]>
Commit | Line | Data |
---|---|---|
fc01f7e7 FB |
1 | /* |
2 | * QEMU System Emulator block driver | |
5fafdf24 | 3 | * |
fc01f7e7 | 4 | * Copyright (c) 2003 Fabrice Bellard |
5fafdf24 | 5 | * |
fc01f7e7 FB |
6 | * Permission is hereby granted, free of charge, to any person obtaining a copy |
7 | * of this software and associated documentation files (the "Software"), to deal | |
8 | * in the Software without restriction, including without limitation the rights | |
9 | * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell | |
10 | * copies of the Software, and to permit persons to whom the Software is | |
11 | * furnished to do so, subject to the following conditions: | |
12 | * | |
13 | * The above copyright notice and this permission notice shall be included in | |
14 | * all copies or substantial portions of the Software. | |
15 | * | |
16 | * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR | |
17 | * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, | |
18 | * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL | |
19 | * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER | |
20 | * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, | |
21 | * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN | |
22 | * THE SOFTWARE. | |
23 | */ | |
3990d09a | 24 | #include "config-host.h" |
faf07963 | 25 | #include "qemu-common.h" |
6d519a5f | 26 | #include "trace.h" |
737e150e PB |
27 | #include "block/block_int.h" |
28 | #include "block/blockjob.h" | |
1de7afc9 | 29 | #include "qemu/module.h" |
7b1b5d19 | 30 | #include "qapi/qmp/qjson.h" |
bfb197e0 | 31 | #include "sysemu/block-backend.h" |
9c17d615 | 32 | #include "sysemu/sysemu.h" |
1de7afc9 | 33 | #include "qemu/notify.h" |
737e150e | 34 | #include "block/coroutine.h" |
c13163fb | 35 | #include "block/qapi.h" |
b2023818 | 36 | #include "qmp-commands.h" |
1de7afc9 | 37 | #include "qemu/timer.h" |
a5ee7bd4 | 38 | #include "qapi-event.h" |
fc01f7e7 | 39 | |
71e72a19 | 40 | #ifdef CONFIG_BSD |
7674e7bf FB |
41 | #include <sys/types.h> |
42 | #include <sys/stat.h> | |
43 | #include <sys/ioctl.h> | |
72cf2d4f | 44 | #include <sys/queue.h> |
c5e97233 | 45 | #ifndef __DragonFly__ |
7674e7bf FB |
46 | #include <sys/disk.h> |
47 | #endif | |
c5e97233 | 48 | #endif |
7674e7bf | 49 | |
49dc768d AL |
50 | #ifdef _WIN32 |
51 | #include <windows.h> | |
52 | #endif | |
53 | ||
e4654d2d FZ |
54 | struct BdrvDirtyBitmap { |
55 | HBitmap *bitmap; | |
56 | QLIST_ENTRY(BdrvDirtyBitmap) list; | |
57 | }; | |
58 | ||
1c9805a3 SH |
59 | #define NOT_DONE 0x7fffffff /* used while emulated sync operation in progress */ |
60 | ||
7c84b1b8 | 61 | static BlockAIOCB *bdrv_aio_readv_em(BlockDriverState *bs, |
f141eafe | 62 | int64_t sector_num, QEMUIOVector *qiov, int nb_sectors, |
097310b5 | 63 | BlockCompletionFunc *cb, void *opaque); |
7c84b1b8 | 64 | static BlockAIOCB *bdrv_aio_writev_em(BlockDriverState *bs, |
f141eafe | 65 | int64_t sector_num, QEMUIOVector *qiov, int nb_sectors, |
097310b5 | 66 | BlockCompletionFunc *cb, void *opaque); |
f9f05dc5 KW |
67 | static int coroutine_fn bdrv_co_readv_em(BlockDriverState *bs, |
68 | int64_t sector_num, int nb_sectors, | |
69 | QEMUIOVector *iov); | |
70 | static int coroutine_fn bdrv_co_writev_em(BlockDriverState *bs, | |
71 | int64_t sector_num, int nb_sectors, | |
72 | QEMUIOVector *iov); | |
775aa8b6 KW |
73 | static int coroutine_fn bdrv_co_do_preadv(BlockDriverState *bs, |
74 | int64_t offset, unsigned int bytes, QEMUIOVector *qiov, | |
470c0504 | 75 | BdrvRequestFlags flags); |
775aa8b6 KW |
76 | static int coroutine_fn bdrv_co_do_pwritev(BlockDriverState *bs, |
77 | int64_t offset, unsigned int bytes, QEMUIOVector *qiov, | |
f08f2dda | 78 | BdrvRequestFlags flags); |
7c84b1b8 MA |
79 | static BlockAIOCB *bdrv_co_aio_rw_vector(BlockDriverState *bs, |
80 | int64_t sector_num, | |
81 | QEMUIOVector *qiov, | |
82 | int nb_sectors, | |
83 | BdrvRequestFlags flags, | |
097310b5 | 84 | BlockCompletionFunc *cb, |
7c84b1b8 MA |
85 | void *opaque, |
86 | bool is_write); | |
b2a61371 | 87 | static void coroutine_fn bdrv_co_do_rw(void *opaque); |
621f0589 | 88 | static int coroutine_fn bdrv_co_do_write_zeroes(BlockDriverState *bs, |
aa7bfbff | 89 | int64_t sector_num, int nb_sectors, BdrvRequestFlags flags); |
ec530c81 | 90 | |
1b7bdbc1 SH |
91 | static QTAILQ_HEAD(, BlockDriverState) bdrv_states = |
92 | QTAILQ_HEAD_INITIALIZER(bdrv_states); | |
7ee930d0 | 93 | |
dc364f4c BC |
94 | static QTAILQ_HEAD(, BlockDriverState) graph_bdrv_states = |
95 | QTAILQ_HEAD_INITIALIZER(graph_bdrv_states); | |
96 | ||
8a22f02a SH |
97 | static QLIST_HEAD(, BlockDriver) bdrv_drivers = |
98 | QLIST_HEAD_INITIALIZER(bdrv_drivers); | |
ea2384d3 | 99 | |
c4237dfa VSO |
100 | static void bdrv_set_dirty(BlockDriverState *bs, int64_t cur_sector, |
101 | int nr_sectors); | |
102 | static void bdrv_reset_dirty(BlockDriverState *bs, int64_t cur_sector, | |
103 | int nr_sectors); | |
eb852011 MA |
104 | /* If non-zero, use only whitelisted block drivers */ |
105 | static int use_bdrv_whitelist; | |
106 | ||
9e0b22f4 SH |
107 | #ifdef _WIN32 |
108 | static int is_windows_drive_prefix(const char *filename) | |
109 | { | |
110 | return (((filename[0] >= 'a' && filename[0] <= 'z') || | |
111 | (filename[0] >= 'A' && filename[0] <= 'Z')) && | |
112 | filename[1] == ':'); | |
113 | } | |
114 | ||
115 | int is_windows_drive(const char *filename) | |
116 | { | |
117 | if (is_windows_drive_prefix(filename) && | |
118 | filename[2] == '\0') | |
119 | return 1; | |
120 | if (strstart(filename, "\\\\.\\", NULL) || | |
121 | strstart(filename, "//./", NULL)) | |
122 | return 1; | |
123 | return 0; | |
124 | } | |
125 | #endif | |
126 | ||
0563e191 | 127 | /* throttling disk I/O limits */ |
cc0681c4 BC |
128 | void bdrv_set_io_limits(BlockDriverState *bs, |
129 | ThrottleConfig *cfg) | |
98f90dba | 130 | { |
cc0681c4 | 131 | int i; |
98f90dba | 132 | |
cc0681c4 | 133 | throttle_config(&bs->throttle_state, cfg); |
98f90dba | 134 | |
cc0681c4 BC |
135 | for (i = 0; i < 2; i++) { |
136 | qemu_co_enter_next(&bs->throttled_reqs[i]); | |
98f90dba | 137 | } |
cc0681c4 BC |
138 | } |
139 | ||
140 | /* this function drain all the throttled IOs */ | |
141 | static bool bdrv_start_throttled_reqs(BlockDriverState *bs) | |
142 | { | |
143 | bool drained = false; | |
144 | bool enabled = bs->io_limits_enabled; | |
145 | int i; | |
146 | ||
147 | bs->io_limits_enabled = false; | |
148 | ||
149 | for (i = 0; i < 2; i++) { | |
150 | while (qemu_co_enter_next(&bs->throttled_reqs[i])) { | |
151 | drained = true; | |
152 | } | |
153 | } | |
154 | ||
155 | bs->io_limits_enabled = enabled; | |
98f90dba | 156 | |
cc0681c4 | 157 | return drained; |
98f90dba ZYW |
158 | } |
159 | ||
cc0681c4 | 160 | void bdrv_io_limits_disable(BlockDriverState *bs) |
0563e191 | 161 | { |
cc0681c4 | 162 | bs->io_limits_enabled = false; |
0563e191 | 163 | |
cc0681c4 BC |
164 | bdrv_start_throttled_reqs(bs); |
165 | ||
166 | throttle_destroy(&bs->throttle_state); | |
0563e191 ZYW |
167 | } |
168 | ||
cc0681c4 | 169 | static void bdrv_throttle_read_timer_cb(void *opaque) |
0563e191 | 170 | { |
cc0681c4 BC |
171 | BlockDriverState *bs = opaque; |
172 | qemu_co_enter_next(&bs->throttled_reqs[0]); | |
0563e191 ZYW |
173 | } |
174 | ||
cc0681c4 | 175 | static void bdrv_throttle_write_timer_cb(void *opaque) |
0563e191 | 176 | { |
cc0681c4 BC |
177 | BlockDriverState *bs = opaque; |
178 | qemu_co_enter_next(&bs->throttled_reqs[1]); | |
0563e191 ZYW |
179 | } |
180 | ||
cc0681c4 BC |
181 | /* should be called before bdrv_set_io_limits if a limit is set */ |
182 | void bdrv_io_limits_enable(BlockDriverState *bs) | |
183 | { | |
184 | assert(!bs->io_limits_enabled); | |
185 | throttle_init(&bs->throttle_state, | |
13af91eb | 186 | bdrv_get_aio_context(bs), |
cc0681c4 BC |
187 | QEMU_CLOCK_VIRTUAL, |
188 | bdrv_throttle_read_timer_cb, | |
189 | bdrv_throttle_write_timer_cb, | |
190 | bs); | |
191 | bs->io_limits_enabled = true; | |
192 | } | |
193 | ||
194 | /* This function makes an IO wait if needed | |
195 | * | |
196 | * @nb_sectors: the number of sectors of the IO | |
197 | * @is_write: is the IO a write | |
198 | */ | |
98f90dba | 199 | static void bdrv_io_limits_intercept(BlockDriverState *bs, |
d5103588 | 200 | unsigned int bytes, |
cc0681c4 | 201 | bool is_write) |
98f90dba | 202 | { |
cc0681c4 BC |
203 | /* does this io must wait */ |
204 | bool must_wait = throttle_schedule_timer(&bs->throttle_state, is_write); | |
98f90dba | 205 | |
cc0681c4 BC |
206 | /* if must wait or any request of this type throttled queue the IO */ |
207 | if (must_wait || | |
208 | !qemu_co_queue_empty(&bs->throttled_reqs[is_write])) { | |
209 | qemu_co_queue_wait(&bs->throttled_reqs[is_write]); | |
98f90dba ZYW |
210 | } |
211 | ||
cc0681c4 | 212 | /* the IO will be executed, do the accounting */ |
d5103588 KW |
213 | throttle_account(&bs->throttle_state, is_write, bytes); |
214 | ||
98f90dba | 215 | |
cc0681c4 BC |
216 | /* if the next request must wait -> do nothing */ |
217 | if (throttle_schedule_timer(&bs->throttle_state, is_write)) { | |
218 | return; | |
98f90dba ZYW |
219 | } |
220 | ||
cc0681c4 BC |
221 | /* else queue next request for execution */ |
222 | qemu_co_queue_next(&bs->throttled_reqs[is_write]); | |
98f90dba ZYW |
223 | } |
224 | ||
339064d5 KW |
225 | size_t bdrv_opt_mem_align(BlockDriverState *bs) |
226 | { | |
227 | if (!bs || !bs->drv) { | |
228 | /* 4k should be on the safe side */ | |
229 | return 4096; | |
230 | } | |
231 | ||
232 | return bs->bl.opt_mem_alignment; | |
233 | } | |
234 | ||
9e0b22f4 | 235 | /* check if the path starts with "<protocol>:" */ |
5c98415b | 236 | int path_has_protocol(const char *path) |
9e0b22f4 | 237 | { |
947995c0 PB |
238 | const char *p; |
239 | ||
9e0b22f4 SH |
240 | #ifdef _WIN32 |
241 | if (is_windows_drive(path) || | |
242 | is_windows_drive_prefix(path)) { | |
243 | return 0; | |
244 | } | |
947995c0 PB |
245 | p = path + strcspn(path, ":/\\"); |
246 | #else | |
247 | p = path + strcspn(path, ":/"); | |
9e0b22f4 SH |
248 | #endif |
249 | ||
947995c0 | 250 | return *p == ':'; |
9e0b22f4 SH |
251 | } |
252 | ||
83f64091 | 253 | int path_is_absolute(const char *path) |
3b0d4f61 | 254 | { |
21664424 FB |
255 | #ifdef _WIN32 |
256 | /* specific case for names like: "\\.\d:" */ | |
f53f4da9 | 257 | if (is_windows_drive(path) || is_windows_drive_prefix(path)) { |
21664424 | 258 | return 1; |
f53f4da9 PB |
259 | } |
260 | return (*path == '/' || *path == '\\'); | |
3b9f94e1 | 261 | #else |
f53f4da9 | 262 | return (*path == '/'); |
3b9f94e1 | 263 | #endif |
3b0d4f61 FB |
264 | } |
265 | ||
83f64091 FB |
266 | /* if filename is absolute, just copy it to dest. Otherwise, build a |
267 | path to it by considering it is relative to base_path. URL are | |
268 | supported. */ | |
269 | void path_combine(char *dest, int dest_size, | |
270 | const char *base_path, | |
271 | const char *filename) | |
3b0d4f61 | 272 | { |
83f64091 FB |
273 | const char *p, *p1; |
274 | int len; | |
275 | ||
276 | if (dest_size <= 0) | |
277 | return; | |
278 | if (path_is_absolute(filename)) { | |
279 | pstrcpy(dest, dest_size, filename); | |
280 | } else { | |
281 | p = strchr(base_path, ':'); | |
282 | if (p) | |
283 | p++; | |
284 | else | |
285 | p = base_path; | |
3b9f94e1 FB |
286 | p1 = strrchr(base_path, '/'); |
287 | #ifdef _WIN32 | |
288 | { | |
289 | const char *p2; | |
290 | p2 = strrchr(base_path, '\\'); | |
291 | if (!p1 || p2 > p1) | |
292 | p1 = p2; | |
293 | } | |
294 | #endif | |
83f64091 FB |
295 | if (p1) |
296 | p1++; | |
297 | else | |
298 | p1 = base_path; | |
299 | if (p1 > p) | |
300 | p = p1; | |
301 | len = p - base_path; | |
302 | if (len > dest_size - 1) | |
303 | len = dest_size - 1; | |
304 | memcpy(dest, base_path, len); | |
305 | dest[len] = '\0'; | |
306 | pstrcat(dest, dest_size, filename); | |
3b0d4f61 | 307 | } |
3b0d4f61 FB |
308 | } |
309 | ||
0a82855a HR |
310 | void bdrv_get_full_backing_filename_from_filename(const char *backed, |
311 | const char *backing, | |
9f07429e HR |
312 | char *dest, size_t sz, |
313 | Error **errp) | |
dc5a1371 | 314 | { |
9f07429e HR |
315 | if (backing[0] == '\0' || path_has_protocol(backing) || |
316 | path_is_absolute(backing)) | |
317 | { | |
0a82855a | 318 | pstrcpy(dest, sz, backing); |
9f07429e HR |
319 | } else if (backed[0] == '\0' || strstart(backed, "json:", NULL)) { |
320 | error_setg(errp, "Cannot use relative backing file names for '%s'", | |
321 | backed); | |
dc5a1371 | 322 | } else { |
0a82855a | 323 | path_combine(dest, sz, backed, backing); |
dc5a1371 PB |
324 | } |
325 | } | |
326 | ||
9f07429e HR |
327 | void bdrv_get_full_backing_filename(BlockDriverState *bs, char *dest, size_t sz, |
328 | Error **errp) | |
0a82855a | 329 | { |
9f07429e HR |
330 | char *backed = bs->exact_filename[0] ? bs->exact_filename : bs->filename; |
331 | ||
332 | bdrv_get_full_backing_filename_from_filename(backed, bs->backing_file, | |
333 | dest, sz, errp); | |
0a82855a HR |
334 | } |
335 | ||
5efa9d5a | 336 | void bdrv_register(BlockDriver *bdrv) |
ea2384d3 | 337 | { |
8c5873d6 SH |
338 | /* Block drivers without coroutine functions need emulation */ |
339 | if (!bdrv->bdrv_co_readv) { | |
f9f05dc5 KW |
340 | bdrv->bdrv_co_readv = bdrv_co_readv_em; |
341 | bdrv->bdrv_co_writev = bdrv_co_writev_em; | |
342 | ||
f8c35c1d SH |
343 | /* bdrv_co_readv_em()/brdv_co_writev_em() work in terms of aio, so if |
344 | * the block driver lacks aio we need to emulate that too. | |
345 | */ | |
f9f05dc5 KW |
346 | if (!bdrv->bdrv_aio_readv) { |
347 | /* add AIO emulation layer */ | |
348 | bdrv->bdrv_aio_readv = bdrv_aio_readv_em; | |
349 | bdrv->bdrv_aio_writev = bdrv_aio_writev_em; | |
f9f05dc5 | 350 | } |
83f64091 | 351 | } |
b2e12bc6 | 352 | |
8a22f02a | 353 | QLIST_INSERT_HEAD(&bdrv_drivers, bdrv, list); |
ea2384d3 | 354 | } |
b338082b | 355 | |
7f06d47e | 356 | BlockDriverState *bdrv_new_root(void) |
b338082b | 357 | { |
7f06d47e | 358 | BlockDriverState *bs = bdrv_new(); |
e4e9986b | 359 | |
e4e9986b | 360 | QTAILQ_INSERT_TAIL(&bdrv_states, bs, device_list); |
e4e9986b MA |
361 | return bs; |
362 | } | |
363 | ||
364 | BlockDriverState *bdrv_new(void) | |
365 | { | |
366 | BlockDriverState *bs; | |
367 | int i; | |
368 | ||
5839e53b | 369 | bs = g_new0(BlockDriverState, 1); |
e4654d2d | 370 | QLIST_INIT(&bs->dirty_bitmaps); |
fbe40ff7 FZ |
371 | for (i = 0; i < BLOCK_OP_TYPE_MAX; i++) { |
372 | QLIST_INIT(&bs->op_blockers[i]); | |
373 | } | |
28a7282a | 374 | bdrv_iostatus_disable(bs); |
d7d512f6 | 375 | notifier_list_init(&bs->close_notifiers); |
d616b224 | 376 | notifier_with_return_list_init(&bs->before_write_notifiers); |
cc0681c4 BC |
377 | qemu_co_queue_init(&bs->throttled_reqs[0]); |
378 | qemu_co_queue_init(&bs->throttled_reqs[1]); | |
9fcb0251 | 379 | bs->refcnt = 1; |
dcd04228 | 380 | bs->aio_context = qemu_get_aio_context(); |
d7d512f6 | 381 | |
b338082b FB |
382 | return bs; |
383 | } | |
384 | ||
d7d512f6 PB |
385 | void bdrv_add_close_notifier(BlockDriverState *bs, Notifier *notify) |
386 | { | |
387 | notifier_list_add(&bs->close_notifiers, notify); | |
388 | } | |
389 | ||
ea2384d3 FB |
390 | BlockDriver *bdrv_find_format(const char *format_name) |
391 | { | |
392 | BlockDriver *drv1; | |
8a22f02a SH |
393 | QLIST_FOREACH(drv1, &bdrv_drivers, list) { |
394 | if (!strcmp(drv1->format_name, format_name)) { | |
ea2384d3 | 395 | return drv1; |
8a22f02a | 396 | } |
ea2384d3 FB |
397 | } |
398 | return NULL; | |
399 | } | |
400 | ||
b64ec4e4 | 401 | static int bdrv_is_whitelisted(BlockDriver *drv, bool read_only) |
eb852011 | 402 | { |
b64ec4e4 FZ |
403 | static const char *whitelist_rw[] = { |
404 | CONFIG_BDRV_RW_WHITELIST | |
405 | }; | |
406 | static const char *whitelist_ro[] = { | |
407 | CONFIG_BDRV_RO_WHITELIST | |
eb852011 MA |
408 | }; |
409 | const char **p; | |
410 | ||
b64ec4e4 | 411 | if (!whitelist_rw[0] && !whitelist_ro[0]) { |
eb852011 | 412 | return 1; /* no whitelist, anything goes */ |
b64ec4e4 | 413 | } |
eb852011 | 414 | |
b64ec4e4 | 415 | for (p = whitelist_rw; *p; p++) { |
eb852011 MA |
416 | if (!strcmp(drv->format_name, *p)) { |
417 | return 1; | |
418 | } | |
419 | } | |
b64ec4e4 FZ |
420 | if (read_only) { |
421 | for (p = whitelist_ro; *p; p++) { | |
422 | if (!strcmp(drv->format_name, *p)) { | |
423 | return 1; | |
424 | } | |
425 | } | |
426 | } | |
eb852011 MA |
427 | return 0; |
428 | } | |
429 | ||
b64ec4e4 FZ |
430 | BlockDriver *bdrv_find_whitelisted_format(const char *format_name, |
431 | bool read_only) | |
eb852011 MA |
432 | { |
433 | BlockDriver *drv = bdrv_find_format(format_name); | |
b64ec4e4 | 434 | return drv && bdrv_is_whitelisted(drv, read_only) ? drv : NULL; |
eb852011 MA |
435 | } |
436 | ||
5b7e1542 ZYW |
437 | typedef struct CreateCo { |
438 | BlockDriver *drv; | |
439 | char *filename; | |
83d0521a | 440 | QemuOpts *opts; |
5b7e1542 | 441 | int ret; |
cc84d90f | 442 | Error *err; |
5b7e1542 ZYW |
443 | } CreateCo; |
444 | ||
445 | static void coroutine_fn bdrv_create_co_entry(void *opaque) | |
446 | { | |
cc84d90f HR |
447 | Error *local_err = NULL; |
448 | int ret; | |
449 | ||
5b7e1542 ZYW |
450 | CreateCo *cco = opaque; |
451 | assert(cco->drv); | |
452 | ||
c282e1fd | 453 | ret = cco->drv->bdrv_create(cco->filename, cco->opts, &local_err); |
84d18f06 | 454 | if (local_err) { |
cc84d90f HR |
455 | error_propagate(&cco->err, local_err); |
456 | } | |
457 | cco->ret = ret; | |
5b7e1542 ZYW |
458 | } |
459 | ||
0e7e1989 | 460 | int bdrv_create(BlockDriver *drv, const char* filename, |
83d0521a | 461 | QemuOpts *opts, Error **errp) |
ea2384d3 | 462 | { |
5b7e1542 ZYW |
463 | int ret; |
464 | ||
465 | Coroutine *co; | |
466 | CreateCo cco = { | |
467 | .drv = drv, | |
468 | .filename = g_strdup(filename), | |
83d0521a | 469 | .opts = opts, |
5b7e1542 | 470 | .ret = NOT_DONE, |
cc84d90f | 471 | .err = NULL, |
5b7e1542 ZYW |
472 | }; |
473 | ||
c282e1fd | 474 | if (!drv->bdrv_create) { |
cc84d90f | 475 | error_setg(errp, "Driver '%s' does not support image creation", drv->format_name); |
80168bff LC |
476 | ret = -ENOTSUP; |
477 | goto out; | |
5b7e1542 ZYW |
478 | } |
479 | ||
480 | if (qemu_in_coroutine()) { | |
481 | /* Fast-path if already in coroutine context */ | |
482 | bdrv_create_co_entry(&cco); | |
483 | } else { | |
484 | co = qemu_coroutine_create(bdrv_create_co_entry); | |
485 | qemu_coroutine_enter(co, &cco); | |
486 | while (cco.ret == NOT_DONE) { | |
b47ec2c4 | 487 | aio_poll(qemu_get_aio_context(), true); |
5b7e1542 ZYW |
488 | } |
489 | } | |
490 | ||
491 | ret = cco.ret; | |
cc84d90f | 492 | if (ret < 0) { |
84d18f06 | 493 | if (cco.err) { |
cc84d90f HR |
494 | error_propagate(errp, cco.err); |
495 | } else { | |
496 | error_setg_errno(errp, -ret, "Could not create image"); | |
497 | } | |
498 | } | |
0e7e1989 | 499 | |
80168bff LC |
500 | out: |
501 | g_free(cco.filename); | |
5b7e1542 | 502 | return ret; |
ea2384d3 FB |
503 | } |
504 | ||
c282e1fd | 505 | int bdrv_create_file(const char *filename, QemuOpts *opts, Error **errp) |
84a12e66 CH |
506 | { |
507 | BlockDriver *drv; | |
cc84d90f HR |
508 | Error *local_err = NULL; |
509 | int ret; | |
84a12e66 | 510 | |
b65a5e12 | 511 | drv = bdrv_find_protocol(filename, true, errp); |
84a12e66 | 512 | if (drv == NULL) { |
16905d71 | 513 | return -ENOENT; |
84a12e66 CH |
514 | } |
515 | ||
c282e1fd | 516 | ret = bdrv_create(drv, filename, opts, &local_err); |
84d18f06 | 517 | if (local_err) { |
cc84d90f HR |
518 | error_propagate(errp, local_err); |
519 | } | |
520 | return ret; | |
84a12e66 CH |
521 | } |
522 | ||
3baca891 | 523 | void bdrv_refresh_limits(BlockDriverState *bs, Error **errp) |
d34682cd KW |
524 | { |
525 | BlockDriver *drv = bs->drv; | |
3baca891 | 526 | Error *local_err = NULL; |
d34682cd KW |
527 | |
528 | memset(&bs->bl, 0, sizeof(bs->bl)); | |
529 | ||
466ad822 | 530 | if (!drv) { |
3baca891 | 531 | return; |
466ad822 KW |
532 | } |
533 | ||
534 | /* Take some limits from the children as a default */ | |
535 | if (bs->file) { | |
3baca891 KW |
536 | bdrv_refresh_limits(bs->file, &local_err); |
537 | if (local_err) { | |
538 | error_propagate(errp, local_err); | |
539 | return; | |
540 | } | |
466ad822 | 541 | bs->bl.opt_transfer_length = bs->file->bl.opt_transfer_length; |
2647fab5 | 542 | bs->bl.max_transfer_length = bs->file->bl.max_transfer_length; |
339064d5 KW |
543 | bs->bl.opt_mem_alignment = bs->file->bl.opt_mem_alignment; |
544 | } else { | |
545 | bs->bl.opt_mem_alignment = 512; | |
466ad822 KW |
546 | } |
547 | ||
548 | if (bs->backing_hd) { | |
3baca891 KW |
549 | bdrv_refresh_limits(bs->backing_hd, &local_err); |
550 | if (local_err) { | |
551 | error_propagate(errp, local_err); | |
552 | return; | |
553 | } | |
466ad822 KW |
554 | bs->bl.opt_transfer_length = |
555 | MAX(bs->bl.opt_transfer_length, | |
556 | bs->backing_hd->bl.opt_transfer_length); | |
2647fab5 PL |
557 | bs->bl.max_transfer_length = |
558 | MIN_NON_ZERO(bs->bl.max_transfer_length, | |
559 | bs->backing_hd->bl.max_transfer_length); | |
339064d5 KW |
560 | bs->bl.opt_mem_alignment = |
561 | MAX(bs->bl.opt_mem_alignment, | |
562 | bs->backing_hd->bl.opt_mem_alignment); | |
466ad822 KW |
563 | } |
564 | ||
565 | /* Then let the driver override it */ | |
566 | if (drv->bdrv_refresh_limits) { | |
3baca891 | 567 | drv->bdrv_refresh_limits(bs, errp); |
d34682cd | 568 | } |
d34682cd KW |
569 | } |
570 | ||
892b7de8 ET |
571 | /** |
572 | * Try to get @bs's logical and physical block size. | |
573 | * On success, store them in @bsz struct and return 0. | |
574 | * On failure return -errno. | |
575 | * @bs must not be empty. | |
576 | */ | |
577 | int bdrv_probe_blocksizes(BlockDriverState *bs, BlockSizes *bsz) | |
578 | { | |
579 | BlockDriver *drv = bs->drv; | |
580 | ||
581 | if (drv && drv->bdrv_probe_blocksizes) { | |
582 | return drv->bdrv_probe_blocksizes(bs, bsz); | |
583 | } | |
584 | ||
585 | return -ENOTSUP; | |
586 | } | |
587 | ||
588 | /** | |
589 | * Try to get @bs's geometry (cyls, heads, sectors). | |
590 | * On success, store them in @geo struct and return 0. | |
591 | * On failure return -errno. | |
592 | * @bs must not be empty. | |
593 | */ | |
594 | int bdrv_probe_geometry(BlockDriverState *bs, HDGeometry *geo) | |
595 | { | |
596 | BlockDriver *drv = bs->drv; | |
597 | ||
598 | if (drv && drv->bdrv_probe_geometry) { | |
599 | return drv->bdrv_probe_geometry(bs, geo); | |
600 | } | |
601 | ||
602 | return -ENOTSUP; | |
603 | } | |
604 | ||
eba25057 JM |
605 | /* |
606 | * Create a uniquely-named empty temporary file. | |
607 | * Return 0 upon success, otherwise a negative errno value. | |
608 | */ | |
609 | int get_tmp_filename(char *filename, int size) | |
d5249393 | 610 | { |
eba25057 | 611 | #ifdef _WIN32 |
3b9f94e1 | 612 | char temp_dir[MAX_PATH]; |
eba25057 JM |
613 | /* GetTempFileName requires that its output buffer (4th param) |
614 | have length MAX_PATH or greater. */ | |
615 | assert(size >= MAX_PATH); | |
616 | return (GetTempPath(MAX_PATH, temp_dir) | |
617 | && GetTempFileName(temp_dir, "qem", 0, filename) | |
618 | ? 0 : -GetLastError()); | |
d5249393 | 619 | #else |
67b915a5 | 620 | int fd; |
7ccfb2eb | 621 | const char *tmpdir; |
0badc1ee | 622 | tmpdir = getenv("TMPDIR"); |
69bef793 AS |
623 | if (!tmpdir) { |
624 | tmpdir = "/var/tmp"; | |
625 | } | |
eba25057 JM |
626 | if (snprintf(filename, size, "%s/vl.XXXXXX", tmpdir) >= size) { |
627 | return -EOVERFLOW; | |
628 | } | |
ea2384d3 | 629 | fd = mkstemp(filename); |
fe235a06 DH |
630 | if (fd < 0) { |
631 | return -errno; | |
632 | } | |
633 | if (close(fd) != 0) { | |
634 | unlink(filename); | |
eba25057 JM |
635 | return -errno; |
636 | } | |
637 | return 0; | |
d5249393 | 638 | #endif |
eba25057 | 639 | } |
fc01f7e7 | 640 | |
84a12e66 CH |
641 | /* |
642 | * Detect host devices. By convention, /dev/cdrom[N] is always | |
643 | * recognized as a host CDROM. | |
644 | */ | |
645 | static BlockDriver *find_hdev_driver(const char *filename) | |
646 | { | |
647 | int score_max = 0, score; | |
648 | BlockDriver *drv = NULL, *d; | |
649 | ||
650 | QLIST_FOREACH(d, &bdrv_drivers, list) { | |
651 | if (d->bdrv_probe_device) { | |
652 | score = d->bdrv_probe_device(filename); | |
653 | if (score > score_max) { | |
654 | score_max = score; | |
655 | drv = d; | |
656 | } | |
657 | } | |
658 | } | |
659 | ||
660 | return drv; | |
661 | } | |
662 | ||
98289620 | 663 | BlockDriver *bdrv_find_protocol(const char *filename, |
b65a5e12 HR |
664 | bool allow_protocol_prefix, |
665 | Error **errp) | |
83f64091 FB |
666 | { |
667 | BlockDriver *drv1; | |
668 | char protocol[128]; | |
1cec71e3 | 669 | int len; |
83f64091 | 670 | const char *p; |
19cb3738 | 671 | |
66f82cee KW |
672 | /* TODO Drivers without bdrv_file_open must be specified explicitly */ |
673 | ||
39508e7a CH |
674 | /* |
675 | * XXX(hch): we really should not let host device detection | |
676 | * override an explicit protocol specification, but moving this | |
677 | * later breaks access to device names with colons in them. | |
678 | * Thanks to the brain-dead persistent naming schemes on udev- | |
679 | * based Linux systems those actually are quite common. | |
680 | */ | |
681 | drv1 = find_hdev_driver(filename); | |
682 | if (drv1) { | |
683 | return drv1; | |
684 | } | |
685 | ||
98289620 | 686 | if (!path_has_protocol(filename) || !allow_protocol_prefix) { |
ef810437 | 687 | return &bdrv_file; |
84a12e66 | 688 | } |
98289620 | 689 | |
9e0b22f4 SH |
690 | p = strchr(filename, ':'); |
691 | assert(p != NULL); | |
1cec71e3 AL |
692 | len = p - filename; |
693 | if (len > sizeof(protocol) - 1) | |
694 | len = sizeof(protocol) - 1; | |
695 | memcpy(protocol, filename, len); | |
696 | protocol[len] = '\0'; | |
8a22f02a | 697 | QLIST_FOREACH(drv1, &bdrv_drivers, list) { |
5fafdf24 | 698 | if (drv1->protocol_name && |
8a22f02a | 699 | !strcmp(drv1->protocol_name, protocol)) { |
83f64091 | 700 | return drv1; |
8a22f02a | 701 | } |
83f64091 | 702 | } |
b65a5e12 HR |
703 | |
704 | error_setg(errp, "Unknown protocol '%s'", protocol); | |
83f64091 FB |
705 | return NULL; |
706 | } | |
707 | ||
c6684249 MA |
708 | /* |
709 | * Guess image format by probing its contents. | |
710 | * This is not a good idea when your image is raw (CVE-2008-2004), but | |
711 | * we do it anyway for backward compatibility. | |
712 | * | |
713 | * @buf contains the image's first @buf_size bytes. | |
7cddd372 KW |
714 | * @buf_size is the buffer size in bytes (generally BLOCK_PROBE_BUF_SIZE, |
715 | * but can be smaller if the image file is smaller) | |
c6684249 MA |
716 | * @filename is its filename. |
717 | * | |
718 | * For all block drivers, call the bdrv_probe() method to get its | |
719 | * probing score. | |
720 | * Return the first block driver with the highest probing score. | |
721 | */ | |
38f3ef57 KW |
722 | BlockDriver *bdrv_probe_all(const uint8_t *buf, int buf_size, |
723 | const char *filename) | |
c6684249 MA |
724 | { |
725 | int score_max = 0, score; | |
726 | BlockDriver *drv = NULL, *d; | |
727 | ||
728 | QLIST_FOREACH(d, &bdrv_drivers, list) { | |
729 | if (d->bdrv_probe) { | |
730 | score = d->bdrv_probe(buf, buf_size, filename); | |
731 | if (score > score_max) { | |
732 | score_max = score; | |
733 | drv = d; | |
734 | } | |
735 | } | |
736 | } | |
737 | ||
738 | return drv; | |
739 | } | |
740 | ||
f500a6d3 | 741 | static int find_image_format(BlockDriverState *bs, const char *filename, |
34b5d2c6 | 742 | BlockDriver **pdrv, Error **errp) |
f3a5d3f8 | 743 | { |
c6684249 | 744 | BlockDriver *drv; |
7cddd372 | 745 | uint8_t buf[BLOCK_PROBE_BUF_SIZE]; |
f500a6d3 | 746 | int ret = 0; |
f8ea0b00 | 747 | |
08a00559 | 748 | /* Return the raw BlockDriver * to scsi-generic devices or empty drives */ |
8e895599 | 749 | if (bs->sg || !bdrv_is_inserted(bs) || bdrv_getlength(bs) == 0) { |
ef810437 | 750 | *pdrv = &bdrv_raw; |
c98ac35d | 751 | return ret; |
1a396859 | 752 | } |
f8ea0b00 | 753 | |
83f64091 | 754 | ret = bdrv_pread(bs, 0, buf, sizeof(buf)); |
83f64091 | 755 | if (ret < 0) { |
34b5d2c6 HR |
756 | error_setg_errno(errp, -ret, "Could not read image for determining its " |
757 | "format"); | |
c98ac35d SW |
758 | *pdrv = NULL; |
759 | return ret; | |
83f64091 FB |
760 | } |
761 | ||
c6684249 | 762 | drv = bdrv_probe_all(buf, ret, filename); |
c98ac35d | 763 | if (!drv) { |
34b5d2c6 HR |
764 | error_setg(errp, "Could not determine image format: No compatible " |
765 | "driver found"); | |
c98ac35d SW |
766 | ret = -ENOENT; |
767 | } | |
768 | *pdrv = drv; | |
769 | return ret; | |
ea2384d3 FB |
770 | } |
771 | ||
51762288 SH |
772 | /** |
773 | * Set the current 'total_sectors' value | |
65a9bb25 | 774 | * Return 0 on success, -errno on error. |
51762288 SH |
775 | */ |
776 | static int refresh_total_sectors(BlockDriverState *bs, int64_t hint) | |
777 | { | |
778 | BlockDriver *drv = bs->drv; | |
779 | ||
396759ad NB |
780 | /* Do not attempt drv->bdrv_getlength() on scsi-generic devices */ |
781 | if (bs->sg) | |
782 | return 0; | |
783 | ||
51762288 SH |
784 | /* query actual device if possible, otherwise just trust the hint */ |
785 | if (drv->bdrv_getlength) { | |
786 | int64_t length = drv->bdrv_getlength(bs); | |
787 | if (length < 0) { | |
788 | return length; | |
789 | } | |
7e382003 | 790 | hint = DIV_ROUND_UP(length, BDRV_SECTOR_SIZE); |
51762288 SH |
791 | } |
792 | ||
793 | bs->total_sectors = hint; | |
794 | return 0; | |
795 | } | |
796 | ||
9e8f1835 PB |
797 | /** |
798 | * Set open flags for a given discard mode | |
799 | * | |
800 | * Return 0 on success, -1 if the discard mode was invalid. | |
801 | */ | |
802 | int bdrv_parse_discard_flags(const char *mode, int *flags) | |
803 | { | |
804 | *flags &= ~BDRV_O_UNMAP; | |
805 | ||
806 | if (!strcmp(mode, "off") || !strcmp(mode, "ignore")) { | |
807 | /* do nothing */ | |
808 | } else if (!strcmp(mode, "on") || !strcmp(mode, "unmap")) { | |
809 | *flags |= BDRV_O_UNMAP; | |
810 | } else { | |
811 | return -1; | |
812 | } | |
813 | ||
814 | return 0; | |
815 | } | |
816 | ||
c3993cdc SH |
817 | /** |
818 | * Set open flags for a given cache mode | |
819 | * | |
820 | * Return 0 on success, -1 if the cache mode was invalid. | |
821 | */ | |
822 | int bdrv_parse_cache_flags(const char *mode, int *flags) | |
823 | { | |
824 | *flags &= ~BDRV_O_CACHE_MASK; | |
825 | ||
826 | if (!strcmp(mode, "off") || !strcmp(mode, "none")) { | |
827 | *flags |= BDRV_O_NOCACHE | BDRV_O_CACHE_WB; | |
92196b2f SH |
828 | } else if (!strcmp(mode, "directsync")) { |
829 | *flags |= BDRV_O_NOCACHE; | |
c3993cdc SH |
830 | } else if (!strcmp(mode, "writeback")) { |
831 | *flags |= BDRV_O_CACHE_WB; | |
832 | } else if (!strcmp(mode, "unsafe")) { | |
833 | *flags |= BDRV_O_CACHE_WB; | |
834 | *flags |= BDRV_O_NO_FLUSH; | |
835 | } else if (!strcmp(mode, "writethrough")) { | |
836 | /* this is the default */ | |
837 | } else { | |
838 | return -1; | |
839 | } | |
840 | ||
841 | return 0; | |
842 | } | |
843 | ||
53fec9d3 SH |
844 | /** |
845 | * The copy-on-read flag is actually a reference count so multiple users may | |
846 | * use the feature without worrying about clobbering its previous state. | |
847 | * Copy-on-read stays enabled until all users have called to disable it. | |
848 | */ | |
849 | void bdrv_enable_copy_on_read(BlockDriverState *bs) | |
850 | { | |
851 | bs->copy_on_read++; | |
852 | } | |
853 | ||
854 | void bdrv_disable_copy_on_read(BlockDriverState *bs) | |
855 | { | |
856 | assert(bs->copy_on_read > 0); | |
857 | bs->copy_on_read--; | |
858 | } | |
859 | ||
b1e6fc08 KW |
860 | /* |
861 | * Returns the flags that a temporary snapshot should get, based on the | |
862 | * originally requested flags (the originally requested image will have flags | |
863 | * like a backing file) | |
864 | */ | |
865 | static int bdrv_temp_snapshot_flags(int flags) | |
866 | { | |
867 | return (flags & ~BDRV_O_SNAPSHOT) | BDRV_O_TEMPORARY; | |
868 | } | |
869 | ||
0b50cc88 KW |
870 | /* |
871 | * Returns the flags that bs->file should get, based on the given flags for | |
872 | * the parent BDS | |
873 | */ | |
874 | static int bdrv_inherited_flags(int flags) | |
875 | { | |
876 | /* Enable protocol handling, disable format probing for bs->file */ | |
877 | flags |= BDRV_O_PROTOCOL; | |
878 | ||
879 | /* Our block drivers take care to send flushes and respect unmap policy, | |
880 | * so we can enable both unconditionally on lower layers. */ | |
881 | flags |= BDRV_O_CACHE_WB | BDRV_O_UNMAP; | |
882 | ||
0b50cc88 | 883 | /* Clear flags that only apply to the top layer */ |
5669b44d | 884 | flags &= ~(BDRV_O_SNAPSHOT | BDRV_O_NO_BACKING | BDRV_O_COPY_ON_READ); |
0b50cc88 KW |
885 | |
886 | return flags; | |
887 | } | |
888 | ||
317fc44e KW |
889 | /* |
890 | * Returns the flags that bs->backing_hd should get, based on the given flags | |
891 | * for the parent BDS | |
892 | */ | |
893 | static int bdrv_backing_flags(int flags) | |
894 | { | |
895 | /* backing files always opened read-only */ | |
896 | flags &= ~(BDRV_O_RDWR | BDRV_O_COPY_ON_READ); | |
897 | ||
898 | /* snapshot=on is handled on the top layer */ | |
8bfea15d | 899 | flags &= ~(BDRV_O_SNAPSHOT | BDRV_O_TEMPORARY); |
317fc44e KW |
900 | |
901 | return flags; | |
902 | } | |
903 | ||
7b272452 KW |
904 | static int bdrv_open_flags(BlockDriverState *bs, int flags) |
905 | { | |
906 | int open_flags = flags | BDRV_O_CACHE_WB; | |
907 | ||
908 | /* | |
909 | * Clear flags that are internal to the block layer before opening the | |
910 | * image. | |
911 | */ | |
20cca275 | 912 | open_flags &= ~(BDRV_O_SNAPSHOT | BDRV_O_NO_BACKING | BDRV_O_PROTOCOL); |
7b272452 KW |
913 | |
914 | /* | |
915 | * Snapshots should be writable. | |
916 | */ | |
8bfea15d | 917 | if (flags & BDRV_O_TEMPORARY) { |
7b272452 KW |
918 | open_flags |= BDRV_O_RDWR; |
919 | } | |
920 | ||
921 | return open_flags; | |
922 | } | |
923 | ||
636ea370 KW |
924 | static void bdrv_assign_node_name(BlockDriverState *bs, |
925 | const char *node_name, | |
926 | Error **errp) | |
6913c0c2 BC |
927 | { |
928 | if (!node_name) { | |
636ea370 | 929 | return; |
6913c0c2 BC |
930 | } |
931 | ||
9aebf3b8 | 932 | /* Check for empty string or invalid characters */ |
f5bebbbb | 933 | if (!id_wellformed(node_name)) { |
9aebf3b8 | 934 | error_setg(errp, "Invalid node name"); |
636ea370 | 935 | return; |
6913c0c2 BC |
936 | } |
937 | ||
0c5e94ee | 938 | /* takes care of avoiding namespaces collisions */ |
7f06d47e | 939 | if (blk_by_name(node_name)) { |
0c5e94ee BC |
940 | error_setg(errp, "node-name=%s is conflicting with a device id", |
941 | node_name); | |
636ea370 | 942 | return; |
0c5e94ee BC |
943 | } |
944 | ||
6913c0c2 BC |
945 | /* takes care of avoiding duplicates node names */ |
946 | if (bdrv_find_node(node_name)) { | |
947 | error_setg(errp, "Duplicate node name"); | |
636ea370 | 948 | return; |
6913c0c2 BC |
949 | } |
950 | ||
951 | /* copy node name into the bs and insert it into the graph list */ | |
952 | pstrcpy(bs->node_name, sizeof(bs->node_name), node_name); | |
953 | QTAILQ_INSERT_TAIL(&graph_bdrv_states, bs, node_list); | |
6913c0c2 BC |
954 | } |
955 | ||
57915332 KW |
956 | /* |
957 | * Common part for opening disk images and files | |
b6ad491a KW |
958 | * |
959 | * Removes all processed options from *options. | |
57915332 | 960 | */ |
f500a6d3 | 961 | static int bdrv_open_common(BlockDriverState *bs, BlockDriverState *file, |
34b5d2c6 | 962 | QDict *options, int flags, BlockDriver *drv, Error **errp) |
57915332 KW |
963 | { |
964 | int ret, open_flags; | |
035fccdf | 965 | const char *filename; |
6913c0c2 | 966 | const char *node_name = NULL; |
34b5d2c6 | 967 | Error *local_err = NULL; |
57915332 KW |
968 | |
969 | assert(drv != NULL); | |
6405875c | 970 | assert(bs->file == NULL); |
707ff828 | 971 | assert(options != NULL && bs->options != options); |
57915332 | 972 | |
45673671 KW |
973 | if (file != NULL) { |
974 | filename = file->filename; | |
975 | } else { | |
976 | filename = qdict_get_try_str(options, "filename"); | |
977 | } | |
978 | ||
765003db KW |
979 | if (drv->bdrv_needs_filename && !filename) { |
980 | error_setg(errp, "The '%s' block driver requires a file name", | |
981 | drv->format_name); | |
982 | return -EINVAL; | |
983 | } | |
984 | ||
45673671 | 985 | trace_bdrv_open_common(bs, filename ?: "", flags, drv->format_name); |
28dcee10 | 986 | |
6913c0c2 | 987 | node_name = qdict_get_try_str(options, "node-name"); |
636ea370 | 988 | bdrv_assign_node_name(bs, node_name, &local_err); |
0fb6395c | 989 | if (local_err) { |
636ea370 KW |
990 | error_propagate(errp, local_err); |
991 | return -EINVAL; | |
6913c0c2 BC |
992 | } |
993 | qdict_del(options, "node-name"); | |
994 | ||
5d186eb0 KW |
995 | /* bdrv_open() with directly using a protocol as drv. This layer is already |
996 | * opened, so assign it to bs (while file becomes a closed BlockDriverState) | |
997 | * and return immediately. */ | |
998 | if (file != NULL && drv->bdrv_file_open) { | |
999 | bdrv_swap(file, bs); | |
1000 | return 0; | |
1001 | } | |
1002 | ||
57915332 | 1003 | bs->open_flags = flags; |
1b7fd729 | 1004 | bs->guest_block_size = 512; |
c25f53b0 | 1005 | bs->request_alignment = 512; |
0d51b4de | 1006 | bs->zero_beyond_eof = true; |
b64ec4e4 FZ |
1007 | open_flags = bdrv_open_flags(bs, flags); |
1008 | bs->read_only = !(open_flags & BDRV_O_RDWR); | |
1009 | ||
1010 | if (use_bdrv_whitelist && !bdrv_is_whitelisted(drv, bs->read_only)) { | |
8f94a6e4 KW |
1011 | error_setg(errp, |
1012 | !bs->read_only && bdrv_is_whitelisted(drv, true) | |
1013 | ? "Driver '%s' can only be used for read-only devices" | |
1014 | : "Driver '%s' is not whitelisted", | |
1015 | drv->format_name); | |
b64ec4e4 FZ |
1016 | return -ENOTSUP; |
1017 | } | |
57915332 | 1018 | |
53fec9d3 | 1019 | assert(bs->copy_on_read == 0); /* bdrv_new() and bdrv_close() make it so */ |
0ebd24e0 KW |
1020 | if (flags & BDRV_O_COPY_ON_READ) { |
1021 | if (!bs->read_only) { | |
1022 | bdrv_enable_copy_on_read(bs); | |
1023 | } else { | |
1024 | error_setg(errp, "Can't use copy-on-read on read-only device"); | |
1025 | return -EINVAL; | |
1026 | } | |
53fec9d3 SH |
1027 | } |
1028 | ||
c2ad1b0c KW |
1029 | if (filename != NULL) { |
1030 | pstrcpy(bs->filename, sizeof(bs->filename), filename); | |
1031 | } else { | |
1032 | bs->filename[0] = '\0'; | |
1033 | } | |
91af7014 | 1034 | pstrcpy(bs->exact_filename, sizeof(bs->exact_filename), bs->filename); |
57915332 | 1035 | |
57915332 | 1036 | bs->drv = drv; |
7267c094 | 1037 | bs->opaque = g_malloc0(drv->instance_size); |
57915332 | 1038 | |
03f541bd | 1039 | bs->enable_write_cache = !!(flags & BDRV_O_CACHE_WB); |
e7c63796 | 1040 | |
66f82cee KW |
1041 | /* Open the image, either directly or using a protocol */ |
1042 | if (drv->bdrv_file_open) { | |
5d186eb0 | 1043 | assert(file == NULL); |
030be321 | 1044 | assert(!drv->bdrv_needs_filename || filename != NULL); |
34b5d2c6 | 1045 | ret = drv->bdrv_file_open(bs, options, open_flags, &local_err); |
f500a6d3 | 1046 | } else { |
2af5ef70 | 1047 | if (file == NULL) { |
34b5d2c6 HR |
1048 | error_setg(errp, "Can't use '%s' as a block driver for the " |
1049 | "protocol level", drv->format_name); | |
2af5ef70 KW |
1050 | ret = -EINVAL; |
1051 | goto free_and_fail; | |
1052 | } | |
f500a6d3 | 1053 | bs->file = file; |
34b5d2c6 | 1054 | ret = drv->bdrv_open(bs, options, open_flags, &local_err); |
66f82cee KW |
1055 | } |
1056 | ||
57915332 | 1057 | if (ret < 0) { |
84d18f06 | 1058 | if (local_err) { |
34b5d2c6 | 1059 | error_propagate(errp, local_err); |
2fa9aa59 DH |
1060 | } else if (bs->filename[0]) { |
1061 | error_setg_errno(errp, -ret, "Could not open '%s'", bs->filename); | |
34b5d2c6 HR |
1062 | } else { |
1063 | error_setg_errno(errp, -ret, "Could not open image"); | |
1064 | } | |
57915332 KW |
1065 | goto free_and_fail; |
1066 | } | |
1067 | ||
a1f688f4 MA |
1068 | if (bs->encrypted) { |
1069 | error_report("Encrypted images are deprecated"); | |
1070 | error_printf("Support for them will be removed in a future release.\n" | |
1071 | "You can use 'qemu-img convert' to convert your image" | |
1072 | " to an unencrypted one.\n"); | |
1073 | } | |
1074 | ||
51762288 SH |
1075 | ret = refresh_total_sectors(bs, bs->total_sectors); |
1076 | if (ret < 0) { | |
34b5d2c6 | 1077 | error_setg_errno(errp, -ret, "Could not refresh total sector count"); |
51762288 | 1078 | goto free_and_fail; |
57915332 | 1079 | } |
51762288 | 1080 | |
3baca891 KW |
1081 | bdrv_refresh_limits(bs, &local_err); |
1082 | if (local_err) { | |
1083 | error_propagate(errp, local_err); | |
1084 | ret = -EINVAL; | |
1085 | goto free_and_fail; | |
1086 | } | |
1087 | ||
c25f53b0 | 1088 | assert(bdrv_opt_mem_align(bs) != 0); |
47ea2de2 | 1089 | assert((bs->request_alignment != 0) || bs->sg); |
57915332 KW |
1090 | return 0; |
1091 | ||
1092 | free_and_fail: | |
f500a6d3 | 1093 | bs->file = NULL; |
7267c094 | 1094 | g_free(bs->opaque); |
57915332 KW |
1095 | bs->opaque = NULL; |
1096 | bs->drv = NULL; | |
1097 | return ret; | |
1098 | } | |
1099 | ||
5e5c4f63 KW |
1100 | static QDict *parse_json_filename(const char *filename, Error **errp) |
1101 | { | |
1102 | QObject *options_obj; | |
1103 | QDict *options; | |
1104 | int ret; | |
1105 | ||
1106 | ret = strstart(filename, "json:", &filename); | |
1107 | assert(ret); | |
1108 | ||
1109 | options_obj = qobject_from_json(filename); | |
1110 | if (!options_obj) { | |
1111 | error_setg(errp, "Could not parse the JSON options"); | |
1112 | return NULL; | |
1113 | } | |
1114 | ||
1115 | if (qobject_type(options_obj) != QTYPE_QDICT) { | |
1116 | qobject_decref(options_obj); | |
1117 | error_setg(errp, "Invalid JSON object given"); | |
1118 | return NULL; | |
1119 | } | |
1120 | ||
1121 | options = qobject_to_qdict(options_obj); | |
1122 | qdict_flatten(options); | |
1123 | ||
1124 | return options; | |
1125 | } | |
1126 | ||
b6ce07aa | 1127 | /* |
f54120ff KW |
1128 | * Fills in default options for opening images and converts the legacy |
1129 | * filename/flags pair to option QDict entries. | |
b6ce07aa | 1130 | */ |
5e5c4f63 | 1131 | static int bdrv_fill_options(QDict **options, const char **pfilename, int flags, |
17b005f1 | 1132 | BlockDriver *drv, Error **errp) |
ea2384d3 | 1133 | { |
5e5c4f63 | 1134 | const char *filename = *pfilename; |
c2ad1b0c | 1135 | const char *drvname; |
462f5bcf | 1136 | bool protocol = flags & BDRV_O_PROTOCOL; |
e3fa4bfa | 1137 | bool parse_filename = false; |
34b5d2c6 | 1138 | Error *local_err = NULL; |
83f64091 | 1139 | |
5e5c4f63 KW |
1140 | /* Parse json: pseudo-protocol */ |
1141 | if (filename && g_str_has_prefix(filename, "json:")) { | |
1142 | QDict *json_options = parse_json_filename(filename, &local_err); | |
1143 | if (local_err) { | |
1144 | error_propagate(errp, local_err); | |
1145 | return -EINVAL; | |
1146 | } | |
1147 | ||
1148 | /* Options given in the filename have lower priority than options | |
1149 | * specified directly */ | |
1150 | qdict_join(*options, json_options, false); | |
1151 | QDECREF(json_options); | |
1152 | *pfilename = filename = NULL; | |
1153 | } | |
1154 | ||
035fccdf | 1155 | /* Fetch the file name from the options QDict if necessary */ |
17b005f1 | 1156 | if (protocol && filename) { |
f54120ff KW |
1157 | if (!qdict_haskey(*options, "filename")) { |
1158 | qdict_put(*options, "filename", qstring_from_str(filename)); | |
1159 | parse_filename = true; | |
1160 | } else { | |
1161 | error_setg(errp, "Can't specify 'file' and 'filename' options at " | |
1162 | "the same time"); | |
1163 | return -EINVAL; | |
1164 | } | |
035fccdf KW |
1165 | } |
1166 | ||
c2ad1b0c | 1167 | /* Find the right block driver */ |
f54120ff | 1168 | filename = qdict_get_try_str(*options, "filename"); |
5acd9d81 | 1169 | drvname = qdict_get_try_str(*options, "driver"); |
f54120ff | 1170 | |
17b005f1 KW |
1171 | if (drv) { |
1172 | if (drvname) { | |
1173 | error_setg(errp, "Driver specified twice"); | |
1174 | return -EINVAL; | |
1175 | } | |
1176 | drvname = drv->format_name; | |
1177 | qdict_put(*options, "driver", qstring_from_str(drvname)); | |
1178 | } else { | |
1179 | if (!drvname && protocol) { | |
1180 | if (filename) { | |
b65a5e12 | 1181 | drv = bdrv_find_protocol(filename, parse_filename, errp); |
17b005f1 | 1182 | if (!drv) { |
17b005f1 KW |
1183 | return -EINVAL; |
1184 | } | |
1185 | ||
1186 | drvname = drv->format_name; | |
1187 | qdict_put(*options, "driver", qstring_from_str(drvname)); | |
1188 | } else { | |
1189 | error_setg(errp, "Must specify either driver or file"); | |
f54120ff KW |
1190 | return -EINVAL; |
1191 | } | |
17b005f1 KW |
1192 | } else if (drvname) { |
1193 | drv = bdrv_find_format(drvname); | |
1194 | if (!drv) { | |
1195 | error_setg(errp, "Unknown driver '%s'", drvname); | |
1196 | return -ENOENT; | |
1197 | } | |
98289620 | 1198 | } |
c2ad1b0c KW |
1199 | } |
1200 | ||
17b005f1 | 1201 | assert(drv || !protocol); |
c2ad1b0c | 1202 | |
f54120ff | 1203 | /* Driver-specific filename parsing */ |
17b005f1 | 1204 | if (drv && drv->bdrv_parse_filename && parse_filename) { |
5acd9d81 | 1205 | drv->bdrv_parse_filename(filename, *options, &local_err); |
84d18f06 | 1206 | if (local_err) { |
34b5d2c6 | 1207 | error_propagate(errp, local_err); |
f54120ff | 1208 | return -EINVAL; |
6963a30d | 1209 | } |
cd5d031e HR |
1210 | |
1211 | if (!drv->bdrv_needs_filename) { | |
1212 | qdict_del(*options, "filename"); | |
cd5d031e | 1213 | } |
6963a30d KW |
1214 | } |
1215 | ||
f54120ff KW |
1216 | return 0; |
1217 | } | |
1218 | ||
8d24cce1 FZ |
1219 | void bdrv_set_backing_hd(BlockDriverState *bs, BlockDriverState *backing_hd) |
1220 | { | |
1221 | ||
826b6ca0 FZ |
1222 | if (bs->backing_hd) { |
1223 | assert(bs->backing_blocker); | |
1224 | bdrv_op_unblock_all(bs->backing_hd, bs->backing_blocker); | |
1225 | } else if (backing_hd) { | |
1226 | error_setg(&bs->backing_blocker, | |
1227 | "device is used as backing hd of '%s'", | |
bfb197e0 | 1228 | bdrv_get_device_name(bs)); |
826b6ca0 FZ |
1229 | } |
1230 | ||
8d24cce1 FZ |
1231 | bs->backing_hd = backing_hd; |
1232 | if (!backing_hd) { | |
826b6ca0 FZ |
1233 | error_free(bs->backing_blocker); |
1234 | bs->backing_blocker = NULL; | |
8d24cce1 FZ |
1235 | goto out; |
1236 | } | |
1237 | bs->open_flags &= ~BDRV_O_NO_BACKING; | |
1238 | pstrcpy(bs->backing_file, sizeof(bs->backing_file), backing_hd->filename); | |
1239 | pstrcpy(bs->backing_format, sizeof(bs->backing_format), | |
1240 | backing_hd->drv ? backing_hd->drv->format_name : ""); | |
826b6ca0 FZ |
1241 | |
1242 | bdrv_op_block_all(bs->backing_hd, bs->backing_blocker); | |
1243 | /* Otherwise we won't be able to commit due to check in bdrv_commit */ | |
bb00021d | 1244 | bdrv_op_unblock(bs->backing_hd, BLOCK_OP_TYPE_COMMIT_TARGET, |
826b6ca0 | 1245 | bs->backing_blocker); |
8d24cce1 | 1246 | out: |
3baca891 | 1247 | bdrv_refresh_limits(bs, NULL); |
8d24cce1 FZ |
1248 | } |
1249 | ||
31ca6d07 KW |
1250 | /* |
1251 | * Opens the backing file for a BlockDriverState if not yet open | |
1252 | * | |
1253 | * options is a QDict of options to pass to the block drivers, or NULL for an | |
1254 | * empty set of options. The reference to the QDict is transferred to this | |
1255 | * function (even on failure), so if the caller intends to reuse the dictionary, | |
1256 | * it needs to use QINCREF() before calling bdrv_file_open. | |
1257 | */ | |
34b5d2c6 | 1258 | int bdrv_open_backing_file(BlockDriverState *bs, QDict *options, Error **errp) |
9156df12 | 1259 | { |
1ba4b6a5 | 1260 | char *backing_filename = g_malloc0(PATH_MAX); |
317fc44e | 1261 | int ret = 0; |
8d24cce1 | 1262 | BlockDriverState *backing_hd; |
34b5d2c6 | 1263 | Error *local_err = NULL; |
9156df12 PB |
1264 | |
1265 | if (bs->backing_hd != NULL) { | |
31ca6d07 | 1266 | QDECREF(options); |
1ba4b6a5 | 1267 | goto free_exit; |
9156df12 PB |
1268 | } |
1269 | ||
31ca6d07 KW |
1270 | /* NULL means an empty set of options */ |
1271 | if (options == NULL) { | |
1272 | options = qdict_new(); | |
1273 | } | |
1274 | ||
9156df12 | 1275 | bs->open_flags &= ~BDRV_O_NO_BACKING; |
1cb6f506 KW |
1276 | if (qdict_haskey(options, "file.filename")) { |
1277 | backing_filename[0] = '\0'; | |
1278 | } else if (bs->backing_file[0] == '\0' && qdict_size(options) == 0) { | |
31ca6d07 | 1279 | QDECREF(options); |
1ba4b6a5 | 1280 | goto free_exit; |
dbecebdd | 1281 | } else { |
9f07429e HR |
1282 | bdrv_get_full_backing_filename(bs, backing_filename, PATH_MAX, |
1283 | &local_err); | |
1284 | if (local_err) { | |
1285 | ret = -EINVAL; | |
1286 | error_propagate(errp, local_err); | |
1287 | QDECREF(options); | |
1288 | goto free_exit; | |
1289 | } | |
9156df12 PB |
1290 | } |
1291 | ||
8ee79e70 KW |
1292 | if (!bs->drv || !bs->drv->supports_backing) { |
1293 | ret = -EINVAL; | |
1294 | error_setg(errp, "Driver doesn't support backing files"); | |
1295 | QDECREF(options); | |
1296 | goto free_exit; | |
1297 | } | |
1298 | ||
e4e9986b | 1299 | backing_hd = bdrv_new(); |
8d24cce1 | 1300 | |
c5f6e493 KW |
1301 | if (bs->backing_format[0] != '\0' && !qdict_haskey(options, "driver")) { |
1302 | qdict_put(options, "driver", qstring_from_str(bs->backing_format)); | |
9156df12 PB |
1303 | } |
1304 | ||
f67503e5 | 1305 | assert(bs->backing_hd == NULL); |
8d24cce1 | 1306 | ret = bdrv_open(&backing_hd, |
ddf5636d | 1307 | *backing_filename ? backing_filename : NULL, NULL, options, |
c5f6e493 | 1308 | bdrv_backing_flags(bs->open_flags), NULL, &local_err); |
9156df12 | 1309 | if (ret < 0) { |
8d24cce1 FZ |
1310 | bdrv_unref(backing_hd); |
1311 | backing_hd = NULL; | |
9156df12 | 1312 | bs->open_flags |= BDRV_O_NO_BACKING; |
b04b6b6e FZ |
1313 | error_setg(errp, "Could not open backing file: %s", |
1314 | error_get_pretty(local_err)); | |
1315 | error_free(local_err); | |
1ba4b6a5 | 1316 | goto free_exit; |
9156df12 | 1317 | } |
8d24cce1 | 1318 | bdrv_set_backing_hd(bs, backing_hd); |
d80ac658 | 1319 | |
1ba4b6a5 BC |
1320 | free_exit: |
1321 | g_free(backing_filename); | |
1322 | return ret; | |
9156df12 PB |
1323 | } |
1324 | ||
da557aac HR |
1325 | /* |
1326 | * Opens a disk image whose options are given as BlockdevRef in another block | |
1327 | * device's options. | |
1328 | * | |
da557aac HR |
1329 | * If allow_none is true, no image will be opened if filename is false and no |
1330 | * BlockdevRef is given. *pbs will remain unchanged and 0 will be returned. | |
1331 | * | |
1332 | * bdrev_key specifies the key for the image's BlockdevRef in the options QDict. | |
1333 | * That QDict has to be flattened; therefore, if the BlockdevRef is a QDict | |
1334 | * itself, all options starting with "${bdref_key}." are considered part of the | |
1335 | * BlockdevRef. | |
1336 | * | |
1337 | * The BlockdevRef will be removed from the options QDict. | |
f67503e5 HR |
1338 | * |
1339 | * To conform with the behavior of bdrv_open(), *pbs has to be NULL. | |
da557aac HR |
1340 | */ |
1341 | int bdrv_open_image(BlockDriverState **pbs, const char *filename, | |
1342 | QDict *options, const char *bdref_key, int flags, | |
f7d9fd8c | 1343 | bool allow_none, Error **errp) |
da557aac HR |
1344 | { |
1345 | QDict *image_options; | |
1346 | int ret; | |
1347 | char *bdref_key_dot; | |
1348 | const char *reference; | |
1349 | ||
f67503e5 HR |
1350 | assert(pbs); |
1351 | assert(*pbs == NULL); | |
1352 | ||
da557aac HR |
1353 | bdref_key_dot = g_strdup_printf("%s.", bdref_key); |
1354 | qdict_extract_subqdict(options, &image_options, bdref_key_dot); | |
1355 | g_free(bdref_key_dot); | |
1356 | ||
1357 | reference = qdict_get_try_str(options, bdref_key); | |
1358 | if (!filename && !reference && !qdict_size(image_options)) { | |
1359 | if (allow_none) { | |
1360 | ret = 0; | |
1361 | } else { | |
1362 | error_setg(errp, "A block device must be specified for \"%s\"", | |
1363 | bdref_key); | |
1364 | ret = -EINVAL; | |
1365 | } | |
b20e61e0 | 1366 | QDECREF(image_options); |
da557aac HR |
1367 | goto done; |
1368 | } | |
1369 | ||
f7d9fd8c | 1370 | ret = bdrv_open(pbs, filename, reference, image_options, flags, NULL, errp); |
da557aac HR |
1371 | |
1372 | done: | |
1373 | qdict_del(options, bdref_key); | |
1374 | return ret; | |
1375 | } | |
1376 | ||
6b8aeca5 | 1377 | int bdrv_append_temp_snapshot(BlockDriverState *bs, int flags, Error **errp) |
b998875d KW |
1378 | { |
1379 | /* TODO: extra byte is a hack to ensure MAX_PATH space on Windows. */ | |
1ba4b6a5 | 1380 | char *tmp_filename = g_malloc0(PATH_MAX + 1); |
b998875d | 1381 | int64_t total_size; |
83d0521a | 1382 | QemuOpts *opts = NULL; |
b998875d KW |
1383 | QDict *snapshot_options; |
1384 | BlockDriverState *bs_snapshot; | |
1385 | Error *local_err; | |
1386 | int ret; | |
1387 | ||
1388 | /* if snapshot, we create a temporary backing file and open it | |
1389 | instead of opening 'filename' directly */ | |
1390 | ||
1391 | /* Get the required size from the image */ | |
f187743a KW |
1392 | total_size = bdrv_getlength(bs); |
1393 | if (total_size < 0) { | |
6b8aeca5 | 1394 | ret = total_size; |
f187743a | 1395 | error_setg_errno(errp, -total_size, "Could not get image size"); |
1ba4b6a5 | 1396 | goto out; |
f187743a | 1397 | } |
b998875d KW |
1398 | |
1399 | /* Create the temporary image */ | |
1ba4b6a5 | 1400 | ret = get_tmp_filename(tmp_filename, PATH_MAX + 1); |
b998875d KW |
1401 | if (ret < 0) { |
1402 | error_setg_errno(errp, -ret, "Could not get temporary filename"); | |
1ba4b6a5 | 1403 | goto out; |
b998875d KW |
1404 | } |
1405 | ||
ef810437 | 1406 | opts = qemu_opts_create(bdrv_qcow2.create_opts, NULL, 0, |
c282e1fd | 1407 | &error_abort); |
39101f25 | 1408 | qemu_opt_set_number(opts, BLOCK_OPT_SIZE, total_size, &error_abort); |
ef810437 | 1409 | ret = bdrv_create(&bdrv_qcow2, tmp_filename, opts, &local_err); |
83d0521a | 1410 | qemu_opts_del(opts); |
b998875d KW |
1411 | if (ret < 0) { |
1412 | error_setg_errno(errp, -ret, "Could not create temporary overlay " | |
1413 | "'%s': %s", tmp_filename, | |
1414 | error_get_pretty(local_err)); | |
1415 | error_free(local_err); | |
1ba4b6a5 | 1416 | goto out; |
b998875d KW |
1417 | } |
1418 | ||
1419 | /* Prepare a new options QDict for the temporary file */ | |
1420 | snapshot_options = qdict_new(); | |
1421 | qdict_put(snapshot_options, "file.driver", | |
1422 | qstring_from_str("file")); | |
1423 | qdict_put(snapshot_options, "file.filename", | |
1424 | qstring_from_str(tmp_filename)); | |
1425 | ||
e4e9986b | 1426 | bs_snapshot = bdrv_new(); |
b998875d KW |
1427 | |
1428 | ret = bdrv_open(&bs_snapshot, NULL, NULL, snapshot_options, | |
ef810437 | 1429 | flags, &bdrv_qcow2, &local_err); |
b998875d KW |
1430 | if (ret < 0) { |
1431 | error_propagate(errp, local_err); | |
1ba4b6a5 | 1432 | goto out; |
b998875d KW |
1433 | } |
1434 | ||
1435 | bdrv_append(bs_snapshot, bs); | |
1ba4b6a5 BC |
1436 | |
1437 | out: | |
1438 | g_free(tmp_filename); | |
6b8aeca5 | 1439 | return ret; |
b998875d KW |
1440 | } |
1441 | ||
b6ce07aa KW |
1442 | /* |
1443 | * Opens a disk image (raw, qcow2, vmdk, ...) | |
de9c0cec KW |
1444 | * |
1445 | * options is a QDict of options to pass to the block drivers, or NULL for an | |
1446 | * empty set of options. The reference to the QDict belongs to the block layer | |
1447 | * after the call (even on failure), so if the caller intends to reuse the | |
1448 | * dictionary, it needs to use QINCREF() before calling bdrv_open. | |
f67503e5 HR |
1449 | * |
1450 | * If *pbs is NULL, a new BDS will be created with a pointer to it stored there. | |
1451 | * If it is not NULL, the referenced BDS will be reused. | |
ddf5636d HR |
1452 | * |
1453 | * The reference parameter may be used to specify an existing block device which | |
1454 | * should be opened. If specified, neither options nor a filename may be given, | |
1455 | * nor can an existing BDS be reused (that is, *pbs has to be NULL). | |
b6ce07aa | 1456 | */ |
ddf5636d HR |
1457 | int bdrv_open(BlockDriverState **pbs, const char *filename, |
1458 | const char *reference, QDict *options, int flags, | |
1459 | BlockDriver *drv, Error **errp) | |
ea2384d3 | 1460 | { |
b6ce07aa | 1461 | int ret; |
f67503e5 | 1462 | BlockDriverState *file = NULL, *bs; |
74fe54f2 | 1463 | const char *drvname; |
34b5d2c6 | 1464 | Error *local_err = NULL; |
b1e6fc08 | 1465 | int snapshot_flags = 0; |
712e7874 | 1466 | |
f67503e5 HR |
1467 | assert(pbs); |
1468 | ||
ddf5636d HR |
1469 | if (reference) { |
1470 | bool options_non_empty = options ? qdict_size(options) : false; | |
1471 | QDECREF(options); | |
1472 | ||
1473 | if (*pbs) { | |
1474 | error_setg(errp, "Cannot reuse an existing BDS when referencing " | |
1475 | "another block device"); | |
1476 | return -EINVAL; | |
1477 | } | |
1478 | ||
1479 | if (filename || options_non_empty) { | |
1480 | error_setg(errp, "Cannot reference an existing block device with " | |
1481 | "additional options or a new filename"); | |
1482 | return -EINVAL; | |
1483 | } | |
1484 | ||
1485 | bs = bdrv_lookup_bs(reference, reference, errp); | |
1486 | if (!bs) { | |
1487 | return -ENODEV; | |
1488 | } | |
1489 | bdrv_ref(bs); | |
1490 | *pbs = bs; | |
1491 | return 0; | |
1492 | } | |
1493 | ||
f67503e5 HR |
1494 | if (*pbs) { |
1495 | bs = *pbs; | |
1496 | } else { | |
e4e9986b | 1497 | bs = bdrv_new(); |
f67503e5 HR |
1498 | } |
1499 | ||
de9c0cec KW |
1500 | /* NULL means an empty set of options */ |
1501 | if (options == NULL) { | |
1502 | options = qdict_new(); | |
1503 | } | |
1504 | ||
17b005f1 | 1505 | ret = bdrv_fill_options(&options, &filename, flags, drv, &local_err); |
462f5bcf KW |
1506 | if (local_err) { |
1507 | goto fail; | |
1508 | } | |
1509 | ||
76c591b0 KW |
1510 | /* Find the right image format driver */ |
1511 | drv = NULL; | |
1512 | drvname = qdict_get_try_str(options, "driver"); | |
1513 | if (drvname) { | |
1514 | drv = bdrv_find_format(drvname); | |
1515 | qdict_del(options, "driver"); | |
1516 | if (!drv) { | |
1517 | error_setg(errp, "Unknown driver: '%s'", drvname); | |
1518 | ret = -EINVAL; | |
1519 | goto fail; | |
1520 | } | |
1521 | } | |
1522 | ||
1523 | assert(drvname || !(flags & BDRV_O_PROTOCOL)); | |
1524 | if (drv && !drv->bdrv_file_open) { | |
1525 | /* If the user explicitly wants a format driver here, we'll need to add | |
1526 | * another layer for the protocol in bs->file */ | |
1527 | flags &= ~BDRV_O_PROTOCOL; | |
1528 | } | |
1529 | ||
de9c0cec | 1530 | bs->options = options; |
b6ad491a | 1531 | options = qdict_clone_shallow(options); |
de9c0cec | 1532 | |
f500a6d3 | 1533 | /* Open image file without format layer */ |
f4788adc KW |
1534 | if ((flags & BDRV_O_PROTOCOL) == 0) { |
1535 | if (flags & BDRV_O_RDWR) { | |
1536 | flags |= BDRV_O_ALLOW_RDWR; | |
1537 | } | |
1538 | if (flags & BDRV_O_SNAPSHOT) { | |
1539 | snapshot_flags = bdrv_temp_snapshot_flags(flags); | |
1540 | flags = bdrv_backing_flags(flags); | |
1541 | } | |
f500a6d3 | 1542 | |
f4788adc KW |
1543 | assert(file == NULL); |
1544 | ret = bdrv_open_image(&file, filename, options, "file", | |
1545 | bdrv_inherited_flags(flags), | |
1546 | true, &local_err); | |
1547 | if (ret < 0) { | |
1548 | goto fail; | |
1549 | } | |
f500a6d3 KW |
1550 | } |
1551 | ||
76c591b0 | 1552 | /* Image format probing */ |
38f3ef57 | 1553 | bs->probed = !drv; |
76c591b0 | 1554 | if (!drv && file) { |
17b005f1 KW |
1555 | ret = find_image_format(file, filename, &drv, &local_err); |
1556 | if (ret < 0) { | |
8bfea15d | 1557 | goto fail; |
2a05cbe4 | 1558 | } |
76c591b0 | 1559 | } else if (!drv) { |
17b005f1 KW |
1560 | error_setg(errp, "Must specify either driver or file"); |
1561 | ret = -EINVAL; | |
8bfea15d | 1562 | goto fail; |
ea2384d3 | 1563 | } |
b6ce07aa KW |
1564 | |
1565 | /* Open the image */ | |
34b5d2c6 | 1566 | ret = bdrv_open_common(bs, file, options, flags, drv, &local_err); |
b6ce07aa | 1567 | if (ret < 0) { |
8bfea15d | 1568 | goto fail; |
6987307c CH |
1569 | } |
1570 | ||
2a05cbe4 | 1571 | if (file && (bs->file != file)) { |
4f6fd349 | 1572 | bdrv_unref(file); |
f500a6d3 KW |
1573 | file = NULL; |
1574 | } | |
1575 | ||
b6ce07aa | 1576 | /* If there is a backing file, use it */ |
9156df12 | 1577 | if ((flags & BDRV_O_NO_BACKING) == 0) { |
31ca6d07 KW |
1578 | QDict *backing_options; |
1579 | ||
5726d872 | 1580 | qdict_extract_subqdict(options, &backing_options, "backing."); |
34b5d2c6 | 1581 | ret = bdrv_open_backing_file(bs, backing_options, &local_err); |
b6ce07aa | 1582 | if (ret < 0) { |
b6ad491a | 1583 | goto close_and_fail; |
b6ce07aa | 1584 | } |
b6ce07aa KW |
1585 | } |
1586 | ||
91af7014 HR |
1587 | bdrv_refresh_filename(bs); |
1588 | ||
b998875d KW |
1589 | /* For snapshot=on, create a temporary qcow2 overlay. bs points to the |
1590 | * temporary snapshot afterwards. */ | |
b1e6fc08 | 1591 | if (snapshot_flags) { |
6b8aeca5 | 1592 | ret = bdrv_append_temp_snapshot(bs, snapshot_flags, &local_err); |
b998875d | 1593 | if (local_err) { |
b998875d KW |
1594 | goto close_and_fail; |
1595 | } | |
1596 | } | |
1597 | ||
b6ad491a | 1598 | /* Check if any unknown options were used */ |
5acd9d81 | 1599 | if (options && (qdict_size(options) != 0)) { |
b6ad491a | 1600 | const QDictEntry *entry = qdict_first(options); |
5acd9d81 HR |
1601 | if (flags & BDRV_O_PROTOCOL) { |
1602 | error_setg(errp, "Block protocol '%s' doesn't support the option " | |
1603 | "'%s'", drv->format_name, entry->key); | |
1604 | } else { | |
1605 | error_setg(errp, "Block format '%s' used by device '%s' doesn't " | |
1606 | "support the option '%s'", drv->format_name, | |
bfb197e0 | 1607 | bdrv_get_device_name(bs), entry->key); |
5acd9d81 | 1608 | } |
b6ad491a KW |
1609 | |
1610 | ret = -EINVAL; | |
1611 | goto close_and_fail; | |
1612 | } | |
b6ad491a | 1613 | |
b6ce07aa | 1614 | if (!bdrv_key_required(bs)) { |
a7f53e26 MA |
1615 | if (bs->blk) { |
1616 | blk_dev_change_media_cb(bs->blk, true); | |
1617 | } | |
c3adb58f MA |
1618 | } else if (!runstate_check(RUN_STATE_PRELAUNCH) |
1619 | && !runstate_check(RUN_STATE_INMIGRATE) | |
1620 | && !runstate_check(RUN_STATE_PAUSED)) { /* HACK */ | |
1621 | error_setg(errp, | |
1622 | "Guest must be stopped for opening of encrypted image"); | |
1623 | ret = -EBUSY; | |
1624 | goto close_and_fail; | |
b6ce07aa KW |
1625 | } |
1626 | ||
c3adb58f | 1627 | QDECREF(options); |
f67503e5 | 1628 | *pbs = bs; |
b6ce07aa KW |
1629 | return 0; |
1630 | ||
8bfea15d | 1631 | fail: |
f500a6d3 | 1632 | if (file != NULL) { |
4f6fd349 | 1633 | bdrv_unref(file); |
f500a6d3 | 1634 | } |
de9c0cec | 1635 | QDECREF(bs->options); |
b6ad491a | 1636 | QDECREF(options); |
de9c0cec | 1637 | bs->options = NULL; |
f67503e5 HR |
1638 | if (!*pbs) { |
1639 | /* If *pbs is NULL, a new BDS has been created in this function and | |
1640 | needs to be freed now. Otherwise, it does not need to be closed, | |
1641 | since it has not really been opened yet. */ | |
1642 | bdrv_unref(bs); | |
1643 | } | |
84d18f06 | 1644 | if (local_err) { |
34b5d2c6 HR |
1645 | error_propagate(errp, local_err); |
1646 | } | |
b6ad491a | 1647 | return ret; |
de9c0cec | 1648 | |
b6ad491a | 1649 | close_and_fail: |
f67503e5 HR |
1650 | /* See fail path, but now the BDS has to be always closed */ |
1651 | if (*pbs) { | |
1652 | bdrv_close(bs); | |
1653 | } else { | |
1654 | bdrv_unref(bs); | |
1655 | } | |
b6ad491a | 1656 | QDECREF(options); |
84d18f06 | 1657 | if (local_err) { |
34b5d2c6 HR |
1658 | error_propagate(errp, local_err); |
1659 | } | |
b6ce07aa KW |
1660 | return ret; |
1661 | } | |
1662 | ||
e971aa12 JC |
1663 | typedef struct BlockReopenQueueEntry { |
1664 | bool prepared; | |
1665 | BDRVReopenState state; | |
1666 | QSIMPLEQ_ENTRY(BlockReopenQueueEntry) entry; | |
1667 | } BlockReopenQueueEntry; | |
1668 | ||
1669 | /* | |
1670 | * Adds a BlockDriverState to a simple queue for an atomic, transactional | |
1671 | * reopen of multiple devices. | |
1672 | * | |
1673 | * bs_queue can either be an existing BlockReopenQueue that has had QSIMPLE_INIT | |
1674 | * already performed, or alternatively may be NULL a new BlockReopenQueue will | |
1675 | * be created and initialized. This newly created BlockReopenQueue should be | |
1676 | * passed back in for subsequent calls that are intended to be of the same | |
1677 | * atomic 'set'. | |
1678 | * | |
1679 | * bs is the BlockDriverState to add to the reopen queue. | |
1680 | * | |
1681 | * flags contains the open flags for the associated bs | |
1682 | * | |
1683 | * returns a pointer to bs_queue, which is either the newly allocated | |
1684 | * bs_queue, or the existing bs_queue being used. | |
1685 | * | |
1686 | */ | |
1687 | BlockReopenQueue *bdrv_reopen_queue(BlockReopenQueue *bs_queue, | |
1688 | BlockDriverState *bs, int flags) | |
1689 | { | |
1690 | assert(bs != NULL); | |
1691 | ||
1692 | BlockReopenQueueEntry *bs_entry; | |
1693 | if (bs_queue == NULL) { | |
1694 | bs_queue = g_new0(BlockReopenQueue, 1); | |
1695 | QSIMPLEQ_INIT(bs_queue); | |
1696 | } | |
1697 | ||
f1f25a2e KW |
1698 | /* bdrv_open() masks this flag out */ |
1699 | flags &= ~BDRV_O_PROTOCOL; | |
1700 | ||
e971aa12 | 1701 | if (bs->file) { |
f1f25a2e | 1702 | bdrv_reopen_queue(bs_queue, bs->file, bdrv_inherited_flags(flags)); |
e971aa12 JC |
1703 | } |
1704 | ||
1705 | bs_entry = g_new0(BlockReopenQueueEntry, 1); | |
1706 | QSIMPLEQ_INSERT_TAIL(bs_queue, bs_entry, entry); | |
1707 | ||
1708 | bs_entry->state.bs = bs; | |
1709 | bs_entry->state.flags = flags; | |
1710 | ||
1711 | return bs_queue; | |
1712 | } | |
1713 | ||
1714 | /* | |
1715 | * Reopen multiple BlockDriverStates atomically & transactionally. | |
1716 | * | |
1717 | * The queue passed in (bs_queue) must have been built up previous | |
1718 | * via bdrv_reopen_queue(). | |
1719 | * | |
1720 | * Reopens all BDS specified in the queue, with the appropriate | |
1721 | * flags. All devices are prepared for reopen, and failure of any | |
1722 | * device will cause all device changes to be abandonded, and intermediate | |
1723 | * data cleaned up. | |
1724 | * | |
1725 | * If all devices prepare successfully, then the changes are committed | |
1726 | * to all devices. | |
1727 | * | |
1728 | */ | |
1729 | int bdrv_reopen_multiple(BlockReopenQueue *bs_queue, Error **errp) | |
1730 | { | |
1731 | int ret = -1; | |
1732 | BlockReopenQueueEntry *bs_entry, *next; | |
1733 | Error *local_err = NULL; | |
1734 | ||
1735 | assert(bs_queue != NULL); | |
1736 | ||
1737 | bdrv_drain_all(); | |
1738 | ||
1739 | QSIMPLEQ_FOREACH(bs_entry, bs_queue, entry) { | |
1740 | if (bdrv_reopen_prepare(&bs_entry->state, bs_queue, &local_err)) { | |
1741 | error_propagate(errp, local_err); | |
1742 | goto cleanup; | |
1743 | } | |
1744 | bs_entry->prepared = true; | |
1745 | } | |
1746 | ||
1747 | /* If we reach this point, we have success and just need to apply the | |
1748 | * changes | |
1749 | */ | |
1750 | QSIMPLEQ_FOREACH(bs_entry, bs_queue, entry) { | |
1751 | bdrv_reopen_commit(&bs_entry->state); | |
1752 | } | |
1753 | ||
1754 | ret = 0; | |
1755 | ||
1756 | cleanup: | |
1757 | QSIMPLEQ_FOREACH_SAFE(bs_entry, bs_queue, entry, next) { | |
1758 | if (ret && bs_entry->prepared) { | |
1759 | bdrv_reopen_abort(&bs_entry->state); | |
1760 | } | |
1761 | g_free(bs_entry); | |
1762 | } | |
1763 | g_free(bs_queue); | |
1764 | return ret; | |
1765 | } | |
1766 | ||
1767 | ||
1768 | /* Reopen a single BlockDriverState with the specified flags. */ | |
1769 | int bdrv_reopen(BlockDriverState *bs, int bdrv_flags, Error **errp) | |
1770 | { | |
1771 | int ret = -1; | |
1772 | Error *local_err = NULL; | |
1773 | BlockReopenQueue *queue = bdrv_reopen_queue(NULL, bs, bdrv_flags); | |
1774 | ||
1775 | ret = bdrv_reopen_multiple(queue, &local_err); | |
1776 | if (local_err != NULL) { | |
1777 | error_propagate(errp, local_err); | |
1778 | } | |
1779 | return ret; | |
1780 | } | |
1781 | ||
1782 | ||
1783 | /* | |
1784 | * Prepares a BlockDriverState for reopen. All changes are staged in the | |
1785 | * 'opaque' field of the BDRVReopenState, which is used and allocated by | |
1786 | * the block driver layer .bdrv_reopen_prepare() | |
1787 | * | |
1788 | * bs is the BlockDriverState to reopen | |
1789 | * flags are the new open flags | |
1790 | * queue is the reopen queue | |
1791 | * | |
1792 | * Returns 0 on success, non-zero on error. On error errp will be set | |
1793 | * as well. | |
1794 | * | |
1795 | * On failure, bdrv_reopen_abort() will be called to clean up any data. | |
1796 | * It is the responsibility of the caller to then call the abort() or | |
1797 | * commit() for any other BDS that have been left in a prepare() state | |
1798 | * | |
1799 | */ | |
1800 | int bdrv_reopen_prepare(BDRVReopenState *reopen_state, BlockReopenQueue *queue, | |
1801 | Error **errp) | |
1802 | { | |
1803 | int ret = -1; | |
1804 | Error *local_err = NULL; | |
1805 | BlockDriver *drv; | |
1806 | ||
1807 | assert(reopen_state != NULL); | |
1808 | assert(reopen_state->bs->drv != NULL); | |
1809 | drv = reopen_state->bs->drv; | |
1810 | ||
1811 | /* if we are to stay read-only, do not allow permission change | |
1812 | * to r/w */ | |
1813 | if (!(reopen_state->bs->open_flags & BDRV_O_ALLOW_RDWR) && | |
1814 | reopen_state->flags & BDRV_O_RDWR) { | |
1815 | error_set(errp, QERR_DEVICE_IS_READ_ONLY, | |
bfb197e0 | 1816 | bdrv_get_device_name(reopen_state->bs)); |
e971aa12 JC |
1817 | goto error; |
1818 | } | |
1819 | ||
1820 | ||
1821 | ret = bdrv_flush(reopen_state->bs); | |
1822 | if (ret) { | |
1823 | error_set(errp, ERROR_CLASS_GENERIC_ERROR, "Error (%s) flushing drive", | |
1824 | strerror(-ret)); | |
1825 | goto error; | |
1826 | } | |
1827 | ||
1828 | if (drv->bdrv_reopen_prepare) { | |
1829 | ret = drv->bdrv_reopen_prepare(reopen_state, queue, &local_err); | |
1830 | if (ret) { | |
1831 | if (local_err != NULL) { | |
1832 | error_propagate(errp, local_err); | |
1833 | } else { | |
d8b6895f LC |
1834 | error_setg(errp, "failed while preparing to reopen image '%s'", |
1835 | reopen_state->bs->filename); | |
e971aa12 JC |
1836 | } |
1837 | goto error; | |
1838 | } | |
1839 | } else { | |
1840 | /* It is currently mandatory to have a bdrv_reopen_prepare() | |
1841 | * handler for each supported drv. */ | |
1842 | error_set(errp, QERR_BLOCK_FORMAT_FEATURE_NOT_SUPPORTED, | |
bfb197e0 | 1843 | drv->format_name, bdrv_get_device_name(reopen_state->bs), |
e971aa12 JC |
1844 | "reopening of file"); |
1845 | ret = -1; | |
1846 | goto error; | |
1847 | } | |
1848 | ||
1849 | ret = 0; | |
1850 | ||
1851 | error: | |
1852 | return ret; | |
1853 | } | |
1854 | ||
1855 | /* | |
1856 | * Takes the staged changes for the reopen from bdrv_reopen_prepare(), and | |
1857 | * makes them final by swapping the staging BlockDriverState contents into | |
1858 | * the active BlockDriverState contents. | |
1859 | */ | |
1860 | void bdrv_reopen_commit(BDRVReopenState *reopen_state) | |
1861 | { | |
1862 | BlockDriver *drv; | |
1863 | ||
1864 | assert(reopen_state != NULL); | |
1865 | drv = reopen_state->bs->drv; | |
1866 | assert(drv != NULL); | |
1867 | ||
1868 | /* If there are any driver level actions to take */ | |
1869 | if (drv->bdrv_reopen_commit) { | |
1870 | drv->bdrv_reopen_commit(reopen_state); | |
1871 | } | |
1872 | ||
1873 | /* set BDS specific flags now */ | |
1874 | reopen_state->bs->open_flags = reopen_state->flags; | |
1875 | reopen_state->bs->enable_write_cache = !!(reopen_state->flags & | |
1876 | BDRV_O_CACHE_WB); | |
1877 | reopen_state->bs->read_only = !(reopen_state->flags & BDRV_O_RDWR); | |
355ef4ac | 1878 | |
3baca891 | 1879 | bdrv_refresh_limits(reopen_state->bs, NULL); |
e971aa12 JC |
1880 | } |
1881 | ||
1882 | /* | |
1883 | * Abort the reopen, and delete and free the staged changes in | |
1884 | * reopen_state | |
1885 | */ | |
1886 | void bdrv_reopen_abort(BDRVReopenState *reopen_state) | |
1887 | { | |
1888 | BlockDriver *drv; | |
1889 | ||
1890 | assert(reopen_state != NULL); | |
1891 | drv = reopen_state->bs->drv; | |
1892 | assert(drv != NULL); | |
1893 | ||
1894 | if (drv->bdrv_reopen_abort) { | |
1895 | drv->bdrv_reopen_abort(reopen_state); | |
1896 | } | |
1897 | } | |
1898 | ||
1899 | ||
fc01f7e7 FB |
1900 | void bdrv_close(BlockDriverState *bs) |
1901 | { | |
33384421 HR |
1902 | BdrvAioNotifier *ban, *ban_next; |
1903 | ||
3cbc002c PB |
1904 | if (bs->job) { |
1905 | block_job_cancel_sync(bs->job); | |
1906 | } | |
58fda173 SH |
1907 | bdrv_drain_all(); /* complete I/O */ |
1908 | bdrv_flush(bs); | |
1909 | bdrv_drain_all(); /* in case flush left pending I/O */ | |
d7d512f6 | 1910 | notifier_list_notify(&bs->close_notifiers, bs); |
7094f12f | 1911 | |
3cbc002c | 1912 | if (bs->drv) { |
557df6ac | 1913 | if (bs->backing_hd) { |
826b6ca0 FZ |
1914 | BlockDriverState *backing_hd = bs->backing_hd; |
1915 | bdrv_set_backing_hd(bs, NULL); | |
1916 | bdrv_unref(backing_hd); | |
557df6ac | 1917 | } |
ea2384d3 | 1918 | bs->drv->bdrv_close(bs); |
7267c094 | 1919 | g_free(bs->opaque); |
ea2384d3 FB |
1920 | bs->opaque = NULL; |
1921 | bs->drv = NULL; | |
53fec9d3 | 1922 | bs->copy_on_read = 0; |
a275fa42 PB |
1923 | bs->backing_file[0] = '\0'; |
1924 | bs->backing_format[0] = '\0'; | |
6405875c PB |
1925 | bs->total_sectors = 0; |
1926 | bs->encrypted = 0; | |
1927 | bs->valid_key = 0; | |
1928 | bs->sg = 0; | |
0d51b4de | 1929 | bs->zero_beyond_eof = false; |
de9c0cec KW |
1930 | QDECREF(bs->options); |
1931 | bs->options = NULL; | |
91af7014 HR |
1932 | QDECREF(bs->full_open_options); |
1933 | bs->full_open_options = NULL; | |
b338082b | 1934 | |
66f82cee | 1935 | if (bs->file != NULL) { |
4f6fd349 | 1936 | bdrv_unref(bs->file); |
0ac9377d | 1937 | bs->file = NULL; |
66f82cee | 1938 | } |
b338082b | 1939 | } |
98f90dba | 1940 | |
a7f53e26 MA |
1941 | if (bs->blk) { |
1942 | blk_dev_change_media_cb(bs->blk, false); | |
1943 | } | |
9ca11154 | 1944 | |
98f90dba ZYW |
1945 | /*throttling disk I/O limits*/ |
1946 | if (bs->io_limits_enabled) { | |
1947 | bdrv_io_limits_disable(bs); | |
1948 | } | |
33384421 HR |
1949 | |
1950 | QLIST_FOREACH_SAFE(ban, &bs->aio_notifiers, list, ban_next) { | |
1951 | g_free(ban); | |
1952 | } | |
1953 | QLIST_INIT(&bs->aio_notifiers); | |
b338082b FB |
1954 | } |
1955 | ||
2bc93fed MK |
1956 | void bdrv_close_all(void) |
1957 | { | |
1958 | BlockDriverState *bs; | |
1959 | ||
dc364f4c | 1960 | QTAILQ_FOREACH(bs, &bdrv_states, device_list) { |
ed78cda3 SH |
1961 | AioContext *aio_context = bdrv_get_aio_context(bs); |
1962 | ||
1963 | aio_context_acquire(aio_context); | |
2bc93fed | 1964 | bdrv_close(bs); |
ed78cda3 | 1965 | aio_context_release(aio_context); |
2bc93fed MK |
1966 | } |
1967 | } | |
1968 | ||
88266f5a SH |
1969 | /* Check if any requests are in-flight (including throttled requests) */ |
1970 | static bool bdrv_requests_pending(BlockDriverState *bs) | |
1971 | { | |
1972 | if (!QLIST_EMPTY(&bs->tracked_requests)) { | |
1973 | return true; | |
1974 | } | |
cc0681c4 BC |
1975 | if (!qemu_co_queue_empty(&bs->throttled_reqs[0])) { |
1976 | return true; | |
1977 | } | |
1978 | if (!qemu_co_queue_empty(&bs->throttled_reqs[1])) { | |
88266f5a SH |
1979 | return true; |
1980 | } | |
1981 | if (bs->file && bdrv_requests_pending(bs->file)) { | |
1982 | return true; | |
1983 | } | |
1984 | if (bs->backing_hd && bdrv_requests_pending(bs->backing_hd)) { | |
1985 | return true; | |
1986 | } | |
1987 | return false; | |
1988 | } | |
1989 | ||
5b98db0a SH |
1990 | static bool bdrv_drain_one(BlockDriverState *bs) |
1991 | { | |
1992 | bool bs_busy; | |
1993 | ||
1994 | bdrv_flush_io_queue(bs); | |
1995 | bdrv_start_throttled_reqs(bs); | |
1996 | bs_busy = bdrv_requests_pending(bs); | |
1997 | bs_busy |= aio_poll(bdrv_get_aio_context(bs), bs_busy); | |
1998 | return bs_busy; | |
1999 | } | |
2000 | ||
2001 | /* | |
2002 | * Wait for pending requests to complete on a single BlockDriverState subtree | |
2003 | * | |
2004 | * See the warning in bdrv_drain_all(). This function can only be called if | |
2005 | * you are sure nothing can generate I/O because you have op blockers | |
2006 | * installed. | |
2007 | * | |
2008 | * Note that unlike bdrv_drain_all(), the caller must hold the BlockDriverState | |
2009 | * AioContext. | |
2010 | */ | |
2011 | void bdrv_drain(BlockDriverState *bs) | |
2012 | { | |
2013 | while (bdrv_drain_one(bs)) { | |
2014 | /* Keep iterating */ | |
2015 | } | |
2016 | } | |
2017 | ||
922453bc SH |
2018 | /* |
2019 | * Wait for pending requests to complete across all BlockDriverStates | |
2020 | * | |
2021 | * This function does not flush data to disk, use bdrv_flush_all() for that | |
2022 | * after calling this function. | |
4c355d53 ZYW |
2023 | * |
2024 | * Note that completion of an asynchronous I/O operation can trigger any | |
2025 | * number of other I/O operations on other devices---for example a coroutine | |
2026 | * can be arbitrarily complex and a constant flow of I/O can come until the | |
2027 | * coroutine is complete. Because of this, it is not possible to have a | |
2028 | * function to drain a single device's I/O queue. | |
922453bc SH |
2029 | */ |
2030 | void bdrv_drain_all(void) | |
2031 | { | |
88266f5a SH |
2032 | /* Always run first iteration so any pending completion BHs run */ |
2033 | bool busy = true; | |
922453bc SH |
2034 | BlockDriverState *bs; |
2035 | ||
88266f5a | 2036 | while (busy) { |
9b536adc SH |
2037 | busy = false; |
2038 | ||
dc364f4c | 2039 | QTAILQ_FOREACH(bs, &bdrv_states, device_list) { |
9b536adc | 2040 | AioContext *aio_context = bdrv_get_aio_context(bs); |
9b536adc SH |
2041 | |
2042 | aio_context_acquire(aio_context); | |
5b98db0a | 2043 | busy |= bdrv_drain_one(bs); |
9b536adc | 2044 | aio_context_release(aio_context); |
9b536adc | 2045 | } |
922453bc SH |
2046 | } |
2047 | } | |
2048 | ||
dc364f4c BC |
2049 | /* make a BlockDriverState anonymous by removing from bdrv_state and |
2050 | * graph_bdrv_state list. | |
d22b2f41 RH |
2051 | Also, NULL terminate the device_name to prevent double remove */ |
2052 | void bdrv_make_anon(BlockDriverState *bs) | |
2053 | { | |
bfb197e0 MA |
2054 | /* |
2055 | * Take care to remove bs from bdrv_states only when it's actually | |
2056 | * in it. Note that bs->device_list.tqe_prev is initially null, | |
2057 | * and gets set to non-null by QTAILQ_INSERT_TAIL(). Establish | |
2058 | * the useful invariant "bs in bdrv_states iff bs->tqe_prev" by | |
2059 | * resetting it to null on remove. | |
2060 | */ | |
2061 | if (bs->device_list.tqe_prev) { | |
dc364f4c | 2062 | QTAILQ_REMOVE(&bdrv_states, bs, device_list); |
bfb197e0 | 2063 | bs->device_list.tqe_prev = NULL; |
d22b2f41 | 2064 | } |
dc364f4c BC |
2065 | if (bs->node_name[0] != '\0') { |
2066 | QTAILQ_REMOVE(&graph_bdrv_states, bs, node_list); | |
2067 | } | |
2068 | bs->node_name[0] = '\0'; | |
d22b2f41 RH |
2069 | } |
2070 | ||
e023b2e2 PB |
2071 | static void bdrv_rebind(BlockDriverState *bs) |
2072 | { | |
2073 | if (bs->drv && bs->drv->bdrv_rebind) { | |
2074 | bs->drv->bdrv_rebind(bs); | |
2075 | } | |
2076 | } | |
2077 | ||
4ddc07ca PB |
2078 | static void bdrv_move_feature_fields(BlockDriverState *bs_dest, |
2079 | BlockDriverState *bs_src) | |
8802d1fd | 2080 | { |
4ddc07ca | 2081 | /* move some fields that need to stay attached to the device */ |
8802d1fd JC |
2082 | |
2083 | /* dev info */ | |
1b7fd729 | 2084 | bs_dest->guest_block_size = bs_src->guest_block_size; |
4ddc07ca | 2085 | bs_dest->copy_on_read = bs_src->copy_on_read; |
8802d1fd | 2086 | |
4ddc07ca | 2087 | bs_dest->enable_write_cache = bs_src->enable_write_cache; |
c4a248a1 | 2088 | |
cc0681c4 BC |
2089 | /* i/o throttled req */ |
2090 | memcpy(&bs_dest->throttle_state, | |
2091 | &bs_src->throttle_state, | |
2092 | sizeof(ThrottleState)); | |
2093 | bs_dest->throttled_reqs[0] = bs_src->throttled_reqs[0]; | |
2094 | bs_dest->throttled_reqs[1] = bs_src->throttled_reqs[1]; | |
4ddc07ca | 2095 | bs_dest->io_limits_enabled = bs_src->io_limits_enabled; |
8802d1fd | 2096 | |
8802d1fd | 2097 | /* r/w error */ |
4ddc07ca PB |
2098 | bs_dest->on_read_error = bs_src->on_read_error; |
2099 | bs_dest->on_write_error = bs_src->on_write_error; | |
8802d1fd JC |
2100 | |
2101 | /* i/o status */ | |
4ddc07ca PB |
2102 | bs_dest->iostatus_enabled = bs_src->iostatus_enabled; |
2103 | bs_dest->iostatus = bs_src->iostatus; | |
8802d1fd | 2104 | |
a9fc4408 | 2105 | /* dirty bitmap */ |
e4654d2d | 2106 | bs_dest->dirty_bitmaps = bs_src->dirty_bitmaps; |
a9fc4408 | 2107 | |
9fcb0251 FZ |
2108 | /* reference count */ |
2109 | bs_dest->refcnt = bs_src->refcnt; | |
2110 | ||
a9fc4408 | 2111 | /* job */ |
4ddc07ca | 2112 | bs_dest->job = bs_src->job; |
a9fc4408 | 2113 | |
8802d1fd | 2114 | /* keep the same entry in bdrv_states */ |
dc364f4c | 2115 | bs_dest->device_list = bs_src->device_list; |
7e7d56d9 MA |
2116 | bs_dest->blk = bs_src->blk; |
2117 | ||
fbe40ff7 FZ |
2118 | memcpy(bs_dest->op_blockers, bs_src->op_blockers, |
2119 | sizeof(bs_dest->op_blockers)); | |
4ddc07ca | 2120 | } |
8802d1fd | 2121 | |
4ddc07ca PB |
2122 | /* |
2123 | * Swap bs contents for two image chains while they are live, | |
2124 | * while keeping required fields on the BlockDriverState that is | |
2125 | * actually attached to a device. | |
2126 | * | |
2127 | * This will modify the BlockDriverState fields, and swap contents | |
2128 | * between bs_new and bs_old. Both bs_new and bs_old are modified. | |
2129 | * | |
bfb197e0 | 2130 | * bs_new must not be attached to a BlockBackend. |
4ddc07ca PB |
2131 | * |
2132 | * This function does not create any image files. | |
2133 | */ | |
2134 | void bdrv_swap(BlockDriverState *bs_new, BlockDriverState *bs_old) | |
2135 | { | |
2136 | BlockDriverState tmp; | |
f6801b83 | 2137 | |
90ce8a06 BC |
2138 | /* The code needs to swap the node_name but simply swapping node_list won't |
2139 | * work so first remove the nodes from the graph list, do the swap then | |
2140 | * insert them back if needed. | |
2141 | */ | |
2142 | if (bs_new->node_name[0] != '\0') { | |
2143 | QTAILQ_REMOVE(&graph_bdrv_states, bs_new, node_list); | |
2144 | } | |
2145 | if (bs_old->node_name[0] != '\0') { | |
2146 | QTAILQ_REMOVE(&graph_bdrv_states, bs_old, node_list); | |
2147 | } | |
2148 | ||
bfb197e0 | 2149 | /* bs_new must be unattached and shouldn't have anything fancy enabled */ |
7e7d56d9 | 2150 | assert(!bs_new->blk); |
e4654d2d | 2151 | assert(QLIST_EMPTY(&bs_new->dirty_bitmaps)); |
4ddc07ca | 2152 | assert(bs_new->job == NULL); |
4ddc07ca | 2153 | assert(bs_new->io_limits_enabled == false); |
cc0681c4 | 2154 | assert(!throttle_have_timer(&bs_new->throttle_state)); |
8802d1fd | 2155 | |
4ddc07ca PB |
2156 | tmp = *bs_new; |
2157 | *bs_new = *bs_old; | |
2158 | *bs_old = tmp; | |
a9fc4408 | 2159 | |
4ddc07ca PB |
2160 | /* there are some fields that should not be swapped, move them back */ |
2161 | bdrv_move_feature_fields(&tmp, bs_old); | |
2162 | bdrv_move_feature_fields(bs_old, bs_new); | |
2163 | bdrv_move_feature_fields(bs_new, &tmp); | |
8802d1fd | 2164 | |
bfb197e0 | 2165 | /* bs_new must remain unattached */ |
7e7d56d9 | 2166 | assert(!bs_new->blk); |
4ddc07ca PB |
2167 | |
2168 | /* Check a few fields that should remain attached to the device */ | |
4ddc07ca | 2169 | assert(bs_new->job == NULL); |
4ddc07ca | 2170 | assert(bs_new->io_limits_enabled == false); |
cc0681c4 | 2171 | assert(!throttle_have_timer(&bs_new->throttle_state)); |
e023b2e2 | 2172 | |
90ce8a06 BC |
2173 | /* insert the nodes back into the graph node list if needed */ |
2174 | if (bs_new->node_name[0] != '\0') { | |
2175 | QTAILQ_INSERT_TAIL(&graph_bdrv_states, bs_new, node_list); | |
2176 | } | |
2177 | if (bs_old->node_name[0] != '\0') { | |
2178 | QTAILQ_INSERT_TAIL(&graph_bdrv_states, bs_old, node_list); | |
2179 | } | |
2180 | ||
e023b2e2 | 2181 | bdrv_rebind(bs_new); |
4ddc07ca PB |
2182 | bdrv_rebind(bs_old); |
2183 | } | |
2184 | ||
2185 | /* | |
2186 | * Add new bs contents at the top of an image chain while the chain is | |
2187 | * live, while keeping required fields on the top layer. | |
2188 | * | |
2189 | * This will modify the BlockDriverState fields, and swap contents | |
2190 | * between bs_new and bs_top. Both bs_new and bs_top are modified. | |
2191 | * | |
bfb197e0 | 2192 | * bs_new must not be attached to a BlockBackend. |
4ddc07ca PB |
2193 | * |
2194 | * This function does not create any image files. | |
2195 | */ | |
2196 | void bdrv_append(BlockDriverState *bs_new, BlockDriverState *bs_top) | |
2197 | { | |
2198 | bdrv_swap(bs_new, bs_top); | |
2199 | ||
2200 | /* The contents of 'tmp' will become bs_top, as we are | |
2201 | * swapping bs_new and bs_top contents. */ | |
8d24cce1 | 2202 | bdrv_set_backing_hd(bs_top, bs_new); |
8802d1fd JC |
2203 | } |
2204 | ||
4f6fd349 | 2205 | static void bdrv_delete(BlockDriverState *bs) |
b338082b | 2206 | { |
3e914655 | 2207 | assert(!bs->job); |
3718d8ab | 2208 | assert(bdrv_op_blocker_is_empty(bs)); |
4f6fd349 | 2209 | assert(!bs->refcnt); |
e4654d2d | 2210 | assert(QLIST_EMPTY(&bs->dirty_bitmaps)); |
18846dee | 2211 | |
e1b5c52e SH |
2212 | bdrv_close(bs); |
2213 | ||
1b7bdbc1 | 2214 | /* remove from list, if necessary */ |
d22b2f41 | 2215 | bdrv_make_anon(bs); |
34c6f050 | 2216 | |
7267c094 | 2217 | g_free(bs); |
fc01f7e7 FB |
2218 | } |
2219 | ||
e97fc193 AL |
2220 | /* |
2221 | * Run consistency checks on an image | |
2222 | * | |
e076f338 | 2223 | * Returns 0 if the check could be completed (it doesn't mean that the image is |
a1c7273b | 2224 | * free of errors) or -errno when an internal error occurred. The results of the |
e076f338 | 2225 | * check are stored in res. |
e97fc193 | 2226 | */ |
4534ff54 | 2227 | int bdrv_check(BlockDriverState *bs, BdrvCheckResult *res, BdrvCheckMode fix) |
e97fc193 | 2228 | { |
908bcd54 HR |
2229 | if (bs->drv == NULL) { |
2230 | return -ENOMEDIUM; | |
2231 | } | |
e97fc193 AL |
2232 | if (bs->drv->bdrv_check == NULL) { |
2233 | return -ENOTSUP; | |
2234 | } | |
2235 | ||
e076f338 | 2236 | memset(res, 0, sizeof(*res)); |
4534ff54 | 2237 | return bs->drv->bdrv_check(bs, res, fix); |
e97fc193 AL |
2238 | } |
2239 | ||
8a426614 KW |
2240 | #define COMMIT_BUF_SECTORS 2048 |
2241 | ||
33e3963e FB |
2242 | /* commit COW file into the raw image */ |
2243 | int bdrv_commit(BlockDriverState *bs) | |
2244 | { | |
19cb3738 | 2245 | BlockDriver *drv = bs->drv; |
72706ea4 | 2246 | int64_t sector, total_sectors, length, backing_length; |
8a426614 | 2247 | int n, ro, open_flags; |
0bce597d | 2248 | int ret = 0; |
72706ea4 | 2249 | uint8_t *buf = NULL; |
33e3963e | 2250 | |
19cb3738 FB |
2251 | if (!drv) |
2252 | return -ENOMEDIUM; | |
6bb45158 | 2253 | |
4dca4b63 NS |
2254 | if (!bs->backing_hd) { |
2255 | return -ENOTSUP; | |
33e3963e FB |
2256 | } |
2257 | ||
bb00021d FZ |
2258 | if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_COMMIT_SOURCE, NULL) || |
2259 | bdrv_op_is_blocked(bs->backing_hd, BLOCK_OP_TYPE_COMMIT_TARGET, NULL)) { | |
2d3735d3 SH |
2260 | return -EBUSY; |
2261 | } | |
2262 | ||
4dca4b63 | 2263 | ro = bs->backing_hd->read_only; |
4dca4b63 NS |
2264 | open_flags = bs->backing_hd->open_flags; |
2265 | ||
2266 | if (ro) { | |
0bce597d JC |
2267 | if (bdrv_reopen(bs->backing_hd, open_flags | BDRV_O_RDWR, NULL)) { |
2268 | return -EACCES; | |
4dca4b63 | 2269 | } |
ea2384d3 | 2270 | } |
33e3963e | 2271 | |
72706ea4 JC |
2272 | length = bdrv_getlength(bs); |
2273 | if (length < 0) { | |
2274 | ret = length; | |
2275 | goto ro_cleanup; | |
2276 | } | |
2277 | ||
2278 | backing_length = bdrv_getlength(bs->backing_hd); | |
2279 | if (backing_length < 0) { | |
2280 | ret = backing_length; | |
2281 | goto ro_cleanup; | |
2282 | } | |
2283 | ||
2284 | /* If our top snapshot is larger than the backing file image, | |
2285 | * grow the backing file image if possible. If not possible, | |
2286 | * we must return an error */ | |
2287 | if (length > backing_length) { | |
2288 | ret = bdrv_truncate(bs->backing_hd, length); | |
2289 | if (ret < 0) { | |
2290 | goto ro_cleanup; | |
2291 | } | |
2292 | } | |
2293 | ||
2294 | total_sectors = length >> BDRV_SECTOR_BITS; | |
857d4f46 KW |
2295 | |
2296 | /* qemu_try_blockalign() for bs will choose an alignment that works for | |
2297 | * bs->backing_hd as well, so no need to compare the alignment manually. */ | |
2298 | buf = qemu_try_blockalign(bs, COMMIT_BUF_SECTORS * BDRV_SECTOR_SIZE); | |
2299 | if (buf == NULL) { | |
2300 | ret = -ENOMEM; | |
2301 | goto ro_cleanup; | |
2302 | } | |
8a426614 KW |
2303 | |
2304 | for (sector = 0; sector < total_sectors; sector += n) { | |
d663640c PB |
2305 | ret = bdrv_is_allocated(bs, sector, COMMIT_BUF_SECTORS, &n); |
2306 | if (ret < 0) { | |
2307 | goto ro_cleanup; | |
2308 | } | |
2309 | if (ret) { | |
dabfa6cc KW |
2310 | ret = bdrv_read(bs, sector, buf, n); |
2311 | if (ret < 0) { | |
8a426614 KW |
2312 | goto ro_cleanup; |
2313 | } | |
2314 | ||
dabfa6cc KW |
2315 | ret = bdrv_write(bs->backing_hd, sector, buf, n); |
2316 | if (ret < 0) { | |
8a426614 KW |
2317 | goto ro_cleanup; |
2318 | } | |
ea2384d3 | 2319 | } |
33e3963e | 2320 | } |
95389c86 | 2321 | |
1d44952f CH |
2322 | if (drv->bdrv_make_empty) { |
2323 | ret = drv->bdrv_make_empty(bs); | |
dabfa6cc KW |
2324 | if (ret < 0) { |
2325 | goto ro_cleanup; | |
2326 | } | |
1d44952f CH |
2327 | bdrv_flush(bs); |
2328 | } | |
95389c86 | 2329 | |
3f5075ae CH |
2330 | /* |
2331 | * Make sure all data we wrote to the backing device is actually | |
2332 | * stable on disk. | |
2333 | */ | |
dabfa6cc | 2334 | if (bs->backing_hd) { |
3f5075ae | 2335 | bdrv_flush(bs->backing_hd); |
dabfa6cc | 2336 | } |
4dca4b63 | 2337 | |
dabfa6cc | 2338 | ret = 0; |
4dca4b63 | 2339 | ro_cleanup: |
857d4f46 | 2340 | qemu_vfree(buf); |
4dca4b63 NS |
2341 | |
2342 | if (ro) { | |
0bce597d JC |
2343 | /* ignoring error return here */ |
2344 | bdrv_reopen(bs->backing_hd, open_flags & ~BDRV_O_RDWR, NULL); | |
4dca4b63 NS |
2345 | } |
2346 | ||
1d44952f | 2347 | return ret; |
33e3963e FB |
2348 | } |
2349 | ||
e8877497 | 2350 | int bdrv_commit_all(void) |
6ab4b5ab MA |
2351 | { |
2352 | BlockDriverState *bs; | |
2353 | ||
dc364f4c | 2354 | QTAILQ_FOREACH(bs, &bdrv_states, device_list) { |
ed78cda3 SH |
2355 | AioContext *aio_context = bdrv_get_aio_context(bs); |
2356 | ||
2357 | aio_context_acquire(aio_context); | |
272d2d8e JC |
2358 | if (bs->drv && bs->backing_hd) { |
2359 | int ret = bdrv_commit(bs); | |
2360 | if (ret < 0) { | |
ed78cda3 | 2361 | aio_context_release(aio_context); |
272d2d8e JC |
2362 | return ret; |
2363 | } | |
e8877497 | 2364 | } |
ed78cda3 | 2365 | aio_context_release(aio_context); |
6ab4b5ab | 2366 | } |
e8877497 | 2367 | return 0; |
6ab4b5ab MA |
2368 | } |
2369 | ||
dbffbdcf SH |
2370 | /** |
2371 | * Remove an active request from the tracked requests list | |
2372 | * | |
2373 | * This function should be called when a tracked request is completing. | |
2374 | */ | |
2375 | static void tracked_request_end(BdrvTrackedRequest *req) | |
2376 | { | |
2dbafdc0 KW |
2377 | if (req->serialising) { |
2378 | req->bs->serialising_in_flight--; | |
2379 | } | |
2380 | ||
dbffbdcf | 2381 | QLIST_REMOVE(req, list); |
f4658285 | 2382 | qemu_co_queue_restart_all(&req->wait_queue); |
dbffbdcf SH |
2383 | } |
2384 | ||
2385 | /** | |
2386 | * Add an active request to the tracked requests list | |
2387 | */ | |
2388 | static void tracked_request_begin(BdrvTrackedRequest *req, | |
2389 | BlockDriverState *bs, | |
793ed47a KW |
2390 | int64_t offset, |
2391 | unsigned int bytes, bool is_write) | |
dbffbdcf SH |
2392 | { |
2393 | *req = (BdrvTrackedRequest){ | |
2394 | .bs = bs, | |
2dbafdc0 KW |
2395 | .offset = offset, |
2396 | .bytes = bytes, | |
2397 | .is_write = is_write, | |
2398 | .co = qemu_coroutine_self(), | |
2399 | .serialising = false, | |
7327145f KW |
2400 | .overlap_offset = offset, |
2401 | .overlap_bytes = bytes, | |
dbffbdcf SH |
2402 | }; |
2403 | ||
f4658285 SH |
2404 | qemu_co_queue_init(&req->wait_queue); |
2405 | ||
dbffbdcf SH |
2406 | QLIST_INSERT_HEAD(&bs->tracked_requests, req, list); |
2407 | } | |
2408 | ||
e96126ff | 2409 | static void mark_request_serialising(BdrvTrackedRequest *req, uint64_t align) |
2dbafdc0 | 2410 | { |
7327145f | 2411 | int64_t overlap_offset = req->offset & ~(align - 1); |
e96126ff KW |
2412 | unsigned int overlap_bytes = ROUND_UP(req->offset + req->bytes, align) |
2413 | - overlap_offset; | |
7327145f | 2414 | |
2dbafdc0 KW |
2415 | if (!req->serialising) { |
2416 | req->bs->serialising_in_flight++; | |
2417 | req->serialising = true; | |
2418 | } | |
7327145f KW |
2419 | |
2420 | req->overlap_offset = MIN(req->overlap_offset, overlap_offset); | |
2421 | req->overlap_bytes = MAX(req->overlap_bytes, overlap_bytes); | |
2dbafdc0 KW |
2422 | } |
2423 | ||
d83947ac SH |
2424 | /** |
2425 | * Round a region to cluster boundaries | |
2426 | */ | |
343bded4 PB |
2427 | void bdrv_round_to_clusters(BlockDriverState *bs, |
2428 | int64_t sector_num, int nb_sectors, | |
2429 | int64_t *cluster_sector_num, | |
2430 | int *cluster_nb_sectors) | |
d83947ac SH |
2431 | { |
2432 | BlockDriverInfo bdi; | |
2433 | ||
2434 | if (bdrv_get_info(bs, &bdi) < 0 || bdi.cluster_size == 0) { | |
2435 | *cluster_sector_num = sector_num; | |
2436 | *cluster_nb_sectors = nb_sectors; | |
2437 | } else { | |
2438 | int64_t c = bdi.cluster_size / BDRV_SECTOR_SIZE; | |
2439 | *cluster_sector_num = QEMU_ALIGN_DOWN(sector_num, c); | |
2440 | *cluster_nb_sectors = QEMU_ALIGN_UP(sector_num - *cluster_sector_num + | |
2441 | nb_sectors, c); | |
2442 | } | |
2443 | } | |
2444 | ||
7327145f | 2445 | static int bdrv_get_cluster_size(BlockDriverState *bs) |
793ed47a KW |
2446 | { |
2447 | BlockDriverInfo bdi; | |
7327145f | 2448 | int ret; |
793ed47a | 2449 | |
7327145f KW |
2450 | ret = bdrv_get_info(bs, &bdi); |
2451 | if (ret < 0 || bdi.cluster_size == 0) { | |
2452 | return bs->request_alignment; | |
793ed47a | 2453 | } else { |
7327145f | 2454 | return bdi.cluster_size; |
793ed47a KW |
2455 | } |
2456 | } | |
2457 | ||
f4658285 | 2458 | static bool tracked_request_overlaps(BdrvTrackedRequest *req, |
793ed47a KW |
2459 | int64_t offset, unsigned int bytes) |
2460 | { | |
d83947ac | 2461 | /* aaaa bbbb */ |
7327145f | 2462 | if (offset >= req->overlap_offset + req->overlap_bytes) { |
d83947ac SH |
2463 | return false; |
2464 | } | |
2465 | /* bbbb aaaa */ | |
7327145f | 2466 | if (req->overlap_offset >= offset + bytes) { |
d83947ac SH |
2467 | return false; |
2468 | } | |
2469 | return true; | |
f4658285 SH |
2470 | } |
2471 | ||
28de2dcd | 2472 | static bool coroutine_fn wait_serialising_requests(BdrvTrackedRequest *self) |
f4658285 | 2473 | { |
2dbafdc0 | 2474 | BlockDriverState *bs = self->bs; |
f4658285 SH |
2475 | BdrvTrackedRequest *req; |
2476 | bool retry; | |
28de2dcd | 2477 | bool waited = false; |
f4658285 | 2478 | |
2dbafdc0 | 2479 | if (!bs->serialising_in_flight) { |
28de2dcd | 2480 | return false; |
2dbafdc0 KW |
2481 | } |
2482 | ||
f4658285 SH |
2483 | do { |
2484 | retry = false; | |
2485 | QLIST_FOREACH(req, &bs->tracked_requests, list) { | |
2dbafdc0 | 2486 | if (req == self || (!req->serialising && !self->serialising)) { |
65afd211 KW |
2487 | continue; |
2488 | } | |
7327145f KW |
2489 | if (tracked_request_overlaps(req, self->overlap_offset, |
2490 | self->overlap_bytes)) | |
2491 | { | |
5f8b6491 SH |
2492 | /* Hitting this means there was a reentrant request, for |
2493 | * example, a block driver issuing nested requests. This must | |
2494 | * never happen since it means deadlock. | |
2495 | */ | |
2496 | assert(qemu_coroutine_self() != req->co); | |
2497 | ||
6460440f KW |
2498 | /* If the request is already (indirectly) waiting for us, or |
2499 | * will wait for us as soon as it wakes up, then just go on | |
2500 | * (instead of producing a deadlock in the former case). */ | |
2501 | if (!req->waiting_for) { | |
2502 | self->waiting_for = req; | |
2503 | qemu_co_queue_wait(&req->wait_queue); | |
2504 | self->waiting_for = NULL; | |
2505 | retry = true; | |
28de2dcd | 2506 | waited = true; |
6460440f KW |
2507 | break; |
2508 | } | |
f4658285 SH |
2509 | } |
2510 | } | |
2511 | } while (retry); | |
28de2dcd KW |
2512 | |
2513 | return waited; | |
f4658285 SH |
2514 | } |
2515 | ||
756e6736 KW |
2516 | /* |
2517 | * Return values: | |
2518 | * 0 - success | |
2519 | * -EINVAL - backing format specified, but no file | |
2520 | * -ENOSPC - can't update the backing file because no space is left in the | |
2521 | * image file header | |
2522 | * -ENOTSUP - format driver doesn't support changing the backing file | |
2523 | */ | |
2524 | int bdrv_change_backing_file(BlockDriverState *bs, | |
2525 | const char *backing_file, const char *backing_fmt) | |
2526 | { | |
2527 | BlockDriver *drv = bs->drv; | |
469ef350 | 2528 | int ret; |
756e6736 | 2529 | |
5f377794 PB |
2530 | /* Backing file format doesn't make sense without a backing file */ |
2531 | if (backing_fmt && !backing_file) { | |
2532 | return -EINVAL; | |
2533 | } | |
2534 | ||
756e6736 | 2535 | if (drv->bdrv_change_backing_file != NULL) { |
469ef350 | 2536 | ret = drv->bdrv_change_backing_file(bs, backing_file, backing_fmt); |
756e6736 | 2537 | } else { |
469ef350 | 2538 | ret = -ENOTSUP; |
756e6736 | 2539 | } |
469ef350 PB |
2540 | |
2541 | if (ret == 0) { | |
2542 | pstrcpy(bs->backing_file, sizeof(bs->backing_file), backing_file ?: ""); | |
2543 | pstrcpy(bs->backing_format, sizeof(bs->backing_format), backing_fmt ?: ""); | |
2544 | } | |
2545 | return ret; | |
756e6736 KW |
2546 | } |
2547 | ||
6ebdcee2 JC |
2548 | /* |
2549 | * Finds the image layer in the chain that has 'bs' as its backing file. | |
2550 | * | |
2551 | * active is the current topmost image. | |
2552 | * | |
2553 | * Returns NULL if bs is not found in active's image chain, | |
2554 | * or if active == bs. | |
4caf0fcd JC |
2555 | * |
2556 | * Returns the bottommost base image if bs == NULL. | |
6ebdcee2 JC |
2557 | */ |
2558 | BlockDriverState *bdrv_find_overlay(BlockDriverState *active, | |
2559 | BlockDriverState *bs) | |
2560 | { | |
4caf0fcd JC |
2561 | while (active && bs != active->backing_hd) { |
2562 | active = active->backing_hd; | |
6ebdcee2 JC |
2563 | } |
2564 | ||
4caf0fcd JC |
2565 | return active; |
2566 | } | |
6ebdcee2 | 2567 | |
4caf0fcd JC |
2568 | /* Given a BDS, searches for the base layer. */ |
2569 | BlockDriverState *bdrv_find_base(BlockDriverState *bs) | |
2570 | { | |
2571 | return bdrv_find_overlay(bs, NULL); | |
6ebdcee2 JC |
2572 | } |
2573 | ||
2574 | typedef struct BlkIntermediateStates { | |
2575 | BlockDriverState *bs; | |
2576 | QSIMPLEQ_ENTRY(BlkIntermediateStates) entry; | |
2577 | } BlkIntermediateStates; | |
2578 | ||
2579 | ||
2580 | /* | |
2581 | * Drops images above 'base' up to and including 'top', and sets the image | |
2582 | * above 'top' to have base as its backing file. | |
2583 | * | |
2584 | * Requires that the overlay to 'top' is opened r/w, so that the backing file | |
2585 | * information in 'bs' can be properly updated. | |
2586 | * | |
2587 | * E.g., this will convert the following chain: | |
2588 | * bottom <- base <- intermediate <- top <- active | |
2589 | * | |
2590 | * to | |
2591 | * | |
2592 | * bottom <- base <- active | |
2593 | * | |
2594 | * It is allowed for bottom==base, in which case it converts: | |
2595 | * | |
2596 | * base <- intermediate <- top <- active | |
2597 | * | |
2598 | * to | |
2599 | * | |
2600 | * base <- active | |
2601 | * | |
54e26900 JC |
2602 | * If backing_file_str is non-NULL, it will be used when modifying top's |
2603 | * overlay image metadata. | |
2604 | * | |
6ebdcee2 JC |
2605 | * Error conditions: |
2606 | * if active == top, that is considered an error | |
2607 | * | |
2608 | */ | |
2609 | int bdrv_drop_intermediate(BlockDriverState *active, BlockDriverState *top, | |
54e26900 | 2610 | BlockDriverState *base, const char *backing_file_str) |
6ebdcee2 JC |
2611 | { |
2612 | BlockDriverState *intermediate; | |
2613 | BlockDriverState *base_bs = NULL; | |
2614 | BlockDriverState *new_top_bs = NULL; | |
2615 | BlkIntermediateStates *intermediate_state, *next; | |
2616 | int ret = -EIO; | |
2617 | ||
2618 | QSIMPLEQ_HEAD(states_to_delete, BlkIntermediateStates) states_to_delete; | |
2619 | QSIMPLEQ_INIT(&states_to_delete); | |
2620 | ||
2621 | if (!top->drv || !base->drv) { | |
2622 | goto exit; | |
2623 | } | |
2624 | ||
2625 | new_top_bs = bdrv_find_overlay(active, top); | |
2626 | ||
2627 | if (new_top_bs == NULL) { | |
2628 | /* we could not find the image above 'top', this is an error */ | |
2629 | goto exit; | |
2630 | } | |
2631 | ||
2632 | /* special case of new_top_bs->backing_hd already pointing to base - nothing | |
2633 | * to do, no intermediate images */ | |
2634 | if (new_top_bs->backing_hd == base) { | |
2635 | ret = 0; | |
2636 | goto exit; | |
2637 | } | |
2638 | ||
2639 | intermediate = top; | |
2640 | ||
2641 | /* now we will go down through the list, and add each BDS we find | |
2642 | * into our deletion queue, until we hit the 'base' | |
2643 | */ | |
2644 | while (intermediate) { | |
5839e53b | 2645 | intermediate_state = g_new0(BlkIntermediateStates, 1); |
6ebdcee2 JC |
2646 | intermediate_state->bs = intermediate; |
2647 | QSIMPLEQ_INSERT_TAIL(&states_to_delete, intermediate_state, entry); | |
2648 | ||
2649 | if (intermediate->backing_hd == base) { | |
2650 | base_bs = intermediate->backing_hd; | |
2651 | break; | |
2652 | } | |
2653 | intermediate = intermediate->backing_hd; | |
2654 | } | |
2655 | if (base_bs == NULL) { | |
2656 | /* something went wrong, we did not end at the base. safely | |
2657 | * unravel everything, and exit with error */ | |
2658 | goto exit; | |
2659 | } | |
2660 | ||
2661 | /* success - we can delete the intermediate states, and link top->base */ | |
54e26900 JC |
2662 | backing_file_str = backing_file_str ? backing_file_str : base_bs->filename; |
2663 | ret = bdrv_change_backing_file(new_top_bs, backing_file_str, | |
6ebdcee2 JC |
2664 | base_bs->drv ? base_bs->drv->format_name : ""); |
2665 | if (ret) { | |
2666 | goto exit; | |
2667 | } | |
920beae1 | 2668 | bdrv_set_backing_hd(new_top_bs, base_bs); |
6ebdcee2 JC |
2669 | |
2670 | QSIMPLEQ_FOREACH_SAFE(intermediate_state, &states_to_delete, entry, next) { | |
2671 | /* so that bdrv_close() does not recursively close the chain */ | |
920beae1 | 2672 | bdrv_set_backing_hd(intermediate_state->bs, NULL); |
4f6fd349 | 2673 | bdrv_unref(intermediate_state->bs); |
6ebdcee2 JC |
2674 | } |
2675 | ret = 0; | |
2676 | ||
2677 | exit: | |
2678 | QSIMPLEQ_FOREACH_SAFE(intermediate_state, &states_to_delete, entry, next) { | |
2679 | g_free(intermediate_state); | |
2680 | } | |
2681 | return ret; | |
2682 | } | |
2683 | ||
2684 | ||
71d0770c AL |
2685 | static int bdrv_check_byte_request(BlockDriverState *bs, int64_t offset, |
2686 | size_t size) | |
2687 | { | |
75af1f34 | 2688 | if (size > BDRV_REQUEST_MAX_SECTORS << BDRV_SECTOR_BITS) { |
1dd3a447 KW |
2689 | return -EIO; |
2690 | } | |
2691 | ||
c0191e76 | 2692 | if (!bdrv_is_inserted(bs)) { |
71d0770c | 2693 | return -ENOMEDIUM; |
c0191e76 | 2694 | } |
71d0770c | 2695 | |
c0191e76 | 2696 | if (offset < 0) { |
71d0770c | 2697 | return -EIO; |
c0191e76 | 2698 | } |
71d0770c AL |
2699 | |
2700 | return 0; | |
2701 | } | |
2702 | ||
2703 | static int bdrv_check_request(BlockDriverState *bs, int64_t sector_num, | |
2704 | int nb_sectors) | |
2705 | { | |
75af1f34 | 2706 | if (nb_sectors < 0 || nb_sectors > BDRV_REQUEST_MAX_SECTORS) { |
8f4754ed KW |
2707 | return -EIO; |
2708 | } | |
2709 | ||
eb5a3165 JS |
2710 | return bdrv_check_byte_request(bs, sector_num * BDRV_SECTOR_SIZE, |
2711 | nb_sectors * BDRV_SECTOR_SIZE); | |
71d0770c AL |
2712 | } |
2713 | ||
1c9805a3 SH |
2714 | typedef struct RwCo { |
2715 | BlockDriverState *bs; | |
775aa8b6 | 2716 | int64_t offset; |
1c9805a3 SH |
2717 | QEMUIOVector *qiov; |
2718 | bool is_write; | |
2719 | int ret; | |
4105eaaa | 2720 | BdrvRequestFlags flags; |
1c9805a3 SH |
2721 | } RwCo; |
2722 | ||
2723 | static void coroutine_fn bdrv_rw_co_entry(void *opaque) | |
fc01f7e7 | 2724 | { |
1c9805a3 | 2725 | RwCo *rwco = opaque; |
ea2384d3 | 2726 | |
1c9805a3 | 2727 | if (!rwco->is_write) { |
775aa8b6 KW |
2728 | rwco->ret = bdrv_co_do_preadv(rwco->bs, rwco->offset, |
2729 | rwco->qiov->size, rwco->qiov, | |
4105eaaa | 2730 | rwco->flags); |
775aa8b6 KW |
2731 | } else { |
2732 | rwco->ret = bdrv_co_do_pwritev(rwco->bs, rwco->offset, | |
2733 | rwco->qiov->size, rwco->qiov, | |
2734 | rwco->flags); | |
1c9805a3 SH |
2735 | } |
2736 | } | |
e7a8a783 | 2737 | |
1c9805a3 | 2738 | /* |
8d3b1a2d | 2739 | * Process a vectored synchronous request using coroutines |
1c9805a3 | 2740 | */ |
775aa8b6 KW |
2741 | static int bdrv_prwv_co(BlockDriverState *bs, int64_t offset, |
2742 | QEMUIOVector *qiov, bool is_write, | |
2743 | BdrvRequestFlags flags) | |
1c9805a3 | 2744 | { |
1c9805a3 SH |
2745 | Coroutine *co; |
2746 | RwCo rwco = { | |
2747 | .bs = bs, | |
775aa8b6 | 2748 | .offset = offset, |
8d3b1a2d | 2749 | .qiov = qiov, |
1c9805a3 SH |
2750 | .is_write = is_write, |
2751 | .ret = NOT_DONE, | |
4105eaaa | 2752 | .flags = flags, |
1c9805a3 | 2753 | }; |
e7a8a783 | 2754 | |
498e386c ZYW |
2755 | /** |
2756 | * In sync call context, when the vcpu is blocked, this throttling timer | |
2757 | * will not fire; so the I/O throttling function has to be disabled here | |
2758 | * if it has been enabled. | |
2759 | */ | |
2760 | if (bs->io_limits_enabled) { | |
2761 | fprintf(stderr, "Disabling I/O throttling on '%s' due " | |
2762 | "to synchronous I/O.\n", bdrv_get_device_name(bs)); | |
2763 | bdrv_io_limits_disable(bs); | |
2764 | } | |
2765 | ||
1c9805a3 SH |
2766 | if (qemu_in_coroutine()) { |
2767 | /* Fast-path if already in coroutine context */ | |
2768 | bdrv_rw_co_entry(&rwco); | |
2769 | } else { | |
2572b37a SH |
2770 | AioContext *aio_context = bdrv_get_aio_context(bs); |
2771 | ||
1c9805a3 SH |
2772 | co = qemu_coroutine_create(bdrv_rw_co_entry); |
2773 | qemu_coroutine_enter(co, &rwco); | |
2774 | while (rwco.ret == NOT_DONE) { | |
2572b37a | 2775 | aio_poll(aio_context, true); |
1c9805a3 SH |
2776 | } |
2777 | } | |
2778 | return rwco.ret; | |
2779 | } | |
b338082b | 2780 | |
8d3b1a2d KW |
2781 | /* |
2782 | * Process a synchronous request using coroutines | |
2783 | */ | |
2784 | static int bdrv_rw_co(BlockDriverState *bs, int64_t sector_num, uint8_t *buf, | |
4105eaaa | 2785 | int nb_sectors, bool is_write, BdrvRequestFlags flags) |
8d3b1a2d KW |
2786 | { |
2787 | QEMUIOVector qiov; | |
2788 | struct iovec iov = { | |
2789 | .iov_base = (void *)buf, | |
2790 | .iov_len = nb_sectors * BDRV_SECTOR_SIZE, | |
2791 | }; | |
2792 | ||
75af1f34 | 2793 | if (nb_sectors < 0 || nb_sectors > BDRV_REQUEST_MAX_SECTORS) { |
da15ee51 KW |
2794 | return -EINVAL; |
2795 | } | |
2796 | ||
8d3b1a2d | 2797 | qemu_iovec_init_external(&qiov, &iov, 1); |
775aa8b6 KW |
2798 | return bdrv_prwv_co(bs, sector_num << BDRV_SECTOR_BITS, |
2799 | &qiov, is_write, flags); | |
8d3b1a2d KW |
2800 | } |
2801 | ||
1c9805a3 SH |
2802 | /* return < 0 if error. See bdrv_write() for the return codes */ |
2803 | int bdrv_read(BlockDriverState *bs, int64_t sector_num, | |
2804 | uint8_t *buf, int nb_sectors) | |
2805 | { | |
4105eaaa | 2806 | return bdrv_rw_co(bs, sector_num, buf, nb_sectors, false, 0); |
fc01f7e7 FB |
2807 | } |
2808 | ||
07d27a44 MA |
2809 | /* Just like bdrv_read(), but with I/O throttling temporarily disabled */ |
2810 | int bdrv_read_unthrottled(BlockDriverState *bs, int64_t sector_num, | |
2811 | uint8_t *buf, int nb_sectors) | |
2812 | { | |
2813 | bool enabled; | |
2814 | int ret; | |
2815 | ||
2816 | enabled = bs->io_limits_enabled; | |
2817 | bs->io_limits_enabled = false; | |
4e7395e8 | 2818 | ret = bdrv_read(bs, sector_num, buf, nb_sectors); |
07d27a44 MA |
2819 | bs->io_limits_enabled = enabled; |
2820 | return ret; | |
2821 | } | |
2822 | ||
5fafdf24 | 2823 | /* Return < 0 if error. Important errors are: |
19cb3738 FB |
2824 | -EIO generic I/O error (may happen for all errors) |
2825 | -ENOMEDIUM No media inserted. | |
2826 | -EINVAL Invalid sector number or nb_sectors | |
2827 | -EACCES Trying to write a read-only device | |
2828 | */ | |
5fafdf24 | 2829 | int bdrv_write(BlockDriverState *bs, int64_t sector_num, |
fc01f7e7 FB |
2830 | const uint8_t *buf, int nb_sectors) |
2831 | { | |
4105eaaa | 2832 | return bdrv_rw_co(bs, sector_num, (uint8_t *)buf, nb_sectors, true, 0); |
83f64091 FB |
2833 | } |
2834 | ||
aa7bfbff PL |
2835 | int bdrv_write_zeroes(BlockDriverState *bs, int64_t sector_num, |
2836 | int nb_sectors, BdrvRequestFlags flags) | |
4105eaaa PL |
2837 | { |
2838 | return bdrv_rw_co(bs, sector_num, NULL, nb_sectors, true, | |
aa7bfbff | 2839 | BDRV_REQ_ZERO_WRITE | flags); |
8d3b1a2d KW |
2840 | } |
2841 | ||
d75cbb5e PL |
2842 | /* |
2843 | * Completely zero out a block device with the help of bdrv_write_zeroes. | |
2844 | * The operation is sped up by checking the block status and only writing | |
2845 | * zeroes to the device if they currently do not return zeroes. Optional | |
2846 | * flags are passed through to bdrv_write_zeroes (e.g. BDRV_REQ_MAY_UNMAP). | |
2847 | * | |
2848 | * Returns < 0 on error, 0 on success. For error codes see bdrv_write(). | |
2849 | */ | |
2850 | int bdrv_make_zero(BlockDriverState *bs, BdrvRequestFlags flags) | |
2851 | { | |
d32f7c10 | 2852 | int64_t target_sectors, ret, nb_sectors, sector_num = 0; |
d75cbb5e PL |
2853 | int n; |
2854 | ||
d32f7c10 MA |
2855 | target_sectors = bdrv_nb_sectors(bs); |
2856 | if (target_sectors < 0) { | |
2857 | return target_sectors; | |
9ce10c0b | 2858 | } |
9ce10c0b | 2859 | |
d75cbb5e | 2860 | for (;;) { |
75af1f34 | 2861 | nb_sectors = MIN(target_sectors - sector_num, BDRV_REQUEST_MAX_SECTORS); |
d75cbb5e PL |
2862 | if (nb_sectors <= 0) { |
2863 | return 0; | |
2864 | } | |
d75cbb5e | 2865 | ret = bdrv_get_block_status(bs, sector_num, nb_sectors, &n); |
3d94ce60 PL |
2866 | if (ret < 0) { |
2867 | error_report("error getting block status at sector %" PRId64 ": %s", | |
2868 | sector_num, strerror(-ret)); | |
2869 | return ret; | |
2870 | } | |
d75cbb5e PL |
2871 | if (ret & BDRV_BLOCK_ZERO) { |
2872 | sector_num += n; | |
2873 | continue; | |
2874 | } | |
2875 | ret = bdrv_write_zeroes(bs, sector_num, n, flags); | |
2876 | if (ret < 0) { | |
2877 | error_report("error writing zeroes at sector %" PRId64 ": %s", | |
2878 | sector_num, strerror(-ret)); | |
2879 | return ret; | |
2880 | } | |
2881 | sector_num += n; | |
2882 | } | |
2883 | } | |
2884 | ||
a3ef6571 | 2885 | int bdrv_pread(BlockDriverState *bs, int64_t offset, void *buf, int bytes) |
83f64091 | 2886 | { |
a3ef6571 KW |
2887 | QEMUIOVector qiov; |
2888 | struct iovec iov = { | |
2889 | .iov_base = (void *)buf, | |
2890 | .iov_len = bytes, | |
2891 | }; | |
9a8c4cce | 2892 | int ret; |
83f64091 | 2893 | |
a3ef6571 KW |
2894 | if (bytes < 0) { |
2895 | return -EINVAL; | |
83f64091 FB |
2896 | } |
2897 | ||
a3ef6571 KW |
2898 | qemu_iovec_init_external(&qiov, &iov, 1); |
2899 | ret = bdrv_prwv_co(bs, offset, &qiov, false, 0); | |
2900 | if (ret < 0) { | |
2901 | return ret; | |
83f64091 | 2902 | } |
a3ef6571 KW |
2903 | |
2904 | return bytes; | |
83f64091 FB |
2905 | } |
2906 | ||
8d3b1a2d | 2907 | int bdrv_pwritev(BlockDriverState *bs, int64_t offset, QEMUIOVector *qiov) |
83f64091 | 2908 | { |
9a8c4cce | 2909 | int ret; |
83f64091 | 2910 | |
8407d5d7 KW |
2911 | ret = bdrv_prwv_co(bs, offset, qiov, true, 0); |
2912 | if (ret < 0) { | |
2913 | return ret; | |
83f64091 FB |
2914 | } |
2915 | ||
8d3b1a2d KW |
2916 | return qiov->size; |
2917 | } | |
2918 | ||
2919 | int bdrv_pwrite(BlockDriverState *bs, int64_t offset, | |
8407d5d7 | 2920 | const void *buf, int bytes) |
8d3b1a2d KW |
2921 | { |
2922 | QEMUIOVector qiov; | |
2923 | struct iovec iov = { | |
2924 | .iov_base = (void *) buf, | |
8407d5d7 | 2925 | .iov_len = bytes, |
8d3b1a2d KW |
2926 | }; |
2927 | ||
8407d5d7 KW |
2928 | if (bytes < 0) { |
2929 | return -EINVAL; | |
2930 | } | |
2931 | ||
8d3b1a2d KW |
2932 | qemu_iovec_init_external(&qiov, &iov, 1); |
2933 | return bdrv_pwritev(bs, offset, &qiov); | |
83f64091 | 2934 | } |
83f64091 | 2935 | |
f08145fe KW |
2936 | /* |
2937 | * Writes to the file and ensures that no writes are reordered across this | |
2938 | * request (acts as a barrier) | |
2939 | * | |
2940 | * Returns 0 on success, -errno in error cases. | |
2941 | */ | |
2942 | int bdrv_pwrite_sync(BlockDriverState *bs, int64_t offset, | |
2943 | const void *buf, int count) | |
2944 | { | |
2945 | int ret; | |
2946 | ||
2947 | ret = bdrv_pwrite(bs, offset, buf, count); | |
2948 | if (ret < 0) { | |
2949 | return ret; | |
2950 | } | |
2951 | ||
f05fa4ad PB |
2952 | /* No flush needed for cache modes that already do it */ |
2953 | if (bs->enable_write_cache) { | |
f08145fe KW |
2954 | bdrv_flush(bs); |
2955 | } | |
2956 | ||
2957 | return 0; | |
2958 | } | |
2959 | ||
470c0504 | 2960 | static int coroutine_fn bdrv_co_do_copy_on_readv(BlockDriverState *bs, |
ab185921 SH |
2961 | int64_t sector_num, int nb_sectors, QEMUIOVector *qiov) |
2962 | { | |
2963 | /* Perform I/O through a temporary buffer so that users who scribble over | |
2964 | * their read buffer while the operation is in progress do not end up | |
2965 | * modifying the image file. This is critical for zero-copy guest I/O | |
2966 | * where anything might happen inside guest memory. | |
2967 | */ | |
2968 | void *bounce_buffer; | |
2969 | ||
79c053bd | 2970 | BlockDriver *drv = bs->drv; |
ab185921 SH |
2971 | struct iovec iov; |
2972 | QEMUIOVector bounce_qiov; | |
2973 | int64_t cluster_sector_num; | |
2974 | int cluster_nb_sectors; | |
2975 | size_t skip_bytes; | |
2976 | int ret; | |
2977 | ||
2978 | /* Cover entire cluster so no additional backing file I/O is required when | |
2979 | * allocating cluster in the image file. | |
2980 | */ | |
343bded4 PB |
2981 | bdrv_round_to_clusters(bs, sector_num, nb_sectors, |
2982 | &cluster_sector_num, &cluster_nb_sectors); | |
ab185921 | 2983 | |
470c0504 SH |
2984 | trace_bdrv_co_do_copy_on_readv(bs, sector_num, nb_sectors, |
2985 | cluster_sector_num, cluster_nb_sectors); | |
ab185921 SH |
2986 | |
2987 | iov.iov_len = cluster_nb_sectors * BDRV_SECTOR_SIZE; | |
857d4f46 KW |
2988 | iov.iov_base = bounce_buffer = qemu_try_blockalign(bs, iov.iov_len); |
2989 | if (bounce_buffer == NULL) { | |
2990 | ret = -ENOMEM; | |
2991 | goto err; | |
2992 | } | |
2993 | ||
ab185921 SH |
2994 | qemu_iovec_init_external(&bounce_qiov, &iov, 1); |
2995 | ||
79c053bd SH |
2996 | ret = drv->bdrv_co_readv(bs, cluster_sector_num, cluster_nb_sectors, |
2997 | &bounce_qiov); | |
ab185921 SH |
2998 | if (ret < 0) { |
2999 | goto err; | |
3000 | } | |
3001 | ||
79c053bd SH |
3002 | if (drv->bdrv_co_write_zeroes && |
3003 | buffer_is_zero(bounce_buffer, iov.iov_len)) { | |
621f0589 | 3004 | ret = bdrv_co_do_write_zeroes(bs, cluster_sector_num, |
aa7bfbff | 3005 | cluster_nb_sectors, 0); |
79c053bd | 3006 | } else { |
f05fa4ad PB |
3007 | /* This does not change the data on the disk, it is not necessary |
3008 | * to flush even in cache=writethrough mode. | |
3009 | */ | |
79c053bd | 3010 | ret = drv->bdrv_co_writev(bs, cluster_sector_num, cluster_nb_sectors, |
ab185921 | 3011 | &bounce_qiov); |
79c053bd SH |
3012 | } |
3013 | ||
ab185921 SH |
3014 | if (ret < 0) { |
3015 | /* It might be okay to ignore write errors for guest requests. If this | |
3016 | * is a deliberate copy-on-read then we don't want to ignore the error. | |
3017 | * Simply report it in all cases. | |
3018 | */ | |
3019 | goto err; | |
3020 | } | |
3021 | ||
3022 | skip_bytes = (sector_num - cluster_sector_num) * BDRV_SECTOR_SIZE; | |
03396148 MT |
3023 | qemu_iovec_from_buf(qiov, 0, bounce_buffer + skip_bytes, |
3024 | nb_sectors * BDRV_SECTOR_SIZE); | |
ab185921 SH |
3025 | |
3026 | err: | |
3027 | qemu_vfree(bounce_buffer); | |
3028 | return ret; | |
3029 | } | |
3030 | ||
c5fbe571 | 3031 | /* |
d0c7f642 KW |
3032 | * Forwards an already correctly aligned request to the BlockDriver. This |
3033 | * handles copy on read and zeroing after EOF; any other features must be | |
3034 | * implemented by the caller. | |
c5fbe571 | 3035 | */ |
d0c7f642 | 3036 | static int coroutine_fn bdrv_aligned_preadv(BlockDriverState *bs, |
65afd211 | 3037 | BdrvTrackedRequest *req, int64_t offset, unsigned int bytes, |
ec746e10 | 3038 | int64_t align, QEMUIOVector *qiov, int flags) |
da1fa91d KW |
3039 | { |
3040 | BlockDriver *drv = bs->drv; | |
dbffbdcf | 3041 | int ret; |
da1fa91d | 3042 | |
d0c7f642 KW |
3043 | int64_t sector_num = offset >> BDRV_SECTOR_BITS; |
3044 | unsigned int nb_sectors = bytes >> BDRV_SECTOR_BITS; | |
da1fa91d | 3045 | |
d0c7f642 KW |
3046 | assert((offset & (BDRV_SECTOR_SIZE - 1)) == 0); |
3047 | assert((bytes & (BDRV_SECTOR_SIZE - 1)) == 0); | |
8eb029c2 | 3048 | assert(!qiov || bytes == qiov->size); |
d0c7f642 KW |
3049 | |
3050 | /* Handle Copy on Read and associated serialisation */ | |
470c0504 | 3051 | if (flags & BDRV_REQ_COPY_ON_READ) { |
7327145f KW |
3052 | /* If we touch the same cluster it counts as an overlap. This |
3053 | * guarantees that allocating writes will be serialized and not race | |
3054 | * with each other for the same cluster. For example, in copy-on-read | |
3055 | * it ensures that the CoR read and write operations are atomic and | |
3056 | * guest writes cannot interleave between them. */ | |
3057 | mark_request_serialising(req, bdrv_get_cluster_size(bs)); | |
470c0504 SH |
3058 | } |
3059 | ||
2dbafdc0 | 3060 | wait_serialising_requests(req); |
f4658285 | 3061 | |
470c0504 | 3062 | if (flags & BDRV_REQ_COPY_ON_READ) { |
ab185921 SH |
3063 | int pnum; |
3064 | ||
bdad13b9 | 3065 | ret = bdrv_is_allocated(bs, sector_num, nb_sectors, &pnum); |
ab185921 SH |
3066 | if (ret < 0) { |
3067 | goto out; | |
3068 | } | |
3069 | ||
3070 | if (!ret || pnum != nb_sectors) { | |
470c0504 | 3071 | ret = bdrv_co_do_copy_on_readv(bs, sector_num, nb_sectors, qiov); |
ab185921 SH |
3072 | goto out; |
3073 | } | |
3074 | } | |
3075 | ||
d0c7f642 | 3076 | /* Forward the request to the BlockDriver */ |
c0191e76 | 3077 | if (!bs->zero_beyond_eof) { |
893a8f62 MK |
3078 | ret = drv->bdrv_co_readv(bs, sector_num, nb_sectors, qiov); |
3079 | } else { | |
c0191e76 | 3080 | /* Read zeros after EOF */ |
4049082c | 3081 | int64_t total_sectors, max_nb_sectors; |
893a8f62 | 3082 | |
4049082c MA |
3083 | total_sectors = bdrv_nb_sectors(bs); |
3084 | if (total_sectors < 0) { | |
3085 | ret = total_sectors; | |
893a8f62 MK |
3086 | goto out; |
3087 | } | |
3088 | ||
5f5bcd80 KW |
3089 | max_nb_sectors = ROUND_UP(MAX(0, total_sectors - sector_num), |
3090 | align >> BDRV_SECTOR_BITS); | |
e012b78c PB |
3091 | if (nb_sectors < max_nb_sectors) { |
3092 | ret = drv->bdrv_co_readv(bs, sector_num, nb_sectors, qiov); | |
3093 | } else if (max_nb_sectors > 0) { | |
33f461e0 | 3094 | QEMUIOVector local_qiov; |
33f461e0 KW |
3095 | |
3096 | qemu_iovec_init(&local_qiov, qiov->niov); | |
3097 | qemu_iovec_concat(&local_qiov, qiov, 0, | |
e012b78c | 3098 | max_nb_sectors * BDRV_SECTOR_SIZE); |
33f461e0 | 3099 | |
e012b78c | 3100 | ret = drv->bdrv_co_readv(bs, sector_num, max_nb_sectors, |
33f461e0 KW |
3101 | &local_qiov); |
3102 | ||
3103 | qemu_iovec_destroy(&local_qiov); | |
893a8f62 MK |
3104 | } else { |
3105 | ret = 0; | |
3106 | } | |
3107 | ||
3108 | /* Reading beyond end of file is supposed to produce zeroes */ | |
3109 | if (ret == 0 && total_sectors < sector_num + nb_sectors) { | |
3110 | uint64_t offset = MAX(0, total_sectors - sector_num); | |
3111 | uint64_t bytes = (sector_num + nb_sectors - offset) * | |
3112 | BDRV_SECTOR_SIZE; | |
3113 | qemu_iovec_memset(qiov, offset * BDRV_SECTOR_SIZE, 0, bytes); | |
3114 | } | |
3115 | } | |
ab185921 SH |
3116 | |
3117 | out: | |
dbffbdcf | 3118 | return ret; |
da1fa91d KW |
3119 | } |
3120 | ||
d0c7f642 KW |
3121 | /* |
3122 | * Handle a read request in coroutine context | |
3123 | */ | |
1b0288ae KW |
3124 | static int coroutine_fn bdrv_co_do_preadv(BlockDriverState *bs, |
3125 | int64_t offset, unsigned int bytes, QEMUIOVector *qiov, | |
d0c7f642 KW |
3126 | BdrvRequestFlags flags) |
3127 | { | |
3128 | BlockDriver *drv = bs->drv; | |
65afd211 KW |
3129 | BdrvTrackedRequest req; |
3130 | ||
1b0288ae KW |
3131 | /* TODO Lift BDRV_SECTOR_SIZE restriction in BlockDriver interface */ |
3132 | uint64_t align = MAX(BDRV_SECTOR_SIZE, bs->request_alignment); | |
3133 | uint8_t *head_buf = NULL; | |
3134 | uint8_t *tail_buf = NULL; | |
3135 | QEMUIOVector local_qiov; | |
3136 | bool use_local_qiov = false; | |
d0c7f642 KW |
3137 | int ret; |
3138 | ||
3139 | if (!drv) { | |
3140 | return -ENOMEDIUM; | |
3141 | } | |
b9c64947 HR |
3142 | |
3143 | ret = bdrv_check_byte_request(bs, offset, bytes); | |
3144 | if (ret < 0) { | |
3145 | return ret; | |
d0c7f642 KW |
3146 | } |
3147 | ||
3148 | if (bs->copy_on_read) { | |
3149 | flags |= BDRV_REQ_COPY_ON_READ; | |
3150 | } | |
3151 | ||
3152 | /* throttling disk I/O */ | |
3153 | if (bs->io_limits_enabled) { | |
d5103588 | 3154 | bdrv_io_limits_intercept(bs, bytes, false); |
1b0288ae KW |
3155 | } |
3156 | ||
3157 | /* Align read if necessary by padding qiov */ | |
3158 | if (offset & (align - 1)) { | |
3159 | head_buf = qemu_blockalign(bs, align); | |
3160 | qemu_iovec_init(&local_qiov, qiov->niov + 2); | |
3161 | qemu_iovec_add(&local_qiov, head_buf, offset & (align - 1)); | |
3162 | qemu_iovec_concat(&local_qiov, qiov, 0, qiov->size); | |
3163 | use_local_qiov = true; | |
3164 | ||
3165 | bytes += offset & (align - 1); | |
3166 | offset = offset & ~(align - 1); | |
3167 | } | |
3168 | ||
3169 | if ((offset + bytes) & (align - 1)) { | |
3170 | if (!use_local_qiov) { | |
3171 | qemu_iovec_init(&local_qiov, qiov->niov + 1); | |
3172 | qemu_iovec_concat(&local_qiov, qiov, 0, qiov->size); | |
3173 | use_local_qiov = true; | |
3174 | } | |
3175 | tail_buf = qemu_blockalign(bs, align); | |
3176 | qemu_iovec_add(&local_qiov, tail_buf, | |
3177 | align - ((offset + bytes) & (align - 1))); | |
3178 | ||
3179 | bytes = ROUND_UP(bytes, align); | |
3180 | } | |
3181 | ||
65afd211 | 3182 | tracked_request_begin(&req, bs, offset, bytes, false); |
ec746e10 | 3183 | ret = bdrv_aligned_preadv(bs, &req, offset, bytes, align, |
1b0288ae KW |
3184 | use_local_qiov ? &local_qiov : qiov, |
3185 | flags); | |
65afd211 | 3186 | tracked_request_end(&req); |
1b0288ae KW |
3187 | |
3188 | if (use_local_qiov) { | |
3189 | qemu_iovec_destroy(&local_qiov); | |
3190 | qemu_vfree(head_buf); | |
3191 | qemu_vfree(tail_buf); | |
d0c7f642 KW |
3192 | } |
3193 | ||
d0c7f642 KW |
3194 | return ret; |
3195 | } | |
3196 | ||
1b0288ae KW |
3197 | static int coroutine_fn bdrv_co_do_readv(BlockDriverState *bs, |
3198 | int64_t sector_num, int nb_sectors, QEMUIOVector *qiov, | |
3199 | BdrvRequestFlags flags) | |
3200 | { | |
75af1f34 | 3201 | if (nb_sectors < 0 || nb_sectors > BDRV_REQUEST_MAX_SECTORS) { |
1b0288ae KW |
3202 | return -EINVAL; |
3203 | } | |
3204 | ||
3205 | return bdrv_co_do_preadv(bs, sector_num << BDRV_SECTOR_BITS, | |
3206 | nb_sectors << BDRV_SECTOR_BITS, qiov, flags); | |
3207 | } | |
3208 | ||
c5fbe571 | 3209 | int coroutine_fn bdrv_co_readv(BlockDriverState *bs, int64_t sector_num, |
da1fa91d KW |
3210 | int nb_sectors, QEMUIOVector *qiov) |
3211 | { | |
c5fbe571 | 3212 | trace_bdrv_co_readv(bs, sector_num, nb_sectors); |
da1fa91d | 3213 | |
470c0504 SH |
3214 | return bdrv_co_do_readv(bs, sector_num, nb_sectors, qiov, 0); |
3215 | } | |
3216 | ||
3217 | int coroutine_fn bdrv_co_copy_on_readv(BlockDriverState *bs, | |
3218 | int64_t sector_num, int nb_sectors, QEMUIOVector *qiov) | |
3219 | { | |
3220 | trace_bdrv_co_copy_on_readv(bs, sector_num, nb_sectors); | |
3221 | ||
3222 | return bdrv_co_do_readv(bs, sector_num, nb_sectors, qiov, | |
3223 | BDRV_REQ_COPY_ON_READ); | |
c5fbe571 SH |
3224 | } |
3225 | ||
98764152 | 3226 | #define MAX_WRITE_ZEROES_BOUNCE_BUFFER 32768 |
c31cb707 | 3227 | |
f08f2dda | 3228 | static int coroutine_fn bdrv_co_do_write_zeroes(BlockDriverState *bs, |
aa7bfbff | 3229 | int64_t sector_num, int nb_sectors, BdrvRequestFlags flags) |
f08f2dda SH |
3230 | { |
3231 | BlockDriver *drv = bs->drv; | |
3232 | QEMUIOVector qiov; | |
c31cb707 PL |
3233 | struct iovec iov = {0}; |
3234 | int ret = 0; | |
f08f2dda | 3235 | |
75af1f34 PL |
3236 | int max_write_zeroes = MIN_NON_ZERO(bs->bl.max_write_zeroes, |
3237 | BDRV_REQUEST_MAX_SECTORS); | |
621f0589 | 3238 | |
c31cb707 PL |
3239 | while (nb_sectors > 0 && !ret) { |
3240 | int num = nb_sectors; | |
3241 | ||
b8d71c09 PB |
3242 | /* Align request. Block drivers can expect the "bulk" of the request |
3243 | * to be aligned. | |
3244 | */ | |
3245 | if (bs->bl.write_zeroes_alignment | |
3246 | && num > bs->bl.write_zeroes_alignment) { | |
3247 | if (sector_num % bs->bl.write_zeroes_alignment != 0) { | |
3248 | /* Make a small request up to the first aligned sector. */ | |
c31cb707 | 3249 | num = bs->bl.write_zeroes_alignment; |
b8d71c09 PB |
3250 | num -= sector_num % bs->bl.write_zeroes_alignment; |
3251 | } else if ((sector_num + num) % bs->bl.write_zeroes_alignment != 0) { | |
3252 | /* Shorten the request to the last aligned sector. num cannot | |
3253 | * underflow because num > bs->bl.write_zeroes_alignment. | |
3254 | */ | |
3255 | num -= (sector_num + num) % bs->bl.write_zeroes_alignment; | |
c31cb707 | 3256 | } |
621f0589 | 3257 | } |
f08f2dda | 3258 | |
c31cb707 PL |
3259 | /* limit request size */ |
3260 | if (num > max_write_zeroes) { | |
3261 | num = max_write_zeroes; | |
3262 | } | |
3263 | ||
3264 | ret = -ENOTSUP; | |
3265 | /* First try the efficient write zeroes operation */ | |
3266 | if (drv->bdrv_co_write_zeroes) { | |
3267 | ret = drv->bdrv_co_write_zeroes(bs, sector_num, num, flags); | |
3268 | } | |
3269 | ||
3270 | if (ret == -ENOTSUP) { | |
3271 | /* Fall back to bounce buffer if write zeroes is unsupported */ | |
095e4fa4 | 3272 | int max_xfer_len = MIN_NON_ZERO(bs->bl.max_transfer_length, |
98764152 | 3273 | MAX_WRITE_ZEROES_BOUNCE_BUFFER); |
095e4fa4 | 3274 | num = MIN(num, max_xfer_len); |
c31cb707 PL |
3275 | iov.iov_len = num * BDRV_SECTOR_SIZE; |
3276 | if (iov.iov_base == NULL) { | |
857d4f46 KW |
3277 | iov.iov_base = qemu_try_blockalign(bs, num * BDRV_SECTOR_SIZE); |
3278 | if (iov.iov_base == NULL) { | |
3279 | ret = -ENOMEM; | |
3280 | goto fail; | |
3281 | } | |
b8d71c09 | 3282 | memset(iov.iov_base, 0, num * BDRV_SECTOR_SIZE); |
c31cb707 PL |
3283 | } |
3284 | qemu_iovec_init_external(&qiov, &iov, 1); | |
f08f2dda | 3285 | |
c31cb707 | 3286 | ret = drv->bdrv_co_writev(bs, sector_num, num, &qiov); |
b8d71c09 PB |
3287 | |
3288 | /* Keep bounce buffer around if it is big enough for all | |
3289 | * all future requests. | |
3290 | */ | |
095e4fa4 | 3291 | if (num < max_xfer_len) { |
b8d71c09 PB |
3292 | qemu_vfree(iov.iov_base); |
3293 | iov.iov_base = NULL; | |
3294 | } | |
c31cb707 PL |
3295 | } |
3296 | ||
3297 | sector_num += num; | |
3298 | nb_sectors -= num; | |
3299 | } | |
f08f2dda | 3300 | |
857d4f46 | 3301 | fail: |
f08f2dda SH |
3302 | qemu_vfree(iov.iov_base); |
3303 | return ret; | |
3304 | } | |
3305 | ||
c5fbe571 | 3306 | /* |
b404f720 | 3307 | * Forwards an already correctly aligned write request to the BlockDriver. |
c5fbe571 | 3308 | */ |
b404f720 | 3309 | static int coroutine_fn bdrv_aligned_pwritev(BlockDriverState *bs, |
65afd211 KW |
3310 | BdrvTrackedRequest *req, int64_t offset, unsigned int bytes, |
3311 | QEMUIOVector *qiov, int flags) | |
c5fbe571 SH |
3312 | { |
3313 | BlockDriver *drv = bs->drv; | |
28de2dcd | 3314 | bool waited; |
6b7cb247 | 3315 | int ret; |
da1fa91d | 3316 | |
b404f720 KW |
3317 | int64_t sector_num = offset >> BDRV_SECTOR_BITS; |
3318 | unsigned int nb_sectors = bytes >> BDRV_SECTOR_BITS; | |
f4658285 | 3319 | |
b404f720 KW |
3320 | assert((offset & (BDRV_SECTOR_SIZE - 1)) == 0); |
3321 | assert((bytes & (BDRV_SECTOR_SIZE - 1)) == 0); | |
8eb029c2 | 3322 | assert(!qiov || bytes == qiov->size); |
cc0681c4 | 3323 | |
28de2dcd KW |
3324 | waited = wait_serialising_requests(req); |
3325 | assert(!waited || !req->serialising); | |
af91f9a7 KW |
3326 | assert(req->overlap_offset <= offset); |
3327 | assert(offset + bytes <= req->overlap_offset + req->overlap_bytes); | |
244eadef | 3328 | |
65afd211 | 3329 | ret = notifier_with_return_list_notify(&bs->before_write_notifiers, req); |
d616b224 | 3330 | |
465bee1d PL |
3331 | if (!ret && bs->detect_zeroes != BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF && |
3332 | !(flags & BDRV_REQ_ZERO_WRITE) && drv->bdrv_co_write_zeroes && | |
3333 | qemu_iovec_is_zero(qiov)) { | |
3334 | flags |= BDRV_REQ_ZERO_WRITE; | |
3335 | if (bs->detect_zeroes == BLOCKDEV_DETECT_ZEROES_OPTIONS_UNMAP) { | |
3336 | flags |= BDRV_REQ_MAY_UNMAP; | |
3337 | } | |
3338 | } | |
3339 | ||
d616b224 SH |
3340 | if (ret < 0) { |
3341 | /* Do nothing, write notifier decided to fail this request */ | |
3342 | } else if (flags & BDRV_REQ_ZERO_WRITE) { | |
9e1cb96d | 3343 | BLKDBG_EVENT(bs, BLKDBG_PWRITEV_ZERO); |
aa7bfbff | 3344 | ret = bdrv_co_do_write_zeroes(bs, sector_num, nb_sectors, flags); |
f08f2dda | 3345 | } else { |
9e1cb96d | 3346 | BLKDBG_EVENT(bs, BLKDBG_PWRITEV); |
f08f2dda SH |
3347 | ret = drv->bdrv_co_writev(bs, sector_num, nb_sectors, qiov); |
3348 | } | |
9e1cb96d | 3349 | BLKDBG_EVENT(bs, BLKDBG_PWRITEV_DONE); |
6b7cb247 | 3350 | |
f05fa4ad PB |
3351 | if (ret == 0 && !bs->enable_write_cache) { |
3352 | ret = bdrv_co_flush(bs); | |
3353 | } | |
3354 | ||
e4654d2d | 3355 | bdrv_set_dirty(bs, sector_num, nb_sectors); |
da1fa91d | 3356 | |
5366d0c8 | 3357 | block_acct_highest_sector(&bs->stats, sector_num, nb_sectors); |
5e5a94b6 | 3358 | |
c0191e76 | 3359 | if (ret >= 0) { |
df2a6f29 PB |
3360 | bs->total_sectors = MAX(bs->total_sectors, sector_num + nb_sectors); |
3361 | } | |
da1fa91d | 3362 | |
6b7cb247 | 3363 | return ret; |
da1fa91d KW |
3364 | } |
3365 | ||
b404f720 KW |
3366 | /* |
3367 | * Handle a write request in coroutine context | |
3368 | */ | |
6601553e KW |
3369 | static int coroutine_fn bdrv_co_do_pwritev(BlockDriverState *bs, |
3370 | int64_t offset, unsigned int bytes, QEMUIOVector *qiov, | |
b404f720 KW |
3371 | BdrvRequestFlags flags) |
3372 | { | |
65afd211 | 3373 | BdrvTrackedRequest req; |
3b8242e0 KW |
3374 | /* TODO Lift BDRV_SECTOR_SIZE restriction in BlockDriver interface */ |
3375 | uint64_t align = MAX(BDRV_SECTOR_SIZE, bs->request_alignment); | |
3376 | uint8_t *head_buf = NULL; | |
3377 | uint8_t *tail_buf = NULL; | |
3378 | QEMUIOVector local_qiov; | |
3379 | bool use_local_qiov = false; | |
b404f720 KW |
3380 | int ret; |
3381 | ||
3382 | if (!bs->drv) { | |
3383 | return -ENOMEDIUM; | |
3384 | } | |
3385 | if (bs->read_only) { | |
3386 | return -EACCES; | |
3387 | } | |
b9c64947 HR |
3388 | |
3389 | ret = bdrv_check_byte_request(bs, offset, bytes); | |
3390 | if (ret < 0) { | |
3391 | return ret; | |
b404f720 KW |
3392 | } |
3393 | ||
b404f720 KW |
3394 | /* throttling disk I/O */ |
3395 | if (bs->io_limits_enabled) { | |
d5103588 | 3396 | bdrv_io_limits_intercept(bs, bytes, true); |
b404f720 KW |
3397 | } |
3398 | ||
3b8242e0 KW |
3399 | /* |
3400 | * Align write if necessary by performing a read-modify-write cycle. | |
3401 | * Pad qiov with the read parts and be sure to have a tracked request not | |
3402 | * only for bdrv_aligned_pwritev, but also for the reads of the RMW cycle. | |
3403 | */ | |
65afd211 | 3404 | tracked_request_begin(&req, bs, offset, bytes, true); |
3b8242e0 KW |
3405 | |
3406 | if (offset & (align - 1)) { | |
3407 | QEMUIOVector head_qiov; | |
3408 | struct iovec head_iov; | |
3409 | ||
3410 | mark_request_serialising(&req, align); | |
3411 | wait_serialising_requests(&req); | |
3412 | ||
3413 | head_buf = qemu_blockalign(bs, align); | |
3414 | head_iov = (struct iovec) { | |
3415 | .iov_base = head_buf, | |
3416 | .iov_len = align, | |
3417 | }; | |
3418 | qemu_iovec_init_external(&head_qiov, &head_iov, 1); | |
3419 | ||
9e1cb96d | 3420 | BLKDBG_EVENT(bs, BLKDBG_PWRITEV_RMW_HEAD); |
3b8242e0 KW |
3421 | ret = bdrv_aligned_preadv(bs, &req, offset & ~(align - 1), align, |
3422 | align, &head_qiov, 0); | |
3423 | if (ret < 0) { | |
3424 | goto fail; | |
3425 | } | |
9e1cb96d | 3426 | BLKDBG_EVENT(bs, BLKDBG_PWRITEV_RMW_AFTER_HEAD); |
3b8242e0 KW |
3427 | |
3428 | qemu_iovec_init(&local_qiov, qiov->niov + 2); | |
3429 | qemu_iovec_add(&local_qiov, head_buf, offset & (align - 1)); | |
3430 | qemu_iovec_concat(&local_qiov, qiov, 0, qiov->size); | |
3431 | use_local_qiov = true; | |
3432 | ||
3433 | bytes += offset & (align - 1); | |
3434 | offset = offset & ~(align - 1); | |
3435 | } | |
3436 | ||
3437 | if ((offset + bytes) & (align - 1)) { | |
3438 | QEMUIOVector tail_qiov; | |
3439 | struct iovec tail_iov; | |
3440 | size_t tail_bytes; | |
28de2dcd | 3441 | bool waited; |
3b8242e0 KW |
3442 | |
3443 | mark_request_serialising(&req, align); | |
28de2dcd KW |
3444 | waited = wait_serialising_requests(&req); |
3445 | assert(!waited || !use_local_qiov); | |
3b8242e0 KW |
3446 | |
3447 | tail_buf = qemu_blockalign(bs, align); | |
3448 | tail_iov = (struct iovec) { | |
3449 | .iov_base = tail_buf, | |
3450 | .iov_len = align, | |
3451 | }; | |
3452 | qemu_iovec_init_external(&tail_qiov, &tail_iov, 1); | |
3453 | ||
9e1cb96d | 3454 | BLKDBG_EVENT(bs, BLKDBG_PWRITEV_RMW_TAIL); |
3b8242e0 KW |
3455 | ret = bdrv_aligned_preadv(bs, &req, (offset + bytes) & ~(align - 1), align, |
3456 | align, &tail_qiov, 0); | |
3457 | if (ret < 0) { | |
3458 | goto fail; | |
3459 | } | |
9e1cb96d | 3460 | BLKDBG_EVENT(bs, BLKDBG_PWRITEV_RMW_AFTER_TAIL); |
3b8242e0 KW |
3461 | |
3462 | if (!use_local_qiov) { | |
3463 | qemu_iovec_init(&local_qiov, qiov->niov + 1); | |
3464 | qemu_iovec_concat(&local_qiov, qiov, 0, qiov->size); | |
3465 | use_local_qiov = true; | |
3466 | } | |
3467 | ||
3468 | tail_bytes = (offset + bytes) & (align - 1); | |
3469 | qemu_iovec_add(&local_qiov, tail_buf + tail_bytes, align - tail_bytes); | |
3470 | ||
3471 | bytes = ROUND_UP(bytes, align); | |
3472 | } | |
3473 | ||
3474 | ret = bdrv_aligned_pwritev(bs, &req, offset, bytes, | |
3475 | use_local_qiov ? &local_qiov : qiov, | |
3476 | flags); | |
3477 | ||
3478 | fail: | |
65afd211 | 3479 | tracked_request_end(&req); |
b404f720 | 3480 | |
3b8242e0 KW |
3481 | if (use_local_qiov) { |
3482 | qemu_iovec_destroy(&local_qiov); | |
3b8242e0 | 3483 | } |
99c4a85c KW |
3484 | qemu_vfree(head_buf); |
3485 | qemu_vfree(tail_buf); | |
3b8242e0 | 3486 | |
b404f720 KW |
3487 | return ret; |
3488 | } | |
3489 | ||
6601553e KW |
3490 | static int coroutine_fn bdrv_co_do_writev(BlockDriverState *bs, |
3491 | int64_t sector_num, int nb_sectors, QEMUIOVector *qiov, | |
3492 | BdrvRequestFlags flags) | |
3493 | { | |
75af1f34 | 3494 | if (nb_sectors < 0 || nb_sectors > BDRV_REQUEST_MAX_SECTORS) { |
6601553e KW |
3495 | return -EINVAL; |
3496 | } | |
3497 | ||
3498 | return bdrv_co_do_pwritev(bs, sector_num << BDRV_SECTOR_BITS, | |
3499 | nb_sectors << BDRV_SECTOR_BITS, qiov, flags); | |
3500 | } | |
3501 | ||
c5fbe571 SH |
3502 | int coroutine_fn bdrv_co_writev(BlockDriverState *bs, int64_t sector_num, |
3503 | int nb_sectors, QEMUIOVector *qiov) | |
3504 | { | |
3505 | trace_bdrv_co_writev(bs, sector_num, nb_sectors); | |
3506 | ||
f08f2dda SH |
3507 | return bdrv_co_do_writev(bs, sector_num, nb_sectors, qiov, 0); |
3508 | } | |
3509 | ||
3510 | int coroutine_fn bdrv_co_write_zeroes(BlockDriverState *bs, | |
aa7bfbff PL |
3511 | int64_t sector_num, int nb_sectors, |
3512 | BdrvRequestFlags flags) | |
f08f2dda | 3513 | { |
94d6ff21 | 3514 | trace_bdrv_co_write_zeroes(bs, sector_num, nb_sectors, flags); |
f08f2dda | 3515 | |
d32f35cb PL |
3516 | if (!(bs->open_flags & BDRV_O_UNMAP)) { |
3517 | flags &= ~BDRV_REQ_MAY_UNMAP; | |
3518 | } | |
3519 | ||
f08f2dda | 3520 | return bdrv_co_do_writev(bs, sector_num, nb_sectors, NULL, |
aa7bfbff | 3521 | BDRV_REQ_ZERO_WRITE | flags); |
c5fbe571 SH |
3522 | } |
3523 | ||
83f64091 FB |
3524 | /** |
3525 | * Truncate file to 'offset' bytes (needed only for file protocols) | |
3526 | */ | |
3527 | int bdrv_truncate(BlockDriverState *bs, int64_t offset) | |
3528 | { | |
3529 | BlockDriver *drv = bs->drv; | |
51762288 | 3530 | int ret; |
83f64091 | 3531 | if (!drv) |
19cb3738 | 3532 | return -ENOMEDIUM; |
83f64091 FB |
3533 | if (!drv->bdrv_truncate) |
3534 | return -ENOTSUP; | |
59f2689d NS |
3535 | if (bs->read_only) |
3536 | return -EACCES; | |
9c75e168 | 3537 | |
51762288 SH |
3538 | ret = drv->bdrv_truncate(bs, offset); |
3539 | if (ret == 0) { | |
3540 | ret = refresh_total_sectors(bs, offset >> BDRV_SECTOR_BITS); | |
a7f53e26 MA |
3541 | if (bs->blk) { |
3542 | blk_dev_resize_cb(bs->blk); | |
3543 | } | |
51762288 SH |
3544 | } |
3545 | return ret; | |
83f64091 FB |
3546 | } |
3547 | ||
4a1d5e1f FZ |
3548 | /** |
3549 | * Length of a allocated file in bytes. Sparse files are counted by actual | |
3550 | * allocated space. Return < 0 if error or unknown. | |
3551 | */ | |
3552 | int64_t bdrv_get_allocated_file_size(BlockDriverState *bs) | |
3553 | { | |
3554 | BlockDriver *drv = bs->drv; | |
3555 | if (!drv) { | |
3556 | return -ENOMEDIUM; | |
3557 | } | |
3558 | if (drv->bdrv_get_allocated_file_size) { | |
3559 | return drv->bdrv_get_allocated_file_size(bs); | |
3560 | } | |
3561 | if (bs->file) { | |
3562 | return bdrv_get_allocated_file_size(bs->file); | |
3563 | } | |
3564 | return -ENOTSUP; | |
3565 | } | |
3566 | ||
83f64091 | 3567 | /** |
65a9bb25 | 3568 | * Return number of sectors on success, -errno on error. |
83f64091 | 3569 | */ |
65a9bb25 | 3570 | int64_t bdrv_nb_sectors(BlockDriverState *bs) |
83f64091 FB |
3571 | { |
3572 | BlockDriver *drv = bs->drv; | |
65a9bb25 | 3573 | |
83f64091 | 3574 | if (!drv) |
19cb3738 | 3575 | return -ENOMEDIUM; |
51762288 | 3576 | |
b94a2610 KW |
3577 | if (drv->has_variable_length) { |
3578 | int ret = refresh_total_sectors(bs, bs->total_sectors); | |
3579 | if (ret < 0) { | |
3580 | return ret; | |
46a4e4e6 | 3581 | } |
83f64091 | 3582 | } |
65a9bb25 MA |
3583 | return bs->total_sectors; |
3584 | } | |
3585 | ||
3586 | /** | |
3587 | * Return length in bytes on success, -errno on error. | |
3588 | * The length is always a multiple of BDRV_SECTOR_SIZE. | |
3589 | */ | |
3590 | int64_t bdrv_getlength(BlockDriverState *bs) | |
3591 | { | |
3592 | int64_t ret = bdrv_nb_sectors(bs); | |
3593 | ||
3594 | return ret < 0 ? ret : ret * BDRV_SECTOR_SIZE; | |
fc01f7e7 FB |
3595 | } |
3596 | ||
19cb3738 | 3597 | /* return 0 as number of sectors if no device present or error */ |
96b8f136 | 3598 | void bdrv_get_geometry(BlockDriverState *bs, uint64_t *nb_sectors_ptr) |
fc01f7e7 | 3599 | { |
65a9bb25 MA |
3600 | int64_t nb_sectors = bdrv_nb_sectors(bs); |
3601 | ||
3602 | *nb_sectors_ptr = nb_sectors < 0 ? 0 : nb_sectors; | |
fc01f7e7 | 3603 | } |
cf98951b | 3604 | |
ff06f5f3 PB |
3605 | void bdrv_set_on_error(BlockDriverState *bs, BlockdevOnError on_read_error, |
3606 | BlockdevOnError on_write_error) | |
abd7f68d MA |
3607 | { |
3608 | bs->on_read_error = on_read_error; | |
3609 | bs->on_write_error = on_write_error; | |
3610 | } | |
3611 | ||
1ceee0d5 | 3612 | BlockdevOnError bdrv_get_on_error(BlockDriverState *bs, bool is_read) |
abd7f68d MA |
3613 | { |
3614 | return is_read ? bs->on_read_error : bs->on_write_error; | |
3615 | } | |
3616 | ||
3e1caa5f PB |
3617 | BlockErrorAction bdrv_get_error_action(BlockDriverState *bs, bool is_read, int error) |
3618 | { | |
3619 | BlockdevOnError on_err = is_read ? bs->on_read_error : bs->on_write_error; | |
3620 | ||
3621 | switch (on_err) { | |
3622 | case BLOCKDEV_ON_ERROR_ENOSPC: | |
a589569f WX |
3623 | return (error == ENOSPC) ? |
3624 | BLOCK_ERROR_ACTION_STOP : BLOCK_ERROR_ACTION_REPORT; | |
3e1caa5f | 3625 | case BLOCKDEV_ON_ERROR_STOP: |
a589569f | 3626 | return BLOCK_ERROR_ACTION_STOP; |
3e1caa5f | 3627 | case BLOCKDEV_ON_ERROR_REPORT: |
a589569f | 3628 | return BLOCK_ERROR_ACTION_REPORT; |
3e1caa5f | 3629 | case BLOCKDEV_ON_ERROR_IGNORE: |
a589569f | 3630 | return BLOCK_ERROR_ACTION_IGNORE; |
3e1caa5f PB |
3631 | default: |
3632 | abort(); | |
3633 | } | |
3634 | } | |
3635 | ||
c7c2ff0c LC |
3636 | static void send_qmp_error_event(BlockDriverState *bs, |
3637 | BlockErrorAction action, | |
3638 | bool is_read, int error) | |
3639 | { | |
573742a5 | 3640 | IoOperationType optype; |
c7c2ff0c | 3641 | |
573742a5 PM |
3642 | optype = is_read ? IO_OPERATION_TYPE_READ : IO_OPERATION_TYPE_WRITE; |
3643 | qapi_event_send_block_io_error(bdrv_get_device_name(bs), optype, action, | |
c7c2ff0c | 3644 | bdrv_iostatus_is_enabled(bs), |
624ff573 LC |
3645 | error == ENOSPC, strerror(error), |
3646 | &error_abort); | |
c7c2ff0c LC |
3647 | } |
3648 | ||
3e1caa5f PB |
3649 | /* This is done by device models because, while the block layer knows |
3650 | * about the error, it does not know whether an operation comes from | |
3651 | * the device or the block layer (from a job, for example). | |
3652 | */ | |
3653 | void bdrv_error_action(BlockDriverState *bs, BlockErrorAction action, | |
3654 | bool is_read, int error) | |
3655 | { | |
3656 | assert(error >= 0); | |
2bd3bce8 | 3657 | |
a589569f | 3658 | if (action == BLOCK_ERROR_ACTION_STOP) { |
2bd3bce8 PB |
3659 | /* First set the iostatus, so that "info block" returns an iostatus |
3660 | * that matches the events raised so far (an additional error iostatus | |
3661 | * is fine, but not a lost one). | |
3662 | */ | |
3e1caa5f | 3663 | bdrv_iostatus_set_err(bs, error); |
2bd3bce8 PB |
3664 | |
3665 | /* Then raise the request to stop the VM and the event. | |
3666 | * qemu_system_vmstop_request_prepare has two effects. First, | |
3667 | * it ensures that the STOP event always comes after the | |
3668 | * BLOCK_IO_ERROR event. Second, it ensures that even if management | |
3669 | * can observe the STOP event and do a "cont" before the STOP | |
3670 | * event is issued, the VM will not stop. In this case, vm_start() | |
3671 | * also ensures that the STOP/RESUME pair of events is emitted. | |
3672 | */ | |
3673 | qemu_system_vmstop_request_prepare(); | |
c7c2ff0c | 3674 | send_qmp_error_event(bs, action, is_read, error); |
2bd3bce8 PB |
3675 | qemu_system_vmstop_request(RUN_STATE_IO_ERROR); |
3676 | } else { | |
c7c2ff0c | 3677 | send_qmp_error_event(bs, action, is_read, error); |
3e1caa5f PB |
3678 | } |
3679 | } | |
3680 | ||
b338082b FB |
3681 | int bdrv_is_read_only(BlockDriverState *bs) |
3682 | { | |
3683 | return bs->read_only; | |
3684 | } | |
3685 | ||
985a03b0 TS |
3686 | int bdrv_is_sg(BlockDriverState *bs) |
3687 | { | |
3688 | return bs->sg; | |
3689 | } | |
3690 | ||
e900a7b7 CH |
3691 | int bdrv_enable_write_cache(BlockDriverState *bs) |
3692 | { | |
3693 | return bs->enable_write_cache; | |
3694 | } | |
3695 | ||
425b0148 PB |
3696 | void bdrv_set_enable_write_cache(BlockDriverState *bs, bool wce) |
3697 | { | |
3698 | bs->enable_write_cache = wce; | |
55b110f2 JC |
3699 | |
3700 | /* so a reopen() will preserve wce */ | |
3701 | if (wce) { | |
3702 | bs->open_flags |= BDRV_O_CACHE_WB; | |
3703 | } else { | |
3704 | bs->open_flags &= ~BDRV_O_CACHE_WB; | |
3705 | } | |
425b0148 PB |
3706 | } |
3707 | ||
ea2384d3 FB |
3708 | int bdrv_is_encrypted(BlockDriverState *bs) |
3709 | { | |
3710 | if (bs->backing_hd && bs->backing_hd->encrypted) | |
3711 | return 1; | |
3712 | return bs->encrypted; | |
3713 | } | |
3714 | ||
c0f4ce77 AL |
3715 | int bdrv_key_required(BlockDriverState *bs) |
3716 | { | |
3717 | BlockDriverState *backing_hd = bs->backing_hd; | |
3718 | ||
3719 | if (backing_hd && backing_hd->encrypted && !backing_hd->valid_key) | |
3720 | return 1; | |
3721 | return (bs->encrypted && !bs->valid_key); | |
3722 | } | |
3723 | ||
ea2384d3 FB |
3724 | int bdrv_set_key(BlockDriverState *bs, const char *key) |
3725 | { | |
3726 | int ret; | |
3727 | if (bs->backing_hd && bs->backing_hd->encrypted) { | |
3728 | ret = bdrv_set_key(bs->backing_hd, key); | |
3729 | if (ret < 0) | |
3730 | return ret; | |
3731 | if (!bs->encrypted) | |
3732 | return 0; | |
3733 | } | |
fd04a2ae SH |
3734 | if (!bs->encrypted) { |
3735 | return -EINVAL; | |
3736 | } else if (!bs->drv || !bs->drv->bdrv_set_key) { | |
3737 | return -ENOMEDIUM; | |
3738 | } | |
c0f4ce77 | 3739 | ret = bs->drv->bdrv_set_key(bs, key); |
bb5fc20f AL |
3740 | if (ret < 0) { |
3741 | bs->valid_key = 0; | |
3742 | } else if (!bs->valid_key) { | |
3743 | bs->valid_key = 1; | |
a7f53e26 MA |
3744 | if (bs->blk) { |
3745 | /* call the change callback now, we skipped it on open */ | |
3746 | blk_dev_change_media_cb(bs->blk, true); | |
3747 | } | |
bb5fc20f | 3748 | } |
c0f4ce77 | 3749 | return ret; |
ea2384d3 FB |
3750 | } |
3751 | ||
4d2855a3 MA |
3752 | /* |
3753 | * Provide an encryption key for @bs. | |
3754 | * If @key is non-null: | |
3755 | * If @bs is not encrypted, fail. | |
3756 | * Else if the key is invalid, fail. | |
3757 | * Else set @bs's key to @key, replacing the existing key, if any. | |
3758 | * If @key is null: | |
3759 | * If @bs is encrypted and still lacks a key, fail. | |
3760 | * Else do nothing. | |
3761 | * On failure, store an error object through @errp if non-null. | |
3762 | */ | |
3763 | void bdrv_add_key(BlockDriverState *bs, const char *key, Error **errp) | |
3764 | { | |
3765 | if (key) { | |
3766 | if (!bdrv_is_encrypted(bs)) { | |
b1ca6391 | 3767 | error_setg(errp, "Device '%s' is not encrypted", |
4d2855a3 MA |
3768 | bdrv_get_device_name(bs)); |
3769 | } else if (bdrv_set_key(bs, key) < 0) { | |
3770 | error_set(errp, QERR_INVALID_PASSWORD); | |
3771 | } | |
3772 | } else { | |
3773 | if (bdrv_key_required(bs)) { | |
b1ca6391 MA |
3774 | error_set(errp, ERROR_CLASS_DEVICE_ENCRYPTED, |
3775 | "'%s' (%s) is encrypted", | |
4d2855a3 MA |
3776 | bdrv_get_device_name(bs), |
3777 | bdrv_get_encrypted_filename(bs)); | |
3778 | } | |
3779 | } | |
3780 | } | |
3781 | ||
f8d6bba1 | 3782 | const char *bdrv_get_format_name(BlockDriverState *bs) |
ea2384d3 | 3783 | { |
f8d6bba1 | 3784 | return bs->drv ? bs->drv->format_name : NULL; |
ea2384d3 FB |
3785 | } |
3786 | ||
ada42401 SH |
3787 | static int qsort_strcmp(const void *a, const void *b) |
3788 | { | |
3789 | return strcmp(a, b); | |
3790 | } | |
3791 | ||
5fafdf24 | 3792 | void bdrv_iterate_format(void (*it)(void *opaque, const char *name), |
ea2384d3 FB |
3793 | void *opaque) |
3794 | { | |
3795 | BlockDriver *drv; | |
e855e4fb | 3796 | int count = 0; |
ada42401 | 3797 | int i; |
e855e4fb | 3798 | const char **formats = NULL; |
ea2384d3 | 3799 | |
8a22f02a | 3800 | QLIST_FOREACH(drv, &bdrv_drivers, list) { |
e855e4fb JC |
3801 | if (drv->format_name) { |
3802 | bool found = false; | |
3803 | int i = count; | |
3804 | while (formats && i && !found) { | |
3805 | found = !strcmp(formats[--i], drv->format_name); | |
3806 | } | |
3807 | ||
3808 | if (!found) { | |
5839e53b | 3809 | formats = g_renew(const char *, formats, count + 1); |
e855e4fb | 3810 | formats[count++] = drv->format_name; |
e855e4fb JC |
3811 | } |
3812 | } | |
ea2384d3 | 3813 | } |
ada42401 SH |
3814 | |
3815 | qsort(formats, count, sizeof(formats[0]), qsort_strcmp); | |
3816 | ||
3817 | for (i = 0; i < count; i++) { | |
3818 | it(opaque, formats[i]); | |
3819 | } | |
3820 | ||
e855e4fb | 3821 | g_free(formats); |
ea2384d3 FB |
3822 | } |
3823 | ||
dc364f4c BC |
3824 | /* This function is to find a node in the bs graph */ |
3825 | BlockDriverState *bdrv_find_node(const char *node_name) | |
3826 | { | |
3827 | BlockDriverState *bs; | |
3828 | ||
3829 | assert(node_name); | |
3830 | ||
3831 | QTAILQ_FOREACH(bs, &graph_bdrv_states, node_list) { | |
3832 | if (!strcmp(node_name, bs->node_name)) { | |
3833 | return bs; | |
3834 | } | |
3835 | } | |
3836 | return NULL; | |
3837 | } | |
3838 | ||
c13163fb BC |
3839 | /* Put this QMP function here so it can access the static graph_bdrv_states. */ |
3840 | BlockDeviceInfoList *bdrv_named_nodes_list(void) | |
3841 | { | |
3842 | BlockDeviceInfoList *list, *entry; | |
3843 | BlockDriverState *bs; | |
3844 | ||
3845 | list = NULL; | |
3846 | QTAILQ_FOREACH(bs, &graph_bdrv_states, node_list) { | |
3847 | entry = g_malloc0(sizeof(*entry)); | |
3848 | entry->value = bdrv_block_device_info(bs); | |
3849 | entry->next = list; | |
3850 | list = entry; | |
3851 | } | |
3852 | ||
3853 | return list; | |
3854 | } | |
3855 | ||
12d3ba82 BC |
3856 | BlockDriverState *bdrv_lookup_bs(const char *device, |
3857 | const char *node_name, | |
3858 | Error **errp) | |
3859 | { | |
7f06d47e MA |
3860 | BlockBackend *blk; |
3861 | BlockDriverState *bs; | |
12d3ba82 | 3862 | |
12d3ba82 | 3863 | if (device) { |
7f06d47e | 3864 | blk = blk_by_name(device); |
12d3ba82 | 3865 | |
7f06d47e MA |
3866 | if (blk) { |
3867 | return blk_bs(blk); | |
12d3ba82 | 3868 | } |
12d3ba82 BC |
3869 | } |
3870 | ||
dd67fa50 BC |
3871 | if (node_name) { |
3872 | bs = bdrv_find_node(node_name); | |
12d3ba82 | 3873 | |
dd67fa50 BC |
3874 | if (bs) { |
3875 | return bs; | |
3876 | } | |
12d3ba82 BC |
3877 | } |
3878 | ||
dd67fa50 BC |
3879 | error_setg(errp, "Cannot find device=%s nor node_name=%s", |
3880 | device ? device : "", | |
3881 | node_name ? node_name : ""); | |
3882 | return NULL; | |
12d3ba82 BC |
3883 | } |
3884 | ||
5a6684d2 JC |
3885 | /* If 'base' is in the same chain as 'top', return true. Otherwise, |
3886 | * return false. If either argument is NULL, return false. */ | |
3887 | bool bdrv_chain_contains(BlockDriverState *top, BlockDriverState *base) | |
3888 | { | |
3889 | while (top && top != base) { | |
3890 | top = top->backing_hd; | |
3891 | } | |
3892 | ||
3893 | return top != NULL; | |
3894 | } | |
3895 | ||
04df765a FZ |
3896 | BlockDriverState *bdrv_next_node(BlockDriverState *bs) |
3897 | { | |
3898 | if (!bs) { | |
3899 | return QTAILQ_FIRST(&graph_bdrv_states); | |
3900 | } | |
3901 | return QTAILQ_NEXT(bs, node_list); | |
3902 | } | |
3903 | ||
2f399b0a MA |
3904 | BlockDriverState *bdrv_next(BlockDriverState *bs) |
3905 | { | |
3906 | if (!bs) { | |
3907 | return QTAILQ_FIRST(&bdrv_states); | |
3908 | } | |
dc364f4c | 3909 | return QTAILQ_NEXT(bs, device_list); |
2f399b0a MA |
3910 | } |
3911 | ||
20a9e77d FZ |
3912 | const char *bdrv_get_node_name(const BlockDriverState *bs) |
3913 | { | |
3914 | return bs->node_name; | |
3915 | } | |
3916 | ||
7f06d47e | 3917 | /* TODO check what callers really want: bs->node_name or blk_name() */ |
bfb197e0 | 3918 | const char *bdrv_get_device_name(const BlockDriverState *bs) |
ea2384d3 | 3919 | { |
bfb197e0 | 3920 | return bs->blk ? blk_name(bs->blk) : ""; |
ea2384d3 FB |
3921 | } |
3922 | ||
c8433287 MA |
3923 | int bdrv_get_flags(BlockDriverState *bs) |
3924 | { | |
3925 | return bs->open_flags; | |
3926 | } | |
3927 | ||
f0f0fdfe | 3928 | int bdrv_flush_all(void) |
c6ca28d6 AL |
3929 | { |
3930 | BlockDriverState *bs; | |
f0f0fdfe | 3931 | int result = 0; |
c6ca28d6 | 3932 | |
dc364f4c | 3933 | QTAILQ_FOREACH(bs, &bdrv_states, device_list) { |
ed78cda3 SH |
3934 | AioContext *aio_context = bdrv_get_aio_context(bs); |
3935 | int ret; | |
3936 | ||
3937 | aio_context_acquire(aio_context); | |
3938 | ret = bdrv_flush(bs); | |
f0f0fdfe KW |
3939 | if (ret < 0 && !result) { |
3940 | result = ret; | |
3941 | } | |
ed78cda3 | 3942 | aio_context_release(aio_context); |
1b7bdbc1 | 3943 | } |
f0f0fdfe KW |
3944 | |
3945 | return result; | |
c6ca28d6 AL |
3946 | } |
3947 | ||
3ac21627 PL |
3948 | int bdrv_has_zero_init_1(BlockDriverState *bs) |
3949 | { | |
3950 | return 1; | |
3951 | } | |
3952 | ||
f2feebbd KW |
3953 | int bdrv_has_zero_init(BlockDriverState *bs) |
3954 | { | |
3955 | assert(bs->drv); | |
3956 | ||
11212d8f PB |
3957 | /* If BS is a copy on write image, it is initialized to |
3958 | the contents of the base image, which may not be zeroes. */ | |
3959 | if (bs->backing_hd) { | |
3960 | return 0; | |
3961 | } | |
336c1c12 KW |
3962 | if (bs->drv->bdrv_has_zero_init) { |
3963 | return bs->drv->bdrv_has_zero_init(bs); | |
f2feebbd KW |
3964 | } |
3965 | ||
3ac21627 PL |
3966 | /* safe default */ |
3967 | return 0; | |
f2feebbd KW |
3968 | } |
3969 | ||
4ce78691 PL |
3970 | bool bdrv_unallocated_blocks_are_zero(BlockDriverState *bs) |
3971 | { | |
3972 | BlockDriverInfo bdi; | |
3973 | ||
3974 | if (bs->backing_hd) { | |
3975 | return false; | |
3976 | } | |
3977 | ||
3978 | if (bdrv_get_info(bs, &bdi) == 0) { | |
3979 | return bdi.unallocated_blocks_are_zero; | |
3980 | } | |
3981 | ||
3982 | return false; | |
3983 | } | |
3984 | ||
3985 | bool bdrv_can_write_zeroes_with_unmap(BlockDriverState *bs) | |
3986 | { | |
3987 | BlockDriverInfo bdi; | |
3988 | ||
3989 | if (bs->backing_hd || !(bs->open_flags & BDRV_O_UNMAP)) { | |
3990 | return false; | |
3991 | } | |
3992 | ||
3993 | if (bdrv_get_info(bs, &bdi) == 0) { | |
3994 | return bdi.can_write_zeroes_with_unmap; | |
3995 | } | |
3996 | ||
3997 | return false; | |
3998 | } | |
3999 | ||
b6b8a333 | 4000 | typedef struct BdrvCoGetBlockStatusData { |
376ae3f1 | 4001 | BlockDriverState *bs; |
b35b2bba | 4002 | BlockDriverState *base; |
376ae3f1 SH |
4003 | int64_t sector_num; |
4004 | int nb_sectors; | |
4005 | int *pnum; | |
b6b8a333 | 4006 | int64_t ret; |
376ae3f1 | 4007 | bool done; |
b6b8a333 | 4008 | } BdrvCoGetBlockStatusData; |
376ae3f1 | 4009 | |
f58c7b35 | 4010 | /* |
705be728 FZ |
4011 | * Returns the allocation status of the specified sectors. |
4012 | * Drivers not implementing the functionality are assumed to not support | |
4013 | * backing files, hence all their sectors are reported as allocated. | |
f58c7b35 | 4014 | * |
bd9533e3 SH |
4015 | * If 'sector_num' is beyond the end of the disk image the return value is 0 |
4016 | * and 'pnum' is set to 0. | |
4017 | * | |
f58c7b35 TS |
4018 | * 'pnum' is set to the number of sectors (including and immediately following |
4019 | * the specified sector) that are known to be in the same | |
4020 | * allocated/unallocated state. | |
4021 | * | |
bd9533e3 SH |
4022 | * 'nb_sectors' is the max value 'pnum' should be set to. If nb_sectors goes |
4023 | * beyond the end of the disk image it will be clamped. | |
f58c7b35 | 4024 | */ |
b6b8a333 PB |
4025 | static int64_t coroutine_fn bdrv_co_get_block_status(BlockDriverState *bs, |
4026 | int64_t sector_num, | |
4027 | int nb_sectors, int *pnum) | |
f58c7b35 | 4028 | { |
30a7f2fc | 4029 | int64_t total_sectors; |
bd9533e3 | 4030 | int64_t n; |
5daa74a6 | 4031 | int64_t ret, ret2; |
bd9533e3 | 4032 | |
30a7f2fc MA |
4033 | total_sectors = bdrv_nb_sectors(bs); |
4034 | if (total_sectors < 0) { | |
4035 | return total_sectors; | |
617ccb46 PB |
4036 | } |
4037 | ||
30a7f2fc | 4038 | if (sector_num >= total_sectors) { |
bd9533e3 SH |
4039 | *pnum = 0; |
4040 | return 0; | |
4041 | } | |
4042 | ||
30a7f2fc | 4043 | n = total_sectors - sector_num; |
bd9533e3 SH |
4044 | if (n < nb_sectors) { |
4045 | nb_sectors = n; | |
4046 | } | |
4047 | ||
b6b8a333 | 4048 | if (!bs->drv->bdrv_co_get_block_status) { |
bd9533e3 | 4049 | *pnum = nb_sectors; |
e88ae226 | 4050 | ret = BDRV_BLOCK_DATA | BDRV_BLOCK_ALLOCATED; |
918e92d7 PB |
4051 | if (bs->drv->protocol_name) { |
4052 | ret |= BDRV_BLOCK_OFFSET_VALID | (sector_num * BDRV_SECTOR_SIZE); | |
4053 | } | |
4054 | return ret; | |
f58c7b35 | 4055 | } |
6aebab14 | 4056 | |
415b5b01 PB |
4057 | ret = bs->drv->bdrv_co_get_block_status(bs, sector_num, nb_sectors, pnum); |
4058 | if (ret < 0) { | |
3e0a233d | 4059 | *pnum = 0; |
415b5b01 PB |
4060 | return ret; |
4061 | } | |
4062 | ||
92bc50a5 PL |
4063 | if (ret & BDRV_BLOCK_RAW) { |
4064 | assert(ret & BDRV_BLOCK_OFFSET_VALID); | |
4065 | return bdrv_get_block_status(bs->file, ret >> BDRV_SECTOR_BITS, | |
4066 | *pnum, pnum); | |
4067 | } | |
4068 | ||
e88ae226 KW |
4069 | if (ret & (BDRV_BLOCK_DATA | BDRV_BLOCK_ZERO)) { |
4070 | ret |= BDRV_BLOCK_ALLOCATED; | |
4071 | } | |
4072 | ||
c3d86884 PL |
4073 | if (!(ret & BDRV_BLOCK_DATA) && !(ret & BDRV_BLOCK_ZERO)) { |
4074 | if (bdrv_unallocated_blocks_are_zero(bs)) { | |
f0ad5712 | 4075 | ret |= BDRV_BLOCK_ZERO; |
1f9db224 | 4076 | } else if (bs->backing_hd) { |
f0ad5712 | 4077 | BlockDriverState *bs2 = bs->backing_hd; |
30a7f2fc MA |
4078 | int64_t nb_sectors2 = bdrv_nb_sectors(bs2); |
4079 | if (nb_sectors2 >= 0 && sector_num >= nb_sectors2) { | |
f0ad5712 PB |
4080 | ret |= BDRV_BLOCK_ZERO; |
4081 | } | |
4082 | } | |
415b5b01 | 4083 | } |
5daa74a6 PB |
4084 | |
4085 | if (bs->file && | |
4086 | (ret & BDRV_BLOCK_DATA) && !(ret & BDRV_BLOCK_ZERO) && | |
4087 | (ret & BDRV_BLOCK_OFFSET_VALID)) { | |
59c9a95f HR |
4088 | int file_pnum; |
4089 | ||
5daa74a6 | 4090 | ret2 = bdrv_co_get_block_status(bs->file, ret >> BDRV_SECTOR_BITS, |
59c9a95f | 4091 | *pnum, &file_pnum); |
5daa74a6 PB |
4092 | if (ret2 >= 0) { |
4093 | /* Ignore errors. This is just providing extra information, it | |
4094 | * is useful but not necessary. | |
4095 | */ | |
59c9a95f HR |
4096 | if (!file_pnum) { |
4097 | /* !file_pnum indicates an offset at or beyond the EOF; it is | |
4098 | * perfectly valid for the format block driver to point to such | |
4099 | * offsets, so catch it and mark everything as zero */ | |
4100 | ret |= BDRV_BLOCK_ZERO; | |
4101 | } else { | |
4102 | /* Limit request to the range reported by the protocol driver */ | |
4103 | *pnum = file_pnum; | |
4104 | ret |= (ret2 & BDRV_BLOCK_ZERO); | |
4105 | } | |
5daa74a6 PB |
4106 | } |
4107 | } | |
4108 | ||
415b5b01 | 4109 | return ret; |
060f51c9 SH |
4110 | } |
4111 | ||
b6b8a333 PB |
4112 | /* Coroutine wrapper for bdrv_get_block_status() */ |
4113 | static void coroutine_fn bdrv_get_block_status_co_entry(void *opaque) | |
060f51c9 | 4114 | { |
b6b8a333 | 4115 | BdrvCoGetBlockStatusData *data = opaque; |
060f51c9 SH |
4116 | BlockDriverState *bs = data->bs; |
4117 | ||
b6b8a333 PB |
4118 | data->ret = bdrv_co_get_block_status(bs, data->sector_num, data->nb_sectors, |
4119 | data->pnum); | |
060f51c9 SH |
4120 | data->done = true; |
4121 | } | |
4122 | ||
4123 | /* | |
b6b8a333 | 4124 | * Synchronous wrapper around bdrv_co_get_block_status(). |
060f51c9 | 4125 | * |
b6b8a333 | 4126 | * See bdrv_co_get_block_status() for details. |
060f51c9 | 4127 | */ |
b6b8a333 PB |
4128 | int64_t bdrv_get_block_status(BlockDriverState *bs, int64_t sector_num, |
4129 | int nb_sectors, int *pnum) | |
060f51c9 | 4130 | { |
6aebab14 | 4131 | Coroutine *co; |
b6b8a333 | 4132 | BdrvCoGetBlockStatusData data = { |
6aebab14 SH |
4133 | .bs = bs, |
4134 | .sector_num = sector_num, | |
4135 | .nb_sectors = nb_sectors, | |
4136 | .pnum = pnum, | |
4137 | .done = false, | |
4138 | }; | |
4139 | ||
bdad13b9 PB |
4140 | if (qemu_in_coroutine()) { |
4141 | /* Fast-path if already in coroutine context */ | |
b6b8a333 | 4142 | bdrv_get_block_status_co_entry(&data); |
bdad13b9 | 4143 | } else { |
2572b37a SH |
4144 | AioContext *aio_context = bdrv_get_aio_context(bs); |
4145 | ||
b6b8a333 | 4146 | co = qemu_coroutine_create(bdrv_get_block_status_co_entry); |
bdad13b9 PB |
4147 | qemu_coroutine_enter(co, &data); |
4148 | while (!data.done) { | |
2572b37a | 4149 | aio_poll(aio_context, true); |
bdad13b9 | 4150 | } |
6aebab14 SH |
4151 | } |
4152 | return data.ret; | |
f58c7b35 TS |
4153 | } |
4154 | ||
b6b8a333 PB |
4155 | int coroutine_fn bdrv_is_allocated(BlockDriverState *bs, int64_t sector_num, |
4156 | int nb_sectors, int *pnum) | |
4157 | { | |
4333bb71 PB |
4158 | int64_t ret = bdrv_get_block_status(bs, sector_num, nb_sectors, pnum); |
4159 | if (ret < 0) { | |
4160 | return ret; | |
4161 | } | |
01fb2705 | 4162 | return !!(ret & BDRV_BLOCK_ALLOCATED); |
b6b8a333 PB |
4163 | } |
4164 | ||
188a7bbf PB |
4165 | /* |
4166 | * Given an image chain: ... -> [BASE] -> [INTER1] -> [INTER2] -> [TOP] | |
4167 | * | |
4168 | * Return true if the given sector is allocated in any image between | |
4169 | * BASE and TOP (inclusive). BASE can be NULL to check if the given | |
4170 | * sector is allocated in any image of the chain. Return false otherwise. | |
4171 | * | |
4172 | * 'pnum' is set to the number of sectors (including and immediately following | |
4173 | * the specified sector) that are known to be in the same | |
4174 | * allocated/unallocated state. | |
4175 | * | |
4176 | */ | |
4f578637 PB |
4177 | int bdrv_is_allocated_above(BlockDriverState *top, |
4178 | BlockDriverState *base, | |
4179 | int64_t sector_num, | |
4180 | int nb_sectors, int *pnum) | |
188a7bbf PB |
4181 | { |
4182 | BlockDriverState *intermediate; | |
4183 | int ret, n = nb_sectors; | |
4184 | ||
4185 | intermediate = top; | |
4186 | while (intermediate && intermediate != base) { | |
4187 | int pnum_inter; | |
bdad13b9 PB |
4188 | ret = bdrv_is_allocated(intermediate, sector_num, nb_sectors, |
4189 | &pnum_inter); | |
188a7bbf PB |
4190 | if (ret < 0) { |
4191 | return ret; | |
4192 | } else if (ret) { | |
4193 | *pnum = pnum_inter; | |
4194 | return 1; | |
4195 | } | |
4196 | ||
4197 | /* | |
4198 | * [sector_num, nb_sectors] is unallocated on top but intermediate | |
4199 | * might have | |
4200 | * | |
4201 | * [sector_num+x, nr_sectors] allocated. | |
4202 | */ | |
63ba17d3 VI |
4203 | if (n > pnum_inter && |
4204 | (intermediate == top || | |
4205 | sector_num + pnum_inter < intermediate->total_sectors)) { | |
188a7bbf PB |
4206 | n = pnum_inter; |
4207 | } | |
4208 | ||
4209 | intermediate = intermediate->backing_hd; | |
4210 | } | |
4211 | ||
4212 | *pnum = n; | |
4213 | return 0; | |
4214 | } | |
4215 | ||
045df330 AL |
4216 | const char *bdrv_get_encrypted_filename(BlockDriverState *bs) |
4217 | { | |
4218 | if (bs->backing_hd && bs->backing_hd->encrypted) | |
4219 | return bs->backing_file; | |
4220 | else if (bs->encrypted) | |
4221 | return bs->filename; | |
4222 | else | |
4223 | return NULL; | |
4224 | } | |
4225 | ||
5fafdf24 | 4226 | void bdrv_get_backing_filename(BlockDriverState *bs, |
83f64091 FB |
4227 | char *filename, int filename_size) |
4228 | { | |
3574c608 | 4229 | pstrcpy(filename, filename_size, bs->backing_file); |
83f64091 FB |
4230 | } |
4231 | ||
5fafdf24 | 4232 | int bdrv_write_compressed(BlockDriverState *bs, int64_t sector_num, |
faea38e7 FB |
4233 | const uint8_t *buf, int nb_sectors) |
4234 | { | |
4235 | BlockDriver *drv = bs->drv; | |
b9c64947 HR |
4236 | int ret; |
4237 | ||
4238 | if (!drv) { | |
19cb3738 | 4239 | return -ENOMEDIUM; |
b9c64947 HR |
4240 | } |
4241 | if (!drv->bdrv_write_compressed) { | |
faea38e7 | 4242 | return -ENOTSUP; |
b9c64947 HR |
4243 | } |
4244 | ret = bdrv_check_request(bs, sector_num, nb_sectors); | |
4245 | if (ret < 0) { | |
4246 | return ret; | |
4247 | } | |
a55eb92c | 4248 | |
e4654d2d | 4249 | assert(QLIST_EMPTY(&bs->dirty_bitmaps)); |
a55eb92c | 4250 | |
faea38e7 FB |
4251 | return drv->bdrv_write_compressed(bs, sector_num, buf, nb_sectors); |
4252 | } | |
3b46e624 | 4253 | |
faea38e7 FB |
4254 | int bdrv_get_info(BlockDriverState *bs, BlockDriverInfo *bdi) |
4255 | { | |
4256 | BlockDriver *drv = bs->drv; | |
4257 | if (!drv) | |
19cb3738 | 4258 | return -ENOMEDIUM; |
faea38e7 FB |
4259 | if (!drv->bdrv_get_info) |
4260 | return -ENOTSUP; | |
4261 | memset(bdi, 0, sizeof(*bdi)); | |
4262 | return drv->bdrv_get_info(bs, bdi); | |
4263 | } | |
4264 | ||
eae041fe HR |
4265 | ImageInfoSpecific *bdrv_get_specific_info(BlockDriverState *bs) |
4266 | { | |
4267 | BlockDriver *drv = bs->drv; | |
4268 | if (drv && drv->bdrv_get_specific_info) { | |
4269 | return drv->bdrv_get_specific_info(bs); | |
4270 | } | |
4271 | return NULL; | |
4272 | } | |
4273 | ||
45566e9c CH |
4274 | int bdrv_save_vmstate(BlockDriverState *bs, const uint8_t *buf, |
4275 | int64_t pos, int size) | |
cf8074b3 KW |
4276 | { |
4277 | QEMUIOVector qiov; | |
4278 | struct iovec iov = { | |
4279 | .iov_base = (void *) buf, | |
4280 | .iov_len = size, | |
4281 | }; | |
4282 | ||
4283 | qemu_iovec_init_external(&qiov, &iov, 1); | |
4284 | return bdrv_writev_vmstate(bs, &qiov, pos); | |
4285 | } | |
4286 | ||
4287 | int bdrv_writev_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos) | |
178e08a5 AL |
4288 | { |
4289 | BlockDriver *drv = bs->drv; | |
cf8074b3 KW |
4290 | |
4291 | if (!drv) { | |
178e08a5 | 4292 | return -ENOMEDIUM; |
cf8074b3 KW |
4293 | } else if (drv->bdrv_save_vmstate) { |
4294 | return drv->bdrv_save_vmstate(bs, qiov, pos); | |
4295 | } else if (bs->file) { | |
4296 | return bdrv_writev_vmstate(bs->file, qiov, pos); | |
4297 | } | |
4298 | ||
7cdb1f6d | 4299 | return -ENOTSUP; |
178e08a5 AL |
4300 | } |
4301 | ||
45566e9c CH |
4302 | int bdrv_load_vmstate(BlockDriverState *bs, uint8_t *buf, |
4303 | int64_t pos, int size) | |
178e08a5 AL |
4304 | { |
4305 | BlockDriver *drv = bs->drv; | |
4306 | if (!drv) | |
4307 | return -ENOMEDIUM; | |
7cdb1f6d MK |
4308 | if (drv->bdrv_load_vmstate) |
4309 | return drv->bdrv_load_vmstate(bs, buf, pos, size); | |
4310 | if (bs->file) | |
4311 | return bdrv_load_vmstate(bs->file, buf, pos, size); | |
4312 | return -ENOTSUP; | |
178e08a5 AL |
4313 | } |
4314 | ||
8b9b0cc2 KW |
4315 | void bdrv_debug_event(BlockDriverState *bs, BlkDebugEvent event) |
4316 | { | |
bf736fe3 | 4317 | if (!bs || !bs->drv || !bs->drv->bdrv_debug_event) { |
8b9b0cc2 KW |
4318 | return; |
4319 | } | |
4320 | ||
bf736fe3 | 4321 | bs->drv->bdrv_debug_event(bs, event); |
41c695c7 KW |
4322 | } |
4323 | ||
4324 | int bdrv_debug_breakpoint(BlockDriverState *bs, const char *event, | |
4325 | const char *tag) | |
4326 | { | |
4327 | while (bs && bs->drv && !bs->drv->bdrv_debug_breakpoint) { | |
4328 | bs = bs->file; | |
4329 | } | |
4330 | ||
4331 | if (bs && bs->drv && bs->drv->bdrv_debug_breakpoint) { | |
4332 | return bs->drv->bdrv_debug_breakpoint(bs, event, tag); | |
4333 | } | |
4334 | ||
4335 | return -ENOTSUP; | |
4336 | } | |
4337 | ||
4cc70e93 FZ |
4338 | int bdrv_debug_remove_breakpoint(BlockDriverState *bs, const char *tag) |
4339 | { | |
4340 | while (bs && bs->drv && !bs->drv->bdrv_debug_remove_breakpoint) { | |
4341 | bs = bs->file; | |
4342 | } | |
4343 | ||
4344 | if (bs && bs->drv && bs->drv->bdrv_debug_remove_breakpoint) { | |
4345 | return bs->drv->bdrv_debug_remove_breakpoint(bs, tag); | |
4346 | } | |
4347 | ||
4348 | return -ENOTSUP; | |
4349 | } | |
4350 | ||
41c695c7 KW |
4351 | int bdrv_debug_resume(BlockDriverState *bs, const char *tag) |
4352 | { | |
938789ea | 4353 | while (bs && (!bs->drv || !bs->drv->bdrv_debug_resume)) { |
41c695c7 KW |
4354 | bs = bs->file; |
4355 | } | |
8b9b0cc2 | 4356 | |
41c695c7 KW |
4357 | if (bs && bs->drv && bs->drv->bdrv_debug_resume) { |
4358 | return bs->drv->bdrv_debug_resume(bs, tag); | |
4359 | } | |
4360 | ||
4361 | return -ENOTSUP; | |
4362 | } | |
4363 | ||
4364 | bool bdrv_debug_is_suspended(BlockDriverState *bs, const char *tag) | |
4365 | { | |
4366 | while (bs && bs->drv && !bs->drv->bdrv_debug_is_suspended) { | |
4367 | bs = bs->file; | |
4368 | } | |
4369 | ||
4370 | if (bs && bs->drv && bs->drv->bdrv_debug_is_suspended) { | |
4371 | return bs->drv->bdrv_debug_is_suspended(bs, tag); | |
4372 | } | |
4373 | ||
4374 | return false; | |
8b9b0cc2 KW |
4375 | } |
4376 | ||
199630b6 BS |
4377 | int bdrv_is_snapshot(BlockDriverState *bs) |
4378 | { | |
4379 | return !!(bs->open_flags & BDRV_O_SNAPSHOT); | |
4380 | } | |
4381 | ||
b1b1d783 JC |
4382 | /* backing_file can either be relative, or absolute, or a protocol. If it is |
4383 | * relative, it must be relative to the chain. So, passing in bs->filename | |
4384 | * from a BDS as backing_file should not be done, as that may be relative to | |
4385 | * the CWD rather than the chain. */ | |
e8a6bb9c MT |
4386 | BlockDriverState *bdrv_find_backing_image(BlockDriverState *bs, |
4387 | const char *backing_file) | |
4388 | { | |
b1b1d783 JC |
4389 | char *filename_full = NULL; |
4390 | char *backing_file_full = NULL; | |
4391 | char *filename_tmp = NULL; | |
4392 | int is_protocol = 0; | |
4393 | BlockDriverState *curr_bs = NULL; | |
4394 | BlockDriverState *retval = NULL; | |
4395 | ||
4396 | if (!bs || !bs->drv || !backing_file) { | |
e8a6bb9c MT |
4397 | return NULL; |
4398 | } | |
4399 | ||
b1b1d783 JC |
4400 | filename_full = g_malloc(PATH_MAX); |
4401 | backing_file_full = g_malloc(PATH_MAX); | |
4402 | filename_tmp = g_malloc(PATH_MAX); | |
4403 | ||
4404 | is_protocol = path_has_protocol(backing_file); | |
4405 | ||
4406 | for (curr_bs = bs; curr_bs->backing_hd; curr_bs = curr_bs->backing_hd) { | |
4407 | ||
4408 | /* If either of the filename paths is actually a protocol, then | |
4409 | * compare unmodified paths; otherwise make paths relative */ | |
4410 | if (is_protocol || path_has_protocol(curr_bs->backing_file)) { | |
4411 | if (strcmp(backing_file, curr_bs->backing_file) == 0) { | |
4412 | retval = curr_bs->backing_hd; | |
4413 | break; | |
4414 | } | |
e8a6bb9c | 4415 | } else { |
b1b1d783 JC |
4416 | /* If not an absolute filename path, make it relative to the current |
4417 | * image's filename path */ | |
4418 | path_combine(filename_tmp, PATH_MAX, curr_bs->filename, | |
4419 | backing_file); | |
4420 | ||
4421 | /* We are going to compare absolute pathnames */ | |
4422 | if (!realpath(filename_tmp, filename_full)) { | |
4423 | continue; | |
4424 | } | |
4425 | ||
4426 | /* We need to make sure the backing filename we are comparing against | |
4427 | * is relative to the current image filename (or absolute) */ | |
4428 | path_combine(filename_tmp, PATH_MAX, curr_bs->filename, | |
4429 | curr_bs->backing_file); | |
4430 | ||
4431 | if (!realpath(filename_tmp, backing_file_full)) { | |
4432 | continue; | |
4433 | } | |
4434 | ||
4435 | if (strcmp(backing_file_full, filename_full) == 0) { | |
4436 | retval = curr_bs->backing_hd; | |
4437 | break; | |
4438 | } | |
e8a6bb9c MT |
4439 | } |
4440 | } | |
4441 | ||
b1b1d783 JC |
4442 | g_free(filename_full); |
4443 | g_free(backing_file_full); | |
4444 | g_free(filename_tmp); | |
4445 | return retval; | |
e8a6bb9c MT |
4446 | } |
4447 | ||
f198fd1c BC |
4448 | int bdrv_get_backing_file_depth(BlockDriverState *bs) |
4449 | { | |
4450 | if (!bs->drv) { | |
4451 | return 0; | |
4452 | } | |
4453 | ||
4454 | if (!bs->backing_hd) { | |
4455 | return 0; | |
4456 | } | |
4457 | ||
4458 | return 1 + bdrv_get_backing_file_depth(bs->backing_hd); | |
4459 | } | |
4460 | ||
ea2384d3 | 4461 | /**************************************************************/ |
83f64091 | 4462 | /* async I/Os */ |
ea2384d3 | 4463 | |
7c84b1b8 MA |
4464 | BlockAIOCB *bdrv_aio_readv(BlockDriverState *bs, int64_t sector_num, |
4465 | QEMUIOVector *qiov, int nb_sectors, | |
097310b5 | 4466 | BlockCompletionFunc *cb, void *opaque) |
83f64091 | 4467 | { |
bbf0a440 SH |
4468 | trace_bdrv_aio_readv(bs, sector_num, nb_sectors, opaque); |
4469 | ||
d20d9b7c | 4470 | return bdrv_co_aio_rw_vector(bs, sector_num, qiov, nb_sectors, 0, |
8c5873d6 | 4471 | cb, opaque, false); |
ea2384d3 FB |
4472 | } |
4473 | ||
7c84b1b8 MA |
4474 | BlockAIOCB *bdrv_aio_writev(BlockDriverState *bs, int64_t sector_num, |
4475 | QEMUIOVector *qiov, int nb_sectors, | |
097310b5 | 4476 | BlockCompletionFunc *cb, void *opaque) |
ea2384d3 | 4477 | { |
bbf0a440 SH |
4478 | trace_bdrv_aio_writev(bs, sector_num, nb_sectors, opaque); |
4479 | ||
d20d9b7c | 4480 | return bdrv_co_aio_rw_vector(bs, sector_num, qiov, nb_sectors, 0, |
8c5873d6 | 4481 | cb, opaque, true); |
83f64091 FB |
4482 | } |
4483 | ||
7c84b1b8 | 4484 | BlockAIOCB *bdrv_aio_write_zeroes(BlockDriverState *bs, |
d5ef94d4 | 4485 | int64_t sector_num, int nb_sectors, BdrvRequestFlags flags, |
097310b5 | 4486 | BlockCompletionFunc *cb, void *opaque) |
d5ef94d4 PB |
4487 | { |
4488 | trace_bdrv_aio_write_zeroes(bs, sector_num, nb_sectors, flags, opaque); | |
4489 | ||
4490 | return bdrv_co_aio_rw_vector(bs, sector_num, NULL, nb_sectors, | |
4491 | BDRV_REQ_ZERO_WRITE | flags, | |
4492 | cb, opaque, true); | |
4493 | } | |
4494 | ||
40b4f539 KW |
4495 | |
4496 | typedef struct MultiwriteCB { | |
4497 | int error; | |
4498 | int num_requests; | |
4499 | int num_callbacks; | |
4500 | struct { | |
097310b5 | 4501 | BlockCompletionFunc *cb; |
40b4f539 KW |
4502 | void *opaque; |
4503 | QEMUIOVector *free_qiov; | |
40b4f539 KW |
4504 | } callbacks[]; |
4505 | } MultiwriteCB; | |
4506 | ||
4507 | static void multiwrite_user_cb(MultiwriteCB *mcb) | |
4508 | { | |
4509 | int i; | |
4510 | ||
4511 | for (i = 0; i < mcb->num_callbacks; i++) { | |
4512 | mcb->callbacks[i].cb(mcb->callbacks[i].opaque, mcb->error); | |
1e1ea48d SH |
4513 | if (mcb->callbacks[i].free_qiov) { |
4514 | qemu_iovec_destroy(mcb->callbacks[i].free_qiov); | |
4515 | } | |
7267c094 | 4516 | g_free(mcb->callbacks[i].free_qiov); |
40b4f539 KW |
4517 | } |
4518 | } | |
4519 | ||
4520 | static void multiwrite_cb(void *opaque, int ret) | |
4521 | { | |
4522 | MultiwriteCB *mcb = opaque; | |
4523 | ||
6d519a5f SH |
4524 | trace_multiwrite_cb(mcb, ret); |
4525 | ||
cb6d3ca0 | 4526 | if (ret < 0 && !mcb->error) { |
40b4f539 | 4527 | mcb->error = ret; |
40b4f539 KW |
4528 | } |
4529 | ||
4530 | mcb->num_requests--; | |
4531 | if (mcb->num_requests == 0) { | |
de189a1b | 4532 | multiwrite_user_cb(mcb); |
7267c094 | 4533 | g_free(mcb); |
40b4f539 KW |
4534 | } |
4535 | } | |
4536 | ||
4537 | static int multiwrite_req_compare(const void *a, const void *b) | |
4538 | { | |
77be4366 CH |
4539 | const BlockRequest *req1 = a, *req2 = b; |
4540 | ||
4541 | /* | |
4542 | * Note that we can't simply subtract req2->sector from req1->sector | |
4543 | * here as that could overflow the return value. | |
4544 | */ | |
4545 | if (req1->sector > req2->sector) { | |
4546 | return 1; | |
4547 | } else if (req1->sector < req2->sector) { | |
4548 | return -1; | |
4549 | } else { | |
4550 | return 0; | |
4551 | } | |
40b4f539 KW |
4552 | } |
4553 | ||
4554 | /* | |
4555 | * Takes a bunch of requests and tries to merge them. Returns the number of | |
4556 | * requests that remain after merging. | |
4557 | */ | |
4558 | static int multiwrite_merge(BlockDriverState *bs, BlockRequest *reqs, | |
4559 | int num_reqs, MultiwriteCB *mcb) | |
4560 | { | |
4561 | int i, outidx; | |
4562 | ||
4563 | // Sort requests by start sector | |
4564 | qsort(reqs, num_reqs, sizeof(*reqs), &multiwrite_req_compare); | |
4565 | ||
4566 | // Check if adjacent requests touch the same clusters. If so, combine them, | |
4567 | // filling up gaps with zero sectors. | |
4568 | outidx = 0; | |
4569 | for (i = 1; i < num_reqs; i++) { | |
4570 | int merge = 0; | |
4571 | int64_t oldreq_last = reqs[outidx].sector + reqs[outidx].nb_sectors; | |
4572 | ||
b6a127a1 | 4573 | // Handle exactly sequential writes and overlapping writes. |
40b4f539 KW |
4574 | if (reqs[i].sector <= oldreq_last) { |
4575 | merge = 1; | |
4576 | } | |
4577 | ||
e2a305fb CH |
4578 | if (reqs[outidx].qiov->niov + reqs[i].qiov->niov + 1 > IOV_MAX) { |
4579 | merge = 0; | |
4580 | } | |
4581 | ||
6c5a42ac PL |
4582 | if (bs->bl.max_transfer_length && reqs[outidx].nb_sectors + |
4583 | reqs[i].nb_sectors > bs->bl.max_transfer_length) { | |
4584 | merge = 0; | |
4585 | } | |
4586 | ||
40b4f539 KW |
4587 | if (merge) { |
4588 | size_t size; | |
7267c094 | 4589 | QEMUIOVector *qiov = g_malloc0(sizeof(*qiov)); |
40b4f539 KW |
4590 | qemu_iovec_init(qiov, |
4591 | reqs[outidx].qiov->niov + reqs[i].qiov->niov + 1); | |
4592 | ||
4593 | // Add the first request to the merged one. If the requests are | |
4594 | // overlapping, drop the last sectors of the first request. | |
4595 | size = (reqs[i].sector - reqs[outidx].sector) << 9; | |
1b093c48 | 4596 | qemu_iovec_concat(qiov, reqs[outidx].qiov, 0, size); |
40b4f539 | 4597 | |
b6a127a1 PB |
4598 | // We should need to add any zeros between the two requests |
4599 | assert (reqs[i].sector <= oldreq_last); | |
40b4f539 KW |
4600 | |
4601 | // Add the second request | |
1b093c48 | 4602 | qemu_iovec_concat(qiov, reqs[i].qiov, 0, reqs[i].qiov->size); |
40b4f539 | 4603 | |
391827eb SH |
4604 | // Add tail of first request, if necessary |
4605 | if (qiov->size < reqs[outidx].qiov->size) { | |
4606 | qemu_iovec_concat(qiov, reqs[outidx].qiov, qiov->size, | |
4607 | reqs[outidx].qiov->size - qiov->size); | |
4608 | } | |
4609 | ||
cbf1dff2 | 4610 | reqs[outidx].nb_sectors = qiov->size >> 9; |
40b4f539 KW |
4611 | reqs[outidx].qiov = qiov; |
4612 | ||
4613 | mcb->callbacks[i].free_qiov = reqs[outidx].qiov; | |
4614 | } else { | |
4615 | outidx++; | |
4616 | reqs[outidx].sector = reqs[i].sector; | |
4617 | reqs[outidx].nb_sectors = reqs[i].nb_sectors; | |
4618 | reqs[outidx].qiov = reqs[i].qiov; | |
4619 | } | |
4620 | } | |
4621 | ||
f4564d53 PL |
4622 | block_acct_merge_done(&bs->stats, BLOCK_ACCT_WRITE, num_reqs - outidx - 1); |
4623 | ||
40b4f539 KW |
4624 | return outidx + 1; |
4625 | } | |
4626 | ||
4627 | /* | |
4628 | * Submit multiple AIO write requests at once. | |
4629 | * | |
4630 | * On success, the function returns 0 and all requests in the reqs array have | |
4631 | * been submitted. In error case this function returns -1, and any of the | |
4632 | * requests may or may not be submitted yet. In particular, this means that the | |
4633 | * callback will be called for some of the requests, for others it won't. The | |
4634 | * caller must check the error field of the BlockRequest to wait for the right | |
4635 | * callbacks (if error != 0, no callback will be called). | |
4636 | * | |
4637 | * The implementation may modify the contents of the reqs array, e.g. to merge | |
4638 | * requests. However, the fields opaque and error are left unmodified as they | |
4639 | * are used to signal failure for a single request to the caller. | |
4640 | */ | |
4641 | int bdrv_aio_multiwrite(BlockDriverState *bs, BlockRequest *reqs, int num_reqs) | |
4642 | { | |
40b4f539 KW |
4643 | MultiwriteCB *mcb; |
4644 | int i; | |
4645 | ||
301db7c2 RH |
4646 | /* don't submit writes if we don't have a medium */ |
4647 | if (bs->drv == NULL) { | |
4648 | for (i = 0; i < num_reqs; i++) { | |
4649 | reqs[i].error = -ENOMEDIUM; | |
4650 | } | |
4651 | return -1; | |
4652 | } | |
4653 | ||
40b4f539 KW |
4654 | if (num_reqs == 0) { |
4655 | return 0; | |
4656 | } | |
4657 | ||
4658 | // Create MultiwriteCB structure | |
7267c094 | 4659 | mcb = g_malloc0(sizeof(*mcb) + num_reqs * sizeof(*mcb->callbacks)); |
40b4f539 KW |
4660 | mcb->num_requests = 0; |
4661 | mcb->num_callbacks = num_reqs; | |
4662 | ||
4663 | for (i = 0; i < num_reqs; i++) { | |
4664 | mcb->callbacks[i].cb = reqs[i].cb; | |
4665 | mcb->callbacks[i].opaque = reqs[i].opaque; | |
4666 | } | |
4667 | ||
4668 | // Check for mergable requests | |
4669 | num_reqs = multiwrite_merge(bs, reqs, num_reqs, mcb); | |
4670 | ||
6d519a5f SH |
4671 | trace_bdrv_aio_multiwrite(mcb, mcb->num_callbacks, num_reqs); |
4672 | ||
df9309fb PB |
4673 | /* Run the aio requests. */ |
4674 | mcb->num_requests = num_reqs; | |
40b4f539 | 4675 | for (i = 0; i < num_reqs; i++) { |
d20d9b7c PB |
4676 | bdrv_co_aio_rw_vector(bs, reqs[i].sector, reqs[i].qiov, |
4677 | reqs[i].nb_sectors, reqs[i].flags, | |
4678 | multiwrite_cb, mcb, | |
4679 | true); | |
40b4f539 KW |
4680 | } |
4681 | ||
4682 | return 0; | |
40b4f539 KW |
4683 | } |
4684 | ||
7c84b1b8 | 4685 | void bdrv_aio_cancel(BlockAIOCB *acb) |
83f64091 | 4686 | { |
ca5fd113 FZ |
4687 | qemu_aio_ref(acb); |
4688 | bdrv_aio_cancel_async(acb); | |
4689 | while (acb->refcnt > 1) { | |
4690 | if (acb->aiocb_info->get_aio_context) { | |
4691 | aio_poll(acb->aiocb_info->get_aio_context(acb), true); | |
4692 | } else if (acb->bs) { | |
4693 | aio_poll(bdrv_get_aio_context(acb->bs), true); | |
4694 | } else { | |
4695 | abort(); | |
02c50efe | 4696 | } |
02c50efe | 4697 | } |
8007429a | 4698 | qemu_aio_unref(acb); |
02c50efe FZ |
4699 | } |
4700 | ||
4701 | /* Async version of aio cancel. The caller is not blocked if the acb implements | |
4702 | * cancel_async, otherwise we do nothing and let the request normally complete. | |
4703 | * In either case the completion callback must be called. */ | |
7c84b1b8 | 4704 | void bdrv_aio_cancel_async(BlockAIOCB *acb) |
02c50efe FZ |
4705 | { |
4706 | if (acb->aiocb_info->cancel_async) { | |
4707 | acb->aiocb_info->cancel_async(acb); | |
4708 | } | |
83f64091 FB |
4709 | } |
4710 | ||
4711 | /**************************************************************/ | |
4712 | /* async block device emulation */ | |
4713 | ||
7c84b1b8 MA |
4714 | typedef struct BlockAIOCBSync { |
4715 | BlockAIOCB common; | |
c16b5a2c CH |
4716 | QEMUBH *bh; |
4717 | int ret; | |
4718 | /* vector translation state */ | |
4719 | QEMUIOVector *qiov; | |
4720 | uint8_t *bounce; | |
4721 | int is_write; | |
7c84b1b8 | 4722 | } BlockAIOCBSync; |
c16b5a2c | 4723 | |
d7331bed | 4724 | static const AIOCBInfo bdrv_em_aiocb_info = { |
7c84b1b8 | 4725 | .aiocb_size = sizeof(BlockAIOCBSync), |
c16b5a2c CH |
4726 | }; |
4727 | ||
ce1a14dc | 4728 | static void bdrv_aio_bh_cb(void *opaque) |
83f64091 | 4729 | { |
7c84b1b8 | 4730 | BlockAIOCBSync *acb = opaque; |
f141eafe | 4731 | |
857d4f46 | 4732 | if (!acb->is_write && acb->ret >= 0) { |
03396148 | 4733 | qemu_iovec_from_buf(acb->qiov, 0, acb->bounce, acb->qiov->size); |
857d4f46 | 4734 | } |
ceb42de8 | 4735 | qemu_vfree(acb->bounce); |
ce1a14dc | 4736 | acb->common.cb(acb->common.opaque, acb->ret); |
6a7ad299 | 4737 | qemu_bh_delete(acb->bh); |
36afc451 | 4738 | acb->bh = NULL; |
8007429a | 4739 | qemu_aio_unref(acb); |
83f64091 | 4740 | } |
beac80cd | 4741 | |
7c84b1b8 MA |
4742 | static BlockAIOCB *bdrv_aio_rw_vector(BlockDriverState *bs, |
4743 | int64_t sector_num, | |
4744 | QEMUIOVector *qiov, | |
4745 | int nb_sectors, | |
097310b5 | 4746 | BlockCompletionFunc *cb, |
7c84b1b8 MA |
4747 | void *opaque, |
4748 | int is_write) | |
f141eafe | 4749 | |
83f64091 | 4750 | { |
7c84b1b8 | 4751 | BlockAIOCBSync *acb; |
ce1a14dc | 4752 | |
d7331bed | 4753 | acb = qemu_aio_get(&bdrv_em_aiocb_info, bs, cb, opaque); |
f141eafe AL |
4754 | acb->is_write = is_write; |
4755 | acb->qiov = qiov; | |
857d4f46 | 4756 | acb->bounce = qemu_try_blockalign(bs, qiov->size); |
2572b37a | 4757 | acb->bh = aio_bh_new(bdrv_get_aio_context(bs), bdrv_aio_bh_cb, acb); |
f141eafe | 4758 | |
857d4f46 KW |
4759 | if (acb->bounce == NULL) { |
4760 | acb->ret = -ENOMEM; | |
4761 | } else if (is_write) { | |
d5e6b161 | 4762 | qemu_iovec_to_buf(acb->qiov, 0, acb->bounce, qiov->size); |
1ed20acf | 4763 | acb->ret = bs->drv->bdrv_write(bs, sector_num, acb->bounce, nb_sectors); |
f141eafe | 4764 | } else { |
1ed20acf | 4765 | acb->ret = bs->drv->bdrv_read(bs, sector_num, acb->bounce, nb_sectors); |
f141eafe AL |
4766 | } |
4767 | ||
ce1a14dc | 4768 | qemu_bh_schedule(acb->bh); |
f141eafe | 4769 | |
ce1a14dc | 4770 | return &acb->common; |
beac80cd FB |
4771 | } |
4772 | ||
7c84b1b8 | 4773 | static BlockAIOCB *bdrv_aio_readv_em(BlockDriverState *bs, |
f141eafe | 4774 | int64_t sector_num, QEMUIOVector *qiov, int nb_sectors, |
097310b5 | 4775 | BlockCompletionFunc *cb, void *opaque) |
beac80cd | 4776 | { |
f141eafe AL |
4777 | return bdrv_aio_rw_vector(bs, sector_num, qiov, nb_sectors, cb, opaque, 0); |
4778 | } | |
83f64091 | 4779 | |
7c84b1b8 | 4780 | static BlockAIOCB *bdrv_aio_writev_em(BlockDriverState *bs, |
f141eafe | 4781 | int64_t sector_num, QEMUIOVector *qiov, int nb_sectors, |
097310b5 | 4782 | BlockCompletionFunc *cb, void *opaque) |
f141eafe AL |
4783 | { |
4784 | return bdrv_aio_rw_vector(bs, sector_num, qiov, nb_sectors, cb, opaque, 1); | |
beac80cd | 4785 | } |
beac80cd | 4786 | |
68485420 | 4787 | |
7c84b1b8 MA |
4788 | typedef struct BlockAIOCBCoroutine { |
4789 | BlockAIOCB common; | |
68485420 KW |
4790 | BlockRequest req; |
4791 | bool is_write; | |
d318aea9 | 4792 | bool *done; |
68485420 | 4793 | QEMUBH* bh; |
7c84b1b8 | 4794 | } BlockAIOCBCoroutine; |
68485420 | 4795 | |
d7331bed | 4796 | static const AIOCBInfo bdrv_em_co_aiocb_info = { |
7c84b1b8 | 4797 | .aiocb_size = sizeof(BlockAIOCBCoroutine), |
68485420 KW |
4798 | }; |
4799 | ||
35246a68 | 4800 | static void bdrv_co_em_bh(void *opaque) |
68485420 | 4801 | { |
7c84b1b8 | 4802 | BlockAIOCBCoroutine *acb = opaque; |
68485420 KW |
4803 | |
4804 | acb->common.cb(acb->common.opaque, acb->req.error); | |
d318aea9 | 4805 | |
68485420 | 4806 | qemu_bh_delete(acb->bh); |
8007429a | 4807 | qemu_aio_unref(acb); |
68485420 KW |
4808 | } |
4809 | ||
b2a61371 SH |
4810 | /* Invoke bdrv_co_do_readv/bdrv_co_do_writev */ |
4811 | static void coroutine_fn bdrv_co_do_rw(void *opaque) | |
4812 | { | |
7c84b1b8 | 4813 | BlockAIOCBCoroutine *acb = opaque; |
b2a61371 SH |
4814 | BlockDriverState *bs = acb->common.bs; |
4815 | ||
4816 | if (!acb->is_write) { | |
4817 | acb->req.error = bdrv_co_do_readv(bs, acb->req.sector, | |
d20d9b7c | 4818 | acb->req.nb_sectors, acb->req.qiov, acb->req.flags); |
b2a61371 SH |
4819 | } else { |
4820 | acb->req.error = bdrv_co_do_writev(bs, acb->req.sector, | |
d20d9b7c | 4821 | acb->req.nb_sectors, acb->req.qiov, acb->req.flags); |
b2a61371 SH |
4822 | } |
4823 | ||
2572b37a | 4824 | acb->bh = aio_bh_new(bdrv_get_aio_context(bs), bdrv_co_em_bh, acb); |
b2a61371 SH |
4825 | qemu_bh_schedule(acb->bh); |
4826 | } | |
4827 | ||
7c84b1b8 MA |
4828 | static BlockAIOCB *bdrv_co_aio_rw_vector(BlockDriverState *bs, |
4829 | int64_t sector_num, | |
4830 | QEMUIOVector *qiov, | |
4831 | int nb_sectors, | |
4832 | BdrvRequestFlags flags, | |
097310b5 | 4833 | BlockCompletionFunc *cb, |
7c84b1b8 MA |
4834 | void *opaque, |
4835 | bool is_write) | |
68485420 KW |
4836 | { |
4837 | Coroutine *co; | |
7c84b1b8 | 4838 | BlockAIOCBCoroutine *acb; |
68485420 | 4839 | |
d7331bed | 4840 | acb = qemu_aio_get(&bdrv_em_co_aiocb_info, bs, cb, opaque); |
68485420 KW |
4841 | acb->req.sector = sector_num; |
4842 | acb->req.nb_sectors = nb_sectors; | |
4843 | acb->req.qiov = qiov; | |
d20d9b7c | 4844 | acb->req.flags = flags; |
68485420 KW |
4845 | acb->is_write = is_write; |
4846 | ||
8c5873d6 | 4847 | co = qemu_coroutine_create(bdrv_co_do_rw); |
68485420 KW |
4848 | qemu_coroutine_enter(co, acb); |
4849 | ||
4850 | return &acb->common; | |
4851 | } | |
4852 | ||
07f07615 | 4853 | static void coroutine_fn bdrv_aio_flush_co_entry(void *opaque) |
b2e12bc6 | 4854 | { |
7c84b1b8 | 4855 | BlockAIOCBCoroutine *acb = opaque; |
07f07615 | 4856 | BlockDriverState *bs = acb->common.bs; |
b2e12bc6 | 4857 | |
07f07615 | 4858 | acb->req.error = bdrv_co_flush(bs); |
2572b37a | 4859 | acb->bh = aio_bh_new(bdrv_get_aio_context(bs), bdrv_co_em_bh, acb); |
b2e12bc6 | 4860 | qemu_bh_schedule(acb->bh); |
b2e12bc6 CH |
4861 | } |
4862 | ||
7c84b1b8 | 4863 | BlockAIOCB *bdrv_aio_flush(BlockDriverState *bs, |
097310b5 | 4864 | BlockCompletionFunc *cb, void *opaque) |
016f5cf6 | 4865 | { |
07f07615 | 4866 | trace_bdrv_aio_flush(bs, opaque); |
016f5cf6 | 4867 | |
07f07615 | 4868 | Coroutine *co; |
7c84b1b8 | 4869 | BlockAIOCBCoroutine *acb; |
016f5cf6 | 4870 | |
d7331bed | 4871 | acb = qemu_aio_get(&bdrv_em_co_aiocb_info, bs, cb, opaque); |
d318aea9 | 4872 | |
07f07615 PB |
4873 | co = qemu_coroutine_create(bdrv_aio_flush_co_entry); |
4874 | qemu_coroutine_enter(co, acb); | |
016f5cf6 | 4875 | |
016f5cf6 AG |
4876 | return &acb->common; |
4877 | } | |
4878 | ||
4265d620 PB |
4879 | static void coroutine_fn bdrv_aio_discard_co_entry(void *opaque) |
4880 | { | |
7c84b1b8 | 4881 | BlockAIOCBCoroutine *acb = opaque; |
4265d620 PB |
4882 | BlockDriverState *bs = acb->common.bs; |
4883 | ||
4884 | acb->req.error = bdrv_co_discard(bs, acb->req.sector, acb->req.nb_sectors); | |
2572b37a | 4885 | acb->bh = aio_bh_new(bdrv_get_aio_context(bs), bdrv_co_em_bh, acb); |
4265d620 PB |
4886 | qemu_bh_schedule(acb->bh); |
4887 | } | |
4888 | ||
7c84b1b8 | 4889 | BlockAIOCB *bdrv_aio_discard(BlockDriverState *bs, |
4265d620 | 4890 | int64_t sector_num, int nb_sectors, |
097310b5 | 4891 | BlockCompletionFunc *cb, void *opaque) |
4265d620 PB |
4892 | { |
4893 | Coroutine *co; | |
7c84b1b8 | 4894 | BlockAIOCBCoroutine *acb; |
4265d620 PB |
4895 | |
4896 | trace_bdrv_aio_discard(bs, sector_num, nb_sectors, opaque); | |
4897 | ||
d7331bed | 4898 | acb = qemu_aio_get(&bdrv_em_co_aiocb_info, bs, cb, opaque); |
4265d620 PB |
4899 | acb->req.sector = sector_num; |
4900 | acb->req.nb_sectors = nb_sectors; | |
4901 | co = qemu_coroutine_create(bdrv_aio_discard_co_entry); | |
4902 | qemu_coroutine_enter(co, acb); | |
4903 | ||
4904 | return &acb->common; | |
4905 | } | |
4906 | ||
ea2384d3 FB |
4907 | void bdrv_init(void) |
4908 | { | |
5efa9d5a | 4909 | module_call_init(MODULE_INIT_BLOCK); |
ea2384d3 | 4910 | } |
ce1a14dc | 4911 | |
eb852011 MA |
4912 | void bdrv_init_with_whitelist(void) |
4913 | { | |
4914 | use_bdrv_whitelist = 1; | |
4915 | bdrv_init(); | |
4916 | } | |
4917 | ||
d7331bed | 4918 | void *qemu_aio_get(const AIOCBInfo *aiocb_info, BlockDriverState *bs, |
097310b5 | 4919 | BlockCompletionFunc *cb, void *opaque) |
ce1a14dc | 4920 | { |
7c84b1b8 | 4921 | BlockAIOCB *acb; |
ce1a14dc | 4922 | |
d7331bed SH |
4923 | acb = g_slice_alloc(aiocb_info->aiocb_size); |
4924 | acb->aiocb_info = aiocb_info; | |
ce1a14dc PB |
4925 | acb->bs = bs; |
4926 | acb->cb = cb; | |
4927 | acb->opaque = opaque; | |
f197fe2b | 4928 | acb->refcnt = 1; |
ce1a14dc PB |
4929 | return acb; |
4930 | } | |
4931 | ||
f197fe2b FZ |
4932 | void qemu_aio_ref(void *p) |
4933 | { | |
7c84b1b8 | 4934 | BlockAIOCB *acb = p; |
f197fe2b FZ |
4935 | acb->refcnt++; |
4936 | } | |
4937 | ||
8007429a | 4938 | void qemu_aio_unref(void *p) |
ce1a14dc | 4939 | { |
7c84b1b8 | 4940 | BlockAIOCB *acb = p; |
f197fe2b FZ |
4941 | assert(acb->refcnt > 0); |
4942 | if (--acb->refcnt == 0) { | |
4943 | g_slice_free1(acb->aiocb_info->aiocb_size, acb); | |
4944 | } | |
ce1a14dc | 4945 | } |
19cb3738 | 4946 | |
f9f05dc5 KW |
4947 | /**************************************************************/ |
4948 | /* Coroutine block device emulation */ | |
4949 | ||
4950 | typedef struct CoroutineIOCompletion { | |
4951 | Coroutine *coroutine; | |
4952 | int ret; | |
4953 | } CoroutineIOCompletion; | |
4954 | ||
4955 | static void bdrv_co_io_em_complete(void *opaque, int ret) | |
4956 | { | |
4957 | CoroutineIOCompletion *co = opaque; | |
4958 | ||
4959 | co->ret = ret; | |
4960 | qemu_coroutine_enter(co->coroutine, NULL); | |
4961 | } | |
4962 | ||
4963 | static int coroutine_fn bdrv_co_io_em(BlockDriverState *bs, int64_t sector_num, | |
4964 | int nb_sectors, QEMUIOVector *iov, | |
4965 | bool is_write) | |
4966 | { | |
4967 | CoroutineIOCompletion co = { | |
4968 | .coroutine = qemu_coroutine_self(), | |
4969 | }; | |
7c84b1b8 | 4970 | BlockAIOCB *acb; |
f9f05dc5 KW |
4971 | |
4972 | if (is_write) { | |
a652d160 SH |
4973 | acb = bs->drv->bdrv_aio_writev(bs, sector_num, iov, nb_sectors, |
4974 | bdrv_co_io_em_complete, &co); | |
f9f05dc5 | 4975 | } else { |
a652d160 SH |
4976 | acb = bs->drv->bdrv_aio_readv(bs, sector_num, iov, nb_sectors, |
4977 | bdrv_co_io_em_complete, &co); | |
f9f05dc5 KW |
4978 | } |
4979 | ||
59370aaa | 4980 | trace_bdrv_co_io_em(bs, sector_num, nb_sectors, is_write, acb); |
f9f05dc5 KW |
4981 | if (!acb) { |
4982 | return -EIO; | |
4983 | } | |
4984 | qemu_coroutine_yield(); | |
4985 | ||
4986 | return co.ret; | |
4987 | } | |
4988 | ||
4989 | static int coroutine_fn bdrv_co_readv_em(BlockDriverState *bs, | |
4990 | int64_t sector_num, int nb_sectors, | |
4991 | QEMUIOVector *iov) | |
4992 | { | |
4993 | return bdrv_co_io_em(bs, sector_num, nb_sectors, iov, false); | |
4994 | } | |
4995 | ||
4996 | static int coroutine_fn bdrv_co_writev_em(BlockDriverState *bs, | |
4997 | int64_t sector_num, int nb_sectors, | |
4998 | QEMUIOVector *iov) | |
4999 | { | |
5000 | return bdrv_co_io_em(bs, sector_num, nb_sectors, iov, true); | |
5001 | } | |
5002 | ||
07f07615 | 5003 | static void coroutine_fn bdrv_flush_co_entry(void *opaque) |
e7a8a783 | 5004 | { |
07f07615 PB |
5005 | RwCo *rwco = opaque; |
5006 | ||
5007 | rwco->ret = bdrv_co_flush(rwco->bs); | |
5008 | } | |
5009 | ||
5010 | int coroutine_fn bdrv_co_flush(BlockDriverState *bs) | |
5011 | { | |
eb489bb1 KW |
5012 | int ret; |
5013 | ||
29cdb251 | 5014 | if (!bs || !bdrv_is_inserted(bs) || bdrv_is_read_only(bs)) { |
07f07615 | 5015 | return 0; |
eb489bb1 KW |
5016 | } |
5017 | ||
ca716364 | 5018 | /* Write back cached data to the OS even with cache=unsafe */ |
bf736fe3 | 5019 | BLKDBG_EVENT(bs->file, BLKDBG_FLUSH_TO_OS); |
eb489bb1 KW |
5020 | if (bs->drv->bdrv_co_flush_to_os) { |
5021 | ret = bs->drv->bdrv_co_flush_to_os(bs); | |
5022 | if (ret < 0) { | |
5023 | return ret; | |
5024 | } | |
5025 | } | |
5026 | ||
ca716364 KW |
5027 | /* But don't actually force it to the disk with cache=unsafe */ |
5028 | if (bs->open_flags & BDRV_O_NO_FLUSH) { | |
d4c82329 | 5029 | goto flush_parent; |
ca716364 KW |
5030 | } |
5031 | ||
bf736fe3 | 5032 | BLKDBG_EVENT(bs->file, BLKDBG_FLUSH_TO_DISK); |
eb489bb1 | 5033 | if (bs->drv->bdrv_co_flush_to_disk) { |
29cdb251 | 5034 | ret = bs->drv->bdrv_co_flush_to_disk(bs); |
07f07615 | 5035 | } else if (bs->drv->bdrv_aio_flush) { |
7c84b1b8 | 5036 | BlockAIOCB *acb; |
07f07615 PB |
5037 | CoroutineIOCompletion co = { |
5038 | .coroutine = qemu_coroutine_self(), | |
5039 | }; | |
5040 | ||
5041 | acb = bs->drv->bdrv_aio_flush(bs, bdrv_co_io_em_complete, &co); | |
5042 | if (acb == NULL) { | |
29cdb251 | 5043 | ret = -EIO; |
07f07615 PB |
5044 | } else { |
5045 | qemu_coroutine_yield(); | |
29cdb251 | 5046 | ret = co.ret; |
07f07615 | 5047 | } |
07f07615 PB |
5048 | } else { |
5049 | /* | |
5050 | * Some block drivers always operate in either writethrough or unsafe | |
5051 | * mode and don't support bdrv_flush therefore. Usually qemu doesn't | |
5052 | * know how the server works (because the behaviour is hardcoded or | |
5053 | * depends on server-side configuration), so we can't ensure that | |
5054 | * everything is safe on disk. Returning an error doesn't work because | |
5055 | * that would break guests even if the server operates in writethrough | |
5056 | * mode. | |
5057 | * | |
5058 | * Let's hope the user knows what he's doing. | |
5059 | */ | |
29cdb251 | 5060 | ret = 0; |
07f07615 | 5061 | } |
29cdb251 PB |
5062 | if (ret < 0) { |
5063 | return ret; | |
5064 | } | |
5065 | ||
5066 | /* Now flush the underlying protocol. It will also have BDRV_O_NO_FLUSH | |
5067 | * in the case of cache=unsafe, so there are no useless flushes. | |
5068 | */ | |
d4c82329 | 5069 | flush_parent: |
29cdb251 | 5070 | return bdrv_co_flush(bs->file); |
07f07615 PB |
5071 | } |
5072 | ||
5a8a30db | 5073 | void bdrv_invalidate_cache(BlockDriverState *bs, Error **errp) |
0f15423c | 5074 | { |
5a8a30db KW |
5075 | Error *local_err = NULL; |
5076 | int ret; | |
5077 | ||
3456a8d1 KW |
5078 | if (!bs->drv) { |
5079 | return; | |
5080 | } | |
5081 | ||
7ea2d269 AK |
5082 | if (!(bs->open_flags & BDRV_O_INCOMING)) { |
5083 | return; | |
5084 | } | |
5085 | bs->open_flags &= ~BDRV_O_INCOMING; | |
5086 | ||
3456a8d1 | 5087 | if (bs->drv->bdrv_invalidate_cache) { |
5a8a30db | 5088 | bs->drv->bdrv_invalidate_cache(bs, &local_err); |
3456a8d1 | 5089 | } else if (bs->file) { |
5a8a30db KW |
5090 | bdrv_invalidate_cache(bs->file, &local_err); |
5091 | } | |
5092 | if (local_err) { | |
5093 | error_propagate(errp, local_err); | |
5094 | return; | |
0f15423c | 5095 | } |
3456a8d1 | 5096 | |
5a8a30db KW |
5097 | ret = refresh_total_sectors(bs, bs->total_sectors); |
5098 | if (ret < 0) { | |
5099 | error_setg_errno(errp, -ret, "Could not refresh total sector count"); | |
5100 | return; | |
5101 | } | |
0f15423c AL |
5102 | } |
5103 | ||
5a8a30db | 5104 | void bdrv_invalidate_cache_all(Error **errp) |
0f15423c AL |
5105 | { |
5106 | BlockDriverState *bs; | |
5a8a30db | 5107 | Error *local_err = NULL; |
0f15423c | 5108 | |
dc364f4c | 5109 | QTAILQ_FOREACH(bs, &bdrv_states, device_list) { |
ed78cda3 SH |
5110 | AioContext *aio_context = bdrv_get_aio_context(bs); |
5111 | ||
5112 | aio_context_acquire(aio_context); | |
5a8a30db | 5113 | bdrv_invalidate_cache(bs, &local_err); |
ed78cda3 | 5114 | aio_context_release(aio_context); |
5a8a30db KW |
5115 | if (local_err) { |
5116 | error_propagate(errp, local_err); | |
5117 | return; | |
5118 | } | |
0f15423c AL |
5119 | } |
5120 | } | |
5121 | ||
07f07615 PB |
5122 | int bdrv_flush(BlockDriverState *bs) |
5123 | { | |
5124 | Coroutine *co; | |
5125 | RwCo rwco = { | |
5126 | .bs = bs, | |
5127 | .ret = NOT_DONE, | |
e7a8a783 | 5128 | }; |
e7a8a783 | 5129 | |
07f07615 PB |
5130 | if (qemu_in_coroutine()) { |
5131 | /* Fast-path if already in coroutine context */ | |
5132 | bdrv_flush_co_entry(&rwco); | |
5133 | } else { | |
2572b37a SH |
5134 | AioContext *aio_context = bdrv_get_aio_context(bs); |
5135 | ||
07f07615 PB |
5136 | co = qemu_coroutine_create(bdrv_flush_co_entry); |
5137 | qemu_coroutine_enter(co, &rwco); | |
5138 | while (rwco.ret == NOT_DONE) { | |
2572b37a | 5139 | aio_poll(aio_context, true); |
07f07615 | 5140 | } |
e7a8a783 | 5141 | } |
07f07615 PB |
5142 | |
5143 | return rwco.ret; | |
e7a8a783 KW |
5144 | } |
5145 | ||
775aa8b6 KW |
5146 | typedef struct DiscardCo { |
5147 | BlockDriverState *bs; | |
5148 | int64_t sector_num; | |
5149 | int nb_sectors; | |
5150 | int ret; | |
5151 | } DiscardCo; | |
4265d620 PB |
5152 | static void coroutine_fn bdrv_discard_co_entry(void *opaque) |
5153 | { | |
775aa8b6 | 5154 | DiscardCo *rwco = opaque; |
4265d620 PB |
5155 | |
5156 | rwco->ret = bdrv_co_discard(rwco->bs, rwco->sector_num, rwco->nb_sectors); | |
5157 | } | |
5158 | ||
5159 | int coroutine_fn bdrv_co_discard(BlockDriverState *bs, int64_t sector_num, | |
5160 | int nb_sectors) | |
5161 | { | |
b9c64947 | 5162 | int max_discard, ret; |
d51e9fe5 | 5163 | |
4265d620 PB |
5164 | if (!bs->drv) { |
5165 | return -ENOMEDIUM; | |
b9c64947 HR |
5166 | } |
5167 | ||
5168 | ret = bdrv_check_request(bs, sector_num, nb_sectors); | |
5169 | if (ret < 0) { | |
5170 | return ret; | |
4265d620 PB |
5171 | } else if (bs->read_only) { |
5172 | return -EROFS; | |
df702c9b PB |
5173 | } |
5174 | ||
e4654d2d | 5175 | bdrv_reset_dirty(bs, sector_num, nb_sectors); |
df702c9b | 5176 | |
9e8f1835 PB |
5177 | /* Do nothing if disabled. */ |
5178 | if (!(bs->open_flags & BDRV_O_UNMAP)) { | |
5179 | return 0; | |
5180 | } | |
5181 | ||
d51e9fe5 PB |
5182 | if (!bs->drv->bdrv_co_discard && !bs->drv->bdrv_aio_discard) { |
5183 | return 0; | |
5184 | } | |
6f14da52 | 5185 | |
75af1f34 | 5186 | max_discard = MIN_NON_ZERO(bs->bl.max_discard, BDRV_REQUEST_MAX_SECTORS); |
d51e9fe5 PB |
5187 | while (nb_sectors > 0) { |
5188 | int ret; | |
5189 | int num = nb_sectors; | |
6f14da52 | 5190 | |
d51e9fe5 PB |
5191 | /* align request */ |
5192 | if (bs->bl.discard_alignment && | |
5193 | num >= bs->bl.discard_alignment && | |
5194 | sector_num % bs->bl.discard_alignment) { | |
5195 | if (num > bs->bl.discard_alignment) { | |
5196 | num = bs->bl.discard_alignment; | |
6f14da52 | 5197 | } |
d51e9fe5 PB |
5198 | num -= sector_num % bs->bl.discard_alignment; |
5199 | } | |
6f14da52 | 5200 | |
d51e9fe5 PB |
5201 | /* limit request size */ |
5202 | if (num > max_discard) { | |
5203 | num = max_discard; | |
5204 | } | |
6f14da52 | 5205 | |
d51e9fe5 | 5206 | if (bs->drv->bdrv_co_discard) { |
6f14da52 | 5207 | ret = bs->drv->bdrv_co_discard(bs, sector_num, num); |
d51e9fe5 | 5208 | } else { |
7c84b1b8 | 5209 | BlockAIOCB *acb; |
d51e9fe5 PB |
5210 | CoroutineIOCompletion co = { |
5211 | .coroutine = qemu_coroutine_self(), | |
5212 | }; | |
5213 | ||
5214 | acb = bs->drv->bdrv_aio_discard(bs, sector_num, nb_sectors, | |
5215 | bdrv_co_io_em_complete, &co); | |
5216 | if (acb == NULL) { | |
5217 | return -EIO; | |
5218 | } else { | |
5219 | qemu_coroutine_yield(); | |
5220 | ret = co.ret; | |
6f14da52 | 5221 | } |
6f14da52 | 5222 | } |
7ce21016 | 5223 | if (ret && ret != -ENOTSUP) { |
d51e9fe5 | 5224 | return ret; |
4265d620 | 5225 | } |
d51e9fe5 PB |
5226 | |
5227 | sector_num += num; | |
5228 | nb_sectors -= num; | |
4265d620 | 5229 | } |
d51e9fe5 | 5230 | return 0; |
4265d620 PB |
5231 | } |
5232 | ||
5233 | int bdrv_discard(BlockDriverState *bs, int64_t sector_num, int nb_sectors) | |
5234 | { | |
5235 | Coroutine *co; | |
775aa8b6 | 5236 | DiscardCo rwco = { |
4265d620 PB |
5237 | .bs = bs, |
5238 | .sector_num = sector_num, | |
5239 | .nb_sectors = nb_sectors, | |
5240 | .ret = NOT_DONE, | |
5241 | }; | |
5242 | ||
5243 | if (qemu_in_coroutine()) { | |
5244 | /* Fast-path if already in coroutine context */ | |
5245 | bdrv_discard_co_entry(&rwco); | |
5246 | } else { | |
2572b37a SH |
5247 | AioContext *aio_context = bdrv_get_aio_context(bs); |
5248 | ||
4265d620 PB |
5249 | co = qemu_coroutine_create(bdrv_discard_co_entry); |
5250 | qemu_coroutine_enter(co, &rwco); | |
5251 | while (rwco.ret == NOT_DONE) { | |
2572b37a | 5252 | aio_poll(aio_context, true); |
4265d620 PB |
5253 | } |
5254 | } | |
5255 | ||
5256 | return rwco.ret; | |
5257 | } | |
5258 | ||
19cb3738 FB |
5259 | /**************************************************************/ |
5260 | /* removable device support */ | |
5261 | ||
5262 | /** | |
5263 | * Return TRUE if the media is present | |
5264 | */ | |
5265 | int bdrv_is_inserted(BlockDriverState *bs) | |
5266 | { | |
5267 | BlockDriver *drv = bs->drv; | |
a1aff5bf | 5268 | |
19cb3738 FB |
5269 | if (!drv) |
5270 | return 0; | |
5271 | if (!drv->bdrv_is_inserted) | |
a1aff5bf MA |
5272 | return 1; |
5273 | return drv->bdrv_is_inserted(bs); | |
19cb3738 FB |
5274 | } |
5275 | ||
5276 | /** | |
8e49ca46 MA |
5277 | * Return whether the media changed since the last call to this |
5278 | * function, or -ENOTSUP if we don't know. Most drivers don't know. | |
19cb3738 FB |
5279 | */ |
5280 | int bdrv_media_changed(BlockDriverState *bs) | |
5281 | { | |
5282 | BlockDriver *drv = bs->drv; | |
19cb3738 | 5283 | |
8e49ca46 MA |
5284 | if (drv && drv->bdrv_media_changed) { |
5285 | return drv->bdrv_media_changed(bs); | |
5286 | } | |
5287 | return -ENOTSUP; | |
19cb3738 FB |
5288 | } |
5289 | ||
5290 | /** | |
5291 | * If eject_flag is TRUE, eject the media. Otherwise, close the tray | |
5292 | */ | |
f36f3949 | 5293 | void bdrv_eject(BlockDriverState *bs, bool eject_flag) |
19cb3738 FB |
5294 | { |
5295 | BlockDriver *drv = bs->drv; | |
bfb197e0 | 5296 | const char *device_name; |
19cb3738 | 5297 | |
822e1cd1 MA |
5298 | if (drv && drv->bdrv_eject) { |
5299 | drv->bdrv_eject(bs, eject_flag); | |
19cb3738 | 5300 | } |
6f382ed2 | 5301 | |
bfb197e0 MA |
5302 | device_name = bdrv_get_device_name(bs); |
5303 | if (device_name[0] != '\0') { | |
5304 | qapi_event_send_device_tray_moved(device_name, | |
a5ee7bd4 | 5305 | eject_flag, &error_abort); |
6f382ed2 | 5306 | } |
19cb3738 FB |
5307 | } |
5308 | ||
19cb3738 FB |
5309 | /** |
5310 | * Lock or unlock the media (if it is locked, the user won't be able | |
5311 | * to eject it manually). | |
5312 | */ | |
025e849a | 5313 | void bdrv_lock_medium(BlockDriverState *bs, bool locked) |
19cb3738 FB |
5314 | { |
5315 | BlockDriver *drv = bs->drv; | |
5316 | ||
025e849a | 5317 | trace_bdrv_lock_medium(bs, locked); |
b8c6d095 | 5318 | |
025e849a MA |
5319 | if (drv && drv->bdrv_lock_medium) { |
5320 | drv->bdrv_lock_medium(bs, locked); | |
19cb3738 FB |
5321 | } |
5322 | } | |
985a03b0 TS |
5323 | |
5324 | /* needed for generic scsi interface */ | |
5325 | ||
5326 | int bdrv_ioctl(BlockDriverState *bs, unsigned long int req, void *buf) | |
5327 | { | |
5328 | BlockDriver *drv = bs->drv; | |
5329 | ||
5330 | if (drv && drv->bdrv_ioctl) | |
5331 | return drv->bdrv_ioctl(bs, req, buf); | |
5332 | return -ENOTSUP; | |
5333 | } | |
7d780669 | 5334 | |
7c84b1b8 | 5335 | BlockAIOCB *bdrv_aio_ioctl(BlockDriverState *bs, |
221f715d | 5336 | unsigned long int req, void *buf, |
097310b5 | 5337 | BlockCompletionFunc *cb, void *opaque) |
7d780669 | 5338 | { |
221f715d | 5339 | BlockDriver *drv = bs->drv; |
7d780669 | 5340 | |
221f715d AL |
5341 | if (drv && drv->bdrv_aio_ioctl) |
5342 | return drv->bdrv_aio_ioctl(bs, req, buf, cb, opaque); | |
5343 | return NULL; | |
7d780669 | 5344 | } |
e268ca52 | 5345 | |
1b7fd729 | 5346 | void bdrv_set_guest_block_size(BlockDriverState *bs, int align) |
7b6f9300 | 5347 | { |
1b7fd729 | 5348 | bs->guest_block_size = align; |
7b6f9300 | 5349 | } |
7cd1e32a | 5350 | |
e268ca52 AL |
5351 | void *qemu_blockalign(BlockDriverState *bs, size_t size) |
5352 | { | |
339064d5 | 5353 | return qemu_memalign(bdrv_opt_mem_align(bs), size); |
e268ca52 | 5354 | } |
7cd1e32a | 5355 | |
9ebd8448 HR |
5356 | void *qemu_blockalign0(BlockDriverState *bs, size_t size) |
5357 | { | |
5358 | return memset(qemu_blockalign(bs, size), 0, size); | |
5359 | } | |
5360 | ||
7d2a35cc KW |
5361 | void *qemu_try_blockalign(BlockDriverState *bs, size_t size) |
5362 | { | |
5363 | size_t align = bdrv_opt_mem_align(bs); | |
5364 | ||
5365 | /* Ensure that NULL is never returned on success */ | |
5366 | assert(align > 0); | |
5367 | if (size == 0) { | |
5368 | size = align; | |
5369 | } | |
5370 | ||
5371 | return qemu_try_memalign(align, size); | |
5372 | } | |
5373 | ||
9ebd8448 HR |
5374 | void *qemu_try_blockalign0(BlockDriverState *bs, size_t size) |
5375 | { | |
5376 | void *mem = qemu_try_blockalign(bs, size); | |
5377 | ||
5378 | if (mem) { | |
5379 | memset(mem, 0, size); | |
5380 | } | |
5381 | ||
5382 | return mem; | |
5383 | } | |
5384 | ||
c53b1c51 SH |
5385 | /* |
5386 | * Check if all memory in this vector is sector aligned. | |
5387 | */ | |
5388 | bool bdrv_qiov_is_aligned(BlockDriverState *bs, QEMUIOVector *qiov) | |
5389 | { | |
5390 | int i; | |
339064d5 | 5391 | size_t alignment = bdrv_opt_mem_align(bs); |
c53b1c51 SH |
5392 | |
5393 | for (i = 0; i < qiov->niov; i++) { | |
339064d5 | 5394 | if ((uintptr_t) qiov->iov[i].iov_base % alignment) { |
c53b1c51 | 5395 | return false; |
1ff735bd | 5396 | } |
339064d5 | 5397 | if (qiov->iov[i].iov_len % alignment) { |
1ff735bd | 5398 | return false; |
c53b1c51 SH |
5399 | } |
5400 | } | |
5401 | ||
5402 | return true; | |
5403 | } | |
5404 | ||
b8afb520 FZ |
5405 | BdrvDirtyBitmap *bdrv_create_dirty_bitmap(BlockDriverState *bs, int granularity, |
5406 | Error **errp) | |
7cd1e32a LS |
5407 | { |
5408 | int64_t bitmap_size; | |
e4654d2d | 5409 | BdrvDirtyBitmap *bitmap; |
a55eb92c | 5410 | |
50717e94 PB |
5411 | assert((granularity & (granularity - 1)) == 0); |
5412 | ||
e4654d2d FZ |
5413 | granularity >>= BDRV_SECTOR_BITS; |
5414 | assert(granularity); | |
57322b78 | 5415 | bitmap_size = bdrv_nb_sectors(bs); |
b8afb520 FZ |
5416 | if (bitmap_size < 0) { |
5417 | error_setg_errno(errp, -bitmap_size, "could not get length of device"); | |
5418 | errno = -bitmap_size; | |
5419 | return NULL; | |
5420 | } | |
5839e53b | 5421 | bitmap = g_new0(BdrvDirtyBitmap, 1); |
e4654d2d FZ |
5422 | bitmap->bitmap = hbitmap_alloc(bitmap_size, ffs(granularity) - 1); |
5423 | QLIST_INSERT_HEAD(&bs->dirty_bitmaps, bitmap, list); | |
5424 | return bitmap; | |
5425 | } | |
5426 | ||
5427 | void bdrv_release_dirty_bitmap(BlockDriverState *bs, BdrvDirtyBitmap *bitmap) | |
5428 | { | |
5429 | BdrvDirtyBitmap *bm, *next; | |
5430 | QLIST_FOREACH_SAFE(bm, &bs->dirty_bitmaps, list, next) { | |
5431 | if (bm == bitmap) { | |
5432 | QLIST_REMOVE(bitmap, list); | |
5433 | hbitmap_free(bitmap->bitmap); | |
5434 | g_free(bitmap); | |
5435 | return; | |
a55eb92c | 5436 | } |
7cd1e32a LS |
5437 | } |
5438 | } | |
5439 | ||
21b56835 FZ |
5440 | BlockDirtyInfoList *bdrv_query_dirty_bitmaps(BlockDriverState *bs) |
5441 | { | |
5442 | BdrvDirtyBitmap *bm; | |
5443 | BlockDirtyInfoList *list = NULL; | |
5444 | BlockDirtyInfoList **plist = &list; | |
5445 | ||
5446 | QLIST_FOREACH(bm, &bs->dirty_bitmaps, list) { | |
5839e53b MA |
5447 | BlockDirtyInfo *info = g_new0(BlockDirtyInfo, 1); |
5448 | BlockDirtyInfoList *entry = g_new0(BlockDirtyInfoList, 1); | |
21b56835 FZ |
5449 | info->count = bdrv_get_dirty_count(bs, bm); |
5450 | info->granularity = | |
5451 | ((int64_t) BDRV_SECTOR_SIZE << hbitmap_granularity(bm->bitmap)); | |
5452 | entry->value = info; | |
5453 | *plist = entry; | |
5454 | plist = &entry->next; | |
5455 | } | |
5456 | ||
5457 | return list; | |
5458 | } | |
5459 | ||
e4654d2d | 5460 | int bdrv_get_dirty(BlockDriverState *bs, BdrvDirtyBitmap *bitmap, int64_t sector) |
7cd1e32a | 5461 | { |
e4654d2d FZ |
5462 | if (bitmap) { |
5463 | return hbitmap_get(bitmap->bitmap, sector); | |
7cd1e32a LS |
5464 | } else { |
5465 | return 0; | |
5466 | } | |
5467 | } | |
5468 | ||
e4654d2d FZ |
5469 | void bdrv_dirty_iter_init(BlockDriverState *bs, |
5470 | BdrvDirtyBitmap *bitmap, HBitmapIter *hbi) | |
1755da16 | 5471 | { |
e4654d2d | 5472 | hbitmap_iter_init(hbi, bitmap->bitmap, 0); |
1755da16 PB |
5473 | } |
5474 | ||
c4237dfa VSO |
5475 | void bdrv_set_dirty_bitmap(BlockDriverState *bs, BdrvDirtyBitmap *bitmap, |
5476 | int64_t cur_sector, int nr_sectors) | |
5477 | { | |
5478 | hbitmap_set(bitmap->bitmap, cur_sector, nr_sectors); | |
5479 | } | |
5480 | ||
5481 | void bdrv_reset_dirty_bitmap(BlockDriverState *bs, BdrvDirtyBitmap *bitmap, | |
5482 | int64_t cur_sector, int nr_sectors) | |
5483 | { | |
5484 | hbitmap_reset(bitmap->bitmap, cur_sector, nr_sectors); | |
5485 | } | |
5486 | ||
5487 | static void bdrv_set_dirty(BlockDriverState *bs, int64_t cur_sector, | |
5488 | int nr_sectors) | |
1755da16 | 5489 | { |
e4654d2d FZ |
5490 | BdrvDirtyBitmap *bitmap; |
5491 | QLIST_FOREACH(bitmap, &bs->dirty_bitmaps, list) { | |
5492 | hbitmap_set(bitmap->bitmap, cur_sector, nr_sectors); | |
5493 | } | |
1755da16 PB |
5494 | } |
5495 | ||
c4237dfa VSO |
5496 | static void bdrv_reset_dirty(BlockDriverState *bs, int64_t cur_sector, |
5497 | int nr_sectors) | |
7cd1e32a | 5498 | { |
e4654d2d FZ |
5499 | BdrvDirtyBitmap *bitmap; |
5500 | QLIST_FOREACH(bitmap, &bs->dirty_bitmaps, list) { | |
5501 | hbitmap_reset(bitmap->bitmap, cur_sector, nr_sectors); | |
5502 | } | |
7cd1e32a | 5503 | } |
aaa0eb75 | 5504 | |
e4654d2d | 5505 | int64_t bdrv_get_dirty_count(BlockDriverState *bs, BdrvDirtyBitmap *bitmap) |
aaa0eb75 | 5506 | { |
e4654d2d | 5507 | return hbitmap_count(bitmap->bitmap); |
aaa0eb75 | 5508 | } |
f88e1a42 | 5509 | |
9fcb0251 FZ |
5510 | /* Get a reference to bs */ |
5511 | void bdrv_ref(BlockDriverState *bs) | |
5512 | { | |
5513 | bs->refcnt++; | |
5514 | } | |
5515 | ||
5516 | /* Release a previously grabbed reference to bs. | |
5517 | * If after releasing, reference count is zero, the BlockDriverState is | |
5518 | * deleted. */ | |
5519 | void bdrv_unref(BlockDriverState *bs) | |
5520 | { | |
9a4d5ca6 JC |
5521 | if (!bs) { |
5522 | return; | |
5523 | } | |
9fcb0251 FZ |
5524 | assert(bs->refcnt > 0); |
5525 | if (--bs->refcnt == 0) { | |
5526 | bdrv_delete(bs); | |
5527 | } | |
5528 | } | |
5529 | ||
fbe40ff7 FZ |
5530 | struct BdrvOpBlocker { |
5531 | Error *reason; | |
5532 | QLIST_ENTRY(BdrvOpBlocker) list; | |
5533 | }; | |
5534 | ||
5535 | bool bdrv_op_is_blocked(BlockDriverState *bs, BlockOpType op, Error **errp) | |
5536 | { | |
5537 | BdrvOpBlocker *blocker; | |
5538 | assert((int) op >= 0 && op < BLOCK_OP_TYPE_MAX); | |
5539 | if (!QLIST_EMPTY(&bs->op_blockers[op])) { | |
5540 | blocker = QLIST_FIRST(&bs->op_blockers[op]); | |
5541 | if (errp) { | |
5542 | error_setg(errp, "Device '%s' is busy: %s", | |
bfb197e0 MA |
5543 | bdrv_get_device_name(bs), |
5544 | error_get_pretty(blocker->reason)); | |
fbe40ff7 FZ |
5545 | } |
5546 | return true; | |
5547 | } | |
5548 | return false; | |
5549 | } | |
5550 | ||
5551 | void bdrv_op_block(BlockDriverState *bs, BlockOpType op, Error *reason) | |
5552 | { | |
5553 | BdrvOpBlocker *blocker; | |
5554 | assert((int) op >= 0 && op < BLOCK_OP_TYPE_MAX); | |
5555 | ||
5839e53b | 5556 | blocker = g_new0(BdrvOpBlocker, 1); |
fbe40ff7 FZ |
5557 | blocker->reason = reason; |
5558 | QLIST_INSERT_HEAD(&bs->op_blockers[op], blocker, list); | |
5559 | } | |
5560 | ||
5561 | void bdrv_op_unblock(BlockDriverState *bs, BlockOpType op, Error *reason) | |
5562 | { | |
5563 | BdrvOpBlocker *blocker, *next; | |
5564 | assert((int) op >= 0 && op < BLOCK_OP_TYPE_MAX); | |
5565 | QLIST_FOREACH_SAFE(blocker, &bs->op_blockers[op], list, next) { | |
5566 | if (blocker->reason == reason) { | |
5567 | QLIST_REMOVE(blocker, list); | |
5568 | g_free(blocker); | |
5569 | } | |
5570 | } | |
5571 | } | |
5572 | ||
5573 | void bdrv_op_block_all(BlockDriverState *bs, Error *reason) | |
5574 | { | |
5575 | int i; | |
5576 | for (i = 0; i < BLOCK_OP_TYPE_MAX; i++) { | |
5577 | bdrv_op_block(bs, i, reason); | |
5578 | } | |
5579 | } | |
5580 | ||
5581 | void bdrv_op_unblock_all(BlockDriverState *bs, Error *reason) | |
5582 | { | |
5583 | int i; | |
5584 | for (i = 0; i < BLOCK_OP_TYPE_MAX; i++) { | |
5585 | bdrv_op_unblock(bs, i, reason); | |
5586 | } | |
5587 | } | |
5588 | ||
5589 | bool bdrv_op_blocker_is_empty(BlockDriverState *bs) | |
5590 | { | |
5591 | int i; | |
5592 | ||
5593 | for (i = 0; i < BLOCK_OP_TYPE_MAX; i++) { | |
5594 | if (!QLIST_EMPTY(&bs->op_blockers[i])) { | |
5595 | return false; | |
5596 | } | |
5597 | } | |
5598 | return true; | |
5599 | } | |
5600 | ||
28a7282a LC |
5601 | void bdrv_iostatus_enable(BlockDriverState *bs) |
5602 | { | |
d6bf279e | 5603 | bs->iostatus_enabled = true; |
58e21ef5 | 5604 | bs->iostatus = BLOCK_DEVICE_IO_STATUS_OK; |
28a7282a LC |
5605 | } |
5606 | ||
5607 | /* The I/O status is only enabled if the drive explicitly | |
5608 | * enables it _and_ the VM is configured to stop on errors */ | |
5609 | bool bdrv_iostatus_is_enabled(const BlockDriverState *bs) | |
5610 | { | |
d6bf279e | 5611 | return (bs->iostatus_enabled && |
92aa5c6d PB |
5612 | (bs->on_write_error == BLOCKDEV_ON_ERROR_ENOSPC || |
5613 | bs->on_write_error == BLOCKDEV_ON_ERROR_STOP || | |
5614 | bs->on_read_error == BLOCKDEV_ON_ERROR_STOP)); | |
28a7282a LC |
5615 | } |
5616 | ||
5617 | void bdrv_iostatus_disable(BlockDriverState *bs) | |
5618 | { | |
d6bf279e | 5619 | bs->iostatus_enabled = false; |
28a7282a LC |
5620 | } |
5621 | ||
5622 | void bdrv_iostatus_reset(BlockDriverState *bs) | |
5623 | { | |
5624 | if (bdrv_iostatus_is_enabled(bs)) { | |
58e21ef5 | 5625 | bs->iostatus = BLOCK_DEVICE_IO_STATUS_OK; |
3bd293c3 PB |
5626 | if (bs->job) { |
5627 | block_job_iostatus_reset(bs->job); | |
5628 | } | |
28a7282a LC |
5629 | } |
5630 | } | |
5631 | ||
28a7282a LC |
5632 | void bdrv_iostatus_set_err(BlockDriverState *bs, int error) |
5633 | { | |
3e1caa5f PB |
5634 | assert(bdrv_iostatus_is_enabled(bs)); |
5635 | if (bs->iostatus == BLOCK_DEVICE_IO_STATUS_OK) { | |
58e21ef5 LC |
5636 | bs->iostatus = error == ENOSPC ? BLOCK_DEVICE_IO_STATUS_NOSPACE : |
5637 | BLOCK_DEVICE_IO_STATUS_FAILED; | |
28a7282a LC |
5638 | } |
5639 | } | |
5640 | ||
d92ada22 LC |
5641 | void bdrv_img_create(const char *filename, const char *fmt, |
5642 | const char *base_filename, const char *base_fmt, | |
f382d43a MR |
5643 | char *options, uint64_t img_size, int flags, |
5644 | Error **errp, bool quiet) | |
f88e1a42 | 5645 | { |
83d0521a CL |
5646 | QemuOptsList *create_opts = NULL; |
5647 | QemuOpts *opts = NULL; | |
5648 | const char *backing_fmt, *backing_file; | |
5649 | int64_t size; | |
f88e1a42 | 5650 | BlockDriver *drv, *proto_drv; |
96df67d1 | 5651 | BlockDriver *backing_drv = NULL; |
cc84d90f | 5652 | Error *local_err = NULL; |
f88e1a42 JS |
5653 | int ret = 0; |
5654 | ||
5655 | /* Find driver and parse its options */ | |
5656 | drv = bdrv_find_format(fmt); | |
5657 | if (!drv) { | |
71c79813 | 5658 | error_setg(errp, "Unknown file format '%s'", fmt); |
d92ada22 | 5659 | return; |
f88e1a42 JS |
5660 | } |
5661 | ||
b65a5e12 | 5662 | proto_drv = bdrv_find_protocol(filename, true, errp); |
f88e1a42 | 5663 | if (!proto_drv) { |
d92ada22 | 5664 | return; |
f88e1a42 JS |
5665 | } |
5666 | ||
c6149724 HR |
5667 | if (!drv->create_opts) { |
5668 | error_setg(errp, "Format driver '%s' does not support image creation", | |
5669 | drv->format_name); | |
5670 | return; | |
5671 | } | |
5672 | ||
5673 | if (!proto_drv->create_opts) { | |
5674 | error_setg(errp, "Protocol driver '%s' does not support image creation", | |
5675 | proto_drv->format_name); | |
5676 | return; | |
5677 | } | |
5678 | ||
c282e1fd CL |
5679 | create_opts = qemu_opts_append(create_opts, drv->create_opts); |
5680 | create_opts = qemu_opts_append(create_opts, proto_drv->create_opts); | |
f88e1a42 JS |
5681 | |
5682 | /* Create parameter list with default values */ | |
83d0521a | 5683 | opts = qemu_opts_create(create_opts, NULL, 0, &error_abort); |
39101f25 | 5684 | qemu_opt_set_number(opts, BLOCK_OPT_SIZE, img_size, &error_abort); |
f88e1a42 JS |
5685 | |
5686 | /* Parse -o options */ | |
5687 | if (options) { | |
dc523cd3 MA |
5688 | qemu_opts_do_parse(opts, options, NULL, &local_err); |
5689 | if (local_err) { | |
5690 | error_report_err(local_err); | |
5691 | local_err = NULL; | |
83d0521a | 5692 | error_setg(errp, "Invalid options for file format '%s'", fmt); |
f88e1a42 JS |
5693 | goto out; |
5694 | } | |
5695 | } | |
5696 | ||
5697 | if (base_filename) { | |
f43e47db | 5698 | qemu_opt_set(opts, BLOCK_OPT_BACKING_FILE, base_filename, &local_err); |
6be4194b | 5699 | if (local_err) { |
71c79813 LC |
5700 | error_setg(errp, "Backing file not supported for file format '%s'", |
5701 | fmt); | |
f88e1a42 JS |
5702 | goto out; |
5703 | } | |
5704 | } | |
5705 | ||
5706 | if (base_fmt) { | |
f43e47db | 5707 | qemu_opt_set(opts, BLOCK_OPT_BACKING_FMT, base_fmt, &local_err); |
6be4194b | 5708 | if (local_err) { |
71c79813 LC |
5709 | error_setg(errp, "Backing file format not supported for file " |
5710 | "format '%s'", fmt); | |
f88e1a42 JS |
5711 | goto out; |
5712 | } | |
5713 | } | |
5714 | ||
83d0521a CL |
5715 | backing_file = qemu_opt_get(opts, BLOCK_OPT_BACKING_FILE); |
5716 | if (backing_file) { | |
5717 | if (!strcmp(filename, backing_file)) { | |
71c79813 LC |
5718 | error_setg(errp, "Error: Trying to create an image with the " |
5719 | "same filename as the backing file"); | |
792da93a JS |
5720 | goto out; |
5721 | } | |
5722 | } | |
5723 | ||
83d0521a CL |
5724 | backing_fmt = qemu_opt_get(opts, BLOCK_OPT_BACKING_FMT); |
5725 | if (backing_fmt) { | |
5726 | backing_drv = bdrv_find_format(backing_fmt); | |
96df67d1 | 5727 | if (!backing_drv) { |
71c79813 | 5728 | error_setg(errp, "Unknown backing file format '%s'", |
83d0521a | 5729 | backing_fmt); |
f88e1a42 JS |
5730 | goto out; |
5731 | } | |
5732 | } | |
5733 | ||
5734 | // The size for the image must always be specified, with one exception: | |
5735 | // If we are using a backing file, we can obtain the size from there | |
83d0521a CL |
5736 | size = qemu_opt_get_size(opts, BLOCK_OPT_SIZE, 0); |
5737 | if (size == -1) { | |
5738 | if (backing_file) { | |
66f6b814 | 5739 | BlockDriverState *bs; |
29168018 | 5740 | char *full_backing = g_new0(char, PATH_MAX); |
52bf1e72 | 5741 | int64_t size; |
63090dac PB |
5742 | int back_flags; |
5743 | ||
29168018 HR |
5744 | bdrv_get_full_backing_filename_from_filename(filename, backing_file, |
5745 | full_backing, PATH_MAX, | |
5746 | &local_err); | |
5747 | if (local_err) { | |
5748 | g_free(full_backing); | |
5749 | goto out; | |
5750 | } | |
5751 | ||
63090dac PB |
5752 | /* backing files always opened read-only */ |
5753 | back_flags = | |
5754 | flags & ~(BDRV_O_RDWR | BDRV_O_SNAPSHOT | BDRV_O_NO_BACKING); | |
f88e1a42 | 5755 | |
f67503e5 | 5756 | bs = NULL; |
29168018 | 5757 | ret = bdrv_open(&bs, full_backing, NULL, NULL, back_flags, |
cc84d90f | 5758 | backing_drv, &local_err); |
29168018 | 5759 | g_free(full_backing); |
f88e1a42 | 5760 | if (ret < 0) { |
f88e1a42 JS |
5761 | goto out; |
5762 | } | |
52bf1e72 MA |
5763 | size = bdrv_getlength(bs); |
5764 | if (size < 0) { | |
5765 | error_setg_errno(errp, -size, "Could not get size of '%s'", | |
5766 | backing_file); | |
5767 | bdrv_unref(bs); | |
5768 | goto out; | |
5769 | } | |
f88e1a42 | 5770 | |
39101f25 | 5771 | qemu_opt_set_number(opts, BLOCK_OPT_SIZE, size, &error_abort); |
66f6b814 HR |
5772 | |
5773 | bdrv_unref(bs); | |
f88e1a42 | 5774 | } else { |
71c79813 | 5775 | error_setg(errp, "Image creation needs a size parameter"); |
f88e1a42 JS |
5776 | goto out; |
5777 | } | |
5778 | } | |
5779 | ||
f382d43a | 5780 | if (!quiet) { |
43c5d8f8 FZ |
5781 | printf("Formatting '%s', fmt=%s", filename, fmt); |
5782 | qemu_opts_print(opts, " "); | |
f382d43a MR |
5783 | puts(""); |
5784 | } | |
83d0521a | 5785 | |
c282e1fd | 5786 | ret = bdrv_create(drv, filename, opts, &local_err); |
83d0521a | 5787 | |
cc84d90f HR |
5788 | if (ret == -EFBIG) { |
5789 | /* This is generally a better message than whatever the driver would | |
5790 | * deliver (especially because of the cluster_size_hint), since that | |
5791 | * is most probably not much different from "image too large". */ | |
5792 | const char *cluster_size_hint = ""; | |
83d0521a | 5793 | if (qemu_opt_get_size(opts, BLOCK_OPT_CLUSTER_SIZE, 0)) { |
cc84d90f | 5794 | cluster_size_hint = " (try using a larger cluster size)"; |
f88e1a42 | 5795 | } |
cc84d90f HR |
5796 | error_setg(errp, "The image size is too large for file format '%s'" |
5797 | "%s", fmt, cluster_size_hint); | |
5798 | error_free(local_err); | |
5799 | local_err = NULL; | |
f88e1a42 JS |
5800 | } |
5801 | ||
5802 | out: | |
83d0521a CL |
5803 | qemu_opts_del(opts); |
5804 | qemu_opts_free(create_opts); | |
84d18f06 | 5805 | if (local_err) { |
cc84d90f HR |
5806 | error_propagate(errp, local_err); |
5807 | } | |
f88e1a42 | 5808 | } |
85d126f3 SH |
5809 | |
5810 | AioContext *bdrv_get_aio_context(BlockDriverState *bs) | |
5811 | { | |
dcd04228 SH |
5812 | return bs->aio_context; |
5813 | } | |
5814 | ||
5815 | void bdrv_detach_aio_context(BlockDriverState *bs) | |
5816 | { | |
33384421 HR |
5817 | BdrvAioNotifier *baf; |
5818 | ||
dcd04228 SH |
5819 | if (!bs->drv) { |
5820 | return; | |
5821 | } | |
5822 | ||
33384421 HR |
5823 | QLIST_FOREACH(baf, &bs->aio_notifiers, list) { |
5824 | baf->detach_aio_context(baf->opaque); | |
5825 | } | |
5826 | ||
13af91eb SH |
5827 | if (bs->io_limits_enabled) { |
5828 | throttle_detach_aio_context(&bs->throttle_state); | |
5829 | } | |
dcd04228 SH |
5830 | if (bs->drv->bdrv_detach_aio_context) { |
5831 | bs->drv->bdrv_detach_aio_context(bs); | |
5832 | } | |
5833 | if (bs->file) { | |
5834 | bdrv_detach_aio_context(bs->file); | |
5835 | } | |
5836 | if (bs->backing_hd) { | |
5837 | bdrv_detach_aio_context(bs->backing_hd); | |
5838 | } | |
5839 | ||
5840 | bs->aio_context = NULL; | |
5841 | } | |
5842 | ||
5843 | void bdrv_attach_aio_context(BlockDriverState *bs, | |
5844 | AioContext *new_context) | |
5845 | { | |
33384421 HR |
5846 | BdrvAioNotifier *ban; |
5847 | ||
dcd04228 SH |
5848 | if (!bs->drv) { |
5849 | return; | |
5850 | } | |
5851 | ||
5852 | bs->aio_context = new_context; | |
5853 | ||
5854 | if (bs->backing_hd) { | |
5855 | bdrv_attach_aio_context(bs->backing_hd, new_context); | |
5856 | } | |
5857 | if (bs->file) { | |
5858 | bdrv_attach_aio_context(bs->file, new_context); | |
5859 | } | |
5860 | if (bs->drv->bdrv_attach_aio_context) { | |
5861 | bs->drv->bdrv_attach_aio_context(bs, new_context); | |
5862 | } | |
13af91eb SH |
5863 | if (bs->io_limits_enabled) { |
5864 | throttle_attach_aio_context(&bs->throttle_state, new_context); | |
5865 | } | |
33384421 HR |
5866 | |
5867 | QLIST_FOREACH(ban, &bs->aio_notifiers, list) { | |
5868 | ban->attached_aio_context(new_context, ban->opaque); | |
5869 | } | |
dcd04228 SH |
5870 | } |
5871 | ||
5872 | void bdrv_set_aio_context(BlockDriverState *bs, AioContext *new_context) | |
5873 | { | |
5874 | bdrv_drain_all(); /* ensure there are no in-flight requests */ | |
5875 | ||
5876 | bdrv_detach_aio_context(bs); | |
5877 | ||
5878 | /* This function executes in the old AioContext so acquire the new one in | |
5879 | * case it runs in a different thread. | |
5880 | */ | |
5881 | aio_context_acquire(new_context); | |
5882 | bdrv_attach_aio_context(bs, new_context); | |
5883 | aio_context_release(new_context); | |
85d126f3 | 5884 | } |
d616b224 | 5885 | |
33384421 HR |
5886 | void bdrv_add_aio_context_notifier(BlockDriverState *bs, |
5887 | void (*attached_aio_context)(AioContext *new_context, void *opaque), | |
5888 | void (*detach_aio_context)(void *opaque), void *opaque) | |
5889 | { | |
5890 | BdrvAioNotifier *ban = g_new(BdrvAioNotifier, 1); | |
5891 | *ban = (BdrvAioNotifier){ | |
5892 | .attached_aio_context = attached_aio_context, | |
5893 | .detach_aio_context = detach_aio_context, | |
5894 | .opaque = opaque | |
5895 | }; | |
5896 | ||
5897 | QLIST_INSERT_HEAD(&bs->aio_notifiers, ban, list); | |
5898 | } | |
5899 | ||
5900 | void bdrv_remove_aio_context_notifier(BlockDriverState *bs, | |
5901 | void (*attached_aio_context)(AioContext *, | |
5902 | void *), | |
5903 | void (*detach_aio_context)(void *), | |
5904 | void *opaque) | |
5905 | { | |
5906 | BdrvAioNotifier *ban, *ban_next; | |
5907 | ||
5908 | QLIST_FOREACH_SAFE(ban, &bs->aio_notifiers, list, ban_next) { | |
5909 | if (ban->attached_aio_context == attached_aio_context && | |
5910 | ban->detach_aio_context == detach_aio_context && | |
5911 | ban->opaque == opaque) | |
5912 | { | |
5913 | QLIST_REMOVE(ban, list); | |
5914 | g_free(ban); | |
5915 | ||
5916 | return; | |
5917 | } | |
5918 | } | |
5919 | ||
5920 | abort(); | |
5921 | } | |
5922 | ||
d616b224 SH |
5923 | void bdrv_add_before_write_notifier(BlockDriverState *bs, |
5924 | NotifierWithReturn *notifier) | |
5925 | { | |
5926 | notifier_with_return_list_add(&bs->before_write_notifiers, notifier); | |
5927 | } | |
6f176b48 | 5928 | |
77485434 HR |
5929 | int bdrv_amend_options(BlockDriverState *bs, QemuOpts *opts, |
5930 | BlockDriverAmendStatusCB *status_cb) | |
6f176b48 | 5931 | { |
c282e1fd | 5932 | if (!bs->drv->bdrv_amend_options) { |
6f176b48 HR |
5933 | return -ENOTSUP; |
5934 | } | |
77485434 | 5935 | return bs->drv->bdrv_amend_options(bs, opts, status_cb); |
6f176b48 | 5936 | } |
f6186f49 | 5937 | |
b5042a36 BC |
5938 | /* This function will be called by the bdrv_recurse_is_first_non_filter method |
5939 | * of block filter and by bdrv_is_first_non_filter. | |
5940 | * It is used to test if the given bs is the candidate or recurse more in the | |
5941 | * node graph. | |
212a5a8f | 5942 | */ |
b5042a36 | 5943 | bool bdrv_recurse_is_first_non_filter(BlockDriverState *bs, |
212a5a8f | 5944 | BlockDriverState *candidate) |
f6186f49 | 5945 | { |
b5042a36 BC |
5946 | /* return false if basic checks fails */ |
5947 | if (!bs || !bs->drv) { | |
212a5a8f | 5948 | return false; |
f6186f49 BC |
5949 | } |
5950 | ||
b5042a36 BC |
5951 | /* the code reached a non block filter driver -> check if the bs is |
5952 | * the same as the candidate. It's the recursion termination condition. | |
5953 | */ | |
5954 | if (!bs->drv->is_filter) { | |
5955 | return bs == candidate; | |
212a5a8f | 5956 | } |
b5042a36 | 5957 | /* Down this path the driver is a block filter driver */ |
212a5a8f | 5958 | |
b5042a36 BC |
5959 | /* If the block filter recursion method is defined use it to recurse down |
5960 | * the node graph. | |
5961 | */ | |
5962 | if (bs->drv->bdrv_recurse_is_first_non_filter) { | |
212a5a8f | 5963 | return bs->drv->bdrv_recurse_is_first_non_filter(bs, candidate); |
f6186f49 BC |
5964 | } |
5965 | ||
b5042a36 BC |
5966 | /* the driver is a block filter but don't allow to recurse -> return false |
5967 | */ | |
5968 | return false; | |
f6186f49 BC |
5969 | } |
5970 | ||
212a5a8f BC |
5971 | /* This function checks if the candidate is the first non filter bs down it's |
5972 | * bs chain. Since we don't have pointers to parents it explore all bs chains | |
5973 | * from the top. Some filters can choose not to pass down the recursion. | |
5974 | */ | |
5975 | bool bdrv_is_first_non_filter(BlockDriverState *candidate) | |
f6186f49 | 5976 | { |
212a5a8f BC |
5977 | BlockDriverState *bs; |
5978 | ||
5979 | /* walk down the bs forest recursively */ | |
5980 | QTAILQ_FOREACH(bs, &bdrv_states, device_list) { | |
5981 | bool perm; | |
5982 | ||
b5042a36 | 5983 | /* try to recurse in this top level bs */ |
e6dc8a1f | 5984 | perm = bdrv_recurse_is_first_non_filter(bs, candidate); |
212a5a8f BC |
5985 | |
5986 | /* candidate is the first non filter */ | |
5987 | if (perm) { | |
5988 | return true; | |
5989 | } | |
5990 | } | |
5991 | ||
5992 | return false; | |
f6186f49 | 5993 | } |
09158f00 BC |
5994 | |
5995 | BlockDriverState *check_to_replace_node(const char *node_name, Error **errp) | |
5996 | { | |
5997 | BlockDriverState *to_replace_bs = bdrv_find_node(node_name); | |
5a7e7a0b SH |
5998 | AioContext *aio_context; |
5999 | ||
09158f00 BC |
6000 | if (!to_replace_bs) { |
6001 | error_setg(errp, "Node name '%s' not found", node_name); | |
6002 | return NULL; | |
6003 | } | |
6004 | ||
5a7e7a0b SH |
6005 | aio_context = bdrv_get_aio_context(to_replace_bs); |
6006 | aio_context_acquire(aio_context); | |
6007 | ||
09158f00 | 6008 | if (bdrv_op_is_blocked(to_replace_bs, BLOCK_OP_TYPE_REPLACE, errp)) { |
5a7e7a0b SH |
6009 | to_replace_bs = NULL; |
6010 | goto out; | |
09158f00 BC |
6011 | } |
6012 | ||
6013 | /* We don't want arbitrary node of the BDS chain to be replaced only the top | |
6014 | * most non filter in order to prevent data corruption. | |
6015 | * Another benefit is that this tests exclude backing files which are | |
6016 | * blocked by the backing blockers. | |
6017 | */ | |
6018 | if (!bdrv_is_first_non_filter(to_replace_bs)) { | |
6019 | error_setg(errp, "Only top most non filter can be replaced"); | |
5a7e7a0b SH |
6020 | to_replace_bs = NULL; |
6021 | goto out; | |
09158f00 BC |
6022 | } |
6023 | ||
5a7e7a0b SH |
6024 | out: |
6025 | aio_context_release(aio_context); | |
09158f00 BC |
6026 | return to_replace_bs; |
6027 | } | |
448ad91d ML |
6028 | |
6029 | void bdrv_io_plug(BlockDriverState *bs) | |
6030 | { | |
6031 | BlockDriver *drv = bs->drv; | |
6032 | if (drv && drv->bdrv_io_plug) { | |
6033 | drv->bdrv_io_plug(bs); | |
6034 | } else if (bs->file) { | |
6035 | bdrv_io_plug(bs->file); | |
6036 | } | |
6037 | } | |
6038 | ||
6039 | void bdrv_io_unplug(BlockDriverState *bs) | |
6040 | { | |
6041 | BlockDriver *drv = bs->drv; | |
6042 | if (drv && drv->bdrv_io_unplug) { | |
6043 | drv->bdrv_io_unplug(bs); | |
6044 | } else if (bs->file) { | |
6045 | bdrv_io_unplug(bs->file); | |
6046 | } | |
6047 | } | |
6048 | ||
6049 | void bdrv_flush_io_queue(BlockDriverState *bs) | |
6050 | { | |
6051 | BlockDriver *drv = bs->drv; | |
6052 | if (drv && drv->bdrv_flush_io_queue) { | |
6053 | drv->bdrv_flush_io_queue(bs); | |
6054 | } else if (bs->file) { | |
6055 | bdrv_flush_io_queue(bs->file); | |
6056 | } | |
6057 | } | |
91af7014 HR |
6058 | |
6059 | static bool append_open_options(QDict *d, BlockDriverState *bs) | |
6060 | { | |
6061 | const QDictEntry *entry; | |
6062 | bool found_any = false; | |
6063 | ||
6064 | for (entry = qdict_first(bs->options); entry; | |
6065 | entry = qdict_next(bs->options, entry)) | |
6066 | { | |
6067 | /* Only take options for this level and exclude all non-driver-specific | |
6068 | * options */ | |
6069 | if (!strchr(qdict_entry_key(entry), '.') && | |
6070 | strcmp(qdict_entry_key(entry), "node-name")) | |
6071 | { | |
6072 | qobject_incref(qdict_entry_value(entry)); | |
6073 | qdict_put_obj(d, qdict_entry_key(entry), qdict_entry_value(entry)); | |
6074 | found_any = true; | |
6075 | } | |
6076 | } | |
6077 | ||
6078 | return found_any; | |
6079 | } | |
6080 | ||
6081 | /* Updates the following BDS fields: | |
6082 | * - exact_filename: A filename which may be used for opening a block device | |
6083 | * which (mostly) equals the given BDS (even without any | |
6084 | * other options; so reading and writing must return the same | |
6085 | * results, but caching etc. may be different) | |
6086 | * - full_open_options: Options which, when given when opening a block device | |
6087 | * (without a filename), result in a BDS (mostly) | |
6088 | * equalling the given one | |
6089 | * - filename: If exact_filename is set, it is copied here. Otherwise, | |
6090 | * full_open_options is converted to a JSON object, prefixed with | |
6091 | * "json:" (for use through the JSON pseudo protocol) and put here. | |
6092 | */ | |
6093 | void bdrv_refresh_filename(BlockDriverState *bs) | |
6094 | { | |
6095 | BlockDriver *drv = bs->drv; | |
6096 | QDict *opts; | |
6097 | ||
6098 | if (!drv) { | |
6099 | return; | |
6100 | } | |
6101 | ||
6102 | /* This BDS's file name will most probably depend on its file's name, so | |
6103 | * refresh that first */ | |
6104 | if (bs->file) { | |
6105 | bdrv_refresh_filename(bs->file); | |
6106 | } | |
6107 | ||
6108 | if (drv->bdrv_refresh_filename) { | |
6109 | /* Obsolete information is of no use here, so drop the old file name | |
6110 | * information before refreshing it */ | |
6111 | bs->exact_filename[0] = '\0'; | |
6112 | if (bs->full_open_options) { | |
6113 | QDECREF(bs->full_open_options); | |
6114 | bs->full_open_options = NULL; | |
6115 | } | |
6116 | ||
6117 | drv->bdrv_refresh_filename(bs); | |
6118 | } else if (bs->file) { | |
6119 | /* Try to reconstruct valid information from the underlying file */ | |
6120 | bool has_open_options; | |
6121 | ||
6122 | bs->exact_filename[0] = '\0'; | |
6123 | if (bs->full_open_options) { | |
6124 | QDECREF(bs->full_open_options); | |
6125 | bs->full_open_options = NULL; | |
6126 | } | |
6127 | ||
6128 | opts = qdict_new(); | |
6129 | has_open_options = append_open_options(opts, bs); | |
6130 | ||
6131 | /* If no specific options have been given for this BDS, the filename of | |
6132 | * the underlying file should suffice for this one as well */ | |
6133 | if (bs->file->exact_filename[0] && !has_open_options) { | |
6134 | strcpy(bs->exact_filename, bs->file->exact_filename); | |
6135 | } | |
6136 | /* Reconstructing the full options QDict is simple for most format block | |
6137 | * drivers, as long as the full options are known for the underlying | |
6138 | * file BDS. The full options QDict of that file BDS should somehow | |
6139 | * contain a representation of the filename, therefore the following | |
6140 | * suffices without querying the (exact_)filename of this BDS. */ | |
6141 | if (bs->file->full_open_options) { | |
6142 | qdict_put_obj(opts, "driver", | |
6143 | QOBJECT(qstring_from_str(drv->format_name))); | |
6144 | QINCREF(bs->file->full_open_options); | |
6145 | qdict_put_obj(opts, "file", QOBJECT(bs->file->full_open_options)); | |
6146 | ||
6147 | bs->full_open_options = opts; | |
6148 | } else { | |
6149 | QDECREF(opts); | |
6150 | } | |
6151 | } else if (!bs->full_open_options && qdict_size(bs->options)) { | |
6152 | /* There is no underlying file BDS (at least referenced by BDS.file), | |
6153 | * so the full options QDict should be equal to the options given | |
6154 | * specifically for this block device when it was opened (plus the | |
6155 | * driver specification). | |
6156 | * Because those options don't change, there is no need to update | |
6157 | * full_open_options when it's already set. */ | |
6158 | ||
6159 | opts = qdict_new(); | |
6160 | append_open_options(opts, bs); | |
6161 | qdict_put_obj(opts, "driver", | |
6162 | QOBJECT(qstring_from_str(drv->format_name))); | |
6163 | ||
6164 | if (bs->exact_filename[0]) { | |
6165 | /* This may not work for all block protocol drivers (some may | |
6166 | * require this filename to be parsed), but we have to find some | |
6167 | * default solution here, so just include it. If some block driver | |
6168 | * does not support pure options without any filename at all or | |
6169 | * needs some special format of the options QDict, it needs to | |
6170 | * implement the driver-specific bdrv_refresh_filename() function. | |
6171 | */ | |
6172 | qdict_put_obj(opts, "filename", | |
6173 | QOBJECT(qstring_from_str(bs->exact_filename))); | |
6174 | } | |
6175 | ||
6176 | bs->full_open_options = opts; | |
6177 | } | |
6178 | ||
6179 | if (bs->exact_filename[0]) { | |
6180 | pstrcpy(bs->filename, sizeof(bs->filename), bs->exact_filename); | |
6181 | } else if (bs->full_open_options) { | |
6182 | QString *json = qobject_to_json(QOBJECT(bs->full_open_options)); | |
6183 | snprintf(bs->filename, sizeof(bs->filename), "json:%s", | |
6184 | qstring_get_str(json)); | |
6185 | QDECREF(json); | |
6186 | } | |
6187 | } | |
5366d0c8 BC |
6188 | |
6189 | /* This accessor function purpose is to allow the device models to access the | |
6190 | * BlockAcctStats structure embedded inside a BlockDriverState without being | |
6191 | * aware of the BlockDriverState structure layout. | |
6192 | * It will go away when the BlockAcctStats structure will be moved inside | |
6193 | * the device models. | |
6194 | */ | |
6195 | BlockAcctStats *bdrv_get_stats(BlockDriverState *bs) | |
6196 | { | |
6197 | return &bs->stats; | |
6198 | } |