]>
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" |
376253ec | 27 | #include "monitor.h" |
ea2384d3 | 28 | #include "block_int.h" |
2f0c9fe6 | 29 | #include "blockjob.h" |
5efa9d5a | 30 | #include "module.h" |
f795e743 | 31 | #include "qjson.h" |
3e1caa5f | 32 | #include "sysemu.h" |
d7d512f6 | 33 | #include "notify.h" |
68485420 | 34 | #include "qemu-coroutine.h" |
b2023818 | 35 | #include "qmp-commands.h" |
0563e191 | 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 | |
f9092b10 MA |
102 | /* The device to use for VM snapshots */ |
103 | static BlockDriverState *bs_snapshots; | |
104 | ||
eb852011 MA |
105 | /* If non-zero, use only whitelisted block drivers */ |
106 | static int use_bdrv_whitelist; | |
107 | ||
9e0b22f4 SH |
108 | #ifdef _WIN32 |
109 | static int is_windows_drive_prefix(const char *filename) | |
110 | { | |
111 | return (((filename[0] >= 'a' && filename[0] <= 'z') || | |
112 | (filename[0] >= 'A' && filename[0] <= 'Z')) && | |
113 | filename[1] == ':'); | |
114 | } | |
115 | ||
116 | int is_windows_drive(const char *filename) | |
117 | { | |
118 | if (is_windows_drive_prefix(filename) && | |
119 | filename[2] == '\0') | |
120 | return 1; | |
121 | if (strstart(filename, "\\\\.\\", NULL) || | |
122 | strstart(filename, "//./", NULL)) | |
123 | return 1; | |
124 | return 0; | |
125 | } | |
126 | #endif | |
127 | ||
0563e191 | 128 | /* throttling disk I/O limits */ |
98f90dba ZYW |
129 | void bdrv_io_limits_disable(BlockDriverState *bs) |
130 | { | |
131 | bs->io_limits_enabled = false; | |
132 | ||
133 | while (qemu_co_queue_next(&bs->throttled_reqs)); | |
134 | ||
135 | if (bs->block_timer) { | |
136 | qemu_del_timer(bs->block_timer); | |
137 | qemu_free_timer(bs->block_timer); | |
138 | bs->block_timer = NULL; | |
139 | } | |
140 | ||
141 | bs->slice_start = 0; | |
142 | bs->slice_end = 0; | |
143 | bs->slice_time = 0; | |
144 | memset(&bs->io_base, 0, sizeof(bs->io_base)); | |
145 | } | |
146 | ||
0563e191 ZYW |
147 | static void bdrv_block_timer(void *opaque) |
148 | { | |
149 | BlockDriverState *bs = opaque; | |
150 | ||
151 | qemu_co_queue_next(&bs->throttled_reqs); | |
152 | } | |
153 | ||
154 | void bdrv_io_limits_enable(BlockDriverState *bs) | |
155 | { | |
156 | qemu_co_queue_init(&bs->throttled_reqs); | |
157 | bs->block_timer = qemu_new_timer_ns(vm_clock, bdrv_block_timer, bs); | |
158 | bs->slice_time = 5 * BLOCK_IO_SLICE_TIME; | |
159 | bs->slice_start = qemu_get_clock_ns(vm_clock); | |
160 | bs->slice_end = bs->slice_start + bs->slice_time; | |
161 | memset(&bs->io_base, 0, sizeof(bs->io_base)); | |
162 | bs->io_limits_enabled = true; | |
163 | } | |
164 | ||
165 | bool bdrv_io_limits_enabled(BlockDriverState *bs) | |
166 | { | |
167 | BlockIOLimit *io_limits = &bs->io_limits; | |
168 | return io_limits->bps[BLOCK_IO_LIMIT_READ] | |
169 | || io_limits->bps[BLOCK_IO_LIMIT_WRITE] | |
170 | || io_limits->bps[BLOCK_IO_LIMIT_TOTAL] | |
171 | || io_limits->iops[BLOCK_IO_LIMIT_READ] | |
172 | || io_limits->iops[BLOCK_IO_LIMIT_WRITE] | |
173 | || io_limits->iops[BLOCK_IO_LIMIT_TOTAL]; | |
174 | } | |
175 | ||
98f90dba ZYW |
176 | static void bdrv_io_limits_intercept(BlockDriverState *bs, |
177 | bool is_write, int nb_sectors) | |
178 | { | |
179 | int64_t wait_time = -1; | |
180 | ||
181 | if (!qemu_co_queue_empty(&bs->throttled_reqs)) { | |
182 | qemu_co_queue_wait(&bs->throttled_reqs); | |
183 | } | |
184 | ||
185 | /* In fact, we hope to keep each request's timing, in FIFO mode. The next | |
186 | * throttled requests will not be dequeued until the current request is | |
187 | * allowed to be serviced. So if the current request still exceeds the | |
188 | * limits, it will be inserted to the head. All requests followed it will | |
189 | * be still in throttled_reqs queue. | |
190 | */ | |
191 | ||
192 | while (bdrv_exceed_io_limits(bs, nb_sectors, is_write, &wait_time)) { | |
193 | qemu_mod_timer(bs->block_timer, | |
194 | wait_time + qemu_get_clock_ns(vm_clock)); | |
195 | qemu_co_queue_wait_insert_head(&bs->throttled_reqs); | |
196 | } | |
197 | ||
198 | qemu_co_queue_next(&bs->throttled_reqs); | |
199 | } | |
200 | ||
9e0b22f4 SH |
201 | /* check if the path starts with "<protocol>:" */ |
202 | static int path_has_protocol(const char *path) | |
203 | { | |
947995c0 PB |
204 | const char *p; |
205 | ||
9e0b22f4 SH |
206 | #ifdef _WIN32 |
207 | if (is_windows_drive(path) || | |
208 | is_windows_drive_prefix(path)) { | |
209 | return 0; | |
210 | } | |
947995c0 PB |
211 | p = path + strcspn(path, ":/\\"); |
212 | #else | |
213 | p = path + strcspn(path, ":/"); | |
9e0b22f4 SH |
214 | #endif |
215 | ||
947995c0 | 216 | return *p == ':'; |
9e0b22f4 SH |
217 | } |
218 | ||
83f64091 | 219 | int path_is_absolute(const char *path) |
3b0d4f61 | 220 | { |
21664424 FB |
221 | #ifdef _WIN32 |
222 | /* specific case for names like: "\\.\d:" */ | |
f53f4da9 | 223 | if (is_windows_drive(path) || is_windows_drive_prefix(path)) { |
21664424 | 224 | return 1; |
f53f4da9 PB |
225 | } |
226 | return (*path == '/' || *path == '\\'); | |
3b9f94e1 | 227 | #else |
f53f4da9 | 228 | return (*path == '/'); |
3b9f94e1 | 229 | #endif |
3b0d4f61 FB |
230 | } |
231 | ||
83f64091 FB |
232 | /* if filename is absolute, just copy it to dest. Otherwise, build a |
233 | path to it by considering it is relative to base_path. URL are | |
234 | supported. */ | |
235 | void path_combine(char *dest, int dest_size, | |
236 | const char *base_path, | |
237 | const char *filename) | |
3b0d4f61 | 238 | { |
83f64091 FB |
239 | const char *p, *p1; |
240 | int len; | |
241 | ||
242 | if (dest_size <= 0) | |
243 | return; | |
244 | if (path_is_absolute(filename)) { | |
245 | pstrcpy(dest, dest_size, filename); | |
246 | } else { | |
247 | p = strchr(base_path, ':'); | |
248 | if (p) | |
249 | p++; | |
250 | else | |
251 | p = base_path; | |
3b9f94e1 FB |
252 | p1 = strrchr(base_path, '/'); |
253 | #ifdef _WIN32 | |
254 | { | |
255 | const char *p2; | |
256 | p2 = strrchr(base_path, '\\'); | |
257 | if (!p1 || p2 > p1) | |
258 | p1 = p2; | |
259 | } | |
260 | #endif | |
83f64091 FB |
261 | if (p1) |
262 | p1++; | |
263 | else | |
264 | p1 = base_path; | |
265 | if (p1 > p) | |
266 | p = p1; | |
267 | len = p - base_path; | |
268 | if (len > dest_size - 1) | |
269 | len = dest_size - 1; | |
270 | memcpy(dest, base_path, len); | |
271 | dest[len] = '\0'; | |
272 | pstrcat(dest, dest_size, filename); | |
3b0d4f61 | 273 | } |
3b0d4f61 FB |
274 | } |
275 | ||
dc5a1371 PB |
276 | void bdrv_get_full_backing_filename(BlockDriverState *bs, char *dest, size_t sz) |
277 | { | |
278 | if (bs->backing_file[0] == '\0' || path_has_protocol(bs->backing_file)) { | |
279 | pstrcpy(dest, sz, bs->backing_file); | |
280 | } else { | |
281 | path_combine(dest, sz, bs->filename, bs->backing_file); | |
282 | } | |
283 | } | |
284 | ||
5efa9d5a | 285 | void bdrv_register(BlockDriver *bdrv) |
ea2384d3 | 286 | { |
8c5873d6 SH |
287 | /* Block drivers without coroutine functions need emulation */ |
288 | if (!bdrv->bdrv_co_readv) { | |
f9f05dc5 KW |
289 | bdrv->bdrv_co_readv = bdrv_co_readv_em; |
290 | bdrv->bdrv_co_writev = bdrv_co_writev_em; | |
291 | ||
f8c35c1d SH |
292 | /* bdrv_co_readv_em()/brdv_co_writev_em() work in terms of aio, so if |
293 | * the block driver lacks aio we need to emulate that too. | |
294 | */ | |
f9f05dc5 KW |
295 | if (!bdrv->bdrv_aio_readv) { |
296 | /* add AIO emulation layer */ | |
297 | bdrv->bdrv_aio_readv = bdrv_aio_readv_em; | |
298 | bdrv->bdrv_aio_writev = bdrv_aio_writev_em; | |
f9f05dc5 | 299 | } |
83f64091 | 300 | } |
b2e12bc6 | 301 | |
8a22f02a | 302 | QLIST_INSERT_HEAD(&bdrv_drivers, bdrv, list); |
ea2384d3 | 303 | } |
b338082b FB |
304 | |
305 | /* create a new block device (by default it is empty) */ | |
306 | BlockDriverState *bdrv_new(const char *device_name) | |
307 | { | |
1b7bdbc1 | 308 | BlockDriverState *bs; |
b338082b | 309 | |
7267c094 | 310 | bs = g_malloc0(sizeof(BlockDriverState)); |
b338082b | 311 | pstrcpy(bs->device_name, sizeof(bs->device_name), device_name); |
ea2384d3 | 312 | if (device_name[0] != '\0') { |
1b7bdbc1 | 313 | QTAILQ_INSERT_TAIL(&bdrv_states, bs, list); |
ea2384d3 | 314 | } |
28a7282a | 315 | bdrv_iostatus_disable(bs); |
d7d512f6 PB |
316 | notifier_list_init(&bs->close_notifiers); |
317 | ||
b338082b FB |
318 | return bs; |
319 | } | |
320 | ||
d7d512f6 PB |
321 | void bdrv_add_close_notifier(BlockDriverState *bs, Notifier *notify) |
322 | { | |
323 | notifier_list_add(&bs->close_notifiers, notify); | |
324 | } | |
325 | ||
ea2384d3 FB |
326 | BlockDriver *bdrv_find_format(const char *format_name) |
327 | { | |
328 | BlockDriver *drv1; | |
8a22f02a SH |
329 | QLIST_FOREACH(drv1, &bdrv_drivers, list) { |
330 | if (!strcmp(drv1->format_name, format_name)) { | |
ea2384d3 | 331 | return drv1; |
8a22f02a | 332 | } |
ea2384d3 FB |
333 | } |
334 | return NULL; | |
335 | } | |
336 | ||
eb852011 MA |
337 | static int bdrv_is_whitelisted(BlockDriver *drv) |
338 | { | |
339 | static const char *whitelist[] = { | |
340 | CONFIG_BDRV_WHITELIST | |
341 | }; | |
342 | const char **p; | |
343 | ||
344 | if (!whitelist[0]) | |
345 | return 1; /* no whitelist, anything goes */ | |
346 | ||
347 | for (p = whitelist; *p; p++) { | |
348 | if (!strcmp(drv->format_name, *p)) { | |
349 | return 1; | |
350 | } | |
351 | } | |
352 | return 0; | |
353 | } | |
354 | ||
355 | BlockDriver *bdrv_find_whitelisted_format(const char *format_name) | |
356 | { | |
357 | BlockDriver *drv = bdrv_find_format(format_name); | |
358 | return drv && bdrv_is_whitelisted(drv) ? drv : NULL; | |
359 | } | |
360 | ||
5b7e1542 ZYW |
361 | typedef struct CreateCo { |
362 | BlockDriver *drv; | |
363 | char *filename; | |
364 | QEMUOptionParameter *options; | |
365 | int ret; | |
366 | } CreateCo; | |
367 | ||
368 | static void coroutine_fn bdrv_create_co_entry(void *opaque) | |
369 | { | |
370 | CreateCo *cco = opaque; | |
371 | assert(cco->drv); | |
372 | ||
373 | cco->ret = cco->drv->bdrv_create(cco->filename, cco->options); | |
374 | } | |
375 | ||
0e7e1989 KW |
376 | int bdrv_create(BlockDriver *drv, const char* filename, |
377 | QEMUOptionParameter *options) | |
ea2384d3 | 378 | { |
5b7e1542 ZYW |
379 | int ret; |
380 | ||
381 | Coroutine *co; | |
382 | CreateCo cco = { | |
383 | .drv = drv, | |
384 | .filename = g_strdup(filename), | |
385 | .options = options, | |
386 | .ret = NOT_DONE, | |
387 | }; | |
388 | ||
389 | if (!drv->bdrv_create) { | |
80168bff LC |
390 | ret = -ENOTSUP; |
391 | goto out; | |
5b7e1542 ZYW |
392 | } |
393 | ||
394 | if (qemu_in_coroutine()) { | |
395 | /* Fast-path if already in coroutine context */ | |
396 | bdrv_create_co_entry(&cco); | |
397 | } else { | |
398 | co = qemu_coroutine_create(bdrv_create_co_entry); | |
399 | qemu_coroutine_enter(co, &cco); | |
400 | while (cco.ret == NOT_DONE) { | |
401 | qemu_aio_wait(); | |
402 | } | |
403 | } | |
404 | ||
405 | ret = cco.ret; | |
0e7e1989 | 406 | |
80168bff LC |
407 | out: |
408 | g_free(cco.filename); | |
5b7e1542 | 409 | return ret; |
ea2384d3 FB |
410 | } |
411 | ||
84a12e66 CH |
412 | int bdrv_create_file(const char* filename, QEMUOptionParameter *options) |
413 | { | |
414 | BlockDriver *drv; | |
415 | ||
b50cbabc | 416 | drv = bdrv_find_protocol(filename); |
84a12e66 | 417 | if (drv == NULL) { |
16905d71 | 418 | return -ENOENT; |
84a12e66 CH |
419 | } |
420 | ||
421 | return bdrv_create(drv, filename, options); | |
422 | } | |
423 | ||
eba25057 JM |
424 | /* |
425 | * Create a uniquely-named empty temporary file. | |
426 | * Return 0 upon success, otherwise a negative errno value. | |
427 | */ | |
428 | int get_tmp_filename(char *filename, int size) | |
d5249393 | 429 | { |
eba25057 | 430 | #ifdef _WIN32 |
3b9f94e1 | 431 | char temp_dir[MAX_PATH]; |
eba25057 JM |
432 | /* GetTempFileName requires that its output buffer (4th param) |
433 | have length MAX_PATH or greater. */ | |
434 | assert(size >= MAX_PATH); | |
435 | return (GetTempPath(MAX_PATH, temp_dir) | |
436 | && GetTempFileName(temp_dir, "qem", 0, filename) | |
437 | ? 0 : -GetLastError()); | |
d5249393 | 438 | #else |
67b915a5 | 439 | int fd; |
7ccfb2eb | 440 | const char *tmpdir; |
0badc1ee AJ |
441 | tmpdir = getenv("TMPDIR"); |
442 | if (!tmpdir) | |
443 | tmpdir = "/tmp"; | |
eba25057 JM |
444 | if (snprintf(filename, size, "%s/vl.XXXXXX", tmpdir) >= size) { |
445 | return -EOVERFLOW; | |
446 | } | |
ea2384d3 | 447 | fd = mkstemp(filename); |
fe235a06 DH |
448 | if (fd < 0) { |
449 | return -errno; | |
450 | } | |
451 | if (close(fd) != 0) { | |
452 | unlink(filename); | |
eba25057 JM |
453 | return -errno; |
454 | } | |
455 | return 0; | |
d5249393 | 456 | #endif |
eba25057 | 457 | } |
fc01f7e7 | 458 | |
84a12e66 CH |
459 | /* |
460 | * Detect host devices. By convention, /dev/cdrom[N] is always | |
461 | * recognized as a host CDROM. | |
462 | */ | |
463 | static BlockDriver *find_hdev_driver(const char *filename) | |
464 | { | |
465 | int score_max = 0, score; | |
466 | BlockDriver *drv = NULL, *d; | |
467 | ||
468 | QLIST_FOREACH(d, &bdrv_drivers, list) { | |
469 | if (d->bdrv_probe_device) { | |
470 | score = d->bdrv_probe_device(filename); | |
471 | if (score > score_max) { | |
472 | score_max = score; | |
473 | drv = d; | |
474 | } | |
475 | } | |
476 | } | |
477 | ||
478 | return drv; | |
479 | } | |
480 | ||
b50cbabc | 481 | BlockDriver *bdrv_find_protocol(const char *filename) |
83f64091 FB |
482 | { |
483 | BlockDriver *drv1; | |
484 | char protocol[128]; | |
1cec71e3 | 485 | int len; |
83f64091 | 486 | const char *p; |
19cb3738 | 487 | |
66f82cee KW |
488 | /* TODO Drivers without bdrv_file_open must be specified explicitly */ |
489 | ||
39508e7a CH |
490 | /* |
491 | * XXX(hch): we really should not let host device detection | |
492 | * override an explicit protocol specification, but moving this | |
493 | * later breaks access to device names with colons in them. | |
494 | * Thanks to the brain-dead persistent naming schemes on udev- | |
495 | * based Linux systems those actually are quite common. | |
496 | */ | |
497 | drv1 = find_hdev_driver(filename); | |
498 | if (drv1) { | |
499 | return drv1; | |
500 | } | |
501 | ||
9e0b22f4 | 502 | if (!path_has_protocol(filename)) { |
39508e7a | 503 | return bdrv_find_format("file"); |
84a12e66 | 504 | } |
9e0b22f4 SH |
505 | p = strchr(filename, ':'); |
506 | assert(p != NULL); | |
1cec71e3 AL |
507 | len = p - filename; |
508 | if (len > sizeof(protocol) - 1) | |
509 | len = sizeof(protocol) - 1; | |
510 | memcpy(protocol, filename, len); | |
511 | protocol[len] = '\0'; | |
8a22f02a | 512 | QLIST_FOREACH(drv1, &bdrv_drivers, list) { |
5fafdf24 | 513 | if (drv1->protocol_name && |
8a22f02a | 514 | !strcmp(drv1->protocol_name, protocol)) { |
83f64091 | 515 | return drv1; |
8a22f02a | 516 | } |
83f64091 FB |
517 | } |
518 | return NULL; | |
519 | } | |
520 | ||
c98ac35d | 521 | static int find_image_format(const char *filename, BlockDriver **pdrv) |
f3a5d3f8 CH |
522 | { |
523 | int ret, score, score_max; | |
524 | BlockDriver *drv1, *drv; | |
525 | uint8_t buf[2048]; | |
526 | BlockDriverState *bs; | |
527 | ||
f5edb014 | 528 | ret = bdrv_file_open(&bs, filename, 0); |
c98ac35d SW |
529 | if (ret < 0) { |
530 | *pdrv = NULL; | |
531 | return ret; | |
532 | } | |
f8ea0b00 | 533 | |
08a00559 KW |
534 | /* Return the raw BlockDriver * to scsi-generic devices or empty drives */ |
535 | if (bs->sg || !bdrv_is_inserted(bs)) { | |
1a396859 | 536 | bdrv_delete(bs); |
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 FB |
545 | ret = bdrv_pread(bs, 0, buf, sizeof(buf)); |
546 | bdrv_delete(bs); | |
547 | if (ret < 0) { | |
c98ac35d SW |
548 | *pdrv = NULL; |
549 | return ret; | |
83f64091 FB |
550 | } |
551 | ||
ea2384d3 | 552 | score_max = 0; |
84a12e66 | 553 | drv = NULL; |
8a22f02a | 554 | QLIST_FOREACH(drv1, &bdrv_drivers, list) { |
83f64091 FB |
555 | if (drv1->bdrv_probe) { |
556 | score = drv1->bdrv_probe(buf, ret, filename); | |
557 | if (score > score_max) { | |
558 | score_max = score; | |
559 | drv = drv1; | |
560 | } | |
0849bf08 | 561 | } |
fc01f7e7 | 562 | } |
c98ac35d SW |
563 | if (!drv) { |
564 | ret = -ENOENT; | |
565 | } | |
566 | *pdrv = drv; | |
567 | return ret; | |
ea2384d3 FB |
568 | } |
569 | ||
51762288 SH |
570 | /** |
571 | * Set the current 'total_sectors' value | |
572 | */ | |
573 | static int refresh_total_sectors(BlockDriverState *bs, int64_t hint) | |
574 | { | |
575 | BlockDriver *drv = bs->drv; | |
576 | ||
396759ad NB |
577 | /* Do not attempt drv->bdrv_getlength() on scsi-generic devices */ |
578 | if (bs->sg) | |
579 | return 0; | |
580 | ||
51762288 SH |
581 | /* query actual device if possible, otherwise just trust the hint */ |
582 | if (drv->bdrv_getlength) { | |
583 | int64_t length = drv->bdrv_getlength(bs); | |
584 | if (length < 0) { | |
585 | return length; | |
586 | } | |
587 | hint = length >> BDRV_SECTOR_BITS; | |
588 | } | |
589 | ||
590 | bs->total_sectors = hint; | |
591 | return 0; | |
592 | } | |
593 | ||
c3993cdc SH |
594 | /** |
595 | * Set open flags for a given cache mode | |
596 | * | |
597 | * Return 0 on success, -1 if the cache mode was invalid. | |
598 | */ | |
599 | int bdrv_parse_cache_flags(const char *mode, int *flags) | |
600 | { | |
601 | *flags &= ~BDRV_O_CACHE_MASK; | |
602 | ||
603 | if (!strcmp(mode, "off") || !strcmp(mode, "none")) { | |
604 | *flags |= BDRV_O_NOCACHE | BDRV_O_CACHE_WB; | |
92196b2f SH |
605 | } else if (!strcmp(mode, "directsync")) { |
606 | *flags |= BDRV_O_NOCACHE; | |
c3993cdc SH |
607 | } else if (!strcmp(mode, "writeback")) { |
608 | *flags |= BDRV_O_CACHE_WB; | |
609 | } else if (!strcmp(mode, "unsafe")) { | |
610 | *flags |= BDRV_O_CACHE_WB; | |
611 | *flags |= BDRV_O_NO_FLUSH; | |
612 | } else if (!strcmp(mode, "writethrough")) { | |
613 | /* this is the default */ | |
614 | } else { | |
615 | return -1; | |
616 | } | |
617 | ||
618 | return 0; | |
619 | } | |
620 | ||
53fec9d3 SH |
621 | /** |
622 | * The copy-on-read flag is actually a reference count so multiple users may | |
623 | * use the feature without worrying about clobbering its previous state. | |
624 | * Copy-on-read stays enabled until all users have called to disable it. | |
625 | */ | |
626 | void bdrv_enable_copy_on_read(BlockDriverState *bs) | |
627 | { | |
628 | bs->copy_on_read++; | |
629 | } | |
630 | ||
631 | void bdrv_disable_copy_on_read(BlockDriverState *bs) | |
632 | { | |
633 | assert(bs->copy_on_read > 0); | |
634 | bs->copy_on_read--; | |
635 | } | |
636 | ||
57915332 KW |
637 | /* |
638 | * Common part for opening disk images and files | |
639 | */ | |
640 | static int bdrv_open_common(BlockDriverState *bs, const char *filename, | |
641 | int flags, BlockDriver *drv) | |
642 | { | |
643 | int ret, open_flags; | |
644 | ||
645 | assert(drv != NULL); | |
6405875c | 646 | assert(bs->file == NULL); |
57915332 | 647 | |
28dcee10 SH |
648 | trace_bdrv_open_common(bs, filename, flags, drv->format_name); |
649 | ||
57915332 | 650 | bs->open_flags = flags; |
57915332 KW |
651 | bs->buffer_alignment = 512; |
652 | ||
53fec9d3 SH |
653 | assert(bs->copy_on_read == 0); /* bdrv_new() and bdrv_close() make it so */ |
654 | if ((flags & BDRV_O_RDWR) && (flags & BDRV_O_COPY_ON_READ)) { | |
655 | bdrv_enable_copy_on_read(bs); | |
656 | } | |
657 | ||
57915332 KW |
658 | pstrcpy(bs->filename, sizeof(bs->filename), filename); |
659 | ||
660 | if (use_bdrv_whitelist && !bdrv_is_whitelisted(drv)) { | |
661 | return -ENOTSUP; | |
662 | } | |
663 | ||
664 | bs->drv = drv; | |
7267c094 | 665 | bs->opaque = g_malloc0(drv->instance_size); |
57915332 | 666 | |
03f541bd | 667 | bs->enable_write_cache = !!(flags & BDRV_O_CACHE_WB); |
e1e9b0ac | 668 | open_flags = flags | BDRV_O_CACHE_WB; |
57915332 KW |
669 | |
670 | /* | |
671 | * Clear flags that are internal to the block layer before opening the | |
672 | * image. | |
673 | */ | |
e1e9b0ac | 674 | open_flags &= ~(BDRV_O_SNAPSHOT | BDRV_O_NO_BACKING); |
57915332 KW |
675 | |
676 | /* | |
ebabb67a | 677 | * Snapshots should be writable. |
57915332 KW |
678 | */ |
679 | if (bs->is_temporary) { | |
680 | open_flags |= BDRV_O_RDWR; | |
681 | } | |
682 | ||
be028adc | 683 | bs->read_only = !(open_flags & BDRV_O_RDWR); |
e7c63796 | 684 | |
66f82cee KW |
685 | /* Open the image, either directly or using a protocol */ |
686 | if (drv->bdrv_file_open) { | |
687 | ret = drv->bdrv_file_open(bs, filename, open_flags); | |
688 | } else { | |
689 | ret = bdrv_file_open(&bs->file, filename, open_flags); | |
690 | if (ret >= 0) { | |
691 | ret = drv->bdrv_open(bs, open_flags); | |
692 | } | |
693 | } | |
694 | ||
57915332 KW |
695 | if (ret < 0) { |
696 | goto free_and_fail; | |
697 | } | |
698 | ||
51762288 SH |
699 | ret = refresh_total_sectors(bs, bs->total_sectors); |
700 | if (ret < 0) { | |
701 | goto free_and_fail; | |
57915332 | 702 | } |
51762288 | 703 | |
57915332 KW |
704 | #ifndef _WIN32 |
705 | if (bs->is_temporary) { | |
706 | unlink(filename); | |
707 | } | |
708 | #endif | |
709 | return 0; | |
710 | ||
711 | free_and_fail: | |
66f82cee KW |
712 | if (bs->file) { |
713 | bdrv_delete(bs->file); | |
714 | bs->file = NULL; | |
715 | } | |
7267c094 | 716 | g_free(bs->opaque); |
57915332 KW |
717 | bs->opaque = NULL; |
718 | bs->drv = NULL; | |
719 | return ret; | |
720 | } | |
721 | ||
b6ce07aa KW |
722 | /* |
723 | * Opens a file using a protocol (file, host_device, nbd, ...) | |
724 | */ | |
83f64091 | 725 | int bdrv_file_open(BlockDriverState **pbs, const char *filename, int flags) |
ea2384d3 | 726 | { |
83f64091 | 727 | BlockDriverState *bs; |
6db95603 | 728 | BlockDriver *drv; |
83f64091 FB |
729 | int ret; |
730 | ||
b50cbabc | 731 | drv = bdrv_find_protocol(filename); |
6db95603 CH |
732 | if (!drv) { |
733 | return -ENOENT; | |
734 | } | |
735 | ||
83f64091 | 736 | bs = bdrv_new(""); |
b6ce07aa | 737 | ret = bdrv_open_common(bs, filename, flags, drv); |
83f64091 FB |
738 | if (ret < 0) { |
739 | bdrv_delete(bs); | |
740 | return ret; | |
3b0d4f61 | 741 | } |
71d0770c | 742 | bs->growable = 1; |
83f64091 FB |
743 | *pbs = bs; |
744 | return 0; | |
745 | } | |
746 | ||
9156df12 PB |
747 | int bdrv_open_backing_file(BlockDriverState *bs) |
748 | { | |
749 | char backing_filename[PATH_MAX]; | |
750 | int back_flags, ret; | |
751 | BlockDriver *back_drv = NULL; | |
752 | ||
753 | if (bs->backing_hd != NULL) { | |
754 | return 0; | |
755 | } | |
756 | ||
757 | bs->open_flags &= ~BDRV_O_NO_BACKING; | |
758 | if (bs->backing_file[0] == '\0') { | |
759 | return 0; | |
760 | } | |
761 | ||
762 | bs->backing_hd = bdrv_new(""); | |
763 | bdrv_get_full_backing_filename(bs, backing_filename, | |
764 | sizeof(backing_filename)); | |
765 | ||
766 | if (bs->backing_format[0] != '\0') { | |
767 | back_drv = bdrv_find_format(bs->backing_format); | |
768 | } | |
769 | ||
770 | /* backing files always opened read-only */ | |
771 | back_flags = bs->open_flags & ~(BDRV_O_RDWR | BDRV_O_SNAPSHOT); | |
772 | ||
773 | ret = bdrv_open(bs->backing_hd, backing_filename, back_flags, back_drv); | |
774 | if (ret < 0) { | |
775 | bdrv_delete(bs->backing_hd); | |
776 | bs->backing_hd = NULL; | |
777 | bs->open_flags |= BDRV_O_NO_BACKING; | |
778 | return ret; | |
779 | } | |
780 | return 0; | |
781 | } | |
782 | ||
b6ce07aa KW |
783 | /* |
784 | * Opens a disk image (raw, qcow2, vmdk, ...) | |
785 | */ | |
d6e9098e KW |
786 | int bdrv_open(BlockDriverState *bs, const char *filename, int flags, |
787 | BlockDriver *drv) | |
ea2384d3 | 788 | { |
b6ce07aa | 789 | int ret; |
2b572816 | 790 | char tmp_filename[PATH_MAX]; |
712e7874 | 791 | |
83f64091 | 792 | if (flags & BDRV_O_SNAPSHOT) { |
ea2384d3 FB |
793 | BlockDriverState *bs1; |
794 | int64_t total_size; | |
7c96d46e | 795 | int is_protocol = 0; |
91a073a9 KW |
796 | BlockDriver *bdrv_qcow2; |
797 | QEMUOptionParameter *options; | |
b6ce07aa | 798 | char backing_filename[PATH_MAX]; |
3b46e624 | 799 | |
ea2384d3 FB |
800 | /* if snapshot, we create a temporary backing file and open it |
801 | instead of opening 'filename' directly */ | |
33e3963e | 802 | |
ea2384d3 FB |
803 | /* if there is a backing file, use it */ |
804 | bs1 = bdrv_new(""); | |
d6e9098e | 805 | ret = bdrv_open(bs1, filename, 0, drv); |
51d7c00c | 806 | if (ret < 0) { |
ea2384d3 | 807 | bdrv_delete(bs1); |
51d7c00c | 808 | return ret; |
ea2384d3 | 809 | } |
3e82990b | 810 | total_size = bdrv_getlength(bs1) & BDRV_SECTOR_MASK; |
7c96d46e AL |
811 | |
812 | if (bs1->drv && bs1->drv->protocol_name) | |
813 | is_protocol = 1; | |
814 | ||
ea2384d3 | 815 | bdrv_delete(bs1); |
3b46e624 | 816 | |
eba25057 JM |
817 | ret = get_tmp_filename(tmp_filename, sizeof(tmp_filename)); |
818 | if (ret < 0) { | |
819 | return ret; | |
820 | } | |
7c96d46e AL |
821 | |
822 | /* Real path is meaningless for protocols */ | |
823 | if (is_protocol) | |
824 | snprintf(backing_filename, sizeof(backing_filename), | |
825 | "%s", filename); | |
114cdfa9 KS |
826 | else if (!realpath(filename, backing_filename)) |
827 | return -errno; | |
7c96d46e | 828 | |
91a073a9 KW |
829 | bdrv_qcow2 = bdrv_find_format("qcow2"); |
830 | options = parse_option_parameters("", bdrv_qcow2->create_options, NULL); | |
831 | ||
3e82990b | 832 | set_option_parameter_int(options, BLOCK_OPT_SIZE, total_size); |
91a073a9 KW |
833 | set_option_parameter(options, BLOCK_OPT_BACKING_FILE, backing_filename); |
834 | if (drv) { | |
835 | set_option_parameter(options, BLOCK_OPT_BACKING_FMT, | |
836 | drv->format_name); | |
837 | } | |
838 | ||
839 | ret = bdrv_create(bdrv_qcow2, tmp_filename, options); | |
d748768c | 840 | free_option_parameters(options); |
51d7c00c AL |
841 | if (ret < 0) { |
842 | return ret; | |
ea2384d3 | 843 | } |
91a073a9 | 844 | |
ea2384d3 | 845 | filename = tmp_filename; |
91a073a9 | 846 | drv = bdrv_qcow2; |
ea2384d3 FB |
847 | bs->is_temporary = 1; |
848 | } | |
712e7874 | 849 | |
b6ce07aa | 850 | /* Find the right image format driver */ |
6db95603 | 851 | if (!drv) { |
c98ac35d | 852 | ret = find_image_format(filename, &drv); |
51d7c00c | 853 | } |
6987307c | 854 | |
51d7c00c | 855 | if (!drv) { |
51d7c00c | 856 | goto unlink_and_fail; |
ea2384d3 | 857 | } |
b6ce07aa | 858 | |
be028adc JC |
859 | if (flags & BDRV_O_RDWR) { |
860 | flags |= BDRV_O_ALLOW_RDWR; | |
861 | } | |
862 | ||
b6ce07aa KW |
863 | /* Open the image */ |
864 | ret = bdrv_open_common(bs, filename, flags, drv); | |
865 | if (ret < 0) { | |
6987307c CH |
866 | goto unlink_and_fail; |
867 | } | |
868 | ||
b6ce07aa | 869 | /* If there is a backing file, use it */ |
9156df12 PB |
870 | if ((flags & BDRV_O_NO_BACKING) == 0) { |
871 | ret = bdrv_open_backing_file(bs); | |
b6ce07aa KW |
872 | if (ret < 0) { |
873 | bdrv_close(bs); | |
874 | return ret; | |
875 | } | |
b6ce07aa KW |
876 | } |
877 | ||
878 | if (!bdrv_key_required(bs)) { | |
7d4b4ba5 | 879 | bdrv_dev_change_media_cb(bs, true); |
b6ce07aa KW |
880 | } |
881 | ||
98f90dba ZYW |
882 | /* throttling disk I/O limits */ |
883 | if (bs->io_limits_enabled) { | |
884 | bdrv_io_limits_enable(bs); | |
885 | } | |
886 | ||
b6ce07aa KW |
887 | return 0; |
888 | ||
889 | unlink_and_fail: | |
890 | if (bs->is_temporary) { | |
891 | unlink(filename); | |
892 | } | |
893 | return ret; | |
894 | } | |
895 | ||
e971aa12 JC |
896 | typedef struct BlockReopenQueueEntry { |
897 | bool prepared; | |
898 | BDRVReopenState state; | |
899 | QSIMPLEQ_ENTRY(BlockReopenQueueEntry) entry; | |
900 | } BlockReopenQueueEntry; | |
901 | ||
902 | /* | |
903 | * Adds a BlockDriverState to a simple queue for an atomic, transactional | |
904 | * reopen of multiple devices. | |
905 | * | |
906 | * bs_queue can either be an existing BlockReopenQueue that has had QSIMPLE_INIT | |
907 | * already performed, or alternatively may be NULL a new BlockReopenQueue will | |
908 | * be created and initialized. This newly created BlockReopenQueue should be | |
909 | * passed back in for subsequent calls that are intended to be of the same | |
910 | * atomic 'set'. | |
911 | * | |
912 | * bs is the BlockDriverState to add to the reopen queue. | |
913 | * | |
914 | * flags contains the open flags for the associated bs | |
915 | * | |
916 | * returns a pointer to bs_queue, which is either the newly allocated | |
917 | * bs_queue, or the existing bs_queue being used. | |
918 | * | |
919 | */ | |
920 | BlockReopenQueue *bdrv_reopen_queue(BlockReopenQueue *bs_queue, | |
921 | BlockDriverState *bs, int flags) | |
922 | { | |
923 | assert(bs != NULL); | |
924 | ||
925 | BlockReopenQueueEntry *bs_entry; | |
926 | if (bs_queue == NULL) { | |
927 | bs_queue = g_new0(BlockReopenQueue, 1); | |
928 | QSIMPLEQ_INIT(bs_queue); | |
929 | } | |
930 | ||
931 | if (bs->file) { | |
932 | bdrv_reopen_queue(bs_queue, bs->file, flags); | |
933 | } | |
934 | ||
935 | bs_entry = g_new0(BlockReopenQueueEntry, 1); | |
936 | QSIMPLEQ_INSERT_TAIL(bs_queue, bs_entry, entry); | |
937 | ||
938 | bs_entry->state.bs = bs; | |
939 | bs_entry->state.flags = flags; | |
940 | ||
941 | return bs_queue; | |
942 | } | |
943 | ||
944 | /* | |
945 | * Reopen multiple BlockDriverStates atomically & transactionally. | |
946 | * | |
947 | * The queue passed in (bs_queue) must have been built up previous | |
948 | * via bdrv_reopen_queue(). | |
949 | * | |
950 | * Reopens all BDS specified in the queue, with the appropriate | |
951 | * flags. All devices are prepared for reopen, and failure of any | |
952 | * device will cause all device changes to be abandonded, and intermediate | |
953 | * data cleaned up. | |
954 | * | |
955 | * If all devices prepare successfully, then the changes are committed | |
956 | * to all devices. | |
957 | * | |
958 | */ | |
959 | int bdrv_reopen_multiple(BlockReopenQueue *bs_queue, Error **errp) | |
960 | { | |
961 | int ret = -1; | |
962 | BlockReopenQueueEntry *bs_entry, *next; | |
963 | Error *local_err = NULL; | |
964 | ||
965 | assert(bs_queue != NULL); | |
966 | ||
967 | bdrv_drain_all(); | |
968 | ||
969 | QSIMPLEQ_FOREACH(bs_entry, bs_queue, entry) { | |
970 | if (bdrv_reopen_prepare(&bs_entry->state, bs_queue, &local_err)) { | |
971 | error_propagate(errp, local_err); | |
972 | goto cleanup; | |
973 | } | |
974 | bs_entry->prepared = true; | |
975 | } | |
976 | ||
977 | /* If we reach this point, we have success and just need to apply the | |
978 | * changes | |
979 | */ | |
980 | QSIMPLEQ_FOREACH(bs_entry, bs_queue, entry) { | |
981 | bdrv_reopen_commit(&bs_entry->state); | |
982 | } | |
983 | ||
984 | ret = 0; | |
985 | ||
986 | cleanup: | |
987 | QSIMPLEQ_FOREACH_SAFE(bs_entry, bs_queue, entry, next) { | |
988 | if (ret && bs_entry->prepared) { | |
989 | bdrv_reopen_abort(&bs_entry->state); | |
990 | } | |
991 | g_free(bs_entry); | |
992 | } | |
993 | g_free(bs_queue); | |
994 | return ret; | |
995 | } | |
996 | ||
997 | ||
998 | /* Reopen a single BlockDriverState with the specified flags. */ | |
999 | int bdrv_reopen(BlockDriverState *bs, int bdrv_flags, Error **errp) | |
1000 | { | |
1001 | int ret = -1; | |
1002 | Error *local_err = NULL; | |
1003 | BlockReopenQueue *queue = bdrv_reopen_queue(NULL, bs, bdrv_flags); | |
1004 | ||
1005 | ret = bdrv_reopen_multiple(queue, &local_err); | |
1006 | if (local_err != NULL) { | |
1007 | error_propagate(errp, local_err); | |
1008 | } | |
1009 | return ret; | |
1010 | } | |
1011 | ||
1012 | ||
1013 | /* | |
1014 | * Prepares a BlockDriverState for reopen. All changes are staged in the | |
1015 | * 'opaque' field of the BDRVReopenState, which is used and allocated by | |
1016 | * the block driver layer .bdrv_reopen_prepare() | |
1017 | * | |
1018 | * bs is the BlockDriverState to reopen | |
1019 | * flags are the new open flags | |
1020 | * queue is the reopen queue | |
1021 | * | |
1022 | * Returns 0 on success, non-zero on error. On error errp will be set | |
1023 | * as well. | |
1024 | * | |
1025 | * On failure, bdrv_reopen_abort() will be called to clean up any data. | |
1026 | * It is the responsibility of the caller to then call the abort() or | |
1027 | * commit() for any other BDS that have been left in a prepare() state | |
1028 | * | |
1029 | */ | |
1030 | int bdrv_reopen_prepare(BDRVReopenState *reopen_state, BlockReopenQueue *queue, | |
1031 | Error **errp) | |
1032 | { | |
1033 | int ret = -1; | |
1034 | Error *local_err = NULL; | |
1035 | BlockDriver *drv; | |
1036 | ||
1037 | assert(reopen_state != NULL); | |
1038 | assert(reopen_state->bs->drv != NULL); | |
1039 | drv = reopen_state->bs->drv; | |
1040 | ||
1041 | /* if we are to stay read-only, do not allow permission change | |
1042 | * to r/w */ | |
1043 | if (!(reopen_state->bs->open_flags & BDRV_O_ALLOW_RDWR) && | |
1044 | reopen_state->flags & BDRV_O_RDWR) { | |
1045 | error_set(errp, QERR_DEVICE_IS_READ_ONLY, | |
1046 | reopen_state->bs->device_name); | |
1047 | goto error; | |
1048 | } | |
1049 | ||
1050 | ||
1051 | ret = bdrv_flush(reopen_state->bs); | |
1052 | if (ret) { | |
1053 | error_set(errp, ERROR_CLASS_GENERIC_ERROR, "Error (%s) flushing drive", | |
1054 | strerror(-ret)); | |
1055 | goto error; | |
1056 | } | |
1057 | ||
1058 | if (drv->bdrv_reopen_prepare) { | |
1059 | ret = drv->bdrv_reopen_prepare(reopen_state, queue, &local_err); | |
1060 | if (ret) { | |
1061 | if (local_err != NULL) { | |
1062 | error_propagate(errp, local_err); | |
1063 | } else { | |
1064 | error_set(errp, QERR_OPEN_FILE_FAILED, | |
1065 | reopen_state->bs->filename); | |
1066 | } | |
1067 | goto error; | |
1068 | } | |
1069 | } else { | |
1070 | /* It is currently mandatory to have a bdrv_reopen_prepare() | |
1071 | * handler for each supported drv. */ | |
1072 | error_set(errp, QERR_BLOCK_FORMAT_FEATURE_NOT_SUPPORTED, | |
1073 | drv->format_name, reopen_state->bs->device_name, | |
1074 | "reopening of file"); | |
1075 | ret = -1; | |
1076 | goto error; | |
1077 | } | |
1078 | ||
1079 | ret = 0; | |
1080 | ||
1081 | error: | |
1082 | return ret; | |
1083 | } | |
1084 | ||
1085 | /* | |
1086 | * Takes the staged changes for the reopen from bdrv_reopen_prepare(), and | |
1087 | * makes them final by swapping the staging BlockDriverState contents into | |
1088 | * the active BlockDriverState contents. | |
1089 | */ | |
1090 | void bdrv_reopen_commit(BDRVReopenState *reopen_state) | |
1091 | { | |
1092 | BlockDriver *drv; | |
1093 | ||
1094 | assert(reopen_state != NULL); | |
1095 | drv = reopen_state->bs->drv; | |
1096 | assert(drv != NULL); | |
1097 | ||
1098 | /* If there are any driver level actions to take */ | |
1099 | if (drv->bdrv_reopen_commit) { | |
1100 | drv->bdrv_reopen_commit(reopen_state); | |
1101 | } | |
1102 | ||
1103 | /* set BDS specific flags now */ | |
1104 | reopen_state->bs->open_flags = reopen_state->flags; | |
1105 | reopen_state->bs->enable_write_cache = !!(reopen_state->flags & | |
1106 | BDRV_O_CACHE_WB); | |
1107 | reopen_state->bs->read_only = !(reopen_state->flags & BDRV_O_RDWR); | |
1108 | } | |
1109 | ||
1110 | /* | |
1111 | * Abort the reopen, and delete and free the staged changes in | |
1112 | * reopen_state | |
1113 | */ | |
1114 | void bdrv_reopen_abort(BDRVReopenState *reopen_state) | |
1115 | { | |
1116 | BlockDriver *drv; | |
1117 | ||
1118 | assert(reopen_state != NULL); | |
1119 | drv = reopen_state->bs->drv; | |
1120 | assert(drv != NULL); | |
1121 | ||
1122 | if (drv->bdrv_reopen_abort) { | |
1123 | drv->bdrv_reopen_abort(reopen_state); | |
1124 | } | |
1125 | } | |
1126 | ||
1127 | ||
fc01f7e7 FB |
1128 | void bdrv_close(BlockDriverState *bs) |
1129 | { | |
80ccf93b | 1130 | bdrv_flush(bs); |
3cbc002c PB |
1131 | if (bs->job) { |
1132 | block_job_cancel_sync(bs->job); | |
1133 | } | |
1134 | bdrv_drain_all(); | |
d7d512f6 | 1135 | notifier_list_notify(&bs->close_notifiers, bs); |
7094f12f | 1136 | |
3cbc002c | 1137 | if (bs->drv) { |
f9092b10 MA |
1138 | if (bs == bs_snapshots) { |
1139 | bs_snapshots = NULL; | |
1140 | } | |
557df6ac | 1141 | if (bs->backing_hd) { |
ea2384d3 | 1142 | bdrv_delete(bs->backing_hd); |
557df6ac SH |
1143 | bs->backing_hd = NULL; |
1144 | } | |
ea2384d3 | 1145 | bs->drv->bdrv_close(bs); |
7267c094 | 1146 | g_free(bs->opaque); |
ea2384d3 FB |
1147 | #ifdef _WIN32 |
1148 | if (bs->is_temporary) { | |
1149 | unlink(bs->filename); | |
1150 | } | |
67b915a5 | 1151 | #endif |
ea2384d3 FB |
1152 | bs->opaque = NULL; |
1153 | bs->drv = NULL; | |
53fec9d3 | 1154 | bs->copy_on_read = 0; |
a275fa42 PB |
1155 | bs->backing_file[0] = '\0'; |
1156 | bs->backing_format[0] = '\0'; | |
6405875c PB |
1157 | bs->total_sectors = 0; |
1158 | bs->encrypted = 0; | |
1159 | bs->valid_key = 0; | |
1160 | bs->sg = 0; | |
1161 | bs->growable = 0; | |
b338082b | 1162 | |
66f82cee | 1163 | if (bs->file != NULL) { |
0ac9377d PB |
1164 | bdrv_delete(bs->file); |
1165 | bs->file = NULL; | |
66f82cee | 1166 | } |
b338082b | 1167 | } |
98f90dba | 1168 | |
9ca11154 PH |
1169 | bdrv_dev_change_media_cb(bs, false); |
1170 | ||
98f90dba ZYW |
1171 | /*throttling disk I/O limits*/ |
1172 | if (bs->io_limits_enabled) { | |
1173 | bdrv_io_limits_disable(bs); | |
1174 | } | |
b338082b FB |
1175 | } |
1176 | ||
2bc93fed MK |
1177 | void bdrv_close_all(void) |
1178 | { | |
1179 | BlockDriverState *bs; | |
1180 | ||
1181 | QTAILQ_FOREACH(bs, &bdrv_states, list) { | |
1182 | bdrv_close(bs); | |
1183 | } | |
1184 | } | |
1185 | ||
922453bc SH |
1186 | /* |
1187 | * Wait for pending requests to complete across all BlockDriverStates | |
1188 | * | |
1189 | * This function does not flush data to disk, use bdrv_flush_all() for that | |
1190 | * after calling this function. | |
4c355d53 ZYW |
1191 | * |
1192 | * Note that completion of an asynchronous I/O operation can trigger any | |
1193 | * number of other I/O operations on other devices---for example a coroutine | |
1194 | * can be arbitrarily complex and a constant flow of I/O can come until the | |
1195 | * coroutine is complete. Because of this, it is not possible to have a | |
1196 | * function to drain a single device's I/O queue. | |
922453bc SH |
1197 | */ |
1198 | void bdrv_drain_all(void) | |
1199 | { | |
1200 | BlockDriverState *bs; | |
4c355d53 ZYW |
1201 | bool busy; |
1202 | ||
1203 | do { | |
1204 | busy = qemu_aio_wait(); | |
922453bc | 1205 | |
4c355d53 ZYW |
1206 | /* FIXME: We do not have timer support here, so this is effectively |
1207 | * a busy wait. | |
1208 | */ | |
1209 | QTAILQ_FOREACH(bs, &bdrv_states, list) { | |
1210 | if (!qemu_co_queue_empty(&bs->throttled_reqs)) { | |
1211 | qemu_co_queue_restart_all(&bs->throttled_reqs); | |
1212 | busy = true; | |
1213 | } | |
1214 | } | |
1215 | } while (busy); | |
922453bc SH |
1216 | |
1217 | /* If requests are still pending there is a bug somewhere */ | |
1218 | QTAILQ_FOREACH(bs, &bdrv_states, list) { | |
1219 | assert(QLIST_EMPTY(&bs->tracked_requests)); | |
1220 | assert(qemu_co_queue_empty(&bs->throttled_reqs)); | |
1221 | } | |
1222 | } | |
1223 | ||
d22b2f41 RH |
1224 | /* make a BlockDriverState anonymous by removing from bdrv_state list. |
1225 | Also, NULL terminate the device_name to prevent double remove */ | |
1226 | void bdrv_make_anon(BlockDriverState *bs) | |
1227 | { | |
1228 | if (bs->device_name[0] != '\0') { | |
1229 | QTAILQ_REMOVE(&bdrv_states, bs, list); | |
1230 | } | |
1231 | bs->device_name[0] = '\0'; | |
1232 | } | |
1233 | ||
e023b2e2 PB |
1234 | static void bdrv_rebind(BlockDriverState *bs) |
1235 | { | |
1236 | if (bs->drv && bs->drv->bdrv_rebind) { | |
1237 | bs->drv->bdrv_rebind(bs); | |
1238 | } | |
1239 | } | |
1240 | ||
4ddc07ca PB |
1241 | static void bdrv_move_feature_fields(BlockDriverState *bs_dest, |
1242 | BlockDriverState *bs_src) | |
8802d1fd | 1243 | { |
4ddc07ca PB |
1244 | /* move some fields that need to stay attached to the device */ |
1245 | bs_dest->open_flags = bs_src->open_flags; | |
8802d1fd JC |
1246 | |
1247 | /* dev info */ | |
4ddc07ca PB |
1248 | bs_dest->dev_ops = bs_src->dev_ops; |
1249 | bs_dest->dev_opaque = bs_src->dev_opaque; | |
1250 | bs_dest->dev = bs_src->dev; | |
1251 | bs_dest->buffer_alignment = bs_src->buffer_alignment; | |
1252 | bs_dest->copy_on_read = bs_src->copy_on_read; | |
8802d1fd | 1253 | |
4ddc07ca | 1254 | bs_dest->enable_write_cache = bs_src->enable_write_cache; |
c4a248a1 | 1255 | |
8802d1fd | 1256 | /* i/o timing parameters */ |
4ddc07ca PB |
1257 | bs_dest->slice_time = bs_src->slice_time; |
1258 | bs_dest->slice_start = bs_src->slice_start; | |
1259 | bs_dest->slice_end = bs_src->slice_end; | |
1260 | bs_dest->io_limits = bs_src->io_limits; | |
1261 | bs_dest->io_base = bs_src->io_base; | |
1262 | bs_dest->throttled_reqs = bs_src->throttled_reqs; | |
1263 | bs_dest->block_timer = bs_src->block_timer; | |
1264 | bs_dest->io_limits_enabled = bs_src->io_limits_enabled; | |
8802d1fd | 1265 | |
8802d1fd | 1266 | /* r/w error */ |
4ddc07ca PB |
1267 | bs_dest->on_read_error = bs_src->on_read_error; |
1268 | bs_dest->on_write_error = bs_src->on_write_error; | |
8802d1fd JC |
1269 | |
1270 | /* i/o status */ | |
4ddc07ca PB |
1271 | bs_dest->iostatus_enabled = bs_src->iostatus_enabled; |
1272 | bs_dest->iostatus = bs_src->iostatus; | |
8802d1fd | 1273 | |
a9fc4408 | 1274 | /* dirty bitmap */ |
4ddc07ca PB |
1275 | bs_dest->dirty_count = bs_src->dirty_count; |
1276 | bs_dest->dirty_bitmap = bs_src->dirty_bitmap; | |
a9fc4408 PB |
1277 | |
1278 | /* job */ | |
4ddc07ca PB |
1279 | bs_dest->in_use = bs_src->in_use; |
1280 | bs_dest->job = bs_src->job; | |
a9fc4408 | 1281 | |
8802d1fd | 1282 | /* keep the same entry in bdrv_states */ |
4ddc07ca PB |
1283 | pstrcpy(bs_dest->device_name, sizeof(bs_dest->device_name), |
1284 | bs_src->device_name); | |
1285 | bs_dest->list = bs_src->list; | |
1286 | } | |
8802d1fd | 1287 | |
4ddc07ca PB |
1288 | /* |
1289 | * Swap bs contents for two image chains while they are live, | |
1290 | * while keeping required fields on the BlockDriverState that is | |
1291 | * actually attached to a device. | |
1292 | * | |
1293 | * This will modify the BlockDriverState fields, and swap contents | |
1294 | * between bs_new and bs_old. Both bs_new and bs_old are modified. | |
1295 | * | |
1296 | * bs_new is required to be anonymous. | |
1297 | * | |
1298 | * This function does not create any image files. | |
1299 | */ | |
1300 | void bdrv_swap(BlockDriverState *bs_new, BlockDriverState *bs_old) | |
1301 | { | |
1302 | BlockDriverState tmp; | |
f6801b83 | 1303 | |
4ddc07ca PB |
1304 | /* bs_new must be anonymous and shouldn't have anything fancy enabled */ |
1305 | assert(bs_new->device_name[0] == '\0'); | |
1306 | assert(bs_new->dirty_bitmap == NULL); | |
1307 | assert(bs_new->job == NULL); | |
1308 | assert(bs_new->dev == NULL); | |
1309 | assert(bs_new->in_use == 0); | |
1310 | assert(bs_new->io_limits_enabled == false); | |
1311 | assert(bs_new->block_timer == NULL); | |
8802d1fd | 1312 | |
4ddc07ca PB |
1313 | tmp = *bs_new; |
1314 | *bs_new = *bs_old; | |
1315 | *bs_old = tmp; | |
a9fc4408 | 1316 | |
4ddc07ca PB |
1317 | /* there are some fields that should not be swapped, move them back */ |
1318 | bdrv_move_feature_fields(&tmp, bs_old); | |
1319 | bdrv_move_feature_fields(bs_old, bs_new); | |
1320 | bdrv_move_feature_fields(bs_new, &tmp); | |
8802d1fd | 1321 | |
4ddc07ca PB |
1322 | /* bs_new shouldn't be in bdrv_states even after the swap! */ |
1323 | assert(bs_new->device_name[0] == '\0'); | |
1324 | ||
1325 | /* Check a few fields that should remain attached to the device */ | |
1326 | assert(bs_new->dev == NULL); | |
1327 | assert(bs_new->job == NULL); | |
1328 | assert(bs_new->in_use == 0); | |
1329 | assert(bs_new->io_limits_enabled == false); | |
1330 | assert(bs_new->block_timer == NULL); | |
e023b2e2 PB |
1331 | |
1332 | bdrv_rebind(bs_new); | |
4ddc07ca PB |
1333 | bdrv_rebind(bs_old); |
1334 | } | |
1335 | ||
1336 | /* | |
1337 | * Add new bs contents at the top of an image chain while the chain is | |
1338 | * live, while keeping required fields on the top layer. | |
1339 | * | |
1340 | * This will modify the BlockDriverState fields, and swap contents | |
1341 | * between bs_new and bs_top. Both bs_new and bs_top are modified. | |
1342 | * | |
1343 | * bs_new is required to be anonymous. | |
1344 | * | |
1345 | * This function does not create any image files. | |
1346 | */ | |
1347 | void bdrv_append(BlockDriverState *bs_new, BlockDriverState *bs_top) | |
1348 | { | |
1349 | bdrv_swap(bs_new, bs_top); | |
1350 | ||
1351 | /* The contents of 'tmp' will become bs_top, as we are | |
1352 | * swapping bs_new and bs_top contents. */ | |
1353 | bs_top->backing_hd = bs_new; | |
1354 | bs_top->open_flags &= ~BDRV_O_NO_BACKING; | |
1355 | pstrcpy(bs_top->backing_file, sizeof(bs_top->backing_file), | |
1356 | bs_new->filename); | |
1357 | pstrcpy(bs_top->backing_format, sizeof(bs_top->backing_format), | |
1358 | bs_new->drv ? bs_new->drv->format_name : ""); | |
8802d1fd JC |
1359 | } |
1360 | ||
b338082b FB |
1361 | void bdrv_delete(BlockDriverState *bs) |
1362 | { | |
fa879d62 | 1363 | assert(!bs->dev); |
3e914655 PB |
1364 | assert(!bs->job); |
1365 | assert(!bs->in_use); | |
18846dee | 1366 | |
1b7bdbc1 | 1367 | /* remove from list, if necessary */ |
d22b2f41 | 1368 | bdrv_make_anon(bs); |
34c6f050 | 1369 | |
b338082b | 1370 | bdrv_close(bs); |
66f82cee | 1371 | |
f9092b10 | 1372 | assert(bs != bs_snapshots); |
7267c094 | 1373 | g_free(bs); |
fc01f7e7 FB |
1374 | } |
1375 | ||
fa879d62 MA |
1376 | int bdrv_attach_dev(BlockDriverState *bs, void *dev) |
1377 | /* TODO change to DeviceState *dev when all users are qdevified */ | |
18846dee | 1378 | { |
fa879d62 | 1379 | if (bs->dev) { |
18846dee MA |
1380 | return -EBUSY; |
1381 | } | |
fa879d62 | 1382 | bs->dev = dev; |
28a7282a | 1383 | bdrv_iostatus_reset(bs); |
18846dee MA |
1384 | return 0; |
1385 | } | |
1386 | ||
fa879d62 MA |
1387 | /* TODO qdevified devices don't use this, remove when devices are qdevified */ |
1388 | void bdrv_attach_dev_nofail(BlockDriverState *bs, void *dev) | |
18846dee | 1389 | { |
fa879d62 MA |
1390 | if (bdrv_attach_dev(bs, dev) < 0) { |
1391 | abort(); | |
1392 | } | |
1393 | } | |
1394 | ||
1395 | void bdrv_detach_dev(BlockDriverState *bs, void *dev) | |
1396 | /* TODO change to DeviceState *dev when all users are qdevified */ | |
1397 | { | |
1398 | assert(bs->dev == dev); | |
1399 | bs->dev = NULL; | |
0e49de52 MA |
1400 | bs->dev_ops = NULL; |
1401 | bs->dev_opaque = NULL; | |
29e05f20 | 1402 | bs->buffer_alignment = 512; |
18846dee MA |
1403 | } |
1404 | ||
fa879d62 MA |
1405 | /* TODO change to return DeviceState * when all users are qdevified */ |
1406 | void *bdrv_get_attached_dev(BlockDriverState *bs) | |
18846dee | 1407 | { |
fa879d62 | 1408 | return bs->dev; |
18846dee MA |
1409 | } |
1410 | ||
0e49de52 MA |
1411 | void bdrv_set_dev_ops(BlockDriverState *bs, const BlockDevOps *ops, |
1412 | void *opaque) | |
1413 | { | |
1414 | bs->dev_ops = ops; | |
1415 | bs->dev_opaque = opaque; | |
2c6942fa MA |
1416 | if (bdrv_dev_has_removable_media(bs) && bs == bs_snapshots) { |
1417 | bs_snapshots = NULL; | |
1418 | } | |
0e49de52 MA |
1419 | } |
1420 | ||
32c81a4a PB |
1421 | void bdrv_emit_qmp_error_event(const BlockDriverState *bdrv, |
1422 | enum MonitorEvent ev, | |
1423 | BlockErrorAction action, bool is_read) | |
329c0a48 LC |
1424 | { |
1425 | QObject *data; | |
1426 | const char *action_str; | |
1427 | ||
1428 | switch (action) { | |
1429 | case BDRV_ACTION_REPORT: | |
1430 | action_str = "report"; | |
1431 | break; | |
1432 | case BDRV_ACTION_IGNORE: | |
1433 | action_str = "ignore"; | |
1434 | break; | |
1435 | case BDRV_ACTION_STOP: | |
1436 | action_str = "stop"; | |
1437 | break; | |
1438 | default: | |
1439 | abort(); | |
1440 | } | |
1441 | ||
1442 | data = qobject_from_jsonf("{ 'device': %s, 'action': %s, 'operation': %s }", | |
1443 | bdrv->device_name, | |
1444 | action_str, | |
1445 | is_read ? "read" : "write"); | |
32c81a4a | 1446 | monitor_protocol_event(ev, data); |
329c0a48 LC |
1447 | |
1448 | qobject_decref(data); | |
1449 | } | |
1450 | ||
6f382ed2 LC |
1451 | static void bdrv_emit_qmp_eject_event(BlockDriverState *bs, bool ejected) |
1452 | { | |
1453 | QObject *data; | |
1454 | ||
1455 | data = qobject_from_jsonf("{ 'device': %s, 'tray-open': %i }", | |
1456 | bdrv_get_device_name(bs), ejected); | |
1457 | monitor_protocol_event(QEVENT_DEVICE_TRAY_MOVED, data); | |
1458 | ||
1459 | qobject_decref(data); | |
1460 | } | |
1461 | ||
7d4b4ba5 | 1462 | static void bdrv_dev_change_media_cb(BlockDriverState *bs, bool load) |
0e49de52 | 1463 | { |
145feb17 | 1464 | if (bs->dev_ops && bs->dev_ops->change_media_cb) { |
6f382ed2 | 1465 | bool tray_was_closed = !bdrv_dev_is_tray_open(bs); |
7d4b4ba5 | 1466 | bs->dev_ops->change_media_cb(bs->dev_opaque, load); |
6f382ed2 LC |
1467 | if (tray_was_closed) { |
1468 | /* tray open */ | |
1469 | bdrv_emit_qmp_eject_event(bs, true); | |
1470 | } | |
1471 | if (load) { | |
1472 | /* tray close */ | |
1473 | bdrv_emit_qmp_eject_event(bs, false); | |
1474 | } | |
145feb17 MA |
1475 | } |
1476 | } | |
1477 | ||
2c6942fa MA |
1478 | bool bdrv_dev_has_removable_media(BlockDriverState *bs) |
1479 | { | |
1480 | return !bs->dev || (bs->dev_ops && bs->dev_ops->change_media_cb); | |
1481 | } | |
1482 | ||
025ccaa7 PB |
1483 | void bdrv_dev_eject_request(BlockDriverState *bs, bool force) |
1484 | { | |
1485 | if (bs->dev_ops && bs->dev_ops->eject_request_cb) { | |
1486 | bs->dev_ops->eject_request_cb(bs->dev_opaque, force); | |
1487 | } | |
1488 | } | |
1489 | ||
e4def80b MA |
1490 | bool bdrv_dev_is_tray_open(BlockDriverState *bs) |
1491 | { | |
1492 | if (bs->dev_ops && bs->dev_ops->is_tray_open) { | |
1493 | return bs->dev_ops->is_tray_open(bs->dev_opaque); | |
1494 | } | |
1495 | return false; | |
1496 | } | |
1497 | ||
145feb17 MA |
1498 | static void bdrv_dev_resize_cb(BlockDriverState *bs) |
1499 | { | |
1500 | if (bs->dev_ops && bs->dev_ops->resize_cb) { | |
1501 | bs->dev_ops->resize_cb(bs->dev_opaque); | |
0e49de52 MA |
1502 | } |
1503 | } | |
1504 | ||
f107639a MA |
1505 | bool bdrv_dev_is_medium_locked(BlockDriverState *bs) |
1506 | { | |
1507 | if (bs->dev_ops && bs->dev_ops->is_medium_locked) { | |
1508 | return bs->dev_ops->is_medium_locked(bs->dev_opaque); | |
1509 | } | |
1510 | return false; | |
1511 | } | |
1512 | ||
e97fc193 AL |
1513 | /* |
1514 | * Run consistency checks on an image | |
1515 | * | |
e076f338 | 1516 | * Returns 0 if the check could be completed (it doesn't mean that the image is |
a1c7273b | 1517 | * free of errors) or -errno when an internal error occurred. The results of the |
e076f338 | 1518 | * check are stored in res. |
e97fc193 | 1519 | */ |
4534ff54 | 1520 | int bdrv_check(BlockDriverState *bs, BdrvCheckResult *res, BdrvCheckMode fix) |
e97fc193 AL |
1521 | { |
1522 | if (bs->drv->bdrv_check == NULL) { | |
1523 | return -ENOTSUP; | |
1524 | } | |
1525 | ||
e076f338 | 1526 | memset(res, 0, sizeof(*res)); |
4534ff54 | 1527 | return bs->drv->bdrv_check(bs, res, fix); |
e97fc193 AL |
1528 | } |
1529 | ||
8a426614 KW |
1530 | #define COMMIT_BUF_SECTORS 2048 |
1531 | ||
33e3963e FB |
1532 | /* commit COW file into the raw image */ |
1533 | int bdrv_commit(BlockDriverState *bs) | |
1534 | { | |
19cb3738 | 1535 | BlockDriver *drv = bs->drv; |
8a426614 KW |
1536 | int64_t sector, total_sectors; |
1537 | int n, ro, open_flags; | |
0bce597d | 1538 | int ret = 0; |
8a426614 | 1539 | uint8_t *buf; |
c2cba3d9 | 1540 | char filename[PATH_MAX]; |
33e3963e | 1541 | |
19cb3738 FB |
1542 | if (!drv) |
1543 | return -ENOMEDIUM; | |
4dca4b63 NS |
1544 | |
1545 | if (!bs->backing_hd) { | |
1546 | return -ENOTSUP; | |
33e3963e FB |
1547 | } |
1548 | ||
2d3735d3 SH |
1549 | if (bdrv_in_use(bs) || bdrv_in_use(bs->backing_hd)) { |
1550 | return -EBUSY; | |
1551 | } | |
1552 | ||
4dca4b63 | 1553 | ro = bs->backing_hd->read_only; |
c2cba3d9 JM |
1554 | /* Use pstrcpy (not strncpy): filename must be NUL-terminated. */ |
1555 | pstrcpy(filename, sizeof(filename), bs->backing_hd->filename); | |
4dca4b63 NS |
1556 | open_flags = bs->backing_hd->open_flags; |
1557 | ||
1558 | if (ro) { | |
0bce597d JC |
1559 | if (bdrv_reopen(bs->backing_hd, open_flags | BDRV_O_RDWR, NULL)) { |
1560 | return -EACCES; | |
4dca4b63 | 1561 | } |
ea2384d3 | 1562 | } |
33e3963e | 1563 | |
6ea44308 | 1564 | total_sectors = bdrv_getlength(bs) >> BDRV_SECTOR_BITS; |
7267c094 | 1565 | buf = g_malloc(COMMIT_BUF_SECTORS * BDRV_SECTOR_SIZE); |
8a426614 KW |
1566 | |
1567 | for (sector = 0; sector < total_sectors; sector += n) { | |
05c4af54 | 1568 | if (bdrv_is_allocated(bs, sector, COMMIT_BUF_SECTORS, &n)) { |
8a426614 KW |
1569 | |
1570 | if (bdrv_read(bs, sector, buf, n) != 0) { | |
1571 | ret = -EIO; | |
1572 | goto ro_cleanup; | |
1573 | } | |
1574 | ||
1575 | if (bdrv_write(bs->backing_hd, sector, buf, n) != 0) { | |
1576 | ret = -EIO; | |
1577 | goto ro_cleanup; | |
1578 | } | |
ea2384d3 | 1579 | } |
33e3963e | 1580 | } |
95389c86 | 1581 | |
1d44952f CH |
1582 | if (drv->bdrv_make_empty) { |
1583 | ret = drv->bdrv_make_empty(bs); | |
1584 | bdrv_flush(bs); | |
1585 | } | |
95389c86 | 1586 | |
3f5075ae CH |
1587 | /* |
1588 | * Make sure all data we wrote to the backing device is actually | |
1589 | * stable on disk. | |
1590 | */ | |
1591 | if (bs->backing_hd) | |
1592 | bdrv_flush(bs->backing_hd); | |
4dca4b63 NS |
1593 | |
1594 | ro_cleanup: | |
7267c094 | 1595 | g_free(buf); |
4dca4b63 NS |
1596 | |
1597 | if (ro) { | |
0bce597d JC |
1598 | /* ignoring error return here */ |
1599 | bdrv_reopen(bs->backing_hd, open_flags & ~BDRV_O_RDWR, NULL); | |
4dca4b63 NS |
1600 | } |
1601 | ||
1d44952f | 1602 | return ret; |
33e3963e FB |
1603 | } |
1604 | ||
e8877497 | 1605 | int bdrv_commit_all(void) |
6ab4b5ab MA |
1606 | { |
1607 | BlockDriverState *bs; | |
1608 | ||
1609 | QTAILQ_FOREACH(bs, &bdrv_states, list) { | |
e8877497 SH |
1610 | int ret = bdrv_commit(bs); |
1611 | if (ret < 0) { | |
1612 | return ret; | |
1613 | } | |
6ab4b5ab | 1614 | } |
e8877497 | 1615 | return 0; |
6ab4b5ab MA |
1616 | } |
1617 | ||
dbffbdcf SH |
1618 | struct BdrvTrackedRequest { |
1619 | BlockDriverState *bs; | |
1620 | int64_t sector_num; | |
1621 | int nb_sectors; | |
1622 | bool is_write; | |
1623 | QLIST_ENTRY(BdrvTrackedRequest) list; | |
5f8b6491 | 1624 | Coroutine *co; /* owner, used for deadlock detection */ |
f4658285 | 1625 | CoQueue wait_queue; /* coroutines blocked on this request */ |
dbffbdcf SH |
1626 | }; |
1627 | ||
1628 | /** | |
1629 | * Remove an active request from the tracked requests list | |
1630 | * | |
1631 | * This function should be called when a tracked request is completing. | |
1632 | */ | |
1633 | static void tracked_request_end(BdrvTrackedRequest *req) | |
1634 | { | |
1635 | QLIST_REMOVE(req, list); | |
f4658285 | 1636 | qemu_co_queue_restart_all(&req->wait_queue); |
dbffbdcf SH |
1637 | } |
1638 | ||
1639 | /** | |
1640 | * Add an active request to the tracked requests list | |
1641 | */ | |
1642 | static void tracked_request_begin(BdrvTrackedRequest *req, | |
1643 | BlockDriverState *bs, | |
1644 | int64_t sector_num, | |
1645 | int nb_sectors, bool is_write) | |
1646 | { | |
1647 | *req = (BdrvTrackedRequest){ | |
1648 | .bs = bs, | |
1649 | .sector_num = sector_num, | |
1650 | .nb_sectors = nb_sectors, | |
1651 | .is_write = is_write, | |
5f8b6491 | 1652 | .co = qemu_coroutine_self(), |
dbffbdcf SH |
1653 | }; |
1654 | ||
f4658285 SH |
1655 | qemu_co_queue_init(&req->wait_queue); |
1656 | ||
dbffbdcf SH |
1657 | QLIST_INSERT_HEAD(&bs->tracked_requests, req, list); |
1658 | } | |
1659 | ||
d83947ac SH |
1660 | /** |
1661 | * Round a region to cluster boundaries | |
1662 | */ | |
1663 | static void round_to_clusters(BlockDriverState *bs, | |
1664 | int64_t sector_num, int nb_sectors, | |
1665 | int64_t *cluster_sector_num, | |
1666 | int *cluster_nb_sectors) | |
1667 | { | |
1668 | BlockDriverInfo bdi; | |
1669 | ||
1670 | if (bdrv_get_info(bs, &bdi) < 0 || bdi.cluster_size == 0) { | |
1671 | *cluster_sector_num = sector_num; | |
1672 | *cluster_nb_sectors = nb_sectors; | |
1673 | } else { | |
1674 | int64_t c = bdi.cluster_size / BDRV_SECTOR_SIZE; | |
1675 | *cluster_sector_num = QEMU_ALIGN_DOWN(sector_num, c); | |
1676 | *cluster_nb_sectors = QEMU_ALIGN_UP(sector_num - *cluster_sector_num + | |
1677 | nb_sectors, c); | |
1678 | } | |
1679 | } | |
1680 | ||
f4658285 SH |
1681 | static bool tracked_request_overlaps(BdrvTrackedRequest *req, |
1682 | int64_t sector_num, int nb_sectors) { | |
d83947ac SH |
1683 | /* aaaa bbbb */ |
1684 | if (sector_num >= req->sector_num + req->nb_sectors) { | |
1685 | return false; | |
1686 | } | |
1687 | /* bbbb aaaa */ | |
1688 | if (req->sector_num >= sector_num + nb_sectors) { | |
1689 | return false; | |
1690 | } | |
1691 | return true; | |
f4658285 SH |
1692 | } |
1693 | ||
1694 | static void coroutine_fn wait_for_overlapping_requests(BlockDriverState *bs, | |
1695 | int64_t sector_num, int nb_sectors) | |
1696 | { | |
1697 | BdrvTrackedRequest *req; | |
d83947ac SH |
1698 | int64_t cluster_sector_num; |
1699 | int cluster_nb_sectors; | |
f4658285 SH |
1700 | bool retry; |
1701 | ||
d83947ac SH |
1702 | /* If we touch the same cluster it counts as an overlap. This guarantees |
1703 | * that allocating writes will be serialized and not race with each other | |
1704 | * for the same cluster. For example, in copy-on-read it ensures that the | |
1705 | * CoR read and write operations are atomic and guest writes cannot | |
1706 | * interleave between them. | |
1707 | */ | |
1708 | round_to_clusters(bs, sector_num, nb_sectors, | |
1709 | &cluster_sector_num, &cluster_nb_sectors); | |
1710 | ||
f4658285 SH |
1711 | do { |
1712 | retry = false; | |
1713 | QLIST_FOREACH(req, &bs->tracked_requests, list) { | |
d83947ac SH |
1714 | if (tracked_request_overlaps(req, cluster_sector_num, |
1715 | cluster_nb_sectors)) { | |
5f8b6491 SH |
1716 | /* Hitting this means there was a reentrant request, for |
1717 | * example, a block driver issuing nested requests. This must | |
1718 | * never happen since it means deadlock. | |
1719 | */ | |
1720 | assert(qemu_coroutine_self() != req->co); | |
1721 | ||
f4658285 SH |
1722 | qemu_co_queue_wait(&req->wait_queue); |
1723 | retry = true; | |
1724 | break; | |
1725 | } | |
1726 | } | |
1727 | } while (retry); | |
1728 | } | |
1729 | ||
756e6736 KW |
1730 | /* |
1731 | * Return values: | |
1732 | * 0 - success | |
1733 | * -EINVAL - backing format specified, but no file | |
1734 | * -ENOSPC - can't update the backing file because no space is left in the | |
1735 | * image file header | |
1736 | * -ENOTSUP - format driver doesn't support changing the backing file | |
1737 | */ | |
1738 | int bdrv_change_backing_file(BlockDriverState *bs, | |
1739 | const char *backing_file, const char *backing_fmt) | |
1740 | { | |
1741 | BlockDriver *drv = bs->drv; | |
469ef350 | 1742 | int ret; |
756e6736 | 1743 | |
5f377794 PB |
1744 | /* Backing file format doesn't make sense without a backing file */ |
1745 | if (backing_fmt && !backing_file) { | |
1746 | return -EINVAL; | |
1747 | } | |
1748 | ||
756e6736 | 1749 | if (drv->bdrv_change_backing_file != NULL) { |
469ef350 | 1750 | ret = drv->bdrv_change_backing_file(bs, backing_file, backing_fmt); |
756e6736 | 1751 | } else { |
469ef350 | 1752 | ret = -ENOTSUP; |
756e6736 | 1753 | } |
469ef350 PB |
1754 | |
1755 | if (ret == 0) { | |
1756 | pstrcpy(bs->backing_file, sizeof(bs->backing_file), backing_file ?: ""); | |
1757 | pstrcpy(bs->backing_format, sizeof(bs->backing_format), backing_fmt ?: ""); | |
1758 | } | |
1759 | return ret; | |
756e6736 KW |
1760 | } |
1761 | ||
6ebdcee2 JC |
1762 | /* |
1763 | * Finds the image layer in the chain that has 'bs' as its backing file. | |
1764 | * | |
1765 | * active is the current topmost image. | |
1766 | * | |
1767 | * Returns NULL if bs is not found in active's image chain, | |
1768 | * or if active == bs. | |
1769 | */ | |
1770 | BlockDriverState *bdrv_find_overlay(BlockDriverState *active, | |
1771 | BlockDriverState *bs) | |
1772 | { | |
1773 | BlockDriverState *overlay = NULL; | |
1774 | BlockDriverState *intermediate; | |
1775 | ||
1776 | assert(active != NULL); | |
1777 | assert(bs != NULL); | |
1778 | ||
1779 | /* if bs is the same as active, then by definition it has no overlay | |
1780 | */ | |
1781 | if (active == bs) { | |
1782 | return NULL; | |
1783 | } | |
1784 | ||
1785 | intermediate = active; | |
1786 | while (intermediate->backing_hd) { | |
1787 | if (intermediate->backing_hd == bs) { | |
1788 | overlay = intermediate; | |
1789 | break; | |
1790 | } | |
1791 | intermediate = intermediate->backing_hd; | |
1792 | } | |
1793 | ||
1794 | return overlay; | |
1795 | } | |
1796 | ||
1797 | typedef struct BlkIntermediateStates { | |
1798 | BlockDriverState *bs; | |
1799 | QSIMPLEQ_ENTRY(BlkIntermediateStates) entry; | |
1800 | } BlkIntermediateStates; | |
1801 | ||
1802 | ||
1803 | /* | |
1804 | * Drops images above 'base' up to and including 'top', and sets the image | |
1805 | * above 'top' to have base as its backing file. | |
1806 | * | |
1807 | * Requires that the overlay to 'top' is opened r/w, so that the backing file | |
1808 | * information in 'bs' can be properly updated. | |
1809 | * | |
1810 | * E.g., this will convert the following chain: | |
1811 | * bottom <- base <- intermediate <- top <- active | |
1812 | * | |
1813 | * to | |
1814 | * | |
1815 | * bottom <- base <- active | |
1816 | * | |
1817 | * It is allowed for bottom==base, in which case it converts: | |
1818 | * | |
1819 | * base <- intermediate <- top <- active | |
1820 | * | |
1821 | * to | |
1822 | * | |
1823 | * base <- active | |
1824 | * | |
1825 | * Error conditions: | |
1826 | * if active == top, that is considered an error | |
1827 | * | |
1828 | */ | |
1829 | int bdrv_drop_intermediate(BlockDriverState *active, BlockDriverState *top, | |
1830 | BlockDriverState *base) | |
1831 | { | |
1832 | BlockDriverState *intermediate; | |
1833 | BlockDriverState *base_bs = NULL; | |
1834 | BlockDriverState *new_top_bs = NULL; | |
1835 | BlkIntermediateStates *intermediate_state, *next; | |
1836 | int ret = -EIO; | |
1837 | ||
1838 | QSIMPLEQ_HEAD(states_to_delete, BlkIntermediateStates) states_to_delete; | |
1839 | QSIMPLEQ_INIT(&states_to_delete); | |
1840 | ||
1841 | if (!top->drv || !base->drv) { | |
1842 | goto exit; | |
1843 | } | |
1844 | ||
1845 | new_top_bs = bdrv_find_overlay(active, top); | |
1846 | ||
1847 | if (new_top_bs == NULL) { | |
1848 | /* we could not find the image above 'top', this is an error */ | |
1849 | goto exit; | |
1850 | } | |
1851 | ||
1852 | /* special case of new_top_bs->backing_hd already pointing to base - nothing | |
1853 | * to do, no intermediate images */ | |
1854 | if (new_top_bs->backing_hd == base) { | |
1855 | ret = 0; | |
1856 | goto exit; | |
1857 | } | |
1858 | ||
1859 | intermediate = top; | |
1860 | ||
1861 | /* now we will go down through the list, and add each BDS we find | |
1862 | * into our deletion queue, until we hit the 'base' | |
1863 | */ | |
1864 | while (intermediate) { | |
1865 | intermediate_state = g_malloc0(sizeof(BlkIntermediateStates)); | |
1866 | intermediate_state->bs = intermediate; | |
1867 | QSIMPLEQ_INSERT_TAIL(&states_to_delete, intermediate_state, entry); | |
1868 | ||
1869 | if (intermediate->backing_hd == base) { | |
1870 | base_bs = intermediate->backing_hd; | |
1871 | break; | |
1872 | } | |
1873 | intermediate = intermediate->backing_hd; | |
1874 | } | |
1875 | if (base_bs == NULL) { | |
1876 | /* something went wrong, we did not end at the base. safely | |
1877 | * unravel everything, and exit with error */ | |
1878 | goto exit; | |
1879 | } | |
1880 | ||
1881 | /* success - we can delete the intermediate states, and link top->base */ | |
1882 | ret = bdrv_change_backing_file(new_top_bs, base_bs->filename, | |
1883 | base_bs->drv ? base_bs->drv->format_name : ""); | |
1884 | if (ret) { | |
1885 | goto exit; | |
1886 | } | |
1887 | new_top_bs->backing_hd = base_bs; | |
1888 | ||
1889 | ||
1890 | QSIMPLEQ_FOREACH_SAFE(intermediate_state, &states_to_delete, entry, next) { | |
1891 | /* so that bdrv_close() does not recursively close the chain */ | |
1892 | intermediate_state->bs->backing_hd = NULL; | |
1893 | bdrv_delete(intermediate_state->bs); | |
1894 | } | |
1895 | ret = 0; | |
1896 | ||
1897 | exit: | |
1898 | QSIMPLEQ_FOREACH_SAFE(intermediate_state, &states_to_delete, entry, next) { | |
1899 | g_free(intermediate_state); | |
1900 | } | |
1901 | return ret; | |
1902 | } | |
1903 | ||
1904 | ||
71d0770c AL |
1905 | static int bdrv_check_byte_request(BlockDriverState *bs, int64_t offset, |
1906 | size_t size) | |
1907 | { | |
1908 | int64_t len; | |
1909 | ||
1910 | if (!bdrv_is_inserted(bs)) | |
1911 | return -ENOMEDIUM; | |
1912 | ||
1913 | if (bs->growable) | |
1914 | return 0; | |
1915 | ||
1916 | len = bdrv_getlength(bs); | |
1917 | ||
fbb7b4e0 KW |
1918 | if (offset < 0) |
1919 | return -EIO; | |
1920 | ||
1921 | if ((offset > len) || (len - offset < size)) | |
71d0770c AL |
1922 | return -EIO; |
1923 | ||
1924 | return 0; | |
1925 | } | |
1926 | ||
1927 | static int bdrv_check_request(BlockDriverState *bs, int64_t sector_num, | |
1928 | int nb_sectors) | |
1929 | { | |
eb5a3165 JS |
1930 | return bdrv_check_byte_request(bs, sector_num * BDRV_SECTOR_SIZE, |
1931 | nb_sectors * BDRV_SECTOR_SIZE); | |
71d0770c AL |
1932 | } |
1933 | ||
1c9805a3 SH |
1934 | typedef struct RwCo { |
1935 | BlockDriverState *bs; | |
1936 | int64_t sector_num; | |
1937 | int nb_sectors; | |
1938 | QEMUIOVector *qiov; | |
1939 | bool is_write; | |
1940 | int ret; | |
1941 | } RwCo; | |
1942 | ||
1943 | static void coroutine_fn bdrv_rw_co_entry(void *opaque) | |
fc01f7e7 | 1944 | { |
1c9805a3 | 1945 | RwCo *rwco = opaque; |
ea2384d3 | 1946 | |
1c9805a3 SH |
1947 | if (!rwco->is_write) { |
1948 | rwco->ret = bdrv_co_do_readv(rwco->bs, rwco->sector_num, | |
470c0504 | 1949 | rwco->nb_sectors, rwco->qiov, 0); |
1c9805a3 SH |
1950 | } else { |
1951 | rwco->ret = bdrv_co_do_writev(rwco->bs, rwco->sector_num, | |
f08f2dda | 1952 | rwco->nb_sectors, rwco->qiov, 0); |
1c9805a3 SH |
1953 | } |
1954 | } | |
e7a8a783 | 1955 | |
1c9805a3 SH |
1956 | /* |
1957 | * Process a synchronous request using coroutines | |
1958 | */ | |
1959 | static int bdrv_rw_co(BlockDriverState *bs, int64_t sector_num, uint8_t *buf, | |
1960 | int nb_sectors, bool is_write) | |
1961 | { | |
1962 | QEMUIOVector qiov; | |
1963 | struct iovec iov = { | |
1964 | .iov_base = (void *)buf, | |
1965 | .iov_len = nb_sectors * BDRV_SECTOR_SIZE, | |
1966 | }; | |
1967 | Coroutine *co; | |
1968 | RwCo rwco = { | |
1969 | .bs = bs, | |
1970 | .sector_num = sector_num, | |
1971 | .nb_sectors = nb_sectors, | |
1972 | .qiov = &qiov, | |
1973 | .is_write = is_write, | |
1974 | .ret = NOT_DONE, | |
1975 | }; | |
e7a8a783 | 1976 | |
1c9805a3 | 1977 | qemu_iovec_init_external(&qiov, &iov, 1); |
e7a8a783 | 1978 | |
498e386c ZYW |
1979 | /** |
1980 | * In sync call context, when the vcpu is blocked, this throttling timer | |
1981 | * will not fire; so the I/O throttling function has to be disabled here | |
1982 | * if it has been enabled. | |
1983 | */ | |
1984 | if (bs->io_limits_enabled) { | |
1985 | fprintf(stderr, "Disabling I/O throttling on '%s' due " | |
1986 | "to synchronous I/O.\n", bdrv_get_device_name(bs)); | |
1987 | bdrv_io_limits_disable(bs); | |
1988 | } | |
1989 | ||
1c9805a3 SH |
1990 | if (qemu_in_coroutine()) { |
1991 | /* Fast-path if already in coroutine context */ | |
1992 | bdrv_rw_co_entry(&rwco); | |
1993 | } else { | |
1994 | co = qemu_coroutine_create(bdrv_rw_co_entry); | |
1995 | qemu_coroutine_enter(co, &rwco); | |
1996 | while (rwco.ret == NOT_DONE) { | |
1997 | qemu_aio_wait(); | |
1998 | } | |
1999 | } | |
2000 | return rwco.ret; | |
2001 | } | |
b338082b | 2002 | |
1c9805a3 SH |
2003 | /* return < 0 if error. See bdrv_write() for the return codes */ |
2004 | int bdrv_read(BlockDriverState *bs, int64_t sector_num, | |
2005 | uint8_t *buf, int nb_sectors) | |
2006 | { | |
2007 | return bdrv_rw_co(bs, sector_num, buf, nb_sectors, false); | |
fc01f7e7 FB |
2008 | } |
2009 | ||
07d27a44 MA |
2010 | /* Just like bdrv_read(), but with I/O throttling temporarily disabled */ |
2011 | int bdrv_read_unthrottled(BlockDriverState *bs, int64_t sector_num, | |
2012 | uint8_t *buf, int nb_sectors) | |
2013 | { | |
2014 | bool enabled; | |
2015 | int ret; | |
2016 | ||
2017 | enabled = bs->io_limits_enabled; | |
2018 | bs->io_limits_enabled = false; | |
2019 | ret = bdrv_read(bs, 0, buf, 1); | |
2020 | bs->io_limits_enabled = enabled; | |
2021 | return ret; | |
2022 | } | |
2023 | ||
71df14fc PB |
2024 | #define BITS_PER_LONG (sizeof(unsigned long) * 8) |
2025 | ||
7cd1e32a | 2026 | static void set_dirty_bitmap(BlockDriverState *bs, int64_t sector_num, |
a55eb92c | 2027 | int nb_sectors, int dirty) |
7cd1e32a LS |
2028 | { |
2029 | int64_t start, end; | |
c6d22830 | 2030 | unsigned long val, idx, bit; |
a55eb92c | 2031 | |
6ea44308 | 2032 | start = sector_num / BDRV_SECTORS_PER_DIRTY_CHUNK; |
c6d22830 | 2033 | end = (sector_num + nb_sectors - 1) / BDRV_SECTORS_PER_DIRTY_CHUNK; |
a55eb92c JK |
2034 | |
2035 | for (; start <= end; start++) { | |
71df14fc PB |
2036 | idx = start / BITS_PER_LONG; |
2037 | bit = start % BITS_PER_LONG; | |
c6d22830 JK |
2038 | val = bs->dirty_bitmap[idx]; |
2039 | if (dirty) { | |
6d59fec1 | 2040 | if (!(val & (1UL << bit))) { |
aaa0eb75 | 2041 | bs->dirty_count++; |
6d59fec1 | 2042 | val |= 1UL << bit; |
aaa0eb75 | 2043 | } |
c6d22830 | 2044 | } else { |
6d59fec1 | 2045 | if (val & (1UL << bit)) { |
aaa0eb75 | 2046 | bs->dirty_count--; |
6d59fec1 | 2047 | val &= ~(1UL << bit); |
aaa0eb75 | 2048 | } |
c6d22830 JK |
2049 | } |
2050 | bs->dirty_bitmap[idx] = val; | |
7cd1e32a LS |
2051 | } |
2052 | } | |
2053 | ||
5fafdf24 | 2054 | /* Return < 0 if error. Important errors are: |
19cb3738 FB |
2055 | -EIO generic I/O error (may happen for all errors) |
2056 | -ENOMEDIUM No media inserted. | |
2057 | -EINVAL Invalid sector number or nb_sectors | |
2058 | -EACCES Trying to write a read-only device | |
2059 | */ | |
5fafdf24 | 2060 | int bdrv_write(BlockDriverState *bs, int64_t sector_num, |
fc01f7e7 FB |
2061 | const uint8_t *buf, int nb_sectors) |
2062 | { | |
1c9805a3 | 2063 | return bdrv_rw_co(bs, sector_num, (uint8_t *)buf, nb_sectors, true); |
83f64091 FB |
2064 | } |
2065 | ||
eda578e5 AL |
2066 | int bdrv_pread(BlockDriverState *bs, int64_t offset, |
2067 | void *buf, int count1) | |
83f64091 | 2068 | { |
6ea44308 | 2069 | uint8_t tmp_buf[BDRV_SECTOR_SIZE]; |
83f64091 FB |
2070 | int len, nb_sectors, count; |
2071 | int64_t sector_num; | |
9a8c4cce | 2072 | int ret; |
83f64091 FB |
2073 | |
2074 | count = count1; | |
2075 | /* first read to align to sector start */ | |
6ea44308 | 2076 | len = (BDRV_SECTOR_SIZE - offset) & (BDRV_SECTOR_SIZE - 1); |
83f64091 FB |
2077 | if (len > count) |
2078 | len = count; | |
6ea44308 | 2079 | sector_num = offset >> BDRV_SECTOR_BITS; |
83f64091 | 2080 | if (len > 0) { |
9a8c4cce KW |
2081 | if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0) |
2082 | return ret; | |
6ea44308 | 2083 | memcpy(buf, tmp_buf + (offset & (BDRV_SECTOR_SIZE - 1)), len); |
83f64091 FB |
2084 | count -= len; |
2085 | if (count == 0) | |
2086 | return count1; | |
2087 | sector_num++; | |
2088 | buf += len; | |
2089 | } | |
2090 | ||
2091 | /* read the sectors "in place" */ | |
6ea44308 | 2092 | nb_sectors = count >> BDRV_SECTOR_BITS; |
83f64091 | 2093 | if (nb_sectors > 0) { |
9a8c4cce KW |
2094 | if ((ret = bdrv_read(bs, sector_num, buf, nb_sectors)) < 0) |
2095 | return ret; | |
83f64091 | 2096 | sector_num += nb_sectors; |
6ea44308 | 2097 | len = nb_sectors << BDRV_SECTOR_BITS; |
83f64091 FB |
2098 | buf += len; |
2099 | count -= len; | |
2100 | } | |
2101 | ||
2102 | /* add data from the last sector */ | |
2103 | if (count > 0) { | |
9a8c4cce KW |
2104 | if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0) |
2105 | return ret; | |
83f64091 FB |
2106 | memcpy(buf, tmp_buf, count); |
2107 | } | |
2108 | return count1; | |
2109 | } | |
2110 | ||
eda578e5 AL |
2111 | int bdrv_pwrite(BlockDriverState *bs, int64_t offset, |
2112 | const void *buf, int count1) | |
83f64091 | 2113 | { |
6ea44308 | 2114 | uint8_t tmp_buf[BDRV_SECTOR_SIZE]; |
83f64091 FB |
2115 | int len, nb_sectors, count; |
2116 | int64_t sector_num; | |
9a8c4cce | 2117 | int ret; |
83f64091 FB |
2118 | |
2119 | count = count1; | |
2120 | /* first write to align to sector start */ | |
6ea44308 | 2121 | len = (BDRV_SECTOR_SIZE - offset) & (BDRV_SECTOR_SIZE - 1); |
83f64091 FB |
2122 | if (len > count) |
2123 | len = count; | |
6ea44308 | 2124 | sector_num = offset >> BDRV_SECTOR_BITS; |
83f64091 | 2125 | if (len > 0) { |
9a8c4cce KW |
2126 | if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0) |
2127 | return ret; | |
6ea44308 | 2128 | memcpy(tmp_buf + (offset & (BDRV_SECTOR_SIZE - 1)), buf, len); |
9a8c4cce KW |
2129 | if ((ret = bdrv_write(bs, sector_num, tmp_buf, 1)) < 0) |
2130 | return ret; | |
83f64091 FB |
2131 | count -= len; |
2132 | if (count == 0) | |
2133 | return count1; | |
2134 | sector_num++; | |
2135 | buf += len; | |
2136 | } | |
2137 | ||
2138 | /* write the sectors "in place" */ | |
6ea44308 | 2139 | nb_sectors = count >> BDRV_SECTOR_BITS; |
83f64091 | 2140 | if (nb_sectors > 0) { |
9a8c4cce KW |
2141 | if ((ret = bdrv_write(bs, sector_num, buf, nb_sectors)) < 0) |
2142 | return ret; | |
83f64091 | 2143 | sector_num += nb_sectors; |
6ea44308 | 2144 | len = nb_sectors << BDRV_SECTOR_BITS; |
83f64091 FB |
2145 | buf += len; |
2146 | count -= len; | |
2147 | } | |
2148 | ||
2149 | /* add data from the last sector */ | |
2150 | if (count > 0) { | |
9a8c4cce KW |
2151 | if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0) |
2152 | return ret; | |
83f64091 | 2153 | memcpy(tmp_buf, buf, count); |
9a8c4cce KW |
2154 | if ((ret = bdrv_write(bs, sector_num, tmp_buf, 1)) < 0) |
2155 | return ret; | |
83f64091 FB |
2156 | } |
2157 | return count1; | |
2158 | } | |
83f64091 | 2159 | |
f08145fe KW |
2160 | /* |
2161 | * Writes to the file and ensures that no writes are reordered across this | |
2162 | * request (acts as a barrier) | |
2163 | * | |
2164 | * Returns 0 on success, -errno in error cases. | |
2165 | */ | |
2166 | int bdrv_pwrite_sync(BlockDriverState *bs, int64_t offset, | |
2167 | const void *buf, int count) | |
2168 | { | |
2169 | int ret; | |
2170 | ||
2171 | ret = bdrv_pwrite(bs, offset, buf, count); | |
2172 | if (ret < 0) { | |
2173 | return ret; | |
2174 | } | |
2175 | ||
f05fa4ad PB |
2176 | /* No flush needed for cache modes that already do it */ |
2177 | if (bs->enable_write_cache) { | |
f08145fe KW |
2178 | bdrv_flush(bs); |
2179 | } | |
2180 | ||
2181 | return 0; | |
2182 | } | |
2183 | ||
470c0504 | 2184 | static int coroutine_fn bdrv_co_do_copy_on_readv(BlockDriverState *bs, |
ab185921 SH |
2185 | int64_t sector_num, int nb_sectors, QEMUIOVector *qiov) |
2186 | { | |
2187 | /* Perform I/O through a temporary buffer so that users who scribble over | |
2188 | * their read buffer while the operation is in progress do not end up | |
2189 | * modifying the image file. This is critical for zero-copy guest I/O | |
2190 | * where anything might happen inside guest memory. | |
2191 | */ | |
2192 | void *bounce_buffer; | |
2193 | ||
79c053bd | 2194 | BlockDriver *drv = bs->drv; |
ab185921 SH |
2195 | struct iovec iov; |
2196 | QEMUIOVector bounce_qiov; | |
2197 | int64_t cluster_sector_num; | |
2198 | int cluster_nb_sectors; | |
2199 | size_t skip_bytes; | |
2200 | int ret; | |
2201 | ||
2202 | /* Cover entire cluster so no additional backing file I/O is required when | |
2203 | * allocating cluster in the image file. | |
2204 | */ | |
2205 | round_to_clusters(bs, sector_num, nb_sectors, | |
2206 | &cluster_sector_num, &cluster_nb_sectors); | |
2207 | ||
470c0504 SH |
2208 | trace_bdrv_co_do_copy_on_readv(bs, sector_num, nb_sectors, |
2209 | cluster_sector_num, cluster_nb_sectors); | |
ab185921 SH |
2210 | |
2211 | iov.iov_len = cluster_nb_sectors * BDRV_SECTOR_SIZE; | |
2212 | iov.iov_base = bounce_buffer = qemu_blockalign(bs, iov.iov_len); | |
2213 | qemu_iovec_init_external(&bounce_qiov, &iov, 1); | |
2214 | ||
79c053bd SH |
2215 | ret = drv->bdrv_co_readv(bs, cluster_sector_num, cluster_nb_sectors, |
2216 | &bounce_qiov); | |
ab185921 SH |
2217 | if (ret < 0) { |
2218 | goto err; | |
2219 | } | |
2220 | ||
79c053bd SH |
2221 | if (drv->bdrv_co_write_zeroes && |
2222 | buffer_is_zero(bounce_buffer, iov.iov_len)) { | |
621f0589 KW |
2223 | ret = bdrv_co_do_write_zeroes(bs, cluster_sector_num, |
2224 | cluster_nb_sectors); | |
79c053bd | 2225 | } else { |
f05fa4ad PB |
2226 | /* This does not change the data on the disk, it is not necessary |
2227 | * to flush even in cache=writethrough mode. | |
2228 | */ | |
79c053bd | 2229 | ret = drv->bdrv_co_writev(bs, cluster_sector_num, cluster_nb_sectors, |
ab185921 | 2230 | &bounce_qiov); |
79c053bd SH |
2231 | } |
2232 | ||
ab185921 SH |
2233 | if (ret < 0) { |
2234 | /* It might be okay to ignore write errors for guest requests. If this | |
2235 | * is a deliberate copy-on-read then we don't want to ignore the error. | |
2236 | * Simply report it in all cases. | |
2237 | */ | |
2238 | goto err; | |
2239 | } | |
2240 | ||
2241 | skip_bytes = (sector_num - cluster_sector_num) * BDRV_SECTOR_SIZE; | |
03396148 MT |
2242 | qemu_iovec_from_buf(qiov, 0, bounce_buffer + skip_bytes, |
2243 | nb_sectors * BDRV_SECTOR_SIZE); | |
ab185921 SH |
2244 | |
2245 | err: | |
2246 | qemu_vfree(bounce_buffer); | |
2247 | return ret; | |
2248 | } | |
2249 | ||
c5fbe571 SH |
2250 | /* |
2251 | * Handle a read request in coroutine context | |
2252 | */ | |
2253 | static int coroutine_fn bdrv_co_do_readv(BlockDriverState *bs, | |
470c0504 SH |
2254 | int64_t sector_num, int nb_sectors, QEMUIOVector *qiov, |
2255 | BdrvRequestFlags flags) | |
da1fa91d KW |
2256 | { |
2257 | BlockDriver *drv = bs->drv; | |
dbffbdcf SH |
2258 | BdrvTrackedRequest req; |
2259 | int ret; | |
da1fa91d | 2260 | |
da1fa91d KW |
2261 | if (!drv) { |
2262 | return -ENOMEDIUM; | |
2263 | } | |
2264 | if (bdrv_check_request(bs, sector_num, nb_sectors)) { | |
2265 | return -EIO; | |
2266 | } | |
2267 | ||
98f90dba ZYW |
2268 | /* throttling disk read I/O */ |
2269 | if (bs->io_limits_enabled) { | |
2270 | bdrv_io_limits_intercept(bs, false, nb_sectors); | |
2271 | } | |
2272 | ||
f4658285 | 2273 | if (bs->copy_on_read) { |
470c0504 SH |
2274 | flags |= BDRV_REQ_COPY_ON_READ; |
2275 | } | |
2276 | if (flags & BDRV_REQ_COPY_ON_READ) { | |
2277 | bs->copy_on_read_in_flight++; | |
2278 | } | |
2279 | ||
2280 | if (bs->copy_on_read_in_flight) { | |
f4658285 SH |
2281 | wait_for_overlapping_requests(bs, sector_num, nb_sectors); |
2282 | } | |
2283 | ||
dbffbdcf | 2284 | tracked_request_begin(&req, bs, sector_num, nb_sectors, false); |
ab185921 | 2285 | |
470c0504 | 2286 | if (flags & BDRV_REQ_COPY_ON_READ) { |
ab185921 SH |
2287 | int pnum; |
2288 | ||
2289 | ret = bdrv_co_is_allocated(bs, sector_num, nb_sectors, &pnum); | |
2290 | if (ret < 0) { | |
2291 | goto out; | |
2292 | } | |
2293 | ||
2294 | if (!ret || pnum != nb_sectors) { | |
470c0504 | 2295 | ret = bdrv_co_do_copy_on_readv(bs, sector_num, nb_sectors, qiov); |
ab185921 SH |
2296 | goto out; |
2297 | } | |
2298 | } | |
2299 | ||
dbffbdcf | 2300 | ret = drv->bdrv_co_readv(bs, sector_num, nb_sectors, qiov); |
ab185921 SH |
2301 | |
2302 | out: | |
dbffbdcf | 2303 | tracked_request_end(&req); |
470c0504 SH |
2304 | |
2305 | if (flags & BDRV_REQ_COPY_ON_READ) { | |
2306 | bs->copy_on_read_in_flight--; | |
2307 | } | |
2308 | ||
dbffbdcf | 2309 | return ret; |
da1fa91d KW |
2310 | } |
2311 | ||
c5fbe571 | 2312 | int coroutine_fn bdrv_co_readv(BlockDriverState *bs, int64_t sector_num, |
da1fa91d KW |
2313 | int nb_sectors, QEMUIOVector *qiov) |
2314 | { | |
c5fbe571 | 2315 | trace_bdrv_co_readv(bs, sector_num, nb_sectors); |
da1fa91d | 2316 | |
470c0504 SH |
2317 | return bdrv_co_do_readv(bs, sector_num, nb_sectors, qiov, 0); |
2318 | } | |
2319 | ||
2320 | int coroutine_fn bdrv_co_copy_on_readv(BlockDriverState *bs, | |
2321 | int64_t sector_num, int nb_sectors, QEMUIOVector *qiov) | |
2322 | { | |
2323 | trace_bdrv_co_copy_on_readv(bs, sector_num, nb_sectors); | |
2324 | ||
2325 | return bdrv_co_do_readv(bs, sector_num, nb_sectors, qiov, | |
2326 | BDRV_REQ_COPY_ON_READ); | |
c5fbe571 SH |
2327 | } |
2328 | ||
f08f2dda SH |
2329 | static int coroutine_fn bdrv_co_do_write_zeroes(BlockDriverState *bs, |
2330 | int64_t sector_num, int nb_sectors) | |
2331 | { | |
2332 | BlockDriver *drv = bs->drv; | |
2333 | QEMUIOVector qiov; | |
2334 | struct iovec iov; | |
2335 | int ret; | |
2336 | ||
621f0589 KW |
2337 | /* TODO Emulate only part of misaligned requests instead of letting block |
2338 | * drivers return -ENOTSUP and emulate everything */ | |
2339 | ||
f08f2dda SH |
2340 | /* First try the efficient write zeroes operation */ |
2341 | if (drv->bdrv_co_write_zeroes) { | |
621f0589 KW |
2342 | ret = drv->bdrv_co_write_zeroes(bs, sector_num, nb_sectors); |
2343 | if (ret != -ENOTSUP) { | |
2344 | return ret; | |
2345 | } | |
f08f2dda SH |
2346 | } |
2347 | ||
2348 | /* Fall back to bounce buffer if write zeroes is unsupported */ | |
2349 | iov.iov_len = nb_sectors * BDRV_SECTOR_SIZE; | |
2350 | iov.iov_base = qemu_blockalign(bs, iov.iov_len); | |
2351 | memset(iov.iov_base, 0, iov.iov_len); | |
2352 | qemu_iovec_init_external(&qiov, &iov, 1); | |
2353 | ||
2354 | ret = drv->bdrv_co_writev(bs, sector_num, nb_sectors, &qiov); | |
2355 | ||
2356 | qemu_vfree(iov.iov_base); | |
2357 | return ret; | |
2358 | } | |
2359 | ||
c5fbe571 SH |
2360 | /* |
2361 | * Handle a write request in coroutine context | |
2362 | */ | |
2363 | static int coroutine_fn bdrv_co_do_writev(BlockDriverState *bs, | |
f08f2dda SH |
2364 | int64_t sector_num, int nb_sectors, QEMUIOVector *qiov, |
2365 | BdrvRequestFlags flags) | |
c5fbe571 SH |
2366 | { |
2367 | BlockDriver *drv = bs->drv; | |
dbffbdcf | 2368 | BdrvTrackedRequest req; |
6b7cb247 | 2369 | int ret; |
da1fa91d KW |
2370 | |
2371 | if (!bs->drv) { | |
2372 | return -ENOMEDIUM; | |
2373 | } | |
2374 | if (bs->read_only) { | |
2375 | return -EACCES; | |
2376 | } | |
2377 | if (bdrv_check_request(bs, sector_num, nb_sectors)) { | |
2378 | return -EIO; | |
2379 | } | |
2380 | ||
98f90dba ZYW |
2381 | /* throttling disk write I/O */ |
2382 | if (bs->io_limits_enabled) { | |
2383 | bdrv_io_limits_intercept(bs, true, nb_sectors); | |
2384 | } | |
2385 | ||
470c0504 | 2386 | if (bs->copy_on_read_in_flight) { |
f4658285 SH |
2387 | wait_for_overlapping_requests(bs, sector_num, nb_sectors); |
2388 | } | |
2389 | ||
dbffbdcf SH |
2390 | tracked_request_begin(&req, bs, sector_num, nb_sectors, true); |
2391 | ||
f08f2dda SH |
2392 | if (flags & BDRV_REQ_ZERO_WRITE) { |
2393 | ret = bdrv_co_do_write_zeroes(bs, sector_num, nb_sectors); | |
2394 | } else { | |
2395 | ret = drv->bdrv_co_writev(bs, sector_num, nb_sectors, qiov); | |
2396 | } | |
6b7cb247 | 2397 | |
f05fa4ad PB |
2398 | if (ret == 0 && !bs->enable_write_cache) { |
2399 | ret = bdrv_co_flush(bs); | |
2400 | } | |
2401 | ||
da1fa91d | 2402 | if (bs->dirty_bitmap) { |
1755da16 | 2403 | bdrv_set_dirty(bs, sector_num, nb_sectors); |
da1fa91d KW |
2404 | } |
2405 | ||
2406 | if (bs->wr_highest_sector < sector_num + nb_sectors - 1) { | |
2407 | bs->wr_highest_sector = sector_num + nb_sectors - 1; | |
2408 | } | |
2409 | ||
dbffbdcf SH |
2410 | tracked_request_end(&req); |
2411 | ||
6b7cb247 | 2412 | return ret; |
da1fa91d KW |
2413 | } |
2414 | ||
c5fbe571 SH |
2415 | int coroutine_fn bdrv_co_writev(BlockDriverState *bs, int64_t sector_num, |
2416 | int nb_sectors, QEMUIOVector *qiov) | |
2417 | { | |
2418 | trace_bdrv_co_writev(bs, sector_num, nb_sectors); | |
2419 | ||
f08f2dda SH |
2420 | return bdrv_co_do_writev(bs, sector_num, nb_sectors, qiov, 0); |
2421 | } | |
2422 | ||
2423 | int coroutine_fn bdrv_co_write_zeroes(BlockDriverState *bs, | |
2424 | int64_t sector_num, int nb_sectors) | |
2425 | { | |
2426 | trace_bdrv_co_write_zeroes(bs, sector_num, nb_sectors); | |
2427 | ||
2428 | return bdrv_co_do_writev(bs, sector_num, nb_sectors, NULL, | |
2429 | BDRV_REQ_ZERO_WRITE); | |
c5fbe571 SH |
2430 | } |
2431 | ||
83f64091 FB |
2432 | /** |
2433 | * Truncate file to 'offset' bytes (needed only for file protocols) | |
2434 | */ | |
2435 | int bdrv_truncate(BlockDriverState *bs, int64_t offset) | |
2436 | { | |
2437 | BlockDriver *drv = bs->drv; | |
51762288 | 2438 | int ret; |
83f64091 | 2439 | if (!drv) |
19cb3738 | 2440 | return -ENOMEDIUM; |
83f64091 FB |
2441 | if (!drv->bdrv_truncate) |
2442 | return -ENOTSUP; | |
59f2689d NS |
2443 | if (bs->read_only) |
2444 | return -EACCES; | |
8591675f MT |
2445 | if (bdrv_in_use(bs)) |
2446 | return -EBUSY; | |
51762288 SH |
2447 | ret = drv->bdrv_truncate(bs, offset); |
2448 | if (ret == 0) { | |
2449 | ret = refresh_total_sectors(bs, offset >> BDRV_SECTOR_BITS); | |
145feb17 | 2450 | bdrv_dev_resize_cb(bs); |
51762288 SH |
2451 | } |
2452 | return ret; | |
83f64091 FB |
2453 | } |
2454 | ||
4a1d5e1f FZ |
2455 | /** |
2456 | * Length of a allocated file in bytes. Sparse files are counted by actual | |
2457 | * allocated space. Return < 0 if error or unknown. | |
2458 | */ | |
2459 | int64_t bdrv_get_allocated_file_size(BlockDriverState *bs) | |
2460 | { | |
2461 | BlockDriver *drv = bs->drv; | |
2462 | if (!drv) { | |
2463 | return -ENOMEDIUM; | |
2464 | } | |
2465 | if (drv->bdrv_get_allocated_file_size) { | |
2466 | return drv->bdrv_get_allocated_file_size(bs); | |
2467 | } | |
2468 | if (bs->file) { | |
2469 | return bdrv_get_allocated_file_size(bs->file); | |
2470 | } | |
2471 | return -ENOTSUP; | |
2472 | } | |
2473 | ||
83f64091 FB |
2474 | /** |
2475 | * Length of a file in bytes. Return < 0 if error or unknown. | |
2476 | */ | |
2477 | int64_t bdrv_getlength(BlockDriverState *bs) | |
2478 | { | |
2479 | BlockDriver *drv = bs->drv; | |
2480 | if (!drv) | |
19cb3738 | 2481 | return -ENOMEDIUM; |
51762288 | 2482 | |
2c6942fa | 2483 | if (bs->growable || bdrv_dev_has_removable_media(bs)) { |
46a4e4e6 SH |
2484 | if (drv->bdrv_getlength) { |
2485 | return drv->bdrv_getlength(bs); | |
2486 | } | |
83f64091 | 2487 | } |
46a4e4e6 | 2488 | return bs->total_sectors * BDRV_SECTOR_SIZE; |
fc01f7e7 FB |
2489 | } |
2490 | ||
19cb3738 | 2491 | /* return 0 as number of sectors if no device present or error */ |
96b8f136 | 2492 | void bdrv_get_geometry(BlockDriverState *bs, uint64_t *nb_sectors_ptr) |
fc01f7e7 | 2493 | { |
19cb3738 FB |
2494 | int64_t length; |
2495 | length = bdrv_getlength(bs); | |
2496 | if (length < 0) | |
2497 | length = 0; | |
2498 | else | |
6ea44308 | 2499 | length = length >> BDRV_SECTOR_BITS; |
19cb3738 | 2500 | *nb_sectors_ptr = length; |
fc01f7e7 | 2501 | } |
cf98951b | 2502 | |
0563e191 ZYW |
2503 | /* throttling disk io limits */ |
2504 | void bdrv_set_io_limits(BlockDriverState *bs, | |
2505 | BlockIOLimit *io_limits) | |
2506 | { | |
2507 | bs->io_limits = *io_limits; | |
2508 | bs->io_limits_enabled = bdrv_io_limits_enabled(bs); | |
2509 | } | |
2510 | ||
ff06f5f3 PB |
2511 | void bdrv_set_on_error(BlockDriverState *bs, BlockdevOnError on_read_error, |
2512 | BlockdevOnError on_write_error) | |
abd7f68d MA |
2513 | { |
2514 | bs->on_read_error = on_read_error; | |
2515 | bs->on_write_error = on_write_error; | |
2516 | } | |
2517 | ||
1ceee0d5 | 2518 | BlockdevOnError bdrv_get_on_error(BlockDriverState *bs, bool is_read) |
abd7f68d MA |
2519 | { |
2520 | return is_read ? bs->on_read_error : bs->on_write_error; | |
2521 | } | |
2522 | ||
3e1caa5f PB |
2523 | BlockErrorAction bdrv_get_error_action(BlockDriverState *bs, bool is_read, int error) |
2524 | { | |
2525 | BlockdevOnError on_err = is_read ? bs->on_read_error : bs->on_write_error; | |
2526 | ||
2527 | switch (on_err) { | |
2528 | case BLOCKDEV_ON_ERROR_ENOSPC: | |
2529 | return (error == ENOSPC) ? BDRV_ACTION_STOP : BDRV_ACTION_REPORT; | |
2530 | case BLOCKDEV_ON_ERROR_STOP: | |
2531 | return BDRV_ACTION_STOP; | |
2532 | case BLOCKDEV_ON_ERROR_REPORT: | |
2533 | return BDRV_ACTION_REPORT; | |
2534 | case BLOCKDEV_ON_ERROR_IGNORE: | |
2535 | return BDRV_ACTION_IGNORE; | |
2536 | default: | |
2537 | abort(); | |
2538 | } | |
2539 | } | |
2540 | ||
2541 | /* This is done by device models because, while the block layer knows | |
2542 | * about the error, it does not know whether an operation comes from | |
2543 | * the device or the block layer (from a job, for example). | |
2544 | */ | |
2545 | void bdrv_error_action(BlockDriverState *bs, BlockErrorAction action, | |
2546 | bool is_read, int error) | |
2547 | { | |
2548 | assert(error >= 0); | |
32c81a4a | 2549 | bdrv_emit_qmp_error_event(bs, QEVENT_BLOCK_IO_ERROR, action, is_read); |
3e1caa5f PB |
2550 | if (action == BDRV_ACTION_STOP) { |
2551 | vm_stop(RUN_STATE_IO_ERROR); | |
2552 | bdrv_iostatus_set_err(bs, error); | |
2553 | } | |
2554 | } | |
2555 | ||
b338082b FB |
2556 | int bdrv_is_read_only(BlockDriverState *bs) |
2557 | { | |
2558 | return bs->read_only; | |
2559 | } | |
2560 | ||
985a03b0 TS |
2561 | int bdrv_is_sg(BlockDriverState *bs) |
2562 | { | |
2563 | return bs->sg; | |
2564 | } | |
2565 | ||
e900a7b7 CH |
2566 | int bdrv_enable_write_cache(BlockDriverState *bs) |
2567 | { | |
2568 | return bs->enable_write_cache; | |
2569 | } | |
2570 | ||
425b0148 PB |
2571 | void bdrv_set_enable_write_cache(BlockDriverState *bs, bool wce) |
2572 | { | |
2573 | bs->enable_write_cache = wce; | |
55b110f2 JC |
2574 | |
2575 | /* so a reopen() will preserve wce */ | |
2576 | if (wce) { | |
2577 | bs->open_flags |= BDRV_O_CACHE_WB; | |
2578 | } else { | |
2579 | bs->open_flags &= ~BDRV_O_CACHE_WB; | |
2580 | } | |
425b0148 PB |
2581 | } |
2582 | ||
ea2384d3 FB |
2583 | int bdrv_is_encrypted(BlockDriverState *bs) |
2584 | { | |
2585 | if (bs->backing_hd && bs->backing_hd->encrypted) | |
2586 | return 1; | |
2587 | return bs->encrypted; | |
2588 | } | |
2589 | ||
c0f4ce77 AL |
2590 | int bdrv_key_required(BlockDriverState *bs) |
2591 | { | |
2592 | BlockDriverState *backing_hd = bs->backing_hd; | |
2593 | ||
2594 | if (backing_hd && backing_hd->encrypted && !backing_hd->valid_key) | |
2595 | return 1; | |
2596 | return (bs->encrypted && !bs->valid_key); | |
2597 | } | |
2598 | ||
ea2384d3 FB |
2599 | int bdrv_set_key(BlockDriverState *bs, const char *key) |
2600 | { | |
2601 | int ret; | |
2602 | if (bs->backing_hd && bs->backing_hd->encrypted) { | |
2603 | ret = bdrv_set_key(bs->backing_hd, key); | |
2604 | if (ret < 0) | |
2605 | return ret; | |
2606 | if (!bs->encrypted) | |
2607 | return 0; | |
2608 | } | |
fd04a2ae SH |
2609 | if (!bs->encrypted) { |
2610 | return -EINVAL; | |
2611 | } else if (!bs->drv || !bs->drv->bdrv_set_key) { | |
2612 | return -ENOMEDIUM; | |
2613 | } | |
c0f4ce77 | 2614 | ret = bs->drv->bdrv_set_key(bs, key); |
bb5fc20f AL |
2615 | if (ret < 0) { |
2616 | bs->valid_key = 0; | |
2617 | } else if (!bs->valid_key) { | |
2618 | bs->valid_key = 1; | |
2619 | /* call the change callback now, we skipped it on open */ | |
7d4b4ba5 | 2620 | bdrv_dev_change_media_cb(bs, true); |
bb5fc20f | 2621 | } |
c0f4ce77 | 2622 | return ret; |
ea2384d3 FB |
2623 | } |
2624 | ||
f8d6bba1 | 2625 | const char *bdrv_get_format_name(BlockDriverState *bs) |
ea2384d3 | 2626 | { |
f8d6bba1 | 2627 | return bs->drv ? bs->drv->format_name : NULL; |
ea2384d3 FB |
2628 | } |
2629 | ||
5fafdf24 | 2630 | void bdrv_iterate_format(void (*it)(void *opaque, const char *name), |
ea2384d3 FB |
2631 | void *opaque) |
2632 | { | |
2633 | BlockDriver *drv; | |
2634 | ||
8a22f02a | 2635 | QLIST_FOREACH(drv, &bdrv_drivers, list) { |
ea2384d3 FB |
2636 | it(opaque, drv->format_name); |
2637 | } | |
2638 | } | |
2639 | ||
b338082b FB |
2640 | BlockDriverState *bdrv_find(const char *name) |
2641 | { | |
2642 | BlockDriverState *bs; | |
2643 | ||
1b7bdbc1 SH |
2644 | QTAILQ_FOREACH(bs, &bdrv_states, list) { |
2645 | if (!strcmp(name, bs->device_name)) { | |
b338082b | 2646 | return bs; |
1b7bdbc1 | 2647 | } |
b338082b FB |
2648 | } |
2649 | return NULL; | |
2650 | } | |
2651 | ||
2f399b0a MA |
2652 | BlockDriverState *bdrv_next(BlockDriverState *bs) |
2653 | { | |
2654 | if (!bs) { | |
2655 | return QTAILQ_FIRST(&bdrv_states); | |
2656 | } | |
2657 | return QTAILQ_NEXT(bs, list); | |
2658 | } | |
2659 | ||
51de9760 | 2660 | void bdrv_iterate(void (*it)(void *opaque, BlockDriverState *bs), void *opaque) |
81d0912d FB |
2661 | { |
2662 | BlockDriverState *bs; | |
2663 | ||
1b7bdbc1 | 2664 | QTAILQ_FOREACH(bs, &bdrv_states, list) { |
51de9760 | 2665 | it(opaque, bs); |
81d0912d FB |
2666 | } |
2667 | } | |
2668 | ||
ea2384d3 FB |
2669 | const char *bdrv_get_device_name(BlockDriverState *bs) |
2670 | { | |
2671 | return bs->device_name; | |
2672 | } | |
2673 | ||
c8433287 MA |
2674 | int bdrv_get_flags(BlockDriverState *bs) |
2675 | { | |
2676 | return bs->open_flags; | |
2677 | } | |
2678 | ||
c6ca28d6 AL |
2679 | void bdrv_flush_all(void) |
2680 | { | |
2681 | BlockDriverState *bs; | |
2682 | ||
1b7bdbc1 | 2683 | QTAILQ_FOREACH(bs, &bdrv_states, list) { |
29cdb251 | 2684 | bdrv_flush(bs); |
1b7bdbc1 | 2685 | } |
c6ca28d6 AL |
2686 | } |
2687 | ||
f2feebbd KW |
2688 | int bdrv_has_zero_init(BlockDriverState *bs) |
2689 | { | |
2690 | assert(bs->drv); | |
2691 | ||
336c1c12 KW |
2692 | if (bs->drv->bdrv_has_zero_init) { |
2693 | return bs->drv->bdrv_has_zero_init(bs); | |
f2feebbd KW |
2694 | } |
2695 | ||
2696 | return 1; | |
2697 | } | |
2698 | ||
376ae3f1 SH |
2699 | typedef struct BdrvCoIsAllocatedData { |
2700 | BlockDriverState *bs; | |
2701 | int64_t sector_num; | |
2702 | int nb_sectors; | |
2703 | int *pnum; | |
2704 | int ret; | |
2705 | bool done; | |
2706 | } BdrvCoIsAllocatedData; | |
2707 | ||
f58c7b35 TS |
2708 | /* |
2709 | * Returns true iff the specified sector is present in the disk image. Drivers | |
2710 | * not implementing the functionality are assumed to not support backing files, | |
2711 | * hence all their sectors are reported as allocated. | |
2712 | * | |
bd9533e3 SH |
2713 | * If 'sector_num' is beyond the end of the disk image the return value is 0 |
2714 | * and 'pnum' is set to 0. | |
2715 | * | |
f58c7b35 TS |
2716 | * 'pnum' is set to the number of sectors (including and immediately following |
2717 | * the specified sector) that are known to be in the same | |
2718 | * allocated/unallocated state. | |
2719 | * | |
bd9533e3 SH |
2720 | * 'nb_sectors' is the max value 'pnum' should be set to. If nb_sectors goes |
2721 | * beyond the end of the disk image it will be clamped. | |
f58c7b35 | 2722 | */ |
060f51c9 SH |
2723 | int coroutine_fn bdrv_co_is_allocated(BlockDriverState *bs, int64_t sector_num, |
2724 | int nb_sectors, int *pnum) | |
f58c7b35 | 2725 | { |
bd9533e3 SH |
2726 | int64_t n; |
2727 | ||
2728 | if (sector_num >= bs->total_sectors) { | |
2729 | *pnum = 0; | |
2730 | return 0; | |
2731 | } | |
2732 | ||
2733 | n = bs->total_sectors - sector_num; | |
2734 | if (n < nb_sectors) { | |
2735 | nb_sectors = n; | |
2736 | } | |
2737 | ||
6aebab14 | 2738 | if (!bs->drv->bdrv_co_is_allocated) { |
bd9533e3 | 2739 | *pnum = nb_sectors; |
f58c7b35 TS |
2740 | return 1; |
2741 | } | |
6aebab14 | 2742 | |
060f51c9 SH |
2743 | return bs->drv->bdrv_co_is_allocated(bs, sector_num, nb_sectors, pnum); |
2744 | } | |
2745 | ||
2746 | /* Coroutine wrapper for bdrv_is_allocated() */ | |
2747 | static void coroutine_fn bdrv_is_allocated_co_entry(void *opaque) | |
2748 | { | |
2749 | BdrvCoIsAllocatedData *data = opaque; | |
2750 | BlockDriverState *bs = data->bs; | |
2751 | ||
2752 | data->ret = bdrv_co_is_allocated(bs, data->sector_num, data->nb_sectors, | |
2753 | data->pnum); | |
2754 | data->done = true; | |
2755 | } | |
2756 | ||
2757 | /* | |
2758 | * Synchronous wrapper around bdrv_co_is_allocated(). | |
2759 | * | |
2760 | * See bdrv_co_is_allocated() for details. | |
2761 | */ | |
2762 | int bdrv_is_allocated(BlockDriverState *bs, int64_t sector_num, int nb_sectors, | |
2763 | int *pnum) | |
2764 | { | |
6aebab14 SH |
2765 | Coroutine *co; |
2766 | BdrvCoIsAllocatedData data = { | |
2767 | .bs = bs, | |
2768 | .sector_num = sector_num, | |
2769 | .nb_sectors = nb_sectors, | |
2770 | .pnum = pnum, | |
2771 | .done = false, | |
2772 | }; | |
2773 | ||
2774 | co = qemu_coroutine_create(bdrv_is_allocated_co_entry); | |
2775 | qemu_coroutine_enter(co, &data); | |
2776 | while (!data.done) { | |
2777 | qemu_aio_wait(); | |
2778 | } | |
2779 | return data.ret; | |
f58c7b35 TS |
2780 | } |
2781 | ||
188a7bbf PB |
2782 | /* |
2783 | * Given an image chain: ... -> [BASE] -> [INTER1] -> [INTER2] -> [TOP] | |
2784 | * | |
2785 | * Return true if the given sector is allocated in any image between | |
2786 | * BASE and TOP (inclusive). BASE can be NULL to check if the given | |
2787 | * sector is allocated in any image of the chain. Return false otherwise. | |
2788 | * | |
2789 | * 'pnum' is set to the number of sectors (including and immediately following | |
2790 | * the specified sector) that are known to be in the same | |
2791 | * allocated/unallocated state. | |
2792 | * | |
2793 | */ | |
2794 | int coroutine_fn bdrv_co_is_allocated_above(BlockDriverState *top, | |
2795 | BlockDriverState *base, | |
2796 | int64_t sector_num, | |
2797 | int nb_sectors, int *pnum) | |
2798 | { | |
2799 | BlockDriverState *intermediate; | |
2800 | int ret, n = nb_sectors; | |
2801 | ||
2802 | intermediate = top; | |
2803 | while (intermediate && intermediate != base) { | |
2804 | int pnum_inter; | |
2805 | ret = bdrv_co_is_allocated(intermediate, sector_num, nb_sectors, | |
2806 | &pnum_inter); | |
2807 | if (ret < 0) { | |
2808 | return ret; | |
2809 | } else if (ret) { | |
2810 | *pnum = pnum_inter; | |
2811 | return 1; | |
2812 | } | |
2813 | ||
2814 | /* | |
2815 | * [sector_num, nb_sectors] is unallocated on top but intermediate | |
2816 | * might have | |
2817 | * | |
2818 | * [sector_num+x, nr_sectors] allocated. | |
2819 | */ | |
2820 | if (n > pnum_inter) { | |
2821 | n = pnum_inter; | |
2822 | } | |
2823 | ||
2824 | intermediate = intermediate->backing_hd; | |
2825 | } | |
2826 | ||
2827 | *pnum = n; | |
2828 | return 0; | |
2829 | } | |
2830 | ||
ac84adac | 2831 | BlockInfo *bdrv_query_info(BlockDriverState *bs) |
b338082b | 2832 | { |
ac84adac PB |
2833 | BlockInfo *info = g_malloc0(sizeof(*info)); |
2834 | info->device = g_strdup(bs->device_name); | |
2835 | info->type = g_strdup("unknown"); | |
2836 | info->locked = bdrv_dev_is_medium_locked(bs); | |
2837 | info->removable = bdrv_dev_has_removable_media(bs); | |
b338082b | 2838 | |
ac84adac PB |
2839 | if (bdrv_dev_has_removable_media(bs)) { |
2840 | info->has_tray_open = true; | |
2841 | info->tray_open = bdrv_dev_is_tray_open(bs); | |
2842 | } | |
d15e5465 | 2843 | |
ac84adac PB |
2844 | if (bdrv_iostatus_is_enabled(bs)) { |
2845 | info->has_io_status = true; | |
2846 | info->io_status = bs->iostatus; | |
2847 | } | |
d15e5465 | 2848 | |
b9a9b3a4 PB |
2849 | if (bs->dirty_bitmap) { |
2850 | info->has_dirty = true; | |
2851 | info->dirty = g_malloc0(sizeof(*info->dirty)); | |
2852 | info->dirty->count = bdrv_get_dirty_count(bs) * | |
2853 | BDRV_SECTORS_PER_DIRTY_CHUNK * BDRV_SECTOR_SIZE; | |
2854 | } | |
d15e5465 | 2855 | |
ac84adac PB |
2856 | if (bs->drv) { |
2857 | info->has_inserted = true; | |
2858 | info->inserted = g_malloc0(sizeof(*info->inserted)); | |
2859 | info->inserted->file = g_strdup(bs->filename); | |
2860 | info->inserted->ro = bs->read_only; | |
2861 | info->inserted->drv = g_strdup(bs->drv->format_name); | |
2862 | info->inserted->encrypted = bs->encrypted; | |
2863 | info->inserted->encryption_key_missing = bdrv_key_required(bs); | |
2864 | ||
2865 | if (bs->backing_file[0]) { | |
2866 | info->inserted->has_backing_file = true; | |
2867 | info->inserted->backing_file = g_strdup(bs->backing_file); | |
e4def80b | 2868 | } |
f04ef601 | 2869 | |
ac84adac PB |
2870 | info->inserted->backing_file_depth = bdrv_get_backing_file_depth(bs); |
2871 | ||
2872 | if (bs->io_limits_enabled) { | |
2873 | info->inserted->bps = | |
2874 | bs->io_limits.bps[BLOCK_IO_LIMIT_TOTAL]; | |
2875 | info->inserted->bps_rd = | |
2876 | bs->io_limits.bps[BLOCK_IO_LIMIT_READ]; | |
2877 | info->inserted->bps_wr = | |
2878 | bs->io_limits.bps[BLOCK_IO_LIMIT_WRITE]; | |
2879 | info->inserted->iops = | |
2880 | bs->io_limits.iops[BLOCK_IO_LIMIT_TOTAL]; | |
2881 | info->inserted->iops_rd = | |
2882 | bs->io_limits.iops[BLOCK_IO_LIMIT_READ]; | |
2883 | info->inserted->iops_wr = | |
2884 | bs->io_limits.iops[BLOCK_IO_LIMIT_WRITE]; | |
f04ef601 | 2885 | } |
ac84adac PB |
2886 | } |
2887 | return info; | |
2888 | } | |
f04ef601 | 2889 | |
ac84adac PB |
2890 | BlockInfoList *qmp_query_block(Error **errp) |
2891 | { | |
2892 | BlockInfoList *head = NULL, **p_next = &head; | |
2893 | BlockDriverState *bs; | |
727f005e | 2894 | |
ac84adac PB |
2895 | QTAILQ_FOREACH(bs, &bdrv_states, list) { |
2896 | BlockInfoList *info = g_malloc0(sizeof(*info)); | |
2897 | info->value = bdrv_query_info(bs); | |
d15e5465 | 2898 | |
ac84adac PB |
2899 | *p_next = info; |
2900 | p_next = &info->next; | |
b338082b | 2901 | } |
d15e5465 | 2902 | |
b2023818 | 2903 | return head; |
b338082b | 2904 | } |
a36e69dd | 2905 | |
9887b616 | 2906 | BlockStats *bdrv_query_stats(const BlockDriverState *bs) |
f11f57e4 LC |
2907 | { |
2908 | BlockStats *s; | |
2909 | ||
2910 | s = g_malloc0(sizeof(*s)); | |
2911 | ||
2912 | if (bs->device_name[0]) { | |
2913 | s->has_device = true; | |
2914 | s->device = g_strdup(bs->device_name); | |
294cc35f KW |
2915 | } |
2916 | ||
f11f57e4 LC |
2917 | s->stats = g_malloc0(sizeof(*s->stats)); |
2918 | s->stats->rd_bytes = bs->nr_bytes[BDRV_ACCT_READ]; | |
2919 | s->stats->wr_bytes = bs->nr_bytes[BDRV_ACCT_WRITE]; | |
2920 | s->stats->rd_operations = bs->nr_ops[BDRV_ACCT_READ]; | |
2921 | s->stats->wr_operations = bs->nr_ops[BDRV_ACCT_WRITE]; | |
2922 | s->stats->wr_highest_offset = bs->wr_highest_sector * BDRV_SECTOR_SIZE; | |
2923 | s->stats->flush_operations = bs->nr_ops[BDRV_ACCT_FLUSH]; | |
2924 | s->stats->wr_total_time_ns = bs->total_time_ns[BDRV_ACCT_WRITE]; | |
2925 | s->stats->rd_total_time_ns = bs->total_time_ns[BDRV_ACCT_READ]; | |
2926 | s->stats->flush_total_time_ns = bs->total_time_ns[BDRV_ACCT_FLUSH]; | |
2927 | ||
294cc35f | 2928 | if (bs->file) { |
f11f57e4 | 2929 | s->has_parent = true; |
9887b616 | 2930 | s->parent = bdrv_query_stats(bs->file); |
294cc35f KW |
2931 | } |
2932 | ||
f11f57e4 | 2933 | return s; |
294cc35f KW |
2934 | } |
2935 | ||
f11f57e4 | 2936 | BlockStatsList *qmp_query_blockstats(Error **errp) |
218a536a | 2937 | { |
9887b616 | 2938 | BlockStatsList *head = NULL, **p_next = &head; |
a36e69dd TS |
2939 | BlockDriverState *bs; |
2940 | ||
1b7bdbc1 | 2941 | QTAILQ_FOREACH(bs, &bdrv_states, list) { |
f11f57e4 | 2942 | BlockStatsList *info = g_malloc0(sizeof(*info)); |
9887b616 | 2943 | info->value = bdrv_query_stats(bs); |
f11f57e4 | 2944 | |
9887b616 PB |
2945 | *p_next = info; |
2946 | p_next = &info->next; | |
a36e69dd | 2947 | } |
218a536a | 2948 | |
f11f57e4 | 2949 | return head; |
a36e69dd | 2950 | } |
ea2384d3 | 2951 | |
045df330 AL |
2952 | const char *bdrv_get_encrypted_filename(BlockDriverState *bs) |
2953 | { | |
2954 | if (bs->backing_hd && bs->backing_hd->encrypted) | |
2955 | return bs->backing_file; | |
2956 | else if (bs->encrypted) | |
2957 | return bs->filename; | |
2958 | else | |
2959 | return NULL; | |
2960 | } | |
2961 | ||
5fafdf24 | 2962 | void bdrv_get_backing_filename(BlockDriverState *bs, |
83f64091 FB |
2963 | char *filename, int filename_size) |
2964 | { | |
3574c608 | 2965 | pstrcpy(filename, filename_size, bs->backing_file); |
83f64091 FB |
2966 | } |
2967 | ||
5fafdf24 | 2968 | int bdrv_write_compressed(BlockDriverState *bs, int64_t sector_num, |
faea38e7 FB |
2969 | const uint8_t *buf, int nb_sectors) |
2970 | { | |
2971 | BlockDriver *drv = bs->drv; | |
2972 | if (!drv) | |
19cb3738 | 2973 | return -ENOMEDIUM; |
faea38e7 FB |
2974 | if (!drv->bdrv_write_compressed) |
2975 | return -ENOTSUP; | |
fbb7b4e0 KW |
2976 | if (bdrv_check_request(bs, sector_num, nb_sectors)) |
2977 | return -EIO; | |
a55eb92c | 2978 | |
1755da16 | 2979 | assert(!bs->dirty_bitmap); |
a55eb92c | 2980 | |
faea38e7 FB |
2981 | return drv->bdrv_write_compressed(bs, sector_num, buf, nb_sectors); |
2982 | } | |
3b46e624 | 2983 | |
faea38e7 FB |
2984 | int bdrv_get_info(BlockDriverState *bs, BlockDriverInfo *bdi) |
2985 | { | |
2986 | BlockDriver *drv = bs->drv; | |
2987 | if (!drv) | |
19cb3738 | 2988 | return -ENOMEDIUM; |
faea38e7 FB |
2989 | if (!drv->bdrv_get_info) |
2990 | return -ENOTSUP; | |
2991 | memset(bdi, 0, sizeof(*bdi)); | |
2992 | return drv->bdrv_get_info(bs, bdi); | |
2993 | } | |
2994 | ||
45566e9c CH |
2995 | int bdrv_save_vmstate(BlockDriverState *bs, const uint8_t *buf, |
2996 | int64_t pos, int size) | |
178e08a5 AL |
2997 | { |
2998 | BlockDriver *drv = bs->drv; | |
2999 | if (!drv) | |
3000 | return -ENOMEDIUM; | |
7cdb1f6d MK |
3001 | if (drv->bdrv_save_vmstate) |
3002 | return drv->bdrv_save_vmstate(bs, buf, pos, size); | |
3003 | if (bs->file) | |
3004 | return bdrv_save_vmstate(bs->file, buf, pos, size); | |
3005 | return -ENOTSUP; | |
178e08a5 AL |
3006 | } |
3007 | ||
45566e9c CH |
3008 | int bdrv_load_vmstate(BlockDriverState *bs, uint8_t *buf, |
3009 | int64_t pos, int size) | |
178e08a5 AL |
3010 | { |
3011 | BlockDriver *drv = bs->drv; | |
3012 | if (!drv) | |
3013 | return -ENOMEDIUM; | |
7cdb1f6d MK |
3014 | if (drv->bdrv_load_vmstate) |
3015 | return drv->bdrv_load_vmstate(bs, buf, pos, size); | |
3016 | if (bs->file) | |
3017 | return bdrv_load_vmstate(bs->file, buf, pos, size); | |
3018 | return -ENOTSUP; | |
178e08a5 AL |
3019 | } |
3020 | ||
8b9b0cc2 KW |
3021 | void bdrv_debug_event(BlockDriverState *bs, BlkDebugEvent event) |
3022 | { | |
3023 | BlockDriver *drv = bs->drv; | |
3024 | ||
3025 | if (!drv || !drv->bdrv_debug_event) { | |
3026 | return; | |
3027 | } | |
3028 | ||
0ed8b6f6 | 3029 | drv->bdrv_debug_event(bs, event); |
8b9b0cc2 KW |
3030 | |
3031 | } | |
3032 | ||
faea38e7 FB |
3033 | /**************************************************************/ |
3034 | /* handling of snapshots */ | |
3035 | ||
feeee5ac MDCF |
3036 | int bdrv_can_snapshot(BlockDriverState *bs) |
3037 | { | |
3038 | BlockDriver *drv = bs->drv; | |
07b70bfb | 3039 | if (!drv || !bdrv_is_inserted(bs) || bdrv_is_read_only(bs)) { |
feeee5ac MDCF |
3040 | return 0; |
3041 | } | |
3042 | ||
3043 | if (!drv->bdrv_snapshot_create) { | |
3044 | if (bs->file != NULL) { | |
3045 | return bdrv_can_snapshot(bs->file); | |
3046 | } | |
3047 | return 0; | |
3048 | } | |
3049 | ||
3050 | return 1; | |
3051 | } | |
3052 | ||
199630b6 BS |
3053 | int bdrv_is_snapshot(BlockDriverState *bs) |
3054 | { | |
3055 | return !!(bs->open_flags & BDRV_O_SNAPSHOT); | |
3056 | } | |
3057 | ||
f9092b10 MA |
3058 | BlockDriverState *bdrv_snapshots(void) |
3059 | { | |
3060 | BlockDriverState *bs; | |
3061 | ||
3ac906f7 | 3062 | if (bs_snapshots) { |
f9092b10 | 3063 | return bs_snapshots; |
3ac906f7 | 3064 | } |
f9092b10 MA |
3065 | |
3066 | bs = NULL; | |
3067 | while ((bs = bdrv_next(bs))) { | |
3068 | if (bdrv_can_snapshot(bs)) { | |
3ac906f7 MA |
3069 | bs_snapshots = bs; |
3070 | return bs; | |
f9092b10 MA |
3071 | } |
3072 | } | |
3073 | return NULL; | |
f9092b10 MA |
3074 | } |
3075 | ||
5fafdf24 | 3076 | int bdrv_snapshot_create(BlockDriverState *bs, |
faea38e7 FB |
3077 | QEMUSnapshotInfo *sn_info) |
3078 | { | |
3079 | BlockDriver *drv = bs->drv; | |
3080 | if (!drv) | |
19cb3738 | 3081 | return -ENOMEDIUM; |
7cdb1f6d MK |
3082 | if (drv->bdrv_snapshot_create) |
3083 | return drv->bdrv_snapshot_create(bs, sn_info); | |
3084 | if (bs->file) | |
3085 | return bdrv_snapshot_create(bs->file, sn_info); | |
3086 | return -ENOTSUP; | |
faea38e7 FB |
3087 | } |
3088 | ||
5fafdf24 | 3089 | int bdrv_snapshot_goto(BlockDriverState *bs, |
faea38e7 FB |
3090 | const char *snapshot_id) |
3091 | { | |
3092 | BlockDriver *drv = bs->drv; | |
7cdb1f6d MK |
3093 | int ret, open_ret; |
3094 | ||
faea38e7 | 3095 | if (!drv) |
19cb3738 | 3096 | return -ENOMEDIUM; |
7cdb1f6d MK |
3097 | if (drv->bdrv_snapshot_goto) |
3098 | return drv->bdrv_snapshot_goto(bs, snapshot_id); | |
3099 | ||
3100 | if (bs->file) { | |
3101 | drv->bdrv_close(bs); | |
3102 | ret = bdrv_snapshot_goto(bs->file, snapshot_id); | |
3103 | open_ret = drv->bdrv_open(bs, bs->open_flags); | |
3104 | if (open_ret < 0) { | |
3105 | bdrv_delete(bs->file); | |
3106 | bs->drv = NULL; | |
3107 | return open_ret; | |
3108 | } | |
3109 | return ret; | |
3110 | } | |
3111 | ||
3112 | return -ENOTSUP; | |
faea38e7 FB |
3113 | } |
3114 | ||
3115 | int bdrv_snapshot_delete(BlockDriverState *bs, const char *snapshot_id) | |
3116 | { | |
3117 | BlockDriver *drv = bs->drv; | |
3118 | if (!drv) | |
19cb3738 | 3119 | return -ENOMEDIUM; |
7cdb1f6d MK |
3120 | if (drv->bdrv_snapshot_delete) |
3121 | return drv->bdrv_snapshot_delete(bs, snapshot_id); | |
3122 | if (bs->file) | |
3123 | return bdrv_snapshot_delete(bs->file, snapshot_id); | |
3124 | return -ENOTSUP; | |
faea38e7 FB |
3125 | } |
3126 | ||
5fafdf24 | 3127 | int bdrv_snapshot_list(BlockDriverState *bs, |
faea38e7 FB |
3128 | QEMUSnapshotInfo **psn_info) |
3129 | { | |
3130 | BlockDriver *drv = bs->drv; | |
3131 | if (!drv) | |
19cb3738 | 3132 | return -ENOMEDIUM; |
7cdb1f6d MK |
3133 | if (drv->bdrv_snapshot_list) |
3134 | return drv->bdrv_snapshot_list(bs, psn_info); | |
3135 | if (bs->file) | |
3136 | return bdrv_snapshot_list(bs->file, psn_info); | |
3137 | return -ENOTSUP; | |
faea38e7 FB |
3138 | } |
3139 | ||
51ef6727 | 3140 | int bdrv_snapshot_load_tmp(BlockDriverState *bs, |
3141 | const char *snapshot_name) | |
3142 | { | |
3143 | BlockDriver *drv = bs->drv; | |
3144 | if (!drv) { | |
3145 | return -ENOMEDIUM; | |
3146 | } | |
3147 | if (!bs->read_only) { | |
3148 | return -EINVAL; | |
3149 | } | |
3150 | if (drv->bdrv_snapshot_load_tmp) { | |
3151 | return drv->bdrv_snapshot_load_tmp(bs, snapshot_name); | |
3152 | } | |
3153 | return -ENOTSUP; | |
3154 | } | |
3155 | ||
b1b1d783 JC |
3156 | /* backing_file can either be relative, or absolute, or a protocol. If it is |
3157 | * relative, it must be relative to the chain. So, passing in bs->filename | |
3158 | * from a BDS as backing_file should not be done, as that may be relative to | |
3159 | * the CWD rather than the chain. */ | |
e8a6bb9c MT |
3160 | BlockDriverState *bdrv_find_backing_image(BlockDriverState *bs, |
3161 | const char *backing_file) | |
3162 | { | |
b1b1d783 JC |
3163 | char *filename_full = NULL; |
3164 | char *backing_file_full = NULL; | |
3165 | char *filename_tmp = NULL; | |
3166 | int is_protocol = 0; | |
3167 | BlockDriverState *curr_bs = NULL; | |
3168 | BlockDriverState *retval = NULL; | |
3169 | ||
3170 | if (!bs || !bs->drv || !backing_file) { | |
e8a6bb9c MT |
3171 | return NULL; |
3172 | } | |
3173 | ||
b1b1d783 JC |
3174 | filename_full = g_malloc(PATH_MAX); |
3175 | backing_file_full = g_malloc(PATH_MAX); | |
3176 | filename_tmp = g_malloc(PATH_MAX); | |
3177 | ||
3178 | is_protocol = path_has_protocol(backing_file); | |
3179 | ||
3180 | for (curr_bs = bs; curr_bs->backing_hd; curr_bs = curr_bs->backing_hd) { | |
3181 | ||
3182 | /* If either of the filename paths is actually a protocol, then | |
3183 | * compare unmodified paths; otherwise make paths relative */ | |
3184 | if (is_protocol || path_has_protocol(curr_bs->backing_file)) { | |
3185 | if (strcmp(backing_file, curr_bs->backing_file) == 0) { | |
3186 | retval = curr_bs->backing_hd; | |
3187 | break; | |
3188 | } | |
e8a6bb9c | 3189 | } else { |
b1b1d783 JC |
3190 | /* If not an absolute filename path, make it relative to the current |
3191 | * image's filename path */ | |
3192 | path_combine(filename_tmp, PATH_MAX, curr_bs->filename, | |
3193 | backing_file); | |
3194 | ||
3195 | /* We are going to compare absolute pathnames */ | |
3196 | if (!realpath(filename_tmp, filename_full)) { | |
3197 | continue; | |
3198 | } | |
3199 | ||
3200 | /* We need to make sure the backing filename we are comparing against | |
3201 | * is relative to the current image filename (or absolute) */ | |
3202 | path_combine(filename_tmp, PATH_MAX, curr_bs->filename, | |
3203 | curr_bs->backing_file); | |
3204 | ||
3205 | if (!realpath(filename_tmp, backing_file_full)) { | |
3206 | continue; | |
3207 | } | |
3208 | ||
3209 | if (strcmp(backing_file_full, filename_full) == 0) { | |
3210 | retval = curr_bs->backing_hd; | |
3211 | break; | |
3212 | } | |
e8a6bb9c MT |
3213 | } |
3214 | } | |
3215 | ||
b1b1d783 JC |
3216 | g_free(filename_full); |
3217 | g_free(backing_file_full); | |
3218 | g_free(filename_tmp); | |
3219 | return retval; | |
e8a6bb9c MT |
3220 | } |
3221 | ||
f198fd1c BC |
3222 | int bdrv_get_backing_file_depth(BlockDriverState *bs) |
3223 | { | |
3224 | if (!bs->drv) { | |
3225 | return 0; | |
3226 | } | |
3227 | ||
3228 | if (!bs->backing_hd) { | |
3229 | return 0; | |
3230 | } | |
3231 | ||
3232 | return 1 + bdrv_get_backing_file_depth(bs->backing_hd); | |
3233 | } | |
3234 | ||
79fac568 JC |
3235 | BlockDriverState *bdrv_find_base(BlockDriverState *bs) |
3236 | { | |
3237 | BlockDriverState *curr_bs = NULL; | |
3238 | ||
3239 | if (!bs) { | |
3240 | return NULL; | |
3241 | } | |
3242 | ||
3243 | curr_bs = bs; | |
3244 | ||
3245 | while (curr_bs->backing_hd) { | |
3246 | curr_bs = curr_bs->backing_hd; | |
3247 | } | |
3248 | return curr_bs; | |
3249 | } | |
3250 | ||
faea38e7 FB |
3251 | #define NB_SUFFIXES 4 |
3252 | ||
3253 | char *get_human_readable_size(char *buf, int buf_size, int64_t size) | |
3254 | { | |
3255 | static const char suffixes[NB_SUFFIXES] = "KMGT"; | |
3256 | int64_t base; | |
3257 | int i; | |
3258 | ||
3259 | if (size <= 999) { | |
3260 | snprintf(buf, buf_size, "%" PRId64, size); | |
3261 | } else { | |
3262 | base = 1024; | |
3263 | for(i = 0; i < NB_SUFFIXES; i++) { | |
3264 | if (size < (10 * base)) { | |
5fafdf24 | 3265 | snprintf(buf, buf_size, "%0.1f%c", |
faea38e7 FB |
3266 | (double)size / base, |
3267 | suffixes[i]); | |
3268 | break; | |
3269 | } else if (size < (1000 * base) || i == (NB_SUFFIXES - 1)) { | |
5fafdf24 | 3270 | snprintf(buf, buf_size, "%" PRId64 "%c", |
faea38e7 FB |
3271 | ((size + (base >> 1)) / base), |
3272 | suffixes[i]); | |
3273 | break; | |
3274 | } | |
3275 | base = base * 1024; | |
3276 | } | |
3277 | } | |
3278 | return buf; | |
3279 | } | |
3280 | ||
3281 | char *bdrv_snapshot_dump(char *buf, int buf_size, QEMUSnapshotInfo *sn) | |
3282 | { | |
3283 | char buf1[128], date_buf[128], clock_buf[128]; | |
3b9f94e1 FB |
3284 | #ifdef _WIN32 |
3285 | struct tm *ptm; | |
3286 | #else | |
faea38e7 | 3287 | struct tm tm; |
3b9f94e1 | 3288 | #endif |
faea38e7 FB |
3289 | time_t ti; |
3290 | int64_t secs; | |
3291 | ||
3292 | if (!sn) { | |
5fafdf24 TS |
3293 | snprintf(buf, buf_size, |
3294 | "%-10s%-20s%7s%20s%15s", | |
faea38e7 FB |
3295 | "ID", "TAG", "VM SIZE", "DATE", "VM CLOCK"); |
3296 | } else { | |
3297 | ti = sn->date_sec; | |
3b9f94e1 FB |
3298 | #ifdef _WIN32 |
3299 | ptm = localtime(&ti); | |
3300 | strftime(date_buf, sizeof(date_buf), | |
3301 | "%Y-%m-%d %H:%M:%S", ptm); | |
3302 | #else | |
faea38e7 FB |
3303 | localtime_r(&ti, &tm); |
3304 | strftime(date_buf, sizeof(date_buf), | |
3305 | "%Y-%m-%d %H:%M:%S", &tm); | |
3b9f94e1 | 3306 | #endif |
faea38e7 FB |
3307 | secs = sn->vm_clock_nsec / 1000000000; |
3308 | snprintf(clock_buf, sizeof(clock_buf), | |
3309 | "%02d:%02d:%02d.%03d", | |
3310 | (int)(secs / 3600), | |
3311 | (int)((secs / 60) % 60), | |
5fafdf24 | 3312 | (int)(secs % 60), |
faea38e7 FB |
3313 | (int)((sn->vm_clock_nsec / 1000000) % 1000)); |
3314 | snprintf(buf, buf_size, | |
5fafdf24 | 3315 | "%-10s%-20s%7s%20s%15s", |
faea38e7 FB |
3316 | sn->id_str, sn->name, |
3317 | get_human_readable_size(buf1, sizeof(buf1), sn->vm_state_size), | |
3318 | date_buf, | |
3319 | clock_buf); | |
3320 | } | |
3321 | return buf; | |
3322 | } | |
3323 | ||
ea2384d3 | 3324 | /**************************************************************/ |
83f64091 | 3325 | /* async I/Os */ |
ea2384d3 | 3326 | |
3b69e4b9 | 3327 | BlockDriverAIOCB *bdrv_aio_readv(BlockDriverState *bs, int64_t sector_num, |
f141eafe | 3328 | QEMUIOVector *qiov, int nb_sectors, |
3b69e4b9 | 3329 | BlockDriverCompletionFunc *cb, void *opaque) |
83f64091 | 3330 | { |
bbf0a440 SH |
3331 | trace_bdrv_aio_readv(bs, sector_num, nb_sectors, opaque); |
3332 | ||
b2a61371 | 3333 | return bdrv_co_aio_rw_vector(bs, sector_num, qiov, nb_sectors, |
8c5873d6 | 3334 | cb, opaque, false); |
ea2384d3 FB |
3335 | } |
3336 | ||
f141eafe AL |
3337 | BlockDriverAIOCB *bdrv_aio_writev(BlockDriverState *bs, int64_t sector_num, |
3338 | QEMUIOVector *qiov, int nb_sectors, | |
3339 | BlockDriverCompletionFunc *cb, void *opaque) | |
ea2384d3 | 3340 | { |
bbf0a440 SH |
3341 | trace_bdrv_aio_writev(bs, sector_num, nb_sectors, opaque); |
3342 | ||
1a6e115b | 3343 | return bdrv_co_aio_rw_vector(bs, sector_num, qiov, nb_sectors, |
8c5873d6 | 3344 | cb, opaque, true); |
83f64091 FB |
3345 | } |
3346 | ||
40b4f539 KW |
3347 | |
3348 | typedef struct MultiwriteCB { | |
3349 | int error; | |
3350 | int num_requests; | |
3351 | int num_callbacks; | |
3352 | struct { | |
3353 | BlockDriverCompletionFunc *cb; | |
3354 | void *opaque; | |
3355 | QEMUIOVector *free_qiov; | |
40b4f539 KW |
3356 | } callbacks[]; |
3357 | } MultiwriteCB; | |
3358 | ||
3359 | static void multiwrite_user_cb(MultiwriteCB *mcb) | |
3360 | { | |
3361 | int i; | |
3362 | ||
3363 | for (i = 0; i < mcb->num_callbacks; i++) { | |
3364 | mcb->callbacks[i].cb(mcb->callbacks[i].opaque, mcb->error); | |
1e1ea48d SH |
3365 | if (mcb->callbacks[i].free_qiov) { |
3366 | qemu_iovec_destroy(mcb->callbacks[i].free_qiov); | |
3367 | } | |
7267c094 | 3368 | g_free(mcb->callbacks[i].free_qiov); |
40b4f539 KW |
3369 | } |
3370 | } | |
3371 | ||
3372 | static void multiwrite_cb(void *opaque, int ret) | |
3373 | { | |
3374 | MultiwriteCB *mcb = opaque; | |
3375 | ||
6d519a5f SH |
3376 | trace_multiwrite_cb(mcb, ret); |
3377 | ||
cb6d3ca0 | 3378 | if (ret < 0 && !mcb->error) { |
40b4f539 | 3379 | mcb->error = ret; |
40b4f539 KW |
3380 | } |
3381 | ||
3382 | mcb->num_requests--; | |
3383 | if (mcb->num_requests == 0) { | |
de189a1b | 3384 | multiwrite_user_cb(mcb); |
7267c094 | 3385 | g_free(mcb); |
40b4f539 KW |
3386 | } |
3387 | } | |
3388 | ||
3389 | static int multiwrite_req_compare(const void *a, const void *b) | |
3390 | { | |
77be4366 CH |
3391 | const BlockRequest *req1 = a, *req2 = b; |
3392 | ||
3393 | /* | |
3394 | * Note that we can't simply subtract req2->sector from req1->sector | |
3395 | * here as that could overflow the return value. | |
3396 | */ | |
3397 | if (req1->sector > req2->sector) { | |
3398 | return 1; | |
3399 | } else if (req1->sector < req2->sector) { | |
3400 | return -1; | |
3401 | } else { | |
3402 | return 0; | |
3403 | } | |
40b4f539 KW |
3404 | } |
3405 | ||
3406 | /* | |
3407 | * Takes a bunch of requests and tries to merge them. Returns the number of | |
3408 | * requests that remain after merging. | |
3409 | */ | |
3410 | static int multiwrite_merge(BlockDriverState *bs, BlockRequest *reqs, | |
3411 | int num_reqs, MultiwriteCB *mcb) | |
3412 | { | |
3413 | int i, outidx; | |
3414 | ||
3415 | // Sort requests by start sector | |
3416 | qsort(reqs, num_reqs, sizeof(*reqs), &multiwrite_req_compare); | |
3417 | ||
3418 | // Check if adjacent requests touch the same clusters. If so, combine them, | |
3419 | // filling up gaps with zero sectors. | |
3420 | outidx = 0; | |
3421 | for (i = 1; i < num_reqs; i++) { | |
3422 | int merge = 0; | |
3423 | int64_t oldreq_last = reqs[outidx].sector + reqs[outidx].nb_sectors; | |
3424 | ||
b6a127a1 | 3425 | // Handle exactly sequential writes and overlapping writes. |
40b4f539 KW |
3426 | if (reqs[i].sector <= oldreq_last) { |
3427 | merge = 1; | |
3428 | } | |
3429 | ||
e2a305fb CH |
3430 | if (reqs[outidx].qiov->niov + reqs[i].qiov->niov + 1 > IOV_MAX) { |
3431 | merge = 0; | |
3432 | } | |
3433 | ||
40b4f539 KW |
3434 | if (merge) { |
3435 | size_t size; | |
7267c094 | 3436 | QEMUIOVector *qiov = g_malloc0(sizeof(*qiov)); |
40b4f539 KW |
3437 | qemu_iovec_init(qiov, |
3438 | reqs[outidx].qiov->niov + reqs[i].qiov->niov + 1); | |
3439 | ||
3440 | // Add the first request to the merged one. If the requests are | |
3441 | // overlapping, drop the last sectors of the first request. | |
3442 | size = (reqs[i].sector - reqs[outidx].sector) << 9; | |
1b093c48 | 3443 | qemu_iovec_concat(qiov, reqs[outidx].qiov, 0, size); |
40b4f539 | 3444 | |
b6a127a1 PB |
3445 | // We should need to add any zeros between the two requests |
3446 | assert (reqs[i].sector <= oldreq_last); | |
40b4f539 KW |
3447 | |
3448 | // Add the second request | |
1b093c48 | 3449 | qemu_iovec_concat(qiov, reqs[i].qiov, 0, reqs[i].qiov->size); |
40b4f539 | 3450 | |
cbf1dff2 | 3451 | reqs[outidx].nb_sectors = qiov->size >> 9; |
40b4f539 KW |
3452 | reqs[outidx].qiov = qiov; |
3453 | ||
3454 | mcb->callbacks[i].free_qiov = reqs[outidx].qiov; | |
3455 | } else { | |
3456 | outidx++; | |
3457 | reqs[outidx].sector = reqs[i].sector; | |
3458 | reqs[outidx].nb_sectors = reqs[i].nb_sectors; | |
3459 | reqs[outidx].qiov = reqs[i].qiov; | |
3460 | } | |
3461 | } | |
3462 | ||
3463 | return outidx + 1; | |
3464 | } | |
3465 | ||
3466 | /* | |
3467 | * Submit multiple AIO write requests at once. | |
3468 | * | |
3469 | * On success, the function returns 0 and all requests in the reqs array have | |
3470 | * been submitted. In error case this function returns -1, and any of the | |
3471 | * requests may or may not be submitted yet. In particular, this means that the | |
3472 | * callback will be called for some of the requests, for others it won't. The | |
3473 | * caller must check the error field of the BlockRequest to wait for the right | |
3474 | * callbacks (if error != 0, no callback will be called). | |
3475 | * | |
3476 | * The implementation may modify the contents of the reqs array, e.g. to merge | |
3477 | * requests. However, the fields opaque and error are left unmodified as they | |
3478 | * are used to signal failure for a single request to the caller. | |
3479 | */ | |
3480 | int bdrv_aio_multiwrite(BlockDriverState *bs, BlockRequest *reqs, int num_reqs) | |
3481 | { | |
40b4f539 KW |
3482 | MultiwriteCB *mcb; |
3483 | int i; | |
3484 | ||
301db7c2 RH |
3485 | /* don't submit writes if we don't have a medium */ |
3486 | if (bs->drv == NULL) { | |
3487 | for (i = 0; i < num_reqs; i++) { | |
3488 | reqs[i].error = -ENOMEDIUM; | |
3489 | } | |
3490 | return -1; | |
3491 | } | |
3492 | ||
40b4f539 KW |
3493 | if (num_reqs == 0) { |
3494 | return 0; | |
3495 | } | |
3496 | ||
3497 | // Create MultiwriteCB structure | |
7267c094 | 3498 | mcb = g_malloc0(sizeof(*mcb) + num_reqs * sizeof(*mcb->callbacks)); |
40b4f539 KW |
3499 | mcb->num_requests = 0; |
3500 | mcb->num_callbacks = num_reqs; | |
3501 | ||
3502 | for (i = 0; i < num_reqs; i++) { | |
3503 | mcb->callbacks[i].cb = reqs[i].cb; | |
3504 | mcb->callbacks[i].opaque = reqs[i].opaque; | |
3505 | } | |
3506 | ||
3507 | // Check for mergable requests | |
3508 | num_reqs = multiwrite_merge(bs, reqs, num_reqs, mcb); | |
3509 | ||
6d519a5f SH |
3510 | trace_bdrv_aio_multiwrite(mcb, mcb->num_callbacks, num_reqs); |
3511 | ||
df9309fb PB |
3512 | /* Run the aio requests. */ |
3513 | mcb->num_requests = num_reqs; | |
40b4f539 | 3514 | for (i = 0; i < num_reqs; i++) { |
ad54ae80 | 3515 | bdrv_aio_writev(bs, reqs[i].sector, reqs[i].qiov, |
40b4f539 | 3516 | reqs[i].nb_sectors, multiwrite_cb, mcb); |
40b4f539 KW |
3517 | } |
3518 | ||
3519 | return 0; | |
40b4f539 KW |
3520 | } |
3521 | ||
83f64091 | 3522 | void bdrv_aio_cancel(BlockDriverAIOCB *acb) |
83f64091 | 3523 | { |
d7331bed | 3524 | acb->aiocb_info->cancel(acb); |
83f64091 FB |
3525 | } |
3526 | ||
98f90dba ZYW |
3527 | /* block I/O throttling */ |
3528 | static bool bdrv_exceed_bps_limits(BlockDriverState *bs, int nb_sectors, | |
3529 | bool is_write, double elapsed_time, uint64_t *wait) | |
3530 | { | |
3531 | uint64_t bps_limit = 0; | |
3532 | double bytes_limit, bytes_base, bytes_res; | |
3533 | double slice_time, wait_time; | |
3534 | ||
3535 | if (bs->io_limits.bps[BLOCK_IO_LIMIT_TOTAL]) { | |
3536 | bps_limit = bs->io_limits.bps[BLOCK_IO_LIMIT_TOTAL]; | |
3537 | } else if (bs->io_limits.bps[is_write]) { | |
3538 | bps_limit = bs->io_limits.bps[is_write]; | |
3539 | } else { | |
3540 | if (wait) { | |
3541 | *wait = 0; | |
3542 | } | |
3543 | ||
3544 | return false; | |
3545 | } | |
3546 | ||
3547 | slice_time = bs->slice_end - bs->slice_start; | |
3548 | slice_time /= (NANOSECONDS_PER_SECOND); | |
3549 | bytes_limit = bps_limit * slice_time; | |
3550 | bytes_base = bs->nr_bytes[is_write] - bs->io_base.bytes[is_write]; | |
3551 | if (bs->io_limits.bps[BLOCK_IO_LIMIT_TOTAL]) { | |
3552 | bytes_base += bs->nr_bytes[!is_write] - bs->io_base.bytes[!is_write]; | |
3553 | } | |
3554 | ||
3555 | /* bytes_base: the bytes of data which have been read/written; and | |
3556 | * it is obtained from the history statistic info. | |
3557 | * bytes_res: the remaining bytes of data which need to be read/written. | |
3558 | * (bytes_base + bytes_res) / bps_limit: used to calcuate | |
3559 | * the total time for completing reading/writting all data. | |
3560 | */ | |
3561 | bytes_res = (unsigned) nb_sectors * BDRV_SECTOR_SIZE; | |
3562 | ||
3563 | if (bytes_base + bytes_res <= bytes_limit) { | |
3564 | if (wait) { | |
3565 | *wait = 0; | |
3566 | } | |
3567 | ||
3568 | return false; | |
3569 | } | |
3570 | ||
3571 | /* Calc approx time to dispatch */ | |
3572 | wait_time = (bytes_base + bytes_res) / bps_limit - elapsed_time; | |
3573 | ||
3574 | /* When the I/O rate at runtime exceeds the limits, | |
3575 | * bs->slice_end need to be extended in order that the current statistic | |
3576 | * info can be kept until the timer fire, so it is increased and tuned | |
3577 | * based on the result of experiment. | |
3578 | */ | |
3579 | bs->slice_time = wait_time * BLOCK_IO_SLICE_TIME * 10; | |
3580 | bs->slice_end += bs->slice_time - 3 * BLOCK_IO_SLICE_TIME; | |
3581 | if (wait) { | |
3582 | *wait = wait_time * BLOCK_IO_SLICE_TIME * 10; | |
3583 | } | |
3584 | ||
3585 | return true; | |
3586 | } | |
3587 | ||
3588 | static bool bdrv_exceed_iops_limits(BlockDriverState *bs, bool is_write, | |
3589 | double elapsed_time, uint64_t *wait) | |
3590 | { | |
3591 | uint64_t iops_limit = 0; | |
3592 | double ios_limit, ios_base; | |
3593 | double slice_time, wait_time; | |
3594 | ||
3595 | if (bs->io_limits.iops[BLOCK_IO_LIMIT_TOTAL]) { | |
3596 | iops_limit = bs->io_limits.iops[BLOCK_IO_LIMIT_TOTAL]; | |
3597 | } else if (bs->io_limits.iops[is_write]) { | |
3598 | iops_limit = bs->io_limits.iops[is_write]; | |
3599 | } else { | |
3600 | if (wait) { | |
3601 | *wait = 0; | |
3602 | } | |
3603 | ||
3604 | return false; | |
3605 | } | |
3606 | ||
3607 | slice_time = bs->slice_end - bs->slice_start; | |
3608 | slice_time /= (NANOSECONDS_PER_SECOND); | |
3609 | ios_limit = iops_limit * slice_time; | |
3610 | ios_base = bs->nr_ops[is_write] - bs->io_base.ios[is_write]; | |
3611 | if (bs->io_limits.iops[BLOCK_IO_LIMIT_TOTAL]) { | |
3612 | ios_base += bs->nr_ops[!is_write] - bs->io_base.ios[!is_write]; | |
3613 | } | |
3614 | ||
3615 | if (ios_base + 1 <= ios_limit) { | |
3616 | if (wait) { | |
3617 | *wait = 0; | |
3618 | } | |
3619 | ||
3620 | return false; | |
3621 | } | |
3622 | ||
3623 | /* Calc approx time to dispatch */ | |
3624 | wait_time = (ios_base + 1) / iops_limit; | |
3625 | if (wait_time > elapsed_time) { | |
3626 | wait_time = wait_time - elapsed_time; | |
3627 | } else { | |
3628 | wait_time = 0; | |
3629 | } | |
3630 | ||
3631 | bs->slice_time = wait_time * BLOCK_IO_SLICE_TIME * 10; | |
3632 | bs->slice_end += bs->slice_time - 3 * BLOCK_IO_SLICE_TIME; | |
3633 | if (wait) { | |
3634 | *wait = wait_time * BLOCK_IO_SLICE_TIME * 10; | |
3635 | } | |
3636 | ||
3637 | return true; | |
3638 | } | |
3639 | ||
3640 | static bool bdrv_exceed_io_limits(BlockDriverState *bs, int nb_sectors, | |
3641 | bool is_write, int64_t *wait) | |
3642 | { | |
3643 | int64_t now, max_wait; | |
3644 | uint64_t bps_wait = 0, iops_wait = 0; | |
3645 | double elapsed_time; | |
3646 | int bps_ret, iops_ret; | |
3647 | ||
3648 | now = qemu_get_clock_ns(vm_clock); | |
3649 | if ((bs->slice_start < now) | |
3650 | && (bs->slice_end > now)) { | |
3651 | bs->slice_end = now + bs->slice_time; | |
3652 | } else { | |
3653 | bs->slice_time = 5 * BLOCK_IO_SLICE_TIME; | |
3654 | bs->slice_start = now; | |
3655 | bs->slice_end = now + bs->slice_time; | |
3656 | ||
3657 | bs->io_base.bytes[is_write] = bs->nr_bytes[is_write]; | |
3658 | bs->io_base.bytes[!is_write] = bs->nr_bytes[!is_write]; | |
3659 | ||
3660 | bs->io_base.ios[is_write] = bs->nr_ops[is_write]; | |
3661 | bs->io_base.ios[!is_write] = bs->nr_ops[!is_write]; | |
3662 | } | |
3663 | ||
3664 | elapsed_time = now - bs->slice_start; | |
3665 | elapsed_time /= (NANOSECONDS_PER_SECOND); | |
3666 | ||
3667 | bps_ret = bdrv_exceed_bps_limits(bs, nb_sectors, | |
3668 | is_write, elapsed_time, &bps_wait); | |
3669 | iops_ret = bdrv_exceed_iops_limits(bs, is_write, | |
3670 | elapsed_time, &iops_wait); | |
3671 | if (bps_ret || iops_ret) { | |
3672 | max_wait = bps_wait > iops_wait ? bps_wait : iops_wait; | |
3673 | if (wait) { | |
3674 | *wait = max_wait; | |
3675 | } | |
3676 | ||
3677 | now = qemu_get_clock_ns(vm_clock); | |
3678 | if (bs->slice_end < now + max_wait) { | |
3679 | bs->slice_end = now + max_wait; | |
3680 | } | |
3681 | ||
3682 | return true; | |
3683 | } | |
3684 | ||
3685 | if (wait) { | |
3686 | *wait = 0; | |
3687 | } | |
3688 | ||
3689 | return false; | |
3690 | } | |
ce1a14dc | 3691 | |
83f64091 FB |
3692 | /**************************************************************/ |
3693 | /* async block device emulation */ | |
3694 | ||
c16b5a2c CH |
3695 | typedef struct BlockDriverAIOCBSync { |
3696 | BlockDriverAIOCB common; | |
3697 | QEMUBH *bh; | |
3698 | int ret; | |
3699 | /* vector translation state */ | |
3700 | QEMUIOVector *qiov; | |
3701 | uint8_t *bounce; | |
3702 | int is_write; | |
3703 | } BlockDriverAIOCBSync; | |
3704 | ||
3705 | static void bdrv_aio_cancel_em(BlockDriverAIOCB *blockacb) | |
3706 | { | |
b666d239 KW |
3707 | BlockDriverAIOCBSync *acb = |
3708 | container_of(blockacb, BlockDriverAIOCBSync, common); | |
6a7ad299 | 3709 | qemu_bh_delete(acb->bh); |
36afc451 | 3710 | acb->bh = NULL; |
c16b5a2c CH |
3711 | qemu_aio_release(acb); |
3712 | } | |
3713 | ||
d7331bed | 3714 | static const AIOCBInfo bdrv_em_aiocb_info = { |
c16b5a2c CH |
3715 | .aiocb_size = sizeof(BlockDriverAIOCBSync), |
3716 | .cancel = bdrv_aio_cancel_em, | |
3717 | }; | |
3718 | ||
ce1a14dc | 3719 | static void bdrv_aio_bh_cb(void *opaque) |
83f64091 | 3720 | { |
ce1a14dc | 3721 | BlockDriverAIOCBSync *acb = opaque; |
f141eafe | 3722 | |
f141eafe | 3723 | if (!acb->is_write) |
03396148 | 3724 | qemu_iovec_from_buf(acb->qiov, 0, acb->bounce, acb->qiov->size); |
ceb42de8 | 3725 | qemu_vfree(acb->bounce); |
ce1a14dc | 3726 | acb->common.cb(acb->common.opaque, acb->ret); |
6a7ad299 | 3727 | qemu_bh_delete(acb->bh); |
36afc451 | 3728 | acb->bh = NULL; |
ce1a14dc | 3729 | qemu_aio_release(acb); |
83f64091 | 3730 | } |
beac80cd | 3731 | |
f141eafe AL |
3732 | static BlockDriverAIOCB *bdrv_aio_rw_vector(BlockDriverState *bs, |
3733 | int64_t sector_num, | |
3734 | QEMUIOVector *qiov, | |
3735 | int nb_sectors, | |
3736 | BlockDriverCompletionFunc *cb, | |
3737 | void *opaque, | |
3738 | int is_write) | |
3739 | ||
83f64091 | 3740 | { |
ce1a14dc | 3741 | BlockDriverAIOCBSync *acb; |
ce1a14dc | 3742 | |
d7331bed | 3743 | acb = qemu_aio_get(&bdrv_em_aiocb_info, bs, cb, opaque); |
f141eafe AL |
3744 | acb->is_write = is_write; |
3745 | acb->qiov = qiov; | |
e268ca52 | 3746 | acb->bounce = qemu_blockalign(bs, qiov->size); |
3f3aace8 | 3747 | acb->bh = qemu_bh_new(bdrv_aio_bh_cb, acb); |
f141eafe AL |
3748 | |
3749 | if (is_write) { | |
d5e6b161 | 3750 | qemu_iovec_to_buf(acb->qiov, 0, acb->bounce, qiov->size); |
1ed20acf | 3751 | acb->ret = bs->drv->bdrv_write(bs, sector_num, acb->bounce, nb_sectors); |
f141eafe | 3752 | } else { |
1ed20acf | 3753 | acb->ret = bs->drv->bdrv_read(bs, sector_num, acb->bounce, nb_sectors); |
f141eafe AL |
3754 | } |
3755 | ||
ce1a14dc | 3756 | qemu_bh_schedule(acb->bh); |
f141eafe | 3757 | |
ce1a14dc | 3758 | return &acb->common; |
beac80cd FB |
3759 | } |
3760 | ||
f141eafe AL |
3761 | static BlockDriverAIOCB *bdrv_aio_readv_em(BlockDriverState *bs, |
3762 | int64_t sector_num, QEMUIOVector *qiov, int nb_sectors, | |
ce1a14dc | 3763 | BlockDriverCompletionFunc *cb, void *opaque) |
beac80cd | 3764 | { |
f141eafe AL |
3765 | return bdrv_aio_rw_vector(bs, sector_num, qiov, nb_sectors, cb, opaque, 0); |
3766 | } | |
83f64091 | 3767 | |
f141eafe AL |
3768 | static BlockDriverAIOCB *bdrv_aio_writev_em(BlockDriverState *bs, |
3769 | int64_t sector_num, QEMUIOVector *qiov, int nb_sectors, | |
3770 | BlockDriverCompletionFunc *cb, void *opaque) | |
3771 | { | |
3772 | return bdrv_aio_rw_vector(bs, sector_num, qiov, nb_sectors, cb, opaque, 1); | |
beac80cd | 3773 | } |
beac80cd | 3774 | |
68485420 KW |
3775 | |
3776 | typedef struct BlockDriverAIOCBCoroutine { | |
3777 | BlockDriverAIOCB common; | |
3778 | BlockRequest req; | |
3779 | bool is_write; | |
3780 | QEMUBH* bh; | |
3781 | } BlockDriverAIOCBCoroutine; | |
3782 | ||
3783 | static void bdrv_aio_co_cancel_em(BlockDriverAIOCB *blockacb) | |
3784 | { | |
3785 | qemu_aio_flush(); | |
3786 | } | |
3787 | ||
d7331bed | 3788 | static const AIOCBInfo bdrv_em_co_aiocb_info = { |
68485420 KW |
3789 | .aiocb_size = sizeof(BlockDriverAIOCBCoroutine), |
3790 | .cancel = bdrv_aio_co_cancel_em, | |
3791 | }; | |
3792 | ||
35246a68 | 3793 | static void bdrv_co_em_bh(void *opaque) |
68485420 KW |
3794 | { |
3795 | BlockDriverAIOCBCoroutine *acb = opaque; | |
3796 | ||
3797 | acb->common.cb(acb->common.opaque, acb->req.error); | |
3798 | qemu_bh_delete(acb->bh); | |
3799 | qemu_aio_release(acb); | |
3800 | } | |
3801 | ||
b2a61371 SH |
3802 | /* Invoke bdrv_co_do_readv/bdrv_co_do_writev */ |
3803 | static void coroutine_fn bdrv_co_do_rw(void *opaque) | |
3804 | { | |
3805 | BlockDriverAIOCBCoroutine *acb = opaque; | |
3806 | BlockDriverState *bs = acb->common.bs; | |
3807 | ||
3808 | if (!acb->is_write) { | |
3809 | acb->req.error = bdrv_co_do_readv(bs, acb->req.sector, | |
470c0504 | 3810 | acb->req.nb_sectors, acb->req.qiov, 0); |
b2a61371 SH |
3811 | } else { |
3812 | acb->req.error = bdrv_co_do_writev(bs, acb->req.sector, | |
f08f2dda | 3813 | acb->req.nb_sectors, acb->req.qiov, 0); |
b2a61371 SH |
3814 | } |
3815 | ||
35246a68 | 3816 | acb->bh = qemu_bh_new(bdrv_co_em_bh, acb); |
b2a61371 SH |
3817 | qemu_bh_schedule(acb->bh); |
3818 | } | |
3819 | ||
68485420 KW |
3820 | static BlockDriverAIOCB *bdrv_co_aio_rw_vector(BlockDriverState *bs, |
3821 | int64_t sector_num, | |
3822 | QEMUIOVector *qiov, | |
3823 | int nb_sectors, | |
3824 | BlockDriverCompletionFunc *cb, | |
3825 | void *opaque, | |
8c5873d6 | 3826 | bool is_write) |
68485420 KW |
3827 | { |
3828 | Coroutine *co; | |
3829 | BlockDriverAIOCBCoroutine *acb; | |
3830 | ||
d7331bed | 3831 | acb = qemu_aio_get(&bdrv_em_co_aiocb_info, bs, cb, opaque); |
68485420 KW |
3832 | acb->req.sector = sector_num; |
3833 | acb->req.nb_sectors = nb_sectors; | |
3834 | acb->req.qiov = qiov; | |
3835 | acb->is_write = is_write; | |
3836 | ||
8c5873d6 | 3837 | co = qemu_coroutine_create(bdrv_co_do_rw); |
68485420 KW |
3838 | qemu_coroutine_enter(co, acb); |
3839 | ||
3840 | return &acb->common; | |
3841 | } | |
3842 | ||
07f07615 | 3843 | static void coroutine_fn bdrv_aio_flush_co_entry(void *opaque) |
b2e12bc6 | 3844 | { |
07f07615 PB |
3845 | BlockDriverAIOCBCoroutine *acb = opaque; |
3846 | BlockDriverState *bs = acb->common.bs; | |
b2e12bc6 | 3847 | |
07f07615 PB |
3848 | acb->req.error = bdrv_co_flush(bs); |
3849 | acb->bh = qemu_bh_new(bdrv_co_em_bh, acb); | |
b2e12bc6 | 3850 | qemu_bh_schedule(acb->bh); |
b2e12bc6 CH |
3851 | } |
3852 | ||
07f07615 | 3853 | BlockDriverAIOCB *bdrv_aio_flush(BlockDriverState *bs, |
016f5cf6 AG |
3854 | BlockDriverCompletionFunc *cb, void *opaque) |
3855 | { | |
07f07615 | 3856 | trace_bdrv_aio_flush(bs, opaque); |
016f5cf6 | 3857 | |
07f07615 PB |
3858 | Coroutine *co; |
3859 | BlockDriverAIOCBCoroutine *acb; | |
016f5cf6 | 3860 | |
d7331bed | 3861 | acb = qemu_aio_get(&bdrv_em_co_aiocb_info, bs, cb, opaque); |
07f07615 PB |
3862 | co = qemu_coroutine_create(bdrv_aio_flush_co_entry); |
3863 | qemu_coroutine_enter(co, acb); | |
016f5cf6 | 3864 | |
016f5cf6 AG |
3865 | return &acb->common; |
3866 | } | |
3867 | ||
4265d620 PB |
3868 | static void coroutine_fn bdrv_aio_discard_co_entry(void *opaque) |
3869 | { | |
3870 | BlockDriverAIOCBCoroutine *acb = opaque; | |
3871 | BlockDriverState *bs = acb->common.bs; | |
3872 | ||
3873 | acb->req.error = bdrv_co_discard(bs, acb->req.sector, acb->req.nb_sectors); | |
3874 | acb->bh = qemu_bh_new(bdrv_co_em_bh, acb); | |
3875 | qemu_bh_schedule(acb->bh); | |
3876 | } | |
3877 | ||
3878 | BlockDriverAIOCB *bdrv_aio_discard(BlockDriverState *bs, | |
3879 | int64_t sector_num, int nb_sectors, | |
3880 | BlockDriverCompletionFunc *cb, void *opaque) | |
3881 | { | |
3882 | Coroutine *co; | |
3883 | BlockDriverAIOCBCoroutine *acb; | |
3884 | ||
3885 | trace_bdrv_aio_discard(bs, sector_num, nb_sectors, opaque); | |
3886 | ||
d7331bed | 3887 | acb = qemu_aio_get(&bdrv_em_co_aiocb_info, bs, cb, opaque); |
4265d620 PB |
3888 | acb->req.sector = sector_num; |
3889 | acb->req.nb_sectors = nb_sectors; | |
3890 | co = qemu_coroutine_create(bdrv_aio_discard_co_entry); | |
3891 | qemu_coroutine_enter(co, acb); | |
3892 | ||
3893 | return &acb->common; | |
3894 | } | |
3895 | ||
ea2384d3 FB |
3896 | void bdrv_init(void) |
3897 | { | |
5efa9d5a | 3898 | module_call_init(MODULE_INIT_BLOCK); |
ea2384d3 | 3899 | } |
ce1a14dc | 3900 | |
eb852011 MA |
3901 | void bdrv_init_with_whitelist(void) |
3902 | { | |
3903 | use_bdrv_whitelist = 1; | |
3904 | bdrv_init(); | |
3905 | } | |
3906 | ||
d7331bed | 3907 | void *qemu_aio_get(const AIOCBInfo *aiocb_info, BlockDriverState *bs, |
c16b5a2c | 3908 | BlockDriverCompletionFunc *cb, void *opaque) |
ce1a14dc | 3909 | { |
ce1a14dc PB |
3910 | BlockDriverAIOCB *acb; |
3911 | ||
d7331bed SH |
3912 | acb = g_slice_alloc(aiocb_info->aiocb_size); |
3913 | acb->aiocb_info = aiocb_info; | |
ce1a14dc PB |
3914 | acb->bs = bs; |
3915 | acb->cb = cb; | |
3916 | acb->opaque = opaque; | |
3917 | return acb; | |
3918 | } | |
3919 | ||
3920 | void qemu_aio_release(void *p) | |
3921 | { | |
d37c975f | 3922 | BlockDriverAIOCB *acb = p; |
d7331bed | 3923 | g_slice_free1(acb->aiocb_info->aiocb_size, acb); |
ce1a14dc | 3924 | } |
19cb3738 | 3925 | |
f9f05dc5 KW |
3926 | /**************************************************************/ |
3927 | /* Coroutine block device emulation */ | |
3928 | ||
3929 | typedef struct CoroutineIOCompletion { | |
3930 | Coroutine *coroutine; | |
3931 | int ret; | |
3932 | } CoroutineIOCompletion; | |
3933 | ||
3934 | static void bdrv_co_io_em_complete(void *opaque, int ret) | |
3935 | { | |
3936 | CoroutineIOCompletion *co = opaque; | |
3937 | ||
3938 | co->ret = ret; | |
3939 | qemu_coroutine_enter(co->coroutine, NULL); | |
3940 | } | |
3941 | ||
3942 | static int coroutine_fn bdrv_co_io_em(BlockDriverState *bs, int64_t sector_num, | |
3943 | int nb_sectors, QEMUIOVector *iov, | |
3944 | bool is_write) | |
3945 | { | |
3946 | CoroutineIOCompletion co = { | |
3947 | .coroutine = qemu_coroutine_self(), | |
3948 | }; | |
3949 | BlockDriverAIOCB *acb; | |
3950 | ||
3951 | if (is_write) { | |
a652d160 SH |
3952 | acb = bs->drv->bdrv_aio_writev(bs, sector_num, iov, nb_sectors, |
3953 | bdrv_co_io_em_complete, &co); | |
f9f05dc5 | 3954 | } else { |
a652d160 SH |
3955 | acb = bs->drv->bdrv_aio_readv(bs, sector_num, iov, nb_sectors, |
3956 | bdrv_co_io_em_complete, &co); | |
f9f05dc5 KW |
3957 | } |
3958 | ||
59370aaa | 3959 | trace_bdrv_co_io_em(bs, sector_num, nb_sectors, is_write, acb); |
f9f05dc5 KW |
3960 | if (!acb) { |
3961 | return -EIO; | |
3962 | } | |
3963 | qemu_coroutine_yield(); | |
3964 | ||
3965 | return co.ret; | |
3966 | } | |
3967 | ||
3968 | static int coroutine_fn bdrv_co_readv_em(BlockDriverState *bs, | |
3969 | int64_t sector_num, int nb_sectors, | |
3970 | QEMUIOVector *iov) | |
3971 | { | |
3972 | return bdrv_co_io_em(bs, sector_num, nb_sectors, iov, false); | |
3973 | } | |
3974 | ||
3975 | static int coroutine_fn bdrv_co_writev_em(BlockDriverState *bs, | |
3976 | int64_t sector_num, int nb_sectors, | |
3977 | QEMUIOVector *iov) | |
3978 | { | |
3979 | return bdrv_co_io_em(bs, sector_num, nb_sectors, iov, true); | |
3980 | } | |
3981 | ||
07f07615 | 3982 | static void coroutine_fn bdrv_flush_co_entry(void *opaque) |
e7a8a783 | 3983 | { |
07f07615 PB |
3984 | RwCo *rwco = opaque; |
3985 | ||
3986 | rwco->ret = bdrv_co_flush(rwco->bs); | |
3987 | } | |
3988 | ||
3989 | int coroutine_fn bdrv_co_flush(BlockDriverState *bs) | |
3990 | { | |
eb489bb1 KW |
3991 | int ret; |
3992 | ||
29cdb251 | 3993 | if (!bs || !bdrv_is_inserted(bs) || bdrv_is_read_only(bs)) { |
07f07615 | 3994 | return 0; |
eb489bb1 KW |
3995 | } |
3996 | ||
ca716364 | 3997 | /* Write back cached data to the OS even with cache=unsafe */ |
eb489bb1 KW |
3998 | if (bs->drv->bdrv_co_flush_to_os) { |
3999 | ret = bs->drv->bdrv_co_flush_to_os(bs); | |
4000 | if (ret < 0) { | |
4001 | return ret; | |
4002 | } | |
4003 | } | |
4004 | ||
ca716364 KW |
4005 | /* But don't actually force it to the disk with cache=unsafe */ |
4006 | if (bs->open_flags & BDRV_O_NO_FLUSH) { | |
d4c82329 | 4007 | goto flush_parent; |
ca716364 KW |
4008 | } |
4009 | ||
eb489bb1 | 4010 | if (bs->drv->bdrv_co_flush_to_disk) { |
29cdb251 | 4011 | ret = bs->drv->bdrv_co_flush_to_disk(bs); |
07f07615 PB |
4012 | } else if (bs->drv->bdrv_aio_flush) { |
4013 | BlockDriverAIOCB *acb; | |
4014 | CoroutineIOCompletion co = { | |
4015 | .coroutine = qemu_coroutine_self(), | |
4016 | }; | |
4017 | ||
4018 | acb = bs->drv->bdrv_aio_flush(bs, bdrv_co_io_em_complete, &co); | |
4019 | if (acb == NULL) { | |
29cdb251 | 4020 | ret = -EIO; |
07f07615 PB |
4021 | } else { |
4022 | qemu_coroutine_yield(); | |
29cdb251 | 4023 | ret = co.ret; |
07f07615 | 4024 | } |
07f07615 PB |
4025 | } else { |
4026 | /* | |
4027 | * Some block drivers always operate in either writethrough or unsafe | |
4028 | * mode and don't support bdrv_flush therefore. Usually qemu doesn't | |
4029 | * know how the server works (because the behaviour is hardcoded or | |
4030 | * depends on server-side configuration), so we can't ensure that | |
4031 | * everything is safe on disk. Returning an error doesn't work because | |
4032 | * that would break guests even if the server operates in writethrough | |
4033 | * mode. | |
4034 | * | |
4035 | * Let's hope the user knows what he's doing. | |
4036 | */ | |
29cdb251 | 4037 | ret = 0; |
07f07615 | 4038 | } |
29cdb251 PB |
4039 | if (ret < 0) { |
4040 | return ret; | |
4041 | } | |
4042 | ||
4043 | /* Now flush the underlying protocol. It will also have BDRV_O_NO_FLUSH | |
4044 | * in the case of cache=unsafe, so there are no useless flushes. | |
4045 | */ | |
d4c82329 | 4046 | flush_parent: |
29cdb251 | 4047 | return bdrv_co_flush(bs->file); |
07f07615 PB |
4048 | } |
4049 | ||
0f15423c AL |
4050 | void bdrv_invalidate_cache(BlockDriverState *bs) |
4051 | { | |
4052 | if (bs->drv && bs->drv->bdrv_invalidate_cache) { | |
4053 | bs->drv->bdrv_invalidate_cache(bs); | |
4054 | } | |
4055 | } | |
4056 | ||
4057 | void bdrv_invalidate_cache_all(void) | |
4058 | { | |
4059 | BlockDriverState *bs; | |
4060 | ||
4061 | QTAILQ_FOREACH(bs, &bdrv_states, list) { | |
4062 | bdrv_invalidate_cache(bs); | |
4063 | } | |
4064 | } | |
4065 | ||
07789269 BC |
4066 | void bdrv_clear_incoming_migration_all(void) |
4067 | { | |
4068 | BlockDriverState *bs; | |
4069 | ||
4070 | QTAILQ_FOREACH(bs, &bdrv_states, list) { | |
4071 | bs->open_flags = bs->open_flags & ~(BDRV_O_INCOMING); | |
4072 | } | |
4073 | } | |
4074 | ||
07f07615 PB |
4075 | int bdrv_flush(BlockDriverState *bs) |
4076 | { | |
4077 | Coroutine *co; | |
4078 | RwCo rwco = { | |
4079 | .bs = bs, | |
4080 | .ret = NOT_DONE, | |
e7a8a783 | 4081 | }; |
e7a8a783 | 4082 | |
07f07615 PB |
4083 | if (qemu_in_coroutine()) { |
4084 | /* Fast-path if already in coroutine context */ | |
4085 | bdrv_flush_co_entry(&rwco); | |
4086 | } else { | |
4087 | co = qemu_coroutine_create(bdrv_flush_co_entry); | |
4088 | qemu_coroutine_enter(co, &rwco); | |
4089 | while (rwco.ret == NOT_DONE) { | |
4090 | qemu_aio_wait(); | |
4091 | } | |
e7a8a783 | 4092 | } |
07f07615 PB |
4093 | |
4094 | return rwco.ret; | |
e7a8a783 KW |
4095 | } |
4096 | ||
4265d620 PB |
4097 | static void coroutine_fn bdrv_discard_co_entry(void *opaque) |
4098 | { | |
4099 | RwCo *rwco = opaque; | |
4100 | ||
4101 | rwco->ret = bdrv_co_discard(rwco->bs, rwco->sector_num, rwco->nb_sectors); | |
4102 | } | |
4103 | ||
4104 | int coroutine_fn bdrv_co_discard(BlockDriverState *bs, int64_t sector_num, | |
4105 | int nb_sectors) | |
4106 | { | |
4107 | if (!bs->drv) { | |
4108 | return -ENOMEDIUM; | |
4109 | } else if (bdrv_check_request(bs, sector_num, nb_sectors)) { | |
4110 | return -EIO; | |
4111 | } else if (bs->read_only) { | |
4112 | return -EROFS; | |
4113 | } else if (bs->drv->bdrv_co_discard) { | |
4114 | return bs->drv->bdrv_co_discard(bs, sector_num, nb_sectors); | |
4115 | } else if (bs->drv->bdrv_aio_discard) { | |
4116 | BlockDriverAIOCB *acb; | |
4117 | CoroutineIOCompletion co = { | |
4118 | .coroutine = qemu_coroutine_self(), | |
4119 | }; | |
4120 | ||
4121 | acb = bs->drv->bdrv_aio_discard(bs, sector_num, nb_sectors, | |
4122 | bdrv_co_io_em_complete, &co); | |
4123 | if (acb == NULL) { | |
4124 | return -EIO; | |
4125 | } else { | |
4126 | qemu_coroutine_yield(); | |
4127 | return co.ret; | |
4128 | } | |
4265d620 PB |
4129 | } else { |
4130 | return 0; | |
4131 | } | |
4132 | } | |
4133 | ||
4134 | int bdrv_discard(BlockDriverState *bs, int64_t sector_num, int nb_sectors) | |
4135 | { | |
4136 | Coroutine *co; | |
4137 | RwCo rwco = { | |
4138 | .bs = bs, | |
4139 | .sector_num = sector_num, | |
4140 | .nb_sectors = nb_sectors, | |
4141 | .ret = NOT_DONE, | |
4142 | }; | |
4143 | ||
4144 | if (qemu_in_coroutine()) { | |
4145 | /* Fast-path if already in coroutine context */ | |
4146 | bdrv_discard_co_entry(&rwco); | |
4147 | } else { | |
4148 | co = qemu_coroutine_create(bdrv_discard_co_entry); | |
4149 | qemu_coroutine_enter(co, &rwco); | |
4150 | while (rwco.ret == NOT_DONE) { | |
4151 | qemu_aio_wait(); | |
4152 | } | |
4153 | } | |
4154 | ||
4155 | return rwco.ret; | |
4156 | } | |
4157 | ||
19cb3738 FB |
4158 | /**************************************************************/ |
4159 | /* removable device support */ | |
4160 | ||
4161 | /** | |
4162 | * Return TRUE if the media is present | |
4163 | */ | |
4164 | int bdrv_is_inserted(BlockDriverState *bs) | |
4165 | { | |
4166 | BlockDriver *drv = bs->drv; | |
a1aff5bf | 4167 | |
19cb3738 FB |
4168 | if (!drv) |
4169 | return 0; | |
4170 | if (!drv->bdrv_is_inserted) | |
a1aff5bf MA |
4171 | return 1; |
4172 | return drv->bdrv_is_inserted(bs); | |
19cb3738 FB |
4173 | } |
4174 | ||
4175 | /** | |
8e49ca46 MA |
4176 | * Return whether the media changed since the last call to this |
4177 | * function, or -ENOTSUP if we don't know. Most drivers don't know. | |
19cb3738 FB |
4178 | */ |
4179 | int bdrv_media_changed(BlockDriverState *bs) | |
4180 | { | |
4181 | BlockDriver *drv = bs->drv; | |
19cb3738 | 4182 | |
8e49ca46 MA |
4183 | if (drv && drv->bdrv_media_changed) { |
4184 | return drv->bdrv_media_changed(bs); | |
4185 | } | |
4186 | return -ENOTSUP; | |
19cb3738 FB |
4187 | } |
4188 | ||
4189 | /** | |
4190 | * If eject_flag is TRUE, eject the media. Otherwise, close the tray | |
4191 | */ | |
f36f3949 | 4192 | void bdrv_eject(BlockDriverState *bs, bool eject_flag) |
19cb3738 FB |
4193 | { |
4194 | BlockDriver *drv = bs->drv; | |
19cb3738 | 4195 | |
822e1cd1 MA |
4196 | if (drv && drv->bdrv_eject) { |
4197 | drv->bdrv_eject(bs, eject_flag); | |
19cb3738 | 4198 | } |
6f382ed2 LC |
4199 | |
4200 | if (bs->device_name[0] != '\0') { | |
4201 | bdrv_emit_qmp_eject_event(bs, eject_flag); | |
4202 | } | |
19cb3738 FB |
4203 | } |
4204 | ||
19cb3738 FB |
4205 | /** |
4206 | * Lock or unlock the media (if it is locked, the user won't be able | |
4207 | * to eject it manually). | |
4208 | */ | |
025e849a | 4209 | void bdrv_lock_medium(BlockDriverState *bs, bool locked) |
19cb3738 FB |
4210 | { |
4211 | BlockDriver *drv = bs->drv; | |
4212 | ||
025e849a | 4213 | trace_bdrv_lock_medium(bs, locked); |
b8c6d095 | 4214 | |
025e849a MA |
4215 | if (drv && drv->bdrv_lock_medium) { |
4216 | drv->bdrv_lock_medium(bs, locked); | |
19cb3738 FB |
4217 | } |
4218 | } | |
985a03b0 TS |
4219 | |
4220 | /* needed for generic scsi interface */ | |
4221 | ||
4222 | int bdrv_ioctl(BlockDriverState *bs, unsigned long int req, void *buf) | |
4223 | { | |
4224 | BlockDriver *drv = bs->drv; | |
4225 | ||
4226 | if (drv && drv->bdrv_ioctl) | |
4227 | return drv->bdrv_ioctl(bs, req, buf); | |
4228 | return -ENOTSUP; | |
4229 | } | |
7d780669 | 4230 | |
221f715d AL |
4231 | BlockDriverAIOCB *bdrv_aio_ioctl(BlockDriverState *bs, |
4232 | unsigned long int req, void *buf, | |
4233 | BlockDriverCompletionFunc *cb, void *opaque) | |
7d780669 | 4234 | { |
221f715d | 4235 | BlockDriver *drv = bs->drv; |
7d780669 | 4236 | |
221f715d AL |
4237 | if (drv && drv->bdrv_aio_ioctl) |
4238 | return drv->bdrv_aio_ioctl(bs, req, buf, cb, opaque); | |
4239 | return NULL; | |
7d780669 | 4240 | } |
e268ca52 | 4241 | |
7b6f9300 MA |
4242 | void bdrv_set_buffer_alignment(BlockDriverState *bs, int align) |
4243 | { | |
4244 | bs->buffer_alignment = align; | |
4245 | } | |
7cd1e32a | 4246 | |
e268ca52 AL |
4247 | void *qemu_blockalign(BlockDriverState *bs, size_t size) |
4248 | { | |
4249 | return qemu_memalign((bs && bs->buffer_alignment) ? bs->buffer_alignment : 512, size); | |
4250 | } | |
7cd1e32a LS |
4251 | |
4252 | void bdrv_set_dirty_tracking(BlockDriverState *bs, int enable) | |
4253 | { | |
4254 | int64_t bitmap_size; | |
a55eb92c | 4255 | |
aaa0eb75 | 4256 | bs->dirty_count = 0; |
a55eb92c | 4257 | if (enable) { |
c6d22830 JK |
4258 | if (!bs->dirty_bitmap) { |
4259 | bitmap_size = (bdrv_getlength(bs) >> BDRV_SECTOR_BITS) + | |
71df14fc PB |
4260 | BDRV_SECTORS_PER_DIRTY_CHUNK * BITS_PER_LONG - 1; |
4261 | bitmap_size /= BDRV_SECTORS_PER_DIRTY_CHUNK * BITS_PER_LONG; | |
a55eb92c | 4262 | |
71df14fc | 4263 | bs->dirty_bitmap = g_new0(unsigned long, bitmap_size); |
a55eb92c | 4264 | } |
7cd1e32a | 4265 | } else { |
c6d22830 | 4266 | if (bs->dirty_bitmap) { |
7267c094 | 4267 | g_free(bs->dirty_bitmap); |
c6d22830 | 4268 | bs->dirty_bitmap = NULL; |
a55eb92c | 4269 | } |
7cd1e32a LS |
4270 | } |
4271 | } | |
4272 | ||
4273 | int bdrv_get_dirty(BlockDriverState *bs, int64_t sector) | |
4274 | { | |
6ea44308 | 4275 | int64_t chunk = sector / (int64_t)BDRV_SECTORS_PER_DIRTY_CHUNK; |
a55eb92c | 4276 | |
c6d22830 JK |
4277 | if (bs->dirty_bitmap && |
4278 | (sector << BDRV_SECTOR_BITS) < bdrv_getlength(bs)) { | |
1755da16 PB |
4279 | return !!(bs->dirty_bitmap[chunk / BITS_PER_LONG] & |
4280 | (1UL << (chunk % BITS_PER_LONG))); | |
7cd1e32a LS |
4281 | } else { |
4282 | return 0; | |
4283 | } | |
4284 | } | |
4285 | ||
1755da16 PB |
4286 | int64_t bdrv_get_next_dirty(BlockDriverState *bs, int64_t sector) |
4287 | { | |
4288 | int64_t chunk; | |
4289 | int bit, elem; | |
4290 | ||
4291 | /* Avoid an infinite loop. */ | |
4292 | assert(bs->dirty_count > 0); | |
4293 | ||
4294 | sector = (sector | (BDRV_SECTORS_PER_DIRTY_CHUNK - 1)) + 1; | |
4295 | chunk = sector / (int64_t)BDRV_SECTORS_PER_DIRTY_CHUNK; | |
4296 | ||
4297 | QEMU_BUILD_BUG_ON(sizeof(bs->dirty_bitmap[0]) * 8 != BITS_PER_LONG); | |
4298 | elem = chunk / BITS_PER_LONG; | |
4299 | bit = chunk % BITS_PER_LONG; | |
4300 | for (;;) { | |
4301 | if (sector >= bs->total_sectors) { | |
4302 | sector = 0; | |
4303 | bit = elem = 0; | |
4304 | } | |
4305 | if (bit == 0 && bs->dirty_bitmap[elem] == 0) { | |
4306 | sector += BDRV_SECTORS_PER_DIRTY_CHUNK * BITS_PER_LONG; | |
4307 | elem++; | |
4308 | } else { | |
4309 | if (bs->dirty_bitmap[elem] & (1UL << bit)) { | |
4310 | return sector; | |
4311 | } | |
4312 | sector += BDRV_SECTORS_PER_DIRTY_CHUNK; | |
4313 | if (++bit == BITS_PER_LONG) { | |
4314 | bit = 0; | |
4315 | elem++; | |
4316 | } | |
4317 | } | |
4318 | } | |
4319 | } | |
4320 | ||
4321 | void bdrv_set_dirty(BlockDriverState *bs, int64_t cur_sector, | |
4322 | int nr_sectors) | |
4323 | { | |
4324 | set_dirty_bitmap(bs, cur_sector, nr_sectors, 1); | |
4325 | } | |
4326 | ||
a55eb92c JK |
4327 | void bdrv_reset_dirty(BlockDriverState *bs, int64_t cur_sector, |
4328 | int nr_sectors) | |
7cd1e32a LS |
4329 | { |
4330 | set_dirty_bitmap(bs, cur_sector, nr_sectors, 0); | |
4331 | } | |
aaa0eb75 LS |
4332 | |
4333 | int64_t bdrv_get_dirty_count(BlockDriverState *bs) | |
4334 | { | |
4335 | return bs->dirty_count; | |
4336 | } | |
f88e1a42 | 4337 | |
db593f25 MT |
4338 | void bdrv_set_in_use(BlockDriverState *bs, int in_use) |
4339 | { | |
4340 | assert(bs->in_use != in_use); | |
4341 | bs->in_use = in_use; | |
4342 | } | |
4343 | ||
4344 | int bdrv_in_use(BlockDriverState *bs) | |
4345 | { | |
4346 | return bs->in_use; | |
4347 | } | |
4348 | ||
28a7282a LC |
4349 | void bdrv_iostatus_enable(BlockDriverState *bs) |
4350 | { | |
d6bf279e | 4351 | bs->iostatus_enabled = true; |
58e21ef5 | 4352 | bs->iostatus = BLOCK_DEVICE_IO_STATUS_OK; |
28a7282a LC |
4353 | } |
4354 | ||
4355 | /* The I/O status is only enabled if the drive explicitly | |
4356 | * enables it _and_ the VM is configured to stop on errors */ | |
4357 | bool bdrv_iostatus_is_enabled(const BlockDriverState *bs) | |
4358 | { | |
d6bf279e | 4359 | return (bs->iostatus_enabled && |
92aa5c6d PB |
4360 | (bs->on_write_error == BLOCKDEV_ON_ERROR_ENOSPC || |
4361 | bs->on_write_error == BLOCKDEV_ON_ERROR_STOP || | |
4362 | bs->on_read_error == BLOCKDEV_ON_ERROR_STOP)); | |
28a7282a LC |
4363 | } |
4364 | ||
4365 | void bdrv_iostatus_disable(BlockDriverState *bs) | |
4366 | { | |
d6bf279e | 4367 | bs->iostatus_enabled = false; |
28a7282a LC |
4368 | } |
4369 | ||
4370 | void bdrv_iostatus_reset(BlockDriverState *bs) | |
4371 | { | |
4372 | if (bdrv_iostatus_is_enabled(bs)) { | |
58e21ef5 | 4373 | bs->iostatus = BLOCK_DEVICE_IO_STATUS_OK; |
3bd293c3 PB |
4374 | if (bs->job) { |
4375 | block_job_iostatus_reset(bs->job); | |
4376 | } | |
28a7282a LC |
4377 | } |
4378 | } | |
4379 | ||
28a7282a LC |
4380 | void bdrv_iostatus_set_err(BlockDriverState *bs, int error) |
4381 | { | |
3e1caa5f PB |
4382 | assert(bdrv_iostatus_is_enabled(bs)); |
4383 | if (bs->iostatus == BLOCK_DEVICE_IO_STATUS_OK) { | |
58e21ef5 LC |
4384 | bs->iostatus = error == ENOSPC ? BLOCK_DEVICE_IO_STATUS_NOSPACE : |
4385 | BLOCK_DEVICE_IO_STATUS_FAILED; | |
28a7282a LC |
4386 | } |
4387 | } | |
4388 | ||
a597e79c CH |
4389 | void |
4390 | bdrv_acct_start(BlockDriverState *bs, BlockAcctCookie *cookie, int64_t bytes, | |
4391 | enum BlockAcctType type) | |
4392 | { | |
4393 | assert(type < BDRV_MAX_IOTYPE); | |
4394 | ||
4395 | cookie->bytes = bytes; | |
c488c7f6 | 4396 | cookie->start_time_ns = get_clock(); |
a597e79c CH |
4397 | cookie->type = type; |
4398 | } | |
4399 | ||
4400 | void | |
4401 | bdrv_acct_done(BlockDriverState *bs, BlockAcctCookie *cookie) | |
4402 | { | |
4403 | assert(cookie->type < BDRV_MAX_IOTYPE); | |
4404 | ||
4405 | bs->nr_bytes[cookie->type] += cookie->bytes; | |
4406 | bs->nr_ops[cookie->type]++; | |
c488c7f6 | 4407 | bs->total_time_ns[cookie->type] += get_clock() - cookie->start_time_ns; |
a597e79c CH |
4408 | } |
4409 | ||
f88e1a42 JS |
4410 | int bdrv_img_create(const char *filename, const char *fmt, |
4411 | const char *base_filename, const char *base_fmt, | |
4412 | char *options, uint64_t img_size, int flags) | |
4413 | { | |
4414 | QEMUOptionParameter *param = NULL, *create_options = NULL; | |
d220894e | 4415 | QEMUOptionParameter *backing_fmt, *backing_file, *size; |
f88e1a42 JS |
4416 | BlockDriverState *bs = NULL; |
4417 | BlockDriver *drv, *proto_drv; | |
96df67d1 | 4418 | BlockDriver *backing_drv = NULL; |
f88e1a42 JS |
4419 | int ret = 0; |
4420 | ||
4421 | /* Find driver and parse its options */ | |
4422 | drv = bdrv_find_format(fmt); | |
4423 | if (!drv) { | |
4424 | error_report("Unknown file format '%s'", fmt); | |
4f70f249 | 4425 | ret = -EINVAL; |
f88e1a42 JS |
4426 | goto out; |
4427 | } | |
4428 | ||
4429 | proto_drv = bdrv_find_protocol(filename); | |
4430 | if (!proto_drv) { | |
4431 | error_report("Unknown protocol '%s'", filename); | |
4f70f249 | 4432 | ret = -EINVAL; |
f88e1a42 JS |
4433 | goto out; |
4434 | } | |
4435 | ||
4436 | create_options = append_option_parameters(create_options, | |
4437 | drv->create_options); | |
4438 | create_options = append_option_parameters(create_options, | |
4439 | proto_drv->create_options); | |
4440 | ||
4441 | /* Create parameter list with default values */ | |
4442 | param = parse_option_parameters("", create_options, param); | |
4443 | ||
4444 | set_option_parameter_int(param, BLOCK_OPT_SIZE, img_size); | |
4445 | ||
4446 | /* Parse -o options */ | |
4447 | if (options) { | |
4448 | param = parse_option_parameters(options, create_options, param); | |
4449 | if (param == NULL) { | |
4450 | error_report("Invalid options for file format '%s'.", fmt); | |
4f70f249 | 4451 | ret = -EINVAL; |
f88e1a42 JS |
4452 | goto out; |
4453 | } | |
4454 | } | |
4455 | ||
4456 | if (base_filename) { | |
4457 | if (set_option_parameter(param, BLOCK_OPT_BACKING_FILE, | |
4458 | base_filename)) { | |
4459 | error_report("Backing file not supported for file format '%s'", | |
4460 | fmt); | |
4f70f249 | 4461 | ret = -EINVAL; |
f88e1a42 JS |
4462 | goto out; |
4463 | } | |
4464 | } | |
4465 | ||
4466 | if (base_fmt) { | |
4467 | if (set_option_parameter(param, BLOCK_OPT_BACKING_FMT, base_fmt)) { | |
4468 | error_report("Backing file format not supported for file " | |
4469 | "format '%s'", fmt); | |
4f70f249 | 4470 | ret = -EINVAL; |
f88e1a42 JS |
4471 | goto out; |
4472 | } | |
4473 | } | |
4474 | ||
792da93a JS |
4475 | backing_file = get_option_parameter(param, BLOCK_OPT_BACKING_FILE); |
4476 | if (backing_file && backing_file->value.s) { | |
4477 | if (!strcmp(filename, backing_file->value.s)) { | |
4478 | error_report("Error: Trying to create an image with the " | |
4479 | "same filename as the backing file"); | |
4f70f249 | 4480 | ret = -EINVAL; |
792da93a JS |
4481 | goto out; |
4482 | } | |
4483 | } | |
4484 | ||
f88e1a42 JS |
4485 | backing_fmt = get_option_parameter(param, BLOCK_OPT_BACKING_FMT); |
4486 | if (backing_fmt && backing_fmt->value.s) { | |
96df67d1 SH |
4487 | backing_drv = bdrv_find_format(backing_fmt->value.s); |
4488 | if (!backing_drv) { | |
f88e1a42 JS |
4489 | error_report("Unknown backing file format '%s'", |
4490 | backing_fmt->value.s); | |
4f70f249 | 4491 | ret = -EINVAL; |
f88e1a42 JS |
4492 | goto out; |
4493 | } | |
4494 | } | |
4495 | ||
4496 | // The size for the image must always be specified, with one exception: | |
4497 | // If we are using a backing file, we can obtain the size from there | |
d220894e KW |
4498 | size = get_option_parameter(param, BLOCK_OPT_SIZE); |
4499 | if (size && size->value.n == -1) { | |
f88e1a42 JS |
4500 | if (backing_file && backing_file->value.s) { |
4501 | uint64_t size; | |
f88e1a42 | 4502 | char buf[32]; |
63090dac PB |
4503 | int back_flags; |
4504 | ||
4505 | /* backing files always opened read-only */ | |
4506 | back_flags = | |
4507 | flags & ~(BDRV_O_RDWR | BDRV_O_SNAPSHOT | BDRV_O_NO_BACKING); | |
f88e1a42 | 4508 | |
f88e1a42 JS |
4509 | bs = bdrv_new(""); |
4510 | ||
63090dac | 4511 | ret = bdrv_open(bs, backing_file->value.s, back_flags, backing_drv); |
f88e1a42 | 4512 | if (ret < 0) { |
96df67d1 | 4513 | error_report("Could not open '%s'", backing_file->value.s); |
f88e1a42 JS |
4514 | goto out; |
4515 | } | |
4516 | bdrv_get_geometry(bs, &size); | |
4517 | size *= 512; | |
4518 | ||
4519 | snprintf(buf, sizeof(buf), "%" PRId64, size); | |
4520 | set_option_parameter(param, BLOCK_OPT_SIZE, buf); | |
4521 | } else { | |
4522 | error_report("Image creation needs a size parameter"); | |
4f70f249 | 4523 | ret = -EINVAL; |
f88e1a42 JS |
4524 | goto out; |
4525 | } | |
4526 | } | |
4527 | ||
4528 | printf("Formatting '%s', fmt=%s ", filename, fmt); | |
4529 | print_option_parameters(param); | |
4530 | puts(""); | |
4531 | ||
4532 | ret = bdrv_create(drv, filename, param); | |
4533 | ||
4534 | if (ret < 0) { | |
4535 | if (ret == -ENOTSUP) { | |
4536 | error_report("Formatting or formatting option not supported for " | |
4537 | "file format '%s'", fmt); | |
4538 | } else if (ret == -EFBIG) { | |
4539 | error_report("The image size is too large for file format '%s'", | |
4540 | fmt); | |
4541 | } else { | |
4542 | error_report("%s: error while creating %s: %s", filename, fmt, | |
4543 | strerror(-ret)); | |
4544 | } | |
4545 | } | |
4546 | ||
4547 | out: | |
4548 | free_option_parameters(create_options); | |
4549 | free_option_parameters(param); | |
4550 | ||
4551 | if (bs) { | |
4552 | bdrv_delete(bs); | |
4553 | } | |
4f70f249 JS |
4554 | |
4555 | return ret; | |
f88e1a42 | 4556 | } |