]>
Commit | Line | Data |
---|---|---|
fc01f7e7 FB |
1 | /* |
2 | * QEMU System Emulator block driver | |
5fafdf24 | 3 | * |
fc01f7e7 | 4 | * Copyright (c) 2003 Fabrice Bellard |
5fafdf24 | 5 | * |
fc01f7e7 FB |
6 | * Permission is hereby granted, free of charge, to any person obtaining a copy |
7 | * of this software and associated documentation files (the "Software"), to deal | |
8 | * in the Software without restriction, including without limitation the rights | |
9 | * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell | |
10 | * copies of the Software, and to permit persons to whom the Software is | |
11 | * furnished to do so, subject to the following conditions: | |
12 | * | |
13 | * The above copyright notice and this permission notice shall be included in | |
14 | * all copies or substantial portions of the Software. | |
15 | * | |
16 | * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR | |
17 | * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, | |
18 | * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL | |
19 | * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER | |
20 | * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, | |
21 | * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN | |
22 | * THE SOFTWARE. | |
23 | */ | |
3990d09a | 24 | #include "config-host.h" |
faf07963 | 25 | #include "qemu-common.h" |
6d519a5f | 26 | #include "trace.h" |
83c9089e | 27 | #include "monitor/monitor.h" |
737e150e PB |
28 | #include "block/block_int.h" |
29 | #include "block/blockjob.h" | |
1de7afc9 | 30 | #include "qemu/module.h" |
7b1b5d19 | 31 | #include "qapi/qmp/qjson.h" |
9c17d615 | 32 | #include "sysemu/sysemu.h" |
1de7afc9 | 33 | #include "qemu/notify.h" |
737e150e | 34 | #include "block/coroutine.h" |
b2023818 | 35 | #include "qmp-commands.h" |
1de7afc9 | 36 | #include "qemu/timer.h" |
fc01f7e7 | 37 | |
71e72a19 | 38 | #ifdef CONFIG_BSD |
7674e7bf FB |
39 | #include <sys/types.h> |
40 | #include <sys/stat.h> | |
41 | #include <sys/ioctl.h> | |
72cf2d4f | 42 | #include <sys/queue.h> |
c5e97233 | 43 | #ifndef __DragonFly__ |
7674e7bf FB |
44 | #include <sys/disk.h> |
45 | #endif | |
c5e97233 | 46 | #endif |
7674e7bf | 47 | |
49dc768d AL |
48 | #ifdef _WIN32 |
49 | #include <windows.h> | |
50 | #endif | |
51 | ||
1c9805a3 SH |
52 | #define NOT_DONE 0x7fffffff /* used while emulated sync operation in progress */ |
53 | ||
470c0504 SH |
54 | typedef enum { |
55 | BDRV_REQ_COPY_ON_READ = 0x1, | |
f08f2dda | 56 | BDRV_REQ_ZERO_WRITE = 0x2, |
470c0504 SH |
57 | } BdrvRequestFlags; |
58 | ||
7d4b4ba5 | 59 | static void bdrv_dev_change_media_cb(BlockDriverState *bs, bool load); |
f141eafe AL |
60 | static BlockDriverAIOCB *bdrv_aio_readv_em(BlockDriverState *bs, |
61 | int64_t sector_num, QEMUIOVector *qiov, int nb_sectors, | |
c87c0672 | 62 | BlockDriverCompletionFunc *cb, void *opaque); |
f141eafe AL |
63 | static BlockDriverAIOCB *bdrv_aio_writev_em(BlockDriverState *bs, |
64 | int64_t sector_num, QEMUIOVector *qiov, int nb_sectors, | |
ce1a14dc | 65 | BlockDriverCompletionFunc *cb, void *opaque); |
f9f05dc5 KW |
66 | static int coroutine_fn bdrv_co_readv_em(BlockDriverState *bs, |
67 | int64_t sector_num, int nb_sectors, | |
68 | QEMUIOVector *iov); | |
69 | static int coroutine_fn bdrv_co_writev_em(BlockDriverState *bs, | |
70 | int64_t sector_num, int nb_sectors, | |
71 | QEMUIOVector *iov); | |
c5fbe571 | 72 | static int coroutine_fn bdrv_co_do_readv(BlockDriverState *bs, |
470c0504 SH |
73 | int64_t sector_num, int nb_sectors, QEMUIOVector *qiov, |
74 | BdrvRequestFlags flags); | |
1c9805a3 | 75 | static int coroutine_fn bdrv_co_do_writev(BlockDriverState *bs, |
f08f2dda SH |
76 | int64_t sector_num, int nb_sectors, QEMUIOVector *qiov, |
77 | BdrvRequestFlags flags); | |
b2a61371 SH |
78 | static BlockDriverAIOCB *bdrv_co_aio_rw_vector(BlockDriverState *bs, |
79 | int64_t sector_num, | |
80 | QEMUIOVector *qiov, | |
81 | int nb_sectors, | |
82 | BlockDriverCompletionFunc *cb, | |
83 | void *opaque, | |
8c5873d6 | 84 | bool is_write); |
b2a61371 | 85 | static void coroutine_fn bdrv_co_do_rw(void *opaque); |
621f0589 KW |
86 | static int coroutine_fn bdrv_co_do_write_zeroes(BlockDriverState *bs, |
87 | int64_t sector_num, int nb_sectors); | |
ec530c81 | 88 | |
98f90dba ZYW |
89 | static bool bdrv_exceed_bps_limits(BlockDriverState *bs, int nb_sectors, |
90 | bool is_write, double elapsed_time, uint64_t *wait); | |
91 | static bool bdrv_exceed_iops_limits(BlockDriverState *bs, bool is_write, | |
92 | double elapsed_time, uint64_t *wait); | |
93 | static bool bdrv_exceed_io_limits(BlockDriverState *bs, int nb_sectors, | |
94 | bool is_write, int64_t *wait); | |
95 | ||
1b7bdbc1 SH |
96 | static QTAILQ_HEAD(, BlockDriverState) bdrv_states = |
97 | QTAILQ_HEAD_INITIALIZER(bdrv_states); | |
7ee930d0 | 98 | |
8a22f02a SH |
99 | static QLIST_HEAD(, BlockDriver) bdrv_drivers = |
100 | QLIST_HEAD_INITIALIZER(bdrv_drivers); | |
ea2384d3 | 101 | |
f9092b10 MA |
102 | /* The device to use for VM snapshots */ |
103 | static BlockDriverState *bs_snapshots; | |
104 | ||
eb852011 MA |
105 | /* If non-zero, use only whitelisted block drivers */ |
106 | static int use_bdrv_whitelist; | |
107 | ||
9e0b22f4 SH |
108 | #ifdef _WIN32 |
109 | static int is_windows_drive_prefix(const char *filename) | |
110 | { | |
111 | return (((filename[0] >= 'a' && filename[0] <= 'z') || | |
112 | (filename[0] >= 'A' && filename[0] <= 'Z')) && | |
113 | filename[1] == ':'); | |
114 | } | |
115 | ||
116 | int is_windows_drive(const char *filename) | |
117 | { | |
118 | if (is_windows_drive_prefix(filename) && | |
119 | filename[2] == '\0') | |
120 | return 1; | |
121 | if (strstart(filename, "\\\\.\\", NULL) || | |
122 | strstart(filename, "//./", NULL)) | |
123 | return 1; | |
124 | return 0; | |
125 | } | |
126 | #endif | |
127 | ||
0563e191 | 128 | /* throttling disk I/O limits */ |
98f90dba ZYW |
129 | void bdrv_io_limits_disable(BlockDriverState *bs) |
130 | { | |
131 | bs->io_limits_enabled = false; | |
132 | ||
133 | while (qemu_co_queue_next(&bs->throttled_reqs)); | |
134 | ||
135 | if (bs->block_timer) { | |
136 | qemu_del_timer(bs->block_timer); | |
137 | qemu_free_timer(bs->block_timer); | |
138 | bs->block_timer = NULL; | |
139 | } | |
140 | ||
141 | bs->slice_start = 0; | |
142 | bs->slice_end = 0; | |
143 | bs->slice_time = 0; | |
144 | memset(&bs->io_base, 0, sizeof(bs->io_base)); | |
145 | } | |
146 | ||
0563e191 ZYW |
147 | static void bdrv_block_timer(void *opaque) |
148 | { | |
149 | BlockDriverState *bs = opaque; | |
150 | ||
151 | qemu_co_queue_next(&bs->throttled_reqs); | |
152 | } | |
153 | ||
154 | void bdrv_io_limits_enable(BlockDriverState *bs) | |
155 | { | |
156 | qemu_co_queue_init(&bs->throttled_reqs); | |
157 | bs->block_timer = qemu_new_timer_ns(vm_clock, bdrv_block_timer, bs); | |
0563e191 ZYW |
158 | bs->io_limits_enabled = true; |
159 | } | |
160 | ||
161 | bool bdrv_io_limits_enabled(BlockDriverState *bs) | |
162 | { | |
163 | BlockIOLimit *io_limits = &bs->io_limits; | |
164 | return io_limits->bps[BLOCK_IO_LIMIT_READ] | |
165 | || io_limits->bps[BLOCK_IO_LIMIT_WRITE] | |
166 | || io_limits->bps[BLOCK_IO_LIMIT_TOTAL] | |
167 | || io_limits->iops[BLOCK_IO_LIMIT_READ] | |
168 | || io_limits->iops[BLOCK_IO_LIMIT_WRITE] | |
169 | || io_limits->iops[BLOCK_IO_LIMIT_TOTAL]; | |
170 | } | |
171 | ||
98f90dba ZYW |
172 | static void bdrv_io_limits_intercept(BlockDriverState *bs, |
173 | bool is_write, int nb_sectors) | |
174 | { | |
175 | int64_t wait_time = -1; | |
176 | ||
177 | if (!qemu_co_queue_empty(&bs->throttled_reqs)) { | |
178 | qemu_co_queue_wait(&bs->throttled_reqs); | |
179 | } | |
180 | ||
181 | /* In fact, we hope to keep each request's timing, in FIFO mode. The next | |
182 | * throttled requests will not be dequeued until the current request is | |
183 | * allowed to be serviced. So if the current request still exceeds the | |
184 | * limits, it will be inserted to the head. All requests followed it will | |
185 | * be still in throttled_reqs queue. | |
186 | */ | |
187 | ||
188 | while (bdrv_exceed_io_limits(bs, nb_sectors, is_write, &wait_time)) { | |
189 | qemu_mod_timer(bs->block_timer, | |
190 | wait_time + qemu_get_clock_ns(vm_clock)); | |
191 | qemu_co_queue_wait_insert_head(&bs->throttled_reqs); | |
192 | } | |
193 | ||
194 | qemu_co_queue_next(&bs->throttled_reqs); | |
195 | } | |
196 | ||
9e0b22f4 SH |
197 | /* check if the path starts with "<protocol>:" */ |
198 | static int path_has_protocol(const char *path) | |
199 | { | |
947995c0 PB |
200 | const char *p; |
201 | ||
9e0b22f4 SH |
202 | #ifdef _WIN32 |
203 | if (is_windows_drive(path) || | |
204 | is_windows_drive_prefix(path)) { | |
205 | return 0; | |
206 | } | |
947995c0 PB |
207 | p = path + strcspn(path, ":/\\"); |
208 | #else | |
209 | p = path + strcspn(path, ":/"); | |
9e0b22f4 SH |
210 | #endif |
211 | ||
947995c0 | 212 | return *p == ':'; |
9e0b22f4 SH |
213 | } |
214 | ||
83f64091 | 215 | int path_is_absolute(const char *path) |
3b0d4f61 | 216 | { |
21664424 FB |
217 | #ifdef _WIN32 |
218 | /* specific case for names like: "\\.\d:" */ | |
f53f4da9 | 219 | if (is_windows_drive(path) || is_windows_drive_prefix(path)) { |
21664424 | 220 | return 1; |
f53f4da9 PB |
221 | } |
222 | return (*path == '/' || *path == '\\'); | |
3b9f94e1 | 223 | #else |
f53f4da9 | 224 | return (*path == '/'); |
3b9f94e1 | 225 | #endif |
3b0d4f61 FB |
226 | } |
227 | ||
83f64091 FB |
228 | /* if filename is absolute, just copy it to dest. Otherwise, build a |
229 | path to it by considering it is relative to base_path. URL are | |
230 | supported. */ | |
231 | void path_combine(char *dest, int dest_size, | |
232 | const char *base_path, | |
233 | const char *filename) | |
3b0d4f61 | 234 | { |
83f64091 FB |
235 | const char *p, *p1; |
236 | int len; | |
237 | ||
238 | if (dest_size <= 0) | |
239 | return; | |
240 | if (path_is_absolute(filename)) { | |
241 | pstrcpy(dest, dest_size, filename); | |
242 | } else { | |
243 | p = strchr(base_path, ':'); | |
244 | if (p) | |
245 | p++; | |
246 | else | |
247 | p = base_path; | |
3b9f94e1 FB |
248 | p1 = strrchr(base_path, '/'); |
249 | #ifdef _WIN32 | |
250 | { | |
251 | const char *p2; | |
252 | p2 = strrchr(base_path, '\\'); | |
253 | if (!p1 || p2 > p1) | |
254 | p1 = p2; | |
255 | } | |
256 | #endif | |
83f64091 FB |
257 | if (p1) |
258 | p1++; | |
259 | else | |
260 | p1 = base_path; | |
261 | if (p1 > p) | |
262 | p = p1; | |
263 | len = p - base_path; | |
264 | if (len > dest_size - 1) | |
265 | len = dest_size - 1; | |
266 | memcpy(dest, base_path, len); | |
267 | dest[len] = '\0'; | |
268 | pstrcat(dest, dest_size, filename); | |
3b0d4f61 | 269 | } |
3b0d4f61 FB |
270 | } |
271 | ||
dc5a1371 PB |
272 | void bdrv_get_full_backing_filename(BlockDriverState *bs, char *dest, size_t sz) |
273 | { | |
274 | if (bs->backing_file[0] == '\0' || path_has_protocol(bs->backing_file)) { | |
275 | pstrcpy(dest, sz, bs->backing_file); | |
276 | } else { | |
277 | path_combine(dest, sz, bs->filename, bs->backing_file); | |
278 | } | |
279 | } | |
280 | ||
5efa9d5a | 281 | void bdrv_register(BlockDriver *bdrv) |
ea2384d3 | 282 | { |
8c5873d6 SH |
283 | /* Block drivers without coroutine functions need emulation */ |
284 | if (!bdrv->bdrv_co_readv) { | |
f9f05dc5 KW |
285 | bdrv->bdrv_co_readv = bdrv_co_readv_em; |
286 | bdrv->bdrv_co_writev = bdrv_co_writev_em; | |
287 | ||
f8c35c1d SH |
288 | /* bdrv_co_readv_em()/brdv_co_writev_em() work in terms of aio, so if |
289 | * the block driver lacks aio we need to emulate that too. | |
290 | */ | |
f9f05dc5 KW |
291 | if (!bdrv->bdrv_aio_readv) { |
292 | /* add AIO emulation layer */ | |
293 | bdrv->bdrv_aio_readv = bdrv_aio_readv_em; | |
294 | bdrv->bdrv_aio_writev = bdrv_aio_writev_em; | |
f9f05dc5 | 295 | } |
83f64091 | 296 | } |
b2e12bc6 | 297 | |
8a22f02a | 298 | QLIST_INSERT_HEAD(&bdrv_drivers, bdrv, list); |
ea2384d3 | 299 | } |
b338082b FB |
300 | |
301 | /* create a new block device (by default it is empty) */ | |
302 | BlockDriverState *bdrv_new(const char *device_name) | |
303 | { | |
1b7bdbc1 | 304 | BlockDriverState *bs; |
b338082b | 305 | |
7267c094 | 306 | bs = g_malloc0(sizeof(BlockDriverState)); |
b338082b | 307 | pstrcpy(bs->device_name, sizeof(bs->device_name), device_name); |
ea2384d3 | 308 | if (device_name[0] != '\0') { |
1b7bdbc1 | 309 | QTAILQ_INSERT_TAIL(&bdrv_states, bs, list); |
ea2384d3 | 310 | } |
28a7282a | 311 | bdrv_iostatus_disable(bs); |
d7d512f6 PB |
312 | notifier_list_init(&bs->close_notifiers); |
313 | ||
b338082b FB |
314 | return bs; |
315 | } | |
316 | ||
d7d512f6 PB |
317 | void bdrv_add_close_notifier(BlockDriverState *bs, Notifier *notify) |
318 | { | |
319 | notifier_list_add(&bs->close_notifiers, notify); | |
320 | } | |
321 | ||
ea2384d3 FB |
322 | BlockDriver *bdrv_find_format(const char *format_name) |
323 | { | |
324 | BlockDriver *drv1; | |
8a22f02a SH |
325 | QLIST_FOREACH(drv1, &bdrv_drivers, list) { |
326 | if (!strcmp(drv1->format_name, format_name)) { | |
ea2384d3 | 327 | return drv1; |
8a22f02a | 328 | } |
ea2384d3 FB |
329 | } |
330 | return NULL; | |
331 | } | |
332 | ||
eb852011 MA |
333 | static int bdrv_is_whitelisted(BlockDriver *drv) |
334 | { | |
335 | static const char *whitelist[] = { | |
336 | CONFIG_BDRV_WHITELIST | |
337 | }; | |
338 | const char **p; | |
339 | ||
340 | if (!whitelist[0]) | |
341 | return 1; /* no whitelist, anything goes */ | |
342 | ||
343 | for (p = whitelist; *p; p++) { | |
344 | if (!strcmp(drv->format_name, *p)) { | |
345 | return 1; | |
346 | } | |
347 | } | |
348 | return 0; | |
349 | } | |
350 | ||
351 | BlockDriver *bdrv_find_whitelisted_format(const char *format_name) | |
352 | { | |
353 | BlockDriver *drv = bdrv_find_format(format_name); | |
354 | return drv && bdrv_is_whitelisted(drv) ? drv : NULL; | |
355 | } | |
356 | ||
5b7e1542 ZYW |
357 | typedef struct CreateCo { |
358 | BlockDriver *drv; | |
359 | char *filename; | |
360 | QEMUOptionParameter *options; | |
361 | int ret; | |
362 | } CreateCo; | |
363 | ||
364 | static void coroutine_fn bdrv_create_co_entry(void *opaque) | |
365 | { | |
366 | CreateCo *cco = opaque; | |
367 | assert(cco->drv); | |
368 | ||
369 | cco->ret = cco->drv->bdrv_create(cco->filename, cco->options); | |
370 | } | |
371 | ||
0e7e1989 KW |
372 | int bdrv_create(BlockDriver *drv, const char* filename, |
373 | QEMUOptionParameter *options) | |
ea2384d3 | 374 | { |
5b7e1542 ZYW |
375 | int ret; |
376 | ||
377 | Coroutine *co; | |
378 | CreateCo cco = { | |
379 | .drv = drv, | |
380 | .filename = g_strdup(filename), | |
381 | .options = options, | |
382 | .ret = NOT_DONE, | |
383 | }; | |
384 | ||
385 | if (!drv->bdrv_create) { | |
80168bff LC |
386 | ret = -ENOTSUP; |
387 | goto out; | |
5b7e1542 ZYW |
388 | } |
389 | ||
390 | if (qemu_in_coroutine()) { | |
391 | /* Fast-path if already in coroutine context */ | |
392 | bdrv_create_co_entry(&cco); | |
393 | } else { | |
394 | co = qemu_coroutine_create(bdrv_create_co_entry); | |
395 | qemu_coroutine_enter(co, &cco); | |
396 | while (cco.ret == NOT_DONE) { | |
397 | qemu_aio_wait(); | |
398 | } | |
399 | } | |
400 | ||
401 | ret = cco.ret; | |
0e7e1989 | 402 | |
80168bff LC |
403 | out: |
404 | g_free(cco.filename); | |
5b7e1542 | 405 | return ret; |
ea2384d3 FB |
406 | } |
407 | ||
84a12e66 CH |
408 | int bdrv_create_file(const char* filename, QEMUOptionParameter *options) |
409 | { | |
410 | BlockDriver *drv; | |
411 | ||
b50cbabc | 412 | drv = bdrv_find_protocol(filename); |
84a12e66 | 413 | if (drv == NULL) { |
16905d71 | 414 | return -ENOENT; |
84a12e66 CH |
415 | } |
416 | ||
417 | return bdrv_create(drv, filename, options); | |
418 | } | |
419 | ||
eba25057 JM |
420 | /* |
421 | * Create a uniquely-named empty temporary file. | |
422 | * Return 0 upon success, otherwise a negative errno value. | |
423 | */ | |
424 | int get_tmp_filename(char *filename, int size) | |
d5249393 | 425 | { |
eba25057 | 426 | #ifdef _WIN32 |
3b9f94e1 | 427 | char temp_dir[MAX_PATH]; |
eba25057 JM |
428 | /* GetTempFileName requires that its output buffer (4th param) |
429 | have length MAX_PATH or greater. */ | |
430 | assert(size >= MAX_PATH); | |
431 | return (GetTempPath(MAX_PATH, temp_dir) | |
432 | && GetTempFileName(temp_dir, "qem", 0, filename) | |
433 | ? 0 : -GetLastError()); | |
d5249393 | 434 | #else |
67b915a5 | 435 | int fd; |
7ccfb2eb | 436 | const char *tmpdir; |
0badc1ee AJ |
437 | tmpdir = getenv("TMPDIR"); |
438 | if (!tmpdir) | |
439 | tmpdir = "/tmp"; | |
eba25057 JM |
440 | if (snprintf(filename, size, "%s/vl.XXXXXX", tmpdir) >= size) { |
441 | return -EOVERFLOW; | |
442 | } | |
ea2384d3 | 443 | fd = mkstemp(filename); |
fe235a06 DH |
444 | if (fd < 0) { |
445 | return -errno; | |
446 | } | |
447 | if (close(fd) != 0) { | |
448 | unlink(filename); | |
eba25057 JM |
449 | return -errno; |
450 | } | |
451 | return 0; | |
d5249393 | 452 | #endif |
eba25057 | 453 | } |
fc01f7e7 | 454 | |
84a12e66 CH |
455 | /* |
456 | * Detect host devices. By convention, /dev/cdrom[N] is always | |
457 | * recognized as a host CDROM. | |
458 | */ | |
459 | static BlockDriver *find_hdev_driver(const char *filename) | |
460 | { | |
461 | int score_max = 0, score; | |
462 | BlockDriver *drv = NULL, *d; | |
463 | ||
464 | QLIST_FOREACH(d, &bdrv_drivers, list) { | |
465 | if (d->bdrv_probe_device) { | |
466 | score = d->bdrv_probe_device(filename); | |
467 | if (score > score_max) { | |
468 | score_max = score; | |
469 | drv = d; | |
470 | } | |
471 | } | |
472 | } | |
473 | ||
474 | return drv; | |
475 | } | |
476 | ||
b50cbabc | 477 | BlockDriver *bdrv_find_protocol(const char *filename) |
83f64091 FB |
478 | { |
479 | BlockDriver *drv1; | |
480 | char protocol[128]; | |
1cec71e3 | 481 | int len; |
83f64091 | 482 | const char *p; |
19cb3738 | 483 | |
66f82cee KW |
484 | /* TODO Drivers without bdrv_file_open must be specified explicitly */ |
485 | ||
39508e7a CH |
486 | /* |
487 | * XXX(hch): we really should not let host device detection | |
488 | * override an explicit protocol specification, but moving this | |
489 | * later breaks access to device names with colons in them. | |
490 | * Thanks to the brain-dead persistent naming schemes on udev- | |
491 | * based Linux systems those actually are quite common. | |
492 | */ | |
493 | drv1 = find_hdev_driver(filename); | |
494 | if (drv1) { | |
495 | return drv1; | |
496 | } | |
497 | ||
9e0b22f4 | 498 | if (!path_has_protocol(filename)) { |
39508e7a | 499 | return bdrv_find_format("file"); |
84a12e66 | 500 | } |
9e0b22f4 SH |
501 | p = strchr(filename, ':'); |
502 | assert(p != NULL); | |
1cec71e3 AL |
503 | len = p - filename; |
504 | if (len > sizeof(protocol) - 1) | |
505 | len = sizeof(protocol) - 1; | |
506 | memcpy(protocol, filename, len); | |
507 | protocol[len] = '\0'; | |
8a22f02a | 508 | QLIST_FOREACH(drv1, &bdrv_drivers, list) { |
5fafdf24 | 509 | if (drv1->protocol_name && |
8a22f02a | 510 | !strcmp(drv1->protocol_name, protocol)) { |
83f64091 | 511 | return drv1; |
8a22f02a | 512 | } |
83f64091 FB |
513 | } |
514 | return NULL; | |
515 | } | |
516 | ||
f500a6d3 KW |
517 | static int find_image_format(BlockDriverState *bs, const char *filename, |
518 | BlockDriver **pdrv) | |
f3a5d3f8 | 519 | { |
f500a6d3 | 520 | int score, score_max; |
f3a5d3f8 CH |
521 | BlockDriver *drv1, *drv; |
522 | uint8_t buf[2048]; | |
f500a6d3 | 523 | int ret = 0; |
f8ea0b00 | 524 | |
08a00559 | 525 | /* Return the raw BlockDriver * to scsi-generic devices or empty drives */ |
8e895599 | 526 | if (bs->sg || !bdrv_is_inserted(bs) || bdrv_getlength(bs) == 0) { |
c98ac35d SW |
527 | drv = bdrv_find_format("raw"); |
528 | if (!drv) { | |
529 | ret = -ENOENT; | |
530 | } | |
531 | *pdrv = drv; | |
532 | return ret; | |
1a396859 | 533 | } |
f8ea0b00 | 534 | |
83f64091 | 535 | ret = bdrv_pread(bs, 0, buf, sizeof(buf)); |
83f64091 | 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 | ||
9e8f1835 PB |
583 | /** |
584 | * Set open flags for a given discard mode | |
585 | * | |
586 | * Return 0 on success, -1 if the discard mode was invalid. | |
587 | */ | |
588 | int bdrv_parse_discard_flags(const char *mode, int *flags) | |
589 | { | |
590 | *flags &= ~BDRV_O_UNMAP; | |
591 | ||
592 | if (!strcmp(mode, "off") || !strcmp(mode, "ignore")) { | |
593 | /* do nothing */ | |
594 | } else if (!strcmp(mode, "on") || !strcmp(mode, "unmap")) { | |
595 | *flags |= BDRV_O_UNMAP; | |
596 | } else { | |
597 | return -1; | |
598 | } | |
599 | ||
600 | return 0; | |
601 | } | |
602 | ||
c3993cdc SH |
603 | /** |
604 | * Set open flags for a given cache mode | |
605 | * | |
606 | * Return 0 on success, -1 if the cache mode was invalid. | |
607 | */ | |
608 | int bdrv_parse_cache_flags(const char *mode, int *flags) | |
609 | { | |
610 | *flags &= ~BDRV_O_CACHE_MASK; | |
611 | ||
612 | if (!strcmp(mode, "off") || !strcmp(mode, "none")) { | |
613 | *flags |= BDRV_O_NOCACHE | BDRV_O_CACHE_WB; | |
92196b2f SH |
614 | } else if (!strcmp(mode, "directsync")) { |
615 | *flags |= BDRV_O_NOCACHE; | |
c3993cdc SH |
616 | } else if (!strcmp(mode, "writeback")) { |
617 | *flags |= BDRV_O_CACHE_WB; | |
618 | } else if (!strcmp(mode, "unsafe")) { | |
619 | *flags |= BDRV_O_CACHE_WB; | |
620 | *flags |= BDRV_O_NO_FLUSH; | |
621 | } else if (!strcmp(mode, "writethrough")) { | |
622 | /* this is the default */ | |
623 | } else { | |
624 | return -1; | |
625 | } | |
626 | ||
627 | return 0; | |
628 | } | |
629 | ||
53fec9d3 SH |
630 | /** |
631 | * The copy-on-read flag is actually a reference count so multiple users may | |
632 | * use the feature without worrying about clobbering its previous state. | |
633 | * Copy-on-read stays enabled until all users have called to disable it. | |
634 | */ | |
635 | void bdrv_enable_copy_on_read(BlockDriverState *bs) | |
636 | { | |
637 | bs->copy_on_read++; | |
638 | } | |
639 | ||
640 | void bdrv_disable_copy_on_read(BlockDriverState *bs) | |
641 | { | |
642 | assert(bs->copy_on_read > 0); | |
643 | bs->copy_on_read--; | |
644 | } | |
645 | ||
7b272452 KW |
646 | static int bdrv_open_flags(BlockDriverState *bs, int flags) |
647 | { | |
648 | int open_flags = flags | BDRV_O_CACHE_WB; | |
649 | ||
650 | /* | |
651 | * Clear flags that are internal to the block layer before opening the | |
652 | * image. | |
653 | */ | |
654 | open_flags &= ~(BDRV_O_SNAPSHOT | BDRV_O_NO_BACKING); | |
655 | ||
656 | /* | |
657 | * Snapshots should be writable. | |
658 | */ | |
659 | if (bs->is_temporary) { | |
660 | open_flags |= BDRV_O_RDWR; | |
661 | } | |
662 | ||
663 | return open_flags; | |
664 | } | |
665 | ||
57915332 KW |
666 | /* |
667 | * Common part for opening disk images and files | |
b6ad491a KW |
668 | * |
669 | * Removes all processed options from *options. | |
57915332 | 670 | */ |
f500a6d3 | 671 | static int bdrv_open_common(BlockDriverState *bs, BlockDriverState *file, |
b6ad491a | 672 | const char *filename, QDict *options, |
57915332 KW |
673 | int flags, BlockDriver *drv) |
674 | { | |
675 | int ret, open_flags; | |
676 | ||
677 | assert(drv != NULL); | |
6405875c | 678 | assert(bs->file == NULL); |
707ff828 | 679 | assert(options != NULL && bs->options != options); |
57915332 | 680 | |
28dcee10 SH |
681 | trace_bdrv_open_common(bs, filename, flags, drv->format_name); |
682 | ||
57915332 | 683 | bs->open_flags = flags; |
57915332 KW |
684 | bs->buffer_alignment = 512; |
685 | ||
53fec9d3 SH |
686 | assert(bs->copy_on_read == 0); /* bdrv_new() and bdrv_close() make it so */ |
687 | if ((flags & BDRV_O_RDWR) && (flags & BDRV_O_COPY_ON_READ)) { | |
688 | bdrv_enable_copy_on_read(bs); | |
689 | } | |
690 | ||
57915332 KW |
691 | pstrcpy(bs->filename, sizeof(bs->filename), filename); |
692 | ||
693 | if (use_bdrv_whitelist && !bdrv_is_whitelisted(drv)) { | |
694 | return -ENOTSUP; | |
695 | } | |
696 | ||
697 | bs->drv = drv; | |
7267c094 | 698 | bs->opaque = g_malloc0(drv->instance_size); |
57915332 | 699 | |
03f541bd | 700 | bs->enable_write_cache = !!(flags & BDRV_O_CACHE_WB); |
7b272452 | 701 | open_flags = bdrv_open_flags(bs, flags); |
57915332 | 702 | |
be028adc | 703 | bs->read_only = !(open_flags & BDRV_O_RDWR); |
e7c63796 | 704 | |
66f82cee KW |
705 | /* Open the image, either directly or using a protocol */ |
706 | if (drv->bdrv_file_open) { | |
f500a6d3 KW |
707 | if (file != NULL) { |
708 | bdrv_swap(file, bs); | |
709 | ret = 0; | |
710 | } else { | |
787e4a85 | 711 | ret = drv->bdrv_file_open(bs, filename, options, open_flags); |
66f82cee | 712 | } |
f500a6d3 KW |
713 | } else { |
714 | assert(file != NULL); | |
715 | bs->file = file; | |
b6ad491a | 716 | ret = drv->bdrv_open(bs, options, open_flags); |
66f82cee KW |
717 | } |
718 | ||
57915332 KW |
719 | if (ret < 0) { |
720 | goto free_and_fail; | |
721 | } | |
722 | ||
51762288 SH |
723 | ret = refresh_total_sectors(bs, bs->total_sectors); |
724 | if (ret < 0) { | |
725 | goto free_and_fail; | |
57915332 | 726 | } |
51762288 | 727 | |
57915332 KW |
728 | #ifndef _WIN32 |
729 | if (bs->is_temporary) { | |
730 | unlink(filename); | |
731 | } | |
732 | #endif | |
733 | return 0; | |
734 | ||
735 | free_and_fail: | |
f500a6d3 | 736 | bs->file = NULL; |
7267c094 | 737 | g_free(bs->opaque); |
57915332 KW |
738 | bs->opaque = NULL; |
739 | bs->drv = NULL; | |
740 | return ret; | |
741 | } | |
742 | ||
b6ce07aa KW |
743 | /* |
744 | * Opens a file using a protocol (file, host_device, nbd, ...) | |
787e4a85 KW |
745 | * |
746 | * options is a QDict of options to pass to the block drivers, or NULL for an | |
747 | * empty set of options. The reference to the QDict belongs to the block layer | |
748 | * after the call (even on failure), so if the caller intends to reuse the | |
749 | * dictionary, it needs to use QINCREF() before calling bdrv_file_open. | |
b6ce07aa | 750 | */ |
787e4a85 KW |
751 | int bdrv_file_open(BlockDriverState **pbs, const char *filename, |
752 | QDict *options, int flags) | |
ea2384d3 | 753 | { |
83f64091 | 754 | BlockDriverState *bs; |
6db95603 | 755 | BlockDriver *drv; |
83f64091 FB |
756 | int ret; |
757 | ||
b50cbabc | 758 | drv = bdrv_find_protocol(filename); |
6db95603 | 759 | if (!drv) { |
707ff828 | 760 | QDECREF(options); |
6db95603 CH |
761 | return -ENOENT; |
762 | } | |
763 | ||
707ff828 KW |
764 | /* NULL means an empty set of options */ |
765 | if (options == NULL) { | |
766 | options = qdict_new(); | |
767 | } | |
768 | ||
83f64091 | 769 | bs = bdrv_new(""); |
707ff828 KW |
770 | bs->options = options; |
771 | options = qdict_clone_shallow(options); | |
772 | ||
6963a30d KW |
773 | if (drv->bdrv_parse_filename) { |
774 | Error *local_err = NULL; | |
775 | drv->bdrv_parse_filename(filename, options, &local_err); | |
776 | if (error_is_set(&local_err)) { | |
777 | qerror_report_err(local_err); | |
778 | error_free(local_err); | |
779 | ret = -EINVAL; | |
780 | goto fail; | |
781 | } | |
782 | } | |
783 | ||
707ff828 | 784 | ret = bdrv_open_common(bs, NULL, filename, options, flags, drv); |
83f64091 | 785 | if (ret < 0) { |
707ff828 KW |
786 | goto fail; |
787 | } | |
788 | ||
789 | /* Check if any unknown options were used */ | |
790 | if (qdict_size(options) != 0) { | |
791 | const QDictEntry *entry = qdict_first(options); | |
792 | qerror_report(ERROR_CLASS_GENERIC_ERROR, "Block protocol '%s' doesn't " | |
793 | "support the option '%s'", | |
794 | drv->format_name, entry->key); | |
795 | ret = -EINVAL; | |
796 | goto fail; | |
3b0d4f61 | 797 | } |
707ff828 KW |
798 | QDECREF(options); |
799 | ||
71d0770c | 800 | bs->growable = 1; |
83f64091 FB |
801 | *pbs = bs; |
802 | return 0; | |
707ff828 KW |
803 | |
804 | fail: | |
805 | QDECREF(options); | |
806 | if (!bs->drv) { | |
807 | QDECREF(bs->options); | |
808 | } | |
809 | bdrv_delete(bs); | |
810 | return ret; | |
83f64091 FB |
811 | } |
812 | ||
9156df12 PB |
813 | int bdrv_open_backing_file(BlockDriverState *bs) |
814 | { | |
815 | char backing_filename[PATH_MAX]; | |
816 | int back_flags, ret; | |
817 | BlockDriver *back_drv = NULL; | |
818 | ||
819 | if (bs->backing_hd != NULL) { | |
820 | return 0; | |
821 | } | |
822 | ||
823 | bs->open_flags &= ~BDRV_O_NO_BACKING; | |
824 | if (bs->backing_file[0] == '\0') { | |
825 | return 0; | |
826 | } | |
827 | ||
828 | bs->backing_hd = bdrv_new(""); | |
829 | bdrv_get_full_backing_filename(bs, backing_filename, | |
830 | sizeof(backing_filename)); | |
831 | ||
832 | if (bs->backing_format[0] != '\0') { | |
833 | back_drv = bdrv_find_format(bs->backing_format); | |
834 | } | |
835 | ||
836 | /* backing files always opened read-only */ | |
837 | back_flags = bs->open_flags & ~(BDRV_O_RDWR | BDRV_O_SNAPSHOT); | |
838 | ||
de9c0cec KW |
839 | ret = bdrv_open(bs->backing_hd, backing_filename, NULL, |
840 | back_flags, back_drv); | |
9156df12 PB |
841 | if (ret < 0) { |
842 | bdrv_delete(bs->backing_hd); | |
843 | bs->backing_hd = NULL; | |
844 | bs->open_flags |= BDRV_O_NO_BACKING; | |
845 | return ret; | |
846 | } | |
847 | return 0; | |
848 | } | |
849 | ||
707ff828 KW |
850 | static void extract_subqdict(QDict *src, QDict **dst, const char *start) |
851 | { | |
852 | const QDictEntry *entry, *next; | |
853 | const char *p; | |
854 | ||
855 | *dst = qdict_new(); | |
856 | entry = qdict_first(src); | |
857 | ||
858 | while (entry != NULL) { | |
859 | next = qdict_next(src, entry); | |
860 | if (strstart(entry->key, start, &p)) { | |
861 | qobject_incref(entry->value); | |
862 | qdict_put_obj(*dst, p, entry->value); | |
863 | qdict_del(src, entry->key); | |
864 | } | |
865 | entry = next; | |
866 | } | |
867 | } | |
868 | ||
b6ce07aa KW |
869 | /* |
870 | * Opens a disk image (raw, qcow2, vmdk, ...) | |
de9c0cec KW |
871 | * |
872 | * options is a QDict of options to pass to the block drivers, or NULL for an | |
873 | * empty set of options. The reference to the QDict belongs to the block layer | |
874 | * after the call (even on failure), so if the caller intends to reuse the | |
875 | * dictionary, it needs to use QINCREF() before calling bdrv_open. | |
b6ce07aa | 876 | */ |
de9c0cec KW |
877 | int bdrv_open(BlockDriverState *bs, const char *filename, QDict *options, |
878 | int flags, BlockDriver *drv) | |
ea2384d3 | 879 | { |
b6ce07aa | 880 | int ret; |
89c9bc3d SW |
881 | /* TODO: extra byte is a hack to ensure MAX_PATH space on Windows. */ |
882 | char tmp_filename[PATH_MAX + 1]; | |
f500a6d3 | 883 | BlockDriverState *file = NULL; |
707ff828 | 884 | QDict *file_options = NULL; |
712e7874 | 885 | |
de9c0cec KW |
886 | /* NULL means an empty set of options */ |
887 | if (options == NULL) { | |
888 | options = qdict_new(); | |
889 | } | |
890 | ||
891 | bs->options = options; | |
b6ad491a | 892 | options = qdict_clone_shallow(options); |
de9c0cec KW |
893 | |
894 | /* For snapshot=on, create a temporary qcow2 overlay */ | |
83f64091 | 895 | if (flags & BDRV_O_SNAPSHOT) { |
ea2384d3 FB |
896 | BlockDriverState *bs1; |
897 | int64_t total_size; | |
91a073a9 | 898 | BlockDriver *bdrv_qcow2; |
08b392e1 | 899 | QEMUOptionParameter *create_options; |
b6ce07aa | 900 | char backing_filename[PATH_MAX]; |
3b46e624 | 901 | |
ea2384d3 FB |
902 | /* if snapshot, we create a temporary backing file and open it |
903 | instead of opening 'filename' directly */ | |
33e3963e | 904 | |
ea2384d3 FB |
905 | /* if there is a backing file, use it */ |
906 | bs1 = bdrv_new(""); | |
de9c0cec | 907 | ret = bdrv_open(bs1, filename, NULL, 0, drv); |
51d7c00c | 908 | if (ret < 0) { |
ea2384d3 | 909 | bdrv_delete(bs1); |
de9c0cec | 910 | goto fail; |
ea2384d3 | 911 | } |
3e82990b | 912 | total_size = bdrv_getlength(bs1) & BDRV_SECTOR_MASK; |
7c96d46e | 913 | |
ea2384d3 | 914 | bdrv_delete(bs1); |
3b46e624 | 915 | |
eba25057 JM |
916 | ret = get_tmp_filename(tmp_filename, sizeof(tmp_filename)); |
917 | if (ret < 0) { | |
de9c0cec | 918 | goto fail; |
eba25057 | 919 | } |
7c96d46e AL |
920 | |
921 | /* Real path is meaningless for protocols */ | |
4d70655b | 922 | if (path_has_protocol(filename)) { |
7c96d46e AL |
923 | snprintf(backing_filename, sizeof(backing_filename), |
924 | "%s", filename); | |
de9c0cec KW |
925 | } else if (!realpath(filename, backing_filename)) { |
926 | ret = -errno; | |
927 | goto fail; | |
928 | } | |
7c96d46e | 929 | |
91a073a9 | 930 | bdrv_qcow2 = bdrv_find_format("qcow2"); |
08b392e1 KW |
931 | create_options = parse_option_parameters("", bdrv_qcow2->create_options, |
932 | NULL); | |
91a073a9 | 933 | |
08b392e1 KW |
934 | set_option_parameter_int(create_options, BLOCK_OPT_SIZE, total_size); |
935 | set_option_parameter(create_options, BLOCK_OPT_BACKING_FILE, | |
936 | backing_filename); | |
91a073a9 | 937 | if (drv) { |
08b392e1 | 938 | set_option_parameter(create_options, BLOCK_OPT_BACKING_FMT, |
91a073a9 KW |
939 | drv->format_name); |
940 | } | |
941 | ||
08b392e1 KW |
942 | ret = bdrv_create(bdrv_qcow2, tmp_filename, create_options); |
943 | free_option_parameters(create_options); | |
51d7c00c | 944 | if (ret < 0) { |
de9c0cec | 945 | goto fail; |
ea2384d3 | 946 | } |
91a073a9 | 947 | |
ea2384d3 | 948 | filename = tmp_filename; |
91a073a9 | 949 | drv = bdrv_qcow2; |
ea2384d3 FB |
950 | bs->is_temporary = 1; |
951 | } | |
712e7874 | 952 | |
f500a6d3 KW |
953 | /* Open image file without format layer */ |
954 | if (flags & BDRV_O_RDWR) { | |
955 | flags |= BDRV_O_ALLOW_RDWR; | |
956 | } | |
957 | ||
707ff828 KW |
958 | extract_subqdict(options, &file_options, "file."); |
959 | ||
960 | ret = bdrv_file_open(&file, filename, file_options, | |
961 | bdrv_open_flags(bs, flags)); | |
f500a6d3 | 962 | if (ret < 0) { |
de9c0cec | 963 | goto fail; |
f500a6d3 KW |
964 | } |
965 | ||
b6ce07aa | 966 | /* Find the right image format driver */ |
6db95603 | 967 | if (!drv) { |
f500a6d3 | 968 | ret = find_image_format(file, filename, &drv); |
51d7c00c | 969 | } |
6987307c | 970 | |
51d7c00c | 971 | if (!drv) { |
51d7c00c | 972 | goto unlink_and_fail; |
ea2384d3 | 973 | } |
b6ce07aa KW |
974 | |
975 | /* Open the image */ | |
b6ad491a | 976 | ret = bdrv_open_common(bs, file, filename, options, flags, drv); |
b6ce07aa | 977 | if (ret < 0) { |
6987307c CH |
978 | goto unlink_and_fail; |
979 | } | |
980 | ||
f500a6d3 KW |
981 | if (bs->file != file) { |
982 | bdrv_delete(file); | |
983 | file = NULL; | |
984 | } | |
985 | ||
b6ce07aa | 986 | /* If there is a backing file, use it */ |
9156df12 PB |
987 | if ((flags & BDRV_O_NO_BACKING) == 0) { |
988 | ret = bdrv_open_backing_file(bs); | |
b6ce07aa | 989 | if (ret < 0) { |
b6ad491a | 990 | goto close_and_fail; |
b6ce07aa | 991 | } |
b6ce07aa KW |
992 | } |
993 | ||
b6ad491a KW |
994 | /* Check if any unknown options were used */ |
995 | if (qdict_size(options) != 0) { | |
996 | const QDictEntry *entry = qdict_first(options); | |
997 | qerror_report(ERROR_CLASS_GENERIC_ERROR, "Block format '%s' used by " | |
998 | "device '%s' doesn't support the option '%s'", | |
999 | drv->format_name, bs->device_name, entry->key); | |
1000 | ||
1001 | ret = -EINVAL; | |
1002 | goto close_and_fail; | |
1003 | } | |
1004 | QDECREF(options); | |
1005 | ||
b6ce07aa | 1006 | if (!bdrv_key_required(bs)) { |
7d4b4ba5 | 1007 | bdrv_dev_change_media_cb(bs, true); |
b6ce07aa KW |
1008 | } |
1009 | ||
98f90dba ZYW |
1010 | /* throttling disk I/O limits */ |
1011 | if (bs->io_limits_enabled) { | |
1012 | bdrv_io_limits_enable(bs); | |
1013 | } | |
1014 | ||
b6ce07aa KW |
1015 | return 0; |
1016 | ||
1017 | unlink_and_fail: | |
f500a6d3 KW |
1018 | if (file != NULL) { |
1019 | bdrv_delete(file); | |
1020 | } | |
b6ce07aa KW |
1021 | if (bs->is_temporary) { |
1022 | unlink(filename); | |
1023 | } | |
de9c0cec KW |
1024 | fail: |
1025 | QDECREF(bs->options); | |
b6ad491a | 1026 | QDECREF(options); |
de9c0cec | 1027 | bs->options = NULL; |
b6ad491a | 1028 | return ret; |
de9c0cec | 1029 | |
b6ad491a KW |
1030 | close_and_fail: |
1031 | bdrv_close(bs); | |
1032 | QDECREF(options); | |
b6ce07aa KW |
1033 | return ret; |
1034 | } | |
1035 | ||
e971aa12 JC |
1036 | typedef struct BlockReopenQueueEntry { |
1037 | bool prepared; | |
1038 | BDRVReopenState state; | |
1039 | QSIMPLEQ_ENTRY(BlockReopenQueueEntry) entry; | |
1040 | } BlockReopenQueueEntry; | |
1041 | ||
1042 | /* | |
1043 | * Adds a BlockDriverState to a simple queue for an atomic, transactional | |
1044 | * reopen of multiple devices. | |
1045 | * | |
1046 | * bs_queue can either be an existing BlockReopenQueue that has had QSIMPLE_INIT | |
1047 | * already performed, or alternatively may be NULL a new BlockReopenQueue will | |
1048 | * be created and initialized. This newly created BlockReopenQueue should be | |
1049 | * passed back in for subsequent calls that are intended to be of the same | |
1050 | * atomic 'set'. | |
1051 | * | |
1052 | * bs is the BlockDriverState to add to the reopen queue. | |
1053 | * | |
1054 | * flags contains the open flags for the associated bs | |
1055 | * | |
1056 | * returns a pointer to bs_queue, which is either the newly allocated | |
1057 | * bs_queue, or the existing bs_queue being used. | |
1058 | * | |
1059 | */ | |
1060 | BlockReopenQueue *bdrv_reopen_queue(BlockReopenQueue *bs_queue, | |
1061 | BlockDriverState *bs, int flags) | |
1062 | { | |
1063 | assert(bs != NULL); | |
1064 | ||
1065 | BlockReopenQueueEntry *bs_entry; | |
1066 | if (bs_queue == NULL) { | |
1067 | bs_queue = g_new0(BlockReopenQueue, 1); | |
1068 | QSIMPLEQ_INIT(bs_queue); | |
1069 | } | |
1070 | ||
1071 | if (bs->file) { | |
1072 | bdrv_reopen_queue(bs_queue, bs->file, flags); | |
1073 | } | |
1074 | ||
1075 | bs_entry = g_new0(BlockReopenQueueEntry, 1); | |
1076 | QSIMPLEQ_INSERT_TAIL(bs_queue, bs_entry, entry); | |
1077 | ||
1078 | bs_entry->state.bs = bs; | |
1079 | bs_entry->state.flags = flags; | |
1080 | ||
1081 | return bs_queue; | |
1082 | } | |
1083 | ||
1084 | /* | |
1085 | * Reopen multiple BlockDriverStates atomically & transactionally. | |
1086 | * | |
1087 | * The queue passed in (bs_queue) must have been built up previous | |
1088 | * via bdrv_reopen_queue(). | |
1089 | * | |
1090 | * Reopens all BDS specified in the queue, with the appropriate | |
1091 | * flags. All devices are prepared for reopen, and failure of any | |
1092 | * device will cause all device changes to be abandonded, and intermediate | |
1093 | * data cleaned up. | |
1094 | * | |
1095 | * If all devices prepare successfully, then the changes are committed | |
1096 | * to all devices. | |
1097 | * | |
1098 | */ | |
1099 | int bdrv_reopen_multiple(BlockReopenQueue *bs_queue, Error **errp) | |
1100 | { | |
1101 | int ret = -1; | |
1102 | BlockReopenQueueEntry *bs_entry, *next; | |
1103 | Error *local_err = NULL; | |
1104 | ||
1105 | assert(bs_queue != NULL); | |
1106 | ||
1107 | bdrv_drain_all(); | |
1108 | ||
1109 | QSIMPLEQ_FOREACH(bs_entry, bs_queue, entry) { | |
1110 | if (bdrv_reopen_prepare(&bs_entry->state, bs_queue, &local_err)) { | |
1111 | error_propagate(errp, local_err); | |
1112 | goto cleanup; | |
1113 | } | |
1114 | bs_entry->prepared = true; | |
1115 | } | |
1116 | ||
1117 | /* If we reach this point, we have success and just need to apply the | |
1118 | * changes | |
1119 | */ | |
1120 | QSIMPLEQ_FOREACH(bs_entry, bs_queue, entry) { | |
1121 | bdrv_reopen_commit(&bs_entry->state); | |
1122 | } | |
1123 | ||
1124 | ret = 0; | |
1125 | ||
1126 | cleanup: | |
1127 | QSIMPLEQ_FOREACH_SAFE(bs_entry, bs_queue, entry, next) { | |
1128 | if (ret && bs_entry->prepared) { | |
1129 | bdrv_reopen_abort(&bs_entry->state); | |
1130 | } | |
1131 | g_free(bs_entry); | |
1132 | } | |
1133 | g_free(bs_queue); | |
1134 | return ret; | |
1135 | } | |
1136 | ||
1137 | ||
1138 | /* Reopen a single BlockDriverState with the specified flags. */ | |
1139 | int bdrv_reopen(BlockDriverState *bs, int bdrv_flags, Error **errp) | |
1140 | { | |
1141 | int ret = -1; | |
1142 | Error *local_err = NULL; | |
1143 | BlockReopenQueue *queue = bdrv_reopen_queue(NULL, bs, bdrv_flags); | |
1144 | ||
1145 | ret = bdrv_reopen_multiple(queue, &local_err); | |
1146 | if (local_err != NULL) { | |
1147 | error_propagate(errp, local_err); | |
1148 | } | |
1149 | return ret; | |
1150 | } | |
1151 | ||
1152 | ||
1153 | /* | |
1154 | * Prepares a BlockDriverState for reopen. All changes are staged in the | |
1155 | * 'opaque' field of the BDRVReopenState, which is used and allocated by | |
1156 | * the block driver layer .bdrv_reopen_prepare() | |
1157 | * | |
1158 | * bs is the BlockDriverState to reopen | |
1159 | * flags are the new open flags | |
1160 | * queue is the reopen queue | |
1161 | * | |
1162 | * Returns 0 on success, non-zero on error. On error errp will be set | |
1163 | * as well. | |
1164 | * | |
1165 | * On failure, bdrv_reopen_abort() will be called to clean up any data. | |
1166 | * It is the responsibility of the caller to then call the abort() or | |
1167 | * commit() for any other BDS that have been left in a prepare() state | |
1168 | * | |
1169 | */ | |
1170 | int bdrv_reopen_prepare(BDRVReopenState *reopen_state, BlockReopenQueue *queue, | |
1171 | Error **errp) | |
1172 | { | |
1173 | int ret = -1; | |
1174 | Error *local_err = NULL; | |
1175 | BlockDriver *drv; | |
1176 | ||
1177 | assert(reopen_state != NULL); | |
1178 | assert(reopen_state->bs->drv != NULL); | |
1179 | drv = reopen_state->bs->drv; | |
1180 | ||
1181 | /* if we are to stay read-only, do not allow permission change | |
1182 | * to r/w */ | |
1183 | if (!(reopen_state->bs->open_flags & BDRV_O_ALLOW_RDWR) && | |
1184 | reopen_state->flags & BDRV_O_RDWR) { | |
1185 | error_set(errp, QERR_DEVICE_IS_READ_ONLY, | |
1186 | reopen_state->bs->device_name); | |
1187 | goto error; | |
1188 | } | |
1189 | ||
1190 | ||
1191 | ret = bdrv_flush(reopen_state->bs); | |
1192 | if (ret) { | |
1193 | error_set(errp, ERROR_CLASS_GENERIC_ERROR, "Error (%s) flushing drive", | |
1194 | strerror(-ret)); | |
1195 | goto error; | |
1196 | } | |
1197 | ||
1198 | if (drv->bdrv_reopen_prepare) { | |
1199 | ret = drv->bdrv_reopen_prepare(reopen_state, queue, &local_err); | |
1200 | if (ret) { | |
1201 | if (local_err != NULL) { | |
1202 | error_propagate(errp, local_err); | |
1203 | } else { | |
1204 | error_set(errp, QERR_OPEN_FILE_FAILED, | |
1205 | reopen_state->bs->filename); | |
1206 | } | |
1207 | goto error; | |
1208 | } | |
1209 | } else { | |
1210 | /* It is currently mandatory to have a bdrv_reopen_prepare() | |
1211 | * handler for each supported drv. */ | |
1212 | error_set(errp, QERR_BLOCK_FORMAT_FEATURE_NOT_SUPPORTED, | |
1213 | drv->format_name, reopen_state->bs->device_name, | |
1214 | "reopening of file"); | |
1215 | ret = -1; | |
1216 | goto error; | |
1217 | } | |
1218 | ||
1219 | ret = 0; | |
1220 | ||
1221 | error: | |
1222 | return ret; | |
1223 | } | |
1224 | ||
1225 | /* | |
1226 | * Takes the staged changes for the reopen from bdrv_reopen_prepare(), and | |
1227 | * makes them final by swapping the staging BlockDriverState contents into | |
1228 | * the active BlockDriverState contents. | |
1229 | */ | |
1230 | void bdrv_reopen_commit(BDRVReopenState *reopen_state) | |
1231 | { | |
1232 | BlockDriver *drv; | |
1233 | ||
1234 | assert(reopen_state != NULL); | |
1235 | drv = reopen_state->bs->drv; | |
1236 | assert(drv != NULL); | |
1237 | ||
1238 | /* If there are any driver level actions to take */ | |
1239 | if (drv->bdrv_reopen_commit) { | |
1240 | drv->bdrv_reopen_commit(reopen_state); | |
1241 | } | |
1242 | ||
1243 | /* set BDS specific flags now */ | |
1244 | reopen_state->bs->open_flags = reopen_state->flags; | |
1245 | reopen_state->bs->enable_write_cache = !!(reopen_state->flags & | |
1246 | BDRV_O_CACHE_WB); | |
1247 | reopen_state->bs->read_only = !(reopen_state->flags & BDRV_O_RDWR); | |
1248 | } | |
1249 | ||
1250 | /* | |
1251 | * Abort the reopen, and delete and free the staged changes in | |
1252 | * reopen_state | |
1253 | */ | |
1254 | void bdrv_reopen_abort(BDRVReopenState *reopen_state) | |
1255 | { | |
1256 | BlockDriver *drv; | |
1257 | ||
1258 | assert(reopen_state != NULL); | |
1259 | drv = reopen_state->bs->drv; | |
1260 | assert(drv != NULL); | |
1261 | ||
1262 | if (drv->bdrv_reopen_abort) { | |
1263 | drv->bdrv_reopen_abort(reopen_state); | |
1264 | } | |
1265 | } | |
1266 | ||
1267 | ||
fc01f7e7 FB |
1268 | void bdrv_close(BlockDriverState *bs) |
1269 | { | |
80ccf93b | 1270 | bdrv_flush(bs); |
3cbc002c PB |
1271 | if (bs->job) { |
1272 | block_job_cancel_sync(bs->job); | |
1273 | } | |
1274 | bdrv_drain_all(); | |
d7d512f6 | 1275 | notifier_list_notify(&bs->close_notifiers, bs); |
7094f12f | 1276 | |
3cbc002c | 1277 | if (bs->drv) { |
f9092b10 MA |
1278 | if (bs == bs_snapshots) { |
1279 | bs_snapshots = NULL; | |
1280 | } | |
557df6ac | 1281 | if (bs->backing_hd) { |
ea2384d3 | 1282 | bdrv_delete(bs->backing_hd); |
557df6ac SH |
1283 | bs->backing_hd = NULL; |
1284 | } | |
ea2384d3 | 1285 | bs->drv->bdrv_close(bs); |
7267c094 | 1286 | g_free(bs->opaque); |
ea2384d3 FB |
1287 | #ifdef _WIN32 |
1288 | if (bs->is_temporary) { | |
1289 | unlink(bs->filename); | |
1290 | } | |
67b915a5 | 1291 | #endif |
ea2384d3 FB |
1292 | bs->opaque = NULL; |
1293 | bs->drv = NULL; | |
53fec9d3 | 1294 | bs->copy_on_read = 0; |
a275fa42 PB |
1295 | bs->backing_file[0] = '\0'; |
1296 | bs->backing_format[0] = '\0'; | |
6405875c PB |
1297 | bs->total_sectors = 0; |
1298 | bs->encrypted = 0; | |
1299 | bs->valid_key = 0; | |
1300 | bs->sg = 0; | |
1301 | bs->growable = 0; | |
de9c0cec KW |
1302 | QDECREF(bs->options); |
1303 | bs->options = NULL; | |
b338082b | 1304 | |
66f82cee | 1305 | if (bs->file != NULL) { |
0ac9377d PB |
1306 | bdrv_delete(bs->file); |
1307 | bs->file = NULL; | |
66f82cee | 1308 | } |
b338082b | 1309 | } |
98f90dba | 1310 | |
9ca11154 PH |
1311 | bdrv_dev_change_media_cb(bs, false); |
1312 | ||
98f90dba ZYW |
1313 | /*throttling disk I/O limits*/ |
1314 | if (bs->io_limits_enabled) { | |
1315 | bdrv_io_limits_disable(bs); | |
1316 | } | |
b338082b FB |
1317 | } |
1318 | ||
2bc93fed MK |
1319 | void bdrv_close_all(void) |
1320 | { | |
1321 | BlockDriverState *bs; | |
1322 | ||
1323 | QTAILQ_FOREACH(bs, &bdrv_states, list) { | |
1324 | bdrv_close(bs); | |
1325 | } | |
1326 | } | |
1327 | ||
922453bc SH |
1328 | /* |
1329 | * Wait for pending requests to complete across all BlockDriverStates | |
1330 | * | |
1331 | * This function does not flush data to disk, use bdrv_flush_all() for that | |
1332 | * after calling this function. | |
4c355d53 ZYW |
1333 | * |
1334 | * Note that completion of an asynchronous I/O operation can trigger any | |
1335 | * number of other I/O operations on other devices---for example a coroutine | |
1336 | * can be arbitrarily complex and a constant flow of I/O can come until the | |
1337 | * coroutine is complete. Because of this, it is not possible to have a | |
1338 | * function to drain a single device's I/O queue. | |
922453bc SH |
1339 | */ |
1340 | void bdrv_drain_all(void) | |
1341 | { | |
1342 | BlockDriverState *bs; | |
4c355d53 ZYW |
1343 | bool busy; |
1344 | ||
1345 | do { | |
1346 | busy = qemu_aio_wait(); | |
922453bc | 1347 | |
4c355d53 ZYW |
1348 | /* FIXME: We do not have timer support here, so this is effectively |
1349 | * a busy wait. | |
1350 | */ | |
1351 | QTAILQ_FOREACH(bs, &bdrv_states, list) { | |
1352 | if (!qemu_co_queue_empty(&bs->throttled_reqs)) { | |
1353 | qemu_co_queue_restart_all(&bs->throttled_reqs); | |
1354 | busy = true; | |
1355 | } | |
1356 | } | |
1357 | } while (busy); | |
922453bc SH |
1358 | |
1359 | /* If requests are still pending there is a bug somewhere */ | |
1360 | QTAILQ_FOREACH(bs, &bdrv_states, list) { | |
1361 | assert(QLIST_EMPTY(&bs->tracked_requests)); | |
1362 | assert(qemu_co_queue_empty(&bs->throttled_reqs)); | |
1363 | } | |
1364 | } | |
1365 | ||
d22b2f41 RH |
1366 | /* make a BlockDriverState anonymous by removing from bdrv_state list. |
1367 | Also, NULL terminate the device_name to prevent double remove */ | |
1368 | void bdrv_make_anon(BlockDriverState *bs) | |
1369 | { | |
1370 | if (bs->device_name[0] != '\0') { | |
1371 | QTAILQ_REMOVE(&bdrv_states, bs, list); | |
1372 | } | |
1373 | bs->device_name[0] = '\0'; | |
1374 | } | |
1375 | ||
e023b2e2 PB |
1376 | static void bdrv_rebind(BlockDriverState *bs) |
1377 | { | |
1378 | if (bs->drv && bs->drv->bdrv_rebind) { | |
1379 | bs->drv->bdrv_rebind(bs); | |
1380 | } | |
1381 | } | |
1382 | ||
4ddc07ca PB |
1383 | static void bdrv_move_feature_fields(BlockDriverState *bs_dest, |
1384 | BlockDriverState *bs_src) | |
8802d1fd | 1385 | { |
4ddc07ca PB |
1386 | /* move some fields that need to stay attached to the device */ |
1387 | bs_dest->open_flags = bs_src->open_flags; | |
8802d1fd JC |
1388 | |
1389 | /* dev info */ | |
4ddc07ca PB |
1390 | bs_dest->dev_ops = bs_src->dev_ops; |
1391 | bs_dest->dev_opaque = bs_src->dev_opaque; | |
1392 | bs_dest->dev = bs_src->dev; | |
1393 | bs_dest->buffer_alignment = bs_src->buffer_alignment; | |
1394 | bs_dest->copy_on_read = bs_src->copy_on_read; | |
8802d1fd | 1395 | |
4ddc07ca | 1396 | bs_dest->enable_write_cache = bs_src->enable_write_cache; |
c4a248a1 | 1397 | |
8802d1fd | 1398 | /* i/o timing parameters */ |
4ddc07ca PB |
1399 | bs_dest->slice_time = bs_src->slice_time; |
1400 | bs_dest->slice_start = bs_src->slice_start; | |
1401 | bs_dest->slice_end = bs_src->slice_end; | |
1402 | bs_dest->io_limits = bs_src->io_limits; | |
1403 | bs_dest->io_base = bs_src->io_base; | |
1404 | bs_dest->throttled_reqs = bs_src->throttled_reqs; | |
1405 | bs_dest->block_timer = bs_src->block_timer; | |
1406 | bs_dest->io_limits_enabled = bs_src->io_limits_enabled; | |
8802d1fd | 1407 | |
8802d1fd | 1408 | /* r/w error */ |
4ddc07ca PB |
1409 | bs_dest->on_read_error = bs_src->on_read_error; |
1410 | bs_dest->on_write_error = bs_src->on_write_error; | |
8802d1fd JC |
1411 | |
1412 | /* i/o status */ | |
4ddc07ca PB |
1413 | bs_dest->iostatus_enabled = bs_src->iostatus_enabled; |
1414 | bs_dest->iostatus = bs_src->iostatus; | |
8802d1fd | 1415 | |
a9fc4408 | 1416 | /* dirty bitmap */ |
4ddc07ca | 1417 | bs_dest->dirty_bitmap = bs_src->dirty_bitmap; |
a9fc4408 PB |
1418 | |
1419 | /* job */ | |
4ddc07ca PB |
1420 | bs_dest->in_use = bs_src->in_use; |
1421 | bs_dest->job = bs_src->job; | |
a9fc4408 | 1422 | |
8802d1fd | 1423 | /* keep the same entry in bdrv_states */ |
4ddc07ca PB |
1424 | pstrcpy(bs_dest->device_name, sizeof(bs_dest->device_name), |
1425 | bs_src->device_name); | |
1426 | bs_dest->list = bs_src->list; | |
1427 | } | |
8802d1fd | 1428 | |
4ddc07ca PB |
1429 | /* |
1430 | * Swap bs contents for two image chains while they are live, | |
1431 | * while keeping required fields on the BlockDriverState that is | |
1432 | * actually attached to a device. | |
1433 | * | |
1434 | * This will modify the BlockDriverState fields, and swap contents | |
1435 | * between bs_new and bs_old. Both bs_new and bs_old are modified. | |
1436 | * | |
1437 | * bs_new is required to be anonymous. | |
1438 | * | |
1439 | * This function does not create any image files. | |
1440 | */ | |
1441 | void bdrv_swap(BlockDriverState *bs_new, BlockDriverState *bs_old) | |
1442 | { | |
1443 | BlockDriverState tmp; | |
f6801b83 | 1444 | |
4ddc07ca PB |
1445 | /* bs_new must be anonymous and shouldn't have anything fancy enabled */ |
1446 | assert(bs_new->device_name[0] == '\0'); | |
1447 | assert(bs_new->dirty_bitmap == NULL); | |
1448 | assert(bs_new->job == NULL); | |
1449 | assert(bs_new->dev == NULL); | |
1450 | assert(bs_new->in_use == 0); | |
1451 | assert(bs_new->io_limits_enabled == false); | |
1452 | assert(bs_new->block_timer == NULL); | |
8802d1fd | 1453 | |
4ddc07ca PB |
1454 | tmp = *bs_new; |
1455 | *bs_new = *bs_old; | |
1456 | *bs_old = tmp; | |
a9fc4408 | 1457 | |
4ddc07ca PB |
1458 | /* there are some fields that should not be swapped, move them back */ |
1459 | bdrv_move_feature_fields(&tmp, bs_old); | |
1460 | bdrv_move_feature_fields(bs_old, bs_new); | |
1461 | bdrv_move_feature_fields(bs_new, &tmp); | |
8802d1fd | 1462 | |
4ddc07ca PB |
1463 | /* bs_new shouldn't be in bdrv_states even after the swap! */ |
1464 | assert(bs_new->device_name[0] == '\0'); | |
1465 | ||
1466 | /* Check a few fields that should remain attached to the device */ | |
1467 | assert(bs_new->dev == NULL); | |
1468 | assert(bs_new->job == NULL); | |
1469 | assert(bs_new->in_use == 0); | |
1470 | assert(bs_new->io_limits_enabled == false); | |
1471 | assert(bs_new->block_timer == NULL); | |
e023b2e2 PB |
1472 | |
1473 | bdrv_rebind(bs_new); | |
4ddc07ca PB |
1474 | bdrv_rebind(bs_old); |
1475 | } | |
1476 | ||
1477 | /* | |
1478 | * Add new bs contents at the top of an image chain while the chain is | |
1479 | * live, while keeping required fields on the top layer. | |
1480 | * | |
1481 | * This will modify the BlockDriverState fields, and swap contents | |
1482 | * between bs_new and bs_top. Both bs_new and bs_top are modified. | |
1483 | * | |
1484 | * bs_new is required to be anonymous. | |
1485 | * | |
1486 | * This function does not create any image files. | |
1487 | */ | |
1488 | void bdrv_append(BlockDriverState *bs_new, BlockDriverState *bs_top) | |
1489 | { | |
1490 | bdrv_swap(bs_new, bs_top); | |
1491 | ||
1492 | /* The contents of 'tmp' will become bs_top, as we are | |
1493 | * swapping bs_new and bs_top contents. */ | |
1494 | bs_top->backing_hd = bs_new; | |
1495 | bs_top->open_flags &= ~BDRV_O_NO_BACKING; | |
1496 | pstrcpy(bs_top->backing_file, sizeof(bs_top->backing_file), | |
1497 | bs_new->filename); | |
1498 | pstrcpy(bs_top->backing_format, sizeof(bs_top->backing_format), | |
1499 | bs_new->drv ? bs_new->drv->format_name : ""); | |
8802d1fd JC |
1500 | } |
1501 | ||
b338082b FB |
1502 | void bdrv_delete(BlockDriverState *bs) |
1503 | { | |
fa879d62 | 1504 | assert(!bs->dev); |
3e914655 PB |
1505 | assert(!bs->job); |
1506 | assert(!bs->in_use); | |
18846dee | 1507 | |
1b7bdbc1 | 1508 | /* remove from list, if necessary */ |
d22b2f41 | 1509 | bdrv_make_anon(bs); |
34c6f050 | 1510 | |
b338082b | 1511 | bdrv_close(bs); |
66f82cee | 1512 | |
f9092b10 | 1513 | assert(bs != bs_snapshots); |
7267c094 | 1514 | g_free(bs); |
fc01f7e7 FB |
1515 | } |
1516 | ||
fa879d62 MA |
1517 | int bdrv_attach_dev(BlockDriverState *bs, void *dev) |
1518 | /* TODO change to DeviceState *dev when all users are qdevified */ | |
18846dee | 1519 | { |
fa879d62 | 1520 | if (bs->dev) { |
18846dee MA |
1521 | return -EBUSY; |
1522 | } | |
fa879d62 | 1523 | bs->dev = dev; |
28a7282a | 1524 | bdrv_iostatus_reset(bs); |
18846dee MA |
1525 | return 0; |
1526 | } | |
1527 | ||
fa879d62 MA |
1528 | /* TODO qdevified devices don't use this, remove when devices are qdevified */ |
1529 | void bdrv_attach_dev_nofail(BlockDriverState *bs, void *dev) | |
18846dee | 1530 | { |
fa879d62 MA |
1531 | if (bdrv_attach_dev(bs, dev) < 0) { |
1532 | abort(); | |
1533 | } | |
1534 | } | |
1535 | ||
1536 | void bdrv_detach_dev(BlockDriverState *bs, void *dev) | |
1537 | /* TODO change to DeviceState *dev when all users are qdevified */ | |
1538 | { | |
1539 | assert(bs->dev == dev); | |
1540 | bs->dev = NULL; | |
0e49de52 MA |
1541 | bs->dev_ops = NULL; |
1542 | bs->dev_opaque = NULL; | |
29e05f20 | 1543 | bs->buffer_alignment = 512; |
18846dee MA |
1544 | } |
1545 | ||
fa879d62 MA |
1546 | /* TODO change to return DeviceState * when all users are qdevified */ |
1547 | void *bdrv_get_attached_dev(BlockDriverState *bs) | |
18846dee | 1548 | { |
fa879d62 | 1549 | return bs->dev; |
18846dee MA |
1550 | } |
1551 | ||
0e49de52 MA |
1552 | void bdrv_set_dev_ops(BlockDriverState *bs, const BlockDevOps *ops, |
1553 | void *opaque) | |
1554 | { | |
1555 | bs->dev_ops = ops; | |
1556 | bs->dev_opaque = opaque; | |
2c6942fa MA |
1557 | if (bdrv_dev_has_removable_media(bs) && bs == bs_snapshots) { |
1558 | bs_snapshots = NULL; | |
1559 | } | |
0e49de52 MA |
1560 | } |
1561 | ||
32c81a4a PB |
1562 | void bdrv_emit_qmp_error_event(const BlockDriverState *bdrv, |
1563 | enum MonitorEvent ev, | |
1564 | BlockErrorAction action, bool is_read) | |
329c0a48 LC |
1565 | { |
1566 | QObject *data; | |
1567 | const char *action_str; | |
1568 | ||
1569 | switch (action) { | |
1570 | case BDRV_ACTION_REPORT: | |
1571 | action_str = "report"; | |
1572 | break; | |
1573 | case BDRV_ACTION_IGNORE: | |
1574 | action_str = "ignore"; | |
1575 | break; | |
1576 | case BDRV_ACTION_STOP: | |
1577 | action_str = "stop"; | |
1578 | break; | |
1579 | default: | |
1580 | abort(); | |
1581 | } | |
1582 | ||
1583 | data = qobject_from_jsonf("{ 'device': %s, 'action': %s, 'operation': %s }", | |
1584 | bdrv->device_name, | |
1585 | action_str, | |
1586 | is_read ? "read" : "write"); | |
32c81a4a | 1587 | monitor_protocol_event(ev, data); |
329c0a48 LC |
1588 | |
1589 | qobject_decref(data); | |
1590 | } | |
1591 | ||
6f382ed2 LC |
1592 | static void bdrv_emit_qmp_eject_event(BlockDriverState *bs, bool ejected) |
1593 | { | |
1594 | QObject *data; | |
1595 | ||
1596 | data = qobject_from_jsonf("{ 'device': %s, 'tray-open': %i }", | |
1597 | bdrv_get_device_name(bs), ejected); | |
1598 | monitor_protocol_event(QEVENT_DEVICE_TRAY_MOVED, data); | |
1599 | ||
1600 | qobject_decref(data); | |
1601 | } | |
1602 | ||
7d4b4ba5 | 1603 | static void bdrv_dev_change_media_cb(BlockDriverState *bs, bool load) |
0e49de52 | 1604 | { |
145feb17 | 1605 | if (bs->dev_ops && bs->dev_ops->change_media_cb) { |
6f382ed2 | 1606 | bool tray_was_closed = !bdrv_dev_is_tray_open(bs); |
7d4b4ba5 | 1607 | bs->dev_ops->change_media_cb(bs->dev_opaque, load); |
6f382ed2 LC |
1608 | if (tray_was_closed) { |
1609 | /* tray open */ | |
1610 | bdrv_emit_qmp_eject_event(bs, true); | |
1611 | } | |
1612 | if (load) { | |
1613 | /* tray close */ | |
1614 | bdrv_emit_qmp_eject_event(bs, false); | |
1615 | } | |
145feb17 MA |
1616 | } |
1617 | } | |
1618 | ||
2c6942fa MA |
1619 | bool bdrv_dev_has_removable_media(BlockDriverState *bs) |
1620 | { | |
1621 | return !bs->dev || (bs->dev_ops && bs->dev_ops->change_media_cb); | |
1622 | } | |
1623 | ||
025ccaa7 PB |
1624 | void bdrv_dev_eject_request(BlockDriverState *bs, bool force) |
1625 | { | |
1626 | if (bs->dev_ops && bs->dev_ops->eject_request_cb) { | |
1627 | bs->dev_ops->eject_request_cb(bs->dev_opaque, force); | |
1628 | } | |
1629 | } | |
1630 | ||
e4def80b MA |
1631 | bool bdrv_dev_is_tray_open(BlockDriverState *bs) |
1632 | { | |
1633 | if (bs->dev_ops && bs->dev_ops->is_tray_open) { | |
1634 | return bs->dev_ops->is_tray_open(bs->dev_opaque); | |
1635 | } | |
1636 | return false; | |
1637 | } | |
1638 | ||
145feb17 MA |
1639 | static void bdrv_dev_resize_cb(BlockDriverState *bs) |
1640 | { | |
1641 | if (bs->dev_ops && bs->dev_ops->resize_cb) { | |
1642 | bs->dev_ops->resize_cb(bs->dev_opaque); | |
0e49de52 MA |
1643 | } |
1644 | } | |
1645 | ||
f107639a MA |
1646 | bool bdrv_dev_is_medium_locked(BlockDriverState *bs) |
1647 | { | |
1648 | if (bs->dev_ops && bs->dev_ops->is_medium_locked) { | |
1649 | return bs->dev_ops->is_medium_locked(bs->dev_opaque); | |
1650 | } | |
1651 | return false; | |
1652 | } | |
1653 | ||
e97fc193 AL |
1654 | /* |
1655 | * Run consistency checks on an image | |
1656 | * | |
e076f338 | 1657 | * Returns 0 if the check could be completed (it doesn't mean that the image is |
a1c7273b | 1658 | * free of errors) or -errno when an internal error occurred. The results of the |
e076f338 | 1659 | * check are stored in res. |
e97fc193 | 1660 | */ |
4534ff54 | 1661 | int bdrv_check(BlockDriverState *bs, BdrvCheckResult *res, BdrvCheckMode fix) |
e97fc193 AL |
1662 | { |
1663 | if (bs->drv->bdrv_check == NULL) { | |
1664 | return -ENOTSUP; | |
1665 | } | |
1666 | ||
e076f338 | 1667 | memset(res, 0, sizeof(*res)); |
4534ff54 | 1668 | return bs->drv->bdrv_check(bs, res, fix); |
e97fc193 AL |
1669 | } |
1670 | ||
8a426614 KW |
1671 | #define COMMIT_BUF_SECTORS 2048 |
1672 | ||
33e3963e FB |
1673 | /* commit COW file into the raw image */ |
1674 | int bdrv_commit(BlockDriverState *bs) | |
1675 | { | |
19cb3738 | 1676 | BlockDriver *drv = bs->drv; |
8a426614 KW |
1677 | int64_t sector, total_sectors; |
1678 | int n, ro, open_flags; | |
0bce597d | 1679 | int ret = 0; |
8a426614 | 1680 | uint8_t *buf; |
c2cba3d9 | 1681 | char filename[PATH_MAX]; |
33e3963e | 1682 | |
19cb3738 FB |
1683 | if (!drv) |
1684 | return -ENOMEDIUM; | |
4dca4b63 NS |
1685 | |
1686 | if (!bs->backing_hd) { | |
1687 | return -ENOTSUP; | |
33e3963e FB |
1688 | } |
1689 | ||
2d3735d3 SH |
1690 | if (bdrv_in_use(bs) || bdrv_in_use(bs->backing_hd)) { |
1691 | return -EBUSY; | |
1692 | } | |
1693 | ||
4dca4b63 | 1694 | ro = bs->backing_hd->read_only; |
c2cba3d9 JM |
1695 | /* Use pstrcpy (not strncpy): filename must be NUL-terminated. */ |
1696 | pstrcpy(filename, sizeof(filename), bs->backing_hd->filename); | |
4dca4b63 NS |
1697 | open_flags = bs->backing_hd->open_flags; |
1698 | ||
1699 | if (ro) { | |
0bce597d JC |
1700 | if (bdrv_reopen(bs->backing_hd, open_flags | BDRV_O_RDWR, NULL)) { |
1701 | return -EACCES; | |
4dca4b63 | 1702 | } |
ea2384d3 | 1703 | } |
33e3963e | 1704 | |
6ea44308 | 1705 | total_sectors = bdrv_getlength(bs) >> BDRV_SECTOR_BITS; |
7267c094 | 1706 | buf = g_malloc(COMMIT_BUF_SECTORS * BDRV_SECTOR_SIZE); |
8a426614 KW |
1707 | |
1708 | for (sector = 0; sector < total_sectors; sector += n) { | |
05c4af54 | 1709 | if (bdrv_is_allocated(bs, sector, COMMIT_BUF_SECTORS, &n)) { |
8a426614 KW |
1710 | |
1711 | if (bdrv_read(bs, sector, buf, n) != 0) { | |
1712 | ret = -EIO; | |
1713 | goto ro_cleanup; | |
1714 | } | |
1715 | ||
1716 | if (bdrv_write(bs->backing_hd, sector, buf, n) != 0) { | |
1717 | ret = -EIO; | |
1718 | goto ro_cleanup; | |
1719 | } | |
ea2384d3 | 1720 | } |
33e3963e | 1721 | } |
95389c86 | 1722 | |
1d44952f CH |
1723 | if (drv->bdrv_make_empty) { |
1724 | ret = drv->bdrv_make_empty(bs); | |
1725 | bdrv_flush(bs); | |
1726 | } | |
95389c86 | 1727 | |
3f5075ae CH |
1728 | /* |
1729 | * Make sure all data we wrote to the backing device is actually | |
1730 | * stable on disk. | |
1731 | */ | |
1732 | if (bs->backing_hd) | |
1733 | bdrv_flush(bs->backing_hd); | |
4dca4b63 NS |
1734 | |
1735 | ro_cleanup: | |
7267c094 | 1736 | g_free(buf); |
4dca4b63 NS |
1737 | |
1738 | if (ro) { | |
0bce597d JC |
1739 | /* ignoring error return here */ |
1740 | bdrv_reopen(bs->backing_hd, open_flags & ~BDRV_O_RDWR, NULL); | |
4dca4b63 NS |
1741 | } |
1742 | ||
1d44952f | 1743 | return ret; |
33e3963e FB |
1744 | } |
1745 | ||
e8877497 | 1746 | int bdrv_commit_all(void) |
6ab4b5ab MA |
1747 | { |
1748 | BlockDriverState *bs; | |
1749 | ||
1750 | QTAILQ_FOREACH(bs, &bdrv_states, list) { | |
272d2d8e JC |
1751 | if (bs->drv && bs->backing_hd) { |
1752 | int ret = bdrv_commit(bs); | |
1753 | if (ret < 0) { | |
1754 | return ret; | |
1755 | } | |
e8877497 | 1756 | } |
6ab4b5ab | 1757 | } |
e8877497 | 1758 | return 0; |
6ab4b5ab MA |
1759 | } |
1760 | ||
dbffbdcf SH |
1761 | struct BdrvTrackedRequest { |
1762 | BlockDriverState *bs; | |
1763 | int64_t sector_num; | |
1764 | int nb_sectors; | |
1765 | bool is_write; | |
1766 | QLIST_ENTRY(BdrvTrackedRequest) list; | |
5f8b6491 | 1767 | Coroutine *co; /* owner, used for deadlock detection */ |
f4658285 | 1768 | CoQueue wait_queue; /* coroutines blocked on this request */ |
dbffbdcf SH |
1769 | }; |
1770 | ||
1771 | /** | |
1772 | * Remove an active request from the tracked requests list | |
1773 | * | |
1774 | * This function should be called when a tracked request is completing. | |
1775 | */ | |
1776 | static void tracked_request_end(BdrvTrackedRequest *req) | |
1777 | { | |
1778 | QLIST_REMOVE(req, list); | |
f4658285 | 1779 | qemu_co_queue_restart_all(&req->wait_queue); |
dbffbdcf SH |
1780 | } |
1781 | ||
1782 | /** | |
1783 | * Add an active request to the tracked requests list | |
1784 | */ | |
1785 | static void tracked_request_begin(BdrvTrackedRequest *req, | |
1786 | BlockDriverState *bs, | |
1787 | int64_t sector_num, | |
1788 | int nb_sectors, bool is_write) | |
1789 | { | |
1790 | *req = (BdrvTrackedRequest){ | |
1791 | .bs = bs, | |
1792 | .sector_num = sector_num, | |
1793 | .nb_sectors = nb_sectors, | |
1794 | .is_write = is_write, | |
5f8b6491 | 1795 | .co = qemu_coroutine_self(), |
dbffbdcf SH |
1796 | }; |
1797 | ||
f4658285 SH |
1798 | qemu_co_queue_init(&req->wait_queue); |
1799 | ||
dbffbdcf SH |
1800 | QLIST_INSERT_HEAD(&bs->tracked_requests, req, list); |
1801 | } | |
1802 | ||
d83947ac SH |
1803 | /** |
1804 | * Round a region to cluster boundaries | |
1805 | */ | |
343bded4 PB |
1806 | void bdrv_round_to_clusters(BlockDriverState *bs, |
1807 | int64_t sector_num, int nb_sectors, | |
1808 | int64_t *cluster_sector_num, | |
1809 | int *cluster_nb_sectors) | |
d83947ac SH |
1810 | { |
1811 | BlockDriverInfo bdi; | |
1812 | ||
1813 | if (bdrv_get_info(bs, &bdi) < 0 || bdi.cluster_size == 0) { | |
1814 | *cluster_sector_num = sector_num; | |
1815 | *cluster_nb_sectors = nb_sectors; | |
1816 | } else { | |
1817 | int64_t c = bdi.cluster_size / BDRV_SECTOR_SIZE; | |
1818 | *cluster_sector_num = QEMU_ALIGN_DOWN(sector_num, c); | |
1819 | *cluster_nb_sectors = QEMU_ALIGN_UP(sector_num - *cluster_sector_num + | |
1820 | nb_sectors, c); | |
1821 | } | |
1822 | } | |
1823 | ||
f4658285 SH |
1824 | static bool tracked_request_overlaps(BdrvTrackedRequest *req, |
1825 | int64_t sector_num, int nb_sectors) { | |
d83947ac SH |
1826 | /* aaaa bbbb */ |
1827 | if (sector_num >= req->sector_num + req->nb_sectors) { | |
1828 | return false; | |
1829 | } | |
1830 | /* bbbb aaaa */ | |
1831 | if (req->sector_num >= sector_num + nb_sectors) { | |
1832 | return false; | |
1833 | } | |
1834 | return true; | |
f4658285 SH |
1835 | } |
1836 | ||
1837 | static void coroutine_fn wait_for_overlapping_requests(BlockDriverState *bs, | |
1838 | int64_t sector_num, int nb_sectors) | |
1839 | { | |
1840 | BdrvTrackedRequest *req; | |
d83947ac SH |
1841 | int64_t cluster_sector_num; |
1842 | int cluster_nb_sectors; | |
f4658285 SH |
1843 | bool retry; |
1844 | ||
d83947ac SH |
1845 | /* If we touch the same cluster it counts as an overlap. This guarantees |
1846 | * that allocating writes will be serialized and not race with each other | |
1847 | * for the same cluster. For example, in copy-on-read it ensures that the | |
1848 | * CoR read and write operations are atomic and guest writes cannot | |
1849 | * interleave between them. | |
1850 | */ | |
343bded4 PB |
1851 | bdrv_round_to_clusters(bs, sector_num, nb_sectors, |
1852 | &cluster_sector_num, &cluster_nb_sectors); | |
d83947ac | 1853 | |
f4658285 SH |
1854 | do { |
1855 | retry = false; | |
1856 | QLIST_FOREACH(req, &bs->tracked_requests, list) { | |
d83947ac SH |
1857 | if (tracked_request_overlaps(req, cluster_sector_num, |
1858 | cluster_nb_sectors)) { | |
5f8b6491 SH |
1859 | /* Hitting this means there was a reentrant request, for |
1860 | * example, a block driver issuing nested requests. This must | |
1861 | * never happen since it means deadlock. | |
1862 | */ | |
1863 | assert(qemu_coroutine_self() != req->co); | |
1864 | ||
f4658285 SH |
1865 | qemu_co_queue_wait(&req->wait_queue); |
1866 | retry = true; | |
1867 | break; | |
1868 | } | |
1869 | } | |
1870 | } while (retry); | |
1871 | } | |
1872 | ||
756e6736 KW |
1873 | /* |
1874 | * Return values: | |
1875 | * 0 - success | |
1876 | * -EINVAL - backing format specified, but no file | |
1877 | * -ENOSPC - can't update the backing file because no space is left in the | |
1878 | * image file header | |
1879 | * -ENOTSUP - format driver doesn't support changing the backing file | |
1880 | */ | |
1881 | int bdrv_change_backing_file(BlockDriverState *bs, | |
1882 | const char *backing_file, const char *backing_fmt) | |
1883 | { | |
1884 | BlockDriver *drv = bs->drv; | |
469ef350 | 1885 | int ret; |
756e6736 | 1886 | |
5f377794 PB |
1887 | /* Backing file format doesn't make sense without a backing file */ |
1888 | if (backing_fmt && !backing_file) { | |
1889 | return -EINVAL; | |
1890 | } | |
1891 | ||
756e6736 | 1892 | if (drv->bdrv_change_backing_file != NULL) { |
469ef350 | 1893 | ret = drv->bdrv_change_backing_file(bs, backing_file, backing_fmt); |
756e6736 | 1894 | } else { |
469ef350 | 1895 | ret = -ENOTSUP; |
756e6736 | 1896 | } |
469ef350 PB |
1897 | |
1898 | if (ret == 0) { | |
1899 | pstrcpy(bs->backing_file, sizeof(bs->backing_file), backing_file ?: ""); | |
1900 | pstrcpy(bs->backing_format, sizeof(bs->backing_format), backing_fmt ?: ""); | |
1901 | } | |
1902 | return ret; | |
756e6736 KW |
1903 | } |
1904 | ||
6ebdcee2 JC |
1905 | /* |
1906 | * Finds the image layer in the chain that has 'bs' as its backing file. | |
1907 | * | |
1908 | * active is the current topmost image. | |
1909 | * | |
1910 | * Returns NULL if bs is not found in active's image chain, | |
1911 | * or if active == bs. | |
1912 | */ | |
1913 | BlockDriverState *bdrv_find_overlay(BlockDriverState *active, | |
1914 | BlockDriverState *bs) | |
1915 | { | |
1916 | BlockDriverState *overlay = NULL; | |
1917 | BlockDriverState *intermediate; | |
1918 | ||
1919 | assert(active != NULL); | |
1920 | assert(bs != NULL); | |
1921 | ||
1922 | /* if bs is the same as active, then by definition it has no overlay | |
1923 | */ | |
1924 | if (active == bs) { | |
1925 | return NULL; | |
1926 | } | |
1927 | ||
1928 | intermediate = active; | |
1929 | while (intermediate->backing_hd) { | |
1930 | if (intermediate->backing_hd == bs) { | |
1931 | overlay = intermediate; | |
1932 | break; | |
1933 | } | |
1934 | intermediate = intermediate->backing_hd; | |
1935 | } | |
1936 | ||
1937 | return overlay; | |
1938 | } | |
1939 | ||
1940 | typedef struct BlkIntermediateStates { | |
1941 | BlockDriverState *bs; | |
1942 | QSIMPLEQ_ENTRY(BlkIntermediateStates) entry; | |
1943 | } BlkIntermediateStates; | |
1944 | ||
1945 | ||
1946 | /* | |
1947 | * Drops images above 'base' up to and including 'top', and sets the image | |
1948 | * above 'top' to have base as its backing file. | |
1949 | * | |
1950 | * Requires that the overlay to 'top' is opened r/w, so that the backing file | |
1951 | * information in 'bs' can be properly updated. | |
1952 | * | |
1953 | * E.g., this will convert the following chain: | |
1954 | * bottom <- base <- intermediate <- top <- active | |
1955 | * | |
1956 | * to | |
1957 | * | |
1958 | * bottom <- base <- active | |
1959 | * | |
1960 | * It is allowed for bottom==base, in which case it converts: | |
1961 | * | |
1962 | * base <- intermediate <- top <- active | |
1963 | * | |
1964 | * to | |
1965 | * | |
1966 | * base <- active | |
1967 | * | |
1968 | * Error conditions: | |
1969 | * if active == top, that is considered an error | |
1970 | * | |
1971 | */ | |
1972 | int bdrv_drop_intermediate(BlockDriverState *active, BlockDriverState *top, | |
1973 | BlockDriverState *base) | |
1974 | { | |
1975 | BlockDriverState *intermediate; | |
1976 | BlockDriverState *base_bs = NULL; | |
1977 | BlockDriverState *new_top_bs = NULL; | |
1978 | BlkIntermediateStates *intermediate_state, *next; | |
1979 | int ret = -EIO; | |
1980 | ||
1981 | QSIMPLEQ_HEAD(states_to_delete, BlkIntermediateStates) states_to_delete; | |
1982 | QSIMPLEQ_INIT(&states_to_delete); | |
1983 | ||
1984 | if (!top->drv || !base->drv) { | |
1985 | goto exit; | |
1986 | } | |
1987 | ||
1988 | new_top_bs = bdrv_find_overlay(active, top); | |
1989 | ||
1990 | if (new_top_bs == NULL) { | |
1991 | /* we could not find the image above 'top', this is an error */ | |
1992 | goto exit; | |
1993 | } | |
1994 | ||
1995 | /* special case of new_top_bs->backing_hd already pointing to base - nothing | |
1996 | * to do, no intermediate images */ | |
1997 | if (new_top_bs->backing_hd == base) { | |
1998 | ret = 0; | |
1999 | goto exit; | |
2000 | } | |
2001 | ||
2002 | intermediate = top; | |
2003 | ||
2004 | /* now we will go down through the list, and add each BDS we find | |
2005 | * into our deletion queue, until we hit the 'base' | |
2006 | */ | |
2007 | while (intermediate) { | |
2008 | intermediate_state = g_malloc0(sizeof(BlkIntermediateStates)); | |
2009 | intermediate_state->bs = intermediate; | |
2010 | QSIMPLEQ_INSERT_TAIL(&states_to_delete, intermediate_state, entry); | |
2011 | ||
2012 | if (intermediate->backing_hd == base) { | |
2013 | base_bs = intermediate->backing_hd; | |
2014 | break; | |
2015 | } | |
2016 | intermediate = intermediate->backing_hd; | |
2017 | } | |
2018 | if (base_bs == NULL) { | |
2019 | /* something went wrong, we did not end at the base. safely | |
2020 | * unravel everything, and exit with error */ | |
2021 | goto exit; | |
2022 | } | |
2023 | ||
2024 | /* success - we can delete the intermediate states, and link top->base */ | |
2025 | ret = bdrv_change_backing_file(new_top_bs, base_bs->filename, | |
2026 | base_bs->drv ? base_bs->drv->format_name : ""); | |
2027 | if (ret) { | |
2028 | goto exit; | |
2029 | } | |
2030 | new_top_bs->backing_hd = base_bs; | |
2031 | ||
2032 | ||
2033 | QSIMPLEQ_FOREACH_SAFE(intermediate_state, &states_to_delete, entry, next) { | |
2034 | /* so that bdrv_close() does not recursively close the chain */ | |
2035 | intermediate_state->bs->backing_hd = NULL; | |
2036 | bdrv_delete(intermediate_state->bs); | |
2037 | } | |
2038 | ret = 0; | |
2039 | ||
2040 | exit: | |
2041 | QSIMPLEQ_FOREACH_SAFE(intermediate_state, &states_to_delete, entry, next) { | |
2042 | g_free(intermediate_state); | |
2043 | } | |
2044 | return ret; | |
2045 | } | |
2046 | ||
2047 | ||
71d0770c AL |
2048 | static int bdrv_check_byte_request(BlockDriverState *bs, int64_t offset, |
2049 | size_t size) | |
2050 | { | |
2051 | int64_t len; | |
2052 | ||
2053 | if (!bdrv_is_inserted(bs)) | |
2054 | return -ENOMEDIUM; | |
2055 | ||
2056 | if (bs->growable) | |
2057 | return 0; | |
2058 | ||
2059 | len = bdrv_getlength(bs); | |
2060 | ||
fbb7b4e0 KW |
2061 | if (offset < 0) |
2062 | return -EIO; | |
2063 | ||
2064 | if ((offset > len) || (len - offset < size)) | |
71d0770c AL |
2065 | return -EIO; |
2066 | ||
2067 | return 0; | |
2068 | } | |
2069 | ||
2070 | static int bdrv_check_request(BlockDriverState *bs, int64_t sector_num, | |
2071 | int nb_sectors) | |
2072 | { | |
eb5a3165 JS |
2073 | return bdrv_check_byte_request(bs, sector_num * BDRV_SECTOR_SIZE, |
2074 | nb_sectors * BDRV_SECTOR_SIZE); | |
71d0770c AL |
2075 | } |
2076 | ||
1c9805a3 SH |
2077 | typedef struct RwCo { |
2078 | BlockDriverState *bs; | |
2079 | int64_t sector_num; | |
2080 | int nb_sectors; | |
2081 | QEMUIOVector *qiov; | |
2082 | bool is_write; | |
2083 | int ret; | |
2084 | } RwCo; | |
2085 | ||
2086 | static void coroutine_fn bdrv_rw_co_entry(void *opaque) | |
fc01f7e7 | 2087 | { |
1c9805a3 | 2088 | RwCo *rwco = opaque; |
ea2384d3 | 2089 | |
1c9805a3 SH |
2090 | if (!rwco->is_write) { |
2091 | rwco->ret = bdrv_co_do_readv(rwco->bs, rwco->sector_num, | |
470c0504 | 2092 | rwco->nb_sectors, rwco->qiov, 0); |
1c9805a3 SH |
2093 | } else { |
2094 | rwco->ret = bdrv_co_do_writev(rwco->bs, rwco->sector_num, | |
f08f2dda | 2095 | rwco->nb_sectors, rwco->qiov, 0); |
1c9805a3 SH |
2096 | } |
2097 | } | |
e7a8a783 | 2098 | |
1c9805a3 SH |
2099 | /* |
2100 | * Process a synchronous request using coroutines | |
2101 | */ | |
2102 | static int bdrv_rw_co(BlockDriverState *bs, int64_t sector_num, uint8_t *buf, | |
2103 | int nb_sectors, bool is_write) | |
2104 | { | |
2105 | QEMUIOVector qiov; | |
2106 | struct iovec iov = { | |
2107 | .iov_base = (void *)buf, | |
2108 | .iov_len = nb_sectors * BDRV_SECTOR_SIZE, | |
2109 | }; | |
2110 | Coroutine *co; | |
2111 | RwCo rwco = { | |
2112 | .bs = bs, | |
2113 | .sector_num = sector_num, | |
2114 | .nb_sectors = nb_sectors, | |
2115 | .qiov = &qiov, | |
2116 | .is_write = is_write, | |
2117 | .ret = NOT_DONE, | |
2118 | }; | |
e7a8a783 | 2119 | |
1c9805a3 | 2120 | qemu_iovec_init_external(&qiov, &iov, 1); |
e7a8a783 | 2121 | |
498e386c ZYW |
2122 | /** |
2123 | * In sync call context, when the vcpu is blocked, this throttling timer | |
2124 | * will not fire; so the I/O throttling function has to be disabled here | |
2125 | * if it has been enabled. | |
2126 | */ | |
2127 | if (bs->io_limits_enabled) { | |
2128 | fprintf(stderr, "Disabling I/O throttling on '%s' due " | |
2129 | "to synchronous I/O.\n", bdrv_get_device_name(bs)); | |
2130 | bdrv_io_limits_disable(bs); | |
2131 | } | |
2132 | ||
1c9805a3 SH |
2133 | if (qemu_in_coroutine()) { |
2134 | /* Fast-path if already in coroutine context */ | |
2135 | bdrv_rw_co_entry(&rwco); | |
2136 | } else { | |
2137 | co = qemu_coroutine_create(bdrv_rw_co_entry); | |
2138 | qemu_coroutine_enter(co, &rwco); | |
2139 | while (rwco.ret == NOT_DONE) { | |
2140 | qemu_aio_wait(); | |
2141 | } | |
2142 | } | |
2143 | return rwco.ret; | |
2144 | } | |
b338082b | 2145 | |
1c9805a3 SH |
2146 | /* return < 0 if error. See bdrv_write() for the return codes */ |
2147 | int bdrv_read(BlockDriverState *bs, int64_t sector_num, | |
2148 | uint8_t *buf, int nb_sectors) | |
2149 | { | |
2150 | return bdrv_rw_co(bs, sector_num, buf, nb_sectors, false); | |
fc01f7e7 FB |
2151 | } |
2152 | ||
07d27a44 MA |
2153 | /* Just like bdrv_read(), but with I/O throttling temporarily disabled */ |
2154 | int bdrv_read_unthrottled(BlockDriverState *bs, int64_t sector_num, | |
2155 | uint8_t *buf, int nb_sectors) | |
2156 | { | |
2157 | bool enabled; | |
2158 | int ret; | |
2159 | ||
2160 | enabled = bs->io_limits_enabled; | |
2161 | bs->io_limits_enabled = false; | |
2162 | ret = bdrv_read(bs, 0, buf, 1); | |
2163 | bs->io_limits_enabled = enabled; | |
2164 | return ret; | |
2165 | } | |
2166 | ||
5fafdf24 | 2167 | /* Return < 0 if error. Important errors are: |
19cb3738 FB |
2168 | -EIO generic I/O error (may happen for all errors) |
2169 | -ENOMEDIUM No media inserted. | |
2170 | -EINVAL Invalid sector number or nb_sectors | |
2171 | -EACCES Trying to write a read-only device | |
2172 | */ | |
5fafdf24 | 2173 | int bdrv_write(BlockDriverState *bs, int64_t sector_num, |
fc01f7e7 FB |
2174 | const uint8_t *buf, int nb_sectors) |
2175 | { | |
1c9805a3 | 2176 | return bdrv_rw_co(bs, sector_num, (uint8_t *)buf, nb_sectors, true); |
83f64091 FB |
2177 | } |
2178 | ||
eda578e5 AL |
2179 | int bdrv_pread(BlockDriverState *bs, int64_t offset, |
2180 | void *buf, int count1) | |
83f64091 | 2181 | { |
6ea44308 | 2182 | uint8_t tmp_buf[BDRV_SECTOR_SIZE]; |
83f64091 FB |
2183 | int len, nb_sectors, count; |
2184 | int64_t sector_num; | |
9a8c4cce | 2185 | int ret; |
83f64091 FB |
2186 | |
2187 | count = count1; | |
2188 | /* first read to align to sector start */ | |
6ea44308 | 2189 | len = (BDRV_SECTOR_SIZE - offset) & (BDRV_SECTOR_SIZE - 1); |
83f64091 FB |
2190 | if (len > count) |
2191 | len = count; | |
6ea44308 | 2192 | sector_num = offset >> BDRV_SECTOR_BITS; |
83f64091 | 2193 | if (len > 0) { |
9a8c4cce KW |
2194 | if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0) |
2195 | return ret; | |
6ea44308 | 2196 | memcpy(buf, tmp_buf + (offset & (BDRV_SECTOR_SIZE - 1)), len); |
83f64091 FB |
2197 | count -= len; |
2198 | if (count == 0) | |
2199 | return count1; | |
2200 | sector_num++; | |
2201 | buf += len; | |
2202 | } | |
2203 | ||
2204 | /* read the sectors "in place" */ | |
6ea44308 | 2205 | nb_sectors = count >> BDRV_SECTOR_BITS; |
83f64091 | 2206 | if (nb_sectors > 0) { |
9a8c4cce KW |
2207 | if ((ret = bdrv_read(bs, sector_num, buf, nb_sectors)) < 0) |
2208 | return ret; | |
83f64091 | 2209 | sector_num += nb_sectors; |
6ea44308 | 2210 | len = nb_sectors << BDRV_SECTOR_BITS; |
83f64091 FB |
2211 | buf += len; |
2212 | count -= len; | |
2213 | } | |
2214 | ||
2215 | /* add data from the last sector */ | |
2216 | if (count > 0) { | |
9a8c4cce KW |
2217 | if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0) |
2218 | return ret; | |
83f64091 FB |
2219 | memcpy(buf, tmp_buf, count); |
2220 | } | |
2221 | return count1; | |
2222 | } | |
2223 | ||
eda578e5 AL |
2224 | int bdrv_pwrite(BlockDriverState *bs, int64_t offset, |
2225 | const void *buf, int count1) | |
83f64091 | 2226 | { |
6ea44308 | 2227 | uint8_t tmp_buf[BDRV_SECTOR_SIZE]; |
83f64091 FB |
2228 | int len, nb_sectors, count; |
2229 | int64_t sector_num; | |
9a8c4cce | 2230 | int ret; |
83f64091 FB |
2231 | |
2232 | count = count1; | |
2233 | /* first write to align to sector start */ | |
6ea44308 | 2234 | len = (BDRV_SECTOR_SIZE - offset) & (BDRV_SECTOR_SIZE - 1); |
83f64091 FB |
2235 | if (len > count) |
2236 | len = count; | |
6ea44308 | 2237 | sector_num = offset >> BDRV_SECTOR_BITS; |
83f64091 | 2238 | if (len > 0) { |
9a8c4cce KW |
2239 | if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0) |
2240 | return ret; | |
6ea44308 | 2241 | memcpy(tmp_buf + (offset & (BDRV_SECTOR_SIZE - 1)), buf, len); |
9a8c4cce KW |
2242 | if ((ret = bdrv_write(bs, sector_num, tmp_buf, 1)) < 0) |
2243 | return ret; | |
83f64091 FB |
2244 | count -= len; |
2245 | if (count == 0) | |
2246 | return count1; | |
2247 | sector_num++; | |
2248 | buf += len; | |
2249 | } | |
2250 | ||
2251 | /* write the sectors "in place" */ | |
6ea44308 | 2252 | nb_sectors = count >> BDRV_SECTOR_BITS; |
83f64091 | 2253 | if (nb_sectors > 0) { |
9a8c4cce KW |
2254 | if ((ret = bdrv_write(bs, sector_num, buf, nb_sectors)) < 0) |
2255 | return ret; | |
83f64091 | 2256 | sector_num += nb_sectors; |
6ea44308 | 2257 | len = nb_sectors << BDRV_SECTOR_BITS; |
83f64091 FB |
2258 | buf += len; |
2259 | count -= len; | |
2260 | } | |
2261 | ||
2262 | /* add data from the last sector */ | |
2263 | if (count > 0) { | |
9a8c4cce KW |
2264 | if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0) |
2265 | return ret; | |
83f64091 | 2266 | memcpy(tmp_buf, buf, count); |
9a8c4cce KW |
2267 | if ((ret = bdrv_write(bs, sector_num, tmp_buf, 1)) < 0) |
2268 | return ret; | |
83f64091 FB |
2269 | } |
2270 | return count1; | |
2271 | } | |
83f64091 | 2272 | |
f08145fe KW |
2273 | /* |
2274 | * Writes to the file and ensures that no writes are reordered across this | |
2275 | * request (acts as a barrier) | |
2276 | * | |
2277 | * Returns 0 on success, -errno in error cases. | |
2278 | */ | |
2279 | int bdrv_pwrite_sync(BlockDriverState *bs, int64_t offset, | |
2280 | const void *buf, int count) | |
2281 | { | |
2282 | int ret; | |
2283 | ||
2284 | ret = bdrv_pwrite(bs, offset, buf, count); | |
2285 | if (ret < 0) { | |
2286 | return ret; | |
2287 | } | |
2288 | ||
f05fa4ad PB |
2289 | /* No flush needed for cache modes that already do it */ |
2290 | if (bs->enable_write_cache) { | |
f08145fe KW |
2291 | bdrv_flush(bs); |
2292 | } | |
2293 | ||
2294 | return 0; | |
2295 | } | |
2296 | ||
470c0504 | 2297 | static int coroutine_fn bdrv_co_do_copy_on_readv(BlockDriverState *bs, |
ab185921 SH |
2298 | int64_t sector_num, int nb_sectors, QEMUIOVector *qiov) |
2299 | { | |
2300 | /* Perform I/O through a temporary buffer so that users who scribble over | |
2301 | * their read buffer while the operation is in progress do not end up | |
2302 | * modifying the image file. This is critical for zero-copy guest I/O | |
2303 | * where anything might happen inside guest memory. | |
2304 | */ | |
2305 | void *bounce_buffer; | |
2306 | ||
79c053bd | 2307 | BlockDriver *drv = bs->drv; |
ab185921 SH |
2308 | struct iovec iov; |
2309 | QEMUIOVector bounce_qiov; | |
2310 | int64_t cluster_sector_num; | |
2311 | int cluster_nb_sectors; | |
2312 | size_t skip_bytes; | |
2313 | int ret; | |
2314 | ||
2315 | /* Cover entire cluster so no additional backing file I/O is required when | |
2316 | * allocating cluster in the image file. | |
2317 | */ | |
343bded4 PB |
2318 | bdrv_round_to_clusters(bs, sector_num, nb_sectors, |
2319 | &cluster_sector_num, &cluster_nb_sectors); | |
ab185921 | 2320 | |
470c0504 SH |
2321 | trace_bdrv_co_do_copy_on_readv(bs, sector_num, nb_sectors, |
2322 | cluster_sector_num, cluster_nb_sectors); | |
ab185921 SH |
2323 | |
2324 | iov.iov_len = cluster_nb_sectors * BDRV_SECTOR_SIZE; | |
2325 | iov.iov_base = bounce_buffer = qemu_blockalign(bs, iov.iov_len); | |
2326 | qemu_iovec_init_external(&bounce_qiov, &iov, 1); | |
2327 | ||
79c053bd SH |
2328 | ret = drv->bdrv_co_readv(bs, cluster_sector_num, cluster_nb_sectors, |
2329 | &bounce_qiov); | |
ab185921 SH |
2330 | if (ret < 0) { |
2331 | goto err; | |
2332 | } | |
2333 | ||
79c053bd SH |
2334 | if (drv->bdrv_co_write_zeroes && |
2335 | buffer_is_zero(bounce_buffer, iov.iov_len)) { | |
621f0589 KW |
2336 | ret = bdrv_co_do_write_zeroes(bs, cluster_sector_num, |
2337 | cluster_nb_sectors); | |
79c053bd | 2338 | } else { |
f05fa4ad PB |
2339 | /* This does not change the data on the disk, it is not necessary |
2340 | * to flush even in cache=writethrough mode. | |
2341 | */ | |
79c053bd | 2342 | ret = drv->bdrv_co_writev(bs, cluster_sector_num, cluster_nb_sectors, |
ab185921 | 2343 | &bounce_qiov); |
79c053bd SH |
2344 | } |
2345 | ||
ab185921 SH |
2346 | if (ret < 0) { |
2347 | /* It might be okay to ignore write errors for guest requests. If this | |
2348 | * is a deliberate copy-on-read then we don't want to ignore the error. | |
2349 | * Simply report it in all cases. | |
2350 | */ | |
2351 | goto err; | |
2352 | } | |
2353 | ||
2354 | skip_bytes = (sector_num - cluster_sector_num) * BDRV_SECTOR_SIZE; | |
03396148 MT |
2355 | qemu_iovec_from_buf(qiov, 0, bounce_buffer + skip_bytes, |
2356 | nb_sectors * BDRV_SECTOR_SIZE); | |
ab185921 SH |
2357 | |
2358 | err: | |
2359 | qemu_vfree(bounce_buffer); | |
2360 | return ret; | |
2361 | } | |
2362 | ||
c5fbe571 SH |
2363 | /* |
2364 | * Handle a read request in coroutine context | |
2365 | */ | |
2366 | static int coroutine_fn bdrv_co_do_readv(BlockDriverState *bs, | |
470c0504 SH |
2367 | int64_t sector_num, int nb_sectors, QEMUIOVector *qiov, |
2368 | BdrvRequestFlags flags) | |
da1fa91d KW |
2369 | { |
2370 | BlockDriver *drv = bs->drv; | |
dbffbdcf SH |
2371 | BdrvTrackedRequest req; |
2372 | int ret; | |
da1fa91d | 2373 | |
da1fa91d KW |
2374 | if (!drv) { |
2375 | return -ENOMEDIUM; | |
2376 | } | |
2377 | if (bdrv_check_request(bs, sector_num, nb_sectors)) { | |
2378 | return -EIO; | |
2379 | } | |
2380 | ||
98f90dba ZYW |
2381 | /* throttling disk read I/O */ |
2382 | if (bs->io_limits_enabled) { | |
2383 | bdrv_io_limits_intercept(bs, false, nb_sectors); | |
2384 | } | |
2385 | ||
f4658285 | 2386 | if (bs->copy_on_read) { |
470c0504 SH |
2387 | flags |= BDRV_REQ_COPY_ON_READ; |
2388 | } | |
2389 | if (flags & BDRV_REQ_COPY_ON_READ) { | |
2390 | bs->copy_on_read_in_flight++; | |
2391 | } | |
2392 | ||
2393 | if (bs->copy_on_read_in_flight) { | |
f4658285 SH |
2394 | wait_for_overlapping_requests(bs, sector_num, nb_sectors); |
2395 | } | |
2396 | ||
dbffbdcf | 2397 | tracked_request_begin(&req, bs, sector_num, nb_sectors, false); |
ab185921 | 2398 | |
470c0504 | 2399 | if (flags & BDRV_REQ_COPY_ON_READ) { |
ab185921 SH |
2400 | int pnum; |
2401 | ||
2402 | ret = bdrv_co_is_allocated(bs, sector_num, nb_sectors, &pnum); | |
2403 | if (ret < 0) { | |
2404 | goto out; | |
2405 | } | |
2406 | ||
2407 | if (!ret || pnum != nb_sectors) { | |
470c0504 | 2408 | ret = bdrv_co_do_copy_on_readv(bs, sector_num, nb_sectors, qiov); |
ab185921 SH |
2409 | goto out; |
2410 | } | |
2411 | } | |
2412 | ||
dbffbdcf | 2413 | ret = drv->bdrv_co_readv(bs, sector_num, nb_sectors, qiov); |
ab185921 SH |
2414 | |
2415 | out: | |
dbffbdcf | 2416 | tracked_request_end(&req); |
470c0504 SH |
2417 | |
2418 | if (flags & BDRV_REQ_COPY_ON_READ) { | |
2419 | bs->copy_on_read_in_flight--; | |
2420 | } | |
2421 | ||
dbffbdcf | 2422 | return ret; |
da1fa91d KW |
2423 | } |
2424 | ||
c5fbe571 | 2425 | int coroutine_fn bdrv_co_readv(BlockDriverState *bs, int64_t sector_num, |
da1fa91d KW |
2426 | int nb_sectors, QEMUIOVector *qiov) |
2427 | { | |
c5fbe571 | 2428 | trace_bdrv_co_readv(bs, sector_num, nb_sectors); |
da1fa91d | 2429 | |
470c0504 SH |
2430 | return bdrv_co_do_readv(bs, sector_num, nb_sectors, qiov, 0); |
2431 | } | |
2432 | ||
2433 | int coroutine_fn bdrv_co_copy_on_readv(BlockDriverState *bs, | |
2434 | int64_t sector_num, int nb_sectors, QEMUIOVector *qiov) | |
2435 | { | |
2436 | trace_bdrv_co_copy_on_readv(bs, sector_num, nb_sectors); | |
2437 | ||
2438 | return bdrv_co_do_readv(bs, sector_num, nb_sectors, qiov, | |
2439 | BDRV_REQ_COPY_ON_READ); | |
c5fbe571 SH |
2440 | } |
2441 | ||
f08f2dda SH |
2442 | static int coroutine_fn bdrv_co_do_write_zeroes(BlockDriverState *bs, |
2443 | int64_t sector_num, int nb_sectors) | |
2444 | { | |
2445 | BlockDriver *drv = bs->drv; | |
2446 | QEMUIOVector qiov; | |
2447 | struct iovec iov; | |
2448 | int ret; | |
2449 | ||
621f0589 KW |
2450 | /* TODO Emulate only part of misaligned requests instead of letting block |
2451 | * drivers return -ENOTSUP and emulate everything */ | |
2452 | ||
f08f2dda SH |
2453 | /* First try the efficient write zeroes operation */ |
2454 | if (drv->bdrv_co_write_zeroes) { | |
621f0589 KW |
2455 | ret = drv->bdrv_co_write_zeroes(bs, sector_num, nb_sectors); |
2456 | if (ret != -ENOTSUP) { | |
2457 | return ret; | |
2458 | } | |
f08f2dda SH |
2459 | } |
2460 | ||
2461 | /* Fall back to bounce buffer if write zeroes is unsupported */ | |
2462 | iov.iov_len = nb_sectors * BDRV_SECTOR_SIZE; | |
2463 | iov.iov_base = qemu_blockalign(bs, iov.iov_len); | |
2464 | memset(iov.iov_base, 0, iov.iov_len); | |
2465 | qemu_iovec_init_external(&qiov, &iov, 1); | |
2466 | ||
2467 | ret = drv->bdrv_co_writev(bs, sector_num, nb_sectors, &qiov); | |
2468 | ||
2469 | qemu_vfree(iov.iov_base); | |
2470 | return ret; | |
2471 | } | |
2472 | ||
c5fbe571 SH |
2473 | /* |
2474 | * Handle a write request in coroutine context | |
2475 | */ | |
2476 | static int coroutine_fn bdrv_co_do_writev(BlockDriverState *bs, | |
f08f2dda SH |
2477 | int64_t sector_num, int nb_sectors, QEMUIOVector *qiov, |
2478 | BdrvRequestFlags flags) | |
c5fbe571 SH |
2479 | { |
2480 | BlockDriver *drv = bs->drv; | |
dbffbdcf | 2481 | BdrvTrackedRequest req; |
6b7cb247 | 2482 | int ret; |
da1fa91d KW |
2483 | |
2484 | if (!bs->drv) { | |
2485 | return -ENOMEDIUM; | |
2486 | } | |
2487 | if (bs->read_only) { | |
2488 | return -EACCES; | |
2489 | } | |
2490 | if (bdrv_check_request(bs, sector_num, nb_sectors)) { | |
2491 | return -EIO; | |
2492 | } | |
2493 | ||
98f90dba ZYW |
2494 | /* throttling disk write I/O */ |
2495 | if (bs->io_limits_enabled) { | |
2496 | bdrv_io_limits_intercept(bs, true, nb_sectors); | |
2497 | } | |
2498 | ||
470c0504 | 2499 | if (bs->copy_on_read_in_flight) { |
f4658285 SH |
2500 | wait_for_overlapping_requests(bs, sector_num, nb_sectors); |
2501 | } | |
2502 | ||
dbffbdcf SH |
2503 | tracked_request_begin(&req, bs, sector_num, nb_sectors, true); |
2504 | ||
f08f2dda SH |
2505 | if (flags & BDRV_REQ_ZERO_WRITE) { |
2506 | ret = bdrv_co_do_write_zeroes(bs, sector_num, nb_sectors); | |
2507 | } else { | |
2508 | ret = drv->bdrv_co_writev(bs, sector_num, nb_sectors, qiov); | |
2509 | } | |
6b7cb247 | 2510 | |
f05fa4ad PB |
2511 | if (ret == 0 && !bs->enable_write_cache) { |
2512 | ret = bdrv_co_flush(bs); | |
2513 | } | |
2514 | ||
da1fa91d | 2515 | if (bs->dirty_bitmap) { |
1755da16 | 2516 | bdrv_set_dirty(bs, sector_num, nb_sectors); |
da1fa91d KW |
2517 | } |
2518 | ||
2519 | if (bs->wr_highest_sector < sector_num + nb_sectors - 1) { | |
2520 | bs->wr_highest_sector = sector_num + nb_sectors - 1; | |
2521 | } | |
2522 | ||
dbffbdcf SH |
2523 | tracked_request_end(&req); |
2524 | ||
6b7cb247 | 2525 | return ret; |
da1fa91d KW |
2526 | } |
2527 | ||
c5fbe571 SH |
2528 | int coroutine_fn bdrv_co_writev(BlockDriverState *bs, int64_t sector_num, |
2529 | int nb_sectors, QEMUIOVector *qiov) | |
2530 | { | |
2531 | trace_bdrv_co_writev(bs, sector_num, nb_sectors); | |
2532 | ||
f08f2dda SH |
2533 | return bdrv_co_do_writev(bs, sector_num, nb_sectors, qiov, 0); |
2534 | } | |
2535 | ||
2536 | int coroutine_fn bdrv_co_write_zeroes(BlockDriverState *bs, | |
2537 | int64_t sector_num, int nb_sectors) | |
2538 | { | |
2539 | trace_bdrv_co_write_zeroes(bs, sector_num, nb_sectors); | |
2540 | ||
2541 | return bdrv_co_do_writev(bs, sector_num, nb_sectors, NULL, | |
2542 | BDRV_REQ_ZERO_WRITE); | |
c5fbe571 SH |
2543 | } |
2544 | ||
83f64091 FB |
2545 | /** |
2546 | * Truncate file to 'offset' bytes (needed only for file protocols) | |
2547 | */ | |
2548 | int bdrv_truncate(BlockDriverState *bs, int64_t offset) | |
2549 | { | |
2550 | BlockDriver *drv = bs->drv; | |
51762288 | 2551 | int ret; |
83f64091 | 2552 | if (!drv) |
19cb3738 | 2553 | return -ENOMEDIUM; |
83f64091 FB |
2554 | if (!drv->bdrv_truncate) |
2555 | return -ENOTSUP; | |
59f2689d NS |
2556 | if (bs->read_only) |
2557 | return -EACCES; | |
8591675f MT |
2558 | if (bdrv_in_use(bs)) |
2559 | return -EBUSY; | |
51762288 SH |
2560 | ret = drv->bdrv_truncate(bs, offset); |
2561 | if (ret == 0) { | |
2562 | ret = refresh_total_sectors(bs, offset >> BDRV_SECTOR_BITS); | |
145feb17 | 2563 | bdrv_dev_resize_cb(bs); |
51762288 SH |
2564 | } |
2565 | return ret; | |
83f64091 FB |
2566 | } |
2567 | ||
4a1d5e1f FZ |
2568 | /** |
2569 | * Length of a allocated file in bytes. Sparse files are counted by actual | |
2570 | * allocated space. Return < 0 if error or unknown. | |
2571 | */ | |
2572 | int64_t bdrv_get_allocated_file_size(BlockDriverState *bs) | |
2573 | { | |
2574 | BlockDriver *drv = bs->drv; | |
2575 | if (!drv) { | |
2576 | return -ENOMEDIUM; | |
2577 | } | |
2578 | if (drv->bdrv_get_allocated_file_size) { | |
2579 | return drv->bdrv_get_allocated_file_size(bs); | |
2580 | } | |
2581 | if (bs->file) { | |
2582 | return bdrv_get_allocated_file_size(bs->file); | |
2583 | } | |
2584 | return -ENOTSUP; | |
2585 | } | |
2586 | ||
83f64091 FB |
2587 | /** |
2588 | * Length of a file in bytes. Return < 0 if error or unknown. | |
2589 | */ | |
2590 | int64_t bdrv_getlength(BlockDriverState *bs) | |
2591 | { | |
2592 | BlockDriver *drv = bs->drv; | |
2593 | if (!drv) | |
19cb3738 | 2594 | return -ENOMEDIUM; |
51762288 | 2595 | |
2c6942fa | 2596 | if (bs->growable || bdrv_dev_has_removable_media(bs)) { |
46a4e4e6 SH |
2597 | if (drv->bdrv_getlength) { |
2598 | return drv->bdrv_getlength(bs); | |
2599 | } | |
83f64091 | 2600 | } |
46a4e4e6 | 2601 | return bs->total_sectors * BDRV_SECTOR_SIZE; |
fc01f7e7 FB |
2602 | } |
2603 | ||
19cb3738 | 2604 | /* return 0 as number of sectors if no device present or error */ |
96b8f136 | 2605 | void bdrv_get_geometry(BlockDriverState *bs, uint64_t *nb_sectors_ptr) |
fc01f7e7 | 2606 | { |
19cb3738 FB |
2607 | int64_t length; |
2608 | length = bdrv_getlength(bs); | |
2609 | if (length < 0) | |
2610 | length = 0; | |
2611 | else | |
6ea44308 | 2612 | length = length >> BDRV_SECTOR_BITS; |
19cb3738 | 2613 | *nb_sectors_ptr = length; |
fc01f7e7 | 2614 | } |
cf98951b | 2615 | |
0563e191 ZYW |
2616 | /* throttling disk io limits */ |
2617 | void bdrv_set_io_limits(BlockDriverState *bs, | |
2618 | BlockIOLimit *io_limits) | |
2619 | { | |
2620 | bs->io_limits = *io_limits; | |
2621 | bs->io_limits_enabled = bdrv_io_limits_enabled(bs); | |
2622 | } | |
2623 | ||
ff06f5f3 PB |
2624 | void bdrv_set_on_error(BlockDriverState *bs, BlockdevOnError on_read_error, |
2625 | BlockdevOnError on_write_error) | |
abd7f68d MA |
2626 | { |
2627 | bs->on_read_error = on_read_error; | |
2628 | bs->on_write_error = on_write_error; | |
2629 | } | |
2630 | ||
1ceee0d5 | 2631 | BlockdevOnError bdrv_get_on_error(BlockDriverState *bs, bool is_read) |
abd7f68d MA |
2632 | { |
2633 | return is_read ? bs->on_read_error : bs->on_write_error; | |
2634 | } | |
2635 | ||
3e1caa5f PB |
2636 | BlockErrorAction bdrv_get_error_action(BlockDriverState *bs, bool is_read, int error) |
2637 | { | |
2638 | BlockdevOnError on_err = is_read ? bs->on_read_error : bs->on_write_error; | |
2639 | ||
2640 | switch (on_err) { | |
2641 | case BLOCKDEV_ON_ERROR_ENOSPC: | |
2642 | return (error == ENOSPC) ? BDRV_ACTION_STOP : BDRV_ACTION_REPORT; | |
2643 | case BLOCKDEV_ON_ERROR_STOP: | |
2644 | return BDRV_ACTION_STOP; | |
2645 | case BLOCKDEV_ON_ERROR_REPORT: | |
2646 | return BDRV_ACTION_REPORT; | |
2647 | case BLOCKDEV_ON_ERROR_IGNORE: | |
2648 | return BDRV_ACTION_IGNORE; | |
2649 | default: | |
2650 | abort(); | |
2651 | } | |
2652 | } | |
2653 | ||
2654 | /* This is done by device models because, while the block layer knows | |
2655 | * about the error, it does not know whether an operation comes from | |
2656 | * the device or the block layer (from a job, for example). | |
2657 | */ | |
2658 | void bdrv_error_action(BlockDriverState *bs, BlockErrorAction action, | |
2659 | bool is_read, int error) | |
2660 | { | |
2661 | assert(error >= 0); | |
32c81a4a | 2662 | bdrv_emit_qmp_error_event(bs, QEVENT_BLOCK_IO_ERROR, action, is_read); |
3e1caa5f PB |
2663 | if (action == BDRV_ACTION_STOP) { |
2664 | vm_stop(RUN_STATE_IO_ERROR); | |
2665 | bdrv_iostatus_set_err(bs, error); | |
2666 | } | |
2667 | } | |
2668 | ||
b338082b FB |
2669 | int bdrv_is_read_only(BlockDriverState *bs) |
2670 | { | |
2671 | return bs->read_only; | |
2672 | } | |
2673 | ||
985a03b0 TS |
2674 | int bdrv_is_sg(BlockDriverState *bs) |
2675 | { | |
2676 | return bs->sg; | |
2677 | } | |
2678 | ||
e900a7b7 CH |
2679 | int bdrv_enable_write_cache(BlockDriverState *bs) |
2680 | { | |
2681 | return bs->enable_write_cache; | |
2682 | } | |
2683 | ||
425b0148 PB |
2684 | void bdrv_set_enable_write_cache(BlockDriverState *bs, bool wce) |
2685 | { | |
2686 | bs->enable_write_cache = wce; | |
55b110f2 JC |
2687 | |
2688 | /* so a reopen() will preserve wce */ | |
2689 | if (wce) { | |
2690 | bs->open_flags |= BDRV_O_CACHE_WB; | |
2691 | } else { | |
2692 | bs->open_flags &= ~BDRV_O_CACHE_WB; | |
2693 | } | |
425b0148 PB |
2694 | } |
2695 | ||
ea2384d3 FB |
2696 | int bdrv_is_encrypted(BlockDriverState *bs) |
2697 | { | |
2698 | if (bs->backing_hd && bs->backing_hd->encrypted) | |
2699 | return 1; | |
2700 | return bs->encrypted; | |
2701 | } | |
2702 | ||
c0f4ce77 AL |
2703 | int bdrv_key_required(BlockDriverState *bs) |
2704 | { | |
2705 | BlockDriverState *backing_hd = bs->backing_hd; | |
2706 | ||
2707 | if (backing_hd && backing_hd->encrypted && !backing_hd->valid_key) | |
2708 | return 1; | |
2709 | return (bs->encrypted && !bs->valid_key); | |
2710 | } | |
2711 | ||
ea2384d3 FB |
2712 | int bdrv_set_key(BlockDriverState *bs, const char *key) |
2713 | { | |
2714 | int ret; | |
2715 | if (bs->backing_hd && bs->backing_hd->encrypted) { | |
2716 | ret = bdrv_set_key(bs->backing_hd, key); | |
2717 | if (ret < 0) | |
2718 | return ret; | |
2719 | if (!bs->encrypted) | |
2720 | return 0; | |
2721 | } | |
fd04a2ae SH |
2722 | if (!bs->encrypted) { |
2723 | return -EINVAL; | |
2724 | } else if (!bs->drv || !bs->drv->bdrv_set_key) { | |
2725 | return -ENOMEDIUM; | |
2726 | } | |
c0f4ce77 | 2727 | ret = bs->drv->bdrv_set_key(bs, key); |
bb5fc20f AL |
2728 | if (ret < 0) { |
2729 | bs->valid_key = 0; | |
2730 | } else if (!bs->valid_key) { | |
2731 | bs->valid_key = 1; | |
2732 | /* call the change callback now, we skipped it on open */ | |
7d4b4ba5 | 2733 | bdrv_dev_change_media_cb(bs, true); |
bb5fc20f | 2734 | } |
c0f4ce77 | 2735 | return ret; |
ea2384d3 FB |
2736 | } |
2737 | ||
f8d6bba1 | 2738 | const char *bdrv_get_format_name(BlockDriverState *bs) |
ea2384d3 | 2739 | { |
f8d6bba1 | 2740 | return bs->drv ? bs->drv->format_name : NULL; |
ea2384d3 FB |
2741 | } |
2742 | ||
5fafdf24 | 2743 | void bdrv_iterate_format(void (*it)(void *opaque, const char *name), |
ea2384d3 FB |
2744 | void *opaque) |
2745 | { | |
2746 | BlockDriver *drv; | |
2747 | ||
8a22f02a | 2748 | QLIST_FOREACH(drv, &bdrv_drivers, list) { |
ea2384d3 FB |
2749 | it(opaque, drv->format_name); |
2750 | } | |
2751 | } | |
2752 | ||
b338082b FB |
2753 | BlockDriverState *bdrv_find(const char *name) |
2754 | { | |
2755 | BlockDriverState *bs; | |
2756 | ||
1b7bdbc1 SH |
2757 | QTAILQ_FOREACH(bs, &bdrv_states, list) { |
2758 | if (!strcmp(name, bs->device_name)) { | |
b338082b | 2759 | return bs; |
1b7bdbc1 | 2760 | } |
b338082b FB |
2761 | } |
2762 | return NULL; | |
2763 | } | |
2764 | ||
2f399b0a MA |
2765 | BlockDriverState *bdrv_next(BlockDriverState *bs) |
2766 | { | |
2767 | if (!bs) { | |
2768 | return QTAILQ_FIRST(&bdrv_states); | |
2769 | } | |
2770 | return QTAILQ_NEXT(bs, list); | |
2771 | } | |
2772 | ||
51de9760 | 2773 | void bdrv_iterate(void (*it)(void *opaque, BlockDriverState *bs), void *opaque) |
81d0912d FB |
2774 | { |
2775 | BlockDriverState *bs; | |
2776 | ||
1b7bdbc1 | 2777 | QTAILQ_FOREACH(bs, &bdrv_states, list) { |
51de9760 | 2778 | it(opaque, bs); |
81d0912d FB |
2779 | } |
2780 | } | |
2781 | ||
ea2384d3 FB |
2782 | const char *bdrv_get_device_name(BlockDriverState *bs) |
2783 | { | |
2784 | return bs->device_name; | |
2785 | } | |
2786 | ||
c8433287 MA |
2787 | int bdrv_get_flags(BlockDriverState *bs) |
2788 | { | |
2789 | return bs->open_flags; | |
2790 | } | |
2791 | ||
c6ca28d6 AL |
2792 | void bdrv_flush_all(void) |
2793 | { | |
2794 | BlockDriverState *bs; | |
2795 | ||
1b7bdbc1 | 2796 | QTAILQ_FOREACH(bs, &bdrv_states, list) { |
29cdb251 | 2797 | bdrv_flush(bs); |
1b7bdbc1 | 2798 | } |
c6ca28d6 AL |
2799 | } |
2800 | ||
f2feebbd KW |
2801 | int bdrv_has_zero_init(BlockDriverState *bs) |
2802 | { | |
2803 | assert(bs->drv); | |
2804 | ||
336c1c12 KW |
2805 | if (bs->drv->bdrv_has_zero_init) { |
2806 | return bs->drv->bdrv_has_zero_init(bs); | |
f2feebbd KW |
2807 | } |
2808 | ||
2809 | return 1; | |
2810 | } | |
2811 | ||
376ae3f1 SH |
2812 | typedef struct BdrvCoIsAllocatedData { |
2813 | BlockDriverState *bs; | |
b35b2bba | 2814 | BlockDriverState *base; |
376ae3f1 SH |
2815 | int64_t sector_num; |
2816 | int nb_sectors; | |
2817 | int *pnum; | |
2818 | int ret; | |
2819 | bool done; | |
2820 | } BdrvCoIsAllocatedData; | |
2821 | ||
f58c7b35 TS |
2822 | /* |
2823 | * Returns true iff the specified sector is present in the disk image. Drivers | |
2824 | * not implementing the functionality are assumed to not support backing files, | |
2825 | * hence all their sectors are reported as allocated. | |
2826 | * | |
bd9533e3 SH |
2827 | * If 'sector_num' is beyond the end of the disk image the return value is 0 |
2828 | * and 'pnum' is set to 0. | |
2829 | * | |
f58c7b35 TS |
2830 | * 'pnum' is set to the number of sectors (including and immediately following |
2831 | * the specified sector) that are known to be in the same | |
2832 | * allocated/unallocated state. | |
2833 | * | |
bd9533e3 SH |
2834 | * 'nb_sectors' is the max value 'pnum' should be set to. If nb_sectors goes |
2835 | * beyond the end of the disk image it will be clamped. | |
f58c7b35 | 2836 | */ |
060f51c9 SH |
2837 | int coroutine_fn bdrv_co_is_allocated(BlockDriverState *bs, int64_t sector_num, |
2838 | int nb_sectors, int *pnum) | |
f58c7b35 | 2839 | { |
bd9533e3 SH |
2840 | int64_t n; |
2841 | ||
2842 | if (sector_num >= bs->total_sectors) { | |
2843 | *pnum = 0; | |
2844 | return 0; | |
2845 | } | |
2846 | ||
2847 | n = bs->total_sectors - sector_num; | |
2848 | if (n < nb_sectors) { | |
2849 | nb_sectors = n; | |
2850 | } | |
2851 | ||
6aebab14 | 2852 | if (!bs->drv->bdrv_co_is_allocated) { |
bd9533e3 | 2853 | *pnum = nb_sectors; |
f58c7b35 TS |
2854 | return 1; |
2855 | } | |
6aebab14 | 2856 | |
060f51c9 SH |
2857 | return bs->drv->bdrv_co_is_allocated(bs, sector_num, nb_sectors, pnum); |
2858 | } | |
2859 | ||
2860 | /* Coroutine wrapper for bdrv_is_allocated() */ | |
2861 | static void coroutine_fn bdrv_is_allocated_co_entry(void *opaque) | |
2862 | { | |
2863 | BdrvCoIsAllocatedData *data = opaque; | |
2864 | BlockDriverState *bs = data->bs; | |
2865 | ||
2866 | data->ret = bdrv_co_is_allocated(bs, data->sector_num, data->nb_sectors, | |
2867 | data->pnum); | |
2868 | data->done = true; | |
2869 | } | |
2870 | ||
2871 | /* | |
2872 | * Synchronous wrapper around bdrv_co_is_allocated(). | |
2873 | * | |
2874 | * See bdrv_co_is_allocated() for details. | |
2875 | */ | |
2876 | int bdrv_is_allocated(BlockDriverState *bs, int64_t sector_num, int nb_sectors, | |
2877 | int *pnum) | |
2878 | { | |
6aebab14 SH |
2879 | Coroutine *co; |
2880 | BdrvCoIsAllocatedData data = { | |
2881 | .bs = bs, | |
2882 | .sector_num = sector_num, | |
2883 | .nb_sectors = nb_sectors, | |
2884 | .pnum = pnum, | |
2885 | .done = false, | |
2886 | }; | |
2887 | ||
2888 | co = qemu_coroutine_create(bdrv_is_allocated_co_entry); | |
2889 | qemu_coroutine_enter(co, &data); | |
2890 | while (!data.done) { | |
2891 | qemu_aio_wait(); | |
2892 | } | |
2893 | return data.ret; | |
f58c7b35 TS |
2894 | } |
2895 | ||
188a7bbf PB |
2896 | /* |
2897 | * Given an image chain: ... -> [BASE] -> [INTER1] -> [INTER2] -> [TOP] | |
2898 | * | |
2899 | * Return true if the given sector is allocated in any image between | |
2900 | * BASE and TOP (inclusive). BASE can be NULL to check if the given | |
2901 | * sector is allocated in any image of the chain. Return false otherwise. | |
2902 | * | |
2903 | * 'pnum' is set to the number of sectors (including and immediately following | |
2904 | * the specified sector) that are known to be in the same | |
2905 | * allocated/unallocated state. | |
2906 | * | |
2907 | */ | |
2908 | int coroutine_fn bdrv_co_is_allocated_above(BlockDriverState *top, | |
2909 | BlockDriverState *base, | |
2910 | int64_t sector_num, | |
2911 | int nb_sectors, int *pnum) | |
2912 | { | |
2913 | BlockDriverState *intermediate; | |
2914 | int ret, n = nb_sectors; | |
2915 | ||
2916 | intermediate = top; | |
2917 | while (intermediate && intermediate != base) { | |
2918 | int pnum_inter; | |
2919 | ret = bdrv_co_is_allocated(intermediate, sector_num, nb_sectors, | |
2920 | &pnum_inter); | |
2921 | if (ret < 0) { | |
2922 | return ret; | |
2923 | } else if (ret) { | |
2924 | *pnum = pnum_inter; | |
2925 | return 1; | |
2926 | } | |
2927 | ||
2928 | /* | |
2929 | * [sector_num, nb_sectors] is unallocated on top but intermediate | |
2930 | * might have | |
2931 | * | |
2932 | * [sector_num+x, nr_sectors] allocated. | |
2933 | */ | |
63ba17d3 VI |
2934 | if (n > pnum_inter && |
2935 | (intermediate == top || | |
2936 | sector_num + pnum_inter < intermediate->total_sectors)) { | |
188a7bbf PB |
2937 | n = pnum_inter; |
2938 | } | |
2939 | ||
2940 | intermediate = intermediate->backing_hd; | |
2941 | } | |
2942 | ||
2943 | *pnum = n; | |
2944 | return 0; | |
2945 | } | |
2946 | ||
b35b2bba MR |
2947 | /* Coroutine wrapper for bdrv_is_allocated_above() */ |
2948 | static void coroutine_fn bdrv_is_allocated_above_co_entry(void *opaque) | |
2949 | { | |
2950 | BdrvCoIsAllocatedData *data = opaque; | |
2951 | BlockDriverState *top = data->bs; | |
2952 | BlockDriverState *base = data->base; | |
2953 | ||
2954 | data->ret = bdrv_co_is_allocated_above(top, base, data->sector_num, | |
2955 | data->nb_sectors, data->pnum); | |
2956 | data->done = true; | |
2957 | } | |
2958 | ||
2959 | /* | |
2960 | * Synchronous wrapper around bdrv_co_is_allocated_above(). | |
2961 | * | |
2962 | * See bdrv_co_is_allocated_above() for details. | |
2963 | */ | |
2964 | int bdrv_is_allocated_above(BlockDriverState *top, BlockDriverState *base, | |
2965 | int64_t sector_num, int nb_sectors, int *pnum) | |
2966 | { | |
2967 | Coroutine *co; | |
2968 | BdrvCoIsAllocatedData data = { | |
2969 | .bs = top, | |
2970 | .base = base, | |
2971 | .sector_num = sector_num, | |
2972 | .nb_sectors = nb_sectors, | |
2973 | .pnum = pnum, | |
2974 | .done = false, | |
2975 | }; | |
2976 | ||
2977 | co = qemu_coroutine_create(bdrv_is_allocated_above_co_entry); | |
2978 | qemu_coroutine_enter(co, &data); | |
2979 | while (!data.done) { | |
2980 | qemu_aio_wait(); | |
2981 | } | |
2982 | return data.ret; | |
2983 | } | |
2984 | ||
ac84adac | 2985 | BlockInfo *bdrv_query_info(BlockDriverState *bs) |
b338082b | 2986 | { |
ac84adac PB |
2987 | BlockInfo *info = g_malloc0(sizeof(*info)); |
2988 | info->device = g_strdup(bs->device_name); | |
2989 | info->type = g_strdup("unknown"); | |
2990 | info->locked = bdrv_dev_is_medium_locked(bs); | |
2991 | info->removable = bdrv_dev_has_removable_media(bs); | |
b338082b | 2992 | |
ac84adac PB |
2993 | if (bdrv_dev_has_removable_media(bs)) { |
2994 | info->has_tray_open = true; | |
2995 | info->tray_open = bdrv_dev_is_tray_open(bs); | |
2996 | } | |
d15e5465 | 2997 | |
ac84adac PB |
2998 | if (bdrv_iostatus_is_enabled(bs)) { |
2999 | info->has_io_status = true; | |
3000 | info->io_status = bs->iostatus; | |
3001 | } | |
d15e5465 | 3002 | |
b9a9b3a4 PB |
3003 | if (bs->dirty_bitmap) { |
3004 | info->has_dirty = true; | |
3005 | info->dirty = g_malloc0(sizeof(*info->dirty)); | |
acc906c6 | 3006 | info->dirty->count = bdrv_get_dirty_count(bs) * BDRV_SECTOR_SIZE; |
50717e94 PB |
3007 | info->dirty->granularity = |
3008 | ((int64_t) BDRV_SECTOR_SIZE << hbitmap_granularity(bs->dirty_bitmap)); | |
b9a9b3a4 | 3009 | } |
d15e5465 | 3010 | |
ac84adac PB |
3011 | if (bs->drv) { |
3012 | info->has_inserted = true; | |
3013 | info->inserted = g_malloc0(sizeof(*info->inserted)); | |
3014 | info->inserted->file = g_strdup(bs->filename); | |
3015 | info->inserted->ro = bs->read_only; | |
3016 | info->inserted->drv = g_strdup(bs->drv->format_name); | |
3017 | info->inserted->encrypted = bs->encrypted; | |
3018 | info->inserted->encryption_key_missing = bdrv_key_required(bs); | |
3019 | ||
3020 | if (bs->backing_file[0]) { | |
3021 | info->inserted->has_backing_file = true; | |
3022 | info->inserted->backing_file = g_strdup(bs->backing_file); | |
e4def80b | 3023 | } |
f04ef601 | 3024 | |
ac84adac PB |
3025 | info->inserted->backing_file_depth = bdrv_get_backing_file_depth(bs); |
3026 | ||
3027 | if (bs->io_limits_enabled) { | |
3028 | info->inserted->bps = | |
3029 | bs->io_limits.bps[BLOCK_IO_LIMIT_TOTAL]; | |
3030 | info->inserted->bps_rd = | |
3031 | bs->io_limits.bps[BLOCK_IO_LIMIT_READ]; | |
3032 | info->inserted->bps_wr = | |
3033 | bs->io_limits.bps[BLOCK_IO_LIMIT_WRITE]; | |
3034 | info->inserted->iops = | |
3035 | bs->io_limits.iops[BLOCK_IO_LIMIT_TOTAL]; | |
3036 | info->inserted->iops_rd = | |
3037 | bs->io_limits.iops[BLOCK_IO_LIMIT_READ]; | |
3038 | info->inserted->iops_wr = | |
3039 | bs->io_limits.iops[BLOCK_IO_LIMIT_WRITE]; | |
f04ef601 | 3040 | } |
ac84adac PB |
3041 | } |
3042 | return info; | |
3043 | } | |
f04ef601 | 3044 | |
ac84adac PB |
3045 | BlockInfoList *qmp_query_block(Error **errp) |
3046 | { | |
3047 | BlockInfoList *head = NULL, **p_next = &head; | |
3048 | BlockDriverState *bs; | |
727f005e | 3049 | |
ac84adac PB |
3050 | QTAILQ_FOREACH(bs, &bdrv_states, list) { |
3051 | BlockInfoList *info = g_malloc0(sizeof(*info)); | |
3052 | info->value = bdrv_query_info(bs); | |
d15e5465 | 3053 | |
ac84adac PB |
3054 | *p_next = info; |
3055 | p_next = &info->next; | |
b338082b | 3056 | } |
d15e5465 | 3057 | |
b2023818 | 3058 | return head; |
b338082b | 3059 | } |
a36e69dd | 3060 | |
9887b616 | 3061 | BlockStats *bdrv_query_stats(const BlockDriverState *bs) |
f11f57e4 LC |
3062 | { |
3063 | BlockStats *s; | |
3064 | ||
3065 | s = g_malloc0(sizeof(*s)); | |
3066 | ||
3067 | if (bs->device_name[0]) { | |
3068 | s->has_device = true; | |
3069 | s->device = g_strdup(bs->device_name); | |
294cc35f KW |
3070 | } |
3071 | ||
f11f57e4 LC |
3072 | s->stats = g_malloc0(sizeof(*s->stats)); |
3073 | s->stats->rd_bytes = bs->nr_bytes[BDRV_ACCT_READ]; | |
3074 | s->stats->wr_bytes = bs->nr_bytes[BDRV_ACCT_WRITE]; | |
3075 | s->stats->rd_operations = bs->nr_ops[BDRV_ACCT_READ]; | |
3076 | s->stats->wr_operations = bs->nr_ops[BDRV_ACCT_WRITE]; | |
3077 | s->stats->wr_highest_offset = bs->wr_highest_sector * BDRV_SECTOR_SIZE; | |
3078 | s->stats->flush_operations = bs->nr_ops[BDRV_ACCT_FLUSH]; | |
3079 | s->stats->wr_total_time_ns = bs->total_time_ns[BDRV_ACCT_WRITE]; | |
3080 | s->stats->rd_total_time_ns = bs->total_time_ns[BDRV_ACCT_READ]; | |
3081 | s->stats->flush_total_time_ns = bs->total_time_ns[BDRV_ACCT_FLUSH]; | |
3082 | ||
294cc35f | 3083 | if (bs->file) { |
f11f57e4 | 3084 | s->has_parent = true; |
9887b616 | 3085 | s->parent = bdrv_query_stats(bs->file); |
294cc35f KW |
3086 | } |
3087 | ||
f11f57e4 | 3088 | return s; |
294cc35f KW |
3089 | } |
3090 | ||
f11f57e4 | 3091 | BlockStatsList *qmp_query_blockstats(Error **errp) |
218a536a | 3092 | { |
9887b616 | 3093 | BlockStatsList *head = NULL, **p_next = &head; |
a36e69dd TS |
3094 | BlockDriverState *bs; |
3095 | ||
1b7bdbc1 | 3096 | QTAILQ_FOREACH(bs, &bdrv_states, list) { |
f11f57e4 | 3097 | BlockStatsList *info = g_malloc0(sizeof(*info)); |
9887b616 | 3098 | info->value = bdrv_query_stats(bs); |
f11f57e4 | 3099 | |
9887b616 PB |
3100 | *p_next = info; |
3101 | p_next = &info->next; | |
a36e69dd | 3102 | } |
218a536a | 3103 | |
f11f57e4 | 3104 | return head; |
a36e69dd | 3105 | } |
ea2384d3 | 3106 | |
045df330 AL |
3107 | const char *bdrv_get_encrypted_filename(BlockDriverState *bs) |
3108 | { | |
3109 | if (bs->backing_hd && bs->backing_hd->encrypted) | |
3110 | return bs->backing_file; | |
3111 | else if (bs->encrypted) | |
3112 | return bs->filename; | |
3113 | else | |
3114 | return NULL; | |
3115 | } | |
3116 | ||
5fafdf24 | 3117 | void bdrv_get_backing_filename(BlockDriverState *bs, |
83f64091 FB |
3118 | char *filename, int filename_size) |
3119 | { | |
3574c608 | 3120 | pstrcpy(filename, filename_size, bs->backing_file); |
83f64091 FB |
3121 | } |
3122 | ||
5fafdf24 | 3123 | int bdrv_write_compressed(BlockDriverState *bs, int64_t sector_num, |
faea38e7 FB |
3124 | const uint8_t *buf, int nb_sectors) |
3125 | { | |
3126 | BlockDriver *drv = bs->drv; | |
3127 | if (!drv) | |
19cb3738 | 3128 | return -ENOMEDIUM; |
faea38e7 FB |
3129 | if (!drv->bdrv_write_compressed) |
3130 | return -ENOTSUP; | |
fbb7b4e0 KW |
3131 | if (bdrv_check_request(bs, sector_num, nb_sectors)) |
3132 | return -EIO; | |
a55eb92c | 3133 | |
1755da16 | 3134 | assert(!bs->dirty_bitmap); |
a55eb92c | 3135 | |
faea38e7 FB |
3136 | return drv->bdrv_write_compressed(bs, sector_num, buf, nb_sectors); |
3137 | } | |
3b46e624 | 3138 | |
faea38e7 FB |
3139 | int bdrv_get_info(BlockDriverState *bs, BlockDriverInfo *bdi) |
3140 | { | |
3141 | BlockDriver *drv = bs->drv; | |
3142 | if (!drv) | |
19cb3738 | 3143 | return -ENOMEDIUM; |
faea38e7 FB |
3144 | if (!drv->bdrv_get_info) |
3145 | return -ENOTSUP; | |
3146 | memset(bdi, 0, sizeof(*bdi)); | |
3147 | return drv->bdrv_get_info(bs, bdi); | |
3148 | } | |
3149 | ||
45566e9c CH |
3150 | int bdrv_save_vmstate(BlockDriverState *bs, const uint8_t *buf, |
3151 | int64_t pos, int size) | |
178e08a5 AL |
3152 | { |
3153 | BlockDriver *drv = bs->drv; | |
3154 | if (!drv) | |
3155 | return -ENOMEDIUM; | |
7cdb1f6d MK |
3156 | if (drv->bdrv_save_vmstate) |
3157 | return drv->bdrv_save_vmstate(bs, buf, pos, size); | |
3158 | if (bs->file) | |
3159 | return bdrv_save_vmstate(bs->file, buf, pos, size); | |
3160 | return -ENOTSUP; | |
178e08a5 AL |
3161 | } |
3162 | ||
45566e9c CH |
3163 | int bdrv_load_vmstate(BlockDriverState *bs, uint8_t *buf, |
3164 | int64_t pos, int size) | |
178e08a5 AL |
3165 | { |
3166 | BlockDriver *drv = bs->drv; | |
3167 | if (!drv) | |
3168 | return -ENOMEDIUM; | |
7cdb1f6d MK |
3169 | if (drv->bdrv_load_vmstate) |
3170 | return drv->bdrv_load_vmstate(bs, buf, pos, size); | |
3171 | if (bs->file) | |
3172 | return bdrv_load_vmstate(bs->file, buf, pos, size); | |
3173 | return -ENOTSUP; | |
178e08a5 AL |
3174 | } |
3175 | ||
8b9b0cc2 KW |
3176 | void bdrv_debug_event(BlockDriverState *bs, BlkDebugEvent event) |
3177 | { | |
3178 | BlockDriver *drv = bs->drv; | |
3179 | ||
3180 | if (!drv || !drv->bdrv_debug_event) { | |
3181 | return; | |
3182 | } | |
3183 | ||
0ed8b6f6 | 3184 | drv->bdrv_debug_event(bs, event); |
41c695c7 KW |
3185 | } |
3186 | ||
3187 | int bdrv_debug_breakpoint(BlockDriverState *bs, const char *event, | |
3188 | const char *tag) | |
3189 | { | |
3190 | while (bs && bs->drv && !bs->drv->bdrv_debug_breakpoint) { | |
3191 | bs = bs->file; | |
3192 | } | |
3193 | ||
3194 | if (bs && bs->drv && bs->drv->bdrv_debug_breakpoint) { | |
3195 | return bs->drv->bdrv_debug_breakpoint(bs, event, tag); | |
3196 | } | |
3197 | ||
3198 | return -ENOTSUP; | |
3199 | } | |
3200 | ||
3201 | int bdrv_debug_resume(BlockDriverState *bs, const char *tag) | |
3202 | { | |
3203 | while (bs && bs->drv && !bs->drv->bdrv_debug_resume) { | |
3204 | bs = bs->file; | |
3205 | } | |
8b9b0cc2 | 3206 | |
41c695c7 KW |
3207 | if (bs && bs->drv && bs->drv->bdrv_debug_resume) { |
3208 | return bs->drv->bdrv_debug_resume(bs, tag); | |
3209 | } | |
3210 | ||
3211 | return -ENOTSUP; | |
3212 | } | |
3213 | ||
3214 | bool bdrv_debug_is_suspended(BlockDriverState *bs, const char *tag) | |
3215 | { | |
3216 | while (bs && bs->drv && !bs->drv->bdrv_debug_is_suspended) { | |
3217 | bs = bs->file; | |
3218 | } | |
3219 | ||
3220 | if (bs && bs->drv && bs->drv->bdrv_debug_is_suspended) { | |
3221 | return bs->drv->bdrv_debug_is_suspended(bs, tag); | |
3222 | } | |
3223 | ||
3224 | return false; | |
8b9b0cc2 KW |
3225 | } |
3226 | ||
faea38e7 FB |
3227 | /**************************************************************/ |
3228 | /* handling of snapshots */ | |
3229 | ||
feeee5ac MDCF |
3230 | int bdrv_can_snapshot(BlockDriverState *bs) |
3231 | { | |
3232 | BlockDriver *drv = bs->drv; | |
07b70bfb | 3233 | if (!drv || !bdrv_is_inserted(bs) || bdrv_is_read_only(bs)) { |
feeee5ac MDCF |
3234 | return 0; |
3235 | } | |
3236 | ||
3237 | if (!drv->bdrv_snapshot_create) { | |
3238 | if (bs->file != NULL) { | |
3239 | return bdrv_can_snapshot(bs->file); | |
3240 | } | |
3241 | return 0; | |
3242 | } | |
3243 | ||
3244 | return 1; | |
3245 | } | |
3246 | ||
199630b6 BS |
3247 | int bdrv_is_snapshot(BlockDriverState *bs) |
3248 | { | |
3249 | return !!(bs->open_flags & BDRV_O_SNAPSHOT); | |
3250 | } | |
3251 | ||
f9092b10 MA |
3252 | BlockDriverState *bdrv_snapshots(void) |
3253 | { | |
3254 | BlockDriverState *bs; | |
3255 | ||
3ac906f7 | 3256 | if (bs_snapshots) { |
f9092b10 | 3257 | return bs_snapshots; |
3ac906f7 | 3258 | } |
f9092b10 MA |
3259 | |
3260 | bs = NULL; | |
3261 | while ((bs = bdrv_next(bs))) { | |
3262 | if (bdrv_can_snapshot(bs)) { | |
3ac906f7 MA |
3263 | bs_snapshots = bs; |
3264 | return bs; | |
f9092b10 MA |
3265 | } |
3266 | } | |
3267 | return NULL; | |
f9092b10 MA |
3268 | } |
3269 | ||
5fafdf24 | 3270 | int bdrv_snapshot_create(BlockDriverState *bs, |
faea38e7 FB |
3271 | QEMUSnapshotInfo *sn_info) |
3272 | { | |
3273 | BlockDriver *drv = bs->drv; | |
3274 | if (!drv) | |
19cb3738 | 3275 | return -ENOMEDIUM; |
7cdb1f6d MK |
3276 | if (drv->bdrv_snapshot_create) |
3277 | return drv->bdrv_snapshot_create(bs, sn_info); | |
3278 | if (bs->file) | |
3279 | return bdrv_snapshot_create(bs->file, sn_info); | |
3280 | return -ENOTSUP; | |
faea38e7 FB |
3281 | } |
3282 | ||
5fafdf24 | 3283 | int bdrv_snapshot_goto(BlockDriverState *bs, |
faea38e7 FB |
3284 | const char *snapshot_id) |
3285 | { | |
3286 | BlockDriver *drv = bs->drv; | |
7cdb1f6d MK |
3287 | int ret, open_ret; |
3288 | ||
faea38e7 | 3289 | if (!drv) |
19cb3738 | 3290 | return -ENOMEDIUM; |
7cdb1f6d MK |
3291 | if (drv->bdrv_snapshot_goto) |
3292 | return drv->bdrv_snapshot_goto(bs, snapshot_id); | |
3293 | ||
3294 | if (bs->file) { | |
3295 | drv->bdrv_close(bs); | |
3296 | ret = bdrv_snapshot_goto(bs->file, snapshot_id); | |
1a86938f | 3297 | open_ret = drv->bdrv_open(bs, NULL, bs->open_flags); |
7cdb1f6d MK |
3298 | if (open_ret < 0) { |
3299 | bdrv_delete(bs->file); | |
3300 | bs->drv = NULL; | |
3301 | return open_ret; | |
3302 | } | |
3303 | return ret; | |
3304 | } | |
3305 | ||
3306 | return -ENOTSUP; | |
faea38e7 FB |
3307 | } |
3308 | ||
3309 | int bdrv_snapshot_delete(BlockDriverState *bs, const char *snapshot_id) | |
3310 | { | |
3311 | BlockDriver *drv = bs->drv; | |
3312 | if (!drv) | |
19cb3738 | 3313 | return -ENOMEDIUM; |
7cdb1f6d MK |
3314 | if (drv->bdrv_snapshot_delete) |
3315 | return drv->bdrv_snapshot_delete(bs, snapshot_id); | |
3316 | if (bs->file) | |
3317 | return bdrv_snapshot_delete(bs->file, snapshot_id); | |
3318 | return -ENOTSUP; | |
faea38e7 FB |
3319 | } |
3320 | ||
5fafdf24 | 3321 | int bdrv_snapshot_list(BlockDriverState *bs, |
faea38e7 FB |
3322 | QEMUSnapshotInfo **psn_info) |
3323 | { | |
3324 | BlockDriver *drv = bs->drv; | |
3325 | if (!drv) | |
19cb3738 | 3326 | return -ENOMEDIUM; |
7cdb1f6d MK |
3327 | if (drv->bdrv_snapshot_list) |
3328 | return drv->bdrv_snapshot_list(bs, psn_info); | |
3329 | if (bs->file) | |
3330 | return bdrv_snapshot_list(bs->file, psn_info); | |
3331 | return -ENOTSUP; | |
faea38e7 FB |
3332 | } |
3333 | ||
51ef6727 | 3334 | int bdrv_snapshot_load_tmp(BlockDriverState *bs, |
3335 | const char *snapshot_name) | |
3336 | { | |
3337 | BlockDriver *drv = bs->drv; | |
3338 | if (!drv) { | |
3339 | return -ENOMEDIUM; | |
3340 | } | |
3341 | if (!bs->read_only) { | |
3342 | return -EINVAL; | |
3343 | } | |
3344 | if (drv->bdrv_snapshot_load_tmp) { | |
3345 | return drv->bdrv_snapshot_load_tmp(bs, snapshot_name); | |
3346 | } | |
3347 | return -ENOTSUP; | |
3348 | } | |
3349 | ||
b1b1d783 JC |
3350 | /* backing_file can either be relative, or absolute, or a protocol. If it is |
3351 | * relative, it must be relative to the chain. So, passing in bs->filename | |
3352 | * from a BDS as backing_file should not be done, as that may be relative to | |
3353 | * the CWD rather than the chain. */ | |
e8a6bb9c MT |
3354 | BlockDriverState *bdrv_find_backing_image(BlockDriverState *bs, |
3355 | const char *backing_file) | |
3356 | { | |
b1b1d783 JC |
3357 | char *filename_full = NULL; |
3358 | char *backing_file_full = NULL; | |
3359 | char *filename_tmp = NULL; | |
3360 | int is_protocol = 0; | |
3361 | BlockDriverState *curr_bs = NULL; | |
3362 | BlockDriverState *retval = NULL; | |
3363 | ||
3364 | if (!bs || !bs->drv || !backing_file) { | |
e8a6bb9c MT |
3365 | return NULL; |
3366 | } | |
3367 | ||
b1b1d783 JC |
3368 | filename_full = g_malloc(PATH_MAX); |
3369 | backing_file_full = g_malloc(PATH_MAX); | |
3370 | filename_tmp = g_malloc(PATH_MAX); | |
3371 | ||
3372 | is_protocol = path_has_protocol(backing_file); | |
3373 | ||
3374 | for (curr_bs = bs; curr_bs->backing_hd; curr_bs = curr_bs->backing_hd) { | |
3375 | ||
3376 | /* If either of the filename paths is actually a protocol, then | |
3377 | * compare unmodified paths; otherwise make paths relative */ | |
3378 | if (is_protocol || path_has_protocol(curr_bs->backing_file)) { | |
3379 | if (strcmp(backing_file, curr_bs->backing_file) == 0) { | |
3380 | retval = curr_bs->backing_hd; | |
3381 | break; | |
3382 | } | |
e8a6bb9c | 3383 | } else { |
b1b1d783 JC |
3384 | /* If not an absolute filename path, make it relative to the current |
3385 | * image's filename path */ | |
3386 | path_combine(filename_tmp, PATH_MAX, curr_bs->filename, | |
3387 | backing_file); | |
3388 | ||
3389 | /* We are going to compare absolute pathnames */ | |
3390 | if (!realpath(filename_tmp, filename_full)) { | |
3391 | continue; | |
3392 | } | |
3393 | ||
3394 | /* We need to make sure the backing filename we are comparing against | |
3395 | * is relative to the current image filename (or absolute) */ | |
3396 | path_combine(filename_tmp, PATH_MAX, curr_bs->filename, | |
3397 | curr_bs->backing_file); | |
3398 | ||
3399 | if (!realpath(filename_tmp, backing_file_full)) { | |
3400 | continue; | |
3401 | } | |
3402 | ||
3403 | if (strcmp(backing_file_full, filename_full) == 0) { | |
3404 | retval = curr_bs->backing_hd; | |
3405 | break; | |
3406 | } | |
e8a6bb9c MT |
3407 | } |
3408 | } | |
3409 | ||
b1b1d783 JC |
3410 | g_free(filename_full); |
3411 | g_free(backing_file_full); | |
3412 | g_free(filename_tmp); | |
3413 | return retval; | |
e8a6bb9c MT |
3414 | } |
3415 | ||
f198fd1c BC |
3416 | int bdrv_get_backing_file_depth(BlockDriverState *bs) |
3417 | { | |
3418 | if (!bs->drv) { | |
3419 | return 0; | |
3420 | } | |
3421 | ||
3422 | if (!bs->backing_hd) { | |
3423 | return 0; | |
3424 | } | |
3425 | ||
3426 | return 1 + bdrv_get_backing_file_depth(bs->backing_hd); | |
3427 | } | |
3428 | ||
79fac568 JC |
3429 | BlockDriverState *bdrv_find_base(BlockDriverState *bs) |
3430 | { | |
3431 | BlockDriverState *curr_bs = NULL; | |
3432 | ||
3433 | if (!bs) { | |
3434 | return NULL; | |
3435 | } | |
3436 | ||
3437 | curr_bs = bs; | |
3438 | ||
3439 | while (curr_bs->backing_hd) { | |
3440 | curr_bs = curr_bs->backing_hd; | |
3441 | } | |
3442 | return curr_bs; | |
3443 | } | |
3444 | ||
faea38e7 FB |
3445 | #define NB_SUFFIXES 4 |
3446 | ||
3447 | char *get_human_readable_size(char *buf, int buf_size, int64_t size) | |
3448 | { | |
3449 | static const char suffixes[NB_SUFFIXES] = "KMGT"; | |
3450 | int64_t base; | |
3451 | int i; | |
3452 | ||
3453 | if (size <= 999) { | |
3454 | snprintf(buf, buf_size, "%" PRId64, size); | |
3455 | } else { | |
3456 | base = 1024; | |
3457 | for(i = 0; i < NB_SUFFIXES; i++) { | |
3458 | if (size < (10 * base)) { | |
5fafdf24 | 3459 | snprintf(buf, buf_size, "%0.1f%c", |
faea38e7 FB |
3460 | (double)size / base, |
3461 | suffixes[i]); | |
3462 | break; | |
3463 | } else if (size < (1000 * base) || i == (NB_SUFFIXES - 1)) { | |
5fafdf24 | 3464 | snprintf(buf, buf_size, "%" PRId64 "%c", |
faea38e7 FB |
3465 | ((size + (base >> 1)) / base), |
3466 | suffixes[i]); | |
3467 | break; | |
3468 | } | |
3469 | base = base * 1024; | |
3470 | } | |
3471 | } | |
3472 | return buf; | |
3473 | } | |
3474 | ||
3475 | char *bdrv_snapshot_dump(char *buf, int buf_size, QEMUSnapshotInfo *sn) | |
3476 | { | |
3477 | char buf1[128], date_buf[128], clock_buf[128]; | |
3478 | struct tm tm; | |
3479 | time_t ti; | |
3480 | int64_t secs; | |
3481 | ||
3482 | if (!sn) { | |
5fafdf24 TS |
3483 | snprintf(buf, buf_size, |
3484 | "%-10s%-20s%7s%20s%15s", | |
faea38e7 FB |
3485 | "ID", "TAG", "VM SIZE", "DATE", "VM CLOCK"); |
3486 | } else { | |
3487 | ti = sn->date_sec; | |
3488 | localtime_r(&ti, &tm); | |
3489 | strftime(date_buf, sizeof(date_buf), | |
3490 | "%Y-%m-%d %H:%M:%S", &tm); | |
3491 | secs = sn->vm_clock_nsec / 1000000000; | |
3492 | snprintf(clock_buf, sizeof(clock_buf), | |
3493 | "%02d:%02d:%02d.%03d", | |
3494 | (int)(secs / 3600), | |
3495 | (int)((secs / 60) % 60), | |
5fafdf24 | 3496 | (int)(secs % 60), |
faea38e7 FB |
3497 | (int)((sn->vm_clock_nsec / 1000000) % 1000)); |
3498 | snprintf(buf, buf_size, | |
5fafdf24 | 3499 | "%-10s%-20s%7s%20s%15s", |
faea38e7 FB |
3500 | sn->id_str, sn->name, |
3501 | get_human_readable_size(buf1, sizeof(buf1), sn->vm_state_size), | |
3502 | date_buf, | |
3503 | clock_buf); | |
3504 | } | |
3505 | return buf; | |
3506 | } | |
3507 | ||
ea2384d3 | 3508 | /**************************************************************/ |
83f64091 | 3509 | /* async I/Os */ |
ea2384d3 | 3510 | |
3b69e4b9 | 3511 | BlockDriverAIOCB *bdrv_aio_readv(BlockDriverState *bs, int64_t sector_num, |
f141eafe | 3512 | QEMUIOVector *qiov, int nb_sectors, |
3b69e4b9 | 3513 | BlockDriverCompletionFunc *cb, void *opaque) |
83f64091 | 3514 | { |
bbf0a440 SH |
3515 | trace_bdrv_aio_readv(bs, sector_num, nb_sectors, opaque); |
3516 | ||
b2a61371 | 3517 | return bdrv_co_aio_rw_vector(bs, sector_num, qiov, nb_sectors, |
8c5873d6 | 3518 | cb, opaque, false); |
ea2384d3 FB |
3519 | } |
3520 | ||
f141eafe AL |
3521 | BlockDriverAIOCB *bdrv_aio_writev(BlockDriverState *bs, int64_t sector_num, |
3522 | QEMUIOVector *qiov, int nb_sectors, | |
3523 | BlockDriverCompletionFunc *cb, void *opaque) | |
ea2384d3 | 3524 | { |
bbf0a440 SH |
3525 | trace_bdrv_aio_writev(bs, sector_num, nb_sectors, opaque); |
3526 | ||
1a6e115b | 3527 | return bdrv_co_aio_rw_vector(bs, sector_num, qiov, nb_sectors, |
8c5873d6 | 3528 | cb, opaque, true); |
83f64091 FB |
3529 | } |
3530 | ||
40b4f539 KW |
3531 | |
3532 | typedef struct MultiwriteCB { | |
3533 | int error; | |
3534 | int num_requests; | |
3535 | int num_callbacks; | |
3536 | struct { | |
3537 | BlockDriverCompletionFunc *cb; | |
3538 | void *opaque; | |
3539 | QEMUIOVector *free_qiov; | |
40b4f539 KW |
3540 | } callbacks[]; |
3541 | } MultiwriteCB; | |
3542 | ||
3543 | static void multiwrite_user_cb(MultiwriteCB *mcb) | |
3544 | { | |
3545 | int i; | |
3546 | ||
3547 | for (i = 0; i < mcb->num_callbacks; i++) { | |
3548 | mcb->callbacks[i].cb(mcb->callbacks[i].opaque, mcb->error); | |
1e1ea48d SH |
3549 | if (mcb->callbacks[i].free_qiov) { |
3550 | qemu_iovec_destroy(mcb->callbacks[i].free_qiov); | |
3551 | } | |
7267c094 | 3552 | g_free(mcb->callbacks[i].free_qiov); |
40b4f539 KW |
3553 | } |
3554 | } | |
3555 | ||
3556 | static void multiwrite_cb(void *opaque, int ret) | |
3557 | { | |
3558 | MultiwriteCB *mcb = opaque; | |
3559 | ||
6d519a5f SH |
3560 | trace_multiwrite_cb(mcb, ret); |
3561 | ||
cb6d3ca0 | 3562 | if (ret < 0 && !mcb->error) { |
40b4f539 | 3563 | mcb->error = ret; |
40b4f539 KW |
3564 | } |
3565 | ||
3566 | mcb->num_requests--; | |
3567 | if (mcb->num_requests == 0) { | |
de189a1b | 3568 | multiwrite_user_cb(mcb); |
7267c094 | 3569 | g_free(mcb); |
40b4f539 KW |
3570 | } |
3571 | } | |
3572 | ||
3573 | static int multiwrite_req_compare(const void *a, const void *b) | |
3574 | { | |
77be4366 CH |
3575 | const BlockRequest *req1 = a, *req2 = b; |
3576 | ||
3577 | /* | |
3578 | * Note that we can't simply subtract req2->sector from req1->sector | |
3579 | * here as that could overflow the return value. | |
3580 | */ | |
3581 | if (req1->sector > req2->sector) { | |
3582 | return 1; | |
3583 | } else if (req1->sector < req2->sector) { | |
3584 | return -1; | |
3585 | } else { | |
3586 | return 0; | |
3587 | } | |
40b4f539 KW |
3588 | } |
3589 | ||
3590 | /* | |
3591 | * Takes a bunch of requests and tries to merge them. Returns the number of | |
3592 | * requests that remain after merging. | |
3593 | */ | |
3594 | static int multiwrite_merge(BlockDriverState *bs, BlockRequest *reqs, | |
3595 | int num_reqs, MultiwriteCB *mcb) | |
3596 | { | |
3597 | int i, outidx; | |
3598 | ||
3599 | // Sort requests by start sector | |
3600 | qsort(reqs, num_reqs, sizeof(*reqs), &multiwrite_req_compare); | |
3601 | ||
3602 | // Check if adjacent requests touch the same clusters. If so, combine them, | |
3603 | // filling up gaps with zero sectors. | |
3604 | outidx = 0; | |
3605 | for (i = 1; i < num_reqs; i++) { | |
3606 | int merge = 0; | |
3607 | int64_t oldreq_last = reqs[outidx].sector + reqs[outidx].nb_sectors; | |
3608 | ||
b6a127a1 | 3609 | // Handle exactly sequential writes and overlapping writes. |
40b4f539 KW |
3610 | if (reqs[i].sector <= oldreq_last) { |
3611 | merge = 1; | |
3612 | } | |
3613 | ||
e2a305fb CH |
3614 | if (reqs[outidx].qiov->niov + reqs[i].qiov->niov + 1 > IOV_MAX) { |
3615 | merge = 0; | |
3616 | } | |
3617 | ||
40b4f539 KW |
3618 | if (merge) { |
3619 | size_t size; | |
7267c094 | 3620 | QEMUIOVector *qiov = g_malloc0(sizeof(*qiov)); |
40b4f539 KW |
3621 | qemu_iovec_init(qiov, |
3622 | reqs[outidx].qiov->niov + reqs[i].qiov->niov + 1); | |
3623 | ||
3624 | // Add the first request to the merged one. If the requests are | |
3625 | // overlapping, drop the last sectors of the first request. | |
3626 | size = (reqs[i].sector - reqs[outidx].sector) << 9; | |
1b093c48 | 3627 | qemu_iovec_concat(qiov, reqs[outidx].qiov, 0, size); |
40b4f539 | 3628 | |
b6a127a1 PB |
3629 | // We should need to add any zeros between the two requests |
3630 | assert (reqs[i].sector <= oldreq_last); | |
40b4f539 KW |
3631 | |
3632 | // Add the second request | |
1b093c48 | 3633 | qemu_iovec_concat(qiov, reqs[i].qiov, 0, reqs[i].qiov->size); |
40b4f539 | 3634 | |
cbf1dff2 | 3635 | reqs[outidx].nb_sectors = qiov->size >> 9; |
40b4f539 KW |
3636 | reqs[outidx].qiov = qiov; |
3637 | ||
3638 | mcb->callbacks[i].free_qiov = reqs[outidx].qiov; | |
3639 | } else { | |
3640 | outidx++; | |
3641 | reqs[outidx].sector = reqs[i].sector; | |
3642 | reqs[outidx].nb_sectors = reqs[i].nb_sectors; | |
3643 | reqs[outidx].qiov = reqs[i].qiov; | |
3644 | } | |
3645 | } | |
3646 | ||
3647 | return outidx + 1; | |
3648 | } | |
3649 | ||
3650 | /* | |
3651 | * Submit multiple AIO write requests at once. | |
3652 | * | |
3653 | * On success, the function returns 0 and all requests in the reqs array have | |
3654 | * been submitted. In error case this function returns -1, and any of the | |
3655 | * requests may or may not be submitted yet. In particular, this means that the | |
3656 | * callback will be called for some of the requests, for others it won't. The | |
3657 | * caller must check the error field of the BlockRequest to wait for the right | |
3658 | * callbacks (if error != 0, no callback will be called). | |
3659 | * | |
3660 | * The implementation may modify the contents of the reqs array, e.g. to merge | |
3661 | * requests. However, the fields opaque and error are left unmodified as they | |
3662 | * are used to signal failure for a single request to the caller. | |
3663 | */ | |
3664 | int bdrv_aio_multiwrite(BlockDriverState *bs, BlockRequest *reqs, int num_reqs) | |
3665 | { | |
40b4f539 KW |
3666 | MultiwriteCB *mcb; |
3667 | int i; | |
3668 | ||
301db7c2 RH |
3669 | /* don't submit writes if we don't have a medium */ |
3670 | if (bs->drv == NULL) { | |
3671 | for (i = 0; i < num_reqs; i++) { | |
3672 | reqs[i].error = -ENOMEDIUM; | |
3673 | } | |
3674 | return -1; | |
3675 | } | |
3676 | ||
40b4f539 KW |
3677 | if (num_reqs == 0) { |
3678 | return 0; | |
3679 | } | |
3680 | ||
3681 | // Create MultiwriteCB structure | |
7267c094 | 3682 | mcb = g_malloc0(sizeof(*mcb) + num_reqs * sizeof(*mcb->callbacks)); |
40b4f539 KW |
3683 | mcb->num_requests = 0; |
3684 | mcb->num_callbacks = num_reqs; | |
3685 | ||
3686 | for (i = 0; i < num_reqs; i++) { | |
3687 | mcb->callbacks[i].cb = reqs[i].cb; | |
3688 | mcb->callbacks[i].opaque = reqs[i].opaque; | |
3689 | } | |
3690 | ||
3691 | // Check for mergable requests | |
3692 | num_reqs = multiwrite_merge(bs, reqs, num_reqs, mcb); | |
3693 | ||
6d519a5f SH |
3694 | trace_bdrv_aio_multiwrite(mcb, mcb->num_callbacks, num_reqs); |
3695 | ||
df9309fb PB |
3696 | /* Run the aio requests. */ |
3697 | mcb->num_requests = num_reqs; | |
40b4f539 | 3698 | for (i = 0; i < num_reqs; i++) { |
ad54ae80 | 3699 | bdrv_aio_writev(bs, reqs[i].sector, reqs[i].qiov, |
40b4f539 | 3700 | reqs[i].nb_sectors, multiwrite_cb, mcb); |
40b4f539 KW |
3701 | } |
3702 | ||
3703 | return 0; | |
40b4f539 KW |
3704 | } |
3705 | ||
83f64091 | 3706 | void bdrv_aio_cancel(BlockDriverAIOCB *acb) |
83f64091 | 3707 | { |
d7331bed | 3708 | acb->aiocb_info->cancel(acb); |
83f64091 FB |
3709 | } |
3710 | ||
98f90dba ZYW |
3711 | /* block I/O throttling */ |
3712 | static bool bdrv_exceed_bps_limits(BlockDriverState *bs, int nb_sectors, | |
3713 | bool is_write, double elapsed_time, uint64_t *wait) | |
3714 | { | |
3715 | uint64_t bps_limit = 0; | |
3716 | double bytes_limit, bytes_base, bytes_res; | |
3717 | double slice_time, wait_time; | |
3718 | ||
3719 | if (bs->io_limits.bps[BLOCK_IO_LIMIT_TOTAL]) { | |
3720 | bps_limit = bs->io_limits.bps[BLOCK_IO_LIMIT_TOTAL]; | |
3721 | } else if (bs->io_limits.bps[is_write]) { | |
3722 | bps_limit = bs->io_limits.bps[is_write]; | |
3723 | } else { | |
3724 | if (wait) { | |
3725 | *wait = 0; | |
3726 | } | |
3727 | ||
3728 | return false; | |
3729 | } | |
3730 | ||
3731 | slice_time = bs->slice_end - bs->slice_start; | |
3732 | slice_time /= (NANOSECONDS_PER_SECOND); | |
3733 | bytes_limit = bps_limit * slice_time; | |
3734 | bytes_base = bs->nr_bytes[is_write] - bs->io_base.bytes[is_write]; | |
3735 | if (bs->io_limits.bps[BLOCK_IO_LIMIT_TOTAL]) { | |
3736 | bytes_base += bs->nr_bytes[!is_write] - bs->io_base.bytes[!is_write]; | |
3737 | } | |
3738 | ||
3739 | /* bytes_base: the bytes of data which have been read/written; and | |
3740 | * it is obtained from the history statistic info. | |
3741 | * bytes_res: the remaining bytes of data which need to be read/written. | |
3742 | * (bytes_base + bytes_res) / bps_limit: used to calcuate | |
3743 | * the total time for completing reading/writting all data. | |
3744 | */ | |
3745 | bytes_res = (unsigned) nb_sectors * BDRV_SECTOR_SIZE; | |
3746 | ||
3747 | if (bytes_base + bytes_res <= bytes_limit) { | |
3748 | if (wait) { | |
3749 | *wait = 0; | |
3750 | } | |
3751 | ||
3752 | return false; | |
3753 | } | |
3754 | ||
3755 | /* Calc approx time to dispatch */ | |
3756 | wait_time = (bytes_base + bytes_res) / bps_limit - elapsed_time; | |
3757 | ||
3758 | /* When the I/O rate at runtime exceeds the limits, | |
3759 | * bs->slice_end need to be extended in order that the current statistic | |
3760 | * info can be kept until the timer fire, so it is increased and tuned | |
3761 | * based on the result of experiment. | |
3762 | */ | |
3763 | bs->slice_time = wait_time * BLOCK_IO_SLICE_TIME * 10; | |
3764 | bs->slice_end += bs->slice_time - 3 * BLOCK_IO_SLICE_TIME; | |
3765 | if (wait) { | |
3766 | *wait = wait_time * BLOCK_IO_SLICE_TIME * 10; | |
3767 | } | |
3768 | ||
3769 | return true; | |
3770 | } | |
3771 | ||
3772 | static bool bdrv_exceed_iops_limits(BlockDriverState *bs, bool is_write, | |
3773 | double elapsed_time, uint64_t *wait) | |
3774 | { | |
3775 | uint64_t iops_limit = 0; | |
3776 | double ios_limit, ios_base; | |
3777 | double slice_time, wait_time; | |
3778 | ||
3779 | if (bs->io_limits.iops[BLOCK_IO_LIMIT_TOTAL]) { | |
3780 | iops_limit = bs->io_limits.iops[BLOCK_IO_LIMIT_TOTAL]; | |
3781 | } else if (bs->io_limits.iops[is_write]) { | |
3782 | iops_limit = bs->io_limits.iops[is_write]; | |
3783 | } else { | |
3784 | if (wait) { | |
3785 | *wait = 0; | |
3786 | } | |
3787 | ||
3788 | return false; | |
3789 | } | |
3790 | ||
3791 | slice_time = bs->slice_end - bs->slice_start; | |
3792 | slice_time /= (NANOSECONDS_PER_SECOND); | |
3793 | ios_limit = iops_limit * slice_time; | |
3794 | ios_base = bs->nr_ops[is_write] - bs->io_base.ios[is_write]; | |
3795 | if (bs->io_limits.iops[BLOCK_IO_LIMIT_TOTAL]) { | |
3796 | ios_base += bs->nr_ops[!is_write] - bs->io_base.ios[!is_write]; | |
3797 | } | |
3798 | ||
3799 | if (ios_base + 1 <= ios_limit) { | |
3800 | if (wait) { | |
3801 | *wait = 0; | |
3802 | } | |
3803 | ||
3804 | return false; | |
3805 | } | |
3806 | ||
3807 | /* Calc approx time to dispatch */ | |
3808 | wait_time = (ios_base + 1) / iops_limit; | |
3809 | if (wait_time > elapsed_time) { | |
3810 | wait_time = wait_time - elapsed_time; | |
3811 | } else { | |
3812 | wait_time = 0; | |
3813 | } | |
3814 | ||
3815 | bs->slice_time = wait_time * BLOCK_IO_SLICE_TIME * 10; | |
3816 | bs->slice_end += bs->slice_time - 3 * BLOCK_IO_SLICE_TIME; | |
3817 | if (wait) { | |
3818 | *wait = wait_time * BLOCK_IO_SLICE_TIME * 10; | |
3819 | } | |
3820 | ||
3821 | return true; | |
3822 | } | |
3823 | ||
3824 | static bool bdrv_exceed_io_limits(BlockDriverState *bs, int nb_sectors, | |
3825 | bool is_write, int64_t *wait) | |
3826 | { | |
3827 | int64_t now, max_wait; | |
3828 | uint64_t bps_wait = 0, iops_wait = 0; | |
3829 | double elapsed_time; | |
3830 | int bps_ret, iops_ret; | |
3831 | ||
3832 | now = qemu_get_clock_ns(vm_clock); | |
3833 | if ((bs->slice_start < now) | |
3834 | && (bs->slice_end > now)) { | |
3835 | bs->slice_end = now + bs->slice_time; | |
3836 | } else { | |
3837 | bs->slice_time = 5 * BLOCK_IO_SLICE_TIME; | |
3838 | bs->slice_start = now; | |
3839 | bs->slice_end = now + bs->slice_time; | |
3840 | ||
3841 | bs->io_base.bytes[is_write] = bs->nr_bytes[is_write]; | |
3842 | bs->io_base.bytes[!is_write] = bs->nr_bytes[!is_write]; | |
3843 | ||
3844 | bs->io_base.ios[is_write] = bs->nr_ops[is_write]; | |
3845 | bs->io_base.ios[!is_write] = bs->nr_ops[!is_write]; | |
3846 | } | |
3847 | ||
3848 | elapsed_time = now - bs->slice_start; | |
3849 | elapsed_time /= (NANOSECONDS_PER_SECOND); | |
3850 | ||
3851 | bps_ret = bdrv_exceed_bps_limits(bs, nb_sectors, | |
3852 | is_write, elapsed_time, &bps_wait); | |
3853 | iops_ret = bdrv_exceed_iops_limits(bs, is_write, | |
3854 | elapsed_time, &iops_wait); | |
3855 | if (bps_ret || iops_ret) { | |
3856 | max_wait = bps_wait > iops_wait ? bps_wait : iops_wait; | |
3857 | if (wait) { | |
3858 | *wait = max_wait; | |
3859 | } | |
3860 | ||
3861 | now = qemu_get_clock_ns(vm_clock); | |
3862 | if (bs->slice_end < now + max_wait) { | |
3863 | bs->slice_end = now + max_wait; | |
3864 | } | |
3865 | ||
3866 | return true; | |
3867 | } | |
3868 | ||
3869 | if (wait) { | |
3870 | *wait = 0; | |
3871 | } | |
3872 | ||
3873 | return false; | |
3874 | } | |
ce1a14dc | 3875 | |
83f64091 FB |
3876 | /**************************************************************/ |
3877 | /* async block device emulation */ | |
3878 | ||
c16b5a2c CH |
3879 | typedef struct BlockDriverAIOCBSync { |
3880 | BlockDriverAIOCB common; | |
3881 | QEMUBH *bh; | |
3882 | int ret; | |
3883 | /* vector translation state */ | |
3884 | QEMUIOVector *qiov; | |
3885 | uint8_t *bounce; | |
3886 | int is_write; | |
3887 | } BlockDriverAIOCBSync; | |
3888 | ||
3889 | static void bdrv_aio_cancel_em(BlockDriverAIOCB *blockacb) | |
3890 | { | |
b666d239 KW |
3891 | BlockDriverAIOCBSync *acb = |
3892 | container_of(blockacb, BlockDriverAIOCBSync, common); | |
6a7ad299 | 3893 | qemu_bh_delete(acb->bh); |
36afc451 | 3894 | acb->bh = NULL; |
c16b5a2c CH |
3895 | qemu_aio_release(acb); |
3896 | } | |
3897 | ||
d7331bed | 3898 | static const AIOCBInfo bdrv_em_aiocb_info = { |
c16b5a2c CH |
3899 | .aiocb_size = sizeof(BlockDriverAIOCBSync), |
3900 | .cancel = bdrv_aio_cancel_em, | |
3901 | }; | |
3902 | ||
ce1a14dc | 3903 | static void bdrv_aio_bh_cb(void *opaque) |
83f64091 | 3904 | { |
ce1a14dc | 3905 | BlockDriverAIOCBSync *acb = opaque; |
f141eafe | 3906 | |
f141eafe | 3907 | if (!acb->is_write) |
03396148 | 3908 | qemu_iovec_from_buf(acb->qiov, 0, acb->bounce, acb->qiov->size); |
ceb42de8 | 3909 | qemu_vfree(acb->bounce); |
ce1a14dc | 3910 | acb->common.cb(acb->common.opaque, acb->ret); |
6a7ad299 | 3911 | qemu_bh_delete(acb->bh); |
36afc451 | 3912 | acb->bh = NULL; |
ce1a14dc | 3913 | qemu_aio_release(acb); |
83f64091 | 3914 | } |
beac80cd | 3915 | |
f141eafe AL |
3916 | static BlockDriverAIOCB *bdrv_aio_rw_vector(BlockDriverState *bs, |
3917 | int64_t sector_num, | |
3918 | QEMUIOVector *qiov, | |
3919 | int nb_sectors, | |
3920 | BlockDriverCompletionFunc *cb, | |
3921 | void *opaque, | |
3922 | int is_write) | |
3923 | ||
83f64091 | 3924 | { |
ce1a14dc | 3925 | BlockDriverAIOCBSync *acb; |
ce1a14dc | 3926 | |
d7331bed | 3927 | acb = qemu_aio_get(&bdrv_em_aiocb_info, bs, cb, opaque); |
f141eafe AL |
3928 | acb->is_write = is_write; |
3929 | acb->qiov = qiov; | |
e268ca52 | 3930 | acb->bounce = qemu_blockalign(bs, qiov->size); |
3f3aace8 | 3931 | acb->bh = qemu_bh_new(bdrv_aio_bh_cb, acb); |
f141eafe AL |
3932 | |
3933 | if (is_write) { | |
d5e6b161 | 3934 | qemu_iovec_to_buf(acb->qiov, 0, acb->bounce, qiov->size); |
1ed20acf | 3935 | acb->ret = bs->drv->bdrv_write(bs, sector_num, acb->bounce, nb_sectors); |
f141eafe | 3936 | } else { |
1ed20acf | 3937 | acb->ret = bs->drv->bdrv_read(bs, sector_num, acb->bounce, nb_sectors); |
f141eafe AL |
3938 | } |
3939 | ||
ce1a14dc | 3940 | qemu_bh_schedule(acb->bh); |
f141eafe | 3941 | |
ce1a14dc | 3942 | return &acb->common; |
beac80cd FB |
3943 | } |
3944 | ||
f141eafe AL |
3945 | static BlockDriverAIOCB *bdrv_aio_readv_em(BlockDriverState *bs, |
3946 | int64_t sector_num, QEMUIOVector *qiov, int nb_sectors, | |
ce1a14dc | 3947 | BlockDriverCompletionFunc *cb, void *opaque) |
beac80cd | 3948 | { |
f141eafe AL |
3949 | return bdrv_aio_rw_vector(bs, sector_num, qiov, nb_sectors, cb, opaque, 0); |
3950 | } | |
83f64091 | 3951 | |
f141eafe AL |
3952 | static BlockDriverAIOCB *bdrv_aio_writev_em(BlockDriverState *bs, |
3953 | int64_t sector_num, QEMUIOVector *qiov, int nb_sectors, | |
3954 | BlockDriverCompletionFunc *cb, void *opaque) | |
3955 | { | |
3956 | return bdrv_aio_rw_vector(bs, sector_num, qiov, nb_sectors, cb, opaque, 1); | |
beac80cd | 3957 | } |
beac80cd | 3958 | |
68485420 KW |
3959 | |
3960 | typedef struct BlockDriverAIOCBCoroutine { | |
3961 | BlockDriverAIOCB common; | |
3962 | BlockRequest req; | |
3963 | bool is_write; | |
d318aea9 | 3964 | bool *done; |
68485420 KW |
3965 | QEMUBH* bh; |
3966 | } BlockDriverAIOCBCoroutine; | |
3967 | ||
3968 | static void bdrv_aio_co_cancel_em(BlockDriverAIOCB *blockacb) | |
3969 | { | |
d318aea9 KW |
3970 | BlockDriverAIOCBCoroutine *acb = |
3971 | container_of(blockacb, BlockDriverAIOCBCoroutine, common); | |
3972 | bool done = false; | |
3973 | ||
3974 | acb->done = &done; | |
3975 | while (!done) { | |
3976 | qemu_aio_wait(); | |
3977 | } | |
68485420 KW |
3978 | } |
3979 | ||
d7331bed | 3980 | static const AIOCBInfo bdrv_em_co_aiocb_info = { |
68485420 KW |
3981 | .aiocb_size = sizeof(BlockDriverAIOCBCoroutine), |
3982 | .cancel = bdrv_aio_co_cancel_em, | |
3983 | }; | |
3984 | ||
35246a68 | 3985 | static void bdrv_co_em_bh(void *opaque) |
68485420 KW |
3986 | { |
3987 | BlockDriverAIOCBCoroutine *acb = opaque; | |
3988 | ||
3989 | acb->common.cb(acb->common.opaque, acb->req.error); | |
d318aea9 KW |
3990 | |
3991 | if (acb->done) { | |
3992 | *acb->done = true; | |
3993 | } | |
3994 | ||
68485420 KW |
3995 | qemu_bh_delete(acb->bh); |
3996 | qemu_aio_release(acb); | |
3997 | } | |
3998 | ||
b2a61371 SH |
3999 | /* Invoke bdrv_co_do_readv/bdrv_co_do_writev */ |
4000 | static void coroutine_fn bdrv_co_do_rw(void *opaque) | |
4001 | { | |
4002 | BlockDriverAIOCBCoroutine *acb = opaque; | |
4003 | BlockDriverState *bs = acb->common.bs; | |
4004 | ||
4005 | if (!acb->is_write) { | |
4006 | acb->req.error = bdrv_co_do_readv(bs, acb->req.sector, | |
470c0504 | 4007 | acb->req.nb_sectors, acb->req.qiov, 0); |
b2a61371 SH |
4008 | } else { |
4009 | acb->req.error = bdrv_co_do_writev(bs, acb->req.sector, | |
f08f2dda | 4010 | acb->req.nb_sectors, acb->req.qiov, 0); |
b2a61371 SH |
4011 | } |
4012 | ||
35246a68 | 4013 | acb->bh = qemu_bh_new(bdrv_co_em_bh, acb); |
b2a61371 SH |
4014 | qemu_bh_schedule(acb->bh); |
4015 | } | |
4016 | ||
68485420 KW |
4017 | static BlockDriverAIOCB *bdrv_co_aio_rw_vector(BlockDriverState *bs, |
4018 | int64_t sector_num, | |
4019 | QEMUIOVector *qiov, | |
4020 | int nb_sectors, | |
4021 | BlockDriverCompletionFunc *cb, | |
4022 | void *opaque, | |
8c5873d6 | 4023 | bool is_write) |
68485420 KW |
4024 | { |
4025 | Coroutine *co; | |
4026 | BlockDriverAIOCBCoroutine *acb; | |
4027 | ||
d7331bed | 4028 | acb = qemu_aio_get(&bdrv_em_co_aiocb_info, bs, cb, opaque); |
68485420 KW |
4029 | acb->req.sector = sector_num; |
4030 | acb->req.nb_sectors = nb_sectors; | |
4031 | acb->req.qiov = qiov; | |
4032 | acb->is_write = is_write; | |
d318aea9 | 4033 | acb->done = NULL; |
68485420 | 4034 | |
8c5873d6 | 4035 | co = qemu_coroutine_create(bdrv_co_do_rw); |
68485420 KW |
4036 | qemu_coroutine_enter(co, acb); |
4037 | ||
4038 | return &acb->common; | |
4039 | } | |
4040 | ||
07f07615 | 4041 | static void coroutine_fn bdrv_aio_flush_co_entry(void *opaque) |
b2e12bc6 | 4042 | { |
07f07615 PB |
4043 | BlockDriverAIOCBCoroutine *acb = opaque; |
4044 | BlockDriverState *bs = acb->common.bs; | |
b2e12bc6 | 4045 | |
07f07615 PB |
4046 | acb->req.error = bdrv_co_flush(bs); |
4047 | acb->bh = qemu_bh_new(bdrv_co_em_bh, acb); | |
b2e12bc6 | 4048 | qemu_bh_schedule(acb->bh); |
b2e12bc6 CH |
4049 | } |
4050 | ||
07f07615 | 4051 | BlockDriverAIOCB *bdrv_aio_flush(BlockDriverState *bs, |
016f5cf6 AG |
4052 | BlockDriverCompletionFunc *cb, void *opaque) |
4053 | { | |
07f07615 | 4054 | trace_bdrv_aio_flush(bs, opaque); |
016f5cf6 | 4055 | |
07f07615 PB |
4056 | Coroutine *co; |
4057 | BlockDriverAIOCBCoroutine *acb; | |
016f5cf6 | 4058 | |
d7331bed | 4059 | acb = qemu_aio_get(&bdrv_em_co_aiocb_info, bs, cb, opaque); |
d318aea9 KW |
4060 | acb->done = NULL; |
4061 | ||
07f07615 PB |
4062 | co = qemu_coroutine_create(bdrv_aio_flush_co_entry); |
4063 | qemu_coroutine_enter(co, acb); | |
016f5cf6 | 4064 | |
016f5cf6 AG |
4065 | return &acb->common; |
4066 | } | |
4067 | ||
4265d620 PB |
4068 | static void coroutine_fn bdrv_aio_discard_co_entry(void *opaque) |
4069 | { | |
4070 | BlockDriverAIOCBCoroutine *acb = opaque; | |
4071 | BlockDriverState *bs = acb->common.bs; | |
4072 | ||
4073 | acb->req.error = bdrv_co_discard(bs, acb->req.sector, acb->req.nb_sectors); | |
4074 | acb->bh = qemu_bh_new(bdrv_co_em_bh, acb); | |
4075 | qemu_bh_schedule(acb->bh); | |
4076 | } | |
4077 | ||
4078 | BlockDriverAIOCB *bdrv_aio_discard(BlockDriverState *bs, | |
4079 | int64_t sector_num, int nb_sectors, | |
4080 | BlockDriverCompletionFunc *cb, void *opaque) | |
4081 | { | |
4082 | Coroutine *co; | |
4083 | BlockDriverAIOCBCoroutine *acb; | |
4084 | ||
4085 | trace_bdrv_aio_discard(bs, sector_num, nb_sectors, opaque); | |
4086 | ||
d7331bed | 4087 | acb = qemu_aio_get(&bdrv_em_co_aiocb_info, bs, cb, opaque); |
4265d620 PB |
4088 | acb->req.sector = sector_num; |
4089 | acb->req.nb_sectors = nb_sectors; | |
d318aea9 | 4090 | acb->done = NULL; |
4265d620 PB |
4091 | co = qemu_coroutine_create(bdrv_aio_discard_co_entry); |
4092 | qemu_coroutine_enter(co, acb); | |
4093 | ||
4094 | return &acb->common; | |
4095 | } | |
4096 | ||
ea2384d3 FB |
4097 | void bdrv_init(void) |
4098 | { | |
5efa9d5a | 4099 | module_call_init(MODULE_INIT_BLOCK); |
ea2384d3 | 4100 | } |
ce1a14dc | 4101 | |
eb852011 MA |
4102 | void bdrv_init_with_whitelist(void) |
4103 | { | |
4104 | use_bdrv_whitelist = 1; | |
4105 | bdrv_init(); | |
4106 | } | |
4107 | ||
d7331bed | 4108 | void *qemu_aio_get(const AIOCBInfo *aiocb_info, BlockDriverState *bs, |
c16b5a2c | 4109 | BlockDriverCompletionFunc *cb, void *opaque) |
ce1a14dc | 4110 | { |
ce1a14dc PB |
4111 | BlockDriverAIOCB *acb; |
4112 | ||
d7331bed SH |
4113 | acb = g_slice_alloc(aiocb_info->aiocb_size); |
4114 | acb->aiocb_info = aiocb_info; | |
ce1a14dc PB |
4115 | acb->bs = bs; |
4116 | acb->cb = cb; | |
4117 | acb->opaque = opaque; | |
4118 | return acb; | |
4119 | } | |
4120 | ||
4121 | void qemu_aio_release(void *p) | |
4122 | { | |
d37c975f | 4123 | BlockDriverAIOCB *acb = p; |
d7331bed | 4124 | g_slice_free1(acb->aiocb_info->aiocb_size, acb); |
ce1a14dc | 4125 | } |
19cb3738 | 4126 | |
f9f05dc5 KW |
4127 | /**************************************************************/ |
4128 | /* Coroutine block device emulation */ | |
4129 | ||
4130 | typedef struct CoroutineIOCompletion { | |
4131 | Coroutine *coroutine; | |
4132 | int ret; | |
4133 | } CoroutineIOCompletion; | |
4134 | ||
4135 | static void bdrv_co_io_em_complete(void *opaque, int ret) | |
4136 | { | |
4137 | CoroutineIOCompletion *co = opaque; | |
4138 | ||
4139 | co->ret = ret; | |
4140 | qemu_coroutine_enter(co->coroutine, NULL); | |
4141 | } | |
4142 | ||
4143 | static int coroutine_fn bdrv_co_io_em(BlockDriverState *bs, int64_t sector_num, | |
4144 | int nb_sectors, QEMUIOVector *iov, | |
4145 | bool is_write) | |
4146 | { | |
4147 | CoroutineIOCompletion co = { | |
4148 | .coroutine = qemu_coroutine_self(), | |
4149 | }; | |
4150 | BlockDriverAIOCB *acb; | |
4151 | ||
4152 | if (is_write) { | |
a652d160 SH |
4153 | acb = bs->drv->bdrv_aio_writev(bs, sector_num, iov, nb_sectors, |
4154 | bdrv_co_io_em_complete, &co); | |
f9f05dc5 | 4155 | } else { |
a652d160 SH |
4156 | acb = bs->drv->bdrv_aio_readv(bs, sector_num, iov, nb_sectors, |
4157 | bdrv_co_io_em_complete, &co); | |
f9f05dc5 KW |
4158 | } |
4159 | ||
59370aaa | 4160 | trace_bdrv_co_io_em(bs, sector_num, nb_sectors, is_write, acb); |
f9f05dc5 KW |
4161 | if (!acb) { |
4162 | return -EIO; | |
4163 | } | |
4164 | qemu_coroutine_yield(); | |
4165 | ||
4166 | return co.ret; | |
4167 | } | |
4168 | ||
4169 | static int coroutine_fn bdrv_co_readv_em(BlockDriverState *bs, | |
4170 | int64_t sector_num, int nb_sectors, | |
4171 | QEMUIOVector *iov) | |
4172 | { | |
4173 | return bdrv_co_io_em(bs, sector_num, nb_sectors, iov, false); | |
4174 | } | |
4175 | ||
4176 | static int coroutine_fn bdrv_co_writev_em(BlockDriverState *bs, | |
4177 | int64_t sector_num, int nb_sectors, | |
4178 | QEMUIOVector *iov) | |
4179 | { | |
4180 | return bdrv_co_io_em(bs, sector_num, nb_sectors, iov, true); | |
4181 | } | |
4182 | ||
07f07615 | 4183 | static void coroutine_fn bdrv_flush_co_entry(void *opaque) |
e7a8a783 | 4184 | { |
07f07615 PB |
4185 | RwCo *rwco = opaque; |
4186 | ||
4187 | rwco->ret = bdrv_co_flush(rwco->bs); | |
4188 | } | |
4189 | ||
4190 | int coroutine_fn bdrv_co_flush(BlockDriverState *bs) | |
4191 | { | |
eb489bb1 KW |
4192 | int ret; |
4193 | ||
29cdb251 | 4194 | if (!bs || !bdrv_is_inserted(bs) || bdrv_is_read_only(bs)) { |
07f07615 | 4195 | return 0; |
eb489bb1 KW |
4196 | } |
4197 | ||
ca716364 | 4198 | /* Write back cached data to the OS even with cache=unsafe */ |
eb489bb1 KW |
4199 | if (bs->drv->bdrv_co_flush_to_os) { |
4200 | ret = bs->drv->bdrv_co_flush_to_os(bs); | |
4201 | if (ret < 0) { | |
4202 | return ret; | |
4203 | } | |
4204 | } | |
4205 | ||
ca716364 KW |
4206 | /* But don't actually force it to the disk with cache=unsafe */ |
4207 | if (bs->open_flags & BDRV_O_NO_FLUSH) { | |
d4c82329 | 4208 | goto flush_parent; |
ca716364 KW |
4209 | } |
4210 | ||
eb489bb1 | 4211 | if (bs->drv->bdrv_co_flush_to_disk) { |
29cdb251 | 4212 | ret = bs->drv->bdrv_co_flush_to_disk(bs); |
07f07615 PB |
4213 | } else if (bs->drv->bdrv_aio_flush) { |
4214 | BlockDriverAIOCB *acb; | |
4215 | CoroutineIOCompletion co = { | |
4216 | .coroutine = qemu_coroutine_self(), | |
4217 | }; | |
4218 | ||
4219 | acb = bs->drv->bdrv_aio_flush(bs, bdrv_co_io_em_complete, &co); | |
4220 | if (acb == NULL) { | |
29cdb251 | 4221 | ret = -EIO; |
07f07615 PB |
4222 | } else { |
4223 | qemu_coroutine_yield(); | |
29cdb251 | 4224 | ret = co.ret; |
07f07615 | 4225 | } |
07f07615 PB |
4226 | } else { |
4227 | /* | |
4228 | * Some block drivers always operate in either writethrough or unsafe | |
4229 | * mode and don't support bdrv_flush therefore. Usually qemu doesn't | |
4230 | * know how the server works (because the behaviour is hardcoded or | |
4231 | * depends on server-side configuration), so we can't ensure that | |
4232 | * everything is safe on disk. Returning an error doesn't work because | |
4233 | * that would break guests even if the server operates in writethrough | |
4234 | * mode. | |
4235 | * | |
4236 | * Let's hope the user knows what he's doing. | |
4237 | */ | |
29cdb251 | 4238 | ret = 0; |
07f07615 | 4239 | } |
29cdb251 PB |
4240 | if (ret < 0) { |
4241 | return ret; | |
4242 | } | |
4243 | ||
4244 | /* Now flush the underlying protocol. It will also have BDRV_O_NO_FLUSH | |
4245 | * in the case of cache=unsafe, so there are no useless flushes. | |
4246 | */ | |
d4c82329 | 4247 | flush_parent: |
29cdb251 | 4248 | return bdrv_co_flush(bs->file); |
07f07615 PB |
4249 | } |
4250 | ||
0f15423c AL |
4251 | void bdrv_invalidate_cache(BlockDriverState *bs) |
4252 | { | |
4253 | if (bs->drv && bs->drv->bdrv_invalidate_cache) { | |
4254 | bs->drv->bdrv_invalidate_cache(bs); | |
4255 | } | |
4256 | } | |
4257 | ||
4258 | void bdrv_invalidate_cache_all(void) | |
4259 | { | |
4260 | BlockDriverState *bs; | |
4261 | ||
4262 | QTAILQ_FOREACH(bs, &bdrv_states, list) { | |
4263 | bdrv_invalidate_cache(bs); | |
4264 | } | |
4265 | } | |
4266 | ||
07789269 BC |
4267 | void bdrv_clear_incoming_migration_all(void) |
4268 | { | |
4269 | BlockDriverState *bs; | |
4270 | ||
4271 | QTAILQ_FOREACH(bs, &bdrv_states, list) { | |
4272 | bs->open_flags = bs->open_flags & ~(BDRV_O_INCOMING); | |
4273 | } | |
4274 | } | |
4275 | ||
07f07615 PB |
4276 | int bdrv_flush(BlockDriverState *bs) |
4277 | { | |
4278 | Coroutine *co; | |
4279 | RwCo rwco = { | |
4280 | .bs = bs, | |
4281 | .ret = NOT_DONE, | |
e7a8a783 | 4282 | }; |
e7a8a783 | 4283 | |
07f07615 PB |
4284 | if (qemu_in_coroutine()) { |
4285 | /* Fast-path if already in coroutine context */ | |
4286 | bdrv_flush_co_entry(&rwco); | |
4287 | } else { | |
4288 | co = qemu_coroutine_create(bdrv_flush_co_entry); | |
4289 | qemu_coroutine_enter(co, &rwco); | |
4290 | while (rwco.ret == NOT_DONE) { | |
4291 | qemu_aio_wait(); | |
4292 | } | |
e7a8a783 | 4293 | } |
07f07615 PB |
4294 | |
4295 | return rwco.ret; | |
e7a8a783 KW |
4296 | } |
4297 | ||
4265d620 PB |
4298 | static void coroutine_fn bdrv_discard_co_entry(void *opaque) |
4299 | { | |
4300 | RwCo *rwco = opaque; | |
4301 | ||
4302 | rwco->ret = bdrv_co_discard(rwco->bs, rwco->sector_num, rwco->nb_sectors); | |
4303 | } | |
4304 | ||
4305 | int coroutine_fn bdrv_co_discard(BlockDriverState *bs, int64_t sector_num, | |
4306 | int nb_sectors) | |
4307 | { | |
4308 | if (!bs->drv) { | |
4309 | return -ENOMEDIUM; | |
4310 | } else if (bdrv_check_request(bs, sector_num, nb_sectors)) { | |
4311 | return -EIO; | |
4312 | } else if (bs->read_only) { | |
4313 | return -EROFS; | |
df702c9b PB |
4314 | } |
4315 | ||
4316 | if (bs->dirty_bitmap) { | |
8f0720ec | 4317 | bdrv_reset_dirty(bs, sector_num, nb_sectors); |
df702c9b PB |
4318 | } |
4319 | ||
9e8f1835 PB |
4320 | /* Do nothing if disabled. */ |
4321 | if (!(bs->open_flags & BDRV_O_UNMAP)) { | |
4322 | return 0; | |
4323 | } | |
4324 | ||
df702c9b | 4325 | if (bs->drv->bdrv_co_discard) { |
4265d620 PB |
4326 | return bs->drv->bdrv_co_discard(bs, sector_num, nb_sectors); |
4327 | } else if (bs->drv->bdrv_aio_discard) { | |
4328 | BlockDriverAIOCB *acb; | |
4329 | CoroutineIOCompletion co = { | |
4330 | .coroutine = qemu_coroutine_self(), | |
4331 | }; | |
4332 | ||
4333 | acb = bs->drv->bdrv_aio_discard(bs, sector_num, nb_sectors, | |
4334 | bdrv_co_io_em_complete, &co); | |
4335 | if (acb == NULL) { | |
4336 | return -EIO; | |
4337 | } else { | |
4338 | qemu_coroutine_yield(); | |
4339 | return co.ret; | |
4340 | } | |
4265d620 PB |
4341 | } else { |
4342 | return 0; | |
4343 | } | |
4344 | } | |
4345 | ||
4346 | int bdrv_discard(BlockDriverState *bs, int64_t sector_num, int nb_sectors) | |
4347 | { | |
4348 | Coroutine *co; | |
4349 | RwCo rwco = { | |
4350 | .bs = bs, | |
4351 | .sector_num = sector_num, | |
4352 | .nb_sectors = nb_sectors, | |
4353 | .ret = NOT_DONE, | |
4354 | }; | |
4355 | ||
4356 | if (qemu_in_coroutine()) { | |
4357 | /* Fast-path if already in coroutine context */ | |
4358 | bdrv_discard_co_entry(&rwco); | |
4359 | } else { | |
4360 | co = qemu_coroutine_create(bdrv_discard_co_entry); | |
4361 | qemu_coroutine_enter(co, &rwco); | |
4362 | while (rwco.ret == NOT_DONE) { | |
4363 | qemu_aio_wait(); | |
4364 | } | |
4365 | } | |
4366 | ||
4367 | return rwco.ret; | |
4368 | } | |
4369 | ||
19cb3738 FB |
4370 | /**************************************************************/ |
4371 | /* removable device support */ | |
4372 | ||
4373 | /** | |
4374 | * Return TRUE if the media is present | |
4375 | */ | |
4376 | int bdrv_is_inserted(BlockDriverState *bs) | |
4377 | { | |
4378 | BlockDriver *drv = bs->drv; | |
a1aff5bf | 4379 | |
19cb3738 FB |
4380 | if (!drv) |
4381 | return 0; | |
4382 | if (!drv->bdrv_is_inserted) | |
a1aff5bf MA |
4383 | return 1; |
4384 | return drv->bdrv_is_inserted(bs); | |
19cb3738 FB |
4385 | } |
4386 | ||
4387 | /** | |
8e49ca46 MA |
4388 | * Return whether the media changed since the last call to this |
4389 | * function, or -ENOTSUP if we don't know. Most drivers don't know. | |
19cb3738 FB |
4390 | */ |
4391 | int bdrv_media_changed(BlockDriverState *bs) | |
4392 | { | |
4393 | BlockDriver *drv = bs->drv; | |
19cb3738 | 4394 | |
8e49ca46 MA |
4395 | if (drv && drv->bdrv_media_changed) { |
4396 | return drv->bdrv_media_changed(bs); | |
4397 | } | |
4398 | return -ENOTSUP; | |
19cb3738 FB |
4399 | } |
4400 | ||
4401 | /** | |
4402 | * If eject_flag is TRUE, eject the media. Otherwise, close the tray | |
4403 | */ | |
f36f3949 | 4404 | void bdrv_eject(BlockDriverState *bs, bool eject_flag) |
19cb3738 FB |
4405 | { |
4406 | BlockDriver *drv = bs->drv; | |
19cb3738 | 4407 | |
822e1cd1 MA |
4408 | if (drv && drv->bdrv_eject) { |
4409 | drv->bdrv_eject(bs, eject_flag); | |
19cb3738 | 4410 | } |
6f382ed2 LC |
4411 | |
4412 | if (bs->device_name[0] != '\0') { | |
4413 | bdrv_emit_qmp_eject_event(bs, eject_flag); | |
4414 | } | |
19cb3738 FB |
4415 | } |
4416 | ||
19cb3738 FB |
4417 | /** |
4418 | * Lock or unlock the media (if it is locked, the user won't be able | |
4419 | * to eject it manually). | |
4420 | */ | |
025e849a | 4421 | void bdrv_lock_medium(BlockDriverState *bs, bool locked) |
19cb3738 FB |
4422 | { |
4423 | BlockDriver *drv = bs->drv; | |
4424 | ||
025e849a | 4425 | trace_bdrv_lock_medium(bs, locked); |
b8c6d095 | 4426 | |
025e849a MA |
4427 | if (drv && drv->bdrv_lock_medium) { |
4428 | drv->bdrv_lock_medium(bs, locked); | |
19cb3738 FB |
4429 | } |
4430 | } | |
985a03b0 TS |
4431 | |
4432 | /* needed for generic scsi interface */ | |
4433 | ||
4434 | int bdrv_ioctl(BlockDriverState *bs, unsigned long int req, void *buf) | |
4435 | { | |
4436 | BlockDriver *drv = bs->drv; | |
4437 | ||
4438 | if (drv && drv->bdrv_ioctl) | |
4439 | return drv->bdrv_ioctl(bs, req, buf); | |
4440 | return -ENOTSUP; | |
4441 | } | |
7d780669 | 4442 | |
221f715d AL |
4443 | BlockDriverAIOCB *bdrv_aio_ioctl(BlockDriverState *bs, |
4444 | unsigned long int req, void *buf, | |
4445 | BlockDriverCompletionFunc *cb, void *opaque) | |
7d780669 | 4446 | { |
221f715d | 4447 | BlockDriver *drv = bs->drv; |
7d780669 | 4448 | |
221f715d AL |
4449 | if (drv && drv->bdrv_aio_ioctl) |
4450 | return drv->bdrv_aio_ioctl(bs, req, buf, cb, opaque); | |
4451 | return NULL; | |
7d780669 | 4452 | } |
e268ca52 | 4453 | |
7b6f9300 MA |
4454 | void bdrv_set_buffer_alignment(BlockDriverState *bs, int align) |
4455 | { | |
4456 | bs->buffer_alignment = align; | |
4457 | } | |
7cd1e32a | 4458 | |
e268ca52 AL |
4459 | void *qemu_blockalign(BlockDriverState *bs, size_t size) |
4460 | { | |
4461 | return qemu_memalign((bs && bs->buffer_alignment) ? bs->buffer_alignment : 512, size); | |
4462 | } | |
7cd1e32a | 4463 | |
c53b1c51 SH |
4464 | /* |
4465 | * Check if all memory in this vector is sector aligned. | |
4466 | */ | |
4467 | bool bdrv_qiov_is_aligned(BlockDriverState *bs, QEMUIOVector *qiov) | |
4468 | { | |
4469 | int i; | |
4470 | ||
4471 | for (i = 0; i < qiov->niov; i++) { | |
4472 | if ((uintptr_t) qiov->iov[i].iov_base % bs->buffer_alignment) { | |
4473 | return false; | |
4474 | } | |
4475 | } | |
4476 | ||
4477 | return true; | |
4478 | } | |
4479 | ||
50717e94 | 4480 | void bdrv_set_dirty_tracking(BlockDriverState *bs, int granularity) |
7cd1e32a | 4481 | { |
4482 | int64_t bitmap_size; | |
a55eb92c | 4483 | |
50717e94 PB |
4484 | assert((granularity & (granularity - 1)) == 0); |
4485 | ||
4486 | if (granularity) { | |
4487 | granularity >>= BDRV_SECTOR_BITS; | |
4488 | assert(!bs->dirty_bitmap); | |
4489 | bitmap_size = (bdrv_getlength(bs) >> BDRV_SECTOR_BITS); | |
4490 | bs->dirty_bitmap = hbitmap_alloc(bitmap_size, ffs(granularity) - 1); | |
7cd1e32a | 4491 | } else { |
c6d22830 | 4492 | if (bs->dirty_bitmap) { |
8f0720ec | 4493 | hbitmap_free(bs->dirty_bitmap); |
c6d22830 | 4494 | bs->dirty_bitmap = NULL; |
a55eb92c | 4495 | } |
7cd1e32a | 4496 | } |
4497 | } | |
4498 | ||
4499 | int bdrv_get_dirty(BlockDriverState *bs, int64_t sector) | |
4500 | { | |
8f0720ec PB |
4501 | if (bs->dirty_bitmap) { |
4502 | return hbitmap_get(bs->dirty_bitmap, sector); | |
7cd1e32a | 4503 | } else { |
4504 | return 0; | |
4505 | } | |
4506 | } | |
4507 | ||
8f0720ec | 4508 | void bdrv_dirty_iter_init(BlockDriverState *bs, HBitmapIter *hbi) |
1755da16 | 4509 | { |
8f0720ec | 4510 | hbitmap_iter_init(hbi, bs->dirty_bitmap, 0); |
1755da16 PB |
4511 | } |
4512 | ||
4513 | void bdrv_set_dirty(BlockDriverState *bs, int64_t cur_sector, | |
4514 | int nr_sectors) | |
4515 | { | |
8f0720ec | 4516 | hbitmap_set(bs->dirty_bitmap, cur_sector, nr_sectors); |
1755da16 PB |
4517 | } |
4518 | ||
a55eb92c JK |
4519 | void bdrv_reset_dirty(BlockDriverState *bs, int64_t cur_sector, |
4520 | int nr_sectors) | |
7cd1e32a | 4521 | { |
8f0720ec | 4522 | hbitmap_reset(bs->dirty_bitmap, cur_sector, nr_sectors); |
7cd1e32a | 4523 | } |
aaa0eb75 LS |
4524 | |
4525 | int64_t bdrv_get_dirty_count(BlockDriverState *bs) | |
4526 | { | |
8f0720ec | 4527 | if (bs->dirty_bitmap) { |
acc906c6 | 4528 | return hbitmap_count(bs->dirty_bitmap); |
8f0720ec PB |
4529 | } else { |
4530 | return 0; | |
4531 | } | |
aaa0eb75 | 4532 | } |
f88e1a42 | 4533 | |
db593f25 MT |
4534 | void bdrv_set_in_use(BlockDriverState *bs, int in_use) |
4535 | { | |
4536 | assert(bs->in_use != in_use); | |
4537 | bs->in_use = in_use; | |
4538 | } | |
4539 | ||
4540 | int bdrv_in_use(BlockDriverState *bs) | |
4541 | { | |
4542 | return bs->in_use; | |
4543 | } | |
4544 | ||
28a7282a LC |
4545 | void bdrv_iostatus_enable(BlockDriverState *bs) |
4546 | { | |
d6bf279e | 4547 | bs->iostatus_enabled = true; |
58e21ef5 | 4548 | bs->iostatus = BLOCK_DEVICE_IO_STATUS_OK; |
28a7282a LC |
4549 | } |
4550 | ||
4551 | /* The I/O status is only enabled if the drive explicitly | |
4552 | * enables it _and_ the VM is configured to stop on errors */ | |
4553 | bool bdrv_iostatus_is_enabled(const BlockDriverState *bs) | |
4554 | { | |
d6bf279e | 4555 | return (bs->iostatus_enabled && |
92aa5c6d PB |
4556 | (bs->on_write_error == BLOCKDEV_ON_ERROR_ENOSPC || |
4557 | bs->on_write_error == BLOCKDEV_ON_ERROR_STOP || | |
4558 | bs->on_read_error == BLOCKDEV_ON_ERROR_STOP)); | |
28a7282a LC |
4559 | } |
4560 | ||
4561 | void bdrv_iostatus_disable(BlockDriverState *bs) | |
4562 | { | |
d6bf279e | 4563 | bs->iostatus_enabled = false; |
28a7282a LC |
4564 | } |
4565 | ||
4566 | void bdrv_iostatus_reset(BlockDriverState *bs) | |
4567 | { | |
4568 | if (bdrv_iostatus_is_enabled(bs)) { | |
58e21ef5 | 4569 | bs->iostatus = BLOCK_DEVICE_IO_STATUS_OK; |
3bd293c3 PB |
4570 | if (bs->job) { |
4571 | block_job_iostatus_reset(bs->job); | |
4572 | } | |
28a7282a LC |
4573 | } |
4574 | } | |
4575 | ||
28a7282a LC |
4576 | void bdrv_iostatus_set_err(BlockDriverState *bs, int error) |
4577 | { | |
3e1caa5f PB |
4578 | assert(bdrv_iostatus_is_enabled(bs)); |
4579 | if (bs->iostatus == BLOCK_DEVICE_IO_STATUS_OK) { | |
58e21ef5 LC |
4580 | bs->iostatus = error == ENOSPC ? BLOCK_DEVICE_IO_STATUS_NOSPACE : |
4581 | BLOCK_DEVICE_IO_STATUS_FAILED; | |
28a7282a LC |
4582 | } |
4583 | } | |
4584 | ||
a597e79c CH |
4585 | void |
4586 | bdrv_acct_start(BlockDriverState *bs, BlockAcctCookie *cookie, int64_t bytes, | |
4587 | enum BlockAcctType type) | |
4588 | { | |
4589 | assert(type < BDRV_MAX_IOTYPE); | |
4590 | ||
4591 | cookie->bytes = bytes; | |
c488c7f6 | 4592 | cookie->start_time_ns = get_clock(); |
a597e79c CH |
4593 | cookie->type = type; |
4594 | } | |
4595 | ||
4596 | void | |
4597 | bdrv_acct_done(BlockDriverState *bs, BlockAcctCookie *cookie) | |
4598 | { | |
4599 | assert(cookie->type < BDRV_MAX_IOTYPE); | |
4600 | ||
4601 | bs->nr_bytes[cookie->type] += cookie->bytes; | |
4602 | bs->nr_ops[cookie->type]++; | |
c488c7f6 | 4603 | bs->total_time_ns[cookie->type] += get_clock() - cookie->start_time_ns; |
a597e79c CH |
4604 | } |
4605 | ||
d92ada22 LC |
4606 | void bdrv_img_create(const char *filename, const char *fmt, |
4607 | const char *base_filename, const char *base_fmt, | |
f382d43a MR |
4608 | char *options, uint64_t img_size, int flags, |
4609 | Error **errp, bool quiet) | |
f88e1a42 JS |
4610 | { |
4611 | QEMUOptionParameter *param = NULL, *create_options = NULL; | |
d220894e | 4612 | QEMUOptionParameter *backing_fmt, *backing_file, *size; |
f88e1a42 JS |
4613 | BlockDriverState *bs = NULL; |
4614 | BlockDriver *drv, *proto_drv; | |
96df67d1 | 4615 | BlockDriver *backing_drv = NULL; |
f88e1a42 JS |
4616 | int ret = 0; |
4617 | ||
4618 | /* Find driver and parse its options */ | |
4619 | drv = bdrv_find_format(fmt); | |
4620 | if (!drv) { | |
71c79813 | 4621 | error_setg(errp, "Unknown file format '%s'", fmt); |
d92ada22 | 4622 | return; |
f88e1a42 JS |
4623 | } |
4624 | ||
4625 | proto_drv = bdrv_find_protocol(filename); | |
4626 | if (!proto_drv) { | |
71c79813 | 4627 | error_setg(errp, "Unknown protocol '%s'", filename); |
d92ada22 | 4628 | return; |
f88e1a42 JS |
4629 | } |
4630 | ||
4631 | create_options = append_option_parameters(create_options, | |
4632 | drv->create_options); | |
4633 | create_options = append_option_parameters(create_options, | |
4634 | proto_drv->create_options); | |
4635 | ||
4636 | /* Create parameter list with default values */ | |
4637 | param = parse_option_parameters("", create_options, param); | |
4638 | ||
4639 | set_option_parameter_int(param, BLOCK_OPT_SIZE, img_size); | |
4640 | ||
4641 | /* Parse -o options */ | |
4642 | if (options) { | |
4643 | param = parse_option_parameters(options, create_options, param); | |
4644 | if (param == NULL) { | |
71c79813 | 4645 | error_setg(errp, "Invalid options for file format '%s'.", fmt); |
f88e1a42 JS |
4646 | goto out; |
4647 | } | |
4648 | } | |
4649 | ||
4650 | if (base_filename) { | |
4651 | if (set_option_parameter(param, BLOCK_OPT_BACKING_FILE, | |
4652 | base_filename)) { | |
71c79813 LC |
4653 | error_setg(errp, "Backing file not supported for file format '%s'", |
4654 | fmt); | |
f88e1a42 JS |
4655 | goto out; |
4656 | } | |
4657 | } | |
4658 | ||
4659 | if (base_fmt) { | |
4660 | if (set_option_parameter(param, BLOCK_OPT_BACKING_FMT, base_fmt)) { | |
71c79813 LC |
4661 | error_setg(errp, "Backing file format not supported for file " |
4662 | "format '%s'", fmt); | |
f88e1a42 JS |
4663 | goto out; |
4664 | } | |
4665 | } | |
4666 | ||
792da93a JS |
4667 | backing_file = get_option_parameter(param, BLOCK_OPT_BACKING_FILE); |
4668 | if (backing_file && backing_file->value.s) { | |
4669 | if (!strcmp(filename, backing_file->value.s)) { | |
71c79813 LC |
4670 | error_setg(errp, "Error: Trying to create an image with the " |
4671 | "same filename as the backing file"); | |
792da93a JS |
4672 | goto out; |
4673 | } | |
4674 | } | |
4675 | ||
f88e1a42 JS |
4676 | backing_fmt = get_option_parameter(param, BLOCK_OPT_BACKING_FMT); |
4677 | if (backing_fmt && backing_fmt->value.s) { | |
96df67d1 SH |
4678 | backing_drv = bdrv_find_format(backing_fmt->value.s); |
4679 | if (!backing_drv) { | |
71c79813 LC |
4680 | error_setg(errp, "Unknown backing file format '%s'", |
4681 | backing_fmt->value.s); | |
f88e1a42 JS |
4682 | goto out; |
4683 | } | |
4684 | } | |
4685 | ||
4686 | // The size for the image must always be specified, with one exception: | |
4687 | // If we are using a backing file, we can obtain the size from there | |
d220894e KW |
4688 | size = get_option_parameter(param, BLOCK_OPT_SIZE); |
4689 | if (size && size->value.n == -1) { | |
f88e1a42 JS |
4690 | if (backing_file && backing_file->value.s) { |
4691 | uint64_t size; | |
f88e1a42 | 4692 | char buf[32]; |
63090dac PB |
4693 | int back_flags; |
4694 | ||
4695 | /* backing files always opened read-only */ | |
4696 | back_flags = | |
4697 | flags & ~(BDRV_O_RDWR | BDRV_O_SNAPSHOT | BDRV_O_NO_BACKING); | |
f88e1a42 | 4698 | |
f88e1a42 JS |
4699 | bs = bdrv_new(""); |
4700 | ||
de9c0cec KW |
4701 | ret = bdrv_open(bs, backing_file->value.s, NULL, back_flags, |
4702 | backing_drv); | |
f88e1a42 | 4703 | if (ret < 0) { |
71c79813 LC |
4704 | error_setg_errno(errp, -ret, "Could not open '%s'", |
4705 | backing_file->value.s); | |
f88e1a42 JS |
4706 | goto out; |
4707 | } | |
4708 | bdrv_get_geometry(bs, &size); | |
4709 | size *= 512; | |
4710 | ||
4711 | snprintf(buf, sizeof(buf), "%" PRId64, size); | |
4712 | set_option_parameter(param, BLOCK_OPT_SIZE, buf); | |
4713 | } else { | |
71c79813 | 4714 | error_setg(errp, "Image creation needs a size parameter"); |
f88e1a42 JS |
4715 | goto out; |
4716 | } | |
4717 | } | |
4718 | ||
f382d43a MR |
4719 | if (!quiet) { |
4720 | printf("Formatting '%s', fmt=%s ", filename, fmt); | |
4721 | print_option_parameters(param); | |
4722 | puts(""); | |
4723 | } | |
f88e1a42 | 4724 | ret = bdrv_create(drv, filename, param); |
f88e1a42 JS |
4725 | if (ret < 0) { |
4726 | if (ret == -ENOTSUP) { | |
71c79813 LC |
4727 | error_setg(errp,"Formatting or formatting option not supported for " |
4728 | "file format '%s'", fmt); | |
f88e1a42 | 4729 | } else if (ret == -EFBIG) { |
71c79813 LC |
4730 | error_setg(errp, "The image size is too large for file format '%s'", |
4731 | fmt); | |
f88e1a42 | 4732 | } else { |
71c79813 LC |
4733 | error_setg(errp, "%s: error while creating %s: %s", filename, fmt, |
4734 | strerror(-ret)); | |
f88e1a42 JS |
4735 | } |
4736 | } | |
4737 | ||
4738 | out: | |
4739 | free_option_parameters(create_options); | |
4740 | free_option_parameters(param); | |
4741 | ||
4742 | if (bs) { | |
4743 | bdrv_delete(bs); | |
4744 | } | |
f88e1a42 | 4745 | } |
85d126f3 SH |
4746 | |
4747 | AioContext *bdrv_get_aio_context(BlockDriverState *bs) | |
4748 | { | |
4749 | /* Currently BlockDriverState always uses the main loop AioContext */ | |
4750 | return qemu_get_aio_context(); | |
4751 | } |