]>
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 | } | |
7e382003 | 643 | hint = DIV_ROUND_UP(length, BDRV_SECTOR_SIZE); |
51762288 SH |
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)) { | |
8f94a6e4 KW |
772 | error_setg(errp, |
773 | !bs->read_only && bdrv_is_whitelisted(drv, true) | |
774 | ? "Driver '%s' can only be used for read-only devices" | |
775 | : "Driver '%s' is not whitelisted", | |
776 | drv->format_name); | |
b64ec4e4 FZ |
777 | return -ENOTSUP; |
778 | } | |
57915332 | 779 | |
53fec9d3 | 780 | assert(bs->copy_on_read == 0); /* bdrv_new() and bdrv_close() make it so */ |
0ebd24e0 KW |
781 | if (flags & BDRV_O_COPY_ON_READ) { |
782 | if (!bs->read_only) { | |
783 | bdrv_enable_copy_on_read(bs); | |
784 | } else { | |
785 | error_setg(errp, "Can't use copy-on-read on read-only device"); | |
786 | return -EINVAL; | |
787 | } | |
53fec9d3 SH |
788 | } |
789 | ||
c2ad1b0c KW |
790 | if (filename != NULL) { |
791 | pstrcpy(bs->filename, sizeof(bs->filename), filename); | |
792 | } else { | |
793 | bs->filename[0] = '\0'; | |
794 | } | |
57915332 | 795 | |
57915332 | 796 | bs->drv = drv; |
7267c094 | 797 | bs->opaque = g_malloc0(drv->instance_size); |
57915332 | 798 | |
03f541bd | 799 | bs->enable_write_cache = !!(flags & BDRV_O_CACHE_WB); |
e7c63796 | 800 | |
66f82cee KW |
801 | /* Open the image, either directly or using a protocol */ |
802 | if (drv->bdrv_file_open) { | |
5d186eb0 | 803 | assert(file == NULL); |
030be321 | 804 | assert(!drv->bdrv_needs_filename || filename != NULL); |
34b5d2c6 | 805 | ret = drv->bdrv_file_open(bs, options, open_flags, &local_err); |
f500a6d3 | 806 | } else { |
2af5ef70 | 807 | if (file == NULL) { |
34b5d2c6 MR |
808 | error_setg(errp, "Can't use '%s' as a block driver for the " |
809 | "protocol level", drv->format_name); | |
2af5ef70 KW |
810 | ret = -EINVAL; |
811 | goto free_and_fail; | |
812 | } | |
f500a6d3 | 813 | bs->file = file; |
34b5d2c6 | 814 | ret = drv->bdrv_open(bs, options, open_flags, &local_err); |
66f82cee KW |
815 | } |
816 | ||
57915332 | 817 | if (ret < 0) { |
34b5d2c6 MR |
818 | if (error_is_set(&local_err)) { |
819 | error_propagate(errp, local_err); | |
2fa9aa59 DH |
820 | } else if (bs->filename[0]) { |
821 | error_setg_errno(errp, -ret, "Could not open '%s'", bs->filename); | |
34b5d2c6 MR |
822 | } else { |
823 | error_setg_errno(errp, -ret, "Could not open image"); | |
824 | } | |
57915332 KW |
825 | goto free_and_fail; |
826 | } | |
827 | ||
51762288 SH |
828 | ret = refresh_total_sectors(bs, bs->total_sectors); |
829 | if (ret < 0) { | |
34b5d2c6 | 830 | error_setg_errno(errp, -ret, "Could not refresh total sector count"); |
51762288 | 831 | goto free_and_fail; |
57915332 | 832 | } |
51762288 | 833 | |
57915332 KW |
834 | #ifndef _WIN32 |
835 | if (bs->is_temporary) { | |
d4cea8df DH |
836 | assert(bs->filename[0] != '\0'); |
837 | unlink(bs->filename); | |
57915332 KW |
838 | } |
839 | #endif | |
840 | return 0; | |
841 | ||
842 | free_and_fail: | |
f500a6d3 | 843 | bs->file = NULL; |
7267c094 | 844 | g_free(bs->opaque); |
57915332 KW |
845 | bs->opaque = NULL; |
846 | bs->drv = NULL; | |
847 | return ret; | |
848 | } | |
849 | ||
b6ce07aa KW |
850 | /* |
851 | * Opens a file using a protocol (file, host_device, nbd, ...) | |
787e4a85 KW |
852 | * |
853 | * options is a QDict of options to pass to the block drivers, or NULL for an | |
854 | * empty set of options. The reference to the QDict belongs to the block layer | |
855 | * after the call (even on failure), so if the caller intends to reuse the | |
856 | * dictionary, it needs to use QINCREF() before calling bdrv_file_open. | |
b6ce07aa | 857 | */ |
787e4a85 | 858 | int bdrv_file_open(BlockDriverState **pbs, const char *filename, |
34b5d2c6 | 859 | QDict *options, int flags, Error **errp) |
ea2384d3 | 860 | { |
83f64091 | 861 | BlockDriverState *bs; |
6db95603 | 862 | BlockDriver *drv; |
c2ad1b0c | 863 | const char *drvname; |
98289620 | 864 | bool allow_protocol_prefix = false; |
34b5d2c6 | 865 | Error *local_err = NULL; |
83f64091 FB |
866 | int ret; |
867 | ||
707ff828 KW |
868 | /* NULL means an empty set of options */ |
869 | if (options == NULL) { | |
870 | options = qdict_new(); | |
871 | } | |
872 | ||
83f64091 | 873 | bs = bdrv_new(""); |
707ff828 KW |
874 | bs->options = options; |
875 | options = qdict_clone_shallow(options); | |
876 | ||
035fccdf KW |
877 | /* Fetch the file name from the options QDict if necessary */ |
878 | if (!filename) { | |
879 | filename = qdict_get_try_str(options, "filename"); | |
880 | } else if (filename && !qdict_haskey(options, "filename")) { | |
881 | qdict_put(options, "filename", qstring_from_str(filename)); | |
98289620 | 882 | allow_protocol_prefix = true; |
035fccdf | 883 | } else { |
34b5d2c6 MR |
884 | error_setg(errp, "Can't specify 'file' and 'filename' options at the " |
885 | "same time"); | |
035fccdf KW |
886 | ret = -EINVAL; |
887 | goto fail; | |
888 | } | |
889 | ||
c2ad1b0c KW |
890 | /* Find the right block driver */ |
891 | drvname = qdict_get_try_str(options, "driver"); | |
892 | if (drvname) { | |
8f94a6e4 | 893 | drv = bdrv_find_format(drvname); |
34b5d2c6 MR |
894 | if (!drv) { |
895 | error_setg(errp, "Unknown driver '%s'", drvname); | |
896 | } | |
c2ad1b0c KW |
897 | qdict_del(options, "driver"); |
898 | } else if (filename) { | |
98289620 KW |
899 | drv = bdrv_find_protocol(filename, allow_protocol_prefix); |
900 | if (!drv) { | |
34b5d2c6 | 901 | error_setg(errp, "Unknown protocol"); |
98289620 | 902 | } |
c2ad1b0c | 903 | } else { |
34b5d2c6 | 904 | error_setg(errp, "Must specify either driver or file"); |
c2ad1b0c KW |
905 | drv = NULL; |
906 | } | |
907 | ||
908 | if (!drv) { | |
34b5d2c6 | 909 | /* errp has been set already */ |
c2ad1b0c KW |
910 | ret = -ENOENT; |
911 | goto fail; | |
912 | } | |
913 | ||
914 | /* Parse the filename and open it */ | |
915 | if (drv->bdrv_parse_filename && filename) { | |
6963a30d KW |
916 | drv->bdrv_parse_filename(filename, options, &local_err); |
917 | if (error_is_set(&local_err)) { | |
34b5d2c6 | 918 | error_propagate(errp, local_err); |
6963a30d KW |
919 | ret = -EINVAL; |
920 | goto fail; | |
921 | } | |
56d1b4d2 | 922 | qdict_del(options, "filename"); |
030be321 | 923 | } else if (drv->bdrv_needs_filename && !filename) { |
34b5d2c6 MR |
924 | error_setg(errp, "The '%s' block driver requires a file name", |
925 | drv->format_name); | |
c2ad1b0c KW |
926 | ret = -EINVAL; |
927 | goto fail; | |
6963a30d KW |
928 | } |
929 | ||
34b5d2c6 | 930 | ret = bdrv_open_common(bs, NULL, options, flags, drv, &local_err); |
83f64091 | 931 | if (ret < 0) { |
34b5d2c6 | 932 | error_propagate(errp, local_err); |
707ff828 KW |
933 | goto fail; |
934 | } | |
935 | ||
936 | /* Check if any unknown options were used */ | |
937 | if (qdict_size(options) != 0) { | |
938 | const QDictEntry *entry = qdict_first(options); | |
34b5d2c6 MR |
939 | error_setg(errp, "Block protocol '%s' doesn't support the option '%s'", |
940 | drv->format_name, entry->key); | |
707ff828 KW |
941 | ret = -EINVAL; |
942 | goto fail; | |
3b0d4f61 | 943 | } |
707ff828 KW |
944 | QDECREF(options); |
945 | ||
71d0770c | 946 | bs->growable = 1; |
83f64091 FB |
947 | *pbs = bs; |
948 | return 0; | |
707ff828 KW |
949 | |
950 | fail: | |
951 | QDECREF(options); | |
952 | if (!bs->drv) { | |
953 | QDECREF(bs->options); | |
954 | } | |
4f6fd349 | 955 | bdrv_unref(bs); |
707ff828 | 956 | return ret; |
83f64091 FB |
957 | } |
958 | ||
31ca6d07 KW |
959 | /* |
960 | * Opens the backing file for a BlockDriverState if not yet open | |
961 | * | |
962 | * options is a QDict of options to pass to the block drivers, or NULL for an | |
963 | * empty set of options. The reference to the QDict is transferred to this | |
964 | * function (even on failure), so if the caller intends to reuse the dictionary, | |
965 | * it needs to use QINCREF() before calling bdrv_file_open. | |
966 | */ | |
34b5d2c6 | 967 | int bdrv_open_backing_file(BlockDriverState *bs, QDict *options, Error **errp) |
9156df12 PB |
968 | { |
969 | char backing_filename[PATH_MAX]; | |
970 | int back_flags, ret; | |
971 | BlockDriver *back_drv = NULL; | |
34b5d2c6 | 972 | Error *local_err = NULL; |
9156df12 PB |
973 | |
974 | if (bs->backing_hd != NULL) { | |
31ca6d07 | 975 | QDECREF(options); |
9156df12 PB |
976 | return 0; |
977 | } | |
978 | ||
31ca6d07 KW |
979 | /* NULL means an empty set of options */ |
980 | if (options == NULL) { | |
981 | options = qdict_new(); | |
982 | } | |
983 | ||
9156df12 | 984 | bs->open_flags &= ~BDRV_O_NO_BACKING; |
1cb6f506 KW |
985 | if (qdict_haskey(options, "file.filename")) { |
986 | backing_filename[0] = '\0'; | |
987 | } else if (bs->backing_file[0] == '\0' && qdict_size(options) == 0) { | |
31ca6d07 | 988 | QDECREF(options); |
9156df12 | 989 | return 0; |
dbecebdd FZ |
990 | } else { |
991 | bdrv_get_full_backing_filename(bs, backing_filename, | |
992 | sizeof(backing_filename)); | |
9156df12 PB |
993 | } |
994 | ||
995 | bs->backing_hd = bdrv_new(""); | |
9156df12 PB |
996 | |
997 | if (bs->backing_format[0] != '\0') { | |
998 | back_drv = bdrv_find_format(bs->backing_format); | |
999 | } | |
1000 | ||
1001 | /* backing files always opened read-only */ | |
87a5debd TL |
1002 | back_flags = bs->open_flags & ~(BDRV_O_RDWR | BDRV_O_SNAPSHOT | |
1003 | BDRV_O_COPY_ON_READ); | |
9156df12 | 1004 | |
31ca6d07 KW |
1005 | ret = bdrv_open(bs->backing_hd, |
1006 | *backing_filename ? backing_filename : NULL, options, | |
34b5d2c6 | 1007 | back_flags, back_drv, &local_err); |
9156df12 | 1008 | if (ret < 0) { |
4f6fd349 | 1009 | bdrv_unref(bs->backing_hd); |
9156df12 PB |
1010 | bs->backing_hd = NULL; |
1011 | bs->open_flags |= BDRV_O_NO_BACKING; | |
b04b6b6e FZ |
1012 | error_setg(errp, "Could not open backing file: %s", |
1013 | error_get_pretty(local_err)); | |
1014 | error_free(local_err); | |
9156df12 PB |
1015 | return ret; |
1016 | } | |
61ed2684 MR |
1017 | pstrcpy(bs->backing_file, sizeof(bs->backing_file), |
1018 | bs->backing_hd->file->filename); | |
9156df12 PB |
1019 | return 0; |
1020 | } | |
1021 | ||
b6ce07aa KW |
1022 | /* |
1023 | * Opens a disk image (raw, qcow2, vmdk, ...) | |
de9c0cec KW |
1024 | * |
1025 | * options is a QDict of options to pass to the block drivers, or NULL for an | |
1026 | * empty set of options. The reference to the QDict belongs to the block layer | |
1027 | * after the call (even on failure), so if the caller intends to reuse the | |
1028 | * dictionary, it needs to use QINCREF() before calling bdrv_open. | |
b6ce07aa | 1029 | */ |
de9c0cec | 1030 | int bdrv_open(BlockDriverState *bs, const char *filename, QDict *options, |
34b5d2c6 | 1031 | int flags, BlockDriver *drv, Error **errp) |
ea2384d3 | 1032 | { |
b6ce07aa | 1033 | int ret; |
89c9bc3d SW |
1034 | /* TODO: extra byte is a hack to ensure MAX_PATH space on Windows. */ |
1035 | char tmp_filename[PATH_MAX + 1]; | |
f500a6d3 | 1036 | BlockDriverState *file = NULL; |
707ff828 | 1037 | QDict *file_options = NULL; |
74fe54f2 | 1038 | const char *drvname; |
34b5d2c6 | 1039 | Error *local_err = NULL; |
712e7874 | 1040 | |
de9c0cec KW |
1041 | /* NULL means an empty set of options */ |
1042 | if (options == NULL) { | |
1043 | options = qdict_new(); | |
1044 | } | |
1045 | ||
1046 | bs->options = options; | |
b6ad491a | 1047 | options = qdict_clone_shallow(options); |
de9c0cec KW |
1048 | |
1049 | /* For snapshot=on, create a temporary qcow2 overlay */ | |
83f64091 | 1050 | if (flags & BDRV_O_SNAPSHOT) { |
ea2384d3 FB |
1051 | BlockDriverState *bs1; |
1052 | int64_t total_size; | |
91a073a9 | 1053 | BlockDriver *bdrv_qcow2; |
08b392e1 | 1054 | QEMUOptionParameter *create_options; |
b6ce07aa | 1055 | char backing_filename[PATH_MAX]; |
3b46e624 | 1056 | |
c2ad1b0c | 1057 | if (qdict_size(options) != 0) { |
34b5d2c6 | 1058 | error_setg(errp, "Can't use snapshot=on with driver-specific options"); |
c2ad1b0c KW |
1059 | ret = -EINVAL; |
1060 | goto fail; | |
1061 | } | |
1062 | assert(filename != NULL); | |
1063 | ||
ea2384d3 FB |
1064 | /* if snapshot, we create a temporary backing file and open it |
1065 | instead of opening 'filename' directly */ | |
33e3963e | 1066 | |
ea2384d3 FB |
1067 | /* if there is a backing file, use it */ |
1068 | bs1 = bdrv_new(""); | |
34b5d2c6 | 1069 | ret = bdrv_open(bs1, filename, NULL, 0, drv, &local_err); |
51d7c00c | 1070 | if (ret < 0) { |
4f6fd349 | 1071 | bdrv_unref(bs1); |
de9c0cec | 1072 | goto fail; |
ea2384d3 | 1073 | } |
3e82990b | 1074 | total_size = bdrv_getlength(bs1) & BDRV_SECTOR_MASK; |
7c96d46e | 1075 | |
4f6fd349 | 1076 | bdrv_unref(bs1); |
3b46e624 | 1077 | |
eba25057 JM |
1078 | ret = get_tmp_filename(tmp_filename, sizeof(tmp_filename)); |
1079 | if (ret < 0) { | |
34b5d2c6 | 1080 | error_setg_errno(errp, -ret, "Could not get temporary filename"); |
de9c0cec | 1081 | goto fail; |
eba25057 | 1082 | } |
7c96d46e AL |
1083 | |
1084 | /* Real path is meaningless for protocols */ | |
4d70655b | 1085 | if (path_has_protocol(filename)) { |
7c96d46e AL |
1086 | snprintf(backing_filename, sizeof(backing_filename), |
1087 | "%s", filename); | |
de9c0cec KW |
1088 | } else if (!realpath(filename, backing_filename)) { |
1089 | ret = -errno; | |
17826bc1 | 1090 | error_setg_errno(errp, errno, "Could not resolve path '%s'", filename); |
de9c0cec KW |
1091 | goto fail; |
1092 | } | |
7c96d46e | 1093 | |
91a073a9 | 1094 | bdrv_qcow2 = bdrv_find_format("qcow2"); |
08b392e1 KW |
1095 | create_options = parse_option_parameters("", bdrv_qcow2->create_options, |
1096 | NULL); | |
91a073a9 | 1097 | |
08b392e1 KW |
1098 | set_option_parameter_int(create_options, BLOCK_OPT_SIZE, total_size); |
1099 | set_option_parameter(create_options, BLOCK_OPT_BACKING_FILE, | |
1100 | backing_filename); | |
91a073a9 | 1101 | if (drv) { |
08b392e1 | 1102 | set_option_parameter(create_options, BLOCK_OPT_BACKING_FMT, |
91a073a9 KW |
1103 | drv->format_name); |
1104 | } | |
1105 | ||
cc84d90f | 1106 | ret = bdrv_create(bdrv_qcow2, tmp_filename, create_options, &local_err); |
08b392e1 | 1107 | free_option_parameters(create_options); |
51d7c00c | 1108 | if (ret < 0) { |
34b5d2c6 | 1109 | error_setg_errno(errp, -ret, "Could not create temporary overlay " |
cc84d90f MR |
1110 | "'%s': %s", tmp_filename, |
1111 | error_get_pretty(local_err)); | |
1112 | error_free(local_err); | |
1113 | local_err = NULL; | |
de9c0cec | 1114 | goto fail; |
ea2384d3 | 1115 | } |
91a073a9 | 1116 | |
ea2384d3 | 1117 | filename = tmp_filename; |
91a073a9 | 1118 | drv = bdrv_qcow2; |
ea2384d3 FB |
1119 | bs->is_temporary = 1; |
1120 | } | |
712e7874 | 1121 | |
f500a6d3 KW |
1122 | /* Open image file without format layer */ |
1123 | if (flags & BDRV_O_RDWR) { | |
1124 | flags |= BDRV_O_ALLOW_RDWR; | |
1125 | } | |
1126 | ||
5726d872 | 1127 | qdict_extract_subqdict(options, &file_options, "file."); |
707ff828 KW |
1128 | |
1129 | ret = bdrv_file_open(&file, filename, file_options, | |
34b5d2c6 | 1130 | bdrv_open_flags(bs, flags | BDRV_O_UNMAP), &local_err); |
f500a6d3 | 1131 | if (ret < 0) { |
de9c0cec | 1132 | goto fail; |
f500a6d3 KW |
1133 | } |
1134 | ||
b6ce07aa | 1135 | /* Find the right image format driver */ |
74fe54f2 KW |
1136 | drvname = qdict_get_try_str(options, "driver"); |
1137 | if (drvname) { | |
8f94a6e4 | 1138 | drv = bdrv_find_format(drvname); |
74fe54f2 KW |
1139 | qdict_del(options, "driver"); |
1140 | } | |
1141 | ||
6db95603 | 1142 | if (!drv) { |
34b5d2c6 | 1143 | ret = find_image_format(file, filename, &drv, &local_err); |
51d7c00c | 1144 | } |
6987307c | 1145 | |
51d7c00c | 1146 | if (!drv) { |
51d7c00c | 1147 | goto unlink_and_fail; |
ea2384d3 | 1148 | } |
b6ce07aa KW |
1149 | |
1150 | /* Open the image */ | |
34b5d2c6 | 1151 | ret = bdrv_open_common(bs, file, options, flags, drv, &local_err); |
b6ce07aa | 1152 | if (ret < 0) { |
6987307c CH |
1153 | goto unlink_and_fail; |
1154 | } | |
1155 | ||
f500a6d3 | 1156 | if (bs->file != file) { |
4f6fd349 | 1157 | bdrv_unref(file); |
f500a6d3 KW |
1158 | file = NULL; |
1159 | } | |
1160 | ||
b6ce07aa | 1161 | /* If there is a backing file, use it */ |
9156df12 | 1162 | if ((flags & BDRV_O_NO_BACKING) == 0) { |
31ca6d07 KW |
1163 | QDict *backing_options; |
1164 | ||
5726d872 | 1165 | qdict_extract_subqdict(options, &backing_options, "backing."); |
34b5d2c6 | 1166 | ret = bdrv_open_backing_file(bs, backing_options, &local_err); |
b6ce07aa | 1167 | if (ret < 0) { |
b6ad491a | 1168 | goto close_and_fail; |
b6ce07aa | 1169 | } |
b6ce07aa KW |
1170 | } |
1171 | ||
b6ad491a KW |
1172 | /* Check if any unknown options were used */ |
1173 | if (qdict_size(options) != 0) { | |
1174 | const QDictEntry *entry = qdict_first(options); | |
34b5d2c6 MR |
1175 | error_setg(errp, "Block format '%s' used by device '%s' doesn't " |
1176 | "support the option '%s'", drv->format_name, bs->device_name, | |
1177 | entry->key); | |
b6ad491a KW |
1178 | |
1179 | ret = -EINVAL; | |
1180 | goto close_and_fail; | |
1181 | } | |
1182 | QDECREF(options); | |
1183 | ||
b6ce07aa | 1184 | if (!bdrv_key_required(bs)) { |
7d4b4ba5 | 1185 | bdrv_dev_change_media_cb(bs, true); |
b6ce07aa KW |
1186 | } |
1187 | ||
1188 | return 0; | |
1189 | ||
1190 | unlink_and_fail: | |
f500a6d3 | 1191 | if (file != NULL) { |
4f6fd349 | 1192 | bdrv_unref(file); |
f500a6d3 | 1193 | } |
b6ce07aa KW |
1194 | if (bs->is_temporary) { |
1195 | unlink(filename); | |
1196 | } | |
de9c0cec KW |
1197 | fail: |
1198 | QDECREF(bs->options); | |
b6ad491a | 1199 | QDECREF(options); |
de9c0cec | 1200 | bs->options = NULL; |
34b5d2c6 MR |
1201 | if (error_is_set(&local_err)) { |
1202 | error_propagate(errp, local_err); | |
1203 | } | |
b6ad491a | 1204 | return ret; |
de9c0cec | 1205 | |
b6ad491a KW |
1206 | close_and_fail: |
1207 | bdrv_close(bs); | |
1208 | QDECREF(options); | |
34b5d2c6 MR |
1209 | if (error_is_set(&local_err)) { |
1210 | error_propagate(errp, local_err); | |
1211 | } | |
b6ce07aa KW |
1212 | return ret; |
1213 | } | |
1214 | ||
e971aa12 JC |
1215 | typedef struct BlockReopenQueueEntry { |
1216 | bool prepared; | |
1217 | BDRVReopenState state; | |
1218 | QSIMPLEQ_ENTRY(BlockReopenQueueEntry) entry; | |
1219 | } BlockReopenQueueEntry; | |
1220 | ||
1221 | /* | |
1222 | * Adds a BlockDriverState to a simple queue for an atomic, transactional | |
1223 | * reopen of multiple devices. | |
1224 | * | |
1225 | * bs_queue can either be an existing BlockReopenQueue that has had QSIMPLE_INIT | |
1226 | * already performed, or alternatively may be NULL a new BlockReopenQueue will | |
1227 | * be created and initialized. This newly created BlockReopenQueue should be | |
1228 | * passed back in for subsequent calls that are intended to be of the same | |
1229 | * atomic 'set'. | |
1230 | * | |
1231 | * bs is the BlockDriverState to add to the reopen queue. | |
1232 | * | |
1233 | * flags contains the open flags for the associated bs | |
1234 | * | |
1235 | * returns a pointer to bs_queue, which is either the newly allocated | |
1236 | * bs_queue, or the existing bs_queue being used. | |
1237 | * | |
1238 | */ | |
1239 | BlockReopenQueue *bdrv_reopen_queue(BlockReopenQueue *bs_queue, | |
1240 | BlockDriverState *bs, int flags) | |
1241 | { | |
1242 | assert(bs != NULL); | |
1243 | ||
1244 | BlockReopenQueueEntry *bs_entry; | |
1245 | if (bs_queue == NULL) { | |
1246 | bs_queue = g_new0(BlockReopenQueue, 1); | |
1247 | QSIMPLEQ_INIT(bs_queue); | |
1248 | } | |
1249 | ||
1250 | if (bs->file) { | |
1251 | bdrv_reopen_queue(bs_queue, bs->file, flags); | |
1252 | } | |
1253 | ||
1254 | bs_entry = g_new0(BlockReopenQueueEntry, 1); | |
1255 | QSIMPLEQ_INSERT_TAIL(bs_queue, bs_entry, entry); | |
1256 | ||
1257 | bs_entry->state.bs = bs; | |
1258 | bs_entry->state.flags = flags; | |
1259 | ||
1260 | return bs_queue; | |
1261 | } | |
1262 | ||
1263 | /* | |
1264 | * Reopen multiple BlockDriverStates atomically & transactionally. | |
1265 | * | |
1266 | * The queue passed in (bs_queue) must have been built up previous | |
1267 | * via bdrv_reopen_queue(). | |
1268 | * | |
1269 | * Reopens all BDS specified in the queue, with the appropriate | |
1270 | * flags. All devices are prepared for reopen, and failure of any | |
1271 | * device will cause all device changes to be abandonded, and intermediate | |
1272 | * data cleaned up. | |
1273 | * | |
1274 | * If all devices prepare successfully, then the changes are committed | |
1275 | * to all devices. | |
1276 | * | |
1277 | */ | |
1278 | int bdrv_reopen_multiple(BlockReopenQueue *bs_queue, Error **errp) | |
1279 | { | |
1280 | int ret = -1; | |
1281 | BlockReopenQueueEntry *bs_entry, *next; | |
1282 | Error *local_err = NULL; | |
1283 | ||
1284 | assert(bs_queue != NULL); | |
1285 | ||
1286 | bdrv_drain_all(); | |
1287 | ||
1288 | QSIMPLEQ_FOREACH(bs_entry, bs_queue, entry) { | |
1289 | if (bdrv_reopen_prepare(&bs_entry->state, bs_queue, &local_err)) { | |
1290 | error_propagate(errp, local_err); | |
1291 | goto cleanup; | |
1292 | } | |
1293 | bs_entry->prepared = true; | |
1294 | } | |
1295 | ||
1296 | /* If we reach this point, we have success and just need to apply the | |
1297 | * changes | |
1298 | */ | |
1299 | QSIMPLEQ_FOREACH(bs_entry, bs_queue, entry) { | |
1300 | bdrv_reopen_commit(&bs_entry->state); | |
1301 | } | |
1302 | ||
1303 | ret = 0; | |
1304 | ||
1305 | cleanup: | |
1306 | QSIMPLEQ_FOREACH_SAFE(bs_entry, bs_queue, entry, next) { | |
1307 | if (ret && bs_entry->prepared) { | |
1308 | bdrv_reopen_abort(&bs_entry->state); | |
1309 | } | |
1310 | g_free(bs_entry); | |
1311 | } | |
1312 | g_free(bs_queue); | |
1313 | return ret; | |
1314 | } | |
1315 | ||
1316 | ||
1317 | /* Reopen a single BlockDriverState with the specified flags. */ | |
1318 | int bdrv_reopen(BlockDriverState *bs, int bdrv_flags, Error **errp) | |
1319 | { | |
1320 | int ret = -1; | |
1321 | Error *local_err = NULL; | |
1322 | BlockReopenQueue *queue = bdrv_reopen_queue(NULL, bs, bdrv_flags); | |
1323 | ||
1324 | ret = bdrv_reopen_multiple(queue, &local_err); | |
1325 | if (local_err != NULL) { | |
1326 | error_propagate(errp, local_err); | |
1327 | } | |
1328 | return ret; | |
1329 | } | |
1330 | ||
1331 | ||
1332 | /* | |
1333 | * Prepares a BlockDriverState for reopen. All changes are staged in the | |
1334 | * 'opaque' field of the BDRVReopenState, which is used and allocated by | |
1335 | * the block driver layer .bdrv_reopen_prepare() | |
1336 | * | |
1337 | * bs is the BlockDriverState to reopen | |
1338 | * flags are the new open flags | |
1339 | * queue is the reopen queue | |
1340 | * | |
1341 | * Returns 0 on success, non-zero on error. On error errp will be set | |
1342 | * as well. | |
1343 | * | |
1344 | * On failure, bdrv_reopen_abort() will be called to clean up any data. | |
1345 | * It is the responsibility of the caller to then call the abort() or | |
1346 | * commit() for any other BDS that have been left in a prepare() state | |
1347 | * | |
1348 | */ | |
1349 | int bdrv_reopen_prepare(BDRVReopenState *reopen_state, BlockReopenQueue *queue, | |
1350 | Error **errp) | |
1351 | { | |
1352 | int ret = -1; | |
1353 | Error *local_err = NULL; | |
1354 | BlockDriver *drv; | |
1355 | ||
1356 | assert(reopen_state != NULL); | |
1357 | assert(reopen_state->bs->drv != NULL); | |
1358 | drv = reopen_state->bs->drv; | |
1359 | ||
1360 | /* if we are to stay read-only, do not allow permission change | |
1361 | * to r/w */ | |
1362 | if (!(reopen_state->bs->open_flags & BDRV_O_ALLOW_RDWR) && | |
1363 | reopen_state->flags & BDRV_O_RDWR) { | |
1364 | error_set(errp, QERR_DEVICE_IS_READ_ONLY, | |
1365 | reopen_state->bs->device_name); | |
1366 | goto error; | |
1367 | } | |
1368 | ||
1369 | ||
1370 | ret = bdrv_flush(reopen_state->bs); | |
1371 | if (ret) { | |
1372 | error_set(errp, ERROR_CLASS_GENERIC_ERROR, "Error (%s) flushing drive", | |
1373 | strerror(-ret)); | |
1374 | goto error; | |
1375 | } | |
1376 | ||
1377 | if (drv->bdrv_reopen_prepare) { | |
1378 | ret = drv->bdrv_reopen_prepare(reopen_state, queue, &local_err); | |
1379 | if (ret) { | |
1380 | if (local_err != NULL) { | |
1381 | error_propagate(errp, local_err); | |
1382 | } else { | |
d8b6895f LC |
1383 | error_setg(errp, "failed while preparing to reopen image '%s'", |
1384 | reopen_state->bs->filename); | |
e971aa12 JC |
1385 | } |
1386 | goto error; | |
1387 | } | |
1388 | } else { | |
1389 | /* It is currently mandatory to have a bdrv_reopen_prepare() | |
1390 | * handler for each supported drv. */ | |
1391 | error_set(errp, QERR_BLOCK_FORMAT_FEATURE_NOT_SUPPORTED, | |
1392 | drv->format_name, reopen_state->bs->device_name, | |
1393 | "reopening of file"); | |
1394 | ret = -1; | |
1395 | goto error; | |
1396 | } | |
1397 | ||
1398 | ret = 0; | |
1399 | ||
1400 | error: | |
1401 | return ret; | |
1402 | } | |
1403 | ||
1404 | /* | |
1405 | * Takes the staged changes for the reopen from bdrv_reopen_prepare(), and | |
1406 | * makes them final by swapping the staging BlockDriverState contents into | |
1407 | * the active BlockDriverState contents. | |
1408 | */ | |
1409 | void bdrv_reopen_commit(BDRVReopenState *reopen_state) | |
1410 | { | |
1411 | BlockDriver *drv; | |
1412 | ||
1413 | assert(reopen_state != NULL); | |
1414 | drv = reopen_state->bs->drv; | |
1415 | assert(drv != NULL); | |
1416 | ||
1417 | /* If there are any driver level actions to take */ | |
1418 | if (drv->bdrv_reopen_commit) { | |
1419 | drv->bdrv_reopen_commit(reopen_state); | |
1420 | } | |
1421 | ||
1422 | /* set BDS specific flags now */ | |
1423 | reopen_state->bs->open_flags = reopen_state->flags; | |
1424 | reopen_state->bs->enable_write_cache = !!(reopen_state->flags & | |
1425 | BDRV_O_CACHE_WB); | |
1426 | reopen_state->bs->read_only = !(reopen_state->flags & BDRV_O_RDWR); | |
1427 | } | |
1428 | ||
1429 | /* | |
1430 | * Abort the reopen, and delete and free the staged changes in | |
1431 | * reopen_state | |
1432 | */ | |
1433 | void bdrv_reopen_abort(BDRVReopenState *reopen_state) | |
1434 | { | |
1435 | BlockDriver *drv; | |
1436 | ||
1437 | assert(reopen_state != NULL); | |
1438 | drv = reopen_state->bs->drv; | |
1439 | assert(drv != NULL); | |
1440 | ||
1441 | if (drv->bdrv_reopen_abort) { | |
1442 | drv->bdrv_reopen_abort(reopen_state); | |
1443 | } | |
1444 | } | |
1445 | ||
1446 | ||
fc01f7e7 FB |
1447 | void bdrv_close(BlockDriverState *bs) |
1448 | { | |
3cbc002c PB |
1449 | if (bs->job) { |
1450 | block_job_cancel_sync(bs->job); | |
1451 | } | |
58fda173 SH |
1452 | bdrv_drain_all(); /* complete I/O */ |
1453 | bdrv_flush(bs); | |
1454 | bdrv_drain_all(); /* in case flush left pending I/O */ | |
d7d512f6 | 1455 | notifier_list_notify(&bs->close_notifiers, bs); |
7094f12f | 1456 | |
3cbc002c | 1457 | if (bs->drv) { |
557df6ac | 1458 | if (bs->backing_hd) { |
4f6fd349 | 1459 | bdrv_unref(bs->backing_hd); |
557df6ac SH |
1460 | bs->backing_hd = NULL; |
1461 | } | |
ea2384d3 | 1462 | bs->drv->bdrv_close(bs); |
7267c094 | 1463 | g_free(bs->opaque); |
ea2384d3 FB |
1464 | #ifdef _WIN32 |
1465 | if (bs->is_temporary) { | |
1466 | unlink(bs->filename); | |
1467 | } | |
67b915a5 | 1468 | #endif |
ea2384d3 FB |
1469 | bs->opaque = NULL; |
1470 | bs->drv = NULL; | |
53fec9d3 | 1471 | bs->copy_on_read = 0; |
a275fa42 PB |
1472 | bs->backing_file[0] = '\0'; |
1473 | bs->backing_format[0] = '\0'; | |
6405875c PB |
1474 | bs->total_sectors = 0; |
1475 | bs->encrypted = 0; | |
1476 | bs->valid_key = 0; | |
1477 | bs->sg = 0; | |
1478 | bs->growable = 0; | |
0d51b4de | 1479 | bs->zero_beyond_eof = false; |
de9c0cec KW |
1480 | QDECREF(bs->options); |
1481 | bs->options = NULL; | |
b338082b | 1482 | |
66f82cee | 1483 | if (bs->file != NULL) { |
4f6fd349 | 1484 | bdrv_unref(bs->file); |
0ac9377d | 1485 | bs->file = NULL; |
66f82cee | 1486 | } |
b338082b | 1487 | } |
98f90dba | 1488 | |
9ca11154 PH |
1489 | bdrv_dev_change_media_cb(bs, false); |
1490 | ||
98f90dba ZYW |
1491 | /*throttling disk I/O limits*/ |
1492 | if (bs->io_limits_enabled) { | |
1493 | bdrv_io_limits_disable(bs); | |
1494 | } | |
b338082b FB |
1495 | } |
1496 | ||
2bc93fed MK |
1497 | void bdrv_close_all(void) |
1498 | { | |
1499 | BlockDriverState *bs; | |
1500 | ||
1501 | QTAILQ_FOREACH(bs, &bdrv_states, list) { | |
1502 | bdrv_close(bs); | |
1503 | } | |
1504 | } | |
1505 | ||
88266f5a SH |
1506 | /* Check if any requests are in-flight (including throttled requests) */ |
1507 | static bool bdrv_requests_pending(BlockDriverState *bs) | |
1508 | { | |
1509 | if (!QLIST_EMPTY(&bs->tracked_requests)) { | |
1510 | return true; | |
1511 | } | |
cc0681c4 BC |
1512 | if (!qemu_co_queue_empty(&bs->throttled_reqs[0])) { |
1513 | return true; | |
1514 | } | |
1515 | if (!qemu_co_queue_empty(&bs->throttled_reqs[1])) { | |
88266f5a SH |
1516 | return true; |
1517 | } | |
1518 | if (bs->file && bdrv_requests_pending(bs->file)) { | |
1519 | return true; | |
1520 | } | |
1521 | if (bs->backing_hd && bdrv_requests_pending(bs->backing_hd)) { | |
1522 | return true; | |
1523 | } | |
1524 | return false; | |
1525 | } | |
1526 | ||
1527 | static bool bdrv_requests_pending_all(void) | |
1528 | { | |
1529 | BlockDriverState *bs; | |
1530 | QTAILQ_FOREACH(bs, &bdrv_states, list) { | |
1531 | if (bdrv_requests_pending(bs)) { | |
1532 | return true; | |
1533 | } | |
1534 | } | |
1535 | return false; | |
1536 | } | |
1537 | ||
922453bc SH |
1538 | /* |
1539 | * Wait for pending requests to complete across all BlockDriverStates | |
1540 | * | |
1541 | * This function does not flush data to disk, use bdrv_flush_all() for that | |
1542 | * after calling this function. | |
4c355d53 ZYW |
1543 | * |
1544 | * Note that completion of an asynchronous I/O operation can trigger any | |
1545 | * number of other I/O operations on other devices---for example a coroutine | |
1546 | * can be arbitrarily complex and a constant flow of I/O can come until the | |
1547 | * coroutine is complete. Because of this, it is not possible to have a | |
1548 | * function to drain a single device's I/O queue. | |
922453bc SH |
1549 | */ |
1550 | void bdrv_drain_all(void) | |
1551 | { | |
88266f5a SH |
1552 | /* Always run first iteration so any pending completion BHs run */ |
1553 | bool busy = true; | |
922453bc SH |
1554 | BlockDriverState *bs; |
1555 | ||
88266f5a | 1556 | while (busy) { |
4c355d53 ZYW |
1557 | /* FIXME: We do not have timer support here, so this is effectively |
1558 | * a busy wait. | |
1559 | */ | |
1560 | QTAILQ_FOREACH(bs, &bdrv_states, list) { | |
cc0681c4 | 1561 | if (bdrv_start_throttled_reqs(bs)) { |
4c355d53 ZYW |
1562 | busy = true; |
1563 | } | |
1564 | } | |
922453bc | 1565 | |
88266f5a SH |
1566 | busy = bdrv_requests_pending_all(); |
1567 | busy |= aio_poll(qemu_get_aio_context(), busy); | |
922453bc SH |
1568 | } |
1569 | } | |
1570 | ||
d22b2f41 RH |
1571 | /* make a BlockDriverState anonymous by removing from bdrv_state list. |
1572 | Also, NULL terminate the device_name to prevent double remove */ | |
1573 | void bdrv_make_anon(BlockDriverState *bs) | |
1574 | { | |
1575 | if (bs->device_name[0] != '\0') { | |
1576 | QTAILQ_REMOVE(&bdrv_states, bs, list); | |
1577 | } | |
1578 | bs->device_name[0] = '\0'; | |
1579 | } | |
1580 | ||
e023b2e2 PB |
1581 | static void bdrv_rebind(BlockDriverState *bs) |
1582 | { | |
1583 | if (bs->drv && bs->drv->bdrv_rebind) { | |
1584 | bs->drv->bdrv_rebind(bs); | |
1585 | } | |
1586 | } | |
1587 | ||
4ddc07ca PB |
1588 | static void bdrv_move_feature_fields(BlockDriverState *bs_dest, |
1589 | BlockDriverState *bs_src) | |
8802d1fd | 1590 | { |
4ddc07ca PB |
1591 | /* move some fields that need to stay attached to the device */ |
1592 | bs_dest->open_flags = bs_src->open_flags; | |
8802d1fd JC |
1593 | |
1594 | /* dev info */ | |
4ddc07ca PB |
1595 | bs_dest->dev_ops = bs_src->dev_ops; |
1596 | bs_dest->dev_opaque = bs_src->dev_opaque; | |
1597 | bs_dest->dev = bs_src->dev; | |
1598 | bs_dest->buffer_alignment = bs_src->buffer_alignment; | |
1599 | bs_dest->copy_on_read = bs_src->copy_on_read; | |
8802d1fd | 1600 | |
4ddc07ca | 1601 | bs_dest->enable_write_cache = bs_src->enable_write_cache; |
c4a248a1 | 1602 | |
cc0681c4 BC |
1603 | /* i/o throttled req */ |
1604 | memcpy(&bs_dest->throttle_state, | |
1605 | &bs_src->throttle_state, | |
1606 | sizeof(ThrottleState)); | |
1607 | bs_dest->throttled_reqs[0] = bs_src->throttled_reqs[0]; | |
1608 | bs_dest->throttled_reqs[1] = bs_src->throttled_reqs[1]; | |
4ddc07ca | 1609 | bs_dest->io_limits_enabled = bs_src->io_limits_enabled; |
8802d1fd | 1610 | |
8802d1fd | 1611 | /* r/w error */ |
4ddc07ca PB |
1612 | bs_dest->on_read_error = bs_src->on_read_error; |
1613 | bs_dest->on_write_error = bs_src->on_write_error; | |
8802d1fd JC |
1614 | |
1615 | /* i/o status */ | |
4ddc07ca PB |
1616 | bs_dest->iostatus_enabled = bs_src->iostatus_enabled; |
1617 | bs_dest->iostatus = bs_src->iostatus; | |
8802d1fd | 1618 | |
a9fc4408 | 1619 | /* dirty bitmap */ |
4ddc07ca | 1620 | bs_dest->dirty_bitmap = bs_src->dirty_bitmap; |
a9fc4408 | 1621 | |
9fcb0251 FZ |
1622 | /* reference count */ |
1623 | bs_dest->refcnt = bs_src->refcnt; | |
1624 | ||
a9fc4408 | 1625 | /* job */ |
4ddc07ca PB |
1626 | bs_dest->in_use = bs_src->in_use; |
1627 | bs_dest->job = bs_src->job; | |
a9fc4408 | 1628 | |
8802d1fd | 1629 | /* keep the same entry in bdrv_states */ |
4ddc07ca PB |
1630 | pstrcpy(bs_dest->device_name, sizeof(bs_dest->device_name), |
1631 | bs_src->device_name); | |
1632 | bs_dest->list = bs_src->list; | |
1633 | } | |
8802d1fd | 1634 | |
4ddc07ca PB |
1635 | /* |
1636 | * Swap bs contents for two image chains while they are live, | |
1637 | * while keeping required fields on the BlockDriverState that is | |
1638 | * actually attached to a device. | |
1639 | * | |
1640 | * This will modify the BlockDriverState fields, and swap contents | |
1641 | * between bs_new and bs_old. Both bs_new and bs_old are modified. | |
1642 | * | |
1643 | * bs_new is required to be anonymous. | |
1644 | * | |
1645 | * This function does not create any image files. | |
1646 | */ | |
1647 | void bdrv_swap(BlockDriverState *bs_new, BlockDriverState *bs_old) | |
1648 | { | |
1649 | BlockDriverState tmp; | |
f6801b83 | 1650 | |
4ddc07ca PB |
1651 | /* bs_new must be anonymous and shouldn't have anything fancy enabled */ |
1652 | assert(bs_new->device_name[0] == '\0'); | |
1653 | assert(bs_new->dirty_bitmap == NULL); | |
1654 | assert(bs_new->job == NULL); | |
1655 | assert(bs_new->dev == NULL); | |
1656 | assert(bs_new->in_use == 0); | |
1657 | assert(bs_new->io_limits_enabled == false); | |
cc0681c4 | 1658 | assert(!throttle_have_timer(&bs_new->throttle_state)); |
8802d1fd | 1659 | |
4ddc07ca PB |
1660 | tmp = *bs_new; |
1661 | *bs_new = *bs_old; | |
1662 | *bs_old = tmp; | |
a9fc4408 | 1663 | |
4ddc07ca PB |
1664 | /* there are some fields that should not be swapped, move them back */ |
1665 | bdrv_move_feature_fields(&tmp, bs_old); | |
1666 | bdrv_move_feature_fields(bs_old, bs_new); | |
1667 | bdrv_move_feature_fields(bs_new, &tmp); | |
8802d1fd | 1668 | |
4ddc07ca PB |
1669 | /* bs_new shouldn't be in bdrv_states even after the swap! */ |
1670 | assert(bs_new->device_name[0] == '\0'); | |
1671 | ||
1672 | /* Check a few fields that should remain attached to the device */ | |
1673 | assert(bs_new->dev == NULL); | |
1674 | assert(bs_new->job == NULL); | |
1675 | assert(bs_new->in_use == 0); | |
1676 | assert(bs_new->io_limits_enabled == false); | |
cc0681c4 | 1677 | assert(!throttle_have_timer(&bs_new->throttle_state)); |
e023b2e2 PB |
1678 | |
1679 | bdrv_rebind(bs_new); | |
4ddc07ca PB |
1680 | bdrv_rebind(bs_old); |
1681 | } | |
1682 | ||
1683 | /* | |
1684 | * Add new bs contents at the top of an image chain while the chain is | |
1685 | * live, while keeping required fields on the top layer. | |
1686 | * | |
1687 | * This will modify the BlockDriverState fields, and swap contents | |
1688 | * between bs_new and bs_top. Both bs_new and bs_top are modified. | |
1689 | * | |
1690 | * bs_new is required to be anonymous. | |
1691 | * | |
1692 | * This function does not create any image files. | |
1693 | */ | |
1694 | void bdrv_append(BlockDriverState *bs_new, BlockDriverState *bs_top) | |
1695 | { | |
1696 | bdrv_swap(bs_new, bs_top); | |
1697 | ||
1698 | /* The contents of 'tmp' will become bs_top, as we are | |
1699 | * swapping bs_new and bs_top contents. */ | |
1700 | bs_top->backing_hd = bs_new; | |
1701 | bs_top->open_flags &= ~BDRV_O_NO_BACKING; | |
1702 | pstrcpy(bs_top->backing_file, sizeof(bs_top->backing_file), | |
1703 | bs_new->filename); | |
1704 | pstrcpy(bs_top->backing_format, sizeof(bs_top->backing_format), | |
1705 | bs_new->drv ? bs_new->drv->format_name : ""); | |
8802d1fd JC |
1706 | } |
1707 | ||
4f6fd349 | 1708 | static void bdrv_delete(BlockDriverState *bs) |
b338082b | 1709 | { |
fa879d62 | 1710 | assert(!bs->dev); |
3e914655 PB |
1711 | assert(!bs->job); |
1712 | assert(!bs->in_use); | |
4f6fd349 | 1713 | assert(!bs->refcnt); |
18846dee | 1714 | |
e1b5c52e SH |
1715 | bdrv_close(bs); |
1716 | ||
1b7bdbc1 | 1717 | /* remove from list, if necessary */ |
d22b2f41 | 1718 | bdrv_make_anon(bs); |
34c6f050 | 1719 | |
7267c094 | 1720 | g_free(bs); |
fc01f7e7 FB |
1721 | } |
1722 | ||
fa879d62 MA |
1723 | int bdrv_attach_dev(BlockDriverState *bs, void *dev) |
1724 | /* TODO change to DeviceState *dev when all users are qdevified */ | |
18846dee | 1725 | { |
fa879d62 | 1726 | if (bs->dev) { |
18846dee MA |
1727 | return -EBUSY; |
1728 | } | |
fa879d62 | 1729 | bs->dev = dev; |
28a7282a | 1730 | bdrv_iostatus_reset(bs); |
18846dee MA |
1731 | return 0; |
1732 | } | |
1733 | ||
fa879d62 MA |
1734 | /* TODO qdevified devices don't use this, remove when devices are qdevified */ |
1735 | void bdrv_attach_dev_nofail(BlockDriverState *bs, void *dev) | |
18846dee | 1736 | { |
fa879d62 MA |
1737 | if (bdrv_attach_dev(bs, dev) < 0) { |
1738 | abort(); | |
1739 | } | |
1740 | } | |
1741 | ||
1742 | void bdrv_detach_dev(BlockDriverState *bs, void *dev) | |
1743 | /* TODO change to DeviceState *dev when all users are qdevified */ | |
1744 | { | |
1745 | assert(bs->dev == dev); | |
1746 | bs->dev = NULL; | |
0e49de52 MA |
1747 | bs->dev_ops = NULL; |
1748 | bs->dev_opaque = NULL; | |
29e05f20 | 1749 | bs->buffer_alignment = 512; |
18846dee MA |
1750 | } |
1751 | ||
fa879d62 MA |
1752 | /* TODO change to return DeviceState * when all users are qdevified */ |
1753 | void *bdrv_get_attached_dev(BlockDriverState *bs) | |
18846dee | 1754 | { |
fa879d62 | 1755 | return bs->dev; |
18846dee MA |
1756 | } |
1757 | ||
0e49de52 MA |
1758 | void bdrv_set_dev_ops(BlockDriverState *bs, const BlockDevOps *ops, |
1759 | void *opaque) | |
1760 | { | |
1761 | bs->dev_ops = ops; | |
1762 | bs->dev_opaque = opaque; | |
1763 | } | |
1764 | ||
32c81a4a PB |
1765 | void bdrv_emit_qmp_error_event(const BlockDriverState *bdrv, |
1766 | enum MonitorEvent ev, | |
1767 | BlockErrorAction action, bool is_read) | |
329c0a48 LC |
1768 | { |
1769 | QObject *data; | |
1770 | const char *action_str; | |
1771 | ||
1772 | switch (action) { | |
1773 | case BDRV_ACTION_REPORT: | |
1774 | action_str = "report"; | |
1775 | break; | |
1776 | case BDRV_ACTION_IGNORE: | |
1777 | action_str = "ignore"; | |
1778 | break; | |
1779 | case BDRV_ACTION_STOP: | |
1780 | action_str = "stop"; | |
1781 | break; | |
1782 | default: | |
1783 | abort(); | |
1784 | } | |
1785 | ||
1786 | data = qobject_from_jsonf("{ 'device': %s, 'action': %s, 'operation': %s }", | |
1787 | bdrv->device_name, | |
1788 | action_str, | |
1789 | is_read ? "read" : "write"); | |
32c81a4a | 1790 | monitor_protocol_event(ev, data); |
329c0a48 LC |
1791 | |
1792 | qobject_decref(data); | |
1793 | } | |
1794 | ||
6f382ed2 LC |
1795 | static void bdrv_emit_qmp_eject_event(BlockDriverState *bs, bool ejected) |
1796 | { | |
1797 | QObject *data; | |
1798 | ||
1799 | data = qobject_from_jsonf("{ 'device': %s, 'tray-open': %i }", | |
1800 | bdrv_get_device_name(bs), ejected); | |
1801 | monitor_protocol_event(QEVENT_DEVICE_TRAY_MOVED, data); | |
1802 | ||
1803 | qobject_decref(data); | |
1804 | } | |
1805 | ||
7d4b4ba5 | 1806 | static void bdrv_dev_change_media_cb(BlockDriverState *bs, bool load) |
0e49de52 | 1807 | { |
145feb17 | 1808 | if (bs->dev_ops && bs->dev_ops->change_media_cb) { |
6f382ed2 | 1809 | bool tray_was_closed = !bdrv_dev_is_tray_open(bs); |
7d4b4ba5 | 1810 | bs->dev_ops->change_media_cb(bs->dev_opaque, load); |
6f382ed2 LC |
1811 | if (tray_was_closed) { |
1812 | /* tray open */ | |
1813 | bdrv_emit_qmp_eject_event(bs, true); | |
1814 | } | |
1815 | if (load) { | |
1816 | /* tray close */ | |
1817 | bdrv_emit_qmp_eject_event(bs, false); | |
1818 | } | |
145feb17 MA |
1819 | } |
1820 | } | |
1821 | ||
2c6942fa MA |
1822 | bool bdrv_dev_has_removable_media(BlockDriverState *bs) |
1823 | { | |
1824 | return !bs->dev || (bs->dev_ops && bs->dev_ops->change_media_cb); | |
1825 | } | |
1826 | ||
025ccaa7 PB |
1827 | void bdrv_dev_eject_request(BlockDriverState *bs, bool force) |
1828 | { | |
1829 | if (bs->dev_ops && bs->dev_ops->eject_request_cb) { | |
1830 | bs->dev_ops->eject_request_cb(bs->dev_opaque, force); | |
1831 | } | |
1832 | } | |
1833 | ||
e4def80b MA |
1834 | bool bdrv_dev_is_tray_open(BlockDriverState *bs) |
1835 | { | |
1836 | if (bs->dev_ops && bs->dev_ops->is_tray_open) { | |
1837 | return bs->dev_ops->is_tray_open(bs->dev_opaque); | |
1838 | } | |
1839 | return false; | |
1840 | } | |
1841 | ||
145feb17 MA |
1842 | static void bdrv_dev_resize_cb(BlockDriverState *bs) |
1843 | { | |
1844 | if (bs->dev_ops && bs->dev_ops->resize_cb) { | |
1845 | bs->dev_ops->resize_cb(bs->dev_opaque); | |
0e49de52 MA |
1846 | } |
1847 | } | |
1848 | ||
f107639a MA |
1849 | bool bdrv_dev_is_medium_locked(BlockDriverState *bs) |
1850 | { | |
1851 | if (bs->dev_ops && bs->dev_ops->is_medium_locked) { | |
1852 | return bs->dev_ops->is_medium_locked(bs->dev_opaque); | |
1853 | } | |
1854 | return false; | |
1855 | } | |
1856 | ||
e97fc193 AL |
1857 | /* |
1858 | * Run consistency checks on an image | |
1859 | * | |
e076f338 | 1860 | * Returns 0 if the check could be completed (it doesn't mean that the image is |
a1c7273b | 1861 | * free of errors) or -errno when an internal error occurred. The results of the |
e076f338 | 1862 | * check are stored in res. |
e97fc193 | 1863 | */ |
4534ff54 | 1864 | int bdrv_check(BlockDriverState *bs, BdrvCheckResult *res, BdrvCheckMode fix) |
e97fc193 AL |
1865 | { |
1866 | if (bs->drv->bdrv_check == NULL) { | |
1867 | return -ENOTSUP; | |
1868 | } | |
1869 | ||
e076f338 | 1870 | memset(res, 0, sizeof(*res)); |
4534ff54 | 1871 | return bs->drv->bdrv_check(bs, res, fix); |
e97fc193 AL |
1872 | } |
1873 | ||
8a426614 KW |
1874 | #define COMMIT_BUF_SECTORS 2048 |
1875 | ||
33e3963e FB |
1876 | /* commit COW file into the raw image */ |
1877 | int bdrv_commit(BlockDriverState *bs) | |
1878 | { | |
19cb3738 | 1879 | BlockDriver *drv = bs->drv; |
8a426614 KW |
1880 | int64_t sector, total_sectors; |
1881 | int n, ro, open_flags; | |
0bce597d | 1882 | int ret = 0; |
8a426614 | 1883 | uint8_t *buf; |
c2cba3d9 | 1884 | char filename[PATH_MAX]; |
33e3963e | 1885 | |
19cb3738 FB |
1886 | if (!drv) |
1887 | return -ENOMEDIUM; | |
4dca4b63 NS |
1888 | |
1889 | if (!bs->backing_hd) { | |
1890 | return -ENOTSUP; | |
33e3963e FB |
1891 | } |
1892 | ||
2d3735d3 SH |
1893 | if (bdrv_in_use(bs) || bdrv_in_use(bs->backing_hd)) { |
1894 | return -EBUSY; | |
1895 | } | |
1896 | ||
4dca4b63 | 1897 | ro = bs->backing_hd->read_only; |
c2cba3d9 JM |
1898 | /* Use pstrcpy (not strncpy): filename must be NUL-terminated. */ |
1899 | pstrcpy(filename, sizeof(filename), bs->backing_hd->filename); | |
4dca4b63 NS |
1900 | open_flags = bs->backing_hd->open_flags; |
1901 | ||
1902 | if (ro) { | |
0bce597d JC |
1903 | if (bdrv_reopen(bs->backing_hd, open_flags | BDRV_O_RDWR, NULL)) { |
1904 | return -EACCES; | |
4dca4b63 | 1905 | } |
ea2384d3 | 1906 | } |
33e3963e | 1907 | |
6ea44308 | 1908 | total_sectors = bdrv_getlength(bs) >> BDRV_SECTOR_BITS; |
7267c094 | 1909 | buf = g_malloc(COMMIT_BUF_SECTORS * BDRV_SECTOR_SIZE); |
8a426614 KW |
1910 | |
1911 | for (sector = 0; sector < total_sectors; sector += n) { | |
d663640c PB |
1912 | ret = bdrv_is_allocated(bs, sector, COMMIT_BUF_SECTORS, &n); |
1913 | if (ret < 0) { | |
1914 | goto ro_cleanup; | |
1915 | } | |
1916 | if (ret) { | |
8a426614 KW |
1917 | if (bdrv_read(bs, sector, buf, n) != 0) { |
1918 | ret = -EIO; | |
1919 | goto ro_cleanup; | |
1920 | } | |
1921 | ||
1922 | if (bdrv_write(bs->backing_hd, sector, buf, n) != 0) { | |
1923 | ret = -EIO; | |
1924 | goto ro_cleanup; | |
1925 | } | |
ea2384d3 | 1926 | } |
33e3963e | 1927 | } |
95389c86 | 1928 | |
1d44952f CH |
1929 | if (drv->bdrv_make_empty) { |
1930 | ret = drv->bdrv_make_empty(bs); | |
1931 | bdrv_flush(bs); | |
1932 | } | |
95389c86 | 1933 | |
3f5075ae CH |
1934 | /* |
1935 | * Make sure all data we wrote to the backing device is actually | |
1936 | * stable on disk. | |
1937 | */ | |
1938 | if (bs->backing_hd) | |
1939 | bdrv_flush(bs->backing_hd); | |
4dca4b63 NS |
1940 | |
1941 | ro_cleanup: | |
7267c094 | 1942 | g_free(buf); |
4dca4b63 NS |
1943 | |
1944 | if (ro) { | |
0bce597d JC |
1945 | /* ignoring error return here */ |
1946 | bdrv_reopen(bs->backing_hd, open_flags & ~BDRV_O_RDWR, NULL); | |
4dca4b63 NS |
1947 | } |
1948 | ||
1d44952f | 1949 | return ret; |
33e3963e FB |
1950 | } |
1951 | ||
e8877497 | 1952 | int bdrv_commit_all(void) |
6ab4b5ab MA |
1953 | { |
1954 | BlockDriverState *bs; | |
1955 | ||
1956 | QTAILQ_FOREACH(bs, &bdrv_states, list) { | |
272d2d8e JC |
1957 | if (bs->drv && bs->backing_hd) { |
1958 | int ret = bdrv_commit(bs); | |
1959 | if (ret < 0) { | |
1960 | return ret; | |
1961 | } | |
e8877497 | 1962 | } |
6ab4b5ab | 1963 | } |
e8877497 | 1964 | return 0; |
6ab4b5ab MA |
1965 | } |
1966 | ||
dbffbdcf SH |
1967 | /** |
1968 | * Remove an active request from the tracked requests list | |
1969 | * | |
1970 | * This function should be called when a tracked request is completing. | |
1971 | */ | |
1972 | static void tracked_request_end(BdrvTrackedRequest *req) | |
1973 | { | |
1974 | QLIST_REMOVE(req, list); | |
f4658285 | 1975 | qemu_co_queue_restart_all(&req->wait_queue); |
dbffbdcf SH |
1976 | } |
1977 | ||
1978 | /** | |
1979 | * Add an active request to the tracked requests list | |
1980 | */ | |
1981 | static void tracked_request_begin(BdrvTrackedRequest *req, | |
1982 | BlockDriverState *bs, | |
1983 | int64_t sector_num, | |
1984 | int nb_sectors, bool is_write) | |
1985 | { | |
1986 | *req = (BdrvTrackedRequest){ | |
1987 | .bs = bs, | |
1988 | .sector_num = sector_num, | |
1989 | .nb_sectors = nb_sectors, | |
1990 | .is_write = is_write, | |
5f8b6491 | 1991 | .co = qemu_coroutine_self(), |
dbffbdcf SH |
1992 | }; |
1993 | ||
f4658285 SH |
1994 | qemu_co_queue_init(&req->wait_queue); |
1995 | ||
dbffbdcf SH |
1996 | QLIST_INSERT_HEAD(&bs->tracked_requests, req, list); |
1997 | } | |
1998 | ||
d83947ac SH |
1999 | /** |
2000 | * Round a region to cluster boundaries | |
2001 | */ | |
343bded4 PB |
2002 | void bdrv_round_to_clusters(BlockDriverState *bs, |
2003 | int64_t sector_num, int nb_sectors, | |
2004 | int64_t *cluster_sector_num, | |
2005 | int *cluster_nb_sectors) | |
d83947ac SH |
2006 | { |
2007 | BlockDriverInfo bdi; | |
2008 | ||
2009 | if (bdrv_get_info(bs, &bdi) < 0 || bdi.cluster_size == 0) { | |
2010 | *cluster_sector_num = sector_num; | |
2011 | *cluster_nb_sectors = nb_sectors; | |
2012 | } else { | |
2013 | int64_t c = bdi.cluster_size / BDRV_SECTOR_SIZE; | |
2014 | *cluster_sector_num = QEMU_ALIGN_DOWN(sector_num, c); | |
2015 | *cluster_nb_sectors = QEMU_ALIGN_UP(sector_num - *cluster_sector_num + | |
2016 | nb_sectors, c); | |
2017 | } | |
2018 | } | |
2019 | ||
f4658285 SH |
2020 | static bool tracked_request_overlaps(BdrvTrackedRequest *req, |
2021 | int64_t sector_num, int nb_sectors) { | |
d83947ac SH |
2022 | /* aaaa bbbb */ |
2023 | if (sector_num >= req->sector_num + req->nb_sectors) { | |
2024 | return false; | |
2025 | } | |
2026 | /* bbbb aaaa */ | |
2027 | if (req->sector_num >= sector_num + nb_sectors) { | |
2028 | return false; | |
2029 | } | |
2030 | return true; | |
f4658285 SH |
2031 | } |
2032 | ||
2033 | static void coroutine_fn wait_for_overlapping_requests(BlockDriverState *bs, | |
2034 | int64_t sector_num, int nb_sectors) | |
2035 | { | |
2036 | BdrvTrackedRequest *req; | |
d83947ac SH |
2037 | int64_t cluster_sector_num; |
2038 | int cluster_nb_sectors; | |
f4658285 SH |
2039 | bool retry; |
2040 | ||
d83947ac SH |
2041 | /* If we touch the same cluster it counts as an overlap. This guarantees |
2042 | * that allocating writes will be serialized and not race with each other | |
2043 | * for the same cluster. For example, in copy-on-read it ensures that the | |
2044 | * CoR read and write operations are atomic and guest writes cannot | |
2045 | * interleave between them. | |
2046 | */ | |
343bded4 PB |
2047 | bdrv_round_to_clusters(bs, sector_num, nb_sectors, |
2048 | &cluster_sector_num, &cluster_nb_sectors); | |
d83947ac | 2049 | |
f4658285 SH |
2050 | do { |
2051 | retry = false; | |
2052 | QLIST_FOREACH(req, &bs->tracked_requests, list) { | |
d83947ac SH |
2053 | if (tracked_request_overlaps(req, cluster_sector_num, |
2054 | cluster_nb_sectors)) { | |
5f8b6491 SH |
2055 | /* Hitting this means there was a reentrant request, for |
2056 | * example, a block driver issuing nested requests. This must | |
2057 | * never happen since it means deadlock. | |
2058 | */ | |
2059 | assert(qemu_coroutine_self() != req->co); | |
2060 | ||
f4658285 SH |
2061 | qemu_co_queue_wait(&req->wait_queue); |
2062 | retry = true; | |
2063 | break; | |
2064 | } | |
2065 | } | |
2066 | } while (retry); | |
2067 | } | |
2068 | ||
756e6736 KW |
2069 | /* |
2070 | * Return values: | |
2071 | * 0 - success | |
2072 | * -EINVAL - backing format specified, but no file | |
2073 | * -ENOSPC - can't update the backing file because no space is left in the | |
2074 | * image file header | |
2075 | * -ENOTSUP - format driver doesn't support changing the backing file | |
2076 | */ | |
2077 | int bdrv_change_backing_file(BlockDriverState *bs, | |
2078 | const char *backing_file, const char *backing_fmt) | |
2079 | { | |
2080 | BlockDriver *drv = bs->drv; | |
469ef350 | 2081 | int ret; |
756e6736 | 2082 | |
5f377794 PB |
2083 | /* Backing file format doesn't make sense without a backing file */ |
2084 | if (backing_fmt && !backing_file) { | |
2085 | return -EINVAL; | |
2086 | } | |
2087 | ||
756e6736 | 2088 | if (drv->bdrv_change_backing_file != NULL) { |
469ef350 | 2089 | ret = drv->bdrv_change_backing_file(bs, backing_file, backing_fmt); |
756e6736 | 2090 | } else { |
469ef350 | 2091 | ret = -ENOTSUP; |
756e6736 | 2092 | } |
469ef350 PB |
2093 | |
2094 | if (ret == 0) { | |
2095 | pstrcpy(bs->backing_file, sizeof(bs->backing_file), backing_file ?: ""); | |
2096 | pstrcpy(bs->backing_format, sizeof(bs->backing_format), backing_fmt ?: ""); | |
2097 | } | |
2098 | return ret; | |
756e6736 KW |
2099 | } |
2100 | ||
6ebdcee2 JC |
2101 | /* |
2102 | * Finds the image layer in the chain that has 'bs' as its backing file. | |
2103 | * | |
2104 | * active is the current topmost image. | |
2105 | * | |
2106 | * Returns NULL if bs is not found in active's image chain, | |
2107 | * or if active == bs. | |
2108 | */ | |
2109 | BlockDriverState *bdrv_find_overlay(BlockDriverState *active, | |
2110 | BlockDriverState *bs) | |
2111 | { | |
2112 | BlockDriverState *overlay = NULL; | |
2113 | BlockDriverState *intermediate; | |
2114 | ||
2115 | assert(active != NULL); | |
2116 | assert(bs != NULL); | |
2117 | ||
2118 | /* if bs is the same as active, then by definition it has no overlay | |
2119 | */ | |
2120 | if (active == bs) { | |
2121 | return NULL; | |
2122 | } | |
2123 | ||
2124 | intermediate = active; | |
2125 | while (intermediate->backing_hd) { | |
2126 | if (intermediate->backing_hd == bs) { | |
2127 | overlay = intermediate; | |
2128 | break; | |
2129 | } | |
2130 | intermediate = intermediate->backing_hd; | |
2131 | } | |
2132 | ||
2133 | return overlay; | |
2134 | } | |
2135 | ||
2136 | typedef struct BlkIntermediateStates { | |
2137 | BlockDriverState *bs; | |
2138 | QSIMPLEQ_ENTRY(BlkIntermediateStates) entry; | |
2139 | } BlkIntermediateStates; | |
2140 | ||
2141 | ||
2142 | /* | |
2143 | * Drops images above 'base' up to and including 'top', and sets the image | |
2144 | * above 'top' to have base as its backing file. | |
2145 | * | |
2146 | * Requires that the overlay to 'top' is opened r/w, so that the backing file | |
2147 | * information in 'bs' can be properly updated. | |
2148 | * | |
2149 | * E.g., this will convert the following chain: | |
2150 | * bottom <- base <- intermediate <- top <- active | |
2151 | * | |
2152 | * to | |
2153 | * | |
2154 | * bottom <- base <- active | |
2155 | * | |
2156 | * It is allowed for bottom==base, in which case it converts: | |
2157 | * | |
2158 | * base <- intermediate <- top <- active | |
2159 | * | |
2160 | * to | |
2161 | * | |
2162 | * base <- active | |
2163 | * | |
2164 | * Error conditions: | |
2165 | * if active == top, that is considered an error | |
2166 | * | |
2167 | */ | |
2168 | int bdrv_drop_intermediate(BlockDriverState *active, BlockDriverState *top, | |
2169 | BlockDriverState *base) | |
2170 | { | |
2171 | BlockDriverState *intermediate; | |
2172 | BlockDriverState *base_bs = NULL; | |
2173 | BlockDriverState *new_top_bs = NULL; | |
2174 | BlkIntermediateStates *intermediate_state, *next; | |
2175 | int ret = -EIO; | |
2176 | ||
2177 | QSIMPLEQ_HEAD(states_to_delete, BlkIntermediateStates) states_to_delete; | |
2178 | QSIMPLEQ_INIT(&states_to_delete); | |
2179 | ||
2180 | if (!top->drv || !base->drv) { | |
2181 | goto exit; | |
2182 | } | |
2183 | ||
2184 | new_top_bs = bdrv_find_overlay(active, top); | |
2185 | ||
2186 | if (new_top_bs == NULL) { | |
2187 | /* we could not find the image above 'top', this is an error */ | |
2188 | goto exit; | |
2189 | } | |
2190 | ||
2191 | /* special case of new_top_bs->backing_hd already pointing to base - nothing | |
2192 | * to do, no intermediate images */ | |
2193 | if (new_top_bs->backing_hd == base) { | |
2194 | ret = 0; | |
2195 | goto exit; | |
2196 | } | |
2197 | ||
2198 | intermediate = top; | |
2199 | ||
2200 | /* now we will go down through the list, and add each BDS we find | |
2201 | * into our deletion queue, until we hit the 'base' | |
2202 | */ | |
2203 | while (intermediate) { | |
2204 | intermediate_state = g_malloc0(sizeof(BlkIntermediateStates)); | |
2205 | intermediate_state->bs = intermediate; | |
2206 | QSIMPLEQ_INSERT_TAIL(&states_to_delete, intermediate_state, entry); | |
2207 | ||
2208 | if (intermediate->backing_hd == base) { | |
2209 | base_bs = intermediate->backing_hd; | |
2210 | break; | |
2211 | } | |
2212 | intermediate = intermediate->backing_hd; | |
2213 | } | |
2214 | if (base_bs == NULL) { | |
2215 | /* something went wrong, we did not end at the base. safely | |
2216 | * unravel everything, and exit with error */ | |
2217 | goto exit; | |
2218 | } | |
2219 | ||
2220 | /* success - we can delete the intermediate states, and link top->base */ | |
2221 | ret = bdrv_change_backing_file(new_top_bs, base_bs->filename, | |
2222 | base_bs->drv ? base_bs->drv->format_name : ""); | |
2223 | if (ret) { | |
2224 | goto exit; | |
2225 | } | |
2226 | new_top_bs->backing_hd = base_bs; | |
2227 | ||
2228 | ||
2229 | QSIMPLEQ_FOREACH_SAFE(intermediate_state, &states_to_delete, entry, next) { | |
2230 | /* so that bdrv_close() does not recursively close the chain */ | |
2231 | intermediate_state->bs->backing_hd = NULL; | |
4f6fd349 | 2232 | bdrv_unref(intermediate_state->bs); |
6ebdcee2 JC |
2233 | } |
2234 | ret = 0; | |
2235 | ||
2236 | exit: | |
2237 | QSIMPLEQ_FOREACH_SAFE(intermediate_state, &states_to_delete, entry, next) { | |
2238 | g_free(intermediate_state); | |
2239 | } | |
2240 | return ret; | |
2241 | } | |
2242 | ||
2243 | ||
71d0770c AL |
2244 | static int bdrv_check_byte_request(BlockDriverState *bs, int64_t offset, |
2245 | size_t size) | |
2246 | { | |
2247 | int64_t len; | |
2248 | ||
2249 | if (!bdrv_is_inserted(bs)) | |
2250 | return -ENOMEDIUM; | |
2251 | ||
2252 | if (bs->growable) | |
2253 | return 0; | |
2254 | ||
2255 | len = bdrv_getlength(bs); | |
2256 | ||
fbb7b4e0 KW |
2257 | if (offset < 0) |
2258 | return -EIO; | |
2259 | ||
2260 | if ((offset > len) || (len - offset < size)) | |
71d0770c AL |
2261 | return -EIO; |
2262 | ||
2263 | return 0; | |
2264 | } | |
2265 | ||
2266 | static int bdrv_check_request(BlockDriverState *bs, int64_t sector_num, | |
2267 | int nb_sectors) | |
2268 | { | |
eb5a3165 JS |
2269 | return bdrv_check_byte_request(bs, sector_num * BDRV_SECTOR_SIZE, |
2270 | nb_sectors * BDRV_SECTOR_SIZE); | |
71d0770c AL |
2271 | } |
2272 | ||
1c9805a3 SH |
2273 | typedef struct RwCo { |
2274 | BlockDriverState *bs; | |
2275 | int64_t sector_num; | |
2276 | int nb_sectors; | |
2277 | QEMUIOVector *qiov; | |
2278 | bool is_write; | |
2279 | int ret; | |
4105eaaa | 2280 | BdrvRequestFlags flags; |
1c9805a3 SH |
2281 | } RwCo; |
2282 | ||
2283 | static void coroutine_fn bdrv_rw_co_entry(void *opaque) | |
fc01f7e7 | 2284 | { |
1c9805a3 | 2285 | RwCo *rwco = opaque; |
ea2384d3 | 2286 | |
1c9805a3 SH |
2287 | if (!rwco->is_write) { |
2288 | rwco->ret = bdrv_co_do_readv(rwco->bs, rwco->sector_num, | |
4105eaaa PL |
2289 | rwco->nb_sectors, rwco->qiov, |
2290 | rwco->flags); | |
1c9805a3 SH |
2291 | } else { |
2292 | rwco->ret = bdrv_co_do_writev(rwco->bs, rwco->sector_num, | |
4105eaaa PL |
2293 | rwco->nb_sectors, rwco->qiov, |
2294 | rwco->flags); | |
1c9805a3 SH |
2295 | } |
2296 | } | |
e7a8a783 | 2297 | |
1c9805a3 | 2298 | /* |
8d3b1a2d | 2299 | * Process a vectored synchronous request using coroutines |
1c9805a3 | 2300 | */ |
8d3b1a2d | 2301 | static int bdrv_rwv_co(BlockDriverState *bs, int64_t sector_num, |
4105eaaa PL |
2302 | QEMUIOVector *qiov, bool is_write, |
2303 | BdrvRequestFlags flags) | |
1c9805a3 | 2304 | { |
1c9805a3 SH |
2305 | Coroutine *co; |
2306 | RwCo rwco = { | |
2307 | .bs = bs, | |
2308 | .sector_num = sector_num, | |
8d3b1a2d KW |
2309 | .nb_sectors = qiov->size >> BDRV_SECTOR_BITS, |
2310 | .qiov = qiov, | |
1c9805a3 SH |
2311 | .is_write = is_write, |
2312 | .ret = NOT_DONE, | |
4105eaaa | 2313 | .flags = flags, |
1c9805a3 | 2314 | }; |
8d3b1a2d | 2315 | assert((qiov->size & (BDRV_SECTOR_SIZE - 1)) == 0); |
e7a8a783 | 2316 | |
498e386c ZYW |
2317 | /** |
2318 | * In sync call context, when the vcpu is blocked, this throttling timer | |
2319 | * will not fire; so the I/O throttling function has to be disabled here | |
2320 | * if it has been enabled. | |
2321 | */ | |
2322 | if (bs->io_limits_enabled) { | |
2323 | fprintf(stderr, "Disabling I/O throttling on '%s' due " | |
2324 | "to synchronous I/O.\n", bdrv_get_device_name(bs)); | |
2325 | bdrv_io_limits_disable(bs); | |
2326 | } | |
2327 | ||
1c9805a3 SH |
2328 | if (qemu_in_coroutine()) { |
2329 | /* Fast-path if already in coroutine context */ | |
2330 | bdrv_rw_co_entry(&rwco); | |
2331 | } else { | |
2332 | co = qemu_coroutine_create(bdrv_rw_co_entry); | |
2333 | qemu_coroutine_enter(co, &rwco); | |
2334 | while (rwco.ret == NOT_DONE) { | |
2335 | qemu_aio_wait(); | |
2336 | } | |
2337 | } | |
2338 | return rwco.ret; | |
2339 | } | |
b338082b | 2340 | |
8d3b1a2d KW |
2341 | /* |
2342 | * Process a synchronous request using coroutines | |
2343 | */ | |
2344 | static int bdrv_rw_co(BlockDriverState *bs, int64_t sector_num, uint8_t *buf, | |
4105eaaa | 2345 | int nb_sectors, bool is_write, BdrvRequestFlags flags) |
8d3b1a2d KW |
2346 | { |
2347 | QEMUIOVector qiov; | |
2348 | struct iovec iov = { | |
2349 | .iov_base = (void *)buf, | |
2350 | .iov_len = nb_sectors * BDRV_SECTOR_SIZE, | |
2351 | }; | |
2352 | ||
2353 | qemu_iovec_init_external(&qiov, &iov, 1); | |
4105eaaa | 2354 | return bdrv_rwv_co(bs, sector_num, &qiov, is_write, flags); |
8d3b1a2d KW |
2355 | } |
2356 | ||
1c9805a3 SH |
2357 | /* return < 0 if error. See bdrv_write() for the return codes */ |
2358 | int bdrv_read(BlockDriverState *bs, int64_t sector_num, | |
2359 | uint8_t *buf, int nb_sectors) | |
2360 | { | |
4105eaaa | 2361 | return bdrv_rw_co(bs, sector_num, buf, nb_sectors, false, 0); |
fc01f7e7 FB |
2362 | } |
2363 | ||
07d27a44 MA |
2364 | /* Just like bdrv_read(), but with I/O throttling temporarily disabled */ |
2365 | int bdrv_read_unthrottled(BlockDriverState *bs, int64_t sector_num, | |
2366 | uint8_t *buf, int nb_sectors) | |
2367 | { | |
2368 | bool enabled; | |
2369 | int ret; | |
2370 | ||
2371 | enabled = bs->io_limits_enabled; | |
2372 | bs->io_limits_enabled = false; | |
4e7395e8 | 2373 | ret = bdrv_read(bs, sector_num, buf, nb_sectors); |
07d27a44 MA |
2374 | bs->io_limits_enabled = enabled; |
2375 | return ret; | |
2376 | } | |
2377 | ||
5fafdf24 | 2378 | /* Return < 0 if error. Important errors are: |
19cb3738 FB |
2379 | -EIO generic I/O error (may happen for all errors) |
2380 | -ENOMEDIUM No media inserted. | |
2381 | -EINVAL Invalid sector number or nb_sectors | |
2382 | -EACCES Trying to write a read-only device | |
2383 | */ | |
5fafdf24 | 2384 | int bdrv_write(BlockDriverState *bs, int64_t sector_num, |
fc01f7e7 FB |
2385 | const uint8_t *buf, int nb_sectors) |
2386 | { | |
4105eaaa | 2387 | return bdrv_rw_co(bs, sector_num, (uint8_t *)buf, nb_sectors, true, 0); |
83f64091 FB |
2388 | } |
2389 | ||
8d3b1a2d KW |
2390 | int bdrv_writev(BlockDriverState *bs, int64_t sector_num, QEMUIOVector *qiov) |
2391 | { | |
4105eaaa PL |
2392 | return bdrv_rwv_co(bs, sector_num, qiov, true, 0); |
2393 | } | |
2394 | ||
2395 | int bdrv_write_zeroes(BlockDriverState *bs, int64_t sector_num, int nb_sectors) | |
2396 | { | |
2397 | return bdrv_rw_co(bs, sector_num, NULL, nb_sectors, true, | |
2398 | BDRV_REQ_ZERO_WRITE); | |
8d3b1a2d KW |
2399 | } |
2400 | ||
eda578e5 AL |
2401 | int bdrv_pread(BlockDriverState *bs, int64_t offset, |
2402 | void *buf, int count1) | |
83f64091 | 2403 | { |
6ea44308 | 2404 | uint8_t tmp_buf[BDRV_SECTOR_SIZE]; |
83f64091 FB |
2405 | int len, nb_sectors, count; |
2406 | int64_t sector_num; | |
9a8c4cce | 2407 | int ret; |
83f64091 FB |
2408 | |
2409 | count = count1; | |
2410 | /* first read to align to sector start */ | |
6ea44308 | 2411 | len = (BDRV_SECTOR_SIZE - offset) & (BDRV_SECTOR_SIZE - 1); |
83f64091 FB |
2412 | if (len > count) |
2413 | len = count; | |
6ea44308 | 2414 | sector_num = offset >> BDRV_SECTOR_BITS; |
83f64091 | 2415 | if (len > 0) { |
9a8c4cce KW |
2416 | if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0) |
2417 | return ret; | |
6ea44308 | 2418 | memcpy(buf, tmp_buf + (offset & (BDRV_SECTOR_SIZE - 1)), len); |
83f64091 FB |
2419 | count -= len; |
2420 | if (count == 0) | |
2421 | return count1; | |
2422 | sector_num++; | |
2423 | buf += len; | |
2424 | } | |
2425 | ||
2426 | /* read the sectors "in place" */ | |
6ea44308 | 2427 | nb_sectors = count >> BDRV_SECTOR_BITS; |
83f64091 | 2428 | if (nb_sectors > 0) { |
9a8c4cce KW |
2429 | if ((ret = bdrv_read(bs, sector_num, buf, nb_sectors)) < 0) |
2430 | return ret; | |
83f64091 | 2431 | sector_num += nb_sectors; |
6ea44308 | 2432 | len = nb_sectors << BDRV_SECTOR_BITS; |
83f64091 FB |
2433 | buf += len; |
2434 | count -= len; | |
2435 | } | |
2436 | ||
2437 | /* add data from the last sector */ | |
2438 | if (count > 0) { | |
9a8c4cce KW |
2439 | if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0) |
2440 | return ret; | |
83f64091 FB |
2441 | memcpy(buf, tmp_buf, count); |
2442 | } | |
2443 | return count1; | |
2444 | } | |
2445 | ||
8d3b1a2d | 2446 | int bdrv_pwritev(BlockDriverState *bs, int64_t offset, QEMUIOVector *qiov) |
83f64091 | 2447 | { |
6ea44308 | 2448 | uint8_t tmp_buf[BDRV_SECTOR_SIZE]; |
83f64091 FB |
2449 | int len, nb_sectors, count; |
2450 | int64_t sector_num; | |
9a8c4cce | 2451 | int ret; |
83f64091 | 2452 | |
8d3b1a2d KW |
2453 | count = qiov->size; |
2454 | ||
83f64091 | 2455 | /* first write to align to sector start */ |
6ea44308 | 2456 | len = (BDRV_SECTOR_SIZE - offset) & (BDRV_SECTOR_SIZE - 1); |
83f64091 FB |
2457 | if (len > count) |
2458 | len = count; | |
6ea44308 | 2459 | sector_num = offset >> BDRV_SECTOR_BITS; |
83f64091 | 2460 | if (len > 0) { |
9a8c4cce KW |
2461 | if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0) |
2462 | return ret; | |
8d3b1a2d KW |
2463 | qemu_iovec_to_buf(qiov, 0, tmp_buf + (offset & (BDRV_SECTOR_SIZE - 1)), |
2464 | len); | |
9a8c4cce KW |
2465 | if ((ret = bdrv_write(bs, sector_num, tmp_buf, 1)) < 0) |
2466 | return ret; | |
83f64091 FB |
2467 | count -= len; |
2468 | if (count == 0) | |
8d3b1a2d | 2469 | return qiov->size; |
83f64091 | 2470 | sector_num++; |
83f64091 FB |
2471 | } |
2472 | ||
2473 | /* write the sectors "in place" */ | |
6ea44308 | 2474 | nb_sectors = count >> BDRV_SECTOR_BITS; |
83f64091 | 2475 | if (nb_sectors > 0) { |
8d3b1a2d KW |
2476 | QEMUIOVector qiov_inplace; |
2477 | ||
2478 | qemu_iovec_init(&qiov_inplace, qiov->niov); | |
2479 | qemu_iovec_concat(&qiov_inplace, qiov, len, | |
2480 | nb_sectors << BDRV_SECTOR_BITS); | |
2481 | ret = bdrv_writev(bs, sector_num, &qiov_inplace); | |
2482 | qemu_iovec_destroy(&qiov_inplace); | |
2483 | if (ret < 0) { | |
9a8c4cce | 2484 | return ret; |
8d3b1a2d KW |
2485 | } |
2486 | ||
83f64091 | 2487 | sector_num += nb_sectors; |
6ea44308 | 2488 | len = nb_sectors << BDRV_SECTOR_BITS; |
83f64091 FB |
2489 | count -= len; |
2490 | } | |
2491 | ||
2492 | /* add data from the last sector */ | |
2493 | if (count > 0) { | |
9a8c4cce KW |
2494 | if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0) |
2495 | return ret; | |
8d3b1a2d | 2496 | qemu_iovec_to_buf(qiov, qiov->size - count, tmp_buf, count); |
9a8c4cce KW |
2497 | if ((ret = bdrv_write(bs, sector_num, tmp_buf, 1)) < 0) |
2498 | return ret; | |
83f64091 | 2499 | } |
8d3b1a2d KW |
2500 | return qiov->size; |
2501 | } | |
2502 | ||
2503 | int bdrv_pwrite(BlockDriverState *bs, int64_t offset, | |
2504 | const void *buf, int count1) | |
2505 | { | |
2506 | QEMUIOVector qiov; | |
2507 | struct iovec iov = { | |
2508 | .iov_base = (void *) buf, | |
2509 | .iov_len = count1, | |
2510 | }; | |
2511 | ||
2512 | qemu_iovec_init_external(&qiov, &iov, 1); | |
2513 | return bdrv_pwritev(bs, offset, &qiov); | |
83f64091 | 2514 | } |
83f64091 | 2515 | |
f08145fe KW |
2516 | /* |
2517 | * Writes to the file and ensures that no writes are reordered across this | |
2518 | * request (acts as a barrier) | |
2519 | * | |
2520 | * Returns 0 on success, -errno in error cases. | |
2521 | */ | |
2522 | int bdrv_pwrite_sync(BlockDriverState *bs, int64_t offset, | |
2523 | const void *buf, int count) | |
2524 | { | |
2525 | int ret; | |
2526 | ||
2527 | ret = bdrv_pwrite(bs, offset, buf, count); | |
2528 | if (ret < 0) { | |
2529 | return ret; | |
2530 | } | |
2531 | ||
f05fa4ad PB |
2532 | /* No flush needed for cache modes that already do it */ |
2533 | if (bs->enable_write_cache) { | |
f08145fe KW |
2534 | bdrv_flush(bs); |
2535 | } | |
2536 | ||
2537 | return 0; | |
2538 | } | |
2539 | ||
470c0504 | 2540 | static int coroutine_fn bdrv_co_do_copy_on_readv(BlockDriverState *bs, |
ab185921 SH |
2541 | int64_t sector_num, int nb_sectors, QEMUIOVector *qiov) |
2542 | { | |
2543 | /* Perform I/O through a temporary buffer so that users who scribble over | |
2544 | * their read buffer while the operation is in progress do not end up | |
2545 | * modifying the image file. This is critical for zero-copy guest I/O | |
2546 | * where anything might happen inside guest memory. | |
2547 | */ | |
2548 | void *bounce_buffer; | |
2549 | ||
79c053bd | 2550 | BlockDriver *drv = bs->drv; |
ab185921 SH |
2551 | struct iovec iov; |
2552 | QEMUIOVector bounce_qiov; | |
2553 | int64_t cluster_sector_num; | |
2554 | int cluster_nb_sectors; | |
2555 | size_t skip_bytes; | |
2556 | int ret; | |
2557 | ||
2558 | /* Cover entire cluster so no additional backing file I/O is required when | |
2559 | * allocating cluster in the image file. | |
2560 | */ | |
343bded4 PB |
2561 | bdrv_round_to_clusters(bs, sector_num, nb_sectors, |
2562 | &cluster_sector_num, &cluster_nb_sectors); | |
ab185921 | 2563 | |
470c0504 SH |
2564 | trace_bdrv_co_do_copy_on_readv(bs, sector_num, nb_sectors, |
2565 | cluster_sector_num, cluster_nb_sectors); | |
ab185921 SH |
2566 | |
2567 | iov.iov_len = cluster_nb_sectors * BDRV_SECTOR_SIZE; | |
2568 | iov.iov_base = bounce_buffer = qemu_blockalign(bs, iov.iov_len); | |
2569 | qemu_iovec_init_external(&bounce_qiov, &iov, 1); | |
2570 | ||
79c053bd SH |
2571 | ret = drv->bdrv_co_readv(bs, cluster_sector_num, cluster_nb_sectors, |
2572 | &bounce_qiov); | |
ab185921 SH |
2573 | if (ret < 0) { |
2574 | goto err; | |
2575 | } | |
2576 | ||
79c053bd SH |
2577 | if (drv->bdrv_co_write_zeroes && |
2578 | buffer_is_zero(bounce_buffer, iov.iov_len)) { | |
621f0589 KW |
2579 | ret = bdrv_co_do_write_zeroes(bs, cluster_sector_num, |
2580 | cluster_nb_sectors); | |
79c053bd | 2581 | } else { |
f05fa4ad PB |
2582 | /* This does not change the data on the disk, it is not necessary |
2583 | * to flush even in cache=writethrough mode. | |
2584 | */ | |
79c053bd | 2585 | ret = drv->bdrv_co_writev(bs, cluster_sector_num, cluster_nb_sectors, |
ab185921 | 2586 | &bounce_qiov); |
79c053bd SH |
2587 | } |
2588 | ||
ab185921 SH |
2589 | if (ret < 0) { |
2590 | /* It might be okay to ignore write errors for guest requests. If this | |
2591 | * is a deliberate copy-on-read then we don't want to ignore the error. | |
2592 | * Simply report it in all cases. | |
2593 | */ | |
2594 | goto err; | |
2595 | } | |
2596 | ||
2597 | skip_bytes = (sector_num - cluster_sector_num) * BDRV_SECTOR_SIZE; | |
03396148 MT |
2598 | qemu_iovec_from_buf(qiov, 0, bounce_buffer + skip_bytes, |
2599 | nb_sectors * BDRV_SECTOR_SIZE); | |
ab185921 SH |
2600 | |
2601 | err: | |
2602 | qemu_vfree(bounce_buffer); | |
2603 | return ret; | |
2604 | } | |
2605 | ||
c5fbe571 SH |
2606 | /* |
2607 | * Handle a read request in coroutine context | |
2608 | */ | |
2609 | static int coroutine_fn bdrv_co_do_readv(BlockDriverState *bs, | |
470c0504 SH |
2610 | int64_t sector_num, int nb_sectors, QEMUIOVector *qiov, |
2611 | BdrvRequestFlags flags) | |
da1fa91d KW |
2612 | { |
2613 | BlockDriver *drv = bs->drv; | |
dbffbdcf SH |
2614 | BdrvTrackedRequest req; |
2615 | int ret; | |
da1fa91d | 2616 | |
da1fa91d KW |
2617 | if (!drv) { |
2618 | return -ENOMEDIUM; | |
2619 | } | |
2620 | if (bdrv_check_request(bs, sector_num, nb_sectors)) { | |
2621 | return -EIO; | |
2622 | } | |
2623 | ||
f4658285 | 2624 | if (bs->copy_on_read) { |
470c0504 SH |
2625 | flags |= BDRV_REQ_COPY_ON_READ; |
2626 | } | |
2627 | if (flags & BDRV_REQ_COPY_ON_READ) { | |
2628 | bs->copy_on_read_in_flight++; | |
2629 | } | |
2630 | ||
2631 | if (bs->copy_on_read_in_flight) { | |
f4658285 SH |
2632 | wait_for_overlapping_requests(bs, sector_num, nb_sectors); |
2633 | } | |
2634 | ||
cc0681c4 BC |
2635 | /* throttling disk I/O */ |
2636 | if (bs->io_limits_enabled) { | |
2637 | bdrv_io_limits_intercept(bs, nb_sectors, false); | |
2638 | } | |
2639 | ||
dbffbdcf | 2640 | tracked_request_begin(&req, bs, sector_num, nb_sectors, false); |
ab185921 | 2641 | |
470c0504 | 2642 | if (flags & BDRV_REQ_COPY_ON_READ) { |
ab185921 SH |
2643 | int pnum; |
2644 | ||
bdad13b9 | 2645 | ret = bdrv_is_allocated(bs, sector_num, nb_sectors, &pnum); |
ab185921 SH |
2646 | if (ret < 0) { |
2647 | goto out; | |
2648 | } | |
2649 | ||
2650 | if (!ret || pnum != nb_sectors) { | |
470c0504 | 2651 | ret = bdrv_co_do_copy_on_readv(bs, sector_num, nb_sectors, qiov); |
ab185921 SH |
2652 | goto out; |
2653 | } | |
2654 | } | |
2655 | ||
893a8f62 MK |
2656 | if (!(bs->zero_beyond_eof && bs->growable)) { |
2657 | ret = drv->bdrv_co_readv(bs, sector_num, nb_sectors, qiov); | |
2658 | } else { | |
2659 | /* Read zeros after EOF of growable BDSes */ | |
2660 | int64_t len, total_sectors, max_nb_sectors; | |
2661 | ||
2662 | len = bdrv_getlength(bs); | |
2663 | if (len < 0) { | |
2664 | ret = len; | |
2665 | goto out; | |
2666 | } | |
2667 | ||
d055a1fe | 2668 | total_sectors = DIV_ROUND_UP(len, BDRV_SECTOR_SIZE); |
893a8f62 MK |
2669 | max_nb_sectors = MAX(0, total_sectors - sector_num); |
2670 | if (max_nb_sectors > 0) { | |
2671 | ret = drv->bdrv_co_readv(bs, sector_num, | |
2672 | MIN(nb_sectors, max_nb_sectors), qiov); | |
2673 | } else { | |
2674 | ret = 0; | |
2675 | } | |
2676 | ||
2677 | /* Reading beyond end of file is supposed to produce zeroes */ | |
2678 | if (ret == 0 && total_sectors < sector_num + nb_sectors) { | |
2679 | uint64_t offset = MAX(0, total_sectors - sector_num); | |
2680 | uint64_t bytes = (sector_num + nb_sectors - offset) * | |
2681 | BDRV_SECTOR_SIZE; | |
2682 | qemu_iovec_memset(qiov, offset * BDRV_SECTOR_SIZE, 0, bytes); | |
2683 | } | |
2684 | } | |
ab185921 SH |
2685 | |
2686 | out: | |
dbffbdcf | 2687 | tracked_request_end(&req); |
470c0504 SH |
2688 | |
2689 | if (flags & BDRV_REQ_COPY_ON_READ) { | |
2690 | bs->copy_on_read_in_flight--; | |
2691 | } | |
2692 | ||
dbffbdcf | 2693 | return ret; |
da1fa91d KW |
2694 | } |
2695 | ||
c5fbe571 | 2696 | int coroutine_fn bdrv_co_readv(BlockDriverState *bs, int64_t sector_num, |
da1fa91d KW |
2697 | int nb_sectors, QEMUIOVector *qiov) |
2698 | { | |
c5fbe571 | 2699 | trace_bdrv_co_readv(bs, sector_num, nb_sectors); |
da1fa91d | 2700 | |
470c0504 SH |
2701 | return bdrv_co_do_readv(bs, sector_num, nb_sectors, qiov, 0); |
2702 | } | |
2703 | ||
2704 | int coroutine_fn bdrv_co_copy_on_readv(BlockDriverState *bs, | |
2705 | int64_t sector_num, int nb_sectors, QEMUIOVector *qiov) | |
2706 | { | |
2707 | trace_bdrv_co_copy_on_readv(bs, sector_num, nb_sectors); | |
2708 | ||
2709 | return bdrv_co_do_readv(bs, sector_num, nb_sectors, qiov, | |
2710 | BDRV_REQ_COPY_ON_READ); | |
c5fbe571 SH |
2711 | } |
2712 | ||
f08f2dda SH |
2713 | static int coroutine_fn bdrv_co_do_write_zeroes(BlockDriverState *bs, |
2714 | int64_t sector_num, int nb_sectors) | |
2715 | { | |
2716 | BlockDriver *drv = bs->drv; | |
2717 | QEMUIOVector qiov; | |
2718 | struct iovec iov; | |
2719 | int ret; | |
2720 | ||
621f0589 KW |
2721 | /* TODO Emulate only part of misaligned requests instead of letting block |
2722 | * drivers return -ENOTSUP and emulate everything */ | |
2723 | ||
f08f2dda SH |
2724 | /* First try the efficient write zeroes operation */ |
2725 | if (drv->bdrv_co_write_zeroes) { | |
621f0589 KW |
2726 | ret = drv->bdrv_co_write_zeroes(bs, sector_num, nb_sectors); |
2727 | if (ret != -ENOTSUP) { | |
2728 | return ret; | |
2729 | } | |
f08f2dda SH |
2730 | } |
2731 | ||
2732 | /* Fall back to bounce buffer if write zeroes is unsupported */ | |
2733 | iov.iov_len = nb_sectors * BDRV_SECTOR_SIZE; | |
2734 | iov.iov_base = qemu_blockalign(bs, iov.iov_len); | |
2735 | memset(iov.iov_base, 0, iov.iov_len); | |
2736 | qemu_iovec_init_external(&qiov, &iov, 1); | |
2737 | ||
2738 | ret = drv->bdrv_co_writev(bs, sector_num, nb_sectors, &qiov); | |
2739 | ||
2740 | qemu_vfree(iov.iov_base); | |
2741 | return ret; | |
2742 | } | |
2743 | ||
c5fbe571 SH |
2744 | /* |
2745 | * Handle a write request in coroutine context | |
2746 | */ | |
2747 | static int coroutine_fn bdrv_co_do_writev(BlockDriverState *bs, | |
f08f2dda SH |
2748 | int64_t sector_num, int nb_sectors, QEMUIOVector *qiov, |
2749 | BdrvRequestFlags flags) | |
c5fbe571 SH |
2750 | { |
2751 | BlockDriver *drv = bs->drv; | |
dbffbdcf | 2752 | BdrvTrackedRequest req; |
6b7cb247 | 2753 | int ret; |
da1fa91d KW |
2754 | |
2755 | if (!bs->drv) { | |
2756 | return -ENOMEDIUM; | |
2757 | } | |
2758 | if (bs->read_only) { | |
2759 | return -EACCES; | |
2760 | } | |
2761 | if (bdrv_check_request(bs, sector_num, nb_sectors)) { | |
2762 | return -EIO; | |
2763 | } | |
2764 | ||
470c0504 | 2765 | if (bs->copy_on_read_in_flight) { |
f4658285 SH |
2766 | wait_for_overlapping_requests(bs, sector_num, nb_sectors); |
2767 | } | |
2768 | ||
cc0681c4 BC |
2769 | /* throttling disk I/O */ |
2770 | if (bs->io_limits_enabled) { | |
2771 | bdrv_io_limits_intercept(bs, nb_sectors, true); | |
2772 | } | |
2773 | ||
dbffbdcf SH |
2774 | tracked_request_begin(&req, bs, sector_num, nb_sectors, true); |
2775 | ||
d616b224 SH |
2776 | ret = notifier_with_return_list_notify(&bs->before_write_notifiers, &req); |
2777 | ||
2778 | if (ret < 0) { | |
2779 | /* Do nothing, write notifier decided to fail this request */ | |
2780 | } else if (flags & BDRV_REQ_ZERO_WRITE) { | |
f08f2dda SH |
2781 | ret = bdrv_co_do_write_zeroes(bs, sector_num, nb_sectors); |
2782 | } else { | |
2783 | ret = drv->bdrv_co_writev(bs, sector_num, nb_sectors, qiov); | |
2784 | } | |
6b7cb247 | 2785 | |
f05fa4ad PB |
2786 | if (ret == 0 && !bs->enable_write_cache) { |
2787 | ret = bdrv_co_flush(bs); | |
2788 | } | |
2789 | ||
da1fa91d | 2790 | if (bs->dirty_bitmap) { |
1755da16 | 2791 | bdrv_set_dirty(bs, sector_num, nb_sectors); |
da1fa91d KW |
2792 | } |
2793 | ||
2794 | if (bs->wr_highest_sector < sector_num + nb_sectors - 1) { | |
2795 | bs->wr_highest_sector = sector_num + nb_sectors - 1; | |
2796 | } | |
df2a6f29 PB |
2797 | if (bs->growable && ret >= 0) { |
2798 | bs->total_sectors = MAX(bs->total_sectors, sector_num + nb_sectors); | |
2799 | } | |
da1fa91d | 2800 | |
dbffbdcf SH |
2801 | tracked_request_end(&req); |
2802 | ||
6b7cb247 | 2803 | return ret; |
da1fa91d KW |
2804 | } |
2805 | ||
c5fbe571 SH |
2806 | int coroutine_fn bdrv_co_writev(BlockDriverState *bs, int64_t sector_num, |
2807 | int nb_sectors, QEMUIOVector *qiov) | |
2808 | { | |
2809 | trace_bdrv_co_writev(bs, sector_num, nb_sectors); | |
2810 | ||
f08f2dda SH |
2811 | return bdrv_co_do_writev(bs, sector_num, nb_sectors, qiov, 0); |
2812 | } | |
2813 | ||
2814 | int coroutine_fn bdrv_co_write_zeroes(BlockDriverState *bs, | |
2815 | int64_t sector_num, int nb_sectors) | |
2816 | { | |
2817 | trace_bdrv_co_write_zeroes(bs, sector_num, nb_sectors); | |
2818 | ||
2819 | return bdrv_co_do_writev(bs, sector_num, nb_sectors, NULL, | |
2820 | BDRV_REQ_ZERO_WRITE); | |
c5fbe571 SH |
2821 | } |
2822 | ||
83f64091 FB |
2823 | /** |
2824 | * Truncate file to 'offset' bytes (needed only for file protocols) | |
2825 | */ | |
2826 | int bdrv_truncate(BlockDriverState *bs, int64_t offset) | |
2827 | { | |
2828 | BlockDriver *drv = bs->drv; | |
51762288 | 2829 | int ret; |
83f64091 | 2830 | if (!drv) |
19cb3738 | 2831 | return -ENOMEDIUM; |
83f64091 FB |
2832 | if (!drv->bdrv_truncate) |
2833 | return -ENOTSUP; | |
59f2689d NS |
2834 | if (bs->read_only) |
2835 | return -EACCES; | |
8591675f MT |
2836 | if (bdrv_in_use(bs)) |
2837 | return -EBUSY; | |
51762288 SH |
2838 | ret = drv->bdrv_truncate(bs, offset); |
2839 | if (ret == 0) { | |
2840 | ret = refresh_total_sectors(bs, offset >> BDRV_SECTOR_BITS); | |
145feb17 | 2841 | bdrv_dev_resize_cb(bs); |
51762288 SH |
2842 | } |
2843 | return ret; | |
83f64091 FB |
2844 | } |
2845 | ||
4a1d5e1f FZ |
2846 | /** |
2847 | * Length of a allocated file in bytes. Sparse files are counted by actual | |
2848 | * allocated space. Return < 0 if error or unknown. | |
2849 | */ | |
2850 | int64_t bdrv_get_allocated_file_size(BlockDriverState *bs) | |
2851 | { | |
2852 | BlockDriver *drv = bs->drv; | |
2853 | if (!drv) { | |
2854 | return -ENOMEDIUM; | |
2855 | } | |
2856 | if (drv->bdrv_get_allocated_file_size) { | |
2857 | return drv->bdrv_get_allocated_file_size(bs); | |
2858 | } | |
2859 | if (bs->file) { | |
2860 | return bdrv_get_allocated_file_size(bs->file); | |
2861 | } | |
2862 | return -ENOTSUP; | |
2863 | } | |
2864 | ||
83f64091 FB |
2865 | /** |
2866 | * Length of a file in bytes. Return < 0 if error or unknown. | |
2867 | */ | |
2868 | int64_t bdrv_getlength(BlockDriverState *bs) | |
2869 | { | |
2870 | BlockDriver *drv = bs->drv; | |
2871 | if (!drv) | |
19cb3738 | 2872 | return -ENOMEDIUM; |
51762288 | 2873 | |
b94a2610 KW |
2874 | if (drv->has_variable_length) { |
2875 | int ret = refresh_total_sectors(bs, bs->total_sectors); | |
2876 | if (ret < 0) { | |
2877 | return ret; | |
46a4e4e6 | 2878 | } |
83f64091 | 2879 | } |
46a4e4e6 | 2880 | return bs->total_sectors * BDRV_SECTOR_SIZE; |
fc01f7e7 FB |
2881 | } |
2882 | ||
19cb3738 | 2883 | /* return 0 as number of sectors if no device present or error */ |
96b8f136 | 2884 | void bdrv_get_geometry(BlockDriverState *bs, uint64_t *nb_sectors_ptr) |
fc01f7e7 | 2885 | { |
19cb3738 FB |
2886 | int64_t length; |
2887 | length = bdrv_getlength(bs); | |
2888 | if (length < 0) | |
2889 | length = 0; | |
2890 | else | |
6ea44308 | 2891 | length = length >> BDRV_SECTOR_BITS; |
19cb3738 | 2892 | *nb_sectors_ptr = length; |
fc01f7e7 | 2893 | } |
cf98951b | 2894 | |
ff06f5f3 PB |
2895 | void bdrv_set_on_error(BlockDriverState *bs, BlockdevOnError on_read_error, |
2896 | BlockdevOnError on_write_error) | |
abd7f68d MA |
2897 | { |
2898 | bs->on_read_error = on_read_error; | |
2899 | bs->on_write_error = on_write_error; | |
2900 | } | |
2901 | ||
1ceee0d5 | 2902 | BlockdevOnError bdrv_get_on_error(BlockDriverState *bs, bool is_read) |
abd7f68d MA |
2903 | { |
2904 | return is_read ? bs->on_read_error : bs->on_write_error; | |
2905 | } | |
2906 | ||
3e1caa5f PB |
2907 | BlockErrorAction bdrv_get_error_action(BlockDriverState *bs, bool is_read, int error) |
2908 | { | |
2909 | BlockdevOnError on_err = is_read ? bs->on_read_error : bs->on_write_error; | |
2910 | ||
2911 | switch (on_err) { | |
2912 | case BLOCKDEV_ON_ERROR_ENOSPC: | |
2913 | return (error == ENOSPC) ? BDRV_ACTION_STOP : BDRV_ACTION_REPORT; | |
2914 | case BLOCKDEV_ON_ERROR_STOP: | |
2915 | return BDRV_ACTION_STOP; | |
2916 | case BLOCKDEV_ON_ERROR_REPORT: | |
2917 | return BDRV_ACTION_REPORT; | |
2918 | case BLOCKDEV_ON_ERROR_IGNORE: | |
2919 | return BDRV_ACTION_IGNORE; | |
2920 | default: | |
2921 | abort(); | |
2922 | } | |
2923 | } | |
2924 | ||
2925 | /* This is done by device models because, while the block layer knows | |
2926 | * about the error, it does not know whether an operation comes from | |
2927 | * the device or the block layer (from a job, for example). | |
2928 | */ | |
2929 | void bdrv_error_action(BlockDriverState *bs, BlockErrorAction action, | |
2930 | bool is_read, int error) | |
2931 | { | |
2932 | assert(error >= 0); | |
32c81a4a | 2933 | bdrv_emit_qmp_error_event(bs, QEVENT_BLOCK_IO_ERROR, action, is_read); |
3e1caa5f PB |
2934 | if (action == BDRV_ACTION_STOP) { |
2935 | vm_stop(RUN_STATE_IO_ERROR); | |
2936 | bdrv_iostatus_set_err(bs, error); | |
2937 | } | |
2938 | } | |
2939 | ||
b338082b FB |
2940 | int bdrv_is_read_only(BlockDriverState *bs) |
2941 | { | |
2942 | return bs->read_only; | |
2943 | } | |
2944 | ||
985a03b0 TS |
2945 | int bdrv_is_sg(BlockDriverState *bs) |
2946 | { | |
2947 | return bs->sg; | |
2948 | } | |
2949 | ||
e900a7b7 CH |
2950 | int bdrv_enable_write_cache(BlockDriverState *bs) |
2951 | { | |
2952 | return bs->enable_write_cache; | |
2953 | } | |
2954 | ||
425b0148 PB |
2955 | void bdrv_set_enable_write_cache(BlockDriverState *bs, bool wce) |
2956 | { | |
2957 | bs->enable_write_cache = wce; | |
55b110f2 JC |
2958 | |
2959 | /* so a reopen() will preserve wce */ | |
2960 | if (wce) { | |
2961 | bs->open_flags |= BDRV_O_CACHE_WB; | |
2962 | } else { | |
2963 | bs->open_flags &= ~BDRV_O_CACHE_WB; | |
2964 | } | |
425b0148 PB |
2965 | } |
2966 | ||
ea2384d3 FB |
2967 | int bdrv_is_encrypted(BlockDriverState *bs) |
2968 | { | |
2969 | if (bs->backing_hd && bs->backing_hd->encrypted) | |
2970 | return 1; | |
2971 | return bs->encrypted; | |
2972 | } | |
2973 | ||
c0f4ce77 AL |
2974 | int bdrv_key_required(BlockDriverState *bs) |
2975 | { | |
2976 | BlockDriverState *backing_hd = bs->backing_hd; | |
2977 | ||
2978 | if (backing_hd && backing_hd->encrypted && !backing_hd->valid_key) | |
2979 | return 1; | |
2980 | return (bs->encrypted && !bs->valid_key); | |
2981 | } | |
2982 | ||
ea2384d3 FB |
2983 | int bdrv_set_key(BlockDriverState *bs, const char *key) |
2984 | { | |
2985 | int ret; | |
2986 | if (bs->backing_hd && bs->backing_hd->encrypted) { | |
2987 | ret = bdrv_set_key(bs->backing_hd, key); | |
2988 | if (ret < 0) | |
2989 | return ret; | |
2990 | if (!bs->encrypted) | |
2991 | return 0; | |
2992 | } | |
fd04a2ae SH |
2993 | if (!bs->encrypted) { |
2994 | return -EINVAL; | |
2995 | } else if (!bs->drv || !bs->drv->bdrv_set_key) { | |
2996 | return -ENOMEDIUM; | |
2997 | } | |
c0f4ce77 | 2998 | ret = bs->drv->bdrv_set_key(bs, key); |
bb5fc20f AL |
2999 | if (ret < 0) { |
3000 | bs->valid_key = 0; | |
3001 | } else if (!bs->valid_key) { | |
3002 | bs->valid_key = 1; | |
3003 | /* call the change callback now, we skipped it on open */ | |
7d4b4ba5 | 3004 | bdrv_dev_change_media_cb(bs, true); |
bb5fc20f | 3005 | } |
c0f4ce77 | 3006 | return ret; |
ea2384d3 FB |
3007 | } |
3008 | ||
f8d6bba1 | 3009 | const char *bdrv_get_format_name(BlockDriverState *bs) |
ea2384d3 | 3010 | { |
f8d6bba1 | 3011 | return bs->drv ? bs->drv->format_name : NULL; |
ea2384d3 FB |
3012 | } |
3013 | ||
5fafdf24 | 3014 | void bdrv_iterate_format(void (*it)(void *opaque, const char *name), |
ea2384d3 FB |
3015 | void *opaque) |
3016 | { | |
3017 | BlockDriver *drv; | |
3018 | ||
8a22f02a | 3019 | QLIST_FOREACH(drv, &bdrv_drivers, list) { |
ea2384d3 FB |
3020 | it(opaque, drv->format_name); |
3021 | } | |
3022 | } | |
3023 | ||
b338082b FB |
3024 | BlockDriverState *bdrv_find(const char *name) |
3025 | { | |
3026 | BlockDriverState *bs; | |
3027 | ||
1b7bdbc1 SH |
3028 | QTAILQ_FOREACH(bs, &bdrv_states, list) { |
3029 | if (!strcmp(name, bs->device_name)) { | |
b338082b | 3030 | return bs; |
1b7bdbc1 | 3031 | } |
b338082b FB |
3032 | } |
3033 | return NULL; | |
3034 | } | |
3035 | ||
2f399b0a MA |
3036 | BlockDriverState *bdrv_next(BlockDriverState *bs) |
3037 | { | |
3038 | if (!bs) { | |
3039 | return QTAILQ_FIRST(&bdrv_states); | |
3040 | } | |
3041 | return QTAILQ_NEXT(bs, list); | |
3042 | } | |
3043 | ||
51de9760 | 3044 | void bdrv_iterate(void (*it)(void *opaque, BlockDriverState *bs), void *opaque) |
81d0912d FB |
3045 | { |
3046 | BlockDriverState *bs; | |
3047 | ||
1b7bdbc1 | 3048 | QTAILQ_FOREACH(bs, &bdrv_states, list) { |
51de9760 | 3049 | it(opaque, bs); |
81d0912d FB |
3050 | } |
3051 | } | |
3052 | ||
ea2384d3 FB |
3053 | const char *bdrv_get_device_name(BlockDriverState *bs) |
3054 | { | |
3055 | return bs->device_name; | |
3056 | } | |
3057 | ||
c8433287 MA |
3058 | int bdrv_get_flags(BlockDriverState *bs) |
3059 | { | |
3060 | return bs->open_flags; | |
3061 | } | |
3062 | ||
f0f0fdfe | 3063 | int bdrv_flush_all(void) |
c6ca28d6 AL |
3064 | { |
3065 | BlockDriverState *bs; | |
f0f0fdfe | 3066 | int result = 0; |
c6ca28d6 | 3067 | |
1b7bdbc1 | 3068 | QTAILQ_FOREACH(bs, &bdrv_states, list) { |
f0f0fdfe KW |
3069 | int ret = bdrv_flush(bs); |
3070 | if (ret < 0 && !result) { | |
3071 | result = ret; | |
3072 | } | |
1b7bdbc1 | 3073 | } |
f0f0fdfe KW |
3074 | |
3075 | return result; | |
c6ca28d6 AL |
3076 | } |
3077 | ||
3ac21627 PL |
3078 | int bdrv_has_zero_init_1(BlockDriverState *bs) |
3079 | { | |
3080 | return 1; | |
3081 | } | |
3082 | ||
f2feebbd KW |
3083 | int bdrv_has_zero_init(BlockDriverState *bs) |
3084 | { | |
3085 | assert(bs->drv); | |
3086 | ||
11212d8f PB |
3087 | /* If BS is a copy on write image, it is initialized to |
3088 | the contents of the base image, which may not be zeroes. */ | |
3089 | if (bs->backing_hd) { | |
3090 | return 0; | |
3091 | } | |
336c1c12 KW |
3092 | if (bs->drv->bdrv_has_zero_init) { |
3093 | return bs->drv->bdrv_has_zero_init(bs); | |
f2feebbd KW |
3094 | } |
3095 | ||
3ac21627 PL |
3096 | /* safe default */ |
3097 | return 0; | |
f2feebbd KW |
3098 | } |
3099 | ||
b6b8a333 | 3100 | typedef struct BdrvCoGetBlockStatusData { |
376ae3f1 | 3101 | BlockDriverState *bs; |
b35b2bba | 3102 | BlockDriverState *base; |
376ae3f1 SH |
3103 | int64_t sector_num; |
3104 | int nb_sectors; | |
3105 | int *pnum; | |
b6b8a333 | 3106 | int64_t ret; |
376ae3f1 | 3107 | bool done; |
b6b8a333 | 3108 | } BdrvCoGetBlockStatusData; |
376ae3f1 | 3109 | |
f58c7b35 TS |
3110 | /* |
3111 | * Returns true iff the specified sector is present in the disk image. Drivers | |
3112 | * not implementing the functionality are assumed to not support backing files, | |
3113 | * hence all their sectors are reported as allocated. | |
3114 | * | |
bd9533e3 SH |
3115 | * If 'sector_num' is beyond the end of the disk image the return value is 0 |
3116 | * and 'pnum' is set to 0. | |
3117 | * | |
f58c7b35 TS |
3118 | * 'pnum' is set to the number of sectors (including and immediately following |
3119 | * the specified sector) that are known to be in the same | |
3120 | * allocated/unallocated state. | |
3121 | * | |
bd9533e3 SH |
3122 | * 'nb_sectors' is the max value 'pnum' should be set to. If nb_sectors goes |
3123 | * beyond the end of the disk image it will be clamped. | |
f58c7b35 | 3124 | */ |
b6b8a333 PB |
3125 | static int64_t coroutine_fn bdrv_co_get_block_status(BlockDriverState *bs, |
3126 | int64_t sector_num, | |
3127 | int nb_sectors, int *pnum) | |
f58c7b35 | 3128 | { |
617ccb46 | 3129 | int64_t length; |
bd9533e3 | 3130 | int64_t n; |
5daa74a6 | 3131 | int64_t ret, ret2; |
bd9533e3 | 3132 | |
617ccb46 PB |
3133 | length = bdrv_getlength(bs); |
3134 | if (length < 0) { | |
3135 | return length; | |
3136 | } | |
3137 | ||
3138 | if (sector_num >= (length >> BDRV_SECTOR_BITS)) { | |
bd9533e3 SH |
3139 | *pnum = 0; |
3140 | return 0; | |
3141 | } | |
3142 | ||
3143 | n = bs->total_sectors - sector_num; | |
3144 | if (n < nb_sectors) { | |
3145 | nb_sectors = n; | |
3146 | } | |
3147 | ||
b6b8a333 | 3148 | if (!bs->drv->bdrv_co_get_block_status) { |
bd9533e3 | 3149 | *pnum = nb_sectors; |
918e92d7 PB |
3150 | ret = BDRV_BLOCK_DATA; |
3151 | if (bs->drv->protocol_name) { | |
3152 | ret |= BDRV_BLOCK_OFFSET_VALID | (sector_num * BDRV_SECTOR_SIZE); | |
3153 | } | |
3154 | return ret; | |
f58c7b35 | 3155 | } |
6aebab14 | 3156 | |
415b5b01 PB |
3157 | ret = bs->drv->bdrv_co_get_block_status(bs, sector_num, nb_sectors, pnum); |
3158 | if (ret < 0) { | |
3e0a233d | 3159 | *pnum = 0; |
415b5b01 PB |
3160 | return ret; |
3161 | } | |
3162 | ||
92bc50a5 PL |
3163 | if (ret & BDRV_BLOCK_RAW) { |
3164 | assert(ret & BDRV_BLOCK_OFFSET_VALID); | |
3165 | return bdrv_get_block_status(bs->file, ret >> BDRV_SECTOR_BITS, | |
3166 | *pnum, pnum); | |
3167 | } | |
3168 | ||
f0ad5712 PB |
3169 | if (!(ret & BDRV_BLOCK_DATA)) { |
3170 | if (bdrv_has_zero_init(bs)) { | |
3171 | ret |= BDRV_BLOCK_ZERO; | |
1f9db224 | 3172 | } else if (bs->backing_hd) { |
f0ad5712 PB |
3173 | BlockDriverState *bs2 = bs->backing_hd; |
3174 | int64_t length2 = bdrv_getlength(bs2); | |
3175 | if (length2 >= 0 && sector_num >= (length2 >> BDRV_SECTOR_BITS)) { | |
3176 | ret |= BDRV_BLOCK_ZERO; | |
3177 | } | |
3178 | } | |
415b5b01 | 3179 | } |
5daa74a6 PB |
3180 | |
3181 | if (bs->file && | |
3182 | (ret & BDRV_BLOCK_DATA) && !(ret & BDRV_BLOCK_ZERO) && | |
3183 | (ret & BDRV_BLOCK_OFFSET_VALID)) { | |
3184 | ret2 = bdrv_co_get_block_status(bs->file, ret >> BDRV_SECTOR_BITS, | |
3185 | *pnum, pnum); | |
3186 | if (ret2 >= 0) { | |
3187 | /* Ignore errors. This is just providing extra information, it | |
3188 | * is useful but not necessary. | |
3189 | */ | |
3190 | ret |= (ret2 & BDRV_BLOCK_ZERO); | |
3191 | } | |
3192 | } | |
3193 | ||
415b5b01 | 3194 | return ret; |
060f51c9 SH |
3195 | } |
3196 | ||
b6b8a333 PB |
3197 | /* Coroutine wrapper for bdrv_get_block_status() */ |
3198 | static void coroutine_fn bdrv_get_block_status_co_entry(void *opaque) | |
060f51c9 | 3199 | { |
b6b8a333 | 3200 | BdrvCoGetBlockStatusData *data = opaque; |
060f51c9 SH |
3201 | BlockDriverState *bs = data->bs; |
3202 | ||
b6b8a333 PB |
3203 | data->ret = bdrv_co_get_block_status(bs, data->sector_num, data->nb_sectors, |
3204 | data->pnum); | |
060f51c9 SH |
3205 | data->done = true; |
3206 | } | |
3207 | ||
3208 | /* | |
b6b8a333 | 3209 | * Synchronous wrapper around bdrv_co_get_block_status(). |
060f51c9 | 3210 | * |
b6b8a333 | 3211 | * See bdrv_co_get_block_status() for details. |
060f51c9 | 3212 | */ |
b6b8a333 PB |
3213 | int64_t bdrv_get_block_status(BlockDriverState *bs, int64_t sector_num, |
3214 | int nb_sectors, int *pnum) | |
060f51c9 | 3215 | { |
6aebab14 | 3216 | Coroutine *co; |
b6b8a333 | 3217 | BdrvCoGetBlockStatusData data = { |
6aebab14 SH |
3218 | .bs = bs, |
3219 | .sector_num = sector_num, | |
3220 | .nb_sectors = nb_sectors, | |
3221 | .pnum = pnum, | |
3222 | .done = false, | |
3223 | }; | |
3224 | ||
bdad13b9 PB |
3225 | if (qemu_in_coroutine()) { |
3226 | /* Fast-path if already in coroutine context */ | |
b6b8a333 | 3227 | bdrv_get_block_status_co_entry(&data); |
bdad13b9 | 3228 | } else { |
b6b8a333 | 3229 | co = qemu_coroutine_create(bdrv_get_block_status_co_entry); |
bdad13b9 PB |
3230 | qemu_coroutine_enter(co, &data); |
3231 | while (!data.done) { | |
3232 | qemu_aio_wait(); | |
3233 | } | |
6aebab14 SH |
3234 | } |
3235 | return data.ret; | |
f58c7b35 TS |
3236 | } |
3237 | ||
b6b8a333 PB |
3238 | int coroutine_fn bdrv_is_allocated(BlockDriverState *bs, int64_t sector_num, |
3239 | int nb_sectors, int *pnum) | |
3240 | { | |
4333bb71 PB |
3241 | int64_t ret = bdrv_get_block_status(bs, sector_num, nb_sectors, pnum); |
3242 | if (ret < 0) { | |
3243 | return ret; | |
3244 | } | |
3245 | return | |
3246 | (ret & BDRV_BLOCK_DATA) || | |
3247 | ((ret & BDRV_BLOCK_ZERO) && !bdrv_has_zero_init(bs)); | |
b6b8a333 PB |
3248 | } |
3249 | ||
188a7bbf PB |
3250 | /* |
3251 | * Given an image chain: ... -> [BASE] -> [INTER1] -> [INTER2] -> [TOP] | |
3252 | * | |
3253 | * Return true if the given sector is allocated in any image between | |
3254 | * BASE and TOP (inclusive). BASE can be NULL to check if the given | |
3255 | * sector is allocated in any image of the chain. Return false otherwise. | |
3256 | * | |
3257 | * 'pnum' is set to the number of sectors (including and immediately following | |
3258 | * the specified sector) that are known to be in the same | |
3259 | * allocated/unallocated state. | |
3260 | * | |
3261 | */ | |
4f578637 PB |
3262 | int bdrv_is_allocated_above(BlockDriverState *top, |
3263 | BlockDriverState *base, | |
3264 | int64_t sector_num, | |
3265 | int nb_sectors, int *pnum) | |
188a7bbf PB |
3266 | { |
3267 | BlockDriverState *intermediate; | |
3268 | int ret, n = nb_sectors; | |
3269 | ||
3270 | intermediate = top; | |
3271 | while (intermediate && intermediate != base) { | |
3272 | int pnum_inter; | |
bdad13b9 PB |
3273 | ret = bdrv_is_allocated(intermediate, sector_num, nb_sectors, |
3274 | &pnum_inter); | |
188a7bbf PB |
3275 | if (ret < 0) { |
3276 | return ret; | |
3277 | } else if (ret) { | |
3278 | *pnum = pnum_inter; | |
3279 | return 1; | |
3280 | } | |
3281 | ||
3282 | /* | |
3283 | * [sector_num, nb_sectors] is unallocated on top but intermediate | |
3284 | * might have | |
3285 | * | |
3286 | * [sector_num+x, nr_sectors] allocated. | |
3287 | */ | |
63ba17d3 VI |
3288 | if (n > pnum_inter && |
3289 | (intermediate == top || | |
3290 | sector_num + pnum_inter < intermediate->total_sectors)) { | |
188a7bbf PB |
3291 | n = pnum_inter; |
3292 | } | |
3293 | ||
3294 | intermediate = intermediate->backing_hd; | |
3295 | } | |
3296 | ||
3297 | *pnum = n; | |
3298 | return 0; | |
3299 | } | |
3300 | ||
045df330 AL |
3301 | const char *bdrv_get_encrypted_filename(BlockDriverState *bs) |
3302 | { | |
3303 | if (bs->backing_hd && bs->backing_hd->encrypted) | |
3304 | return bs->backing_file; | |
3305 | else if (bs->encrypted) | |
3306 | return bs->filename; | |
3307 | else | |
3308 | return NULL; | |
3309 | } | |
3310 | ||
5fafdf24 | 3311 | void bdrv_get_backing_filename(BlockDriverState *bs, |
83f64091 FB |
3312 | char *filename, int filename_size) |
3313 | { | |
3574c608 | 3314 | pstrcpy(filename, filename_size, bs->backing_file); |
83f64091 FB |
3315 | } |
3316 | ||
5fafdf24 | 3317 | int bdrv_write_compressed(BlockDriverState *bs, int64_t sector_num, |
faea38e7 FB |
3318 | const uint8_t *buf, int nb_sectors) |
3319 | { | |
3320 | BlockDriver *drv = bs->drv; | |
3321 | if (!drv) | |
19cb3738 | 3322 | return -ENOMEDIUM; |
faea38e7 FB |
3323 | if (!drv->bdrv_write_compressed) |
3324 | return -ENOTSUP; | |
fbb7b4e0 KW |
3325 | if (bdrv_check_request(bs, sector_num, nb_sectors)) |
3326 | return -EIO; | |
a55eb92c | 3327 | |
1755da16 | 3328 | assert(!bs->dirty_bitmap); |
a55eb92c | 3329 | |
faea38e7 FB |
3330 | return drv->bdrv_write_compressed(bs, sector_num, buf, nb_sectors); |
3331 | } | |
3b46e624 | 3332 | |
faea38e7 FB |
3333 | int bdrv_get_info(BlockDriverState *bs, BlockDriverInfo *bdi) |
3334 | { | |
3335 | BlockDriver *drv = bs->drv; | |
3336 | if (!drv) | |
19cb3738 | 3337 | return -ENOMEDIUM; |
faea38e7 FB |
3338 | if (!drv->bdrv_get_info) |
3339 | return -ENOTSUP; | |
3340 | memset(bdi, 0, sizeof(*bdi)); | |
3341 | return drv->bdrv_get_info(bs, bdi); | |
3342 | } | |
3343 | ||
eae041fe MR |
3344 | ImageInfoSpecific *bdrv_get_specific_info(BlockDriverState *bs) |
3345 | { | |
3346 | BlockDriver *drv = bs->drv; | |
3347 | if (drv && drv->bdrv_get_specific_info) { | |
3348 | return drv->bdrv_get_specific_info(bs); | |
3349 | } | |
3350 | return NULL; | |
3351 | } | |
3352 | ||
45566e9c CH |
3353 | int bdrv_save_vmstate(BlockDriverState *bs, const uint8_t *buf, |
3354 | int64_t pos, int size) | |
cf8074b3 KW |
3355 | { |
3356 | QEMUIOVector qiov; | |
3357 | struct iovec iov = { | |
3358 | .iov_base = (void *) buf, | |
3359 | .iov_len = size, | |
3360 | }; | |
3361 | ||
3362 | qemu_iovec_init_external(&qiov, &iov, 1); | |
3363 | return bdrv_writev_vmstate(bs, &qiov, pos); | |
3364 | } | |
3365 | ||
3366 | int bdrv_writev_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos) | |
178e08a5 AL |
3367 | { |
3368 | BlockDriver *drv = bs->drv; | |
cf8074b3 KW |
3369 | |
3370 | if (!drv) { | |
178e08a5 | 3371 | return -ENOMEDIUM; |
cf8074b3 KW |
3372 | } else if (drv->bdrv_save_vmstate) { |
3373 | return drv->bdrv_save_vmstate(bs, qiov, pos); | |
3374 | } else if (bs->file) { | |
3375 | return bdrv_writev_vmstate(bs->file, qiov, pos); | |
3376 | } | |
3377 | ||
7cdb1f6d | 3378 | return -ENOTSUP; |
178e08a5 AL |
3379 | } |
3380 | ||
45566e9c CH |
3381 | int bdrv_load_vmstate(BlockDriverState *bs, uint8_t *buf, |
3382 | int64_t pos, int size) | |
178e08a5 AL |
3383 | { |
3384 | BlockDriver *drv = bs->drv; | |
3385 | if (!drv) | |
3386 | return -ENOMEDIUM; | |
7cdb1f6d MK |
3387 | if (drv->bdrv_load_vmstate) |
3388 | return drv->bdrv_load_vmstate(bs, buf, pos, size); | |
3389 | if (bs->file) | |
3390 | return bdrv_load_vmstate(bs->file, buf, pos, size); | |
3391 | return -ENOTSUP; | |
178e08a5 AL |
3392 | } |
3393 | ||
8b9b0cc2 KW |
3394 | void bdrv_debug_event(BlockDriverState *bs, BlkDebugEvent event) |
3395 | { | |
bf736fe3 | 3396 | if (!bs || !bs->drv || !bs->drv->bdrv_debug_event) { |
8b9b0cc2 KW |
3397 | return; |
3398 | } | |
3399 | ||
bf736fe3 | 3400 | bs->drv->bdrv_debug_event(bs, event); |
41c695c7 KW |
3401 | } |
3402 | ||
3403 | int bdrv_debug_breakpoint(BlockDriverState *bs, const char *event, | |
3404 | const char *tag) | |
3405 | { | |
3406 | while (bs && bs->drv && !bs->drv->bdrv_debug_breakpoint) { | |
3407 | bs = bs->file; | |
3408 | } | |
3409 | ||
3410 | if (bs && bs->drv && bs->drv->bdrv_debug_breakpoint) { | |
3411 | return bs->drv->bdrv_debug_breakpoint(bs, event, tag); | |
3412 | } | |
3413 | ||
3414 | return -ENOTSUP; | |
3415 | } | |
3416 | ||
3417 | int bdrv_debug_resume(BlockDriverState *bs, const char *tag) | |
3418 | { | |
3419 | while (bs && bs->drv && !bs->drv->bdrv_debug_resume) { | |
3420 | bs = bs->file; | |
3421 | } | |
8b9b0cc2 | 3422 | |
41c695c7 KW |
3423 | if (bs && bs->drv && bs->drv->bdrv_debug_resume) { |
3424 | return bs->drv->bdrv_debug_resume(bs, tag); | |
3425 | } | |
3426 | ||
3427 | return -ENOTSUP; | |
3428 | } | |
3429 | ||
3430 | bool bdrv_debug_is_suspended(BlockDriverState *bs, const char *tag) | |
3431 | { | |
3432 | while (bs && bs->drv && !bs->drv->bdrv_debug_is_suspended) { | |
3433 | bs = bs->file; | |
3434 | } | |
3435 | ||
3436 | if (bs && bs->drv && bs->drv->bdrv_debug_is_suspended) { | |
3437 | return bs->drv->bdrv_debug_is_suspended(bs, tag); | |
3438 | } | |
3439 | ||
3440 | return false; | |
8b9b0cc2 KW |
3441 | } |
3442 | ||
199630b6 BS |
3443 | int bdrv_is_snapshot(BlockDriverState *bs) |
3444 | { | |
3445 | return !!(bs->open_flags & BDRV_O_SNAPSHOT); | |
3446 | } | |
3447 | ||
b1b1d783 JC |
3448 | /* backing_file can either be relative, or absolute, or a protocol. If it is |
3449 | * relative, it must be relative to the chain. So, passing in bs->filename | |
3450 | * from a BDS as backing_file should not be done, as that may be relative to | |
3451 | * the CWD rather than the chain. */ | |
e8a6bb9c MT |
3452 | BlockDriverState *bdrv_find_backing_image(BlockDriverState *bs, |
3453 | const char *backing_file) | |
3454 | { | |
b1b1d783 JC |
3455 | char *filename_full = NULL; |
3456 | char *backing_file_full = NULL; | |
3457 | char *filename_tmp = NULL; | |
3458 | int is_protocol = 0; | |
3459 | BlockDriverState *curr_bs = NULL; | |
3460 | BlockDriverState *retval = NULL; | |
3461 | ||
3462 | if (!bs || !bs->drv || !backing_file) { | |
e8a6bb9c MT |
3463 | return NULL; |
3464 | } | |
3465 | ||
b1b1d783 JC |
3466 | filename_full = g_malloc(PATH_MAX); |
3467 | backing_file_full = g_malloc(PATH_MAX); | |
3468 | filename_tmp = g_malloc(PATH_MAX); | |
3469 | ||
3470 | is_protocol = path_has_protocol(backing_file); | |
3471 | ||
3472 | for (curr_bs = bs; curr_bs->backing_hd; curr_bs = curr_bs->backing_hd) { | |
3473 | ||
3474 | /* If either of the filename paths is actually a protocol, then | |
3475 | * compare unmodified paths; otherwise make paths relative */ | |
3476 | if (is_protocol || path_has_protocol(curr_bs->backing_file)) { | |
3477 | if (strcmp(backing_file, curr_bs->backing_file) == 0) { | |
3478 | retval = curr_bs->backing_hd; | |
3479 | break; | |
3480 | } | |
e8a6bb9c | 3481 | } else { |
b1b1d783 JC |
3482 | /* If not an absolute filename path, make it relative to the current |
3483 | * image's filename path */ | |
3484 | path_combine(filename_tmp, PATH_MAX, curr_bs->filename, | |
3485 | backing_file); | |
3486 | ||
3487 | /* We are going to compare absolute pathnames */ | |
3488 | if (!realpath(filename_tmp, filename_full)) { | |
3489 | continue; | |
3490 | } | |
3491 | ||
3492 | /* We need to make sure the backing filename we are comparing against | |
3493 | * is relative to the current image filename (or absolute) */ | |
3494 | path_combine(filename_tmp, PATH_MAX, curr_bs->filename, | |
3495 | curr_bs->backing_file); | |
3496 | ||
3497 | if (!realpath(filename_tmp, backing_file_full)) { | |
3498 | continue; | |
3499 | } | |
3500 | ||
3501 | if (strcmp(backing_file_full, filename_full) == 0) { | |
3502 | retval = curr_bs->backing_hd; | |
3503 | break; | |
3504 | } | |
e8a6bb9c MT |
3505 | } |
3506 | } | |
3507 | ||
b1b1d783 JC |
3508 | g_free(filename_full); |
3509 | g_free(backing_file_full); | |
3510 | g_free(filename_tmp); | |
3511 | return retval; | |
e8a6bb9c MT |
3512 | } |
3513 | ||
f198fd1c BC |
3514 | int bdrv_get_backing_file_depth(BlockDriverState *bs) |
3515 | { | |
3516 | if (!bs->drv) { | |
3517 | return 0; | |
3518 | } | |
3519 | ||
3520 | if (!bs->backing_hd) { | |
3521 | return 0; | |
3522 | } | |
3523 | ||
3524 | return 1 + bdrv_get_backing_file_depth(bs->backing_hd); | |
3525 | } | |
3526 | ||
79fac568 JC |
3527 | BlockDriverState *bdrv_find_base(BlockDriverState *bs) |
3528 | { | |
3529 | BlockDriverState *curr_bs = NULL; | |
3530 | ||
3531 | if (!bs) { | |
3532 | return NULL; | |
3533 | } | |
3534 | ||
3535 | curr_bs = bs; | |
3536 | ||
3537 | while (curr_bs->backing_hd) { | |
3538 | curr_bs = curr_bs->backing_hd; | |
3539 | } | |
3540 | return curr_bs; | |
3541 | } | |
3542 | ||
ea2384d3 | 3543 | /**************************************************************/ |
83f64091 | 3544 | /* async I/Os */ |
ea2384d3 | 3545 | |
3b69e4b9 | 3546 | BlockDriverAIOCB *bdrv_aio_readv(BlockDriverState *bs, int64_t sector_num, |
f141eafe | 3547 | QEMUIOVector *qiov, int nb_sectors, |
3b69e4b9 | 3548 | BlockDriverCompletionFunc *cb, void *opaque) |
83f64091 | 3549 | { |
bbf0a440 SH |
3550 | trace_bdrv_aio_readv(bs, sector_num, nb_sectors, opaque); |
3551 | ||
b2a61371 | 3552 | return bdrv_co_aio_rw_vector(bs, sector_num, qiov, nb_sectors, |
8c5873d6 | 3553 | cb, opaque, false); |
ea2384d3 FB |
3554 | } |
3555 | ||
f141eafe AL |
3556 | BlockDriverAIOCB *bdrv_aio_writev(BlockDriverState *bs, int64_t sector_num, |
3557 | QEMUIOVector *qiov, int nb_sectors, | |
3558 | BlockDriverCompletionFunc *cb, void *opaque) | |
ea2384d3 | 3559 | { |
bbf0a440 SH |
3560 | trace_bdrv_aio_writev(bs, sector_num, nb_sectors, opaque); |
3561 | ||
1a6e115b | 3562 | return bdrv_co_aio_rw_vector(bs, sector_num, qiov, nb_sectors, |
8c5873d6 | 3563 | cb, opaque, true); |
83f64091 FB |
3564 | } |
3565 | ||
40b4f539 KW |
3566 | |
3567 | typedef struct MultiwriteCB { | |
3568 | int error; | |
3569 | int num_requests; | |
3570 | int num_callbacks; | |
3571 | struct { | |
3572 | BlockDriverCompletionFunc *cb; | |
3573 | void *opaque; | |
3574 | QEMUIOVector *free_qiov; | |
40b4f539 KW |
3575 | } callbacks[]; |
3576 | } MultiwriteCB; | |
3577 | ||
3578 | static void multiwrite_user_cb(MultiwriteCB *mcb) | |
3579 | { | |
3580 | int i; | |
3581 | ||
3582 | for (i = 0; i < mcb->num_callbacks; i++) { | |
3583 | mcb->callbacks[i].cb(mcb->callbacks[i].opaque, mcb->error); | |
1e1ea48d SH |
3584 | if (mcb->callbacks[i].free_qiov) { |
3585 | qemu_iovec_destroy(mcb->callbacks[i].free_qiov); | |
3586 | } | |
7267c094 | 3587 | g_free(mcb->callbacks[i].free_qiov); |
40b4f539 KW |
3588 | } |
3589 | } | |
3590 | ||
3591 | static void multiwrite_cb(void *opaque, int ret) | |
3592 | { | |
3593 | MultiwriteCB *mcb = opaque; | |
3594 | ||
6d519a5f SH |
3595 | trace_multiwrite_cb(mcb, ret); |
3596 | ||
cb6d3ca0 | 3597 | if (ret < 0 && !mcb->error) { |
40b4f539 | 3598 | mcb->error = ret; |
40b4f539 KW |
3599 | } |
3600 | ||
3601 | mcb->num_requests--; | |
3602 | if (mcb->num_requests == 0) { | |
de189a1b | 3603 | multiwrite_user_cb(mcb); |
7267c094 | 3604 | g_free(mcb); |
40b4f539 KW |
3605 | } |
3606 | } | |
3607 | ||
3608 | static int multiwrite_req_compare(const void *a, const void *b) | |
3609 | { | |
77be4366 CH |
3610 | const BlockRequest *req1 = a, *req2 = b; |
3611 | ||
3612 | /* | |
3613 | * Note that we can't simply subtract req2->sector from req1->sector | |
3614 | * here as that could overflow the return value. | |
3615 | */ | |
3616 | if (req1->sector > req2->sector) { | |
3617 | return 1; | |
3618 | } else if (req1->sector < req2->sector) { | |
3619 | return -1; | |
3620 | } else { | |
3621 | return 0; | |
3622 | } | |
40b4f539 KW |
3623 | } |
3624 | ||
3625 | /* | |
3626 | * Takes a bunch of requests and tries to merge them. Returns the number of | |
3627 | * requests that remain after merging. | |
3628 | */ | |
3629 | static int multiwrite_merge(BlockDriverState *bs, BlockRequest *reqs, | |
3630 | int num_reqs, MultiwriteCB *mcb) | |
3631 | { | |
3632 | int i, outidx; | |
3633 | ||
3634 | // Sort requests by start sector | |
3635 | qsort(reqs, num_reqs, sizeof(*reqs), &multiwrite_req_compare); | |
3636 | ||
3637 | // Check if adjacent requests touch the same clusters. If so, combine them, | |
3638 | // filling up gaps with zero sectors. | |
3639 | outidx = 0; | |
3640 | for (i = 1; i < num_reqs; i++) { | |
3641 | int merge = 0; | |
3642 | int64_t oldreq_last = reqs[outidx].sector + reqs[outidx].nb_sectors; | |
3643 | ||
b6a127a1 | 3644 | // Handle exactly sequential writes and overlapping writes. |
40b4f539 KW |
3645 | if (reqs[i].sector <= oldreq_last) { |
3646 | merge = 1; | |
3647 | } | |
3648 | ||
e2a305fb CH |
3649 | if (reqs[outidx].qiov->niov + reqs[i].qiov->niov + 1 > IOV_MAX) { |
3650 | merge = 0; | |
3651 | } | |
3652 | ||
40b4f539 KW |
3653 | if (merge) { |
3654 | size_t size; | |
7267c094 | 3655 | QEMUIOVector *qiov = g_malloc0(sizeof(*qiov)); |
40b4f539 KW |
3656 | qemu_iovec_init(qiov, |
3657 | reqs[outidx].qiov->niov + reqs[i].qiov->niov + 1); | |
3658 | ||
3659 | // Add the first request to the merged one. If the requests are | |
3660 | // overlapping, drop the last sectors of the first request. | |
3661 | size = (reqs[i].sector - reqs[outidx].sector) << 9; | |
1b093c48 | 3662 | qemu_iovec_concat(qiov, reqs[outidx].qiov, 0, size); |
40b4f539 | 3663 | |
b6a127a1 PB |
3664 | // We should need to add any zeros between the two requests |
3665 | assert (reqs[i].sector <= oldreq_last); | |
40b4f539 KW |
3666 | |
3667 | // Add the second request | |
1b093c48 | 3668 | qemu_iovec_concat(qiov, reqs[i].qiov, 0, reqs[i].qiov->size); |
40b4f539 | 3669 | |
cbf1dff2 | 3670 | reqs[outidx].nb_sectors = qiov->size >> 9; |
40b4f539 KW |
3671 | reqs[outidx].qiov = qiov; |
3672 | ||
3673 | mcb->callbacks[i].free_qiov = reqs[outidx].qiov; | |
3674 | } else { | |
3675 | outidx++; | |
3676 | reqs[outidx].sector = reqs[i].sector; | |
3677 | reqs[outidx].nb_sectors = reqs[i].nb_sectors; | |
3678 | reqs[outidx].qiov = reqs[i].qiov; | |
3679 | } | |
3680 | } | |
3681 | ||
3682 | return outidx + 1; | |
3683 | } | |
3684 | ||
3685 | /* | |
3686 | * Submit multiple AIO write requests at once. | |
3687 | * | |
3688 | * On success, the function returns 0 and all requests in the reqs array have | |
3689 | * been submitted. In error case this function returns -1, and any of the | |
3690 | * requests may or may not be submitted yet. In particular, this means that the | |
3691 | * callback will be called for some of the requests, for others it won't. The | |
3692 | * caller must check the error field of the BlockRequest to wait for the right | |
3693 | * callbacks (if error != 0, no callback will be called). | |
3694 | * | |
3695 | * The implementation may modify the contents of the reqs array, e.g. to merge | |
3696 | * requests. However, the fields opaque and error are left unmodified as they | |
3697 | * are used to signal failure for a single request to the caller. | |
3698 | */ | |
3699 | int bdrv_aio_multiwrite(BlockDriverState *bs, BlockRequest *reqs, int num_reqs) | |
3700 | { | |
40b4f539 KW |
3701 | MultiwriteCB *mcb; |
3702 | int i; | |
3703 | ||
301db7c2 RH |
3704 | /* don't submit writes if we don't have a medium */ |
3705 | if (bs->drv == NULL) { | |
3706 | for (i = 0; i < num_reqs; i++) { | |
3707 | reqs[i].error = -ENOMEDIUM; | |
3708 | } | |
3709 | return -1; | |
3710 | } | |
3711 | ||
40b4f539 KW |
3712 | if (num_reqs == 0) { |
3713 | return 0; | |
3714 | } | |
3715 | ||
3716 | // Create MultiwriteCB structure | |
7267c094 | 3717 | mcb = g_malloc0(sizeof(*mcb) + num_reqs * sizeof(*mcb->callbacks)); |
40b4f539 KW |
3718 | mcb->num_requests = 0; |
3719 | mcb->num_callbacks = num_reqs; | |
3720 | ||
3721 | for (i = 0; i < num_reqs; i++) { | |
3722 | mcb->callbacks[i].cb = reqs[i].cb; | |
3723 | mcb->callbacks[i].opaque = reqs[i].opaque; | |
3724 | } | |
3725 | ||
3726 | // Check for mergable requests | |
3727 | num_reqs = multiwrite_merge(bs, reqs, num_reqs, mcb); | |
3728 | ||
6d519a5f SH |
3729 | trace_bdrv_aio_multiwrite(mcb, mcb->num_callbacks, num_reqs); |
3730 | ||
df9309fb PB |
3731 | /* Run the aio requests. */ |
3732 | mcb->num_requests = num_reqs; | |
40b4f539 | 3733 | for (i = 0; i < num_reqs; i++) { |
ad54ae80 | 3734 | bdrv_aio_writev(bs, reqs[i].sector, reqs[i].qiov, |
40b4f539 | 3735 | reqs[i].nb_sectors, multiwrite_cb, mcb); |
40b4f539 KW |
3736 | } |
3737 | ||
3738 | return 0; | |
40b4f539 KW |
3739 | } |
3740 | ||
83f64091 | 3741 | void bdrv_aio_cancel(BlockDriverAIOCB *acb) |
83f64091 | 3742 | { |
d7331bed | 3743 | acb->aiocb_info->cancel(acb); |
83f64091 FB |
3744 | } |
3745 | ||
3746 | /**************************************************************/ | |
3747 | /* async block device emulation */ | |
3748 | ||
c16b5a2c CH |
3749 | typedef struct BlockDriverAIOCBSync { |
3750 | BlockDriverAIOCB common; | |
3751 | QEMUBH *bh; | |
3752 | int ret; | |
3753 | /* vector translation state */ | |
3754 | QEMUIOVector *qiov; | |
3755 | uint8_t *bounce; | |
3756 | int is_write; | |
3757 | } BlockDriverAIOCBSync; | |
3758 | ||
3759 | static void bdrv_aio_cancel_em(BlockDriverAIOCB *blockacb) | |
3760 | { | |
b666d239 KW |
3761 | BlockDriverAIOCBSync *acb = |
3762 | container_of(blockacb, BlockDriverAIOCBSync, common); | |
6a7ad299 | 3763 | qemu_bh_delete(acb->bh); |
36afc451 | 3764 | acb->bh = NULL; |
c16b5a2c CH |
3765 | qemu_aio_release(acb); |
3766 | } | |
3767 | ||
d7331bed | 3768 | static const AIOCBInfo bdrv_em_aiocb_info = { |
c16b5a2c CH |
3769 | .aiocb_size = sizeof(BlockDriverAIOCBSync), |
3770 | .cancel = bdrv_aio_cancel_em, | |
3771 | }; | |
3772 | ||
ce1a14dc | 3773 | static void bdrv_aio_bh_cb(void *opaque) |
83f64091 | 3774 | { |
ce1a14dc | 3775 | BlockDriverAIOCBSync *acb = opaque; |
f141eafe | 3776 | |
f141eafe | 3777 | if (!acb->is_write) |
03396148 | 3778 | qemu_iovec_from_buf(acb->qiov, 0, acb->bounce, acb->qiov->size); |
ceb42de8 | 3779 | qemu_vfree(acb->bounce); |
ce1a14dc | 3780 | acb->common.cb(acb->common.opaque, acb->ret); |
6a7ad299 | 3781 | qemu_bh_delete(acb->bh); |
36afc451 | 3782 | acb->bh = NULL; |
ce1a14dc | 3783 | qemu_aio_release(acb); |
83f64091 | 3784 | } |
beac80cd | 3785 | |
f141eafe AL |
3786 | static BlockDriverAIOCB *bdrv_aio_rw_vector(BlockDriverState *bs, |
3787 | int64_t sector_num, | |
3788 | QEMUIOVector *qiov, | |
3789 | int nb_sectors, | |
3790 | BlockDriverCompletionFunc *cb, | |
3791 | void *opaque, | |
3792 | int is_write) | |
3793 | ||
83f64091 | 3794 | { |
ce1a14dc | 3795 | BlockDriverAIOCBSync *acb; |
ce1a14dc | 3796 | |
d7331bed | 3797 | acb = qemu_aio_get(&bdrv_em_aiocb_info, bs, cb, opaque); |
f141eafe AL |
3798 | acb->is_write = is_write; |
3799 | acb->qiov = qiov; | |
e268ca52 | 3800 | acb->bounce = qemu_blockalign(bs, qiov->size); |
3f3aace8 | 3801 | acb->bh = qemu_bh_new(bdrv_aio_bh_cb, acb); |
f141eafe AL |
3802 | |
3803 | if (is_write) { | |
d5e6b161 | 3804 | qemu_iovec_to_buf(acb->qiov, 0, acb->bounce, qiov->size); |
1ed20acf | 3805 | acb->ret = bs->drv->bdrv_write(bs, sector_num, acb->bounce, nb_sectors); |
f141eafe | 3806 | } else { |
1ed20acf | 3807 | acb->ret = bs->drv->bdrv_read(bs, sector_num, acb->bounce, nb_sectors); |
f141eafe AL |
3808 | } |
3809 | ||
ce1a14dc | 3810 | qemu_bh_schedule(acb->bh); |
f141eafe | 3811 | |
ce1a14dc | 3812 | return &acb->common; |
beac80cd FB |
3813 | } |
3814 | ||
f141eafe AL |
3815 | static BlockDriverAIOCB *bdrv_aio_readv_em(BlockDriverState *bs, |
3816 | int64_t sector_num, QEMUIOVector *qiov, int nb_sectors, | |
ce1a14dc | 3817 | BlockDriverCompletionFunc *cb, void *opaque) |
beac80cd | 3818 | { |
f141eafe AL |
3819 | return bdrv_aio_rw_vector(bs, sector_num, qiov, nb_sectors, cb, opaque, 0); |
3820 | } | |
83f64091 | 3821 | |
f141eafe AL |
3822 | static BlockDriverAIOCB *bdrv_aio_writev_em(BlockDriverState *bs, |
3823 | int64_t sector_num, QEMUIOVector *qiov, int nb_sectors, | |
3824 | BlockDriverCompletionFunc *cb, void *opaque) | |
3825 | { | |
3826 | return bdrv_aio_rw_vector(bs, sector_num, qiov, nb_sectors, cb, opaque, 1); | |
beac80cd | 3827 | } |
beac80cd | 3828 | |
68485420 KW |
3829 | |
3830 | typedef struct BlockDriverAIOCBCoroutine { | |
3831 | BlockDriverAIOCB common; | |
3832 | BlockRequest req; | |
3833 | bool is_write; | |
d318aea9 | 3834 | bool *done; |
68485420 KW |
3835 | QEMUBH* bh; |
3836 | } BlockDriverAIOCBCoroutine; | |
3837 | ||
3838 | static void bdrv_aio_co_cancel_em(BlockDriverAIOCB *blockacb) | |
3839 | { | |
d318aea9 KW |
3840 | BlockDriverAIOCBCoroutine *acb = |
3841 | container_of(blockacb, BlockDriverAIOCBCoroutine, common); | |
3842 | bool done = false; | |
3843 | ||
3844 | acb->done = &done; | |
3845 | while (!done) { | |
3846 | qemu_aio_wait(); | |
3847 | } | |
68485420 KW |
3848 | } |
3849 | ||
d7331bed | 3850 | static const AIOCBInfo bdrv_em_co_aiocb_info = { |
68485420 KW |
3851 | .aiocb_size = sizeof(BlockDriverAIOCBCoroutine), |
3852 | .cancel = bdrv_aio_co_cancel_em, | |
3853 | }; | |
3854 | ||
35246a68 | 3855 | static void bdrv_co_em_bh(void *opaque) |
68485420 KW |
3856 | { |
3857 | BlockDriverAIOCBCoroutine *acb = opaque; | |
3858 | ||
3859 | acb->common.cb(acb->common.opaque, acb->req.error); | |
d318aea9 KW |
3860 | |
3861 | if (acb->done) { | |
3862 | *acb->done = true; | |
3863 | } | |
3864 | ||
68485420 KW |
3865 | qemu_bh_delete(acb->bh); |
3866 | qemu_aio_release(acb); | |
3867 | } | |
3868 | ||
b2a61371 SH |
3869 | /* Invoke bdrv_co_do_readv/bdrv_co_do_writev */ |
3870 | static void coroutine_fn bdrv_co_do_rw(void *opaque) | |
3871 | { | |
3872 | BlockDriverAIOCBCoroutine *acb = opaque; | |
3873 | BlockDriverState *bs = acb->common.bs; | |
3874 | ||
3875 | if (!acb->is_write) { | |
3876 | acb->req.error = bdrv_co_do_readv(bs, acb->req.sector, | |
470c0504 | 3877 | acb->req.nb_sectors, acb->req.qiov, 0); |
b2a61371 SH |
3878 | } else { |
3879 | acb->req.error = bdrv_co_do_writev(bs, acb->req.sector, | |
f08f2dda | 3880 | acb->req.nb_sectors, acb->req.qiov, 0); |
b2a61371 SH |
3881 | } |
3882 | ||
35246a68 | 3883 | acb->bh = qemu_bh_new(bdrv_co_em_bh, acb); |
b2a61371 SH |
3884 | qemu_bh_schedule(acb->bh); |
3885 | } | |
3886 | ||
68485420 KW |
3887 | static BlockDriverAIOCB *bdrv_co_aio_rw_vector(BlockDriverState *bs, |
3888 | int64_t sector_num, | |
3889 | QEMUIOVector *qiov, | |
3890 | int nb_sectors, | |
3891 | BlockDriverCompletionFunc *cb, | |
3892 | void *opaque, | |
8c5873d6 | 3893 | bool is_write) |
68485420 KW |
3894 | { |
3895 | Coroutine *co; | |
3896 | BlockDriverAIOCBCoroutine *acb; | |
3897 | ||
d7331bed | 3898 | acb = qemu_aio_get(&bdrv_em_co_aiocb_info, bs, cb, opaque); |
68485420 KW |
3899 | acb->req.sector = sector_num; |
3900 | acb->req.nb_sectors = nb_sectors; | |
3901 | acb->req.qiov = qiov; | |
3902 | acb->is_write = is_write; | |
d318aea9 | 3903 | acb->done = NULL; |
68485420 | 3904 | |
8c5873d6 | 3905 | co = qemu_coroutine_create(bdrv_co_do_rw); |
68485420 KW |
3906 | qemu_coroutine_enter(co, acb); |
3907 | ||
3908 | return &acb->common; | |
3909 | } | |
3910 | ||
07f07615 | 3911 | static void coroutine_fn bdrv_aio_flush_co_entry(void *opaque) |
b2e12bc6 | 3912 | { |
07f07615 PB |
3913 | BlockDriverAIOCBCoroutine *acb = opaque; |
3914 | BlockDriverState *bs = acb->common.bs; | |
b2e12bc6 | 3915 | |
07f07615 PB |
3916 | acb->req.error = bdrv_co_flush(bs); |
3917 | acb->bh = qemu_bh_new(bdrv_co_em_bh, acb); | |
b2e12bc6 | 3918 | qemu_bh_schedule(acb->bh); |
b2e12bc6 CH |
3919 | } |
3920 | ||
07f07615 | 3921 | BlockDriverAIOCB *bdrv_aio_flush(BlockDriverState *bs, |
016f5cf6 AG |
3922 | BlockDriverCompletionFunc *cb, void *opaque) |
3923 | { | |
07f07615 | 3924 | trace_bdrv_aio_flush(bs, opaque); |
016f5cf6 | 3925 | |
07f07615 PB |
3926 | Coroutine *co; |
3927 | BlockDriverAIOCBCoroutine *acb; | |
016f5cf6 | 3928 | |
d7331bed | 3929 | acb = qemu_aio_get(&bdrv_em_co_aiocb_info, bs, cb, opaque); |
d318aea9 KW |
3930 | acb->done = NULL; |
3931 | ||
07f07615 PB |
3932 | co = qemu_coroutine_create(bdrv_aio_flush_co_entry); |
3933 | qemu_coroutine_enter(co, acb); | |
016f5cf6 | 3934 | |
016f5cf6 AG |
3935 | return &acb->common; |
3936 | } | |
3937 | ||
4265d620 PB |
3938 | static void coroutine_fn bdrv_aio_discard_co_entry(void *opaque) |
3939 | { | |
3940 | BlockDriverAIOCBCoroutine *acb = opaque; | |
3941 | BlockDriverState *bs = acb->common.bs; | |
3942 | ||
3943 | acb->req.error = bdrv_co_discard(bs, acb->req.sector, acb->req.nb_sectors); | |
3944 | acb->bh = qemu_bh_new(bdrv_co_em_bh, acb); | |
3945 | qemu_bh_schedule(acb->bh); | |
3946 | } | |
3947 | ||
3948 | BlockDriverAIOCB *bdrv_aio_discard(BlockDriverState *bs, | |
3949 | int64_t sector_num, int nb_sectors, | |
3950 | BlockDriverCompletionFunc *cb, void *opaque) | |
3951 | { | |
3952 | Coroutine *co; | |
3953 | BlockDriverAIOCBCoroutine *acb; | |
3954 | ||
3955 | trace_bdrv_aio_discard(bs, sector_num, nb_sectors, opaque); | |
3956 | ||
d7331bed | 3957 | acb = qemu_aio_get(&bdrv_em_co_aiocb_info, bs, cb, opaque); |
4265d620 PB |
3958 | acb->req.sector = sector_num; |
3959 | acb->req.nb_sectors = nb_sectors; | |
d318aea9 | 3960 | acb->done = NULL; |
4265d620 PB |
3961 | co = qemu_coroutine_create(bdrv_aio_discard_co_entry); |
3962 | qemu_coroutine_enter(co, acb); | |
3963 | ||
3964 | return &acb->common; | |
3965 | } | |
3966 | ||
ea2384d3 FB |
3967 | void bdrv_init(void) |
3968 | { | |
5efa9d5a | 3969 | module_call_init(MODULE_INIT_BLOCK); |
ea2384d3 | 3970 | } |
ce1a14dc | 3971 | |
eb852011 MA |
3972 | void bdrv_init_with_whitelist(void) |
3973 | { | |
3974 | use_bdrv_whitelist = 1; | |
3975 | bdrv_init(); | |
3976 | } | |
3977 | ||
d7331bed | 3978 | void *qemu_aio_get(const AIOCBInfo *aiocb_info, BlockDriverState *bs, |
c16b5a2c | 3979 | BlockDriverCompletionFunc *cb, void *opaque) |
ce1a14dc | 3980 | { |
ce1a14dc PB |
3981 | BlockDriverAIOCB *acb; |
3982 | ||
d7331bed SH |
3983 | acb = g_slice_alloc(aiocb_info->aiocb_size); |
3984 | acb->aiocb_info = aiocb_info; | |
ce1a14dc PB |
3985 | acb->bs = bs; |
3986 | acb->cb = cb; | |
3987 | acb->opaque = opaque; | |
3988 | return acb; | |
3989 | } | |
3990 | ||
3991 | void qemu_aio_release(void *p) | |
3992 | { | |
d37c975f | 3993 | BlockDriverAIOCB *acb = p; |
d7331bed | 3994 | g_slice_free1(acb->aiocb_info->aiocb_size, acb); |
ce1a14dc | 3995 | } |
19cb3738 | 3996 | |
f9f05dc5 KW |
3997 | /**************************************************************/ |
3998 | /* Coroutine block device emulation */ | |
3999 | ||
4000 | typedef struct CoroutineIOCompletion { | |
4001 | Coroutine *coroutine; | |
4002 | int ret; | |
4003 | } CoroutineIOCompletion; | |
4004 | ||
4005 | static void bdrv_co_io_em_complete(void *opaque, int ret) | |
4006 | { | |
4007 | CoroutineIOCompletion *co = opaque; | |
4008 | ||
4009 | co->ret = ret; | |
4010 | qemu_coroutine_enter(co->coroutine, NULL); | |
4011 | } | |
4012 | ||
4013 | static int coroutine_fn bdrv_co_io_em(BlockDriverState *bs, int64_t sector_num, | |
4014 | int nb_sectors, QEMUIOVector *iov, | |
4015 | bool is_write) | |
4016 | { | |
4017 | CoroutineIOCompletion co = { | |
4018 | .coroutine = qemu_coroutine_self(), | |
4019 | }; | |
4020 | BlockDriverAIOCB *acb; | |
4021 | ||
4022 | if (is_write) { | |
a652d160 SH |
4023 | acb = bs->drv->bdrv_aio_writev(bs, sector_num, iov, nb_sectors, |
4024 | bdrv_co_io_em_complete, &co); | |
f9f05dc5 | 4025 | } else { |
a652d160 SH |
4026 | acb = bs->drv->bdrv_aio_readv(bs, sector_num, iov, nb_sectors, |
4027 | bdrv_co_io_em_complete, &co); | |
f9f05dc5 KW |
4028 | } |
4029 | ||
59370aaa | 4030 | trace_bdrv_co_io_em(bs, sector_num, nb_sectors, is_write, acb); |
f9f05dc5 KW |
4031 | if (!acb) { |
4032 | return -EIO; | |
4033 | } | |
4034 | qemu_coroutine_yield(); | |
4035 | ||
4036 | return co.ret; | |
4037 | } | |
4038 | ||
4039 | static int coroutine_fn bdrv_co_readv_em(BlockDriverState *bs, | |
4040 | int64_t sector_num, int nb_sectors, | |
4041 | QEMUIOVector *iov) | |
4042 | { | |
4043 | return bdrv_co_io_em(bs, sector_num, nb_sectors, iov, false); | |
4044 | } | |
4045 | ||
4046 | static int coroutine_fn bdrv_co_writev_em(BlockDriverState *bs, | |
4047 | int64_t sector_num, int nb_sectors, | |
4048 | QEMUIOVector *iov) | |
4049 | { | |
4050 | return bdrv_co_io_em(bs, sector_num, nb_sectors, iov, true); | |
4051 | } | |
4052 | ||
07f07615 | 4053 | static void coroutine_fn bdrv_flush_co_entry(void *opaque) |
e7a8a783 | 4054 | { |
07f07615 PB |
4055 | RwCo *rwco = opaque; |
4056 | ||
4057 | rwco->ret = bdrv_co_flush(rwco->bs); | |
4058 | } | |
4059 | ||
4060 | int coroutine_fn bdrv_co_flush(BlockDriverState *bs) | |
4061 | { | |
eb489bb1 KW |
4062 | int ret; |
4063 | ||
29cdb251 | 4064 | if (!bs || !bdrv_is_inserted(bs) || bdrv_is_read_only(bs)) { |
07f07615 | 4065 | return 0; |
eb489bb1 KW |
4066 | } |
4067 | ||
ca716364 | 4068 | /* Write back cached data to the OS even with cache=unsafe */ |
bf736fe3 | 4069 | BLKDBG_EVENT(bs->file, BLKDBG_FLUSH_TO_OS); |
eb489bb1 KW |
4070 | if (bs->drv->bdrv_co_flush_to_os) { |
4071 | ret = bs->drv->bdrv_co_flush_to_os(bs); | |
4072 | if (ret < 0) { | |
4073 | return ret; | |
4074 | } | |
4075 | } | |
4076 | ||
ca716364 KW |
4077 | /* But don't actually force it to the disk with cache=unsafe */ |
4078 | if (bs->open_flags & BDRV_O_NO_FLUSH) { | |
d4c82329 | 4079 | goto flush_parent; |
ca716364 KW |
4080 | } |
4081 | ||
bf736fe3 | 4082 | BLKDBG_EVENT(bs->file, BLKDBG_FLUSH_TO_DISK); |
eb489bb1 | 4083 | if (bs->drv->bdrv_co_flush_to_disk) { |
29cdb251 | 4084 | ret = bs->drv->bdrv_co_flush_to_disk(bs); |
07f07615 PB |
4085 | } else if (bs->drv->bdrv_aio_flush) { |
4086 | BlockDriverAIOCB *acb; | |
4087 | CoroutineIOCompletion co = { | |
4088 | .coroutine = qemu_coroutine_self(), | |
4089 | }; | |
4090 | ||
4091 | acb = bs->drv->bdrv_aio_flush(bs, bdrv_co_io_em_complete, &co); | |
4092 | if (acb == NULL) { | |
29cdb251 | 4093 | ret = -EIO; |
07f07615 PB |
4094 | } else { |
4095 | qemu_coroutine_yield(); | |
29cdb251 | 4096 | ret = co.ret; |
07f07615 | 4097 | } |
07f07615 PB |
4098 | } else { |
4099 | /* | |
4100 | * Some block drivers always operate in either writethrough or unsafe | |
4101 | * mode and don't support bdrv_flush therefore. Usually qemu doesn't | |
4102 | * know how the server works (because the behaviour is hardcoded or | |
4103 | * depends on server-side configuration), so we can't ensure that | |
4104 | * everything is safe on disk. Returning an error doesn't work because | |
4105 | * that would break guests even if the server operates in writethrough | |
4106 | * mode. | |
4107 | * | |
4108 | * Let's hope the user knows what he's doing. | |
4109 | */ | |
29cdb251 | 4110 | ret = 0; |
07f07615 | 4111 | } |
29cdb251 PB |
4112 | if (ret < 0) { |
4113 | return ret; | |
4114 | } | |
4115 | ||
4116 | /* Now flush the underlying protocol. It will also have BDRV_O_NO_FLUSH | |
4117 | * in the case of cache=unsafe, so there are no useless flushes. | |
4118 | */ | |
d4c82329 | 4119 | flush_parent: |
29cdb251 | 4120 | return bdrv_co_flush(bs->file); |
07f07615 PB |
4121 | } |
4122 | ||
0f15423c AL |
4123 | void bdrv_invalidate_cache(BlockDriverState *bs) |
4124 | { | |
4125 | if (bs->drv && bs->drv->bdrv_invalidate_cache) { | |
4126 | bs->drv->bdrv_invalidate_cache(bs); | |
4127 | } | |
4128 | } | |
4129 | ||
4130 | void bdrv_invalidate_cache_all(void) | |
4131 | { | |
4132 | BlockDriverState *bs; | |
4133 | ||
4134 | QTAILQ_FOREACH(bs, &bdrv_states, list) { | |
4135 | bdrv_invalidate_cache(bs); | |
4136 | } | |
4137 | } | |
4138 | ||
07789269 BC |
4139 | void bdrv_clear_incoming_migration_all(void) |
4140 | { | |
4141 | BlockDriverState *bs; | |
4142 | ||
4143 | QTAILQ_FOREACH(bs, &bdrv_states, list) { | |
4144 | bs->open_flags = bs->open_flags & ~(BDRV_O_INCOMING); | |
4145 | } | |
4146 | } | |
4147 | ||
07f07615 PB |
4148 | int bdrv_flush(BlockDriverState *bs) |
4149 | { | |
4150 | Coroutine *co; | |
4151 | RwCo rwco = { | |
4152 | .bs = bs, | |
4153 | .ret = NOT_DONE, | |
e7a8a783 | 4154 | }; |
e7a8a783 | 4155 | |
07f07615 PB |
4156 | if (qemu_in_coroutine()) { |
4157 | /* Fast-path if already in coroutine context */ | |
4158 | bdrv_flush_co_entry(&rwco); | |
4159 | } else { | |
4160 | co = qemu_coroutine_create(bdrv_flush_co_entry); | |
4161 | qemu_coroutine_enter(co, &rwco); | |
4162 | while (rwco.ret == NOT_DONE) { | |
4163 | qemu_aio_wait(); | |
4164 | } | |
e7a8a783 | 4165 | } |
07f07615 PB |
4166 | |
4167 | return rwco.ret; | |
e7a8a783 KW |
4168 | } |
4169 | ||
4265d620 PB |
4170 | static void coroutine_fn bdrv_discard_co_entry(void *opaque) |
4171 | { | |
4172 | RwCo *rwco = opaque; | |
4173 | ||
4174 | rwco->ret = bdrv_co_discard(rwco->bs, rwco->sector_num, rwco->nb_sectors); | |
4175 | } | |
4176 | ||
4177 | int coroutine_fn bdrv_co_discard(BlockDriverState *bs, int64_t sector_num, | |
4178 | int nb_sectors) | |
4179 | { | |
4180 | if (!bs->drv) { | |
4181 | return -ENOMEDIUM; | |
4182 | } else if (bdrv_check_request(bs, sector_num, nb_sectors)) { | |
4183 | return -EIO; | |
4184 | } else if (bs->read_only) { | |
4185 | return -EROFS; | |
df702c9b PB |
4186 | } |
4187 | ||
4188 | if (bs->dirty_bitmap) { | |
8f0720ec | 4189 | bdrv_reset_dirty(bs, sector_num, nb_sectors); |
df702c9b PB |
4190 | } |
4191 | ||
9e8f1835 PB |
4192 | /* Do nothing if disabled. */ |
4193 | if (!(bs->open_flags & BDRV_O_UNMAP)) { | |
4194 | return 0; | |
4195 | } | |
4196 | ||
df702c9b | 4197 | if (bs->drv->bdrv_co_discard) { |
4265d620 PB |
4198 | return bs->drv->bdrv_co_discard(bs, sector_num, nb_sectors); |
4199 | } else if (bs->drv->bdrv_aio_discard) { | |
4200 | BlockDriverAIOCB *acb; | |
4201 | CoroutineIOCompletion co = { | |
4202 | .coroutine = qemu_coroutine_self(), | |
4203 | }; | |
4204 | ||
4205 | acb = bs->drv->bdrv_aio_discard(bs, sector_num, nb_sectors, | |
4206 | bdrv_co_io_em_complete, &co); | |
4207 | if (acb == NULL) { | |
4208 | return -EIO; | |
4209 | } else { | |
4210 | qemu_coroutine_yield(); | |
4211 | return co.ret; | |
4212 | } | |
4265d620 PB |
4213 | } else { |
4214 | return 0; | |
4215 | } | |
4216 | } | |
4217 | ||
4218 | int bdrv_discard(BlockDriverState *bs, int64_t sector_num, int nb_sectors) | |
4219 | { | |
4220 | Coroutine *co; | |
4221 | RwCo rwco = { | |
4222 | .bs = bs, | |
4223 | .sector_num = sector_num, | |
4224 | .nb_sectors = nb_sectors, | |
4225 | .ret = NOT_DONE, | |
4226 | }; | |
4227 | ||
4228 | if (qemu_in_coroutine()) { | |
4229 | /* Fast-path if already in coroutine context */ | |
4230 | bdrv_discard_co_entry(&rwco); | |
4231 | } else { | |
4232 | co = qemu_coroutine_create(bdrv_discard_co_entry); | |
4233 | qemu_coroutine_enter(co, &rwco); | |
4234 | while (rwco.ret == NOT_DONE) { | |
4235 | qemu_aio_wait(); | |
4236 | } | |
4237 | } | |
4238 | ||
4239 | return rwco.ret; | |
4240 | } | |
4241 | ||
19cb3738 FB |
4242 | /**************************************************************/ |
4243 | /* removable device support */ | |
4244 | ||
4245 | /** | |
4246 | * Return TRUE if the media is present | |
4247 | */ | |
4248 | int bdrv_is_inserted(BlockDriverState *bs) | |
4249 | { | |
4250 | BlockDriver *drv = bs->drv; | |
a1aff5bf | 4251 | |
19cb3738 FB |
4252 | if (!drv) |
4253 | return 0; | |
4254 | if (!drv->bdrv_is_inserted) | |
a1aff5bf MA |
4255 | return 1; |
4256 | return drv->bdrv_is_inserted(bs); | |
19cb3738 FB |
4257 | } |
4258 | ||
4259 | /** | |
8e49ca46 MA |
4260 | * Return whether the media changed since the last call to this |
4261 | * function, or -ENOTSUP if we don't know. Most drivers don't know. | |
19cb3738 FB |
4262 | */ |
4263 | int bdrv_media_changed(BlockDriverState *bs) | |
4264 | { | |
4265 | BlockDriver *drv = bs->drv; | |
19cb3738 | 4266 | |
8e49ca46 MA |
4267 | if (drv && drv->bdrv_media_changed) { |
4268 | return drv->bdrv_media_changed(bs); | |
4269 | } | |
4270 | return -ENOTSUP; | |
19cb3738 FB |
4271 | } |
4272 | ||
4273 | /** | |
4274 | * If eject_flag is TRUE, eject the media. Otherwise, close the tray | |
4275 | */ | |
f36f3949 | 4276 | void bdrv_eject(BlockDriverState *bs, bool eject_flag) |
19cb3738 FB |
4277 | { |
4278 | BlockDriver *drv = bs->drv; | |
19cb3738 | 4279 | |
822e1cd1 MA |
4280 | if (drv && drv->bdrv_eject) { |
4281 | drv->bdrv_eject(bs, eject_flag); | |
19cb3738 | 4282 | } |
6f382ed2 LC |
4283 | |
4284 | if (bs->device_name[0] != '\0') { | |
4285 | bdrv_emit_qmp_eject_event(bs, eject_flag); | |
4286 | } | |
19cb3738 FB |
4287 | } |
4288 | ||
19cb3738 FB |
4289 | /** |
4290 | * Lock or unlock the media (if it is locked, the user won't be able | |
4291 | * to eject it manually). | |
4292 | */ | |
025e849a | 4293 | void bdrv_lock_medium(BlockDriverState *bs, bool locked) |
19cb3738 FB |
4294 | { |
4295 | BlockDriver *drv = bs->drv; | |
4296 | ||
025e849a | 4297 | trace_bdrv_lock_medium(bs, locked); |
b8c6d095 | 4298 | |
025e849a MA |
4299 | if (drv && drv->bdrv_lock_medium) { |
4300 | drv->bdrv_lock_medium(bs, locked); | |
19cb3738 FB |
4301 | } |
4302 | } | |
985a03b0 TS |
4303 | |
4304 | /* needed for generic scsi interface */ | |
4305 | ||
4306 | int bdrv_ioctl(BlockDriverState *bs, unsigned long int req, void *buf) | |
4307 | { | |
4308 | BlockDriver *drv = bs->drv; | |
4309 | ||
4310 | if (drv && drv->bdrv_ioctl) | |
4311 | return drv->bdrv_ioctl(bs, req, buf); | |
4312 | return -ENOTSUP; | |
4313 | } | |
7d780669 | 4314 | |
221f715d AL |
4315 | BlockDriverAIOCB *bdrv_aio_ioctl(BlockDriverState *bs, |
4316 | unsigned long int req, void *buf, | |
4317 | BlockDriverCompletionFunc *cb, void *opaque) | |
7d780669 | 4318 | { |
221f715d | 4319 | BlockDriver *drv = bs->drv; |
7d780669 | 4320 | |
221f715d AL |
4321 | if (drv && drv->bdrv_aio_ioctl) |
4322 | return drv->bdrv_aio_ioctl(bs, req, buf, cb, opaque); | |
4323 | return NULL; | |
7d780669 | 4324 | } |
e268ca52 | 4325 | |
7b6f9300 MA |
4326 | void bdrv_set_buffer_alignment(BlockDriverState *bs, int align) |
4327 | { | |
4328 | bs->buffer_alignment = align; | |
4329 | } | |
7cd1e32a | 4330 | |
e268ca52 AL |
4331 | void *qemu_blockalign(BlockDriverState *bs, size_t size) |
4332 | { | |
4333 | return qemu_memalign((bs && bs->buffer_alignment) ? bs->buffer_alignment : 512, size); | |
4334 | } | |
7cd1e32a | 4335 | |
c53b1c51 SH |
4336 | /* |
4337 | * Check if all memory in this vector is sector aligned. | |
4338 | */ | |
4339 | bool bdrv_qiov_is_aligned(BlockDriverState *bs, QEMUIOVector *qiov) | |
4340 | { | |
4341 | int i; | |
4342 | ||
4343 | for (i = 0; i < qiov->niov; i++) { | |
4344 | if ((uintptr_t) qiov->iov[i].iov_base % bs->buffer_alignment) { | |
4345 | return false; | |
4346 | } | |
4347 | } | |
4348 | ||
4349 | return true; | |
4350 | } | |
4351 | ||
50717e94 | 4352 | void bdrv_set_dirty_tracking(BlockDriverState *bs, int granularity) |
7cd1e32a | 4353 | { |
4354 | int64_t bitmap_size; | |
a55eb92c | 4355 | |
50717e94 PB |
4356 | assert((granularity & (granularity - 1)) == 0); |
4357 | ||
4358 | if (granularity) { | |
4359 | granularity >>= BDRV_SECTOR_BITS; | |
4360 | assert(!bs->dirty_bitmap); | |
4361 | bitmap_size = (bdrv_getlength(bs) >> BDRV_SECTOR_BITS); | |
4362 | bs->dirty_bitmap = hbitmap_alloc(bitmap_size, ffs(granularity) - 1); | |
7cd1e32a | 4363 | } else { |
c6d22830 | 4364 | if (bs->dirty_bitmap) { |
8f0720ec | 4365 | hbitmap_free(bs->dirty_bitmap); |
c6d22830 | 4366 | bs->dirty_bitmap = NULL; |
a55eb92c | 4367 | } |
7cd1e32a | 4368 | } |
4369 | } | |
4370 | ||
4371 | int bdrv_get_dirty(BlockDriverState *bs, int64_t sector) | |
4372 | { | |
8f0720ec PB |
4373 | if (bs->dirty_bitmap) { |
4374 | return hbitmap_get(bs->dirty_bitmap, sector); | |
7cd1e32a | 4375 | } else { |
4376 | return 0; | |
4377 | } | |
4378 | } | |
4379 | ||
8f0720ec | 4380 | void bdrv_dirty_iter_init(BlockDriverState *bs, HBitmapIter *hbi) |
1755da16 | 4381 | { |
8f0720ec | 4382 | hbitmap_iter_init(hbi, bs->dirty_bitmap, 0); |
1755da16 PB |
4383 | } |
4384 | ||
4385 | void bdrv_set_dirty(BlockDriverState *bs, int64_t cur_sector, | |
4386 | int nr_sectors) | |
4387 | { | |
8f0720ec | 4388 | hbitmap_set(bs->dirty_bitmap, cur_sector, nr_sectors); |
1755da16 PB |
4389 | } |
4390 | ||
a55eb92c JK |
4391 | void bdrv_reset_dirty(BlockDriverState *bs, int64_t cur_sector, |
4392 | int nr_sectors) | |
7cd1e32a | 4393 | { |
8f0720ec | 4394 | hbitmap_reset(bs->dirty_bitmap, cur_sector, nr_sectors); |
7cd1e32a | 4395 | } |
aaa0eb75 LS |
4396 | |
4397 | int64_t bdrv_get_dirty_count(BlockDriverState *bs) | |
4398 | { | |
8f0720ec | 4399 | if (bs->dirty_bitmap) { |
acc906c6 | 4400 | return hbitmap_count(bs->dirty_bitmap); |
8f0720ec PB |
4401 | } else { |
4402 | return 0; | |
4403 | } | |
aaa0eb75 | 4404 | } |
f88e1a42 | 4405 | |
9fcb0251 FZ |
4406 | /* Get a reference to bs */ |
4407 | void bdrv_ref(BlockDriverState *bs) | |
4408 | { | |
4409 | bs->refcnt++; | |
4410 | } | |
4411 | ||
4412 | /* Release a previously grabbed reference to bs. | |
4413 | * If after releasing, reference count is zero, the BlockDriverState is | |
4414 | * deleted. */ | |
4415 | void bdrv_unref(BlockDriverState *bs) | |
4416 | { | |
4417 | assert(bs->refcnt > 0); | |
4418 | if (--bs->refcnt == 0) { | |
4419 | bdrv_delete(bs); | |
4420 | } | |
4421 | } | |
4422 | ||
db593f25 MT |
4423 | void bdrv_set_in_use(BlockDriverState *bs, int in_use) |
4424 | { | |
4425 | assert(bs->in_use != in_use); | |
4426 | bs->in_use = in_use; | |
4427 | } | |
4428 | ||
4429 | int bdrv_in_use(BlockDriverState *bs) | |
4430 | { | |
4431 | return bs->in_use; | |
4432 | } | |
4433 | ||
28a7282a LC |
4434 | void bdrv_iostatus_enable(BlockDriverState *bs) |
4435 | { | |
d6bf279e | 4436 | bs->iostatus_enabled = true; |
58e21ef5 | 4437 | bs->iostatus = BLOCK_DEVICE_IO_STATUS_OK; |
28a7282a LC |
4438 | } |
4439 | ||
4440 | /* The I/O status is only enabled if the drive explicitly | |
4441 | * enables it _and_ the VM is configured to stop on errors */ | |
4442 | bool bdrv_iostatus_is_enabled(const BlockDriverState *bs) | |
4443 | { | |
d6bf279e | 4444 | return (bs->iostatus_enabled && |
92aa5c6d PB |
4445 | (bs->on_write_error == BLOCKDEV_ON_ERROR_ENOSPC || |
4446 | bs->on_write_error == BLOCKDEV_ON_ERROR_STOP || | |
4447 | bs->on_read_error == BLOCKDEV_ON_ERROR_STOP)); | |
28a7282a LC |
4448 | } |
4449 | ||
4450 | void bdrv_iostatus_disable(BlockDriverState *bs) | |
4451 | { | |
d6bf279e | 4452 | bs->iostatus_enabled = false; |
28a7282a LC |
4453 | } |
4454 | ||
4455 | void bdrv_iostatus_reset(BlockDriverState *bs) | |
4456 | { | |
4457 | if (bdrv_iostatus_is_enabled(bs)) { | |
58e21ef5 | 4458 | bs->iostatus = BLOCK_DEVICE_IO_STATUS_OK; |
3bd293c3 PB |
4459 | if (bs->job) { |
4460 | block_job_iostatus_reset(bs->job); | |
4461 | } | |
28a7282a LC |
4462 | } |
4463 | } | |
4464 | ||
28a7282a LC |
4465 | void bdrv_iostatus_set_err(BlockDriverState *bs, int error) |
4466 | { | |
3e1caa5f PB |
4467 | assert(bdrv_iostatus_is_enabled(bs)); |
4468 | if (bs->iostatus == BLOCK_DEVICE_IO_STATUS_OK) { | |
58e21ef5 LC |
4469 | bs->iostatus = error == ENOSPC ? BLOCK_DEVICE_IO_STATUS_NOSPACE : |
4470 | BLOCK_DEVICE_IO_STATUS_FAILED; | |
28a7282a LC |
4471 | } |
4472 | } | |
4473 | ||
a597e79c CH |
4474 | void |
4475 | bdrv_acct_start(BlockDriverState *bs, BlockAcctCookie *cookie, int64_t bytes, | |
4476 | enum BlockAcctType type) | |
4477 | { | |
4478 | assert(type < BDRV_MAX_IOTYPE); | |
4479 | ||
4480 | cookie->bytes = bytes; | |
c488c7f6 | 4481 | cookie->start_time_ns = get_clock(); |
a597e79c CH |
4482 | cookie->type = type; |
4483 | } | |
4484 | ||
4485 | void | |
4486 | bdrv_acct_done(BlockDriverState *bs, BlockAcctCookie *cookie) | |
4487 | { | |
4488 | assert(cookie->type < BDRV_MAX_IOTYPE); | |
4489 | ||
4490 | bs->nr_bytes[cookie->type] += cookie->bytes; | |
4491 | bs->nr_ops[cookie->type]++; | |
c488c7f6 | 4492 | bs->total_time_ns[cookie->type] += get_clock() - cookie->start_time_ns; |
a597e79c CH |
4493 | } |
4494 | ||
d92ada22 LC |
4495 | void bdrv_img_create(const char *filename, const char *fmt, |
4496 | const char *base_filename, const char *base_fmt, | |
f382d43a MR |
4497 | char *options, uint64_t img_size, int flags, |
4498 | Error **errp, bool quiet) | |
f88e1a42 JS |
4499 | { |
4500 | QEMUOptionParameter *param = NULL, *create_options = NULL; | |
d220894e | 4501 | QEMUOptionParameter *backing_fmt, *backing_file, *size; |
f88e1a42 JS |
4502 | BlockDriverState *bs = NULL; |
4503 | BlockDriver *drv, *proto_drv; | |
96df67d1 | 4504 | BlockDriver *backing_drv = NULL; |
cc84d90f | 4505 | Error *local_err = NULL; |
f88e1a42 JS |
4506 | int ret = 0; |
4507 | ||
4508 | /* Find driver and parse its options */ | |
4509 | drv = bdrv_find_format(fmt); | |
4510 | if (!drv) { | |
71c79813 | 4511 | error_setg(errp, "Unknown file format '%s'", fmt); |
d92ada22 | 4512 | return; |
f88e1a42 JS |
4513 | } |
4514 | ||
98289620 | 4515 | proto_drv = bdrv_find_protocol(filename, true); |
f88e1a42 | 4516 | if (!proto_drv) { |
71c79813 | 4517 | error_setg(errp, "Unknown protocol '%s'", filename); |
d92ada22 | 4518 | return; |
f88e1a42 JS |
4519 | } |
4520 | ||
4521 | create_options = append_option_parameters(create_options, | |
4522 | drv->create_options); | |
4523 | create_options = append_option_parameters(create_options, | |
4524 | proto_drv->create_options); | |
4525 | ||
4526 | /* Create parameter list with default values */ | |
4527 | param = parse_option_parameters("", create_options, param); | |
4528 | ||
4529 | set_option_parameter_int(param, BLOCK_OPT_SIZE, img_size); | |
4530 | ||
4531 | /* Parse -o options */ | |
4532 | if (options) { | |
4533 | param = parse_option_parameters(options, create_options, param); | |
4534 | if (param == NULL) { | |
71c79813 | 4535 | error_setg(errp, "Invalid options for file format '%s'.", fmt); |
f88e1a42 JS |
4536 | goto out; |
4537 | } | |
4538 | } | |
4539 | ||
4540 | if (base_filename) { | |
4541 | if (set_option_parameter(param, BLOCK_OPT_BACKING_FILE, | |
4542 | base_filename)) { | |
71c79813 LC |
4543 | error_setg(errp, "Backing file not supported for file format '%s'", |
4544 | fmt); | |
f88e1a42 JS |
4545 | goto out; |
4546 | } | |
4547 | } | |
4548 | ||
4549 | if (base_fmt) { | |
4550 | if (set_option_parameter(param, BLOCK_OPT_BACKING_FMT, base_fmt)) { | |
71c79813 LC |
4551 | error_setg(errp, "Backing file format not supported for file " |
4552 | "format '%s'", fmt); | |
f88e1a42 JS |
4553 | goto out; |
4554 | } | |
4555 | } | |
4556 | ||
792da93a JS |
4557 | backing_file = get_option_parameter(param, BLOCK_OPT_BACKING_FILE); |
4558 | if (backing_file && backing_file->value.s) { | |
4559 | if (!strcmp(filename, backing_file->value.s)) { | |
71c79813 LC |
4560 | error_setg(errp, "Error: Trying to create an image with the " |
4561 | "same filename as the backing file"); | |
792da93a JS |
4562 | goto out; |
4563 | } | |
4564 | } | |
4565 | ||
f88e1a42 JS |
4566 | backing_fmt = get_option_parameter(param, BLOCK_OPT_BACKING_FMT); |
4567 | if (backing_fmt && backing_fmt->value.s) { | |
96df67d1 SH |
4568 | backing_drv = bdrv_find_format(backing_fmt->value.s); |
4569 | if (!backing_drv) { | |
71c79813 LC |
4570 | error_setg(errp, "Unknown backing file format '%s'", |
4571 | backing_fmt->value.s); | |
f88e1a42 JS |
4572 | goto out; |
4573 | } | |
4574 | } | |
4575 | ||
4576 | // The size for the image must always be specified, with one exception: | |
4577 | // If we are using a backing file, we can obtain the size from there | |
d220894e KW |
4578 | size = get_option_parameter(param, BLOCK_OPT_SIZE); |
4579 | if (size && size->value.n == -1) { | |
f88e1a42 JS |
4580 | if (backing_file && backing_file->value.s) { |
4581 | uint64_t size; | |
f88e1a42 | 4582 | char buf[32]; |
63090dac PB |
4583 | int back_flags; |
4584 | ||
4585 | /* backing files always opened read-only */ | |
4586 | back_flags = | |
4587 | flags & ~(BDRV_O_RDWR | BDRV_O_SNAPSHOT | BDRV_O_NO_BACKING); | |
f88e1a42 | 4588 | |
f88e1a42 JS |
4589 | bs = bdrv_new(""); |
4590 | ||
de9c0cec | 4591 | ret = bdrv_open(bs, backing_file->value.s, NULL, back_flags, |
cc84d90f | 4592 | backing_drv, &local_err); |
f88e1a42 | 4593 | if (ret < 0) { |
cc84d90f MR |
4594 | error_setg_errno(errp, -ret, "Could not open '%s': %s", |
4595 | backing_file->value.s, | |
4596 | error_get_pretty(local_err)); | |
4597 | error_free(local_err); | |
4598 | local_err = NULL; | |
f88e1a42 JS |
4599 | goto out; |
4600 | } | |
4601 | bdrv_get_geometry(bs, &size); | |
4602 | size *= 512; | |
4603 | ||
4604 | snprintf(buf, sizeof(buf), "%" PRId64, size); | |
4605 | set_option_parameter(param, BLOCK_OPT_SIZE, buf); | |
4606 | } else { | |
71c79813 | 4607 | error_setg(errp, "Image creation needs a size parameter"); |
f88e1a42 JS |
4608 | goto out; |
4609 | } | |
4610 | } | |
4611 | ||
f382d43a MR |
4612 | if (!quiet) { |
4613 | printf("Formatting '%s', fmt=%s ", filename, fmt); | |
4614 | print_option_parameters(param); | |
4615 | puts(""); | |
4616 | } | |
cc84d90f MR |
4617 | ret = bdrv_create(drv, filename, param, &local_err); |
4618 | if (ret == -EFBIG) { | |
4619 | /* This is generally a better message than whatever the driver would | |
4620 | * deliver (especially because of the cluster_size_hint), since that | |
4621 | * is most probably not much different from "image too large". */ | |
4622 | const char *cluster_size_hint = ""; | |
4623 | if (get_option_parameter(create_options, BLOCK_OPT_CLUSTER_SIZE)) { | |
4624 | cluster_size_hint = " (try using a larger cluster size)"; | |
f88e1a42 | 4625 | } |
cc84d90f MR |
4626 | error_setg(errp, "The image size is too large for file format '%s'" |
4627 | "%s", fmt, cluster_size_hint); | |
4628 | error_free(local_err); | |
4629 | local_err = NULL; | |
f88e1a42 JS |
4630 | } |
4631 | ||
4632 | out: | |
4633 | free_option_parameters(create_options); | |
4634 | free_option_parameters(param); | |
4635 | ||
4636 | if (bs) { | |
4f6fd349 | 4637 | bdrv_unref(bs); |
f88e1a42 | 4638 | } |
cc84d90f MR |
4639 | if (error_is_set(&local_err)) { |
4640 | error_propagate(errp, local_err); | |
4641 | } | |
f88e1a42 | 4642 | } |
85d126f3 SH |
4643 | |
4644 | AioContext *bdrv_get_aio_context(BlockDriverState *bs) | |
4645 | { | |
4646 | /* Currently BlockDriverState always uses the main loop AioContext */ | |
4647 | return qemu_get_aio_context(); | |
4648 | } | |
d616b224 SH |
4649 | |
4650 | void bdrv_add_before_write_notifier(BlockDriverState *bs, | |
4651 | NotifierWithReturn *notifier) | |
4652 | { | |
4653 | notifier_with_return_list_add(&bs->before_write_notifiers, notifier); | |
4654 | } | |
6f176b48 MR |
4655 | |
4656 | int bdrv_amend_options(BlockDriverState *bs, QEMUOptionParameter *options) | |
4657 | { | |
4658 | if (bs->drv->bdrv_amend_options == NULL) { | |
4659 | return -ENOTSUP; | |
4660 | } | |
4661 | return bs->drv->bdrv_amend_options(bs, options); | |
4662 | } | |
f6186f49 BC |
4663 | |
4664 | ExtSnapshotPerm bdrv_check_ext_snapshot(BlockDriverState *bs) | |
4665 | { | |
4666 | if (bs->drv->bdrv_check_ext_snapshot) { | |
4667 | return bs->drv->bdrv_check_ext_snapshot(bs); | |
4668 | } | |
4669 | ||
4670 | if (bs->file && bs->file->drv && bs->file->drv->bdrv_check_ext_snapshot) { | |
4671 | return bs->file->drv->bdrv_check_ext_snapshot(bs); | |
4672 | } | |
4673 | ||
4674 | /* external snapshots are allowed by default */ | |
4675 | return EXT_SNAPSHOT_ALLOWED; | |
4676 | } | |
4677 | ||
4678 | ExtSnapshotPerm bdrv_check_ext_snapshot_forbidden(BlockDriverState *bs) | |
4679 | { | |
4680 | return EXT_SNAPSHOT_FORBIDDEN; | |
4681 | } |