]>
Commit | Line | Data |
---|---|---|
fc01f7e7 FB |
1 | /* |
2 | * QEMU System Emulator block driver | |
5fafdf24 | 3 | * |
fc01f7e7 | 4 | * Copyright (c) 2003 Fabrice Bellard |
5fafdf24 | 5 | * |
fc01f7e7 FB |
6 | * Permission is hereby granted, free of charge, to any person obtaining a copy |
7 | * of this software and associated documentation files (the "Software"), to deal | |
8 | * in the Software without restriction, including without limitation the rights | |
9 | * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell | |
10 | * copies of the Software, and to permit persons to whom the Software is | |
11 | * furnished to do so, subject to the following conditions: | |
12 | * | |
13 | * The above copyright notice and this permission notice shall be included in | |
14 | * all copies or substantial portions of the Software. | |
15 | * | |
16 | * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR | |
17 | * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, | |
18 | * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL | |
19 | * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER | |
20 | * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, | |
21 | * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN | |
22 | * THE SOFTWARE. | |
23 | */ | |
3990d09a | 24 | #include "config-host.h" |
faf07963 | 25 | #include "qemu-common.h" |
6d519a5f | 26 | #include "trace.h" |
83c9089e | 27 | #include "monitor/monitor.h" |
737e150e PB |
28 | #include "block/block_int.h" |
29 | #include "block/blockjob.h" | |
1de7afc9 | 30 | #include "qemu/module.h" |
7b1b5d19 | 31 | #include "qapi/qmp/qjson.h" |
9c17d615 | 32 | #include "sysemu/sysemu.h" |
1de7afc9 | 33 | #include "qemu/notify.h" |
737e150e | 34 | #include "block/coroutine.h" |
b2023818 | 35 | #include "qmp-commands.h" |
1de7afc9 | 36 | #include "qemu/timer.h" |
fc01f7e7 | 37 | |
71e72a19 | 38 | #ifdef CONFIG_BSD |
7674e7bf FB |
39 | #include <sys/types.h> |
40 | #include <sys/stat.h> | |
41 | #include <sys/ioctl.h> | |
72cf2d4f | 42 | #include <sys/queue.h> |
c5e97233 | 43 | #ifndef __DragonFly__ |
7674e7bf FB |
44 | #include <sys/disk.h> |
45 | #endif | |
c5e97233 | 46 | #endif |
7674e7bf | 47 | |
49dc768d AL |
48 | #ifdef _WIN32 |
49 | #include <windows.h> | |
50 | #endif | |
51 | ||
1c9805a3 SH |
52 | #define NOT_DONE 0x7fffffff /* used while emulated sync operation in progress */ |
53 | ||
470c0504 SH |
54 | typedef enum { |
55 | BDRV_REQ_COPY_ON_READ = 0x1, | |
f08f2dda | 56 | BDRV_REQ_ZERO_WRITE = 0x2, |
470c0504 SH |
57 | } BdrvRequestFlags; |
58 | ||
7d4b4ba5 | 59 | static void bdrv_dev_change_media_cb(BlockDriverState *bs, bool load); |
f141eafe AL |
60 | static BlockDriverAIOCB *bdrv_aio_readv_em(BlockDriverState *bs, |
61 | int64_t sector_num, QEMUIOVector *qiov, int nb_sectors, | |
c87c0672 | 62 | BlockDriverCompletionFunc *cb, void *opaque); |
f141eafe AL |
63 | static BlockDriverAIOCB *bdrv_aio_writev_em(BlockDriverState *bs, |
64 | int64_t sector_num, QEMUIOVector *qiov, int nb_sectors, | |
ce1a14dc | 65 | BlockDriverCompletionFunc *cb, void *opaque); |
f9f05dc5 KW |
66 | static int coroutine_fn bdrv_co_readv_em(BlockDriverState *bs, |
67 | int64_t sector_num, int nb_sectors, | |
68 | QEMUIOVector *iov); | |
69 | static int coroutine_fn bdrv_co_writev_em(BlockDriverState *bs, | |
70 | int64_t sector_num, int nb_sectors, | |
71 | QEMUIOVector *iov); | |
c5fbe571 | 72 | static int coroutine_fn bdrv_co_do_readv(BlockDriverState *bs, |
470c0504 SH |
73 | int64_t sector_num, int nb_sectors, QEMUIOVector *qiov, |
74 | BdrvRequestFlags flags); | |
1c9805a3 | 75 | static int coroutine_fn bdrv_co_do_writev(BlockDriverState *bs, |
f08f2dda SH |
76 | int64_t sector_num, int nb_sectors, QEMUIOVector *qiov, |
77 | BdrvRequestFlags flags); | |
b2a61371 SH |
78 | static BlockDriverAIOCB *bdrv_co_aio_rw_vector(BlockDriverState *bs, |
79 | int64_t sector_num, | |
80 | QEMUIOVector *qiov, | |
81 | int nb_sectors, | |
82 | BlockDriverCompletionFunc *cb, | |
83 | void *opaque, | |
8c5873d6 | 84 | bool is_write); |
b2a61371 | 85 | static void coroutine_fn bdrv_co_do_rw(void *opaque); |
621f0589 KW |
86 | static int coroutine_fn bdrv_co_do_write_zeroes(BlockDriverState *bs, |
87 | int64_t sector_num, int nb_sectors); | |
ec530c81 | 88 | |
98f90dba ZYW |
89 | static bool bdrv_exceed_bps_limits(BlockDriverState *bs, int nb_sectors, |
90 | bool is_write, double elapsed_time, uint64_t *wait); | |
91 | static bool bdrv_exceed_iops_limits(BlockDriverState *bs, bool is_write, | |
92 | double elapsed_time, uint64_t *wait); | |
93 | static bool bdrv_exceed_io_limits(BlockDriverState *bs, int nb_sectors, | |
94 | bool is_write, int64_t *wait); | |
95 | ||
1b7bdbc1 SH |
96 | static QTAILQ_HEAD(, BlockDriverState) bdrv_states = |
97 | QTAILQ_HEAD_INITIALIZER(bdrv_states); | |
7ee930d0 | 98 | |
8a22f02a SH |
99 | static QLIST_HEAD(, BlockDriver) bdrv_drivers = |
100 | QLIST_HEAD_INITIALIZER(bdrv_drivers); | |
ea2384d3 | 101 | |
eb852011 MA |
102 | /* If non-zero, use only whitelisted block drivers */ |
103 | static int use_bdrv_whitelist; | |
104 | ||
9e0b22f4 SH |
105 | #ifdef _WIN32 |
106 | static int is_windows_drive_prefix(const char *filename) | |
107 | { | |
108 | return (((filename[0] >= 'a' && filename[0] <= 'z') || | |
109 | (filename[0] >= 'A' && filename[0] <= 'Z')) && | |
110 | filename[1] == ':'); | |
111 | } | |
112 | ||
113 | int is_windows_drive(const char *filename) | |
114 | { | |
115 | if (is_windows_drive_prefix(filename) && | |
116 | filename[2] == '\0') | |
117 | return 1; | |
118 | if (strstart(filename, "\\\\.\\", NULL) || | |
119 | strstart(filename, "//./", NULL)) | |
120 | return 1; | |
121 | return 0; | |
122 | } | |
123 | #endif | |
124 | ||
0563e191 | 125 | /* throttling disk I/O limits */ |
98f90dba ZYW |
126 | void bdrv_io_limits_disable(BlockDriverState *bs) |
127 | { | |
128 | bs->io_limits_enabled = false; | |
129 | ||
130 | while (qemu_co_queue_next(&bs->throttled_reqs)); | |
131 | ||
132 | if (bs->block_timer) { | |
133 | qemu_del_timer(bs->block_timer); | |
134 | qemu_free_timer(bs->block_timer); | |
135 | bs->block_timer = NULL; | |
136 | } | |
137 | ||
138 | bs->slice_start = 0; | |
139 | bs->slice_end = 0; | |
98f90dba ZYW |
140 | } |
141 | ||
0563e191 ZYW |
142 | static void bdrv_block_timer(void *opaque) |
143 | { | |
144 | BlockDriverState *bs = opaque; | |
145 | ||
146 | qemu_co_queue_next(&bs->throttled_reqs); | |
147 | } | |
148 | ||
149 | void bdrv_io_limits_enable(BlockDriverState *bs) | |
150 | { | |
151 | qemu_co_queue_init(&bs->throttled_reqs); | |
152 | bs->block_timer = qemu_new_timer_ns(vm_clock, bdrv_block_timer, bs); | |
0563e191 ZYW |
153 | bs->io_limits_enabled = true; |
154 | } | |
155 | ||
156 | bool bdrv_io_limits_enabled(BlockDriverState *bs) | |
157 | { | |
158 | BlockIOLimit *io_limits = &bs->io_limits; | |
159 | return io_limits->bps[BLOCK_IO_LIMIT_READ] | |
160 | || io_limits->bps[BLOCK_IO_LIMIT_WRITE] | |
161 | || io_limits->bps[BLOCK_IO_LIMIT_TOTAL] | |
162 | || io_limits->iops[BLOCK_IO_LIMIT_READ] | |
163 | || io_limits->iops[BLOCK_IO_LIMIT_WRITE] | |
164 | || io_limits->iops[BLOCK_IO_LIMIT_TOTAL]; | |
165 | } | |
166 | ||
98f90dba ZYW |
167 | static void bdrv_io_limits_intercept(BlockDriverState *bs, |
168 | bool is_write, int nb_sectors) | |
169 | { | |
170 | int64_t wait_time = -1; | |
171 | ||
172 | if (!qemu_co_queue_empty(&bs->throttled_reqs)) { | |
173 | qemu_co_queue_wait(&bs->throttled_reqs); | |
174 | } | |
175 | ||
176 | /* In fact, we hope to keep each request's timing, in FIFO mode. The next | |
177 | * throttled requests will not be dequeued until the current request is | |
178 | * allowed to be serviced. So if the current request still exceeds the | |
179 | * limits, it will be inserted to the head. All requests followed it will | |
180 | * be still in throttled_reqs queue. | |
181 | */ | |
182 | ||
183 | while (bdrv_exceed_io_limits(bs, nb_sectors, is_write, &wait_time)) { | |
184 | qemu_mod_timer(bs->block_timer, | |
185 | wait_time + qemu_get_clock_ns(vm_clock)); | |
186 | qemu_co_queue_wait_insert_head(&bs->throttled_reqs); | |
187 | } | |
188 | ||
189 | qemu_co_queue_next(&bs->throttled_reqs); | |
190 | } | |
191 | ||
9e0b22f4 SH |
192 | /* check if the path starts with "<protocol>:" */ |
193 | static int path_has_protocol(const char *path) | |
194 | { | |
947995c0 PB |
195 | const char *p; |
196 | ||
9e0b22f4 SH |
197 | #ifdef _WIN32 |
198 | if (is_windows_drive(path) || | |
199 | is_windows_drive_prefix(path)) { | |
200 | return 0; | |
201 | } | |
947995c0 PB |
202 | p = path + strcspn(path, ":/\\"); |
203 | #else | |
204 | p = path + strcspn(path, ":/"); | |
9e0b22f4 SH |
205 | #endif |
206 | ||
947995c0 | 207 | return *p == ':'; |
9e0b22f4 SH |
208 | } |
209 | ||
83f64091 | 210 | int path_is_absolute(const char *path) |
3b0d4f61 | 211 | { |
21664424 FB |
212 | #ifdef _WIN32 |
213 | /* specific case for names like: "\\.\d:" */ | |
f53f4da9 | 214 | if (is_windows_drive(path) || is_windows_drive_prefix(path)) { |
21664424 | 215 | return 1; |
f53f4da9 PB |
216 | } |
217 | return (*path == '/' || *path == '\\'); | |
3b9f94e1 | 218 | #else |
f53f4da9 | 219 | return (*path == '/'); |
3b9f94e1 | 220 | #endif |
3b0d4f61 FB |
221 | } |
222 | ||
83f64091 FB |
223 | /* if filename is absolute, just copy it to dest. Otherwise, build a |
224 | path to it by considering it is relative to base_path. URL are | |
225 | supported. */ | |
226 | void path_combine(char *dest, int dest_size, | |
227 | const char *base_path, | |
228 | const char *filename) | |
3b0d4f61 | 229 | { |
83f64091 FB |
230 | const char *p, *p1; |
231 | int len; | |
232 | ||
233 | if (dest_size <= 0) | |
234 | return; | |
235 | if (path_is_absolute(filename)) { | |
236 | pstrcpy(dest, dest_size, filename); | |
237 | } else { | |
238 | p = strchr(base_path, ':'); | |
239 | if (p) | |
240 | p++; | |
241 | else | |
242 | p = base_path; | |
3b9f94e1 FB |
243 | p1 = strrchr(base_path, '/'); |
244 | #ifdef _WIN32 | |
245 | { | |
246 | const char *p2; | |
247 | p2 = strrchr(base_path, '\\'); | |
248 | if (!p1 || p2 > p1) | |
249 | p1 = p2; | |
250 | } | |
251 | #endif | |
83f64091 FB |
252 | if (p1) |
253 | p1++; | |
254 | else | |
255 | p1 = base_path; | |
256 | if (p1 > p) | |
257 | p = p1; | |
258 | len = p - base_path; | |
259 | if (len > dest_size - 1) | |
260 | len = dest_size - 1; | |
261 | memcpy(dest, base_path, len); | |
262 | dest[len] = '\0'; | |
263 | pstrcat(dest, dest_size, filename); | |
3b0d4f61 | 264 | } |
3b0d4f61 FB |
265 | } |
266 | ||
dc5a1371 PB |
267 | void bdrv_get_full_backing_filename(BlockDriverState *bs, char *dest, size_t sz) |
268 | { | |
269 | if (bs->backing_file[0] == '\0' || path_has_protocol(bs->backing_file)) { | |
270 | pstrcpy(dest, sz, bs->backing_file); | |
271 | } else { | |
272 | path_combine(dest, sz, bs->filename, bs->backing_file); | |
273 | } | |
274 | } | |
275 | ||
5efa9d5a | 276 | void bdrv_register(BlockDriver *bdrv) |
ea2384d3 | 277 | { |
8c5873d6 SH |
278 | /* Block drivers without coroutine functions need emulation */ |
279 | if (!bdrv->bdrv_co_readv) { | |
f9f05dc5 KW |
280 | bdrv->bdrv_co_readv = bdrv_co_readv_em; |
281 | bdrv->bdrv_co_writev = bdrv_co_writev_em; | |
282 | ||
f8c35c1d SH |
283 | /* bdrv_co_readv_em()/brdv_co_writev_em() work in terms of aio, so if |
284 | * the block driver lacks aio we need to emulate that too. | |
285 | */ | |
f9f05dc5 KW |
286 | if (!bdrv->bdrv_aio_readv) { |
287 | /* add AIO emulation layer */ | |
288 | bdrv->bdrv_aio_readv = bdrv_aio_readv_em; | |
289 | bdrv->bdrv_aio_writev = bdrv_aio_writev_em; | |
f9f05dc5 | 290 | } |
83f64091 | 291 | } |
b2e12bc6 | 292 | |
8a22f02a | 293 | QLIST_INSERT_HEAD(&bdrv_drivers, bdrv, list); |
ea2384d3 | 294 | } |
b338082b FB |
295 | |
296 | /* create a new block device (by default it is empty) */ | |
297 | BlockDriverState *bdrv_new(const char *device_name) | |
298 | { | |
1b7bdbc1 | 299 | BlockDriverState *bs; |
b338082b | 300 | |
7267c094 | 301 | bs = g_malloc0(sizeof(BlockDriverState)); |
b338082b | 302 | pstrcpy(bs->device_name, sizeof(bs->device_name), device_name); |
ea2384d3 | 303 | if (device_name[0] != '\0') { |
1b7bdbc1 | 304 | QTAILQ_INSERT_TAIL(&bdrv_states, bs, list); |
ea2384d3 | 305 | } |
28a7282a | 306 | bdrv_iostatus_disable(bs); |
d7d512f6 | 307 | notifier_list_init(&bs->close_notifiers); |
d616b224 | 308 | notifier_with_return_list_init(&bs->before_write_notifiers); |
d7d512f6 | 309 | |
b338082b FB |
310 | return bs; |
311 | } | |
312 | ||
d7d512f6 PB |
313 | void bdrv_add_close_notifier(BlockDriverState *bs, Notifier *notify) |
314 | { | |
315 | notifier_list_add(&bs->close_notifiers, notify); | |
316 | } | |
317 | ||
ea2384d3 FB |
318 | BlockDriver *bdrv_find_format(const char *format_name) |
319 | { | |
320 | BlockDriver *drv1; | |
8a22f02a SH |
321 | QLIST_FOREACH(drv1, &bdrv_drivers, list) { |
322 | if (!strcmp(drv1->format_name, format_name)) { | |
ea2384d3 | 323 | return drv1; |
8a22f02a | 324 | } |
ea2384d3 FB |
325 | } |
326 | return NULL; | |
327 | } | |
328 | ||
b64ec4e4 | 329 | static int bdrv_is_whitelisted(BlockDriver *drv, bool read_only) |
eb852011 | 330 | { |
b64ec4e4 FZ |
331 | static const char *whitelist_rw[] = { |
332 | CONFIG_BDRV_RW_WHITELIST | |
333 | }; | |
334 | static const char *whitelist_ro[] = { | |
335 | CONFIG_BDRV_RO_WHITELIST | |
eb852011 MA |
336 | }; |
337 | const char **p; | |
338 | ||
b64ec4e4 | 339 | if (!whitelist_rw[0] && !whitelist_ro[0]) { |
eb852011 | 340 | return 1; /* no whitelist, anything goes */ |
b64ec4e4 | 341 | } |
eb852011 | 342 | |
b64ec4e4 | 343 | for (p = whitelist_rw; *p; p++) { |
eb852011 MA |
344 | if (!strcmp(drv->format_name, *p)) { |
345 | return 1; | |
346 | } | |
347 | } | |
b64ec4e4 FZ |
348 | if (read_only) { |
349 | for (p = whitelist_ro; *p; p++) { | |
350 | if (!strcmp(drv->format_name, *p)) { | |
351 | return 1; | |
352 | } | |
353 | } | |
354 | } | |
eb852011 MA |
355 | return 0; |
356 | } | |
357 | ||
b64ec4e4 FZ |
358 | BlockDriver *bdrv_find_whitelisted_format(const char *format_name, |
359 | bool read_only) | |
eb852011 MA |
360 | { |
361 | BlockDriver *drv = bdrv_find_format(format_name); | |
b64ec4e4 | 362 | return drv && bdrv_is_whitelisted(drv, read_only) ? drv : NULL; |
eb852011 MA |
363 | } |
364 | ||
5b7e1542 ZYW |
365 | typedef struct CreateCo { |
366 | BlockDriver *drv; | |
367 | char *filename; | |
368 | QEMUOptionParameter *options; | |
369 | int ret; | |
370 | } CreateCo; | |
371 | ||
372 | static void coroutine_fn bdrv_create_co_entry(void *opaque) | |
373 | { | |
374 | CreateCo *cco = opaque; | |
375 | assert(cco->drv); | |
376 | ||
377 | cco->ret = cco->drv->bdrv_create(cco->filename, cco->options); | |
378 | } | |
379 | ||
0e7e1989 KW |
380 | int bdrv_create(BlockDriver *drv, const char* filename, |
381 | QEMUOptionParameter *options) | |
ea2384d3 | 382 | { |
5b7e1542 ZYW |
383 | int ret; |
384 | ||
385 | Coroutine *co; | |
386 | CreateCo cco = { | |
387 | .drv = drv, | |
388 | .filename = g_strdup(filename), | |
389 | .options = options, | |
390 | .ret = NOT_DONE, | |
391 | }; | |
392 | ||
393 | if (!drv->bdrv_create) { | |
80168bff LC |
394 | ret = -ENOTSUP; |
395 | goto out; | |
5b7e1542 ZYW |
396 | } |
397 | ||
398 | if (qemu_in_coroutine()) { | |
399 | /* Fast-path if already in coroutine context */ | |
400 | bdrv_create_co_entry(&cco); | |
401 | } else { | |
402 | co = qemu_coroutine_create(bdrv_create_co_entry); | |
403 | qemu_coroutine_enter(co, &cco); | |
404 | while (cco.ret == NOT_DONE) { | |
405 | qemu_aio_wait(); | |
406 | } | |
407 | } | |
408 | ||
409 | ret = cco.ret; | |
0e7e1989 | 410 | |
80168bff LC |
411 | out: |
412 | g_free(cco.filename); | |
5b7e1542 | 413 | return ret; |
ea2384d3 FB |
414 | } |
415 | ||
84a12e66 CH |
416 | int bdrv_create_file(const char* filename, QEMUOptionParameter *options) |
417 | { | |
418 | BlockDriver *drv; | |
419 | ||
98289620 | 420 | drv = bdrv_find_protocol(filename, true); |
84a12e66 | 421 | if (drv == NULL) { |
16905d71 | 422 | return -ENOENT; |
84a12e66 CH |
423 | } |
424 | ||
425 | return bdrv_create(drv, filename, options); | |
426 | } | |
427 | ||
eba25057 JM |
428 | /* |
429 | * Create a uniquely-named empty temporary file. | |
430 | * Return 0 upon success, otherwise a negative errno value. | |
431 | */ | |
432 | int get_tmp_filename(char *filename, int size) | |
d5249393 | 433 | { |
eba25057 | 434 | #ifdef _WIN32 |
3b9f94e1 | 435 | char temp_dir[MAX_PATH]; |
eba25057 JM |
436 | /* GetTempFileName requires that its output buffer (4th param) |
437 | have length MAX_PATH or greater. */ | |
438 | assert(size >= MAX_PATH); | |
439 | return (GetTempPath(MAX_PATH, temp_dir) | |
440 | && GetTempFileName(temp_dir, "qem", 0, filename) | |
441 | ? 0 : -GetLastError()); | |
d5249393 | 442 | #else |
67b915a5 | 443 | int fd; |
7ccfb2eb | 444 | const char *tmpdir; |
0badc1ee AJ |
445 | tmpdir = getenv("TMPDIR"); |
446 | if (!tmpdir) | |
447 | tmpdir = "/tmp"; | |
eba25057 JM |
448 | if (snprintf(filename, size, "%s/vl.XXXXXX", tmpdir) >= size) { |
449 | return -EOVERFLOW; | |
450 | } | |
ea2384d3 | 451 | fd = mkstemp(filename); |
fe235a06 DH |
452 | if (fd < 0) { |
453 | return -errno; | |
454 | } | |
455 | if (close(fd) != 0) { | |
456 | unlink(filename); | |
eba25057 JM |
457 | return -errno; |
458 | } | |
459 | return 0; | |
d5249393 | 460 | #endif |
eba25057 | 461 | } |
fc01f7e7 | 462 | |
84a12e66 CH |
463 | /* |
464 | * Detect host devices. By convention, /dev/cdrom[N] is always | |
465 | * recognized as a host CDROM. | |
466 | */ | |
467 | static BlockDriver *find_hdev_driver(const char *filename) | |
468 | { | |
469 | int score_max = 0, score; | |
470 | BlockDriver *drv = NULL, *d; | |
471 | ||
472 | QLIST_FOREACH(d, &bdrv_drivers, list) { | |
473 | if (d->bdrv_probe_device) { | |
474 | score = d->bdrv_probe_device(filename); | |
475 | if (score > score_max) { | |
476 | score_max = score; | |
477 | drv = d; | |
478 | } | |
479 | } | |
480 | } | |
481 | ||
482 | return drv; | |
483 | } | |
484 | ||
98289620 KW |
485 | BlockDriver *bdrv_find_protocol(const char *filename, |
486 | bool allow_protocol_prefix) | |
83f64091 FB |
487 | { |
488 | BlockDriver *drv1; | |
489 | char protocol[128]; | |
1cec71e3 | 490 | int len; |
83f64091 | 491 | const char *p; |
19cb3738 | 492 | |
66f82cee KW |
493 | /* TODO Drivers without bdrv_file_open must be specified explicitly */ |
494 | ||
39508e7a CH |
495 | /* |
496 | * XXX(hch): we really should not let host device detection | |
497 | * override an explicit protocol specification, but moving this | |
498 | * later breaks access to device names with colons in them. | |
499 | * Thanks to the brain-dead persistent naming schemes on udev- | |
500 | * based Linux systems those actually are quite common. | |
501 | */ | |
502 | drv1 = find_hdev_driver(filename); | |
503 | if (drv1) { | |
504 | return drv1; | |
505 | } | |
506 | ||
98289620 | 507 | if (!path_has_protocol(filename) || !allow_protocol_prefix) { |
39508e7a | 508 | return bdrv_find_format("file"); |
84a12e66 | 509 | } |
98289620 | 510 | |
9e0b22f4 SH |
511 | p = strchr(filename, ':'); |
512 | assert(p != NULL); | |
1cec71e3 AL |
513 | len = p - filename; |
514 | if (len > sizeof(protocol) - 1) | |
515 | len = sizeof(protocol) - 1; | |
516 | memcpy(protocol, filename, len); | |
517 | protocol[len] = '\0'; | |
8a22f02a | 518 | QLIST_FOREACH(drv1, &bdrv_drivers, list) { |
5fafdf24 | 519 | if (drv1->protocol_name && |
8a22f02a | 520 | !strcmp(drv1->protocol_name, protocol)) { |
83f64091 | 521 | return drv1; |
8a22f02a | 522 | } |
83f64091 FB |
523 | } |
524 | return NULL; | |
525 | } | |
526 | ||
f500a6d3 KW |
527 | static int find_image_format(BlockDriverState *bs, const char *filename, |
528 | BlockDriver **pdrv) | |
f3a5d3f8 | 529 | { |
f500a6d3 | 530 | int score, score_max; |
f3a5d3f8 CH |
531 | BlockDriver *drv1, *drv; |
532 | uint8_t buf[2048]; | |
f500a6d3 | 533 | int ret = 0; |
f8ea0b00 | 534 | |
08a00559 | 535 | /* Return the raw BlockDriver * to scsi-generic devices or empty drives */ |
8e895599 | 536 | if (bs->sg || !bdrv_is_inserted(bs) || bdrv_getlength(bs) == 0) { |
c98ac35d SW |
537 | drv = bdrv_find_format("raw"); |
538 | if (!drv) { | |
539 | ret = -ENOENT; | |
540 | } | |
541 | *pdrv = drv; | |
542 | return ret; | |
1a396859 | 543 | } |
f8ea0b00 | 544 | |
83f64091 | 545 | ret = bdrv_pread(bs, 0, buf, sizeof(buf)); |
83f64091 | 546 | if (ret < 0) { |
c98ac35d SW |
547 | *pdrv = NULL; |
548 | return ret; | |
83f64091 FB |
549 | } |
550 | ||
ea2384d3 | 551 | score_max = 0; |
84a12e66 | 552 | drv = NULL; |
8a22f02a | 553 | QLIST_FOREACH(drv1, &bdrv_drivers, list) { |
83f64091 FB |
554 | if (drv1->bdrv_probe) { |
555 | score = drv1->bdrv_probe(buf, ret, filename); | |
556 | if (score > score_max) { | |
557 | score_max = score; | |
558 | drv = drv1; | |
559 | } | |
0849bf08 | 560 | } |
fc01f7e7 | 561 | } |
c98ac35d SW |
562 | if (!drv) { |
563 | ret = -ENOENT; | |
564 | } | |
565 | *pdrv = drv; | |
566 | return ret; | |
ea2384d3 FB |
567 | } |
568 | ||
51762288 SH |
569 | /** |
570 | * Set the current 'total_sectors' value | |
571 | */ | |
572 | static int refresh_total_sectors(BlockDriverState *bs, int64_t hint) | |
573 | { | |
574 | BlockDriver *drv = bs->drv; | |
575 | ||
396759ad NB |
576 | /* Do not attempt drv->bdrv_getlength() on scsi-generic devices */ |
577 | if (bs->sg) | |
578 | return 0; | |
579 | ||
51762288 SH |
580 | /* query actual device if possible, otherwise just trust the hint */ |
581 | if (drv->bdrv_getlength) { | |
582 | int64_t length = drv->bdrv_getlength(bs); | |
583 | if (length < 0) { | |
584 | return length; | |
585 | } | |
586 | hint = length >> BDRV_SECTOR_BITS; | |
587 | } | |
588 | ||
589 | bs->total_sectors = hint; | |
590 | return 0; | |
591 | } | |
592 | ||
9e8f1835 PB |
593 | /** |
594 | * Set open flags for a given discard mode | |
595 | * | |
596 | * Return 0 on success, -1 if the discard mode was invalid. | |
597 | */ | |
598 | int bdrv_parse_discard_flags(const char *mode, int *flags) | |
599 | { | |
600 | *flags &= ~BDRV_O_UNMAP; | |
601 | ||
602 | if (!strcmp(mode, "off") || !strcmp(mode, "ignore")) { | |
603 | /* do nothing */ | |
604 | } else if (!strcmp(mode, "on") || !strcmp(mode, "unmap")) { | |
605 | *flags |= BDRV_O_UNMAP; | |
606 | } else { | |
607 | return -1; | |
608 | } | |
609 | ||
610 | return 0; | |
611 | } | |
612 | ||
c3993cdc SH |
613 | /** |
614 | * Set open flags for a given cache mode | |
615 | * | |
616 | * Return 0 on success, -1 if the cache mode was invalid. | |
617 | */ | |
618 | int bdrv_parse_cache_flags(const char *mode, int *flags) | |
619 | { | |
620 | *flags &= ~BDRV_O_CACHE_MASK; | |
621 | ||
622 | if (!strcmp(mode, "off") || !strcmp(mode, "none")) { | |
623 | *flags |= BDRV_O_NOCACHE | BDRV_O_CACHE_WB; | |
92196b2f SH |
624 | } else if (!strcmp(mode, "directsync")) { |
625 | *flags |= BDRV_O_NOCACHE; | |
c3993cdc SH |
626 | } else if (!strcmp(mode, "writeback")) { |
627 | *flags |= BDRV_O_CACHE_WB; | |
628 | } else if (!strcmp(mode, "unsafe")) { | |
629 | *flags |= BDRV_O_CACHE_WB; | |
630 | *flags |= BDRV_O_NO_FLUSH; | |
631 | } else if (!strcmp(mode, "writethrough")) { | |
632 | /* this is the default */ | |
633 | } else { | |
634 | return -1; | |
635 | } | |
636 | ||
637 | return 0; | |
638 | } | |
639 | ||
53fec9d3 SH |
640 | /** |
641 | * The copy-on-read flag is actually a reference count so multiple users may | |
642 | * use the feature without worrying about clobbering its previous state. | |
643 | * Copy-on-read stays enabled until all users have called to disable it. | |
644 | */ | |
645 | void bdrv_enable_copy_on_read(BlockDriverState *bs) | |
646 | { | |
647 | bs->copy_on_read++; | |
648 | } | |
649 | ||
650 | void bdrv_disable_copy_on_read(BlockDriverState *bs) | |
651 | { | |
652 | assert(bs->copy_on_read > 0); | |
653 | bs->copy_on_read--; | |
654 | } | |
655 | ||
7b272452 KW |
656 | static int bdrv_open_flags(BlockDriverState *bs, int flags) |
657 | { | |
658 | int open_flags = flags | BDRV_O_CACHE_WB; | |
659 | ||
660 | /* | |
661 | * Clear flags that are internal to the block layer before opening the | |
662 | * image. | |
663 | */ | |
664 | open_flags &= ~(BDRV_O_SNAPSHOT | BDRV_O_NO_BACKING); | |
665 | ||
666 | /* | |
667 | * Snapshots should be writable. | |
668 | */ | |
669 | if (bs->is_temporary) { | |
670 | open_flags |= BDRV_O_RDWR; | |
671 | } | |
672 | ||
673 | return open_flags; | |
674 | } | |
675 | ||
57915332 KW |
676 | /* |
677 | * Common part for opening disk images and files | |
b6ad491a KW |
678 | * |
679 | * Removes all processed options from *options. | |
57915332 | 680 | */ |
f500a6d3 | 681 | static int bdrv_open_common(BlockDriverState *bs, BlockDriverState *file, |
035fccdf | 682 | QDict *options, int flags, BlockDriver *drv) |
57915332 KW |
683 | { |
684 | int ret, open_flags; | |
035fccdf | 685 | const char *filename; |
57915332 KW |
686 | |
687 | assert(drv != NULL); | |
6405875c | 688 | assert(bs->file == NULL); |
707ff828 | 689 | assert(options != NULL && bs->options != options); |
57915332 | 690 | |
45673671 KW |
691 | if (file != NULL) { |
692 | filename = file->filename; | |
693 | } else { | |
694 | filename = qdict_get_try_str(options, "filename"); | |
695 | } | |
696 | ||
697 | trace_bdrv_open_common(bs, filename ?: "", flags, drv->format_name); | |
28dcee10 | 698 | |
5d186eb0 KW |
699 | /* bdrv_open() with directly using a protocol as drv. This layer is already |
700 | * opened, so assign it to bs (while file becomes a closed BlockDriverState) | |
701 | * and return immediately. */ | |
702 | if (file != NULL && drv->bdrv_file_open) { | |
703 | bdrv_swap(file, bs); | |
704 | return 0; | |
705 | } | |
706 | ||
57915332 | 707 | bs->open_flags = flags; |
57915332 | 708 | bs->buffer_alignment = 512; |
b64ec4e4 FZ |
709 | open_flags = bdrv_open_flags(bs, flags); |
710 | bs->read_only = !(open_flags & BDRV_O_RDWR); | |
711 | ||
712 | if (use_bdrv_whitelist && !bdrv_is_whitelisted(drv, bs->read_only)) { | |
713 | return -ENOTSUP; | |
714 | } | |
57915332 | 715 | |
53fec9d3 | 716 | assert(bs->copy_on_read == 0); /* bdrv_new() and bdrv_close() make it so */ |
b64ec4e4 | 717 | if (!bs->read_only && (flags & BDRV_O_COPY_ON_READ)) { |
53fec9d3 SH |
718 | bdrv_enable_copy_on_read(bs); |
719 | } | |
720 | ||
c2ad1b0c KW |
721 | if (filename != NULL) { |
722 | pstrcpy(bs->filename, sizeof(bs->filename), filename); | |
723 | } else { | |
724 | bs->filename[0] = '\0'; | |
725 | } | |
57915332 | 726 | |
57915332 | 727 | bs->drv = drv; |
7267c094 | 728 | bs->opaque = g_malloc0(drv->instance_size); |
57915332 | 729 | |
03f541bd | 730 | bs->enable_write_cache = !!(flags & BDRV_O_CACHE_WB); |
e7c63796 | 731 | |
66f82cee KW |
732 | /* Open the image, either directly or using a protocol */ |
733 | if (drv->bdrv_file_open) { | |
5d186eb0 KW |
734 | assert(file == NULL); |
735 | assert(drv->bdrv_parse_filename || filename != NULL); | |
56d1b4d2 | 736 | ret = drv->bdrv_file_open(bs, options, open_flags); |
f500a6d3 | 737 | } else { |
2af5ef70 KW |
738 | if (file == NULL) { |
739 | qerror_report(ERROR_CLASS_GENERIC_ERROR, "Can't use '%s' as a " | |
740 | "block driver for the protocol level", | |
741 | drv->format_name); | |
742 | ret = -EINVAL; | |
743 | goto free_and_fail; | |
744 | } | |
f500a6d3 KW |
745 | assert(file != NULL); |
746 | bs->file = file; | |
b6ad491a | 747 | ret = drv->bdrv_open(bs, options, open_flags); |
66f82cee KW |
748 | } |
749 | ||
57915332 KW |
750 | if (ret < 0) { |
751 | goto free_and_fail; | |
752 | } | |
753 | ||
51762288 SH |
754 | ret = refresh_total_sectors(bs, bs->total_sectors); |
755 | if (ret < 0) { | |
756 | goto free_and_fail; | |
57915332 | 757 | } |
51762288 | 758 | |
57915332 KW |
759 | #ifndef _WIN32 |
760 | if (bs->is_temporary) { | |
c2ad1b0c | 761 | assert(filename != NULL); |
57915332 KW |
762 | unlink(filename); |
763 | } | |
764 | #endif | |
765 | return 0; | |
766 | ||
767 | free_and_fail: | |
f500a6d3 | 768 | bs->file = NULL; |
7267c094 | 769 | g_free(bs->opaque); |
57915332 KW |
770 | bs->opaque = NULL; |
771 | bs->drv = NULL; | |
772 | return ret; | |
773 | } | |
774 | ||
b6ce07aa KW |
775 | /* |
776 | * Opens a file using a protocol (file, host_device, nbd, ...) | |
787e4a85 KW |
777 | * |
778 | * options is a QDict of options to pass to the block drivers, or NULL for an | |
779 | * empty set of options. The reference to the QDict belongs to the block layer | |
780 | * after the call (even on failure), so if the caller intends to reuse the | |
781 | * dictionary, it needs to use QINCREF() before calling bdrv_file_open. | |
b6ce07aa | 782 | */ |
787e4a85 KW |
783 | int bdrv_file_open(BlockDriverState **pbs, const char *filename, |
784 | QDict *options, int flags) | |
ea2384d3 | 785 | { |
83f64091 | 786 | BlockDriverState *bs; |
6db95603 | 787 | BlockDriver *drv; |
c2ad1b0c | 788 | const char *drvname; |
98289620 | 789 | bool allow_protocol_prefix = false; |
83f64091 FB |
790 | int ret; |
791 | ||
707ff828 KW |
792 | /* NULL means an empty set of options */ |
793 | if (options == NULL) { | |
794 | options = qdict_new(); | |
795 | } | |
796 | ||
83f64091 | 797 | bs = bdrv_new(""); |
707ff828 KW |
798 | bs->options = options; |
799 | options = qdict_clone_shallow(options); | |
800 | ||
035fccdf KW |
801 | /* Fetch the file name from the options QDict if necessary */ |
802 | if (!filename) { | |
803 | filename = qdict_get_try_str(options, "filename"); | |
804 | } else if (filename && !qdict_haskey(options, "filename")) { | |
805 | qdict_put(options, "filename", qstring_from_str(filename)); | |
98289620 | 806 | allow_protocol_prefix = true; |
035fccdf KW |
807 | } else { |
808 | qerror_report(ERROR_CLASS_GENERIC_ERROR, "Can't specify 'file' and " | |
809 | "'filename' options at the same time"); | |
810 | ret = -EINVAL; | |
811 | goto fail; | |
812 | } | |
813 | ||
c2ad1b0c KW |
814 | /* Find the right block driver */ |
815 | drvname = qdict_get_try_str(options, "driver"); | |
816 | if (drvname) { | |
b64ec4e4 | 817 | drv = bdrv_find_whitelisted_format(drvname, !(flags & BDRV_O_RDWR)); |
c2ad1b0c KW |
818 | qdict_del(options, "driver"); |
819 | } else if (filename) { | |
98289620 KW |
820 | drv = bdrv_find_protocol(filename, allow_protocol_prefix); |
821 | if (!drv) { | |
822 | qerror_report(ERROR_CLASS_GENERIC_ERROR, "Unknown protocol"); | |
823 | } | |
c2ad1b0c KW |
824 | } else { |
825 | qerror_report(ERROR_CLASS_GENERIC_ERROR, | |
826 | "Must specify either driver or file"); | |
827 | drv = NULL; | |
828 | } | |
829 | ||
830 | if (!drv) { | |
831 | ret = -ENOENT; | |
832 | goto fail; | |
833 | } | |
834 | ||
835 | /* Parse the filename and open it */ | |
836 | if (drv->bdrv_parse_filename && filename) { | |
6963a30d KW |
837 | Error *local_err = NULL; |
838 | drv->bdrv_parse_filename(filename, options, &local_err); | |
839 | if (error_is_set(&local_err)) { | |
840 | qerror_report_err(local_err); | |
841 | error_free(local_err); | |
842 | ret = -EINVAL; | |
843 | goto fail; | |
844 | } | |
56d1b4d2 | 845 | qdict_del(options, "filename"); |
c2ad1b0c KW |
846 | } else if (!drv->bdrv_parse_filename && !filename) { |
847 | qerror_report(ERROR_CLASS_GENERIC_ERROR, | |
848 | "The '%s' block driver requires a file name", | |
849 | drv->format_name); | |
850 | ret = -EINVAL; | |
851 | goto fail; | |
6963a30d KW |
852 | } |
853 | ||
035fccdf | 854 | ret = bdrv_open_common(bs, NULL, options, flags, drv); |
83f64091 | 855 | if (ret < 0) { |
707ff828 KW |
856 | goto fail; |
857 | } | |
858 | ||
859 | /* Check if any unknown options were used */ | |
860 | if (qdict_size(options) != 0) { | |
861 | const QDictEntry *entry = qdict_first(options); | |
862 | qerror_report(ERROR_CLASS_GENERIC_ERROR, "Block protocol '%s' doesn't " | |
863 | "support the option '%s'", | |
864 | drv->format_name, entry->key); | |
865 | ret = -EINVAL; | |
866 | goto fail; | |
3b0d4f61 | 867 | } |
707ff828 KW |
868 | QDECREF(options); |
869 | ||
71d0770c | 870 | bs->growable = 1; |
83f64091 FB |
871 | *pbs = bs; |
872 | return 0; | |
707ff828 KW |
873 | |
874 | fail: | |
875 | QDECREF(options); | |
876 | if (!bs->drv) { | |
877 | QDECREF(bs->options); | |
878 | } | |
879 | bdrv_delete(bs); | |
880 | return ret; | |
83f64091 FB |
881 | } |
882 | ||
31ca6d07 KW |
883 | /* |
884 | * Opens the backing file for a BlockDriverState if not yet open | |
885 | * | |
886 | * options is a QDict of options to pass to the block drivers, or NULL for an | |
887 | * empty set of options. The reference to the QDict is transferred to this | |
888 | * function (even on failure), so if the caller intends to reuse the dictionary, | |
889 | * it needs to use QINCREF() before calling bdrv_file_open. | |
890 | */ | |
891 | int bdrv_open_backing_file(BlockDriverState *bs, QDict *options) | |
9156df12 PB |
892 | { |
893 | char backing_filename[PATH_MAX]; | |
894 | int back_flags, ret; | |
895 | BlockDriver *back_drv = NULL; | |
896 | ||
897 | if (bs->backing_hd != NULL) { | |
31ca6d07 | 898 | QDECREF(options); |
9156df12 PB |
899 | return 0; |
900 | } | |
901 | ||
31ca6d07 KW |
902 | /* NULL means an empty set of options */ |
903 | if (options == NULL) { | |
904 | options = qdict_new(); | |
905 | } | |
906 | ||
9156df12 | 907 | bs->open_flags &= ~BDRV_O_NO_BACKING; |
1cb6f506 KW |
908 | if (qdict_haskey(options, "file.filename")) { |
909 | backing_filename[0] = '\0'; | |
910 | } else if (bs->backing_file[0] == '\0' && qdict_size(options) == 0) { | |
31ca6d07 | 911 | QDECREF(options); |
9156df12 PB |
912 | return 0; |
913 | } | |
914 | ||
915 | bs->backing_hd = bdrv_new(""); | |
916 | bdrv_get_full_backing_filename(bs, backing_filename, | |
917 | sizeof(backing_filename)); | |
918 | ||
919 | if (bs->backing_format[0] != '\0') { | |
920 | back_drv = bdrv_find_format(bs->backing_format); | |
921 | } | |
922 | ||
923 | /* backing files always opened read-only */ | |
924 | back_flags = bs->open_flags & ~(BDRV_O_RDWR | BDRV_O_SNAPSHOT); | |
925 | ||
31ca6d07 KW |
926 | ret = bdrv_open(bs->backing_hd, |
927 | *backing_filename ? backing_filename : NULL, options, | |
de9c0cec | 928 | back_flags, back_drv); |
9156df12 PB |
929 | if (ret < 0) { |
930 | bdrv_delete(bs->backing_hd); | |
931 | bs->backing_hd = NULL; | |
932 | bs->open_flags |= BDRV_O_NO_BACKING; | |
933 | return ret; | |
934 | } | |
935 | return 0; | |
936 | } | |
937 | ||
707ff828 KW |
938 | static void extract_subqdict(QDict *src, QDict **dst, const char *start) |
939 | { | |
940 | const QDictEntry *entry, *next; | |
941 | const char *p; | |
942 | ||
943 | *dst = qdict_new(); | |
944 | entry = qdict_first(src); | |
945 | ||
946 | while (entry != NULL) { | |
947 | next = qdict_next(src, entry); | |
948 | if (strstart(entry->key, start, &p)) { | |
949 | qobject_incref(entry->value); | |
950 | qdict_put_obj(*dst, p, entry->value); | |
951 | qdict_del(src, entry->key); | |
952 | } | |
953 | entry = next; | |
954 | } | |
955 | } | |
956 | ||
b6ce07aa KW |
957 | /* |
958 | * Opens a disk image (raw, qcow2, vmdk, ...) | |
de9c0cec KW |
959 | * |
960 | * options is a QDict of options to pass to the block drivers, or NULL for an | |
961 | * empty set of options. The reference to the QDict belongs to the block layer | |
962 | * after the call (even on failure), so if the caller intends to reuse the | |
963 | * dictionary, it needs to use QINCREF() before calling bdrv_open. | |
b6ce07aa | 964 | */ |
de9c0cec KW |
965 | int bdrv_open(BlockDriverState *bs, const char *filename, QDict *options, |
966 | int flags, BlockDriver *drv) | |
ea2384d3 | 967 | { |
b6ce07aa | 968 | int ret; |
89c9bc3d SW |
969 | /* TODO: extra byte is a hack to ensure MAX_PATH space on Windows. */ |
970 | char tmp_filename[PATH_MAX + 1]; | |
f500a6d3 | 971 | BlockDriverState *file = NULL; |
707ff828 | 972 | QDict *file_options = NULL; |
74fe54f2 | 973 | const char *drvname; |
712e7874 | 974 | |
de9c0cec KW |
975 | /* NULL means an empty set of options */ |
976 | if (options == NULL) { | |
977 | options = qdict_new(); | |
978 | } | |
979 | ||
980 | bs->options = options; | |
b6ad491a | 981 | options = qdict_clone_shallow(options); |
de9c0cec KW |
982 | |
983 | /* For snapshot=on, create a temporary qcow2 overlay */ | |
83f64091 | 984 | if (flags & BDRV_O_SNAPSHOT) { |
ea2384d3 FB |
985 | BlockDriverState *bs1; |
986 | int64_t total_size; | |
91a073a9 | 987 | BlockDriver *bdrv_qcow2; |
08b392e1 | 988 | QEMUOptionParameter *create_options; |
b6ce07aa | 989 | char backing_filename[PATH_MAX]; |
3b46e624 | 990 | |
c2ad1b0c KW |
991 | if (qdict_size(options) != 0) { |
992 | error_report("Can't use snapshot=on with driver-specific options"); | |
993 | ret = -EINVAL; | |
994 | goto fail; | |
995 | } | |
996 | assert(filename != NULL); | |
997 | ||
ea2384d3 FB |
998 | /* if snapshot, we create a temporary backing file and open it |
999 | instead of opening 'filename' directly */ | |
33e3963e | 1000 | |
ea2384d3 FB |
1001 | /* if there is a backing file, use it */ |
1002 | bs1 = bdrv_new(""); | |
de9c0cec | 1003 | ret = bdrv_open(bs1, filename, NULL, 0, drv); |
51d7c00c | 1004 | if (ret < 0) { |
ea2384d3 | 1005 | bdrv_delete(bs1); |
de9c0cec | 1006 | goto fail; |
ea2384d3 | 1007 | } |
3e82990b | 1008 | total_size = bdrv_getlength(bs1) & BDRV_SECTOR_MASK; |
7c96d46e | 1009 | |
ea2384d3 | 1010 | bdrv_delete(bs1); |
3b46e624 | 1011 | |
eba25057 JM |
1012 | ret = get_tmp_filename(tmp_filename, sizeof(tmp_filename)); |
1013 | if (ret < 0) { | |
de9c0cec | 1014 | goto fail; |
eba25057 | 1015 | } |
7c96d46e AL |
1016 | |
1017 | /* Real path is meaningless for protocols */ | |
4d70655b | 1018 | if (path_has_protocol(filename)) { |
7c96d46e AL |
1019 | snprintf(backing_filename, sizeof(backing_filename), |
1020 | "%s", filename); | |
de9c0cec KW |
1021 | } else if (!realpath(filename, backing_filename)) { |
1022 | ret = -errno; | |
1023 | goto fail; | |
1024 | } | |
7c96d46e | 1025 | |
91a073a9 | 1026 | bdrv_qcow2 = bdrv_find_format("qcow2"); |
08b392e1 KW |
1027 | create_options = parse_option_parameters("", bdrv_qcow2->create_options, |
1028 | NULL); | |
91a073a9 | 1029 | |
08b392e1 KW |
1030 | set_option_parameter_int(create_options, BLOCK_OPT_SIZE, total_size); |
1031 | set_option_parameter(create_options, BLOCK_OPT_BACKING_FILE, | |
1032 | backing_filename); | |
91a073a9 | 1033 | if (drv) { |
08b392e1 | 1034 | set_option_parameter(create_options, BLOCK_OPT_BACKING_FMT, |
91a073a9 KW |
1035 | drv->format_name); |
1036 | } | |
1037 | ||
08b392e1 KW |
1038 | ret = bdrv_create(bdrv_qcow2, tmp_filename, create_options); |
1039 | free_option_parameters(create_options); | |
51d7c00c | 1040 | if (ret < 0) { |
de9c0cec | 1041 | goto fail; |
ea2384d3 | 1042 | } |
91a073a9 | 1043 | |
ea2384d3 | 1044 | filename = tmp_filename; |
91a073a9 | 1045 | drv = bdrv_qcow2; |
ea2384d3 FB |
1046 | bs->is_temporary = 1; |
1047 | } | |
712e7874 | 1048 | |
f500a6d3 KW |
1049 | /* Open image file without format layer */ |
1050 | if (flags & BDRV_O_RDWR) { | |
1051 | flags |= BDRV_O_ALLOW_RDWR; | |
1052 | } | |
1053 | ||
707ff828 KW |
1054 | extract_subqdict(options, &file_options, "file."); |
1055 | ||
1056 | ret = bdrv_file_open(&file, filename, file_options, | |
50b05b6f | 1057 | bdrv_open_flags(bs, flags | BDRV_O_UNMAP)); |
f500a6d3 | 1058 | if (ret < 0) { |
de9c0cec | 1059 | goto fail; |
f500a6d3 KW |
1060 | } |
1061 | ||
b6ce07aa | 1062 | /* Find the right image format driver */ |
74fe54f2 KW |
1063 | drvname = qdict_get_try_str(options, "driver"); |
1064 | if (drvname) { | |
1065 | drv = bdrv_find_whitelisted_format(drvname, !(flags & BDRV_O_RDWR)); | |
1066 | qdict_del(options, "driver"); | |
1067 | } | |
1068 | ||
6db95603 | 1069 | if (!drv) { |
f500a6d3 | 1070 | ret = find_image_format(file, filename, &drv); |
51d7c00c | 1071 | } |
6987307c | 1072 | |
51d7c00c | 1073 | if (!drv) { |
51d7c00c | 1074 | goto unlink_and_fail; |
ea2384d3 | 1075 | } |
b6ce07aa KW |
1076 | |
1077 | /* Open the image */ | |
035fccdf | 1078 | ret = bdrv_open_common(bs, file, options, flags, drv); |
b6ce07aa | 1079 | if (ret < 0) { |
6987307c CH |
1080 | goto unlink_and_fail; |
1081 | } | |
1082 | ||
f500a6d3 KW |
1083 | if (bs->file != file) { |
1084 | bdrv_delete(file); | |
1085 | file = NULL; | |
1086 | } | |
1087 | ||
b6ce07aa | 1088 | /* If there is a backing file, use it */ |
9156df12 | 1089 | if ((flags & BDRV_O_NO_BACKING) == 0) { |
31ca6d07 KW |
1090 | QDict *backing_options; |
1091 | ||
1092 | extract_subqdict(options, &backing_options, "backing."); | |
1093 | ret = bdrv_open_backing_file(bs, backing_options); | |
b6ce07aa | 1094 | if (ret < 0) { |
b6ad491a | 1095 | goto close_and_fail; |
b6ce07aa | 1096 | } |
b6ce07aa KW |
1097 | } |
1098 | ||
b6ad491a KW |
1099 | /* Check if any unknown options were used */ |
1100 | if (qdict_size(options) != 0) { | |
1101 | const QDictEntry *entry = qdict_first(options); | |
1102 | qerror_report(ERROR_CLASS_GENERIC_ERROR, "Block format '%s' used by " | |
1103 | "device '%s' doesn't support the option '%s'", | |
1104 | drv->format_name, bs->device_name, entry->key); | |
1105 | ||
1106 | ret = -EINVAL; | |
1107 | goto close_and_fail; | |
1108 | } | |
1109 | QDECREF(options); | |
1110 | ||
b6ce07aa | 1111 | if (!bdrv_key_required(bs)) { |
7d4b4ba5 | 1112 | bdrv_dev_change_media_cb(bs, true); |
b6ce07aa KW |
1113 | } |
1114 | ||
98f90dba ZYW |
1115 | /* throttling disk I/O limits */ |
1116 | if (bs->io_limits_enabled) { | |
1117 | bdrv_io_limits_enable(bs); | |
1118 | } | |
1119 | ||
b6ce07aa KW |
1120 | return 0; |
1121 | ||
1122 | unlink_and_fail: | |
f500a6d3 KW |
1123 | if (file != NULL) { |
1124 | bdrv_delete(file); | |
1125 | } | |
b6ce07aa KW |
1126 | if (bs->is_temporary) { |
1127 | unlink(filename); | |
1128 | } | |
de9c0cec KW |
1129 | fail: |
1130 | QDECREF(bs->options); | |
b6ad491a | 1131 | QDECREF(options); |
de9c0cec | 1132 | bs->options = NULL; |
b6ad491a | 1133 | return ret; |
de9c0cec | 1134 | |
b6ad491a KW |
1135 | close_and_fail: |
1136 | bdrv_close(bs); | |
1137 | QDECREF(options); | |
b6ce07aa KW |
1138 | return ret; |
1139 | } | |
1140 | ||
e971aa12 JC |
1141 | typedef struct BlockReopenQueueEntry { |
1142 | bool prepared; | |
1143 | BDRVReopenState state; | |
1144 | QSIMPLEQ_ENTRY(BlockReopenQueueEntry) entry; | |
1145 | } BlockReopenQueueEntry; | |
1146 | ||
1147 | /* | |
1148 | * Adds a BlockDriverState to a simple queue for an atomic, transactional | |
1149 | * reopen of multiple devices. | |
1150 | * | |
1151 | * bs_queue can either be an existing BlockReopenQueue that has had QSIMPLE_INIT | |
1152 | * already performed, or alternatively may be NULL a new BlockReopenQueue will | |
1153 | * be created and initialized. This newly created BlockReopenQueue should be | |
1154 | * passed back in for subsequent calls that are intended to be of the same | |
1155 | * atomic 'set'. | |
1156 | * | |
1157 | * bs is the BlockDriverState to add to the reopen queue. | |
1158 | * | |
1159 | * flags contains the open flags for the associated bs | |
1160 | * | |
1161 | * returns a pointer to bs_queue, which is either the newly allocated | |
1162 | * bs_queue, or the existing bs_queue being used. | |
1163 | * | |
1164 | */ | |
1165 | BlockReopenQueue *bdrv_reopen_queue(BlockReopenQueue *bs_queue, | |
1166 | BlockDriverState *bs, int flags) | |
1167 | { | |
1168 | assert(bs != NULL); | |
1169 | ||
1170 | BlockReopenQueueEntry *bs_entry; | |
1171 | if (bs_queue == NULL) { | |
1172 | bs_queue = g_new0(BlockReopenQueue, 1); | |
1173 | QSIMPLEQ_INIT(bs_queue); | |
1174 | } | |
1175 | ||
1176 | if (bs->file) { | |
1177 | bdrv_reopen_queue(bs_queue, bs->file, flags); | |
1178 | } | |
1179 | ||
1180 | bs_entry = g_new0(BlockReopenQueueEntry, 1); | |
1181 | QSIMPLEQ_INSERT_TAIL(bs_queue, bs_entry, entry); | |
1182 | ||
1183 | bs_entry->state.bs = bs; | |
1184 | bs_entry->state.flags = flags; | |
1185 | ||
1186 | return bs_queue; | |
1187 | } | |
1188 | ||
1189 | /* | |
1190 | * Reopen multiple BlockDriverStates atomically & transactionally. | |
1191 | * | |
1192 | * The queue passed in (bs_queue) must have been built up previous | |
1193 | * via bdrv_reopen_queue(). | |
1194 | * | |
1195 | * Reopens all BDS specified in the queue, with the appropriate | |
1196 | * flags. All devices are prepared for reopen, and failure of any | |
1197 | * device will cause all device changes to be abandonded, and intermediate | |
1198 | * data cleaned up. | |
1199 | * | |
1200 | * If all devices prepare successfully, then the changes are committed | |
1201 | * to all devices. | |
1202 | * | |
1203 | */ | |
1204 | int bdrv_reopen_multiple(BlockReopenQueue *bs_queue, Error **errp) | |
1205 | { | |
1206 | int ret = -1; | |
1207 | BlockReopenQueueEntry *bs_entry, *next; | |
1208 | Error *local_err = NULL; | |
1209 | ||
1210 | assert(bs_queue != NULL); | |
1211 | ||
1212 | bdrv_drain_all(); | |
1213 | ||
1214 | QSIMPLEQ_FOREACH(bs_entry, bs_queue, entry) { | |
1215 | if (bdrv_reopen_prepare(&bs_entry->state, bs_queue, &local_err)) { | |
1216 | error_propagate(errp, local_err); | |
1217 | goto cleanup; | |
1218 | } | |
1219 | bs_entry->prepared = true; | |
1220 | } | |
1221 | ||
1222 | /* If we reach this point, we have success and just need to apply the | |
1223 | * changes | |
1224 | */ | |
1225 | QSIMPLEQ_FOREACH(bs_entry, bs_queue, entry) { | |
1226 | bdrv_reopen_commit(&bs_entry->state); | |
1227 | } | |
1228 | ||
1229 | ret = 0; | |
1230 | ||
1231 | cleanup: | |
1232 | QSIMPLEQ_FOREACH_SAFE(bs_entry, bs_queue, entry, next) { | |
1233 | if (ret && bs_entry->prepared) { | |
1234 | bdrv_reopen_abort(&bs_entry->state); | |
1235 | } | |
1236 | g_free(bs_entry); | |
1237 | } | |
1238 | g_free(bs_queue); | |
1239 | return ret; | |
1240 | } | |
1241 | ||
1242 | ||
1243 | /* Reopen a single BlockDriverState with the specified flags. */ | |
1244 | int bdrv_reopen(BlockDriverState *bs, int bdrv_flags, Error **errp) | |
1245 | { | |
1246 | int ret = -1; | |
1247 | Error *local_err = NULL; | |
1248 | BlockReopenQueue *queue = bdrv_reopen_queue(NULL, bs, bdrv_flags); | |
1249 | ||
1250 | ret = bdrv_reopen_multiple(queue, &local_err); | |
1251 | if (local_err != NULL) { | |
1252 | error_propagate(errp, local_err); | |
1253 | } | |
1254 | return ret; | |
1255 | } | |
1256 | ||
1257 | ||
1258 | /* | |
1259 | * Prepares a BlockDriverState for reopen. All changes are staged in the | |
1260 | * 'opaque' field of the BDRVReopenState, which is used and allocated by | |
1261 | * the block driver layer .bdrv_reopen_prepare() | |
1262 | * | |
1263 | * bs is the BlockDriverState to reopen | |
1264 | * flags are the new open flags | |
1265 | * queue is the reopen queue | |
1266 | * | |
1267 | * Returns 0 on success, non-zero on error. On error errp will be set | |
1268 | * as well. | |
1269 | * | |
1270 | * On failure, bdrv_reopen_abort() will be called to clean up any data. | |
1271 | * It is the responsibility of the caller to then call the abort() or | |
1272 | * commit() for any other BDS that have been left in a prepare() state | |
1273 | * | |
1274 | */ | |
1275 | int bdrv_reopen_prepare(BDRVReopenState *reopen_state, BlockReopenQueue *queue, | |
1276 | Error **errp) | |
1277 | { | |
1278 | int ret = -1; | |
1279 | Error *local_err = NULL; | |
1280 | BlockDriver *drv; | |
1281 | ||
1282 | assert(reopen_state != NULL); | |
1283 | assert(reopen_state->bs->drv != NULL); | |
1284 | drv = reopen_state->bs->drv; | |
1285 | ||
1286 | /* if we are to stay read-only, do not allow permission change | |
1287 | * to r/w */ | |
1288 | if (!(reopen_state->bs->open_flags & BDRV_O_ALLOW_RDWR) && | |
1289 | reopen_state->flags & BDRV_O_RDWR) { | |
1290 | error_set(errp, QERR_DEVICE_IS_READ_ONLY, | |
1291 | reopen_state->bs->device_name); | |
1292 | goto error; | |
1293 | } | |
1294 | ||
1295 | ||
1296 | ret = bdrv_flush(reopen_state->bs); | |
1297 | if (ret) { | |
1298 | error_set(errp, ERROR_CLASS_GENERIC_ERROR, "Error (%s) flushing drive", | |
1299 | strerror(-ret)); | |
1300 | goto error; | |
1301 | } | |
1302 | ||
1303 | if (drv->bdrv_reopen_prepare) { | |
1304 | ret = drv->bdrv_reopen_prepare(reopen_state, queue, &local_err); | |
1305 | if (ret) { | |
1306 | if (local_err != NULL) { | |
1307 | error_propagate(errp, local_err); | |
1308 | } else { | |
d8b6895f LC |
1309 | error_setg(errp, "failed while preparing to reopen image '%s'", |
1310 | reopen_state->bs->filename); | |
e971aa12 JC |
1311 | } |
1312 | goto error; | |
1313 | } | |
1314 | } else { | |
1315 | /* It is currently mandatory to have a bdrv_reopen_prepare() | |
1316 | * handler for each supported drv. */ | |
1317 | error_set(errp, QERR_BLOCK_FORMAT_FEATURE_NOT_SUPPORTED, | |
1318 | drv->format_name, reopen_state->bs->device_name, | |
1319 | "reopening of file"); | |
1320 | ret = -1; | |
1321 | goto error; | |
1322 | } | |
1323 | ||
1324 | ret = 0; | |
1325 | ||
1326 | error: | |
1327 | return ret; | |
1328 | } | |
1329 | ||
1330 | /* | |
1331 | * Takes the staged changes for the reopen from bdrv_reopen_prepare(), and | |
1332 | * makes them final by swapping the staging BlockDriverState contents into | |
1333 | * the active BlockDriverState contents. | |
1334 | */ | |
1335 | void bdrv_reopen_commit(BDRVReopenState *reopen_state) | |
1336 | { | |
1337 | BlockDriver *drv; | |
1338 | ||
1339 | assert(reopen_state != NULL); | |
1340 | drv = reopen_state->bs->drv; | |
1341 | assert(drv != NULL); | |
1342 | ||
1343 | /* If there are any driver level actions to take */ | |
1344 | if (drv->bdrv_reopen_commit) { | |
1345 | drv->bdrv_reopen_commit(reopen_state); | |
1346 | } | |
1347 | ||
1348 | /* set BDS specific flags now */ | |
1349 | reopen_state->bs->open_flags = reopen_state->flags; | |
1350 | reopen_state->bs->enable_write_cache = !!(reopen_state->flags & | |
1351 | BDRV_O_CACHE_WB); | |
1352 | reopen_state->bs->read_only = !(reopen_state->flags & BDRV_O_RDWR); | |
1353 | } | |
1354 | ||
1355 | /* | |
1356 | * Abort the reopen, and delete and free the staged changes in | |
1357 | * reopen_state | |
1358 | */ | |
1359 | void bdrv_reopen_abort(BDRVReopenState *reopen_state) | |
1360 | { | |
1361 | BlockDriver *drv; | |
1362 | ||
1363 | assert(reopen_state != NULL); | |
1364 | drv = reopen_state->bs->drv; | |
1365 | assert(drv != NULL); | |
1366 | ||
1367 | if (drv->bdrv_reopen_abort) { | |
1368 | drv->bdrv_reopen_abort(reopen_state); | |
1369 | } | |
1370 | } | |
1371 | ||
1372 | ||
fc01f7e7 FB |
1373 | void bdrv_close(BlockDriverState *bs) |
1374 | { | |
3cbc002c PB |
1375 | if (bs->job) { |
1376 | block_job_cancel_sync(bs->job); | |
1377 | } | |
58fda173 SH |
1378 | bdrv_drain_all(); /* complete I/O */ |
1379 | bdrv_flush(bs); | |
1380 | bdrv_drain_all(); /* in case flush left pending I/O */ | |
d7d512f6 | 1381 | notifier_list_notify(&bs->close_notifiers, bs); |
7094f12f | 1382 | |
3cbc002c | 1383 | if (bs->drv) { |
557df6ac | 1384 | if (bs->backing_hd) { |
ea2384d3 | 1385 | bdrv_delete(bs->backing_hd); |
557df6ac SH |
1386 | bs->backing_hd = NULL; |
1387 | } | |
ea2384d3 | 1388 | bs->drv->bdrv_close(bs); |
7267c094 | 1389 | g_free(bs->opaque); |
ea2384d3 FB |
1390 | #ifdef _WIN32 |
1391 | if (bs->is_temporary) { | |
1392 | unlink(bs->filename); | |
1393 | } | |
67b915a5 | 1394 | #endif |
ea2384d3 FB |
1395 | bs->opaque = NULL; |
1396 | bs->drv = NULL; | |
53fec9d3 | 1397 | bs->copy_on_read = 0; |
a275fa42 PB |
1398 | bs->backing_file[0] = '\0'; |
1399 | bs->backing_format[0] = '\0'; | |
6405875c PB |
1400 | bs->total_sectors = 0; |
1401 | bs->encrypted = 0; | |
1402 | bs->valid_key = 0; | |
1403 | bs->sg = 0; | |
1404 | bs->growable = 0; | |
de9c0cec KW |
1405 | QDECREF(bs->options); |
1406 | bs->options = NULL; | |
b338082b | 1407 | |
66f82cee | 1408 | if (bs->file != NULL) { |
0ac9377d PB |
1409 | bdrv_delete(bs->file); |
1410 | bs->file = NULL; | |
66f82cee | 1411 | } |
b338082b | 1412 | } |
98f90dba | 1413 | |
9ca11154 PH |
1414 | bdrv_dev_change_media_cb(bs, false); |
1415 | ||
98f90dba ZYW |
1416 | /*throttling disk I/O limits*/ |
1417 | if (bs->io_limits_enabled) { | |
1418 | bdrv_io_limits_disable(bs); | |
1419 | } | |
b338082b FB |
1420 | } |
1421 | ||
2bc93fed MK |
1422 | void bdrv_close_all(void) |
1423 | { | |
1424 | BlockDriverState *bs; | |
1425 | ||
1426 | QTAILQ_FOREACH(bs, &bdrv_states, list) { | |
1427 | bdrv_close(bs); | |
1428 | } | |
1429 | } | |
1430 | ||
922453bc SH |
1431 | /* |
1432 | * Wait for pending requests to complete across all BlockDriverStates | |
1433 | * | |
1434 | * This function does not flush data to disk, use bdrv_flush_all() for that | |
1435 | * after calling this function. | |
4c355d53 ZYW |
1436 | * |
1437 | * Note that completion of an asynchronous I/O operation can trigger any | |
1438 | * number of other I/O operations on other devices---for example a coroutine | |
1439 | * can be arbitrarily complex and a constant flow of I/O can come until the | |
1440 | * coroutine is complete. Because of this, it is not possible to have a | |
1441 | * function to drain a single device's I/O queue. | |
922453bc SH |
1442 | */ |
1443 | void bdrv_drain_all(void) | |
1444 | { | |
1445 | BlockDriverState *bs; | |
4c355d53 ZYW |
1446 | bool busy; |
1447 | ||
1448 | do { | |
1449 | busy = qemu_aio_wait(); | |
922453bc | 1450 | |
4c355d53 ZYW |
1451 | /* FIXME: We do not have timer support here, so this is effectively |
1452 | * a busy wait. | |
1453 | */ | |
1454 | QTAILQ_FOREACH(bs, &bdrv_states, list) { | |
1455 | if (!qemu_co_queue_empty(&bs->throttled_reqs)) { | |
1456 | qemu_co_queue_restart_all(&bs->throttled_reqs); | |
1457 | busy = true; | |
1458 | } | |
1459 | } | |
1460 | } while (busy); | |
922453bc SH |
1461 | |
1462 | /* If requests are still pending there is a bug somewhere */ | |
1463 | QTAILQ_FOREACH(bs, &bdrv_states, list) { | |
1464 | assert(QLIST_EMPTY(&bs->tracked_requests)); | |
1465 | assert(qemu_co_queue_empty(&bs->throttled_reqs)); | |
1466 | } | |
1467 | } | |
1468 | ||
d22b2f41 RH |
1469 | /* make a BlockDriverState anonymous by removing from bdrv_state list. |
1470 | Also, NULL terminate the device_name to prevent double remove */ | |
1471 | void bdrv_make_anon(BlockDriverState *bs) | |
1472 | { | |
1473 | if (bs->device_name[0] != '\0') { | |
1474 | QTAILQ_REMOVE(&bdrv_states, bs, list); | |
1475 | } | |
1476 | bs->device_name[0] = '\0'; | |
1477 | } | |
1478 | ||
e023b2e2 PB |
1479 | static void bdrv_rebind(BlockDriverState *bs) |
1480 | { | |
1481 | if (bs->drv && bs->drv->bdrv_rebind) { | |
1482 | bs->drv->bdrv_rebind(bs); | |
1483 | } | |
1484 | } | |
1485 | ||
4ddc07ca PB |
1486 | static void bdrv_move_feature_fields(BlockDriverState *bs_dest, |
1487 | BlockDriverState *bs_src) | |
8802d1fd | 1488 | { |
4ddc07ca PB |
1489 | /* move some fields that need to stay attached to the device */ |
1490 | bs_dest->open_flags = bs_src->open_flags; | |
8802d1fd JC |
1491 | |
1492 | /* dev info */ | |
4ddc07ca PB |
1493 | bs_dest->dev_ops = bs_src->dev_ops; |
1494 | bs_dest->dev_opaque = bs_src->dev_opaque; | |
1495 | bs_dest->dev = bs_src->dev; | |
1496 | bs_dest->buffer_alignment = bs_src->buffer_alignment; | |
1497 | bs_dest->copy_on_read = bs_src->copy_on_read; | |
8802d1fd | 1498 | |
4ddc07ca | 1499 | bs_dest->enable_write_cache = bs_src->enable_write_cache; |
c4a248a1 | 1500 | |
8802d1fd | 1501 | /* i/o timing parameters */ |
4ddc07ca PB |
1502 | bs_dest->slice_start = bs_src->slice_start; |
1503 | bs_dest->slice_end = bs_src->slice_end; | |
5905fbc9 | 1504 | bs_dest->slice_submitted = bs_src->slice_submitted; |
4ddc07ca | 1505 | bs_dest->io_limits = bs_src->io_limits; |
4ddc07ca PB |
1506 | bs_dest->throttled_reqs = bs_src->throttled_reqs; |
1507 | bs_dest->block_timer = bs_src->block_timer; | |
1508 | bs_dest->io_limits_enabled = bs_src->io_limits_enabled; | |
8802d1fd | 1509 | |
8802d1fd | 1510 | /* r/w error */ |
4ddc07ca PB |
1511 | bs_dest->on_read_error = bs_src->on_read_error; |
1512 | bs_dest->on_write_error = bs_src->on_write_error; | |
8802d1fd JC |
1513 | |
1514 | /* i/o status */ | |
4ddc07ca PB |
1515 | bs_dest->iostatus_enabled = bs_src->iostatus_enabled; |
1516 | bs_dest->iostatus = bs_src->iostatus; | |
8802d1fd | 1517 | |
a9fc4408 | 1518 | /* dirty bitmap */ |
4ddc07ca | 1519 | bs_dest->dirty_bitmap = bs_src->dirty_bitmap; |
a9fc4408 PB |
1520 | |
1521 | /* job */ | |
4ddc07ca PB |
1522 | bs_dest->in_use = bs_src->in_use; |
1523 | bs_dest->job = bs_src->job; | |
a9fc4408 | 1524 | |
8802d1fd | 1525 | /* keep the same entry in bdrv_states */ |
4ddc07ca PB |
1526 | pstrcpy(bs_dest->device_name, sizeof(bs_dest->device_name), |
1527 | bs_src->device_name); | |
1528 | bs_dest->list = bs_src->list; | |
1529 | } | |
8802d1fd | 1530 | |
4ddc07ca PB |
1531 | /* |
1532 | * Swap bs contents for two image chains while they are live, | |
1533 | * while keeping required fields on the BlockDriverState that is | |
1534 | * actually attached to a device. | |
1535 | * | |
1536 | * This will modify the BlockDriverState fields, and swap contents | |
1537 | * between bs_new and bs_old. Both bs_new and bs_old are modified. | |
1538 | * | |
1539 | * bs_new is required to be anonymous. | |
1540 | * | |
1541 | * This function does not create any image files. | |
1542 | */ | |
1543 | void bdrv_swap(BlockDriverState *bs_new, BlockDriverState *bs_old) | |
1544 | { | |
1545 | BlockDriverState tmp; | |
f6801b83 | 1546 | |
4ddc07ca PB |
1547 | /* bs_new must be anonymous and shouldn't have anything fancy enabled */ |
1548 | assert(bs_new->device_name[0] == '\0'); | |
1549 | assert(bs_new->dirty_bitmap == NULL); | |
1550 | assert(bs_new->job == NULL); | |
1551 | assert(bs_new->dev == NULL); | |
1552 | assert(bs_new->in_use == 0); | |
1553 | assert(bs_new->io_limits_enabled == false); | |
1554 | assert(bs_new->block_timer == NULL); | |
8802d1fd | 1555 | |
4ddc07ca PB |
1556 | tmp = *bs_new; |
1557 | *bs_new = *bs_old; | |
1558 | *bs_old = tmp; | |
a9fc4408 | 1559 | |
4ddc07ca PB |
1560 | /* there are some fields that should not be swapped, move them back */ |
1561 | bdrv_move_feature_fields(&tmp, bs_old); | |
1562 | bdrv_move_feature_fields(bs_old, bs_new); | |
1563 | bdrv_move_feature_fields(bs_new, &tmp); | |
8802d1fd | 1564 | |
4ddc07ca PB |
1565 | /* bs_new shouldn't be in bdrv_states even after the swap! */ |
1566 | assert(bs_new->device_name[0] == '\0'); | |
1567 | ||
1568 | /* Check a few fields that should remain attached to the device */ | |
1569 | assert(bs_new->dev == NULL); | |
1570 | assert(bs_new->job == NULL); | |
1571 | assert(bs_new->in_use == 0); | |
1572 | assert(bs_new->io_limits_enabled == false); | |
1573 | assert(bs_new->block_timer == NULL); | |
e023b2e2 PB |
1574 | |
1575 | bdrv_rebind(bs_new); | |
4ddc07ca PB |
1576 | bdrv_rebind(bs_old); |
1577 | } | |
1578 | ||
1579 | /* | |
1580 | * Add new bs contents at the top of an image chain while the chain is | |
1581 | * live, while keeping required fields on the top layer. | |
1582 | * | |
1583 | * This will modify the BlockDriverState fields, and swap contents | |
1584 | * between bs_new and bs_top. Both bs_new and bs_top are modified. | |
1585 | * | |
1586 | * bs_new is required to be anonymous. | |
1587 | * | |
1588 | * This function does not create any image files. | |
1589 | */ | |
1590 | void bdrv_append(BlockDriverState *bs_new, BlockDriverState *bs_top) | |
1591 | { | |
1592 | bdrv_swap(bs_new, bs_top); | |
1593 | ||
1594 | /* The contents of 'tmp' will become bs_top, as we are | |
1595 | * swapping bs_new and bs_top contents. */ | |
1596 | bs_top->backing_hd = bs_new; | |
1597 | bs_top->open_flags &= ~BDRV_O_NO_BACKING; | |
1598 | pstrcpy(bs_top->backing_file, sizeof(bs_top->backing_file), | |
1599 | bs_new->filename); | |
1600 | pstrcpy(bs_top->backing_format, sizeof(bs_top->backing_format), | |
1601 | bs_new->drv ? bs_new->drv->format_name : ""); | |
8802d1fd JC |
1602 | } |
1603 | ||
b338082b FB |
1604 | void bdrv_delete(BlockDriverState *bs) |
1605 | { | |
fa879d62 | 1606 | assert(!bs->dev); |
3e914655 PB |
1607 | assert(!bs->job); |
1608 | assert(!bs->in_use); | |
18846dee | 1609 | |
1b7bdbc1 | 1610 | /* remove from list, if necessary */ |
d22b2f41 | 1611 | bdrv_make_anon(bs); |
34c6f050 | 1612 | |
b338082b | 1613 | bdrv_close(bs); |
66f82cee | 1614 | |
7267c094 | 1615 | g_free(bs); |
fc01f7e7 FB |
1616 | } |
1617 | ||
fa879d62 MA |
1618 | int bdrv_attach_dev(BlockDriverState *bs, void *dev) |
1619 | /* TODO change to DeviceState *dev when all users are qdevified */ | |
18846dee | 1620 | { |
fa879d62 | 1621 | if (bs->dev) { |
18846dee MA |
1622 | return -EBUSY; |
1623 | } | |
fa879d62 | 1624 | bs->dev = dev; |
28a7282a | 1625 | bdrv_iostatus_reset(bs); |
18846dee MA |
1626 | return 0; |
1627 | } | |
1628 | ||
fa879d62 MA |
1629 | /* TODO qdevified devices don't use this, remove when devices are qdevified */ |
1630 | void bdrv_attach_dev_nofail(BlockDriverState *bs, void *dev) | |
18846dee | 1631 | { |
fa879d62 MA |
1632 | if (bdrv_attach_dev(bs, dev) < 0) { |
1633 | abort(); | |
1634 | } | |
1635 | } | |
1636 | ||
1637 | void bdrv_detach_dev(BlockDriverState *bs, void *dev) | |
1638 | /* TODO change to DeviceState *dev when all users are qdevified */ | |
1639 | { | |
1640 | assert(bs->dev == dev); | |
1641 | bs->dev = NULL; | |
0e49de52 MA |
1642 | bs->dev_ops = NULL; |
1643 | bs->dev_opaque = NULL; | |
29e05f20 | 1644 | bs->buffer_alignment = 512; |
18846dee MA |
1645 | } |
1646 | ||
fa879d62 MA |
1647 | /* TODO change to return DeviceState * when all users are qdevified */ |
1648 | void *bdrv_get_attached_dev(BlockDriverState *bs) | |
18846dee | 1649 | { |
fa879d62 | 1650 | return bs->dev; |
18846dee MA |
1651 | } |
1652 | ||
0e49de52 MA |
1653 | void bdrv_set_dev_ops(BlockDriverState *bs, const BlockDevOps *ops, |
1654 | void *opaque) | |
1655 | { | |
1656 | bs->dev_ops = ops; | |
1657 | bs->dev_opaque = opaque; | |
1658 | } | |
1659 | ||
32c81a4a PB |
1660 | void bdrv_emit_qmp_error_event(const BlockDriverState *bdrv, |
1661 | enum MonitorEvent ev, | |
1662 | BlockErrorAction action, bool is_read) | |
329c0a48 LC |
1663 | { |
1664 | QObject *data; | |
1665 | const char *action_str; | |
1666 | ||
1667 | switch (action) { | |
1668 | case BDRV_ACTION_REPORT: | |
1669 | action_str = "report"; | |
1670 | break; | |
1671 | case BDRV_ACTION_IGNORE: | |
1672 | action_str = "ignore"; | |
1673 | break; | |
1674 | case BDRV_ACTION_STOP: | |
1675 | action_str = "stop"; | |
1676 | break; | |
1677 | default: | |
1678 | abort(); | |
1679 | } | |
1680 | ||
1681 | data = qobject_from_jsonf("{ 'device': %s, 'action': %s, 'operation': %s }", | |
1682 | bdrv->device_name, | |
1683 | action_str, | |
1684 | is_read ? "read" : "write"); | |
32c81a4a | 1685 | monitor_protocol_event(ev, data); |
329c0a48 LC |
1686 | |
1687 | qobject_decref(data); | |
1688 | } | |
1689 | ||
6f382ed2 LC |
1690 | static void bdrv_emit_qmp_eject_event(BlockDriverState *bs, bool ejected) |
1691 | { | |
1692 | QObject *data; | |
1693 | ||
1694 | data = qobject_from_jsonf("{ 'device': %s, 'tray-open': %i }", | |
1695 | bdrv_get_device_name(bs), ejected); | |
1696 | monitor_protocol_event(QEVENT_DEVICE_TRAY_MOVED, data); | |
1697 | ||
1698 | qobject_decref(data); | |
1699 | } | |
1700 | ||
7d4b4ba5 | 1701 | static void bdrv_dev_change_media_cb(BlockDriverState *bs, bool load) |
0e49de52 | 1702 | { |
145feb17 | 1703 | if (bs->dev_ops && bs->dev_ops->change_media_cb) { |
6f382ed2 | 1704 | bool tray_was_closed = !bdrv_dev_is_tray_open(bs); |
7d4b4ba5 | 1705 | bs->dev_ops->change_media_cb(bs->dev_opaque, load); |
6f382ed2 LC |
1706 | if (tray_was_closed) { |
1707 | /* tray open */ | |
1708 | bdrv_emit_qmp_eject_event(bs, true); | |
1709 | } | |
1710 | if (load) { | |
1711 | /* tray close */ | |
1712 | bdrv_emit_qmp_eject_event(bs, false); | |
1713 | } | |
145feb17 MA |
1714 | } |
1715 | } | |
1716 | ||
2c6942fa MA |
1717 | bool bdrv_dev_has_removable_media(BlockDriverState *bs) |
1718 | { | |
1719 | return !bs->dev || (bs->dev_ops && bs->dev_ops->change_media_cb); | |
1720 | } | |
1721 | ||
025ccaa7 PB |
1722 | void bdrv_dev_eject_request(BlockDriverState *bs, bool force) |
1723 | { | |
1724 | if (bs->dev_ops && bs->dev_ops->eject_request_cb) { | |
1725 | bs->dev_ops->eject_request_cb(bs->dev_opaque, force); | |
1726 | } | |
1727 | } | |
1728 | ||
e4def80b MA |
1729 | bool bdrv_dev_is_tray_open(BlockDriverState *bs) |
1730 | { | |
1731 | if (bs->dev_ops && bs->dev_ops->is_tray_open) { | |
1732 | return bs->dev_ops->is_tray_open(bs->dev_opaque); | |
1733 | } | |
1734 | return false; | |
1735 | } | |
1736 | ||
145feb17 MA |
1737 | static void bdrv_dev_resize_cb(BlockDriverState *bs) |
1738 | { | |
1739 | if (bs->dev_ops && bs->dev_ops->resize_cb) { | |
1740 | bs->dev_ops->resize_cb(bs->dev_opaque); | |
0e49de52 MA |
1741 | } |
1742 | } | |
1743 | ||
f107639a MA |
1744 | bool bdrv_dev_is_medium_locked(BlockDriverState *bs) |
1745 | { | |
1746 | if (bs->dev_ops && bs->dev_ops->is_medium_locked) { | |
1747 | return bs->dev_ops->is_medium_locked(bs->dev_opaque); | |
1748 | } | |
1749 | return false; | |
1750 | } | |
1751 | ||
e97fc193 AL |
1752 | /* |
1753 | * Run consistency checks on an image | |
1754 | * | |
e076f338 | 1755 | * Returns 0 if the check could be completed (it doesn't mean that the image is |
a1c7273b | 1756 | * free of errors) or -errno when an internal error occurred. The results of the |
e076f338 | 1757 | * check are stored in res. |
e97fc193 | 1758 | */ |
4534ff54 | 1759 | int bdrv_check(BlockDriverState *bs, BdrvCheckResult *res, BdrvCheckMode fix) |
e97fc193 AL |
1760 | { |
1761 | if (bs->drv->bdrv_check == NULL) { | |
1762 | return -ENOTSUP; | |
1763 | } | |
1764 | ||
e076f338 | 1765 | memset(res, 0, sizeof(*res)); |
4534ff54 | 1766 | return bs->drv->bdrv_check(bs, res, fix); |
e97fc193 AL |
1767 | } |
1768 | ||
8a426614 KW |
1769 | #define COMMIT_BUF_SECTORS 2048 |
1770 | ||
33e3963e FB |
1771 | /* commit COW file into the raw image */ |
1772 | int bdrv_commit(BlockDriverState *bs) | |
1773 | { | |
19cb3738 | 1774 | BlockDriver *drv = bs->drv; |
8a426614 KW |
1775 | int64_t sector, total_sectors; |
1776 | int n, ro, open_flags; | |
0bce597d | 1777 | int ret = 0; |
8a426614 | 1778 | uint8_t *buf; |
c2cba3d9 | 1779 | char filename[PATH_MAX]; |
33e3963e | 1780 | |
19cb3738 FB |
1781 | if (!drv) |
1782 | return -ENOMEDIUM; | |
4dca4b63 NS |
1783 | |
1784 | if (!bs->backing_hd) { | |
1785 | return -ENOTSUP; | |
33e3963e FB |
1786 | } |
1787 | ||
2d3735d3 SH |
1788 | if (bdrv_in_use(bs) || bdrv_in_use(bs->backing_hd)) { |
1789 | return -EBUSY; | |
1790 | } | |
1791 | ||
4dca4b63 | 1792 | ro = bs->backing_hd->read_only; |
c2cba3d9 JM |
1793 | /* Use pstrcpy (not strncpy): filename must be NUL-terminated. */ |
1794 | pstrcpy(filename, sizeof(filename), bs->backing_hd->filename); | |
4dca4b63 NS |
1795 | open_flags = bs->backing_hd->open_flags; |
1796 | ||
1797 | if (ro) { | |
0bce597d JC |
1798 | if (bdrv_reopen(bs->backing_hd, open_flags | BDRV_O_RDWR, NULL)) { |
1799 | return -EACCES; | |
4dca4b63 | 1800 | } |
ea2384d3 | 1801 | } |
33e3963e | 1802 | |
6ea44308 | 1803 | total_sectors = bdrv_getlength(bs) >> BDRV_SECTOR_BITS; |
7267c094 | 1804 | buf = g_malloc(COMMIT_BUF_SECTORS * BDRV_SECTOR_SIZE); |
8a426614 KW |
1805 | |
1806 | for (sector = 0; sector < total_sectors; sector += n) { | |
05c4af54 | 1807 | if (bdrv_is_allocated(bs, sector, COMMIT_BUF_SECTORS, &n)) { |
8a426614 KW |
1808 | |
1809 | if (bdrv_read(bs, sector, buf, n) != 0) { | |
1810 | ret = -EIO; | |
1811 | goto ro_cleanup; | |
1812 | } | |
1813 | ||
1814 | if (bdrv_write(bs->backing_hd, sector, buf, n) != 0) { | |
1815 | ret = -EIO; | |
1816 | goto ro_cleanup; | |
1817 | } | |
ea2384d3 | 1818 | } |
33e3963e | 1819 | } |
95389c86 | 1820 | |
1d44952f CH |
1821 | if (drv->bdrv_make_empty) { |
1822 | ret = drv->bdrv_make_empty(bs); | |
1823 | bdrv_flush(bs); | |
1824 | } | |
95389c86 | 1825 | |
3f5075ae CH |
1826 | /* |
1827 | * Make sure all data we wrote to the backing device is actually | |
1828 | * stable on disk. | |
1829 | */ | |
1830 | if (bs->backing_hd) | |
1831 | bdrv_flush(bs->backing_hd); | |
4dca4b63 NS |
1832 | |
1833 | ro_cleanup: | |
7267c094 | 1834 | g_free(buf); |
4dca4b63 NS |
1835 | |
1836 | if (ro) { | |
0bce597d JC |
1837 | /* ignoring error return here */ |
1838 | bdrv_reopen(bs->backing_hd, open_flags & ~BDRV_O_RDWR, NULL); | |
4dca4b63 NS |
1839 | } |
1840 | ||
1d44952f | 1841 | return ret; |
33e3963e FB |
1842 | } |
1843 | ||
e8877497 | 1844 | int bdrv_commit_all(void) |
6ab4b5ab MA |
1845 | { |
1846 | BlockDriverState *bs; | |
1847 | ||
1848 | QTAILQ_FOREACH(bs, &bdrv_states, list) { | |
272d2d8e JC |
1849 | if (bs->drv && bs->backing_hd) { |
1850 | int ret = bdrv_commit(bs); | |
1851 | if (ret < 0) { | |
1852 | return ret; | |
1853 | } | |
e8877497 | 1854 | } |
6ab4b5ab | 1855 | } |
e8877497 | 1856 | return 0; |
6ab4b5ab MA |
1857 | } |
1858 | ||
dbffbdcf SH |
1859 | /** |
1860 | * Remove an active request from the tracked requests list | |
1861 | * | |
1862 | * This function should be called when a tracked request is completing. | |
1863 | */ | |
1864 | static void tracked_request_end(BdrvTrackedRequest *req) | |
1865 | { | |
1866 | QLIST_REMOVE(req, list); | |
f4658285 | 1867 | qemu_co_queue_restart_all(&req->wait_queue); |
dbffbdcf SH |
1868 | } |
1869 | ||
1870 | /** | |
1871 | * Add an active request to the tracked requests list | |
1872 | */ | |
1873 | static void tracked_request_begin(BdrvTrackedRequest *req, | |
1874 | BlockDriverState *bs, | |
1875 | int64_t sector_num, | |
1876 | int nb_sectors, bool is_write) | |
1877 | { | |
1878 | *req = (BdrvTrackedRequest){ | |
1879 | .bs = bs, | |
1880 | .sector_num = sector_num, | |
1881 | .nb_sectors = nb_sectors, | |
1882 | .is_write = is_write, | |
5f8b6491 | 1883 | .co = qemu_coroutine_self(), |
dbffbdcf SH |
1884 | }; |
1885 | ||
f4658285 SH |
1886 | qemu_co_queue_init(&req->wait_queue); |
1887 | ||
dbffbdcf SH |
1888 | QLIST_INSERT_HEAD(&bs->tracked_requests, req, list); |
1889 | } | |
1890 | ||
d83947ac SH |
1891 | /** |
1892 | * Round a region to cluster boundaries | |
1893 | */ | |
343bded4 PB |
1894 | void bdrv_round_to_clusters(BlockDriverState *bs, |
1895 | int64_t sector_num, int nb_sectors, | |
1896 | int64_t *cluster_sector_num, | |
1897 | int *cluster_nb_sectors) | |
d83947ac SH |
1898 | { |
1899 | BlockDriverInfo bdi; | |
1900 | ||
1901 | if (bdrv_get_info(bs, &bdi) < 0 || bdi.cluster_size == 0) { | |
1902 | *cluster_sector_num = sector_num; | |
1903 | *cluster_nb_sectors = nb_sectors; | |
1904 | } else { | |
1905 | int64_t c = bdi.cluster_size / BDRV_SECTOR_SIZE; | |
1906 | *cluster_sector_num = QEMU_ALIGN_DOWN(sector_num, c); | |
1907 | *cluster_nb_sectors = QEMU_ALIGN_UP(sector_num - *cluster_sector_num + | |
1908 | nb_sectors, c); | |
1909 | } | |
1910 | } | |
1911 | ||
f4658285 SH |
1912 | static bool tracked_request_overlaps(BdrvTrackedRequest *req, |
1913 | int64_t sector_num, int nb_sectors) { | |
d83947ac SH |
1914 | /* aaaa bbbb */ |
1915 | if (sector_num >= req->sector_num + req->nb_sectors) { | |
1916 | return false; | |
1917 | } | |
1918 | /* bbbb aaaa */ | |
1919 | if (req->sector_num >= sector_num + nb_sectors) { | |
1920 | return false; | |
1921 | } | |
1922 | return true; | |
f4658285 SH |
1923 | } |
1924 | ||
1925 | static void coroutine_fn wait_for_overlapping_requests(BlockDriverState *bs, | |
1926 | int64_t sector_num, int nb_sectors) | |
1927 | { | |
1928 | BdrvTrackedRequest *req; | |
d83947ac SH |
1929 | int64_t cluster_sector_num; |
1930 | int cluster_nb_sectors; | |
f4658285 SH |
1931 | bool retry; |
1932 | ||
d83947ac SH |
1933 | /* If we touch the same cluster it counts as an overlap. This guarantees |
1934 | * that allocating writes will be serialized and not race with each other | |
1935 | * for the same cluster. For example, in copy-on-read it ensures that the | |
1936 | * CoR read and write operations are atomic and guest writes cannot | |
1937 | * interleave between them. | |
1938 | */ | |
343bded4 PB |
1939 | bdrv_round_to_clusters(bs, sector_num, nb_sectors, |
1940 | &cluster_sector_num, &cluster_nb_sectors); | |
d83947ac | 1941 | |
f4658285 SH |
1942 | do { |
1943 | retry = false; | |
1944 | QLIST_FOREACH(req, &bs->tracked_requests, list) { | |
d83947ac SH |
1945 | if (tracked_request_overlaps(req, cluster_sector_num, |
1946 | cluster_nb_sectors)) { | |
5f8b6491 SH |
1947 | /* Hitting this means there was a reentrant request, for |
1948 | * example, a block driver issuing nested requests. This must | |
1949 | * never happen since it means deadlock. | |
1950 | */ | |
1951 | assert(qemu_coroutine_self() != req->co); | |
1952 | ||
f4658285 SH |
1953 | qemu_co_queue_wait(&req->wait_queue); |
1954 | retry = true; | |
1955 | break; | |
1956 | } | |
1957 | } | |
1958 | } while (retry); | |
1959 | } | |
1960 | ||
756e6736 KW |
1961 | /* |
1962 | * Return values: | |
1963 | * 0 - success | |
1964 | * -EINVAL - backing format specified, but no file | |
1965 | * -ENOSPC - can't update the backing file because no space is left in the | |
1966 | * image file header | |
1967 | * -ENOTSUP - format driver doesn't support changing the backing file | |
1968 | */ | |
1969 | int bdrv_change_backing_file(BlockDriverState *bs, | |
1970 | const char *backing_file, const char *backing_fmt) | |
1971 | { | |
1972 | BlockDriver *drv = bs->drv; | |
469ef350 | 1973 | int ret; |
756e6736 | 1974 | |
5f377794 PB |
1975 | /* Backing file format doesn't make sense without a backing file */ |
1976 | if (backing_fmt && !backing_file) { | |
1977 | return -EINVAL; | |
1978 | } | |
1979 | ||
756e6736 | 1980 | if (drv->bdrv_change_backing_file != NULL) { |
469ef350 | 1981 | ret = drv->bdrv_change_backing_file(bs, backing_file, backing_fmt); |
756e6736 | 1982 | } else { |
469ef350 | 1983 | ret = -ENOTSUP; |
756e6736 | 1984 | } |
469ef350 PB |
1985 | |
1986 | if (ret == 0) { | |
1987 | pstrcpy(bs->backing_file, sizeof(bs->backing_file), backing_file ?: ""); | |
1988 | pstrcpy(bs->backing_format, sizeof(bs->backing_format), backing_fmt ?: ""); | |
1989 | } | |
1990 | return ret; | |
756e6736 KW |
1991 | } |
1992 | ||
6ebdcee2 JC |
1993 | /* |
1994 | * Finds the image layer in the chain that has 'bs' as its backing file. | |
1995 | * | |
1996 | * active is the current topmost image. | |
1997 | * | |
1998 | * Returns NULL if bs is not found in active's image chain, | |
1999 | * or if active == bs. | |
2000 | */ | |
2001 | BlockDriverState *bdrv_find_overlay(BlockDriverState *active, | |
2002 | BlockDriverState *bs) | |
2003 | { | |
2004 | BlockDriverState *overlay = NULL; | |
2005 | BlockDriverState *intermediate; | |
2006 | ||
2007 | assert(active != NULL); | |
2008 | assert(bs != NULL); | |
2009 | ||
2010 | /* if bs is the same as active, then by definition it has no overlay | |
2011 | */ | |
2012 | if (active == bs) { | |
2013 | return NULL; | |
2014 | } | |
2015 | ||
2016 | intermediate = active; | |
2017 | while (intermediate->backing_hd) { | |
2018 | if (intermediate->backing_hd == bs) { | |
2019 | overlay = intermediate; | |
2020 | break; | |
2021 | } | |
2022 | intermediate = intermediate->backing_hd; | |
2023 | } | |
2024 | ||
2025 | return overlay; | |
2026 | } | |
2027 | ||
2028 | typedef struct BlkIntermediateStates { | |
2029 | BlockDriverState *bs; | |
2030 | QSIMPLEQ_ENTRY(BlkIntermediateStates) entry; | |
2031 | } BlkIntermediateStates; | |
2032 | ||
2033 | ||
2034 | /* | |
2035 | * Drops images above 'base' up to and including 'top', and sets the image | |
2036 | * above 'top' to have base as its backing file. | |
2037 | * | |
2038 | * Requires that the overlay to 'top' is opened r/w, so that the backing file | |
2039 | * information in 'bs' can be properly updated. | |
2040 | * | |
2041 | * E.g., this will convert the following chain: | |
2042 | * bottom <- base <- intermediate <- top <- active | |
2043 | * | |
2044 | * to | |
2045 | * | |
2046 | * bottom <- base <- active | |
2047 | * | |
2048 | * It is allowed for bottom==base, in which case it converts: | |
2049 | * | |
2050 | * base <- intermediate <- top <- active | |
2051 | * | |
2052 | * to | |
2053 | * | |
2054 | * base <- active | |
2055 | * | |
2056 | * Error conditions: | |
2057 | * if active == top, that is considered an error | |
2058 | * | |
2059 | */ | |
2060 | int bdrv_drop_intermediate(BlockDriverState *active, BlockDriverState *top, | |
2061 | BlockDriverState *base) | |
2062 | { | |
2063 | BlockDriverState *intermediate; | |
2064 | BlockDriverState *base_bs = NULL; | |
2065 | BlockDriverState *new_top_bs = NULL; | |
2066 | BlkIntermediateStates *intermediate_state, *next; | |
2067 | int ret = -EIO; | |
2068 | ||
2069 | QSIMPLEQ_HEAD(states_to_delete, BlkIntermediateStates) states_to_delete; | |
2070 | QSIMPLEQ_INIT(&states_to_delete); | |
2071 | ||
2072 | if (!top->drv || !base->drv) { | |
2073 | goto exit; | |
2074 | } | |
2075 | ||
2076 | new_top_bs = bdrv_find_overlay(active, top); | |
2077 | ||
2078 | if (new_top_bs == NULL) { | |
2079 | /* we could not find the image above 'top', this is an error */ | |
2080 | goto exit; | |
2081 | } | |
2082 | ||
2083 | /* special case of new_top_bs->backing_hd already pointing to base - nothing | |
2084 | * to do, no intermediate images */ | |
2085 | if (new_top_bs->backing_hd == base) { | |
2086 | ret = 0; | |
2087 | goto exit; | |
2088 | } | |
2089 | ||
2090 | intermediate = top; | |
2091 | ||
2092 | /* now we will go down through the list, and add each BDS we find | |
2093 | * into our deletion queue, until we hit the 'base' | |
2094 | */ | |
2095 | while (intermediate) { | |
2096 | intermediate_state = g_malloc0(sizeof(BlkIntermediateStates)); | |
2097 | intermediate_state->bs = intermediate; | |
2098 | QSIMPLEQ_INSERT_TAIL(&states_to_delete, intermediate_state, entry); | |
2099 | ||
2100 | if (intermediate->backing_hd == base) { | |
2101 | base_bs = intermediate->backing_hd; | |
2102 | break; | |
2103 | } | |
2104 | intermediate = intermediate->backing_hd; | |
2105 | } | |
2106 | if (base_bs == NULL) { | |
2107 | /* something went wrong, we did not end at the base. safely | |
2108 | * unravel everything, and exit with error */ | |
2109 | goto exit; | |
2110 | } | |
2111 | ||
2112 | /* success - we can delete the intermediate states, and link top->base */ | |
2113 | ret = bdrv_change_backing_file(new_top_bs, base_bs->filename, | |
2114 | base_bs->drv ? base_bs->drv->format_name : ""); | |
2115 | if (ret) { | |
2116 | goto exit; | |
2117 | } | |
2118 | new_top_bs->backing_hd = base_bs; | |
2119 | ||
2120 | ||
2121 | QSIMPLEQ_FOREACH_SAFE(intermediate_state, &states_to_delete, entry, next) { | |
2122 | /* so that bdrv_close() does not recursively close the chain */ | |
2123 | intermediate_state->bs->backing_hd = NULL; | |
2124 | bdrv_delete(intermediate_state->bs); | |
2125 | } | |
2126 | ret = 0; | |
2127 | ||
2128 | exit: | |
2129 | QSIMPLEQ_FOREACH_SAFE(intermediate_state, &states_to_delete, entry, next) { | |
2130 | g_free(intermediate_state); | |
2131 | } | |
2132 | return ret; | |
2133 | } | |
2134 | ||
2135 | ||
71d0770c AL |
2136 | static int bdrv_check_byte_request(BlockDriverState *bs, int64_t offset, |
2137 | size_t size) | |
2138 | { | |
2139 | int64_t len; | |
2140 | ||
2141 | if (!bdrv_is_inserted(bs)) | |
2142 | return -ENOMEDIUM; | |
2143 | ||
2144 | if (bs->growable) | |
2145 | return 0; | |
2146 | ||
2147 | len = bdrv_getlength(bs); | |
2148 | ||
fbb7b4e0 KW |
2149 | if (offset < 0) |
2150 | return -EIO; | |
2151 | ||
2152 | if ((offset > len) || (len - offset < size)) | |
71d0770c AL |
2153 | return -EIO; |
2154 | ||
2155 | return 0; | |
2156 | } | |
2157 | ||
2158 | static int bdrv_check_request(BlockDriverState *bs, int64_t sector_num, | |
2159 | int nb_sectors) | |
2160 | { | |
eb5a3165 JS |
2161 | return bdrv_check_byte_request(bs, sector_num * BDRV_SECTOR_SIZE, |
2162 | nb_sectors * BDRV_SECTOR_SIZE); | |
71d0770c AL |
2163 | } |
2164 | ||
1c9805a3 SH |
2165 | typedef struct RwCo { |
2166 | BlockDriverState *bs; | |
2167 | int64_t sector_num; | |
2168 | int nb_sectors; | |
2169 | QEMUIOVector *qiov; | |
2170 | bool is_write; | |
2171 | int ret; | |
4105eaaa | 2172 | BdrvRequestFlags flags; |
1c9805a3 SH |
2173 | } RwCo; |
2174 | ||
2175 | static void coroutine_fn bdrv_rw_co_entry(void *opaque) | |
fc01f7e7 | 2176 | { |
1c9805a3 | 2177 | RwCo *rwco = opaque; |
ea2384d3 | 2178 | |
1c9805a3 SH |
2179 | if (!rwco->is_write) { |
2180 | rwco->ret = bdrv_co_do_readv(rwco->bs, rwco->sector_num, | |
4105eaaa PL |
2181 | rwco->nb_sectors, rwco->qiov, |
2182 | rwco->flags); | |
1c9805a3 SH |
2183 | } else { |
2184 | rwco->ret = bdrv_co_do_writev(rwco->bs, rwco->sector_num, | |
4105eaaa PL |
2185 | rwco->nb_sectors, rwco->qiov, |
2186 | rwco->flags); | |
1c9805a3 SH |
2187 | } |
2188 | } | |
e7a8a783 | 2189 | |
1c9805a3 | 2190 | /* |
8d3b1a2d | 2191 | * Process a vectored synchronous request using coroutines |
1c9805a3 | 2192 | */ |
8d3b1a2d | 2193 | static int bdrv_rwv_co(BlockDriverState *bs, int64_t sector_num, |
4105eaaa PL |
2194 | QEMUIOVector *qiov, bool is_write, |
2195 | BdrvRequestFlags flags) | |
1c9805a3 | 2196 | { |
1c9805a3 SH |
2197 | Coroutine *co; |
2198 | RwCo rwco = { | |
2199 | .bs = bs, | |
2200 | .sector_num = sector_num, | |
8d3b1a2d KW |
2201 | .nb_sectors = qiov->size >> BDRV_SECTOR_BITS, |
2202 | .qiov = qiov, | |
1c9805a3 SH |
2203 | .is_write = is_write, |
2204 | .ret = NOT_DONE, | |
4105eaaa | 2205 | .flags = flags, |
1c9805a3 | 2206 | }; |
8d3b1a2d | 2207 | assert((qiov->size & (BDRV_SECTOR_SIZE - 1)) == 0); |
e7a8a783 | 2208 | |
498e386c ZYW |
2209 | /** |
2210 | * In sync call context, when the vcpu is blocked, this throttling timer | |
2211 | * will not fire; so the I/O throttling function has to be disabled here | |
2212 | * if it has been enabled. | |
2213 | */ | |
2214 | if (bs->io_limits_enabled) { | |
2215 | fprintf(stderr, "Disabling I/O throttling on '%s' due " | |
2216 | "to synchronous I/O.\n", bdrv_get_device_name(bs)); | |
2217 | bdrv_io_limits_disable(bs); | |
2218 | } | |
2219 | ||
1c9805a3 SH |
2220 | if (qemu_in_coroutine()) { |
2221 | /* Fast-path if already in coroutine context */ | |
2222 | bdrv_rw_co_entry(&rwco); | |
2223 | } else { | |
2224 | co = qemu_coroutine_create(bdrv_rw_co_entry); | |
2225 | qemu_coroutine_enter(co, &rwco); | |
2226 | while (rwco.ret == NOT_DONE) { | |
2227 | qemu_aio_wait(); | |
2228 | } | |
2229 | } | |
2230 | return rwco.ret; | |
2231 | } | |
b338082b | 2232 | |
8d3b1a2d KW |
2233 | /* |
2234 | * Process a synchronous request using coroutines | |
2235 | */ | |
2236 | static int bdrv_rw_co(BlockDriverState *bs, int64_t sector_num, uint8_t *buf, | |
4105eaaa | 2237 | int nb_sectors, bool is_write, BdrvRequestFlags flags) |
8d3b1a2d KW |
2238 | { |
2239 | QEMUIOVector qiov; | |
2240 | struct iovec iov = { | |
2241 | .iov_base = (void *)buf, | |
2242 | .iov_len = nb_sectors * BDRV_SECTOR_SIZE, | |
2243 | }; | |
2244 | ||
2245 | qemu_iovec_init_external(&qiov, &iov, 1); | |
4105eaaa | 2246 | return bdrv_rwv_co(bs, sector_num, &qiov, is_write, flags); |
8d3b1a2d KW |
2247 | } |
2248 | ||
1c9805a3 SH |
2249 | /* return < 0 if error. See bdrv_write() for the return codes */ |
2250 | int bdrv_read(BlockDriverState *bs, int64_t sector_num, | |
2251 | uint8_t *buf, int nb_sectors) | |
2252 | { | |
4105eaaa | 2253 | return bdrv_rw_co(bs, sector_num, buf, nb_sectors, false, 0); |
fc01f7e7 FB |
2254 | } |
2255 | ||
07d27a44 MA |
2256 | /* Just like bdrv_read(), but with I/O throttling temporarily disabled */ |
2257 | int bdrv_read_unthrottled(BlockDriverState *bs, int64_t sector_num, | |
2258 | uint8_t *buf, int nb_sectors) | |
2259 | { | |
2260 | bool enabled; | |
2261 | int ret; | |
2262 | ||
2263 | enabled = bs->io_limits_enabled; | |
2264 | bs->io_limits_enabled = false; | |
4e7395e8 | 2265 | ret = bdrv_read(bs, sector_num, buf, nb_sectors); |
07d27a44 MA |
2266 | bs->io_limits_enabled = enabled; |
2267 | return ret; | |
2268 | } | |
2269 | ||
5fafdf24 | 2270 | /* Return < 0 if error. Important errors are: |
19cb3738 FB |
2271 | -EIO generic I/O error (may happen for all errors) |
2272 | -ENOMEDIUM No media inserted. | |
2273 | -EINVAL Invalid sector number or nb_sectors | |
2274 | -EACCES Trying to write a read-only device | |
2275 | */ | |
5fafdf24 | 2276 | int bdrv_write(BlockDriverState *bs, int64_t sector_num, |
fc01f7e7 FB |
2277 | const uint8_t *buf, int nb_sectors) |
2278 | { | |
4105eaaa | 2279 | return bdrv_rw_co(bs, sector_num, (uint8_t *)buf, nb_sectors, true, 0); |
83f64091 FB |
2280 | } |
2281 | ||
8d3b1a2d KW |
2282 | int bdrv_writev(BlockDriverState *bs, int64_t sector_num, QEMUIOVector *qiov) |
2283 | { | |
4105eaaa PL |
2284 | return bdrv_rwv_co(bs, sector_num, qiov, true, 0); |
2285 | } | |
2286 | ||
2287 | int bdrv_write_zeroes(BlockDriverState *bs, int64_t sector_num, int nb_sectors) | |
2288 | { | |
2289 | return bdrv_rw_co(bs, sector_num, NULL, nb_sectors, true, | |
2290 | BDRV_REQ_ZERO_WRITE); | |
8d3b1a2d KW |
2291 | } |
2292 | ||
eda578e5 AL |
2293 | int bdrv_pread(BlockDriverState *bs, int64_t offset, |
2294 | void *buf, int count1) | |
83f64091 | 2295 | { |
6ea44308 | 2296 | uint8_t tmp_buf[BDRV_SECTOR_SIZE]; |
83f64091 FB |
2297 | int len, nb_sectors, count; |
2298 | int64_t sector_num; | |
9a8c4cce | 2299 | int ret; |
83f64091 FB |
2300 | |
2301 | count = count1; | |
2302 | /* first read to align to sector start */ | |
6ea44308 | 2303 | len = (BDRV_SECTOR_SIZE - offset) & (BDRV_SECTOR_SIZE - 1); |
83f64091 FB |
2304 | if (len > count) |
2305 | len = count; | |
6ea44308 | 2306 | sector_num = offset >> BDRV_SECTOR_BITS; |
83f64091 | 2307 | if (len > 0) { |
9a8c4cce KW |
2308 | if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0) |
2309 | return ret; | |
6ea44308 | 2310 | memcpy(buf, tmp_buf + (offset & (BDRV_SECTOR_SIZE - 1)), len); |
83f64091 FB |
2311 | count -= len; |
2312 | if (count == 0) | |
2313 | return count1; | |
2314 | sector_num++; | |
2315 | buf += len; | |
2316 | } | |
2317 | ||
2318 | /* read the sectors "in place" */ | |
6ea44308 | 2319 | nb_sectors = count >> BDRV_SECTOR_BITS; |
83f64091 | 2320 | if (nb_sectors > 0) { |
9a8c4cce KW |
2321 | if ((ret = bdrv_read(bs, sector_num, buf, nb_sectors)) < 0) |
2322 | return ret; | |
83f64091 | 2323 | sector_num += nb_sectors; |
6ea44308 | 2324 | len = nb_sectors << BDRV_SECTOR_BITS; |
83f64091 FB |
2325 | buf += len; |
2326 | count -= len; | |
2327 | } | |
2328 | ||
2329 | /* add data from the last sector */ | |
2330 | if (count > 0) { | |
9a8c4cce KW |
2331 | if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0) |
2332 | return ret; | |
83f64091 FB |
2333 | memcpy(buf, tmp_buf, count); |
2334 | } | |
2335 | return count1; | |
2336 | } | |
2337 | ||
8d3b1a2d | 2338 | int bdrv_pwritev(BlockDriverState *bs, int64_t offset, QEMUIOVector *qiov) |
83f64091 | 2339 | { |
6ea44308 | 2340 | uint8_t tmp_buf[BDRV_SECTOR_SIZE]; |
83f64091 FB |
2341 | int len, nb_sectors, count; |
2342 | int64_t sector_num; | |
9a8c4cce | 2343 | int ret; |
83f64091 | 2344 | |
8d3b1a2d KW |
2345 | count = qiov->size; |
2346 | ||
83f64091 | 2347 | /* first write to align to sector start */ |
6ea44308 | 2348 | len = (BDRV_SECTOR_SIZE - offset) & (BDRV_SECTOR_SIZE - 1); |
83f64091 FB |
2349 | if (len > count) |
2350 | len = count; | |
6ea44308 | 2351 | sector_num = offset >> BDRV_SECTOR_BITS; |
83f64091 | 2352 | if (len > 0) { |
9a8c4cce KW |
2353 | if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0) |
2354 | return ret; | |
8d3b1a2d KW |
2355 | qemu_iovec_to_buf(qiov, 0, tmp_buf + (offset & (BDRV_SECTOR_SIZE - 1)), |
2356 | len); | |
9a8c4cce KW |
2357 | if ((ret = bdrv_write(bs, sector_num, tmp_buf, 1)) < 0) |
2358 | return ret; | |
83f64091 FB |
2359 | count -= len; |
2360 | if (count == 0) | |
8d3b1a2d | 2361 | return qiov->size; |
83f64091 | 2362 | sector_num++; |
83f64091 FB |
2363 | } |
2364 | ||
2365 | /* write the sectors "in place" */ | |
6ea44308 | 2366 | nb_sectors = count >> BDRV_SECTOR_BITS; |
83f64091 | 2367 | if (nb_sectors > 0) { |
8d3b1a2d KW |
2368 | QEMUIOVector qiov_inplace; |
2369 | ||
2370 | qemu_iovec_init(&qiov_inplace, qiov->niov); | |
2371 | qemu_iovec_concat(&qiov_inplace, qiov, len, | |
2372 | nb_sectors << BDRV_SECTOR_BITS); | |
2373 | ret = bdrv_writev(bs, sector_num, &qiov_inplace); | |
2374 | qemu_iovec_destroy(&qiov_inplace); | |
2375 | if (ret < 0) { | |
9a8c4cce | 2376 | return ret; |
8d3b1a2d KW |
2377 | } |
2378 | ||
83f64091 | 2379 | sector_num += nb_sectors; |
6ea44308 | 2380 | len = nb_sectors << BDRV_SECTOR_BITS; |
83f64091 FB |
2381 | count -= len; |
2382 | } | |
2383 | ||
2384 | /* add data from the last sector */ | |
2385 | if (count > 0) { | |
9a8c4cce KW |
2386 | if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0) |
2387 | return ret; | |
8d3b1a2d | 2388 | qemu_iovec_to_buf(qiov, qiov->size - count, tmp_buf, count); |
9a8c4cce KW |
2389 | if ((ret = bdrv_write(bs, sector_num, tmp_buf, 1)) < 0) |
2390 | return ret; | |
83f64091 | 2391 | } |
8d3b1a2d KW |
2392 | return qiov->size; |
2393 | } | |
2394 | ||
2395 | int bdrv_pwrite(BlockDriverState *bs, int64_t offset, | |
2396 | const void *buf, int count1) | |
2397 | { | |
2398 | QEMUIOVector qiov; | |
2399 | struct iovec iov = { | |
2400 | .iov_base = (void *) buf, | |
2401 | .iov_len = count1, | |
2402 | }; | |
2403 | ||
2404 | qemu_iovec_init_external(&qiov, &iov, 1); | |
2405 | return bdrv_pwritev(bs, offset, &qiov); | |
83f64091 | 2406 | } |
83f64091 | 2407 | |
f08145fe KW |
2408 | /* |
2409 | * Writes to the file and ensures that no writes are reordered across this | |
2410 | * request (acts as a barrier) | |
2411 | * | |
2412 | * Returns 0 on success, -errno in error cases. | |
2413 | */ | |
2414 | int bdrv_pwrite_sync(BlockDriverState *bs, int64_t offset, | |
2415 | const void *buf, int count) | |
2416 | { | |
2417 | int ret; | |
2418 | ||
2419 | ret = bdrv_pwrite(bs, offset, buf, count); | |
2420 | if (ret < 0) { | |
2421 | return ret; | |
2422 | } | |
2423 | ||
f05fa4ad PB |
2424 | /* No flush needed for cache modes that already do it */ |
2425 | if (bs->enable_write_cache) { | |
f08145fe KW |
2426 | bdrv_flush(bs); |
2427 | } | |
2428 | ||
2429 | return 0; | |
2430 | } | |
2431 | ||
470c0504 | 2432 | static int coroutine_fn bdrv_co_do_copy_on_readv(BlockDriverState *bs, |
ab185921 SH |
2433 | int64_t sector_num, int nb_sectors, QEMUIOVector *qiov) |
2434 | { | |
2435 | /* Perform I/O through a temporary buffer so that users who scribble over | |
2436 | * their read buffer while the operation is in progress do not end up | |
2437 | * modifying the image file. This is critical for zero-copy guest I/O | |
2438 | * where anything might happen inside guest memory. | |
2439 | */ | |
2440 | void *bounce_buffer; | |
2441 | ||
79c053bd | 2442 | BlockDriver *drv = bs->drv; |
ab185921 SH |
2443 | struct iovec iov; |
2444 | QEMUIOVector bounce_qiov; | |
2445 | int64_t cluster_sector_num; | |
2446 | int cluster_nb_sectors; | |
2447 | size_t skip_bytes; | |
2448 | int ret; | |
2449 | ||
2450 | /* Cover entire cluster so no additional backing file I/O is required when | |
2451 | * allocating cluster in the image file. | |
2452 | */ | |
343bded4 PB |
2453 | bdrv_round_to_clusters(bs, sector_num, nb_sectors, |
2454 | &cluster_sector_num, &cluster_nb_sectors); | |
ab185921 | 2455 | |
470c0504 SH |
2456 | trace_bdrv_co_do_copy_on_readv(bs, sector_num, nb_sectors, |
2457 | cluster_sector_num, cluster_nb_sectors); | |
ab185921 SH |
2458 | |
2459 | iov.iov_len = cluster_nb_sectors * BDRV_SECTOR_SIZE; | |
2460 | iov.iov_base = bounce_buffer = qemu_blockalign(bs, iov.iov_len); | |
2461 | qemu_iovec_init_external(&bounce_qiov, &iov, 1); | |
2462 | ||
79c053bd SH |
2463 | ret = drv->bdrv_co_readv(bs, cluster_sector_num, cluster_nb_sectors, |
2464 | &bounce_qiov); | |
ab185921 SH |
2465 | if (ret < 0) { |
2466 | goto err; | |
2467 | } | |
2468 | ||
79c053bd SH |
2469 | if (drv->bdrv_co_write_zeroes && |
2470 | buffer_is_zero(bounce_buffer, iov.iov_len)) { | |
621f0589 KW |
2471 | ret = bdrv_co_do_write_zeroes(bs, cluster_sector_num, |
2472 | cluster_nb_sectors); | |
79c053bd | 2473 | } else { |
f05fa4ad PB |
2474 | /* This does not change the data on the disk, it is not necessary |
2475 | * to flush even in cache=writethrough mode. | |
2476 | */ | |
79c053bd | 2477 | ret = drv->bdrv_co_writev(bs, cluster_sector_num, cluster_nb_sectors, |
ab185921 | 2478 | &bounce_qiov); |
79c053bd SH |
2479 | } |
2480 | ||
ab185921 SH |
2481 | if (ret < 0) { |
2482 | /* It might be okay to ignore write errors for guest requests. If this | |
2483 | * is a deliberate copy-on-read then we don't want to ignore the error. | |
2484 | * Simply report it in all cases. | |
2485 | */ | |
2486 | goto err; | |
2487 | } | |
2488 | ||
2489 | skip_bytes = (sector_num - cluster_sector_num) * BDRV_SECTOR_SIZE; | |
03396148 MT |
2490 | qemu_iovec_from_buf(qiov, 0, bounce_buffer + skip_bytes, |
2491 | nb_sectors * BDRV_SECTOR_SIZE); | |
ab185921 SH |
2492 | |
2493 | err: | |
2494 | qemu_vfree(bounce_buffer); | |
2495 | return ret; | |
2496 | } | |
2497 | ||
c5fbe571 SH |
2498 | /* |
2499 | * Handle a read request in coroutine context | |
2500 | */ | |
2501 | static int coroutine_fn bdrv_co_do_readv(BlockDriverState *bs, | |
470c0504 SH |
2502 | int64_t sector_num, int nb_sectors, QEMUIOVector *qiov, |
2503 | BdrvRequestFlags flags) | |
da1fa91d KW |
2504 | { |
2505 | BlockDriver *drv = bs->drv; | |
dbffbdcf SH |
2506 | BdrvTrackedRequest req; |
2507 | int ret; | |
da1fa91d | 2508 | |
da1fa91d KW |
2509 | if (!drv) { |
2510 | return -ENOMEDIUM; | |
2511 | } | |
2512 | if (bdrv_check_request(bs, sector_num, nb_sectors)) { | |
2513 | return -EIO; | |
2514 | } | |
2515 | ||
98f90dba ZYW |
2516 | /* throttling disk read I/O */ |
2517 | if (bs->io_limits_enabled) { | |
2518 | bdrv_io_limits_intercept(bs, false, nb_sectors); | |
2519 | } | |
2520 | ||
f4658285 | 2521 | if (bs->copy_on_read) { |
470c0504 SH |
2522 | flags |= BDRV_REQ_COPY_ON_READ; |
2523 | } | |
2524 | if (flags & BDRV_REQ_COPY_ON_READ) { | |
2525 | bs->copy_on_read_in_flight++; | |
2526 | } | |
2527 | ||
2528 | if (bs->copy_on_read_in_flight) { | |
f4658285 SH |
2529 | wait_for_overlapping_requests(bs, sector_num, nb_sectors); |
2530 | } | |
2531 | ||
dbffbdcf | 2532 | tracked_request_begin(&req, bs, sector_num, nb_sectors, false); |
ab185921 | 2533 | |
470c0504 | 2534 | if (flags & BDRV_REQ_COPY_ON_READ) { |
ab185921 SH |
2535 | int pnum; |
2536 | ||
2537 | ret = bdrv_co_is_allocated(bs, sector_num, nb_sectors, &pnum); | |
2538 | if (ret < 0) { | |
2539 | goto out; | |
2540 | } | |
2541 | ||
2542 | if (!ret || pnum != nb_sectors) { | |
470c0504 | 2543 | ret = bdrv_co_do_copy_on_readv(bs, sector_num, nb_sectors, qiov); |
ab185921 SH |
2544 | goto out; |
2545 | } | |
2546 | } | |
2547 | ||
dbffbdcf | 2548 | ret = drv->bdrv_co_readv(bs, sector_num, nb_sectors, qiov); |
ab185921 SH |
2549 | |
2550 | out: | |
dbffbdcf | 2551 | tracked_request_end(&req); |
470c0504 SH |
2552 | |
2553 | if (flags & BDRV_REQ_COPY_ON_READ) { | |
2554 | bs->copy_on_read_in_flight--; | |
2555 | } | |
2556 | ||
dbffbdcf | 2557 | return ret; |
da1fa91d KW |
2558 | } |
2559 | ||
c5fbe571 | 2560 | int coroutine_fn bdrv_co_readv(BlockDriverState *bs, int64_t sector_num, |
da1fa91d KW |
2561 | int nb_sectors, QEMUIOVector *qiov) |
2562 | { | |
c5fbe571 | 2563 | trace_bdrv_co_readv(bs, sector_num, nb_sectors); |
da1fa91d | 2564 | |
470c0504 SH |
2565 | return bdrv_co_do_readv(bs, sector_num, nb_sectors, qiov, 0); |
2566 | } | |
2567 | ||
2568 | int coroutine_fn bdrv_co_copy_on_readv(BlockDriverState *bs, | |
2569 | int64_t sector_num, int nb_sectors, QEMUIOVector *qiov) | |
2570 | { | |
2571 | trace_bdrv_co_copy_on_readv(bs, sector_num, nb_sectors); | |
2572 | ||
2573 | return bdrv_co_do_readv(bs, sector_num, nb_sectors, qiov, | |
2574 | BDRV_REQ_COPY_ON_READ); | |
c5fbe571 SH |
2575 | } |
2576 | ||
f08f2dda SH |
2577 | static int coroutine_fn bdrv_co_do_write_zeroes(BlockDriverState *bs, |
2578 | int64_t sector_num, int nb_sectors) | |
2579 | { | |
2580 | BlockDriver *drv = bs->drv; | |
2581 | QEMUIOVector qiov; | |
2582 | struct iovec iov; | |
2583 | int ret; | |
2584 | ||
621f0589 KW |
2585 | /* TODO Emulate only part of misaligned requests instead of letting block |
2586 | * drivers return -ENOTSUP and emulate everything */ | |
2587 | ||
f08f2dda SH |
2588 | /* First try the efficient write zeroes operation */ |
2589 | if (drv->bdrv_co_write_zeroes) { | |
621f0589 KW |
2590 | ret = drv->bdrv_co_write_zeroes(bs, sector_num, nb_sectors); |
2591 | if (ret != -ENOTSUP) { | |
2592 | return ret; | |
2593 | } | |
f08f2dda SH |
2594 | } |
2595 | ||
2596 | /* Fall back to bounce buffer if write zeroes is unsupported */ | |
2597 | iov.iov_len = nb_sectors * BDRV_SECTOR_SIZE; | |
2598 | iov.iov_base = qemu_blockalign(bs, iov.iov_len); | |
2599 | memset(iov.iov_base, 0, iov.iov_len); | |
2600 | qemu_iovec_init_external(&qiov, &iov, 1); | |
2601 | ||
2602 | ret = drv->bdrv_co_writev(bs, sector_num, nb_sectors, &qiov); | |
2603 | ||
2604 | qemu_vfree(iov.iov_base); | |
2605 | return ret; | |
2606 | } | |
2607 | ||
c5fbe571 SH |
2608 | /* |
2609 | * Handle a write request in coroutine context | |
2610 | */ | |
2611 | static int coroutine_fn bdrv_co_do_writev(BlockDriverState *bs, | |
f08f2dda SH |
2612 | int64_t sector_num, int nb_sectors, QEMUIOVector *qiov, |
2613 | BdrvRequestFlags flags) | |
c5fbe571 SH |
2614 | { |
2615 | BlockDriver *drv = bs->drv; | |
dbffbdcf | 2616 | BdrvTrackedRequest req; |
6b7cb247 | 2617 | int ret; |
da1fa91d KW |
2618 | |
2619 | if (!bs->drv) { | |
2620 | return -ENOMEDIUM; | |
2621 | } | |
2622 | if (bs->read_only) { | |
2623 | return -EACCES; | |
2624 | } | |
2625 | if (bdrv_check_request(bs, sector_num, nb_sectors)) { | |
2626 | return -EIO; | |
2627 | } | |
2628 | ||
98f90dba ZYW |
2629 | /* throttling disk write I/O */ |
2630 | if (bs->io_limits_enabled) { | |
2631 | bdrv_io_limits_intercept(bs, true, nb_sectors); | |
2632 | } | |
2633 | ||
470c0504 | 2634 | if (bs->copy_on_read_in_flight) { |
f4658285 SH |
2635 | wait_for_overlapping_requests(bs, sector_num, nb_sectors); |
2636 | } | |
2637 | ||
dbffbdcf SH |
2638 | tracked_request_begin(&req, bs, sector_num, nb_sectors, true); |
2639 | ||
d616b224 SH |
2640 | ret = notifier_with_return_list_notify(&bs->before_write_notifiers, &req); |
2641 | ||
2642 | if (ret < 0) { | |
2643 | /* Do nothing, write notifier decided to fail this request */ | |
2644 | } else if (flags & BDRV_REQ_ZERO_WRITE) { | |
f08f2dda SH |
2645 | ret = bdrv_co_do_write_zeroes(bs, sector_num, nb_sectors); |
2646 | } else { | |
2647 | ret = drv->bdrv_co_writev(bs, sector_num, nb_sectors, qiov); | |
2648 | } | |
6b7cb247 | 2649 | |
f05fa4ad PB |
2650 | if (ret == 0 && !bs->enable_write_cache) { |
2651 | ret = bdrv_co_flush(bs); | |
2652 | } | |
2653 | ||
da1fa91d | 2654 | if (bs->dirty_bitmap) { |
1755da16 | 2655 | bdrv_set_dirty(bs, sector_num, nb_sectors); |
da1fa91d KW |
2656 | } |
2657 | ||
2658 | if (bs->wr_highest_sector < sector_num + nb_sectors - 1) { | |
2659 | bs->wr_highest_sector = sector_num + nb_sectors - 1; | |
2660 | } | |
2661 | ||
dbffbdcf SH |
2662 | tracked_request_end(&req); |
2663 | ||
6b7cb247 | 2664 | return ret; |
da1fa91d KW |
2665 | } |
2666 | ||
c5fbe571 SH |
2667 | int coroutine_fn bdrv_co_writev(BlockDriverState *bs, int64_t sector_num, |
2668 | int nb_sectors, QEMUIOVector *qiov) | |
2669 | { | |
2670 | trace_bdrv_co_writev(bs, sector_num, nb_sectors); | |
2671 | ||
f08f2dda SH |
2672 | return bdrv_co_do_writev(bs, sector_num, nb_sectors, qiov, 0); |
2673 | } | |
2674 | ||
2675 | int coroutine_fn bdrv_co_write_zeroes(BlockDriverState *bs, | |
2676 | int64_t sector_num, int nb_sectors) | |
2677 | { | |
2678 | trace_bdrv_co_write_zeroes(bs, sector_num, nb_sectors); | |
2679 | ||
2680 | return bdrv_co_do_writev(bs, sector_num, nb_sectors, NULL, | |
2681 | BDRV_REQ_ZERO_WRITE); | |
c5fbe571 SH |
2682 | } |
2683 | ||
83f64091 FB |
2684 | /** |
2685 | * Truncate file to 'offset' bytes (needed only for file protocols) | |
2686 | */ | |
2687 | int bdrv_truncate(BlockDriverState *bs, int64_t offset) | |
2688 | { | |
2689 | BlockDriver *drv = bs->drv; | |
51762288 | 2690 | int ret; |
83f64091 | 2691 | if (!drv) |
19cb3738 | 2692 | return -ENOMEDIUM; |
83f64091 FB |
2693 | if (!drv->bdrv_truncate) |
2694 | return -ENOTSUP; | |
59f2689d NS |
2695 | if (bs->read_only) |
2696 | return -EACCES; | |
8591675f MT |
2697 | if (bdrv_in_use(bs)) |
2698 | return -EBUSY; | |
51762288 SH |
2699 | ret = drv->bdrv_truncate(bs, offset); |
2700 | if (ret == 0) { | |
2701 | ret = refresh_total_sectors(bs, offset >> BDRV_SECTOR_BITS); | |
145feb17 | 2702 | bdrv_dev_resize_cb(bs); |
51762288 SH |
2703 | } |
2704 | return ret; | |
83f64091 FB |
2705 | } |
2706 | ||
4a1d5e1f FZ |
2707 | /** |
2708 | * Length of a allocated file in bytes. Sparse files are counted by actual | |
2709 | * allocated space. Return < 0 if error or unknown. | |
2710 | */ | |
2711 | int64_t bdrv_get_allocated_file_size(BlockDriverState *bs) | |
2712 | { | |
2713 | BlockDriver *drv = bs->drv; | |
2714 | if (!drv) { | |
2715 | return -ENOMEDIUM; | |
2716 | } | |
2717 | if (drv->bdrv_get_allocated_file_size) { | |
2718 | return drv->bdrv_get_allocated_file_size(bs); | |
2719 | } | |
2720 | if (bs->file) { | |
2721 | return bdrv_get_allocated_file_size(bs->file); | |
2722 | } | |
2723 | return -ENOTSUP; | |
2724 | } | |
2725 | ||
83f64091 FB |
2726 | /** |
2727 | * Length of a file in bytes. Return < 0 if error or unknown. | |
2728 | */ | |
2729 | int64_t bdrv_getlength(BlockDriverState *bs) | |
2730 | { | |
2731 | BlockDriver *drv = bs->drv; | |
2732 | if (!drv) | |
19cb3738 | 2733 | return -ENOMEDIUM; |
51762288 | 2734 | |
2c6942fa | 2735 | if (bs->growable || bdrv_dev_has_removable_media(bs)) { |
46a4e4e6 SH |
2736 | if (drv->bdrv_getlength) { |
2737 | return drv->bdrv_getlength(bs); | |
2738 | } | |
83f64091 | 2739 | } |
46a4e4e6 | 2740 | return bs->total_sectors * BDRV_SECTOR_SIZE; |
fc01f7e7 FB |
2741 | } |
2742 | ||
19cb3738 | 2743 | /* return 0 as number of sectors if no device present or error */ |
96b8f136 | 2744 | void bdrv_get_geometry(BlockDriverState *bs, uint64_t *nb_sectors_ptr) |
fc01f7e7 | 2745 | { |
19cb3738 FB |
2746 | int64_t length; |
2747 | length = bdrv_getlength(bs); | |
2748 | if (length < 0) | |
2749 | length = 0; | |
2750 | else | |
6ea44308 | 2751 | length = length >> BDRV_SECTOR_BITS; |
19cb3738 | 2752 | *nb_sectors_ptr = length; |
fc01f7e7 | 2753 | } |
cf98951b | 2754 | |
0563e191 ZYW |
2755 | /* throttling disk io limits */ |
2756 | void bdrv_set_io_limits(BlockDriverState *bs, | |
2757 | BlockIOLimit *io_limits) | |
2758 | { | |
2759 | bs->io_limits = *io_limits; | |
2760 | bs->io_limits_enabled = bdrv_io_limits_enabled(bs); | |
2761 | } | |
2762 | ||
ff06f5f3 PB |
2763 | void bdrv_set_on_error(BlockDriverState *bs, BlockdevOnError on_read_error, |
2764 | BlockdevOnError on_write_error) | |
abd7f68d MA |
2765 | { |
2766 | bs->on_read_error = on_read_error; | |
2767 | bs->on_write_error = on_write_error; | |
2768 | } | |
2769 | ||
1ceee0d5 | 2770 | BlockdevOnError bdrv_get_on_error(BlockDriverState *bs, bool is_read) |
abd7f68d MA |
2771 | { |
2772 | return is_read ? bs->on_read_error : bs->on_write_error; | |
2773 | } | |
2774 | ||
3e1caa5f PB |
2775 | BlockErrorAction bdrv_get_error_action(BlockDriverState *bs, bool is_read, int error) |
2776 | { | |
2777 | BlockdevOnError on_err = is_read ? bs->on_read_error : bs->on_write_error; | |
2778 | ||
2779 | switch (on_err) { | |
2780 | case BLOCKDEV_ON_ERROR_ENOSPC: | |
2781 | return (error == ENOSPC) ? BDRV_ACTION_STOP : BDRV_ACTION_REPORT; | |
2782 | case BLOCKDEV_ON_ERROR_STOP: | |
2783 | return BDRV_ACTION_STOP; | |
2784 | case BLOCKDEV_ON_ERROR_REPORT: | |
2785 | return BDRV_ACTION_REPORT; | |
2786 | case BLOCKDEV_ON_ERROR_IGNORE: | |
2787 | return BDRV_ACTION_IGNORE; | |
2788 | default: | |
2789 | abort(); | |
2790 | } | |
2791 | } | |
2792 | ||
2793 | /* This is done by device models because, while the block layer knows | |
2794 | * about the error, it does not know whether an operation comes from | |
2795 | * the device or the block layer (from a job, for example). | |
2796 | */ | |
2797 | void bdrv_error_action(BlockDriverState *bs, BlockErrorAction action, | |
2798 | bool is_read, int error) | |
2799 | { | |
2800 | assert(error >= 0); | |
32c81a4a | 2801 | bdrv_emit_qmp_error_event(bs, QEVENT_BLOCK_IO_ERROR, action, is_read); |
3e1caa5f PB |
2802 | if (action == BDRV_ACTION_STOP) { |
2803 | vm_stop(RUN_STATE_IO_ERROR); | |
2804 | bdrv_iostatus_set_err(bs, error); | |
2805 | } | |
2806 | } | |
2807 | ||
b338082b FB |
2808 | int bdrv_is_read_only(BlockDriverState *bs) |
2809 | { | |
2810 | return bs->read_only; | |
2811 | } | |
2812 | ||
985a03b0 TS |
2813 | int bdrv_is_sg(BlockDriverState *bs) |
2814 | { | |
2815 | return bs->sg; | |
2816 | } | |
2817 | ||
e900a7b7 CH |
2818 | int bdrv_enable_write_cache(BlockDriverState *bs) |
2819 | { | |
2820 | return bs->enable_write_cache; | |
2821 | } | |
2822 | ||
425b0148 PB |
2823 | void bdrv_set_enable_write_cache(BlockDriverState *bs, bool wce) |
2824 | { | |
2825 | bs->enable_write_cache = wce; | |
55b110f2 JC |
2826 | |
2827 | /* so a reopen() will preserve wce */ | |
2828 | if (wce) { | |
2829 | bs->open_flags |= BDRV_O_CACHE_WB; | |
2830 | } else { | |
2831 | bs->open_flags &= ~BDRV_O_CACHE_WB; | |
2832 | } | |
425b0148 PB |
2833 | } |
2834 | ||
ea2384d3 FB |
2835 | int bdrv_is_encrypted(BlockDriverState *bs) |
2836 | { | |
2837 | if (bs->backing_hd && bs->backing_hd->encrypted) | |
2838 | return 1; | |
2839 | return bs->encrypted; | |
2840 | } | |
2841 | ||
c0f4ce77 AL |
2842 | int bdrv_key_required(BlockDriverState *bs) |
2843 | { | |
2844 | BlockDriverState *backing_hd = bs->backing_hd; | |
2845 | ||
2846 | if (backing_hd && backing_hd->encrypted && !backing_hd->valid_key) | |
2847 | return 1; | |
2848 | return (bs->encrypted && !bs->valid_key); | |
2849 | } | |
2850 | ||
ea2384d3 FB |
2851 | int bdrv_set_key(BlockDriverState *bs, const char *key) |
2852 | { | |
2853 | int ret; | |
2854 | if (bs->backing_hd && bs->backing_hd->encrypted) { | |
2855 | ret = bdrv_set_key(bs->backing_hd, key); | |
2856 | if (ret < 0) | |
2857 | return ret; | |
2858 | if (!bs->encrypted) | |
2859 | return 0; | |
2860 | } | |
fd04a2ae SH |
2861 | if (!bs->encrypted) { |
2862 | return -EINVAL; | |
2863 | } else if (!bs->drv || !bs->drv->bdrv_set_key) { | |
2864 | return -ENOMEDIUM; | |
2865 | } | |
c0f4ce77 | 2866 | ret = bs->drv->bdrv_set_key(bs, key); |
bb5fc20f AL |
2867 | if (ret < 0) { |
2868 | bs->valid_key = 0; | |
2869 | } else if (!bs->valid_key) { | |
2870 | bs->valid_key = 1; | |
2871 | /* call the change callback now, we skipped it on open */ | |
7d4b4ba5 | 2872 | bdrv_dev_change_media_cb(bs, true); |
bb5fc20f | 2873 | } |
c0f4ce77 | 2874 | return ret; |
ea2384d3 FB |
2875 | } |
2876 | ||
f8d6bba1 | 2877 | const char *bdrv_get_format_name(BlockDriverState *bs) |
ea2384d3 | 2878 | { |
f8d6bba1 | 2879 | return bs->drv ? bs->drv->format_name : NULL; |
ea2384d3 FB |
2880 | } |
2881 | ||
5fafdf24 | 2882 | void bdrv_iterate_format(void (*it)(void *opaque, const char *name), |
ea2384d3 FB |
2883 | void *opaque) |
2884 | { | |
2885 | BlockDriver *drv; | |
2886 | ||
8a22f02a | 2887 | QLIST_FOREACH(drv, &bdrv_drivers, list) { |
ea2384d3 FB |
2888 | it(opaque, drv->format_name); |
2889 | } | |
2890 | } | |
2891 | ||
b338082b FB |
2892 | BlockDriverState *bdrv_find(const char *name) |
2893 | { | |
2894 | BlockDriverState *bs; | |
2895 | ||
1b7bdbc1 SH |
2896 | QTAILQ_FOREACH(bs, &bdrv_states, list) { |
2897 | if (!strcmp(name, bs->device_name)) { | |
b338082b | 2898 | return bs; |
1b7bdbc1 | 2899 | } |
b338082b FB |
2900 | } |
2901 | return NULL; | |
2902 | } | |
2903 | ||
2f399b0a MA |
2904 | BlockDriverState *bdrv_next(BlockDriverState *bs) |
2905 | { | |
2906 | if (!bs) { | |
2907 | return QTAILQ_FIRST(&bdrv_states); | |
2908 | } | |
2909 | return QTAILQ_NEXT(bs, list); | |
2910 | } | |
2911 | ||
51de9760 | 2912 | void bdrv_iterate(void (*it)(void *opaque, BlockDriverState *bs), void *opaque) |
81d0912d FB |
2913 | { |
2914 | BlockDriverState *bs; | |
2915 | ||
1b7bdbc1 | 2916 | QTAILQ_FOREACH(bs, &bdrv_states, list) { |
51de9760 | 2917 | it(opaque, bs); |
81d0912d FB |
2918 | } |
2919 | } | |
2920 | ||
ea2384d3 FB |
2921 | const char *bdrv_get_device_name(BlockDriverState *bs) |
2922 | { | |
2923 | return bs->device_name; | |
2924 | } | |
2925 | ||
c8433287 MA |
2926 | int bdrv_get_flags(BlockDriverState *bs) |
2927 | { | |
2928 | return bs->open_flags; | |
2929 | } | |
2930 | ||
f0f0fdfe | 2931 | int bdrv_flush_all(void) |
c6ca28d6 AL |
2932 | { |
2933 | BlockDriverState *bs; | |
f0f0fdfe | 2934 | int result = 0; |
c6ca28d6 | 2935 | |
1b7bdbc1 | 2936 | QTAILQ_FOREACH(bs, &bdrv_states, list) { |
f0f0fdfe KW |
2937 | int ret = bdrv_flush(bs); |
2938 | if (ret < 0 && !result) { | |
2939 | result = ret; | |
2940 | } | |
1b7bdbc1 | 2941 | } |
f0f0fdfe KW |
2942 | |
2943 | return result; | |
c6ca28d6 AL |
2944 | } |
2945 | ||
3ac21627 PL |
2946 | int bdrv_has_zero_init_1(BlockDriverState *bs) |
2947 | { | |
2948 | return 1; | |
2949 | } | |
2950 | ||
f2feebbd KW |
2951 | int bdrv_has_zero_init(BlockDriverState *bs) |
2952 | { | |
2953 | assert(bs->drv); | |
2954 | ||
336c1c12 KW |
2955 | if (bs->drv->bdrv_has_zero_init) { |
2956 | return bs->drv->bdrv_has_zero_init(bs); | |
f2feebbd KW |
2957 | } |
2958 | ||
3ac21627 PL |
2959 | /* safe default */ |
2960 | return 0; | |
f2feebbd KW |
2961 | } |
2962 | ||
376ae3f1 SH |
2963 | typedef struct BdrvCoIsAllocatedData { |
2964 | BlockDriverState *bs; | |
b35b2bba | 2965 | BlockDriverState *base; |
376ae3f1 SH |
2966 | int64_t sector_num; |
2967 | int nb_sectors; | |
2968 | int *pnum; | |
2969 | int ret; | |
2970 | bool done; | |
2971 | } BdrvCoIsAllocatedData; | |
2972 | ||
f58c7b35 TS |
2973 | /* |
2974 | * Returns true iff the specified sector is present in the disk image. Drivers | |
2975 | * not implementing the functionality are assumed to not support backing files, | |
2976 | * hence all their sectors are reported as allocated. | |
2977 | * | |
bd9533e3 SH |
2978 | * If 'sector_num' is beyond the end of the disk image the return value is 0 |
2979 | * and 'pnum' is set to 0. | |
2980 | * | |
f58c7b35 TS |
2981 | * 'pnum' is set to the number of sectors (including and immediately following |
2982 | * the specified sector) that are known to be in the same | |
2983 | * allocated/unallocated state. | |
2984 | * | |
bd9533e3 SH |
2985 | * 'nb_sectors' is the max value 'pnum' should be set to. If nb_sectors goes |
2986 | * beyond the end of the disk image it will be clamped. | |
f58c7b35 | 2987 | */ |
060f51c9 SH |
2988 | int coroutine_fn bdrv_co_is_allocated(BlockDriverState *bs, int64_t sector_num, |
2989 | int nb_sectors, int *pnum) | |
f58c7b35 | 2990 | { |
bd9533e3 SH |
2991 | int64_t n; |
2992 | ||
2993 | if (sector_num >= bs->total_sectors) { | |
2994 | *pnum = 0; | |
2995 | return 0; | |
2996 | } | |
2997 | ||
2998 | n = bs->total_sectors - sector_num; | |
2999 | if (n < nb_sectors) { | |
3000 | nb_sectors = n; | |
3001 | } | |
3002 | ||
6aebab14 | 3003 | if (!bs->drv->bdrv_co_is_allocated) { |
bd9533e3 | 3004 | *pnum = nb_sectors; |
f58c7b35 TS |
3005 | return 1; |
3006 | } | |
6aebab14 | 3007 | |
060f51c9 SH |
3008 | return bs->drv->bdrv_co_is_allocated(bs, sector_num, nb_sectors, pnum); |
3009 | } | |
3010 | ||
3011 | /* Coroutine wrapper for bdrv_is_allocated() */ | |
3012 | static void coroutine_fn bdrv_is_allocated_co_entry(void *opaque) | |
3013 | { | |
3014 | BdrvCoIsAllocatedData *data = opaque; | |
3015 | BlockDriverState *bs = data->bs; | |
3016 | ||
3017 | data->ret = bdrv_co_is_allocated(bs, data->sector_num, data->nb_sectors, | |
3018 | data->pnum); | |
3019 | data->done = true; | |
3020 | } | |
3021 | ||
3022 | /* | |
3023 | * Synchronous wrapper around bdrv_co_is_allocated(). | |
3024 | * | |
3025 | * See bdrv_co_is_allocated() for details. | |
3026 | */ | |
3027 | int bdrv_is_allocated(BlockDriverState *bs, int64_t sector_num, int nb_sectors, | |
3028 | int *pnum) | |
3029 | { | |
6aebab14 SH |
3030 | Coroutine *co; |
3031 | BdrvCoIsAllocatedData data = { | |
3032 | .bs = bs, | |
3033 | .sector_num = sector_num, | |
3034 | .nb_sectors = nb_sectors, | |
3035 | .pnum = pnum, | |
3036 | .done = false, | |
3037 | }; | |
3038 | ||
3039 | co = qemu_coroutine_create(bdrv_is_allocated_co_entry); | |
3040 | qemu_coroutine_enter(co, &data); | |
3041 | while (!data.done) { | |
3042 | qemu_aio_wait(); | |
3043 | } | |
3044 | return data.ret; | |
f58c7b35 TS |
3045 | } |
3046 | ||
188a7bbf PB |
3047 | /* |
3048 | * Given an image chain: ... -> [BASE] -> [INTER1] -> [INTER2] -> [TOP] | |
3049 | * | |
3050 | * Return true if the given sector is allocated in any image between | |
3051 | * BASE and TOP (inclusive). BASE can be NULL to check if the given | |
3052 | * sector is allocated in any image of the chain. Return false otherwise. | |
3053 | * | |
3054 | * 'pnum' is set to the number of sectors (including and immediately following | |
3055 | * the specified sector) that are known to be in the same | |
3056 | * allocated/unallocated state. | |
3057 | * | |
3058 | */ | |
3059 | int coroutine_fn bdrv_co_is_allocated_above(BlockDriverState *top, | |
3060 | BlockDriverState *base, | |
3061 | int64_t sector_num, | |
3062 | int nb_sectors, int *pnum) | |
3063 | { | |
3064 | BlockDriverState *intermediate; | |
3065 | int ret, n = nb_sectors; | |
3066 | ||
3067 | intermediate = top; | |
3068 | while (intermediate && intermediate != base) { | |
3069 | int pnum_inter; | |
3070 | ret = bdrv_co_is_allocated(intermediate, sector_num, nb_sectors, | |
3071 | &pnum_inter); | |
3072 | if (ret < 0) { | |
3073 | return ret; | |
3074 | } else if (ret) { | |
3075 | *pnum = pnum_inter; | |
3076 | return 1; | |
3077 | } | |
3078 | ||
3079 | /* | |
3080 | * [sector_num, nb_sectors] is unallocated on top but intermediate | |
3081 | * might have | |
3082 | * | |
3083 | * [sector_num+x, nr_sectors] allocated. | |
3084 | */ | |
63ba17d3 VI |
3085 | if (n > pnum_inter && |
3086 | (intermediate == top || | |
3087 | sector_num + pnum_inter < intermediate->total_sectors)) { | |
188a7bbf PB |
3088 | n = pnum_inter; |
3089 | } | |
3090 | ||
3091 | intermediate = intermediate->backing_hd; | |
3092 | } | |
3093 | ||
3094 | *pnum = n; | |
3095 | return 0; | |
3096 | } | |
3097 | ||
b35b2bba MR |
3098 | /* Coroutine wrapper for bdrv_is_allocated_above() */ |
3099 | static void coroutine_fn bdrv_is_allocated_above_co_entry(void *opaque) | |
3100 | { | |
3101 | BdrvCoIsAllocatedData *data = opaque; | |
3102 | BlockDriverState *top = data->bs; | |
3103 | BlockDriverState *base = data->base; | |
3104 | ||
3105 | data->ret = bdrv_co_is_allocated_above(top, base, data->sector_num, | |
3106 | data->nb_sectors, data->pnum); | |
3107 | data->done = true; | |
3108 | } | |
3109 | ||
3110 | /* | |
3111 | * Synchronous wrapper around bdrv_co_is_allocated_above(). | |
3112 | * | |
3113 | * See bdrv_co_is_allocated_above() for details. | |
3114 | */ | |
3115 | int bdrv_is_allocated_above(BlockDriverState *top, BlockDriverState *base, | |
3116 | int64_t sector_num, int nb_sectors, int *pnum) | |
3117 | { | |
3118 | Coroutine *co; | |
3119 | BdrvCoIsAllocatedData data = { | |
3120 | .bs = top, | |
3121 | .base = base, | |
3122 | .sector_num = sector_num, | |
3123 | .nb_sectors = nb_sectors, | |
3124 | .pnum = pnum, | |
3125 | .done = false, | |
3126 | }; | |
3127 | ||
3128 | co = qemu_coroutine_create(bdrv_is_allocated_above_co_entry); | |
3129 | qemu_coroutine_enter(co, &data); | |
3130 | while (!data.done) { | |
3131 | qemu_aio_wait(); | |
3132 | } | |
3133 | return data.ret; | |
3134 | } | |
3135 | ||
045df330 AL |
3136 | const char *bdrv_get_encrypted_filename(BlockDriverState *bs) |
3137 | { | |
3138 | if (bs->backing_hd && bs->backing_hd->encrypted) | |
3139 | return bs->backing_file; | |
3140 | else if (bs->encrypted) | |
3141 | return bs->filename; | |
3142 | else | |
3143 | return NULL; | |
3144 | } | |
3145 | ||
5fafdf24 | 3146 | void bdrv_get_backing_filename(BlockDriverState *bs, |
83f64091 FB |
3147 | char *filename, int filename_size) |
3148 | { | |
3574c608 | 3149 | pstrcpy(filename, filename_size, bs->backing_file); |
83f64091 FB |
3150 | } |
3151 | ||
5fafdf24 | 3152 | int bdrv_write_compressed(BlockDriverState *bs, int64_t sector_num, |
faea38e7 FB |
3153 | const uint8_t *buf, int nb_sectors) |
3154 | { | |
3155 | BlockDriver *drv = bs->drv; | |
3156 | if (!drv) | |
19cb3738 | 3157 | return -ENOMEDIUM; |
faea38e7 FB |
3158 | if (!drv->bdrv_write_compressed) |
3159 | return -ENOTSUP; | |
fbb7b4e0 KW |
3160 | if (bdrv_check_request(bs, sector_num, nb_sectors)) |
3161 | return -EIO; | |
a55eb92c | 3162 | |
1755da16 | 3163 | assert(!bs->dirty_bitmap); |
a55eb92c | 3164 | |
faea38e7 FB |
3165 | return drv->bdrv_write_compressed(bs, sector_num, buf, nb_sectors); |
3166 | } | |
3b46e624 | 3167 | |
faea38e7 FB |
3168 | int bdrv_get_info(BlockDriverState *bs, BlockDriverInfo *bdi) |
3169 | { | |
3170 | BlockDriver *drv = bs->drv; | |
3171 | if (!drv) | |
19cb3738 | 3172 | return -ENOMEDIUM; |
faea38e7 FB |
3173 | if (!drv->bdrv_get_info) |
3174 | return -ENOTSUP; | |
3175 | memset(bdi, 0, sizeof(*bdi)); | |
3176 | return drv->bdrv_get_info(bs, bdi); | |
3177 | } | |
3178 | ||
45566e9c CH |
3179 | int bdrv_save_vmstate(BlockDriverState *bs, const uint8_t *buf, |
3180 | int64_t pos, int size) | |
cf8074b3 KW |
3181 | { |
3182 | QEMUIOVector qiov; | |
3183 | struct iovec iov = { | |
3184 | .iov_base = (void *) buf, | |
3185 | .iov_len = size, | |
3186 | }; | |
3187 | ||
3188 | qemu_iovec_init_external(&qiov, &iov, 1); | |
3189 | return bdrv_writev_vmstate(bs, &qiov, pos); | |
3190 | } | |
3191 | ||
3192 | int bdrv_writev_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos) | |
178e08a5 AL |
3193 | { |
3194 | BlockDriver *drv = bs->drv; | |
cf8074b3 KW |
3195 | |
3196 | if (!drv) { | |
178e08a5 | 3197 | return -ENOMEDIUM; |
cf8074b3 KW |
3198 | } else if (drv->bdrv_save_vmstate) { |
3199 | return drv->bdrv_save_vmstate(bs, qiov, pos); | |
3200 | } else if (bs->file) { | |
3201 | return bdrv_writev_vmstate(bs->file, qiov, pos); | |
3202 | } | |
3203 | ||
7cdb1f6d | 3204 | return -ENOTSUP; |
178e08a5 AL |
3205 | } |
3206 | ||
45566e9c CH |
3207 | int bdrv_load_vmstate(BlockDriverState *bs, uint8_t *buf, |
3208 | int64_t pos, int size) | |
178e08a5 AL |
3209 | { |
3210 | BlockDriver *drv = bs->drv; | |
3211 | if (!drv) | |
3212 | return -ENOMEDIUM; | |
7cdb1f6d MK |
3213 | if (drv->bdrv_load_vmstate) |
3214 | return drv->bdrv_load_vmstate(bs, buf, pos, size); | |
3215 | if (bs->file) | |
3216 | return bdrv_load_vmstate(bs->file, buf, pos, size); | |
3217 | return -ENOTSUP; | |
178e08a5 AL |
3218 | } |
3219 | ||
8b9b0cc2 KW |
3220 | void bdrv_debug_event(BlockDriverState *bs, BlkDebugEvent event) |
3221 | { | |
bf736fe3 | 3222 | if (!bs || !bs->drv || !bs->drv->bdrv_debug_event) { |
8b9b0cc2 KW |
3223 | return; |
3224 | } | |
3225 | ||
bf736fe3 | 3226 | bs->drv->bdrv_debug_event(bs, event); |
41c695c7 KW |
3227 | } |
3228 | ||
3229 | int bdrv_debug_breakpoint(BlockDriverState *bs, const char *event, | |
3230 | const char *tag) | |
3231 | { | |
3232 | while (bs && bs->drv && !bs->drv->bdrv_debug_breakpoint) { | |
3233 | bs = bs->file; | |
3234 | } | |
3235 | ||
3236 | if (bs && bs->drv && bs->drv->bdrv_debug_breakpoint) { | |
3237 | return bs->drv->bdrv_debug_breakpoint(bs, event, tag); | |
3238 | } | |
3239 | ||
3240 | return -ENOTSUP; | |
3241 | } | |
3242 | ||
3243 | int bdrv_debug_resume(BlockDriverState *bs, const char *tag) | |
3244 | { | |
3245 | while (bs && bs->drv && !bs->drv->bdrv_debug_resume) { | |
3246 | bs = bs->file; | |
3247 | } | |
8b9b0cc2 | 3248 | |
41c695c7 KW |
3249 | if (bs && bs->drv && bs->drv->bdrv_debug_resume) { |
3250 | return bs->drv->bdrv_debug_resume(bs, tag); | |
3251 | } | |
3252 | ||
3253 | return -ENOTSUP; | |
3254 | } | |
3255 | ||
3256 | bool bdrv_debug_is_suspended(BlockDriverState *bs, const char *tag) | |
3257 | { | |
3258 | while (bs && bs->drv && !bs->drv->bdrv_debug_is_suspended) { | |
3259 | bs = bs->file; | |
3260 | } | |
3261 | ||
3262 | if (bs && bs->drv && bs->drv->bdrv_debug_is_suspended) { | |
3263 | return bs->drv->bdrv_debug_is_suspended(bs, tag); | |
3264 | } | |
3265 | ||
3266 | return false; | |
8b9b0cc2 KW |
3267 | } |
3268 | ||
199630b6 BS |
3269 | int bdrv_is_snapshot(BlockDriverState *bs) |
3270 | { | |
3271 | return !!(bs->open_flags & BDRV_O_SNAPSHOT); | |
3272 | } | |
3273 | ||
b1b1d783 JC |
3274 | /* backing_file can either be relative, or absolute, or a protocol. If it is |
3275 | * relative, it must be relative to the chain. So, passing in bs->filename | |
3276 | * from a BDS as backing_file should not be done, as that may be relative to | |
3277 | * the CWD rather than the chain. */ | |
e8a6bb9c MT |
3278 | BlockDriverState *bdrv_find_backing_image(BlockDriverState *bs, |
3279 | const char *backing_file) | |
3280 | { | |
b1b1d783 JC |
3281 | char *filename_full = NULL; |
3282 | char *backing_file_full = NULL; | |
3283 | char *filename_tmp = NULL; | |
3284 | int is_protocol = 0; | |
3285 | BlockDriverState *curr_bs = NULL; | |
3286 | BlockDriverState *retval = NULL; | |
3287 | ||
3288 | if (!bs || !bs->drv || !backing_file) { | |
e8a6bb9c MT |
3289 | return NULL; |
3290 | } | |
3291 | ||
b1b1d783 JC |
3292 | filename_full = g_malloc(PATH_MAX); |
3293 | backing_file_full = g_malloc(PATH_MAX); | |
3294 | filename_tmp = g_malloc(PATH_MAX); | |
3295 | ||
3296 | is_protocol = path_has_protocol(backing_file); | |
3297 | ||
3298 | for (curr_bs = bs; curr_bs->backing_hd; curr_bs = curr_bs->backing_hd) { | |
3299 | ||
3300 | /* If either of the filename paths is actually a protocol, then | |
3301 | * compare unmodified paths; otherwise make paths relative */ | |
3302 | if (is_protocol || path_has_protocol(curr_bs->backing_file)) { | |
3303 | if (strcmp(backing_file, curr_bs->backing_file) == 0) { | |
3304 | retval = curr_bs->backing_hd; | |
3305 | break; | |
3306 | } | |
e8a6bb9c | 3307 | } else { |
b1b1d783 JC |
3308 | /* If not an absolute filename path, make it relative to the current |
3309 | * image's filename path */ | |
3310 | path_combine(filename_tmp, PATH_MAX, curr_bs->filename, | |
3311 | backing_file); | |
3312 | ||
3313 | /* We are going to compare absolute pathnames */ | |
3314 | if (!realpath(filename_tmp, filename_full)) { | |
3315 | continue; | |
3316 | } | |
3317 | ||
3318 | /* We need to make sure the backing filename we are comparing against | |
3319 | * is relative to the current image filename (or absolute) */ | |
3320 | path_combine(filename_tmp, PATH_MAX, curr_bs->filename, | |
3321 | curr_bs->backing_file); | |
3322 | ||
3323 | if (!realpath(filename_tmp, backing_file_full)) { | |
3324 | continue; | |
3325 | } | |
3326 | ||
3327 | if (strcmp(backing_file_full, filename_full) == 0) { | |
3328 | retval = curr_bs->backing_hd; | |
3329 | break; | |
3330 | } | |
e8a6bb9c MT |
3331 | } |
3332 | } | |
3333 | ||
b1b1d783 JC |
3334 | g_free(filename_full); |
3335 | g_free(backing_file_full); | |
3336 | g_free(filename_tmp); | |
3337 | return retval; | |
e8a6bb9c MT |
3338 | } |
3339 | ||
f198fd1c BC |
3340 | int bdrv_get_backing_file_depth(BlockDriverState *bs) |
3341 | { | |
3342 | if (!bs->drv) { | |
3343 | return 0; | |
3344 | } | |
3345 | ||
3346 | if (!bs->backing_hd) { | |
3347 | return 0; | |
3348 | } | |
3349 | ||
3350 | return 1 + bdrv_get_backing_file_depth(bs->backing_hd); | |
3351 | } | |
3352 | ||
79fac568 JC |
3353 | BlockDriverState *bdrv_find_base(BlockDriverState *bs) |
3354 | { | |
3355 | BlockDriverState *curr_bs = NULL; | |
3356 | ||
3357 | if (!bs) { | |
3358 | return NULL; | |
3359 | } | |
3360 | ||
3361 | curr_bs = bs; | |
3362 | ||
3363 | while (curr_bs->backing_hd) { | |
3364 | curr_bs = curr_bs->backing_hd; | |
3365 | } | |
3366 | return curr_bs; | |
3367 | } | |
3368 | ||
ea2384d3 | 3369 | /**************************************************************/ |
83f64091 | 3370 | /* async I/Os */ |
ea2384d3 | 3371 | |
3b69e4b9 | 3372 | BlockDriverAIOCB *bdrv_aio_readv(BlockDriverState *bs, int64_t sector_num, |
f141eafe | 3373 | QEMUIOVector *qiov, int nb_sectors, |
3b69e4b9 | 3374 | BlockDriverCompletionFunc *cb, void *opaque) |
83f64091 | 3375 | { |
bbf0a440 SH |
3376 | trace_bdrv_aio_readv(bs, sector_num, nb_sectors, opaque); |
3377 | ||
b2a61371 | 3378 | return bdrv_co_aio_rw_vector(bs, sector_num, qiov, nb_sectors, |
8c5873d6 | 3379 | cb, opaque, false); |
ea2384d3 FB |
3380 | } |
3381 | ||
f141eafe AL |
3382 | BlockDriverAIOCB *bdrv_aio_writev(BlockDriverState *bs, int64_t sector_num, |
3383 | QEMUIOVector *qiov, int nb_sectors, | |
3384 | BlockDriverCompletionFunc *cb, void *opaque) | |
ea2384d3 | 3385 | { |
bbf0a440 SH |
3386 | trace_bdrv_aio_writev(bs, sector_num, nb_sectors, opaque); |
3387 | ||
1a6e115b | 3388 | return bdrv_co_aio_rw_vector(bs, sector_num, qiov, nb_sectors, |
8c5873d6 | 3389 | cb, opaque, true); |
83f64091 FB |
3390 | } |
3391 | ||
40b4f539 KW |
3392 | |
3393 | typedef struct MultiwriteCB { | |
3394 | int error; | |
3395 | int num_requests; | |
3396 | int num_callbacks; | |
3397 | struct { | |
3398 | BlockDriverCompletionFunc *cb; | |
3399 | void *opaque; | |
3400 | QEMUIOVector *free_qiov; | |
40b4f539 KW |
3401 | } callbacks[]; |
3402 | } MultiwriteCB; | |
3403 | ||
3404 | static void multiwrite_user_cb(MultiwriteCB *mcb) | |
3405 | { | |
3406 | int i; | |
3407 | ||
3408 | for (i = 0; i < mcb->num_callbacks; i++) { | |
3409 | mcb->callbacks[i].cb(mcb->callbacks[i].opaque, mcb->error); | |
1e1ea48d SH |
3410 | if (mcb->callbacks[i].free_qiov) { |
3411 | qemu_iovec_destroy(mcb->callbacks[i].free_qiov); | |
3412 | } | |
7267c094 | 3413 | g_free(mcb->callbacks[i].free_qiov); |
40b4f539 KW |
3414 | } |
3415 | } | |
3416 | ||
3417 | static void multiwrite_cb(void *opaque, int ret) | |
3418 | { | |
3419 | MultiwriteCB *mcb = opaque; | |
3420 | ||
6d519a5f SH |
3421 | trace_multiwrite_cb(mcb, ret); |
3422 | ||
cb6d3ca0 | 3423 | if (ret < 0 && !mcb->error) { |
40b4f539 | 3424 | mcb->error = ret; |
40b4f539 KW |
3425 | } |
3426 | ||
3427 | mcb->num_requests--; | |
3428 | if (mcb->num_requests == 0) { | |
de189a1b | 3429 | multiwrite_user_cb(mcb); |
7267c094 | 3430 | g_free(mcb); |
40b4f539 KW |
3431 | } |
3432 | } | |
3433 | ||
3434 | static int multiwrite_req_compare(const void *a, const void *b) | |
3435 | { | |
77be4366 CH |
3436 | const BlockRequest *req1 = a, *req2 = b; |
3437 | ||
3438 | /* | |
3439 | * Note that we can't simply subtract req2->sector from req1->sector | |
3440 | * here as that could overflow the return value. | |
3441 | */ | |
3442 | if (req1->sector > req2->sector) { | |
3443 | return 1; | |
3444 | } else if (req1->sector < req2->sector) { | |
3445 | return -1; | |
3446 | } else { | |
3447 | return 0; | |
3448 | } | |
40b4f539 KW |
3449 | } |
3450 | ||
3451 | /* | |
3452 | * Takes a bunch of requests and tries to merge them. Returns the number of | |
3453 | * requests that remain after merging. | |
3454 | */ | |
3455 | static int multiwrite_merge(BlockDriverState *bs, BlockRequest *reqs, | |
3456 | int num_reqs, MultiwriteCB *mcb) | |
3457 | { | |
3458 | int i, outidx; | |
3459 | ||
3460 | // Sort requests by start sector | |
3461 | qsort(reqs, num_reqs, sizeof(*reqs), &multiwrite_req_compare); | |
3462 | ||
3463 | // Check if adjacent requests touch the same clusters. If so, combine them, | |
3464 | // filling up gaps with zero sectors. | |
3465 | outidx = 0; | |
3466 | for (i = 1; i < num_reqs; i++) { | |
3467 | int merge = 0; | |
3468 | int64_t oldreq_last = reqs[outidx].sector + reqs[outidx].nb_sectors; | |
3469 | ||
b6a127a1 | 3470 | // Handle exactly sequential writes and overlapping writes. |
40b4f539 KW |
3471 | if (reqs[i].sector <= oldreq_last) { |
3472 | merge = 1; | |
3473 | } | |
3474 | ||
e2a305fb CH |
3475 | if (reqs[outidx].qiov->niov + reqs[i].qiov->niov + 1 > IOV_MAX) { |
3476 | merge = 0; | |
3477 | } | |
3478 | ||
40b4f539 KW |
3479 | if (merge) { |
3480 | size_t size; | |
7267c094 | 3481 | QEMUIOVector *qiov = g_malloc0(sizeof(*qiov)); |
40b4f539 KW |
3482 | qemu_iovec_init(qiov, |
3483 | reqs[outidx].qiov->niov + reqs[i].qiov->niov + 1); | |
3484 | ||
3485 | // Add the first request to the merged one. If the requests are | |
3486 | // overlapping, drop the last sectors of the first request. | |
3487 | size = (reqs[i].sector - reqs[outidx].sector) << 9; | |
1b093c48 | 3488 | qemu_iovec_concat(qiov, reqs[outidx].qiov, 0, size); |
40b4f539 | 3489 | |
b6a127a1 PB |
3490 | // We should need to add any zeros between the two requests |
3491 | assert (reqs[i].sector <= oldreq_last); | |
40b4f539 KW |
3492 | |
3493 | // Add the second request | |
1b093c48 | 3494 | qemu_iovec_concat(qiov, reqs[i].qiov, 0, reqs[i].qiov->size); |
40b4f539 | 3495 | |
cbf1dff2 | 3496 | reqs[outidx].nb_sectors = qiov->size >> 9; |
40b4f539 KW |
3497 | reqs[outidx].qiov = qiov; |
3498 | ||
3499 | mcb->callbacks[i].free_qiov = reqs[outidx].qiov; | |
3500 | } else { | |
3501 | outidx++; | |
3502 | reqs[outidx].sector = reqs[i].sector; | |
3503 | reqs[outidx].nb_sectors = reqs[i].nb_sectors; | |
3504 | reqs[outidx].qiov = reqs[i].qiov; | |
3505 | } | |
3506 | } | |
3507 | ||
3508 | return outidx + 1; | |
3509 | } | |
3510 | ||
3511 | /* | |
3512 | * Submit multiple AIO write requests at once. | |
3513 | * | |
3514 | * On success, the function returns 0 and all requests in the reqs array have | |
3515 | * been submitted. In error case this function returns -1, and any of the | |
3516 | * requests may or may not be submitted yet. In particular, this means that the | |
3517 | * callback will be called for some of the requests, for others it won't. The | |
3518 | * caller must check the error field of the BlockRequest to wait for the right | |
3519 | * callbacks (if error != 0, no callback will be called). | |
3520 | * | |
3521 | * The implementation may modify the contents of the reqs array, e.g. to merge | |
3522 | * requests. However, the fields opaque and error are left unmodified as they | |
3523 | * are used to signal failure for a single request to the caller. | |
3524 | */ | |
3525 | int bdrv_aio_multiwrite(BlockDriverState *bs, BlockRequest *reqs, int num_reqs) | |
3526 | { | |
40b4f539 KW |
3527 | MultiwriteCB *mcb; |
3528 | int i; | |
3529 | ||
301db7c2 RH |
3530 | /* don't submit writes if we don't have a medium */ |
3531 | if (bs->drv == NULL) { | |
3532 | for (i = 0; i < num_reqs; i++) { | |
3533 | reqs[i].error = -ENOMEDIUM; | |
3534 | } | |
3535 | return -1; | |
3536 | } | |
3537 | ||
40b4f539 KW |
3538 | if (num_reqs == 0) { |
3539 | return 0; | |
3540 | } | |
3541 | ||
3542 | // Create MultiwriteCB structure | |
7267c094 | 3543 | mcb = g_malloc0(sizeof(*mcb) + num_reqs * sizeof(*mcb->callbacks)); |
40b4f539 KW |
3544 | mcb->num_requests = 0; |
3545 | mcb->num_callbacks = num_reqs; | |
3546 | ||
3547 | for (i = 0; i < num_reqs; i++) { | |
3548 | mcb->callbacks[i].cb = reqs[i].cb; | |
3549 | mcb->callbacks[i].opaque = reqs[i].opaque; | |
3550 | } | |
3551 | ||
3552 | // Check for mergable requests | |
3553 | num_reqs = multiwrite_merge(bs, reqs, num_reqs, mcb); | |
3554 | ||
6d519a5f SH |
3555 | trace_bdrv_aio_multiwrite(mcb, mcb->num_callbacks, num_reqs); |
3556 | ||
df9309fb PB |
3557 | /* Run the aio requests. */ |
3558 | mcb->num_requests = num_reqs; | |
40b4f539 | 3559 | for (i = 0; i < num_reqs; i++) { |
ad54ae80 | 3560 | bdrv_aio_writev(bs, reqs[i].sector, reqs[i].qiov, |
40b4f539 | 3561 | reqs[i].nb_sectors, multiwrite_cb, mcb); |
40b4f539 KW |
3562 | } |
3563 | ||
3564 | return 0; | |
40b4f539 KW |
3565 | } |
3566 | ||
83f64091 | 3567 | void bdrv_aio_cancel(BlockDriverAIOCB *acb) |
83f64091 | 3568 | { |
d7331bed | 3569 | acb->aiocb_info->cancel(acb); |
83f64091 FB |
3570 | } |
3571 | ||
98f90dba ZYW |
3572 | /* block I/O throttling */ |
3573 | static bool bdrv_exceed_bps_limits(BlockDriverState *bs, int nb_sectors, | |
3574 | bool is_write, double elapsed_time, uint64_t *wait) | |
3575 | { | |
3576 | uint64_t bps_limit = 0; | |
ae29d6c6 | 3577 | uint64_t extension; |
98f90dba ZYW |
3578 | double bytes_limit, bytes_base, bytes_res; |
3579 | double slice_time, wait_time; | |
3580 | ||
3581 | if (bs->io_limits.bps[BLOCK_IO_LIMIT_TOTAL]) { | |
3582 | bps_limit = bs->io_limits.bps[BLOCK_IO_LIMIT_TOTAL]; | |
3583 | } else if (bs->io_limits.bps[is_write]) { | |
3584 | bps_limit = bs->io_limits.bps[is_write]; | |
3585 | } else { | |
3586 | if (wait) { | |
3587 | *wait = 0; | |
3588 | } | |
3589 | ||
3590 | return false; | |
3591 | } | |
3592 | ||
3593 | slice_time = bs->slice_end - bs->slice_start; | |
3594 | slice_time /= (NANOSECONDS_PER_SECOND); | |
3595 | bytes_limit = bps_limit * slice_time; | |
5905fbc9 | 3596 | bytes_base = bs->slice_submitted.bytes[is_write]; |
98f90dba | 3597 | if (bs->io_limits.bps[BLOCK_IO_LIMIT_TOTAL]) { |
5905fbc9 | 3598 | bytes_base += bs->slice_submitted.bytes[!is_write]; |
98f90dba ZYW |
3599 | } |
3600 | ||
3601 | /* bytes_base: the bytes of data which have been read/written; and | |
3602 | * it is obtained from the history statistic info. | |
3603 | * bytes_res: the remaining bytes of data which need to be read/written. | |
3604 | * (bytes_base + bytes_res) / bps_limit: used to calcuate | |
3605 | * the total time for completing reading/writting all data. | |
3606 | */ | |
3607 | bytes_res = (unsigned) nb_sectors * BDRV_SECTOR_SIZE; | |
3608 | ||
3609 | if (bytes_base + bytes_res <= bytes_limit) { | |
3610 | if (wait) { | |
3611 | *wait = 0; | |
3612 | } | |
3613 | ||
3614 | return false; | |
3615 | } | |
3616 | ||
3617 | /* Calc approx time to dispatch */ | |
3618 | wait_time = (bytes_base + bytes_res) / bps_limit - elapsed_time; | |
3619 | ||
3620 | /* When the I/O rate at runtime exceeds the limits, | |
3621 | * bs->slice_end need to be extended in order that the current statistic | |
3622 | * info can be kept until the timer fire, so it is increased and tuned | |
3623 | * based on the result of experiment. | |
3624 | */ | |
ae29d6c6 SH |
3625 | extension = wait_time * NANOSECONDS_PER_SECOND; |
3626 | extension = DIV_ROUND_UP(extension, BLOCK_IO_SLICE_TIME) * | |
3627 | BLOCK_IO_SLICE_TIME; | |
3628 | bs->slice_end += extension; | |
98f90dba | 3629 | if (wait) { |
0775437f | 3630 | *wait = wait_time * NANOSECONDS_PER_SECOND; |
98f90dba ZYW |
3631 | } |
3632 | ||
3633 | return true; | |
3634 | } | |
3635 | ||
3636 | static bool bdrv_exceed_iops_limits(BlockDriverState *bs, bool is_write, | |
3637 | double elapsed_time, uint64_t *wait) | |
3638 | { | |
3639 | uint64_t iops_limit = 0; | |
3640 | double ios_limit, ios_base; | |
3641 | double slice_time, wait_time; | |
3642 | ||
3643 | if (bs->io_limits.iops[BLOCK_IO_LIMIT_TOTAL]) { | |
3644 | iops_limit = bs->io_limits.iops[BLOCK_IO_LIMIT_TOTAL]; | |
3645 | } else if (bs->io_limits.iops[is_write]) { | |
3646 | iops_limit = bs->io_limits.iops[is_write]; | |
3647 | } else { | |
3648 | if (wait) { | |
3649 | *wait = 0; | |
3650 | } | |
3651 | ||
3652 | return false; | |
3653 | } | |
3654 | ||
3655 | slice_time = bs->slice_end - bs->slice_start; | |
3656 | slice_time /= (NANOSECONDS_PER_SECOND); | |
3657 | ios_limit = iops_limit * slice_time; | |
5905fbc9 | 3658 | ios_base = bs->slice_submitted.ios[is_write]; |
98f90dba | 3659 | if (bs->io_limits.iops[BLOCK_IO_LIMIT_TOTAL]) { |
5905fbc9 | 3660 | ios_base += bs->slice_submitted.ios[!is_write]; |
98f90dba ZYW |
3661 | } |
3662 | ||
3663 | if (ios_base + 1 <= ios_limit) { | |
3664 | if (wait) { | |
3665 | *wait = 0; | |
3666 | } | |
3667 | ||
3668 | return false; | |
3669 | } | |
3670 | ||
0775437f | 3671 | /* Calc approx time to dispatch, in seconds */ |
98f90dba ZYW |
3672 | wait_time = (ios_base + 1) / iops_limit; |
3673 | if (wait_time > elapsed_time) { | |
3674 | wait_time = wait_time - elapsed_time; | |
3675 | } else { | |
3676 | wait_time = 0; | |
3677 | } | |
3678 | ||
ae29d6c6 SH |
3679 | /* Exceeded current slice, extend it by another slice time */ |
3680 | bs->slice_end += BLOCK_IO_SLICE_TIME; | |
98f90dba | 3681 | if (wait) { |
0775437f | 3682 | *wait = wait_time * NANOSECONDS_PER_SECOND; |
98f90dba ZYW |
3683 | } |
3684 | ||
3685 | return true; | |
3686 | } | |
3687 | ||
3688 | static bool bdrv_exceed_io_limits(BlockDriverState *bs, int nb_sectors, | |
3689 | bool is_write, int64_t *wait) | |
3690 | { | |
3691 | int64_t now, max_wait; | |
3692 | uint64_t bps_wait = 0, iops_wait = 0; | |
3693 | double elapsed_time; | |
3694 | int bps_ret, iops_ret; | |
3695 | ||
3696 | now = qemu_get_clock_ns(vm_clock); | |
e660fb8b | 3697 | if (now > bs->slice_end) { |
98f90dba | 3698 | bs->slice_start = now; |
ae29d6c6 | 3699 | bs->slice_end = now + BLOCK_IO_SLICE_TIME; |
5905fbc9 | 3700 | memset(&bs->slice_submitted, 0, sizeof(bs->slice_submitted)); |
98f90dba ZYW |
3701 | } |
3702 | ||
3703 | elapsed_time = now - bs->slice_start; | |
3704 | elapsed_time /= (NANOSECONDS_PER_SECOND); | |
3705 | ||
3706 | bps_ret = bdrv_exceed_bps_limits(bs, nb_sectors, | |
3707 | is_write, elapsed_time, &bps_wait); | |
3708 | iops_ret = bdrv_exceed_iops_limits(bs, is_write, | |
3709 | elapsed_time, &iops_wait); | |
3710 | if (bps_ret || iops_ret) { | |
3711 | max_wait = bps_wait > iops_wait ? bps_wait : iops_wait; | |
3712 | if (wait) { | |
3713 | *wait = max_wait; | |
3714 | } | |
3715 | ||
3716 | now = qemu_get_clock_ns(vm_clock); | |
3717 | if (bs->slice_end < now + max_wait) { | |
3718 | bs->slice_end = now + max_wait; | |
3719 | } | |
3720 | ||
3721 | return true; | |
3722 | } | |
3723 | ||
3724 | if (wait) { | |
3725 | *wait = 0; | |
3726 | } | |
3727 | ||
5905fbc9 SH |
3728 | bs->slice_submitted.bytes[is_write] += (int64_t)nb_sectors * |
3729 | BDRV_SECTOR_SIZE; | |
3730 | bs->slice_submitted.ios[is_write]++; | |
3731 | ||
98f90dba ZYW |
3732 | return false; |
3733 | } | |
ce1a14dc | 3734 | |
83f64091 FB |
3735 | /**************************************************************/ |
3736 | /* async block device emulation */ | |
3737 | ||
c16b5a2c CH |
3738 | typedef struct BlockDriverAIOCBSync { |
3739 | BlockDriverAIOCB common; | |
3740 | QEMUBH *bh; | |
3741 | int ret; | |
3742 | /* vector translation state */ | |
3743 | QEMUIOVector *qiov; | |
3744 | uint8_t *bounce; | |
3745 | int is_write; | |
3746 | } BlockDriverAIOCBSync; | |
3747 | ||
3748 | static void bdrv_aio_cancel_em(BlockDriverAIOCB *blockacb) | |
3749 | { | |
b666d239 KW |
3750 | BlockDriverAIOCBSync *acb = |
3751 | container_of(blockacb, BlockDriverAIOCBSync, common); | |
6a7ad299 | 3752 | qemu_bh_delete(acb->bh); |
36afc451 | 3753 | acb->bh = NULL; |
c16b5a2c CH |
3754 | qemu_aio_release(acb); |
3755 | } | |
3756 | ||
d7331bed | 3757 | static const AIOCBInfo bdrv_em_aiocb_info = { |
c16b5a2c CH |
3758 | .aiocb_size = sizeof(BlockDriverAIOCBSync), |
3759 | .cancel = bdrv_aio_cancel_em, | |
3760 | }; | |
3761 | ||
ce1a14dc | 3762 | static void bdrv_aio_bh_cb(void *opaque) |
83f64091 | 3763 | { |
ce1a14dc | 3764 | BlockDriverAIOCBSync *acb = opaque; |
f141eafe | 3765 | |
f141eafe | 3766 | if (!acb->is_write) |
03396148 | 3767 | qemu_iovec_from_buf(acb->qiov, 0, acb->bounce, acb->qiov->size); |
ceb42de8 | 3768 | qemu_vfree(acb->bounce); |
ce1a14dc | 3769 | acb->common.cb(acb->common.opaque, acb->ret); |
6a7ad299 | 3770 | qemu_bh_delete(acb->bh); |
36afc451 | 3771 | acb->bh = NULL; |
ce1a14dc | 3772 | qemu_aio_release(acb); |
83f64091 | 3773 | } |
beac80cd | 3774 | |
f141eafe AL |
3775 | static BlockDriverAIOCB *bdrv_aio_rw_vector(BlockDriverState *bs, |
3776 | int64_t sector_num, | |
3777 | QEMUIOVector *qiov, | |
3778 | int nb_sectors, | |
3779 | BlockDriverCompletionFunc *cb, | |
3780 | void *opaque, | |
3781 | int is_write) | |
3782 | ||
83f64091 | 3783 | { |
ce1a14dc | 3784 | BlockDriverAIOCBSync *acb; |
ce1a14dc | 3785 | |
d7331bed | 3786 | acb = qemu_aio_get(&bdrv_em_aiocb_info, bs, cb, opaque); |
f141eafe AL |
3787 | acb->is_write = is_write; |
3788 | acb->qiov = qiov; | |
e268ca52 | 3789 | acb->bounce = qemu_blockalign(bs, qiov->size); |
3f3aace8 | 3790 | acb->bh = qemu_bh_new(bdrv_aio_bh_cb, acb); |
f141eafe AL |
3791 | |
3792 | if (is_write) { | |
d5e6b161 | 3793 | qemu_iovec_to_buf(acb->qiov, 0, acb->bounce, qiov->size); |
1ed20acf | 3794 | acb->ret = bs->drv->bdrv_write(bs, sector_num, acb->bounce, nb_sectors); |
f141eafe | 3795 | } else { |
1ed20acf | 3796 | acb->ret = bs->drv->bdrv_read(bs, sector_num, acb->bounce, nb_sectors); |
f141eafe AL |
3797 | } |
3798 | ||
ce1a14dc | 3799 | qemu_bh_schedule(acb->bh); |
f141eafe | 3800 | |
ce1a14dc | 3801 | return &acb->common; |
beac80cd FB |
3802 | } |
3803 | ||
f141eafe AL |
3804 | static BlockDriverAIOCB *bdrv_aio_readv_em(BlockDriverState *bs, |
3805 | int64_t sector_num, QEMUIOVector *qiov, int nb_sectors, | |
ce1a14dc | 3806 | BlockDriverCompletionFunc *cb, void *opaque) |
beac80cd | 3807 | { |
f141eafe AL |
3808 | return bdrv_aio_rw_vector(bs, sector_num, qiov, nb_sectors, cb, opaque, 0); |
3809 | } | |
83f64091 | 3810 | |
f141eafe AL |
3811 | static BlockDriverAIOCB *bdrv_aio_writev_em(BlockDriverState *bs, |
3812 | int64_t sector_num, QEMUIOVector *qiov, int nb_sectors, | |
3813 | BlockDriverCompletionFunc *cb, void *opaque) | |
3814 | { | |
3815 | return bdrv_aio_rw_vector(bs, sector_num, qiov, nb_sectors, cb, opaque, 1); | |
beac80cd | 3816 | } |
beac80cd | 3817 | |
68485420 KW |
3818 | |
3819 | typedef struct BlockDriverAIOCBCoroutine { | |
3820 | BlockDriverAIOCB common; | |
3821 | BlockRequest req; | |
3822 | bool is_write; | |
d318aea9 | 3823 | bool *done; |
68485420 KW |
3824 | QEMUBH* bh; |
3825 | } BlockDriverAIOCBCoroutine; | |
3826 | ||
3827 | static void bdrv_aio_co_cancel_em(BlockDriverAIOCB *blockacb) | |
3828 | { | |
d318aea9 KW |
3829 | BlockDriverAIOCBCoroutine *acb = |
3830 | container_of(blockacb, BlockDriverAIOCBCoroutine, common); | |
3831 | bool done = false; | |
3832 | ||
3833 | acb->done = &done; | |
3834 | while (!done) { | |
3835 | qemu_aio_wait(); | |
3836 | } | |
68485420 KW |
3837 | } |
3838 | ||
d7331bed | 3839 | static const AIOCBInfo bdrv_em_co_aiocb_info = { |
68485420 KW |
3840 | .aiocb_size = sizeof(BlockDriverAIOCBCoroutine), |
3841 | .cancel = bdrv_aio_co_cancel_em, | |
3842 | }; | |
3843 | ||
35246a68 | 3844 | static void bdrv_co_em_bh(void *opaque) |
68485420 KW |
3845 | { |
3846 | BlockDriverAIOCBCoroutine *acb = opaque; | |
3847 | ||
3848 | acb->common.cb(acb->common.opaque, acb->req.error); | |
d318aea9 KW |
3849 | |
3850 | if (acb->done) { | |
3851 | *acb->done = true; | |
3852 | } | |
3853 | ||
68485420 KW |
3854 | qemu_bh_delete(acb->bh); |
3855 | qemu_aio_release(acb); | |
3856 | } | |
3857 | ||
b2a61371 SH |
3858 | /* Invoke bdrv_co_do_readv/bdrv_co_do_writev */ |
3859 | static void coroutine_fn bdrv_co_do_rw(void *opaque) | |
3860 | { | |
3861 | BlockDriverAIOCBCoroutine *acb = opaque; | |
3862 | BlockDriverState *bs = acb->common.bs; | |
3863 | ||
3864 | if (!acb->is_write) { | |
3865 | acb->req.error = bdrv_co_do_readv(bs, acb->req.sector, | |
470c0504 | 3866 | acb->req.nb_sectors, acb->req.qiov, 0); |
b2a61371 SH |
3867 | } else { |
3868 | acb->req.error = bdrv_co_do_writev(bs, acb->req.sector, | |
f08f2dda | 3869 | acb->req.nb_sectors, acb->req.qiov, 0); |
b2a61371 SH |
3870 | } |
3871 | ||
35246a68 | 3872 | acb->bh = qemu_bh_new(bdrv_co_em_bh, acb); |
b2a61371 SH |
3873 | qemu_bh_schedule(acb->bh); |
3874 | } | |
3875 | ||
68485420 KW |
3876 | static BlockDriverAIOCB *bdrv_co_aio_rw_vector(BlockDriverState *bs, |
3877 | int64_t sector_num, | |
3878 | QEMUIOVector *qiov, | |
3879 | int nb_sectors, | |
3880 | BlockDriverCompletionFunc *cb, | |
3881 | void *opaque, | |
8c5873d6 | 3882 | bool is_write) |
68485420 KW |
3883 | { |
3884 | Coroutine *co; | |
3885 | BlockDriverAIOCBCoroutine *acb; | |
3886 | ||
d7331bed | 3887 | acb = qemu_aio_get(&bdrv_em_co_aiocb_info, bs, cb, opaque); |
68485420 KW |
3888 | acb->req.sector = sector_num; |
3889 | acb->req.nb_sectors = nb_sectors; | |
3890 | acb->req.qiov = qiov; | |
3891 | acb->is_write = is_write; | |
d318aea9 | 3892 | acb->done = NULL; |
68485420 | 3893 | |
8c5873d6 | 3894 | co = qemu_coroutine_create(bdrv_co_do_rw); |
68485420 KW |
3895 | qemu_coroutine_enter(co, acb); |
3896 | ||
3897 | return &acb->common; | |
3898 | } | |
3899 | ||
07f07615 | 3900 | static void coroutine_fn bdrv_aio_flush_co_entry(void *opaque) |
b2e12bc6 | 3901 | { |
07f07615 PB |
3902 | BlockDriverAIOCBCoroutine *acb = opaque; |
3903 | BlockDriverState *bs = acb->common.bs; | |
b2e12bc6 | 3904 | |
07f07615 PB |
3905 | acb->req.error = bdrv_co_flush(bs); |
3906 | acb->bh = qemu_bh_new(bdrv_co_em_bh, acb); | |
b2e12bc6 | 3907 | qemu_bh_schedule(acb->bh); |
b2e12bc6 CH |
3908 | } |
3909 | ||
07f07615 | 3910 | BlockDriverAIOCB *bdrv_aio_flush(BlockDriverState *bs, |
016f5cf6 AG |
3911 | BlockDriverCompletionFunc *cb, void *opaque) |
3912 | { | |
07f07615 | 3913 | trace_bdrv_aio_flush(bs, opaque); |
016f5cf6 | 3914 | |
07f07615 PB |
3915 | Coroutine *co; |
3916 | BlockDriverAIOCBCoroutine *acb; | |
016f5cf6 | 3917 | |
d7331bed | 3918 | acb = qemu_aio_get(&bdrv_em_co_aiocb_info, bs, cb, opaque); |
d318aea9 KW |
3919 | acb->done = NULL; |
3920 | ||
07f07615 PB |
3921 | co = qemu_coroutine_create(bdrv_aio_flush_co_entry); |
3922 | qemu_coroutine_enter(co, acb); | |
016f5cf6 | 3923 | |
016f5cf6 AG |
3924 | return &acb->common; |
3925 | } | |
3926 | ||
4265d620 PB |
3927 | static void coroutine_fn bdrv_aio_discard_co_entry(void *opaque) |
3928 | { | |
3929 | BlockDriverAIOCBCoroutine *acb = opaque; | |
3930 | BlockDriverState *bs = acb->common.bs; | |
3931 | ||
3932 | acb->req.error = bdrv_co_discard(bs, acb->req.sector, acb->req.nb_sectors); | |
3933 | acb->bh = qemu_bh_new(bdrv_co_em_bh, acb); | |
3934 | qemu_bh_schedule(acb->bh); | |
3935 | } | |
3936 | ||
3937 | BlockDriverAIOCB *bdrv_aio_discard(BlockDriverState *bs, | |
3938 | int64_t sector_num, int nb_sectors, | |
3939 | BlockDriverCompletionFunc *cb, void *opaque) | |
3940 | { | |
3941 | Coroutine *co; | |
3942 | BlockDriverAIOCBCoroutine *acb; | |
3943 | ||
3944 | trace_bdrv_aio_discard(bs, sector_num, nb_sectors, opaque); | |
3945 | ||
d7331bed | 3946 | acb = qemu_aio_get(&bdrv_em_co_aiocb_info, bs, cb, opaque); |
4265d620 PB |
3947 | acb->req.sector = sector_num; |
3948 | acb->req.nb_sectors = nb_sectors; | |
d318aea9 | 3949 | acb->done = NULL; |
4265d620 PB |
3950 | co = qemu_coroutine_create(bdrv_aio_discard_co_entry); |
3951 | qemu_coroutine_enter(co, acb); | |
3952 | ||
3953 | return &acb->common; | |
3954 | } | |
3955 | ||
ea2384d3 FB |
3956 | void bdrv_init(void) |
3957 | { | |
5efa9d5a | 3958 | module_call_init(MODULE_INIT_BLOCK); |
ea2384d3 | 3959 | } |
ce1a14dc | 3960 | |
eb852011 MA |
3961 | void bdrv_init_with_whitelist(void) |
3962 | { | |
3963 | use_bdrv_whitelist = 1; | |
3964 | bdrv_init(); | |
3965 | } | |
3966 | ||
d7331bed | 3967 | void *qemu_aio_get(const AIOCBInfo *aiocb_info, BlockDriverState *bs, |
c16b5a2c | 3968 | BlockDriverCompletionFunc *cb, void *opaque) |
ce1a14dc | 3969 | { |
ce1a14dc PB |
3970 | BlockDriverAIOCB *acb; |
3971 | ||
d7331bed SH |
3972 | acb = g_slice_alloc(aiocb_info->aiocb_size); |
3973 | acb->aiocb_info = aiocb_info; | |
ce1a14dc PB |
3974 | acb->bs = bs; |
3975 | acb->cb = cb; | |
3976 | acb->opaque = opaque; | |
3977 | return acb; | |
3978 | } | |
3979 | ||
3980 | void qemu_aio_release(void *p) | |
3981 | { | |
d37c975f | 3982 | BlockDriverAIOCB *acb = p; |
d7331bed | 3983 | g_slice_free1(acb->aiocb_info->aiocb_size, acb); |
ce1a14dc | 3984 | } |
19cb3738 | 3985 | |
f9f05dc5 KW |
3986 | /**************************************************************/ |
3987 | /* Coroutine block device emulation */ | |
3988 | ||
3989 | typedef struct CoroutineIOCompletion { | |
3990 | Coroutine *coroutine; | |
3991 | int ret; | |
3992 | } CoroutineIOCompletion; | |
3993 | ||
3994 | static void bdrv_co_io_em_complete(void *opaque, int ret) | |
3995 | { | |
3996 | CoroutineIOCompletion *co = opaque; | |
3997 | ||
3998 | co->ret = ret; | |
3999 | qemu_coroutine_enter(co->coroutine, NULL); | |
4000 | } | |
4001 | ||
4002 | static int coroutine_fn bdrv_co_io_em(BlockDriverState *bs, int64_t sector_num, | |
4003 | int nb_sectors, QEMUIOVector *iov, | |
4004 | bool is_write) | |
4005 | { | |
4006 | CoroutineIOCompletion co = { | |
4007 | .coroutine = qemu_coroutine_self(), | |
4008 | }; | |
4009 | BlockDriverAIOCB *acb; | |
4010 | ||
4011 | if (is_write) { | |
a652d160 SH |
4012 | acb = bs->drv->bdrv_aio_writev(bs, sector_num, iov, nb_sectors, |
4013 | bdrv_co_io_em_complete, &co); | |
f9f05dc5 | 4014 | } else { |
a652d160 SH |
4015 | acb = bs->drv->bdrv_aio_readv(bs, sector_num, iov, nb_sectors, |
4016 | bdrv_co_io_em_complete, &co); | |
f9f05dc5 KW |
4017 | } |
4018 | ||
59370aaa | 4019 | trace_bdrv_co_io_em(bs, sector_num, nb_sectors, is_write, acb); |
f9f05dc5 KW |
4020 | if (!acb) { |
4021 | return -EIO; | |
4022 | } | |
4023 | qemu_coroutine_yield(); | |
4024 | ||
4025 | return co.ret; | |
4026 | } | |
4027 | ||
4028 | static int coroutine_fn bdrv_co_readv_em(BlockDriverState *bs, | |
4029 | int64_t sector_num, int nb_sectors, | |
4030 | QEMUIOVector *iov) | |
4031 | { | |
4032 | return bdrv_co_io_em(bs, sector_num, nb_sectors, iov, false); | |
4033 | } | |
4034 | ||
4035 | static int coroutine_fn bdrv_co_writev_em(BlockDriverState *bs, | |
4036 | int64_t sector_num, int nb_sectors, | |
4037 | QEMUIOVector *iov) | |
4038 | { | |
4039 | return bdrv_co_io_em(bs, sector_num, nb_sectors, iov, true); | |
4040 | } | |
4041 | ||
07f07615 | 4042 | static void coroutine_fn bdrv_flush_co_entry(void *opaque) |
e7a8a783 | 4043 | { |
07f07615 PB |
4044 | RwCo *rwco = opaque; |
4045 | ||
4046 | rwco->ret = bdrv_co_flush(rwco->bs); | |
4047 | } | |
4048 | ||
4049 | int coroutine_fn bdrv_co_flush(BlockDriverState *bs) | |
4050 | { | |
eb489bb1 KW |
4051 | int ret; |
4052 | ||
29cdb251 | 4053 | if (!bs || !bdrv_is_inserted(bs) || bdrv_is_read_only(bs)) { |
07f07615 | 4054 | return 0; |
eb489bb1 KW |
4055 | } |
4056 | ||
ca716364 | 4057 | /* Write back cached data to the OS even with cache=unsafe */ |
bf736fe3 | 4058 | BLKDBG_EVENT(bs->file, BLKDBG_FLUSH_TO_OS); |
eb489bb1 KW |
4059 | if (bs->drv->bdrv_co_flush_to_os) { |
4060 | ret = bs->drv->bdrv_co_flush_to_os(bs); | |
4061 | if (ret < 0) { | |
4062 | return ret; | |
4063 | } | |
4064 | } | |
4065 | ||
ca716364 KW |
4066 | /* But don't actually force it to the disk with cache=unsafe */ |
4067 | if (bs->open_flags & BDRV_O_NO_FLUSH) { | |
d4c82329 | 4068 | goto flush_parent; |
ca716364 KW |
4069 | } |
4070 | ||
bf736fe3 | 4071 | BLKDBG_EVENT(bs->file, BLKDBG_FLUSH_TO_DISK); |
eb489bb1 | 4072 | if (bs->drv->bdrv_co_flush_to_disk) { |
29cdb251 | 4073 | ret = bs->drv->bdrv_co_flush_to_disk(bs); |
07f07615 PB |
4074 | } else if (bs->drv->bdrv_aio_flush) { |
4075 | BlockDriverAIOCB *acb; | |
4076 | CoroutineIOCompletion co = { | |
4077 | .coroutine = qemu_coroutine_self(), | |
4078 | }; | |
4079 | ||
4080 | acb = bs->drv->bdrv_aio_flush(bs, bdrv_co_io_em_complete, &co); | |
4081 | if (acb == NULL) { | |
29cdb251 | 4082 | ret = -EIO; |
07f07615 PB |
4083 | } else { |
4084 | qemu_coroutine_yield(); | |
29cdb251 | 4085 | ret = co.ret; |
07f07615 | 4086 | } |
07f07615 PB |
4087 | } else { |
4088 | /* | |
4089 | * Some block drivers always operate in either writethrough or unsafe | |
4090 | * mode and don't support bdrv_flush therefore. Usually qemu doesn't | |
4091 | * know how the server works (because the behaviour is hardcoded or | |
4092 | * depends on server-side configuration), so we can't ensure that | |
4093 | * everything is safe on disk. Returning an error doesn't work because | |
4094 | * that would break guests even if the server operates in writethrough | |
4095 | * mode. | |
4096 | * | |
4097 | * Let's hope the user knows what he's doing. | |
4098 | */ | |
29cdb251 | 4099 | ret = 0; |
07f07615 | 4100 | } |
29cdb251 PB |
4101 | if (ret < 0) { |
4102 | return ret; | |
4103 | } | |
4104 | ||
4105 | /* Now flush the underlying protocol. It will also have BDRV_O_NO_FLUSH | |
4106 | * in the case of cache=unsafe, so there are no useless flushes. | |
4107 | */ | |
d4c82329 | 4108 | flush_parent: |
29cdb251 | 4109 | return bdrv_co_flush(bs->file); |
07f07615 PB |
4110 | } |
4111 | ||
0f15423c AL |
4112 | void bdrv_invalidate_cache(BlockDriverState *bs) |
4113 | { | |
4114 | if (bs->drv && bs->drv->bdrv_invalidate_cache) { | |
4115 | bs->drv->bdrv_invalidate_cache(bs); | |
4116 | } | |
4117 | } | |
4118 | ||
4119 | void bdrv_invalidate_cache_all(void) | |
4120 | { | |
4121 | BlockDriverState *bs; | |
4122 | ||
4123 | QTAILQ_FOREACH(bs, &bdrv_states, list) { | |
4124 | bdrv_invalidate_cache(bs); | |
4125 | } | |
4126 | } | |
4127 | ||
07789269 BC |
4128 | void bdrv_clear_incoming_migration_all(void) |
4129 | { | |
4130 | BlockDriverState *bs; | |
4131 | ||
4132 | QTAILQ_FOREACH(bs, &bdrv_states, list) { | |
4133 | bs->open_flags = bs->open_flags & ~(BDRV_O_INCOMING); | |
4134 | } | |
4135 | } | |
4136 | ||
07f07615 PB |
4137 | int bdrv_flush(BlockDriverState *bs) |
4138 | { | |
4139 | Coroutine *co; | |
4140 | RwCo rwco = { | |
4141 | .bs = bs, | |
4142 | .ret = NOT_DONE, | |
e7a8a783 | 4143 | }; |
e7a8a783 | 4144 | |
07f07615 PB |
4145 | if (qemu_in_coroutine()) { |
4146 | /* Fast-path if already in coroutine context */ | |
4147 | bdrv_flush_co_entry(&rwco); | |
4148 | } else { | |
4149 | co = qemu_coroutine_create(bdrv_flush_co_entry); | |
4150 | qemu_coroutine_enter(co, &rwco); | |
4151 | while (rwco.ret == NOT_DONE) { | |
4152 | qemu_aio_wait(); | |
4153 | } | |
e7a8a783 | 4154 | } |
07f07615 PB |
4155 | |
4156 | return rwco.ret; | |
e7a8a783 KW |
4157 | } |
4158 | ||
4265d620 PB |
4159 | static void coroutine_fn bdrv_discard_co_entry(void *opaque) |
4160 | { | |
4161 | RwCo *rwco = opaque; | |
4162 | ||
4163 | rwco->ret = bdrv_co_discard(rwco->bs, rwco->sector_num, rwco->nb_sectors); | |
4164 | } | |
4165 | ||
4166 | int coroutine_fn bdrv_co_discard(BlockDriverState *bs, int64_t sector_num, | |
4167 | int nb_sectors) | |
4168 | { | |
4169 | if (!bs->drv) { | |
4170 | return -ENOMEDIUM; | |
4171 | } else if (bdrv_check_request(bs, sector_num, nb_sectors)) { | |
4172 | return -EIO; | |
4173 | } else if (bs->read_only) { | |
4174 | return -EROFS; | |
df702c9b PB |
4175 | } |
4176 | ||
4177 | if (bs->dirty_bitmap) { | |
8f0720ec | 4178 | bdrv_reset_dirty(bs, sector_num, nb_sectors); |
df702c9b PB |
4179 | } |
4180 | ||
9e8f1835 PB |
4181 | /* Do nothing if disabled. */ |
4182 | if (!(bs->open_flags & BDRV_O_UNMAP)) { | |
4183 | return 0; | |
4184 | } | |
4185 | ||
df702c9b | 4186 | if (bs->drv->bdrv_co_discard) { |
4265d620 PB |
4187 | return bs->drv->bdrv_co_discard(bs, sector_num, nb_sectors); |
4188 | } else if (bs->drv->bdrv_aio_discard) { | |
4189 | BlockDriverAIOCB *acb; | |
4190 | CoroutineIOCompletion co = { | |
4191 | .coroutine = qemu_coroutine_self(), | |
4192 | }; | |
4193 | ||
4194 | acb = bs->drv->bdrv_aio_discard(bs, sector_num, nb_sectors, | |
4195 | bdrv_co_io_em_complete, &co); | |
4196 | if (acb == NULL) { | |
4197 | return -EIO; | |
4198 | } else { | |
4199 | qemu_coroutine_yield(); | |
4200 | return co.ret; | |
4201 | } | |
4265d620 PB |
4202 | } else { |
4203 | return 0; | |
4204 | } | |
4205 | } | |
4206 | ||
4207 | int bdrv_discard(BlockDriverState *bs, int64_t sector_num, int nb_sectors) | |
4208 | { | |
4209 | Coroutine *co; | |
4210 | RwCo rwco = { | |
4211 | .bs = bs, | |
4212 | .sector_num = sector_num, | |
4213 | .nb_sectors = nb_sectors, | |
4214 | .ret = NOT_DONE, | |
4215 | }; | |
4216 | ||
4217 | if (qemu_in_coroutine()) { | |
4218 | /* Fast-path if already in coroutine context */ | |
4219 | bdrv_discard_co_entry(&rwco); | |
4220 | } else { | |
4221 | co = qemu_coroutine_create(bdrv_discard_co_entry); | |
4222 | qemu_coroutine_enter(co, &rwco); | |
4223 | while (rwco.ret == NOT_DONE) { | |
4224 | qemu_aio_wait(); | |
4225 | } | |
4226 | } | |
4227 | ||
4228 | return rwco.ret; | |
4229 | } | |
4230 | ||
19cb3738 FB |
4231 | /**************************************************************/ |
4232 | /* removable device support */ | |
4233 | ||
4234 | /** | |
4235 | * Return TRUE if the media is present | |
4236 | */ | |
4237 | int bdrv_is_inserted(BlockDriverState *bs) | |
4238 | { | |
4239 | BlockDriver *drv = bs->drv; | |
a1aff5bf | 4240 | |
19cb3738 FB |
4241 | if (!drv) |
4242 | return 0; | |
4243 | if (!drv->bdrv_is_inserted) | |
a1aff5bf MA |
4244 | return 1; |
4245 | return drv->bdrv_is_inserted(bs); | |
19cb3738 FB |
4246 | } |
4247 | ||
4248 | /** | |
8e49ca46 MA |
4249 | * Return whether the media changed since the last call to this |
4250 | * function, or -ENOTSUP if we don't know. Most drivers don't know. | |
19cb3738 FB |
4251 | */ |
4252 | int bdrv_media_changed(BlockDriverState *bs) | |
4253 | { | |
4254 | BlockDriver *drv = bs->drv; | |
19cb3738 | 4255 | |
8e49ca46 MA |
4256 | if (drv && drv->bdrv_media_changed) { |
4257 | return drv->bdrv_media_changed(bs); | |
4258 | } | |
4259 | return -ENOTSUP; | |
19cb3738 FB |
4260 | } |
4261 | ||
4262 | /** | |
4263 | * If eject_flag is TRUE, eject the media. Otherwise, close the tray | |
4264 | */ | |
f36f3949 | 4265 | void bdrv_eject(BlockDriverState *bs, bool eject_flag) |
19cb3738 FB |
4266 | { |
4267 | BlockDriver *drv = bs->drv; | |
19cb3738 | 4268 | |
822e1cd1 MA |
4269 | if (drv && drv->bdrv_eject) { |
4270 | drv->bdrv_eject(bs, eject_flag); | |
19cb3738 | 4271 | } |
6f382ed2 LC |
4272 | |
4273 | if (bs->device_name[0] != '\0') { | |
4274 | bdrv_emit_qmp_eject_event(bs, eject_flag); | |
4275 | } | |
19cb3738 FB |
4276 | } |
4277 | ||
19cb3738 FB |
4278 | /** |
4279 | * Lock or unlock the media (if it is locked, the user won't be able | |
4280 | * to eject it manually). | |
4281 | */ | |
025e849a | 4282 | void bdrv_lock_medium(BlockDriverState *bs, bool locked) |
19cb3738 FB |
4283 | { |
4284 | BlockDriver *drv = bs->drv; | |
4285 | ||
025e849a | 4286 | trace_bdrv_lock_medium(bs, locked); |
b8c6d095 | 4287 | |
025e849a MA |
4288 | if (drv && drv->bdrv_lock_medium) { |
4289 | drv->bdrv_lock_medium(bs, locked); | |
19cb3738 FB |
4290 | } |
4291 | } | |
985a03b0 TS |
4292 | |
4293 | /* needed for generic scsi interface */ | |
4294 | ||
4295 | int bdrv_ioctl(BlockDriverState *bs, unsigned long int req, void *buf) | |
4296 | { | |
4297 | BlockDriver *drv = bs->drv; | |
4298 | ||
4299 | if (drv && drv->bdrv_ioctl) | |
4300 | return drv->bdrv_ioctl(bs, req, buf); | |
4301 | return -ENOTSUP; | |
4302 | } | |
7d780669 | 4303 | |
221f715d AL |
4304 | BlockDriverAIOCB *bdrv_aio_ioctl(BlockDriverState *bs, |
4305 | unsigned long int req, void *buf, | |
4306 | BlockDriverCompletionFunc *cb, void *opaque) | |
7d780669 | 4307 | { |
221f715d | 4308 | BlockDriver *drv = bs->drv; |
7d780669 | 4309 | |
221f715d AL |
4310 | if (drv && drv->bdrv_aio_ioctl) |
4311 | return drv->bdrv_aio_ioctl(bs, req, buf, cb, opaque); | |
4312 | return NULL; | |
7d780669 | 4313 | } |
e268ca52 | 4314 | |
7b6f9300 MA |
4315 | void bdrv_set_buffer_alignment(BlockDriverState *bs, int align) |
4316 | { | |
4317 | bs->buffer_alignment = align; | |
4318 | } | |
7cd1e32a | 4319 | |
e268ca52 AL |
4320 | void *qemu_blockalign(BlockDriverState *bs, size_t size) |
4321 | { | |
4322 | return qemu_memalign((bs && bs->buffer_alignment) ? bs->buffer_alignment : 512, size); | |
4323 | } | |
7cd1e32a | 4324 | |
c53b1c51 SH |
4325 | /* |
4326 | * Check if all memory in this vector is sector aligned. | |
4327 | */ | |
4328 | bool bdrv_qiov_is_aligned(BlockDriverState *bs, QEMUIOVector *qiov) | |
4329 | { | |
4330 | int i; | |
4331 | ||
4332 | for (i = 0; i < qiov->niov; i++) { | |
4333 | if ((uintptr_t) qiov->iov[i].iov_base % bs->buffer_alignment) { | |
4334 | return false; | |
4335 | } | |
4336 | } | |
4337 | ||
4338 | return true; | |
4339 | } | |
4340 | ||
50717e94 | 4341 | void bdrv_set_dirty_tracking(BlockDriverState *bs, int granularity) |
7cd1e32a | 4342 | { |
4343 | int64_t bitmap_size; | |
a55eb92c | 4344 | |
50717e94 PB |
4345 | assert((granularity & (granularity - 1)) == 0); |
4346 | ||
4347 | if (granularity) { | |
4348 | granularity >>= BDRV_SECTOR_BITS; | |
4349 | assert(!bs->dirty_bitmap); | |
4350 | bitmap_size = (bdrv_getlength(bs) >> BDRV_SECTOR_BITS); | |
4351 | bs->dirty_bitmap = hbitmap_alloc(bitmap_size, ffs(granularity) - 1); | |
7cd1e32a | 4352 | } else { |
c6d22830 | 4353 | if (bs->dirty_bitmap) { |
8f0720ec | 4354 | hbitmap_free(bs->dirty_bitmap); |
c6d22830 | 4355 | bs->dirty_bitmap = NULL; |
a55eb92c | 4356 | } |
7cd1e32a | 4357 | } |
4358 | } | |
4359 | ||
4360 | int bdrv_get_dirty(BlockDriverState *bs, int64_t sector) | |
4361 | { | |
8f0720ec PB |
4362 | if (bs->dirty_bitmap) { |
4363 | return hbitmap_get(bs->dirty_bitmap, sector); | |
7cd1e32a | 4364 | } else { |
4365 | return 0; | |
4366 | } | |
4367 | } | |
4368 | ||
8f0720ec | 4369 | void bdrv_dirty_iter_init(BlockDriverState *bs, HBitmapIter *hbi) |
1755da16 | 4370 | { |
8f0720ec | 4371 | hbitmap_iter_init(hbi, bs->dirty_bitmap, 0); |
1755da16 PB |
4372 | } |
4373 | ||
4374 | void bdrv_set_dirty(BlockDriverState *bs, int64_t cur_sector, | |
4375 | int nr_sectors) | |
4376 | { | |
8f0720ec | 4377 | hbitmap_set(bs->dirty_bitmap, cur_sector, nr_sectors); |
1755da16 PB |
4378 | } |
4379 | ||
a55eb92c JK |
4380 | void bdrv_reset_dirty(BlockDriverState *bs, int64_t cur_sector, |
4381 | int nr_sectors) | |
7cd1e32a | 4382 | { |
8f0720ec | 4383 | hbitmap_reset(bs->dirty_bitmap, cur_sector, nr_sectors); |
7cd1e32a | 4384 | } |
aaa0eb75 LS |
4385 | |
4386 | int64_t bdrv_get_dirty_count(BlockDriverState *bs) | |
4387 | { | |
8f0720ec | 4388 | if (bs->dirty_bitmap) { |
acc906c6 | 4389 | return hbitmap_count(bs->dirty_bitmap); |
8f0720ec PB |
4390 | } else { |
4391 | return 0; | |
4392 | } | |
aaa0eb75 | 4393 | } |
f88e1a42 | 4394 | |
db593f25 MT |
4395 | void bdrv_set_in_use(BlockDriverState *bs, int in_use) |
4396 | { | |
4397 | assert(bs->in_use != in_use); | |
4398 | bs->in_use = in_use; | |
4399 | } | |
4400 | ||
4401 | int bdrv_in_use(BlockDriverState *bs) | |
4402 | { | |
4403 | return bs->in_use; | |
4404 | } | |
4405 | ||
28a7282a LC |
4406 | void bdrv_iostatus_enable(BlockDriverState *bs) |
4407 | { | |
d6bf279e | 4408 | bs->iostatus_enabled = true; |
58e21ef5 | 4409 | bs->iostatus = BLOCK_DEVICE_IO_STATUS_OK; |
28a7282a LC |
4410 | } |
4411 | ||
4412 | /* The I/O status is only enabled if the drive explicitly | |
4413 | * enables it _and_ the VM is configured to stop on errors */ | |
4414 | bool bdrv_iostatus_is_enabled(const BlockDriverState *bs) | |
4415 | { | |
d6bf279e | 4416 | return (bs->iostatus_enabled && |
92aa5c6d PB |
4417 | (bs->on_write_error == BLOCKDEV_ON_ERROR_ENOSPC || |
4418 | bs->on_write_error == BLOCKDEV_ON_ERROR_STOP || | |
4419 | bs->on_read_error == BLOCKDEV_ON_ERROR_STOP)); | |
28a7282a LC |
4420 | } |
4421 | ||
4422 | void bdrv_iostatus_disable(BlockDriverState *bs) | |
4423 | { | |
d6bf279e | 4424 | bs->iostatus_enabled = false; |
28a7282a LC |
4425 | } |
4426 | ||
4427 | void bdrv_iostatus_reset(BlockDriverState *bs) | |
4428 | { | |
4429 | if (bdrv_iostatus_is_enabled(bs)) { | |
58e21ef5 | 4430 | bs->iostatus = BLOCK_DEVICE_IO_STATUS_OK; |
3bd293c3 PB |
4431 | if (bs->job) { |
4432 | block_job_iostatus_reset(bs->job); | |
4433 | } | |
28a7282a LC |
4434 | } |
4435 | } | |
4436 | ||
28a7282a LC |
4437 | void bdrv_iostatus_set_err(BlockDriverState *bs, int error) |
4438 | { | |
3e1caa5f PB |
4439 | assert(bdrv_iostatus_is_enabled(bs)); |
4440 | if (bs->iostatus == BLOCK_DEVICE_IO_STATUS_OK) { | |
58e21ef5 LC |
4441 | bs->iostatus = error == ENOSPC ? BLOCK_DEVICE_IO_STATUS_NOSPACE : |
4442 | BLOCK_DEVICE_IO_STATUS_FAILED; | |
28a7282a LC |
4443 | } |
4444 | } | |
4445 | ||
a597e79c CH |
4446 | void |
4447 | bdrv_acct_start(BlockDriverState *bs, BlockAcctCookie *cookie, int64_t bytes, | |
4448 | enum BlockAcctType type) | |
4449 | { | |
4450 | assert(type < BDRV_MAX_IOTYPE); | |
4451 | ||
4452 | cookie->bytes = bytes; | |
c488c7f6 | 4453 | cookie->start_time_ns = get_clock(); |
a597e79c CH |
4454 | cookie->type = type; |
4455 | } | |
4456 | ||
4457 | void | |
4458 | bdrv_acct_done(BlockDriverState *bs, BlockAcctCookie *cookie) | |
4459 | { | |
4460 | assert(cookie->type < BDRV_MAX_IOTYPE); | |
4461 | ||
4462 | bs->nr_bytes[cookie->type] += cookie->bytes; | |
4463 | bs->nr_ops[cookie->type]++; | |
c488c7f6 | 4464 | bs->total_time_ns[cookie->type] += get_clock() - cookie->start_time_ns; |
a597e79c CH |
4465 | } |
4466 | ||
d92ada22 LC |
4467 | void bdrv_img_create(const char *filename, const char *fmt, |
4468 | const char *base_filename, const char *base_fmt, | |
f382d43a MR |
4469 | char *options, uint64_t img_size, int flags, |
4470 | Error **errp, bool quiet) | |
f88e1a42 JS |
4471 | { |
4472 | QEMUOptionParameter *param = NULL, *create_options = NULL; | |
d220894e | 4473 | QEMUOptionParameter *backing_fmt, *backing_file, *size; |
f88e1a42 JS |
4474 | BlockDriverState *bs = NULL; |
4475 | BlockDriver *drv, *proto_drv; | |
96df67d1 | 4476 | BlockDriver *backing_drv = NULL; |
f88e1a42 JS |
4477 | int ret = 0; |
4478 | ||
4479 | /* Find driver and parse its options */ | |
4480 | drv = bdrv_find_format(fmt); | |
4481 | if (!drv) { | |
71c79813 | 4482 | error_setg(errp, "Unknown file format '%s'", fmt); |
d92ada22 | 4483 | return; |
f88e1a42 JS |
4484 | } |
4485 | ||
98289620 | 4486 | proto_drv = bdrv_find_protocol(filename, true); |
f88e1a42 | 4487 | if (!proto_drv) { |
71c79813 | 4488 | error_setg(errp, "Unknown protocol '%s'", filename); |
d92ada22 | 4489 | return; |
f88e1a42 JS |
4490 | } |
4491 | ||
4492 | create_options = append_option_parameters(create_options, | |
4493 | drv->create_options); | |
4494 | create_options = append_option_parameters(create_options, | |
4495 | proto_drv->create_options); | |
4496 | ||
4497 | /* Create parameter list with default values */ | |
4498 | param = parse_option_parameters("", create_options, param); | |
4499 | ||
4500 | set_option_parameter_int(param, BLOCK_OPT_SIZE, img_size); | |
4501 | ||
4502 | /* Parse -o options */ | |
4503 | if (options) { | |
4504 | param = parse_option_parameters(options, create_options, param); | |
4505 | if (param == NULL) { | |
71c79813 | 4506 | error_setg(errp, "Invalid options for file format '%s'.", fmt); |
f88e1a42 JS |
4507 | goto out; |
4508 | } | |
4509 | } | |
4510 | ||
4511 | if (base_filename) { | |
4512 | if (set_option_parameter(param, BLOCK_OPT_BACKING_FILE, | |
4513 | base_filename)) { | |
71c79813 LC |
4514 | error_setg(errp, "Backing file not supported for file format '%s'", |
4515 | fmt); | |
f88e1a42 JS |
4516 | goto out; |
4517 | } | |
4518 | } | |
4519 | ||
4520 | if (base_fmt) { | |
4521 | if (set_option_parameter(param, BLOCK_OPT_BACKING_FMT, base_fmt)) { | |
71c79813 LC |
4522 | error_setg(errp, "Backing file format not supported for file " |
4523 | "format '%s'", fmt); | |
f88e1a42 JS |
4524 | goto out; |
4525 | } | |
4526 | } | |
4527 | ||
792da93a JS |
4528 | backing_file = get_option_parameter(param, BLOCK_OPT_BACKING_FILE); |
4529 | if (backing_file && backing_file->value.s) { | |
4530 | if (!strcmp(filename, backing_file->value.s)) { | |
71c79813 LC |
4531 | error_setg(errp, "Error: Trying to create an image with the " |
4532 | "same filename as the backing file"); | |
792da93a JS |
4533 | goto out; |
4534 | } | |
4535 | } | |
4536 | ||
f88e1a42 JS |
4537 | backing_fmt = get_option_parameter(param, BLOCK_OPT_BACKING_FMT); |
4538 | if (backing_fmt && backing_fmt->value.s) { | |
96df67d1 SH |
4539 | backing_drv = bdrv_find_format(backing_fmt->value.s); |
4540 | if (!backing_drv) { | |
71c79813 LC |
4541 | error_setg(errp, "Unknown backing file format '%s'", |
4542 | backing_fmt->value.s); | |
f88e1a42 JS |
4543 | goto out; |
4544 | } | |
4545 | } | |
4546 | ||
4547 | // The size for the image must always be specified, with one exception: | |
4548 | // If we are using a backing file, we can obtain the size from there | |
d220894e KW |
4549 | size = get_option_parameter(param, BLOCK_OPT_SIZE); |
4550 | if (size && size->value.n == -1) { | |
f88e1a42 JS |
4551 | if (backing_file && backing_file->value.s) { |
4552 | uint64_t size; | |
f88e1a42 | 4553 | char buf[32]; |
63090dac PB |
4554 | int back_flags; |
4555 | ||
4556 | /* backing files always opened read-only */ | |
4557 | back_flags = | |
4558 | flags & ~(BDRV_O_RDWR | BDRV_O_SNAPSHOT | BDRV_O_NO_BACKING); | |
f88e1a42 | 4559 | |
f88e1a42 JS |
4560 | bs = bdrv_new(""); |
4561 | ||
de9c0cec KW |
4562 | ret = bdrv_open(bs, backing_file->value.s, NULL, back_flags, |
4563 | backing_drv); | |
f88e1a42 | 4564 | if (ret < 0) { |
71c79813 LC |
4565 | error_setg_errno(errp, -ret, "Could not open '%s'", |
4566 | backing_file->value.s); | |
f88e1a42 JS |
4567 | goto out; |
4568 | } | |
4569 | bdrv_get_geometry(bs, &size); | |
4570 | size *= 512; | |
4571 | ||
4572 | snprintf(buf, sizeof(buf), "%" PRId64, size); | |
4573 | set_option_parameter(param, BLOCK_OPT_SIZE, buf); | |
4574 | } else { | |
71c79813 | 4575 | error_setg(errp, "Image creation needs a size parameter"); |
f88e1a42 JS |
4576 | goto out; |
4577 | } | |
4578 | } | |
4579 | ||
f382d43a MR |
4580 | if (!quiet) { |
4581 | printf("Formatting '%s', fmt=%s ", filename, fmt); | |
4582 | print_option_parameters(param); | |
4583 | puts(""); | |
4584 | } | |
f88e1a42 | 4585 | ret = bdrv_create(drv, filename, param); |
f88e1a42 JS |
4586 | if (ret < 0) { |
4587 | if (ret == -ENOTSUP) { | |
71c79813 LC |
4588 | error_setg(errp,"Formatting or formatting option not supported for " |
4589 | "file format '%s'", fmt); | |
f88e1a42 | 4590 | } else if (ret == -EFBIG) { |
f3f4d2c0 KW |
4591 | const char *cluster_size_hint = ""; |
4592 | if (get_option_parameter(create_options, BLOCK_OPT_CLUSTER_SIZE)) { | |
4593 | cluster_size_hint = " (try using a larger cluster size)"; | |
4594 | } | |
4595 | error_setg(errp, "The image size is too large for file format '%s'%s", | |
4596 | fmt, cluster_size_hint); | |
f88e1a42 | 4597 | } else { |
71c79813 LC |
4598 | error_setg(errp, "%s: error while creating %s: %s", filename, fmt, |
4599 | strerror(-ret)); | |
f88e1a42 JS |
4600 | } |
4601 | } | |
4602 | ||
4603 | out: | |
4604 | free_option_parameters(create_options); | |
4605 | free_option_parameters(param); | |
4606 | ||
4607 | if (bs) { | |
4608 | bdrv_delete(bs); | |
4609 | } | |
f88e1a42 | 4610 | } |
85d126f3 SH |
4611 | |
4612 | AioContext *bdrv_get_aio_context(BlockDriverState *bs) | |
4613 | { | |
4614 | /* Currently BlockDriverState always uses the main loop AioContext */ | |
4615 | return qemu_get_aio_context(); | |
4616 | } | |
d616b224 SH |
4617 | |
4618 | void bdrv_add_before_write_notifier(BlockDriverState *bs, | |
4619 | NotifierWithReturn *notifier) | |
4620 | { | |
4621 | notifier_with_return_list_add(&bs->before_write_notifiers, notifier); | |
4622 | } |