 | #ifndef __DragonFly__ |
7674e7bf FB |
44 | #include <sys/disk.h> |
45 | #endif | |
c5e97233 | 46 | #endif |
7674e7bf | 47 | |
49dc768d AL |
48 | #ifdef _WIN32 |
49 | #include <windows.h> | |
50 | #endif | |
51 | ||
1c9805a3 SH |
52 | #define NOT_DONE 0x7fffffff /* used while emulated sync operation in progress */ |
53 | ||
470c0504 SH |
54 | typedef enum { |
55 | BDRV_REQ_COPY_ON_READ = 0x1, | |
f08f2dda | 56 | BDRV_REQ_ZERO_WRITE = 0x2, |
470c0504 SH |
57 | } BdrvRequestFlags; |
58 | ||
7d4b4ba5 | 59 | static void bdrv_dev_change_media_cb(BlockDriverState *bs, bool load); |
f141eafe AL |
60 | static BlockDriverAIOCB *bdrv_aio_readv_em(BlockDriverState *bs, |
61 | int64_t sector_num, QEMUIOVector *qiov, int nb_sectors, | |
c87c0672 | 62 | BlockDriverCompletionFunc *cb, void *opaque); |
f141eafe AL |
63 | static BlockDriverAIOCB *bdrv_aio_writev_em(BlockDriverState *bs, |
64 | int64_t sector_num, QEMUIOVector *qiov, int nb_sectors, | |
ce1a14dc | 65 | BlockDriverCompletionFunc *cb, void *opaque); |
f9f05dc5 KW |
66 | static int coroutine_fn bdrv_co_readv_em(BlockDriverState *bs, |
67 | int64_t sector_num, int nb_sectors, | |
68 | QEMUIOVector *iov); | |
69 | static int coroutine_fn bdrv_co_writev_em(BlockDriverState *bs, | |
70 | int64_t sector_num, int nb_sectors, | |
71 | QEMUIOVector *iov); | |
c5fbe571 | 72 | static int coroutine_fn bdrv_co_do_readv(BlockDriverState *bs, |
470c0504 SH |
73 | int64_t sector_num, int nb_sectors, QEMUIOVector *qiov, |
74 | BdrvRequestFlags flags); | |
1c9805a3 | 75 | static int coroutine_fn bdrv_co_do_writev(BlockDriverState *bs, |
f08f2dda SH |
76 | int64_t sector_num, int nb_sectors, QEMUIOVector *qiov, |
77 | BdrvRequestFlags flags); | |
b2a61371 SH |
78 | static BlockDriverAIOCB *bdrv_co_aio_rw_vector(BlockDriverState *bs, |
79 | int64_t sector_num, | |
80 | QEMUIOVector *qiov, | |
81 | int nb_sectors, | |
82 | BlockDriverCompletionFunc *cb, | |
83 | void *opaque, | |
8c5873d6 | 84 | bool is_write); |
b2a61371 | 85 | static void coroutine_fn bdrv_co_do_rw(void *opaque); |
621f0589 KW |
86 | static int coroutine_fn bdrv_co_do_write_zeroes(BlockDriverState *bs, |
87 | int64_t sector_num, int nb_sectors); | |
ec530c81 | 88 | |
1b7bdbc1 SH |
89 | static QTAILQ_HEAD(, BlockDriverState) bdrv_states = |
90 | QTAILQ_HEAD_INITIALIZER(bdrv_states); | |
7ee930d0 | 91 | |
8a22f02a SH |
92 | static QLIST_HEAD(, BlockDriver) bdrv_drivers = |
93 | QLIST_HEAD_INITIALIZER(bdrv_drivers); | |
ea2384d3 | 94 | |
eb852011 MA |
95 | /* If non-zero, use only whitelisted block drivers */ |
96 | static int use_bdrv_whitelist; | |
97 | ||
9e0b22f4 SH |
98 | #ifdef _WIN32 |
99 | static int is_windows_drive_prefix(const char *filename) | |
100 | { | |
101 | return (((filename[0] >= 'a' && filename[0] <= 'z') || | |
102 | (filename[0] >= 'A' && filename[0] <= 'Z')) && | |
103 | filename[1] == ':'); | |
104 | } | |
105 | ||
106 | int is_windows_drive(const char *filename) | |
107 | { | |
108 | if (is_windows_drive_prefix(filename) && | |
109 | filename[2] == '\0') | |
110 | return 1; | |
111 | if (strstart(filename, "\\\\.\\", NULL) || | |
112 | strstart(filename, "//./", NULL)) | |
113 | return 1; | |
114 | return 0; | |
115 | } | |
116 | #endif | |
117 | ||
0563e191 | 118 | /* throttling disk I/O limits */ |
cc0681c4 BC |
119 | void bdrv_set_io_limits(BlockDriverState *bs, |
120 | ThrottleConfig *cfg) | |
98f90dba | 121 | { |
cc0681c4 | 122 | int i; |
98f90dba | 123 | |
cc0681c4 | 124 | throttle_config(&bs->throttle_state, cfg); |
98f90dba | 125 | |
cc0681c4 BC |
126 | for (i = 0; i < 2; i++) { |
127 | qemu_co_enter_next(&bs->throttled_reqs[i]); | |
98f90dba | 128 | } |
cc0681c4 BC |
129 | } |
130 | ||
131 | /* this function drain all the throttled IOs */ | |
132 | static bool bdrv_start_throttled_reqs(BlockDriverState *bs) | |
133 | { | |
134 | bool drained = false; | |
135 | bool enabled = bs->io_limits_enabled; | |
136 | int i; | |
137 | ||
138 | bs->io_limits_enabled = false; | |
139 | ||
140 | for (i = 0; i < 2; i++) { | |
141 | while (qemu_co_enter_next(&bs->throttled_reqs[i])) { | |
142 | drained = true; | |
143 | } | |
144 | } | |
145 | ||
146 | bs->io_limits_enabled = enabled; | |
98f90dba | 147 | |
cc0681c4 | 148 | return drained; |
98f90dba ZYW |
149 | } |
150 | ||
cc0681c4 | 151 | void bdrv_io_limits_disable(BlockDriverState *bs) |
0563e191 | 152 | { |
cc0681c4 | 153 | bs->io_limits_enabled = false; |
0563e191 | 154 | |
cc0681c4 BC |
155 | bdrv_start_throttled_reqs(bs); |
156 | ||
157 | throttle_destroy(&bs->throttle_state); | |
0563e191 ZYW |
158 | } |
159 | ||
cc0681c4 | 160 | static void bdrv_throttle_read_timer_cb(void *opaque) |
0563e191 | 161 | { |
cc0681c4 BC |
162 | BlockDriverState *bs = opaque; |
163 | qemu_co_enter_next(&bs->throttled_reqs[0]); | |
0563e191 ZYW |
164 | } |
165 | ||
cc0681c4 | 166 | static void bdrv_throttle_write_timer_cb(void *opaque) |
0563e191 | 167 | { |
cc0681c4 BC |
168 | BlockDriverState *bs = opaque; |
169 | qemu_co_enter_next(&bs->throttled_reqs[1]); | |
0563e191 ZYW |
170 | } |
171 | ||
cc0681c4 BC |
172 | /* should be called before bdrv_set_io_limits if a limit is set */ |
173 | void bdrv_io_limits_enable(BlockDriverState *bs) | |
174 | { | |
175 | assert(!bs->io_limits_enabled); | |
176 | throttle_init(&bs->throttle_state, | |
177 | QEMU_CLOCK_VIRTUAL, | |
178 | bdrv_throttle_read_timer_cb, | |
179 | bdrv_throttle_write_timer_cb, | |
180 | bs); | |
181 | bs->io_limits_enabled = true; | |
182 | } | |
183 | ||
184 | /* This function makes an IO wait if needed | |
185 | * | |
186 | * @nb_sectors: the number of sectors of the IO | |
187 | * @is_write: is the IO a write | |
188 | */ | |
98f90dba | 189 | static void bdrv_io_limits_intercept(BlockDriverState *bs, |
cc0681c4 BC |
190 | int nb_sectors, |
191 | bool is_write) | |
98f90dba | 192 | { |
cc0681c4 BC |
193 | /* does this io must wait */ |
194 | bool must_wait = throttle_schedule_timer(&bs->throttle_state, is_write); | |
98f90dba | 195 | |
cc0681c4 BC |
196 | /* if must wait or any request of this type throttled queue the IO */ |
197 | if (must_wait || | |
198 | !qemu_co_queue_empty(&bs->throttled_reqs[is_write])) { | |
199 | qemu_co_queue_wait(&bs->throttled_reqs[is_write]); | |
98f90dba ZYW |
200 | } |
201 | ||
cc0681c4 BC |
202 | /* the IO will be executed, do the accounting */ |
203 | throttle_account(&bs->throttle_state, | |
204 | is_write, | |
205 | nb_sectors * BDRV_SECTOR_SIZE); | |
98f90dba | 206 | |
cc0681c4 BC |
207 | /* if the next request must wait -> do nothing */ |
208 | if (throttle_schedule_timer(&bs->throttle_state, is_write)) { | |
209 | return; | |
98f90dba ZYW |
210 | } |
211 | ||
cc0681c4 BC |
212 | /* else queue next request for execution */ |
213 | qemu_co_queue_next(&bs->throttled_reqs[is_write]); | |
98f90dba ZYW |
214 | } |
215 | ||
9e0b22f4 SH |
216 | /* check if the path starts with "<protocol>:" */ |
217 | static int path_has_protocol(const char *path) | |
218 | { | |
947995c0 PB |
219 | const char *p; |
220 | ||
9e0b22f4 SH |
221 | #ifdef _WIN32 |
222 | if (is_windows_drive(path) || | |
223 | is_windows_drive_prefix(path)) { | |
224 | return 0; | |
225 | } | |
947995c0 PB |
226 | p = path + strcspn(path, ":/\\"); |
227 | #else | |
228 | p = path + strcspn(path, ":/"); | |
9e0b22f4 SH |
229 | #endif |
230 | ||
947995c0 | 231 | return *p == ':'; |
9e0b22f4 SH |
232 | } |
233 | ||
83f64091 | 234 | int path_is_absolute(const char *path) |
3b0d4f61 | 235 | { |
21664424 FB |
236 | #ifdef _WIN32 |
237 | /* specific case for names like: "\\.\d:" */ | |
f53f4da9 | 238 | if (is_windows_drive(path) || is_windows_drive_prefix(path)) { |
21664424 | 239 | return 1; |
f53f4da9 PB |
240 | } |
241 | return (*path == '/' || *path == '\\'); | |
3b9f94e1 | 242 | #else |
f53f4da9 | 243 | return (*path == '/'); |
3b9f94e1 | 244 | #endif |
3b0d4f61 FB |
245 | } |
246 | ||
83f64091 FB |
247 | /* if filename is absolute, just copy it to dest. Otherwise, build a |
248 | path to it by considering it is relative to base_path. URL are | |
249 | supported. */ | |
250 | void path_combine(char *dest, int dest_size, | |
251 | const char *base_path, | |
252 | const char *filename) | |
3b0d4f61 | 253 | { |
83f64091 FB |
254 | const char *p, *p1; |
255 | int len; | |
256 | ||
257 | if (dest_size <= 0) | |
258 | return; | |
259 | if (path_is_absolute(filename)) { | |
260 | pstrcpy(dest, dest_size, filename); | |
261 | } else { | |
262 | p = strchr(base_path, ':'); | |
263 | if (p) | |
264 | p++; | |
265 | else | |
266 | p = base_path; | |
3b9f94e1 FB |
267 | p1 = strrchr(base_path, '/'); |
268 | #ifdef _WIN32 | |
269 | { | |
270 | const char *p2; | |
271 | p2 = strrchr(base_path, '\\'); | |
272 | if (!p1 || p2 > p1) | |
273 | p1 = p2; | |
274 | } | |
275 | #endif | |
83f64091 FB |
276 | if (p1) |
277 | p1++; | |
278 | else | |
279 | p1 = base_path; | |
280 | if (p1 > p) | |
281 | p = p1; | |
282 | len = p - base_path; | |
283 | if (len > dest_size - 1) | |
284 | len = dest_size - 1; | |
285 | memcpy(dest, base_path, len); | |
286 | dest[len] = '\0'; | |
287 | pstrcat(dest, dest_size, filename); | |
3b0d4f61 | 288 | } |
3b0d4f61 FB |
289 | } |
290 | ||
dc5a1371 PB |
291 | void bdrv_get_full_backing_filename(BlockDriverState *bs, char *dest, size_t sz) |
292 | { | |
293 | if (bs->backing_file[0] == '\0' || path_has_protocol(bs->backing_file)) { | |
294 | pstrcpy(dest, sz, bs->backing_file); | |
295 | } else { | |
296 | path_combine(dest, sz, bs->filename, bs->backing_file); | |
297 | } | |
298 | } | |
299 | ||
5efa9d5a | 300 | void bdrv_register(BlockDriver *bdrv) |
ea2384d3 | 301 | { |
8c5873d6 SH |
302 | /* Block drivers without coroutine functions need emulation */ |
303 | if (!bdrv->bdrv_co_readv) { | |
f9f05dc5 KW |
304 | bdrv->bdrv_co_readv = bdrv_co_readv_em; |
305 | bdrv->bdrv_co_writev = bdrv_co_writev_em; | |
306 | ||
f8c35c1d SH |
307 | /* bdrv_co_readv_em()/brdv_co_writev_em() work in terms of aio, so if |
308 | * the block driver lacks aio we need to emulate that too. | |
309 | */ | |
f9f05dc5 KW |
310 | if (!bdrv->bdrv_aio_readv) { |
311 | /* add AIO emulation layer */ | |
312 | bdrv->bdrv_aio_readv = bdrv_aio_readv_em; | |
313 | bdrv->bdrv_aio_writev = bdrv_aio_writev_em; | |
f9f05dc5 | 314 | } |
83f64091 | 315 | } |
b2e12bc6 | 316 | |
8a22f02a | 317 | QLIST_INSERT_HEAD(&bdrv_drivers, bdrv, list); |
ea2384d3 | 318 | } |
b338082b FB |
319 | |
320 | /* create a new block device (by default it is empty) */ | |
321 | BlockDriverState *bdrv_new(const char *device_name) | |
322 | { | |
1b7bdbc1 | 323 | BlockDriverState *bs; |
b338082b | 324 | |
7267c094 | 325 | bs = g_malloc0(sizeof(BlockDriverState)); |
b338082b | 326 | pstrcpy(bs->device_name, sizeof(bs->device_name), device_name); |
ea2384d3 | 327 | if (device_name[0] != '\0') { |
1b7bdbc1 | 328 | QTAILQ_INSERT_TAIL(&bdrv_states, bs, list); |
ea2384d3 | 329 | } |
28a7282a | 330 | bdrv_iostatus_disable(bs); |
d7d512f6 | 331 | notifier_list_init(&bs->close_notifiers); |
d616b224 | 332 | notifier_with_return_list_init(&bs->before_write_notifiers); |
cc0681c4 BC |
333 | qemu_co_queue_init(&bs->throttled_reqs[0]); |
334 | qemu_co_queue_init(&bs->throttled_reqs[1]); | |
9fcb0251 | 335 | bs->refcnt = 1; |
d7d512f6 | 336 | |
b338082b FB |
337 | return bs; |
338 | } | |
339 | ||
d7d512f6 PB |
340 | void bdrv_add_close_notifier(BlockDriverState *bs, Notifier *notify) |
341 | { | |
342 | notifier_list_add(&bs->close_notifiers, notify); | |
343 | } | |
344 | ||
ea2384d3 FB |
345 | BlockDriver *bdrv_find_format(const char *format_name) |
346 | { | |
347 | BlockDriver *drv1; | |
8a22f02a SH |
348 | QLIST_FOREACH(drv1, &bdrv_drivers, list) { |
349 | if (!strcmp(drv1->format_name, format_name)) { | |
ea2384d3 | 350 | return drv1; |
8a22f02a | 351 | } |
ea2384d3 FB |
352 | } |
353 | return NULL; | |
354 | } | |
355 | ||
b64ec4e4 | 356 | static int bdrv_is_whitelisted(BlockDriver *drv, bool read_only) |
eb852011 | 357 | { |
b64ec4e4 FZ |
358 | static const char *whitelist_rw[] = { |
359 | CONFIG_BDRV_RW_WHITELIST | |
360 | }; | |
361 | static const char *whitelist_ro[] = { | |
362 | CONFIG_BDRV_RO_WHITELIST | |
eb852011 MA |
363 | }; |
364 | const char **p; | |
365 | ||
b64ec4e4 | 366 | if (!whitelist_rw[0] && !whitelist_ro[0]) { |
eb852011 | 367 | return 1; /* no whitelist, anything goes */ |
b64ec4e4 | 368 | } |
eb852011 | 369 | |
b64ec4e4 | 370 | for (p = whitelist_rw; *p; p++) { |
eb852011 MA |
371 | if (!strcmp(drv->format_name, *p)) { |
372 | return 1; | |
373 | } | |
374 | } | |
b64ec4e4 FZ |
375 | if (read_only) { |
376 | for (p = whitelist_ro; *p; p++) { | |
377 | if (!strcmp(drv->format_name, *p)) { | |
378 | return 1; | |
379 | } | |
380 | } | |
381 | } | |
eb852011 MA |
382 | return 0; |
383 | } | |
384 | ||
b64ec4e4 FZ |
385 | BlockDriver *bdrv_find_whitelisted_format(const char *format_name, |
386 | bool read_only) | |
eb852011 MA |
387 | { |
388 | BlockDriver *drv = bdrv_find_format(format_name); | |
b64ec4e4 | 389 | return drv && bdrv_is_whitelisted(drv, read_only) ? drv : NULL; |
eb852011 MA |
390 | } |
391 | ||
5b7e1542 ZYW |
392 | typedef struct CreateCo { |
393 | BlockDriver *drv; | |
394 | char *filename; | |
395 | QEMUOptionParameter *options; | |
396 | int ret; | |
cc84d90f | 397 | Error *err; |
5b7e1542 ZYW |
398 | } CreateCo; |
399 | ||
400 | static void coroutine_fn bdrv_create_co_entry(void *opaque) | |
401 | { | |
cc84d90f MR |
402 | Error *local_err = NULL; |
403 | int ret; | |
404 | ||
5b7e1542 ZYW |
405 | CreateCo *cco = opaque; |
406 | assert(cco->drv); | |
407 | ||
cc84d90f MR |
408 | ret = cco->drv->bdrv_create(cco->filename, cco->options, &local_err); |
409 | if (error_is_set(&local_err)) { | |
410 | error_propagate(&cco->err, local_err); | |
411 | } | |
412 | cco->ret = ret; | |
5b7e1542 ZYW |
413 | } |
414 | ||
0e7e1989 | 415 | int bdrv_create(BlockDriver *drv, const char* filename, |
cc84d90f | 416 | QEMUOptionParameter *options, Error **errp) |
ea2384d3 | 417 | { |
5b7e1542 ZYW |
418 | int ret; |
419 | ||
420 | Coroutine *co; | |
421 | CreateCo cco = { | |
422 | .drv = drv, | |
423 | .filename = g_strdup(filename), | |
424 | .options = options, | |
425 | .ret = NOT_DONE, | |
cc84d90f | 426 | .err = NULL, |
5b7e1542 ZYW |
427 | }; |
428 | ||
429 | if (!drv->bdrv_create) { | |
cc84d90f | 430 | error_setg(errp, "Driver '%s' does not support image creation", drv->format_name); |
80168bff LC |
431 | ret = -ENOTSUP; |
432 | goto out; | |
5b7e1542 ZYW |
433 | } |
434 | ||
435 | if (qemu_in_coroutine()) { | |
436 | /* Fast-path if already in coroutine context */ | |
437 | bdrv_create_co_entry(&cco); | |
438 | } else { | |
439 | co = qemu_coroutine_create(bdrv_create_co_entry); | |
440 | qemu_coroutine_enter(co, &cco); | |
441 | while (cco.ret == NOT_DONE) { | |
442 | qemu_aio_wait(); | |
443 | } | |
444 | } | |
445 | ||
446 | ret = cco.ret; | |
cc84d90f MR |
447 | if (ret < 0) { |
448 | if (error_is_set(&cco.err)) { | |
449 | error_propagate(errp, cco.err); | |
450 | } else { | |
451 | error_setg_errno(errp, -ret, "Could not create image"); | |
452 | } | |
453 | } | |
0e7e1989 | 454 | |
80168bff LC |
455 | out: |
456 | g_free(cco.filename); | |
5b7e1542 | 457 | return ret; |
ea2384d3 FB |
458 | } |
459 | ||
cc84d90f MR |
460 | int bdrv_create_file(const char* filename, QEMUOptionParameter *options, |
461 | Error **errp) | |
84a12e66 CH |
462 | { |
463 | BlockDriver *drv; | |
cc84d90f MR |
464 | Error *local_err = NULL; |
465 | int ret; | |
84a12e66 | 466 | |
98289620 | 467 | drv = bdrv_find_protocol(filename, true); |
84a12e66 | 468 | if (drv == NULL) { |
cc84d90f | 469 | error_setg(errp, "Could not find protocol for file '%s'", filename); |
16905d71 | 470 | return -ENOENT; |
84a12e66 CH |
471 | } |
472 | ||
cc84d90f MR |
473 | ret = bdrv_create(drv, filename, options, &local_err); |
474 | if (error_is_set(&local_err)) { | |
475 | error_propagate(errp, local_err); | |
476 | } | |
477 | return ret; | |
84a12e66 CH |
478 | } |
479 | ||
eba25057 JM |
480 | /* |
481 | * Create a uniquely-named empty temporary file. | |
482 | * Return 0 upon success, otherwise a negative errno value. | |
483 | */ | |
484 | int get_tmp_filename(char *filename, int size) | |
d5249393 | 485 | { |
eba25057 | 486 | #ifdef _WIN32 |
3b9f94e1 | 487 | char temp_dir[MAX_PATH]; |
eba25057 JM |
488 | /* GetTempFileName requires that its output buffer (4th param) |
489 | have length MAX_PATH or greater. */ | |
490 | assert(size >= MAX_PATH); | |
491 | return (GetTempPath(MAX_PATH, temp_dir) | |
492 | && GetTempFileName(temp_dir, "qem", 0, filename) | |
493 | ? 0 : -GetLastError()); | |
d5249393 | 494 | #else |
67b915a5 | 495 | int fd; |
7ccfb2eb | 496 | const char *tmpdir; |
0badc1ee AJ |
497 | tmpdir = getenv("TMPDIR"); |
498 | if (!tmpdir) | |
499 | tmpdir = "/tmp"; | |
eba25057 JM |
500 | if (snprintf(filename, size, "%s/vl.XXXXXX", tmpdir) >= size) { |
501 | return -EOVERFLOW; | |
502 | } | |
ea2384d3 | 503 | fd = mkstemp(filename); |
fe235a06 DH |
504 | if (fd < 0) { |
505 | return -errno; | |
506 | } | |
507 | if (close(fd) != 0) { | |
508 | unlink(filename); | |
eba25057 JM |
509 | return -errno; |
510 | } | |
511 | return 0; | |
d5249393 | 512 | #endif |
eba25057 | 513 | } |
fc01f7e7 | 514 | |
84a12e66 CH |
515 | /* |
516 | * Detect host devices. By convention, /dev/cdrom[N] is always | |
517 | * recognized as a host CDROM. | |
518 | */ | |
519 | static BlockDriver *find_hdev_driver(const char *filename) | |
520 | { | |
521 | int score_max = 0, score; | |
522 | BlockDriver *drv = NULL, *d; | |
523 | ||
524 | QLIST_FOREACH(d, &bdrv_drivers, list) { | |
525 | if (d->bdrv_probe_device) { | |
526 | score = d->bdrv_probe_device(filename); | |
527 | if (score > score_max) { | |
528 | score_max = score; | |
529 | drv = d; | |
530 | } | |
531 | } | |
532 | } | |
533 | ||
534 | return drv; | |
535 | } | |
536 | ||
98289620 KW |
537 | BlockDriver *bdrv_find_protocol(const char *filename, |
538 | bool allow_protocol_prefix) | |
83f64091 FB |
539 | { |
540 | BlockDriver *drv1; | |
541 | char protocol[128]; | |
1cec71e3 | 542 | int len; |
83f64091 | 543 | const char *p; |
19cb3738 | 544 | |
66f82cee KW |
545 | /* TODO Drivers without bdrv_file_open must be specified explicitly */ |
546 | ||
39508e7a CH |
547 | /* |
548 | * XXX(hch): we really should not let host device detection | |
549 | * override an explicit protocol specification, but moving this | |
550 | * later breaks access to device names with colons in them. | |
551 | * Thanks to the brain-dead persistent naming schemes on udev- | |
552 | * based Linux systems those actually are quite common. | |
553 | */ | |
554 | drv1 = find_hdev_driver(filename); | |
555 | if (drv1) { | |
556 | return drv1; | |
557 | } | |
558 | ||
98289620 | 559 | if (!path_has_protocol(filename) || !allow_protocol_prefix) { |
39508e7a | 560 | return bdrv_find_format("file"); |
84a12e66 | 561 | } |
98289620 | 562 | |
9e0b22f4 SH |
563 | p = strchr(filename, ':'); |
564 | assert(p != NULL); | |
1cec71e3 AL |
565 | len = p - filename; |
566 | if (len > sizeof(protocol) - 1) | |
567 | len = sizeof(protocol) - 1; | |
568 | memcpy(protocol, filename, len); | |
569 | protocol[len] = '\0'; | |
8a22f02a | 570 | QLIST_FOREACH(drv1, &bdrv_drivers, list) { |
5fafdf24 | 571 | if (drv1->protocol_name && |
8a22f02a | 572 | !strcmp(drv1->protocol_name, protocol)) { |
83f64091 | 573 | return drv1; |
8a22f02a | 574 | } |
83f64091 FB |
575 | } |
576 | return NULL; | |
577 | } | |
578 | ||
f500a6d3 | 579 | static int find_image_format(BlockDriverState *bs, const char *filename, |
34b5d2c6 | 580 | BlockDriver **pdrv, Error **errp) |
f3a5d3f8 | 581 | { |
f500a6d3 | 582 | int score, score_max; |
f3a5d3f8 CH |
583 | BlockDriver *drv1, *drv; |
584 | uint8_t buf[2048]; | |
f500a6d3 | 585 | int ret = 0; |
f8ea0b00 | 586 | |
08a00559 | 587 | /* Return the raw BlockDriver * to scsi-generic devices or empty drives */ |
8e895599 | 588 | if (bs->sg || !bdrv_is_inserted(bs) || bdrv_getlength(bs) == 0) { |
c98ac35d SW |
589 | drv = bdrv_find_format("raw"); |
590 | if (!drv) { | |
34b5d2c6 | 591 | error_setg(errp, "Could not find raw image format"); |
c98ac35d SW |
592 | ret = -ENOENT; |
593 | } | |
594 | *pdrv = drv; | |
595 | return ret; | |
1a396859 | 596 | } |
f8ea0b00 | 597 | |
83f64091 | 598 | ret = bdrv_pread(bs, 0, buf, sizeof(buf)); |
83f64091 | 599 | if (ret < 0) { |
34b5d2c6 MR |
600 | error_setg_errno(errp, -ret, "Could not read image for determining its " |
601 | "format"); | |
c98ac35d SW |
602 | *pdrv = NULL; |
603 | return ret; | |
83f64091 FB |
604 | } |
605 | ||
ea2384d3 | 606 | score_max = 0; |
84a12e66 | 607 | drv = NULL; |
8a22f02a | 608 | QLIST_FOREACH(drv1, &bdrv_drivers, list) { |
83f64091 FB |
609 | if (drv1->bdrv_probe) { |
610 | score = drv1->bdrv_probe(buf, ret, filename); | |
611 | if (score > score_max) { | |
612 | score_max = score; | |
613 | drv = drv1; | |
614 | } | |
0849bf08 | 615 | } |
fc01f7e7 | 616 | } |
c98ac35d | 617 | if (!drv) { |
34b5d2c6 MR |
618 | error_setg(errp, "Could not determine image format: No compatible " |
619 | "driver found"); | |
c98ac35d SW |
620 | ret = -ENOENT; |
621 | } | |
622 | *pdrv = drv; | |
623 | return ret; | |
ea2384d3 FB |
624 | } |
625 | ||
51762288 SH |
626 | /** |
627 | * Set the current 'total_sectors' value | |
628 | */ | |
629 | static int refresh_total_sectors(BlockDriverState *bs, int64_t hint) | |
630 | { | |
631 | BlockDriver *drv = bs->drv; | |
632 | ||
396759ad NB |
633 | /* Do not attempt drv->bdrv_getlength() on scsi-generic devices */ |
634 | if (bs->sg) | |
635 | return 0; | |
636 | ||
51762288 SH |
637 | /* query actual device if possible, otherwise just trust the hint */ |
638 | if (drv->bdrv_getlength) { | |
639 | int64_t length = drv->bdrv_getlength(bs); | |
640 | if (length < 0) { | |
641 | return length; | |
642 | } | |
643 | hint = length >> BDRV_SECTOR_BITS; | |
644 | } | |
645 | ||
646 | bs->total_sectors = hint; | |
647 | return 0; | |
648 | } | |
649 | ||
9e8f1835 PB |
650 | /** |
651 | * Set open flags for a given discard mode | |
652 | * | |
653 | * Return 0 on success, -1 if the discard mode was invalid. | |
654 | */ | |
655 | int bdrv_parse_discard_flags(const char *mode, int *flags) | |
656 | { | |
657 | *flags &= ~BDRV_O_UNMAP; | |
658 | ||
659 | if (!strcmp(mode, "off") || !strcmp(mode, "ignore")) { | |
660 | /* do nothing */ | |
661 | } else if (!strcmp(mode, "on") || !strcmp(mode, "unmap")) { | |
662 | *flags |= BDRV_O_UNMAP; | |
663 | } else { | |
664 | return -1; | |
665 | } | |
666 | ||
667 | return 0; | |
668 | } | |
669 | ||
c3993cdc SH |
670 | /** |
671 | * Set open flags for a given cache mode | |
672 | * | |
673 | * Return 0 on success, -1 if the cache mode was invalid. | |
674 | */ | |
675 | int bdrv_parse_cache_flags(const char *mode, int *flags) | |
676 | { | |
677 | *flags &= ~BDRV_O_CACHE_MASK; | |
678 | ||
679 | if (!strcmp(mode, "off") || !strcmp(mode, "none")) { | |
680 | *flags |= BDRV_O_NOCACHE | BDRV_O_CACHE_WB; | |
92196b2f SH |
681 | } else if (!strcmp(mode, "directsync")) { |
682 | *flags |= BDRV_O_NOCACHE; | |
c3993cdc SH |
683 | } else if (!strcmp(mode, "writeback")) { |
684 | *flags |= BDRV_O_CACHE_WB; | |
685 | } else if (!strcmp(mode, "unsafe")) { | |
686 | *flags |= BDRV_O_CACHE_WB; | |
687 | *flags |= BDRV_O_NO_FLUSH; | |
688 | } else if (!strcmp(mode, "writethrough")) { | |
689 | /* this is the default */ | |
690 | } else { | |
691 | return -1; | |
692 | } | |
693 | ||
694 | return 0; | |
695 | } | |
696 | ||
53fec9d3 SH |
697 | /** |
698 | * The copy-on-read flag is actually a reference count so multiple users may | |
699 | * use the feature without worrying about clobbering its previous state. | |
700 | * Copy-on-read stays enabled until all users have called to disable it. | |
701 | */ | |
702 | void bdrv_enable_copy_on_read(BlockDriverState *bs) | |
703 | { | |
704 | bs->copy_on_read++; | |
705 | } | |
706 | ||
707 | void bdrv_disable_copy_on_read(BlockDriverState *bs) | |
708 | { | |
709 | assert(bs->copy_on_read > 0); | |
710 | bs->copy_on_read--; | |
711 | } | |
712 | ||
7b272452 KW |
713 | static int bdrv_open_flags(BlockDriverState *bs, int flags) |
714 | { | |
715 | int open_flags = flags | BDRV_O_CACHE_WB; | |
716 | ||
717 | /* | |
718 | * Clear flags that are internal to the block layer before opening the | |
719 | * image. | |
720 | */ | |
721 | open_flags &= ~(BDRV_O_SNAPSHOT | BDRV_O_NO_BACKING); | |
722 | ||
723 | /* | |
724 | * Snapshots should be writable. | |
725 | */ | |
726 | if (bs->is_temporary) { | |
727 | open_flags |= BDRV_O_RDWR; | |
728 | } | |
729 | ||
730 | return open_flags; | |
731 | } | |
732 | ||
57915332 KW |
733 | /* |
734 | * Common part for opening disk images and files | |
b6ad491a KW |
735 | * |
736 | * Removes all processed options from *options. | |
57915332 | 737 | */ |
f500a6d3 | 738 | static int bdrv_open_common(BlockDriverState *bs, BlockDriverState *file, |
34b5d2c6 | 739 | QDict *options, int flags, BlockDriver *drv, Error **errp) |
57915332 KW |
740 | { |
741 | int ret, open_flags; | |
035fccdf | 742 | const char *filename; |
34b5d2c6 | 743 | Error *local_err = NULL; |
57915332 KW |
744 | |
745 | assert(drv != NULL); | |
6405875c | 746 | assert(bs->file == NULL); |
707ff828 | 747 | assert(options != NULL && bs->options != options); |
57915332 | 748 | |
45673671 KW |
749 | if (file != NULL) { |
750 | filename = file->filename; | |
751 | } else { | |
752 | filename = qdict_get_try_str(options, "filename"); | |
753 | } | |
754 | ||
755 | trace_bdrv_open_common(bs, filename ?: "", flags, drv->format_name); | |
28dcee10 | 756 | |
5d186eb0 KW |
757 | /* bdrv_open() with directly using a protocol as drv. This layer is already |
758 | * opened, so assign it to bs (while file becomes a closed BlockDriverState) | |
759 | * and return immediately. */ | |
760 | if (file != NULL && drv->bdrv_file_open) { | |
761 | bdrv_swap(file, bs); | |
762 | return 0; | |
763 | } | |
764 | ||
57915332 | 765 | bs->open_flags = flags; |
57915332 | 766 | bs->buffer_alignment = 512; |
0d51b4de | 767 | bs->zero_beyond_eof = true; |
b64ec4e4 FZ |
768 | open_flags = bdrv_open_flags(bs, flags); |
769 | bs->read_only = !(open_flags & BDRV_O_RDWR); | |
770 | ||
771 | if (use_bdrv_whitelist && !bdrv_is_whitelisted(drv, bs->read_only)) { | |
34b5d2c6 | 772 | error_setg(errp, "Driver '%s' is not whitelisted", drv->format_name); |
b64ec4e4 FZ |
773 | return -ENOTSUP; |
774 | } | |
57915332 | 775 | |
53fec9d3 | 776 | assert(bs->copy_on_read == 0); /* bdrv_new() and bdrv_close() make it so */ |
b64ec4e4 | 777 | if (!bs->read_only && (flags & BDRV_O_COPY_ON_READ)) { |
53fec9d3 SH |
778 | bdrv_enable_copy_on_read(bs); |
779 | } | |
780 | ||
c2ad1b0c KW |
781 | if (filename != NULL) { |
782 | pstrcpy(bs->filename, sizeof(bs->filename), filename); | |
783 | } else { | |
784 | bs->filename[0] = '\0'; | |
785 | } | |
57915332 | 786 | |
57915332 | 787 | bs->drv = drv; |
7267c094 | 788 | bs->opaque = g_malloc0(drv->instance_size); |
57915332 | 789 | |
03f541bd | 790 | bs->enable_write_cache = !!(flags & BDRV_O_CACHE_WB); |
e7c63796 | 791 | |
66f82cee KW |
792 | /* Open the image, either directly or using a protocol */ |
793 | if (drv->bdrv_file_open) { | |
5d186eb0 | 794 | assert(file == NULL); |
030be321 | 795 | assert(!drv->bdrv_needs_filename || filename != NULL); |
34b5d2c6 | 796 | ret = drv->bdrv_file_open(bs, options, open_flags, &local_err); |
f500a6d3 | 797 | } else { |
2af5ef70 | 798 | if (file == NULL) { |
34b5d2c6 MR |
799 | error_setg(errp, "Can't use '%s' as a block driver for the " |
800 | "protocol level", drv->format_name); | |
2af5ef70 KW |
801 | ret = -EINVAL; |
802 | goto free_and_fail; | |
803 | } | |
f500a6d3 | 804 | bs->file = file; |
34b5d2c6 | 805 | ret = drv->bdrv_open(bs, options, open_flags, &local_err); |
66f82cee KW |
806 | } |
807 | ||
57915332 | 808 | if (ret < 0) { |
34b5d2c6 MR |
809 | if (error_is_set(&local_err)) { |
810 | error_propagate(errp, local_err); | |
811 | } else if (filename) { | |
812 | error_setg_errno(errp, -ret, "Could not open '%s'", filename); | |
813 | } else { | |
814 | error_setg_errno(errp, -ret, "Could not open image"); | |
815 | } | |
57915332 KW |
816 | goto free_and_fail; |
817 | } | |
818 | ||
51762288 SH |
819 | ret = refresh_total_sectors(bs, bs->total_sectors); |
820 | if (ret < 0) { | |
34b5d2c6 | 821 | error_setg_errno(errp, -ret, "Could not refresh total sector count"); |
51762288 | 822 | goto free_and_fail; |
57915332 | 823 | } |
51762288 | 824 | |
57915332 KW |
825 | #ifndef _WIN32 |
826 | if (bs->is_temporary) { | |
c2ad1b0c | 827 | assert(filename != NULL); |
57915332 KW |
828 | unlink(filename); |
829 | } | |
830 | #endif | |
831 | return 0; | |
832 | ||
833 | free_and_fail: | |
f500a6d3 | 834 | bs->file = NULL; |
7267c094 | 835 | g_free(bs->opaque); |
57915332 KW |
836 | bs->opaque = NULL; |
837 | bs->drv = NULL; | |
838 | return ret; | |
839 | } | |
840 | ||
b6ce07aa KW |
841 | /* |
842 | * Opens a file using a protocol (file, host_device, nbd, ...) | |
787e4a85 KW |
843 | * |
844 | * options is a QDict of options to pass to the block drivers, or NULL for an | |
845 | * empty set of options. The reference to the QDict belongs to the block layer | |
846 | * after the call (even on failure), so if the caller intends to reuse the | |
847 | * dictionary, it needs to use QINCREF() before calling bdrv_file_open. | |
b6ce07aa | 848 | */ |
787e4a85 | 849 | int bdrv_file_open(BlockDriverState **pbs, const char *filename, |
34b5d2c6 | 850 | QDict *options, int flags, Error **errp) |
ea2384d3 | 851 | { |
83f64091 | 852 | BlockDriverState *bs; |
6db95603 | 853 | BlockDriver *drv; |
c2ad1b0c | 854 | const char *drvname; |
98289620 | 855 | bool allow_protocol_prefix = false; |
34b5d2c6 | 856 | Error *local_err = NULL; |
83f64091 FB |
857 | int ret; |
858 | ||
707ff828 KW |
859 | /* NULL means an empty set of options */ |
860 | if (options == NULL) { | |
861 | options = qdict_new(); | |
862 | } | |
863 | ||
83f64091 | 864 | bs = bdrv_new(""); |
707ff828 KW |
865 | bs->options = options; |
866 | options = qdict_clone_shallow(options); | |
867 | ||
035fccdf KW |
868 | /* Fetch the file name from the options QDict if necessary */ |
869 | if (!filename) { | |
870 | filename = qdict_get_try_str(options, "filename"); | |
871 | } else if (filename && !qdict_haskey(options, "filename")) { | |
872 | qdict_put(options, "filename", qstring_from_str(filename)); | |
98289620 | 873 | allow_protocol_prefix = true; |
035fccdf | 874 | } else { |
34b5d2c6 MR |
875 | error_setg(errp, "Can't specify 'file' and 'filename' options at the " |
876 | "same time"); | |
035fccdf KW |
877 | ret = -EINVAL; |
878 | goto fail; | |
879 | } | |
880 | ||
c2ad1b0c KW |
881 | /* Find the right block driver */ |
882 | drvname = qdict_get_try_str(options, "driver"); | |
883 | if (drvname) { | |
b64ec4e4 | 884 | drv = bdrv_find_whitelisted_format(drvname, !(flags & BDRV_O_RDWR)); |
34b5d2c6 MR |
885 | if (!drv) { |
886 | error_setg(errp, "Unknown driver '%s'", drvname); | |
887 | } | |
c2ad1b0c KW |
888 | qdict_del(options, "driver"); |
889 | } else if (filename) { | |
98289620 KW |
890 | drv = bdrv_find_protocol(filename, allow_protocol_prefix); |
891 | if (!drv) { | |
34b5d2c6 | 892 | error_setg(errp, "Unknown protocol"); |
98289620 | 893 | } |
c2ad1b0c | 894 | } else { |
34b5d2c6 | 895 | error_setg(errp, "Must specify either driver or file"); |
c2ad1b0c KW |
896 | drv = NULL; |
897 | } | |
898 | ||
899 | if (!drv) { | |
34b5d2c6 | 900 | /* errp has been set already */ |
c2ad1b0c KW |
901 | ret = -ENOENT; |
902 | goto fail; | |
903 | } | |
904 | ||
905 | /* Parse the filename and open it */ | |
906 | if (drv->bdrv_parse_filename && filename) { | |
6963a30d KW |
907 | drv->bdrv_parse_filename(filename, options, &local_err); |
908 | if (error_is_set(&local_err)) { | |
34b5d2c6 | 909 | error_propagate(errp, local_err); |
6963a30d KW |
910 | ret = -EINVAL; |
911 | goto fail; | |
912 | } | |
56d1b4d2 | 913 | qdict_del(options, "filename"); |
030be321 | 914 | } else if (drv->bdrv_needs_filename && !filename) { |
34b5d2c6 MR |
915 | error_setg(errp, "The '%s' block driver requires a file name", |
916 | drv->format_name); | |
c2ad1b0c KW |
917 | ret = -EINVAL; |
918 | goto fail; | |
6963a30d KW |
919 | } |
920 | ||
34b5d2c6 | 921 | ret = bdrv_open_common(bs, NULL, options, flags, drv, &local_err); |
83f64091 | 922 | if (ret < 0) { |
34b5d2c6 | 923 | error_propagate(errp, local_err); |
707ff828 KW |
924 | goto fail; |
925 | } | |
926 | ||
927 | /* Check if any unknown options were used */ | |
928 | if (qdict_size(options) != 0) { | |
929 | const QDictEntry *entry = qdict_first(options); | |
34b5d2c6 MR |
930 | error_setg(errp, "Block protocol '%s' doesn't support the option '%s'", |
931 | drv->format_name, entry->key); | |
707ff828 KW |
932 | ret = -EINVAL; |
933 | goto fail; | |
3b0d4f61 | 934 | } |
707ff828 KW |
935 | QDECREF(options); |
936 | ||
71d0770c | 937 | bs->growable = 1; |
83f64091 FB |
938 | *pbs = bs; |
939 | return 0; | |
707ff828 KW |
940 | |
941 | fail: | |
942 | QDECREF(options); | |
943 | if (!bs->drv) { | |
944 | QDECREF(bs->options); | |
945 | } | |
4f6fd349 | 946 | bdrv_unref(bs); |
707ff828 | 947 | return ret; |
83f64091 FB |
948 | } |
949 | ||
31ca6d07 KW |
950 | /* |
951 | * Opens the backing file for a BlockDriverState if not yet open | |
952 | * | |
953 | * options is a QDict of options to pass to the block drivers, or NULL for an | |
954 | * empty set of options. The reference to the QDict is transferred to this | |
955 | * function (even on failure), so if the caller intends to reuse the dictionary, | |
956 | * it needs to use QINCREF() before calling bdrv_file_open. | |
957 | */ | |
34b5d2c6 | 958 | int bdrv_open_backing_file(BlockDriverState *bs, QDict *options, Error **errp) |
9156df12 PB |
959 | { |
960 | char backing_filename[PATH_MAX]; | |
961 | int back_flags, ret; | |
962 | BlockDriver *back_drv = NULL; | |
34b5d2c6 | 963 | Error *local_err = NULL; |
9156df12 PB |
964 | |
965 | if (bs->backing_hd != NULL) { | |
31ca6d07 | 966 | QDECREF(options); |
9156df12 PB |
967 | return 0; |
968 | } | |
969 | ||
31ca6d07 KW |
970 | /* NULL means an empty set of options */ |
971 | if (options == NULL) { | |
972 | options = qdict_new(); | |
973 | } | |
974 | ||
9156df12 | 975 | bs->open_flags &= ~BDRV_O_NO_BACKING; |
1cb6f506 KW |
976 | if (qdict_haskey(options, "file.filename")) { |
977 | backing_filename[0] = '\0'; | |
978 | } else if (bs->backing_file[0] == '\0' && qdict_size(options) == 0) { | |
31ca6d07 | 979 | QDECREF(options); |
9156df12 | 980 | return 0; |
dbecebdd FZ |
981 | } else { |
982 | bdrv_get_full_backing_filename(bs, backing_filename, | |
983 | sizeof(backing_filename)); | |
9156df12 PB |
984 | } |
985 | ||
986 | bs->backing_hd = bdrv_new(""); | |
9156df12 PB |
987 | |
988 | if (bs->backing_format[0] != '\0') { | |
989 | back_drv = bdrv_find_format(bs->backing_format); | |
990 | } | |
991 | ||
992 | /* backing files always opened read-only */ | |
993 | back_flags = bs->open_flags & ~(BDRV_O_RDWR | BDRV_O_SNAPSHOT); | |
994 | ||
31ca6d07 KW |
995 | ret = bdrv_open(bs->backing_hd, |
996 | *backing_filename ? backing_filename : NULL, options, | |
34b5d2c6 | 997 | back_flags, back_drv, &local_err); |
dbecebdd FZ |
998 | pstrcpy(bs->backing_file, sizeof(bs->backing_file), |
999 | bs->backing_hd->file->filename); | |
9156df12 | 1000 | if (ret < 0) { |
4f6fd349 | 1001 | bdrv_unref(bs->backing_hd); |
9156df12 PB |
1002 | bs->backing_hd = NULL; |
1003 | bs->open_flags |= BDRV_O_NO_BACKING; | |
34b5d2c6 | 1004 | error_propagate(errp, local_err); |
9156df12 PB |
1005 | return ret; |
1006 | } | |
1007 | return 0; | |
1008 | } | |
1009 | ||
b6ce07aa KW |
1010 | /* |
1011 | * Opens a disk image (raw, qcow2, vmdk, ...) | |
de9c0cec KW |
1012 | * |
1013 | * options is a QDict of options to pass to the block drivers, or NULL for an | |
1014 | * empty set of options. The reference to the QDict belongs to the block layer | |
1015 | * after the call (even on failure), so if the caller intends to reuse the | |
1016 | * dictionary, it needs to use QINCREF() before calling bdrv_open. | |
b6ce07aa | 1017 | */ |
de9c0cec | 1018 | int bdrv_open(BlockDriverState *bs, const char *filename, QDict *options, |
34b5d2c6 | 1019 | int flags, BlockDriver *drv, Error **errp) |
ea2384d3 | 1020 | { |
b6ce07aa | 1021 | int ret; |
89c9bc3d SW |
1022 | /* TODO: extra byte is a hack to ensure MAX_PATH space on Windows. */ |
1023 | char tmp_filename[PATH_MAX + 1]; | |
f500a6d3 | 1024 | BlockDriverState *file = NULL; |
707ff828 | 1025 | QDict *file_options = NULL; |
74fe54f2 | 1026 | const char *drvname; |
34b5d2c6 | 1027 | Error *local_err = NULL; |
712e7874 | 1028 | |
de9c0cec KW |
1029 | /* NULL means an empty set of options */ |
1030 | if (options == NULL) { | |
1031 | options = qdict_new(); | |
1032 | } | |
1033 | ||
1034 | bs->options = options; | |
b6ad491a | 1035 | options = qdict_clone_shallow(options); |
de9c0cec KW |
1036 | |
1037 | /* For snapshot=on, create a temporary qcow2 overlay */ | |
83f64091 | 1038 | if (flags & BDRV_O_SNAPSHOT) { |
ea2384d3 FB |
1039 | BlockDriverState *bs1; |
1040 | int64_t total_size; | |
91a073a9 | 1041 | BlockDriver *bdrv_qcow2; |
08b392e1 | 1042 | QEMUOptionParameter *create_options; |
b6ce07aa | 1043 | char backing_filename[PATH_MAX]; |
3b46e624 | 1044 | |
c2ad1b0c | 1045 | if (qdict_size(options) != 0) { |
34b5d2c6 | 1046 | error_setg(errp, "Can't use snapshot=on with driver-specific options"); |
c2ad1b0c KW |
1047 | ret = -EINVAL; |
1048 | goto fail; | |
1049 | } | |
1050 | assert(filename != NULL); | |
1051 | ||
ea2384d3 FB |
1052 | /* if snapshot, we create a temporary backing file and open it |
1053 | instead of opening 'filename' directly */ | |
33e3963e | 1054 | |
ea2384d3 FB |
1055 | /* if there is a backing file, use it */ |
1056 | bs1 = bdrv_new(""); | |
34b5d2c6 | 1057 | ret = bdrv_open(bs1, filename, NULL, 0, drv, &local_err); |
51d7c00c | 1058 | if (ret < 0) { |
4f6fd349 | 1059 | bdrv_unref(bs1); |
de9c0cec | 1060 | goto fail; |
ea2384d3 | 1061 | } |
3e82990b | 1062 | total_size = bdrv_getlength(bs1) & BDRV_SECTOR_MASK; |
7c96d46e | 1063 | |
4f6fd349 | 1064 | bdrv_unref(bs1); |
3b46e624 | 1065 | |
eba25057 JM |
1066 | ret = get_tmp_filename(tmp_filename, sizeof(tmp_filename)); |
1067 | if (ret < 0) { | |
34b5d2c6 | 1068 | error_setg_errno(errp, -ret, "Could not get temporary filename"); |
de9c0cec | 1069 | goto fail; |
eba25057 | 1070 | } |
7c96d46e AL |
1071 | |
1072 | /* Real path is meaningless for protocols */ | |
4d70655b | 1073 | if (path_has_protocol(filename)) { |
7c96d46e AL |
1074 | snprintf(backing_filename, sizeof(backing_filename), |
1075 | "%s", filename); | |
de9c0cec | 1076 | } else if (!realpath(filename, backing_filename)) { |
34b5d2c6 | 1077 | error_setg_errno(errp, errno, "Could not resolve path '%s'", filename); |
de9c0cec KW |
1078 | ret = -errno; |
1079 | goto fail; | |
1080 | } | |
7c96d46e | 1081 | |
91a073a9 | 1082 | bdrv_qcow2 = bdrv_find_format("qcow2"); |
08b392e1 KW |
1083 | create_options = parse_option_parameters("", bdrv_qcow2->create_options, |
1084 | NULL); | |
91a073a9 | 1085 | |
08b392e1 KW |
1086 | set_option_parameter_int(create_options, BLOCK_OPT_SIZE, total_size); |
1087 | set_option_parameter(create_options, BLOCK_OPT_BACKING_FILE, | |
1088 | backing_filename); | |
91a073a9 | 1089 | if (drv) { |
08b392e1 | 1090 | set_option_parameter(create_options, BLOCK_OPT_BACKING_FMT, |
91a073a9 KW |
1091 | drv->format_name); |
1092 | } | |
1093 | ||
cc84d90f | 1094 | ret = bdrv_create(bdrv_qcow2, tmp_filename, create_options, &local_err); |
08b392e1 | 1095 | free_option_parameters(create_options); |
51d7c00c | 1096 | if (ret < 0) { |
34b5d2c6 | 1097 | error_setg_errno(errp, -ret, "Could not create temporary overlay " |
cc84d90f MR |
1098 | "'%s': %s", tmp_filename, |
1099 | error_get_pretty(local_err)); | |
1100 | error_free(local_err); | |
1101 | local_err = NULL; | |
de9c0cec | 1102 | goto fail; |
ea2384d3 | 1103 | } |
91a073a9 | 1104 | |
ea2384d3 | 1105 | filename = tmp_filename; |
91a073a9 | 1106 | drv = bdrv_qcow2; |
ea2384d3 FB |
1107 | bs->is_temporary = 1; |
1108 | } | |
712e7874 | 1109 | |
f500a6d3 KW |
1110 | /* Open image file without format layer */ |
1111 | if (flags & BDRV_O_RDWR) { | |
1112 | flags |= BDRV_O_ALLOW_RDWR; | |
1113 | } | |
1114 | ||
5726d872 | 1115 | qdict_extract_subqdict(options, &file_options, "file."); |
707ff828 KW |
1116 | |
1117 | ret = bdrv_file_open(&file, filename, file_options, | |
34b5d2c6 | 1118 | bdrv_open_flags(bs, flags | BDRV_O_UNMAP), &local_err); |
f500a6d3 | 1119 | if (ret < 0) { |
de9c0cec | 1120 | goto fail; |
f500a6d3 KW |
1121 | } |
1122 | ||
b6ce07aa | 1123 | /* Find the right image format driver */ |
74fe54f2 KW |
1124 | drvname = qdict_get_try_str(options, "driver"); |
1125 | if (drvname) { | |
1126 | drv = bdrv_find_whitelisted_format(drvname, !(flags & BDRV_O_RDWR)); | |
1127 | qdict_del(options, "driver"); | |
1128 | } | |
1129 | ||
6db95603 | 1130 | if (!drv) { |
34b5d2c6 | 1131 | ret = find_image_format(file, filename, &drv, &local_err); |
51d7c00c | 1132 | } |
6987307c | 1133 | |
51d7c00c | 1134 | if (!drv) { |
51d7c00c | 1135 | goto unlink_and_fail; |
ea2384d3 | 1136 | } |
b6ce07aa KW |
1137 | |
1138 | /* Open the image */ | |
34b5d2c6 | 1139 | ret = bdrv_open_common(bs, file, options, flags, drv, &local_err); |
b6ce07aa | 1140 | if (ret < 0) { |
6987307c CH |
1141 | goto unlink_and_fail; |
1142 | } | |
1143 | ||
f500a6d3 | 1144 | if (bs->file != file) { |
4f6fd349 | 1145 | bdrv_unref(file); |
f500a6d3 KW |
1146 | file = NULL; |
1147 | } | |
1148 | ||
b6ce07aa | 1149 | /* If there is a backing file, use it */ |
9156df12 | 1150 | if ((flags & BDRV_O_NO_BACKING) == 0) { |
31ca6d07 KW |
1151 | QDict *backing_options; |
1152 | ||
5726d872 | 1153 | qdict_extract_subqdict(options, &backing_options, "backing."); |
34b5d2c6 | 1154 | ret = bdrv_open_backing_file(bs, backing_options, &local_err); |
b6ce07aa | 1155 | if (ret < 0) { |
b6ad491a | 1156 | goto close_and_fail; |
b6ce07aa | 1157 | } |
b6ce07aa KW |
1158 | } |
1159 | ||
b6ad491a KW |
1160 | /* Check if any unknown options were used */ |
1161 | if (qdict_size(options) != 0) { | |
1162 | const QDictEntry *entry = qdict_first(options); | |
34b5d2c6 MR |
1163 | error_setg(errp, "Block format '%s' used by device '%s' doesn't " |
1164 | "support the option '%s'", drv->format_name, bs->device_name, | |
1165 | entry->key); | |
b6ad491a KW |
1166 | |
1167 | ret = -EINVAL; | |
1168 | goto close_and_fail; | |
1169 | } | |
1170 | QDECREF(options); | |
1171 | ||
b6ce07aa | 1172 | if (!bdrv_key_required(bs)) { |
7d4b4ba5 | 1173 | bdrv_dev_change_media_cb(bs, true); |
b6ce07aa KW |
1174 | } |
1175 | ||
1176 | return 0; | |
1177 | ||
1178 | unlink_and_fail: | |
f500a6d3 | 1179 | if (file != NULL) { |
4f6fd349 | 1180 | bdrv_unref(file); |
f500a6d3 | 1181 | } |
b6ce07aa KW |
1182 | if (bs->is_temporary) { |
1183 | unlink(filename); | |
1184 | } | |
de9c0cec KW |
1185 | fail: |
1186 | QDECREF(bs->options); | |
b6ad491a | 1187 | QDECREF(options); |
de9c0cec | 1188 | bs->options = NULL; |
34b5d2c6 MR |
1189 | if (error_is_set(&local_err)) { |
1190 | error_propagate(errp, local_err); | |
1191 | } | |
b6ad491a | 1192 | return ret; |
de9c0cec | 1193 | |
b6ad491a KW |
1194 | close_and_fail: |
1195 | bdrv_close(bs); | |
1196 | QDECREF(options); | |
34b5d2c6 MR |
1197 | if (error_is_set(&local_err)) { |
1198 | error_propagate(errp, local_err); | |
1199 | } | |
b6ce07aa KW |
1200 | return ret; |
1201 | } | |
1202 | ||
e971aa12 JC |
1203 | typedef struct BlockReopenQueueEntry { |
1204 | bool prepared; | |
1205 | BDRVReopenState state; | |
1206 | QSIMPLEQ_ENTRY(BlockReopenQueueEntry) entry; | |
1207 | } BlockReopenQueueEntry; | |
1208 | ||
1209 | /* | |
1210 | * Adds a BlockDriverState to a simple queue for an atomic, transactional | |
1211 | * reopen of multiple devices. | |
1212 | * | |
1213 | * bs_queue can either be an existing BlockReopenQueue that has had QSIMPLE_INIT | |
1214 | * already performed, or alternatively may be NULL a new BlockReopenQueue will | |
1215 | * be created and initialized. This newly created BlockReopenQueue should be | |
1216 | * passed back in for subsequent calls that are intended to be of the same | |
1217 | * atomic 'set'. | |
1218 | * | |
1219 | * bs is the BlockDriverState to add to the reopen queue. | |
1220 | * | |
1221 | * flags contains the open flags for the associated bs | |
1222 | * | |
1223 | * returns a pointer to bs_queue, which is either the newly allocated | |
1224 | * bs_queue, or the existing bs_queue being used. | |
1225 | * | |
1226 | */ | |
1227 | BlockReopenQueue *bdrv_reopen_queue(BlockReopenQueue *bs_queue, | |
1228 | BlockDriverState *bs, int flags) | |
1229 | { | |
1230 | assert(bs != NULL); | |
1231 | ||
1232 | BlockReopenQueueEntry *bs_entry; | |
1233 | if (bs_queue == NULL) { | |
1234 | bs_queue = g_new0(BlockReopenQueue, 1); | |
1235 | QSIMPLEQ_INIT(bs_queue); | |
1236 | } | |
1237 | ||
1238 | if (bs->file) { | |
1239 | bdrv_reopen_queue(bs_queue, bs->file, flags); | |
1240 | } | |
1241 | ||
1242 | bs_entry = g_new0(BlockReopenQueueEntry, 1); | |
1243 | QSIMPLEQ_INSERT_TAIL(bs_queue, bs_entry, entry); | |
1244 | ||
1245 | bs_entry->state.bs = bs; | |
1246 | bs_entry->state.flags = flags; | |
1247 | ||
1248 | return bs_queue; | |
1249 | } | |
1250 | ||
1251 | /* | |
1252 | * Reopen multiple BlockDriverStates atomically & transactionally. | |
1253 | * | |
1254 | * The queue passed in (bs_queue) must have been built up previous | |
1255 | * via bdrv_reopen_queue(). | |
1256 | * | |
1257 | * Reopens all BDS specified in the queue, with the appropriate | |
1258 | * flags. All devices are prepared for reopen, and failure of any | |
1259 | * device will cause all device changes to be abandonded, and intermediate | |
1260 | * data cleaned up. | |
1261 | * | |
1262 | * If all devices prepare successfully, then the changes are committed | |
1263 | * to all devices. | |
1264 | * | |
1265 | */ | |
1266 | int bdrv_reopen_multiple(BlockReopenQueue *bs_queue, Error **errp) | |
1267 | { | |
1268 | int ret = -1; | |
1269 | BlockReopenQueueEntry *bs_entry, *next; | |
1270 | Error *local_err = NULL; | |
1271 | ||
1272 | assert(bs_queue != NULL); | |
1273 | ||
1274 | bdrv_drain_all(); | |
1275 | ||
1276 | QSIMPLEQ_FOREACH(bs_entry, bs_queue, entry) { | |
1277 | if (bdrv_reopen_prepare(&bs_entry->state, bs_queue, &local_err)) { | |
1278 | error_propagate(errp, local_err); | |
1279 | goto cleanup; | |
1280 | } | |
1281 | bs_entry->prepared = true; | |
1282 | } | |
1283 | ||
1284 | /* If we reach this point, we have success and just need to apply the | |
1285 | * changes | |
1286 | */ | |
1287 | QSIMPLEQ_FOREACH(bs_entry, bs_queue, entry) { | |
1288 | bdrv_reopen_commit(&bs_entry->state); | |
1289 | } | |
1290 | ||
1291 | ret = 0; | |
1292 | ||
1293 | cleanup: | |
1294 | QSIMPLEQ_FOREACH_SAFE(bs_entry, bs_queue, entry, next) { | |
1295 | if (ret && bs_entry->prepared) { | |
1296 | bdrv_reopen_abort(&bs_entry->state); | |
1297 | } | |
1298 | g_free(bs_entry); | |
1299 | } | |
1300 | g_free(bs_queue); | |
1301 | return ret; | |
1302 | } | |
1303 | ||
1304 | ||
1305 | /* Reopen a single BlockDriverState with the specified flags. */ | |
1306 | int bdrv_reopen(BlockDriverState *bs, int bdrv_flags, Error **errp) | |
1307 | { | |
1308 | int ret = -1; | |
1309 | Error *local_err = NULL; | |
1310 | BlockReopenQueue *queue = bdrv_reopen_queue(NULL, bs, bdrv_flags); | |
1311 | ||
1312 | ret = bdrv_reopen_multiple(queue, &local_err); | |
1313 | if (local_err != NULL) { | |
1314 | error_propagate(errp, local_err); | |
1315 | } | |
1316 | return ret; | |
1317 | } | |
1318 | ||
1319 | ||
1320 | /* | |
1321 | * Prepares a BlockDriverState for reopen. All changes are staged in the | |
1322 | * 'opaque' field of the BDRVReopenState, which is used and allocated by | |
1323 | * the block driver layer .bdrv_reopen_prepare() | |
1324 | * | |
1325 | * bs is the BlockDriverState to reopen | |
1326 | * flags are the new open flags | |
1327 | * queue is the reopen queue | |
1328 | * | |
1329 | * Returns 0 on success, non-zero on error. On error errp will be set | |
1330 | * as well. | |
1331 | * | |
1332 | * On failure, bdrv_reopen_abort() will be called to clean up any data. | |
1333 | * It is the responsibility of the caller to then call the abort() or | |
1334 | * commit() for any other BDS that have been left in a prepare() state | |
1335 | * | |
1336 | */ | |
1337 | int bdrv_reopen_prepare(BDRVReopenState *reopen_state, BlockReopenQueue *queue, | |
1338 | Error **errp) | |
1339 | { | |
1340 | int ret = -1; | |
1341 | Error *local_err = NULL; | |
1342 | BlockDriver *drv; | |
1343 | ||
1344 | assert(reopen_state != NULL); | |
1345 | assert(reopen_state->bs->drv != NULL); | |
1346 | drv = reopen_state->bs->drv; | |
1347 | ||
1348 | /* if we are to stay read-only, do not allow permission change | |
1349 | * to r/w */ | |
1350 | if (!(reopen_state->bs->open_flags & BDRV_O_ALLOW_RDWR) && | |
1351 | reopen_state->flags & BDRV_O_RDWR) { | |
1352 | error_set(errp, QERR_DEVICE_IS_READ_ONLY, | |
1353 | reopen_state->bs->device_name); | |
1354 | goto error; | |
1355 | } | |
1356 | ||
1357 | ||
1358 | ret = bdrv_flush(reopen_state->bs); | |
1359 | if (ret) { | |
1360 | error_set(errp, ERROR_CLASS_GENERIC_ERROR, "Error (%s) flushing drive", | |
1361 | strerror(-ret)); | |
1362 | goto error; | |
1363 | } | |
1364 | ||
1365 | if (drv->bdrv_reopen_prepare) { | |
1366 | ret = drv->bdrv_reopen_prepare(reopen_state, queue, &local_err); | |
1367 | if (ret) { | |
1368 | if (local_err != NULL) { | |
1369 | error_propagate(errp, local_err); | |
1370 | } else { | |
d8b6895f LC |
1371 | error_setg(errp, "failed while preparing to reopen image '%s'", |
1372 | reopen_state->bs->filename); | |
e971aa12 JC |
1373 | } |
1374 | goto error; | |
1375 | } | |
1376 | } else { | |
1377 | /* It is currently mandatory to have a bdrv_reopen_prepare() | |
1378 | * handler for each supported drv. */ | |
1379 | error_set(errp, QERR_BLOCK_FORMAT_FEATURE_NOT_SUPPORTED, | |
1380 | drv->format_name, reopen_state->bs->device_name, | |
1381 | "reopening of file"); | |
1382 | ret = -1; | |
1383 | goto error; | |
1384 | } | |
1385 | ||
1386 | ret = 0; | |
1387 | ||
1388 | error: | |
1389 | return ret; | |
1390 | } | |
1391 | ||
1392 | /* | |
1393 | * Takes the staged changes for the reopen from bdrv_reopen_prepare(), and | |
1394 | * makes them final by swapping the staging BlockDriverState contents into | |
1395 | * the active BlockDriverState contents. | |
1396 | */ | |
1397 | void bdrv_reopen_commit(BDRVReopenState *reopen_state) | |
1398 | { | |
1399 | BlockDriver *drv; | |
1400 | ||
1401 | assert(reopen_state != NULL); | |
1402 | drv = reopen_state->bs->drv; | |
1403 | assert(drv != NULL); | |
1404 | ||
1405 | /* If there are any driver level actions to take */ | |
1406 | if (drv->bdrv_reopen_commit) { | |
1407 | drv->bdrv_reopen_commit(reopen_state); | |
1408 | } | |
1409 | ||
1410 | /* set BDS specific flags now */ | |
1411 | reopen_state->bs->open_flags = reopen_state->flags; | |
1412 | reopen_state->bs->enable_write_cache = !!(reopen_state->flags & | |
1413 | BDRV_O_CACHE_WB); | |
1414 | reopen_state->bs->read_only = !(reopen_state->flags & BDRV_O_RDWR); | |
1415 | } | |
1416 | ||
1417 | /* | |
1418 | * Abort the reopen, and delete and free the staged changes in | |
1419 | * reopen_state | |
1420 | */ | |
1421 | void bdrv_reopen_abort(BDRVReopenState *reopen_state) | |
1422 | { | |
1423 | BlockDriver *drv; | |
1424 | ||
1425 | assert(reopen_state != NULL); | |
1426 | drv = reopen_state->bs->drv; | |
1427 | assert(drv != NULL); | |
1428 | ||
1429 | if (drv->bdrv_reopen_abort) { | |
1430 | drv->bdrv_reopen_abort(reopen_state); | |
1431 | } | |
1432 | } | |
1433 | ||
1434 | ||
fc01f7e7 FB |
1435 | void bdrv_close(BlockDriverState *bs) |
1436 | { | |
3cbc002c PB |
1437 | if (bs->job) { |
1438 | block_job_cancel_sync(bs->job); | |
1439 | } | |
58fda173 SH |
1440 | bdrv_drain_all(); /* complete I/O */ |
1441 | bdrv_flush(bs); | |
1442 | bdrv_drain_all(); /* in case flush left pending I/O */ | |
d7d512f6 | 1443 | notifier_list_notify(&bs->close_notifiers, bs); |
7094f12f | 1444 | |
3cbc002c | 1445 | if (bs->drv) { |
557df6ac | 1446 | if (bs->backing_hd) { |
4f6fd349 | 1447 | bdrv_unref(bs->backing_hd); |
557df6ac SH |
1448 | bs->backing_hd = NULL; |
1449 | } | |
ea2384d3 | 1450 | bs->drv->bdrv_close(bs); |
7267c094 | 1451 | g_free(bs->opaque); |
ea2384d3 FB |
1452 | #ifdef _WIN32 |
1453 | if (bs->is_temporary) { | |
1454 | unlink(bs->filename); | |
1455 | } | |
67b915a5 | 1456 | #endif |
ea2384d3 FB |
1457 | bs->opaque = NULL; |
1458 | bs->drv = NULL; | |
53fec9d3 | 1459 | bs->copy_on_read = 0; |
a275fa42 PB |
1460 | bs->backing_file[0] = '\0'; |
1461 | bs->backing_format[0] = '\0'; | |
6405875c PB |
1462 | bs->total_sectors = 0; |
1463 | bs->encrypted = 0; | |
1464 | bs->valid_key = 0; | |
1465 | bs->sg = 0; | |
1466 | bs->growable = 0; | |
0d51b4de | 1467 | bs->zero_beyond_eof = false; |
de9c0cec KW |
1468 | QDECREF(bs->options); |
1469 | bs->options = NULL; | |
b338082b | 1470 | |
66f82cee | 1471 | if (bs->file != NULL) { |
4f6fd349 | 1472 | bdrv_unref(bs->file); |
0ac9377d | 1473 | bs->file = NULL; |
66f82cee | 1474 | } |
b338082b | 1475 | } |
98f90dba | 1476 | |
9ca11154 PH |
1477 | bdrv_dev_change_media_cb(bs, false); |
1478 | ||
98f90dba ZYW |
1479 | /*throttling disk I/O limits*/ |
1480 | if (bs->io_limits_enabled) { | |
1481 | bdrv_io_limits_disable(bs); | |
1482 | } | |
b338082b FB |
1483 | } |
1484 | ||
2bc93fed MK |
1485 | void bdrv_close_all(void) |
1486 | { | |
1487 | BlockDriverState *bs; | |
1488 | ||
1489 | QTAILQ_FOREACH(bs, &bdrv_states, list) { | |
1490 | bdrv_close(bs); | |
1491 | } | |
1492 | } | |
1493 | ||
88266f5a SH |
1494 | /* Check if any requests are in-flight (including throttled requests) */ |
1495 | static bool bdrv_requests_pending(BlockDriverState *bs) | |
1496 | { | |
1497 | if (!QLIST_EMPTY(&bs->tracked_requests)) { | |
1498 | return true; | |
1499 | } | |
cc0681c4 BC |
1500 | if (!qemu_co_queue_empty(&bs->throttled_reqs[0])) { |
1501 | return true; | |
1502 | } | |
1503 | if (!qemu_co_queue_empty(&bs->throttled_reqs[1])) { | |
88266f5a SH |
1504 | return true; |
1505 | } | |
1506 | if (bs->file && bdrv_requests_pending(bs->file)) { | |
1507 | return true; | |
1508 | } | |
1509 | if (bs->backing_hd && bdrv_requests_pending(bs->backing_hd)) { | |
1510 | return true; | |
1511 | } | |
1512 | return false; | |
1513 | } | |
1514 | ||
1515 | static bool bdrv_requests_pending_all(void) | |
1516 | { | |
1517 | BlockDriverState *bs; | |
1518 | QTAILQ_FOREACH(bs, &bdrv_states, list) { | |
1519 | if (bdrv_requests_pending(bs)) { | |
1520 | return true; | |
1521 | } | |
1522 | } | |
1523 | return false; | |
1524 | } | |
1525 | ||
922453bc SH |
1526 | /* |
1527 | * Wait for pending requests to complete across all BlockDriverStates | |
1528 | * | |
1529 | * This function does not flush data to disk, use bdrv_flush_all() for that | |
1530 | * after calling this function. | |
4c355d53 ZYW |
1531 | * |
1532 | * Note that completion of an asynchronous I/O operation can trigger any | |
1533 | * number of other I/O operations on other devices---for example a coroutine | |
1534 | * can be arbitrarily complex and a constant flow of I/O can come until the | |
1535 | * coroutine is complete. Because of this, it is not possible to have a | |
1536 | * function to drain a single device's I/O queue. | |
922453bc SH |
1537 | */ |
1538 | void bdrv_drain_all(void) | |
1539 | { | |
88266f5a SH |
1540 | /* Always run first iteration so any pending completion BHs run */ |
1541 | bool busy = true; | |
922453bc SH |
1542 | BlockDriverState *bs; |
1543 | ||
88266f5a | 1544 | while (busy) { |
4c355d53 ZYW |
1545 | /* FIXME: We do not have timer support here, so this is effectively |
1546 | * a busy wait. | |
1547 | */ | |
1548 | QTAILQ_FOREACH(bs, &bdrv_states, list) { | |
cc0681c4 | 1549 | if (bdrv_start_throttled_reqs(bs)) { |
4c355d53 ZYW |
1550 | busy = true; |
1551 | } | |
1552 | } | |
922453bc | 1553 | |
88266f5a SH |
1554 | busy = bdrv_requests_pending_all(); |
1555 | busy |= aio_poll(qemu_get_aio_context(), busy); | |
922453bc SH |
1556 | } |
1557 | } | |
1558 | ||
d22b2f41 RH |
1559 | /* make a BlockDriverState anonymous by removing from bdrv_state list. |
1560 | Also, NULL terminate the device_name to prevent double remove */ | |
1561 | void bdrv_make_anon(BlockDriverState *bs) | |
1562 | { | |
1563 | if (bs->device_name[0] != '\0') { | |
1564 | QTAILQ_REMOVE(&bdrv_states, bs, list); | |
1565 | } | |
1566 | bs->device_name[0] = '\0'; | |
1567 | } | |
1568 | ||
e023b2e2 PB |
1569 | static void bdrv_rebind(BlockDriverState *bs) |
1570 | { | |
1571 | if (bs->drv && bs->drv->bdrv_rebind) { | |
1572 | bs->drv->bdrv_rebind(bs); | |
1573 | } | |
1574 | } | |
1575 | ||
4ddc07ca PB |
1576 | static void bdrv_move_feature_fields(BlockDriverState *bs_dest, |
1577 | BlockDriverState *bs_src) | |
8802d1fd | 1578 | { |
4ddc07ca PB |
1579 | /* move some fields that need to stay attached to the device */ |
1580 | bs_dest->open_flags = bs_src->open_flags; | |
8802d1fd JC |
1581 | |
1582 | /* dev info */ | |
4ddc07ca PB |
1583 | bs_dest->dev_ops = bs_src->dev_ops; |
1584 | bs_dest->dev_opaque = bs_src->dev_opaque; | |
1585 | bs_dest->dev = bs_src->dev; | |
1586 | bs_dest->buffer_alignment = bs_src->buffer_alignment; | |
1587 | bs_dest->copy_on_read = bs_src->copy_on_read; | |
8802d1fd | 1588 | |
4ddc07ca | 1589 | bs_dest->enable_write_cache = bs_src->enable_write_cache; |
c4a248a1 | 1590 | |
cc0681c4 BC |
1591 | /* i/o throttled req */ |
1592 | memcpy(&bs_dest->throttle_state, | |
1593 | &bs_src->throttle_state, | |
1594 | sizeof(ThrottleState)); | |
1595 | bs_dest->throttled_reqs[0] = bs_src->throttled_reqs[0]; | |
1596 | bs_dest->throttled_reqs[1] = bs_src->throttled_reqs[1]; | |
4ddc07ca | 1597 | bs_dest->io_limits_enabled = bs_src->io_limits_enabled; |
8802d1fd | 1598 | |
8802d1fd | 1599 | /* r/w error */ |
4ddc07ca PB |
1600 | bs_dest->on_read_error = bs_src->on_read_error; |
1601 | bs_dest->on_write_error = bs_src->on_write_error; | |
8802d1fd JC |
1602 | |
1603 | /* i/o status */ | |
4ddc07ca PB |
1604 | bs_dest->iostatus_enabled = bs_src->iostatus_enabled; |
1605 | bs_dest->iostatus = bs_src->iostatus; | |
8802d1fd | 1606 | |
a9fc4408 | 1607 | /* dirty bitmap */ |
4ddc07ca | 1608 | bs_dest->dirty_bitmap = bs_src->dirty_bitmap; |
a9fc4408 | 1609 | |
9fcb0251 FZ |
1610 | /* reference count */ |
1611 | bs_dest->refcnt = bs_src->refcnt; | |
1612 | ||
a9fc4408 | 1613 | /* job */ |
4ddc07ca PB |
1614 | bs_dest->in_use = bs_src->in_use; |
1615 | bs_dest->job = bs_src->job; | |
a9fc4408 | 1616 | |
8802d1fd | 1617 | /* keep the same entry in bdrv_states */ |
4ddc07ca PB |
1618 | pstrcpy(bs_dest->device_name, sizeof(bs_dest->device_name), |
1619 | bs_src->device_name); | |
1620 | bs_dest->list = bs_src->list; | |
1621 | } | |
8802d1fd | 1622 | |
4ddc07ca PB |
1623 | /* |
1624 | * Swap bs contents for two image chains while they are live, | |
1625 | * while keeping required fields on the BlockDriverState that is | |
1626 | * actually attached to a device. | |
1627 | * | |
1628 | * This will modify the BlockDriverState fields, and swap contents | |
1629 | * between bs_new and bs_old. Both bs_new and bs_old are modified. | |
1630 | * | |
1631 | * bs_new is required to be anonymous. | |
1632 | * | |
1633 | * This function does not create any image files. | |
1634 | */ | |
1635 | void bdrv_swap(BlockDriverState *bs_new, BlockDriverState *bs_old) | |
1636 | { | |
1637 | BlockDriverState tmp; | |
f6801b83 | 1638 | |
4ddc07ca PB |
1639 | /* bs_new must be anonymous and shouldn't have anything fancy enabled */ |
1640 | assert(bs_new->device_name[0] == '\0'); | |
1641 | assert(bs_new->dirty_bitmap == NULL); | |
1642 | assert(bs_new->job == NULL); | |
1643 | assert(bs_new->dev == NULL); | |
1644 | assert(bs_new->in_use == 0); | |
1645 | assert(bs_new->io_limits_enabled == false); | |
cc0681c4 | 1646 | assert(!throttle_have_timer(&bs_new->throttle_state)); |
8802d1fd | 1647 | |
4ddc07ca PB |
1648 | tmp = *bs_new; |
1649 | *bs_new = *bs_old; | |
1650 | *bs_old = tmp; | |
a9fc4408 | 1651 | |
4ddc07ca PB |
1652 | /* there are some fields that should not be swapped, move them back */ |
1653 | bdrv_move_feature_fields(&tmp, bs_old); | |
1654 | bdrv_move_feature_fields(bs_old, bs_new); | |
1655 | bdrv_move_feature_fields(bs_new, &tmp); | |
8802d1fd | 1656 | |
4ddc07ca PB |
1657 | /* bs_new shouldn't be in bdrv_states even after the swap! */ |
1658 | assert(bs_new->device_name[0] == '\0'); | |
1659 | ||
1660 | /* Check a few fields that should remain attached to the device */ | |
1661 | assert(bs_new->dev == NULL); | |
1662 | assert(bs_new->job == NULL); | |
1663 | assert(bs_new->in_use == 0); | |
1664 | assert(bs_new->io_limits_enabled == false); | |
cc0681c4 | 1665 | assert(!throttle_have_timer(&bs_new->throttle_state)); |
e023b2e2 PB |
1666 | |
1667 | bdrv_rebind(bs_new); | |
4ddc07ca PB |
1668 | bdrv_rebind(bs_old); |
1669 | } | |
1670 | ||
1671 | /* | |
1672 | * Add new bs contents at the top of an image chain while the chain is | |
1673 | * live, while keeping required fields on the top layer. | |
1674 | * | |
1675 | * This will modify the BlockDriverState fields, and swap contents | |
1676 | * between bs_new and bs_top. Both bs_new and bs_top are modified. | |
1677 | * | |
1678 | * bs_new is required to be anonymous. | |
1679 | * | |
1680 | * This function does not create any image files. | |
1681 | */ | |
1682 | void bdrv_append(BlockDriverState *bs_new, BlockDriverState *bs_top) | |
1683 | { | |
1684 | bdrv_swap(bs_new, bs_top); | |
1685 | ||
1686 | /* The contents of 'tmp' will become bs_top, as we are | |
1687 | * swapping bs_new and bs_top contents. */ | |
1688 | bs_top->backing_hd = bs_new; | |
1689 | bs_top->open_flags &= ~BDRV_O_NO_BACKING; | |
1690 | pstrcpy(bs_top->backing_file, sizeof(bs_top->backing_file), | |
1691 | bs_new->filename); | |
1692 | pstrcpy(bs_top->backing_format, sizeof(bs_top->backing_format), | |
1693 | bs_new->drv ? bs_new->drv->format_name : ""); | |
8802d1fd JC |
1694 | } |
1695 | ||
4f6fd349 | 1696 | static void bdrv_delete(BlockDriverState *bs) |
b338082b | 1697 | { |
fa879d62 | 1698 | assert(!bs->dev); |
3e914655 PB |
1699 | assert(!bs->job); |
1700 | assert(!bs->in_use); | |
4f6fd349 | 1701 | assert(!bs->refcnt); |
18846dee | 1702 | |
e1b5c52e SH |
1703 | bdrv_close(bs); |
1704 | ||
1b7bdbc1 | 1705 | /* remove from list, if necessary */ |
d22b2f41 | 1706 | bdrv_make_anon(bs); |
34c6f050 | 1707 | |
7267c094 | 1708 | g_free(bs); |
fc01f7e7 FB |
1709 | } |
1710 | ||
fa879d62 MA |
1711 | int bdrv_attach_dev(BlockDriverState *bs, void *dev) |
1712 | /* TODO change to DeviceState *dev when all users are qdevified */ | |
18846dee | 1713 | { |
fa879d62 | 1714 | if (bs->dev) { |
18846dee MA |
1715 | return -EBUSY; |
1716 | } | |
fa879d62 | 1717 | bs->dev = dev; |
28a7282a | 1718 | bdrv_iostatus_reset(bs); |
18846dee MA |
1719 | return 0; |
1720 | } | |
1721 | ||
fa879d62 MA |
1722 | /* TODO qdevified devices don't use this, remove when devices are qdevified */ |
1723 | void bdrv_attach_dev_nofail(BlockDriverState *bs, void *dev) | |
18846dee | 1724 | { |
fa879d62 MA |
1725 | if (bdrv_attach_dev(bs, dev) < 0) { |
1726 | abort(); | |
1727 | } | |
1728 | } | |
1729 | ||
1730 | void bdrv_detach_dev(BlockDriverState *bs, void *dev) | |
1731 | /* TODO change to DeviceState *dev when all users are qdevified */ | |
1732 | { | |
1733 | assert(bs->dev == dev); | |
1734 | bs->dev = NULL; | |
0e49de52 MA |
1735 | bs->dev_ops = NULL; |
1736 | bs->dev_opaque = NULL; | |
29e05f20 | 1737 | bs->buffer_alignment = 512; |
18846dee MA |
1738 | } |
1739 | ||
fa879d62 MA |
1740 | /* TODO change to return DeviceState * when all users are qdevified */ |
1741 | void *bdrv_get_attached_dev(BlockDriverState *bs) | |
18846dee | 1742 | { |
fa879d62 | 1743 | return bs->dev; |
18846dee MA |
1744 | } |
1745 | ||
0e49de52 MA |
1746 | void bdrv_set_dev_ops(BlockDriverState *bs, const BlockDevOps *ops, |
1747 | void *opaque) | |
1748 | { | |
1749 | bs->dev_ops = ops; | |
1750 | bs->dev_opaque = opaque; | |
1751 | } | |
1752 | ||
32c81a4a PB |
1753 | void bdrv_emit_qmp_error_event(const BlockDriverState *bdrv, |
1754 | enum MonitorEvent ev, | |
1755 | BlockErrorAction action, bool is_read) | |
329c0a48 LC |
1756 | { |
1757 | QObject *data; | |
1758 | const char *action_str; | |
1759 | ||
1760 | switch (action) { | |
1761 | case BDRV_ACTION_REPORT: | |
1762 | action_str = "report"; | |
1763 | break; | |
1764 | case BDRV_ACTION_IGNORE: | |
1765 | action_str = "ignore"; | |
1766 | break; | |
1767 | case BDRV_ACTION_STOP: | |
1768 | action_str = "stop"; | |
1769 | break; | |
1770 | default: | |
1771 | abort(); | |
1772 | } | |
1773 | ||
1774 | data = qobject_from_jsonf("{ 'device': %s, 'action': %s, 'operation': %s }", | |
1775 | bdrv->device_name, | |
1776 | action_str, | |
1777 | is_read ? "read" : "write"); | |
32c81a4a | 1778 | monitor_protocol_event(ev, data); |
329c0a48 LC |
1779 | |
1780 | qobject_decref(data); | |
1781 | } | |
1782 | ||
6f382ed2 LC |
1783 | static void bdrv_emit_qmp_eject_event(BlockDriverState *bs, bool ejected) |
1784 | { | |
1785 | QObject *data; | |
1786 | ||
1787 | data = qobject_from_jsonf("{ 'device': %s, 'tray-open': %i }", | |
1788 | bdrv_get_device_name(bs), ejected); | |
1789 | monitor_protocol_event(QEVENT_DEVICE_TRAY_MOVED, data); | |
1790 | ||
1791 | qobject_decref(data); | |
1792 | } | |
1793 | ||
7d4b4ba5 | 1794 | static void bdrv_dev_change_media_cb(BlockDriverState *bs, bool load) |
0e49de52 | 1795 | { |
145feb17 | 1796 | if (bs->dev_ops && bs->dev_ops->change_media_cb) { |
6f382ed2 | 1797 | bool tray_was_closed = !bdrv_dev_is_tray_open(bs); |
7d4b4ba5 | 1798 | bs->dev_ops->change_media_cb(bs->dev_opaque, load); |
6f382ed2 LC |
1799 | if (tray_was_closed) { |
1800 | /* tray open */ | |
1801 | bdrv_emit_qmp_eject_event(bs, true); | |
1802 | } | |
1803 | if (load) { | |
1804 | /* tray close */ | |
1805 | bdrv_emit_qmp_eject_event(bs, false); | |
1806 | } | |
145feb17 MA |
1807 | } |
1808 | } | |
1809 | ||
2c6942fa MA |
1810 | bool bdrv_dev_has_removable_media(BlockDriverState *bs) |
1811 | { | |
1812 | return !bs->dev || (bs->dev_ops && bs->dev_ops->change_media_cb); | |
1813 | } | |
1814 | ||
025ccaa7 PB |
1815 | void bdrv_dev_eject_request(BlockDriverState *bs, bool force) |
1816 | { | |
1817 | if (bs->dev_ops && bs->dev_ops->eject_request_cb) { | |
1818 | bs->dev_ops->eject_request_cb(bs->dev_opaque, force); | |
1819 | } | |
1820 | } | |
1821 | ||
e4def80b MA |
1822 | bool bdrv_dev_is_tray_open(BlockDriverState *bs) |
1823 | { | |
1824 | if (bs->dev_ops && bs->dev_ops->is_tray_open) { | |
1825 | return bs->dev_ops->is_tray_open(bs->dev_opaque); | |
1826 | } | |
1827 | return false; | |
1828 | } | |
1829 | ||
145feb17 MA |
1830 | static void bdrv_dev_resize_cb(BlockDriverState *bs) |
1831 | { | |
1832 | if (bs->dev_ops && bs->dev_ops->resize_cb) { | |
1833 | bs->dev_ops->resize_cb(bs->dev_opaque); | |
0e49de52 MA |
1834 | } |
1835 | } | |
1836 | ||
f107639a MA |
1837 | bool bdrv_dev_is_medium_locked(BlockDriverState *bs) |
1838 | { | |
1839 | if (bs->dev_ops && bs->dev_ops->is_medium_locked) { | |
1840 | return bs->dev_ops->is_medium_locked(bs->dev_opaque); | |
1841 | } | |
1842 | return false; | |
1843 | } | |
1844 | ||
e97fc193 AL |
1845 | /* |
1846 | * Run consistency checks on an image | |
1847 | * | |
e076f338 | 1848 | * Returns 0 if the check could be completed (it doesn't mean that the image is |
a1c7273b | 1849 | * free of errors) or -errno when an internal error occurred. The results of the |
e076f338 | 1850 | * check are stored in res. |
e97fc193 | 1851 | */ |
4534ff54 | 1852 | int bdrv_check(BlockDriverState *bs, BdrvCheckResult *res, BdrvCheckMode fix) |
e97fc193 AL |
1853 | { |
1854 | if (bs->drv->bdrv_check == NULL) { | |
1855 | return -ENOTSUP; | |
1856 | } | |
1857 | ||
e076f338 | 1858 | memset(res, 0, sizeof(*res)); |
4534ff54 | 1859 | return bs->drv->bdrv_check(bs, res, fix); |
e97fc193 AL |
1860 | } |
1861 | ||
8a426614 KW |
1862 | #define COMMIT_BUF_SECTORS 2048 |
1863 | ||
33e3963e FB |
1864 | /* commit COW file into the raw image */ |
1865 | int bdrv_commit(BlockDriverState *bs) | |
1866 | { | |
19cb3738 | 1867 | BlockDriver *drv = bs->drv; |
8a426614 KW |
1868 | int64_t sector, total_sectors; |
1869 | int n, ro, open_flags; | |
0bce597d | 1870 | int ret = 0; |
8a426614 | 1871 | uint8_t *buf; |
c2cba3d9 | 1872 | char filename[PATH_MAX]; |
33e3963e | 1873 | |
19cb3738 FB |
1874 | if (!drv) |
1875 | return -ENOMEDIUM; | |
4dca4b63 NS |
1876 | |
1877 | if (!bs->backing_hd) { | |
1878 | return -ENOTSUP; | |
33e3963e FB |
1879 | } |
1880 | ||
2d3735d3 SH |
1881 | if (bdrv_in_use(bs) || bdrv_in_use(bs->backing_hd)) { |
1882 | return -EBUSY; | |
1883 | } | |
1884 | ||
4dca4b63 | 1885 | ro = bs->backing_hd->read_only; |
c2cba3d9 JM |
1886 | /* Use pstrcpy (not strncpy): filename must be NUL-terminated. */ |
1887 | pstrcpy(filename, sizeof(filename), bs->backing_hd->filename); | |
4dca4b63 NS |
1888 | open_flags = bs->backing_hd->open_flags; |
1889 | ||
1890 | if (ro) { | |
0bce597d JC |
1891 | if (bdrv_reopen(bs->backing_hd, open_flags | BDRV_O_RDWR, NULL)) { |
1892 | return -EACCES; | |
4dca4b63 | 1893 | } |
ea2384d3 | 1894 | } |
33e3963e | 1895 | |
6ea44308 | 1896 | total_sectors = bdrv_getlength(bs) >> BDRV_SECTOR_BITS; |
7267c094 | 1897 | buf = g_malloc(COMMIT_BUF_SECTORS * BDRV_SECTOR_SIZE); |
8a426614 KW |
1898 | |
1899 | for (sector = 0; sector < total_sectors; sector += n) { | |
d663640c PB |
1900 | ret = bdrv_is_allocated(bs, sector, COMMIT_BUF_SECTORS, &n); |
1901 | if (ret < 0) { | |
1902 | goto ro_cleanup; | |
1903 | } | |
1904 | if (ret) { | |
8a426614 KW |
1905 | if (bdrv_read(bs, sector, buf, n) != 0) { |
1906 | ret = -EIO; | |
1907 | goto ro_cleanup; | |
1908 | } | |
1909 | ||
1910 | if (bdrv_write(bs->backing_hd, sector, buf, n) != 0) { | |
1911 | ret = -EIO; | |
1912 | goto ro_cleanup; | |
1913 | } | |
ea2384d3 | 1914 | } |
33e3963e | 1915 | } |
95389c86 | 1916 | |
1d44952f CH |
1917 | if (drv->bdrv_make_empty) { |
1918 | ret = drv->bdrv_make_empty(bs); | |
1919 | bdrv_flush(bs); | |
1920 | } | |
95389c86 | 1921 | |
3f5075ae CH |
1922 | /* |
1923 | * Make sure all data we wrote to the backing device is actually | |
1924 | * stable on disk. | |
1925 | */ | |
1926 | if (bs->backing_hd) | |
1927 | bdrv_flush(bs->backing_hd); | |
4dca4b63 NS |
1928 | |
1929 | ro_cleanup: | |
7267c094 | 1930 | g_free(buf); |
4dca4b63 NS |
1931 | |
1932 | if (ro) { | |
0bce597d JC |
1933 | /* ignoring error return here */ |
1934 | bdrv_reopen(bs->backing_hd, open_flags & ~BDRV_O_RDWR, NULL); | |
4dca4b63 NS |
1935 | } |
1936 | ||
1d44952f | 1937 | return ret; |
33e3963e FB |
1938 | } |
1939 | ||
e8877497 | 1940 | int bdrv_commit_all(void) |
6ab4b5ab MA |
1941 | { |
1942 | BlockDriverState *bs; | |
1943 | ||
1944 | QTAILQ_FOREACH(bs, &bdrv_states, list) { | |
272d2d8e JC |
1945 | if (bs->drv && bs->backing_hd) { |
1946 | int ret = bdrv_commit(bs); | |
1947 | if (ret < 0) { | |
1948 | return ret; | |
1949 | } | |
e8877497 | 1950 | } |
6ab4b5ab | 1951 | } |
e8877497 | 1952 | return 0; |
6ab4b5ab MA |
1953 | } |
1954 | ||
dbffbdcf SH |
1955 | /** |
1956 | * Remove an active request from the tracked requests list | |
1957 | * | |
1958 | * This function should be called when a tracked request is completing. | |
1959 | */ | |
1960 | static void tracked_request_end(BdrvTrackedRequest *req) | |
1961 | { | |
1962 | QLIST_REMOVE(req, list); | |
f4658285 | 1963 | qemu_co_queue_restart_all(&req->wait_queue); |
dbffbdcf SH |
1964 | } |
1965 | ||
1966 | /** | |
1967 | * Add an active request to the tracked requests list | |
1968 | */ | |
1969 | static void tracked_request_begin(BdrvTrackedRequest *req, | |
1970 | BlockDriverState *bs, | |
1971 | int64_t sector_num, | |
1972 | int nb_sectors, bool is_write) | |
1973 | { | |
1974 | *req = (BdrvTrackedRequest){ | |
1975 | .bs = bs, | |
1976 | .sector_num = sector_num, | |
1977 | .nb_sectors = nb_sectors, | |
1978 | .is_write = is_write, | |
5f8b6491 | 1979 | .co = qemu_coroutine_self(), |
dbffbdcf SH |
1980 | }; |
1981 | ||
f4658285 SH |
1982 | qemu_co_queue_init(&req->wait_queue); |
1983 | ||
dbffbdcf SH |
1984 | QLIST_INSERT_HEAD(&bs->tracked_requests, req, list); |
1985 | } | |
1986 | ||
d83947ac SH |
1987 | /** |
1988 | * Round a region to cluster boundaries | |
1989 | */ | |
343bded4 PB |
1990 | void bdrv_round_to_clusters(BlockDriverState *bs, |
1991 | int64_t sector_num, int nb_sectors, | |
1992 | int64_t *cluster_sector_num, | |
1993 | int *cluster_nb_sectors) | |
d83947ac SH |
1994 | { |
1995 | BlockDriverInfo bdi; | |
1996 | ||
1997 | if (bdrv_get_info(bs, &bdi) < 0 || bdi.cluster_size == 0) { | |
1998 | *cluster_sector_num = sector_num; | |
1999 | *cluster_nb_sectors = nb_sectors; | |
2000 | } else { | |
2001 | int64_t c = bdi.cluster_size / BDRV_SECTOR_SIZE; | |
2002 | *cluster_sector_num = QEMU_ALIGN_DOWN(sector_num, c); | |
2003 | *cluster_nb_sectors = QEMU_ALIGN_UP(sector_num - *cluster_sector_num + | |
2004 | nb_sectors, c); | |
2005 | } | |
2006 | } | |
2007 | ||
f4658285 SH |
2008 | static bool tracked_request_overlaps(BdrvTrackedRequest *req, |
2009 | int64_t sector_num, int nb_sectors) { | |
d83947ac SH |
2010 | /* aaaa bbbb */ |
2011 | if (sector_num >= req->sector_num + req->nb_sectors) { | |
2012 | return false; | |
2013 | } | |
2014 | /* bbbb aaaa */ | |
2015 | if (req->sector_num >= sector_num + nb_sectors) { | |
2016 | return false; | |
2017 | } | |
2018 | return true; | |
f4658285 SH |
2019 | } |
2020 | ||
2021 | static void coroutine_fn wait_for_overlapping_requests(BlockDriverState *bs, | |
2022 | int64_t sector_num, int nb_sectors) | |
2023 | { | |
2024 | BdrvTrackedRequest *req; | |
d83947ac SH |
2025 | int64_t cluster_sector_num; |
2026 | int cluster_nb_sectors; | |
f4658285 SH |
2027 | bool retry; |
2028 | ||
d83947ac SH |
2029 | /* If we touch the same cluster it counts as an overlap. This guarantees |
2030 | * that allocating writes will be serialized and not race with each other | |
2031 | * for the same cluster. For example, in copy-on-read it ensures that the | |
2032 | * CoR read and write operations are atomic and guest writes cannot | |
2033 | * interleave between them. | |
2034 | */ | |
343bded4 PB |
2035 | bdrv_round_to_clusters(bs, sector_num, nb_sectors, |
2036 | &cluster_sector_num, &cluster_nb_sectors); | |
d83947ac | 2037 | |
f4658285 SH |
2038 | do { |
2039 | retry = false; | |
2040 | QLIST_FOREACH(req, &bs->tracked_requests, list) { | |
d83947ac SH |
2041 | if (tracked_request_overlaps(req, cluster_sector_num, |
2042 | cluster_nb_sectors)) { | |
5f8b6491 SH |
2043 | /* Hitting this means there was a reentrant request, for |
2044 | * example, a block driver issuing nested requests. This must | |
2045 | * never happen since it means deadlock. | |
2046 | */ | |
2047 | assert(qemu_coroutine_self() != req->co); | |
2048 | ||
f4658285 SH |
2049 | qemu_co_queue_wait(&req->wait_queue); |
2050 | retry = true; | |
2051 | break; | |
2052 | } | |
2053 | } | |
2054 | } while (retry); | |
2055 | } | |
2056 | ||
756e6736 KW |
2057 | /* |
2058 | * Return values: | |
2059 | * 0 - success | |
2060 | * -EINVAL - backing format specified, but no file | |
2061 | * -ENOSPC - can't update the backing file because no space is left in the | |
2062 | * image file header | |
2063 | * -ENOTSUP - format driver doesn't support changing the backing file | |
2064 | */ | |
2065 | int bdrv_change_backing_file(BlockDriverState *bs, | |
2066 | const char *backing_file, const char *backing_fmt) | |
2067 | { | |
2068 | BlockDriver *drv = bs->drv; | |
469ef350 | 2069 | int ret; |
756e6736 | 2070 | |
5f377794 PB |
2071 | /* Backing file format doesn't make sense without a backing file */ |
2072 | if (backing_fmt && !backing_file) { | |
2073 | return -EINVAL; | |
2074 | } | |
2075 | ||
756e6736 | 2076 | if (drv->bdrv_change_backing_file != NULL) { |
469ef350 | 2077 | ret = drv->bdrv_change_backing_file(bs, backing_file, backing_fmt); |
756e6736 | 2078 | } else { |
469ef350 | 2079 | ret = -ENOTSUP; |
756e6736 | 2080 | } |
469ef350 PB |
2081 | |
2082 | if (ret == 0) { | |
2083 | pstrcpy(bs->backing_file, sizeof(bs->backing_file), backing_file ?: ""); | |
2084 | pstrcpy(bs->backing_format, sizeof(bs->backing_format), backing_fmt ?: ""); | |
2085 | } | |
2086 | return ret; | |
756e6736 KW |
2087 | } |
2088 | ||
6ebdcee2 JC |
2089 | /* |
2090 | * Finds the image layer in the chain that has 'bs' as its backing file. | |
2091 | * | |
2092 | * active is the current topmost image. | |
2093 | * | |
2094 | * Returns NULL if bs is not found in active's image chain, | |
2095 | * or if active == bs. | |
2096 | */ | |
2097 | BlockDriverState *bdrv_find_overlay(BlockDriverState *active, | |
2098 | BlockDriverState *bs) | |
2099 | { | |
2100 | BlockDriverState *overlay = NULL; | |
2101 | BlockDriverState *intermediate; | |
2102 | ||
2103 | assert(active != NULL); | |
2104 | assert(bs != NULL); | |
2105 | ||
2106 | /* if bs is the same as active, then by definition it has no overlay | |
2107 | */ | |
2108 | if (active == bs) { | |
2109 | return NULL; | |
2110 | } | |
2111 | ||
2112 | intermediate = active; | |
2113 | while (intermediate->backing_hd) { | |
2114 | if (intermediate->backing_hd == bs) { | |
2115 | overlay = intermediate; | |
2116 | break; | |
2117 | } | |
2118 | intermediate = intermediate->backing_hd; | |
2119 | } | |
2120 | ||
2121 | return overlay; | |
2122 | } | |
2123 | ||
2124 | typedef struct BlkIntermediateStates { | |
2125 | BlockDriverState *bs; | |
2126 | QSIMPLEQ_ENTRY(BlkIntermediateStates) entry; | |
2127 | } BlkIntermediateStates; | |
2128 | ||
2129 | ||
2130 | /* | |
2131 | * Drops images above 'base' up to and including 'top', and sets the image | |
2132 | * above 'top' to have base as its backing file. | |
2133 | * | |
2134 | * Requires that the overlay to 'top' is opened r/w, so that the backing file | |
2135 | * information in 'bs' can be properly updated. | |
2136 | * | |
2137 | * E.g., this will convert the following chain: | |
2138 | * bottom <- base <- intermediate <- top <- active | |
2139 | * | |
2140 | * to | |
2141 | * | |
2142 | * bottom <- base <- active | |
2143 | * | |
2144 | * It is allowed for bottom==base, in which case it converts: | |
2145 | * | |
2146 | * base <- intermediate <- top <- active | |
2147 | * | |
2148 | * to | |
2149 | * | |
2150 | * base <- active | |
2151 | * | |
2152 | * Error conditions: | |
2153 | * if active == top, that is considered an error | |
2154 | * | |
2155 | */ | |
2156 | int bdrv_drop_intermediate(BlockDriverState *active, BlockDriverState *top, | |
2157 | BlockDriverState *base) | |
2158 | { | |
2159 | BlockDriverState *intermediate; | |
2160 | BlockDriverState *base_bs = NULL; | |
2161 | BlockDriverState *new_top_bs = NULL; | |
2162 | BlkIntermediateStates *intermediate_state, *next; | |
2163 | int ret = -EIO; | |
2164 | ||
2165 | QSIMPLEQ_HEAD(states_to_delete, BlkIntermediateStates) states_to_delete; | |
2166 | QSIMPLEQ_INIT(&states_to_delete); | |
2167 | ||
2168 | if (!top->drv || !base->drv) { | |
2169 | goto exit; | |
2170 | } | |
2171 | ||
2172 | new_top_bs = bdrv_find_overlay(active, top); | |
2173 | ||
2174 | if (new_top_bs == NULL) { | |
2175 | /* we could not find the image above 'top', this is an error */ | |
2176 | goto exit; | |
2177 | } | |
2178 | ||
2179 | /* special case of new_top_bs->backing_hd already pointing to base - nothing | |
2180 | * to do, no intermediate images */ | |
2181 | if (new_top_bs->backing_hd == base) { | |
2182 | ret = 0; | |
2183 | goto exit; | |
2184 | } | |
2185 | ||
2186 | intermediate = top; | |
2187 | ||
2188 | /* now we will go down through the list, and add each BDS we find | |
2189 | * into our deletion queue, until we hit the 'base' | |
2190 | */ | |
2191 | while (intermediate) { | |
2192 | intermediate_state = g_malloc0(sizeof(BlkIntermediateStates)); | |
2193 | intermediate_state->bs = intermediate; | |
2194 | QSIMPLEQ_INSERT_TAIL(&states_to_delete, intermediate_state, entry); | |
2195 | ||
2196 | if (intermediate->backing_hd == base) { | |
2197 | base_bs = intermediate->backing_hd; | |
2198 | break; | |
2199 | } | |
2200 | intermediate = intermediate->backing_hd; | |
2201 | } | |
2202 | if (base_bs == NULL) { | |
2203 | /* something went wrong, we did not end at the base. safely | |
2204 | * unravel everything, and exit with error */ | |
2205 | goto exit; | |
2206 | } | |
2207 | ||
2208 | /* success - we can delete the intermediate states, and link top->base */ | |
2209 | ret = bdrv_change_backing_file(new_top_bs, base_bs->filename, | |
2210 | base_bs->drv ? base_bs->drv->format_name : ""); | |
2211 | if (ret) { | |
2212 | goto exit; | |
2213 | } | |
2214 | new_top_bs->backing_hd = base_bs; | |
2215 | ||
2216 | ||
2217 | QSIMPLEQ_FOREACH_SAFE(intermediate_state, &states_to_delete, entry, next) { | |
2218 | /* so that bdrv_close() does not recursively close the chain */ | |
2219 | intermediate_state->bs->backing_hd = NULL; | |
4f6fd349 | 2220 | bdrv_unref(intermediate_state->bs); |
6ebdcee2 JC |
2221 | } |
2222 | ret = 0; | |
2223 | ||
2224 | exit: | |
2225 | QSIMPLEQ_FOREACH_SAFE(intermediate_state, &states_to_delete, entry, next) { | |
2226 | g_free(intermediate_state); | |
2227 | } | |
2228 | return ret; | |
2229 | } | |
2230 | ||
2231 | ||
71d0770c AL |
2232 | static int bdrv_check_byte_request(BlockDriverState *bs, int64_t offset, |
2233 | size_t size) | |
2234 | { | |
2235 | int64_t len; | |
2236 | ||
2237 | if (!bdrv_is_inserted(bs)) | |
2238 | return -ENOMEDIUM; | |
2239 | ||
2240 | if (bs->growable) | |
2241 | return 0; | |
2242 | ||
2243 | len = bdrv_getlength(bs); | |
2244 | ||
fbb7b4e0 KW |
2245 | if (offset < 0) |
2246 | return -EIO; | |
2247 | ||
2248 | if ((offset > len) || (len - offset < size)) | |
71d0770c AL |
2249 | return -EIO; |
2250 | ||
2251 | return 0; | |
2252 | } | |
2253 | ||
2254 | static int bdrv_check_request(BlockDriverState *bs, int64_t sector_num, | |
2255 | int nb_sectors) | |
2256 | { | |
eb5a3165 JS |
2257 | return bdrv_check_byte_request(bs, sector_num * BDRV_SECTOR_SIZE, |
2258 | nb_sectors * BDRV_SECTOR_SIZE); | |
71d0770c AL |
2259 | } |
2260 | ||
1c9805a3 SH |
2261 | typedef struct RwCo { |
2262 | BlockDriverState *bs; | |
2263 | int64_t sector_num; | |
2264 | int nb_sectors; | |
2265 | QEMUIOVector *qiov; | |
2266 | bool is_write; | |
2267 | int ret; | |
4105eaaa | 2268 | BdrvRequestFlags flags; |
1c9805a3 SH |
2269 | } RwCo; |
2270 | ||
2271 | static void coroutine_fn bdrv_rw_co_entry(void *opaque) | |
fc01f7e7 | 2272 | { |
1c9805a3 | 2273 | RwCo *rwco = opaque; |
ea2384d3 | 2274 | |
1c9805a3 SH |
2275 | if (!rwco->is_write) { |
2276 | rwco->ret = bdrv_co_do_readv(rwco->bs, rwco->sector_num, | |
4105eaaa PL |
2277 | rwco->nb_sectors, rwco->qiov, |
2278 | rwco->flags); | |
1c9805a3 SH |
2279 | } else { |
2280 | rwco->ret = bdrv_co_do_writev(rwco->bs, rwco->sector_num, | |
4105eaaa PL |
2281 | rwco->nb_sectors, rwco->qiov, |
2282 | rwco->flags); | |
1c9805a3 SH |
2283 | } |
2284 | } | |
e7a8a783 | 2285 | |
1c9805a3 | 2286 | /* |
8d3b1a2d | 2287 | * Process a vectored synchronous request using coroutines |
1c9805a3 | 2288 | */ |
8d3b1a2d | 2289 | static int bdrv_rwv_co(BlockDriverState *bs, int64_t sector_num, |
4105eaaa PL |
2290 | QEMUIOVector *qiov, bool is_write, |
2291 | BdrvRequestFlags flags) | |
1c9805a3 | 2292 | { |
1c9805a3 SH |
2293 | Coroutine *co; |
2294 | RwCo rwco = { | |
2295 | .bs = bs, | |
2296 | .sector_num = sector_num, | |
8d3b1a2d KW |
2297 | .nb_sectors = qiov->size >> BDRV_SECTOR_BITS, |
2298 | .qiov = qiov, | |
1c9805a3 SH |
2299 | .is_write = is_write, |
2300 | .ret = NOT_DONE, | |
4105eaaa | 2301 | .flags = flags, |
1c9805a3 | 2302 | }; |
8d3b1a2d | 2303 | assert((qiov->size & (BDRV_SECTOR_SIZE - 1)) == 0); |
e7a8a783 | 2304 | |
498e386c ZYW |
2305 | /** |
2306 | * In sync call context, when the vcpu is blocked, this throttling timer | |
2307 | * will not fire; so the I/O throttling function has to be disabled here | |
2308 | * if it has been enabled. | |
2309 | */ | |
2310 | if (bs->io_limits_enabled) { | |
2311 | fprintf(stderr, "Disabling I/O throttling on '%s' due " | |
2312 | "to synchronous I/O.\n", bdrv_get_device_name(bs)); | |
2313 | bdrv_io_limits_disable(bs); | |
2314 | } | |
2315 | ||
1c9805a3 SH |
2316 | if (qemu_in_coroutine()) { |
2317 | /* Fast-path if already in coroutine context */ | |
2318 | bdrv_rw_co_entry(&rwco); | |
2319 | } else { | |
2320 | co = qemu_coroutine_create(bdrv_rw_co_entry); | |
2321 | qemu_coroutine_enter(co, &rwco); | |
2322 | while (rwco.ret == NOT_DONE) { | |
2323 | qemu_aio_wait(); | |
2324 | } | |
2325 | } | |
2326 | return rwco.ret; | |
2327 | } | |
b338082b | 2328 | |
8d3b1a2d KW |
2329 | /* |
2330 | * Process a synchronous request using coroutines | |
2331 | */ | |
2332 | static int bdrv_rw_co(BlockDriverState *bs, int64_t sector_num, uint8_t *buf, | |
4105eaaa | 2333 | int nb_sectors, bool is_write, BdrvRequestFlags flags) |
8d3b1a2d KW |
2334 | { |
2335 | QEMUIOVector qiov; | |
2336 | struct iovec iov = { | |
2337 | .iov_base = (void *)buf, | |
2338 | .iov_len = nb_sectors * BDRV_SECTOR_SIZE, | |
2339 | }; | |
2340 | ||
2341 | qemu_iovec_init_external(&qiov, &iov, 1); | |
4105eaaa | 2342 | return bdrv_rwv_co(bs, sector_num, &qiov, is_write, flags); |
8d3b1a2d KW |
2343 | } |
2344 | ||
1c9805a3 SH |
2345 | /* return < 0 if error. See bdrv_write() for the return codes */ |
2346 | int bdrv_read(BlockDriverState *bs, int64_t sector_num, | |
2347 | uint8_t *buf, int nb_sectors) | |
2348 | { | |
4105eaaa | 2349 | return bdrv_rw_co(bs, sector_num, buf, nb_sectors, false, 0); |
fc01f7e7 FB |
2350 | } |
2351 | ||
07d27a44 MA |
2352 | /* Just like bdrv_read(), but with I/O throttling temporarily disabled */ |
2353 | int bdrv_read_unthrottled(BlockDriverState *bs, int64_t sector_num, | |
2354 | uint8_t *buf, int nb_sectors) | |
2355 | { | |
2356 | bool enabled; | |
2357 | int ret; | |
2358 | ||
2359 | enabled = bs->io_limits_enabled; | |
2360 | bs->io_limits_enabled = false; | |
4e7395e8 | 2361 | ret = bdrv_read(bs, sector_num, buf, nb_sectors); |
07d27a44 MA |
2362 | bs->io_limits_enabled = enabled; |
2363 | return ret; | |
2364 | } | |
2365 | ||
5fafdf24 | 2366 | /* Return < 0 if error. Important errors are: |
19cb3738 FB |
2367 | -EIO generic I/O error (may happen for all errors) |
2368 | -ENOMEDIUM No media inserted. | |
2369 | -EINVAL Invalid sector number or nb_sectors | |
2370 | -EACCES Trying to write a read-only device | |
2371 | */ | |
5fafdf24 | 2372 | int bdrv_write(BlockDriverState *bs, int64_t sector_num, |
fc01f7e7 FB |
2373 | const uint8_t *buf, int nb_sectors) |
2374 | { | |
4105eaaa | 2375 | return bdrv_rw_co(bs, sector_num, (uint8_t *)buf, nb_sectors, true, 0); |
83f64091 FB |
2376 | } |
2377 | ||
8d3b1a2d KW |
2378 | int bdrv_writev(BlockDriverState *bs, int64_t sector_num, QEMUIOVector *qiov) |
2379 | { | |
4105eaaa PL |
2380 | return bdrv_rwv_co(bs, sector_num, qiov, true, 0); |
2381 | } | |
2382 | ||
2383 | int bdrv_write_zeroes(BlockDriverState *bs, int64_t sector_num, int nb_sectors) | |
2384 | { | |
2385 | return bdrv_rw_co(bs, sector_num, NULL, nb_sectors, true, | |
2386 | BDRV_REQ_ZERO_WRITE); | |
8d3b1a2d KW |
2387 | } |
2388 | ||
eda578e5 AL |
2389 | int bdrv_pread(BlockDriverState *bs, int64_t offset, |
2390 | void *buf, int count1) | |
83f64091 | 2391 | { |
6ea44308 | 2392 | uint8_t tmp_buf[BDRV_SECTOR_SIZE]; |
83f64091 FB |
2393 | int len, nb_sectors, count; |
2394 | int64_t sector_num; | |
9a8c4cce | 2395 | int ret; |
83f64091 FB |
2396 | |
2397 | count = count1; | |
2398 | /* first read to align to sector start */ | |
6ea44308 | 2399 | len = (BDRV_SECTOR_SIZE - offset) & (BDRV_SECTOR_SIZE - 1); |
83f64091 FB |
2400 | if (len > count) |
2401 | len = count; | |
6ea44308 | 2402 | sector_num = offset >> BDRV_SECTOR_BITS; |
83f64091 | 2403 | if (len > 0) { |
9a8c4cce KW |
2404 | if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0) |
2405 | return ret; | |
6ea44308 | 2406 | memcpy(buf, tmp_buf + (offset & (BDRV_SECTOR_SIZE - 1)), len); |
83f64091 FB |
2407 | count -= len; |
2408 | if (count == 0) | |
2409 | return count1; | |
2410 | sector_num++; | |
2411 | buf += len; | |
2412 | } | |
2413 | ||
2414 | /* read the sectors "in place" */ | |
6ea44308 | 2415 | nb_sectors = count >> BDRV_SECTOR_BITS; |
83f64091 | 2416 | if (nb_sectors > 0) { |
9a8c4cce KW |
2417 | if ((ret = bdrv_read(bs, sector_num, buf, nb_sectors)) < 0) |
2418 | return ret; | |
83f64091 | 2419 | sector_num += nb_sectors; |
6ea44308 | 2420 | len = nb_sectors << BDRV_SECTOR_BITS; |
83f64091 FB |
2421 | buf += len; |
2422 | count -= len; | |
2423 | } | |
2424 | ||
2425 | /* add data from the last sector */ | |
2426 | if (count > 0) { | |
9a8c4cce KW |
2427 | if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0) |
2428 | return ret; | |
83f64091 FB |
2429 | memcpy(buf, tmp_buf, count); |
2430 | } | |
2431 | return count1; | |
2432 | } | |
2433 | ||
8d3b1a2d | 2434 | int bdrv_pwritev(BlockDriverState *bs, int64_t offset, QEMUIOVector *qiov) |
83f64091 | 2435 | { |
6ea44308 | 2436 | uint8_t tmp_buf[BDRV_SECTOR_SIZE]; |
83f64091 FB |
2437 | int len, nb_sectors, count; |
2438 | int64_t sector_num; | |
9a8c4cce | 2439 | int ret; |
83f64091 | 2440 | |
8d3b1a2d KW |
2441 | count = qiov->size; |
2442 | ||
83f64091 | 2443 | /* first write to align to sector start */ |
6ea44308 | 2444 | len = (BDRV_SECTOR_SIZE - offset) & (BDRV_SECTOR_SIZE - 1); |
83f64091 FB |
2445 | if (len > count) |
2446 | len = count; | |
6ea44308 | 2447 | sector_num = offset >> BDRV_SECTOR_BITS; |
83f64091 | 2448 | if (len > 0) { |
9a8c4cce KW |
2449 | if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0) |
2450 | return ret; | |
8d3b1a2d KW |
2451 | qemu_iovec_to_buf(qiov, 0, tmp_buf + (offset & (BDRV_SECTOR_SIZE - 1)), |
2452 | len); | |
9a8c4cce KW |
2453 | if ((ret = bdrv_write(bs, sector_num, tmp_buf, 1)) < 0) |
2454 | return ret; | |
83f64091 FB |
2455 | count -= len; |
2456 | if (count == 0) | |
8d3b1a2d | 2457 | return qiov->size; |
83f64091 | 2458 | sector_num++; |
83f64091 FB |
2459 | } |
2460 | ||
2461 | /* write the sectors "in place" */ | |
6ea44308 | 2462 | nb_sectors = count >> BDRV_SECTOR_BITS; |
83f64091 | 2463 | if (nb_sectors > 0) { |
8d3b1a2d KW |
2464 | QEMUIOVector qiov_inplace; |
2465 | ||
2466 | qemu_iovec_init(&qiov_inplace, qiov->niov); | |
2467 | qemu_iovec_concat(&qiov_inplace, qiov, len, | |
2468 | nb_sectors << BDRV_SECTOR_BITS); | |
2469 | ret = bdrv_writev(bs, sector_num, &qiov_inplace); | |
2470 | qemu_iovec_destroy(&qiov_inplace); | |
2471 | if (ret < 0) { | |
9a8c4cce | 2472 | return ret; |
8d3b1a2d KW |
2473 | } |
2474 | ||
83f64091 | 2475 | sector_num += nb_sectors; |
6ea44308 | 2476 | len = nb_sectors << BDRV_SECTOR_BITS; |
83f64091 FB |
2477 | count -= len; |
2478 | } | |
2479 | ||
2480 | /* add data from the last sector */ | |
2481 | if (count > 0) { | |
9a8c4cce KW |
2482 | if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0) |
2483 | return ret; | |
8d3b1a2d | 2484 | qemu_iovec_to_buf(qiov, qiov->size - count, tmp_buf, count); |
9a8c4cce KW |
2485 | if ((ret = bdrv_write(bs, sector_num, tmp_buf, 1)) < 0) |
2486 | return ret; | |
83f64091 | 2487 | } |
8d3b1a2d KW |
2488 | return qiov->size; |
2489 | } | |
2490 | ||
2491 | int bdrv_pwrite(BlockDriverState *bs, int64_t offset, | |
2492 | const void *buf, int count1) | |
2493 | { | |
2494 | QEMUIOVector qiov; | |
2495 | struct iovec iov = { | |
2496 | .iov_base = (void *) buf, | |
2497 | .iov_len = count1, | |
2498 | }; | |
2499 | ||
2500 | qemu_iovec_init_external(&qiov, &iov, 1); | |
2501 | return bdrv_pwritev(bs, offset, &qiov); | |
83f64091 | 2502 | } |
83f64091 | 2503 | |
f08145fe KW |
2504 | /* |
2505 | * Writes to the file and ensures that no writes are reordered across this | |
2506 | * request (acts as a barrier) | |
2507 | * | |
2508 | * Returns 0 on success, -errno in error cases. | |
2509 | */ | |
2510 | int bdrv_pwrite_sync(BlockDriverState *bs, int64_t offset, | |
2511 | const void *buf, int count) | |
2512 | { | |
2513 | int ret; | |
2514 | ||
2515 | ret = bdrv_pwrite(bs, offset, buf, count); | |
2516 | if (ret < 0) { | |
2517 | return ret; | |
2518 | } | |
2519 | ||
f05fa4ad PB |
2520 | /* No flush needed for cache modes that already do it */ |
2521 | if (bs->enable_write_cache) { | |
f08145fe KW |
2522 | bdrv_flush(bs); |
2523 | } | |
2524 | ||
2525 | return 0; | |
2526 | } | |
2527 | ||
470c0504 | 2528 | static int coroutine_fn bdrv_co_do_copy_on_readv(BlockDriverState *bs, |
ab185921 SH |
2529 | int64_t sector_num, int nb_sectors, QEMUIOVector *qiov) |
2530 | { | |
2531 | /* Perform I/O through a temporary buffer so that users who scribble over | |
2532 | * their read buffer while the operation is in progress do not end up | |
2533 | * modifying the image file. This is critical for zero-copy guest I/O | |
2534 | * where anything might happen inside guest memory. | |
2535 | */ | |
2536 | void *bounce_buffer; | |
2537 | ||
79c053bd | 2538 | BlockDriver *drv = bs->drv; |
ab185921 SH |
2539 | struct iovec iov; |
2540 | QEMUIOVector bounce_qiov; | |
2541 | int64_t cluster_sector_num; | |
2542 | int cluster_nb_sectors; | |
2543 | size_t skip_bytes; | |
2544 | int ret; | |
2545 | ||
2546 | /* Cover entire cluster so no additional backing file I/O is required when | |
2547 | * allocating cluster in the image file. | |
2548 | */ | |
343bded4 PB |
2549 | bdrv_round_to_clusters(bs, sector_num, nb_sectors, |
2550 | &cluster_sector_num, &cluster_nb_sectors); | |
ab185921 | 2551 | |
470c0504 SH |
2552 | trace_bdrv_co_do_copy_on_readv(bs, sector_num, nb_sectors, |
2553 | cluster_sector_num, cluster_nb_sectors); | |
ab185921 SH |
2554 | |
2555 | iov.iov_len = cluster_nb_sectors * BDRV_SECTOR_SIZE; | |
2556 | iov.iov_base = bounce_buffer = qemu_blockalign(bs, iov.iov_len); | |
2557 | qemu_iovec_init_external(&bounce_qiov, &iov, 1); | |
2558 | ||
79c053bd SH |
2559 | ret = drv->bdrv_co_readv(bs, cluster_sector_num, cluster_nb_sectors, |
2560 | &bounce_qiov); | |
ab185921 SH |
2561 | if (ret < 0) { |
2562 | goto err; | |
2563 | } | |
2564 | ||
79c053bd SH |
2565 | if (drv->bdrv_co_write_zeroes && |
2566 | buffer_is_zero(bounce_buffer, iov.iov_len)) { | |
621f0589 KW |
2567 | ret = bdrv_co_do_write_zeroes(bs, cluster_sector_num, |
2568 | cluster_nb_sectors); | |
79c053bd | 2569 | } else { |
f05fa4ad PB |
2570 | /* This does not change the data on the disk, it is not necessary |
2571 | * to flush even in cache=writethrough mode. | |
2572 | */ | |
79c053bd | 2573 | ret = drv->bdrv_co_writev(bs, cluster_sector_num, cluster_nb_sectors, |
ab185921 | 2574 | &bounce_qiov); |
79c053bd SH |
2575 | } |
2576 | ||
ab185921 SH |
2577 | if (ret < 0) { |
2578 | /* It might be okay to ignore write errors for guest requests. If this | |
2579 | * is a deliberate copy-on-read then we don't want to ignore the error. | |
2580 | * Simply report it in all cases. | |
2581 | */ | |
2582 | goto err; | |
2583 | } | |
2584 | ||
2585 | skip_bytes = (sector_num - cluster_sector_num) * BDRV_SECTOR_SIZE; | |
03396148 MT |
2586 | qemu_iovec_from_buf(qiov, 0, bounce_buffer + skip_bytes, |
2587 | nb_sectors * BDRV_SECTOR_SIZE); | |
ab185921 SH |
2588 | |
2589 | err: | |
2590 | qemu_vfree(bounce_buffer); | |
2591 | return ret; | |
2592 | } | |
2593 | ||
c5fbe571 SH |
2594 | /* |
2595 | * Handle a read request in coroutine context | |
2596 | */ | |
2597 | static int coroutine_fn bdrv_co_do_readv(BlockDriverState *bs, | |
470c0504 SH |
2598 | int64_t sector_num, int nb_sectors, QEMUIOVector *qiov, |
2599 | BdrvRequestFlags flags) | |
da1fa91d KW |
2600 | { |
2601 | BlockDriver *drv = bs->drv; | |
dbffbdcf SH |
2602 | BdrvTrackedRequest req; |
2603 | int ret; | |
da1fa91d | 2604 | |
da1fa91d KW |
2605 | if (!drv) { |
2606 | return -ENOMEDIUM; | |
2607 | } | |
2608 | if (bdrv_check_request(bs, sector_num, nb_sectors)) { | |
2609 | return -EIO; | |
2610 | } | |
2611 | ||
f4658285 | 2612 | if (bs->copy_on_read) { |
470c0504 SH |
2613 | flags |= BDRV_REQ_COPY_ON_READ; |
2614 | } | |
2615 | if (flags & BDRV_REQ_COPY_ON_READ) { | |
2616 | bs->copy_on_read_in_flight++; | |
2617 | } | |
2618 | ||
2619 | if (bs->copy_on_read_in_flight) { | |
f4658285 SH |
2620 | wait_for_overlapping_requests(bs, sector_num, nb_sectors); |
2621 | } | |
2622 | ||
cc0681c4 BC |
2623 | /* throttling disk I/O */ |
2624 | if (bs->io_limits_enabled) { | |
2625 | bdrv_io_limits_intercept(bs, nb_sectors, false); | |
2626 | } | |
2627 | ||
dbffbdcf | 2628 | tracked_request_begin(&req, bs, sector_num, nb_sectors, false); |
ab185921 | 2629 | |
470c0504 | 2630 | if (flags & BDRV_REQ_COPY_ON_READ) { |
ab185921 SH |
2631 | int pnum; |
2632 | ||
bdad13b9 | 2633 | ret = bdrv_is_allocated(bs, sector_num, nb_sectors, &pnum); |
ab185921 SH |
2634 | if (ret < 0) { |
2635 | goto out; | |
2636 | } | |
2637 | ||
2638 | if (!ret || pnum != nb_sectors) { | |
470c0504 | 2639 | ret = bdrv_co_do_copy_on_readv(bs, sector_num, nb_sectors, qiov); |
ab185921 SH |
2640 | goto out; |
2641 | } | |
2642 | } | |
2643 | ||
893a8f62 MK |
2644 | if (!(bs->zero_beyond_eof && bs->growable)) { |
2645 | ret = drv->bdrv_co_readv(bs, sector_num, nb_sectors, qiov); | |
2646 | } else { | |
2647 | /* Read zeros after EOF of growable BDSes */ | |
2648 | int64_t len, total_sectors, max_nb_sectors; | |
2649 | ||
2650 | len = bdrv_getlength(bs); | |
2651 | if (len < 0) { | |
2652 | ret = len; | |
2653 | goto out; | |
2654 | } | |
2655 | ||
d055a1fe | 2656 | total_sectors = DIV_ROUND_UP(len, BDRV_SECTOR_SIZE); |
893a8f62 MK |
2657 | max_nb_sectors = MAX(0, total_sectors - sector_num); |
2658 | if (max_nb_sectors > 0) { | |
2659 | ret = drv->bdrv_co_readv(bs, sector_num, | |
2660 | MIN(nb_sectors, max_nb_sectors), qiov); | |
2661 | } else { | |
2662 | ret = 0; | |
2663 | } | |
2664 | ||
2665 | /* Reading beyond end of file is supposed to produce zeroes */ | |
2666 | if (ret == 0 && total_sectors < sector_num + nb_sectors) { | |
2667 | uint64_t offset = MAX(0, total_sectors - sector_num); | |
2668 | uint64_t bytes = (sector_num + nb_sectors - offset) * | |
2669 | BDRV_SECTOR_SIZE; | |
2670 | qemu_iovec_memset(qiov, offset * BDRV_SECTOR_SIZE, 0, bytes); | |
2671 | } | |
2672 | } | |
ab185921 SH |
2673 | |
2674 | out: | |
dbffbdcf | 2675 | tracked_request_end(&req); |
470c0504 SH |
2676 | |
2677 | if (flags & BDRV_REQ_COPY_ON_READ) { | |
2678 | bs->copy_on_read_in_flight--; | |
2679 | } | |
2680 | ||
dbffbdcf | 2681 | return ret; |
da1fa91d KW |
2682 | } |
2683 | ||
c5fbe571 | 2684 | int coroutine_fn bdrv_co_readv(BlockDriverState *bs, int64_t sector_num, |
da1fa91d KW |
2685 | int nb_sectors, QEMUIOVector *qiov) |
2686 | { | |
c5fbe571 | 2687 | trace_bdrv_co_readv(bs, sector_num, nb_sectors); |
da1fa91d | 2688 | |
470c0504 SH |
2689 | return bdrv_co_do_readv(bs, sector_num, nb_sectors, qiov, 0); |
2690 | } | |
2691 | ||
2692 | int coroutine_fn bdrv_co_copy_on_readv(BlockDriverState *bs, | |
2693 | int64_t sector_num, int nb_sectors, QEMUIOVector *qiov) | |
2694 | { | |
2695 | trace_bdrv_co_copy_on_readv(bs, sector_num, nb_sectors); | |
2696 | ||
2697 | return bdrv_co_do_readv(bs, sector_num, nb_sectors, qiov, | |
2698 | BDRV_REQ_COPY_ON_READ); | |
c5fbe571 SH |
2699 | } |
2700 | ||
f08f2dda SH |
2701 | static int coroutine_fn bdrv_co_do_write_zeroes(BlockDriverState *bs, |
2702 | int64_t sector_num, int nb_sectors) | |
2703 | { | |
2704 | BlockDriver *drv = bs->drv; | |
2705 | QEMUIOVector qiov; | |
2706 | struct iovec iov; | |
2707 | int ret; | |
2708 | ||
621f0589 KW |
2709 | /* TODO Emulate only part of misaligned requests instead of letting block |
2710 | * drivers return -ENOTSUP and emulate everything */ | |
2711 | ||
f08f2dda SH |
2712 | /* First try the efficient write zeroes operation */ |
2713 | if (drv->bdrv_co_write_zeroes) { | |
621f0589 KW |
2714 | ret = drv->bdrv_co_write_zeroes(bs, sector_num, nb_sectors); |
2715 | if (ret != -ENOTSUP) { | |
2716 | return ret; | |
2717 | } | |
f08f2dda SH |
2718 | } |
2719 | ||
2720 | /* Fall back to bounce buffer if write zeroes is unsupported */ | |
2721 | iov.iov_len = nb_sectors * BDRV_SECTOR_SIZE; | |
2722 | iov.iov_base = qemu_blockalign(bs, iov.iov_len); | |
2723 | memset(iov.iov_base, 0, iov.iov_len); | |
2724 | qemu_iovec_init_external(&qiov, &iov, 1); | |
2725 | ||
2726 | ret = drv->bdrv_co_writev(bs, sector_num, nb_sectors, &qiov); | |
2727 | ||
2728 | qemu_vfree(iov.iov_base); | |
2729 | return ret; | |
2730 | } | |
2731 | ||
c5fbe571 SH |
2732 | /* |
2733 | * Handle a write request in coroutine context | |
2734 | */ | |
2735 | static int coroutine_fn bdrv_co_do_writev(BlockDriverState *bs, | |
f08f2dda SH |
2736 | int64_t sector_num, int nb_sectors, QEMUIOVector *qiov, |
2737 | BdrvRequestFlags flags) | |
c5fbe571 SH |
2738 | { |
2739 | BlockDriver *drv = bs->drv; | |
dbffbdcf | 2740 | BdrvTrackedRequest req; |
6b7cb247 | 2741 | int ret; |
da1fa91d KW |
2742 | |
2743 | if (!bs->drv) { | |
2744 | return -ENOMEDIUM; | |
2745 | } | |
2746 | if (bs->read_only) { | |
2747 | return -EACCES; | |
2748 | } | |
2749 | if (bdrv_check_request(bs, sector_num, nb_sectors)) { | |
2750 | return -EIO; | |
2751 | } | |
2752 | ||
470c0504 | 2753 | if (bs->copy_on_read_in_flight) { |
f4658285 SH |
2754 | wait_for_overlapping_requests(bs, sector_num, nb_sectors); |
2755 | } | |
2756 | ||
cc0681c4 BC |
2757 | /* throttling disk I/O */ |
2758 | if (bs->io_limits_enabled) { | |
2759 | bdrv_io_limits_intercept(bs, nb_sectors, true); | |
2760 | } | |
2761 | ||
dbffbdcf SH |
2762 | tracked_request_begin(&req, bs, sector_num, nb_sectors, true); |
2763 | ||
d616b224 SH |
2764 | ret = notifier_with_return_list_notify(&bs->before_write_notifiers, &req); |
2765 | ||
2766 | if (ret < 0) { | |
2767 | /* Do nothing, write notifier decided to fail this request */ | |
2768 | } else if (flags & BDRV_REQ_ZERO_WRITE) { | |
f08f2dda SH |
2769 | ret = bdrv_co_do_write_zeroes(bs, sector_num, nb_sectors); |
2770 | } else { | |
2771 | ret = drv->bdrv_co_writev(bs, sector_num, nb_sectors, qiov); | |
2772 | } | |
6b7cb247 | 2773 | |
f05fa4ad PB |
2774 | if (ret == 0 && !bs->enable_write_cache) { |
2775 | ret = bdrv_co_flush(bs); | |
2776 | } | |
2777 | ||
da1fa91d | 2778 | if (bs->dirty_bitmap) { |
1755da16 | 2779 | bdrv_set_dirty(bs, sector_num, nb_sectors); |
da1fa91d KW |
2780 | } |
2781 | ||
2782 | if (bs->wr_highest_sector < sector_num + nb_sectors - 1) { | |
2783 | bs->wr_highest_sector = sector_num + nb_sectors - 1; | |
2784 | } | |
df2a6f29 PB |
2785 | if (bs->growable && ret >= 0) { |
2786 | bs->total_sectors = MAX(bs->total_sectors, sector_num + nb_sectors); | |
2787 | } | |
da1fa91d | 2788 | |
dbffbdcf SH |
2789 | tracked_request_end(&req); |
2790 | ||
6b7cb247 | 2791 | return ret; |
da1fa91d KW |
2792 | } |
2793 | ||
c5fbe571 SH |
2794 | int coroutine_fn bdrv_co_writev(BlockDriverState *bs, int64_t sector_num, |
2795 | int nb_sectors, QEMUIOVector *qiov) | |
2796 | { | |
2797 | trace_bdrv_co_writev(bs, sector_num, nb_sectors); | |
2798 | ||
f08f2dda SH |
2799 | return bdrv_co_do_writev(bs, sector_num, nb_sectors, qiov, 0); |
2800 | } | |
2801 | ||
2802 | int coroutine_fn bdrv_co_write_zeroes(BlockDriverState *bs, | |
2803 | int64_t sector_num, int nb_sectors) | |
2804 | { | |
2805 | trace_bdrv_co_write_zeroes(bs, sector_num, nb_sectors); | |
2806 | ||
2807 | return bdrv_co_do_writev(bs, sector_num, nb_sectors, NULL, | |
2808 | BDRV_REQ_ZERO_WRITE); | |
c5fbe571 SH |
2809 | } |
2810 | ||
83f64091 FB |
2811 | /** |
2812 | * Truncate file to 'offset' bytes (needed only for file protocols) | |
2813 | */ | |
2814 | int bdrv_truncate(BlockDriverState *bs, int64_t offset) | |
2815 | { | |
2816 | BlockDriver *drv = bs->drv; | |
51762288 | 2817 | int ret; |
83f64091 | 2818 | if (!drv) |
19cb3738 | 2819 | return -ENOMEDIUM; |
83f64091 FB |
2820 | if (!drv->bdrv_truncate) |
2821 | return -ENOTSUP; | |
59f2689d NS |
2822 | if (bs->read_only) |
2823 | return -EACCES; | |
8591675f MT |
2824 | if (bdrv_in_use(bs)) |
2825 | return -EBUSY; | |
51762288 SH |
2826 | ret = drv->bdrv_truncate(bs, offset); |
2827 | if (ret == 0) { | |
2828 | ret = refresh_total_sectors(bs, offset >> BDRV_SECTOR_BITS); | |
145feb17 | 2829 | bdrv_dev_resize_cb(bs); |
51762288 SH |
2830 | } |
2831 | return ret; | |
83f64091 FB |
2832 | } |
2833 | ||
4a1d5e1f FZ |
2834 | /** |
2835 | * Length of a allocated file in bytes. Sparse files are counted by actual | |
2836 | * allocated space. Return < 0 if error or unknown. | |
2837 | */ | |
2838 | int64_t bdrv_get_allocated_file_size(BlockDriverState *bs) | |
2839 | { | |
2840 | BlockDriver *drv = bs->drv; | |
2841 | if (!drv) { | |
2842 | return -ENOMEDIUM; | |
2843 | } | |
2844 | if (drv->bdrv_get_allocated_file_size) { | |
2845 | return drv->bdrv_get_allocated_file_size(bs); | |
2846 | } | |
2847 | if (bs->file) { | |
2848 | return bdrv_get_allocated_file_size(bs->file); | |
2849 | } | |
2850 | return -ENOTSUP; | |
2851 | } | |
2852 | ||
83f64091 FB |
2853 | /** |
2854 | * Length of a file in bytes. Return < 0 if error or unknown. | |
2855 | */ | |
2856 | int64_t bdrv_getlength(BlockDriverState *bs) | |
2857 | { | |
2858 | BlockDriver *drv = bs->drv; | |
2859 | if (!drv) | |
19cb3738 | 2860 | return -ENOMEDIUM; |
51762288 | 2861 | |
df2a6f29 | 2862 | if (bdrv_dev_has_removable_media(bs)) { |
46a4e4e6 SH |
2863 | if (drv->bdrv_getlength) { |
2864 | return drv->bdrv_getlength(bs); | |
2865 | } | |
83f64091 | 2866 | } |
46a4e4e6 | 2867 | return bs->total_sectors * BDRV_SECTOR_SIZE; |
fc01f7e7 FB |
2868 | } |
2869 | ||
19cb3738 | 2870 | /* return 0 as number of sectors if no device present or error */ |
96b8f136 | 2871 | void bdrv_get_geometry(BlockDriverState *bs, uint64_t *nb_sectors_ptr) |
fc01f7e7 | 2872 | { |
19cb3738 FB |
2873 | int64_t length; |
2874 | length = bdrv_getlength(bs); | |
2875 | if (length < 0) | |
2876 | length = 0; | |
2877 | else | |
6ea44308 | 2878 | length = length >> BDRV_SECTOR_BITS; |
19cb3738 | 2879 | *nb_sectors_ptr = length; |
fc01f7e7 | 2880 | } |
cf98951b | 2881 | |
ff06f5f3 PB |
2882 | void bdrv_set_on_error(BlockDriverState *bs, BlockdevOnError on_read_error, |
2883 | BlockdevOnError on_write_error) | |
abd7f68d MA |
2884 | { |
2885 | bs->on_read_error = on_read_error; | |
2886 | bs->on_write_error = on_write_error; | |
2887 | } | |
2888 | ||
1ceee0d5 | 2889 | BlockdevOnError bdrv_get_on_error(BlockDriverState *bs, bool is_read) |
abd7f68d MA |
2890 | { |
2891 | return is_read ? bs->on_read_error : bs->on_write_error; | |
2892 | } | |
2893 | ||
3e1caa5f PB |
2894 | BlockErrorAction bdrv_get_error_action(BlockDriverState *bs, bool is_read, int error) |
2895 | { | |
2896 | BlockdevOnError on_err = is_read ? bs->on_read_error : bs->on_write_error; | |
2897 | ||
2898 | switch (on_err) { | |
2899 | case BLOCKDEV_ON_ERROR_ENOSPC: | |
2900 | return (error == ENOSPC) ? BDRV_ACTION_STOP : BDRV_ACTION_REPORT; | |
2901 | case BLOCKDEV_ON_ERROR_STOP: | |
2902 | return BDRV_ACTION_STOP; | |
2903 | case BLOCKDEV_ON_ERROR_REPORT: | |
2904 | return BDRV_ACTION_REPORT; | |
2905 | case BLOCKDEV_ON_ERROR_IGNORE: | |
2906 | return BDRV_ACTION_IGNORE; | |
2907 | default: | |
2908 | abort(); | |
2909 | } | |
2910 | } | |
2911 | ||
2912 | /* This is done by device models because, while the block layer knows | |
2913 | * about the error, it does not know whether an operation comes from | |
2914 | * the device or the block layer (from a job, for example). | |
2915 | */ | |
2916 | void bdrv_error_action(BlockDriverState *bs, BlockErrorAction action, | |
2917 | bool is_read, int error) | |
2918 | { | |
2919 | assert(error >= 0); | |
32c81a4a | 2920 | bdrv_emit_qmp_error_event(bs, QEVENT_BLOCK_IO_ERROR, action, is_read); |
3e1caa5f PB |
2921 | if (action == BDRV_ACTION_STOP) { |
2922 | vm_stop(RUN_STATE_IO_ERROR); | |
2923 | bdrv_iostatus_set_err(bs, error); | |
2924 | } | |
2925 | } | |
2926 | ||
b338082b FB |
2927 | int bdrv_is_read_only(BlockDriverState *bs) |
2928 | { | |
2929 | return bs->read_only; | |
2930 | } | |
2931 | ||
985a03b0 TS |
2932 | int bdrv_is_sg(BlockDriverState *bs) |
2933 | { | |
2934 | return bs->sg; | |
2935 | } | |
2936 | ||
e900a7b7 CH |
2937 | int bdrv_enable_write_cache(BlockDriverState *bs) |
2938 | { | |
2939 | return bs->enable_write_cache; | |
2940 | } | |
2941 | ||
425b0148 PB |
2942 | void bdrv_set_enable_write_cache(BlockDriverState *bs, bool wce) |
2943 | { | |
2944 | bs->enable_write_cache = wce; | |
55b110f2 JC |
2945 | |
2946 | /* so a reopen() will preserve wce */ | |
2947 | if (wce) { | |
2948 | bs->open_flags |= BDRV_O_CACHE_WB; | |
2949 | } else { | |
2950 | bs->open_flags &= ~BDRV_O_CACHE_WB; | |
2951 | } | |
425b0148 PB |
2952 | } |
2953 | ||
ea2384d3 FB |
2954 | int bdrv_is_encrypted(BlockDriverState *bs) |
2955 | { | |
2956 | if (bs->backing_hd && bs->backing_hd->encrypted) | |
2957 | return 1; | |
2958 | return bs->encrypted; | |
2959 | } | |
2960 | ||
c0f4ce77 AL |
2961 | int bdrv_key_required(BlockDriverState *bs) |
2962 | { | |
2963 | BlockDriverState *backing_hd = bs->backing_hd; | |
2964 | ||
2965 | if (backing_hd && backing_hd->encrypted && !backing_hd->valid_key) | |
2966 | return 1; | |
2967 | return (bs->encrypted && !bs->valid_key); | |
2968 | } | |
2969 | ||
ea2384d3 FB |
2970 | int bdrv_set_key(BlockDriverState *bs, const char *key) |
2971 | { | |
2972 | int ret; | |
2973 | if (bs->backing_hd && bs->backing_hd->encrypted) { | |
2974 | ret = bdrv_set_key(bs->backing_hd, key); | |
2975 | if (ret < 0) | |
2976 | return ret; | |
2977 | if (!bs->encrypted) | |
2978 | return 0; | |
2979 | } | |
fd04a2ae SH |
2980 | if (!bs->encrypted) { |
2981 | return -EINVAL; | |
2982 | } else if (!bs->drv || !bs->drv->bdrv_set_key) { | |
2983 | return -ENOMEDIUM; | |
2984 | } | |
c0f4ce77 | 2985 | ret = bs->drv->bdrv_set_key(bs, key); |
bb5fc20f AL |
2986 | if (ret < 0) { |
2987 | bs->valid_key = 0; | |
2988 | } else if (!bs->valid_key) { | |
2989 | bs->valid_key = 1; | |
2990 | /* call the change callback now, we skipped it on open */ | |
7d4b4ba5 | 2991 | bdrv_dev_change_media_cb(bs, true); |
bb5fc20f | 2992 | } |
c0f4ce77 | 2993 | return ret; |
ea2384d3 FB |
2994 | } |
2995 | ||
f8d6bba1 | 2996 | const char *bdrv_get_format_name(BlockDriverState *bs) |
ea2384d3 | 2997 | { |
f8d6bba1 | 2998 | return bs->drv ? bs->drv->format_name : NULL; |
ea2384d3 FB |
2999 | } |
3000 | ||
5fafdf24 | 3001 | void bdrv_iterate_format(void (*it)(void *opaque, const char *name), |
ea2384d3 FB |
3002 | void *opaque) |
3003 | { | |
3004 | BlockDriver *drv; | |
3005 | ||
8a22f02a | 3006 | QLIST_FOREACH(drv, &bdrv_drivers, list) { |
ea2384d3 FB |
3007 | it(opaque, drv->format_name); |
3008 | } | |
3009 | } | |
3010 | ||
b338082b FB |
3011 | BlockDriverState *bdrv_find(const char *name) |
3012 | { | |
3013 | BlockDriverState *bs; | |
3014 | ||
1b7bdbc1 SH |
3015 | QTAILQ_FOREACH(bs, &bdrv_states, list) { |
3016 | if (!strcmp(name, bs->device_name)) { | |
b338082b | 3017 | return bs; |
1b7bdbc1 | 3018 | } |
b338082b FB |
3019 | } |
3020 | return NULL; | |
3021 | } | |
3022 | ||
2f399b0a MA |
3023 | BlockDriverState *bdrv_next(BlockDriverState *bs) |
3024 | { | |
3025 | if (!bs) { | |
3026 | return QTAILQ_FIRST(&bdrv_states); | |
3027 | } | |
3028 | return QTAILQ_NEXT(bs, list); | |
3029 | } | |
3030 | ||
51de9760 | 3031 | void bdrv_iterate(void (*it)(void *opaque, BlockDriverState *bs), void *opaque) |
81d0912d FB |
3032 | { |
3033 | BlockDriverState *bs; | |
3034 | ||
1b7bdbc1 | 3035 | QTAILQ_FOREACH(bs, &bdrv_states, list) { |
51de9760 | 3036 | it(opaque, bs); |
81d0912d FB |
3037 | } |
3038 | } | |
3039 | ||
ea2384d3 FB |
3040 | const char *bdrv_get_device_name(BlockDriverState *bs) |
3041 | { | |
3042 | return bs->device_name; | |
3043 | } | |
3044 | ||
c8433287 MA |
3045 | int bdrv_get_flags(BlockDriverState *bs) |
3046 | { | |
3047 | return bs->open_flags; | |
3048 | } | |
3049 | ||
f0f0fdfe | 3050 | int bdrv_flush_all(void) |
c6ca28d6 AL |
3051 | { |
3052 | BlockDriverState *bs; | |
f0f0fdfe | 3053 | int result = 0; |
c6ca28d6 | 3054 | |
1b7bdbc1 | 3055 | QTAILQ_FOREACH(bs, &bdrv_states, list) { |
f0f0fdfe KW |
3056 | int ret = bdrv_flush(bs); |
3057 | if (ret < 0 && !result) { | |
3058 | result = ret; | |
3059 | } | |
1b7bdbc1 | 3060 | } |
f0f0fdfe KW |
3061 | |
3062 | return result; | |
c6ca28d6 AL |
3063 | } |
3064 | ||
3ac21627 PL |
3065 | int bdrv_has_zero_init_1(BlockDriverState *bs) |
3066 | { | |
3067 | return 1; | |
3068 | } | |
3069 | ||
f2feebbd KW |
3070 | int bdrv_has_zero_init(BlockDriverState *bs) |
3071 | { | |
3072 | assert(bs->drv); | |
3073 | ||
11212d8f PB |
3074 | /* If BS is a copy on write image, it is initialized to |
3075 | the contents of the base image, which may not be zeroes. */ | |
3076 | if (bs->backing_hd) { | |
3077 | return 0; | |
3078 | } | |
336c1c12 KW |
3079 | if (bs->drv->bdrv_has_zero_init) { |
3080 | return bs->drv->bdrv_has_zero_init(bs); | |
f2feebbd KW |
3081 | } |
3082 | ||
3ac21627 PL |
3083 | /* safe default */ |
3084 | return 0; | |
f2feebbd KW |
3085 | } |
3086 | ||
b6b8a333 | 3087 | typedef struct BdrvCoGetBlockStatusData { |
376ae3f1 | 3088 | BlockDriverState *bs; |
b35b2bba | 3089 | BlockDriverState *base; |
376ae3f1 SH |
3090 | int64_t sector_num; |
3091 | int nb_sectors; | |
3092 | int *pnum; | |
b6b8a333 | 3093 | int64_t ret; |
376ae3f1 | 3094 | bool done; |
b6b8a333 | 3095 | } BdrvCoGetBlockStatusData; |
376ae3f1 | 3096 | |
f58c7b35 TS |
3097 | /* |
3098 | * Returns true iff the specified sector is present in the disk image. Drivers | |
3099 | * not implementing the functionality are assumed to not support backing files, | |
3100 | * hence all their sectors are reported as allocated. | |
3101 | * | |
bd9533e3 SH |
3102 | * If 'sector_num' is beyond the end of the disk image the return value is 0 |
3103 | * and 'pnum' is set to 0. | |
3104 | * | |
f58c7b35 TS |
3105 | * 'pnum' is set to the number of sectors (including and immediately following |
3106 | * the specified sector) that are known to be in the same | |
3107 | * allocated/unallocated state. | |
3108 | * | |
bd9533e3 SH |
3109 | * 'nb_sectors' is the max value 'pnum' should be set to. If nb_sectors goes |
3110 | * beyond the end of the disk image it will be clamped. | |
f58c7b35 | 3111 | */ |
b6b8a333 PB |
3112 | static int64_t coroutine_fn bdrv_co_get_block_status(BlockDriverState *bs, |
3113 | int64_t sector_num, | |
3114 | int nb_sectors, int *pnum) | |
f58c7b35 | 3115 | { |
617ccb46 | 3116 | int64_t length; |
bd9533e3 | 3117 | int64_t n; |
5daa74a6 | 3118 | int64_t ret, ret2; |
bd9533e3 | 3119 | |
617ccb46 PB |
3120 | length = bdrv_getlength(bs); |
3121 | if (length < 0) { | |
3122 | return length; | |
3123 | } | |
3124 | ||
3125 | if (sector_num >= (length >> BDRV_SECTOR_BITS)) { | |
bd9533e3 SH |
3126 | *pnum = 0; |
3127 | return 0; | |
3128 | } | |
3129 | ||
3130 | n = bs->total_sectors - sector_num; | |
3131 | if (n < nb_sectors) { | |
3132 | nb_sectors = n; | |
3133 | } | |
3134 | ||
b6b8a333 | 3135 | if (!bs->drv->bdrv_co_get_block_status) { |
bd9533e3 | 3136 | *pnum = nb_sectors; |
918e92d7 PB |
3137 | ret = BDRV_BLOCK_DATA; |
3138 | if (bs->drv->protocol_name) { | |
3139 | ret |= BDRV_BLOCK_OFFSET_VALID | (sector_num * BDRV_SECTOR_SIZE); | |
3140 | } | |
3141 | return ret; | |
f58c7b35 | 3142 | } |
6aebab14 | 3143 | |
415b5b01 PB |
3144 | ret = bs->drv->bdrv_co_get_block_status(bs, sector_num, nb_sectors, pnum); |
3145 | if (ret < 0) { | |
3e0a233d | 3146 | *pnum = 0; |
415b5b01 PB |
3147 | return ret; |
3148 | } | |
3149 | ||
f0ad5712 PB |
3150 | if (!(ret & BDRV_BLOCK_DATA)) { |
3151 | if (bdrv_has_zero_init(bs)) { | |
3152 | ret |= BDRV_BLOCK_ZERO; | |
1f9db224 | 3153 | } else if (bs->backing_hd) { |
f0ad5712 PB |
3154 | BlockDriverState *bs2 = bs->backing_hd; |
3155 | int64_t length2 = bdrv_getlength(bs2); | |
3156 | if (length2 >= 0 && sector_num >= (length2 >> BDRV_SECTOR_BITS)) { | |
3157 | ret |= BDRV_BLOCK_ZERO; | |
3158 | } | |
3159 | } | |
415b5b01 | 3160 | } |
5daa74a6 PB |
3161 | |
3162 | if (bs->file && | |
3163 | (ret & BDRV_BLOCK_DATA) && !(ret & BDRV_BLOCK_ZERO) && | |
3164 | (ret & BDRV_BLOCK_OFFSET_VALID)) { | |
3165 | ret2 = bdrv_co_get_block_status(bs->file, ret >> BDRV_SECTOR_BITS, | |
3166 | *pnum, pnum); | |
3167 | if (ret2 >= 0) { | |
3168 | /* Ignore errors. This is just providing extra information, it | |
3169 | * is useful but not necessary. | |
3170 | */ | |
3171 | ret |= (ret2 & BDRV_BLOCK_ZERO); | |
3172 | } | |
3173 | } | |
3174 | ||
415b5b01 | 3175 | return ret; |
060f51c9 SH |
3176 | } |
3177 | ||
b6b8a333 PB |
3178 | /* Coroutine wrapper for bdrv_get_block_status() */ |
3179 | static void coroutine_fn bdrv_get_block_status_co_entry(void *opaque) | |
060f51c9 | 3180 | { |
b6b8a333 | 3181 | BdrvCoGetBlockStatusData *data = opaque; |
060f51c9 SH |
3182 | BlockDriverState *bs = data->bs; |
3183 | ||
b6b8a333 PB |
3184 | data->ret = bdrv_co_get_block_status(bs, data->sector_num, data->nb_sectors, |
3185 | data->pnum); | |
060f51c9 SH |
3186 | data->done = true; |
3187 | } | |
3188 | ||
3189 | /* | |
b6b8a333 | 3190 | * Synchronous wrapper around bdrv_co_get_block_status(). |
060f51c9 | 3191 | * |
b6b8a333 | 3192 | * See bdrv_co_get_block_status() for details. |
060f51c9 | 3193 | */ |
b6b8a333 PB |
3194 | int64_t bdrv_get_block_status(BlockDriverState *bs, int64_t sector_num, |
3195 | int nb_sectors, int *pnum) | |
060f51c9 | 3196 | { |
6aebab14 | 3197 | Coroutine *co; |
b6b8a333 | 3198 | BdrvCoGetBlockStatusData data = { |
6aebab14 SH |
3199 | .bs = bs, |
3200 | .sector_num = sector_num, | |
3201 | .nb_sectors = nb_sectors, | |
3202 | .pnum = pnum, | |
3203 | .done = false, | |
3204 | }; | |
3205 | ||
bdad13b9 PB |
3206 | if (qemu_in_coroutine()) { |
3207 | /* Fast-path if already in coroutine context */ | |
b6b8a333 | 3208 | bdrv_get_block_status_co_entry(&data); |
bdad13b9 | 3209 | } else { |
b6b8a333 | 3210 | co = qemu_coroutine_create(bdrv_get_block_status_co_entry); |
bdad13b9 PB |
3211 | qemu_coroutine_enter(co, &data); |
3212 | while (!data.done) { | |
3213 | qemu_aio_wait(); | |
3214 | } | |
6aebab14 SH |
3215 | } |
3216 | return data.ret; | |
f58c7b35 TS |
3217 | } |
3218 | ||
b6b8a333 PB |
3219 | int coroutine_fn bdrv_is_allocated(BlockDriverState *bs, int64_t sector_num, |
3220 | int nb_sectors, int *pnum) | |
3221 | { | |
4333bb71 PB |
3222 | int64_t ret = bdrv_get_block_status(bs, sector_num, nb_sectors, pnum); |
3223 | if (ret < 0) { | |
3224 | return ret; | |
3225 | } | |
3226 | return | |
3227 | (ret & BDRV_BLOCK_DATA) || | |
3228 | ((ret & BDRV_BLOCK_ZERO) && !bdrv_has_zero_init(bs)); | |
b6b8a333 PB |
3229 | } |
3230 | ||
188a7bbf PB |
3231 | /* |
3232 | * Given an image chain: ... -> [BASE] -> [INTER1] -> [INTER2] -> [TOP] | |
3233 | * | |
3234 | * Return true if the given sector is allocated in any image between | |
3235 | * BASE and TOP (inclusive). BASE can be NULL to check if the given | |
3236 | * sector is allocated in any image of the chain. Return false otherwise. | |
3237 | * | |
3238 | * 'pnum' is set to the number of sectors (including and immediately following | |
3239 | * the specified sector) that are known to be in the same | |
3240 | * allocated/unallocated state. | |
3241 | * | |
3242 | */ | |
4f578637 PB |
3243 | int bdrv_is_allocated_above(BlockDriverState *top, |
3244 | BlockDriverState *base, | |
3245 | int64_t sector_num, | |
3246 | int nb_sectors, int *pnum) | |
188a7bbf PB |
3247 | { |
3248 | BlockDriverState *intermediate; | |
3249 | int ret, n = nb_sectors; | |
3250 | ||
3251 | intermediate = top; | |
3252 | while (intermediate && intermediate != base) { | |
3253 | int pnum_inter; | |
bdad13b9 PB |
3254 | ret = bdrv_is_allocated(intermediate, sector_num, nb_sectors, |
3255 | &pnum_inter); | |
188a7bbf PB |
3256 | if (ret < 0) { |
3257 | return ret; | |
3258 | } else if (ret) { | |
3259 | *pnum = pnum_inter; | |
3260 | return 1; | |
3261 | } | |
3262 | ||
3263 | /* | |
3264 | * [sector_num, nb_sectors] is unallocated on top but intermediate | |
3265 | * might have | |
3266 | * | |
3267 | * [sector_num+x, nr_sectors] allocated. | |
3268 | */ | |
63ba17d3 VI |
3269 | if (n > pnum_inter && |
3270 | (intermediate == top || | |
3271 | sector_num + pnum_inter < intermediate->total_sectors)) { | |
188a7bbf PB |
3272 | n = pnum_inter; |
3273 | } | |
3274 | ||
3275 | intermediate = intermediate->backing_hd; | |
3276 | } | |
3277 | ||
3278 | *pnum = n; | |
3279 | return 0; | |
3280 | } | |
3281 | ||
045df330 AL |
3282 | const char *bdrv_get_encrypted_filename(BlockDriverState *bs) |
3283 | { | |
3284 | if (bs->backing_hd && bs->backing_hd->encrypted) | |
3285 | return bs->backing_file; | |
3286 | else if (bs->encrypted) | |
3287 | return bs->filename; | |
3288 | else | |
3289 | return NULL; | |
3290 | } | |
3291 | ||
5fafdf24 | 3292 | void bdrv_get_backing_filename(BlockDriverState *bs, |
83f64091 FB |
3293 | char *filename, int filename_size) |
3294 | { | |
3574c608 | 3295 | pstrcpy(filename, filename_size, bs->backing_file); |
83f64091 FB |
3296 | } |
3297 | ||
5fafdf24 | 3298 | int bdrv_write_compressed(BlockDriverState *bs, int64_t sector_num, |
faea38e7 FB |
3299 | const uint8_t *buf, int nb_sectors) |
3300 | { | |
3301 | BlockDriver *drv = bs->drv; | |
3302 | if (!drv) | |
19cb3738 | 3303 | return -ENOMEDIUM; |
faea38e7 FB |
3304 | if (!drv->bdrv_write_compressed) |
3305 | return -ENOTSUP; | |
fbb7b4e0 KW |
3306 | if (bdrv_check_request(bs, sector_num, nb_sectors)) |
3307 | return -EIO; | |
a55eb92c | 3308 | |
1755da16 | 3309 | assert(!bs->dirty_bitmap); |
a55eb92c | 3310 | |
faea38e7 FB |
3311 | return drv->bdrv_write_compressed(bs, sector_num, buf, nb_sectors); |
3312 | } | |
3b46e624 | 3313 | |
faea38e7 FB |
3314 | int bdrv_get_info(BlockDriverState *bs, BlockDriverInfo *bdi) |
3315 | { | |
3316 | BlockDriver *drv = bs->drv; | |
3317 | if (!drv) | |
19cb3738 | 3318 | return -ENOMEDIUM; |
faea38e7 FB |
3319 | if (!drv->bdrv_get_info) |
3320 | return -ENOTSUP; | |
3321 | memset(bdi, 0, sizeof(*bdi)); | |
3322 | return drv->bdrv_get_info(bs, bdi); | |
3323 | } | |
3324 | ||
45566e9c CH |
3325 | int bdrv_save_vmstate(BlockDriverState *bs, const uint8_t *buf, |
3326 | int64_t pos, int size) | |
cf8074b3 KW |
3327 | { |
3328 | QEMUIOVector qiov; | |
3329 | struct iovec iov = { | |
3330 | .iov_base = (void *) buf, | |
3331 | .iov_len = size, | |
3332 | }; | |
3333 | ||
3334 | qemu_iovec_init_external(&qiov, &iov, 1); | |
3335 | return bdrv_writev_vmstate(bs, &qiov, pos); | |
3336 | } | |
3337 | ||
3338 | int bdrv_writev_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos) | |
178e08a5 AL |
3339 | { |
3340 | BlockDriver *drv = bs->drv; | |
cf8074b3 KW |
3341 | |
3342 | if (!drv) { | |
178e08a5 | 3343 | return -ENOMEDIUM; |
cf8074b3 KW |
3344 | } else if (drv->bdrv_save_vmstate) { |
3345 | return drv->bdrv_save_vmstate(bs, qiov, pos); | |
3346 | } else if (bs->file) { | |
3347 | return bdrv_writev_vmstate(bs->file, qiov, pos); | |
3348 | } | |
3349 | ||
7cdb1f6d | 3350 | return -ENOTSUP; |
178e08a5 AL |
3351 | } |
3352 | ||
45566e9c CH |
3353 | int bdrv_load_vmstate(BlockDriverState *bs, uint8_t *buf, |
3354 | int64_t pos, int size) | |
178e08a5 AL |
3355 | { |
3356 | BlockDriver *drv = bs->drv; | |
3357 | if (!drv) | |
3358 | return -ENOMEDIUM; | |
7cdb1f6d MK |
3359 | if (drv->bdrv_load_vmstate) |
3360 | return drv->bdrv_load_vmstate(bs, buf, pos, size); | |
3361 | if (bs->file) | |
3362 | return bdrv_load_vmstate(bs->file, buf, pos, size); | |
3363 | return -ENOTSUP; | |
178e08a5 AL |
3364 | } |
3365 | ||
8b9b0cc2 KW |
3366 | void bdrv_debug_event(BlockDriverState *bs, BlkDebugEvent event) |
3367 | { | |
bf736fe3 | 3368 | if (!bs || !bs->drv || !bs->drv->bdrv_debug_event) { |
8b9b0cc2 KW |
3369 | return; |
3370 | } | |
3371 | ||
bf736fe3 | 3372 | bs->drv->bdrv_debug_event(bs, event); |
41c695c7 KW |
3373 | } |
3374 | ||
3375 | int bdrv_debug_breakpoint(BlockDriverState *bs, const char *event, | |
3376 | const char *tag) | |
3377 | { | |
3378 | while (bs && bs->drv && !bs->drv->bdrv_debug_breakpoint) { | |
3379 | bs = bs->file; | |
3380 | } | |
3381 | ||
3382 | if (bs && bs->drv && bs->drv->bdrv_debug_breakpoint) { | |
3383 | return bs->drv->bdrv_debug_breakpoint(bs, event, tag); | |
3384 | } | |
3385 | ||
3386 | return -ENOTSUP; | |
3387 | } | |
3388 | ||
3389 | int bdrv_debug_resume(BlockDriverState *bs, const char *tag) | |
3390 | { | |
3391 | while (bs && bs->drv && !bs->drv->bdrv_debug_resume) { | |
3392 | bs = bs->file; | |
3393 | } | |
8b9b0cc2 | 3394 | |
41c695c7 KW |
3395 | if (bs && bs->drv && bs->drv->bdrv_debug_resume) { |
3396 | return bs->drv->bdrv_debug_resume(bs, tag); | |
3397 | } | |
3398 | ||
3399 | return -ENOTSUP; | |
3400 | } | |
3401 | ||
3402 | bool bdrv_debug_is_suspended(BlockDriverState *bs, const char *tag) | |
3403 | { | |
3404 | while (bs && bs->drv && !bs->drv->bdrv_debug_is_suspended) { | |
3405 | bs = bs->file; | |
3406 | } | |
3407 | ||
3408 | if (bs && bs->drv && bs->drv->bdrv_debug_is_suspended) { | |
3409 | return bs->drv->bdrv_debug_is_suspended(bs, tag); | |
3410 | } | |
3411 | ||
3412 | return false; | |
8b9b0cc2 KW |
3413 | } |
3414 | ||
199630b6 BS |
3415 | int bdrv_is_snapshot(BlockDriverState *bs) |
3416 | { | |
3417 | return !!(bs->open_flags & BDRV_O_SNAPSHOT); | |
3418 | } | |
3419 | ||
b1b1d783 JC |
3420 | /* backing_file can either be relative, or absolute, or a protocol. If it is |
3421 | * relative, it must be relative to the chain. So, passing in bs->filename | |
3422 | * from a BDS as backing_file should not be done, as that may be relative to | |
3423 | * the CWD rather than the chain. */ | |
e8a6bb9c MT |
3424 | BlockDriverState *bdrv_find_backing_image(BlockDriverState *bs, |
3425 | const char *backing_file) | |
3426 | { | |
b1b1d783 JC |
3427 | char *filename_full = NULL; |
3428 | char *backing_file_full = NULL; | |
3429 | char *filename_tmp = NULL; | |
3430 | int is_protocol = 0; | |
3431 | BlockDriverState *curr_bs = NULL; | |
3432 | BlockDriverState *retval = NULL; | |
3433 | ||
3434 | if (!bs || !bs->drv || !backing_file) { | |
e8a6bb9c MT |
3435 | return NULL; |
3436 | } | |
3437 | ||
b1b1d783 JC |
3438 | filename_full = g_malloc(PATH_MAX); |
3439 | backing_file_full = g_malloc(PATH_MAX); | |
3440 | filename_tmp = g_malloc(PATH_MAX); | |
3441 | ||
3442 | is_protocol = path_has_protocol(backing_file); | |
3443 | ||
3444 | for (curr_bs = bs; curr_bs->backing_hd; curr_bs = curr_bs->backing_hd) { | |
3445 | ||
3446 | /* If either of the filename paths is actually a protocol, then | |
3447 | * compare unmodified paths; otherwise make paths relative */ | |
3448 | if (is_protocol || path_has_protocol(curr_bs->backing_file)) { | |
3449 | if (strcmp(backing_file, curr_bs->backing_file) == 0) { | |
3450 | retval = curr_bs->backing_hd; | |
3451 | break; | |
3452 | } | |
e8a6bb9c | 3453 | } else { |
b1b1d783 JC |
3454 | /* If not an absolute filename path, make it relative to the current |
3455 | * image's filename path */ | |
3456 | path_combine(filename_tmp, PATH_MAX, curr_bs->filename, | |
3457 | backing_file); | |
3458 | ||
3459 | /* We are going to compare absolute pathnames */ | |
3460 | if (!realpath(filename_tmp, filename_full)) { | |
3461 | continue; | |
3462 | } | |
3463 | ||
3464 | /* We need to make sure the backing filename we are comparing against | |
3465 | * is relative to the current image filename (or absolute) */ | |
3466 | path_combine(filename_tmp, PATH_MAX, curr_bs->filename, | |
3467 | curr_bs->backing_file); | |
3468 | ||
3469 | if (!realpath(filename_tmp, backing_file_full)) { | |
3470 | continue; | |
3471 | } | |
3472 | ||
3473 | if (strcmp(backing_file_full, filename_full) == 0) { | |
3474 | retval = curr_bs->backing_hd; | |
3475 | break; | |
3476 | } | |
e8a6bb9c MT |
3477 | } |
3478 | } | |
3479 | ||
b1b1d783 JC |
3480 | g_free(filename_full); |
3481 | g_free(backing_file_full); | |
3482 | g_free(filename_tmp); | |
3483 | return retval; | |
e8a6bb9c MT |
3484 | } |
3485 | ||
f198fd1c BC |
3486 | int bdrv_get_backing_file_depth(BlockDriverState *bs) |
3487 | { | |
3488 | if (!bs->drv) { | |
3489 | return 0; | |
3490 | } | |
3491 | ||
3492 | if (!bs->backing_hd) { | |
3493 | return 0; | |
3494 | } | |
3495 | ||
3496 | return 1 + bdrv_get_backing_file_depth(bs->backing_hd); | |
3497 | } | |
3498 | ||
79fac568 JC |
3499 | BlockDriverState *bdrv_find_base(BlockDriverState *bs) |
3500 | { | |
3501 | BlockDriverState *curr_bs = NULL; | |
3502 | ||
3503 | if (!bs) { | |
3504 | return NULL; | |
3505 | } | |
3506 | ||
3507 | curr_bs = bs; | |
3508 | ||
3509 | while (curr_bs->backing_hd) { | |
3510 | curr_bs = curr_bs->backing_hd; | |
3511 | } | |
3512 | return curr_bs; | |
3513 | } | |
3514 | ||
ea2384d3 | 3515 | /**************************************************************/ |
83f64091 | 3516 | /* async I/Os */ |
ea2384d3 | 3517 | |
3b69e4b9 | 3518 | BlockDriverAIOCB *bdrv_aio_readv(BlockDriverState *bs, int64_t sector_num, |
f141eafe | 3519 | QEMUIOVector *qiov, int nb_sectors, |
3b69e4b9 | 3520 | BlockDriverCompletionFunc *cb, void *opaque) |
83f64091 | 3521 | { |
bbf0a440 SH |
3522 | trace_bdrv_aio_readv(bs, sector_num, nb_sectors, opaque); |
3523 | ||
b2a61371 | 3524 | return bdrv_co_aio_rw_vector(bs, sector_num, qiov, nb_sectors, |
8c5873d6 | 3525 | cb, opaque, false); |
ea2384d3 FB |
3526 | } |
3527 | ||
f141eafe AL |
3528 | BlockDriverAIOCB *bdrv_aio_writev(BlockDriverState *bs, int64_t sector_num, |
3529 | QEMUIOVector *qiov, int nb_sectors, | |
3530 | BlockDriverCompletionFunc *cb, void *opaque) | |
ea2384d3 | 3531 | { |
bbf0a440 SH |
3532 | trace_bdrv_aio_writev(bs, sector_num, nb_sectors, opaque); |
3533 | ||
1a6e115b | 3534 | return bdrv_co_aio_rw_vector(bs, sector_num, qiov, nb_sectors, |
8c5873d6 | 3535 | cb, opaque, true); |
83f64091 FB |
3536 | } |
3537 | ||
40b4f539 KW |
3538 | |
3539 | typedef struct MultiwriteCB { | |
3540 | int error; | |
3541 | int num_requests; | |
3542 | int num_callbacks; | |
3543 | struct { | |
3544 | BlockDriverCompletionFunc *cb; | |
3545 | void *opaque; | |
3546 | QEMUIOVector *free_qiov; | |
40b4f539 KW |
3547 | } callbacks[]; |
3548 | } MultiwriteCB; | |
3549 | ||
3550 | static void multiwrite_user_cb(MultiwriteCB *mcb) | |
3551 | { | |
3552 | int i; | |
3553 | ||
3554 | for (i = 0; i < mcb->num_callbacks; i++) { | |
3555 | mcb->callbacks[i].cb(mcb->callbacks[i].opaque, mcb->error); | |
1e1ea48d SH |
3556 | if (mcb->callbacks[i].free_qiov) { |
3557 | qemu_iovec_destroy(mcb->callbacks[i].free_qiov); | |
3558 | } | |
7267c094 | 3559 | g_free(mcb->callbacks[i].free_qiov); |
40b4f539 KW |
3560 | } |
3561 | } | |
3562 | ||
3563 | static void multiwrite_cb(void *opaque, int ret) | |
3564 | { | |
3565 | MultiwriteCB *mcb = opaque; | |
3566 | ||
6d519a5f SH |
3567 | trace_multiwrite_cb(mcb, ret); |
3568 | ||
cb6d3ca0 | 3569 | if (ret < 0 && !mcb->error) { |
40b4f539 | 3570 | mcb->error = ret; |
40b4f539 KW |
3571 | } |
3572 | ||
3573 | mcb->num_requests--; | |
3574 | if (mcb->num_requests == 0) { | |
de189a1b | 3575 | multiwrite_user_cb(mcb); |
7267c094 | 3576 | g_free(mcb); |
40b4f539 KW |
3577 | } |
3578 | } | |
3579 | ||
3580 | static int multiwrite_req_compare(const void *a, const void *b) | |
3581 | { | |
77be4366 CH |
3582 | const BlockRequest *req1 = a, *req2 = b; |
3583 | ||
3584 | /* | |
3585 | * Note that we can't simply subtract req2->sector from req1->sector | |
3586 | * here as that could overflow the return value. | |
3587 | */ | |
3588 | if (req1->sector > req2->sector) { | |
3589 | return 1; | |
3590 | } else if (req1->sector < req2->sector) { | |
3591 | return -1; | |
3592 | } else { | |
3593 | return 0; | |
3594 | } | |
40b4f539 KW |
3595 | } |
3596 | ||
3597 | /* | |
3598 | * Takes a bunch of requests and tries to merge them. Returns the number of | |
3599 | * requests that remain after merging. | |
3600 | */ | |
3601 | static int multiwrite_merge(BlockDriverState *bs, BlockRequest *reqs, | |
3602 | int num_reqs, MultiwriteCB *mcb) | |
3603 | { | |
3604 | int i, outidx; | |
3605 | ||
3606 | // Sort requests by start sector | |
3607 | qsort(reqs, num_reqs, sizeof(*reqs), &multiwrite_req_compare); | |
3608 | ||
3609 | // Check if adjacent requests touch the same clusters. If so, combine them, | |
3610 | // filling up gaps with zero sectors. | |
3611 | outidx = 0; | |
3612 | for (i = 1; i < num_reqs; i++) { | |
3613 | int merge = 0; | |
3614 | int64_t oldreq_last = reqs[outidx].sector + reqs[outidx].nb_sectors; | |
3615 | ||
b6a127a1 | 3616 | // Handle exactly sequential writes and overlapping writes. |
40b4f539 KW |
3617 | if (reqs[i].sector <= oldreq_last) { |
3618 | merge = 1; | |
3619 | } | |
3620 | ||
e2a305fb CH |
3621 | if (reqs[outidx].qiov->niov + reqs[i].qiov->niov + 1 > IOV_MAX) { |
3622 | merge = 0; | |
3623 | } | |
3624 | ||
40b4f539 KW |
3625 | if (merge) { |
3626 | size_t size; | |
7267c094 | 3627 | QEMUIOVector *qiov = g_malloc0(sizeof(*qiov)); |
40b4f539 KW |
3628 | qemu_iovec_init(qiov, |
3629 | reqs[outidx].qiov->niov + reqs[i].qiov->niov + 1); | |
3630 | ||
3631 | // Add the first request to the merged one. If the requests are | |
3632 | // overlapping, drop the last sectors of the first request. | |
3633 | size = (reqs[i].sector - reqs[outidx].sector) << 9; | |
1b093c48 | 3634 | qemu_iovec_concat(qiov, reqs[outidx].qiov, 0, size); |
40b4f539 | 3635 | |
b6a127a1 PB |
3636 | // We should need to add any zeros between the two requests |
3637 | assert (reqs[i].sector <= oldreq_last); | |
40b4f539 KW |
3638 | |
3639 | // Add the second request | |
1b093c48 | 3640 | qemu_iovec_concat(qiov, reqs[i].qiov, 0, reqs[i].qiov->size); |
40b4f539 | 3641 | |
cbf1dff2 | 3642 | reqs[outidx].nb_sectors = qiov->size >> 9; |
40b4f539 KW |
3643 | reqs[outidx].qiov = qiov; |
3644 | ||
3645 | mcb->callbacks[i].free_qiov = reqs[outidx].qiov; | |
3646 | } else { | |
3647 | outidx++; | |
3648 | reqs[outidx].sector = reqs[i].sector; | |
3649 | reqs[outidx].nb_sectors = reqs[i].nb_sectors; | |
3650 | reqs[outidx].qiov = reqs[i].qiov; | |
3651 | } | |
3652 | } | |
3653 | ||
3654 | return outidx + 1; | |
3655 | } | |
3656 | ||
3657 | /* | |
3658 | * Submit multiple AIO write requests at once. | |
3659 | * | |
3660 | * On success, the function returns 0 and all requests in the reqs array have | |
3661 | * been submitted. In error case this function returns -1, and any of the | |
3662 | * requests may or may not be submitted yet. In particular, this means that the | |
3663 | * callback will be called for some of the requests, for others it won't. The | |
3664 | * caller must check the error field of the BlockRequest to wait for the right | |
3665 | * callbacks (if error != 0, no callback will be called). | |
3666 | * | |
3667 | * The implementation may modify the contents of the reqs array, e.g. to merge | |
3668 | * requests. However, the fields opaque and error are left unmodified as they | |
3669 | * are used to signal failure for a single request to the caller. | |
3670 | */ | |
3671 | int bdrv_aio_multiwrite(BlockDriverState *bs, BlockRequest *reqs, int num_reqs) | |
3672 | { | |
40b4f539 KW |
3673 | MultiwriteCB *mcb; |
3674 | int i; | |
3675 | ||
301db7c2 RH |
3676 | /* don't submit writes if we don't have a medium */ |
3677 | if (bs->drv == NULL) { | |
3678 | for (i = 0; i < num_reqs; i++) { | |
3679 | reqs[i].error = -ENOMEDIUM; | |
3680 | } | |
3681 | return -1; | |
3682 | } | |
3683 | ||
40b4f539 KW |
3684 | if (num_reqs == 0) { |
3685 | return 0; | |
3686 | } | |
3687 | ||
3688 | // Create MultiwriteCB structure | |
7267c094 | 3689 | mcb = g_malloc0(sizeof(*mcb) + num_reqs * sizeof(*mcb->callbacks)); |
40b4f539 KW |
3690 | mcb->num_requests = 0; |
3691 | mcb->num_callbacks = num_reqs; | |
3692 | ||
3693 | for (i = 0; i < num_reqs; i++) { | |
3694 | mcb->callbacks[i].cb = reqs[i].cb; | |
3695 | mcb->callbacks[i].opaque = reqs[i].opaque; | |
3696 | } | |
3697 | ||
3698 | // Check for mergable requests | |
3699 | num_reqs = multiwrite_merge(bs, reqs, num_reqs, mcb); | |
3700 | ||
6d519a5f SH |
3701 | trace_bdrv_aio_multiwrite(mcb, mcb->num_callbacks, num_reqs); |
3702 | ||
df9309fb PB |
3703 | /* Run the aio requests. */ |
3704 | mcb->num_requests = num_reqs; | |
40b4f539 | 3705 | for (i = 0; i < num_reqs; i++) { |
ad54ae80 | 3706 | bdrv_aio_writev(bs, reqs[i].sector, reqs[i].qiov, |
40b4f539 | 3707 | reqs[i].nb_sectors, multiwrite_cb, mcb); |
40b4f539 KW |
3708 | } |
3709 | ||
3710 | return 0; | |
40b4f539 KW |
3711 | } |
3712 | ||
83f64091 | 3713 | void bdrv_aio_cancel(BlockDriverAIOCB *acb) |
83f64091 | 3714 | { |
d7331bed | 3715 | acb->aiocb_info->cancel(acb); |
83f64091 FB |
3716 | } |
3717 | ||
3718 | /**************************************************************/ | |
3719 | /* async block device emulation */ | |
3720 | ||
c16b5a2c CH |
3721 | typedef struct BlockDriverAIOCBSync { |
3722 | BlockDriverAIOCB common; | |
3723 | QEMUBH *bh; | |
3724 | int ret; | |
3725 | /* vector translation state */ | |
3726 | QEMUIOVector *qiov; | |
3727 | uint8_t *bounce; | |
3728 | int is_write; | |
3729 | } BlockDriverAIOCBSync; | |
3730 | ||
3731 | static void bdrv_aio_cancel_em(BlockDriverAIOCB *blockacb) | |
3732 | { | |
b666d239 KW |
3733 | BlockDriverAIOCBSync *acb = |
3734 | container_of(blockacb, BlockDriverAIOCBSync, common); | |
6a7ad299 | 3735 | qemu_bh_delete(acb->bh); |
36afc451 | 3736 | acb->bh = NULL; |
c16b5a2c CH |
3737 | qemu_aio_release(acb); |
3738 | } | |
3739 | ||
d7331bed | 3740 | static const AIOCBInfo bdrv_em_aiocb_info = { |
c16b5a2c CH |
3741 | .aiocb_size = sizeof(BlockDriverAIOCBSync), |
3742 | .cancel = bdrv_aio_cancel_em, | |
3743 | }; | |
3744 | ||
ce1a14dc | 3745 | static void bdrv_aio_bh_cb(void *opaque) |
83f64091 | 3746 | { |
ce1a14dc | 3747 | BlockDriverAIOCBSync *acb = opaque; |
f141eafe | 3748 | |
f141eafe | 3749 | if (!acb->is_write) |
03396148 | 3750 | qemu_iovec_from_buf(acb->qiov, 0, acb->bounce, acb->qiov->size); |
ceb42de8 | 3751 | qemu_vfree(acb->bounce); |
ce1a14dc | 3752 | acb->common.cb(acb->common.opaque, acb->ret); |
6a7ad299 | 3753 | qemu_bh_delete(acb->bh); |
36afc451 | 3754 | acb->bh = NULL; |
ce1a14dc | 3755 | qemu_aio_release(acb); |
83f64091 | 3756 | } |
beac80cd | 3757 | |
f141eafe AL |
3758 | static BlockDriverAIOCB *bdrv_aio_rw_vector(BlockDriverState *bs, |
3759 | int64_t sector_num, | |
3760 | QEMUIOVector *qiov, | |
3761 | int nb_sectors, | |
3762 | BlockDriverCompletionFunc *cb, | |
3763 | void *opaque, | |
3764 | int is_write) | |
3765 | ||
83f64091 | 3766 | { |
ce1a14dc | 3767 | BlockDriverAIOCBSync *acb; |
ce1a14dc | 3768 | |
d7331bed | 3769 | acb = qemu_aio_get(&bdrv_em_aiocb_info, bs, cb, opaque); |
f141eafe AL |
3770 | acb->is_write = is_write; |
3771 | acb->qiov = qiov; | |
e268ca52 | 3772 | acb->bounce = qemu_blockalign(bs, qiov->size); |
3f3aace8 | 3773 | acb->bh = qemu_bh_new(bdrv_aio_bh_cb, acb); |
f141eafe AL |
3774 | |
3775 | if (is_write) { | |
d5e6b161 | 3776 | qemu_iovec_to_buf(acb->qiov, 0, acb->bounce, qiov->size); |
1ed20acf | 3777 | acb->ret = bs->drv->bdrv_write(bs, sector_num, acb->bounce, nb_sectors); |
f141eafe | 3778 | } else { |
1ed20acf | 3779 | acb->ret = bs->drv->bdrv_read(bs, sector_num, acb->bounce, nb_sectors); |
f141eafe AL |
3780 | } |
3781 | ||
ce1a14dc | 3782 | qemu_bh_schedule(acb->bh); |
f141eafe | 3783 | |
ce1a14dc | 3784 | return &acb->common; |
beac80cd FB |
3785 | } |
3786 | ||
f141eafe AL |
3787 | static BlockDriverAIOCB *bdrv_aio_readv_em(BlockDriverState *bs, |
3788 | int64_t sector_num, QEMUIOVector *qiov, int nb_sectors, | |
ce1a14dc | 3789 | BlockDriverCompletionFunc *cb, void *opaque) |
beac80cd | 3790 | { |
f141eafe AL |
3791 | return bdrv_aio_rw_vector(bs, sector_num, qiov, nb_sectors, cb, opaque, 0); |
3792 | } | |
83f64091 | 3793 | |
f141eafe AL |
3794 | static BlockDriverAIOCB *bdrv_aio_writev_em(BlockDriverState *bs, |
3795 | int64_t sector_num, QEMUIOVector *qiov, int nb_sectors, | |
3796 | BlockDriverCompletionFunc *cb, void *opaque) | |
3797 | { | |
3798 | return bdrv_aio_rw_vector(bs, sector_num, qiov, nb_sectors, cb, opaque, 1); | |
beac80cd | 3799 | } |
beac80cd | 3800 | |
68485420 KW |
3801 | |
3802 | typedef struct BlockDriverAIOCBCoroutine { | |
3803 | BlockDriverAIOCB common; | |
3804 | BlockRequest req; | |
3805 | bool is_write; | |
d318aea9 | 3806 | bool *done; |
68485420 KW |
3807 | QEMUBH* bh; |
3808 | } BlockDriverAIOCBCoroutine; | |
3809 | ||
3810 | static void bdrv_aio_co_cancel_em(BlockDriverAIOCB *blockacb) | |
3811 | { | |
d318aea9 KW |
3812 | BlockDriverAIOCBCoroutine *acb = |
3813 | container_of(blockacb, BlockDriverAIOCBCoroutine, common); | |
3814 | bool done = false; | |
3815 | ||
3816 | acb->done = &done; | |
3817 | while (!done) { | |
3818 | qemu_aio_wait(); | |
3819 | } | |
68485420 KW |
3820 | } |
3821 | ||
d7331bed | 3822 | static const AIOCBInfo bdrv_em_co_aiocb_info = { |
68485420 KW |
3823 | .aiocb_size = sizeof(BlockDriverAIOCBCoroutine), |
3824 | .cancel = bdrv_aio_co_cancel_em, | |
3825 | }; | |
3826 | ||
35246a68 | 3827 | static void bdrv_co_em_bh(void *opaque) |
68485420 KW |
3828 | { |
3829 | BlockDriverAIOCBCoroutine *acb = opaque; | |
3830 | ||
3831 | acb->common.cb(acb->common.opaque, acb->req.error); | |
d318aea9 KW |
3832 | |
3833 | if (acb->done) { | |
3834 | *acb->done = true; | |
3835 | } | |
3836 | ||
68485420 KW |
3837 | qemu_bh_delete(acb->bh); |
3838 | qemu_aio_release(acb); | |
3839 | } | |
3840 | ||
b2a61371 SH |
3841 | /* Invoke bdrv_co_do_readv/bdrv_co_do_writev */ |
3842 | static void coroutine_fn bdrv_co_do_rw(void *opaque) | |
3843 | { | |
3844 | BlockDriverAIOCBCoroutine *acb = opaque; | |
3845 | BlockDriverState *bs = acb->common.bs; | |
3846 | ||
3847 | if (!acb->is_write) { | |
3848 | acb->req.error = bdrv_co_do_readv(bs, acb->req.sector, | |
470c0504 | 3849 | acb->req.nb_sectors, acb->req.qiov, 0); |
b2a61371 SH |
3850 | } else { |
3851 | acb->req.error = bdrv_co_do_writev(bs, acb->req.sector, | |
f08f2dda | 3852 | acb->req.nb_sectors, acb->req.qiov, 0); |
b2a61371 SH |
3853 | } |
3854 | ||
35246a68 | 3855 | acb->bh = qemu_bh_new(bdrv_co_em_bh, acb); |
b2a61371 SH |
3856 | qemu_bh_schedule(acb->bh); |
3857 | } | |
3858 | ||
68485420 KW |
3859 | static BlockDriverAIOCB *bdrv_co_aio_rw_vector(BlockDriverState *bs, |
3860 | int64_t sector_num, | |
3861 | QEMUIOVector *qiov, | |
3862 | int nb_sectors, | |
3863 | BlockDriverCompletionFunc *cb, | |
3864 | void *opaque, | |
8c5873d6 | 3865 | bool is_write) |
68485420 KW |
3866 | { |
3867 | Coroutine *co; | |
3868 | BlockDriverAIOCBCoroutine *acb; | |
3869 | ||
d7331bed | 3870 | acb = qemu_aio_get(&bdrv_em_co_aiocb_info, bs, cb, opaque); |
68485420 KW |
3871 | acb->req.sector = sector_num; |
3872 | acb->req.nb_sectors = nb_sectors; | |
3873 | acb->req.qiov = qiov; | |
3874 | acb->is_write = is_write; | |
d318aea9 | 3875 | acb->done = NULL; |
68485420 | 3876 | |
8c5873d6 | 3877 | co = qemu_coroutine_create(bdrv_co_do_rw); |
68485420 KW |
3878 | qemu_coroutine_enter(co, acb); |
3879 | ||
3880 | return &acb->common; | |
3881 | } | |
3882 | ||
07f07615 | 3883 | static void coroutine_fn bdrv_aio_flush_co_entry(void *opaque) |
b2e12bc6 | 3884 | { |
07f07615 PB |
3885 | BlockDriverAIOCBCoroutine *acb = opaque; |
3886 | BlockDriverState *bs = acb->common.bs; | |
b2e12bc6 | 3887 | |
07f07615 PB |
3888 | acb->req.error = bdrv_co_flush(bs); |
3889 | acb->bh = qemu_bh_new(bdrv_co_em_bh, acb); | |
b2e12bc6 | 3890 | qemu_bh_schedule(acb->bh); |
b2e12bc6 CH |
3891 | } |
3892 | ||
07f07615 | 3893 | BlockDriverAIOCB *bdrv_aio_flush(BlockDriverState *bs, |
016f5cf6 AG |
3894 | BlockDriverCompletionFunc *cb, void *opaque) |
3895 | { | |
07f07615 | 3896 | trace_bdrv_aio_flush(bs, opaque); |
016f5cf6 | 3897 | |
07f07615 PB |
3898 | Coroutine *co; |
3899 | BlockDriverAIOCBCoroutine *acb; | |
016f5cf6 | 3900 | |
d7331bed | 3901 | acb = qemu_aio_get(&bdrv_em_co_aiocb_info, bs, cb, opaque); |
d318aea9 KW |
3902 | acb->done = NULL; |
3903 | ||
07f07615 PB |
3904 | co = qemu_coroutine_create(bdrv_aio_flush_co_entry); |
3905 | qemu_coroutine_enter(co, acb); | |
016f5cf6 | 3906 | |
016f5cf6 AG |
3907 | return &acb->common; |
3908 | } | |
3909 | ||
4265d620 PB |
3910 | static void coroutine_fn bdrv_aio_discard_co_entry(void *opaque) |
3911 | { | |
3912 | BlockDriverAIOCBCoroutine *acb = opaque; | |
3913 | BlockDriverState *bs = acb->common.bs; | |
3914 | ||
3915 | acb->req.error = bdrv_co_discard(bs, acb->req.sector, acb->req.nb_sectors); | |
3916 | acb->bh = qemu_bh_new(bdrv_co_em_bh, acb); | |
3917 | qemu_bh_schedule(acb->bh); | |
3918 | } | |
3919 | ||
3920 | BlockDriverAIOCB *bdrv_aio_discard(BlockDriverState *bs, | |
3921 | int64_t sector_num, int nb_sectors, | |
3922 | BlockDriverCompletionFunc *cb, void *opaque) | |
3923 | { | |
3924 | Coroutine *co; | |
3925 | BlockDriverAIOCBCoroutine *acb; | |
3926 | ||
3927 | trace_bdrv_aio_discard(bs, sector_num, nb_sectors, opaque); | |
3928 | ||
d7331bed | 3929 | acb = qemu_aio_get(&bdrv_em_co_aiocb_info, bs, cb, opaque); |
4265d620 PB |
3930 | acb->req.sector = sector_num; |
3931 | acb->req.nb_sectors = nb_sectors; | |
d318aea9 | 3932 | acb->done = NULL; |
4265d620 PB |
3933 | co = qemu_coroutine_create(bdrv_aio_discard_co_entry); |
3934 | qemu_coroutine_enter(co, acb); | |
3935 | ||
3936 | return &acb->common; | |
3937 | } | |
3938 | ||
ea2384d3 FB |
3939 | void bdrv_init(void) |
3940 | { | |
5efa9d5a | 3941 | module_call_init(MODULE_INIT_BLOCK); |
ea2384d3 | 3942 | } |
ce1a14dc | 3943 | |
eb852011 MA |
3944 | void bdrv_init_with_whitelist(void) |
3945 | { | |
3946 | use_bdrv_whitelist = 1; | |
3947 | bdrv_init(); | |
3948 | } | |
3949 | ||
d7331bed | 3950 | void *qemu_aio_get(const AIOCBInfo *aiocb_info, BlockDriverState *bs, |
c16b5a2c | 3951 | BlockDriverCompletionFunc *cb, void *opaque) |
ce1a14dc | 3952 | { |
ce1a14dc PB |
3953 | BlockDriverAIOCB *acb; |
3954 | ||
d7331bed SH |
3955 | acb = g_slice_alloc(aiocb_info->aiocb_size); |
3956 | acb->aiocb_info = aiocb_info; | |
ce1a14dc PB |
3957 | acb->bs = bs; |
3958 | acb->cb = cb; | |
3959 | acb->opaque = opaque; | |
3960 | return acb; | |
3961 | } | |
3962 | ||
3963 | void qemu_aio_release(void *p) | |
3964 | { | |
d37c975f | 3965 | BlockDriverAIOCB *acb = p; |
d7331bed | 3966 | g_slice_free1(acb->aiocb_info->aiocb_size, acb); |
ce1a14dc | 3967 | } |
19cb3738 | 3968 | |
f9f05dc5 KW |
3969 | /**************************************************************/ |
3970 | /* Coroutine block device emulation */ | |
3971 | ||
3972 | typedef struct CoroutineIOCompletion { | |
3973 | Coroutine *coroutine; | |
3974 | int ret; | |
3975 | } CoroutineIOCompletion; | |
3976 | ||
3977 | static void bdrv_co_io_em_complete(void *opaque, int ret) | |
3978 | { | |
3979 | CoroutineIOCompletion *co = opaque; | |
3980 | ||
3981 | co->ret = ret; | |
3982 | qemu_coroutine_enter(co->coroutine, NULL); | |
3983 | } | |
3984 | ||
3985 | static int coroutine_fn bdrv_co_io_em(BlockDriverState *bs, int64_t sector_num, | |
3986 | int nb_sectors, QEMUIOVector *iov, | |
3987 | bool is_write) | |
3988 | { | |
3989 | CoroutineIOCompletion co = { | |
3990 | .coroutine = qemu_coroutine_self(), | |
3991 | }; | |
3992 | BlockDriverAIOCB *acb; | |
3993 | ||
3994 | if (is_write) { | |
a652d160 SH |
3995 | acb = bs->drv->bdrv_aio_writev(bs, sector_num, iov, nb_sectors, |
3996 | bdrv_co_io_em_complete, &co); | |
f9f05dc5 | 3997 | } else { |
a652d160 SH |
3998 | acb = bs->drv->bdrv_aio_readv(bs, sector_num, iov, nb_sectors, |
3999 | bdrv_co_io_em_complete, &co); | |
f9f05dc5 KW |
4000 | } |
4001 | ||
59370aaa | 4002 | trace_bdrv_co_io_em(bs, sector_num, nb_sectors, is_write, acb); |
f9f05dc5 KW |
4003 | if (!acb) { |
4004 | return -EIO; | |
4005 | } | |
4006 | qemu_coroutine_yield(); | |
4007 | ||
4008 | return co.ret; | |
4009 | } | |
4010 | ||
4011 | static int coroutine_fn bdrv_co_readv_em(BlockDriverState *bs, | |
4012 | int64_t sector_num, int nb_sectors, | |
4013 | QEMUIOVector *iov) | |
4014 | { | |
4015 | return bdrv_co_io_em(bs, sector_num, nb_sectors, iov, false); | |
4016 | } | |
4017 | ||
4018 | static int coroutine_fn bdrv_co_writev_em(BlockDriverState *bs, | |
4019 | int64_t sector_num, int nb_sectors, | |
4020 | QEMUIOVector *iov) | |
4021 | { | |
4022 | return bdrv_co_io_em(bs, sector_num, nb_sectors, iov, true); | |
4023 | } | |
4024 | ||
07f07615 | 4025 | static void coroutine_fn bdrv_flush_co_entry(void *opaque) |
e7a8a783 | 4026 | { |
07f07615 PB |
4027 | RwCo *rwco = opaque; |
4028 | ||
4029 | rwco->ret = bdrv_co_flush(rwco->bs); | |
4030 | } | |
4031 | ||
4032 | int coroutine_fn bdrv_co_flush(BlockDriverState *bs) | |
4033 | { | |
eb489bb1 KW |
4034 | int ret; |
4035 | ||
29cdb251 | 4036 | if (!bs || !bdrv_is_inserted(bs) || bdrv_is_read_only(bs)) { |
07f07615 | 4037 | return 0; |
eb489bb1 KW |
4038 | } |
4039 | ||
ca716364 | 4040 | /* Write back cached data to the OS even with cache=unsafe */ |
bf736fe3 | 4041 | BLKDBG_EVENT(bs->file, BLKDBG_FLUSH_TO_OS); |
eb489bb1 KW |
4042 | if (bs->drv->bdrv_co_flush_to_os) { |
4043 | ret = bs->drv->bdrv_co_flush_to_os(bs); | |
4044 | if (ret < 0) { | |
4045 | return ret; | |
4046 | } | |
4047 | } | |
4048 | ||
ca716364 KW |
4049 | /* But don't actually force it to the disk with cache=unsafe */ |
4050 | if (bs->open_flags & BDRV_O_NO_FLUSH) { | |
d4c82329 | 4051 | goto flush_parent; |
ca716364 KW |
4052 | } |
4053 | ||
bf736fe3 | 4054 | BLKDBG_EVENT(bs->file, BLKDBG_FLUSH_TO_DISK); |
eb489bb1 | 4055 | if (bs->drv->bdrv_co_flush_to_disk) { |
29cdb251 | 4056 | ret = bs->drv->bdrv_co_flush_to_disk(bs); |
07f07615 PB |
4057 | } else if (bs->drv->bdrv_aio_flush) { |
4058 | BlockDriverAIOCB *acb; | |
4059 | CoroutineIOCompletion co = { | |
4060 | .coroutine = qemu_coroutine_self(), | |
4061 | }; | |
4062 | ||
4063 | acb = bs->drv->bdrv_aio_flush(bs, bdrv_co_io_em_complete, &co); | |
4064 | if (acb == NULL) { | |
29cdb251 | 4065 | ret = -EIO; |
07f07615 PB |
4066 | } else { |
4067 | qemu_coroutine_yield(); | |
29cdb251 | 4068 | ret = co.ret; |
07f07615 | 4069 | } |
07f07615 PB |
4070 | } else { |
4071 | /* | |
4072 | * Some block drivers always operate in either writethrough or unsafe | |
4073 | * mode and don't support bdrv_flush therefore. Usually qemu doesn't | |
4074 | * know how the server works (because the behaviour is hardcoded or | |
4075 | * depends on server-side configuration), so we can't ensure that | |
4076 | * everything is safe on disk. Returning an error doesn't work because | |
4077 | * that would break guests even if the server operates in writethrough | |
4078 | * mode. | |
4079 | * | |
4080 | * Let's hope the user knows what he's doing. | |
4081 | */ | |
29cdb251 | 4082 | ret = 0; |
07f07615 | 4083 | } |
29cdb251 PB |
4084 | if (ret < 0) { |
4085 | return ret; | |
4086 | } | |
4087 | ||
4088 | /* Now flush the underlying protocol. It will also have BDRV_O_NO_FLUSH | |
4089 | * in the case of cache=unsafe, so there are no useless flushes. | |
4090 | */ | |
d4c82329 | 4091 | flush_parent: |
29cdb251 | 4092 | return bdrv_co_flush(bs->file); |
07f07615 PB |
4093 | } |
4094 | ||
0f15423c AL |
4095 | void bdrv_invalidate_cache(BlockDriverState *bs) |
4096 | { | |
4097 | if (bs->drv && bs->drv->bdrv_invalidate_cache) { | |
4098 | bs->drv->bdrv_invalidate_cache(bs); | |
4099 | } | |
4100 | } | |
4101 | ||
4102 | void bdrv_invalidate_cache_all(void) | |
4103 | { | |
4104 | BlockDriverState *bs; | |
4105 | ||
4106 | QTAILQ_FOREACH(bs, &bdrv_states, list) { | |
4107 | bdrv_invalidate_cache(bs); | |
4108 | } | |
4109 | } | |
4110 | ||
07789269 BC |
4111 | void bdrv_clear_incoming_migration_all(void) |
4112 | { | |
4113 | BlockDriverState *bs; | |
4114 | ||
4115 | QTAILQ_FOREACH(bs, &bdrv_states, list) { | |
4116 | bs->open_flags = bs->open_flags & ~(BDRV_O_INCOMING); | |
4117 | } | |
4118 | } | |
4119 | ||
07f07615 PB |
4120 | int bdrv_flush(BlockDriverState *bs) |
4121 | { | |
4122 | Coroutine *co; | |
4123 | RwCo rwco = { | |
4124 | .bs = bs, | |
4125 | .ret = NOT_DONE, | |
e7a8a783 | 4126 | }; |
e7a8a783 | 4127 | |
07f07615 PB |
4128 | if (qemu_in_coroutine()) { |
4129 | /* Fast-path if already in coroutine context */ | |
4130 | bdrv_flush_co_entry(&rwco); | |
4131 | } else { | |
4132 | co = qemu_coroutine_create(bdrv_flush_co_entry); | |
4133 | qemu_coroutine_enter(co, &rwco); | |
4134 | while (rwco.ret == NOT_DONE) { | |
4135 | qemu_aio_wait(); | |
4136 | } | |
e7a8a783 | 4137 | } |
07f07615 PB |
4138 | |
4139 | return rwco.ret; | |
e7a8a783 KW |
4140 | } |
4141 | ||
4265d620 PB |
4142 | static void coroutine_fn bdrv_discard_co_entry(void *opaque) |
4143 | { | |
4144 | RwCo *rwco = opaque; | |
4145 | ||
4146 | rwco->ret = bdrv_co_discard(rwco->bs, rwco->sector_num, rwco->nb_sectors); | |
4147 | } | |
4148 | ||
4149 | int coroutine_fn bdrv_co_discard(BlockDriverState *bs, int64_t sector_num, | |
4150 | int nb_sectors) | |
4151 | { | |
4152 | if (!bs->drv) { | |
4153 | return -ENOMEDIUM; | |
4154 | } else if (bdrv_check_request(bs, sector_num, nb_sectors)) { | |
4155 | return -EIO; | |
4156 | } else if (bs->read_only) { | |
4157 | return -EROFS; | |
df702c9b PB |
4158 | } |
4159 | ||
4160 | if (bs->dirty_bitmap) { | |
8f0720ec | 4161 | bdrv_reset_dirty(bs, sector_num, nb_sectors); |
df702c9b PB |
4162 | } |
4163 | ||
9e8f1835 PB |
4164 | /* Do nothing if disabled. */ |
4165 | if (!(bs->open_flags & BDRV_O_UNMAP)) { | |
4166 | return 0; | |
4167 | } | |
4168 | ||
df702c9b | 4169 | if (bs->drv->bdrv_co_discard) { |
4265d620 PB |
4170 | return bs->drv->bdrv_co_discard(bs, sector_num, nb_sectors); |
4171 | } else if (bs->drv->bdrv_aio_discard) { | |
4172 | BlockDriverAIOCB *acb; | |
4173 | CoroutineIOCompletion co = { | |
4174 | .coroutine = qemu_coroutine_self(), | |
4175 | }; | |
4176 | ||
4177 | acb = bs->drv->bdrv_aio_discard(bs, sector_num, nb_sectors, | |
4178 | bdrv_co_io_em_complete, &co); | |
4179 | if (acb == NULL) { | |
4180 | return -EIO; | |
4181 | } else { | |
4182 | qemu_coroutine_yield(); | |
4183 | return co.ret; | |
4184 | } | |
4265d620 PB |
4185 | } else { |
4186 | return 0; | |
4187 | } | |
4188 | } | |
4189 | ||
4190 | int bdrv_discard(BlockDriverState *bs, int64_t sector_num, int nb_sectors) | |
4191 | { | |
4192 | Coroutine *co; | |
4193 | RwCo rwco = { | |
4194 | .bs = bs, | |
4195 | .sector_num = sector_num, | |
4196 | .nb_sectors = nb_sectors, | |
4197 | .ret = NOT_DONE, | |
4198 | }; | |
4199 | ||
4200 | if (qemu_in_coroutine()) { | |
4201 | /* Fast-path if already in coroutine context */ | |
4202 | bdrv_discard_co_entry(&rwco); | |
4203 | } else { | |
4204 | co = qemu_coroutine_create(bdrv_discard_co_entry); | |
4205 | qemu_coroutine_enter(co, &rwco); | |
4206 | while (rwco.ret == NOT_DONE) { | |
4207 | qemu_aio_wait(); | |
4208 | } | |
4209 | } | |
4210 | ||
4211 | return rwco.ret; | |
4212 | } | |
4213 | ||
19cb3738 FB |
4214 | /**************************************************************/ |
4215 | /* removable device support */ | |
4216 | ||
4217 | /** | |
4218 | * Return TRUE if the media is present | |
4219 | */ | |
4220 | int bdrv_is_inserted(BlockDriverState *bs) | |
4221 | { | |
4222 | BlockDriver *drv = bs->drv; | |
a1aff5bf | 4223 | |
19cb3738 FB |
4224 | if (!drv) |
4225 | return 0; | |
4226 | if (!drv->bdrv_is_inserted) | |
a1aff5bf MA |
4227 | return 1; |
4228 | return drv->bdrv_is_inserted(bs); | |
19cb3738 FB |
4229 | } |
4230 | ||
4231 | /** | |
8e49ca46 MA |
4232 | * Return whether the media changed since the last call to this |
4233 | * function, or -ENOTSUP if we don't know. Most drivers don't know. | |
19cb3738 FB |
4234 | */ |
4235 | int bdrv_media_changed(BlockDriverState *bs) | |
4236 | { | |
4237 | BlockDriver *drv = bs->drv; | |
19cb3738 | 4238 | |
8e49ca46 MA |
4239 | if (drv && drv->bdrv_media_changed) { |
4240 | return drv->bdrv_media_changed(bs); | |
4241 | } | |
4242 | return -ENOTSUP; | |
19cb3738 FB |
4243 | } |
4244 | ||
4245 | /** | |
4246 | * If eject_flag is TRUE, eject the media. Otherwise, close the tray | |
4247 | */ | |
f36f3949 | 4248 | void bdrv_eject(BlockDriverState *bs, bool eject_flag) |
19cb3738 FB |
4249 | { |
4250 | BlockDriver *drv = bs->drv; | |
19cb3738 | 4251 | |
822e1cd1 MA |
4252 | if (drv && drv->bdrv_eject) { |
4253 | drv->bdrv_eject(bs, eject_flag); | |
19cb3738 | 4254 | } |
6f382ed2 LC |
4255 | |
4256 | if (bs->device_name[0] != '\0') { | |
4257 | bdrv_emit_qmp_eject_event(bs, eject_flag); | |
4258 | } | |
19cb3738 FB |
4259 | } |
4260 | ||
19cb3738 FB |
4261 | /** |
4262 | * Lock or unlock the media (if it is locked, the user won't be able | |
4263 | * to eject it manually). | |
4264 | */ | |
025e849a | 4265 | void bdrv_lock_medium(BlockDriverState *bs, bool locked) |
19cb3738 FB |
4266 | { |
4267 | BlockDriver *drv = bs->drv; | |
4268 | ||
025e849a | 4269 | trace_bdrv_lock_medium(bs, locked); |
b8c6d095 | 4270 | |
025e849a MA |
4271 | if (drv && drv->bdrv_lock_medium) { |
4272 | drv->bdrv_lock_medium(bs, locked); | |
19cb3738 FB |
4273 | } |
4274 | } | |
985a03b0 TS |
4275 | |
4276 | /* needed for generic scsi interface */ | |
4277 | ||
4278 | int bdrv_ioctl(BlockDriverState *bs, unsigned long int req, void *buf) | |
4279 | { | |
4280 | BlockDriver *drv = bs->drv; | |
4281 | ||
4282 | if (drv && drv->bdrv_ioctl) | |
4283 | return drv->bdrv_ioctl(bs, req, buf); | |
4284 | return -ENOTSUP; | |
4285 | } | |
7d780669 | 4286 | |
221f715d AL |
4287 | BlockDriverAIOCB *bdrv_aio_ioctl(BlockDriverState *bs, |
4288 | unsigned long int req, void *buf, | |
4289 | BlockDriverCompletionFunc *cb, void *opaque) | |
7d780669 | 4290 | { |
221f715d | 4291 | BlockDriver *drv = bs->drv; |
7d780669 | 4292 | |
221f715d AL |
4293 | if (drv && drv->bdrv_aio_ioctl) |
4294 | return drv->bdrv_aio_ioctl(bs, req, buf, cb, opaque); | |
4295 | return NULL; | |
7d780669 | 4296 | } |
e268ca52 | 4297 | |
7b6f9300 MA |
4298 | void bdrv_set_buffer_alignment(BlockDriverState *bs, int align) |
4299 | { | |
4300 | bs->buffer_alignment = align; | |
4301 | } | |
7cd1e32a | 4302 | |
e268ca52 AL |
4303 | void *qemu_blockalign(BlockDriverState *bs, size_t size) |
4304 | { | |
4305 | return qemu_memalign((bs && bs->buffer_alignment) ? bs->buffer_alignment : 512, size); | |
4306 | } | |
7cd1e32a | 4307 | |
c53b1c51 SH |
4308 | /* |
4309 | * Check if all memory in this vector is sector aligned. | |
4310 | */ | |
4311 | bool bdrv_qiov_is_aligned(BlockDriverState *bs, QEMUIOVector *qiov) | |
4312 | { | |
4313 | int i; | |
4314 | ||
4315 | for (i = 0; i < qiov->niov; i++) { | |
4316 | if ((uintptr_t) qiov->iov[i].iov_base % bs->buffer_alignment) { | |
4317 | return false; | |
4318 | } | |
4319 | } | |
4320 | ||
4321 | return true; | |
4322 | } | |
4323 | ||
50717e94 | 4324 | void bdrv_set_dirty_tracking(BlockDriverState *bs, int granularity) |
7cd1e32a | 4325 | { |
4326 | int64_t bitmap_size; | |
a55eb92c | 4327 | |
50717e94 PB |
4328 | assert((granularity & (granularity - 1)) == 0); |
4329 | ||
4330 | if (granularity) { | |
4331 | granularity >>= BDRV_SECTOR_BITS; | |
4332 | assert(!bs->dirty_bitmap); | |
4333 | bitmap_size = (bdrv_getlength(bs) >> BDRV_SECTOR_BITS); | |
4334 | bs->dirty_bitmap = hbitmap_alloc(bitmap_size, ffs(granularity) - 1); | |
7cd1e32a | 4335 | } else { |
c6d22830 | 4336 | if (bs->dirty_bitmap) { |
8f0720ec | 4337 | hbitmap_free(bs->dirty_bitmap); |
c6d22830 | 4338 | bs->dirty_bitmap = NULL; |
a55eb92c | 4339 | } |
7cd1e32a | 4340 | } |
4341 | } | |
4342 | ||
4343 | int bdrv_get_dirty(BlockDriverState *bs, int64_t sector) | |
4344 | { | |
8f0720ec PB |
4345 | if (bs->dirty_bitmap) { |
4346 | return hbitmap_get(bs->dirty_bitmap, sector); | |
7cd1e32a | 4347 | } else { |
4348 | return 0; | |
4349 | } | |
4350 | } | |
4351 | ||
8f0720ec | 4352 | void bdrv_dirty_iter_init(BlockDriverState *bs, HBitmapIter *hbi) |
1755da16 | 4353 | { |
8f0720ec | 4354 | hbitmap_iter_init(hbi, bs->dirty_bitmap, 0); |
1755da16 PB |
4355 | } |
4356 | ||
4357 | void bdrv_set_dirty(BlockDriverState *bs, int64_t cur_sector, | |
4358 | int nr_sectors) | |
4359 | { | |
8f0720ec | 4360 | hbitmap_set(bs->dirty_bitmap, cur_sector, nr_sectors); |
1755da16 PB |
4361 | } |
4362 | ||
a55eb92c JK |
4363 | void bdrv_reset_dirty(BlockDriverState *bs, int64_t cur_sector, |
4364 | int nr_sectors) | |
7cd1e32a | 4365 | { |
8f0720ec | 4366 | hbitmap_reset(bs->dirty_bitmap, cur_sector, nr_sectors); |
7cd1e32a | 4367 | } |
aaa0eb75 LS |
4368 | |
4369 | int64_t bdrv_get_dirty_count(BlockDriverState *bs) | |
4370 | { | |
8f0720ec | 4371 | if (bs->dirty_bitmap) { |
acc906c6 | 4372 | return hbitmap_count(bs->dirty_bitmap); |
8f0720ec PB |
4373 | } else { |
4374 | return 0; | |
4375 | } | |
aaa0eb75 | 4376 | } |
f88e1a42 | 4377 | |
9fcb0251 FZ |
4378 | /* Get a reference to bs */ |
4379 | void bdrv_ref(BlockDriverState *bs) | |
4380 | { | |
4381 | bs->refcnt++; | |
4382 | } | |
4383 | ||
4384 | /* Release a previously grabbed reference to bs. | |
4385 | * If after releasing, reference count is zero, the BlockDriverState is | |
4386 | * deleted. */ | |
4387 | void bdrv_unref(BlockDriverState *bs) | |
4388 | { | |
4389 | assert(bs->refcnt > 0); | |
4390 | if (--bs->refcnt == 0) { | |
4391 | bdrv_delete(bs); | |
4392 | } | |
4393 | } | |
4394 | ||
db593f25 MT |
4395 | void bdrv_set_in_use(BlockDriverState *bs, int in_use) |
4396 | { | |
4397 | assert(bs->in_use != in_use); | |
4398 | bs->in_use = in_use; | |
4399 | } | |
4400 | ||
4401 | int bdrv_in_use(BlockDriverState *bs) | |
4402 | { | |
4403 | return bs->in_use; | |
4404 | } | |
4405 | ||
28a7282a LC |
4406 | void bdrv_iostatus_enable(BlockDriverState *bs) |
4407 | { | |
d6bf279e | 4408 | bs->iostatus_enabled = true; |
58e21ef5 | 4409 | bs->iostatus = BLOCK_DEVICE_IO_STATUS_OK; |
28a7282a LC |
4410 | } |
4411 | ||
4412 | /* The I/O status is only enabled if the drive explicitly | |
4413 | * enables it _and_ the VM is configured to stop on errors */ | |
4414 | bool bdrv_iostatus_is_enabled(const BlockDriverState *bs) | |
4415 | { | |
d6bf279e | 4416 | return (bs->iostatus_enabled && |
92aa5c6d PB |
4417 | (bs->on_write_error == BLOCKDEV_ON_ERROR_ENOSPC || |
4418 | bs->on_write_error == BLOCKDEV_ON_ERROR_STOP || | |
4419 | bs->on_read_error == BLOCKDEV_ON_ERROR_STOP)); | |
28a7282a LC |
4420 | } |
4421 | ||
4422 | void bdrv_iostatus_disable(BlockDriverState *bs) | |
4423 | { | |
d6bf279e | 4424 | bs->iostatus_enabled = false; |
28a7282a LC |
4425 | } |
4426 | ||
4427 | void bdrv_iostatus_reset(BlockDriverState *bs) | |
4428 | { | |
4429 | if (bdrv_iostatus_is_enabled(bs)) { | |
58e21ef5 | 4430 | bs->iostatus = BLOCK_DEVICE_IO_STATUS_OK; |
3bd293c3 PB |
4431 | if (bs->job) { |
4432 | block_job_iostatus_reset(bs->job); | |
4433 | } | |
28a7282a LC |
4434 | } |
4435 | } | |
4436 | ||
28a7282a LC |
4437 | void bdrv_iostatus_set_err(BlockDriverState *bs, int error) |
4438 | { | |
3e1caa5f PB |
4439 | assert(bdrv_iostatus_is_enabled(bs)); |
4440 | if (bs->iostatus == BLOCK_DEVICE_IO_STATUS_OK) { | |
58e21ef5 LC |
4441 | bs->iostatus = error == ENOSPC ? BLOCK_DEVICE_IO_STATUS_NOSPACE : |
4442 | BLOCK_DEVICE_IO_STATUS_FAILED; | |
28a7282a LC |
4443 | } |
4444 | } | |
4445 | ||
a597e79c CH |
4446 | void |
4447 | bdrv_acct_start(BlockDriverState *bs, BlockAcctCookie *cookie, int64_t bytes, | |
4448 | enum BlockAcctType type) | |
4449 | { | |
4450 | assert(type < BDRV_MAX_IOTYPE); | |
4451 | ||
4452 | cookie->bytes = bytes; | |
c488c7f6 | 4453 | cookie->start_time_ns = get_clock(); |
a597e79c CH |
4454 | cookie->type = type; |
4455 | } | |
4456 | ||
4457 | void | |
4458 | bdrv_acct_done(BlockDriverState *bs, BlockAcctCookie *cookie) | |
4459 | { | |
4460 | assert(cookie->type < BDRV_MAX_IOTYPE); | |
4461 | ||
4462 | bs->nr_bytes[cookie->type] += cookie->bytes; | |
4463 | bs->nr_ops[cookie->type]++; | |
c488c7f6 | 4464 | bs->total_time_ns[cookie->type] += get_clock() - cookie->start_time_ns; |
a597e79c CH |
4465 | } |
4466 | ||
d92ada22 LC |
4467 | void bdrv_img_create(const char *filename, const char *fmt, |
4468 | const char *base_filename, const char *base_fmt, | |
f382d43a MR |
4469 | char *options, uint64_t img_size, int flags, |
4470 | Error **errp, bool quiet) | |
f88e1a42 JS |
4471 | { |
4472 | QEMUOptionParameter *param = NULL, *create_options = NULL; | |
d220894e | 4473 | QEMUOptionParameter *backing_fmt, *backing_file, *size; |
f88e1a42 JS |
4474 | BlockDriverState *bs = NULL; |
4475 | BlockDriver *drv, *proto_drv; | |
96df67d1 | 4476 | BlockDriver *backing_drv = NULL; |
cc84d90f | 4477 | Error *local_err = NULL; |
f88e1a42 JS |
4478 | int ret = 0; |
4479 | ||
4480 | /* Find driver and parse its options */ | |
4481 | drv = bdrv_find_format(fmt); | |
4482 | if (!drv) { | |
71c79813 | 4483 | error_setg(errp, "Unknown file format '%s'", fmt); |
d92ada22 | 4484 | return; |
f88e1a42 JS |
4485 | } |
4486 | ||
98289620 | 4487 | proto_drv = bdrv_find_protocol(filename, true); |
f88e1a42 | 4488 | if (!proto_drv) { |
71c79813 | 4489 | error_setg(errp, "Unknown protocol '%s'", filename); |
d92ada22 | 4490 | return; |
f88e1a42 JS |
4491 | } |
4492 | ||
4493 | create_options = append_option_parameters(create_options, | |
4494 | drv->create_options); | |
4495 | create_options = append_option_parameters(create_options, | |
4496 | proto_drv->create_options); | |
4497 | ||
4498 | /* Create parameter list with default values */ | |
4499 | param = parse_option_parameters("", create_options, param); | |
4500 | ||
4501 | set_option_parameter_int(param, BLOCK_OPT_SIZE, img_size); | |
4502 | ||
4503 | /* Parse -o options */ | |
4504 | if (options) { | |
4505 | param = parse_option_parameters(options, create_options, param); | |
4506 | if (param == NULL) { | |
71c79813 | 4507 | error_setg(errp, "Invalid options for file format '%s'.", fmt); |
f88e1a42 JS |
4508 | goto out; |
4509 | } | |
4510 | } | |
4511 | ||
4512 | if (base_filename) { | |
4513 | if (set_option_parameter(param, BLOCK_OPT_BACKING_FILE, | |
4514 | base_filename)) { | |
71c79813 LC |
4515 | error_setg(errp, "Backing file not supported for file format '%s'", |
4516 | fmt); | |
f88e1a42 JS |
4517 | goto out; |
4518 | } | |
4519 | } | |
4520 | ||
4521 | if (base_fmt) { | |
4522 | if (set_option_parameter(param, BLOCK_OPT_BACKING_FMT, base_fmt)) { | |
71c79813 LC |
4523 | error_setg(errp, "Backing file format not supported for file " |
4524 | "format '%s'", fmt); | |
f88e1a42 JS |
4525 | goto out; |
4526 | } | |
4527 | } | |
4528 | ||
792da93a JS |
4529 | backing_file = get_option_parameter(param, BLOCK_OPT_BACKING_FILE); |
4530 | if (backing_file && backing_file->value.s) { | |
4531 | if (!strcmp(filename, backing_file->value.s)) { | |
71c79813 LC |
4532 | error_setg(errp, "Error: Trying to create an image with the " |
4533 | "same filename as the backing file"); | |
792da93a JS |
4534 | goto out; |
4535 | } | |
4536 | } | |
4537 | ||
f88e1a42 JS |
4538 | backing_fmt = get_option_parameter(param, BLOCK_OPT_BACKING_FMT); |
4539 | if (backing_fmt && backing_fmt->value.s) { | |
96df67d1 SH |
4540 | backing_drv = bdrv_find_format(backing_fmt->value.s); |
4541 | if (!backing_drv) { | |
71c79813 LC |
4542 | error_setg(errp, "Unknown backing file format '%s'", |
4543 | backing_fmt->value.s); | |
f88e1a42 JS |
4544 | goto out; |
4545 | } | |
4546 | } | |
4547 | ||
4548 | // The size for the image must always be specified, with one exception: | |
4549 | // If we are using a backing file, we can obtain the size from there | |
d220894e KW |
4550 | size = get_option_parameter(param, BLOCK_OPT_SIZE); |
4551 | if (size && size->value.n == -1) { | |
f88e1a42 JS |
4552 | if (backing_file && backing_file->value.s) { |
4553 | uint64_t size; | |
f88e1a42 | 4554 | char buf[32]; |
63090dac PB |
4555 | int back_flags; |
4556 | ||
4557 | /* backing files always opened read-only */ | |
4558 | back_flags = | |
4559 | flags & ~(BDRV_O_RDWR | BDRV_O_SNAPSHOT | BDRV_O_NO_BACKING); | |
f88e1a42 | 4560 | |
f88e1a42 JS |
4561 | bs = bdrv_new(""); |
4562 | ||
de9c0cec | 4563 | ret = bdrv_open(bs, backing_file->value.s, NULL, back_flags, |
cc84d90f | 4564 | backing_drv, &local_err); |
f88e1a42 | 4565 | if (ret < 0) { |
cc84d90f MR |
4566 | error_setg_errno(errp, -ret, "Could not open '%s': %s", |
4567 | backing_file->value.s, | |
4568 | error_get_pretty(local_err)); | |
4569 | error_free(local_err); | |
4570 | local_err = NULL; | |
f88e1a42 JS |
4571 | goto out; |
4572 | } | |
4573 | bdrv_get_geometry(bs, &size); | |
4574 | size *= 512; | |
4575 | ||
4576 | snprintf(buf, sizeof(buf), "%" PRId64, size); | |
4577 | set_option_parameter(param, BLOCK_OPT_SIZE, buf); | |
4578 | } else { | |
71c79813 | 4579 | error_setg(errp, "Image creation needs a size parameter"); |
f88e1a42 JS |
4580 | goto out; |
4581 | } | |
4582 | } | |
4583 | ||
f382d43a MR |
4584 | if (!quiet) { |
4585 | printf("Formatting '%s', fmt=%s ", filename, fmt); | |
4586 | print_option_parameters(param); | |
4587 | puts(""); | |
4588 | } | |
cc84d90f MR |
4589 | ret = bdrv_create(drv, filename, param, &local_err); |
4590 | if (ret == -EFBIG) { | |
4591 | /* This is generally a better message than whatever the driver would | |
4592 | * deliver (especially because of the cluster_size_hint), since that | |
4593 | * is most probably not much different from "image too large". */ | |
4594 | const char *cluster_size_hint = ""; | |
4595 | if (get_option_parameter(create_options, BLOCK_OPT_CLUSTER_SIZE)) { | |
4596 | cluster_size_hint = " (try using a larger cluster size)"; | |
f88e1a42 | 4597 | } |
cc84d90f MR |
4598 | error_setg(errp, "The image size is too large for file format '%s'" |
4599 | "%s", fmt, cluster_size_hint); | |
4600 | error_free(local_err); | |
4601 | local_err = NULL; | |
f88e1a42 JS |
4602 | } |
4603 | ||
4604 | out: | |
4605 | free_option_parameters(create_options); | |
4606 | free_option_parameters(param); | |
4607 | ||
4608 | if (bs) { | |
4f6fd349 | 4609 | bdrv_unref(bs); |
f88e1a42 | 4610 | } |
cc84d90f MR |
4611 | if (error_is_set(&local_err)) { |
4612 | error_propagate(errp, local_err); | |
4613 | } | |
f88e1a42 | 4614 | } |
85d126f3 SH |
4615 | |
4616 | AioContext *bdrv_get_aio_context(BlockDriverState *bs) | |
4617 | { | |
4618 | /* Currently BlockDriverState always uses the main loop AioContext */ | |
4619 | return qemu_get_aio_context(); | |
4620 | } | |
d616b224 SH |
4621 | |
4622 | void bdrv_add_before_write_notifier(BlockDriverState *bs, | |
4623 | NotifierWithReturn *notifier) | |
4624 | { | |
4625 | notifier_with_return_list_add(&bs->before_write_notifiers, notifier); | |
4626 | } | |
6f176b48 MR |
4627 | |
4628 | int bdrv_amend_options(BlockDriverState *bs, QEMUOptionParameter *options) | |
4629 | { | |
4630 | if (bs->drv->bdrv_amend_options == NULL) { | |
4631 | return -ENOTSUP; | |
4632 | } | |
4633 | return bs->drv->bdrv_amend_options(bs, options); | |
4634 | } |