]>
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]); | |
d7d512f6 | 335 | |
b338082b FB |
336 | return bs; |
337 | } | |
338 | ||
d7d512f6 PB |
339 | void bdrv_add_close_notifier(BlockDriverState *bs, Notifier *notify) |
340 | { | |
341 | notifier_list_add(&bs->close_notifiers, notify); | |
342 | } | |
343 | ||
ea2384d3 FB |
344 | BlockDriver *bdrv_find_format(const char *format_name) |
345 | { | |
346 | BlockDriver *drv1; | |
8a22f02a SH |
347 | QLIST_FOREACH(drv1, &bdrv_drivers, list) { |
348 | if (!strcmp(drv1->format_name, format_name)) { | |
ea2384d3 | 349 | return drv1; |
8a22f02a | 350 | } |
ea2384d3 FB |
351 | } |
352 | return NULL; | |
353 | } | |
354 | ||
b64ec4e4 | 355 | static int bdrv_is_whitelisted(BlockDriver *drv, bool read_only) |
eb852011 | 356 | { |
b64ec4e4 FZ |
357 | static const char *whitelist_rw[] = { |
358 | CONFIG_BDRV_RW_WHITELIST | |
359 | }; | |
360 | static const char *whitelist_ro[] = { | |
361 | CONFIG_BDRV_RO_WHITELIST | |
eb852011 MA |
362 | }; |
363 | const char **p; | |
364 | ||
b64ec4e4 | 365 | if (!whitelist_rw[0] && !whitelist_ro[0]) { |
eb852011 | 366 | return 1; /* no whitelist, anything goes */ |
b64ec4e4 | 367 | } |
eb852011 | 368 | |
b64ec4e4 | 369 | for (p = whitelist_rw; *p; p++) { |
eb852011 MA |
370 | if (!strcmp(drv->format_name, *p)) { |
371 | return 1; | |
372 | } | |
373 | } | |
b64ec4e4 FZ |
374 | if (read_only) { |
375 | for (p = whitelist_ro; *p; p++) { | |
376 | if (!strcmp(drv->format_name, *p)) { | |
377 | return 1; | |
378 | } | |
379 | } | |
380 | } | |
eb852011 MA |
381 | return 0; |
382 | } | |
383 | ||
b64ec4e4 FZ |
384 | BlockDriver *bdrv_find_whitelisted_format(const char *format_name, |
385 | bool read_only) | |
eb852011 MA |
386 | { |
387 | BlockDriver *drv = bdrv_find_format(format_name); | |
b64ec4e4 | 388 | return drv && bdrv_is_whitelisted(drv, read_only) ? drv : NULL; |
eb852011 MA |
389 | } |
390 | ||
5b7e1542 ZYW |
391 | typedef struct CreateCo { |
392 | BlockDriver *drv; | |
393 | char *filename; | |
394 | QEMUOptionParameter *options; | |
395 | int ret; | |
396 | } CreateCo; | |
397 | ||
398 | static void coroutine_fn bdrv_create_co_entry(void *opaque) | |
399 | { | |
400 | CreateCo *cco = opaque; | |
401 | assert(cco->drv); | |
402 | ||
403 | cco->ret = cco->drv->bdrv_create(cco->filename, cco->options); | |
404 | } | |
405 | ||
0e7e1989 KW |
406 | int bdrv_create(BlockDriver *drv, const char* filename, |
407 | QEMUOptionParameter *options) | |
ea2384d3 | 408 | { |
5b7e1542 ZYW |
409 | int ret; |
410 | ||
411 | Coroutine *co; | |
412 | CreateCo cco = { | |
413 | .drv = drv, | |
414 | .filename = g_strdup(filename), | |
415 | .options = options, | |
416 | .ret = NOT_DONE, | |
417 | }; | |
418 | ||
419 | if (!drv->bdrv_create) { | |
80168bff LC |
420 | ret = -ENOTSUP; |
421 | goto out; | |
5b7e1542 ZYW |
422 | } |
423 | ||
424 | if (qemu_in_coroutine()) { | |
425 | /* Fast-path if already in coroutine context */ | |
426 | bdrv_create_co_entry(&cco); | |
427 | } else { | |
428 | co = qemu_coroutine_create(bdrv_create_co_entry); | |
429 | qemu_coroutine_enter(co, &cco); | |
430 | while (cco.ret == NOT_DONE) { | |
431 | qemu_aio_wait(); | |
432 | } | |
433 | } | |
434 | ||
435 | ret = cco.ret; | |
0e7e1989 | 436 | |
80168bff LC |
437 | out: |
438 | g_free(cco.filename); | |
5b7e1542 | 439 | return ret; |
ea2384d3 FB |
440 | } |
441 | ||
84a12e66 CH |
442 | int bdrv_create_file(const char* filename, QEMUOptionParameter *options) |
443 | { | |
444 | BlockDriver *drv; | |
445 | ||
98289620 | 446 | drv = bdrv_find_protocol(filename, true); |
84a12e66 | 447 | if (drv == NULL) { |
16905d71 | 448 | return -ENOENT; |
84a12e66 CH |
449 | } |
450 | ||
451 | return bdrv_create(drv, filename, options); | |
452 | } | |
453 | ||
eba25057 JM |
454 | /* |
455 | * Create a uniquely-named empty temporary file. | |
456 | * Return 0 upon success, otherwise a negative errno value. | |
457 | */ | |
458 | int get_tmp_filename(char *filename, int size) | |
d5249393 | 459 | { |
eba25057 | 460 | #ifdef _WIN32 |
3b9f94e1 | 461 | char temp_dir[MAX_PATH]; |
eba25057 JM |
462 | /* GetTempFileName requires that its output buffer (4th param) |
463 | have length MAX_PATH or greater. */ | |
464 | assert(size >= MAX_PATH); | |
465 | return (GetTempPath(MAX_PATH, temp_dir) | |
466 | && GetTempFileName(temp_dir, "qem", 0, filename) | |
467 | ? 0 : -GetLastError()); | |
d5249393 | 468 | #else |
67b915a5 | 469 | int fd; |
7ccfb2eb | 470 | const char *tmpdir; |
0badc1ee AJ |
471 | tmpdir = getenv("TMPDIR"); |
472 | if (!tmpdir) | |
473 | tmpdir = "/tmp"; | |
eba25057 JM |
474 | if (snprintf(filename, size, "%s/vl.XXXXXX", tmpdir) >= size) { |
475 | return -EOVERFLOW; | |
476 | } | |
ea2384d3 | 477 | fd = mkstemp(filename); |
fe235a06 DH |
478 | if (fd < 0) { |
479 | return -errno; | |
480 | } | |
481 | if (close(fd) != 0) { | |
482 | unlink(filename); | |
eba25057 JM |
483 | return -errno; |
484 | } | |
485 | return 0; | |
d5249393 | 486 | #endif |
eba25057 | 487 | } |
fc01f7e7 | 488 | |
84a12e66 CH |
489 | /* |
490 | * Detect host devices. By convention, /dev/cdrom[N] is always | |
491 | * recognized as a host CDROM. | |
492 | */ | |
493 | static BlockDriver *find_hdev_driver(const char *filename) | |
494 | { | |
495 | int score_max = 0, score; | |
496 | BlockDriver *drv = NULL, *d; | |
497 | ||
498 | QLIST_FOREACH(d, &bdrv_drivers, list) { | |
499 | if (d->bdrv_probe_device) { | |
500 | score = d->bdrv_probe_device(filename); | |
501 | if (score > score_max) { | |
502 | score_max = score; | |
503 | drv = d; | |
504 | } | |
505 | } | |
506 | } | |
507 | ||
508 | return drv; | |
509 | } | |
510 | ||
98289620 KW |
511 | BlockDriver *bdrv_find_protocol(const char *filename, |
512 | bool allow_protocol_prefix) | |
83f64091 FB |
513 | { |
514 | BlockDriver *drv1; | |
515 | char protocol[128]; | |
1cec71e3 | 516 | int len; |
83f64091 | 517 | const char *p; |
19cb3738 | 518 | |
66f82cee KW |
519 | /* TODO Drivers without bdrv_file_open must be specified explicitly */ |
520 | ||
39508e7a CH |
521 | /* |
522 | * XXX(hch): we really should not let host device detection | |
523 | * override an explicit protocol specification, but moving this | |
524 | * later breaks access to device names with colons in them. | |
525 | * Thanks to the brain-dead persistent naming schemes on udev- | |
526 | * based Linux systems those actually are quite common. | |
527 | */ | |
528 | drv1 = find_hdev_driver(filename); | |
529 | if (drv1) { | |
530 | return drv1; | |
531 | } | |
532 | ||
98289620 | 533 | if (!path_has_protocol(filename) || !allow_protocol_prefix) { |
39508e7a | 534 | return bdrv_find_format("file"); |
84a12e66 | 535 | } |
98289620 | 536 | |
9e0b22f4 SH |
537 | p = strchr(filename, ':'); |
538 | assert(p != NULL); | |
1cec71e3 AL |
539 | len = p - filename; |
540 | if (len > sizeof(protocol) - 1) | |
541 | len = sizeof(protocol) - 1; | |
542 | memcpy(protocol, filename, len); | |
543 | protocol[len] = '\0'; | |
8a22f02a | 544 | QLIST_FOREACH(drv1, &bdrv_drivers, list) { |
5fafdf24 | 545 | if (drv1->protocol_name && |
8a22f02a | 546 | !strcmp(drv1->protocol_name, protocol)) { |
83f64091 | 547 | return drv1; |
8a22f02a | 548 | } |
83f64091 FB |
549 | } |
550 | return NULL; | |
551 | } | |
552 | ||
f500a6d3 KW |
553 | static int find_image_format(BlockDriverState *bs, const char *filename, |
554 | BlockDriver **pdrv) | |
f3a5d3f8 | 555 | { |
f500a6d3 | 556 | int score, score_max; |
f3a5d3f8 CH |
557 | BlockDriver *drv1, *drv; |
558 | uint8_t buf[2048]; | |
f500a6d3 | 559 | int ret = 0; |
f8ea0b00 | 560 | |
08a00559 | 561 | /* Return the raw BlockDriver * to scsi-generic devices or empty drives */ |
8e895599 | 562 | if (bs->sg || !bdrv_is_inserted(bs) || bdrv_getlength(bs) == 0) { |
c98ac35d SW |
563 | drv = bdrv_find_format("raw"); |
564 | if (!drv) { | |
565 | ret = -ENOENT; | |
566 | } | |
567 | *pdrv = drv; | |
568 | return ret; | |
1a396859 | 569 | } |
f8ea0b00 | 570 | |
83f64091 | 571 | ret = bdrv_pread(bs, 0, buf, sizeof(buf)); |
83f64091 | 572 | if (ret < 0) { |
c98ac35d SW |
573 | *pdrv = NULL; |
574 | return ret; | |
83f64091 FB |
575 | } |
576 | ||
ea2384d3 | 577 | score_max = 0; |
84a12e66 | 578 | drv = NULL; |
8a22f02a | 579 | QLIST_FOREACH(drv1, &bdrv_drivers, list) { |
83f64091 FB |
580 | if (drv1->bdrv_probe) { |
581 | score = drv1->bdrv_probe(buf, ret, filename); | |
582 | if (score > score_max) { | |
583 | score_max = score; | |
584 | drv = drv1; | |
585 | } | |
0849bf08 | 586 | } |
fc01f7e7 | 587 | } |
c98ac35d SW |
588 | if (!drv) { |
589 | ret = -ENOENT; | |
590 | } | |
591 | *pdrv = drv; | |
592 | return ret; | |
ea2384d3 FB |
593 | } |
594 | ||
51762288 SH |
595 | /** |
596 | * Set the current 'total_sectors' value | |
597 | */ | |
598 | static int refresh_total_sectors(BlockDriverState *bs, int64_t hint) | |
599 | { | |
600 | BlockDriver *drv = bs->drv; | |
601 | ||
396759ad NB |
602 | /* Do not attempt drv->bdrv_getlength() on scsi-generic devices */ |
603 | if (bs->sg) | |
604 | return 0; | |
605 | ||
51762288 SH |
606 | /* query actual device if possible, otherwise just trust the hint */ |
607 | if (drv->bdrv_getlength) { | |
608 | int64_t length = drv->bdrv_getlength(bs); | |
609 | if (length < 0) { | |
610 | return length; | |
611 | } | |
612 | hint = length >> BDRV_SECTOR_BITS; | |
613 | } | |
614 | ||
615 | bs->total_sectors = hint; | |
616 | return 0; | |
617 | } | |
618 | ||
9e8f1835 PB |
619 | /** |
620 | * Set open flags for a given discard mode | |
621 | * | |
622 | * Return 0 on success, -1 if the discard mode was invalid. | |
623 | */ | |
624 | int bdrv_parse_discard_flags(const char *mode, int *flags) | |
625 | { | |
626 | *flags &= ~BDRV_O_UNMAP; | |
627 | ||
628 | if (!strcmp(mode, "off") || !strcmp(mode, "ignore")) { | |
629 | /* do nothing */ | |
630 | } else if (!strcmp(mode, "on") || !strcmp(mode, "unmap")) { | |
631 | *flags |= BDRV_O_UNMAP; | |
632 | } else { | |
633 | return -1; | |
634 | } | |
635 | ||
636 | return 0; | |
637 | } | |
638 | ||
c3993cdc SH |
639 | /** |
640 | * Set open flags for a given cache mode | |
641 | * | |
642 | * Return 0 on success, -1 if the cache mode was invalid. | |
643 | */ | |
644 | int bdrv_parse_cache_flags(const char *mode, int *flags) | |
645 | { | |
646 | *flags &= ~BDRV_O_CACHE_MASK; | |
647 | ||
648 | if (!strcmp(mode, "off") || !strcmp(mode, "none")) { | |
649 | *flags |= BDRV_O_NOCACHE | BDRV_O_CACHE_WB; | |
92196b2f SH |
650 | } else if (!strcmp(mode, "directsync")) { |
651 | *flags |= BDRV_O_NOCACHE; | |
c3993cdc SH |
652 | } else if (!strcmp(mode, "writeback")) { |
653 | *flags |= BDRV_O_CACHE_WB; | |
654 | } else if (!strcmp(mode, "unsafe")) { | |
655 | *flags |= BDRV_O_CACHE_WB; | |
656 | *flags |= BDRV_O_NO_FLUSH; | |
657 | } else if (!strcmp(mode, "writethrough")) { | |
658 | /* this is the default */ | |
659 | } else { | |
660 | return -1; | |
661 | } | |
662 | ||
663 | return 0; | |
664 | } | |
665 | ||
53fec9d3 SH |
666 | /** |
667 | * The copy-on-read flag is actually a reference count so multiple users may | |
668 | * use the feature without worrying about clobbering its previous state. | |
669 | * Copy-on-read stays enabled until all users have called to disable it. | |
670 | */ | |
671 | void bdrv_enable_copy_on_read(BlockDriverState *bs) | |
672 | { | |
673 | bs->copy_on_read++; | |
674 | } | |
675 | ||
676 | void bdrv_disable_copy_on_read(BlockDriverState *bs) | |
677 | { | |
678 | assert(bs->copy_on_read > 0); | |
679 | bs->copy_on_read--; | |
680 | } | |
681 | ||
7b272452 KW |
682 | static int bdrv_open_flags(BlockDriverState *bs, int flags) |
683 | { | |
684 | int open_flags = flags | BDRV_O_CACHE_WB; | |
685 | ||
686 | /* | |
687 | * Clear flags that are internal to the block layer before opening the | |
688 | * image. | |
689 | */ | |
690 | open_flags &= ~(BDRV_O_SNAPSHOT | BDRV_O_NO_BACKING); | |
691 | ||
692 | /* | |
693 | * Snapshots should be writable. | |
694 | */ | |
695 | if (bs->is_temporary) { | |
696 | open_flags |= BDRV_O_RDWR; | |
697 | } | |
698 | ||
699 | return open_flags; | |
700 | } | |
701 | ||
57915332 KW |
702 | /* |
703 | * Common part for opening disk images and files | |
b6ad491a KW |
704 | * |
705 | * Removes all processed options from *options. | |
57915332 | 706 | */ |
f500a6d3 | 707 | static int bdrv_open_common(BlockDriverState *bs, BlockDriverState *file, |
035fccdf | 708 | QDict *options, int flags, BlockDriver *drv) |
57915332 KW |
709 | { |
710 | int ret, open_flags; | |
035fccdf | 711 | const char *filename; |
57915332 KW |
712 | |
713 | assert(drv != NULL); | |
6405875c | 714 | assert(bs->file == NULL); |
707ff828 | 715 | assert(options != NULL && bs->options != options); |
57915332 | 716 | |
45673671 KW |
717 | if (file != NULL) { |
718 | filename = file->filename; | |
719 | } else { | |
720 | filename = qdict_get_try_str(options, "filename"); | |
721 | } | |
722 | ||
723 | trace_bdrv_open_common(bs, filename ?: "", flags, drv->format_name); | |
28dcee10 | 724 | |
5d186eb0 KW |
725 | /* bdrv_open() with directly using a protocol as drv. This layer is already |
726 | * opened, so assign it to bs (while file becomes a closed BlockDriverState) | |
727 | * and return immediately. */ | |
728 | if (file != NULL && drv->bdrv_file_open) { | |
729 | bdrv_swap(file, bs); | |
730 | return 0; | |
731 | } | |
732 | ||
57915332 | 733 | bs->open_flags = flags; |
57915332 | 734 | bs->buffer_alignment = 512; |
0d51b4de | 735 | bs->zero_beyond_eof = true; |
b64ec4e4 FZ |
736 | open_flags = bdrv_open_flags(bs, flags); |
737 | bs->read_only = !(open_flags & BDRV_O_RDWR); | |
738 | ||
739 | if (use_bdrv_whitelist && !bdrv_is_whitelisted(drv, bs->read_only)) { | |
740 | return -ENOTSUP; | |
741 | } | |
57915332 | 742 | |
53fec9d3 | 743 | assert(bs->copy_on_read == 0); /* bdrv_new() and bdrv_close() make it so */ |
b64ec4e4 | 744 | if (!bs->read_only && (flags & BDRV_O_COPY_ON_READ)) { |
53fec9d3 SH |
745 | bdrv_enable_copy_on_read(bs); |
746 | } | |
747 | ||
c2ad1b0c KW |
748 | if (filename != NULL) { |
749 | pstrcpy(bs->filename, sizeof(bs->filename), filename); | |
750 | } else { | |
751 | bs->filename[0] = '\0'; | |
752 | } | |
57915332 | 753 | |
57915332 | 754 | bs->drv = drv; |
7267c094 | 755 | bs->opaque = g_malloc0(drv->instance_size); |
57915332 | 756 | |
03f541bd | 757 | bs->enable_write_cache = !!(flags & BDRV_O_CACHE_WB); |
e7c63796 | 758 | |
66f82cee KW |
759 | /* Open the image, either directly or using a protocol */ |
760 | if (drv->bdrv_file_open) { | |
5d186eb0 KW |
761 | assert(file == NULL); |
762 | assert(drv->bdrv_parse_filename || filename != NULL); | |
56d1b4d2 | 763 | ret = drv->bdrv_file_open(bs, options, open_flags); |
f500a6d3 | 764 | } else { |
2af5ef70 KW |
765 | if (file == NULL) { |
766 | qerror_report(ERROR_CLASS_GENERIC_ERROR, "Can't use '%s' as a " | |
767 | "block driver for the protocol level", | |
768 | drv->format_name); | |
769 | ret = -EINVAL; | |
770 | goto free_and_fail; | |
771 | } | |
f500a6d3 | 772 | bs->file = file; |
b6ad491a | 773 | ret = drv->bdrv_open(bs, options, open_flags); |
66f82cee KW |
774 | } |
775 | ||
57915332 KW |
776 | if (ret < 0) { |
777 | goto free_and_fail; | |
778 | } | |
779 | ||
51762288 SH |
780 | ret = refresh_total_sectors(bs, bs->total_sectors); |
781 | if (ret < 0) { | |
782 | goto free_and_fail; | |
57915332 | 783 | } |
51762288 | 784 | |
57915332 KW |
785 | #ifndef _WIN32 |
786 | if (bs->is_temporary) { | |
c2ad1b0c | 787 | assert(filename != NULL); |
57915332 KW |
788 | unlink(filename); |
789 | } | |
790 | #endif | |
791 | return 0; | |
792 | ||
793 | free_and_fail: | |
f500a6d3 | 794 | bs->file = NULL; |
7267c094 | 795 | g_free(bs->opaque); |
57915332 KW |
796 | bs->opaque = NULL; |
797 | bs->drv = NULL; | |
798 | return ret; | |
799 | } | |
800 | ||
b6ce07aa KW |
801 | /* |
802 | * Opens a file using a protocol (file, host_device, nbd, ...) | |
787e4a85 KW |
803 | * |
804 | * options is a QDict of options to pass to the block drivers, or NULL for an | |
805 | * empty set of options. The reference to the QDict belongs to the block layer | |
806 | * after the call (even on failure), so if the caller intends to reuse the | |
807 | * dictionary, it needs to use QINCREF() before calling bdrv_file_open. | |
b6ce07aa | 808 | */ |
787e4a85 KW |
809 | int bdrv_file_open(BlockDriverState **pbs, const char *filename, |
810 | QDict *options, int flags) | |
ea2384d3 | 811 | { |
83f64091 | 812 | BlockDriverState *bs; |
6db95603 | 813 | BlockDriver *drv; |
c2ad1b0c | 814 | const char *drvname; |
98289620 | 815 | bool allow_protocol_prefix = false; |
83f64091 FB |
816 | int ret; |
817 | ||
707ff828 KW |
818 | /* NULL means an empty set of options */ |
819 | if (options == NULL) { | |
820 | options = qdict_new(); | |
821 | } | |
822 | ||
83f64091 | 823 | bs = bdrv_new(""); |
707ff828 KW |
824 | bs->options = options; |
825 | options = qdict_clone_shallow(options); | |
826 | ||
035fccdf KW |
827 | /* Fetch the file name from the options QDict if necessary */ |
828 | if (!filename) { | |
829 | filename = qdict_get_try_str(options, "filename"); | |
830 | } else if (filename && !qdict_haskey(options, "filename")) { | |
831 | qdict_put(options, "filename", qstring_from_str(filename)); | |
98289620 | 832 | allow_protocol_prefix = true; |
035fccdf KW |
833 | } else { |
834 | qerror_report(ERROR_CLASS_GENERIC_ERROR, "Can't specify 'file' and " | |
835 | "'filename' options at the same time"); | |
836 | ret = -EINVAL; | |
837 | goto fail; | |
838 | } | |
839 | ||
c2ad1b0c KW |
840 | /* Find the right block driver */ |
841 | drvname = qdict_get_try_str(options, "driver"); | |
842 | if (drvname) { | |
b64ec4e4 | 843 | drv = bdrv_find_whitelisted_format(drvname, !(flags & BDRV_O_RDWR)); |
c2ad1b0c KW |
844 | qdict_del(options, "driver"); |
845 | } else if (filename) { | |
98289620 KW |
846 | drv = bdrv_find_protocol(filename, allow_protocol_prefix); |
847 | if (!drv) { | |
848 | qerror_report(ERROR_CLASS_GENERIC_ERROR, "Unknown protocol"); | |
849 | } | |
c2ad1b0c KW |
850 | } else { |
851 | qerror_report(ERROR_CLASS_GENERIC_ERROR, | |
852 | "Must specify either driver or file"); | |
853 | drv = NULL; | |
854 | } | |
855 | ||
856 | if (!drv) { | |
857 | ret = -ENOENT; | |
858 | goto fail; | |
859 | } | |
860 | ||
861 | /* Parse the filename and open it */ | |
862 | if (drv->bdrv_parse_filename && filename) { | |
6963a30d KW |
863 | Error *local_err = NULL; |
864 | drv->bdrv_parse_filename(filename, options, &local_err); | |
865 | if (error_is_set(&local_err)) { | |
866 | qerror_report_err(local_err); | |
867 | error_free(local_err); | |
868 | ret = -EINVAL; | |
869 | goto fail; | |
870 | } | |
56d1b4d2 | 871 | qdict_del(options, "filename"); |
c2ad1b0c KW |
872 | } else if (!drv->bdrv_parse_filename && !filename) { |
873 | qerror_report(ERROR_CLASS_GENERIC_ERROR, | |
874 | "The '%s' block driver requires a file name", | |
875 | drv->format_name); | |
876 | ret = -EINVAL; | |
877 | goto fail; | |
6963a30d KW |
878 | } |
879 | ||
035fccdf | 880 | ret = bdrv_open_common(bs, NULL, options, flags, drv); |
83f64091 | 881 | if (ret < 0) { |
707ff828 KW |
882 | goto fail; |
883 | } | |
884 | ||
885 | /* Check if any unknown options were used */ | |
886 | if (qdict_size(options) != 0) { | |
887 | const QDictEntry *entry = qdict_first(options); | |
888 | qerror_report(ERROR_CLASS_GENERIC_ERROR, "Block protocol '%s' doesn't " | |
889 | "support the option '%s'", | |
890 | drv->format_name, entry->key); | |
891 | ret = -EINVAL; | |
892 | goto fail; | |
3b0d4f61 | 893 | } |
707ff828 KW |
894 | QDECREF(options); |
895 | ||
71d0770c | 896 | bs->growable = 1; |
83f64091 FB |
897 | *pbs = bs; |
898 | return 0; | |
707ff828 KW |
899 | |
900 | fail: | |
901 | QDECREF(options); | |
902 | if (!bs->drv) { | |
903 | QDECREF(bs->options); | |
904 | } | |
905 | bdrv_delete(bs); | |
906 | return ret; | |
83f64091 FB |
907 | } |
908 | ||
31ca6d07 KW |
909 | /* |
910 | * Opens the backing file for a BlockDriverState if not yet open | |
911 | * | |
912 | * options is a QDict of options to pass to the block drivers, or NULL for an | |
913 | * empty set of options. The reference to the QDict is transferred to this | |
914 | * function (even on failure), so if the caller intends to reuse the dictionary, | |
915 | * it needs to use QINCREF() before calling bdrv_file_open. | |
916 | */ | |
917 | int bdrv_open_backing_file(BlockDriverState *bs, QDict *options) | |
9156df12 PB |
918 | { |
919 | char backing_filename[PATH_MAX]; | |
920 | int back_flags, ret; | |
921 | BlockDriver *back_drv = NULL; | |
922 | ||
923 | if (bs->backing_hd != NULL) { | |
31ca6d07 | 924 | QDECREF(options); |
9156df12 PB |
925 | return 0; |
926 | } | |
927 | ||
31ca6d07 KW |
928 | /* NULL means an empty set of options */ |
929 | if (options == NULL) { | |
930 | options = qdict_new(); | |
931 | } | |
932 | ||
9156df12 | 933 | bs->open_flags &= ~BDRV_O_NO_BACKING; |
1cb6f506 KW |
934 | if (qdict_haskey(options, "file.filename")) { |
935 | backing_filename[0] = '\0'; | |
936 | } else if (bs->backing_file[0] == '\0' && qdict_size(options) == 0) { | |
31ca6d07 | 937 | QDECREF(options); |
9156df12 PB |
938 | return 0; |
939 | } | |
940 | ||
941 | bs->backing_hd = bdrv_new(""); | |
942 | bdrv_get_full_backing_filename(bs, backing_filename, | |
943 | sizeof(backing_filename)); | |
944 | ||
945 | if (bs->backing_format[0] != '\0') { | |
946 | back_drv = bdrv_find_format(bs->backing_format); | |
947 | } | |
948 | ||
949 | /* backing files always opened read-only */ | |
950 | back_flags = bs->open_flags & ~(BDRV_O_RDWR | BDRV_O_SNAPSHOT); | |
951 | ||
31ca6d07 KW |
952 | ret = bdrv_open(bs->backing_hd, |
953 | *backing_filename ? backing_filename : NULL, options, | |
de9c0cec | 954 | back_flags, back_drv); |
9156df12 PB |
955 | if (ret < 0) { |
956 | bdrv_delete(bs->backing_hd); | |
957 | bs->backing_hd = NULL; | |
958 | bs->open_flags |= BDRV_O_NO_BACKING; | |
959 | return ret; | |
960 | } | |
961 | return 0; | |
962 | } | |
963 | ||
707ff828 KW |
964 | static void extract_subqdict(QDict *src, QDict **dst, const char *start) |
965 | { | |
966 | const QDictEntry *entry, *next; | |
967 | const char *p; | |
968 | ||
969 | *dst = qdict_new(); | |
970 | entry = qdict_first(src); | |
971 | ||
972 | while (entry != NULL) { | |
973 | next = qdict_next(src, entry); | |
974 | if (strstart(entry->key, start, &p)) { | |
975 | qobject_incref(entry->value); | |
976 | qdict_put_obj(*dst, p, entry->value); | |
977 | qdict_del(src, entry->key); | |
978 | } | |
979 | entry = next; | |
980 | } | |
981 | } | |
982 | ||
b6ce07aa KW |
983 | /* |
984 | * Opens a disk image (raw, qcow2, vmdk, ...) | |
de9c0cec KW |
985 | * |
986 | * options is a QDict of options to pass to the block drivers, or NULL for an | |
987 | * empty set of options. The reference to the QDict belongs to the block layer | |
988 | * after the call (even on failure), so if the caller intends to reuse the | |
989 | * dictionary, it needs to use QINCREF() before calling bdrv_open. | |
b6ce07aa | 990 | */ |
de9c0cec KW |
991 | int bdrv_open(BlockDriverState *bs, const char *filename, QDict *options, |
992 | int flags, BlockDriver *drv) | |
ea2384d3 | 993 | { |
b6ce07aa | 994 | int ret; |
89c9bc3d SW |
995 | /* TODO: extra byte is a hack to ensure MAX_PATH space on Windows. */ |
996 | char tmp_filename[PATH_MAX + 1]; | |
f500a6d3 | 997 | BlockDriverState *file = NULL; |
707ff828 | 998 | QDict *file_options = NULL; |
74fe54f2 | 999 | const char *drvname; |
712e7874 | 1000 | |
de9c0cec KW |
1001 | /* NULL means an empty set of options */ |
1002 | if (options == NULL) { | |
1003 | options = qdict_new(); | |
1004 | } | |
1005 | ||
1006 | bs->options = options; | |
b6ad491a | 1007 | options = qdict_clone_shallow(options); |
de9c0cec KW |
1008 | |
1009 | /* For snapshot=on, create a temporary qcow2 overlay */ | |
83f64091 | 1010 | if (flags & BDRV_O_SNAPSHOT) { |
ea2384d3 FB |
1011 | BlockDriverState *bs1; |
1012 | int64_t total_size; | |
91a073a9 | 1013 | BlockDriver *bdrv_qcow2; |
08b392e1 | 1014 | QEMUOptionParameter *create_options; |
b6ce07aa | 1015 | char backing_filename[PATH_MAX]; |
3b46e624 | 1016 | |
c2ad1b0c KW |
1017 | if (qdict_size(options) != 0) { |
1018 | error_report("Can't use snapshot=on with driver-specific options"); | |
1019 | ret = -EINVAL; | |
1020 | goto fail; | |
1021 | } | |
1022 | assert(filename != NULL); | |
1023 | ||
ea2384d3 FB |
1024 | /* if snapshot, we create a temporary backing file and open it |
1025 | instead of opening 'filename' directly */ | |
33e3963e | 1026 | |
ea2384d3 FB |
1027 | /* if there is a backing file, use it */ |
1028 | bs1 = bdrv_new(""); | |
de9c0cec | 1029 | ret = bdrv_open(bs1, filename, NULL, 0, drv); |
51d7c00c | 1030 | if (ret < 0) { |
ea2384d3 | 1031 | bdrv_delete(bs1); |
de9c0cec | 1032 | goto fail; |
ea2384d3 | 1033 | } |
3e82990b | 1034 | total_size = bdrv_getlength(bs1) & BDRV_SECTOR_MASK; |
7c96d46e | 1035 | |
ea2384d3 | 1036 | bdrv_delete(bs1); |
3b46e624 | 1037 | |
eba25057 JM |
1038 | ret = get_tmp_filename(tmp_filename, sizeof(tmp_filename)); |
1039 | if (ret < 0) { | |
de9c0cec | 1040 | goto fail; |
eba25057 | 1041 | } |
7c96d46e AL |
1042 | |
1043 | /* Real path is meaningless for protocols */ | |
4d70655b | 1044 | if (path_has_protocol(filename)) { |
7c96d46e AL |
1045 | snprintf(backing_filename, sizeof(backing_filename), |
1046 | "%s", filename); | |
de9c0cec KW |
1047 | } else if (!realpath(filename, backing_filename)) { |
1048 | ret = -errno; | |
1049 | goto fail; | |
1050 | } | |
7c96d46e | 1051 | |
91a073a9 | 1052 | bdrv_qcow2 = bdrv_find_format("qcow2"); |
08b392e1 KW |
1053 | create_options = parse_option_parameters("", bdrv_qcow2->create_options, |
1054 | NULL); | |
91a073a9 | 1055 | |
08b392e1 KW |
1056 | set_option_parameter_int(create_options, BLOCK_OPT_SIZE, total_size); |
1057 | set_option_parameter(create_options, BLOCK_OPT_BACKING_FILE, | |
1058 | backing_filename); | |
91a073a9 | 1059 | if (drv) { |
08b392e1 | 1060 | set_option_parameter(create_options, BLOCK_OPT_BACKING_FMT, |
91a073a9 KW |
1061 | drv->format_name); |
1062 | } | |
1063 | ||
08b392e1 KW |
1064 | ret = bdrv_create(bdrv_qcow2, tmp_filename, create_options); |
1065 | free_option_parameters(create_options); | |
51d7c00c | 1066 | if (ret < 0) { |
de9c0cec | 1067 | goto fail; |
ea2384d3 | 1068 | } |
91a073a9 | 1069 | |
ea2384d3 | 1070 | filename = tmp_filename; |
91a073a9 | 1071 | drv = bdrv_qcow2; |
ea2384d3 FB |
1072 | bs->is_temporary = 1; |
1073 | } | |
712e7874 | 1074 | |
f500a6d3 KW |
1075 | /* Open image file without format layer */ |
1076 | if (flags & BDRV_O_RDWR) { | |
1077 | flags |= BDRV_O_ALLOW_RDWR; | |
1078 | } | |
1079 | ||
707ff828 KW |
1080 | extract_subqdict(options, &file_options, "file."); |
1081 | ||
1082 | ret = bdrv_file_open(&file, filename, file_options, | |
50b05b6f | 1083 | bdrv_open_flags(bs, flags | BDRV_O_UNMAP)); |
f500a6d3 | 1084 | if (ret < 0) { |
de9c0cec | 1085 | goto fail; |
f500a6d3 KW |
1086 | } |
1087 | ||
b6ce07aa | 1088 | /* Find the right image format driver */ |
74fe54f2 KW |
1089 | drvname = qdict_get_try_str(options, "driver"); |
1090 | if (drvname) { | |
1091 | drv = bdrv_find_whitelisted_format(drvname, !(flags & BDRV_O_RDWR)); | |
1092 | qdict_del(options, "driver"); | |
1093 | } | |
1094 | ||
6db95603 | 1095 | if (!drv) { |
f500a6d3 | 1096 | ret = find_image_format(file, filename, &drv); |
51d7c00c | 1097 | } |
6987307c | 1098 | |
51d7c00c | 1099 | if (!drv) { |
51d7c00c | 1100 | goto unlink_and_fail; |
ea2384d3 | 1101 | } |
b6ce07aa KW |
1102 | |
1103 | /* Open the image */ | |
035fccdf | 1104 | ret = bdrv_open_common(bs, file, options, flags, drv); |
b6ce07aa | 1105 | if (ret < 0) { |
6987307c CH |
1106 | goto unlink_and_fail; |
1107 | } | |
1108 | ||
f500a6d3 KW |
1109 | if (bs->file != file) { |
1110 | bdrv_delete(file); | |
1111 | file = NULL; | |
1112 | } | |
1113 | ||
b6ce07aa | 1114 | /* If there is a backing file, use it */ |
9156df12 | 1115 | if ((flags & BDRV_O_NO_BACKING) == 0) { |
31ca6d07 KW |
1116 | QDict *backing_options; |
1117 | ||
1118 | extract_subqdict(options, &backing_options, "backing."); | |
1119 | ret = bdrv_open_backing_file(bs, backing_options); | |
b6ce07aa | 1120 | if (ret < 0) { |
b6ad491a | 1121 | goto close_and_fail; |
b6ce07aa | 1122 | } |
b6ce07aa KW |
1123 | } |
1124 | ||
b6ad491a KW |
1125 | /* Check if any unknown options were used */ |
1126 | if (qdict_size(options) != 0) { | |
1127 | const QDictEntry *entry = qdict_first(options); | |
1128 | qerror_report(ERROR_CLASS_GENERIC_ERROR, "Block format '%s' used by " | |
1129 | "device '%s' doesn't support the option '%s'", | |
1130 | drv->format_name, bs->device_name, entry->key); | |
1131 | ||
1132 | ret = -EINVAL; | |
1133 | goto close_and_fail; | |
1134 | } | |
1135 | QDECREF(options); | |
1136 | ||
b6ce07aa | 1137 | if (!bdrv_key_required(bs)) { |
7d4b4ba5 | 1138 | bdrv_dev_change_media_cb(bs, true); |
b6ce07aa KW |
1139 | } |
1140 | ||
1141 | return 0; | |
1142 | ||
1143 | unlink_and_fail: | |
f500a6d3 KW |
1144 | if (file != NULL) { |
1145 | bdrv_delete(file); | |
1146 | } | |
b6ce07aa KW |
1147 | if (bs->is_temporary) { |
1148 | unlink(filename); | |
1149 | } | |
de9c0cec KW |
1150 | fail: |
1151 | QDECREF(bs->options); | |
b6ad491a | 1152 | QDECREF(options); |
de9c0cec | 1153 | bs->options = NULL; |
b6ad491a | 1154 | return ret; |
de9c0cec | 1155 | |
b6ad491a KW |
1156 | close_and_fail: |
1157 | bdrv_close(bs); | |
1158 | QDECREF(options); | |
b6ce07aa KW |
1159 | return ret; |
1160 | } | |
1161 | ||
e971aa12 JC |
1162 | typedef struct BlockReopenQueueEntry { |
1163 | bool prepared; | |
1164 | BDRVReopenState state; | |
1165 | QSIMPLEQ_ENTRY(BlockReopenQueueEntry) entry; | |
1166 | } BlockReopenQueueEntry; | |
1167 | ||
1168 | /* | |
1169 | * Adds a BlockDriverState to a simple queue for an atomic, transactional | |
1170 | * reopen of multiple devices. | |
1171 | * | |
1172 | * bs_queue can either be an existing BlockReopenQueue that has had QSIMPLE_INIT | |
1173 | * already performed, or alternatively may be NULL a new BlockReopenQueue will | |
1174 | * be created and initialized. This newly created BlockReopenQueue should be | |
1175 | * passed back in for subsequent calls that are intended to be of the same | |
1176 | * atomic 'set'. | |
1177 | * | |
1178 | * bs is the BlockDriverState to add to the reopen queue. | |
1179 | * | |
1180 | * flags contains the open flags for the associated bs | |
1181 | * | |
1182 | * returns a pointer to bs_queue, which is either the newly allocated | |
1183 | * bs_queue, or the existing bs_queue being used. | |
1184 | * | |
1185 | */ | |
1186 | BlockReopenQueue *bdrv_reopen_queue(BlockReopenQueue *bs_queue, | |
1187 | BlockDriverState *bs, int flags) | |
1188 | { | |
1189 | assert(bs != NULL); | |
1190 | ||
1191 | BlockReopenQueueEntry *bs_entry; | |
1192 | if (bs_queue == NULL) { | |
1193 | bs_queue = g_new0(BlockReopenQueue, 1); | |
1194 | QSIMPLEQ_INIT(bs_queue); | |
1195 | } | |
1196 | ||
1197 | if (bs->file) { | |
1198 | bdrv_reopen_queue(bs_queue, bs->file, flags); | |
1199 | } | |
1200 | ||
1201 | bs_entry = g_new0(BlockReopenQueueEntry, 1); | |
1202 | QSIMPLEQ_INSERT_TAIL(bs_queue, bs_entry, entry); | |
1203 | ||
1204 | bs_entry->state.bs = bs; | |
1205 | bs_entry->state.flags = flags; | |
1206 | ||
1207 | return bs_queue; | |
1208 | } | |
1209 | ||
1210 | /* | |
1211 | * Reopen multiple BlockDriverStates atomically & transactionally. | |
1212 | * | |
1213 | * The queue passed in (bs_queue) must have been built up previous | |
1214 | * via bdrv_reopen_queue(). | |
1215 | * | |
1216 | * Reopens all BDS specified in the queue, with the appropriate | |
1217 | * flags. All devices are prepared for reopen, and failure of any | |
1218 | * device will cause all device changes to be abandonded, and intermediate | |
1219 | * data cleaned up. | |
1220 | * | |
1221 | * If all devices prepare successfully, then the changes are committed | |
1222 | * to all devices. | |
1223 | * | |
1224 | */ | |
1225 | int bdrv_reopen_multiple(BlockReopenQueue *bs_queue, Error **errp) | |
1226 | { | |
1227 | int ret = -1; | |
1228 | BlockReopenQueueEntry *bs_entry, *next; | |
1229 | Error *local_err = NULL; | |
1230 | ||
1231 | assert(bs_queue != NULL); | |
1232 | ||
1233 | bdrv_drain_all(); | |
1234 | ||
1235 | QSIMPLEQ_FOREACH(bs_entry, bs_queue, entry) { | |
1236 | if (bdrv_reopen_prepare(&bs_entry->state, bs_queue, &local_err)) { | |
1237 | error_propagate(errp, local_err); | |
1238 | goto cleanup; | |
1239 | } | |
1240 | bs_entry->prepared = true; | |
1241 | } | |
1242 | ||
1243 | /* If we reach this point, we have success and just need to apply the | |
1244 | * changes | |
1245 | */ | |
1246 | QSIMPLEQ_FOREACH(bs_entry, bs_queue, entry) { | |
1247 | bdrv_reopen_commit(&bs_entry->state); | |
1248 | } | |
1249 | ||
1250 | ret = 0; | |
1251 | ||
1252 | cleanup: | |
1253 | QSIMPLEQ_FOREACH_SAFE(bs_entry, bs_queue, entry, next) { | |
1254 | if (ret && bs_entry->prepared) { | |
1255 | bdrv_reopen_abort(&bs_entry->state); | |
1256 | } | |
1257 | g_free(bs_entry); | |
1258 | } | |
1259 | g_free(bs_queue); | |
1260 | return ret; | |
1261 | } | |
1262 | ||
1263 | ||
1264 | /* Reopen a single BlockDriverState with the specified flags. */ | |
1265 | int bdrv_reopen(BlockDriverState *bs, int bdrv_flags, Error **errp) | |
1266 | { | |
1267 | int ret = -1; | |
1268 | Error *local_err = NULL; | |
1269 | BlockReopenQueue *queue = bdrv_reopen_queue(NULL, bs, bdrv_flags); | |
1270 | ||
1271 | ret = bdrv_reopen_multiple(queue, &local_err); | |
1272 | if (local_err != NULL) { | |
1273 | error_propagate(errp, local_err); | |
1274 | } | |
1275 | return ret; | |
1276 | } | |
1277 | ||
1278 | ||
1279 | /* | |
1280 | * Prepares a BlockDriverState for reopen. All changes are staged in the | |
1281 | * 'opaque' field of the BDRVReopenState, which is used and allocated by | |
1282 | * the block driver layer .bdrv_reopen_prepare() | |
1283 | * | |
1284 | * bs is the BlockDriverState to reopen | |
1285 | * flags are the new open flags | |
1286 | * queue is the reopen queue | |
1287 | * | |
1288 | * Returns 0 on success, non-zero on error. On error errp will be set | |
1289 | * as well. | |
1290 | * | |
1291 | * On failure, bdrv_reopen_abort() will be called to clean up any data. | |
1292 | * It is the responsibility of the caller to then call the abort() or | |
1293 | * commit() for any other BDS that have been left in a prepare() state | |
1294 | * | |
1295 | */ | |
1296 | int bdrv_reopen_prepare(BDRVReopenState *reopen_state, BlockReopenQueue *queue, | |
1297 | Error **errp) | |
1298 | { | |
1299 | int ret = -1; | |
1300 | Error *local_err = NULL; | |
1301 | BlockDriver *drv; | |
1302 | ||
1303 | assert(reopen_state != NULL); | |
1304 | assert(reopen_state->bs->drv != NULL); | |
1305 | drv = reopen_state->bs->drv; | |
1306 | ||
1307 | /* if we are to stay read-only, do not allow permission change | |
1308 | * to r/w */ | |
1309 | if (!(reopen_state->bs->open_flags & BDRV_O_ALLOW_RDWR) && | |
1310 | reopen_state->flags & BDRV_O_RDWR) { | |
1311 | error_set(errp, QERR_DEVICE_IS_READ_ONLY, | |
1312 | reopen_state->bs->device_name); | |
1313 | goto error; | |
1314 | } | |
1315 | ||
1316 | ||
1317 | ret = bdrv_flush(reopen_state->bs); | |
1318 | if (ret) { | |
1319 | error_set(errp, ERROR_CLASS_GENERIC_ERROR, "Error (%s) flushing drive", | |
1320 | strerror(-ret)); | |
1321 | goto error; | |
1322 | } | |
1323 | ||
1324 | if (drv->bdrv_reopen_prepare) { | |
1325 | ret = drv->bdrv_reopen_prepare(reopen_state, queue, &local_err); | |
1326 | if (ret) { | |
1327 | if (local_err != NULL) { | |
1328 | error_propagate(errp, local_err); | |
1329 | } else { | |
d8b6895f LC |
1330 | error_setg(errp, "failed while preparing to reopen image '%s'", |
1331 | reopen_state->bs->filename); | |
e971aa12 JC |
1332 | } |
1333 | goto error; | |
1334 | } | |
1335 | } else { | |
1336 | /* It is currently mandatory to have a bdrv_reopen_prepare() | |
1337 | * handler for each supported drv. */ | |
1338 | error_set(errp, QERR_BLOCK_FORMAT_FEATURE_NOT_SUPPORTED, | |
1339 | drv->format_name, reopen_state->bs->device_name, | |
1340 | "reopening of file"); | |
1341 | ret = -1; | |
1342 | goto error; | |
1343 | } | |
1344 | ||
1345 | ret = 0; | |
1346 | ||
1347 | error: | |
1348 | return ret; | |
1349 | } | |
1350 | ||
1351 | /* | |
1352 | * Takes the staged changes for the reopen from bdrv_reopen_prepare(), and | |
1353 | * makes them final by swapping the staging BlockDriverState contents into | |
1354 | * the active BlockDriverState contents. | |
1355 | */ | |
1356 | void bdrv_reopen_commit(BDRVReopenState *reopen_state) | |
1357 | { | |
1358 | BlockDriver *drv; | |
1359 | ||
1360 | assert(reopen_state != NULL); | |
1361 | drv = reopen_state->bs->drv; | |
1362 | assert(drv != NULL); | |
1363 | ||
1364 | /* If there are any driver level actions to take */ | |
1365 | if (drv->bdrv_reopen_commit) { | |
1366 | drv->bdrv_reopen_commit(reopen_state); | |
1367 | } | |
1368 | ||
1369 | /* set BDS specific flags now */ | |
1370 | reopen_state->bs->open_flags = reopen_state->flags; | |
1371 | reopen_state->bs->enable_write_cache = !!(reopen_state->flags & | |
1372 | BDRV_O_CACHE_WB); | |
1373 | reopen_state->bs->read_only = !(reopen_state->flags & BDRV_O_RDWR); | |
1374 | } | |
1375 | ||
1376 | /* | |
1377 | * Abort the reopen, and delete and free the staged changes in | |
1378 | * reopen_state | |
1379 | */ | |
1380 | void bdrv_reopen_abort(BDRVReopenState *reopen_state) | |
1381 | { | |
1382 | BlockDriver *drv; | |
1383 | ||
1384 | assert(reopen_state != NULL); | |
1385 | drv = reopen_state->bs->drv; | |
1386 | assert(drv != NULL); | |
1387 | ||
1388 | if (drv->bdrv_reopen_abort) { | |
1389 | drv->bdrv_reopen_abort(reopen_state); | |
1390 | } | |
1391 | } | |
1392 | ||
1393 | ||
fc01f7e7 FB |
1394 | void bdrv_close(BlockDriverState *bs) |
1395 | { | |
3cbc002c PB |
1396 | if (bs->job) { |
1397 | block_job_cancel_sync(bs->job); | |
1398 | } | |
58fda173 SH |
1399 | bdrv_drain_all(); /* complete I/O */ |
1400 | bdrv_flush(bs); | |
1401 | bdrv_drain_all(); /* in case flush left pending I/O */ | |
d7d512f6 | 1402 | notifier_list_notify(&bs->close_notifiers, bs); |
7094f12f | 1403 | |
3cbc002c | 1404 | if (bs->drv) { |
557df6ac | 1405 | if (bs->backing_hd) { |
ea2384d3 | 1406 | bdrv_delete(bs->backing_hd); |
557df6ac SH |
1407 | bs->backing_hd = NULL; |
1408 | } | |
ea2384d3 | 1409 | bs->drv->bdrv_close(bs); |
7267c094 | 1410 | g_free(bs->opaque); |
ea2384d3 FB |
1411 | #ifdef _WIN32 |
1412 | if (bs->is_temporary) { | |
1413 | unlink(bs->filename); | |
1414 | } | |
67b915a5 | 1415 | #endif |
ea2384d3 FB |
1416 | bs->opaque = NULL; |
1417 | bs->drv = NULL; | |
53fec9d3 | 1418 | bs->copy_on_read = 0; |
a275fa42 PB |
1419 | bs->backing_file[0] = '\0'; |
1420 | bs->backing_format[0] = '\0'; | |
6405875c PB |
1421 | bs->total_sectors = 0; |
1422 | bs->encrypted = 0; | |
1423 | bs->valid_key = 0; | |
1424 | bs->sg = 0; | |
1425 | bs->growable = 0; | |
0d51b4de | 1426 | bs->zero_beyond_eof = false; |
de9c0cec KW |
1427 | QDECREF(bs->options); |
1428 | bs->options = NULL; | |
b338082b | 1429 | |
66f82cee | 1430 | if (bs->file != NULL) { |
0ac9377d PB |
1431 | bdrv_delete(bs->file); |
1432 | bs->file = NULL; | |
66f82cee | 1433 | } |
b338082b | 1434 | } |
98f90dba | 1435 | |
9ca11154 PH |
1436 | bdrv_dev_change_media_cb(bs, false); |
1437 | ||
98f90dba ZYW |
1438 | /*throttling disk I/O limits*/ |
1439 | if (bs->io_limits_enabled) { | |
1440 | bdrv_io_limits_disable(bs); | |
1441 | } | |
b338082b FB |
1442 | } |
1443 | ||
2bc93fed MK |
1444 | void bdrv_close_all(void) |
1445 | { | |
1446 | BlockDriverState *bs; | |
1447 | ||
1448 | QTAILQ_FOREACH(bs, &bdrv_states, list) { | |
1449 | bdrv_close(bs); | |
1450 | } | |
1451 | } | |
1452 | ||
88266f5a SH |
1453 | /* Check if any requests are in-flight (including throttled requests) */ |
1454 | static bool bdrv_requests_pending(BlockDriverState *bs) | |
1455 | { | |
1456 | if (!QLIST_EMPTY(&bs->tracked_requests)) { | |
1457 | return true; | |
1458 | } | |
cc0681c4 BC |
1459 | if (!qemu_co_queue_empty(&bs->throttled_reqs[0])) { |
1460 | return true; | |
1461 | } | |
1462 | if (!qemu_co_queue_empty(&bs->throttled_reqs[1])) { | |
88266f5a SH |
1463 | return true; |
1464 | } | |
1465 | if (bs->file && bdrv_requests_pending(bs->file)) { | |
1466 | return true; | |
1467 | } | |
1468 | if (bs->backing_hd && bdrv_requests_pending(bs->backing_hd)) { | |
1469 | return true; | |
1470 | } | |
1471 | return false; | |
1472 | } | |
1473 | ||
1474 | static bool bdrv_requests_pending_all(void) | |
1475 | { | |
1476 | BlockDriverState *bs; | |
1477 | QTAILQ_FOREACH(bs, &bdrv_states, list) { | |
1478 | if (bdrv_requests_pending(bs)) { | |
1479 | return true; | |
1480 | } | |
1481 | } | |
1482 | return false; | |
1483 | } | |
1484 | ||
922453bc SH |
1485 | /* |
1486 | * Wait for pending requests to complete across all BlockDriverStates | |
1487 | * | |
1488 | * This function does not flush data to disk, use bdrv_flush_all() for that | |
1489 | * after calling this function. | |
4c355d53 ZYW |
1490 | * |
1491 | * Note that completion of an asynchronous I/O operation can trigger any | |
1492 | * number of other I/O operations on other devices---for example a coroutine | |
1493 | * can be arbitrarily complex and a constant flow of I/O can come until the | |
1494 | * coroutine is complete. Because of this, it is not possible to have a | |
1495 | * function to drain a single device's I/O queue. | |
922453bc SH |
1496 | */ |
1497 | void bdrv_drain_all(void) | |
1498 | { | |
88266f5a SH |
1499 | /* Always run first iteration so any pending completion BHs run */ |
1500 | bool busy = true; | |
922453bc SH |
1501 | BlockDriverState *bs; |
1502 | ||
88266f5a | 1503 | while (busy) { |
4c355d53 ZYW |
1504 | /* FIXME: We do not have timer support here, so this is effectively |
1505 | * a busy wait. | |
1506 | */ | |
1507 | QTAILQ_FOREACH(bs, &bdrv_states, list) { | |
cc0681c4 | 1508 | if (bdrv_start_throttled_reqs(bs)) { |
4c355d53 ZYW |
1509 | busy = true; |
1510 | } | |
1511 | } | |
922453bc | 1512 | |
88266f5a SH |
1513 | busy = bdrv_requests_pending_all(); |
1514 | busy |= aio_poll(qemu_get_aio_context(), busy); | |
922453bc SH |
1515 | } |
1516 | } | |
1517 | ||
d22b2f41 RH |
1518 | /* make a BlockDriverState anonymous by removing from bdrv_state list. |
1519 | Also, NULL terminate the device_name to prevent double remove */ | |
1520 | void bdrv_make_anon(BlockDriverState *bs) | |
1521 | { | |
1522 | if (bs->device_name[0] != '\0') { | |
1523 | QTAILQ_REMOVE(&bdrv_states, bs, list); | |
1524 | } | |
1525 | bs->device_name[0] = '\0'; | |
1526 | } | |
1527 | ||
e023b2e2 PB |
1528 | static void bdrv_rebind(BlockDriverState *bs) |
1529 | { | |
1530 | if (bs->drv && bs->drv->bdrv_rebind) { | |
1531 | bs->drv->bdrv_rebind(bs); | |
1532 | } | |
1533 | } | |
1534 | ||
4ddc07ca PB |
1535 | static void bdrv_move_feature_fields(BlockDriverState *bs_dest, |
1536 | BlockDriverState *bs_src) | |
8802d1fd | 1537 | { |
4ddc07ca PB |
1538 | /* move some fields that need to stay attached to the device */ |
1539 | bs_dest->open_flags = bs_src->open_flags; | |
8802d1fd JC |
1540 | |
1541 | /* dev info */ | |
4ddc07ca PB |
1542 | bs_dest->dev_ops = bs_src->dev_ops; |
1543 | bs_dest->dev_opaque = bs_src->dev_opaque; | |
1544 | bs_dest->dev = bs_src->dev; | |
1545 | bs_dest->buffer_alignment = bs_src->buffer_alignment; | |
1546 | bs_dest->copy_on_read = bs_src->copy_on_read; | |
8802d1fd | 1547 | |
4ddc07ca | 1548 | bs_dest->enable_write_cache = bs_src->enable_write_cache; |
c4a248a1 | 1549 | |
cc0681c4 BC |
1550 | /* i/o throttled req */ |
1551 | memcpy(&bs_dest->throttle_state, | |
1552 | &bs_src->throttle_state, | |
1553 | sizeof(ThrottleState)); | |
1554 | bs_dest->throttled_reqs[0] = bs_src->throttled_reqs[0]; | |
1555 | bs_dest->throttled_reqs[1] = bs_src->throttled_reqs[1]; | |
4ddc07ca | 1556 | bs_dest->io_limits_enabled = bs_src->io_limits_enabled; |
8802d1fd | 1557 | |
8802d1fd | 1558 | /* r/w error */ |
4ddc07ca PB |
1559 | bs_dest->on_read_error = bs_src->on_read_error; |
1560 | bs_dest->on_write_error = bs_src->on_write_error; | |
8802d1fd JC |
1561 | |
1562 | /* i/o status */ | |
4ddc07ca PB |
1563 | bs_dest->iostatus_enabled = bs_src->iostatus_enabled; |
1564 | bs_dest->iostatus = bs_src->iostatus; | |
8802d1fd | 1565 | |
a9fc4408 | 1566 | /* dirty bitmap */ |
4ddc07ca | 1567 | bs_dest->dirty_bitmap = bs_src->dirty_bitmap; |
a9fc4408 PB |
1568 | |
1569 | /* job */ | |
4ddc07ca PB |
1570 | bs_dest->in_use = bs_src->in_use; |
1571 | bs_dest->job = bs_src->job; | |
a9fc4408 | 1572 | |
8802d1fd | 1573 | /* keep the same entry in bdrv_states */ |
4ddc07ca PB |
1574 | pstrcpy(bs_dest->device_name, sizeof(bs_dest->device_name), |
1575 | bs_src->device_name); | |
1576 | bs_dest->list = bs_src->list; | |
1577 | } | |
8802d1fd | 1578 | |
4ddc07ca PB |
1579 | /* |
1580 | * Swap bs contents for two image chains while they are live, | |
1581 | * while keeping required fields on the BlockDriverState that is | |
1582 | * actually attached to a device. | |
1583 | * | |
1584 | * This will modify the BlockDriverState fields, and swap contents | |
1585 | * between bs_new and bs_old. Both bs_new and bs_old are modified. | |
1586 | * | |
1587 | * bs_new is required to be anonymous. | |
1588 | * | |
1589 | * This function does not create any image files. | |
1590 | */ | |
1591 | void bdrv_swap(BlockDriverState *bs_new, BlockDriverState *bs_old) | |
1592 | { | |
1593 | BlockDriverState tmp; | |
f6801b83 | 1594 | |
4ddc07ca PB |
1595 | /* bs_new must be anonymous and shouldn't have anything fancy enabled */ |
1596 | assert(bs_new->device_name[0] == '\0'); | |
1597 | assert(bs_new->dirty_bitmap == NULL); | |
1598 | assert(bs_new->job == NULL); | |
1599 | assert(bs_new->dev == NULL); | |
1600 | assert(bs_new->in_use == 0); | |
1601 | assert(bs_new->io_limits_enabled == false); | |
cc0681c4 | 1602 | assert(!throttle_have_timer(&bs_new->throttle_state)); |
8802d1fd | 1603 | |
4ddc07ca PB |
1604 | tmp = *bs_new; |
1605 | *bs_new = *bs_old; | |
1606 | *bs_old = tmp; | |
a9fc4408 | 1607 | |
4ddc07ca PB |
1608 | /* there are some fields that should not be swapped, move them back */ |
1609 | bdrv_move_feature_fields(&tmp, bs_old); | |
1610 | bdrv_move_feature_fields(bs_old, bs_new); | |
1611 | bdrv_move_feature_fields(bs_new, &tmp); | |
8802d1fd | 1612 | |
4ddc07ca PB |
1613 | /* bs_new shouldn't be in bdrv_states even after the swap! */ |
1614 | assert(bs_new->device_name[0] == '\0'); | |
1615 | ||
1616 | /* Check a few fields that should remain attached to the device */ | |
1617 | assert(bs_new->dev == NULL); | |
1618 | assert(bs_new->job == NULL); | |
1619 | assert(bs_new->in_use == 0); | |
1620 | assert(bs_new->io_limits_enabled == false); | |
cc0681c4 | 1621 | assert(!throttle_have_timer(&bs_new->throttle_state)); |
e023b2e2 PB |
1622 | |
1623 | bdrv_rebind(bs_new); | |
4ddc07ca PB |
1624 | bdrv_rebind(bs_old); |
1625 | } | |
1626 | ||
1627 | /* | |
1628 | * Add new bs contents at the top of an image chain while the chain is | |
1629 | * live, while keeping required fields on the top layer. | |
1630 | * | |
1631 | * This will modify the BlockDriverState fields, and swap contents | |
1632 | * between bs_new and bs_top. Both bs_new and bs_top are modified. | |
1633 | * | |
1634 | * bs_new is required to be anonymous. | |
1635 | * | |
1636 | * This function does not create any image files. | |
1637 | */ | |
1638 | void bdrv_append(BlockDriverState *bs_new, BlockDriverState *bs_top) | |
1639 | { | |
1640 | bdrv_swap(bs_new, bs_top); | |
1641 | ||
1642 | /* The contents of 'tmp' will become bs_top, as we are | |
1643 | * swapping bs_new and bs_top contents. */ | |
1644 | bs_top->backing_hd = bs_new; | |
1645 | bs_top->open_flags &= ~BDRV_O_NO_BACKING; | |
1646 | pstrcpy(bs_top->backing_file, sizeof(bs_top->backing_file), | |
1647 | bs_new->filename); | |
1648 | pstrcpy(bs_top->backing_format, sizeof(bs_top->backing_format), | |
1649 | bs_new->drv ? bs_new->drv->format_name : ""); | |
8802d1fd JC |
1650 | } |
1651 | ||
b338082b FB |
1652 | void bdrv_delete(BlockDriverState *bs) |
1653 | { | |
fa879d62 | 1654 | assert(!bs->dev); |
3e914655 PB |
1655 | assert(!bs->job); |
1656 | assert(!bs->in_use); | |
18846dee | 1657 | |
e1b5c52e SH |
1658 | bdrv_close(bs); |
1659 | ||
1b7bdbc1 | 1660 | /* remove from list, if necessary */ |
d22b2f41 | 1661 | bdrv_make_anon(bs); |
34c6f050 | 1662 | |
7267c094 | 1663 | g_free(bs); |
fc01f7e7 FB |
1664 | } |
1665 | ||
fa879d62 MA |
1666 | int bdrv_attach_dev(BlockDriverState *bs, void *dev) |
1667 | /* TODO change to DeviceState *dev when all users are qdevified */ | |
18846dee | 1668 | { |
fa879d62 | 1669 | if (bs->dev) { |
18846dee MA |
1670 | return -EBUSY; |
1671 | } | |
fa879d62 | 1672 | bs->dev = dev; |
28a7282a | 1673 | bdrv_iostatus_reset(bs); |
18846dee MA |
1674 | return 0; |
1675 | } | |
1676 | ||
fa879d62 MA |
1677 | /* TODO qdevified devices don't use this, remove when devices are qdevified */ |
1678 | void bdrv_attach_dev_nofail(BlockDriverState *bs, void *dev) | |
18846dee | 1679 | { |
fa879d62 MA |
1680 | if (bdrv_attach_dev(bs, dev) < 0) { |
1681 | abort(); | |
1682 | } | |
1683 | } | |
1684 | ||
1685 | void bdrv_detach_dev(BlockDriverState *bs, void *dev) | |
1686 | /* TODO change to DeviceState *dev when all users are qdevified */ | |
1687 | { | |
1688 | assert(bs->dev == dev); | |
1689 | bs->dev = NULL; | |
0e49de52 MA |
1690 | bs->dev_ops = NULL; |
1691 | bs->dev_opaque = NULL; | |
29e05f20 | 1692 | bs->buffer_alignment = 512; |
18846dee MA |
1693 | } |
1694 | ||
fa879d62 MA |
1695 | /* TODO change to return DeviceState * when all users are qdevified */ |
1696 | void *bdrv_get_attached_dev(BlockDriverState *bs) | |
18846dee | 1697 | { |
fa879d62 | 1698 | return bs->dev; |
18846dee MA |
1699 | } |
1700 | ||
0e49de52 MA |
1701 | void bdrv_set_dev_ops(BlockDriverState *bs, const BlockDevOps *ops, |
1702 | void *opaque) | |
1703 | { | |
1704 | bs->dev_ops = ops; | |
1705 | bs->dev_opaque = opaque; | |
1706 | } | |
1707 | ||
32c81a4a PB |
1708 | void bdrv_emit_qmp_error_event(const BlockDriverState *bdrv, |
1709 | enum MonitorEvent ev, | |
1710 | BlockErrorAction action, bool is_read) | |
329c0a48 LC |
1711 | { |
1712 | QObject *data; | |
1713 | const char *action_str; | |
1714 | ||
1715 | switch (action) { | |
1716 | case BDRV_ACTION_REPORT: | |
1717 | action_str = "report"; | |
1718 | break; | |
1719 | case BDRV_ACTION_IGNORE: | |
1720 | action_str = "ignore"; | |
1721 | break; | |
1722 | case BDRV_ACTION_STOP: | |
1723 | action_str = "stop"; | |
1724 | break; | |
1725 | default: | |
1726 | abort(); | |
1727 | } | |
1728 | ||
1729 | data = qobject_from_jsonf("{ 'device': %s, 'action': %s, 'operation': %s }", | |
1730 | bdrv->device_name, | |
1731 | action_str, | |
1732 | is_read ? "read" : "write"); | |
32c81a4a | 1733 | monitor_protocol_event(ev, data); |
329c0a48 LC |
1734 | |
1735 | qobject_decref(data); | |
1736 | } | |
1737 | ||
6f382ed2 LC |
1738 | static void bdrv_emit_qmp_eject_event(BlockDriverState *bs, bool ejected) |
1739 | { | |
1740 | QObject *data; | |
1741 | ||
1742 | data = qobject_from_jsonf("{ 'device': %s, 'tray-open': %i }", | |
1743 | bdrv_get_device_name(bs), ejected); | |
1744 | monitor_protocol_event(QEVENT_DEVICE_TRAY_MOVED, data); | |
1745 | ||
1746 | qobject_decref(data); | |
1747 | } | |
1748 | ||
7d4b4ba5 | 1749 | static void bdrv_dev_change_media_cb(BlockDriverState *bs, bool load) |
0e49de52 | 1750 | { |
145feb17 | 1751 | if (bs->dev_ops && bs->dev_ops->change_media_cb) { |
6f382ed2 | 1752 | bool tray_was_closed = !bdrv_dev_is_tray_open(bs); |
7d4b4ba5 | 1753 | bs->dev_ops->change_media_cb(bs->dev_opaque, load); |
6f382ed2 LC |
1754 | if (tray_was_closed) { |
1755 | /* tray open */ | |
1756 | bdrv_emit_qmp_eject_event(bs, true); | |
1757 | } | |
1758 | if (load) { | |
1759 | /* tray close */ | |
1760 | bdrv_emit_qmp_eject_event(bs, false); | |
1761 | } | |
145feb17 MA |
1762 | } |
1763 | } | |
1764 | ||
2c6942fa MA |
1765 | bool bdrv_dev_has_removable_media(BlockDriverState *bs) |
1766 | { | |
1767 | return !bs->dev || (bs->dev_ops && bs->dev_ops->change_media_cb); | |
1768 | } | |
1769 | ||
025ccaa7 PB |
1770 | void bdrv_dev_eject_request(BlockDriverState *bs, bool force) |
1771 | { | |
1772 | if (bs->dev_ops && bs->dev_ops->eject_request_cb) { | |
1773 | bs->dev_ops->eject_request_cb(bs->dev_opaque, force); | |
1774 | } | |
1775 | } | |
1776 | ||
e4def80b MA |
1777 | bool bdrv_dev_is_tray_open(BlockDriverState *bs) |
1778 | { | |
1779 | if (bs->dev_ops && bs->dev_ops->is_tray_open) { | |
1780 | return bs->dev_ops->is_tray_open(bs->dev_opaque); | |
1781 | } | |
1782 | return false; | |
1783 | } | |
1784 | ||
145feb17 MA |
1785 | static void bdrv_dev_resize_cb(BlockDriverState *bs) |
1786 | { | |
1787 | if (bs->dev_ops && bs->dev_ops->resize_cb) { | |
1788 | bs->dev_ops->resize_cb(bs->dev_opaque); | |
0e49de52 MA |
1789 | } |
1790 | } | |
1791 | ||
f107639a MA |
1792 | bool bdrv_dev_is_medium_locked(BlockDriverState *bs) |
1793 | { | |
1794 | if (bs->dev_ops && bs->dev_ops->is_medium_locked) { | |
1795 | return bs->dev_ops->is_medium_locked(bs->dev_opaque); | |
1796 | } | |
1797 | return false; | |
1798 | } | |
1799 | ||
e97fc193 AL |
1800 | /* |
1801 | * Run consistency checks on an image | |
1802 | * | |
e076f338 | 1803 | * Returns 0 if the check could be completed (it doesn't mean that the image is |
a1c7273b | 1804 | * free of errors) or -errno when an internal error occurred. The results of the |
e076f338 | 1805 | * check are stored in res. |
e97fc193 | 1806 | */ |
4534ff54 | 1807 | int bdrv_check(BlockDriverState *bs, BdrvCheckResult *res, BdrvCheckMode fix) |
e97fc193 AL |
1808 | { |
1809 | if (bs->drv->bdrv_check == NULL) { | |
1810 | return -ENOTSUP; | |
1811 | } | |
1812 | ||
e076f338 | 1813 | memset(res, 0, sizeof(*res)); |
4534ff54 | 1814 | return bs->drv->bdrv_check(bs, res, fix); |
e97fc193 AL |
1815 | } |
1816 | ||
8a426614 KW |
1817 | #define COMMIT_BUF_SECTORS 2048 |
1818 | ||
33e3963e FB |
1819 | /* commit COW file into the raw image */ |
1820 | int bdrv_commit(BlockDriverState *bs) | |
1821 | { | |
19cb3738 | 1822 | BlockDriver *drv = bs->drv; |
8a426614 KW |
1823 | int64_t sector, total_sectors; |
1824 | int n, ro, open_flags; | |
0bce597d | 1825 | int ret = 0; |
8a426614 | 1826 | uint8_t *buf; |
c2cba3d9 | 1827 | char filename[PATH_MAX]; |
33e3963e | 1828 | |
19cb3738 FB |
1829 | if (!drv) |
1830 | return -ENOMEDIUM; | |
4dca4b63 NS |
1831 | |
1832 | if (!bs->backing_hd) { | |
1833 | return -ENOTSUP; | |
33e3963e FB |
1834 | } |
1835 | ||
2d3735d3 SH |
1836 | if (bdrv_in_use(bs) || bdrv_in_use(bs->backing_hd)) { |
1837 | return -EBUSY; | |
1838 | } | |
1839 | ||
4dca4b63 | 1840 | ro = bs->backing_hd->read_only; |
c2cba3d9 JM |
1841 | /* Use pstrcpy (not strncpy): filename must be NUL-terminated. */ |
1842 | pstrcpy(filename, sizeof(filename), bs->backing_hd->filename); | |
4dca4b63 NS |
1843 | open_flags = bs->backing_hd->open_flags; |
1844 | ||
1845 | if (ro) { | |
0bce597d JC |
1846 | if (bdrv_reopen(bs->backing_hd, open_flags | BDRV_O_RDWR, NULL)) { |
1847 | return -EACCES; | |
4dca4b63 | 1848 | } |
ea2384d3 | 1849 | } |
33e3963e | 1850 | |
6ea44308 | 1851 | total_sectors = bdrv_getlength(bs) >> BDRV_SECTOR_BITS; |
7267c094 | 1852 | buf = g_malloc(COMMIT_BUF_SECTORS * BDRV_SECTOR_SIZE); |
8a426614 KW |
1853 | |
1854 | for (sector = 0; sector < total_sectors; sector += n) { | |
05c4af54 | 1855 | if (bdrv_is_allocated(bs, sector, COMMIT_BUF_SECTORS, &n)) { |
8a426614 KW |
1856 | |
1857 | if (bdrv_read(bs, sector, buf, n) != 0) { | |
1858 | ret = -EIO; | |
1859 | goto ro_cleanup; | |
1860 | } | |
1861 | ||
1862 | if (bdrv_write(bs->backing_hd, sector, buf, n) != 0) { | |
1863 | ret = -EIO; | |
1864 | goto ro_cleanup; | |
1865 | } | |
ea2384d3 | 1866 | } |
33e3963e | 1867 | } |
95389c86 | 1868 | |
1d44952f CH |
1869 | if (drv->bdrv_make_empty) { |
1870 | ret = drv->bdrv_make_empty(bs); | |
1871 | bdrv_flush(bs); | |
1872 | } | |
95389c86 | 1873 | |
3f5075ae CH |
1874 | /* |
1875 | * Make sure all data we wrote to the backing device is actually | |
1876 | * stable on disk. | |
1877 | */ | |
1878 | if (bs->backing_hd) | |
1879 | bdrv_flush(bs->backing_hd); | |
4dca4b63 NS |
1880 | |
1881 | ro_cleanup: | |
7267c094 | 1882 | g_free(buf); |
4dca4b63 NS |
1883 | |
1884 | if (ro) { | |
0bce597d JC |
1885 | /* ignoring error return here */ |
1886 | bdrv_reopen(bs->backing_hd, open_flags & ~BDRV_O_RDWR, NULL); | |
4dca4b63 NS |
1887 | } |
1888 | ||
1d44952f | 1889 | return ret; |
33e3963e FB |
1890 | } |
1891 | ||
e8877497 | 1892 | int bdrv_commit_all(void) |
6ab4b5ab MA |
1893 | { |
1894 | BlockDriverState *bs; | |
1895 | ||
1896 | QTAILQ_FOREACH(bs, &bdrv_states, list) { | |
272d2d8e JC |
1897 | if (bs->drv && bs->backing_hd) { |
1898 | int ret = bdrv_commit(bs); | |
1899 | if (ret < 0) { | |
1900 | return ret; | |
1901 | } | |
e8877497 | 1902 | } |
6ab4b5ab | 1903 | } |
e8877497 | 1904 | return 0; |
6ab4b5ab MA |
1905 | } |
1906 | ||
dbffbdcf SH |
1907 | /** |
1908 | * Remove an active request from the tracked requests list | |
1909 | * | |
1910 | * This function should be called when a tracked request is completing. | |
1911 | */ | |
1912 | static void tracked_request_end(BdrvTrackedRequest *req) | |
1913 | { | |
1914 | QLIST_REMOVE(req, list); | |
f4658285 | 1915 | qemu_co_queue_restart_all(&req->wait_queue); |
dbffbdcf SH |
1916 | } |
1917 | ||
1918 | /** | |
1919 | * Add an active request to the tracked requests list | |
1920 | */ | |
1921 | static void tracked_request_begin(BdrvTrackedRequest *req, | |
1922 | BlockDriverState *bs, | |
1923 | int64_t sector_num, | |
1924 | int nb_sectors, bool is_write) | |
1925 | { | |
1926 | *req = (BdrvTrackedRequest){ | |
1927 | .bs = bs, | |
1928 | .sector_num = sector_num, | |
1929 | .nb_sectors = nb_sectors, | |
1930 | .is_write = is_write, | |
5f8b6491 | 1931 | .co = qemu_coroutine_self(), |
dbffbdcf SH |
1932 | }; |
1933 | ||
f4658285 SH |
1934 | qemu_co_queue_init(&req->wait_queue); |
1935 | ||
dbffbdcf SH |
1936 | QLIST_INSERT_HEAD(&bs->tracked_requests, req, list); |
1937 | } | |
1938 | ||
d83947ac SH |
1939 | /** |
1940 | * Round a region to cluster boundaries | |
1941 | */ | |
343bded4 PB |
1942 | void bdrv_round_to_clusters(BlockDriverState *bs, |
1943 | int64_t sector_num, int nb_sectors, | |
1944 | int64_t *cluster_sector_num, | |
1945 | int *cluster_nb_sectors) | |
d83947ac SH |
1946 | { |
1947 | BlockDriverInfo bdi; | |
1948 | ||
1949 | if (bdrv_get_info(bs, &bdi) < 0 || bdi.cluster_size == 0) { | |
1950 | *cluster_sector_num = sector_num; | |
1951 | *cluster_nb_sectors = nb_sectors; | |
1952 | } else { | |
1953 | int64_t c = bdi.cluster_size / BDRV_SECTOR_SIZE; | |
1954 | *cluster_sector_num = QEMU_ALIGN_DOWN(sector_num, c); | |
1955 | *cluster_nb_sectors = QEMU_ALIGN_UP(sector_num - *cluster_sector_num + | |
1956 | nb_sectors, c); | |
1957 | } | |
1958 | } | |
1959 | ||
f4658285 SH |
1960 | static bool tracked_request_overlaps(BdrvTrackedRequest *req, |
1961 | int64_t sector_num, int nb_sectors) { | |
d83947ac SH |
1962 | /* aaaa bbbb */ |
1963 | if (sector_num >= req->sector_num + req->nb_sectors) { | |
1964 | return false; | |
1965 | } | |
1966 | /* bbbb aaaa */ | |
1967 | if (req->sector_num >= sector_num + nb_sectors) { | |
1968 | return false; | |
1969 | } | |
1970 | return true; | |
f4658285 SH |
1971 | } |
1972 | ||
1973 | static void coroutine_fn wait_for_overlapping_requests(BlockDriverState *bs, | |
1974 | int64_t sector_num, int nb_sectors) | |
1975 | { | |
1976 | BdrvTrackedRequest *req; | |
d83947ac SH |
1977 | int64_t cluster_sector_num; |
1978 | int cluster_nb_sectors; | |
f4658285 SH |
1979 | bool retry; |
1980 | ||
d83947ac SH |
1981 | /* If we touch the same cluster it counts as an overlap. This guarantees |
1982 | * that allocating writes will be serialized and not race with each other | |
1983 | * for the same cluster. For example, in copy-on-read it ensures that the | |
1984 | * CoR read and write operations are atomic and guest writes cannot | |
1985 | * interleave between them. | |
1986 | */ | |
343bded4 PB |
1987 | bdrv_round_to_clusters(bs, sector_num, nb_sectors, |
1988 | &cluster_sector_num, &cluster_nb_sectors); | |
d83947ac | 1989 | |
f4658285 SH |
1990 | do { |
1991 | retry = false; | |
1992 | QLIST_FOREACH(req, &bs->tracked_requests, list) { | |
d83947ac SH |
1993 | if (tracked_request_overlaps(req, cluster_sector_num, |
1994 | cluster_nb_sectors)) { | |
5f8b6491 SH |
1995 | /* Hitting this means there was a reentrant request, for |
1996 | * example, a block driver issuing nested requests. This must | |
1997 | * never happen since it means deadlock. | |
1998 | */ | |
1999 | assert(qemu_coroutine_self() != req->co); | |
2000 | ||
f4658285 SH |
2001 | qemu_co_queue_wait(&req->wait_queue); |
2002 | retry = true; | |
2003 | break; | |
2004 | } | |
2005 | } | |
2006 | } while (retry); | |
2007 | } | |
2008 | ||
756e6736 KW |
2009 | /* |
2010 | * Return values: | |
2011 | * 0 - success | |
2012 | * -EINVAL - backing format specified, but no file | |
2013 | * -ENOSPC - can't update the backing file because no space is left in the | |
2014 | * image file header | |
2015 | * -ENOTSUP - format driver doesn't support changing the backing file | |
2016 | */ | |
2017 | int bdrv_change_backing_file(BlockDriverState *bs, | |
2018 | const char *backing_file, const char *backing_fmt) | |
2019 | { | |
2020 | BlockDriver *drv = bs->drv; | |
469ef350 | 2021 | int ret; |
756e6736 | 2022 | |
5f377794 PB |
2023 | /* Backing file format doesn't make sense without a backing file */ |
2024 | if (backing_fmt && !backing_file) { | |
2025 | return -EINVAL; | |
2026 | } | |
2027 | ||
756e6736 | 2028 | if (drv->bdrv_change_backing_file != NULL) { |
469ef350 | 2029 | ret = drv->bdrv_change_backing_file(bs, backing_file, backing_fmt); |
756e6736 | 2030 | } else { |
469ef350 | 2031 | ret = -ENOTSUP; |
756e6736 | 2032 | } |
469ef350 PB |
2033 | |
2034 | if (ret == 0) { | |
2035 | pstrcpy(bs->backing_file, sizeof(bs->backing_file), backing_file ?: ""); | |
2036 | pstrcpy(bs->backing_format, sizeof(bs->backing_format), backing_fmt ?: ""); | |
2037 | } | |
2038 | return ret; | |
756e6736 KW |
2039 | } |
2040 | ||
6ebdcee2 JC |
2041 | /* |
2042 | * Finds the image layer in the chain that has 'bs' as its backing file. | |
2043 | * | |
2044 | * active is the current topmost image. | |
2045 | * | |
2046 | * Returns NULL if bs is not found in active's image chain, | |
2047 | * or if active == bs. | |
2048 | */ | |
2049 | BlockDriverState *bdrv_find_overlay(BlockDriverState *active, | |
2050 | BlockDriverState *bs) | |
2051 | { | |
2052 | BlockDriverState *overlay = NULL; | |
2053 | BlockDriverState *intermediate; | |
2054 | ||
2055 | assert(active != NULL); | |
2056 | assert(bs != NULL); | |
2057 | ||
2058 | /* if bs is the same as active, then by definition it has no overlay | |
2059 | */ | |
2060 | if (active == bs) { | |
2061 | return NULL; | |
2062 | } | |
2063 | ||
2064 | intermediate = active; | |
2065 | while (intermediate->backing_hd) { | |
2066 | if (intermediate->backing_hd == bs) { | |
2067 | overlay = intermediate; | |
2068 | break; | |
2069 | } | |
2070 | intermediate = intermediate->backing_hd; | |
2071 | } | |
2072 | ||
2073 | return overlay; | |
2074 | } | |
2075 | ||
2076 | typedef struct BlkIntermediateStates { | |
2077 | BlockDriverState *bs; | |
2078 | QSIMPLEQ_ENTRY(BlkIntermediateStates) entry; | |
2079 | } BlkIntermediateStates; | |
2080 | ||
2081 | ||
2082 | /* | |
2083 | * Drops images above 'base' up to and including 'top', and sets the image | |
2084 | * above 'top' to have base as its backing file. | |
2085 | * | |
2086 | * Requires that the overlay to 'top' is opened r/w, so that the backing file | |
2087 | * information in 'bs' can be properly updated. | |
2088 | * | |
2089 | * E.g., this will convert the following chain: | |
2090 | * bottom <- base <- intermediate <- top <- active | |
2091 | * | |
2092 | * to | |
2093 | * | |
2094 | * bottom <- base <- active | |
2095 | * | |
2096 | * It is allowed for bottom==base, in which case it converts: | |
2097 | * | |
2098 | * base <- intermediate <- top <- active | |
2099 | * | |
2100 | * to | |
2101 | * | |
2102 | * base <- active | |
2103 | * | |
2104 | * Error conditions: | |
2105 | * if active == top, that is considered an error | |
2106 | * | |
2107 | */ | |
2108 | int bdrv_drop_intermediate(BlockDriverState *active, BlockDriverState *top, | |
2109 | BlockDriverState *base) | |
2110 | { | |
2111 | BlockDriverState *intermediate; | |
2112 | BlockDriverState *base_bs = NULL; | |
2113 | BlockDriverState *new_top_bs = NULL; | |
2114 | BlkIntermediateStates *intermediate_state, *next; | |
2115 | int ret = -EIO; | |
2116 | ||
2117 | QSIMPLEQ_HEAD(states_to_delete, BlkIntermediateStates) states_to_delete; | |
2118 | QSIMPLEQ_INIT(&states_to_delete); | |
2119 | ||
2120 | if (!top->drv || !base->drv) { | |
2121 | goto exit; | |
2122 | } | |
2123 | ||
2124 | new_top_bs = bdrv_find_overlay(active, top); | |
2125 | ||
2126 | if (new_top_bs == NULL) { | |
2127 | /* we could not find the image above 'top', this is an error */ | |
2128 | goto exit; | |
2129 | } | |
2130 | ||
2131 | /* special case of new_top_bs->backing_hd already pointing to base - nothing | |
2132 | * to do, no intermediate images */ | |
2133 | if (new_top_bs->backing_hd == base) { | |
2134 | ret = 0; | |
2135 | goto exit; | |
2136 | } | |
2137 | ||
2138 | intermediate = top; | |
2139 | ||
2140 | /* now we will go down through the list, and add each BDS we find | |
2141 | * into our deletion queue, until we hit the 'base' | |
2142 | */ | |
2143 | while (intermediate) { | |
2144 | intermediate_state = g_malloc0(sizeof(BlkIntermediateStates)); | |
2145 | intermediate_state->bs = intermediate; | |
2146 | QSIMPLEQ_INSERT_TAIL(&states_to_delete, intermediate_state, entry); | |
2147 | ||
2148 | if (intermediate->backing_hd == base) { | |
2149 | base_bs = intermediate->backing_hd; | |
2150 | break; | |
2151 | } | |
2152 | intermediate = intermediate->backing_hd; | |
2153 | } | |
2154 | if (base_bs == NULL) { | |
2155 | /* something went wrong, we did not end at the base. safely | |
2156 | * unravel everything, and exit with error */ | |
2157 | goto exit; | |
2158 | } | |
2159 | ||
2160 | /* success - we can delete the intermediate states, and link top->base */ | |
2161 | ret = bdrv_change_backing_file(new_top_bs, base_bs->filename, | |
2162 | base_bs->drv ? base_bs->drv->format_name : ""); | |
2163 | if (ret) { | |
2164 | goto exit; | |
2165 | } | |
2166 | new_top_bs->backing_hd = base_bs; | |
2167 | ||
2168 | ||
2169 | QSIMPLEQ_FOREACH_SAFE(intermediate_state, &states_to_delete, entry, next) { | |
2170 | /* so that bdrv_close() does not recursively close the chain */ | |
2171 | intermediate_state->bs->backing_hd = NULL; | |
2172 | bdrv_delete(intermediate_state->bs); | |
2173 | } | |
2174 | ret = 0; | |
2175 | ||
2176 | exit: | |
2177 | QSIMPLEQ_FOREACH_SAFE(intermediate_state, &states_to_delete, entry, next) { | |
2178 | g_free(intermediate_state); | |
2179 | } | |
2180 | return ret; | |
2181 | } | |
2182 | ||
2183 | ||
71d0770c AL |
2184 | static int bdrv_check_byte_request(BlockDriverState *bs, int64_t offset, |
2185 | size_t size) | |
2186 | { | |
2187 | int64_t len; | |
2188 | ||
2189 | if (!bdrv_is_inserted(bs)) | |
2190 | return -ENOMEDIUM; | |
2191 | ||
2192 | if (bs->growable) | |
2193 | return 0; | |
2194 | ||
2195 | len = bdrv_getlength(bs); | |
2196 | ||
fbb7b4e0 KW |
2197 | if (offset < 0) |
2198 | return -EIO; | |
2199 | ||
2200 | if ((offset > len) || (len - offset < size)) | |
71d0770c AL |
2201 | return -EIO; |
2202 | ||
2203 | return 0; | |
2204 | } | |
2205 | ||
2206 | static int bdrv_check_request(BlockDriverState *bs, int64_t sector_num, | |
2207 | int nb_sectors) | |
2208 | { | |
eb5a3165 JS |
2209 | return bdrv_check_byte_request(bs, sector_num * BDRV_SECTOR_SIZE, |
2210 | nb_sectors * BDRV_SECTOR_SIZE); | |
71d0770c AL |
2211 | } |
2212 | ||
1c9805a3 SH |
2213 | typedef struct RwCo { |
2214 | BlockDriverState *bs; | |
2215 | int64_t sector_num; | |
2216 | int nb_sectors; | |
2217 | QEMUIOVector *qiov; | |
2218 | bool is_write; | |
2219 | int ret; | |
4105eaaa | 2220 | BdrvRequestFlags flags; |
1c9805a3 SH |
2221 | } RwCo; |
2222 | ||
2223 | static void coroutine_fn bdrv_rw_co_entry(void *opaque) | |
fc01f7e7 | 2224 | { |
1c9805a3 | 2225 | RwCo *rwco = opaque; |
ea2384d3 | 2226 | |
1c9805a3 SH |
2227 | if (!rwco->is_write) { |
2228 | rwco->ret = bdrv_co_do_readv(rwco->bs, rwco->sector_num, | |
4105eaaa PL |
2229 | rwco->nb_sectors, rwco->qiov, |
2230 | rwco->flags); | |
1c9805a3 SH |
2231 | } else { |
2232 | rwco->ret = bdrv_co_do_writev(rwco->bs, rwco->sector_num, | |
4105eaaa PL |
2233 | rwco->nb_sectors, rwco->qiov, |
2234 | rwco->flags); | |
1c9805a3 SH |
2235 | } |
2236 | } | |
e7a8a783 | 2237 | |
1c9805a3 | 2238 | /* |
8d3b1a2d | 2239 | * Process a vectored synchronous request using coroutines |
1c9805a3 | 2240 | */ |
8d3b1a2d | 2241 | static int bdrv_rwv_co(BlockDriverState *bs, int64_t sector_num, |
4105eaaa PL |
2242 | QEMUIOVector *qiov, bool is_write, |
2243 | BdrvRequestFlags flags) | |
1c9805a3 | 2244 | { |
1c9805a3 SH |
2245 | Coroutine *co; |
2246 | RwCo rwco = { | |
2247 | .bs = bs, | |
2248 | .sector_num = sector_num, | |
8d3b1a2d KW |
2249 | .nb_sectors = qiov->size >> BDRV_SECTOR_BITS, |
2250 | .qiov = qiov, | |
1c9805a3 SH |
2251 | .is_write = is_write, |
2252 | .ret = NOT_DONE, | |
4105eaaa | 2253 | .flags = flags, |
1c9805a3 | 2254 | }; |
8d3b1a2d | 2255 | assert((qiov->size & (BDRV_SECTOR_SIZE - 1)) == 0); |
e7a8a783 | 2256 | |
498e386c ZYW |
2257 | /** |
2258 | * In sync call context, when the vcpu is blocked, this throttling timer | |
2259 | * will not fire; so the I/O throttling function has to be disabled here | |
2260 | * if it has been enabled. | |
2261 | */ | |
2262 | if (bs->io_limits_enabled) { | |
2263 | fprintf(stderr, "Disabling I/O throttling on '%s' due " | |
2264 | "to synchronous I/O.\n", bdrv_get_device_name(bs)); | |
2265 | bdrv_io_limits_disable(bs); | |
2266 | } | |
2267 | ||
1c9805a3 SH |
2268 | if (qemu_in_coroutine()) { |
2269 | /* Fast-path if already in coroutine context */ | |
2270 | bdrv_rw_co_entry(&rwco); | |
2271 | } else { | |
2272 | co = qemu_coroutine_create(bdrv_rw_co_entry); | |
2273 | qemu_coroutine_enter(co, &rwco); | |
2274 | while (rwco.ret == NOT_DONE) { | |
2275 | qemu_aio_wait(); | |
2276 | } | |
2277 | } | |
2278 | return rwco.ret; | |
2279 | } | |
b338082b | 2280 | |
8d3b1a2d KW |
2281 | /* |
2282 | * Process a synchronous request using coroutines | |
2283 | */ | |
2284 | static int bdrv_rw_co(BlockDriverState *bs, int64_t sector_num, uint8_t *buf, | |
4105eaaa | 2285 | int nb_sectors, bool is_write, BdrvRequestFlags flags) |
8d3b1a2d KW |
2286 | { |
2287 | QEMUIOVector qiov; | |
2288 | struct iovec iov = { | |
2289 | .iov_base = (void *)buf, | |
2290 | .iov_len = nb_sectors * BDRV_SECTOR_SIZE, | |
2291 | }; | |
2292 | ||
2293 | qemu_iovec_init_external(&qiov, &iov, 1); | |
4105eaaa | 2294 | return bdrv_rwv_co(bs, sector_num, &qiov, is_write, flags); |
8d3b1a2d KW |
2295 | } |
2296 | ||
1c9805a3 SH |
2297 | /* return < 0 if error. See bdrv_write() for the return codes */ |
2298 | int bdrv_read(BlockDriverState *bs, int64_t sector_num, | |
2299 | uint8_t *buf, int nb_sectors) | |
2300 | { | |
4105eaaa | 2301 | return bdrv_rw_co(bs, sector_num, buf, nb_sectors, false, 0); |
fc01f7e7 FB |
2302 | } |
2303 | ||
07d27a44 MA |
2304 | /* Just like bdrv_read(), but with I/O throttling temporarily disabled */ |
2305 | int bdrv_read_unthrottled(BlockDriverState *bs, int64_t sector_num, | |
2306 | uint8_t *buf, int nb_sectors) | |
2307 | { | |
2308 | bool enabled; | |
2309 | int ret; | |
2310 | ||
2311 | enabled = bs->io_limits_enabled; | |
2312 | bs->io_limits_enabled = false; | |
4e7395e8 | 2313 | ret = bdrv_read(bs, sector_num, buf, nb_sectors); |
07d27a44 MA |
2314 | bs->io_limits_enabled = enabled; |
2315 | return ret; | |
2316 | } | |
2317 | ||
5fafdf24 | 2318 | /* Return < 0 if error. Important errors are: |
19cb3738 FB |
2319 | -EIO generic I/O error (may happen for all errors) |
2320 | -ENOMEDIUM No media inserted. | |
2321 | -EINVAL Invalid sector number or nb_sectors | |
2322 | -EACCES Trying to write a read-only device | |
2323 | */ | |
5fafdf24 | 2324 | int bdrv_write(BlockDriverState *bs, int64_t sector_num, |
fc01f7e7 FB |
2325 | const uint8_t *buf, int nb_sectors) |
2326 | { | |
4105eaaa | 2327 | return bdrv_rw_co(bs, sector_num, (uint8_t *)buf, nb_sectors, true, 0); |
83f64091 FB |
2328 | } |
2329 | ||
8d3b1a2d KW |
2330 | int bdrv_writev(BlockDriverState *bs, int64_t sector_num, QEMUIOVector *qiov) |
2331 | { | |
4105eaaa PL |
2332 | return bdrv_rwv_co(bs, sector_num, qiov, true, 0); |
2333 | } | |
2334 | ||
2335 | int bdrv_write_zeroes(BlockDriverState *bs, int64_t sector_num, int nb_sectors) | |
2336 | { | |
2337 | return bdrv_rw_co(bs, sector_num, NULL, nb_sectors, true, | |
2338 | BDRV_REQ_ZERO_WRITE); | |
8d3b1a2d KW |
2339 | } |
2340 | ||
eda578e5 AL |
2341 | int bdrv_pread(BlockDriverState *bs, int64_t offset, |
2342 | void *buf, int count1) | |
83f64091 | 2343 | { |
6ea44308 | 2344 | uint8_t tmp_buf[BDRV_SECTOR_SIZE]; |
83f64091 FB |
2345 | int len, nb_sectors, count; |
2346 | int64_t sector_num; | |
9a8c4cce | 2347 | int ret; |
83f64091 FB |
2348 | |
2349 | count = count1; | |
2350 | /* first read to align to sector start */ | |
6ea44308 | 2351 | len = (BDRV_SECTOR_SIZE - offset) & (BDRV_SECTOR_SIZE - 1); |
83f64091 FB |
2352 | if (len > count) |
2353 | len = count; | |
6ea44308 | 2354 | sector_num = offset >> BDRV_SECTOR_BITS; |
83f64091 | 2355 | if (len > 0) { |
9a8c4cce KW |
2356 | if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0) |
2357 | return ret; | |
6ea44308 | 2358 | memcpy(buf, tmp_buf + (offset & (BDRV_SECTOR_SIZE - 1)), len); |
83f64091 FB |
2359 | count -= len; |
2360 | if (count == 0) | |
2361 | return count1; | |
2362 | sector_num++; | |
2363 | buf += len; | |
2364 | } | |
2365 | ||
2366 | /* read the sectors "in place" */ | |
6ea44308 | 2367 | nb_sectors = count >> BDRV_SECTOR_BITS; |
83f64091 | 2368 | if (nb_sectors > 0) { |
9a8c4cce KW |
2369 | if ((ret = bdrv_read(bs, sector_num, buf, nb_sectors)) < 0) |
2370 | return ret; | |
83f64091 | 2371 | sector_num += nb_sectors; |
6ea44308 | 2372 | len = nb_sectors << BDRV_SECTOR_BITS; |
83f64091 FB |
2373 | buf += len; |
2374 | count -= len; | |
2375 | } | |
2376 | ||
2377 | /* add data from the last sector */ | |
2378 | if (count > 0) { | |
9a8c4cce KW |
2379 | if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0) |
2380 | return ret; | |
83f64091 FB |
2381 | memcpy(buf, tmp_buf, count); |
2382 | } | |
2383 | return count1; | |
2384 | } | |
2385 | ||
8d3b1a2d | 2386 | int bdrv_pwritev(BlockDriverState *bs, int64_t offset, QEMUIOVector *qiov) |
83f64091 | 2387 | { |
6ea44308 | 2388 | uint8_t tmp_buf[BDRV_SECTOR_SIZE]; |
83f64091 FB |
2389 | int len, nb_sectors, count; |
2390 | int64_t sector_num; | |
9a8c4cce | 2391 | int ret; |
83f64091 | 2392 | |
8d3b1a2d KW |
2393 | count = qiov->size; |
2394 | ||
83f64091 | 2395 | /* first write to align to sector start */ |
6ea44308 | 2396 | len = (BDRV_SECTOR_SIZE - offset) & (BDRV_SECTOR_SIZE - 1); |
83f64091 FB |
2397 | if (len > count) |
2398 | len = count; | |
6ea44308 | 2399 | sector_num = offset >> BDRV_SECTOR_BITS; |
83f64091 | 2400 | if (len > 0) { |
9a8c4cce KW |
2401 | if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0) |
2402 | return ret; | |
8d3b1a2d KW |
2403 | qemu_iovec_to_buf(qiov, 0, tmp_buf + (offset & (BDRV_SECTOR_SIZE - 1)), |
2404 | len); | |
9a8c4cce KW |
2405 | if ((ret = bdrv_write(bs, sector_num, tmp_buf, 1)) < 0) |
2406 | return ret; | |
83f64091 FB |
2407 | count -= len; |
2408 | if (count == 0) | |
8d3b1a2d | 2409 | return qiov->size; |
83f64091 | 2410 | sector_num++; |
83f64091 FB |
2411 | } |
2412 | ||
2413 | /* write the sectors "in place" */ | |
6ea44308 | 2414 | nb_sectors = count >> BDRV_SECTOR_BITS; |
83f64091 | 2415 | if (nb_sectors > 0) { |
8d3b1a2d KW |
2416 | QEMUIOVector qiov_inplace; |
2417 | ||
2418 | qemu_iovec_init(&qiov_inplace, qiov->niov); | |
2419 | qemu_iovec_concat(&qiov_inplace, qiov, len, | |
2420 | nb_sectors << BDRV_SECTOR_BITS); | |
2421 | ret = bdrv_writev(bs, sector_num, &qiov_inplace); | |
2422 | qemu_iovec_destroy(&qiov_inplace); | |
2423 | if (ret < 0) { | |
9a8c4cce | 2424 | return ret; |
8d3b1a2d KW |
2425 | } |
2426 | ||
83f64091 | 2427 | sector_num += nb_sectors; |
6ea44308 | 2428 | len = nb_sectors << BDRV_SECTOR_BITS; |
83f64091 FB |
2429 | count -= len; |
2430 | } | |
2431 | ||
2432 | /* add data from the last sector */ | |
2433 | if (count > 0) { | |
9a8c4cce KW |
2434 | if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0) |
2435 | return ret; | |
8d3b1a2d | 2436 | qemu_iovec_to_buf(qiov, qiov->size - count, tmp_buf, count); |
9a8c4cce KW |
2437 | if ((ret = bdrv_write(bs, sector_num, tmp_buf, 1)) < 0) |
2438 | return ret; | |
83f64091 | 2439 | } |
8d3b1a2d KW |
2440 | return qiov->size; |
2441 | } | |
2442 | ||
2443 | int bdrv_pwrite(BlockDriverState *bs, int64_t offset, | |
2444 | const void *buf, int count1) | |
2445 | { | |
2446 | QEMUIOVector qiov; | |
2447 | struct iovec iov = { | |
2448 | .iov_base = (void *) buf, | |
2449 | .iov_len = count1, | |
2450 | }; | |
2451 | ||
2452 | qemu_iovec_init_external(&qiov, &iov, 1); | |
2453 | return bdrv_pwritev(bs, offset, &qiov); | |
83f64091 | 2454 | } |
83f64091 | 2455 | |
f08145fe KW |
2456 | /* |
2457 | * Writes to the file and ensures that no writes are reordered across this | |
2458 | * request (acts as a barrier) | |
2459 | * | |
2460 | * Returns 0 on success, -errno in error cases. | |
2461 | */ | |
2462 | int bdrv_pwrite_sync(BlockDriverState *bs, int64_t offset, | |
2463 | const void *buf, int count) | |
2464 | { | |
2465 | int ret; | |
2466 | ||
2467 | ret = bdrv_pwrite(bs, offset, buf, count); | |
2468 | if (ret < 0) { | |
2469 | return ret; | |
2470 | } | |
2471 | ||
f05fa4ad PB |
2472 | /* No flush needed for cache modes that already do it */ |
2473 | if (bs->enable_write_cache) { | |
f08145fe KW |
2474 | bdrv_flush(bs); |
2475 | } | |
2476 | ||
2477 | return 0; | |
2478 | } | |
2479 | ||
470c0504 | 2480 | static int coroutine_fn bdrv_co_do_copy_on_readv(BlockDriverState *bs, |
ab185921 SH |
2481 | int64_t sector_num, int nb_sectors, QEMUIOVector *qiov) |
2482 | { | |
2483 | /* Perform I/O through a temporary buffer so that users who scribble over | |
2484 | * their read buffer while the operation is in progress do not end up | |
2485 | * modifying the image file. This is critical for zero-copy guest I/O | |
2486 | * where anything might happen inside guest memory. | |
2487 | */ | |
2488 | void *bounce_buffer; | |
2489 | ||
79c053bd | 2490 | BlockDriver *drv = bs->drv; |
ab185921 SH |
2491 | struct iovec iov; |
2492 | QEMUIOVector bounce_qiov; | |
2493 | int64_t cluster_sector_num; | |
2494 | int cluster_nb_sectors; | |
2495 | size_t skip_bytes; | |
2496 | int ret; | |
2497 | ||
2498 | /* Cover entire cluster so no additional backing file I/O is required when | |
2499 | * allocating cluster in the image file. | |
2500 | */ | |
343bded4 PB |
2501 | bdrv_round_to_clusters(bs, sector_num, nb_sectors, |
2502 | &cluster_sector_num, &cluster_nb_sectors); | |
ab185921 | 2503 | |
470c0504 SH |
2504 | trace_bdrv_co_do_copy_on_readv(bs, sector_num, nb_sectors, |
2505 | cluster_sector_num, cluster_nb_sectors); | |
ab185921 SH |
2506 | |
2507 | iov.iov_len = cluster_nb_sectors * BDRV_SECTOR_SIZE; | |
2508 | iov.iov_base = bounce_buffer = qemu_blockalign(bs, iov.iov_len); | |
2509 | qemu_iovec_init_external(&bounce_qiov, &iov, 1); | |
2510 | ||
79c053bd SH |
2511 | ret = drv->bdrv_co_readv(bs, cluster_sector_num, cluster_nb_sectors, |
2512 | &bounce_qiov); | |
ab185921 SH |
2513 | if (ret < 0) { |
2514 | goto err; | |
2515 | } | |
2516 | ||
79c053bd SH |
2517 | if (drv->bdrv_co_write_zeroes && |
2518 | buffer_is_zero(bounce_buffer, iov.iov_len)) { | |
621f0589 KW |
2519 | ret = bdrv_co_do_write_zeroes(bs, cluster_sector_num, |
2520 | cluster_nb_sectors); | |
79c053bd | 2521 | } else { |
f05fa4ad PB |
2522 | /* This does not change the data on the disk, it is not necessary |
2523 | * to flush even in cache=writethrough mode. | |
2524 | */ | |
79c053bd | 2525 | ret = drv->bdrv_co_writev(bs, cluster_sector_num, cluster_nb_sectors, |
ab185921 | 2526 | &bounce_qiov); |
79c053bd SH |
2527 | } |
2528 | ||
ab185921 SH |
2529 | if (ret < 0) { |
2530 | /* It might be okay to ignore write errors for guest requests. If this | |
2531 | * is a deliberate copy-on-read then we don't want to ignore the error. | |
2532 | * Simply report it in all cases. | |
2533 | */ | |
2534 | goto err; | |
2535 | } | |
2536 | ||
2537 | skip_bytes = (sector_num - cluster_sector_num) * BDRV_SECTOR_SIZE; | |
03396148 MT |
2538 | qemu_iovec_from_buf(qiov, 0, bounce_buffer + skip_bytes, |
2539 | nb_sectors * BDRV_SECTOR_SIZE); | |
ab185921 SH |
2540 | |
2541 | err: | |
2542 | qemu_vfree(bounce_buffer); | |
2543 | return ret; | |
2544 | } | |
2545 | ||
c5fbe571 SH |
2546 | /* |
2547 | * Handle a read request in coroutine context | |
2548 | */ | |
2549 | static int coroutine_fn bdrv_co_do_readv(BlockDriverState *bs, | |
470c0504 SH |
2550 | int64_t sector_num, int nb_sectors, QEMUIOVector *qiov, |
2551 | BdrvRequestFlags flags) | |
da1fa91d KW |
2552 | { |
2553 | BlockDriver *drv = bs->drv; | |
dbffbdcf SH |
2554 | BdrvTrackedRequest req; |
2555 | int ret; | |
da1fa91d | 2556 | |
da1fa91d KW |
2557 | if (!drv) { |
2558 | return -ENOMEDIUM; | |
2559 | } | |
2560 | if (bdrv_check_request(bs, sector_num, nb_sectors)) { | |
2561 | return -EIO; | |
2562 | } | |
2563 | ||
f4658285 | 2564 | if (bs->copy_on_read) { |
470c0504 SH |
2565 | flags |= BDRV_REQ_COPY_ON_READ; |
2566 | } | |
2567 | if (flags & BDRV_REQ_COPY_ON_READ) { | |
2568 | bs->copy_on_read_in_flight++; | |
2569 | } | |
2570 | ||
2571 | if (bs->copy_on_read_in_flight) { | |
f4658285 SH |
2572 | wait_for_overlapping_requests(bs, sector_num, nb_sectors); |
2573 | } | |
2574 | ||
cc0681c4 BC |
2575 | /* throttling disk I/O */ |
2576 | if (bs->io_limits_enabled) { | |
2577 | bdrv_io_limits_intercept(bs, nb_sectors, false); | |
2578 | } | |
2579 | ||
dbffbdcf | 2580 | tracked_request_begin(&req, bs, sector_num, nb_sectors, false); |
ab185921 | 2581 | |
470c0504 | 2582 | if (flags & BDRV_REQ_COPY_ON_READ) { |
ab185921 SH |
2583 | int pnum; |
2584 | ||
2585 | ret = bdrv_co_is_allocated(bs, sector_num, nb_sectors, &pnum); | |
2586 | if (ret < 0) { | |
2587 | goto out; | |
2588 | } | |
2589 | ||
2590 | if (!ret || pnum != nb_sectors) { | |
470c0504 | 2591 | ret = bdrv_co_do_copy_on_readv(bs, sector_num, nb_sectors, qiov); |
ab185921 SH |
2592 | goto out; |
2593 | } | |
2594 | } | |
2595 | ||
893a8f62 MK |
2596 | if (!(bs->zero_beyond_eof && bs->growable)) { |
2597 | ret = drv->bdrv_co_readv(bs, sector_num, nb_sectors, qiov); | |
2598 | } else { | |
2599 | /* Read zeros after EOF of growable BDSes */ | |
2600 | int64_t len, total_sectors, max_nb_sectors; | |
2601 | ||
2602 | len = bdrv_getlength(bs); | |
2603 | if (len < 0) { | |
2604 | ret = len; | |
2605 | goto out; | |
2606 | } | |
2607 | ||
2608 | total_sectors = len >> BDRV_SECTOR_BITS; | |
2609 | max_nb_sectors = MAX(0, total_sectors - sector_num); | |
2610 | if (max_nb_sectors > 0) { | |
2611 | ret = drv->bdrv_co_readv(bs, sector_num, | |
2612 | MIN(nb_sectors, max_nb_sectors), qiov); | |
2613 | } else { | |
2614 | ret = 0; | |
2615 | } | |
2616 | ||
2617 | /* Reading beyond end of file is supposed to produce zeroes */ | |
2618 | if (ret == 0 && total_sectors < sector_num + nb_sectors) { | |
2619 | uint64_t offset = MAX(0, total_sectors - sector_num); | |
2620 | uint64_t bytes = (sector_num + nb_sectors - offset) * | |
2621 | BDRV_SECTOR_SIZE; | |
2622 | qemu_iovec_memset(qiov, offset * BDRV_SECTOR_SIZE, 0, bytes); | |
2623 | } | |
2624 | } | |
ab185921 SH |
2625 | |
2626 | out: | |
dbffbdcf | 2627 | tracked_request_end(&req); |
470c0504 SH |
2628 | |
2629 | if (flags & BDRV_REQ_COPY_ON_READ) { | |
2630 | bs->copy_on_read_in_flight--; | |
2631 | } | |
2632 | ||
dbffbdcf | 2633 | return ret; |
da1fa91d KW |
2634 | } |
2635 | ||
c5fbe571 | 2636 | int coroutine_fn bdrv_co_readv(BlockDriverState *bs, int64_t sector_num, |
da1fa91d KW |
2637 | int nb_sectors, QEMUIOVector *qiov) |
2638 | { | |
c5fbe571 | 2639 | trace_bdrv_co_readv(bs, sector_num, nb_sectors); |
da1fa91d | 2640 | |
470c0504 SH |
2641 | return bdrv_co_do_readv(bs, sector_num, nb_sectors, qiov, 0); |
2642 | } | |
2643 | ||
2644 | int coroutine_fn bdrv_co_copy_on_readv(BlockDriverState *bs, | |
2645 | int64_t sector_num, int nb_sectors, QEMUIOVector *qiov) | |
2646 | { | |
2647 | trace_bdrv_co_copy_on_readv(bs, sector_num, nb_sectors); | |
2648 | ||
2649 | return bdrv_co_do_readv(bs, sector_num, nb_sectors, qiov, | |
2650 | BDRV_REQ_COPY_ON_READ); | |
c5fbe571 SH |
2651 | } |
2652 | ||
f08f2dda SH |
2653 | static int coroutine_fn bdrv_co_do_write_zeroes(BlockDriverState *bs, |
2654 | int64_t sector_num, int nb_sectors) | |
2655 | { | |
2656 | BlockDriver *drv = bs->drv; | |
2657 | QEMUIOVector qiov; | |
2658 | struct iovec iov; | |
2659 | int ret; | |
2660 | ||
621f0589 KW |
2661 | /* TODO Emulate only part of misaligned requests instead of letting block |
2662 | * drivers return -ENOTSUP and emulate everything */ | |
2663 | ||
f08f2dda SH |
2664 | /* First try the efficient write zeroes operation */ |
2665 | if (drv->bdrv_co_write_zeroes) { | |
621f0589 KW |
2666 | ret = drv->bdrv_co_write_zeroes(bs, sector_num, nb_sectors); |
2667 | if (ret != -ENOTSUP) { | |
2668 | return ret; | |
2669 | } | |
f08f2dda SH |
2670 | } |
2671 | ||
2672 | /* Fall back to bounce buffer if write zeroes is unsupported */ | |
2673 | iov.iov_len = nb_sectors * BDRV_SECTOR_SIZE; | |
2674 | iov.iov_base = qemu_blockalign(bs, iov.iov_len); | |
2675 | memset(iov.iov_base, 0, iov.iov_len); | |
2676 | qemu_iovec_init_external(&qiov, &iov, 1); | |
2677 | ||
2678 | ret = drv->bdrv_co_writev(bs, sector_num, nb_sectors, &qiov); | |
2679 | ||
2680 | qemu_vfree(iov.iov_base); | |
2681 | return ret; | |
2682 | } | |
2683 | ||
c5fbe571 SH |
2684 | /* |
2685 | * Handle a write request in coroutine context | |
2686 | */ | |
2687 | static int coroutine_fn bdrv_co_do_writev(BlockDriverState *bs, | |
f08f2dda SH |
2688 | int64_t sector_num, int nb_sectors, QEMUIOVector *qiov, |
2689 | BdrvRequestFlags flags) | |
c5fbe571 SH |
2690 | { |
2691 | BlockDriver *drv = bs->drv; | |
dbffbdcf | 2692 | BdrvTrackedRequest req; |
6b7cb247 | 2693 | int ret; |
da1fa91d KW |
2694 | |
2695 | if (!bs->drv) { | |
2696 | return -ENOMEDIUM; | |
2697 | } | |
2698 | if (bs->read_only) { | |
2699 | return -EACCES; | |
2700 | } | |
2701 | if (bdrv_check_request(bs, sector_num, nb_sectors)) { | |
2702 | return -EIO; | |
2703 | } | |
2704 | ||
470c0504 | 2705 | if (bs->copy_on_read_in_flight) { |
f4658285 SH |
2706 | wait_for_overlapping_requests(bs, sector_num, nb_sectors); |
2707 | } | |
2708 | ||
cc0681c4 BC |
2709 | /* throttling disk I/O */ |
2710 | if (bs->io_limits_enabled) { | |
2711 | bdrv_io_limits_intercept(bs, nb_sectors, true); | |
2712 | } | |
2713 | ||
dbffbdcf SH |
2714 | tracked_request_begin(&req, bs, sector_num, nb_sectors, true); |
2715 | ||
d616b224 SH |
2716 | ret = notifier_with_return_list_notify(&bs->before_write_notifiers, &req); |
2717 | ||
2718 | if (ret < 0) { | |
2719 | /* Do nothing, write notifier decided to fail this request */ | |
2720 | } else if (flags & BDRV_REQ_ZERO_WRITE) { | |
f08f2dda SH |
2721 | ret = bdrv_co_do_write_zeroes(bs, sector_num, nb_sectors); |
2722 | } else { | |
2723 | ret = drv->bdrv_co_writev(bs, sector_num, nb_sectors, qiov); | |
2724 | } | |
6b7cb247 | 2725 | |
f05fa4ad PB |
2726 | if (ret == 0 && !bs->enable_write_cache) { |
2727 | ret = bdrv_co_flush(bs); | |
2728 | } | |
2729 | ||
da1fa91d | 2730 | if (bs->dirty_bitmap) { |
1755da16 | 2731 | bdrv_set_dirty(bs, sector_num, nb_sectors); |
da1fa91d KW |
2732 | } |
2733 | ||
2734 | if (bs->wr_highest_sector < sector_num + nb_sectors - 1) { | |
2735 | bs->wr_highest_sector = sector_num + nb_sectors - 1; | |
2736 | } | |
2737 | ||
dbffbdcf SH |
2738 | tracked_request_end(&req); |
2739 | ||
6b7cb247 | 2740 | return ret; |
da1fa91d KW |
2741 | } |
2742 | ||
c5fbe571 SH |
2743 | int coroutine_fn bdrv_co_writev(BlockDriverState *bs, int64_t sector_num, |
2744 | int nb_sectors, QEMUIOVector *qiov) | |
2745 | { | |
2746 | trace_bdrv_co_writev(bs, sector_num, nb_sectors); | |
2747 | ||
f08f2dda SH |
2748 | return bdrv_co_do_writev(bs, sector_num, nb_sectors, qiov, 0); |
2749 | } | |
2750 | ||
2751 | int coroutine_fn bdrv_co_write_zeroes(BlockDriverState *bs, | |
2752 | int64_t sector_num, int nb_sectors) | |
2753 | { | |
2754 | trace_bdrv_co_write_zeroes(bs, sector_num, nb_sectors); | |
2755 | ||
2756 | return bdrv_co_do_writev(bs, sector_num, nb_sectors, NULL, | |
2757 | BDRV_REQ_ZERO_WRITE); | |
c5fbe571 SH |
2758 | } |
2759 | ||
83f64091 FB |
2760 | /** |
2761 | * Truncate file to 'offset' bytes (needed only for file protocols) | |
2762 | */ | |
2763 | int bdrv_truncate(BlockDriverState *bs, int64_t offset) | |
2764 | { | |
2765 | BlockDriver *drv = bs->drv; | |
51762288 | 2766 | int ret; |
83f64091 | 2767 | if (!drv) |
19cb3738 | 2768 | return -ENOMEDIUM; |
83f64091 FB |
2769 | if (!drv->bdrv_truncate) |
2770 | return -ENOTSUP; | |
59f2689d NS |
2771 | if (bs->read_only) |
2772 | return -EACCES; | |
8591675f MT |
2773 | if (bdrv_in_use(bs)) |
2774 | return -EBUSY; | |
51762288 SH |
2775 | ret = drv->bdrv_truncate(bs, offset); |
2776 | if (ret == 0) { | |
2777 | ret = refresh_total_sectors(bs, offset >> BDRV_SECTOR_BITS); | |
145feb17 | 2778 | bdrv_dev_resize_cb(bs); |
51762288 SH |
2779 | } |
2780 | return ret; | |
83f64091 FB |
2781 | } |
2782 | ||
4a1d5e1f FZ |
2783 | /** |
2784 | * Length of a allocated file in bytes. Sparse files are counted by actual | |
2785 | * allocated space. Return < 0 if error or unknown. | |
2786 | */ | |
2787 | int64_t bdrv_get_allocated_file_size(BlockDriverState *bs) | |
2788 | { | |
2789 | BlockDriver *drv = bs->drv; | |
2790 | if (!drv) { | |
2791 | return -ENOMEDIUM; | |
2792 | } | |
2793 | if (drv->bdrv_get_allocated_file_size) { | |
2794 | return drv->bdrv_get_allocated_file_size(bs); | |
2795 | } | |
2796 | if (bs->file) { | |
2797 | return bdrv_get_allocated_file_size(bs->file); | |
2798 | } | |
2799 | return -ENOTSUP; | |
2800 | } | |
2801 | ||
83f64091 FB |
2802 | /** |
2803 | * Length of a file in bytes. Return < 0 if error or unknown. | |
2804 | */ | |
2805 | int64_t bdrv_getlength(BlockDriverState *bs) | |
2806 | { | |
2807 | BlockDriver *drv = bs->drv; | |
2808 | if (!drv) | |
19cb3738 | 2809 | return -ENOMEDIUM; |
51762288 | 2810 | |
2c6942fa | 2811 | if (bs->growable || bdrv_dev_has_removable_media(bs)) { |
46a4e4e6 SH |
2812 | if (drv->bdrv_getlength) { |
2813 | return drv->bdrv_getlength(bs); | |
2814 | } | |
83f64091 | 2815 | } |
46a4e4e6 | 2816 | return bs->total_sectors * BDRV_SECTOR_SIZE; |
fc01f7e7 FB |
2817 | } |
2818 | ||
19cb3738 | 2819 | /* return 0 as number of sectors if no device present or error */ |
96b8f136 | 2820 | void bdrv_get_geometry(BlockDriverState *bs, uint64_t *nb_sectors_ptr) |
fc01f7e7 | 2821 | { |
19cb3738 FB |
2822 | int64_t length; |
2823 | length = bdrv_getlength(bs); | |
2824 | if (length < 0) | |
2825 | length = 0; | |
2826 | else | |
6ea44308 | 2827 | length = length >> BDRV_SECTOR_BITS; |
19cb3738 | 2828 | *nb_sectors_ptr = length; |
fc01f7e7 | 2829 | } |
cf98951b | 2830 | |
ff06f5f3 PB |
2831 | void bdrv_set_on_error(BlockDriverState *bs, BlockdevOnError on_read_error, |
2832 | BlockdevOnError on_write_error) | |
abd7f68d MA |
2833 | { |
2834 | bs->on_read_error = on_read_error; | |
2835 | bs->on_write_error = on_write_error; | |
2836 | } | |
2837 | ||
1ceee0d5 | 2838 | BlockdevOnError bdrv_get_on_error(BlockDriverState *bs, bool is_read) |
abd7f68d MA |
2839 | { |
2840 | return is_read ? bs->on_read_error : bs->on_write_error; | |
2841 | } | |
2842 | ||
3e1caa5f PB |
2843 | BlockErrorAction bdrv_get_error_action(BlockDriverState *bs, bool is_read, int error) |
2844 | { | |
2845 | BlockdevOnError on_err = is_read ? bs->on_read_error : bs->on_write_error; | |
2846 | ||
2847 | switch (on_err) { | |
2848 | case BLOCKDEV_ON_ERROR_ENOSPC: | |
2849 | return (error == ENOSPC) ? BDRV_ACTION_STOP : BDRV_ACTION_REPORT; | |
2850 | case BLOCKDEV_ON_ERROR_STOP: | |
2851 | return BDRV_ACTION_STOP; | |
2852 | case BLOCKDEV_ON_ERROR_REPORT: | |
2853 | return BDRV_ACTION_REPORT; | |
2854 | case BLOCKDEV_ON_ERROR_IGNORE: | |
2855 | return BDRV_ACTION_IGNORE; | |
2856 | default: | |
2857 | abort(); | |
2858 | } | |
2859 | } | |
2860 | ||
2861 | /* This is done by device models because, while the block layer knows | |
2862 | * about the error, it does not know whether an operation comes from | |
2863 | * the device or the block layer (from a job, for example). | |
2864 | */ | |
2865 | void bdrv_error_action(BlockDriverState *bs, BlockErrorAction action, | |
2866 | bool is_read, int error) | |
2867 | { | |
2868 | assert(error >= 0); | |
32c81a4a | 2869 | bdrv_emit_qmp_error_event(bs, QEVENT_BLOCK_IO_ERROR, action, is_read); |
3e1caa5f PB |
2870 | if (action == BDRV_ACTION_STOP) { |
2871 | vm_stop(RUN_STATE_IO_ERROR); | |
2872 | bdrv_iostatus_set_err(bs, error); | |
2873 | } | |
2874 | } | |
2875 | ||
b338082b FB |
2876 | int bdrv_is_read_only(BlockDriverState *bs) |
2877 | { | |
2878 | return bs->read_only; | |
2879 | } | |
2880 | ||
985a03b0 TS |
2881 | int bdrv_is_sg(BlockDriverState *bs) |
2882 | { | |
2883 | return bs->sg; | |
2884 | } | |
2885 | ||
e900a7b7 CH |
2886 | int bdrv_enable_write_cache(BlockDriverState *bs) |
2887 | { | |
2888 | return bs->enable_write_cache; | |
2889 | } | |
2890 | ||
425b0148 PB |
2891 | void bdrv_set_enable_write_cache(BlockDriverState *bs, bool wce) |
2892 | { | |
2893 | bs->enable_write_cache = wce; | |
55b110f2 JC |
2894 | |
2895 | /* so a reopen() will preserve wce */ | |
2896 | if (wce) { | |
2897 | bs->open_flags |= BDRV_O_CACHE_WB; | |
2898 | } else { | |
2899 | bs->open_flags &= ~BDRV_O_CACHE_WB; | |
2900 | } | |
425b0148 PB |
2901 | } |
2902 | ||
ea2384d3 FB |
2903 | int bdrv_is_encrypted(BlockDriverState *bs) |
2904 | { | |
2905 | if (bs->backing_hd && bs->backing_hd->encrypted) | |
2906 | return 1; | |
2907 | return bs->encrypted; | |
2908 | } | |
2909 | ||
c0f4ce77 AL |
2910 | int bdrv_key_required(BlockDriverState *bs) |
2911 | { | |
2912 | BlockDriverState *backing_hd = bs->backing_hd; | |
2913 | ||
2914 | if (backing_hd && backing_hd->encrypted && !backing_hd->valid_key) | |
2915 | return 1; | |
2916 | return (bs->encrypted && !bs->valid_key); | |
2917 | } | |
2918 | ||
ea2384d3 FB |
2919 | int bdrv_set_key(BlockDriverState *bs, const char *key) |
2920 | { | |
2921 | int ret; | |
2922 | if (bs->backing_hd && bs->backing_hd->encrypted) { | |
2923 | ret = bdrv_set_key(bs->backing_hd, key); | |
2924 | if (ret < 0) | |
2925 | return ret; | |
2926 | if (!bs->encrypted) | |
2927 | return 0; | |
2928 | } | |
fd04a2ae SH |
2929 | if (!bs->encrypted) { |
2930 | return -EINVAL; | |
2931 | } else if (!bs->drv || !bs->drv->bdrv_set_key) { | |
2932 | return -ENOMEDIUM; | |
2933 | } | |
c0f4ce77 | 2934 | ret = bs->drv->bdrv_set_key(bs, key); |
bb5fc20f AL |
2935 | if (ret < 0) { |
2936 | bs->valid_key = 0; | |
2937 | } else if (!bs->valid_key) { | |
2938 | bs->valid_key = 1; | |
2939 | /* call the change callback now, we skipped it on open */ | |
7d4b4ba5 | 2940 | bdrv_dev_change_media_cb(bs, true); |
bb5fc20f | 2941 | } |
c0f4ce77 | 2942 | return ret; |
ea2384d3 FB |
2943 | } |
2944 | ||
f8d6bba1 | 2945 | const char *bdrv_get_format_name(BlockDriverState *bs) |
ea2384d3 | 2946 | { |
f8d6bba1 | 2947 | return bs->drv ? bs->drv->format_name : NULL; |
ea2384d3 FB |
2948 | } |
2949 | ||
5fafdf24 | 2950 | void bdrv_iterate_format(void (*it)(void *opaque, const char *name), |
ea2384d3 FB |
2951 | void *opaque) |
2952 | { | |
2953 | BlockDriver *drv; | |
2954 | ||
8a22f02a | 2955 | QLIST_FOREACH(drv, &bdrv_drivers, list) { |
ea2384d3 FB |
2956 | it(opaque, drv->format_name); |
2957 | } | |
2958 | } | |
2959 | ||
b338082b FB |
2960 | BlockDriverState *bdrv_find(const char *name) |
2961 | { | |
2962 | BlockDriverState *bs; | |
2963 | ||
1b7bdbc1 SH |
2964 | QTAILQ_FOREACH(bs, &bdrv_states, list) { |
2965 | if (!strcmp(name, bs->device_name)) { | |
b338082b | 2966 | return bs; |
1b7bdbc1 | 2967 | } |
b338082b FB |
2968 | } |
2969 | return NULL; | |
2970 | } | |
2971 | ||
2f399b0a MA |
2972 | BlockDriverState *bdrv_next(BlockDriverState *bs) |
2973 | { | |
2974 | if (!bs) { | |
2975 | return QTAILQ_FIRST(&bdrv_states); | |
2976 | } | |
2977 | return QTAILQ_NEXT(bs, list); | |
2978 | } | |
2979 | ||
51de9760 | 2980 | void bdrv_iterate(void (*it)(void *opaque, BlockDriverState *bs), void *opaque) |
81d0912d FB |
2981 | { |
2982 | BlockDriverState *bs; | |
2983 | ||
1b7bdbc1 | 2984 | QTAILQ_FOREACH(bs, &bdrv_states, list) { |
51de9760 | 2985 | it(opaque, bs); |
81d0912d FB |
2986 | } |
2987 | } | |
2988 | ||
ea2384d3 FB |
2989 | const char *bdrv_get_device_name(BlockDriverState *bs) |
2990 | { | |
2991 | return bs->device_name; | |
2992 | } | |
2993 | ||
c8433287 MA |
2994 | int bdrv_get_flags(BlockDriverState *bs) |
2995 | { | |
2996 | return bs->open_flags; | |
2997 | } | |
2998 | ||
f0f0fdfe | 2999 | int bdrv_flush_all(void) |
c6ca28d6 AL |
3000 | { |
3001 | BlockDriverState *bs; | |
f0f0fdfe | 3002 | int result = 0; |
c6ca28d6 | 3003 | |
1b7bdbc1 | 3004 | QTAILQ_FOREACH(bs, &bdrv_states, list) { |
f0f0fdfe KW |
3005 | int ret = bdrv_flush(bs); |
3006 | if (ret < 0 && !result) { | |
3007 | result = ret; | |
3008 | } | |
1b7bdbc1 | 3009 | } |
f0f0fdfe KW |
3010 | |
3011 | return result; | |
c6ca28d6 AL |
3012 | } |
3013 | ||
3ac21627 PL |
3014 | int bdrv_has_zero_init_1(BlockDriverState *bs) |
3015 | { | |
3016 | return 1; | |
3017 | } | |
3018 | ||
f2feebbd KW |
3019 | int bdrv_has_zero_init(BlockDriverState *bs) |
3020 | { | |
3021 | assert(bs->drv); | |
3022 | ||
336c1c12 KW |
3023 | if (bs->drv->bdrv_has_zero_init) { |
3024 | return bs->drv->bdrv_has_zero_init(bs); | |
f2feebbd KW |
3025 | } |
3026 | ||
3ac21627 PL |
3027 | /* safe default */ |
3028 | return 0; | |
f2feebbd KW |
3029 | } |
3030 | ||
376ae3f1 SH |
3031 | typedef struct BdrvCoIsAllocatedData { |
3032 | BlockDriverState *bs; | |
b35b2bba | 3033 | BlockDriverState *base; |
376ae3f1 SH |
3034 | int64_t sector_num; |
3035 | int nb_sectors; | |
3036 | int *pnum; | |
3037 | int ret; | |
3038 | bool done; | |
3039 | } BdrvCoIsAllocatedData; | |
3040 | ||
f58c7b35 TS |
3041 | /* |
3042 | * Returns true iff the specified sector is present in the disk image. Drivers | |
3043 | * not implementing the functionality are assumed to not support backing files, | |
3044 | * hence all their sectors are reported as allocated. | |
3045 | * | |
bd9533e3 SH |
3046 | * If 'sector_num' is beyond the end of the disk image the return value is 0 |
3047 | * and 'pnum' is set to 0. | |
3048 | * | |
f58c7b35 TS |
3049 | * 'pnum' is set to the number of sectors (including and immediately following |
3050 | * the specified sector) that are known to be in the same | |
3051 | * allocated/unallocated state. | |
3052 | * | |
bd9533e3 SH |
3053 | * 'nb_sectors' is the max value 'pnum' should be set to. If nb_sectors goes |
3054 | * beyond the end of the disk image it will be clamped. | |
f58c7b35 | 3055 | */ |
060f51c9 SH |
3056 | int coroutine_fn bdrv_co_is_allocated(BlockDriverState *bs, int64_t sector_num, |
3057 | int nb_sectors, int *pnum) | |
f58c7b35 | 3058 | { |
bd9533e3 SH |
3059 | int64_t n; |
3060 | ||
3061 | if (sector_num >= bs->total_sectors) { | |
3062 | *pnum = 0; | |
3063 | return 0; | |
3064 | } | |
3065 | ||
3066 | n = bs->total_sectors - sector_num; | |
3067 | if (n < nb_sectors) { | |
3068 | nb_sectors = n; | |
3069 | } | |
3070 | ||
6aebab14 | 3071 | if (!bs->drv->bdrv_co_is_allocated) { |
bd9533e3 | 3072 | *pnum = nb_sectors; |
f58c7b35 TS |
3073 | return 1; |
3074 | } | |
6aebab14 | 3075 | |
060f51c9 SH |
3076 | return bs->drv->bdrv_co_is_allocated(bs, sector_num, nb_sectors, pnum); |
3077 | } | |
3078 | ||
3079 | /* Coroutine wrapper for bdrv_is_allocated() */ | |
3080 | static void coroutine_fn bdrv_is_allocated_co_entry(void *opaque) | |
3081 | { | |
3082 | BdrvCoIsAllocatedData *data = opaque; | |
3083 | BlockDriverState *bs = data->bs; | |
3084 | ||
3085 | data->ret = bdrv_co_is_allocated(bs, data->sector_num, data->nb_sectors, | |
3086 | data->pnum); | |
3087 | data->done = true; | |
3088 | } | |
3089 | ||
3090 | /* | |
3091 | * Synchronous wrapper around bdrv_co_is_allocated(). | |
3092 | * | |
3093 | * See bdrv_co_is_allocated() for details. | |
3094 | */ | |
3095 | int bdrv_is_allocated(BlockDriverState *bs, int64_t sector_num, int nb_sectors, | |
3096 | int *pnum) | |
3097 | { | |
6aebab14 SH |
3098 | Coroutine *co; |
3099 | BdrvCoIsAllocatedData data = { | |
3100 | .bs = bs, | |
3101 | .sector_num = sector_num, | |
3102 | .nb_sectors = nb_sectors, | |
3103 | .pnum = pnum, | |
3104 | .done = false, | |
3105 | }; | |
3106 | ||
3107 | co = qemu_coroutine_create(bdrv_is_allocated_co_entry); | |
3108 | qemu_coroutine_enter(co, &data); | |
3109 | while (!data.done) { | |
3110 | qemu_aio_wait(); | |
3111 | } | |
3112 | return data.ret; | |
f58c7b35 TS |
3113 | } |
3114 | ||
188a7bbf PB |
3115 | /* |
3116 | * Given an image chain: ... -> [BASE] -> [INTER1] -> [INTER2] -> [TOP] | |
3117 | * | |
3118 | * Return true if the given sector is allocated in any image between | |
3119 | * BASE and TOP (inclusive). BASE can be NULL to check if the given | |
3120 | * sector is allocated in any image of the chain. Return false otherwise. | |
3121 | * | |
3122 | * 'pnum' is set to the number of sectors (including and immediately following | |
3123 | * the specified sector) that are known to be in the same | |
3124 | * allocated/unallocated state. | |
3125 | * | |
3126 | */ | |
3127 | int coroutine_fn bdrv_co_is_allocated_above(BlockDriverState *top, | |
3128 | BlockDriverState *base, | |
3129 | int64_t sector_num, | |
3130 | int nb_sectors, int *pnum) | |
3131 | { | |
3132 | BlockDriverState *intermediate; | |
3133 | int ret, n = nb_sectors; | |
3134 | ||
3135 | intermediate = top; | |
3136 | while (intermediate && intermediate != base) { | |
3137 | int pnum_inter; | |
3138 | ret = bdrv_co_is_allocated(intermediate, sector_num, nb_sectors, | |
3139 | &pnum_inter); | |
3140 | if (ret < 0) { | |
3141 | return ret; | |
3142 | } else if (ret) { | |
3143 | *pnum = pnum_inter; | |
3144 | return 1; | |
3145 | } | |
3146 | ||
3147 | /* | |
3148 | * [sector_num, nb_sectors] is unallocated on top but intermediate | |
3149 | * might have | |
3150 | * | |
3151 | * [sector_num+x, nr_sectors] allocated. | |
3152 | */ | |
63ba17d3 VI |
3153 | if (n > pnum_inter && |
3154 | (intermediate == top || | |
3155 | sector_num + pnum_inter < intermediate->total_sectors)) { | |
188a7bbf PB |
3156 | n = pnum_inter; |
3157 | } | |
3158 | ||
3159 | intermediate = intermediate->backing_hd; | |
3160 | } | |
3161 | ||
3162 | *pnum = n; | |
3163 | return 0; | |
3164 | } | |
3165 | ||
b35b2bba MR |
3166 | /* Coroutine wrapper for bdrv_is_allocated_above() */ |
3167 | static void coroutine_fn bdrv_is_allocated_above_co_entry(void *opaque) | |
3168 | { | |
3169 | BdrvCoIsAllocatedData *data = opaque; | |
3170 | BlockDriverState *top = data->bs; | |
3171 | BlockDriverState *base = data->base; | |
3172 | ||
3173 | data->ret = bdrv_co_is_allocated_above(top, base, data->sector_num, | |
3174 | data->nb_sectors, data->pnum); | |
3175 | data->done = true; | |
3176 | } | |
3177 | ||
3178 | /* | |
3179 | * Synchronous wrapper around bdrv_co_is_allocated_above(). | |
3180 | * | |
3181 | * See bdrv_co_is_allocated_above() for details. | |
3182 | */ | |
3183 | int bdrv_is_allocated_above(BlockDriverState *top, BlockDriverState *base, | |
3184 | int64_t sector_num, int nb_sectors, int *pnum) | |
3185 | { | |
3186 | Coroutine *co; | |
3187 | BdrvCoIsAllocatedData data = { | |
3188 | .bs = top, | |
3189 | .base = base, | |
3190 | .sector_num = sector_num, | |
3191 | .nb_sectors = nb_sectors, | |
3192 | .pnum = pnum, | |
3193 | .done = false, | |
3194 | }; | |
3195 | ||
3196 | co = qemu_coroutine_create(bdrv_is_allocated_above_co_entry); | |
3197 | qemu_coroutine_enter(co, &data); | |
3198 | while (!data.done) { | |
3199 | qemu_aio_wait(); | |
3200 | } | |
3201 | return data.ret; | |
3202 | } | |
3203 | ||
045df330 AL |
3204 | const char *bdrv_get_encrypted_filename(BlockDriverState *bs) |
3205 | { | |
3206 | if (bs->backing_hd && bs->backing_hd->encrypted) | |
3207 | return bs->backing_file; | |
3208 | else if (bs->encrypted) | |
3209 | return bs->filename; | |
3210 | else | |
3211 | return NULL; | |
3212 | } | |
3213 | ||
5fafdf24 | 3214 | void bdrv_get_backing_filename(BlockDriverState *bs, |
83f64091 FB |
3215 | char *filename, int filename_size) |
3216 | { | |
3574c608 | 3217 | pstrcpy(filename, filename_size, bs->backing_file); |
83f64091 FB |
3218 | } |
3219 | ||
5fafdf24 | 3220 | int bdrv_write_compressed(BlockDriverState *bs, int64_t sector_num, |
faea38e7 FB |
3221 | const uint8_t *buf, int nb_sectors) |
3222 | { | |
3223 | BlockDriver *drv = bs->drv; | |
3224 | if (!drv) | |
19cb3738 | 3225 | return -ENOMEDIUM; |
faea38e7 FB |
3226 | if (!drv->bdrv_write_compressed) |
3227 | return -ENOTSUP; | |
fbb7b4e0 KW |
3228 | if (bdrv_check_request(bs, sector_num, nb_sectors)) |
3229 | return -EIO; | |
a55eb92c | 3230 | |
1755da16 | 3231 | assert(!bs->dirty_bitmap); |
a55eb92c | 3232 | |
faea38e7 FB |
3233 | return drv->bdrv_write_compressed(bs, sector_num, buf, nb_sectors); |
3234 | } | |
3b46e624 | 3235 | |
faea38e7 FB |
3236 | int bdrv_get_info(BlockDriverState *bs, BlockDriverInfo *bdi) |
3237 | { | |
3238 | BlockDriver *drv = bs->drv; | |
3239 | if (!drv) | |
19cb3738 | 3240 | return -ENOMEDIUM; |
faea38e7 FB |
3241 | if (!drv->bdrv_get_info) |
3242 | return -ENOTSUP; | |
3243 | memset(bdi, 0, sizeof(*bdi)); | |
3244 | return drv->bdrv_get_info(bs, bdi); | |
3245 | } | |
3246 | ||
45566e9c CH |
3247 | int bdrv_save_vmstate(BlockDriverState *bs, const uint8_t *buf, |
3248 | int64_t pos, int size) | |
cf8074b3 KW |
3249 | { |
3250 | QEMUIOVector qiov; | |
3251 | struct iovec iov = { | |
3252 | .iov_base = (void *) buf, | |
3253 | .iov_len = size, | |
3254 | }; | |
3255 | ||
3256 | qemu_iovec_init_external(&qiov, &iov, 1); | |
3257 | return bdrv_writev_vmstate(bs, &qiov, pos); | |
3258 | } | |
3259 | ||
3260 | int bdrv_writev_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos) | |
178e08a5 AL |
3261 | { |
3262 | BlockDriver *drv = bs->drv; | |
cf8074b3 KW |
3263 | |
3264 | if (!drv) { | |
178e08a5 | 3265 | return -ENOMEDIUM; |
cf8074b3 KW |
3266 | } else if (drv->bdrv_save_vmstate) { |
3267 | return drv->bdrv_save_vmstate(bs, qiov, pos); | |
3268 | } else if (bs->file) { | |
3269 | return bdrv_writev_vmstate(bs->file, qiov, pos); | |
3270 | } | |
3271 | ||
7cdb1f6d | 3272 | return -ENOTSUP; |
178e08a5 AL |
3273 | } |
3274 | ||
45566e9c CH |
3275 | int bdrv_load_vmstate(BlockDriverState *bs, uint8_t *buf, |
3276 | int64_t pos, int size) | |
178e08a5 AL |
3277 | { |
3278 | BlockDriver *drv = bs->drv; | |
3279 | if (!drv) | |
3280 | return -ENOMEDIUM; | |
7cdb1f6d MK |
3281 | if (drv->bdrv_load_vmstate) |
3282 | return drv->bdrv_load_vmstate(bs, buf, pos, size); | |
3283 | if (bs->file) | |
3284 | return bdrv_load_vmstate(bs->file, buf, pos, size); | |
3285 | return -ENOTSUP; | |
178e08a5 AL |
3286 | } |
3287 | ||
8b9b0cc2 KW |
3288 | void bdrv_debug_event(BlockDriverState *bs, BlkDebugEvent event) |
3289 | { | |
bf736fe3 | 3290 | if (!bs || !bs->drv || !bs->drv->bdrv_debug_event) { |
8b9b0cc2 KW |
3291 | return; |
3292 | } | |
3293 | ||
bf736fe3 | 3294 | bs->drv->bdrv_debug_event(bs, event); |
41c695c7 KW |
3295 | } |
3296 | ||
3297 | int bdrv_debug_breakpoint(BlockDriverState *bs, const char *event, | |
3298 | const char *tag) | |
3299 | { | |
3300 | while (bs && bs->drv && !bs->drv->bdrv_debug_breakpoint) { | |
3301 | bs = bs->file; | |
3302 | } | |
3303 | ||
3304 | if (bs && bs->drv && bs->drv->bdrv_debug_breakpoint) { | |
3305 | return bs->drv->bdrv_debug_breakpoint(bs, event, tag); | |
3306 | } | |
3307 | ||
3308 | return -ENOTSUP; | |
3309 | } | |
3310 | ||
3311 | int bdrv_debug_resume(BlockDriverState *bs, const char *tag) | |
3312 | { | |
3313 | while (bs && bs->drv && !bs->drv->bdrv_debug_resume) { | |
3314 | bs = bs->file; | |
3315 | } | |
8b9b0cc2 | 3316 | |
41c695c7 KW |
3317 | if (bs && bs->drv && bs->drv->bdrv_debug_resume) { |
3318 | return bs->drv->bdrv_debug_resume(bs, tag); | |
3319 | } | |
3320 | ||
3321 | return -ENOTSUP; | |
3322 | } | |
3323 | ||
3324 | bool bdrv_debug_is_suspended(BlockDriverState *bs, const char *tag) | |
3325 | { | |
3326 | while (bs && bs->drv && !bs->drv->bdrv_debug_is_suspended) { | |
3327 | bs = bs->file; | |
3328 | } | |
3329 | ||
3330 | if (bs && bs->drv && bs->drv->bdrv_debug_is_suspended) { | |
3331 | return bs->drv->bdrv_debug_is_suspended(bs, tag); | |
3332 | } | |
3333 | ||
3334 | return false; | |
8b9b0cc2 KW |
3335 | } |
3336 | ||
199630b6 BS |
3337 | int bdrv_is_snapshot(BlockDriverState *bs) |
3338 | { | |
3339 | return !!(bs->open_flags & BDRV_O_SNAPSHOT); | |
3340 | } | |
3341 | ||
b1b1d783 JC |
3342 | /* backing_file can either be relative, or absolute, or a protocol. If it is |
3343 | * relative, it must be relative to the chain. So, passing in bs->filename | |
3344 | * from a BDS as backing_file should not be done, as that may be relative to | |
3345 | * the CWD rather than the chain. */ | |
e8a6bb9c MT |
3346 | BlockDriverState *bdrv_find_backing_image(BlockDriverState *bs, |
3347 | const char *backing_file) | |
3348 | { | |
b1b1d783 JC |
3349 | char *filename_full = NULL; |
3350 | char *backing_file_full = NULL; | |
3351 | char *filename_tmp = NULL; | |
3352 | int is_protocol = 0; | |
3353 | BlockDriverState *curr_bs = NULL; | |
3354 | BlockDriverState *retval = NULL; | |
3355 | ||
3356 | if (!bs || !bs->drv || !backing_file) { | |
e8a6bb9c MT |
3357 | return NULL; |
3358 | } | |
3359 | ||
b1b1d783 JC |
3360 | filename_full = g_malloc(PATH_MAX); |
3361 | backing_file_full = g_malloc(PATH_MAX); | |
3362 | filename_tmp = g_malloc(PATH_MAX); | |
3363 | ||
3364 | is_protocol = path_has_protocol(backing_file); | |
3365 | ||
3366 | for (curr_bs = bs; curr_bs->backing_hd; curr_bs = curr_bs->backing_hd) { | |
3367 | ||
3368 | /* If either of the filename paths is actually a protocol, then | |
3369 | * compare unmodified paths; otherwise make paths relative */ | |
3370 | if (is_protocol || path_has_protocol(curr_bs->backing_file)) { | |
3371 | if (strcmp(backing_file, curr_bs->backing_file) == 0) { | |
3372 | retval = curr_bs->backing_hd; | |
3373 | break; | |
3374 | } | |
e8a6bb9c | 3375 | } else { |
b1b1d783 JC |
3376 | /* If not an absolute filename path, make it relative to the current |
3377 | * image's filename path */ | |
3378 | path_combine(filename_tmp, PATH_MAX, curr_bs->filename, | |
3379 | backing_file); | |
3380 | ||
3381 | /* We are going to compare absolute pathnames */ | |
3382 | if (!realpath(filename_tmp, filename_full)) { | |
3383 | continue; | |
3384 | } | |
3385 | ||
3386 | /* We need to make sure the backing filename we are comparing against | |
3387 | * is relative to the current image filename (or absolute) */ | |
3388 | path_combine(filename_tmp, PATH_MAX, curr_bs->filename, | |
3389 | curr_bs->backing_file); | |
3390 | ||
3391 | if (!realpath(filename_tmp, backing_file_full)) { | |
3392 | continue; | |
3393 | } | |
3394 | ||
3395 | if (strcmp(backing_file_full, filename_full) == 0) { | |
3396 | retval = curr_bs->backing_hd; | |
3397 | break; | |
3398 | } | |
e8a6bb9c MT |
3399 | } |
3400 | } | |
3401 | ||
b1b1d783 JC |
3402 | g_free(filename_full); |
3403 | g_free(backing_file_full); | |
3404 | g_free(filename_tmp); | |
3405 | return retval; | |
e8a6bb9c MT |
3406 | } |
3407 | ||
f198fd1c BC |
3408 | int bdrv_get_backing_file_depth(BlockDriverState *bs) |
3409 | { | |
3410 | if (!bs->drv) { | |
3411 | return 0; | |
3412 | } | |
3413 | ||
3414 | if (!bs->backing_hd) { | |
3415 | return 0; | |
3416 | } | |
3417 | ||
3418 | return 1 + bdrv_get_backing_file_depth(bs->backing_hd); | |
3419 | } | |
3420 | ||
79fac568 JC |
3421 | BlockDriverState *bdrv_find_base(BlockDriverState *bs) |
3422 | { | |
3423 | BlockDriverState *curr_bs = NULL; | |
3424 | ||
3425 | if (!bs) { | |
3426 | return NULL; | |
3427 | } | |
3428 | ||
3429 | curr_bs = bs; | |
3430 | ||
3431 | while (curr_bs->backing_hd) { | |
3432 | curr_bs = curr_bs->backing_hd; | |
3433 | } | |
3434 | return curr_bs; | |
3435 | } | |
3436 | ||
ea2384d3 | 3437 | /**************************************************************/ |
83f64091 | 3438 | /* async I/Os */ |
ea2384d3 | 3439 | |
3b69e4b9 | 3440 | BlockDriverAIOCB *bdrv_aio_readv(BlockDriverState *bs, int64_t sector_num, |
f141eafe | 3441 | QEMUIOVector *qiov, int nb_sectors, |
3b69e4b9 | 3442 | BlockDriverCompletionFunc *cb, void *opaque) |
83f64091 | 3443 | { |
bbf0a440 SH |
3444 | trace_bdrv_aio_readv(bs, sector_num, nb_sectors, opaque); |
3445 | ||
b2a61371 | 3446 | return bdrv_co_aio_rw_vector(bs, sector_num, qiov, nb_sectors, |
8c5873d6 | 3447 | cb, opaque, false); |
ea2384d3 FB |
3448 | } |
3449 | ||
f141eafe AL |
3450 | BlockDriverAIOCB *bdrv_aio_writev(BlockDriverState *bs, int64_t sector_num, |
3451 | QEMUIOVector *qiov, int nb_sectors, | |
3452 | BlockDriverCompletionFunc *cb, void *opaque) | |
ea2384d3 | 3453 | { |
bbf0a440 SH |
3454 | trace_bdrv_aio_writev(bs, sector_num, nb_sectors, opaque); |
3455 | ||
1a6e115b | 3456 | return bdrv_co_aio_rw_vector(bs, sector_num, qiov, nb_sectors, |
8c5873d6 | 3457 | cb, opaque, true); |
83f64091 FB |
3458 | } |
3459 | ||
40b4f539 KW |
3460 | |
3461 | typedef struct MultiwriteCB { | |
3462 | int error; | |
3463 | int num_requests; | |
3464 | int num_callbacks; | |
3465 | struct { | |
3466 | BlockDriverCompletionFunc *cb; | |
3467 | void *opaque; | |
3468 | QEMUIOVector *free_qiov; | |
40b4f539 KW |
3469 | } callbacks[]; |
3470 | } MultiwriteCB; | |
3471 | ||
3472 | static void multiwrite_user_cb(MultiwriteCB *mcb) | |
3473 | { | |
3474 | int i; | |
3475 | ||
3476 | for (i = 0; i < mcb->num_callbacks; i++) { | |
3477 | mcb->callbacks[i].cb(mcb->callbacks[i].opaque, mcb->error); | |
1e1ea48d SH |
3478 | if (mcb->callbacks[i].free_qiov) { |
3479 | qemu_iovec_destroy(mcb->callbacks[i].free_qiov); | |
3480 | } | |
7267c094 | 3481 | g_free(mcb->callbacks[i].free_qiov); |
40b4f539 KW |
3482 | } |
3483 | } | |
3484 | ||
3485 | static void multiwrite_cb(void *opaque, int ret) | |
3486 | { | |
3487 | MultiwriteCB *mcb = opaque; | |
3488 | ||
6d519a5f SH |
3489 | trace_multiwrite_cb(mcb, ret); |
3490 | ||
cb6d3ca0 | 3491 | if (ret < 0 && !mcb->error) { |
40b4f539 | 3492 | mcb->error = ret; |
40b4f539 KW |
3493 | } |
3494 | ||
3495 | mcb->num_requests--; | |
3496 | if (mcb->num_requests == 0) { | |
de189a1b | 3497 | multiwrite_user_cb(mcb); |
7267c094 | 3498 | g_free(mcb); |
40b4f539 KW |
3499 | } |
3500 | } | |
3501 | ||
3502 | static int multiwrite_req_compare(const void *a, const void *b) | |
3503 | { | |
77be4366 CH |
3504 | const BlockRequest *req1 = a, *req2 = b; |
3505 | ||
3506 | /* | |
3507 | * Note that we can't simply subtract req2->sector from req1->sector | |
3508 | * here as that could overflow the return value. | |
3509 | */ | |
3510 | if (req1->sector > req2->sector) { | |
3511 | return 1; | |
3512 | } else if (req1->sector < req2->sector) { | |
3513 | return -1; | |
3514 | } else { | |
3515 | return 0; | |
3516 | } | |
40b4f539 KW |
3517 | } |
3518 | ||
3519 | /* | |
3520 | * Takes a bunch of requests and tries to merge them. Returns the number of | |
3521 | * requests that remain after merging. | |
3522 | */ | |
3523 | static int multiwrite_merge(BlockDriverState *bs, BlockRequest *reqs, | |
3524 | int num_reqs, MultiwriteCB *mcb) | |
3525 | { | |
3526 | int i, outidx; | |
3527 | ||
3528 | // Sort requests by start sector | |
3529 | qsort(reqs, num_reqs, sizeof(*reqs), &multiwrite_req_compare); | |
3530 | ||
3531 | // Check if adjacent requests touch the same clusters. If so, combine them, | |
3532 | // filling up gaps with zero sectors. | |
3533 | outidx = 0; | |
3534 | for (i = 1; i < num_reqs; i++) { | |
3535 | int merge = 0; | |
3536 | int64_t oldreq_last = reqs[outidx].sector + reqs[outidx].nb_sectors; | |
3537 | ||
b6a127a1 | 3538 | // Handle exactly sequential writes and overlapping writes. |
40b4f539 KW |
3539 | if (reqs[i].sector <= oldreq_last) { |
3540 | merge = 1; | |
3541 | } | |
3542 | ||
e2a305fb CH |
3543 | if (reqs[outidx].qiov->niov + reqs[i].qiov->niov + 1 > IOV_MAX) { |
3544 | merge = 0; | |
3545 | } | |
3546 | ||
40b4f539 KW |
3547 | if (merge) { |
3548 | size_t size; | |
7267c094 | 3549 | QEMUIOVector *qiov = g_malloc0(sizeof(*qiov)); |
40b4f539 KW |
3550 | qemu_iovec_init(qiov, |
3551 | reqs[outidx].qiov->niov + reqs[i].qiov->niov + 1); | |
3552 | ||
3553 | // Add the first request to the merged one. If the requests are | |
3554 | // overlapping, drop the last sectors of the first request. | |
3555 | size = (reqs[i].sector - reqs[outidx].sector) << 9; | |
1b093c48 | 3556 | qemu_iovec_concat(qiov, reqs[outidx].qiov, 0, size); |
40b4f539 | 3557 | |
b6a127a1 PB |
3558 | // We should need to add any zeros between the two requests |
3559 | assert (reqs[i].sector <= oldreq_last); | |
40b4f539 KW |
3560 | |
3561 | // Add the second request | |
1b093c48 | 3562 | qemu_iovec_concat(qiov, reqs[i].qiov, 0, reqs[i].qiov->size); |
40b4f539 | 3563 | |
cbf1dff2 | 3564 | reqs[outidx].nb_sectors = qiov->size >> 9; |
40b4f539 KW |
3565 | reqs[outidx].qiov = qiov; |
3566 | ||
3567 | mcb->callbacks[i].free_qiov = reqs[outidx].qiov; | |
3568 | } else { | |
3569 | outidx++; | |
3570 | reqs[outidx].sector = reqs[i].sector; | |
3571 | reqs[outidx].nb_sectors = reqs[i].nb_sectors; | |
3572 | reqs[outidx].qiov = reqs[i].qiov; | |
3573 | } | |
3574 | } | |
3575 | ||
3576 | return outidx + 1; | |
3577 | } | |
3578 | ||
3579 | /* | |
3580 | * Submit multiple AIO write requests at once. | |
3581 | * | |
3582 | * On success, the function returns 0 and all requests in the reqs array have | |
3583 | * been submitted. In error case this function returns -1, and any of the | |
3584 | * requests may or may not be submitted yet. In particular, this means that the | |
3585 | * callback will be called for some of the requests, for others it won't. The | |
3586 | * caller must check the error field of the BlockRequest to wait for the right | |
3587 | * callbacks (if error != 0, no callback will be called). | |
3588 | * | |
3589 | * The implementation may modify the contents of the reqs array, e.g. to merge | |
3590 | * requests. However, the fields opaque and error are left unmodified as they | |
3591 | * are used to signal failure for a single request to the caller. | |
3592 | */ | |
3593 | int bdrv_aio_multiwrite(BlockDriverState *bs, BlockRequest *reqs, int num_reqs) | |
3594 | { | |
40b4f539 KW |
3595 | MultiwriteCB *mcb; |
3596 | int i; | |
3597 | ||
301db7c2 RH |
3598 | /* don't submit writes if we don't have a medium */ |
3599 | if (bs->drv == NULL) { | |
3600 | for (i = 0; i < num_reqs; i++) { | |
3601 | reqs[i].error = -ENOMEDIUM; | |
3602 | } | |
3603 | return -1; | |
3604 | } | |
3605 | ||
40b4f539 KW |
3606 | if (num_reqs == 0) { |
3607 | return 0; | |
3608 | } | |
3609 | ||
3610 | // Create MultiwriteCB structure | |
7267c094 | 3611 | mcb = g_malloc0(sizeof(*mcb) + num_reqs * sizeof(*mcb->callbacks)); |
40b4f539 KW |
3612 | mcb->num_requests = 0; |
3613 | mcb->num_callbacks = num_reqs; | |
3614 | ||
3615 | for (i = 0; i < num_reqs; i++) { | |
3616 | mcb->callbacks[i].cb = reqs[i].cb; | |
3617 | mcb->callbacks[i].opaque = reqs[i].opaque; | |
3618 | } | |
3619 | ||
3620 | // Check for mergable requests | |
3621 | num_reqs = multiwrite_merge(bs, reqs, num_reqs, mcb); | |
3622 | ||
6d519a5f SH |
3623 | trace_bdrv_aio_multiwrite(mcb, mcb->num_callbacks, num_reqs); |
3624 | ||
df9309fb PB |
3625 | /* Run the aio requests. */ |
3626 | mcb->num_requests = num_reqs; | |
40b4f539 | 3627 | for (i = 0; i < num_reqs; i++) { |
ad54ae80 | 3628 | bdrv_aio_writev(bs, reqs[i].sector, reqs[i].qiov, |
40b4f539 | 3629 | reqs[i].nb_sectors, multiwrite_cb, mcb); |
40b4f539 KW |
3630 | } |
3631 | ||
3632 | return 0; | |
40b4f539 KW |
3633 | } |
3634 | ||
83f64091 | 3635 | void bdrv_aio_cancel(BlockDriverAIOCB *acb) |
83f64091 | 3636 | { |
d7331bed | 3637 | acb->aiocb_info->cancel(acb); |
83f64091 FB |
3638 | } |
3639 | ||
3640 | /**************************************************************/ | |
3641 | /* async block device emulation */ | |
3642 | ||
c16b5a2c CH |
3643 | typedef struct BlockDriverAIOCBSync { |
3644 | BlockDriverAIOCB common; | |
3645 | QEMUBH *bh; | |
3646 | int ret; | |
3647 | /* vector translation state */ | |
3648 | QEMUIOVector *qiov; | |
3649 | uint8_t *bounce; | |
3650 | int is_write; | |
3651 | } BlockDriverAIOCBSync; | |
3652 | ||
3653 | static void bdrv_aio_cancel_em(BlockDriverAIOCB *blockacb) | |
3654 | { | |
b666d239 KW |
3655 | BlockDriverAIOCBSync *acb = |
3656 | container_of(blockacb, BlockDriverAIOCBSync, common); | |
6a7ad299 | 3657 | qemu_bh_delete(acb->bh); |
36afc451 | 3658 | acb->bh = NULL; |
c16b5a2c CH |
3659 | qemu_aio_release(acb); |
3660 | } | |
3661 | ||
d7331bed | 3662 | static const AIOCBInfo bdrv_em_aiocb_info = { |
c16b5a2c CH |
3663 | .aiocb_size = sizeof(BlockDriverAIOCBSync), |
3664 | .cancel = bdrv_aio_cancel_em, | |
3665 | }; | |
3666 | ||
ce1a14dc | 3667 | static void bdrv_aio_bh_cb(void *opaque) |
83f64091 | 3668 | { |
ce1a14dc | 3669 | BlockDriverAIOCBSync *acb = opaque; |
f141eafe | 3670 | |
f141eafe | 3671 | if (!acb->is_write) |
03396148 | 3672 | qemu_iovec_from_buf(acb->qiov, 0, acb->bounce, acb->qiov->size); |
ceb42de8 | 3673 | qemu_vfree(acb->bounce); |
ce1a14dc | 3674 | acb->common.cb(acb->common.opaque, acb->ret); |
6a7ad299 | 3675 | qemu_bh_delete(acb->bh); |
36afc451 | 3676 | acb->bh = NULL; |
ce1a14dc | 3677 | qemu_aio_release(acb); |
83f64091 | 3678 | } |
beac80cd | 3679 | |
f141eafe AL |
3680 | static BlockDriverAIOCB *bdrv_aio_rw_vector(BlockDriverState *bs, |
3681 | int64_t sector_num, | |
3682 | QEMUIOVector *qiov, | |
3683 | int nb_sectors, | |
3684 | BlockDriverCompletionFunc *cb, | |
3685 | void *opaque, | |
3686 | int is_write) | |
3687 | ||
83f64091 | 3688 | { |
ce1a14dc | 3689 | BlockDriverAIOCBSync *acb; |
ce1a14dc | 3690 | |
d7331bed | 3691 | acb = qemu_aio_get(&bdrv_em_aiocb_info, bs, cb, opaque); |
f141eafe AL |
3692 | acb->is_write = is_write; |
3693 | acb->qiov = qiov; | |
e268ca52 | 3694 | acb->bounce = qemu_blockalign(bs, qiov->size); |
3f3aace8 | 3695 | acb->bh = qemu_bh_new(bdrv_aio_bh_cb, acb); |
f141eafe AL |
3696 | |
3697 | if (is_write) { | |
d5e6b161 | 3698 | qemu_iovec_to_buf(acb->qiov, 0, acb->bounce, qiov->size); |
1ed20acf | 3699 | acb->ret = bs->drv->bdrv_write(bs, sector_num, acb->bounce, nb_sectors); |
f141eafe | 3700 | } else { |
1ed20acf | 3701 | acb->ret = bs->drv->bdrv_read(bs, sector_num, acb->bounce, nb_sectors); |
f141eafe AL |
3702 | } |
3703 | ||
ce1a14dc | 3704 | qemu_bh_schedule(acb->bh); |
f141eafe | 3705 | |
ce1a14dc | 3706 | return &acb->common; |
beac80cd FB |
3707 | } |
3708 | ||
f141eafe AL |
3709 | static BlockDriverAIOCB *bdrv_aio_readv_em(BlockDriverState *bs, |
3710 | int64_t sector_num, QEMUIOVector *qiov, int nb_sectors, | |
ce1a14dc | 3711 | BlockDriverCompletionFunc *cb, void *opaque) |
beac80cd | 3712 | { |
f141eafe AL |
3713 | return bdrv_aio_rw_vector(bs, sector_num, qiov, nb_sectors, cb, opaque, 0); |
3714 | } | |
83f64091 | 3715 | |
f141eafe AL |
3716 | static BlockDriverAIOCB *bdrv_aio_writev_em(BlockDriverState *bs, |
3717 | int64_t sector_num, QEMUIOVector *qiov, int nb_sectors, | |
3718 | BlockDriverCompletionFunc *cb, void *opaque) | |
3719 | { | |
3720 | return bdrv_aio_rw_vector(bs, sector_num, qiov, nb_sectors, cb, opaque, 1); | |
beac80cd | 3721 | } |
beac80cd | 3722 | |
68485420 KW |
3723 | |
3724 | typedef struct BlockDriverAIOCBCoroutine { | |
3725 | BlockDriverAIOCB common; | |
3726 | BlockRequest req; | |
3727 | bool is_write; | |
d318aea9 | 3728 | bool *done; |
68485420 KW |
3729 | QEMUBH* bh; |
3730 | } BlockDriverAIOCBCoroutine; | |
3731 | ||
3732 | static void bdrv_aio_co_cancel_em(BlockDriverAIOCB *blockacb) | |
3733 | { | |
d318aea9 KW |
3734 | BlockDriverAIOCBCoroutine *acb = |
3735 | container_of(blockacb, BlockDriverAIOCBCoroutine, common); | |
3736 | bool done = false; | |
3737 | ||
3738 | acb->done = &done; | |
3739 | while (!done) { | |
3740 | qemu_aio_wait(); | |
3741 | } | |
68485420 KW |
3742 | } |
3743 | ||
d7331bed | 3744 | static const AIOCBInfo bdrv_em_co_aiocb_info = { |
68485420 KW |
3745 | .aiocb_size = sizeof(BlockDriverAIOCBCoroutine), |
3746 | .cancel = bdrv_aio_co_cancel_em, | |
3747 | }; | |
3748 | ||
35246a68 | 3749 | static void bdrv_co_em_bh(void *opaque) |
68485420 KW |
3750 | { |
3751 | BlockDriverAIOCBCoroutine *acb = opaque; | |
3752 | ||
3753 | acb->common.cb(acb->common.opaque, acb->req.error); | |
d318aea9 KW |
3754 | |
3755 | if (acb->done) { | |
3756 | *acb->done = true; | |
3757 | } | |
3758 | ||
68485420 KW |
3759 | qemu_bh_delete(acb->bh); |
3760 | qemu_aio_release(acb); | |
3761 | } | |
3762 | ||
b2a61371 SH |
3763 | /* Invoke bdrv_co_do_readv/bdrv_co_do_writev */ |
3764 | static void coroutine_fn bdrv_co_do_rw(void *opaque) | |
3765 | { | |
3766 | BlockDriverAIOCBCoroutine *acb = opaque; | |
3767 | BlockDriverState *bs = acb->common.bs; | |
3768 | ||
3769 | if (!acb->is_write) { | |
3770 | acb->req.error = bdrv_co_do_readv(bs, acb->req.sector, | |
470c0504 | 3771 | acb->req.nb_sectors, acb->req.qiov, 0); |
b2a61371 SH |
3772 | } else { |
3773 | acb->req.error = bdrv_co_do_writev(bs, acb->req.sector, | |
f08f2dda | 3774 | acb->req.nb_sectors, acb->req.qiov, 0); |
b2a61371 SH |
3775 | } |
3776 | ||
35246a68 | 3777 | acb->bh = qemu_bh_new(bdrv_co_em_bh, acb); |
b2a61371 SH |
3778 | qemu_bh_schedule(acb->bh); |
3779 | } | |
3780 | ||
68485420 KW |
3781 | static BlockDriverAIOCB *bdrv_co_aio_rw_vector(BlockDriverState *bs, |
3782 | int64_t sector_num, | |
3783 | QEMUIOVector *qiov, | |
3784 | int nb_sectors, | |
3785 | BlockDriverCompletionFunc *cb, | |
3786 | void *opaque, | |
8c5873d6 | 3787 | bool is_write) |
68485420 KW |
3788 | { |
3789 | Coroutine *co; | |
3790 | BlockDriverAIOCBCoroutine *acb; | |
3791 | ||
d7331bed | 3792 | acb = qemu_aio_get(&bdrv_em_co_aiocb_info, bs, cb, opaque); |
68485420 KW |
3793 | acb->req.sector = sector_num; |
3794 | acb->req.nb_sectors = nb_sectors; | |
3795 | acb->req.qiov = qiov; | |
3796 | acb->is_write = is_write; | |
d318aea9 | 3797 | acb->done = NULL; |
68485420 | 3798 | |
8c5873d6 | 3799 | co = qemu_coroutine_create(bdrv_co_do_rw); |
68485420 KW |
3800 | qemu_coroutine_enter(co, acb); |
3801 | ||
3802 | return &acb->common; | |
3803 | } | |
3804 | ||
07f07615 | 3805 | static void coroutine_fn bdrv_aio_flush_co_entry(void *opaque) |
b2e12bc6 | 3806 | { |
07f07615 PB |
3807 | BlockDriverAIOCBCoroutine *acb = opaque; |
3808 | BlockDriverState *bs = acb->common.bs; | |
b2e12bc6 | 3809 | |
07f07615 PB |
3810 | acb->req.error = bdrv_co_flush(bs); |
3811 | acb->bh = qemu_bh_new(bdrv_co_em_bh, acb); | |
b2e12bc6 | 3812 | qemu_bh_schedule(acb->bh); |
b2e12bc6 CH |
3813 | } |
3814 | ||
07f07615 | 3815 | BlockDriverAIOCB *bdrv_aio_flush(BlockDriverState *bs, |
016f5cf6 AG |
3816 | BlockDriverCompletionFunc *cb, void *opaque) |
3817 | { | |
07f07615 | 3818 | trace_bdrv_aio_flush(bs, opaque); |
016f5cf6 | 3819 | |
07f07615 PB |
3820 | Coroutine *co; |
3821 | BlockDriverAIOCBCoroutine *acb; | |
016f5cf6 | 3822 | |
d7331bed | 3823 | acb = qemu_aio_get(&bdrv_em_co_aiocb_info, bs, cb, opaque); |
d318aea9 KW |
3824 | acb->done = NULL; |
3825 | ||
07f07615 PB |
3826 | co = qemu_coroutine_create(bdrv_aio_flush_co_entry); |
3827 | qemu_coroutine_enter(co, acb); | |
016f5cf6 | 3828 | |
016f5cf6 AG |
3829 | return &acb->common; |
3830 | } | |
3831 | ||
4265d620 PB |
3832 | static void coroutine_fn bdrv_aio_discard_co_entry(void *opaque) |
3833 | { | |
3834 | BlockDriverAIOCBCoroutine *acb = opaque; | |
3835 | BlockDriverState *bs = acb->common.bs; | |
3836 | ||
3837 | acb->req.error = bdrv_co_discard(bs, acb->req.sector, acb->req.nb_sectors); | |
3838 | acb->bh = qemu_bh_new(bdrv_co_em_bh, acb); | |
3839 | qemu_bh_schedule(acb->bh); | |
3840 | } | |
3841 | ||
3842 | BlockDriverAIOCB *bdrv_aio_discard(BlockDriverState *bs, | |
3843 | int64_t sector_num, int nb_sectors, | |
3844 | BlockDriverCompletionFunc *cb, void *opaque) | |
3845 | { | |
3846 | Coroutine *co; | |
3847 | BlockDriverAIOCBCoroutine *acb; | |
3848 | ||
3849 | trace_bdrv_aio_discard(bs, sector_num, nb_sectors, opaque); | |
3850 | ||
d7331bed | 3851 | acb = qemu_aio_get(&bdrv_em_co_aiocb_info, bs, cb, opaque); |
4265d620 PB |
3852 | acb->req.sector = sector_num; |
3853 | acb->req.nb_sectors = nb_sectors; | |
d318aea9 | 3854 | acb->done = NULL; |
4265d620 PB |
3855 | co = qemu_coroutine_create(bdrv_aio_discard_co_entry); |
3856 | qemu_coroutine_enter(co, acb); | |
3857 | ||
3858 | return &acb->common; | |
3859 | } | |
3860 | ||
ea2384d3 FB |
3861 | void bdrv_init(void) |
3862 | { | |
5efa9d5a | 3863 | module_call_init(MODULE_INIT_BLOCK); |
ea2384d3 | 3864 | } |
ce1a14dc | 3865 | |
eb852011 MA |
3866 | void bdrv_init_with_whitelist(void) |
3867 | { | |
3868 | use_bdrv_whitelist = 1; | |
3869 | bdrv_init(); | |
3870 | } | |
3871 | ||
d7331bed | 3872 | void *qemu_aio_get(const AIOCBInfo *aiocb_info, BlockDriverState *bs, |
c16b5a2c | 3873 | BlockDriverCompletionFunc *cb, void *opaque) |
ce1a14dc | 3874 | { |
ce1a14dc PB |
3875 | BlockDriverAIOCB *acb; |
3876 | ||
d7331bed SH |
3877 | acb = g_slice_alloc(aiocb_info->aiocb_size); |
3878 | acb->aiocb_info = aiocb_info; | |
ce1a14dc PB |
3879 | acb->bs = bs; |
3880 | acb->cb = cb; | |
3881 | acb->opaque = opaque; | |
3882 | return acb; | |
3883 | } | |
3884 | ||
3885 | void qemu_aio_release(void *p) | |
3886 | { | |
d37c975f | 3887 | BlockDriverAIOCB *acb = p; |
d7331bed | 3888 | g_slice_free1(acb->aiocb_info->aiocb_size, acb); |
ce1a14dc | 3889 | } |
19cb3738 | 3890 | |
f9f05dc5 KW |
3891 | /**************************************************************/ |
3892 | /* Coroutine block device emulation */ | |
3893 | ||
3894 | typedef struct CoroutineIOCompletion { | |
3895 | Coroutine *coroutine; | |
3896 | int ret; | |
3897 | } CoroutineIOCompletion; | |
3898 | ||
3899 | static void bdrv_co_io_em_complete(void *opaque, int ret) | |
3900 | { | |
3901 | CoroutineIOCompletion *co = opaque; | |
3902 | ||
3903 | co->ret = ret; | |
3904 | qemu_coroutine_enter(co->coroutine, NULL); | |
3905 | } | |
3906 | ||
3907 | static int coroutine_fn bdrv_co_io_em(BlockDriverState *bs, int64_t sector_num, | |
3908 | int nb_sectors, QEMUIOVector *iov, | |
3909 | bool is_write) | |
3910 | { | |
3911 | CoroutineIOCompletion co = { | |
3912 | .coroutine = qemu_coroutine_self(), | |
3913 | }; | |
3914 | BlockDriverAIOCB *acb; | |
3915 | ||
3916 | if (is_write) { | |
a652d160 SH |
3917 | acb = bs->drv->bdrv_aio_writev(bs, sector_num, iov, nb_sectors, |
3918 | bdrv_co_io_em_complete, &co); | |
f9f05dc5 | 3919 | } else { |
a652d160 SH |
3920 | acb = bs->drv->bdrv_aio_readv(bs, sector_num, iov, nb_sectors, |
3921 | bdrv_co_io_em_complete, &co); | |
f9f05dc5 KW |
3922 | } |
3923 | ||
59370aaa | 3924 | trace_bdrv_co_io_em(bs, sector_num, nb_sectors, is_write, acb); |
f9f05dc5 KW |
3925 | if (!acb) { |
3926 | return -EIO; | |
3927 | } | |
3928 | qemu_coroutine_yield(); | |
3929 | ||
3930 | return co.ret; | |
3931 | } | |
3932 | ||
3933 | static int coroutine_fn bdrv_co_readv_em(BlockDriverState *bs, | |
3934 | int64_t sector_num, int nb_sectors, | |
3935 | QEMUIOVector *iov) | |
3936 | { | |
3937 | return bdrv_co_io_em(bs, sector_num, nb_sectors, iov, false); | |
3938 | } | |
3939 | ||
3940 | static int coroutine_fn bdrv_co_writev_em(BlockDriverState *bs, | |
3941 | int64_t sector_num, int nb_sectors, | |
3942 | QEMUIOVector *iov) | |
3943 | { | |
3944 | return bdrv_co_io_em(bs, sector_num, nb_sectors, iov, true); | |
3945 | } | |
3946 | ||
07f07615 | 3947 | static void coroutine_fn bdrv_flush_co_entry(void *opaque) |
e7a8a783 | 3948 | { |
07f07615 PB |
3949 | RwCo *rwco = opaque; |
3950 | ||
3951 | rwco->ret = bdrv_co_flush(rwco->bs); | |
3952 | } | |
3953 | ||
3954 | int coroutine_fn bdrv_co_flush(BlockDriverState *bs) | |
3955 | { | |
eb489bb1 KW |
3956 | int ret; |
3957 | ||
29cdb251 | 3958 | if (!bs || !bdrv_is_inserted(bs) || bdrv_is_read_only(bs)) { |
07f07615 | 3959 | return 0; |
eb489bb1 KW |
3960 | } |
3961 | ||
ca716364 | 3962 | /* Write back cached data to the OS even with cache=unsafe */ |
bf736fe3 | 3963 | BLKDBG_EVENT(bs->file, BLKDBG_FLUSH_TO_OS); |
eb489bb1 KW |
3964 | if (bs->drv->bdrv_co_flush_to_os) { |
3965 | ret = bs->drv->bdrv_co_flush_to_os(bs); | |
3966 | if (ret < 0) { | |
3967 | return ret; | |
3968 | } | |
3969 | } | |
3970 | ||
ca716364 KW |
3971 | /* But don't actually force it to the disk with cache=unsafe */ |
3972 | if (bs->open_flags & BDRV_O_NO_FLUSH) { | |
d4c82329 | 3973 | goto flush_parent; |
ca716364 KW |
3974 | } |
3975 | ||
bf736fe3 | 3976 | BLKDBG_EVENT(bs->file, BLKDBG_FLUSH_TO_DISK); |
eb489bb1 | 3977 | if (bs->drv->bdrv_co_flush_to_disk) { |
29cdb251 | 3978 | ret = bs->drv->bdrv_co_flush_to_disk(bs); |
07f07615 PB |
3979 | } else if (bs->drv->bdrv_aio_flush) { |
3980 | BlockDriverAIOCB *acb; | |
3981 | CoroutineIOCompletion co = { | |
3982 | .coroutine = qemu_coroutine_self(), | |
3983 | }; | |
3984 | ||
3985 | acb = bs->drv->bdrv_aio_flush(bs, bdrv_co_io_em_complete, &co); | |
3986 | if (acb == NULL) { | |
29cdb251 | 3987 | ret = -EIO; |
07f07615 PB |
3988 | } else { |
3989 | qemu_coroutine_yield(); | |
29cdb251 | 3990 | ret = co.ret; |
07f07615 | 3991 | } |
07f07615 PB |
3992 | } else { |
3993 | /* | |
3994 | * Some block drivers always operate in either writethrough or unsafe | |
3995 | * mode and don't support bdrv_flush therefore. Usually qemu doesn't | |
3996 | * know how the server works (because the behaviour is hardcoded or | |
3997 | * depends on server-side configuration), so we can't ensure that | |
3998 | * everything is safe on disk. Returning an error doesn't work because | |
3999 | * that would break guests even if the server operates in writethrough | |
4000 | * mode. | |
4001 | * | |
4002 | * Let's hope the user knows what he's doing. | |
4003 | */ | |
29cdb251 | 4004 | ret = 0; |
07f07615 | 4005 | } |
29cdb251 PB |
4006 | if (ret < 0) { |
4007 | return ret; | |
4008 | } | |
4009 | ||
4010 | /* Now flush the underlying protocol. It will also have BDRV_O_NO_FLUSH | |
4011 | * in the case of cache=unsafe, so there are no useless flushes. | |
4012 | */ | |
d4c82329 | 4013 | flush_parent: |
29cdb251 | 4014 | return bdrv_co_flush(bs->file); |
07f07615 PB |
4015 | } |
4016 | ||
0f15423c AL |
4017 | void bdrv_invalidate_cache(BlockDriverState *bs) |
4018 | { | |
4019 | if (bs->drv && bs->drv->bdrv_invalidate_cache) { | |
4020 | bs->drv->bdrv_invalidate_cache(bs); | |
4021 | } | |
4022 | } | |
4023 | ||
4024 | void bdrv_invalidate_cache_all(void) | |
4025 | { | |
4026 | BlockDriverState *bs; | |
4027 | ||
4028 | QTAILQ_FOREACH(bs, &bdrv_states, list) { | |
4029 | bdrv_invalidate_cache(bs); | |
4030 | } | |
4031 | } | |
4032 | ||
07789269 BC |
4033 | void bdrv_clear_incoming_migration_all(void) |
4034 | { | |
4035 | BlockDriverState *bs; | |
4036 | ||
4037 | QTAILQ_FOREACH(bs, &bdrv_states, list) { | |
4038 | bs->open_flags = bs->open_flags & ~(BDRV_O_INCOMING); | |
4039 | } | |
4040 | } | |
4041 | ||
07f07615 PB |
4042 | int bdrv_flush(BlockDriverState *bs) |
4043 | { | |
4044 | Coroutine *co; | |
4045 | RwCo rwco = { | |
4046 | .bs = bs, | |
4047 | .ret = NOT_DONE, | |
e7a8a783 | 4048 | }; |
e7a8a783 | 4049 | |
07f07615 PB |
4050 | if (qemu_in_coroutine()) { |
4051 | /* Fast-path if already in coroutine context */ | |
4052 | bdrv_flush_co_entry(&rwco); | |
4053 | } else { | |
4054 | co = qemu_coroutine_create(bdrv_flush_co_entry); | |
4055 | qemu_coroutine_enter(co, &rwco); | |
4056 | while (rwco.ret == NOT_DONE) { | |
4057 | qemu_aio_wait(); | |
4058 | } | |
e7a8a783 | 4059 | } |
07f07615 PB |
4060 | |
4061 | return rwco.ret; | |
e7a8a783 KW |
4062 | } |
4063 | ||
4265d620 PB |
4064 | static void coroutine_fn bdrv_discard_co_entry(void *opaque) |
4065 | { | |
4066 | RwCo *rwco = opaque; | |
4067 | ||
4068 | rwco->ret = bdrv_co_discard(rwco->bs, rwco->sector_num, rwco->nb_sectors); | |
4069 | } | |
4070 | ||
4071 | int coroutine_fn bdrv_co_discard(BlockDriverState *bs, int64_t sector_num, | |
4072 | int nb_sectors) | |
4073 | { | |
4074 | if (!bs->drv) { | |
4075 | return -ENOMEDIUM; | |
4076 | } else if (bdrv_check_request(bs, sector_num, nb_sectors)) { | |
4077 | return -EIO; | |
4078 | } else if (bs->read_only) { | |
4079 | return -EROFS; | |
df702c9b PB |
4080 | } |
4081 | ||
4082 | if (bs->dirty_bitmap) { | |
8f0720ec | 4083 | bdrv_reset_dirty(bs, sector_num, nb_sectors); |
df702c9b PB |
4084 | } |
4085 | ||
9e8f1835 PB |
4086 | /* Do nothing if disabled. */ |
4087 | if (!(bs->open_flags & BDRV_O_UNMAP)) { | |
4088 | return 0; | |
4089 | } | |
4090 | ||
df702c9b | 4091 | if (bs->drv->bdrv_co_discard) { |
4265d620 PB |
4092 | return bs->drv->bdrv_co_discard(bs, sector_num, nb_sectors); |
4093 | } else if (bs->drv->bdrv_aio_discard) { | |
4094 | BlockDriverAIOCB *acb; | |
4095 | CoroutineIOCompletion co = { | |
4096 | .coroutine = qemu_coroutine_self(), | |
4097 | }; | |
4098 | ||
4099 | acb = bs->drv->bdrv_aio_discard(bs, sector_num, nb_sectors, | |
4100 | bdrv_co_io_em_complete, &co); | |
4101 | if (acb == NULL) { | |
4102 | return -EIO; | |
4103 | } else { | |
4104 | qemu_coroutine_yield(); | |
4105 | return co.ret; | |
4106 | } | |
4265d620 PB |
4107 | } else { |
4108 | return 0; | |
4109 | } | |
4110 | } | |
4111 | ||
4112 | int bdrv_discard(BlockDriverState *bs, int64_t sector_num, int nb_sectors) | |
4113 | { | |
4114 | Coroutine *co; | |
4115 | RwCo rwco = { | |
4116 | .bs = bs, | |
4117 | .sector_num = sector_num, | |
4118 | .nb_sectors = nb_sectors, | |
4119 | .ret = NOT_DONE, | |
4120 | }; | |
4121 | ||
4122 | if (qemu_in_coroutine()) { | |
4123 | /* Fast-path if already in coroutine context */ | |
4124 | bdrv_discard_co_entry(&rwco); | |
4125 | } else { | |
4126 | co = qemu_coroutine_create(bdrv_discard_co_entry); | |
4127 | qemu_coroutine_enter(co, &rwco); | |
4128 | while (rwco.ret == NOT_DONE) { | |
4129 | qemu_aio_wait(); | |
4130 | } | |
4131 | } | |
4132 | ||
4133 | return rwco.ret; | |
4134 | } | |
4135 | ||
19cb3738 FB |
4136 | /**************************************************************/ |
4137 | /* removable device support */ | |
4138 | ||
4139 | /** | |
4140 | * Return TRUE if the media is present | |
4141 | */ | |
4142 | int bdrv_is_inserted(BlockDriverState *bs) | |
4143 | { | |
4144 | BlockDriver *drv = bs->drv; | |
a1aff5bf | 4145 | |
19cb3738 FB |
4146 | if (!drv) |
4147 | return 0; | |
4148 | if (!drv->bdrv_is_inserted) | |
a1aff5bf MA |
4149 | return 1; |
4150 | return drv->bdrv_is_inserted(bs); | |
19cb3738 FB |
4151 | } |
4152 | ||
4153 | /** | |
8e49ca46 MA |
4154 | * Return whether the media changed since the last call to this |
4155 | * function, or -ENOTSUP if we don't know. Most drivers don't know. | |
19cb3738 FB |
4156 | */ |
4157 | int bdrv_media_changed(BlockDriverState *bs) | |
4158 | { | |
4159 | BlockDriver *drv = bs->drv; | |
19cb3738 | 4160 | |
8e49ca46 MA |
4161 | if (drv && drv->bdrv_media_changed) { |
4162 | return drv->bdrv_media_changed(bs); | |
4163 | } | |
4164 | return -ENOTSUP; | |
19cb3738 FB |
4165 | } |
4166 | ||
4167 | /** | |
4168 | * If eject_flag is TRUE, eject the media. Otherwise, close the tray | |
4169 | */ | |
f36f3949 | 4170 | void bdrv_eject(BlockDriverState *bs, bool eject_flag) |
19cb3738 FB |
4171 | { |
4172 | BlockDriver *drv = bs->drv; | |
19cb3738 | 4173 | |
822e1cd1 MA |
4174 | if (drv && drv->bdrv_eject) { |
4175 | drv->bdrv_eject(bs, eject_flag); | |
19cb3738 | 4176 | } |
6f382ed2 LC |
4177 | |
4178 | if (bs->device_name[0] != '\0') { | |
4179 | bdrv_emit_qmp_eject_event(bs, eject_flag); | |
4180 | } | |
19cb3738 FB |
4181 | } |
4182 | ||
19cb3738 FB |
4183 | /** |
4184 | * Lock or unlock the media (if it is locked, the user won't be able | |
4185 | * to eject it manually). | |
4186 | */ | |
025e849a | 4187 | void bdrv_lock_medium(BlockDriverState *bs, bool locked) |
19cb3738 FB |
4188 | { |
4189 | BlockDriver *drv = bs->drv; | |
4190 | ||
025e849a | 4191 | trace_bdrv_lock_medium(bs, locked); |
b8c6d095 | 4192 | |
025e849a MA |
4193 | if (drv && drv->bdrv_lock_medium) { |
4194 | drv->bdrv_lock_medium(bs, locked); | |
19cb3738 FB |
4195 | } |
4196 | } | |
985a03b0 TS |
4197 | |
4198 | /* needed for generic scsi interface */ | |
4199 | ||
4200 | int bdrv_ioctl(BlockDriverState *bs, unsigned long int req, void *buf) | |
4201 | { | |
4202 | BlockDriver *drv = bs->drv; | |
4203 | ||
4204 | if (drv && drv->bdrv_ioctl) | |
4205 | return drv->bdrv_ioctl(bs, req, buf); | |
4206 | return -ENOTSUP; | |
4207 | } | |
7d780669 | 4208 | |
221f715d AL |
4209 | BlockDriverAIOCB *bdrv_aio_ioctl(BlockDriverState *bs, |
4210 | unsigned long int req, void *buf, | |
4211 | BlockDriverCompletionFunc *cb, void *opaque) | |
7d780669 | 4212 | { |
221f715d | 4213 | BlockDriver *drv = bs->drv; |
7d780669 | 4214 | |
221f715d AL |
4215 | if (drv && drv->bdrv_aio_ioctl) |
4216 | return drv->bdrv_aio_ioctl(bs, req, buf, cb, opaque); | |
4217 | return NULL; | |
7d780669 | 4218 | } |
e268ca52 | 4219 | |
7b6f9300 MA |
4220 | void bdrv_set_buffer_alignment(BlockDriverState *bs, int align) |
4221 | { | |
4222 | bs->buffer_alignment = align; | |
4223 | } | |
7cd1e32a | 4224 | |
e268ca52 AL |
4225 | void *qemu_blockalign(BlockDriverState *bs, size_t size) |
4226 | { | |
4227 | return qemu_memalign((bs && bs->buffer_alignment) ? bs->buffer_alignment : 512, size); | |
4228 | } | |
7cd1e32a | 4229 | |
c53b1c51 SH |
4230 | /* |
4231 | * Check if all memory in this vector is sector aligned. | |
4232 | */ | |
4233 | bool bdrv_qiov_is_aligned(BlockDriverState *bs, QEMUIOVector *qiov) | |
4234 | { | |
4235 | int i; | |
4236 | ||
4237 | for (i = 0; i < qiov->niov; i++) { | |
4238 | if ((uintptr_t) qiov->iov[i].iov_base % bs->buffer_alignment) { | |
4239 | return false; | |
4240 | } | |
4241 | } | |
4242 | ||
4243 | return true; | |
4244 | } | |
4245 | ||
50717e94 | 4246 | void bdrv_set_dirty_tracking(BlockDriverState *bs, int granularity) |
7cd1e32a LS |
4247 | { |
4248 | int64_t bitmap_size; | |
a55eb92c | 4249 | |
50717e94 PB |
4250 | assert((granularity & (granularity - 1)) == 0); |
4251 | ||
4252 | if (granularity) { | |
4253 | granularity >>= BDRV_SECTOR_BITS; | |
4254 | assert(!bs->dirty_bitmap); | |
4255 | bitmap_size = (bdrv_getlength(bs) >> BDRV_SECTOR_BITS); | |
4256 | bs->dirty_bitmap = hbitmap_alloc(bitmap_size, ffs(granularity) - 1); | |
7cd1e32a | 4257 | } else { |
c6d22830 | 4258 | if (bs->dirty_bitmap) { |
8f0720ec | 4259 | hbitmap_free(bs->dirty_bitmap); |
c6d22830 | 4260 | bs->dirty_bitmap = NULL; |
a55eb92c | 4261 | } |
7cd1e32a LS |
4262 | } |
4263 | } | |
4264 | ||
4265 | int bdrv_get_dirty(BlockDriverState *bs, int64_t sector) | |
4266 | { | |
8f0720ec PB |
4267 | if (bs->dirty_bitmap) { |
4268 | return hbitmap_get(bs->dirty_bitmap, sector); | |
7cd1e32a LS |
4269 | } else { |
4270 | return 0; | |
4271 | } | |
4272 | } | |
4273 | ||
8f0720ec | 4274 | void bdrv_dirty_iter_init(BlockDriverState *bs, HBitmapIter *hbi) |
1755da16 | 4275 | { |
8f0720ec | 4276 | hbitmap_iter_init(hbi, bs->dirty_bitmap, 0); |
1755da16 PB |
4277 | } |
4278 | ||
4279 | void bdrv_set_dirty(BlockDriverState *bs, int64_t cur_sector, | |
4280 | int nr_sectors) | |
4281 | { | |
8f0720ec | 4282 | hbitmap_set(bs->dirty_bitmap, cur_sector, nr_sectors); |
1755da16 PB |
4283 | } |
4284 | ||
a55eb92c JK |
4285 | void bdrv_reset_dirty(BlockDriverState *bs, int64_t cur_sector, |
4286 | int nr_sectors) | |
7cd1e32a | 4287 | { |
8f0720ec | 4288 | hbitmap_reset(bs->dirty_bitmap, cur_sector, nr_sectors); |
7cd1e32a | 4289 | } |
aaa0eb75 LS |
4290 | |
4291 | int64_t bdrv_get_dirty_count(BlockDriverState *bs) | |
4292 | { | |
8f0720ec | 4293 | if (bs->dirty_bitmap) { |
acc906c6 | 4294 | return hbitmap_count(bs->dirty_bitmap); |
8f0720ec PB |
4295 | } else { |
4296 | return 0; | |
4297 | } | |
aaa0eb75 | 4298 | } |
f88e1a42 | 4299 | |
db593f25 MT |
4300 | void bdrv_set_in_use(BlockDriverState *bs, int in_use) |
4301 | { | |
4302 | assert(bs->in_use != in_use); | |
4303 | bs->in_use = in_use; | |
4304 | } | |
4305 | ||
4306 | int bdrv_in_use(BlockDriverState *bs) | |
4307 | { | |
4308 | return bs->in_use; | |
4309 | } | |
4310 | ||
28a7282a LC |
4311 | void bdrv_iostatus_enable(BlockDriverState *bs) |
4312 | { | |
d6bf279e | 4313 | bs->iostatus_enabled = true; |
58e21ef5 | 4314 | bs->iostatus = BLOCK_DEVICE_IO_STATUS_OK; |
28a7282a LC |
4315 | } |
4316 | ||
4317 | /* The I/O status is only enabled if the drive explicitly | |
4318 | * enables it _and_ the VM is configured to stop on errors */ | |
4319 | bool bdrv_iostatus_is_enabled(const BlockDriverState *bs) | |
4320 | { | |
d6bf279e | 4321 | return (bs->iostatus_enabled && |
92aa5c6d PB |
4322 | (bs->on_write_error == BLOCKDEV_ON_ERROR_ENOSPC || |
4323 | bs->on_write_error == BLOCKDEV_ON_ERROR_STOP || | |
4324 | bs->on_read_error == BLOCKDEV_ON_ERROR_STOP)); | |
28a7282a LC |
4325 | } |
4326 | ||
4327 | void bdrv_iostatus_disable(BlockDriverState *bs) | |
4328 | { | |
d6bf279e | 4329 | bs->iostatus_enabled = false; |
28a7282a LC |
4330 | } |
4331 | ||
4332 | void bdrv_iostatus_reset(BlockDriverState *bs) | |
4333 | { | |
4334 | if (bdrv_iostatus_is_enabled(bs)) { | |
58e21ef5 | 4335 | bs->iostatus = BLOCK_DEVICE_IO_STATUS_OK; |
3bd293c3 PB |
4336 | if (bs->job) { |
4337 | block_job_iostatus_reset(bs->job); | |
4338 | } | |
28a7282a LC |
4339 | } |
4340 | } | |
4341 | ||
28a7282a LC |
4342 | void bdrv_iostatus_set_err(BlockDriverState *bs, int error) |
4343 | { | |
3e1caa5f PB |
4344 | assert(bdrv_iostatus_is_enabled(bs)); |
4345 | if (bs->iostatus == BLOCK_DEVICE_IO_STATUS_OK) { | |
58e21ef5 LC |
4346 | bs->iostatus = error == ENOSPC ? BLOCK_DEVICE_IO_STATUS_NOSPACE : |
4347 | BLOCK_DEVICE_IO_STATUS_FAILED; | |
28a7282a LC |
4348 | } |
4349 | } | |
4350 | ||
a597e79c CH |
4351 | void |
4352 | bdrv_acct_start(BlockDriverState *bs, BlockAcctCookie *cookie, int64_t bytes, | |
4353 | enum BlockAcctType type) | |
4354 | { | |
4355 | assert(type < BDRV_MAX_IOTYPE); | |
4356 | ||
4357 | cookie->bytes = bytes; | |
c488c7f6 | 4358 | cookie->start_time_ns = get_clock(); |
a597e79c CH |
4359 | cookie->type = type; |
4360 | } | |
4361 | ||
4362 | void | |
4363 | bdrv_acct_done(BlockDriverState *bs, BlockAcctCookie *cookie) | |
4364 | { | |
4365 | assert(cookie->type < BDRV_MAX_IOTYPE); | |
4366 | ||
4367 | bs->nr_bytes[cookie->type] += cookie->bytes; | |
4368 | bs->nr_ops[cookie->type]++; | |
c488c7f6 | 4369 | bs->total_time_ns[cookie->type] += get_clock() - cookie->start_time_ns; |
a597e79c CH |
4370 | } |
4371 | ||
d92ada22 LC |
4372 | void bdrv_img_create(const char *filename, const char *fmt, |
4373 | const char *base_filename, const char *base_fmt, | |
f382d43a MR |
4374 | char *options, uint64_t img_size, int flags, |
4375 | Error **errp, bool quiet) | |
f88e1a42 JS |
4376 | { |
4377 | QEMUOptionParameter *param = NULL, *create_options = NULL; | |
d220894e | 4378 | QEMUOptionParameter *backing_fmt, *backing_file, *size; |
f88e1a42 JS |
4379 | BlockDriverState *bs = NULL; |
4380 | BlockDriver *drv, *proto_drv; | |
96df67d1 | 4381 | BlockDriver *backing_drv = NULL; |
f88e1a42 JS |
4382 | int ret = 0; |
4383 | ||
4384 | /* Find driver and parse its options */ | |
4385 | drv = bdrv_find_format(fmt); | |
4386 | if (!drv) { | |
71c79813 | 4387 | error_setg(errp, "Unknown file format '%s'", fmt); |
d92ada22 | 4388 | return; |
f88e1a42 JS |
4389 | } |
4390 | ||
98289620 | 4391 | proto_drv = bdrv_find_protocol(filename, true); |
f88e1a42 | 4392 | if (!proto_drv) { |
71c79813 | 4393 | error_setg(errp, "Unknown protocol '%s'", filename); |
d92ada22 | 4394 | return; |
f88e1a42 JS |
4395 | } |
4396 | ||
4397 | create_options = append_option_parameters(create_options, | |
4398 | drv->create_options); | |
4399 | create_options = append_option_parameters(create_options, | |
4400 | proto_drv->create_options); | |
4401 | ||
4402 | /* Create parameter list with default values */ | |
4403 | param = parse_option_parameters("", create_options, param); | |
4404 | ||
4405 | set_option_parameter_int(param, BLOCK_OPT_SIZE, img_size); | |
4406 | ||
4407 | /* Parse -o options */ | |
4408 | if (options) { | |
4409 | param = parse_option_parameters(options, create_options, param); | |
4410 | if (param == NULL) { | |
71c79813 | 4411 | error_setg(errp, "Invalid options for file format '%s'.", fmt); |
f88e1a42 JS |
4412 | goto out; |
4413 | } | |
4414 | } | |
4415 | ||
4416 | if (base_filename) { | |
4417 | if (set_option_parameter(param, BLOCK_OPT_BACKING_FILE, | |
4418 | base_filename)) { | |
71c79813 LC |
4419 | error_setg(errp, "Backing file not supported for file format '%s'", |
4420 | fmt); | |
f88e1a42 JS |
4421 | goto out; |
4422 | } | |
4423 | } | |
4424 | ||
4425 | if (base_fmt) { | |
4426 | if (set_option_parameter(param, BLOCK_OPT_BACKING_FMT, base_fmt)) { | |
71c79813 LC |
4427 | error_setg(errp, "Backing file format not supported for file " |
4428 | "format '%s'", fmt); | |
f88e1a42 JS |
4429 | goto out; |
4430 | } | |
4431 | } | |
4432 | ||
792da93a JS |
4433 | backing_file = get_option_parameter(param, BLOCK_OPT_BACKING_FILE); |
4434 | if (backing_file && backing_file->value.s) { | |
4435 | if (!strcmp(filename, backing_file->value.s)) { | |
71c79813 LC |
4436 | error_setg(errp, "Error: Trying to create an image with the " |
4437 | "same filename as the backing file"); | |
792da93a JS |
4438 | goto out; |
4439 | } | |
4440 | } | |
4441 | ||
f88e1a42 JS |
4442 | backing_fmt = get_option_parameter(param, BLOCK_OPT_BACKING_FMT); |
4443 | if (backing_fmt && backing_fmt->value.s) { | |
96df67d1 SH |
4444 | backing_drv = bdrv_find_format(backing_fmt->value.s); |
4445 | if (!backing_drv) { | |
71c79813 LC |
4446 | error_setg(errp, "Unknown backing file format '%s'", |
4447 | backing_fmt->value.s); | |
f88e1a42 JS |
4448 | goto out; |
4449 | } | |
4450 | } | |
4451 | ||
4452 | // The size for the image must always be specified, with one exception: | |
4453 | // If we are using a backing file, we can obtain the size from there | |
d220894e KW |
4454 | size = get_option_parameter(param, BLOCK_OPT_SIZE); |
4455 | if (size && size->value.n == -1) { | |
f88e1a42 JS |
4456 | if (backing_file && backing_file->value.s) { |
4457 | uint64_t size; | |
f88e1a42 | 4458 | char buf[32]; |
63090dac PB |
4459 | int back_flags; |
4460 | ||
4461 | /* backing files always opened read-only */ | |
4462 | back_flags = | |
4463 | flags & ~(BDRV_O_RDWR | BDRV_O_SNAPSHOT | BDRV_O_NO_BACKING); | |
f88e1a42 | 4464 | |
f88e1a42 JS |
4465 | bs = bdrv_new(""); |
4466 | ||
de9c0cec KW |
4467 | ret = bdrv_open(bs, backing_file->value.s, NULL, back_flags, |
4468 | backing_drv); | |
f88e1a42 | 4469 | if (ret < 0) { |
71c79813 LC |
4470 | error_setg_errno(errp, -ret, "Could not open '%s'", |
4471 | backing_file->value.s); | |
f88e1a42 JS |
4472 | goto out; |
4473 | } | |
4474 | bdrv_get_geometry(bs, &size); | |
4475 | size *= 512; | |
4476 | ||
4477 | snprintf(buf, sizeof(buf), "%" PRId64, size); | |
4478 | set_option_parameter(param, BLOCK_OPT_SIZE, buf); | |
4479 | } else { | |
71c79813 | 4480 | error_setg(errp, "Image creation needs a size parameter"); |
f88e1a42 JS |
4481 | goto out; |
4482 | } | |
4483 | } | |
4484 | ||
f382d43a MR |
4485 | if (!quiet) { |
4486 | printf("Formatting '%s', fmt=%s ", filename, fmt); | |
4487 | print_option_parameters(param); | |
4488 | puts(""); | |
4489 | } | |
f88e1a42 | 4490 | ret = bdrv_create(drv, filename, param); |
f88e1a42 JS |
4491 | if (ret < 0) { |
4492 | if (ret == -ENOTSUP) { | |
71c79813 LC |
4493 | error_setg(errp,"Formatting or formatting option not supported for " |
4494 | "file format '%s'", fmt); | |
f88e1a42 | 4495 | } else if (ret == -EFBIG) { |
f3f4d2c0 KW |
4496 | const char *cluster_size_hint = ""; |
4497 | if (get_option_parameter(create_options, BLOCK_OPT_CLUSTER_SIZE)) { | |
4498 | cluster_size_hint = " (try using a larger cluster size)"; | |
4499 | } | |
4500 | error_setg(errp, "The image size is too large for file format '%s'%s", | |
4501 | fmt, cluster_size_hint); | |
f88e1a42 | 4502 | } else { |
71c79813 LC |
4503 | error_setg(errp, "%s: error while creating %s: %s", filename, fmt, |
4504 | strerror(-ret)); | |
f88e1a42 JS |
4505 | } |
4506 | } | |
4507 | ||
4508 | out: | |
4509 | free_option_parameters(create_options); | |
4510 | free_option_parameters(param); | |
4511 | ||
4512 | if (bs) { | |
4513 | bdrv_delete(bs); | |
4514 | } | |
f88e1a42 | 4515 | } |
85d126f3 SH |
4516 | |
4517 | AioContext *bdrv_get_aio_context(BlockDriverState *bs) | |
4518 | { | |
4519 | /* Currently BlockDriverState always uses the main loop AioContext */ | |
4520 | return qemu_get_aio_context(); | |
4521 | } | |
d616b224 SH |
4522 | |
4523 | void bdrv_add_before_write_notifier(BlockDriverState *bs, | |
4524 | NotifierWithReturn *notifier) | |
4525 | { | |
4526 | notifier_with_return_list_add(&bs->before_write_notifiers, notifier); | |
4527 | } |