]>
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" |
68485420 | 32 | #include "qemu-coroutine.h" |
b2023818 | 33 | #include "qmp-commands.h" |
0563e191 | 34 | #include "qemu-timer.h" |
fc01f7e7 | 35 | |
71e72a19 | 36 | #ifdef CONFIG_BSD |
7674e7bf FB |
37 | #include <sys/types.h> |
38 | #include <sys/stat.h> | |
39 | #include <sys/ioctl.h> | |
72cf2d4f | 40 | #include <sys/queue.h> |
c5e97233 | 41 | #ifndef __DragonFly__ |
7674e7bf FB |
42 | #include <sys/disk.h> |
43 | #endif | |
c5e97233 | 44 | #endif |
7674e7bf | 45 | |
49dc768d AL |
46 | #ifdef _WIN32 |
47 | #include <windows.h> | |
48 | #endif | |
49 | ||
1c9805a3 SH |
50 | #define NOT_DONE 0x7fffffff /* used while emulated sync operation in progress */ |
51 | ||
470c0504 SH |
52 | typedef enum { |
53 | BDRV_REQ_COPY_ON_READ = 0x1, | |
f08f2dda | 54 | BDRV_REQ_ZERO_WRITE = 0x2, |
470c0504 SH |
55 | } BdrvRequestFlags; |
56 | ||
7d4b4ba5 | 57 | static void bdrv_dev_change_media_cb(BlockDriverState *bs, bool load); |
f141eafe AL |
58 | static BlockDriverAIOCB *bdrv_aio_readv_em(BlockDriverState *bs, |
59 | int64_t sector_num, QEMUIOVector *qiov, int nb_sectors, | |
c87c0672 | 60 | BlockDriverCompletionFunc *cb, void *opaque); |
f141eafe AL |
61 | static BlockDriverAIOCB *bdrv_aio_writev_em(BlockDriverState *bs, |
62 | int64_t sector_num, QEMUIOVector *qiov, int nb_sectors, | |
ce1a14dc | 63 | BlockDriverCompletionFunc *cb, void *opaque); |
f9f05dc5 KW |
64 | static int coroutine_fn bdrv_co_readv_em(BlockDriverState *bs, |
65 | int64_t sector_num, int nb_sectors, | |
66 | QEMUIOVector *iov); | |
67 | static int coroutine_fn bdrv_co_writev_em(BlockDriverState *bs, | |
68 | int64_t sector_num, int nb_sectors, | |
69 | QEMUIOVector *iov); | |
c5fbe571 | 70 | static int coroutine_fn bdrv_co_do_readv(BlockDriverState *bs, |
470c0504 SH |
71 | int64_t sector_num, int nb_sectors, QEMUIOVector *qiov, |
72 | BdrvRequestFlags flags); | |
1c9805a3 | 73 | static int coroutine_fn bdrv_co_do_writev(BlockDriverState *bs, |
f08f2dda SH |
74 | int64_t sector_num, int nb_sectors, QEMUIOVector *qiov, |
75 | BdrvRequestFlags flags); | |
b2a61371 SH |
76 | static BlockDriverAIOCB *bdrv_co_aio_rw_vector(BlockDriverState *bs, |
77 | int64_t sector_num, | |
78 | QEMUIOVector *qiov, | |
79 | int nb_sectors, | |
80 | BlockDriverCompletionFunc *cb, | |
81 | void *opaque, | |
8c5873d6 | 82 | bool is_write); |
b2a61371 | 83 | static void coroutine_fn bdrv_co_do_rw(void *opaque); |
621f0589 KW |
84 | static int coroutine_fn bdrv_co_do_write_zeroes(BlockDriverState *bs, |
85 | int64_t sector_num, int nb_sectors); | |
ec530c81 | 86 | |
98f90dba ZYW |
87 | static bool bdrv_exceed_bps_limits(BlockDriverState *bs, int nb_sectors, |
88 | bool is_write, double elapsed_time, uint64_t *wait); | |
89 | static bool bdrv_exceed_iops_limits(BlockDriverState *bs, bool is_write, | |
90 | double elapsed_time, uint64_t *wait); | |
91 | static bool bdrv_exceed_io_limits(BlockDriverState *bs, int nb_sectors, | |
92 | bool is_write, int64_t *wait); | |
93 | ||
1b7bdbc1 SH |
94 | static QTAILQ_HEAD(, BlockDriverState) bdrv_states = |
95 | QTAILQ_HEAD_INITIALIZER(bdrv_states); | |
7ee930d0 | 96 | |
8a22f02a SH |
97 | static QLIST_HEAD(, BlockDriver) bdrv_drivers = |
98 | QLIST_HEAD_INITIALIZER(bdrv_drivers); | |
ea2384d3 | 99 | |
f9092b10 MA |
100 | /* The device to use for VM snapshots */ |
101 | static BlockDriverState *bs_snapshots; | |
102 | ||
eb852011 MA |
103 | /* If non-zero, use only whitelisted block drivers */ |
104 | static int use_bdrv_whitelist; | |
105 | ||
9e0b22f4 SH |
106 | #ifdef _WIN32 |
107 | static int is_windows_drive_prefix(const char *filename) | |
108 | { | |
109 | return (((filename[0] >= 'a' && filename[0] <= 'z') || | |
110 | (filename[0] >= 'A' && filename[0] <= 'Z')) && | |
111 | filename[1] == ':'); | |
112 | } | |
113 | ||
114 | int is_windows_drive(const char *filename) | |
115 | { | |
116 | if (is_windows_drive_prefix(filename) && | |
117 | filename[2] == '\0') | |
118 | return 1; | |
119 | if (strstart(filename, "\\\\.\\", NULL) || | |
120 | strstart(filename, "//./", NULL)) | |
121 | return 1; | |
122 | return 0; | |
123 | } | |
124 | #endif | |
125 | ||
0563e191 | 126 | /* throttling disk I/O limits */ |
98f90dba ZYW |
127 | void bdrv_io_limits_disable(BlockDriverState *bs) |
128 | { | |
129 | bs->io_limits_enabled = false; | |
130 | ||
131 | while (qemu_co_queue_next(&bs->throttled_reqs)); | |
132 | ||
133 | if (bs->block_timer) { | |
134 | qemu_del_timer(bs->block_timer); | |
135 | qemu_free_timer(bs->block_timer); | |
136 | bs->block_timer = NULL; | |
137 | } | |
138 | ||
139 | bs->slice_start = 0; | |
140 | bs->slice_end = 0; | |
141 | bs->slice_time = 0; | |
142 | memset(&bs->io_base, 0, sizeof(bs->io_base)); | |
143 | } | |
144 | ||
0563e191 ZYW |
145 | static void bdrv_block_timer(void *opaque) |
146 | { | |
147 | BlockDriverState *bs = opaque; | |
148 | ||
149 | qemu_co_queue_next(&bs->throttled_reqs); | |
150 | } | |
151 | ||
152 | void bdrv_io_limits_enable(BlockDriverState *bs) | |
153 | { | |
154 | qemu_co_queue_init(&bs->throttled_reqs); | |
155 | bs->block_timer = qemu_new_timer_ns(vm_clock, bdrv_block_timer, bs); | |
156 | bs->slice_time = 5 * BLOCK_IO_SLICE_TIME; | |
157 | bs->slice_start = qemu_get_clock_ns(vm_clock); | |
158 | bs->slice_end = bs->slice_start + bs->slice_time; | |
159 | memset(&bs->io_base, 0, sizeof(bs->io_base)); | |
160 | bs->io_limits_enabled = true; | |
161 | } | |
162 | ||
163 | bool bdrv_io_limits_enabled(BlockDriverState *bs) | |
164 | { | |
165 | BlockIOLimit *io_limits = &bs->io_limits; | |
166 | return io_limits->bps[BLOCK_IO_LIMIT_READ] | |
167 | || io_limits->bps[BLOCK_IO_LIMIT_WRITE] | |
168 | || io_limits->bps[BLOCK_IO_LIMIT_TOTAL] | |
169 | || io_limits->iops[BLOCK_IO_LIMIT_READ] | |
170 | || io_limits->iops[BLOCK_IO_LIMIT_WRITE] | |
171 | || io_limits->iops[BLOCK_IO_LIMIT_TOTAL]; | |
172 | } | |
173 | ||
98f90dba ZYW |
174 | static void bdrv_io_limits_intercept(BlockDriverState *bs, |
175 | bool is_write, int nb_sectors) | |
176 | { | |
177 | int64_t wait_time = -1; | |
178 | ||
179 | if (!qemu_co_queue_empty(&bs->throttled_reqs)) { | |
180 | qemu_co_queue_wait(&bs->throttled_reqs); | |
181 | } | |
182 | ||
183 | /* In fact, we hope to keep each request's timing, in FIFO mode. The next | |
184 | * throttled requests will not be dequeued until the current request is | |
185 | * allowed to be serviced. So if the current request still exceeds the | |
186 | * limits, it will be inserted to the head. All requests followed it will | |
187 | * be still in throttled_reqs queue. | |
188 | */ | |
189 | ||
190 | while (bdrv_exceed_io_limits(bs, nb_sectors, is_write, &wait_time)) { | |
191 | qemu_mod_timer(bs->block_timer, | |
192 | wait_time + qemu_get_clock_ns(vm_clock)); | |
193 | qemu_co_queue_wait_insert_head(&bs->throttled_reqs); | |
194 | } | |
195 | ||
196 | qemu_co_queue_next(&bs->throttled_reqs); | |
197 | } | |
198 | ||
9e0b22f4 SH |
199 | /* check if the path starts with "<protocol>:" */ |
200 | static int path_has_protocol(const char *path) | |
201 | { | |
947995c0 PB |
202 | const char *p; |
203 | ||
9e0b22f4 SH |
204 | #ifdef _WIN32 |
205 | if (is_windows_drive(path) || | |
206 | is_windows_drive_prefix(path)) { | |
207 | return 0; | |
208 | } | |
947995c0 PB |
209 | p = path + strcspn(path, ":/\\"); |
210 | #else | |
211 | p = path + strcspn(path, ":/"); | |
9e0b22f4 SH |
212 | #endif |
213 | ||
947995c0 | 214 | return *p == ':'; |
9e0b22f4 SH |
215 | } |
216 | ||
83f64091 | 217 | int path_is_absolute(const char *path) |
3b0d4f61 | 218 | { |
21664424 FB |
219 | #ifdef _WIN32 |
220 | /* specific case for names like: "\\.\d:" */ | |
f53f4da9 | 221 | if (is_windows_drive(path) || is_windows_drive_prefix(path)) { |
21664424 | 222 | return 1; |
f53f4da9 PB |
223 | } |
224 | return (*path == '/' || *path == '\\'); | |
3b9f94e1 | 225 | #else |
f53f4da9 | 226 | return (*path == '/'); |
3b9f94e1 | 227 | #endif |
3b0d4f61 FB |
228 | } |
229 | ||
83f64091 FB |
230 | /* if filename is absolute, just copy it to dest. Otherwise, build a |
231 | path to it by considering it is relative to base_path. URL are | |
232 | supported. */ | |
233 | void path_combine(char *dest, int dest_size, | |
234 | const char *base_path, | |
235 | const char *filename) | |
3b0d4f61 | 236 | { |
83f64091 FB |
237 | const char *p, *p1; |
238 | int len; | |
239 | ||
240 | if (dest_size <= 0) | |
241 | return; | |
242 | if (path_is_absolute(filename)) { | |
243 | pstrcpy(dest, dest_size, filename); | |
244 | } else { | |
245 | p = strchr(base_path, ':'); | |
246 | if (p) | |
247 | p++; | |
248 | else | |
249 | p = base_path; | |
3b9f94e1 FB |
250 | p1 = strrchr(base_path, '/'); |
251 | #ifdef _WIN32 | |
252 | { | |
253 | const char *p2; | |
254 | p2 = strrchr(base_path, '\\'); | |
255 | if (!p1 || p2 > p1) | |
256 | p1 = p2; | |
257 | } | |
258 | #endif | |
83f64091 FB |
259 | if (p1) |
260 | p1++; | |
261 | else | |
262 | p1 = base_path; | |
263 | if (p1 > p) | |
264 | p = p1; | |
265 | len = p - base_path; | |
266 | if (len > dest_size - 1) | |
267 | len = dest_size - 1; | |
268 | memcpy(dest, base_path, len); | |
269 | dest[len] = '\0'; | |
270 | pstrcat(dest, dest_size, filename); | |
3b0d4f61 | 271 | } |
3b0d4f61 FB |
272 | } |
273 | ||
dc5a1371 PB |
274 | void bdrv_get_full_backing_filename(BlockDriverState *bs, char *dest, size_t sz) |
275 | { | |
276 | if (bs->backing_file[0] == '\0' || path_has_protocol(bs->backing_file)) { | |
277 | pstrcpy(dest, sz, bs->backing_file); | |
278 | } else { | |
279 | path_combine(dest, sz, bs->filename, bs->backing_file); | |
280 | } | |
281 | } | |
282 | ||
5efa9d5a | 283 | void bdrv_register(BlockDriver *bdrv) |
ea2384d3 | 284 | { |
8c5873d6 SH |
285 | /* Block drivers without coroutine functions need emulation */ |
286 | if (!bdrv->bdrv_co_readv) { | |
f9f05dc5 KW |
287 | bdrv->bdrv_co_readv = bdrv_co_readv_em; |
288 | bdrv->bdrv_co_writev = bdrv_co_writev_em; | |
289 | ||
f8c35c1d SH |
290 | /* bdrv_co_readv_em()/brdv_co_writev_em() work in terms of aio, so if |
291 | * the block driver lacks aio we need to emulate that too. | |
292 | */ | |
f9f05dc5 KW |
293 | if (!bdrv->bdrv_aio_readv) { |
294 | /* add AIO emulation layer */ | |
295 | bdrv->bdrv_aio_readv = bdrv_aio_readv_em; | |
296 | bdrv->bdrv_aio_writev = bdrv_aio_writev_em; | |
f9f05dc5 | 297 | } |
83f64091 | 298 | } |
b2e12bc6 | 299 | |
8a22f02a | 300 | QLIST_INSERT_HEAD(&bdrv_drivers, bdrv, list); |
ea2384d3 | 301 | } |
b338082b FB |
302 | |
303 | /* create a new block device (by default it is empty) */ | |
304 | BlockDriverState *bdrv_new(const char *device_name) | |
305 | { | |
1b7bdbc1 | 306 | BlockDriverState *bs; |
b338082b | 307 | |
7267c094 | 308 | bs = g_malloc0(sizeof(BlockDriverState)); |
b338082b | 309 | pstrcpy(bs->device_name, sizeof(bs->device_name), device_name); |
ea2384d3 | 310 | if (device_name[0] != '\0') { |
1b7bdbc1 | 311 | QTAILQ_INSERT_TAIL(&bdrv_states, bs, list); |
ea2384d3 | 312 | } |
28a7282a | 313 | bdrv_iostatus_disable(bs); |
b338082b FB |
314 | return bs; |
315 | } | |
316 | ||
ea2384d3 FB |
317 | BlockDriver *bdrv_find_format(const char *format_name) |
318 | { | |
319 | BlockDriver *drv1; | |
8a22f02a SH |
320 | QLIST_FOREACH(drv1, &bdrv_drivers, list) { |
321 | if (!strcmp(drv1->format_name, format_name)) { | |
ea2384d3 | 322 | return drv1; |
8a22f02a | 323 | } |
ea2384d3 FB |
324 | } |
325 | return NULL; | |
326 | } | |
327 | ||
eb852011 MA |
328 | static int bdrv_is_whitelisted(BlockDriver *drv) |
329 | { | |
330 | static const char *whitelist[] = { | |
331 | CONFIG_BDRV_WHITELIST | |
332 | }; | |
333 | const char **p; | |
334 | ||
335 | if (!whitelist[0]) | |
336 | return 1; /* no whitelist, anything goes */ | |
337 | ||
338 | for (p = whitelist; *p; p++) { | |
339 | if (!strcmp(drv->format_name, *p)) { | |
340 | return 1; | |
341 | } | |
342 | } | |
343 | return 0; | |
344 | } | |
345 | ||
346 | BlockDriver *bdrv_find_whitelisted_format(const char *format_name) | |
347 | { | |
348 | BlockDriver *drv = bdrv_find_format(format_name); | |
349 | return drv && bdrv_is_whitelisted(drv) ? drv : NULL; | |
350 | } | |
351 | ||
5b7e1542 ZYW |
352 | typedef struct CreateCo { |
353 | BlockDriver *drv; | |
354 | char *filename; | |
355 | QEMUOptionParameter *options; | |
356 | int ret; | |
357 | } CreateCo; | |
358 | ||
359 | static void coroutine_fn bdrv_create_co_entry(void *opaque) | |
360 | { | |
361 | CreateCo *cco = opaque; | |
362 | assert(cco->drv); | |
363 | ||
364 | cco->ret = cco->drv->bdrv_create(cco->filename, cco->options); | |
365 | } | |
366 | ||
0e7e1989 KW |
367 | int bdrv_create(BlockDriver *drv, const char* filename, |
368 | QEMUOptionParameter *options) | |
ea2384d3 | 369 | { |
5b7e1542 ZYW |
370 | int ret; |
371 | ||
372 | Coroutine *co; | |
373 | CreateCo cco = { | |
374 | .drv = drv, | |
375 | .filename = g_strdup(filename), | |
376 | .options = options, | |
377 | .ret = NOT_DONE, | |
378 | }; | |
379 | ||
380 | if (!drv->bdrv_create) { | |
ea2384d3 | 381 | return -ENOTSUP; |
5b7e1542 ZYW |
382 | } |
383 | ||
384 | if (qemu_in_coroutine()) { | |
385 | /* Fast-path if already in coroutine context */ | |
386 | bdrv_create_co_entry(&cco); | |
387 | } else { | |
388 | co = qemu_coroutine_create(bdrv_create_co_entry); | |
389 | qemu_coroutine_enter(co, &cco); | |
390 | while (cco.ret == NOT_DONE) { | |
391 | qemu_aio_wait(); | |
392 | } | |
393 | } | |
394 | ||
395 | ret = cco.ret; | |
396 | g_free(cco.filename); | |
0e7e1989 | 397 | |
5b7e1542 | 398 | return ret; |
ea2384d3 FB |
399 | } |
400 | ||
84a12e66 CH |
401 | int bdrv_create_file(const char* filename, QEMUOptionParameter *options) |
402 | { | |
403 | BlockDriver *drv; | |
404 | ||
b50cbabc | 405 | drv = bdrv_find_protocol(filename); |
84a12e66 | 406 | if (drv == NULL) { |
16905d71 | 407 | return -ENOENT; |
84a12e66 CH |
408 | } |
409 | ||
410 | return bdrv_create(drv, filename, options); | |
411 | } | |
412 | ||
eba25057 JM |
413 | /* |
414 | * Create a uniquely-named empty temporary file. | |
415 | * Return 0 upon success, otherwise a negative errno value. | |
416 | */ | |
417 | int get_tmp_filename(char *filename, int size) | |
d5249393 | 418 | { |
eba25057 | 419 | #ifdef _WIN32 |
3b9f94e1 | 420 | char temp_dir[MAX_PATH]; |
eba25057 JM |
421 | /* GetTempFileName requires that its output buffer (4th param) |
422 | have length MAX_PATH or greater. */ | |
423 | assert(size >= MAX_PATH); | |
424 | return (GetTempPath(MAX_PATH, temp_dir) | |
425 | && GetTempFileName(temp_dir, "qem", 0, filename) | |
426 | ? 0 : -GetLastError()); | |
d5249393 | 427 | #else |
67b915a5 | 428 | int fd; |
7ccfb2eb | 429 | const char *tmpdir; |
0badc1ee AJ |
430 | tmpdir = getenv("TMPDIR"); |
431 | if (!tmpdir) | |
432 | tmpdir = "/tmp"; | |
eba25057 JM |
433 | if (snprintf(filename, size, "%s/vl.XXXXXX", tmpdir) >= size) { |
434 | return -EOVERFLOW; | |
435 | } | |
ea2384d3 | 436 | fd = mkstemp(filename); |
fe235a06 DH |
437 | if (fd < 0) { |
438 | return -errno; | |
439 | } | |
440 | if (close(fd) != 0) { | |
441 | unlink(filename); | |
eba25057 JM |
442 | return -errno; |
443 | } | |
444 | return 0; | |
d5249393 | 445 | #endif |
eba25057 | 446 | } |
fc01f7e7 | 447 | |
84a12e66 CH |
448 | /* |
449 | * Detect host devices. By convention, /dev/cdrom[N] is always | |
450 | * recognized as a host CDROM. | |
451 | */ | |
452 | static BlockDriver *find_hdev_driver(const char *filename) | |
453 | { | |
454 | int score_max = 0, score; | |
455 | BlockDriver *drv = NULL, *d; | |
456 | ||
457 | QLIST_FOREACH(d, &bdrv_drivers, list) { | |
458 | if (d->bdrv_probe_device) { | |
459 | score = d->bdrv_probe_device(filename); | |
460 | if (score > score_max) { | |
461 | score_max = score; | |
462 | drv = d; | |
463 | } | |
464 | } | |
465 | } | |
466 | ||
467 | return drv; | |
468 | } | |
469 | ||
b50cbabc | 470 | BlockDriver *bdrv_find_protocol(const char *filename) |
83f64091 FB |
471 | { |
472 | BlockDriver *drv1; | |
473 | char protocol[128]; | |
1cec71e3 | 474 | int len; |
83f64091 | 475 | const char *p; |
19cb3738 | 476 | |
66f82cee KW |
477 | /* TODO Drivers without bdrv_file_open must be specified explicitly */ |
478 | ||
39508e7a CH |
479 | /* |
480 | * XXX(hch): we really should not let host device detection | |
481 | * override an explicit protocol specification, but moving this | |
482 | * later breaks access to device names with colons in them. | |
483 | * Thanks to the brain-dead persistent naming schemes on udev- | |
484 | * based Linux systems those actually are quite common. | |
485 | */ | |
486 | drv1 = find_hdev_driver(filename); | |
487 | if (drv1) { | |
488 | return drv1; | |
489 | } | |
490 | ||
9e0b22f4 | 491 | if (!path_has_protocol(filename)) { |
39508e7a | 492 | return bdrv_find_format("file"); |
84a12e66 | 493 | } |
9e0b22f4 SH |
494 | p = strchr(filename, ':'); |
495 | assert(p != NULL); | |
1cec71e3 AL |
496 | len = p - filename; |
497 | if (len > sizeof(protocol) - 1) | |
498 | len = sizeof(protocol) - 1; | |
499 | memcpy(protocol, filename, len); | |
500 | protocol[len] = '\0'; | |
8a22f02a | 501 | QLIST_FOREACH(drv1, &bdrv_drivers, list) { |
5fafdf24 | 502 | if (drv1->protocol_name && |
8a22f02a | 503 | !strcmp(drv1->protocol_name, protocol)) { |
83f64091 | 504 | return drv1; |
8a22f02a | 505 | } |
83f64091 FB |
506 | } |
507 | return NULL; | |
508 | } | |
509 | ||
c98ac35d | 510 | static int find_image_format(const char *filename, BlockDriver **pdrv) |
f3a5d3f8 CH |
511 | { |
512 | int ret, score, score_max; | |
513 | BlockDriver *drv1, *drv; | |
514 | uint8_t buf[2048]; | |
515 | BlockDriverState *bs; | |
516 | ||
f5edb014 | 517 | ret = bdrv_file_open(&bs, filename, 0); |
c98ac35d SW |
518 | if (ret < 0) { |
519 | *pdrv = NULL; | |
520 | return ret; | |
521 | } | |
f8ea0b00 | 522 | |
08a00559 KW |
523 | /* Return the raw BlockDriver * to scsi-generic devices or empty drives */ |
524 | if (bs->sg || !bdrv_is_inserted(bs)) { | |
1a396859 | 525 | bdrv_delete(bs); |
c98ac35d SW |
526 | drv = bdrv_find_format("raw"); |
527 | if (!drv) { | |
528 | ret = -ENOENT; | |
529 | } | |
530 | *pdrv = drv; | |
531 | return ret; | |
1a396859 | 532 | } |
f8ea0b00 | 533 | |
83f64091 FB |
534 | ret = bdrv_pread(bs, 0, buf, sizeof(buf)); |
535 | bdrv_delete(bs); | |
536 | if (ret < 0) { | |
c98ac35d SW |
537 | *pdrv = NULL; |
538 | return ret; | |
83f64091 FB |
539 | } |
540 | ||
ea2384d3 | 541 | score_max = 0; |
84a12e66 | 542 | drv = NULL; |
8a22f02a | 543 | QLIST_FOREACH(drv1, &bdrv_drivers, list) { |
83f64091 FB |
544 | if (drv1->bdrv_probe) { |
545 | score = drv1->bdrv_probe(buf, ret, filename); | |
546 | if (score > score_max) { | |
547 | score_max = score; | |
548 | drv = drv1; | |
549 | } | |
0849bf08 | 550 | } |
fc01f7e7 | 551 | } |
c98ac35d SW |
552 | if (!drv) { |
553 | ret = -ENOENT; | |
554 | } | |
555 | *pdrv = drv; | |
556 | return ret; | |
ea2384d3 FB |
557 | } |
558 | ||
51762288 SH |
559 | /** |
560 | * Set the current 'total_sectors' value | |
561 | */ | |
562 | static int refresh_total_sectors(BlockDriverState *bs, int64_t hint) | |
563 | { | |
564 | BlockDriver *drv = bs->drv; | |
565 | ||
396759ad NB |
566 | /* Do not attempt drv->bdrv_getlength() on scsi-generic devices */ |
567 | if (bs->sg) | |
568 | return 0; | |
569 | ||
51762288 SH |
570 | /* query actual device if possible, otherwise just trust the hint */ |
571 | if (drv->bdrv_getlength) { | |
572 | int64_t length = drv->bdrv_getlength(bs); | |
573 | if (length < 0) { | |
574 | return length; | |
575 | } | |
576 | hint = length >> BDRV_SECTOR_BITS; | |
577 | } | |
578 | ||
579 | bs->total_sectors = hint; | |
580 | return 0; | |
581 | } | |
582 | ||
c3993cdc SH |
583 | /** |
584 | * Set open flags for a given cache mode | |
585 | * | |
586 | * Return 0 on success, -1 if the cache mode was invalid. | |
587 | */ | |
588 | int bdrv_parse_cache_flags(const char *mode, int *flags) | |
589 | { | |
590 | *flags &= ~BDRV_O_CACHE_MASK; | |
591 | ||
592 | if (!strcmp(mode, "off") || !strcmp(mode, "none")) { | |
593 | *flags |= BDRV_O_NOCACHE | BDRV_O_CACHE_WB; | |
92196b2f SH |
594 | } else if (!strcmp(mode, "directsync")) { |
595 | *flags |= BDRV_O_NOCACHE; | |
c3993cdc SH |
596 | } else if (!strcmp(mode, "writeback")) { |
597 | *flags |= BDRV_O_CACHE_WB; | |
598 | } else if (!strcmp(mode, "unsafe")) { | |
599 | *flags |= BDRV_O_CACHE_WB; | |
600 | *flags |= BDRV_O_NO_FLUSH; | |
601 | } else if (!strcmp(mode, "writethrough")) { | |
602 | /* this is the default */ | |
603 | } else { | |
604 | return -1; | |
605 | } | |
606 | ||
607 | return 0; | |
608 | } | |
609 | ||
53fec9d3 SH |
610 | /** |
611 | * The copy-on-read flag is actually a reference count so multiple users may | |
612 | * use the feature without worrying about clobbering its previous state. | |
613 | * Copy-on-read stays enabled until all users have called to disable it. | |
614 | */ | |
615 | void bdrv_enable_copy_on_read(BlockDriverState *bs) | |
616 | { | |
617 | bs->copy_on_read++; | |
618 | } | |
619 | ||
620 | void bdrv_disable_copy_on_read(BlockDriverState *bs) | |
621 | { | |
622 | assert(bs->copy_on_read > 0); | |
623 | bs->copy_on_read--; | |
624 | } | |
625 | ||
57915332 KW |
626 | /* |
627 | * Common part for opening disk images and files | |
628 | */ | |
629 | static int bdrv_open_common(BlockDriverState *bs, const char *filename, | |
630 | int flags, BlockDriver *drv) | |
631 | { | |
632 | int ret, open_flags; | |
633 | ||
634 | assert(drv != NULL); | |
6405875c | 635 | assert(bs->file == NULL); |
57915332 | 636 | |
28dcee10 SH |
637 | trace_bdrv_open_common(bs, filename, flags, drv->format_name); |
638 | ||
57915332 | 639 | bs->open_flags = flags; |
57915332 KW |
640 | bs->buffer_alignment = 512; |
641 | ||
53fec9d3 SH |
642 | assert(bs->copy_on_read == 0); /* bdrv_new() and bdrv_close() make it so */ |
643 | if ((flags & BDRV_O_RDWR) && (flags & BDRV_O_COPY_ON_READ)) { | |
644 | bdrv_enable_copy_on_read(bs); | |
645 | } | |
646 | ||
57915332 KW |
647 | pstrcpy(bs->filename, sizeof(bs->filename), filename); |
648 | ||
649 | if (use_bdrv_whitelist && !bdrv_is_whitelisted(drv)) { | |
650 | return -ENOTSUP; | |
651 | } | |
652 | ||
653 | bs->drv = drv; | |
7267c094 | 654 | bs->opaque = g_malloc0(drv->instance_size); |
57915332 | 655 | |
03f541bd | 656 | bs->enable_write_cache = !!(flags & BDRV_O_CACHE_WB); |
e1e9b0ac | 657 | open_flags = flags | BDRV_O_CACHE_WB; |
57915332 KW |
658 | |
659 | /* | |
660 | * Clear flags that are internal to the block layer before opening the | |
661 | * image. | |
662 | */ | |
e1e9b0ac | 663 | open_flags &= ~(BDRV_O_SNAPSHOT | BDRV_O_NO_BACKING); |
57915332 KW |
664 | |
665 | /* | |
ebabb67a | 666 | * Snapshots should be writable. |
57915332 KW |
667 | */ |
668 | if (bs->is_temporary) { | |
669 | open_flags |= BDRV_O_RDWR; | |
670 | } | |
671 | ||
be028adc | 672 | bs->read_only = !(open_flags & BDRV_O_RDWR); |
e7c63796 | 673 | |
66f82cee KW |
674 | /* Open the image, either directly or using a protocol */ |
675 | if (drv->bdrv_file_open) { | |
676 | ret = drv->bdrv_file_open(bs, filename, open_flags); | |
677 | } else { | |
678 | ret = bdrv_file_open(&bs->file, filename, open_flags); | |
679 | if (ret >= 0) { | |
680 | ret = drv->bdrv_open(bs, open_flags); | |
681 | } | |
682 | } | |
683 | ||
57915332 KW |
684 | if (ret < 0) { |
685 | goto free_and_fail; | |
686 | } | |
687 | ||
51762288 SH |
688 | ret = refresh_total_sectors(bs, bs->total_sectors); |
689 | if (ret < 0) { | |
690 | goto free_and_fail; | |
57915332 | 691 | } |
51762288 | 692 | |
57915332 KW |
693 | #ifndef _WIN32 |
694 | if (bs->is_temporary) { | |
695 | unlink(filename); | |
696 | } | |
697 | #endif | |
698 | return 0; | |
699 | ||
700 | free_and_fail: | |
66f82cee KW |
701 | if (bs->file) { |
702 | bdrv_delete(bs->file); | |
703 | bs->file = NULL; | |
704 | } | |
7267c094 | 705 | g_free(bs->opaque); |
57915332 KW |
706 | bs->opaque = NULL; |
707 | bs->drv = NULL; | |
708 | return ret; | |
709 | } | |
710 | ||
b6ce07aa KW |
711 | /* |
712 | * Opens a file using a protocol (file, host_device, nbd, ...) | |
713 | */ | |
83f64091 | 714 | int bdrv_file_open(BlockDriverState **pbs, const char *filename, int flags) |
ea2384d3 | 715 | { |
83f64091 | 716 | BlockDriverState *bs; |
6db95603 | 717 | BlockDriver *drv; |
83f64091 FB |
718 | int ret; |
719 | ||
b50cbabc | 720 | drv = bdrv_find_protocol(filename); |
6db95603 CH |
721 | if (!drv) { |
722 | return -ENOENT; | |
723 | } | |
724 | ||
83f64091 | 725 | bs = bdrv_new(""); |
b6ce07aa | 726 | ret = bdrv_open_common(bs, filename, flags, drv); |
83f64091 FB |
727 | if (ret < 0) { |
728 | bdrv_delete(bs); | |
729 | return ret; | |
3b0d4f61 | 730 | } |
71d0770c | 731 | bs->growable = 1; |
83f64091 FB |
732 | *pbs = bs; |
733 | return 0; | |
734 | } | |
735 | ||
b6ce07aa KW |
736 | /* |
737 | * Opens a disk image (raw, qcow2, vmdk, ...) | |
738 | */ | |
d6e9098e KW |
739 | int bdrv_open(BlockDriverState *bs, const char *filename, int flags, |
740 | BlockDriver *drv) | |
ea2384d3 | 741 | { |
b6ce07aa | 742 | int ret; |
2b572816 | 743 | char tmp_filename[PATH_MAX]; |
712e7874 | 744 | |
83f64091 | 745 | if (flags & BDRV_O_SNAPSHOT) { |
ea2384d3 FB |
746 | BlockDriverState *bs1; |
747 | int64_t total_size; | |
7c96d46e | 748 | int is_protocol = 0; |
91a073a9 KW |
749 | BlockDriver *bdrv_qcow2; |
750 | QEMUOptionParameter *options; | |
b6ce07aa | 751 | char backing_filename[PATH_MAX]; |
3b46e624 | 752 | |
ea2384d3 FB |
753 | /* if snapshot, we create a temporary backing file and open it |
754 | instead of opening 'filename' directly */ | |
33e3963e | 755 | |
ea2384d3 FB |
756 | /* if there is a backing file, use it */ |
757 | bs1 = bdrv_new(""); | |
d6e9098e | 758 | ret = bdrv_open(bs1, filename, 0, drv); |
51d7c00c | 759 | if (ret < 0) { |
ea2384d3 | 760 | bdrv_delete(bs1); |
51d7c00c | 761 | return ret; |
ea2384d3 | 762 | } |
3e82990b | 763 | total_size = bdrv_getlength(bs1) & BDRV_SECTOR_MASK; |
7c96d46e AL |
764 | |
765 | if (bs1->drv && bs1->drv->protocol_name) | |
766 | is_protocol = 1; | |
767 | ||
ea2384d3 | 768 | bdrv_delete(bs1); |
3b46e624 | 769 | |
eba25057 JM |
770 | ret = get_tmp_filename(tmp_filename, sizeof(tmp_filename)); |
771 | if (ret < 0) { | |
772 | return ret; | |
773 | } | |
7c96d46e AL |
774 | |
775 | /* Real path is meaningless for protocols */ | |
776 | if (is_protocol) | |
777 | snprintf(backing_filename, sizeof(backing_filename), | |
778 | "%s", filename); | |
114cdfa9 KS |
779 | else if (!realpath(filename, backing_filename)) |
780 | return -errno; | |
7c96d46e | 781 | |
91a073a9 KW |
782 | bdrv_qcow2 = bdrv_find_format("qcow2"); |
783 | options = parse_option_parameters("", bdrv_qcow2->create_options, NULL); | |
784 | ||
3e82990b | 785 | set_option_parameter_int(options, BLOCK_OPT_SIZE, total_size); |
91a073a9 KW |
786 | set_option_parameter(options, BLOCK_OPT_BACKING_FILE, backing_filename); |
787 | if (drv) { | |
788 | set_option_parameter(options, BLOCK_OPT_BACKING_FMT, | |
789 | drv->format_name); | |
790 | } | |
791 | ||
792 | ret = bdrv_create(bdrv_qcow2, tmp_filename, options); | |
d748768c | 793 | free_option_parameters(options); |
51d7c00c AL |
794 | if (ret < 0) { |
795 | return ret; | |
ea2384d3 | 796 | } |
91a073a9 | 797 | |
ea2384d3 | 798 | filename = tmp_filename; |
91a073a9 | 799 | drv = bdrv_qcow2; |
ea2384d3 FB |
800 | bs->is_temporary = 1; |
801 | } | |
712e7874 | 802 | |
b6ce07aa | 803 | /* Find the right image format driver */ |
6db95603 | 804 | if (!drv) { |
c98ac35d | 805 | ret = find_image_format(filename, &drv); |
51d7c00c | 806 | } |
6987307c | 807 | |
51d7c00c | 808 | if (!drv) { |
51d7c00c | 809 | goto unlink_and_fail; |
ea2384d3 | 810 | } |
b6ce07aa | 811 | |
be028adc JC |
812 | if (flags & BDRV_O_RDWR) { |
813 | flags |= BDRV_O_ALLOW_RDWR; | |
814 | } | |
815 | ||
b6ce07aa KW |
816 | /* Open the image */ |
817 | ret = bdrv_open_common(bs, filename, flags, drv); | |
818 | if (ret < 0) { | |
6987307c CH |
819 | goto unlink_and_fail; |
820 | } | |
821 | ||
b6ce07aa KW |
822 | /* If there is a backing file, use it */ |
823 | if ((flags & BDRV_O_NO_BACKING) == 0 && bs->backing_file[0] != '\0') { | |
824 | char backing_filename[PATH_MAX]; | |
825 | int back_flags; | |
826 | BlockDriver *back_drv = NULL; | |
827 | ||
828 | bs->backing_hd = bdrv_new(""); | |
dc5a1371 PB |
829 | bdrv_get_full_backing_filename(bs, backing_filename, |
830 | sizeof(backing_filename)); | |
df2dbb4a SH |
831 | |
832 | if (bs->backing_format[0] != '\0') { | |
b6ce07aa | 833 | back_drv = bdrv_find_format(bs->backing_format); |
df2dbb4a | 834 | } |
b6ce07aa KW |
835 | |
836 | /* backing files always opened read-only */ | |
837 | back_flags = | |
838 | flags & ~(BDRV_O_RDWR | BDRV_O_SNAPSHOT | BDRV_O_NO_BACKING); | |
839 | ||
840 | ret = bdrv_open(bs->backing_hd, backing_filename, back_flags, back_drv); | |
841 | if (ret < 0) { | |
842 | bdrv_close(bs); | |
843 | return ret; | |
844 | } | |
b6ce07aa KW |
845 | } |
846 | ||
847 | if (!bdrv_key_required(bs)) { | |
7d4b4ba5 | 848 | bdrv_dev_change_media_cb(bs, true); |
b6ce07aa KW |
849 | } |
850 | ||
98f90dba ZYW |
851 | /* throttling disk I/O limits */ |
852 | if (bs->io_limits_enabled) { | |
853 | bdrv_io_limits_enable(bs); | |
854 | } | |
855 | ||
b6ce07aa KW |
856 | return 0; |
857 | ||
858 | unlink_and_fail: | |
859 | if (bs->is_temporary) { | |
860 | unlink(filename); | |
861 | } | |
862 | return ret; | |
863 | } | |
864 | ||
e971aa12 JC |
865 | typedef struct BlockReopenQueueEntry { |
866 | bool prepared; | |
867 | BDRVReopenState state; | |
868 | QSIMPLEQ_ENTRY(BlockReopenQueueEntry) entry; | |
869 | } BlockReopenQueueEntry; | |
870 | ||
871 | /* | |
872 | * Adds a BlockDriverState to a simple queue for an atomic, transactional | |
873 | * reopen of multiple devices. | |
874 | * | |
875 | * bs_queue can either be an existing BlockReopenQueue that has had QSIMPLE_INIT | |
876 | * already performed, or alternatively may be NULL a new BlockReopenQueue will | |
877 | * be created and initialized. This newly created BlockReopenQueue should be | |
878 | * passed back in for subsequent calls that are intended to be of the same | |
879 | * atomic 'set'. | |
880 | * | |
881 | * bs is the BlockDriverState to add to the reopen queue. | |
882 | * | |
883 | * flags contains the open flags for the associated bs | |
884 | * | |
885 | * returns a pointer to bs_queue, which is either the newly allocated | |
886 | * bs_queue, or the existing bs_queue being used. | |
887 | * | |
888 | */ | |
889 | BlockReopenQueue *bdrv_reopen_queue(BlockReopenQueue *bs_queue, | |
890 | BlockDriverState *bs, int flags) | |
891 | { | |
892 | assert(bs != NULL); | |
893 | ||
894 | BlockReopenQueueEntry *bs_entry; | |
895 | if (bs_queue == NULL) { | |
896 | bs_queue = g_new0(BlockReopenQueue, 1); | |
897 | QSIMPLEQ_INIT(bs_queue); | |
898 | } | |
899 | ||
900 | if (bs->file) { | |
901 | bdrv_reopen_queue(bs_queue, bs->file, flags); | |
902 | } | |
903 | ||
904 | bs_entry = g_new0(BlockReopenQueueEntry, 1); | |
905 | QSIMPLEQ_INSERT_TAIL(bs_queue, bs_entry, entry); | |
906 | ||
907 | bs_entry->state.bs = bs; | |
908 | bs_entry->state.flags = flags; | |
909 | ||
910 | return bs_queue; | |
911 | } | |
912 | ||
913 | /* | |
914 | * Reopen multiple BlockDriverStates atomically & transactionally. | |
915 | * | |
916 | * The queue passed in (bs_queue) must have been built up previous | |
917 | * via bdrv_reopen_queue(). | |
918 | * | |
919 | * Reopens all BDS specified in the queue, with the appropriate | |
920 | * flags. All devices are prepared for reopen, and failure of any | |
921 | * device will cause all device changes to be abandonded, and intermediate | |
922 | * data cleaned up. | |
923 | * | |
924 | * If all devices prepare successfully, then the changes are committed | |
925 | * to all devices. | |
926 | * | |
927 | */ | |
928 | int bdrv_reopen_multiple(BlockReopenQueue *bs_queue, Error **errp) | |
929 | { | |
930 | int ret = -1; | |
931 | BlockReopenQueueEntry *bs_entry, *next; | |
932 | Error *local_err = NULL; | |
933 | ||
934 | assert(bs_queue != NULL); | |
935 | ||
936 | bdrv_drain_all(); | |
937 | ||
938 | QSIMPLEQ_FOREACH(bs_entry, bs_queue, entry) { | |
939 | if (bdrv_reopen_prepare(&bs_entry->state, bs_queue, &local_err)) { | |
940 | error_propagate(errp, local_err); | |
941 | goto cleanup; | |
942 | } | |
943 | bs_entry->prepared = true; | |
944 | } | |
945 | ||
946 | /* If we reach this point, we have success and just need to apply the | |
947 | * changes | |
948 | */ | |
949 | QSIMPLEQ_FOREACH(bs_entry, bs_queue, entry) { | |
950 | bdrv_reopen_commit(&bs_entry->state); | |
951 | } | |
952 | ||
953 | ret = 0; | |
954 | ||
955 | cleanup: | |
956 | QSIMPLEQ_FOREACH_SAFE(bs_entry, bs_queue, entry, next) { | |
957 | if (ret && bs_entry->prepared) { | |
958 | bdrv_reopen_abort(&bs_entry->state); | |
959 | } | |
960 | g_free(bs_entry); | |
961 | } | |
962 | g_free(bs_queue); | |
963 | return ret; | |
964 | } | |
965 | ||
966 | ||
967 | /* Reopen a single BlockDriverState with the specified flags. */ | |
968 | int bdrv_reopen(BlockDriverState *bs, int bdrv_flags, Error **errp) | |
969 | { | |
970 | int ret = -1; | |
971 | Error *local_err = NULL; | |
972 | BlockReopenQueue *queue = bdrv_reopen_queue(NULL, bs, bdrv_flags); | |
973 | ||
974 | ret = bdrv_reopen_multiple(queue, &local_err); | |
975 | if (local_err != NULL) { | |
976 | error_propagate(errp, local_err); | |
977 | } | |
978 | return ret; | |
979 | } | |
980 | ||
981 | ||
982 | /* | |
983 | * Prepares a BlockDriverState for reopen. All changes are staged in the | |
984 | * 'opaque' field of the BDRVReopenState, which is used and allocated by | |
985 | * the block driver layer .bdrv_reopen_prepare() | |
986 | * | |
987 | * bs is the BlockDriverState to reopen | |
988 | * flags are the new open flags | |
989 | * queue is the reopen queue | |
990 | * | |
991 | * Returns 0 on success, non-zero on error. On error errp will be set | |
992 | * as well. | |
993 | * | |
994 | * On failure, bdrv_reopen_abort() will be called to clean up any data. | |
995 | * It is the responsibility of the caller to then call the abort() or | |
996 | * commit() for any other BDS that have been left in a prepare() state | |
997 | * | |
998 | */ | |
999 | int bdrv_reopen_prepare(BDRVReopenState *reopen_state, BlockReopenQueue *queue, | |
1000 | Error **errp) | |
1001 | { | |
1002 | int ret = -1; | |
1003 | Error *local_err = NULL; | |
1004 | BlockDriver *drv; | |
1005 | ||
1006 | assert(reopen_state != NULL); | |
1007 | assert(reopen_state->bs->drv != NULL); | |
1008 | drv = reopen_state->bs->drv; | |
1009 | ||
1010 | /* if we are to stay read-only, do not allow permission change | |
1011 | * to r/w */ | |
1012 | if (!(reopen_state->bs->open_flags & BDRV_O_ALLOW_RDWR) && | |
1013 | reopen_state->flags & BDRV_O_RDWR) { | |
1014 | error_set(errp, QERR_DEVICE_IS_READ_ONLY, | |
1015 | reopen_state->bs->device_name); | |
1016 | goto error; | |
1017 | } | |
1018 | ||
1019 | ||
1020 | ret = bdrv_flush(reopen_state->bs); | |
1021 | if (ret) { | |
1022 | error_set(errp, ERROR_CLASS_GENERIC_ERROR, "Error (%s) flushing drive", | |
1023 | strerror(-ret)); | |
1024 | goto error; | |
1025 | } | |
1026 | ||
1027 | if (drv->bdrv_reopen_prepare) { | |
1028 | ret = drv->bdrv_reopen_prepare(reopen_state, queue, &local_err); | |
1029 | if (ret) { | |
1030 | if (local_err != NULL) { | |
1031 | error_propagate(errp, local_err); | |
1032 | } else { | |
1033 | error_set(errp, QERR_OPEN_FILE_FAILED, | |
1034 | reopen_state->bs->filename); | |
1035 | } | |
1036 | goto error; | |
1037 | } | |
1038 | } else { | |
1039 | /* It is currently mandatory to have a bdrv_reopen_prepare() | |
1040 | * handler for each supported drv. */ | |
1041 | error_set(errp, QERR_BLOCK_FORMAT_FEATURE_NOT_SUPPORTED, | |
1042 | drv->format_name, reopen_state->bs->device_name, | |
1043 | "reopening of file"); | |
1044 | ret = -1; | |
1045 | goto error; | |
1046 | } | |
1047 | ||
1048 | ret = 0; | |
1049 | ||
1050 | error: | |
1051 | return ret; | |
1052 | } | |
1053 | ||
1054 | /* | |
1055 | * Takes the staged changes for the reopen from bdrv_reopen_prepare(), and | |
1056 | * makes them final by swapping the staging BlockDriverState contents into | |
1057 | * the active BlockDriverState contents. | |
1058 | */ | |
1059 | void bdrv_reopen_commit(BDRVReopenState *reopen_state) | |
1060 | { | |
1061 | BlockDriver *drv; | |
1062 | ||
1063 | assert(reopen_state != NULL); | |
1064 | drv = reopen_state->bs->drv; | |
1065 | assert(drv != NULL); | |
1066 | ||
1067 | /* If there are any driver level actions to take */ | |
1068 | if (drv->bdrv_reopen_commit) { | |
1069 | drv->bdrv_reopen_commit(reopen_state); | |
1070 | } | |
1071 | ||
1072 | /* set BDS specific flags now */ | |
1073 | reopen_state->bs->open_flags = reopen_state->flags; | |
1074 | reopen_state->bs->enable_write_cache = !!(reopen_state->flags & | |
1075 | BDRV_O_CACHE_WB); | |
1076 | reopen_state->bs->read_only = !(reopen_state->flags & BDRV_O_RDWR); | |
1077 | } | |
1078 | ||
1079 | /* | |
1080 | * Abort the reopen, and delete and free the staged changes in | |
1081 | * reopen_state | |
1082 | */ | |
1083 | void bdrv_reopen_abort(BDRVReopenState *reopen_state) | |
1084 | { | |
1085 | BlockDriver *drv; | |
1086 | ||
1087 | assert(reopen_state != NULL); | |
1088 | drv = reopen_state->bs->drv; | |
1089 | assert(drv != NULL); | |
1090 | ||
1091 | if (drv->bdrv_reopen_abort) { | |
1092 | drv->bdrv_reopen_abort(reopen_state); | |
1093 | } | |
1094 | } | |
1095 | ||
1096 | ||
fc01f7e7 FB |
1097 | void bdrv_close(BlockDriverState *bs) |
1098 | { | |
80ccf93b | 1099 | bdrv_flush(bs); |
19cb3738 | 1100 | if (bs->drv) { |
3e914655 PB |
1101 | if (bs->job) { |
1102 | block_job_cancel_sync(bs->job); | |
1103 | } | |
7094f12f KW |
1104 | bdrv_drain_all(); |
1105 | ||
f9092b10 MA |
1106 | if (bs == bs_snapshots) { |
1107 | bs_snapshots = NULL; | |
1108 | } | |
557df6ac | 1109 | if (bs->backing_hd) { |
ea2384d3 | 1110 | bdrv_delete(bs->backing_hd); |
557df6ac SH |
1111 | bs->backing_hd = NULL; |
1112 | } | |
ea2384d3 | 1113 | bs->drv->bdrv_close(bs); |
7267c094 | 1114 | g_free(bs->opaque); |
ea2384d3 FB |
1115 | #ifdef _WIN32 |
1116 | if (bs->is_temporary) { | |
1117 | unlink(bs->filename); | |
1118 | } | |
67b915a5 | 1119 | #endif |
ea2384d3 FB |
1120 | bs->opaque = NULL; |
1121 | bs->drv = NULL; | |
53fec9d3 | 1122 | bs->copy_on_read = 0; |
a275fa42 PB |
1123 | bs->backing_file[0] = '\0'; |
1124 | bs->backing_format[0] = '\0'; | |
6405875c PB |
1125 | bs->total_sectors = 0; |
1126 | bs->encrypted = 0; | |
1127 | bs->valid_key = 0; | |
1128 | bs->sg = 0; | |
1129 | bs->growable = 0; | |
b338082b | 1130 | |
66f82cee | 1131 | if (bs->file != NULL) { |
0ac9377d PB |
1132 | bdrv_delete(bs->file); |
1133 | bs->file = NULL; | |
66f82cee | 1134 | } |
b338082b | 1135 | } |
98f90dba | 1136 | |
9ca11154 PH |
1137 | bdrv_dev_change_media_cb(bs, false); |
1138 | ||
98f90dba ZYW |
1139 | /*throttling disk I/O limits*/ |
1140 | if (bs->io_limits_enabled) { | |
1141 | bdrv_io_limits_disable(bs); | |
1142 | } | |
b338082b FB |
1143 | } |
1144 | ||
2bc93fed MK |
1145 | void bdrv_close_all(void) |
1146 | { | |
1147 | BlockDriverState *bs; | |
1148 | ||
1149 | QTAILQ_FOREACH(bs, &bdrv_states, list) { | |
1150 | bdrv_close(bs); | |
1151 | } | |
1152 | } | |
1153 | ||
922453bc SH |
1154 | /* |
1155 | * Wait for pending requests to complete across all BlockDriverStates | |
1156 | * | |
1157 | * This function does not flush data to disk, use bdrv_flush_all() for that | |
1158 | * after calling this function. | |
4c355d53 ZYW |
1159 | * |
1160 | * Note that completion of an asynchronous I/O operation can trigger any | |
1161 | * number of other I/O operations on other devices---for example a coroutine | |
1162 | * can be arbitrarily complex and a constant flow of I/O can come until the | |
1163 | * coroutine is complete. Because of this, it is not possible to have a | |
1164 | * function to drain a single device's I/O queue. | |
922453bc SH |
1165 | */ |
1166 | void bdrv_drain_all(void) | |
1167 | { | |
1168 | BlockDriverState *bs; | |
4c355d53 ZYW |
1169 | bool busy; |
1170 | ||
1171 | do { | |
1172 | busy = qemu_aio_wait(); | |
922453bc | 1173 | |
4c355d53 ZYW |
1174 | /* FIXME: We do not have timer support here, so this is effectively |
1175 | * a busy wait. | |
1176 | */ | |
1177 | QTAILQ_FOREACH(bs, &bdrv_states, list) { | |
1178 | if (!qemu_co_queue_empty(&bs->throttled_reqs)) { | |
1179 | qemu_co_queue_restart_all(&bs->throttled_reqs); | |
1180 | busy = true; | |
1181 | } | |
1182 | } | |
1183 | } while (busy); | |
922453bc SH |
1184 | |
1185 | /* If requests are still pending there is a bug somewhere */ | |
1186 | QTAILQ_FOREACH(bs, &bdrv_states, list) { | |
1187 | assert(QLIST_EMPTY(&bs->tracked_requests)); | |
1188 | assert(qemu_co_queue_empty(&bs->throttled_reqs)); | |
1189 | } | |
1190 | } | |
1191 | ||
d22b2f41 RH |
1192 | /* make a BlockDriverState anonymous by removing from bdrv_state list. |
1193 | Also, NULL terminate the device_name to prevent double remove */ | |
1194 | void bdrv_make_anon(BlockDriverState *bs) | |
1195 | { | |
1196 | if (bs->device_name[0] != '\0') { | |
1197 | QTAILQ_REMOVE(&bdrv_states, bs, list); | |
1198 | } | |
1199 | bs->device_name[0] = '\0'; | |
1200 | } | |
1201 | ||
e023b2e2 PB |
1202 | static void bdrv_rebind(BlockDriverState *bs) |
1203 | { | |
1204 | if (bs->drv && bs->drv->bdrv_rebind) { | |
1205 | bs->drv->bdrv_rebind(bs); | |
1206 | } | |
1207 | } | |
1208 | ||
4ddc07ca PB |
1209 | static void bdrv_move_feature_fields(BlockDriverState *bs_dest, |
1210 | BlockDriverState *bs_src) | |
8802d1fd | 1211 | { |
4ddc07ca PB |
1212 | /* move some fields that need to stay attached to the device */ |
1213 | bs_dest->open_flags = bs_src->open_flags; | |
8802d1fd JC |
1214 | |
1215 | /* dev info */ | |
4ddc07ca PB |
1216 | bs_dest->dev_ops = bs_src->dev_ops; |
1217 | bs_dest->dev_opaque = bs_src->dev_opaque; | |
1218 | bs_dest->dev = bs_src->dev; | |
1219 | bs_dest->buffer_alignment = bs_src->buffer_alignment; | |
1220 | bs_dest->copy_on_read = bs_src->copy_on_read; | |
8802d1fd | 1221 | |
4ddc07ca | 1222 | bs_dest->enable_write_cache = bs_src->enable_write_cache; |
c4a248a1 | 1223 | |
8802d1fd | 1224 | /* i/o timing parameters */ |
4ddc07ca PB |
1225 | bs_dest->slice_time = bs_src->slice_time; |
1226 | bs_dest->slice_start = bs_src->slice_start; | |
1227 | bs_dest->slice_end = bs_src->slice_end; | |
1228 | bs_dest->io_limits = bs_src->io_limits; | |
1229 | bs_dest->io_base = bs_src->io_base; | |
1230 | bs_dest->throttled_reqs = bs_src->throttled_reqs; | |
1231 | bs_dest->block_timer = bs_src->block_timer; | |
1232 | bs_dest->io_limits_enabled = bs_src->io_limits_enabled; | |
8802d1fd | 1233 | |
8802d1fd | 1234 | /* r/w error */ |
4ddc07ca PB |
1235 | bs_dest->on_read_error = bs_src->on_read_error; |
1236 | bs_dest->on_write_error = bs_src->on_write_error; | |
8802d1fd JC |
1237 | |
1238 | /* i/o status */ | |
4ddc07ca PB |
1239 | bs_dest->iostatus_enabled = bs_src->iostatus_enabled; |
1240 | bs_dest->iostatus = bs_src->iostatus; | |
8802d1fd | 1241 | |
a9fc4408 | 1242 | /* dirty bitmap */ |
4ddc07ca PB |
1243 | bs_dest->dirty_count = bs_src->dirty_count; |
1244 | bs_dest->dirty_bitmap = bs_src->dirty_bitmap; | |
a9fc4408 PB |
1245 | |
1246 | /* job */ | |
4ddc07ca PB |
1247 | bs_dest->in_use = bs_src->in_use; |
1248 | bs_dest->job = bs_src->job; | |
a9fc4408 | 1249 | |
8802d1fd | 1250 | /* keep the same entry in bdrv_states */ |
4ddc07ca PB |
1251 | pstrcpy(bs_dest->device_name, sizeof(bs_dest->device_name), |
1252 | bs_src->device_name); | |
1253 | bs_dest->list = bs_src->list; | |
1254 | } | |
8802d1fd | 1255 | |
4ddc07ca PB |
1256 | /* |
1257 | * Swap bs contents for two image chains while they are live, | |
1258 | * while keeping required fields on the BlockDriverState that is | |
1259 | * actually attached to a device. | |
1260 | * | |
1261 | * This will modify the BlockDriverState fields, and swap contents | |
1262 | * between bs_new and bs_old. Both bs_new and bs_old are modified. | |
1263 | * | |
1264 | * bs_new is required to be anonymous. | |
1265 | * | |
1266 | * This function does not create any image files. | |
1267 | */ | |
1268 | void bdrv_swap(BlockDriverState *bs_new, BlockDriverState *bs_old) | |
1269 | { | |
1270 | BlockDriverState tmp; | |
f6801b83 | 1271 | |
4ddc07ca PB |
1272 | /* bs_new must be anonymous and shouldn't have anything fancy enabled */ |
1273 | assert(bs_new->device_name[0] == '\0'); | |
1274 | assert(bs_new->dirty_bitmap == NULL); | |
1275 | assert(bs_new->job == NULL); | |
1276 | assert(bs_new->dev == NULL); | |
1277 | assert(bs_new->in_use == 0); | |
1278 | assert(bs_new->io_limits_enabled == false); | |
1279 | assert(bs_new->block_timer == NULL); | |
8802d1fd | 1280 | |
4ddc07ca PB |
1281 | tmp = *bs_new; |
1282 | *bs_new = *bs_old; | |
1283 | *bs_old = tmp; | |
a9fc4408 | 1284 | |
4ddc07ca PB |
1285 | /* there are some fields that should not be swapped, move them back */ |
1286 | bdrv_move_feature_fields(&tmp, bs_old); | |
1287 | bdrv_move_feature_fields(bs_old, bs_new); | |
1288 | bdrv_move_feature_fields(bs_new, &tmp); | |
8802d1fd | 1289 | |
4ddc07ca PB |
1290 | /* bs_new shouldn't be in bdrv_states even after the swap! */ |
1291 | assert(bs_new->device_name[0] == '\0'); | |
1292 | ||
1293 | /* Check a few fields that should remain attached to the device */ | |
1294 | assert(bs_new->dev == NULL); | |
1295 | assert(bs_new->job == NULL); | |
1296 | assert(bs_new->in_use == 0); | |
1297 | assert(bs_new->io_limits_enabled == false); | |
1298 | assert(bs_new->block_timer == NULL); | |
e023b2e2 PB |
1299 | |
1300 | bdrv_rebind(bs_new); | |
4ddc07ca PB |
1301 | bdrv_rebind(bs_old); |
1302 | } | |
1303 | ||
1304 | /* | |
1305 | * Add new bs contents at the top of an image chain while the chain is | |
1306 | * live, while keeping required fields on the top layer. | |
1307 | * | |
1308 | * This will modify the BlockDriverState fields, and swap contents | |
1309 | * between bs_new and bs_top. Both bs_new and bs_top are modified. | |
1310 | * | |
1311 | * bs_new is required to be anonymous. | |
1312 | * | |
1313 | * This function does not create any image files. | |
1314 | */ | |
1315 | void bdrv_append(BlockDriverState *bs_new, BlockDriverState *bs_top) | |
1316 | { | |
1317 | bdrv_swap(bs_new, bs_top); | |
1318 | ||
1319 | /* The contents of 'tmp' will become bs_top, as we are | |
1320 | * swapping bs_new and bs_top contents. */ | |
1321 | bs_top->backing_hd = bs_new; | |
1322 | bs_top->open_flags &= ~BDRV_O_NO_BACKING; | |
1323 | pstrcpy(bs_top->backing_file, sizeof(bs_top->backing_file), | |
1324 | bs_new->filename); | |
1325 | pstrcpy(bs_top->backing_format, sizeof(bs_top->backing_format), | |
1326 | bs_new->drv ? bs_new->drv->format_name : ""); | |
8802d1fd JC |
1327 | } |
1328 | ||
b338082b FB |
1329 | void bdrv_delete(BlockDriverState *bs) |
1330 | { | |
fa879d62 | 1331 | assert(!bs->dev); |
3e914655 PB |
1332 | assert(!bs->job); |
1333 | assert(!bs->in_use); | |
18846dee | 1334 | |
1b7bdbc1 | 1335 | /* remove from list, if necessary */ |
d22b2f41 | 1336 | bdrv_make_anon(bs); |
34c6f050 | 1337 | |
b338082b | 1338 | bdrv_close(bs); |
66f82cee | 1339 | |
f9092b10 | 1340 | assert(bs != bs_snapshots); |
7267c094 | 1341 | g_free(bs); |
fc01f7e7 FB |
1342 | } |
1343 | ||
fa879d62 MA |
1344 | int bdrv_attach_dev(BlockDriverState *bs, void *dev) |
1345 | /* TODO change to DeviceState *dev when all users are qdevified */ | |
18846dee | 1346 | { |
fa879d62 | 1347 | if (bs->dev) { |
18846dee MA |
1348 | return -EBUSY; |
1349 | } | |
fa879d62 | 1350 | bs->dev = dev; |
28a7282a | 1351 | bdrv_iostatus_reset(bs); |
18846dee MA |
1352 | return 0; |
1353 | } | |
1354 | ||
fa879d62 MA |
1355 | /* TODO qdevified devices don't use this, remove when devices are qdevified */ |
1356 | void bdrv_attach_dev_nofail(BlockDriverState *bs, void *dev) | |
18846dee | 1357 | { |
fa879d62 MA |
1358 | if (bdrv_attach_dev(bs, dev) < 0) { |
1359 | abort(); | |
1360 | } | |
1361 | } | |
1362 | ||
1363 | void bdrv_detach_dev(BlockDriverState *bs, void *dev) | |
1364 | /* TODO change to DeviceState *dev when all users are qdevified */ | |
1365 | { | |
1366 | assert(bs->dev == dev); | |
1367 | bs->dev = NULL; | |
0e49de52 MA |
1368 | bs->dev_ops = NULL; |
1369 | bs->dev_opaque = NULL; | |
29e05f20 | 1370 | bs->buffer_alignment = 512; |
18846dee MA |
1371 | } |
1372 | ||
fa879d62 MA |
1373 | /* TODO change to return DeviceState * when all users are qdevified */ |
1374 | void *bdrv_get_attached_dev(BlockDriverState *bs) | |
18846dee | 1375 | { |
fa879d62 | 1376 | return bs->dev; |
18846dee MA |
1377 | } |
1378 | ||
0e49de52 MA |
1379 | void bdrv_set_dev_ops(BlockDriverState *bs, const BlockDevOps *ops, |
1380 | void *opaque) | |
1381 | { | |
1382 | bs->dev_ops = ops; | |
1383 | bs->dev_opaque = opaque; | |
2c6942fa MA |
1384 | if (bdrv_dev_has_removable_media(bs) && bs == bs_snapshots) { |
1385 | bs_snapshots = NULL; | |
1386 | } | |
0e49de52 MA |
1387 | } |
1388 | ||
329c0a48 LC |
1389 | void bdrv_emit_qmp_error_event(const BlockDriverState *bdrv, |
1390 | BlockQMPEventAction action, int is_read) | |
1391 | { | |
1392 | QObject *data; | |
1393 | const char *action_str; | |
1394 | ||
1395 | switch (action) { | |
1396 | case BDRV_ACTION_REPORT: | |
1397 | action_str = "report"; | |
1398 | break; | |
1399 | case BDRV_ACTION_IGNORE: | |
1400 | action_str = "ignore"; | |
1401 | break; | |
1402 | case BDRV_ACTION_STOP: | |
1403 | action_str = "stop"; | |
1404 | break; | |
1405 | default: | |
1406 | abort(); | |
1407 | } | |
1408 | ||
1409 | data = qobject_from_jsonf("{ 'device': %s, 'action': %s, 'operation': %s }", | |
1410 | bdrv->device_name, | |
1411 | action_str, | |
1412 | is_read ? "read" : "write"); | |
1413 | monitor_protocol_event(QEVENT_BLOCK_IO_ERROR, data); | |
1414 | ||
1415 | qobject_decref(data); | |
1416 | } | |
1417 | ||
6f382ed2 LC |
1418 | static void bdrv_emit_qmp_eject_event(BlockDriverState *bs, bool ejected) |
1419 | { | |
1420 | QObject *data; | |
1421 | ||
1422 | data = qobject_from_jsonf("{ 'device': %s, 'tray-open': %i }", | |
1423 | bdrv_get_device_name(bs), ejected); | |
1424 | monitor_protocol_event(QEVENT_DEVICE_TRAY_MOVED, data); | |
1425 | ||
1426 | qobject_decref(data); | |
1427 | } | |
1428 | ||
7d4b4ba5 | 1429 | static void bdrv_dev_change_media_cb(BlockDriverState *bs, bool load) |
0e49de52 | 1430 | { |
145feb17 | 1431 | if (bs->dev_ops && bs->dev_ops->change_media_cb) { |
6f382ed2 | 1432 | bool tray_was_closed = !bdrv_dev_is_tray_open(bs); |
7d4b4ba5 | 1433 | bs->dev_ops->change_media_cb(bs->dev_opaque, load); |
6f382ed2 LC |
1434 | if (tray_was_closed) { |
1435 | /* tray open */ | |
1436 | bdrv_emit_qmp_eject_event(bs, true); | |
1437 | } | |
1438 | if (load) { | |
1439 | /* tray close */ | |
1440 | bdrv_emit_qmp_eject_event(bs, false); | |
1441 | } | |
145feb17 MA |
1442 | } |
1443 | } | |
1444 | ||
2c6942fa MA |
1445 | bool bdrv_dev_has_removable_media(BlockDriverState *bs) |
1446 | { | |
1447 | return !bs->dev || (bs->dev_ops && bs->dev_ops->change_media_cb); | |
1448 | } | |
1449 | ||
025ccaa7 PB |
1450 | void bdrv_dev_eject_request(BlockDriverState *bs, bool force) |
1451 | { | |
1452 | if (bs->dev_ops && bs->dev_ops->eject_request_cb) { | |
1453 | bs->dev_ops->eject_request_cb(bs->dev_opaque, force); | |
1454 | } | |
1455 | } | |
1456 | ||
e4def80b MA |
1457 | bool bdrv_dev_is_tray_open(BlockDriverState *bs) |
1458 | { | |
1459 | if (bs->dev_ops && bs->dev_ops->is_tray_open) { | |
1460 | return bs->dev_ops->is_tray_open(bs->dev_opaque); | |
1461 | } | |
1462 | return false; | |
1463 | } | |
1464 | ||
145feb17 MA |
1465 | static void bdrv_dev_resize_cb(BlockDriverState *bs) |
1466 | { | |
1467 | if (bs->dev_ops && bs->dev_ops->resize_cb) { | |
1468 | bs->dev_ops->resize_cb(bs->dev_opaque); | |
0e49de52 MA |
1469 | } |
1470 | } | |
1471 | ||
f107639a MA |
1472 | bool bdrv_dev_is_medium_locked(BlockDriverState *bs) |
1473 | { | |
1474 | if (bs->dev_ops && bs->dev_ops->is_medium_locked) { | |
1475 | return bs->dev_ops->is_medium_locked(bs->dev_opaque); | |
1476 | } | |
1477 | return false; | |
1478 | } | |
1479 | ||
e97fc193 AL |
1480 | /* |
1481 | * Run consistency checks on an image | |
1482 | * | |
e076f338 | 1483 | * Returns 0 if the check could be completed (it doesn't mean that the image is |
a1c7273b | 1484 | * free of errors) or -errno when an internal error occurred. The results of the |
e076f338 | 1485 | * check are stored in res. |
e97fc193 | 1486 | */ |
4534ff54 | 1487 | int bdrv_check(BlockDriverState *bs, BdrvCheckResult *res, BdrvCheckMode fix) |
e97fc193 AL |
1488 | { |
1489 | if (bs->drv->bdrv_check == NULL) { | |
1490 | return -ENOTSUP; | |
1491 | } | |
1492 | ||
e076f338 | 1493 | memset(res, 0, sizeof(*res)); |
4534ff54 | 1494 | return bs->drv->bdrv_check(bs, res, fix); |
e97fc193 AL |
1495 | } |
1496 | ||
8a426614 KW |
1497 | #define COMMIT_BUF_SECTORS 2048 |
1498 | ||
33e3963e FB |
1499 | /* commit COW file into the raw image */ |
1500 | int bdrv_commit(BlockDriverState *bs) | |
1501 | { | |
19cb3738 | 1502 | BlockDriver *drv = bs->drv; |
8a426614 KW |
1503 | int64_t sector, total_sectors; |
1504 | int n, ro, open_flags; | |
0bce597d | 1505 | int ret = 0; |
8a426614 | 1506 | uint8_t *buf; |
4dca4b63 | 1507 | char filename[1024]; |
33e3963e | 1508 | |
19cb3738 FB |
1509 | if (!drv) |
1510 | return -ENOMEDIUM; | |
4dca4b63 NS |
1511 | |
1512 | if (!bs->backing_hd) { | |
1513 | return -ENOTSUP; | |
33e3963e FB |
1514 | } |
1515 | ||
2d3735d3 SH |
1516 | if (bdrv_in_use(bs) || bdrv_in_use(bs->backing_hd)) { |
1517 | return -EBUSY; | |
1518 | } | |
1519 | ||
4dca4b63 NS |
1520 | ro = bs->backing_hd->read_only; |
1521 | strncpy(filename, bs->backing_hd->filename, sizeof(filename)); | |
1522 | open_flags = bs->backing_hd->open_flags; | |
1523 | ||
1524 | if (ro) { | |
0bce597d JC |
1525 | if (bdrv_reopen(bs->backing_hd, open_flags | BDRV_O_RDWR, NULL)) { |
1526 | return -EACCES; | |
4dca4b63 | 1527 | } |
ea2384d3 | 1528 | } |
33e3963e | 1529 | |
6ea44308 | 1530 | total_sectors = bdrv_getlength(bs) >> BDRV_SECTOR_BITS; |
7267c094 | 1531 | buf = g_malloc(COMMIT_BUF_SECTORS * BDRV_SECTOR_SIZE); |
8a426614 KW |
1532 | |
1533 | for (sector = 0; sector < total_sectors; sector += n) { | |
05c4af54 | 1534 | if (bdrv_is_allocated(bs, sector, COMMIT_BUF_SECTORS, &n)) { |
8a426614 KW |
1535 | |
1536 | if (bdrv_read(bs, sector, buf, n) != 0) { | |
1537 | ret = -EIO; | |
1538 | goto ro_cleanup; | |
1539 | } | |
1540 | ||
1541 | if (bdrv_write(bs->backing_hd, sector, buf, n) != 0) { | |
1542 | ret = -EIO; | |
1543 | goto ro_cleanup; | |
1544 | } | |
ea2384d3 | 1545 | } |
33e3963e | 1546 | } |
95389c86 | 1547 | |
1d44952f CH |
1548 | if (drv->bdrv_make_empty) { |
1549 | ret = drv->bdrv_make_empty(bs); | |
1550 | bdrv_flush(bs); | |
1551 | } | |
95389c86 | 1552 | |
3f5075ae CH |
1553 | /* |
1554 | * Make sure all data we wrote to the backing device is actually | |
1555 | * stable on disk. | |
1556 | */ | |
1557 | if (bs->backing_hd) | |
1558 | bdrv_flush(bs->backing_hd); | |
4dca4b63 NS |
1559 | |
1560 | ro_cleanup: | |
7267c094 | 1561 | g_free(buf); |
4dca4b63 NS |
1562 | |
1563 | if (ro) { | |
0bce597d JC |
1564 | /* ignoring error return here */ |
1565 | bdrv_reopen(bs->backing_hd, open_flags & ~BDRV_O_RDWR, NULL); | |
4dca4b63 NS |
1566 | } |
1567 | ||
1d44952f | 1568 | return ret; |
33e3963e FB |
1569 | } |
1570 | ||
e8877497 | 1571 | int bdrv_commit_all(void) |
6ab4b5ab MA |
1572 | { |
1573 | BlockDriverState *bs; | |
1574 | ||
1575 | QTAILQ_FOREACH(bs, &bdrv_states, list) { | |
e8877497 SH |
1576 | int ret = bdrv_commit(bs); |
1577 | if (ret < 0) { | |
1578 | return ret; | |
1579 | } | |
6ab4b5ab | 1580 | } |
e8877497 | 1581 | return 0; |
6ab4b5ab MA |
1582 | } |
1583 | ||
dbffbdcf SH |
1584 | struct BdrvTrackedRequest { |
1585 | BlockDriverState *bs; | |
1586 | int64_t sector_num; | |
1587 | int nb_sectors; | |
1588 | bool is_write; | |
1589 | QLIST_ENTRY(BdrvTrackedRequest) list; | |
5f8b6491 | 1590 | Coroutine *co; /* owner, used for deadlock detection */ |
f4658285 | 1591 | CoQueue wait_queue; /* coroutines blocked on this request */ |
dbffbdcf SH |
1592 | }; |
1593 | ||
1594 | /** | |
1595 | * Remove an active request from the tracked requests list | |
1596 | * | |
1597 | * This function should be called when a tracked request is completing. | |
1598 | */ | |
1599 | static void tracked_request_end(BdrvTrackedRequest *req) | |
1600 | { | |
1601 | QLIST_REMOVE(req, list); | |
f4658285 | 1602 | qemu_co_queue_restart_all(&req->wait_queue); |
dbffbdcf SH |
1603 | } |
1604 | ||
1605 | /** | |
1606 | * Add an active request to the tracked requests list | |
1607 | */ | |
1608 | static void tracked_request_begin(BdrvTrackedRequest *req, | |
1609 | BlockDriverState *bs, | |
1610 | int64_t sector_num, | |
1611 | int nb_sectors, bool is_write) | |
1612 | { | |
1613 | *req = (BdrvTrackedRequest){ | |
1614 | .bs = bs, | |
1615 | .sector_num = sector_num, | |
1616 | .nb_sectors = nb_sectors, | |
1617 | .is_write = is_write, | |
5f8b6491 | 1618 | .co = qemu_coroutine_self(), |
dbffbdcf SH |
1619 | }; |
1620 | ||
f4658285 SH |
1621 | qemu_co_queue_init(&req->wait_queue); |
1622 | ||
dbffbdcf SH |
1623 | QLIST_INSERT_HEAD(&bs->tracked_requests, req, list); |
1624 | } | |
1625 | ||
d83947ac SH |
1626 | /** |
1627 | * Round a region to cluster boundaries | |
1628 | */ | |
1629 | static void round_to_clusters(BlockDriverState *bs, | |
1630 | int64_t sector_num, int nb_sectors, | |
1631 | int64_t *cluster_sector_num, | |
1632 | int *cluster_nb_sectors) | |
1633 | { | |
1634 | BlockDriverInfo bdi; | |
1635 | ||
1636 | if (bdrv_get_info(bs, &bdi) < 0 || bdi.cluster_size == 0) { | |
1637 | *cluster_sector_num = sector_num; | |
1638 | *cluster_nb_sectors = nb_sectors; | |
1639 | } else { | |
1640 | int64_t c = bdi.cluster_size / BDRV_SECTOR_SIZE; | |
1641 | *cluster_sector_num = QEMU_ALIGN_DOWN(sector_num, c); | |
1642 | *cluster_nb_sectors = QEMU_ALIGN_UP(sector_num - *cluster_sector_num + | |
1643 | nb_sectors, c); | |
1644 | } | |
1645 | } | |
1646 | ||
f4658285 SH |
1647 | static bool tracked_request_overlaps(BdrvTrackedRequest *req, |
1648 | int64_t sector_num, int nb_sectors) { | |
d83947ac SH |
1649 | /* aaaa bbbb */ |
1650 | if (sector_num >= req->sector_num + req->nb_sectors) { | |
1651 | return false; | |
1652 | } | |
1653 | /* bbbb aaaa */ | |
1654 | if (req->sector_num >= sector_num + nb_sectors) { | |
1655 | return false; | |
1656 | } | |
1657 | return true; | |
f4658285 SH |
1658 | } |
1659 | ||
1660 | static void coroutine_fn wait_for_overlapping_requests(BlockDriverState *bs, | |
1661 | int64_t sector_num, int nb_sectors) | |
1662 | { | |
1663 | BdrvTrackedRequest *req; | |
d83947ac SH |
1664 | int64_t cluster_sector_num; |
1665 | int cluster_nb_sectors; | |
f4658285 SH |
1666 | bool retry; |
1667 | ||
d83947ac SH |
1668 | /* If we touch the same cluster it counts as an overlap. This guarantees |
1669 | * that allocating writes will be serialized and not race with each other | |
1670 | * for the same cluster. For example, in copy-on-read it ensures that the | |
1671 | * CoR read and write operations are atomic and guest writes cannot | |
1672 | * interleave between them. | |
1673 | */ | |
1674 | round_to_clusters(bs, sector_num, nb_sectors, | |
1675 | &cluster_sector_num, &cluster_nb_sectors); | |
1676 | ||
f4658285 SH |
1677 | do { |
1678 | retry = false; | |
1679 | QLIST_FOREACH(req, &bs->tracked_requests, list) { | |
d83947ac SH |
1680 | if (tracked_request_overlaps(req, cluster_sector_num, |
1681 | cluster_nb_sectors)) { | |
5f8b6491 SH |
1682 | /* Hitting this means there was a reentrant request, for |
1683 | * example, a block driver issuing nested requests. This must | |
1684 | * never happen since it means deadlock. | |
1685 | */ | |
1686 | assert(qemu_coroutine_self() != req->co); | |
1687 | ||
f4658285 SH |
1688 | qemu_co_queue_wait(&req->wait_queue); |
1689 | retry = true; | |
1690 | break; | |
1691 | } | |
1692 | } | |
1693 | } while (retry); | |
1694 | } | |
1695 | ||
756e6736 KW |
1696 | /* |
1697 | * Return values: | |
1698 | * 0 - success | |
1699 | * -EINVAL - backing format specified, but no file | |
1700 | * -ENOSPC - can't update the backing file because no space is left in the | |
1701 | * image file header | |
1702 | * -ENOTSUP - format driver doesn't support changing the backing file | |
1703 | */ | |
1704 | int bdrv_change_backing_file(BlockDriverState *bs, | |
1705 | const char *backing_file, const char *backing_fmt) | |
1706 | { | |
1707 | BlockDriver *drv = bs->drv; | |
469ef350 | 1708 | int ret; |
756e6736 | 1709 | |
5f377794 PB |
1710 | /* Backing file format doesn't make sense without a backing file */ |
1711 | if (backing_fmt && !backing_file) { | |
1712 | return -EINVAL; | |
1713 | } | |
1714 | ||
756e6736 | 1715 | if (drv->bdrv_change_backing_file != NULL) { |
469ef350 | 1716 | ret = drv->bdrv_change_backing_file(bs, backing_file, backing_fmt); |
756e6736 | 1717 | } else { |
469ef350 | 1718 | ret = -ENOTSUP; |
756e6736 | 1719 | } |
469ef350 PB |
1720 | |
1721 | if (ret == 0) { | |
1722 | pstrcpy(bs->backing_file, sizeof(bs->backing_file), backing_file ?: ""); | |
1723 | pstrcpy(bs->backing_format, sizeof(bs->backing_format), backing_fmt ?: ""); | |
1724 | } | |
1725 | return ret; | |
756e6736 KW |
1726 | } |
1727 | ||
6ebdcee2 JC |
1728 | /* |
1729 | * Finds the image layer in the chain that has 'bs' as its backing file. | |
1730 | * | |
1731 | * active is the current topmost image. | |
1732 | * | |
1733 | * Returns NULL if bs is not found in active's image chain, | |
1734 | * or if active == bs. | |
1735 | */ | |
1736 | BlockDriverState *bdrv_find_overlay(BlockDriverState *active, | |
1737 | BlockDriverState *bs) | |
1738 | { | |
1739 | BlockDriverState *overlay = NULL; | |
1740 | BlockDriverState *intermediate; | |
1741 | ||
1742 | assert(active != NULL); | |
1743 | assert(bs != NULL); | |
1744 | ||
1745 | /* if bs is the same as active, then by definition it has no overlay | |
1746 | */ | |
1747 | if (active == bs) { | |
1748 | return NULL; | |
1749 | } | |
1750 | ||
1751 | intermediate = active; | |
1752 | while (intermediate->backing_hd) { | |
1753 | if (intermediate->backing_hd == bs) { | |
1754 | overlay = intermediate; | |
1755 | break; | |
1756 | } | |
1757 | intermediate = intermediate->backing_hd; | |
1758 | } | |
1759 | ||
1760 | return overlay; | |
1761 | } | |
1762 | ||
1763 | typedef struct BlkIntermediateStates { | |
1764 | BlockDriverState *bs; | |
1765 | QSIMPLEQ_ENTRY(BlkIntermediateStates) entry; | |
1766 | } BlkIntermediateStates; | |
1767 | ||
1768 | ||
1769 | /* | |
1770 | * Drops images above 'base' up to and including 'top', and sets the image | |
1771 | * above 'top' to have base as its backing file. | |
1772 | * | |
1773 | * Requires that the overlay to 'top' is opened r/w, so that the backing file | |
1774 | * information in 'bs' can be properly updated. | |
1775 | * | |
1776 | * E.g., this will convert the following chain: | |
1777 | * bottom <- base <- intermediate <- top <- active | |
1778 | * | |
1779 | * to | |
1780 | * | |
1781 | * bottom <- base <- active | |
1782 | * | |
1783 | * It is allowed for bottom==base, in which case it converts: | |
1784 | * | |
1785 | * base <- intermediate <- top <- active | |
1786 | * | |
1787 | * to | |
1788 | * | |
1789 | * base <- active | |
1790 | * | |
1791 | * Error conditions: | |
1792 | * if active == top, that is considered an error | |
1793 | * | |
1794 | */ | |
1795 | int bdrv_drop_intermediate(BlockDriverState *active, BlockDriverState *top, | |
1796 | BlockDriverState *base) | |
1797 | { | |
1798 | BlockDriverState *intermediate; | |
1799 | BlockDriverState *base_bs = NULL; | |
1800 | BlockDriverState *new_top_bs = NULL; | |
1801 | BlkIntermediateStates *intermediate_state, *next; | |
1802 | int ret = -EIO; | |
1803 | ||
1804 | QSIMPLEQ_HEAD(states_to_delete, BlkIntermediateStates) states_to_delete; | |
1805 | QSIMPLEQ_INIT(&states_to_delete); | |
1806 | ||
1807 | if (!top->drv || !base->drv) { | |
1808 | goto exit; | |
1809 | } | |
1810 | ||
1811 | new_top_bs = bdrv_find_overlay(active, top); | |
1812 | ||
1813 | if (new_top_bs == NULL) { | |
1814 | /* we could not find the image above 'top', this is an error */ | |
1815 | goto exit; | |
1816 | } | |
1817 | ||
1818 | /* special case of new_top_bs->backing_hd already pointing to base - nothing | |
1819 | * to do, no intermediate images */ | |
1820 | if (new_top_bs->backing_hd == base) { | |
1821 | ret = 0; | |
1822 | goto exit; | |
1823 | } | |
1824 | ||
1825 | intermediate = top; | |
1826 | ||
1827 | /* now we will go down through the list, and add each BDS we find | |
1828 | * into our deletion queue, until we hit the 'base' | |
1829 | */ | |
1830 | while (intermediate) { | |
1831 | intermediate_state = g_malloc0(sizeof(BlkIntermediateStates)); | |
1832 | intermediate_state->bs = intermediate; | |
1833 | QSIMPLEQ_INSERT_TAIL(&states_to_delete, intermediate_state, entry); | |
1834 | ||
1835 | if (intermediate->backing_hd == base) { | |
1836 | base_bs = intermediate->backing_hd; | |
1837 | break; | |
1838 | } | |
1839 | intermediate = intermediate->backing_hd; | |
1840 | } | |
1841 | if (base_bs == NULL) { | |
1842 | /* something went wrong, we did not end at the base. safely | |
1843 | * unravel everything, and exit with error */ | |
1844 | goto exit; | |
1845 | } | |
1846 | ||
1847 | /* success - we can delete the intermediate states, and link top->base */ | |
1848 | ret = bdrv_change_backing_file(new_top_bs, base_bs->filename, | |
1849 | base_bs->drv ? base_bs->drv->format_name : ""); | |
1850 | if (ret) { | |
1851 | goto exit; | |
1852 | } | |
1853 | new_top_bs->backing_hd = base_bs; | |
1854 | ||
1855 | ||
1856 | QSIMPLEQ_FOREACH_SAFE(intermediate_state, &states_to_delete, entry, next) { | |
1857 | /* so that bdrv_close() does not recursively close the chain */ | |
1858 | intermediate_state->bs->backing_hd = NULL; | |
1859 | bdrv_delete(intermediate_state->bs); | |
1860 | } | |
1861 | ret = 0; | |
1862 | ||
1863 | exit: | |
1864 | QSIMPLEQ_FOREACH_SAFE(intermediate_state, &states_to_delete, entry, next) { | |
1865 | g_free(intermediate_state); | |
1866 | } | |
1867 | return ret; | |
1868 | } | |
1869 | ||
1870 | ||
71d0770c AL |
1871 | static int bdrv_check_byte_request(BlockDriverState *bs, int64_t offset, |
1872 | size_t size) | |
1873 | { | |
1874 | int64_t len; | |
1875 | ||
1876 | if (!bdrv_is_inserted(bs)) | |
1877 | return -ENOMEDIUM; | |
1878 | ||
1879 | if (bs->growable) | |
1880 | return 0; | |
1881 | ||
1882 | len = bdrv_getlength(bs); | |
1883 | ||
fbb7b4e0 KW |
1884 | if (offset < 0) |
1885 | return -EIO; | |
1886 | ||
1887 | if ((offset > len) || (len - offset < size)) | |
71d0770c AL |
1888 | return -EIO; |
1889 | ||
1890 | return 0; | |
1891 | } | |
1892 | ||
1893 | static int bdrv_check_request(BlockDriverState *bs, int64_t sector_num, | |
1894 | int nb_sectors) | |
1895 | { | |
eb5a3165 JS |
1896 | return bdrv_check_byte_request(bs, sector_num * BDRV_SECTOR_SIZE, |
1897 | nb_sectors * BDRV_SECTOR_SIZE); | |
71d0770c AL |
1898 | } |
1899 | ||
1c9805a3 SH |
1900 | typedef struct RwCo { |
1901 | BlockDriverState *bs; | |
1902 | int64_t sector_num; | |
1903 | int nb_sectors; | |
1904 | QEMUIOVector *qiov; | |
1905 | bool is_write; | |
1906 | int ret; | |
1907 | } RwCo; | |
1908 | ||
1909 | static void coroutine_fn bdrv_rw_co_entry(void *opaque) | |
fc01f7e7 | 1910 | { |
1c9805a3 | 1911 | RwCo *rwco = opaque; |
ea2384d3 | 1912 | |
1c9805a3 SH |
1913 | if (!rwco->is_write) { |
1914 | rwco->ret = bdrv_co_do_readv(rwco->bs, rwco->sector_num, | |
470c0504 | 1915 | rwco->nb_sectors, rwco->qiov, 0); |
1c9805a3 SH |
1916 | } else { |
1917 | rwco->ret = bdrv_co_do_writev(rwco->bs, rwco->sector_num, | |
f08f2dda | 1918 | rwco->nb_sectors, rwco->qiov, 0); |
1c9805a3 SH |
1919 | } |
1920 | } | |
e7a8a783 | 1921 | |
1c9805a3 SH |
1922 | /* |
1923 | * Process a synchronous request using coroutines | |
1924 | */ | |
1925 | static int bdrv_rw_co(BlockDriverState *bs, int64_t sector_num, uint8_t *buf, | |
1926 | int nb_sectors, bool is_write) | |
1927 | { | |
1928 | QEMUIOVector qiov; | |
1929 | struct iovec iov = { | |
1930 | .iov_base = (void *)buf, | |
1931 | .iov_len = nb_sectors * BDRV_SECTOR_SIZE, | |
1932 | }; | |
1933 | Coroutine *co; | |
1934 | RwCo rwco = { | |
1935 | .bs = bs, | |
1936 | .sector_num = sector_num, | |
1937 | .nb_sectors = nb_sectors, | |
1938 | .qiov = &qiov, | |
1939 | .is_write = is_write, | |
1940 | .ret = NOT_DONE, | |
1941 | }; | |
e7a8a783 | 1942 | |
1c9805a3 | 1943 | qemu_iovec_init_external(&qiov, &iov, 1); |
e7a8a783 | 1944 | |
498e386c ZYW |
1945 | /** |
1946 | * In sync call context, when the vcpu is blocked, this throttling timer | |
1947 | * will not fire; so the I/O throttling function has to be disabled here | |
1948 | * if it has been enabled. | |
1949 | */ | |
1950 | if (bs->io_limits_enabled) { | |
1951 | fprintf(stderr, "Disabling I/O throttling on '%s' due " | |
1952 | "to synchronous I/O.\n", bdrv_get_device_name(bs)); | |
1953 | bdrv_io_limits_disable(bs); | |
1954 | } | |
1955 | ||
1c9805a3 SH |
1956 | if (qemu_in_coroutine()) { |
1957 | /* Fast-path if already in coroutine context */ | |
1958 | bdrv_rw_co_entry(&rwco); | |
1959 | } else { | |
1960 | co = qemu_coroutine_create(bdrv_rw_co_entry); | |
1961 | qemu_coroutine_enter(co, &rwco); | |
1962 | while (rwco.ret == NOT_DONE) { | |
1963 | qemu_aio_wait(); | |
1964 | } | |
1965 | } | |
1966 | return rwco.ret; | |
1967 | } | |
b338082b | 1968 | |
1c9805a3 SH |
1969 | /* return < 0 if error. See bdrv_write() for the return codes */ |
1970 | int bdrv_read(BlockDriverState *bs, int64_t sector_num, | |
1971 | uint8_t *buf, int nb_sectors) | |
1972 | { | |
1973 | return bdrv_rw_co(bs, sector_num, buf, nb_sectors, false); | |
fc01f7e7 FB |
1974 | } |
1975 | ||
07d27a44 MA |
1976 | /* Just like bdrv_read(), but with I/O throttling temporarily disabled */ |
1977 | int bdrv_read_unthrottled(BlockDriverState *bs, int64_t sector_num, | |
1978 | uint8_t *buf, int nb_sectors) | |
1979 | { | |
1980 | bool enabled; | |
1981 | int ret; | |
1982 | ||
1983 | enabled = bs->io_limits_enabled; | |
1984 | bs->io_limits_enabled = false; | |
1985 | ret = bdrv_read(bs, 0, buf, 1); | |
1986 | bs->io_limits_enabled = enabled; | |
1987 | return ret; | |
1988 | } | |
1989 | ||
71df14fc PB |
1990 | #define BITS_PER_LONG (sizeof(unsigned long) * 8) |
1991 | ||
7cd1e32a | 1992 | static void set_dirty_bitmap(BlockDriverState *bs, int64_t sector_num, |
a55eb92c | 1993 | int nb_sectors, int dirty) |
7cd1e32a | 1994 | { |
1995 | int64_t start, end; | |
c6d22830 | 1996 | unsigned long val, idx, bit; |
a55eb92c | 1997 | |
6ea44308 | 1998 | start = sector_num / BDRV_SECTORS_PER_DIRTY_CHUNK; |
c6d22830 | 1999 | end = (sector_num + nb_sectors - 1) / BDRV_SECTORS_PER_DIRTY_CHUNK; |
a55eb92c JK |
2000 | |
2001 | for (; start <= end; start++) { | |
71df14fc PB |
2002 | idx = start / BITS_PER_LONG; |
2003 | bit = start % BITS_PER_LONG; | |
c6d22830 JK |
2004 | val = bs->dirty_bitmap[idx]; |
2005 | if (dirty) { | |
6d59fec1 | 2006 | if (!(val & (1UL << bit))) { |
aaa0eb75 | 2007 | bs->dirty_count++; |
6d59fec1 | 2008 | val |= 1UL << bit; |
aaa0eb75 | 2009 | } |
c6d22830 | 2010 | } else { |
6d59fec1 | 2011 | if (val & (1UL << bit)) { |
aaa0eb75 | 2012 | bs->dirty_count--; |
6d59fec1 | 2013 | val &= ~(1UL << bit); |
aaa0eb75 | 2014 | } |
c6d22830 JK |
2015 | } |
2016 | bs->dirty_bitmap[idx] = val; | |
7cd1e32a | 2017 | } |
2018 | } | |
2019 | ||
5fafdf24 | 2020 | /* Return < 0 if error. Important errors are: |
19cb3738 FB |
2021 | -EIO generic I/O error (may happen for all errors) |
2022 | -ENOMEDIUM No media inserted. | |
2023 | -EINVAL Invalid sector number or nb_sectors | |
2024 | -EACCES Trying to write a read-only device | |
2025 | */ | |
5fafdf24 | 2026 | int bdrv_write(BlockDriverState *bs, int64_t sector_num, |
fc01f7e7 FB |
2027 | const uint8_t *buf, int nb_sectors) |
2028 | { | |
1c9805a3 | 2029 | return bdrv_rw_co(bs, sector_num, (uint8_t *)buf, nb_sectors, true); |
83f64091 FB |
2030 | } |
2031 | ||
eda578e5 AL |
2032 | int bdrv_pread(BlockDriverState *bs, int64_t offset, |
2033 | void *buf, int count1) | |
83f64091 | 2034 | { |
6ea44308 | 2035 | uint8_t tmp_buf[BDRV_SECTOR_SIZE]; |
83f64091 FB |
2036 | int len, nb_sectors, count; |
2037 | int64_t sector_num; | |
9a8c4cce | 2038 | int ret; |
83f64091 FB |
2039 | |
2040 | count = count1; | |
2041 | /* first read to align to sector start */ | |
6ea44308 | 2042 | len = (BDRV_SECTOR_SIZE - offset) & (BDRV_SECTOR_SIZE - 1); |
83f64091 FB |
2043 | if (len > count) |
2044 | len = count; | |
6ea44308 | 2045 | sector_num = offset >> BDRV_SECTOR_BITS; |
83f64091 | 2046 | if (len > 0) { |
9a8c4cce KW |
2047 | if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0) |
2048 | return ret; | |
6ea44308 | 2049 | memcpy(buf, tmp_buf + (offset & (BDRV_SECTOR_SIZE - 1)), len); |
83f64091 FB |
2050 | count -= len; |
2051 | if (count == 0) | |
2052 | return count1; | |
2053 | sector_num++; | |
2054 | buf += len; | |
2055 | } | |
2056 | ||
2057 | /* read the sectors "in place" */ | |
6ea44308 | 2058 | nb_sectors = count >> BDRV_SECTOR_BITS; |
83f64091 | 2059 | if (nb_sectors > 0) { |
9a8c4cce KW |
2060 | if ((ret = bdrv_read(bs, sector_num, buf, nb_sectors)) < 0) |
2061 | return ret; | |
83f64091 | 2062 | sector_num += nb_sectors; |
6ea44308 | 2063 | len = nb_sectors << BDRV_SECTOR_BITS; |
83f64091 FB |
2064 | buf += len; |
2065 | count -= len; | |
2066 | } | |
2067 | ||
2068 | /* add data from the last sector */ | |
2069 | if (count > 0) { | |
9a8c4cce KW |
2070 | if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0) |
2071 | return ret; | |
83f64091 FB |
2072 | memcpy(buf, tmp_buf, count); |
2073 | } | |
2074 | return count1; | |
2075 | } | |
2076 | ||
eda578e5 AL |
2077 | int bdrv_pwrite(BlockDriverState *bs, int64_t offset, |
2078 | const void *buf, int count1) | |
83f64091 | 2079 | { |
6ea44308 | 2080 | uint8_t tmp_buf[BDRV_SECTOR_SIZE]; |
83f64091 FB |
2081 | int len, nb_sectors, count; |
2082 | int64_t sector_num; | |
9a8c4cce | 2083 | int ret; |
83f64091 FB |
2084 | |
2085 | count = count1; | |
2086 | /* first write to align to sector start */ | |
6ea44308 | 2087 | len = (BDRV_SECTOR_SIZE - offset) & (BDRV_SECTOR_SIZE - 1); |
83f64091 FB |
2088 | if (len > count) |
2089 | len = count; | |
6ea44308 | 2090 | sector_num = offset >> BDRV_SECTOR_BITS; |
83f64091 | 2091 | if (len > 0) { |
9a8c4cce KW |
2092 | if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0) |
2093 | return ret; | |
6ea44308 | 2094 | memcpy(tmp_buf + (offset & (BDRV_SECTOR_SIZE - 1)), buf, len); |
9a8c4cce KW |
2095 | if ((ret = bdrv_write(bs, sector_num, tmp_buf, 1)) < 0) |
2096 | return ret; | |
83f64091 FB |
2097 | count -= len; |
2098 | if (count == 0) | |
2099 | return count1; | |
2100 | sector_num++; | |
2101 | buf += len; | |
2102 | } | |
2103 | ||
2104 | /* write the sectors "in place" */ | |
6ea44308 | 2105 | nb_sectors = count >> BDRV_SECTOR_BITS; |
83f64091 | 2106 | if (nb_sectors > 0) { |
9a8c4cce KW |
2107 | if ((ret = bdrv_write(bs, sector_num, buf, nb_sectors)) < 0) |
2108 | return ret; | |
83f64091 | 2109 | sector_num += nb_sectors; |
6ea44308 | 2110 | len = nb_sectors << BDRV_SECTOR_BITS; |
83f64091 FB |
2111 | buf += len; |
2112 | count -= len; | |
2113 | } | |
2114 | ||
2115 | /* add data from the last sector */ | |
2116 | if (count > 0) { | |
9a8c4cce KW |
2117 | if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0) |
2118 | return ret; | |
83f64091 | 2119 | memcpy(tmp_buf, buf, count); |
9a8c4cce KW |
2120 | if ((ret = bdrv_write(bs, sector_num, tmp_buf, 1)) < 0) |
2121 | return ret; | |
83f64091 FB |
2122 | } |
2123 | return count1; | |
2124 | } | |
83f64091 | 2125 | |
f08145fe KW |
2126 | /* |
2127 | * Writes to the file and ensures that no writes are reordered across this | |
2128 | * request (acts as a barrier) | |
2129 | * | |
2130 | * Returns 0 on success, -errno in error cases. | |
2131 | */ | |
2132 | int bdrv_pwrite_sync(BlockDriverState *bs, int64_t offset, | |
2133 | const void *buf, int count) | |
2134 | { | |
2135 | int ret; | |
2136 | ||
2137 | ret = bdrv_pwrite(bs, offset, buf, count); | |
2138 | if (ret < 0) { | |
2139 | return ret; | |
2140 | } | |
2141 | ||
f05fa4ad PB |
2142 | /* No flush needed for cache modes that already do it */ |
2143 | if (bs->enable_write_cache) { | |
f08145fe KW |
2144 | bdrv_flush(bs); |
2145 | } | |
2146 | ||
2147 | return 0; | |
2148 | } | |
2149 | ||
470c0504 | 2150 | static int coroutine_fn bdrv_co_do_copy_on_readv(BlockDriverState *bs, |
ab185921 SH |
2151 | int64_t sector_num, int nb_sectors, QEMUIOVector *qiov) |
2152 | { | |
2153 | /* Perform I/O through a temporary buffer so that users who scribble over | |
2154 | * their read buffer while the operation is in progress do not end up | |
2155 | * modifying the image file. This is critical for zero-copy guest I/O | |
2156 | * where anything might happen inside guest memory. | |
2157 | */ | |
2158 | void *bounce_buffer; | |
2159 | ||
79c053bd | 2160 | BlockDriver *drv = bs->drv; |
ab185921 SH |
2161 | struct iovec iov; |
2162 | QEMUIOVector bounce_qiov; | |
2163 | int64_t cluster_sector_num; | |
2164 | int cluster_nb_sectors; | |
2165 | size_t skip_bytes; | |
2166 | int ret; | |
2167 | ||
2168 | /* Cover entire cluster so no additional backing file I/O is required when | |
2169 | * allocating cluster in the image file. | |
2170 | */ | |
2171 | round_to_clusters(bs, sector_num, nb_sectors, | |
2172 | &cluster_sector_num, &cluster_nb_sectors); | |
2173 | ||
470c0504 SH |
2174 | trace_bdrv_co_do_copy_on_readv(bs, sector_num, nb_sectors, |
2175 | cluster_sector_num, cluster_nb_sectors); | |
ab185921 SH |
2176 | |
2177 | iov.iov_len = cluster_nb_sectors * BDRV_SECTOR_SIZE; | |
2178 | iov.iov_base = bounce_buffer = qemu_blockalign(bs, iov.iov_len); | |
2179 | qemu_iovec_init_external(&bounce_qiov, &iov, 1); | |
2180 | ||
79c053bd SH |
2181 | ret = drv->bdrv_co_readv(bs, cluster_sector_num, cluster_nb_sectors, |
2182 | &bounce_qiov); | |
ab185921 SH |
2183 | if (ret < 0) { |
2184 | goto err; | |
2185 | } | |
2186 | ||
79c053bd SH |
2187 | if (drv->bdrv_co_write_zeroes && |
2188 | buffer_is_zero(bounce_buffer, iov.iov_len)) { | |
621f0589 KW |
2189 | ret = bdrv_co_do_write_zeroes(bs, cluster_sector_num, |
2190 | cluster_nb_sectors); | |
79c053bd | 2191 | } else { |
f05fa4ad PB |
2192 | /* This does not change the data on the disk, it is not necessary |
2193 | * to flush even in cache=writethrough mode. | |
2194 | */ | |
79c053bd | 2195 | ret = drv->bdrv_co_writev(bs, cluster_sector_num, cluster_nb_sectors, |
ab185921 | 2196 | &bounce_qiov); |
79c053bd SH |
2197 | } |
2198 | ||
ab185921 SH |
2199 | if (ret < 0) { |
2200 | /* It might be okay to ignore write errors for guest requests. If this | |
2201 | * is a deliberate copy-on-read then we don't want to ignore the error. | |
2202 | * Simply report it in all cases. | |
2203 | */ | |
2204 | goto err; | |
2205 | } | |
2206 | ||
2207 | skip_bytes = (sector_num - cluster_sector_num) * BDRV_SECTOR_SIZE; | |
03396148 MT |
2208 | qemu_iovec_from_buf(qiov, 0, bounce_buffer + skip_bytes, |
2209 | nb_sectors * BDRV_SECTOR_SIZE); | |
ab185921 SH |
2210 | |
2211 | err: | |
2212 | qemu_vfree(bounce_buffer); | |
2213 | return ret; | |
2214 | } | |
2215 | ||
c5fbe571 SH |
2216 | /* |
2217 | * Handle a read request in coroutine context | |
2218 | */ | |
2219 | static int coroutine_fn bdrv_co_do_readv(BlockDriverState *bs, | |
470c0504 SH |
2220 | int64_t sector_num, int nb_sectors, QEMUIOVector *qiov, |
2221 | BdrvRequestFlags flags) | |
da1fa91d KW |
2222 | { |
2223 | BlockDriver *drv = bs->drv; | |
dbffbdcf SH |
2224 | BdrvTrackedRequest req; |
2225 | int ret; | |
da1fa91d | 2226 | |
da1fa91d KW |
2227 | if (!drv) { |
2228 | return -ENOMEDIUM; | |
2229 | } | |
2230 | if (bdrv_check_request(bs, sector_num, nb_sectors)) { | |
2231 | return -EIO; | |
2232 | } | |
2233 | ||
98f90dba ZYW |
2234 | /* throttling disk read I/O */ |
2235 | if (bs->io_limits_enabled) { | |
2236 | bdrv_io_limits_intercept(bs, false, nb_sectors); | |
2237 | } | |
2238 | ||
f4658285 | 2239 | if (bs->copy_on_read) { |
470c0504 SH |
2240 | flags |= BDRV_REQ_COPY_ON_READ; |
2241 | } | |
2242 | if (flags & BDRV_REQ_COPY_ON_READ) { | |
2243 | bs->copy_on_read_in_flight++; | |
2244 | } | |
2245 | ||
2246 | if (bs->copy_on_read_in_flight) { | |
f4658285 SH |
2247 | wait_for_overlapping_requests(bs, sector_num, nb_sectors); |
2248 | } | |
2249 | ||
dbffbdcf | 2250 | tracked_request_begin(&req, bs, sector_num, nb_sectors, false); |
ab185921 | 2251 | |
470c0504 | 2252 | if (flags & BDRV_REQ_COPY_ON_READ) { |
ab185921 SH |
2253 | int pnum; |
2254 | ||
2255 | ret = bdrv_co_is_allocated(bs, sector_num, nb_sectors, &pnum); | |
2256 | if (ret < 0) { | |
2257 | goto out; | |
2258 | } | |
2259 | ||
2260 | if (!ret || pnum != nb_sectors) { | |
470c0504 | 2261 | ret = bdrv_co_do_copy_on_readv(bs, sector_num, nb_sectors, qiov); |
ab185921 SH |
2262 | goto out; |
2263 | } | |
2264 | } | |
2265 | ||
dbffbdcf | 2266 | ret = drv->bdrv_co_readv(bs, sector_num, nb_sectors, qiov); |
ab185921 SH |
2267 | |
2268 | out: | |
dbffbdcf | 2269 | tracked_request_end(&req); |
470c0504 SH |
2270 | |
2271 | if (flags & BDRV_REQ_COPY_ON_READ) { | |
2272 | bs->copy_on_read_in_flight--; | |
2273 | } | |
2274 | ||
dbffbdcf | 2275 | return ret; |
da1fa91d KW |
2276 | } |
2277 | ||
c5fbe571 | 2278 | int coroutine_fn bdrv_co_readv(BlockDriverState *bs, int64_t sector_num, |
da1fa91d KW |
2279 | int nb_sectors, QEMUIOVector *qiov) |
2280 | { | |
c5fbe571 | 2281 | trace_bdrv_co_readv(bs, sector_num, nb_sectors); |
da1fa91d | 2282 | |
470c0504 SH |
2283 | return bdrv_co_do_readv(bs, sector_num, nb_sectors, qiov, 0); |
2284 | } | |
2285 | ||
2286 | int coroutine_fn bdrv_co_copy_on_readv(BlockDriverState *bs, | |
2287 | int64_t sector_num, int nb_sectors, QEMUIOVector *qiov) | |
2288 | { | |
2289 | trace_bdrv_co_copy_on_readv(bs, sector_num, nb_sectors); | |
2290 | ||
2291 | return bdrv_co_do_readv(bs, sector_num, nb_sectors, qiov, | |
2292 | BDRV_REQ_COPY_ON_READ); | |
c5fbe571 SH |
2293 | } |
2294 | ||
f08f2dda SH |
2295 | static int coroutine_fn bdrv_co_do_write_zeroes(BlockDriverState *bs, |
2296 | int64_t sector_num, int nb_sectors) | |
2297 | { | |
2298 | BlockDriver *drv = bs->drv; | |
2299 | QEMUIOVector qiov; | |
2300 | struct iovec iov; | |
2301 | int ret; | |
2302 | ||
621f0589 KW |
2303 | /* TODO Emulate only part of misaligned requests instead of letting block |
2304 | * drivers return -ENOTSUP and emulate everything */ | |
2305 | ||
f08f2dda SH |
2306 | /* First try the efficient write zeroes operation */ |
2307 | if (drv->bdrv_co_write_zeroes) { | |
621f0589 KW |
2308 | ret = drv->bdrv_co_write_zeroes(bs, sector_num, nb_sectors); |
2309 | if (ret != -ENOTSUP) { | |
2310 | return ret; | |
2311 | } | |
f08f2dda SH |
2312 | } |
2313 | ||
2314 | /* Fall back to bounce buffer if write zeroes is unsupported */ | |
2315 | iov.iov_len = nb_sectors * BDRV_SECTOR_SIZE; | |
2316 | iov.iov_base = qemu_blockalign(bs, iov.iov_len); | |
2317 | memset(iov.iov_base, 0, iov.iov_len); | |
2318 | qemu_iovec_init_external(&qiov, &iov, 1); | |
2319 | ||
2320 | ret = drv->bdrv_co_writev(bs, sector_num, nb_sectors, &qiov); | |
2321 | ||
2322 | qemu_vfree(iov.iov_base); | |
2323 | return ret; | |
2324 | } | |
2325 | ||
c5fbe571 SH |
2326 | /* |
2327 | * Handle a write request in coroutine context | |
2328 | */ | |
2329 | static int coroutine_fn bdrv_co_do_writev(BlockDriverState *bs, | |
f08f2dda SH |
2330 | int64_t sector_num, int nb_sectors, QEMUIOVector *qiov, |
2331 | BdrvRequestFlags flags) | |
c5fbe571 SH |
2332 | { |
2333 | BlockDriver *drv = bs->drv; | |
dbffbdcf | 2334 | BdrvTrackedRequest req; |
6b7cb247 | 2335 | int ret; |
da1fa91d KW |
2336 | |
2337 | if (!bs->drv) { | |
2338 | return -ENOMEDIUM; | |
2339 | } | |
2340 | if (bs->read_only) { | |
2341 | return -EACCES; | |
2342 | } | |
2343 | if (bdrv_check_request(bs, sector_num, nb_sectors)) { | |
2344 | return -EIO; | |
2345 | } | |
2346 | ||
98f90dba ZYW |
2347 | /* throttling disk write I/O */ |
2348 | if (bs->io_limits_enabled) { | |
2349 | bdrv_io_limits_intercept(bs, true, nb_sectors); | |
2350 | } | |
2351 | ||
470c0504 | 2352 | if (bs->copy_on_read_in_flight) { |
f4658285 SH |
2353 | wait_for_overlapping_requests(bs, sector_num, nb_sectors); |
2354 | } | |
2355 | ||
dbffbdcf SH |
2356 | tracked_request_begin(&req, bs, sector_num, nb_sectors, true); |
2357 | ||
f08f2dda SH |
2358 | if (flags & BDRV_REQ_ZERO_WRITE) { |
2359 | ret = bdrv_co_do_write_zeroes(bs, sector_num, nb_sectors); | |
2360 | } else { | |
2361 | ret = drv->bdrv_co_writev(bs, sector_num, nb_sectors, qiov); | |
2362 | } | |
6b7cb247 | 2363 | |
f05fa4ad PB |
2364 | if (ret == 0 && !bs->enable_write_cache) { |
2365 | ret = bdrv_co_flush(bs); | |
2366 | } | |
2367 | ||
da1fa91d KW |
2368 | if (bs->dirty_bitmap) { |
2369 | set_dirty_bitmap(bs, sector_num, nb_sectors, 1); | |
2370 | } | |
2371 | ||
2372 | if (bs->wr_highest_sector < sector_num + nb_sectors - 1) { | |
2373 | bs->wr_highest_sector = sector_num + nb_sectors - 1; | |
2374 | } | |
2375 | ||
dbffbdcf SH |
2376 | tracked_request_end(&req); |
2377 | ||
6b7cb247 | 2378 | return ret; |
da1fa91d KW |
2379 | } |
2380 | ||
c5fbe571 SH |
2381 | int coroutine_fn bdrv_co_writev(BlockDriverState *bs, int64_t sector_num, |
2382 | int nb_sectors, QEMUIOVector *qiov) | |
2383 | { | |
2384 | trace_bdrv_co_writev(bs, sector_num, nb_sectors); | |
2385 | ||
f08f2dda SH |
2386 | return bdrv_co_do_writev(bs, sector_num, nb_sectors, qiov, 0); |
2387 | } | |
2388 | ||
2389 | int coroutine_fn bdrv_co_write_zeroes(BlockDriverState *bs, | |
2390 | int64_t sector_num, int nb_sectors) | |
2391 | { | |
2392 | trace_bdrv_co_write_zeroes(bs, sector_num, nb_sectors); | |
2393 | ||
2394 | return bdrv_co_do_writev(bs, sector_num, nb_sectors, NULL, | |
2395 | BDRV_REQ_ZERO_WRITE); | |
c5fbe571 SH |
2396 | } |
2397 | ||
83f64091 FB |
2398 | /** |
2399 | * Truncate file to 'offset' bytes (needed only for file protocols) | |
2400 | */ | |
2401 | int bdrv_truncate(BlockDriverState *bs, int64_t offset) | |
2402 | { | |
2403 | BlockDriver *drv = bs->drv; | |
51762288 | 2404 | int ret; |
83f64091 | 2405 | if (!drv) |
19cb3738 | 2406 | return -ENOMEDIUM; |
83f64091 FB |
2407 | if (!drv->bdrv_truncate) |
2408 | return -ENOTSUP; | |
59f2689d NS |
2409 | if (bs->read_only) |
2410 | return -EACCES; | |
8591675f MT |
2411 | if (bdrv_in_use(bs)) |
2412 | return -EBUSY; | |
51762288 SH |
2413 | ret = drv->bdrv_truncate(bs, offset); |
2414 | if (ret == 0) { | |
2415 | ret = refresh_total_sectors(bs, offset >> BDRV_SECTOR_BITS); | |
145feb17 | 2416 | bdrv_dev_resize_cb(bs); |
51762288 SH |
2417 | } |
2418 | return ret; | |
83f64091 FB |
2419 | } |
2420 | ||
4a1d5e1f FZ |
2421 | /** |
2422 | * Length of a allocated file in bytes. Sparse files are counted by actual | |
2423 | * allocated space. Return < 0 if error or unknown. | |
2424 | */ | |
2425 | int64_t bdrv_get_allocated_file_size(BlockDriverState *bs) | |
2426 | { | |
2427 | BlockDriver *drv = bs->drv; | |
2428 | if (!drv) { | |
2429 | return -ENOMEDIUM; | |
2430 | } | |
2431 | if (drv->bdrv_get_allocated_file_size) { | |
2432 | return drv->bdrv_get_allocated_file_size(bs); | |
2433 | } | |
2434 | if (bs->file) { | |
2435 | return bdrv_get_allocated_file_size(bs->file); | |
2436 | } | |
2437 | return -ENOTSUP; | |
2438 | } | |
2439 | ||
83f64091 FB |
2440 | /** |
2441 | * Length of a file in bytes. Return < 0 if error or unknown. | |
2442 | */ | |
2443 | int64_t bdrv_getlength(BlockDriverState *bs) | |
2444 | { | |
2445 | BlockDriver *drv = bs->drv; | |
2446 | if (!drv) | |
19cb3738 | 2447 | return -ENOMEDIUM; |
51762288 | 2448 | |
2c6942fa | 2449 | if (bs->growable || bdrv_dev_has_removable_media(bs)) { |
46a4e4e6 SH |
2450 | if (drv->bdrv_getlength) { |
2451 | return drv->bdrv_getlength(bs); | |
2452 | } | |
83f64091 | 2453 | } |
46a4e4e6 | 2454 | return bs->total_sectors * BDRV_SECTOR_SIZE; |
fc01f7e7 FB |
2455 | } |
2456 | ||
19cb3738 | 2457 | /* return 0 as number of sectors if no device present or error */ |
96b8f136 | 2458 | void bdrv_get_geometry(BlockDriverState *bs, uint64_t *nb_sectors_ptr) |
fc01f7e7 | 2459 | { |
19cb3738 FB |
2460 | int64_t length; |
2461 | length = bdrv_getlength(bs); | |
2462 | if (length < 0) | |
2463 | length = 0; | |
2464 | else | |
6ea44308 | 2465 | length = length >> BDRV_SECTOR_BITS; |
19cb3738 | 2466 | *nb_sectors_ptr = length; |
fc01f7e7 | 2467 | } |
cf98951b | 2468 | |
0563e191 ZYW |
2469 | /* throttling disk io limits */ |
2470 | void bdrv_set_io_limits(BlockDriverState *bs, | |
2471 | BlockIOLimit *io_limits) | |
2472 | { | |
2473 | bs->io_limits = *io_limits; | |
2474 | bs->io_limits_enabled = bdrv_io_limits_enabled(bs); | |
2475 | } | |
2476 | ||
abd7f68d MA |
2477 | void bdrv_set_on_error(BlockDriverState *bs, BlockErrorAction on_read_error, |
2478 | BlockErrorAction on_write_error) | |
2479 | { | |
2480 | bs->on_read_error = on_read_error; | |
2481 | bs->on_write_error = on_write_error; | |
2482 | } | |
2483 | ||
2484 | BlockErrorAction bdrv_get_on_error(BlockDriverState *bs, int is_read) | |
2485 | { | |
2486 | return is_read ? bs->on_read_error : bs->on_write_error; | |
2487 | } | |
2488 | ||
b338082b FB |
2489 | int bdrv_is_read_only(BlockDriverState *bs) |
2490 | { | |
2491 | return bs->read_only; | |
2492 | } | |
2493 | ||
985a03b0 TS |
2494 | int bdrv_is_sg(BlockDriverState *bs) |
2495 | { | |
2496 | return bs->sg; | |
2497 | } | |
2498 | ||
e900a7b7 CH |
2499 | int bdrv_enable_write_cache(BlockDriverState *bs) |
2500 | { | |
2501 | return bs->enable_write_cache; | |
2502 | } | |
2503 | ||
425b0148 PB |
2504 | void bdrv_set_enable_write_cache(BlockDriverState *bs, bool wce) |
2505 | { | |
2506 | bs->enable_write_cache = wce; | |
55b110f2 JC |
2507 | |
2508 | /* so a reopen() will preserve wce */ | |
2509 | if (wce) { | |
2510 | bs->open_flags |= BDRV_O_CACHE_WB; | |
2511 | } else { | |
2512 | bs->open_flags &= ~BDRV_O_CACHE_WB; | |
2513 | } | |
425b0148 PB |
2514 | } |
2515 | ||
ea2384d3 FB |
2516 | int bdrv_is_encrypted(BlockDriverState *bs) |
2517 | { | |
2518 | if (bs->backing_hd && bs->backing_hd->encrypted) | |
2519 | return 1; | |
2520 | return bs->encrypted; | |
2521 | } | |
2522 | ||
c0f4ce77 AL |
2523 | int bdrv_key_required(BlockDriverState *bs) |
2524 | { | |
2525 | BlockDriverState *backing_hd = bs->backing_hd; | |
2526 | ||
2527 | if (backing_hd && backing_hd->encrypted && !backing_hd->valid_key) | |
2528 | return 1; | |
2529 | return (bs->encrypted && !bs->valid_key); | |
2530 | } | |
2531 | ||
ea2384d3 FB |
2532 | int bdrv_set_key(BlockDriverState *bs, const char *key) |
2533 | { | |
2534 | int ret; | |
2535 | if (bs->backing_hd && bs->backing_hd->encrypted) { | |
2536 | ret = bdrv_set_key(bs->backing_hd, key); | |
2537 | if (ret < 0) | |
2538 | return ret; | |
2539 | if (!bs->encrypted) | |
2540 | return 0; | |
2541 | } | |
fd04a2ae SH |
2542 | if (!bs->encrypted) { |
2543 | return -EINVAL; | |
2544 | } else if (!bs->drv || !bs->drv->bdrv_set_key) { | |
2545 | return -ENOMEDIUM; | |
2546 | } | |
c0f4ce77 | 2547 | ret = bs->drv->bdrv_set_key(bs, key); |
bb5fc20f AL |
2548 | if (ret < 0) { |
2549 | bs->valid_key = 0; | |
2550 | } else if (!bs->valid_key) { | |
2551 | bs->valid_key = 1; | |
2552 | /* call the change callback now, we skipped it on open */ | |
7d4b4ba5 | 2553 | bdrv_dev_change_media_cb(bs, true); |
bb5fc20f | 2554 | } |
c0f4ce77 | 2555 | return ret; |
ea2384d3 FB |
2556 | } |
2557 | ||
f8d6bba1 | 2558 | const char *bdrv_get_format_name(BlockDriverState *bs) |
ea2384d3 | 2559 | { |
f8d6bba1 | 2560 | return bs->drv ? bs->drv->format_name : NULL; |
ea2384d3 FB |
2561 | } |
2562 | ||
5fafdf24 | 2563 | void bdrv_iterate_format(void (*it)(void *opaque, const char *name), |
ea2384d3 FB |
2564 | void *opaque) |
2565 | { | |
2566 | BlockDriver *drv; | |
2567 | ||
8a22f02a | 2568 | QLIST_FOREACH(drv, &bdrv_drivers, list) { |
ea2384d3 FB |
2569 | it(opaque, drv->format_name); |
2570 | } | |
2571 | } | |
2572 | ||
b338082b FB |
2573 | BlockDriverState *bdrv_find(const char *name) |
2574 | { | |
2575 | BlockDriverState *bs; | |
2576 | ||
1b7bdbc1 SH |
2577 | QTAILQ_FOREACH(bs, &bdrv_states, list) { |
2578 | if (!strcmp(name, bs->device_name)) { | |
b338082b | 2579 | return bs; |
1b7bdbc1 | 2580 | } |
b338082b FB |
2581 | } |
2582 | return NULL; | |
2583 | } | |
2584 | ||
2f399b0a MA |
2585 | BlockDriverState *bdrv_next(BlockDriverState *bs) |
2586 | { | |
2587 | if (!bs) { | |
2588 | return QTAILQ_FIRST(&bdrv_states); | |
2589 | } | |
2590 | return QTAILQ_NEXT(bs, list); | |
2591 | } | |
2592 | ||
51de9760 | 2593 | void bdrv_iterate(void (*it)(void *opaque, BlockDriverState *bs), void *opaque) |
81d0912d FB |
2594 | { |
2595 | BlockDriverState *bs; | |
2596 | ||
1b7bdbc1 | 2597 | QTAILQ_FOREACH(bs, &bdrv_states, list) { |
51de9760 | 2598 | it(opaque, bs); |
81d0912d FB |
2599 | } |
2600 | } | |
2601 | ||
ea2384d3 FB |
2602 | const char *bdrv_get_device_name(BlockDriverState *bs) |
2603 | { | |
2604 | return bs->device_name; | |
2605 | } | |
2606 | ||
c8433287 MA |
2607 | int bdrv_get_flags(BlockDriverState *bs) |
2608 | { | |
2609 | return bs->open_flags; | |
2610 | } | |
2611 | ||
c6ca28d6 AL |
2612 | void bdrv_flush_all(void) |
2613 | { | |
2614 | BlockDriverState *bs; | |
2615 | ||
1b7bdbc1 | 2616 | QTAILQ_FOREACH(bs, &bdrv_states, list) { |
29cdb251 | 2617 | bdrv_flush(bs); |
1b7bdbc1 | 2618 | } |
c6ca28d6 AL |
2619 | } |
2620 | ||
f2feebbd KW |
2621 | int bdrv_has_zero_init(BlockDriverState *bs) |
2622 | { | |
2623 | assert(bs->drv); | |
2624 | ||
336c1c12 KW |
2625 | if (bs->drv->bdrv_has_zero_init) { |
2626 | return bs->drv->bdrv_has_zero_init(bs); | |
f2feebbd KW |
2627 | } |
2628 | ||
2629 | return 1; | |
2630 | } | |
2631 | ||
376ae3f1 SH |
2632 | typedef struct BdrvCoIsAllocatedData { |
2633 | BlockDriverState *bs; | |
2634 | int64_t sector_num; | |
2635 | int nb_sectors; | |
2636 | int *pnum; | |
2637 | int ret; | |
2638 | bool done; | |
2639 | } BdrvCoIsAllocatedData; | |
2640 | ||
f58c7b35 TS |
2641 | /* |
2642 | * Returns true iff the specified sector is present in the disk image. Drivers | |
2643 | * not implementing the functionality are assumed to not support backing files, | |
2644 | * hence all their sectors are reported as allocated. | |
2645 | * | |
bd9533e3 SH |
2646 | * If 'sector_num' is beyond the end of the disk image the return value is 0 |
2647 | * and 'pnum' is set to 0. | |
2648 | * | |
f58c7b35 TS |
2649 | * 'pnum' is set to the number of sectors (including and immediately following |
2650 | * the specified sector) that are known to be in the same | |
2651 | * allocated/unallocated state. | |
2652 | * | |
bd9533e3 SH |
2653 | * 'nb_sectors' is the max value 'pnum' should be set to. If nb_sectors goes |
2654 | * beyond the end of the disk image it will be clamped. | |
f58c7b35 | 2655 | */ |
060f51c9 SH |
2656 | int coroutine_fn bdrv_co_is_allocated(BlockDriverState *bs, int64_t sector_num, |
2657 | int nb_sectors, int *pnum) | |
f58c7b35 | 2658 | { |
bd9533e3 SH |
2659 | int64_t n; |
2660 | ||
2661 | if (sector_num >= bs->total_sectors) { | |
2662 | *pnum = 0; | |
2663 | return 0; | |
2664 | } | |
2665 | ||
2666 | n = bs->total_sectors - sector_num; | |
2667 | if (n < nb_sectors) { | |
2668 | nb_sectors = n; | |
2669 | } | |
2670 | ||
6aebab14 | 2671 | if (!bs->drv->bdrv_co_is_allocated) { |
bd9533e3 | 2672 | *pnum = nb_sectors; |
f58c7b35 TS |
2673 | return 1; |
2674 | } | |
6aebab14 | 2675 | |
060f51c9 SH |
2676 | return bs->drv->bdrv_co_is_allocated(bs, sector_num, nb_sectors, pnum); |
2677 | } | |
2678 | ||
2679 | /* Coroutine wrapper for bdrv_is_allocated() */ | |
2680 | static void coroutine_fn bdrv_is_allocated_co_entry(void *opaque) | |
2681 | { | |
2682 | BdrvCoIsAllocatedData *data = opaque; | |
2683 | BlockDriverState *bs = data->bs; | |
2684 | ||
2685 | data->ret = bdrv_co_is_allocated(bs, data->sector_num, data->nb_sectors, | |
2686 | data->pnum); | |
2687 | data->done = true; | |
2688 | } | |
2689 | ||
2690 | /* | |
2691 | * Synchronous wrapper around bdrv_co_is_allocated(). | |
2692 | * | |
2693 | * See bdrv_co_is_allocated() for details. | |
2694 | */ | |
2695 | int bdrv_is_allocated(BlockDriverState *bs, int64_t sector_num, int nb_sectors, | |
2696 | int *pnum) | |
2697 | { | |
6aebab14 SH |
2698 | Coroutine *co; |
2699 | BdrvCoIsAllocatedData data = { | |
2700 | .bs = bs, | |
2701 | .sector_num = sector_num, | |
2702 | .nb_sectors = nb_sectors, | |
2703 | .pnum = pnum, | |
2704 | .done = false, | |
2705 | }; | |
2706 | ||
2707 | co = qemu_coroutine_create(bdrv_is_allocated_co_entry); | |
2708 | qemu_coroutine_enter(co, &data); | |
2709 | while (!data.done) { | |
2710 | qemu_aio_wait(); | |
2711 | } | |
2712 | return data.ret; | |
f58c7b35 TS |
2713 | } |
2714 | ||
188a7bbf PB |
2715 | /* |
2716 | * Given an image chain: ... -> [BASE] -> [INTER1] -> [INTER2] -> [TOP] | |
2717 | * | |
2718 | * Return true if the given sector is allocated in any image between | |
2719 | * BASE and TOP (inclusive). BASE can be NULL to check if the given | |
2720 | * sector is allocated in any image of the chain. Return false otherwise. | |
2721 | * | |
2722 | * 'pnum' is set to the number of sectors (including and immediately following | |
2723 | * the specified sector) that are known to be in the same | |
2724 | * allocated/unallocated state. | |
2725 | * | |
2726 | */ | |
2727 | int coroutine_fn bdrv_co_is_allocated_above(BlockDriverState *top, | |
2728 | BlockDriverState *base, | |
2729 | int64_t sector_num, | |
2730 | int nb_sectors, int *pnum) | |
2731 | { | |
2732 | BlockDriverState *intermediate; | |
2733 | int ret, n = nb_sectors; | |
2734 | ||
2735 | intermediate = top; | |
2736 | while (intermediate && intermediate != base) { | |
2737 | int pnum_inter; | |
2738 | ret = bdrv_co_is_allocated(intermediate, sector_num, nb_sectors, | |
2739 | &pnum_inter); | |
2740 | if (ret < 0) { | |
2741 | return ret; | |
2742 | } else if (ret) { | |
2743 | *pnum = pnum_inter; | |
2744 | return 1; | |
2745 | } | |
2746 | ||
2747 | /* | |
2748 | * [sector_num, nb_sectors] is unallocated on top but intermediate | |
2749 | * might have | |
2750 | * | |
2751 | * [sector_num+x, nr_sectors] allocated. | |
2752 | */ | |
2753 | if (n > pnum_inter) { | |
2754 | n = pnum_inter; | |
2755 | } | |
2756 | ||
2757 | intermediate = intermediate->backing_hd; | |
2758 | } | |
2759 | ||
2760 | *pnum = n; | |
2761 | return 0; | |
2762 | } | |
2763 | ||
b2023818 | 2764 | BlockInfoList *qmp_query_block(Error **errp) |
b338082b | 2765 | { |
b2023818 | 2766 | BlockInfoList *head = NULL, *cur_item = NULL; |
b338082b FB |
2767 | BlockDriverState *bs; |
2768 | ||
1b7bdbc1 | 2769 | QTAILQ_FOREACH(bs, &bdrv_states, list) { |
b2023818 | 2770 | BlockInfoList *info = g_malloc0(sizeof(*info)); |
d15e5465 | 2771 | |
b2023818 LC |
2772 | info->value = g_malloc0(sizeof(*info->value)); |
2773 | info->value->device = g_strdup(bs->device_name); | |
2774 | info->value->type = g_strdup("unknown"); | |
2775 | info->value->locked = bdrv_dev_is_medium_locked(bs); | |
2776 | info->value->removable = bdrv_dev_has_removable_media(bs); | |
d15e5465 | 2777 | |
e4def80b | 2778 | if (bdrv_dev_has_removable_media(bs)) { |
b2023818 LC |
2779 | info->value->has_tray_open = true; |
2780 | info->value->tray_open = bdrv_dev_is_tray_open(bs); | |
e4def80b | 2781 | } |
f04ef601 LC |
2782 | |
2783 | if (bdrv_iostatus_is_enabled(bs)) { | |
b2023818 LC |
2784 | info->value->has_io_status = true; |
2785 | info->value->io_status = bs->iostatus; | |
f04ef601 LC |
2786 | } |
2787 | ||
19cb3738 | 2788 | if (bs->drv) { |
b2023818 LC |
2789 | info->value->has_inserted = true; |
2790 | info->value->inserted = g_malloc0(sizeof(*info->value->inserted)); | |
2791 | info->value->inserted->file = g_strdup(bs->filename); | |
2792 | info->value->inserted->ro = bs->read_only; | |
2793 | info->value->inserted->drv = g_strdup(bs->drv->format_name); | |
2794 | info->value->inserted->encrypted = bs->encrypted; | |
c75a1a8a | 2795 | info->value->inserted->encryption_key_missing = bdrv_key_required(bs); |
b2023818 LC |
2796 | if (bs->backing_file[0]) { |
2797 | info->value->inserted->has_backing_file = true; | |
2798 | info->value->inserted->backing_file = g_strdup(bs->backing_file); | |
376253ec | 2799 | } |
727f005e | 2800 | |
2e3e3317 BC |
2801 | info->value->inserted->backing_file_depth = |
2802 | bdrv_get_backing_file_depth(bs); | |
2803 | ||
727f005e ZYW |
2804 | if (bs->io_limits_enabled) { |
2805 | info->value->inserted->bps = | |
2806 | bs->io_limits.bps[BLOCK_IO_LIMIT_TOTAL]; | |
2807 | info->value->inserted->bps_rd = | |
2808 | bs->io_limits.bps[BLOCK_IO_LIMIT_READ]; | |
2809 | info->value->inserted->bps_wr = | |
2810 | bs->io_limits.bps[BLOCK_IO_LIMIT_WRITE]; | |
2811 | info->value->inserted->iops = | |
2812 | bs->io_limits.iops[BLOCK_IO_LIMIT_TOTAL]; | |
2813 | info->value->inserted->iops_rd = | |
2814 | bs->io_limits.iops[BLOCK_IO_LIMIT_READ]; | |
2815 | info->value->inserted->iops_wr = | |
2816 | bs->io_limits.iops[BLOCK_IO_LIMIT_WRITE]; | |
2817 | } | |
b2023818 | 2818 | } |
d15e5465 | 2819 | |
b2023818 LC |
2820 | /* XXX: waiting for the qapi to support GSList */ |
2821 | if (!cur_item) { | |
2822 | head = cur_item = info; | |
2823 | } else { | |
2824 | cur_item->next = info; | |
2825 | cur_item = info; | |
b338082b | 2826 | } |
b338082b | 2827 | } |
d15e5465 | 2828 | |
b2023818 | 2829 | return head; |
b338082b | 2830 | } |
a36e69dd | 2831 | |
f11f57e4 LC |
2832 | /* Consider exposing this as a full fledged QMP command */ |
2833 | static BlockStats *qmp_query_blockstat(const BlockDriverState *bs, Error **errp) | |
2834 | { | |
2835 | BlockStats *s; | |
2836 | ||
2837 | s = g_malloc0(sizeof(*s)); | |
2838 | ||
2839 | if (bs->device_name[0]) { | |
2840 | s->has_device = true; | |
2841 | s->device = g_strdup(bs->device_name); | |
294cc35f KW |
2842 | } |
2843 | ||
f11f57e4 LC |
2844 | s->stats = g_malloc0(sizeof(*s->stats)); |
2845 | s->stats->rd_bytes = bs->nr_bytes[BDRV_ACCT_READ]; | |
2846 | s->stats->wr_bytes = bs->nr_bytes[BDRV_ACCT_WRITE]; | |
2847 | s->stats->rd_operations = bs->nr_ops[BDRV_ACCT_READ]; | |
2848 | s->stats->wr_operations = bs->nr_ops[BDRV_ACCT_WRITE]; | |
2849 | s->stats->wr_highest_offset = bs->wr_highest_sector * BDRV_SECTOR_SIZE; | |
2850 | s->stats->flush_operations = bs->nr_ops[BDRV_ACCT_FLUSH]; | |
2851 | s->stats->wr_total_time_ns = bs->total_time_ns[BDRV_ACCT_WRITE]; | |
2852 | s->stats->rd_total_time_ns = bs->total_time_ns[BDRV_ACCT_READ]; | |
2853 | s->stats->flush_total_time_ns = bs->total_time_ns[BDRV_ACCT_FLUSH]; | |
2854 | ||
294cc35f | 2855 | if (bs->file) { |
f11f57e4 LC |
2856 | s->has_parent = true; |
2857 | s->parent = qmp_query_blockstat(bs->file, NULL); | |
294cc35f KW |
2858 | } |
2859 | ||
f11f57e4 | 2860 | return s; |
294cc35f KW |
2861 | } |
2862 | ||
f11f57e4 | 2863 | BlockStatsList *qmp_query_blockstats(Error **errp) |
218a536a | 2864 | { |
f11f57e4 | 2865 | BlockStatsList *head = NULL, *cur_item = NULL; |
a36e69dd TS |
2866 | BlockDriverState *bs; |
2867 | ||
1b7bdbc1 | 2868 | QTAILQ_FOREACH(bs, &bdrv_states, list) { |
f11f57e4 LC |
2869 | BlockStatsList *info = g_malloc0(sizeof(*info)); |
2870 | info->value = qmp_query_blockstat(bs, NULL); | |
2871 | ||
2872 | /* XXX: waiting for the qapi to support GSList */ | |
2873 | if (!cur_item) { | |
2874 | head = cur_item = info; | |
2875 | } else { | |
2876 | cur_item->next = info; | |
2877 | cur_item = info; | |
2878 | } | |
a36e69dd | 2879 | } |
218a536a | 2880 | |
f11f57e4 | 2881 | return head; |
a36e69dd | 2882 | } |
ea2384d3 | 2883 | |
045df330 AL |
2884 | const char *bdrv_get_encrypted_filename(BlockDriverState *bs) |
2885 | { | |
2886 | if (bs->backing_hd && bs->backing_hd->encrypted) | |
2887 | return bs->backing_file; | |
2888 | else if (bs->encrypted) | |
2889 | return bs->filename; | |
2890 | else | |
2891 | return NULL; | |
2892 | } | |
2893 | ||
5fafdf24 | 2894 | void bdrv_get_backing_filename(BlockDriverState *bs, |
83f64091 FB |
2895 | char *filename, int filename_size) |
2896 | { | |
3574c608 | 2897 | pstrcpy(filename, filename_size, bs->backing_file); |
83f64091 FB |
2898 | } |
2899 | ||
5fafdf24 | 2900 | int bdrv_write_compressed(BlockDriverState *bs, int64_t sector_num, |
faea38e7 FB |
2901 | const uint8_t *buf, int nb_sectors) |
2902 | { | |
2903 | BlockDriver *drv = bs->drv; | |
2904 | if (!drv) | |
19cb3738 | 2905 | return -ENOMEDIUM; |
faea38e7 FB |
2906 | if (!drv->bdrv_write_compressed) |
2907 | return -ENOTSUP; | |
fbb7b4e0 KW |
2908 | if (bdrv_check_request(bs, sector_num, nb_sectors)) |
2909 | return -EIO; | |
a55eb92c | 2910 | |
c6d22830 | 2911 | if (bs->dirty_bitmap) { |
7cd1e32a | 2912 | set_dirty_bitmap(bs, sector_num, nb_sectors, 1); |
2913 | } | |
a55eb92c | 2914 | |
faea38e7 FB |
2915 | return drv->bdrv_write_compressed(bs, sector_num, buf, nb_sectors); |
2916 | } | |
3b46e624 | 2917 | |
faea38e7 FB |
2918 | int bdrv_get_info(BlockDriverState *bs, BlockDriverInfo *bdi) |
2919 | { | |
2920 | BlockDriver *drv = bs->drv; | |
2921 | if (!drv) | |
19cb3738 | 2922 | return -ENOMEDIUM; |
faea38e7 FB |
2923 | if (!drv->bdrv_get_info) |
2924 | return -ENOTSUP; | |
2925 | memset(bdi, 0, sizeof(*bdi)); | |
2926 | return drv->bdrv_get_info(bs, bdi); | |
2927 | } | |
2928 | ||
45566e9c CH |
2929 | int bdrv_save_vmstate(BlockDriverState *bs, const uint8_t *buf, |
2930 | int64_t pos, int size) | |
178e08a5 AL |
2931 | { |
2932 | BlockDriver *drv = bs->drv; | |
2933 | if (!drv) | |
2934 | return -ENOMEDIUM; | |
7cdb1f6d MK |
2935 | if (drv->bdrv_save_vmstate) |
2936 | return drv->bdrv_save_vmstate(bs, buf, pos, size); | |
2937 | if (bs->file) | |
2938 | return bdrv_save_vmstate(bs->file, buf, pos, size); | |
2939 | return -ENOTSUP; | |
178e08a5 AL |
2940 | } |
2941 | ||
45566e9c CH |
2942 | int bdrv_load_vmstate(BlockDriverState *bs, uint8_t *buf, |
2943 | int64_t pos, int size) | |
178e08a5 AL |
2944 | { |
2945 | BlockDriver *drv = bs->drv; | |
2946 | if (!drv) | |
2947 | return -ENOMEDIUM; | |
7cdb1f6d MK |
2948 | if (drv->bdrv_load_vmstate) |
2949 | return drv->bdrv_load_vmstate(bs, buf, pos, size); | |
2950 | if (bs->file) | |
2951 | return bdrv_load_vmstate(bs->file, buf, pos, size); | |
2952 | return -ENOTSUP; | |
178e08a5 AL |
2953 | } |
2954 | ||
8b9b0cc2 KW |
2955 | void bdrv_debug_event(BlockDriverState *bs, BlkDebugEvent event) |
2956 | { | |
2957 | BlockDriver *drv = bs->drv; | |
2958 | ||
2959 | if (!drv || !drv->bdrv_debug_event) { | |
2960 | return; | |
2961 | } | |
2962 | ||
0ed8b6f6 | 2963 | drv->bdrv_debug_event(bs, event); |
8b9b0cc2 KW |
2964 | |
2965 | } | |
2966 | ||
faea38e7 FB |
2967 | /**************************************************************/ |
2968 | /* handling of snapshots */ | |
2969 | ||
feeee5ac MDCF |
2970 | int bdrv_can_snapshot(BlockDriverState *bs) |
2971 | { | |
2972 | BlockDriver *drv = bs->drv; | |
07b70bfb | 2973 | if (!drv || !bdrv_is_inserted(bs) || bdrv_is_read_only(bs)) { |
feeee5ac MDCF |
2974 | return 0; |
2975 | } | |
2976 | ||
2977 | if (!drv->bdrv_snapshot_create) { | |
2978 | if (bs->file != NULL) { | |
2979 | return bdrv_can_snapshot(bs->file); | |
2980 | } | |
2981 | return 0; | |
2982 | } | |
2983 | ||
2984 | return 1; | |
2985 | } | |
2986 | ||
199630b6 BS |
2987 | int bdrv_is_snapshot(BlockDriverState *bs) |
2988 | { | |
2989 | return !!(bs->open_flags & BDRV_O_SNAPSHOT); | |
2990 | } | |
2991 | ||
f9092b10 MA |
2992 | BlockDriverState *bdrv_snapshots(void) |
2993 | { | |
2994 | BlockDriverState *bs; | |
2995 | ||
3ac906f7 | 2996 | if (bs_snapshots) { |
f9092b10 | 2997 | return bs_snapshots; |
3ac906f7 | 2998 | } |
f9092b10 MA |
2999 | |
3000 | bs = NULL; | |
3001 | while ((bs = bdrv_next(bs))) { | |
3002 | if (bdrv_can_snapshot(bs)) { | |
3ac906f7 MA |
3003 | bs_snapshots = bs; |
3004 | return bs; | |
f9092b10 MA |
3005 | } |
3006 | } | |
3007 | return NULL; | |
f9092b10 MA |
3008 | } |
3009 | ||
5fafdf24 | 3010 | int bdrv_snapshot_create(BlockDriverState *bs, |
faea38e7 FB |
3011 | QEMUSnapshotInfo *sn_info) |
3012 | { | |
3013 | BlockDriver *drv = bs->drv; | |
3014 | if (!drv) | |
19cb3738 | 3015 | return -ENOMEDIUM; |
7cdb1f6d MK |
3016 | if (drv->bdrv_snapshot_create) |
3017 | return drv->bdrv_snapshot_create(bs, sn_info); | |
3018 | if (bs->file) | |
3019 | return bdrv_snapshot_create(bs->file, sn_info); | |
3020 | return -ENOTSUP; | |
faea38e7 FB |
3021 | } |
3022 | ||
5fafdf24 | 3023 | int bdrv_snapshot_goto(BlockDriverState *bs, |
faea38e7 FB |
3024 | const char *snapshot_id) |
3025 | { | |
3026 | BlockDriver *drv = bs->drv; | |
7cdb1f6d MK |
3027 | int ret, open_ret; |
3028 | ||
faea38e7 | 3029 | if (!drv) |
19cb3738 | 3030 | return -ENOMEDIUM; |
7cdb1f6d MK |
3031 | if (drv->bdrv_snapshot_goto) |
3032 | return drv->bdrv_snapshot_goto(bs, snapshot_id); | |
3033 | ||
3034 | if (bs->file) { | |
3035 | drv->bdrv_close(bs); | |
3036 | ret = bdrv_snapshot_goto(bs->file, snapshot_id); | |
3037 | open_ret = drv->bdrv_open(bs, bs->open_flags); | |
3038 | if (open_ret < 0) { | |
3039 | bdrv_delete(bs->file); | |
3040 | bs->drv = NULL; | |
3041 | return open_ret; | |
3042 | } | |
3043 | return ret; | |
3044 | } | |
3045 | ||
3046 | return -ENOTSUP; | |
faea38e7 FB |
3047 | } |
3048 | ||
3049 | int bdrv_snapshot_delete(BlockDriverState *bs, const char *snapshot_id) | |
3050 | { | |
3051 | BlockDriver *drv = bs->drv; | |
3052 | if (!drv) | |
19cb3738 | 3053 | return -ENOMEDIUM; |
7cdb1f6d MK |
3054 | if (drv->bdrv_snapshot_delete) |
3055 | return drv->bdrv_snapshot_delete(bs, snapshot_id); | |
3056 | if (bs->file) | |
3057 | return bdrv_snapshot_delete(bs->file, snapshot_id); | |
3058 | return -ENOTSUP; | |
faea38e7 FB |
3059 | } |
3060 | ||
5fafdf24 | 3061 | int bdrv_snapshot_list(BlockDriverState *bs, |
faea38e7 FB |
3062 | QEMUSnapshotInfo **psn_info) |
3063 | { | |
3064 | BlockDriver *drv = bs->drv; | |
3065 | if (!drv) | |
19cb3738 | 3066 | return -ENOMEDIUM; |
7cdb1f6d MK |
3067 | if (drv->bdrv_snapshot_list) |
3068 | return drv->bdrv_snapshot_list(bs, psn_info); | |
3069 | if (bs->file) | |
3070 | return bdrv_snapshot_list(bs->file, psn_info); | |
3071 | return -ENOTSUP; | |
faea38e7 FB |
3072 | } |
3073 | ||
51ef6727 | 3074 | int bdrv_snapshot_load_tmp(BlockDriverState *bs, |
3075 | const char *snapshot_name) | |
3076 | { | |
3077 | BlockDriver *drv = bs->drv; | |
3078 | if (!drv) { | |
3079 | return -ENOMEDIUM; | |
3080 | } | |
3081 | if (!bs->read_only) { | |
3082 | return -EINVAL; | |
3083 | } | |
3084 | if (drv->bdrv_snapshot_load_tmp) { | |
3085 | return drv->bdrv_snapshot_load_tmp(bs, snapshot_name); | |
3086 | } | |
3087 | return -ENOTSUP; | |
3088 | } | |
3089 | ||
e8a6bb9c MT |
3090 | BlockDriverState *bdrv_find_backing_image(BlockDriverState *bs, |
3091 | const char *backing_file) | |
3092 | { | |
3093 | if (!bs->drv) { | |
3094 | return NULL; | |
3095 | } | |
3096 | ||
3097 | if (bs->backing_hd) { | |
3098 | if (strcmp(bs->backing_file, backing_file) == 0) { | |
3099 | return bs->backing_hd; | |
3100 | } else { | |
3101 | return bdrv_find_backing_image(bs->backing_hd, backing_file); | |
3102 | } | |
3103 | } | |
3104 | ||
3105 | return NULL; | |
3106 | } | |
3107 | ||
f198fd1c BC |
3108 | int bdrv_get_backing_file_depth(BlockDriverState *bs) |
3109 | { | |
3110 | if (!bs->drv) { | |
3111 | return 0; | |
3112 | } | |
3113 | ||
3114 | if (!bs->backing_hd) { | |
3115 | return 0; | |
3116 | } | |
3117 | ||
3118 | return 1 + bdrv_get_backing_file_depth(bs->backing_hd); | |
3119 | } | |
3120 | ||
79fac568 JC |
3121 | BlockDriverState *bdrv_find_base(BlockDriverState *bs) |
3122 | { | |
3123 | BlockDriverState *curr_bs = NULL; | |
3124 | ||
3125 | if (!bs) { | |
3126 | return NULL; | |
3127 | } | |
3128 | ||
3129 | curr_bs = bs; | |
3130 | ||
3131 | while (curr_bs->backing_hd) { | |
3132 | curr_bs = curr_bs->backing_hd; | |
3133 | } | |
3134 | return curr_bs; | |
3135 | } | |
3136 | ||
faea38e7 FB |
3137 | #define NB_SUFFIXES 4 |
3138 | ||
3139 | char *get_human_readable_size(char *buf, int buf_size, int64_t size) | |
3140 | { | |
3141 | static const char suffixes[NB_SUFFIXES] = "KMGT"; | |
3142 | int64_t base; | |
3143 | int i; | |
3144 | ||
3145 | if (size <= 999) { | |
3146 | snprintf(buf, buf_size, "%" PRId64, size); | |
3147 | } else { | |
3148 | base = 1024; | |
3149 | for(i = 0; i < NB_SUFFIXES; i++) { | |
3150 | if (size < (10 * base)) { | |
5fafdf24 | 3151 | snprintf(buf, buf_size, "%0.1f%c", |
faea38e7 FB |
3152 | (double)size / base, |
3153 | suffixes[i]); | |
3154 | break; | |
3155 | } else if (size < (1000 * base) || i == (NB_SUFFIXES - 1)) { | |
5fafdf24 | 3156 | snprintf(buf, buf_size, "%" PRId64 "%c", |
faea38e7 FB |
3157 | ((size + (base >> 1)) / base), |
3158 | suffixes[i]); | |
3159 | break; | |
3160 | } | |
3161 | base = base * 1024; | |
3162 | } | |
3163 | } | |
3164 | return buf; | |
3165 | } | |
3166 | ||
3167 | char *bdrv_snapshot_dump(char *buf, int buf_size, QEMUSnapshotInfo *sn) | |
3168 | { | |
3169 | char buf1[128], date_buf[128], clock_buf[128]; | |
3b9f94e1 FB |
3170 | #ifdef _WIN32 |
3171 | struct tm *ptm; | |
3172 | #else | |
faea38e7 | 3173 | struct tm tm; |
3b9f94e1 | 3174 | #endif |
faea38e7 FB |
3175 | time_t ti; |
3176 | int64_t secs; | |
3177 | ||
3178 | if (!sn) { | |
5fafdf24 TS |
3179 | snprintf(buf, buf_size, |
3180 | "%-10s%-20s%7s%20s%15s", | |
faea38e7 FB |
3181 | "ID", "TAG", "VM SIZE", "DATE", "VM CLOCK"); |
3182 | } else { | |
3183 | ti = sn->date_sec; | |
3b9f94e1 FB |
3184 | #ifdef _WIN32 |
3185 | ptm = localtime(&ti); | |
3186 | strftime(date_buf, sizeof(date_buf), | |
3187 | "%Y-%m-%d %H:%M:%S", ptm); | |
3188 | #else | |
faea38e7 FB |
3189 | localtime_r(&ti, &tm); |
3190 | strftime(date_buf, sizeof(date_buf), | |
3191 | "%Y-%m-%d %H:%M:%S", &tm); | |
3b9f94e1 | 3192 | #endif |
faea38e7 FB |
3193 | secs = sn->vm_clock_nsec / 1000000000; |
3194 | snprintf(clock_buf, sizeof(clock_buf), | |
3195 | "%02d:%02d:%02d.%03d", | |
3196 | (int)(secs / 3600), | |
3197 | (int)((secs / 60) % 60), | |
5fafdf24 | 3198 | (int)(secs % 60), |
faea38e7 FB |
3199 | (int)((sn->vm_clock_nsec / 1000000) % 1000)); |
3200 | snprintf(buf, buf_size, | |
5fafdf24 | 3201 | "%-10s%-20s%7s%20s%15s", |
faea38e7 FB |
3202 | sn->id_str, sn->name, |
3203 | get_human_readable_size(buf1, sizeof(buf1), sn->vm_state_size), | |
3204 | date_buf, | |
3205 | clock_buf); | |
3206 | } | |
3207 | return buf; | |
3208 | } | |
3209 | ||
ea2384d3 | 3210 | /**************************************************************/ |
83f64091 | 3211 | /* async I/Os */ |
ea2384d3 | 3212 | |
3b69e4b9 | 3213 | BlockDriverAIOCB *bdrv_aio_readv(BlockDriverState *bs, int64_t sector_num, |
f141eafe | 3214 | QEMUIOVector *qiov, int nb_sectors, |
3b69e4b9 | 3215 | BlockDriverCompletionFunc *cb, void *opaque) |
83f64091 | 3216 | { |
bbf0a440 SH |
3217 | trace_bdrv_aio_readv(bs, sector_num, nb_sectors, opaque); |
3218 | ||
b2a61371 | 3219 | return bdrv_co_aio_rw_vector(bs, sector_num, qiov, nb_sectors, |
8c5873d6 | 3220 | cb, opaque, false); |
ea2384d3 FB |
3221 | } |
3222 | ||
f141eafe AL |
3223 | BlockDriverAIOCB *bdrv_aio_writev(BlockDriverState *bs, int64_t sector_num, |
3224 | QEMUIOVector *qiov, int nb_sectors, | |
3225 | BlockDriverCompletionFunc *cb, void *opaque) | |
ea2384d3 | 3226 | { |
bbf0a440 SH |
3227 | trace_bdrv_aio_writev(bs, sector_num, nb_sectors, opaque); |
3228 | ||
1a6e115b | 3229 | return bdrv_co_aio_rw_vector(bs, sector_num, qiov, nb_sectors, |
8c5873d6 | 3230 | cb, opaque, true); |
83f64091 FB |
3231 | } |
3232 | ||
40b4f539 KW |
3233 | |
3234 | typedef struct MultiwriteCB { | |
3235 | int error; | |
3236 | int num_requests; | |
3237 | int num_callbacks; | |
3238 | struct { | |
3239 | BlockDriverCompletionFunc *cb; | |
3240 | void *opaque; | |
3241 | QEMUIOVector *free_qiov; | |
40b4f539 KW |
3242 | } callbacks[]; |
3243 | } MultiwriteCB; | |
3244 | ||
3245 | static void multiwrite_user_cb(MultiwriteCB *mcb) | |
3246 | { | |
3247 | int i; | |
3248 | ||
3249 | for (i = 0; i < mcb->num_callbacks; i++) { | |
3250 | mcb->callbacks[i].cb(mcb->callbacks[i].opaque, mcb->error); | |
1e1ea48d SH |
3251 | if (mcb->callbacks[i].free_qiov) { |
3252 | qemu_iovec_destroy(mcb->callbacks[i].free_qiov); | |
3253 | } | |
7267c094 | 3254 | g_free(mcb->callbacks[i].free_qiov); |
40b4f539 KW |
3255 | } |
3256 | } | |
3257 | ||
3258 | static void multiwrite_cb(void *opaque, int ret) | |
3259 | { | |
3260 | MultiwriteCB *mcb = opaque; | |
3261 | ||
6d519a5f SH |
3262 | trace_multiwrite_cb(mcb, ret); |
3263 | ||
cb6d3ca0 | 3264 | if (ret < 0 && !mcb->error) { |
40b4f539 | 3265 | mcb->error = ret; |
40b4f539 KW |
3266 | } |
3267 | ||
3268 | mcb->num_requests--; | |
3269 | if (mcb->num_requests == 0) { | |
de189a1b | 3270 | multiwrite_user_cb(mcb); |
7267c094 | 3271 | g_free(mcb); |
40b4f539 KW |
3272 | } |
3273 | } | |
3274 | ||
3275 | static int multiwrite_req_compare(const void *a, const void *b) | |
3276 | { | |
77be4366 CH |
3277 | const BlockRequest *req1 = a, *req2 = b; |
3278 | ||
3279 | /* | |
3280 | * Note that we can't simply subtract req2->sector from req1->sector | |
3281 | * here as that could overflow the return value. | |
3282 | */ | |
3283 | if (req1->sector > req2->sector) { | |
3284 | return 1; | |
3285 | } else if (req1->sector < req2->sector) { | |
3286 | return -1; | |
3287 | } else { | |
3288 | return 0; | |
3289 | } | |
40b4f539 KW |
3290 | } |
3291 | ||
3292 | /* | |
3293 | * Takes a bunch of requests and tries to merge them. Returns the number of | |
3294 | * requests that remain after merging. | |
3295 | */ | |
3296 | static int multiwrite_merge(BlockDriverState *bs, BlockRequest *reqs, | |
3297 | int num_reqs, MultiwriteCB *mcb) | |
3298 | { | |
3299 | int i, outidx; | |
3300 | ||
3301 | // Sort requests by start sector | |
3302 | qsort(reqs, num_reqs, sizeof(*reqs), &multiwrite_req_compare); | |
3303 | ||
3304 | // Check if adjacent requests touch the same clusters. If so, combine them, | |
3305 | // filling up gaps with zero sectors. | |
3306 | outidx = 0; | |
3307 | for (i = 1; i < num_reqs; i++) { | |
3308 | int merge = 0; | |
3309 | int64_t oldreq_last = reqs[outidx].sector + reqs[outidx].nb_sectors; | |
3310 | ||
b6a127a1 | 3311 | // Handle exactly sequential writes and overlapping writes. |
40b4f539 KW |
3312 | if (reqs[i].sector <= oldreq_last) { |
3313 | merge = 1; | |
3314 | } | |
3315 | ||
e2a305fb CH |
3316 | if (reqs[outidx].qiov->niov + reqs[i].qiov->niov + 1 > IOV_MAX) { |
3317 | merge = 0; | |
3318 | } | |
3319 | ||
40b4f539 KW |
3320 | if (merge) { |
3321 | size_t size; | |
7267c094 | 3322 | QEMUIOVector *qiov = g_malloc0(sizeof(*qiov)); |
40b4f539 KW |
3323 | qemu_iovec_init(qiov, |
3324 | reqs[outidx].qiov->niov + reqs[i].qiov->niov + 1); | |
3325 | ||
3326 | // Add the first request to the merged one. If the requests are | |
3327 | // overlapping, drop the last sectors of the first request. | |
3328 | size = (reqs[i].sector - reqs[outidx].sector) << 9; | |
1b093c48 | 3329 | qemu_iovec_concat(qiov, reqs[outidx].qiov, 0, size); |
40b4f539 | 3330 | |
b6a127a1 PB |
3331 | // We should need to add any zeros between the two requests |
3332 | assert (reqs[i].sector <= oldreq_last); | |
40b4f539 KW |
3333 | |
3334 | // Add the second request | |
1b093c48 | 3335 | qemu_iovec_concat(qiov, reqs[i].qiov, 0, reqs[i].qiov->size); |
40b4f539 | 3336 | |
cbf1dff2 | 3337 | reqs[outidx].nb_sectors = qiov->size >> 9; |
40b4f539 KW |
3338 | reqs[outidx].qiov = qiov; |
3339 | ||
3340 | mcb->callbacks[i].free_qiov = reqs[outidx].qiov; | |
3341 | } else { | |
3342 | outidx++; | |
3343 | reqs[outidx].sector = reqs[i].sector; | |
3344 | reqs[outidx].nb_sectors = reqs[i].nb_sectors; | |
3345 | reqs[outidx].qiov = reqs[i].qiov; | |
3346 | } | |
3347 | } | |
3348 | ||
3349 | return outidx + 1; | |
3350 | } | |
3351 | ||
3352 | /* | |
3353 | * Submit multiple AIO write requests at once. | |
3354 | * | |
3355 | * On success, the function returns 0 and all requests in the reqs array have | |
3356 | * been submitted. In error case this function returns -1, and any of the | |
3357 | * requests may or may not be submitted yet. In particular, this means that the | |
3358 | * callback will be called for some of the requests, for others it won't. The | |
3359 | * caller must check the error field of the BlockRequest to wait for the right | |
3360 | * callbacks (if error != 0, no callback will be called). | |
3361 | * | |
3362 | * The implementation may modify the contents of the reqs array, e.g. to merge | |
3363 | * requests. However, the fields opaque and error are left unmodified as they | |
3364 | * are used to signal failure for a single request to the caller. | |
3365 | */ | |
3366 | int bdrv_aio_multiwrite(BlockDriverState *bs, BlockRequest *reqs, int num_reqs) | |
3367 | { | |
40b4f539 KW |
3368 | MultiwriteCB *mcb; |
3369 | int i; | |
3370 | ||
301db7c2 RH |
3371 | /* don't submit writes if we don't have a medium */ |
3372 | if (bs->drv == NULL) { | |
3373 | for (i = 0; i < num_reqs; i++) { | |
3374 | reqs[i].error = -ENOMEDIUM; | |
3375 | } | |
3376 | return -1; | |
3377 | } | |
3378 | ||
40b4f539 KW |
3379 | if (num_reqs == 0) { |
3380 | return 0; | |
3381 | } | |
3382 | ||
3383 | // Create MultiwriteCB structure | |
7267c094 | 3384 | mcb = g_malloc0(sizeof(*mcb) + num_reqs * sizeof(*mcb->callbacks)); |
40b4f539 KW |
3385 | mcb->num_requests = 0; |
3386 | mcb->num_callbacks = num_reqs; | |
3387 | ||
3388 | for (i = 0; i < num_reqs; i++) { | |
3389 | mcb->callbacks[i].cb = reqs[i].cb; | |
3390 | mcb->callbacks[i].opaque = reqs[i].opaque; | |
3391 | } | |
3392 | ||
3393 | // Check for mergable requests | |
3394 | num_reqs = multiwrite_merge(bs, reqs, num_reqs, mcb); | |
3395 | ||
6d519a5f SH |
3396 | trace_bdrv_aio_multiwrite(mcb, mcb->num_callbacks, num_reqs); |
3397 | ||
df9309fb PB |
3398 | /* Run the aio requests. */ |
3399 | mcb->num_requests = num_reqs; | |
40b4f539 | 3400 | for (i = 0; i < num_reqs; i++) { |
ad54ae80 | 3401 | bdrv_aio_writev(bs, reqs[i].sector, reqs[i].qiov, |
40b4f539 | 3402 | reqs[i].nb_sectors, multiwrite_cb, mcb); |
40b4f539 KW |
3403 | } |
3404 | ||
3405 | return 0; | |
40b4f539 KW |
3406 | } |
3407 | ||
83f64091 | 3408 | void bdrv_aio_cancel(BlockDriverAIOCB *acb) |
83f64091 | 3409 | { |
6bbff9a0 | 3410 | acb->pool->cancel(acb); |
83f64091 FB |
3411 | } |
3412 | ||
98f90dba ZYW |
3413 | /* block I/O throttling */ |
3414 | static bool bdrv_exceed_bps_limits(BlockDriverState *bs, int nb_sectors, | |
3415 | bool is_write, double elapsed_time, uint64_t *wait) | |
3416 | { | |
3417 | uint64_t bps_limit = 0; | |
3418 | double bytes_limit, bytes_base, bytes_res; | |
3419 | double slice_time, wait_time; | |
3420 | ||
3421 | if (bs->io_limits.bps[BLOCK_IO_LIMIT_TOTAL]) { | |
3422 | bps_limit = bs->io_limits.bps[BLOCK_IO_LIMIT_TOTAL]; | |
3423 | } else if (bs->io_limits.bps[is_write]) { | |
3424 | bps_limit = bs->io_limits.bps[is_write]; | |
3425 | } else { | |
3426 | if (wait) { | |
3427 | *wait = 0; | |
3428 | } | |
3429 | ||
3430 | return false; | |
3431 | } | |
3432 | ||
3433 | slice_time = bs->slice_end - bs->slice_start; | |
3434 | slice_time /= (NANOSECONDS_PER_SECOND); | |
3435 | bytes_limit = bps_limit * slice_time; | |
3436 | bytes_base = bs->nr_bytes[is_write] - bs->io_base.bytes[is_write]; | |
3437 | if (bs->io_limits.bps[BLOCK_IO_LIMIT_TOTAL]) { | |
3438 | bytes_base += bs->nr_bytes[!is_write] - bs->io_base.bytes[!is_write]; | |
3439 | } | |
3440 | ||
3441 | /* bytes_base: the bytes of data which have been read/written; and | |
3442 | * it is obtained from the history statistic info. | |
3443 | * bytes_res: the remaining bytes of data which need to be read/written. | |
3444 | * (bytes_base + bytes_res) / bps_limit: used to calcuate | |
3445 | * the total time for completing reading/writting all data. | |
3446 | */ | |
3447 | bytes_res = (unsigned) nb_sectors * BDRV_SECTOR_SIZE; | |
3448 | ||
3449 | if (bytes_base + bytes_res <= bytes_limit) { | |
3450 | if (wait) { | |
3451 | *wait = 0; | |
3452 | } | |
3453 | ||
3454 | return false; | |
3455 | } | |
3456 | ||
3457 | /* Calc approx time to dispatch */ | |
3458 | wait_time = (bytes_base + bytes_res) / bps_limit - elapsed_time; | |
3459 | ||
3460 | /* When the I/O rate at runtime exceeds the limits, | |
3461 | * bs->slice_end need to be extended in order that the current statistic | |
3462 | * info can be kept until the timer fire, so it is increased and tuned | |
3463 | * based on the result of experiment. | |
3464 | */ | |
3465 | bs->slice_time = wait_time * BLOCK_IO_SLICE_TIME * 10; | |
3466 | bs->slice_end += bs->slice_time - 3 * BLOCK_IO_SLICE_TIME; | |
3467 | if (wait) { | |
3468 | *wait = wait_time * BLOCK_IO_SLICE_TIME * 10; | |
3469 | } | |
3470 | ||
3471 | return true; | |
3472 | } | |
3473 | ||
3474 | static bool bdrv_exceed_iops_limits(BlockDriverState *bs, bool is_write, | |
3475 | double elapsed_time, uint64_t *wait) | |
3476 | { | |
3477 | uint64_t iops_limit = 0; | |
3478 | double ios_limit, ios_base; | |
3479 | double slice_time, wait_time; | |
3480 | ||
3481 | if (bs->io_limits.iops[BLOCK_IO_LIMIT_TOTAL]) { | |
3482 | iops_limit = bs->io_limits.iops[BLOCK_IO_LIMIT_TOTAL]; | |
3483 | } else if (bs->io_limits.iops[is_write]) { | |
3484 | iops_limit = bs->io_limits.iops[is_write]; | |
3485 | } else { | |
3486 | if (wait) { | |
3487 | *wait = 0; | |
3488 | } | |
3489 | ||
3490 | return false; | |
3491 | } | |
3492 | ||
3493 | slice_time = bs->slice_end - bs->slice_start; | |
3494 | slice_time /= (NANOSECONDS_PER_SECOND); | |
3495 | ios_limit = iops_limit * slice_time; | |
3496 | ios_base = bs->nr_ops[is_write] - bs->io_base.ios[is_write]; | |
3497 | if (bs->io_limits.iops[BLOCK_IO_LIMIT_TOTAL]) { | |
3498 | ios_base += bs->nr_ops[!is_write] - bs->io_base.ios[!is_write]; | |
3499 | } | |
3500 | ||
3501 | if (ios_base + 1 <= ios_limit) { | |
3502 | if (wait) { | |
3503 | *wait = 0; | |
3504 | } | |
3505 | ||
3506 | return false; | |
3507 | } | |
3508 | ||
3509 | /* Calc approx time to dispatch */ | |
3510 | wait_time = (ios_base + 1) / iops_limit; | |
3511 | if (wait_time > elapsed_time) { | |
3512 | wait_time = wait_time - elapsed_time; | |
3513 | } else { | |
3514 | wait_time = 0; | |
3515 | } | |
3516 | ||
3517 | bs->slice_time = wait_time * BLOCK_IO_SLICE_TIME * 10; | |
3518 | bs->slice_end += bs->slice_time - 3 * BLOCK_IO_SLICE_TIME; | |
3519 | if (wait) { | |
3520 | *wait = wait_time * BLOCK_IO_SLICE_TIME * 10; | |
3521 | } | |
3522 | ||
3523 | return true; | |
3524 | } | |
3525 | ||
3526 | static bool bdrv_exceed_io_limits(BlockDriverState *bs, int nb_sectors, | |
3527 | bool is_write, int64_t *wait) | |
3528 | { | |
3529 | int64_t now, max_wait; | |
3530 | uint64_t bps_wait = 0, iops_wait = 0; | |
3531 | double elapsed_time; | |
3532 | int bps_ret, iops_ret; | |
3533 | ||
3534 | now = qemu_get_clock_ns(vm_clock); | |
3535 | if ((bs->slice_start < now) | |
3536 | && (bs->slice_end > now)) { | |
3537 | bs->slice_end = now + bs->slice_time; | |
3538 | } else { | |
3539 | bs->slice_time = 5 * BLOCK_IO_SLICE_TIME; | |
3540 | bs->slice_start = now; | |
3541 | bs->slice_end = now + bs->slice_time; | |
3542 | ||
3543 | bs->io_base.bytes[is_write] = bs->nr_bytes[is_write]; | |
3544 | bs->io_base.bytes[!is_write] = bs->nr_bytes[!is_write]; | |
3545 | ||
3546 | bs->io_base.ios[is_write] = bs->nr_ops[is_write]; | |
3547 | bs->io_base.ios[!is_write] = bs->nr_ops[!is_write]; | |
3548 | } | |
3549 | ||
3550 | elapsed_time = now - bs->slice_start; | |
3551 | elapsed_time /= (NANOSECONDS_PER_SECOND); | |
3552 | ||
3553 | bps_ret = bdrv_exceed_bps_limits(bs, nb_sectors, | |
3554 | is_write, elapsed_time, &bps_wait); | |
3555 | iops_ret = bdrv_exceed_iops_limits(bs, is_write, | |
3556 | elapsed_time, &iops_wait); | |
3557 | if (bps_ret || iops_ret) { | |
3558 | max_wait = bps_wait > iops_wait ? bps_wait : iops_wait; | |
3559 | if (wait) { | |
3560 | *wait = max_wait; | |
3561 | } | |
3562 | ||
3563 | now = qemu_get_clock_ns(vm_clock); | |
3564 | if (bs->slice_end < now + max_wait) { | |
3565 | bs->slice_end = now + max_wait; | |
3566 | } | |
3567 | ||
3568 | return true; | |
3569 | } | |
3570 | ||
3571 | if (wait) { | |
3572 | *wait = 0; | |
3573 | } | |
3574 | ||
3575 | return false; | |
3576 | } | |
ce1a14dc | 3577 | |
83f64091 FB |
3578 | /**************************************************************/ |
3579 | /* async block device emulation */ | |
3580 | ||
c16b5a2c CH |
3581 | typedef struct BlockDriverAIOCBSync { |
3582 | BlockDriverAIOCB common; | |
3583 | QEMUBH *bh; | |
3584 | int ret; | |
3585 | /* vector translation state */ | |
3586 | QEMUIOVector *qiov; | |
3587 | uint8_t *bounce; | |
3588 | int is_write; | |
3589 | } BlockDriverAIOCBSync; | |
3590 | ||
3591 | static void bdrv_aio_cancel_em(BlockDriverAIOCB *blockacb) | |
3592 | { | |
b666d239 KW |
3593 | BlockDriverAIOCBSync *acb = |
3594 | container_of(blockacb, BlockDriverAIOCBSync, common); | |
6a7ad299 | 3595 | qemu_bh_delete(acb->bh); |
36afc451 | 3596 | acb->bh = NULL; |
c16b5a2c CH |
3597 | qemu_aio_release(acb); |
3598 | } | |
3599 | ||
3600 | static AIOPool bdrv_em_aio_pool = { | |
3601 | .aiocb_size = sizeof(BlockDriverAIOCBSync), | |
3602 | .cancel = bdrv_aio_cancel_em, | |
3603 | }; | |
3604 | ||
ce1a14dc | 3605 | static void bdrv_aio_bh_cb(void *opaque) |
83f64091 | 3606 | { |
ce1a14dc | 3607 | BlockDriverAIOCBSync *acb = opaque; |
f141eafe | 3608 | |
f141eafe | 3609 | if (!acb->is_write) |
03396148 | 3610 | qemu_iovec_from_buf(acb->qiov, 0, acb->bounce, acb->qiov->size); |
ceb42de8 | 3611 | qemu_vfree(acb->bounce); |
ce1a14dc | 3612 | acb->common.cb(acb->common.opaque, acb->ret); |
6a7ad299 | 3613 | qemu_bh_delete(acb->bh); |
36afc451 | 3614 | acb->bh = NULL; |
ce1a14dc | 3615 | qemu_aio_release(acb); |
83f64091 | 3616 | } |
beac80cd | 3617 | |
f141eafe AL |
3618 | static BlockDriverAIOCB *bdrv_aio_rw_vector(BlockDriverState *bs, |
3619 | int64_t sector_num, | |
3620 | QEMUIOVector *qiov, | |
3621 | int nb_sectors, | |
3622 | BlockDriverCompletionFunc *cb, | |
3623 | void *opaque, | |
3624 | int is_write) | |
3625 | ||
83f64091 | 3626 | { |
ce1a14dc | 3627 | BlockDriverAIOCBSync *acb; |
ce1a14dc | 3628 | |
c16b5a2c | 3629 | acb = qemu_aio_get(&bdrv_em_aio_pool, bs, cb, opaque); |
f141eafe AL |
3630 | acb->is_write = is_write; |
3631 | acb->qiov = qiov; | |
e268ca52 | 3632 | acb->bounce = qemu_blockalign(bs, qiov->size); |
3f3aace8 | 3633 | acb->bh = qemu_bh_new(bdrv_aio_bh_cb, acb); |
f141eafe AL |
3634 | |
3635 | if (is_write) { | |
d5e6b161 | 3636 | qemu_iovec_to_buf(acb->qiov, 0, acb->bounce, qiov->size); |
1ed20acf | 3637 | acb->ret = bs->drv->bdrv_write(bs, sector_num, acb->bounce, nb_sectors); |
f141eafe | 3638 | } else { |
1ed20acf | 3639 | acb->ret = bs->drv->bdrv_read(bs, sector_num, acb->bounce, nb_sectors); |
f141eafe AL |
3640 | } |
3641 | ||
ce1a14dc | 3642 | qemu_bh_schedule(acb->bh); |
f141eafe | 3643 | |
ce1a14dc | 3644 | return &acb->common; |
beac80cd FB |
3645 | } |
3646 | ||
f141eafe AL |
3647 | static BlockDriverAIOCB *bdrv_aio_readv_em(BlockDriverState *bs, |
3648 | int64_t sector_num, QEMUIOVector *qiov, int nb_sectors, | |
ce1a14dc | 3649 | BlockDriverCompletionFunc *cb, void *opaque) |
beac80cd | 3650 | { |
f141eafe AL |
3651 | return bdrv_aio_rw_vector(bs, sector_num, qiov, nb_sectors, cb, opaque, 0); |
3652 | } | |
83f64091 | 3653 | |
f141eafe AL |
3654 | static BlockDriverAIOCB *bdrv_aio_writev_em(BlockDriverState *bs, |
3655 | int64_t sector_num, QEMUIOVector *qiov, int nb_sectors, | |
3656 | BlockDriverCompletionFunc *cb, void *opaque) | |
3657 | { | |
3658 | return bdrv_aio_rw_vector(bs, sector_num, qiov, nb_sectors, cb, opaque, 1); | |
beac80cd | 3659 | } |
beac80cd | 3660 | |
68485420 KW |
3661 | |
3662 | typedef struct BlockDriverAIOCBCoroutine { | |
3663 | BlockDriverAIOCB common; | |
3664 | BlockRequest req; | |
3665 | bool is_write; | |
3666 | QEMUBH* bh; | |
3667 | } BlockDriverAIOCBCoroutine; | |
3668 | ||
3669 | static void bdrv_aio_co_cancel_em(BlockDriverAIOCB *blockacb) | |
3670 | { | |
3671 | qemu_aio_flush(); | |
3672 | } | |
3673 | ||
3674 | static AIOPool bdrv_em_co_aio_pool = { | |
3675 | .aiocb_size = sizeof(BlockDriverAIOCBCoroutine), | |
3676 | .cancel = bdrv_aio_co_cancel_em, | |
3677 | }; | |
3678 | ||
35246a68 | 3679 | static void bdrv_co_em_bh(void *opaque) |
68485420 KW |
3680 | { |
3681 | BlockDriverAIOCBCoroutine *acb = opaque; | |
3682 | ||
3683 | acb->common.cb(acb->common.opaque, acb->req.error); | |
3684 | qemu_bh_delete(acb->bh); | |
3685 | qemu_aio_release(acb); | |
3686 | } | |
3687 | ||
b2a61371 SH |
3688 | /* Invoke bdrv_co_do_readv/bdrv_co_do_writev */ |
3689 | static void coroutine_fn bdrv_co_do_rw(void *opaque) | |
3690 | { | |
3691 | BlockDriverAIOCBCoroutine *acb = opaque; | |
3692 | BlockDriverState *bs = acb->common.bs; | |
3693 | ||
3694 | if (!acb->is_write) { | |
3695 | acb->req.error = bdrv_co_do_readv(bs, acb->req.sector, | |
470c0504 | 3696 | acb->req.nb_sectors, acb->req.qiov, 0); |
b2a61371 SH |
3697 | } else { |
3698 | acb->req.error = bdrv_co_do_writev(bs, acb->req.sector, | |
f08f2dda | 3699 | acb->req.nb_sectors, acb->req.qiov, 0); |
b2a61371 SH |
3700 | } |
3701 | ||
35246a68 | 3702 | acb->bh = qemu_bh_new(bdrv_co_em_bh, acb); |
b2a61371 SH |
3703 | qemu_bh_schedule(acb->bh); |
3704 | } | |
3705 | ||
68485420 KW |
3706 | static BlockDriverAIOCB *bdrv_co_aio_rw_vector(BlockDriverState *bs, |
3707 | int64_t sector_num, | |
3708 | QEMUIOVector *qiov, | |
3709 | int nb_sectors, | |
3710 | BlockDriverCompletionFunc *cb, | |
3711 | void *opaque, | |
8c5873d6 | 3712 | bool is_write) |
68485420 KW |
3713 | { |
3714 | Coroutine *co; | |
3715 | BlockDriverAIOCBCoroutine *acb; | |
3716 | ||
3717 | acb = qemu_aio_get(&bdrv_em_co_aio_pool, bs, cb, opaque); | |
3718 | acb->req.sector = sector_num; | |
3719 | acb->req.nb_sectors = nb_sectors; | |
3720 | acb->req.qiov = qiov; | |
3721 | acb->is_write = is_write; | |
3722 | ||
8c5873d6 | 3723 | co = qemu_coroutine_create(bdrv_co_do_rw); |
68485420 KW |
3724 | qemu_coroutine_enter(co, acb); |
3725 | ||
3726 | return &acb->common; | |
3727 | } | |
3728 | ||
07f07615 | 3729 | static void coroutine_fn bdrv_aio_flush_co_entry(void *opaque) |
b2e12bc6 | 3730 | { |
07f07615 PB |
3731 | BlockDriverAIOCBCoroutine *acb = opaque; |
3732 | BlockDriverState *bs = acb->common.bs; | |
b2e12bc6 | 3733 | |
07f07615 PB |
3734 | acb->req.error = bdrv_co_flush(bs); |
3735 | acb->bh = qemu_bh_new(bdrv_co_em_bh, acb); | |
b2e12bc6 | 3736 | qemu_bh_schedule(acb->bh); |
b2e12bc6 CH |
3737 | } |
3738 | ||
07f07615 | 3739 | BlockDriverAIOCB *bdrv_aio_flush(BlockDriverState *bs, |
016f5cf6 AG |
3740 | BlockDriverCompletionFunc *cb, void *opaque) |
3741 | { | |
07f07615 | 3742 | trace_bdrv_aio_flush(bs, opaque); |
016f5cf6 | 3743 | |
07f07615 PB |
3744 | Coroutine *co; |
3745 | BlockDriverAIOCBCoroutine *acb; | |
016f5cf6 | 3746 | |
07f07615 PB |
3747 | acb = qemu_aio_get(&bdrv_em_co_aio_pool, bs, cb, opaque); |
3748 | co = qemu_coroutine_create(bdrv_aio_flush_co_entry); | |
3749 | qemu_coroutine_enter(co, acb); | |
016f5cf6 | 3750 | |
016f5cf6 AG |
3751 | return &acb->common; |
3752 | } | |
3753 | ||
4265d620 PB |
3754 | static void coroutine_fn bdrv_aio_discard_co_entry(void *opaque) |
3755 | { | |
3756 | BlockDriverAIOCBCoroutine *acb = opaque; | |
3757 | BlockDriverState *bs = acb->common.bs; | |
3758 | ||
3759 | acb->req.error = bdrv_co_discard(bs, acb->req.sector, acb->req.nb_sectors); | |
3760 | acb->bh = qemu_bh_new(bdrv_co_em_bh, acb); | |
3761 | qemu_bh_schedule(acb->bh); | |
3762 | } | |
3763 | ||
3764 | BlockDriverAIOCB *bdrv_aio_discard(BlockDriverState *bs, | |
3765 | int64_t sector_num, int nb_sectors, | |
3766 | BlockDriverCompletionFunc *cb, void *opaque) | |
3767 | { | |
3768 | Coroutine *co; | |
3769 | BlockDriverAIOCBCoroutine *acb; | |
3770 | ||
3771 | trace_bdrv_aio_discard(bs, sector_num, nb_sectors, opaque); | |
3772 | ||
3773 | acb = qemu_aio_get(&bdrv_em_co_aio_pool, bs, cb, opaque); | |
3774 | acb->req.sector = sector_num; | |
3775 | acb->req.nb_sectors = nb_sectors; | |
3776 | co = qemu_coroutine_create(bdrv_aio_discard_co_entry); | |
3777 | qemu_coroutine_enter(co, acb); | |
3778 | ||
3779 | return &acb->common; | |
3780 | } | |
3781 | ||
ea2384d3 FB |
3782 | void bdrv_init(void) |
3783 | { | |
5efa9d5a | 3784 | module_call_init(MODULE_INIT_BLOCK); |
ea2384d3 | 3785 | } |
ce1a14dc | 3786 | |
eb852011 MA |
3787 | void bdrv_init_with_whitelist(void) |
3788 | { | |
3789 | use_bdrv_whitelist = 1; | |
3790 | bdrv_init(); | |
3791 | } | |
3792 | ||
c16b5a2c CH |
3793 | void *qemu_aio_get(AIOPool *pool, BlockDriverState *bs, |
3794 | BlockDriverCompletionFunc *cb, void *opaque) | |
ce1a14dc | 3795 | { |
ce1a14dc PB |
3796 | BlockDriverAIOCB *acb; |
3797 | ||
6bbff9a0 AL |
3798 | if (pool->free_aiocb) { |
3799 | acb = pool->free_aiocb; | |
3800 | pool->free_aiocb = acb->next; | |
ce1a14dc | 3801 | } else { |
7267c094 | 3802 | acb = g_malloc0(pool->aiocb_size); |
6bbff9a0 | 3803 | acb->pool = pool; |
ce1a14dc PB |
3804 | } |
3805 | acb->bs = bs; | |
3806 | acb->cb = cb; | |
3807 | acb->opaque = opaque; | |
3808 | return acb; | |
3809 | } | |
3810 | ||
3811 | void qemu_aio_release(void *p) | |
3812 | { | |
6bbff9a0 AL |
3813 | BlockDriverAIOCB *acb = (BlockDriverAIOCB *)p; |
3814 | AIOPool *pool = acb->pool; | |
3815 | acb->next = pool->free_aiocb; | |
3816 | pool->free_aiocb = acb; | |
ce1a14dc | 3817 | } |
19cb3738 | 3818 | |
f9f05dc5 KW |
3819 | /**************************************************************/ |
3820 | /* Coroutine block device emulation */ | |
3821 | ||
3822 | typedef struct CoroutineIOCompletion { | |
3823 | Coroutine *coroutine; | |
3824 | int ret; | |
3825 | } CoroutineIOCompletion; | |
3826 | ||
3827 | static void bdrv_co_io_em_complete(void *opaque, int ret) | |
3828 | { | |
3829 | CoroutineIOCompletion *co = opaque; | |
3830 | ||
3831 | co->ret = ret; | |
3832 | qemu_coroutine_enter(co->coroutine, NULL); | |
3833 | } | |
3834 | ||
3835 | static int coroutine_fn bdrv_co_io_em(BlockDriverState *bs, int64_t sector_num, | |
3836 | int nb_sectors, QEMUIOVector *iov, | |
3837 | bool is_write) | |
3838 | { | |
3839 | CoroutineIOCompletion co = { | |
3840 | .coroutine = qemu_coroutine_self(), | |
3841 | }; | |
3842 | BlockDriverAIOCB *acb; | |
3843 | ||
3844 | if (is_write) { | |
a652d160 SH |
3845 | acb = bs->drv->bdrv_aio_writev(bs, sector_num, iov, nb_sectors, |
3846 | bdrv_co_io_em_complete, &co); | |
f9f05dc5 | 3847 | } else { |
a652d160 SH |
3848 | acb = bs->drv->bdrv_aio_readv(bs, sector_num, iov, nb_sectors, |
3849 | bdrv_co_io_em_complete, &co); | |
f9f05dc5 KW |
3850 | } |
3851 | ||
59370aaa | 3852 | trace_bdrv_co_io_em(bs, sector_num, nb_sectors, is_write, acb); |
f9f05dc5 KW |
3853 | if (!acb) { |
3854 | return -EIO; | |
3855 | } | |
3856 | qemu_coroutine_yield(); | |
3857 | ||
3858 | return co.ret; | |
3859 | } | |
3860 | ||
3861 | static int coroutine_fn bdrv_co_readv_em(BlockDriverState *bs, | |
3862 | int64_t sector_num, int nb_sectors, | |
3863 | QEMUIOVector *iov) | |
3864 | { | |
3865 | return bdrv_co_io_em(bs, sector_num, nb_sectors, iov, false); | |
3866 | } | |
3867 | ||
3868 | static int coroutine_fn bdrv_co_writev_em(BlockDriverState *bs, | |
3869 | int64_t sector_num, int nb_sectors, | |
3870 | QEMUIOVector *iov) | |
3871 | { | |
3872 | return bdrv_co_io_em(bs, sector_num, nb_sectors, iov, true); | |
3873 | } | |
3874 | ||
07f07615 | 3875 | static void coroutine_fn bdrv_flush_co_entry(void *opaque) |
e7a8a783 | 3876 | { |
07f07615 PB |
3877 | RwCo *rwco = opaque; |
3878 | ||
3879 | rwco->ret = bdrv_co_flush(rwco->bs); | |
3880 | } | |
3881 | ||
3882 | int coroutine_fn bdrv_co_flush(BlockDriverState *bs) | |
3883 | { | |
eb489bb1 KW |
3884 | int ret; |
3885 | ||
29cdb251 | 3886 | if (!bs || !bdrv_is_inserted(bs) || bdrv_is_read_only(bs)) { |
07f07615 | 3887 | return 0; |
eb489bb1 KW |
3888 | } |
3889 | ||
ca716364 | 3890 | /* Write back cached data to the OS even with cache=unsafe */ |
eb489bb1 KW |
3891 | if (bs->drv->bdrv_co_flush_to_os) { |
3892 | ret = bs->drv->bdrv_co_flush_to_os(bs); | |
3893 | if (ret < 0) { | |
3894 | return ret; | |
3895 | } | |
3896 | } | |
3897 | ||
ca716364 KW |
3898 | /* But don't actually force it to the disk with cache=unsafe */ |
3899 | if (bs->open_flags & BDRV_O_NO_FLUSH) { | |
d4c82329 | 3900 | goto flush_parent; |
ca716364 KW |
3901 | } |
3902 | ||
eb489bb1 | 3903 | if (bs->drv->bdrv_co_flush_to_disk) { |
29cdb251 | 3904 | ret = bs->drv->bdrv_co_flush_to_disk(bs); |
07f07615 PB |
3905 | } else if (bs->drv->bdrv_aio_flush) { |
3906 | BlockDriverAIOCB *acb; | |
3907 | CoroutineIOCompletion co = { | |
3908 | .coroutine = qemu_coroutine_self(), | |
3909 | }; | |
3910 | ||
3911 | acb = bs->drv->bdrv_aio_flush(bs, bdrv_co_io_em_complete, &co); | |
3912 | if (acb == NULL) { | |
29cdb251 | 3913 | ret = -EIO; |
07f07615 PB |
3914 | } else { |
3915 | qemu_coroutine_yield(); | |
29cdb251 | 3916 | ret = co.ret; |
07f07615 | 3917 | } |
07f07615 PB |
3918 | } else { |
3919 | /* | |
3920 | * Some block drivers always operate in either writethrough or unsafe | |
3921 | * mode and don't support bdrv_flush therefore. Usually qemu doesn't | |
3922 | * know how the server works (because the behaviour is hardcoded or | |
3923 | * depends on server-side configuration), so we can't ensure that | |
3924 | * everything is safe on disk. Returning an error doesn't work because | |
3925 | * that would break guests even if the server operates in writethrough | |
3926 | * mode. | |
3927 | * | |
3928 | * Let's hope the user knows what he's doing. | |
3929 | */ | |
29cdb251 | 3930 | ret = 0; |
07f07615 | 3931 | } |
29cdb251 PB |
3932 | if (ret < 0) { |
3933 | return ret; | |
3934 | } | |
3935 | ||
3936 | /* Now flush the underlying protocol. It will also have BDRV_O_NO_FLUSH | |
3937 | * in the case of cache=unsafe, so there are no useless flushes. | |
3938 | */ | |
d4c82329 | 3939 | flush_parent: |
29cdb251 | 3940 | return bdrv_co_flush(bs->file); |
07f07615 PB |
3941 | } |
3942 | ||
0f15423c AL |
3943 | void bdrv_invalidate_cache(BlockDriverState *bs) |
3944 | { | |
3945 | if (bs->drv && bs->drv->bdrv_invalidate_cache) { | |
3946 | bs->drv->bdrv_invalidate_cache(bs); | |
3947 | } | |
3948 | } | |
3949 | ||
3950 | void bdrv_invalidate_cache_all(void) | |
3951 | { | |
3952 | BlockDriverState *bs; | |
3953 | ||
3954 | QTAILQ_FOREACH(bs, &bdrv_states, list) { | |
3955 | bdrv_invalidate_cache(bs); | |
3956 | } | |
3957 | } | |
3958 | ||
07789269 BC |
3959 | void bdrv_clear_incoming_migration_all(void) |
3960 | { | |
3961 | BlockDriverState *bs; | |
3962 | ||
3963 | QTAILQ_FOREACH(bs, &bdrv_states, list) { | |
3964 | bs->open_flags = bs->open_flags & ~(BDRV_O_INCOMING); | |
3965 | } | |
3966 | } | |
3967 | ||
07f07615 PB |
3968 | int bdrv_flush(BlockDriverState *bs) |
3969 | { | |
3970 | Coroutine *co; | |
3971 | RwCo rwco = { | |
3972 | .bs = bs, | |
3973 | .ret = NOT_DONE, | |
e7a8a783 | 3974 | }; |
e7a8a783 | 3975 | |
07f07615 PB |
3976 | if (qemu_in_coroutine()) { |
3977 | /* Fast-path if already in coroutine context */ | |
3978 | bdrv_flush_co_entry(&rwco); | |
3979 | } else { | |
3980 | co = qemu_coroutine_create(bdrv_flush_co_entry); | |
3981 | qemu_coroutine_enter(co, &rwco); | |
3982 | while (rwco.ret == NOT_DONE) { | |
3983 | qemu_aio_wait(); | |
3984 | } | |
e7a8a783 | 3985 | } |
07f07615 PB |
3986 | |
3987 | return rwco.ret; | |
e7a8a783 KW |
3988 | } |
3989 | ||
4265d620 PB |
3990 | static void coroutine_fn bdrv_discard_co_entry(void *opaque) |
3991 | { | |
3992 | RwCo *rwco = opaque; | |
3993 | ||
3994 | rwco->ret = bdrv_co_discard(rwco->bs, rwco->sector_num, rwco->nb_sectors); | |
3995 | } | |
3996 | ||
3997 | int coroutine_fn bdrv_co_discard(BlockDriverState *bs, int64_t sector_num, | |
3998 | int nb_sectors) | |
3999 | { | |
4000 | if (!bs->drv) { | |
4001 | return -ENOMEDIUM; | |
4002 | } else if (bdrv_check_request(bs, sector_num, nb_sectors)) { | |
4003 | return -EIO; | |
4004 | } else if (bs->read_only) { | |
4005 | return -EROFS; | |
4006 | } else if (bs->drv->bdrv_co_discard) { | |
4007 | return bs->drv->bdrv_co_discard(bs, sector_num, nb_sectors); | |
4008 | } else if (bs->drv->bdrv_aio_discard) { | |
4009 | BlockDriverAIOCB *acb; | |
4010 | CoroutineIOCompletion co = { | |
4011 | .coroutine = qemu_coroutine_self(), | |
4012 | }; | |
4013 | ||
4014 | acb = bs->drv->bdrv_aio_discard(bs, sector_num, nb_sectors, | |
4015 | bdrv_co_io_em_complete, &co); | |
4016 | if (acb == NULL) { | |
4017 | return -EIO; | |
4018 | } else { | |
4019 | qemu_coroutine_yield(); | |
4020 | return co.ret; | |
4021 | } | |
4265d620 PB |
4022 | } else { |
4023 | return 0; | |
4024 | } | |
4025 | } | |
4026 | ||
4027 | int bdrv_discard(BlockDriverState *bs, int64_t sector_num, int nb_sectors) | |
4028 | { | |
4029 | Coroutine *co; | |
4030 | RwCo rwco = { | |
4031 | .bs = bs, | |
4032 | .sector_num = sector_num, | |
4033 | .nb_sectors = nb_sectors, | |
4034 | .ret = NOT_DONE, | |
4035 | }; | |
4036 | ||
4037 | if (qemu_in_coroutine()) { | |
4038 | /* Fast-path if already in coroutine context */ | |
4039 | bdrv_discard_co_entry(&rwco); | |
4040 | } else { | |
4041 | co = qemu_coroutine_create(bdrv_discard_co_entry); | |
4042 | qemu_coroutine_enter(co, &rwco); | |
4043 | while (rwco.ret == NOT_DONE) { | |
4044 | qemu_aio_wait(); | |
4045 | } | |
4046 | } | |
4047 | ||
4048 | return rwco.ret; | |
4049 | } | |
4050 | ||
19cb3738 FB |
4051 | /**************************************************************/ |
4052 | /* removable device support */ | |
4053 | ||
4054 | /** | |
4055 | * Return TRUE if the media is present | |
4056 | */ | |
4057 | int bdrv_is_inserted(BlockDriverState *bs) | |
4058 | { | |
4059 | BlockDriver *drv = bs->drv; | |
a1aff5bf | 4060 | |
19cb3738 FB |
4061 | if (!drv) |
4062 | return 0; | |
4063 | if (!drv->bdrv_is_inserted) | |
a1aff5bf MA |
4064 | return 1; |
4065 | return drv->bdrv_is_inserted(bs); | |
19cb3738 FB |
4066 | } |
4067 | ||
4068 | /** | |
8e49ca46 MA |
4069 | * Return whether the media changed since the last call to this |
4070 | * function, or -ENOTSUP if we don't know. Most drivers don't know. | |
19cb3738 FB |
4071 | */ |
4072 | int bdrv_media_changed(BlockDriverState *bs) | |
4073 | { | |
4074 | BlockDriver *drv = bs->drv; | |
19cb3738 | 4075 | |
8e49ca46 MA |
4076 | if (drv && drv->bdrv_media_changed) { |
4077 | return drv->bdrv_media_changed(bs); | |
4078 | } | |
4079 | return -ENOTSUP; | |
19cb3738 FB |
4080 | } |
4081 | ||
4082 | /** | |
4083 | * If eject_flag is TRUE, eject the media. Otherwise, close the tray | |
4084 | */ | |
f36f3949 | 4085 | void bdrv_eject(BlockDriverState *bs, bool eject_flag) |
19cb3738 FB |
4086 | { |
4087 | BlockDriver *drv = bs->drv; | |
19cb3738 | 4088 | |
822e1cd1 MA |
4089 | if (drv && drv->bdrv_eject) { |
4090 | drv->bdrv_eject(bs, eject_flag); | |
19cb3738 | 4091 | } |
6f382ed2 LC |
4092 | |
4093 | if (bs->device_name[0] != '\0') { | |
4094 | bdrv_emit_qmp_eject_event(bs, eject_flag); | |
4095 | } | |
19cb3738 FB |
4096 | } |
4097 | ||
19cb3738 FB |
4098 | /** |
4099 | * Lock or unlock the media (if it is locked, the user won't be able | |
4100 | * to eject it manually). | |
4101 | */ | |
025e849a | 4102 | void bdrv_lock_medium(BlockDriverState *bs, bool locked) |
19cb3738 FB |
4103 | { |
4104 | BlockDriver *drv = bs->drv; | |
4105 | ||
025e849a | 4106 | trace_bdrv_lock_medium(bs, locked); |
b8c6d095 | 4107 | |
025e849a MA |
4108 | if (drv && drv->bdrv_lock_medium) { |
4109 | drv->bdrv_lock_medium(bs, locked); | |
19cb3738 FB |
4110 | } |
4111 | } | |
985a03b0 TS |
4112 | |
4113 | /* needed for generic scsi interface */ | |
4114 | ||
4115 | int bdrv_ioctl(BlockDriverState *bs, unsigned long int req, void *buf) | |
4116 | { | |
4117 | BlockDriver *drv = bs->drv; | |
4118 | ||
4119 | if (drv && drv->bdrv_ioctl) | |
4120 | return drv->bdrv_ioctl(bs, req, buf); | |
4121 | return -ENOTSUP; | |
4122 | } | |
7d780669 | 4123 | |
221f715d AL |
4124 | BlockDriverAIOCB *bdrv_aio_ioctl(BlockDriverState *bs, |
4125 | unsigned long int req, void *buf, | |
4126 | BlockDriverCompletionFunc *cb, void *opaque) | |
7d780669 | 4127 | { |
221f715d | 4128 | BlockDriver *drv = bs->drv; |
7d780669 | 4129 | |
221f715d AL |
4130 | if (drv && drv->bdrv_aio_ioctl) |
4131 | return drv->bdrv_aio_ioctl(bs, req, buf, cb, opaque); | |
4132 | return NULL; | |
7d780669 | 4133 | } |
e268ca52 | 4134 | |
7b6f9300 MA |
4135 | void bdrv_set_buffer_alignment(BlockDriverState *bs, int align) |
4136 | { | |
4137 | bs->buffer_alignment = align; | |
4138 | } | |
7cd1e32a | 4139 | |
e268ca52 AL |
4140 | void *qemu_blockalign(BlockDriverState *bs, size_t size) |
4141 | { | |
4142 | return qemu_memalign((bs && bs->buffer_alignment) ? bs->buffer_alignment : 512, size); | |
4143 | } | |
7cd1e32a | 4144 | |
4145 | void bdrv_set_dirty_tracking(BlockDriverState *bs, int enable) | |
4146 | { | |
4147 | int64_t bitmap_size; | |
a55eb92c | 4148 | |
aaa0eb75 | 4149 | bs->dirty_count = 0; |
a55eb92c | 4150 | if (enable) { |
c6d22830 JK |
4151 | if (!bs->dirty_bitmap) { |
4152 | bitmap_size = (bdrv_getlength(bs) >> BDRV_SECTOR_BITS) + | |
71df14fc PB |
4153 | BDRV_SECTORS_PER_DIRTY_CHUNK * BITS_PER_LONG - 1; |
4154 | bitmap_size /= BDRV_SECTORS_PER_DIRTY_CHUNK * BITS_PER_LONG; | |
a55eb92c | 4155 | |
71df14fc | 4156 | bs->dirty_bitmap = g_new0(unsigned long, bitmap_size); |
a55eb92c | 4157 | } |
7cd1e32a | 4158 | } else { |
c6d22830 | 4159 | if (bs->dirty_bitmap) { |
7267c094 | 4160 | g_free(bs->dirty_bitmap); |
c6d22830 | 4161 | bs->dirty_bitmap = NULL; |
a55eb92c | 4162 | } |
7cd1e32a | 4163 | } |
4164 | } | |
4165 | ||
4166 | int bdrv_get_dirty(BlockDriverState *bs, int64_t sector) | |
4167 | { | |
6ea44308 | 4168 | int64_t chunk = sector / (int64_t)BDRV_SECTORS_PER_DIRTY_CHUNK; |
a55eb92c | 4169 | |
c6d22830 JK |
4170 | if (bs->dirty_bitmap && |
4171 | (sector << BDRV_SECTOR_BITS) < bdrv_getlength(bs)) { | |
6d59fec1 MT |
4172 | return !!(bs->dirty_bitmap[chunk / (sizeof(unsigned long) * 8)] & |
4173 | (1UL << (chunk % (sizeof(unsigned long) * 8)))); | |
7cd1e32a | 4174 | } else { |
4175 | return 0; | |
4176 | } | |
4177 | } | |
4178 | ||
a55eb92c JK |
4179 | void bdrv_reset_dirty(BlockDriverState *bs, int64_t cur_sector, |
4180 | int nr_sectors) | |
7cd1e32a | 4181 | { |
4182 | set_dirty_bitmap(bs, cur_sector, nr_sectors, 0); | |
4183 | } | |
aaa0eb75 LS |
4184 | |
4185 | int64_t bdrv_get_dirty_count(BlockDriverState *bs) | |
4186 | { | |
4187 | return bs->dirty_count; | |
4188 | } | |
f88e1a42 | 4189 | |
db593f25 MT |
4190 | void bdrv_set_in_use(BlockDriverState *bs, int in_use) |
4191 | { | |
4192 | assert(bs->in_use != in_use); | |
4193 | bs->in_use = in_use; | |
4194 | } | |
4195 | ||
4196 | int bdrv_in_use(BlockDriverState *bs) | |
4197 | { | |
4198 | return bs->in_use; | |
4199 | } | |
4200 | ||
28a7282a LC |
4201 | void bdrv_iostatus_enable(BlockDriverState *bs) |
4202 | { | |
d6bf279e | 4203 | bs->iostatus_enabled = true; |
58e21ef5 | 4204 | bs->iostatus = BLOCK_DEVICE_IO_STATUS_OK; |
28a7282a LC |
4205 | } |
4206 | ||
4207 | /* The I/O status is only enabled if the drive explicitly | |
4208 | * enables it _and_ the VM is configured to stop on errors */ | |
4209 | bool bdrv_iostatus_is_enabled(const BlockDriverState *bs) | |
4210 | { | |
d6bf279e | 4211 | return (bs->iostatus_enabled && |
28a7282a LC |
4212 | (bs->on_write_error == BLOCK_ERR_STOP_ENOSPC || |
4213 | bs->on_write_error == BLOCK_ERR_STOP_ANY || | |
4214 | bs->on_read_error == BLOCK_ERR_STOP_ANY)); | |
4215 | } | |
4216 | ||
4217 | void bdrv_iostatus_disable(BlockDriverState *bs) | |
4218 | { | |
d6bf279e | 4219 | bs->iostatus_enabled = false; |
28a7282a LC |
4220 | } |
4221 | ||
4222 | void bdrv_iostatus_reset(BlockDriverState *bs) | |
4223 | { | |
4224 | if (bdrv_iostatus_is_enabled(bs)) { | |
58e21ef5 | 4225 | bs->iostatus = BLOCK_DEVICE_IO_STATUS_OK; |
28a7282a LC |
4226 | } |
4227 | } | |
4228 | ||
4229 | /* XXX: Today this is set by device models because it makes the implementation | |
4230 | quite simple. However, the block layer knows about the error, so it's | |
4231 | possible to implement this without device models being involved */ | |
4232 | void bdrv_iostatus_set_err(BlockDriverState *bs, int error) | |
4233 | { | |
58e21ef5 LC |
4234 | if (bdrv_iostatus_is_enabled(bs) && |
4235 | bs->iostatus == BLOCK_DEVICE_IO_STATUS_OK) { | |
28a7282a | 4236 | assert(error >= 0); |
58e21ef5 LC |
4237 | bs->iostatus = error == ENOSPC ? BLOCK_DEVICE_IO_STATUS_NOSPACE : |
4238 | BLOCK_DEVICE_IO_STATUS_FAILED; | |
28a7282a LC |
4239 | } |
4240 | } | |
4241 | ||
a597e79c CH |
4242 | void |
4243 | bdrv_acct_start(BlockDriverState *bs, BlockAcctCookie *cookie, int64_t bytes, | |
4244 | enum BlockAcctType type) | |
4245 | { | |
4246 | assert(type < BDRV_MAX_IOTYPE); | |
4247 | ||
4248 | cookie->bytes = bytes; | |
c488c7f6 | 4249 | cookie->start_time_ns = get_clock(); |
a597e79c CH |
4250 | cookie->type = type; |
4251 | } | |
4252 | ||
4253 | void | |
4254 | bdrv_acct_done(BlockDriverState *bs, BlockAcctCookie *cookie) | |
4255 | { | |
4256 | assert(cookie->type < BDRV_MAX_IOTYPE); | |
4257 | ||
4258 | bs->nr_bytes[cookie->type] += cookie->bytes; | |
4259 | bs->nr_ops[cookie->type]++; | |
c488c7f6 | 4260 | bs->total_time_ns[cookie->type] += get_clock() - cookie->start_time_ns; |
a597e79c CH |
4261 | } |
4262 | ||
f88e1a42 JS |
4263 | int bdrv_img_create(const char *filename, const char *fmt, |
4264 | const char *base_filename, const char *base_fmt, | |
4265 | char *options, uint64_t img_size, int flags) | |
4266 | { | |
4267 | QEMUOptionParameter *param = NULL, *create_options = NULL; | |
d220894e | 4268 | QEMUOptionParameter *backing_fmt, *backing_file, *size; |
f88e1a42 JS |
4269 | BlockDriverState *bs = NULL; |
4270 | BlockDriver *drv, *proto_drv; | |
96df67d1 | 4271 | BlockDriver *backing_drv = NULL; |
f88e1a42 JS |
4272 | int ret = 0; |
4273 | ||
4274 | /* Find driver and parse its options */ | |
4275 | drv = bdrv_find_format(fmt); | |
4276 | if (!drv) { | |
4277 | error_report("Unknown file format '%s'", fmt); | |
4f70f249 | 4278 | ret = -EINVAL; |
f88e1a42 JS |
4279 | goto out; |
4280 | } | |
4281 | ||
4282 | proto_drv = bdrv_find_protocol(filename); | |
4283 | if (!proto_drv) { | |
4284 | error_report("Unknown protocol '%s'", filename); | |
4f70f249 | 4285 | ret = -EINVAL; |
f88e1a42 JS |
4286 | goto out; |
4287 | } | |
4288 | ||
4289 | create_options = append_option_parameters(create_options, | |
4290 | drv->create_options); | |
4291 | create_options = append_option_parameters(create_options, | |
4292 | proto_drv->create_options); | |
4293 | ||
4294 | /* Create parameter list with default values */ | |
4295 | param = parse_option_parameters("", create_options, param); | |
4296 | ||
4297 | set_option_parameter_int(param, BLOCK_OPT_SIZE, img_size); | |
4298 | ||
4299 | /* Parse -o options */ | |
4300 | if (options) { | |
4301 | param = parse_option_parameters(options, create_options, param); | |
4302 | if (param == NULL) { | |
4303 | error_report("Invalid options for file format '%s'.", fmt); | |
4f70f249 | 4304 | ret = -EINVAL; |
f88e1a42 JS |
4305 | goto out; |
4306 | } | |
4307 | } | |
4308 | ||
4309 | if (base_filename) { | |
4310 | if (set_option_parameter(param, BLOCK_OPT_BACKING_FILE, | |
4311 | base_filename)) { | |
4312 | error_report("Backing file not supported for file format '%s'", | |
4313 | fmt); | |
4f70f249 | 4314 | ret = -EINVAL; |
f88e1a42 JS |
4315 | goto out; |
4316 | } | |
4317 | } | |
4318 | ||
4319 | if (base_fmt) { | |
4320 | if (set_option_parameter(param, BLOCK_OPT_BACKING_FMT, base_fmt)) { | |
4321 | error_report("Backing file format not supported for file " | |
4322 | "format '%s'", fmt); | |
4f70f249 | 4323 | ret = -EINVAL; |
f88e1a42 JS |
4324 | goto out; |
4325 | } | |
4326 | } | |
4327 | ||
792da93a JS |
4328 | backing_file = get_option_parameter(param, BLOCK_OPT_BACKING_FILE); |
4329 | if (backing_file && backing_file->value.s) { | |
4330 | if (!strcmp(filename, backing_file->value.s)) { | |
4331 | error_report("Error: Trying to create an image with the " | |
4332 | "same filename as the backing file"); | |
4f70f249 | 4333 | ret = -EINVAL; |
792da93a JS |
4334 | goto out; |
4335 | } | |
4336 | } | |
4337 | ||
f88e1a42 JS |
4338 | backing_fmt = get_option_parameter(param, BLOCK_OPT_BACKING_FMT); |
4339 | if (backing_fmt && backing_fmt->value.s) { | |
96df67d1 SH |
4340 | backing_drv = bdrv_find_format(backing_fmt->value.s); |
4341 | if (!backing_drv) { | |
f88e1a42 JS |
4342 | error_report("Unknown backing file format '%s'", |
4343 | backing_fmt->value.s); | |
4f70f249 | 4344 | ret = -EINVAL; |
f88e1a42 JS |
4345 | goto out; |
4346 | } | |
4347 | } | |
4348 | ||
4349 | // The size for the image must always be specified, with one exception: | |
4350 | // If we are using a backing file, we can obtain the size from there | |
d220894e KW |
4351 | size = get_option_parameter(param, BLOCK_OPT_SIZE); |
4352 | if (size && size->value.n == -1) { | |
f88e1a42 JS |
4353 | if (backing_file && backing_file->value.s) { |
4354 | uint64_t size; | |
f88e1a42 | 4355 | char buf[32]; |
63090dac PB |
4356 | int back_flags; |
4357 | ||
4358 | /* backing files always opened read-only */ | |
4359 | back_flags = | |
4360 | flags & ~(BDRV_O_RDWR | BDRV_O_SNAPSHOT | BDRV_O_NO_BACKING); | |
f88e1a42 | 4361 | |
f88e1a42 JS |
4362 | bs = bdrv_new(""); |
4363 | ||
63090dac | 4364 | ret = bdrv_open(bs, backing_file->value.s, back_flags, backing_drv); |
f88e1a42 | 4365 | if (ret < 0) { |
96df67d1 | 4366 | error_report("Could not open '%s'", backing_file->value.s); |
f88e1a42 JS |
4367 | goto out; |
4368 | } | |
4369 | bdrv_get_geometry(bs, &size); | |
4370 | size *= 512; | |
4371 | ||
4372 | snprintf(buf, sizeof(buf), "%" PRId64, size); | |
4373 | set_option_parameter(param, BLOCK_OPT_SIZE, buf); | |
4374 | } else { | |
4375 | error_report("Image creation needs a size parameter"); | |
4f70f249 | 4376 | ret = -EINVAL; |
f88e1a42 JS |
4377 | goto out; |
4378 | } | |
4379 | } | |
4380 | ||
4381 | printf("Formatting '%s', fmt=%s ", filename, fmt); | |
4382 | print_option_parameters(param); | |
4383 | puts(""); | |
4384 | ||
4385 | ret = bdrv_create(drv, filename, param); | |
4386 | ||
4387 | if (ret < 0) { | |
4388 | if (ret == -ENOTSUP) { | |
4389 | error_report("Formatting or formatting option not supported for " | |
4390 | "file format '%s'", fmt); | |
4391 | } else if (ret == -EFBIG) { | |
4392 | error_report("The image size is too large for file format '%s'", | |
4393 | fmt); | |
4394 | } else { | |
4395 | error_report("%s: error while creating %s: %s", filename, fmt, | |
4396 | strerror(-ret)); | |
4397 | } | |
4398 | } | |
4399 | ||
4400 | out: | |
4401 | free_option_parameters(create_options); | |
4402 | free_option_parameters(param); | |
4403 | ||
4404 | if (bs) { | |
4405 | bdrv_delete(bs); | |
4406 | } | |
4f70f249 JS |
4407 | |
4408 | return ret; | |
f88e1a42 | 4409 | } |