]>
Commit | Line | Data |
---|---|---|
fc01f7e7 FB |
1 | /* |
2 | * QEMU System Emulator block driver | |
5fafdf24 | 3 | * |
fc01f7e7 | 4 | * Copyright (c) 2003 Fabrice Bellard |
5fafdf24 | 5 | * |
fc01f7e7 FB |
6 | * Permission is hereby granted, free of charge, to any person obtaining a copy |
7 | * of this software and associated documentation files (the "Software"), to deal | |
8 | * in the Software without restriction, including without limitation the rights | |
9 | * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell | |
10 | * copies of the Software, and to permit persons to whom the Software is | |
11 | * furnished to do so, subject to the following conditions: | |
12 | * | |
13 | * The above copyright notice and this permission notice shall be included in | |
14 | * all copies or substantial portions of the Software. | |
15 | * | |
16 | * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR | |
17 | * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, | |
18 | * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL | |
19 | * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER | |
20 | * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, | |
21 | * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN | |
22 | * THE SOFTWARE. | |
23 | */ | |
3990d09a | 24 | #include "config-host.h" |
faf07963 | 25 | #include "qemu-common.h" |
6d519a5f | 26 | #include "trace.h" |
376253ec | 27 | #include "monitor.h" |
ea2384d3 | 28 | #include "block_int.h" |
5efa9d5a | 29 | #include "module.h" |
f795e743 | 30 | #include "qjson.h" |
68485420 | 31 | #include "qemu-coroutine.h" |
b2023818 | 32 | #include "qmp-commands.h" |
0563e191 | 33 | #include "qemu-timer.h" |
fc01f7e7 | 34 | |
71e72a19 | 35 | #ifdef CONFIG_BSD |
7674e7bf FB |
36 | #include <sys/types.h> |
37 | #include <sys/stat.h> | |
38 | #include <sys/ioctl.h> | |
72cf2d4f | 39 | #include <sys/queue.h> |
c5e97233 | 40 | #ifndef __DragonFly__ |
7674e7bf FB |
41 | #include <sys/disk.h> |
42 | #endif | |
c5e97233 | 43 | #endif |
7674e7bf | 44 | |
49dc768d AL |
45 | #ifdef _WIN32 |
46 | #include <windows.h> | |
47 | #endif | |
48 | ||
1c9805a3 SH |
49 | #define NOT_DONE 0x7fffffff /* used while emulated sync operation in progress */ |
50 | ||
470c0504 SH |
51 | typedef enum { |
52 | BDRV_REQ_COPY_ON_READ = 0x1, | |
f08f2dda | 53 | BDRV_REQ_ZERO_WRITE = 0x2, |
470c0504 SH |
54 | } BdrvRequestFlags; |
55 | ||
7d4b4ba5 | 56 | static void bdrv_dev_change_media_cb(BlockDriverState *bs, bool load); |
f141eafe AL |
57 | static BlockDriverAIOCB *bdrv_aio_readv_em(BlockDriverState *bs, |
58 | int64_t sector_num, QEMUIOVector *qiov, int nb_sectors, | |
c87c0672 | 59 | BlockDriverCompletionFunc *cb, void *opaque); |
f141eafe AL |
60 | static BlockDriverAIOCB *bdrv_aio_writev_em(BlockDriverState *bs, |
61 | int64_t sector_num, QEMUIOVector *qiov, int nb_sectors, | |
ce1a14dc | 62 | BlockDriverCompletionFunc *cb, void *opaque); |
f9f05dc5 KW |
63 | static int coroutine_fn bdrv_co_readv_em(BlockDriverState *bs, |
64 | int64_t sector_num, int nb_sectors, | |
65 | QEMUIOVector *iov); | |
66 | static int coroutine_fn bdrv_co_writev_em(BlockDriverState *bs, | |
67 | int64_t sector_num, int nb_sectors, | |
68 | QEMUIOVector *iov); | |
c5fbe571 | 69 | static int coroutine_fn bdrv_co_do_readv(BlockDriverState *bs, |
470c0504 SH |
70 | int64_t sector_num, int nb_sectors, QEMUIOVector *qiov, |
71 | BdrvRequestFlags flags); | |
1c9805a3 | 72 | static int coroutine_fn bdrv_co_do_writev(BlockDriverState *bs, |
f08f2dda SH |
73 | int64_t sector_num, int nb_sectors, QEMUIOVector *qiov, |
74 | BdrvRequestFlags flags); | |
b2a61371 SH |
75 | static BlockDriverAIOCB *bdrv_co_aio_rw_vector(BlockDriverState *bs, |
76 | int64_t sector_num, | |
77 | QEMUIOVector *qiov, | |
78 | int nb_sectors, | |
79 | BlockDriverCompletionFunc *cb, | |
80 | void *opaque, | |
8c5873d6 | 81 | bool is_write); |
b2a61371 | 82 | static void coroutine_fn bdrv_co_do_rw(void *opaque); |
621f0589 KW |
83 | static int coroutine_fn bdrv_co_do_write_zeroes(BlockDriverState *bs, |
84 | int64_t sector_num, int nb_sectors); | |
ec530c81 | 85 | |
98f90dba ZYW |
86 | static bool bdrv_exceed_bps_limits(BlockDriverState *bs, int nb_sectors, |
87 | bool is_write, double elapsed_time, uint64_t *wait); | |
88 | static bool bdrv_exceed_iops_limits(BlockDriverState *bs, bool is_write, | |
89 | double elapsed_time, uint64_t *wait); | |
90 | static bool bdrv_exceed_io_limits(BlockDriverState *bs, int nb_sectors, | |
91 | bool is_write, int64_t *wait); | |
92 | ||
1b7bdbc1 SH |
93 | static QTAILQ_HEAD(, BlockDriverState) bdrv_states = |
94 | QTAILQ_HEAD_INITIALIZER(bdrv_states); | |
7ee930d0 | 95 | |
8a22f02a SH |
96 | static QLIST_HEAD(, BlockDriver) bdrv_drivers = |
97 | QLIST_HEAD_INITIALIZER(bdrv_drivers); | |
ea2384d3 | 98 | |
f9092b10 MA |
99 | /* The device to use for VM snapshots */ |
100 | static BlockDriverState *bs_snapshots; | |
101 | ||
eb852011 MA |
102 | /* If non-zero, use only whitelisted block drivers */ |
103 | static int use_bdrv_whitelist; | |
104 | ||
9e0b22f4 SH |
105 | #ifdef _WIN32 |
106 | static int is_windows_drive_prefix(const char *filename) | |
107 | { | |
108 | return (((filename[0] >= 'a' && filename[0] <= 'z') || | |
109 | (filename[0] >= 'A' && filename[0] <= 'Z')) && | |
110 | filename[1] == ':'); | |
111 | } | |
112 | ||
113 | int is_windows_drive(const char *filename) | |
114 | { | |
115 | if (is_windows_drive_prefix(filename) && | |
116 | filename[2] == '\0') | |
117 | return 1; | |
118 | if (strstart(filename, "\\\\.\\", NULL) || | |
119 | strstart(filename, "//./", NULL)) | |
120 | return 1; | |
121 | return 0; | |
122 | } | |
123 | #endif | |
124 | ||
0563e191 | 125 | /* throttling disk I/O limits */ |
98f90dba ZYW |
126 | void bdrv_io_limits_disable(BlockDriverState *bs) |
127 | { | |
128 | bs->io_limits_enabled = false; | |
129 | ||
130 | while (qemu_co_queue_next(&bs->throttled_reqs)); | |
131 | ||
132 | if (bs->block_timer) { | |
133 | qemu_del_timer(bs->block_timer); | |
134 | qemu_free_timer(bs->block_timer); | |
135 | bs->block_timer = NULL; | |
136 | } | |
137 | ||
138 | bs->slice_start = 0; | |
139 | bs->slice_end = 0; | |
140 | bs->slice_time = 0; | |
141 | memset(&bs->io_base, 0, sizeof(bs->io_base)); | |
142 | } | |
143 | ||
0563e191 ZYW |
144 | static void bdrv_block_timer(void *opaque) |
145 | { | |
146 | BlockDriverState *bs = opaque; | |
147 | ||
148 | qemu_co_queue_next(&bs->throttled_reqs); | |
149 | } | |
150 | ||
151 | void bdrv_io_limits_enable(BlockDriverState *bs) | |
152 | { | |
153 | qemu_co_queue_init(&bs->throttled_reqs); | |
154 | bs->block_timer = qemu_new_timer_ns(vm_clock, bdrv_block_timer, bs); | |
155 | bs->slice_time = 5 * BLOCK_IO_SLICE_TIME; | |
156 | bs->slice_start = qemu_get_clock_ns(vm_clock); | |
157 | bs->slice_end = bs->slice_start + bs->slice_time; | |
158 | memset(&bs->io_base, 0, sizeof(bs->io_base)); | |
159 | bs->io_limits_enabled = true; | |
160 | } | |
161 | ||
162 | bool bdrv_io_limits_enabled(BlockDriverState *bs) | |
163 | { | |
164 | BlockIOLimit *io_limits = &bs->io_limits; | |
165 | return io_limits->bps[BLOCK_IO_LIMIT_READ] | |
166 | || io_limits->bps[BLOCK_IO_LIMIT_WRITE] | |
167 | || io_limits->bps[BLOCK_IO_LIMIT_TOTAL] | |
168 | || io_limits->iops[BLOCK_IO_LIMIT_READ] | |
169 | || io_limits->iops[BLOCK_IO_LIMIT_WRITE] | |
170 | || io_limits->iops[BLOCK_IO_LIMIT_TOTAL]; | |
171 | } | |
172 | ||
98f90dba ZYW |
173 | static void bdrv_io_limits_intercept(BlockDriverState *bs, |
174 | bool is_write, int nb_sectors) | |
175 | { | |
176 | int64_t wait_time = -1; | |
177 | ||
178 | if (!qemu_co_queue_empty(&bs->throttled_reqs)) { | |
179 | qemu_co_queue_wait(&bs->throttled_reqs); | |
180 | } | |
181 | ||
182 | /* In fact, we hope to keep each request's timing, in FIFO mode. The next | |
183 | * throttled requests will not be dequeued until the current request is | |
184 | * allowed to be serviced. So if the current request still exceeds the | |
185 | * limits, it will be inserted to the head. All requests followed it will | |
186 | * be still in throttled_reqs queue. | |
187 | */ | |
188 | ||
189 | while (bdrv_exceed_io_limits(bs, nb_sectors, is_write, &wait_time)) { | |
190 | qemu_mod_timer(bs->block_timer, | |
191 | wait_time + qemu_get_clock_ns(vm_clock)); | |
192 | qemu_co_queue_wait_insert_head(&bs->throttled_reqs); | |
193 | } | |
194 | ||
195 | qemu_co_queue_next(&bs->throttled_reqs); | |
196 | } | |
197 | ||
9e0b22f4 SH |
198 | /* check if the path starts with "<protocol>:" */ |
199 | static int path_has_protocol(const char *path) | |
200 | { | |
201 | #ifdef _WIN32 | |
202 | if (is_windows_drive(path) || | |
203 | is_windows_drive_prefix(path)) { | |
204 | return 0; | |
205 | } | |
206 | #endif | |
207 | ||
208 | return strchr(path, ':') != NULL; | |
209 | } | |
210 | ||
83f64091 | 211 | int path_is_absolute(const char *path) |
3b0d4f61 | 212 | { |
21664424 FB |
213 | #ifdef _WIN32 |
214 | /* specific case for names like: "\\.\d:" */ | |
f53f4da9 | 215 | if (is_windows_drive(path) || is_windows_drive_prefix(path)) { |
21664424 | 216 | return 1; |
f53f4da9 PB |
217 | } |
218 | return (*path == '/' || *path == '\\'); | |
3b9f94e1 | 219 | #else |
f53f4da9 | 220 | return (*path == '/'); |
3b9f94e1 | 221 | #endif |
3b0d4f61 FB |
222 | } |
223 | ||
83f64091 FB |
224 | /* if filename is absolute, just copy it to dest. Otherwise, build a |
225 | path to it by considering it is relative to base_path. URL are | |
226 | supported. */ | |
227 | void path_combine(char *dest, int dest_size, | |
228 | const char *base_path, | |
229 | const char *filename) | |
3b0d4f61 | 230 | { |
83f64091 FB |
231 | const char *p, *p1; |
232 | int len; | |
233 | ||
234 | if (dest_size <= 0) | |
235 | return; | |
236 | if (path_is_absolute(filename)) { | |
237 | pstrcpy(dest, dest_size, filename); | |
238 | } else { | |
239 | p = strchr(base_path, ':'); | |
240 | if (p) | |
241 | p++; | |
242 | else | |
243 | p = base_path; | |
3b9f94e1 FB |
244 | p1 = strrchr(base_path, '/'); |
245 | #ifdef _WIN32 | |
246 | { | |
247 | const char *p2; | |
248 | p2 = strrchr(base_path, '\\'); | |
249 | if (!p1 || p2 > p1) | |
250 | p1 = p2; | |
251 | } | |
252 | #endif | |
83f64091 FB |
253 | if (p1) |
254 | p1++; | |
255 | else | |
256 | p1 = base_path; | |
257 | if (p1 > p) | |
258 | p = p1; | |
259 | len = p - base_path; | |
260 | if (len > dest_size - 1) | |
261 | len = dest_size - 1; | |
262 | memcpy(dest, base_path, len); | |
263 | dest[len] = '\0'; | |
264 | pstrcat(dest, dest_size, filename); | |
3b0d4f61 | 265 | } |
3b0d4f61 FB |
266 | } |
267 | ||
5efa9d5a | 268 | void bdrv_register(BlockDriver *bdrv) |
ea2384d3 | 269 | { |
8c5873d6 SH |
270 | /* Block drivers without coroutine functions need emulation */ |
271 | if (!bdrv->bdrv_co_readv) { | |
f9f05dc5 KW |
272 | bdrv->bdrv_co_readv = bdrv_co_readv_em; |
273 | bdrv->bdrv_co_writev = bdrv_co_writev_em; | |
274 | ||
f8c35c1d SH |
275 | /* bdrv_co_readv_em()/brdv_co_writev_em() work in terms of aio, so if |
276 | * the block driver lacks aio we need to emulate that too. | |
277 | */ | |
f9f05dc5 KW |
278 | if (!bdrv->bdrv_aio_readv) { |
279 | /* add AIO emulation layer */ | |
280 | bdrv->bdrv_aio_readv = bdrv_aio_readv_em; | |
281 | bdrv->bdrv_aio_writev = bdrv_aio_writev_em; | |
f9f05dc5 | 282 | } |
83f64091 | 283 | } |
b2e12bc6 | 284 | |
8a22f02a | 285 | QLIST_INSERT_HEAD(&bdrv_drivers, bdrv, list); |
ea2384d3 | 286 | } |
b338082b FB |
287 | |
288 | /* create a new block device (by default it is empty) */ | |
289 | BlockDriverState *bdrv_new(const char *device_name) | |
290 | { | |
1b7bdbc1 | 291 | BlockDriverState *bs; |
b338082b | 292 | |
7267c094 | 293 | bs = g_malloc0(sizeof(BlockDriverState)); |
b338082b | 294 | pstrcpy(bs->device_name, sizeof(bs->device_name), device_name); |
ea2384d3 | 295 | if (device_name[0] != '\0') { |
1b7bdbc1 | 296 | QTAILQ_INSERT_TAIL(&bdrv_states, bs, list); |
ea2384d3 | 297 | } |
28a7282a | 298 | bdrv_iostatus_disable(bs); |
b338082b FB |
299 | return bs; |
300 | } | |
301 | ||
ea2384d3 FB |
302 | BlockDriver *bdrv_find_format(const char *format_name) |
303 | { | |
304 | BlockDriver *drv1; | |
8a22f02a SH |
305 | QLIST_FOREACH(drv1, &bdrv_drivers, list) { |
306 | if (!strcmp(drv1->format_name, format_name)) { | |
ea2384d3 | 307 | return drv1; |
8a22f02a | 308 | } |
ea2384d3 FB |
309 | } |
310 | return NULL; | |
311 | } | |
312 | ||
eb852011 MA |
313 | static int bdrv_is_whitelisted(BlockDriver *drv) |
314 | { | |
315 | static const char *whitelist[] = { | |
316 | CONFIG_BDRV_WHITELIST | |
317 | }; | |
318 | const char **p; | |
319 | ||
320 | if (!whitelist[0]) | |
321 | return 1; /* no whitelist, anything goes */ | |
322 | ||
323 | for (p = whitelist; *p; p++) { | |
324 | if (!strcmp(drv->format_name, *p)) { | |
325 | return 1; | |
326 | } | |
327 | } | |
328 | return 0; | |
329 | } | |
330 | ||
331 | BlockDriver *bdrv_find_whitelisted_format(const char *format_name) | |
332 | { | |
333 | BlockDriver *drv = bdrv_find_format(format_name); | |
334 | return drv && bdrv_is_whitelisted(drv) ? drv : NULL; | |
335 | } | |
336 | ||
5b7e1542 ZYW |
337 | typedef struct CreateCo { |
338 | BlockDriver *drv; | |
339 | char *filename; | |
340 | QEMUOptionParameter *options; | |
341 | int ret; | |
342 | } CreateCo; | |
343 | ||
344 | static void coroutine_fn bdrv_create_co_entry(void *opaque) | |
345 | { | |
346 | CreateCo *cco = opaque; | |
347 | assert(cco->drv); | |
348 | ||
349 | cco->ret = cco->drv->bdrv_create(cco->filename, cco->options); | |
350 | } | |
351 | ||
0e7e1989 KW |
352 | int bdrv_create(BlockDriver *drv, const char* filename, |
353 | QEMUOptionParameter *options) | |
ea2384d3 | 354 | { |
5b7e1542 ZYW |
355 | int ret; |
356 | ||
357 | Coroutine *co; | |
358 | CreateCo cco = { | |
359 | .drv = drv, | |
360 | .filename = g_strdup(filename), | |
361 | .options = options, | |
362 | .ret = NOT_DONE, | |
363 | }; | |
364 | ||
365 | if (!drv->bdrv_create) { | |
ea2384d3 | 366 | return -ENOTSUP; |
5b7e1542 ZYW |
367 | } |
368 | ||
369 | if (qemu_in_coroutine()) { | |
370 | /* Fast-path if already in coroutine context */ | |
371 | bdrv_create_co_entry(&cco); | |
372 | } else { | |
373 | co = qemu_coroutine_create(bdrv_create_co_entry); | |
374 | qemu_coroutine_enter(co, &cco); | |
375 | while (cco.ret == NOT_DONE) { | |
376 | qemu_aio_wait(); | |
377 | } | |
378 | } | |
379 | ||
380 | ret = cco.ret; | |
381 | g_free(cco.filename); | |
0e7e1989 | 382 | |
5b7e1542 | 383 | return ret; |
ea2384d3 FB |
384 | } |
385 | ||
84a12e66 CH |
386 | int bdrv_create_file(const char* filename, QEMUOptionParameter *options) |
387 | { | |
388 | BlockDriver *drv; | |
389 | ||
b50cbabc | 390 | drv = bdrv_find_protocol(filename); |
84a12e66 | 391 | if (drv == NULL) { |
16905d71 | 392 | return -ENOENT; |
84a12e66 CH |
393 | } |
394 | ||
395 | return bdrv_create(drv, filename, options); | |
396 | } | |
397 | ||
d5249393 | 398 | #ifdef _WIN32 |
95389c86 | 399 | void get_tmp_filename(char *filename, int size) |
d5249393 | 400 | { |
3b9f94e1 | 401 | char temp_dir[MAX_PATH]; |
3b46e624 | 402 | |
3b9f94e1 FB |
403 | GetTempPath(MAX_PATH, temp_dir); |
404 | GetTempFileName(temp_dir, "qem", 0, filename); | |
d5249393 FB |
405 | } |
406 | #else | |
95389c86 | 407 | void get_tmp_filename(char *filename, int size) |
fc01f7e7 | 408 | { |
67b915a5 | 409 | int fd; |
7ccfb2eb | 410 | const char *tmpdir; |
d5249393 | 411 | /* XXX: race condition possible */ |
0badc1ee AJ |
412 | tmpdir = getenv("TMPDIR"); |
413 | if (!tmpdir) | |
414 | tmpdir = "/tmp"; | |
415 | snprintf(filename, size, "%s/vl.XXXXXX", tmpdir); | |
ea2384d3 FB |
416 | fd = mkstemp(filename); |
417 | close(fd); | |
418 | } | |
d5249393 | 419 | #endif |
fc01f7e7 | 420 | |
84a12e66 CH |
421 | /* |
422 | * Detect host devices. By convention, /dev/cdrom[N] is always | |
423 | * recognized as a host CDROM. | |
424 | */ | |
425 | static BlockDriver *find_hdev_driver(const char *filename) | |
426 | { | |
427 | int score_max = 0, score; | |
428 | BlockDriver *drv = NULL, *d; | |
429 | ||
430 | QLIST_FOREACH(d, &bdrv_drivers, list) { | |
431 | if (d->bdrv_probe_device) { | |
432 | score = d->bdrv_probe_device(filename); | |
433 | if (score > score_max) { | |
434 | score_max = score; | |
435 | drv = d; | |
436 | } | |
437 | } | |
438 | } | |
439 | ||
440 | return drv; | |
441 | } | |
442 | ||
b50cbabc | 443 | BlockDriver *bdrv_find_protocol(const char *filename) |
83f64091 FB |
444 | { |
445 | BlockDriver *drv1; | |
446 | char protocol[128]; | |
1cec71e3 | 447 | int len; |
83f64091 | 448 | const char *p; |
19cb3738 | 449 | |
66f82cee KW |
450 | /* TODO Drivers without bdrv_file_open must be specified explicitly */ |
451 | ||
39508e7a CH |
452 | /* |
453 | * XXX(hch): we really should not let host device detection | |
454 | * override an explicit protocol specification, but moving this | |
455 | * later breaks access to device names with colons in them. | |
456 | * Thanks to the brain-dead persistent naming schemes on udev- | |
457 | * based Linux systems those actually are quite common. | |
458 | */ | |
459 | drv1 = find_hdev_driver(filename); | |
460 | if (drv1) { | |
461 | return drv1; | |
462 | } | |
463 | ||
9e0b22f4 | 464 | if (!path_has_protocol(filename)) { |
39508e7a | 465 | return bdrv_find_format("file"); |
84a12e66 | 466 | } |
9e0b22f4 SH |
467 | p = strchr(filename, ':'); |
468 | assert(p != NULL); | |
1cec71e3 AL |
469 | len = p - filename; |
470 | if (len > sizeof(protocol) - 1) | |
471 | len = sizeof(protocol) - 1; | |
472 | memcpy(protocol, filename, len); | |
473 | protocol[len] = '\0'; | |
8a22f02a | 474 | QLIST_FOREACH(drv1, &bdrv_drivers, list) { |
5fafdf24 | 475 | if (drv1->protocol_name && |
8a22f02a | 476 | !strcmp(drv1->protocol_name, protocol)) { |
83f64091 | 477 | return drv1; |
8a22f02a | 478 | } |
83f64091 FB |
479 | } |
480 | return NULL; | |
481 | } | |
482 | ||
c98ac35d | 483 | static int find_image_format(const char *filename, BlockDriver **pdrv) |
f3a5d3f8 CH |
484 | { |
485 | int ret, score, score_max; | |
486 | BlockDriver *drv1, *drv; | |
487 | uint8_t buf[2048]; | |
488 | BlockDriverState *bs; | |
489 | ||
f5edb014 | 490 | ret = bdrv_file_open(&bs, filename, 0); |
c98ac35d SW |
491 | if (ret < 0) { |
492 | *pdrv = NULL; | |
493 | return ret; | |
494 | } | |
f8ea0b00 | 495 | |
08a00559 KW |
496 | /* Return the raw BlockDriver * to scsi-generic devices or empty drives */ |
497 | if (bs->sg || !bdrv_is_inserted(bs)) { | |
1a396859 | 498 | bdrv_delete(bs); |
c98ac35d SW |
499 | drv = bdrv_find_format("raw"); |
500 | if (!drv) { | |
501 | ret = -ENOENT; | |
502 | } | |
503 | *pdrv = drv; | |
504 | return ret; | |
1a396859 | 505 | } |
f8ea0b00 | 506 | |
83f64091 FB |
507 | ret = bdrv_pread(bs, 0, buf, sizeof(buf)); |
508 | bdrv_delete(bs); | |
509 | if (ret < 0) { | |
c98ac35d SW |
510 | *pdrv = NULL; |
511 | return ret; | |
83f64091 FB |
512 | } |
513 | ||
ea2384d3 | 514 | score_max = 0; |
84a12e66 | 515 | drv = NULL; |
8a22f02a | 516 | QLIST_FOREACH(drv1, &bdrv_drivers, list) { |
83f64091 FB |
517 | if (drv1->bdrv_probe) { |
518 | score = drv1->bdrv_probe(buf, ret, filename); | |
519 | if (score > score_max) { | |
520 | score_max = score; | |
521 | drv = drv1; | |
522 | } | |
0849bf08 | 523 | } |
fc01f7e7 | 524 | } |
c98ac35d SW |
525 | if (!drv) { |
526 | ret = -ENOENT; | |
527 | } | |
528 | *pdrv = drv; | |
529 | return ret; | |
ea2384d3 FB |
530 | } |
531 | ||
51762288 SH |
532 | /** |
533 | * Set the current 'total_sectors' value | |
534 | */ | |
535 | static int refresh_total_sectors(BlockDriverState *bs, int64_t hint) | |
536 | { | |
537 | BlockDriver *drv = bs->drv; | |
538 | ||
396759ad NB |
539 | /* Do not attempt drv->bdrv_getlength() on scsi-generic devices */ |
540 | if (bs->sg) | |
541 | return 0; | |
542 | ||
51762288 SH |
543 | /* query actual device if possible, otherwise just trust the hint */ |
544 | if (drv->bdrv_getlength) { | |
545 | int64_t length = drv->bdrv_getlength(bs); | |
546 | if (length < 0) { | |
547 | return length; | |
548 | } | |
549 | hint = length >> BDRV_SECTOR_BITS; | |
550 | } | |
551 | ||
552 | bs->total_sectors = hint; | |
553 | return 0; | |
554 | } | |
555 | ||
c3993cdc SH |
556 | /** |
557 | * Set open flags for a given cache mode | |
558 | * | |
559 | * Return 0 on success, -1 if the cache mode was invalid. | |
560 | */ | |
561 | int bdrv_parse_cache_flags(const char *mode, int *flags) | |
562 | { | |
563 | *flags &= ~BDRV_O_CACHE_MASK; | |
564 | ||
565 | if (!strcmp(mode, "off") || !strcmp(mode, "none")) { | |
566 | *flags |= BDRV_O_NOCACHE | BDRV_O_CACHE_WB; | |
92196b2f SH |
567 | } else if (!strcmp(mode, "directsync")) { |
568 | *flags |= BDRV_O_NOCACHE; | |
c3993cdc SH |
569 | } else if (!strcmp(mode, "writeback")) { |
570 | *flags |= BDRV_O_CACHE_WB; | |
571 | } else if (!strcmp(mode, "unsafe")) { | |
572 | *flags |= BDRV_O_CACHE_WB; | |
573 | *flags |= BDRV_O_NO_FLUSH; | |
574 | } else if (!strcmp(mode, "writethrough")) { | |
575 | /* this is the default */ | |
576 | } else { | |
577 | return -1; | |
578 | } | |
579 | ||
580 | return 0; | |
581 | } | |
582 | ||
53fec9d3 SH |
583 | /** |
584 | * The copy-on-read flag is actually a reference count so multiple users may | |
585 | * use the feature without worrying about clobbering its previous state. | |
586 | * Copy-on-read stays enabled until all users have called to disable it. | |
587 | */ | |
588 | void bdrv_enable_copy_on_read(BlockDriverState *bs) | |
589 | { | |
590 | bs->copy_on_read++; | |
591 | } | |
592 | ||
593 | void bdrv_disable_copy_on_read(BlockDriverState *bs) | |
594 | { | |
595 | assert(bs->copy_on_read > 0); | |
596 | bs->copy_on_read--; | |
597 | } | |
598 | ||
57915332 KW |
599 | /* |
600 | * Common part for opening disk images and files | |
601 | */ | |
602 | static int bdrv_open_common(BlockDriverState *bs, const char *filename, | |
603 | int flags, BlockDriver *drv) | |
604 | { | |
605 | int ret, open_flags; | |
606 | ||
607 | assert(drv != NULL); | |
608 | ||
28dcee10 SH |
609 | trace_bdrv_open_common(bs, filename, flags, drv->format_name); |
610 | ||
66f82cee | 611 | bs->file = NULL; |
51762288 | 612 | bs->total_sectors = 0; |
57915332 KW |
613 | bs->encrypted = 0; |
614 | bs->valid_key = 0; | |
03f541bd | 615 | bs->sg = 0; |
57915332 | 616 | bs->open_flags = flags; |
03f541bd | 617 | bs->growable = 0; |
57915332 KW |
618 | bs->buffer_alignment = 512; |
619 | ||
53fec9d3 SH |
620 | assert(bs->copy_on_read == 0); /* bdrv_new() and bdrv_close() make it so */ |
621 | if ((flags & BDRV_O_RDWR) && (flags & BDRV_O_COPY_ON_READ)) { | |
622 | bdrv_enable_copy_on_read(bs); | |
623 | } | |
624 | ||
57915332 | 625 | pstrcpy(bs->filename, sizeof(bs->filename), filename); |
03f541bd | 626 | bs->backing_file[0] = '\0'; |
57915332 KW |
627 | |
628 | if (use_bdrv_whitelist && !bdrv_is_whitelisted(drv)) { | |
629 | return -ENOTSUP; | |
630 | } | |
631 | ||
632 | bs->drv = drv; | |
7267c094 | 633 | bs->opaque = g_malloc0(drv->instance_size); |
57915332 | 634 | |
03f541bd | 635 | bs->enable_write_cache = !!(flags & BDRV_O_CACHE_WB); |
57915332 KW |
636 | |
637 | /* | |
638 | * Clear flags that are internal to the block layer before opening the | |
639 | * image. | |
640 | */ | |
641 | open_flags = flags & ~(BDRV_O_SNAPSHOT | BDRV_O_NO_BACKING); | |
642 | ||
643 | /* | |
ebabb67a | 644 | * Snapshots should be writable. |
57915332 KW |
645 | */ |
646 | if (bs->is_temporary) { | |
647 | open_flags |= BDRV_O_RDWR; | |
648 | } | |
649 | ||
e7c63796 SH |
650 | bs->keep_read_only = bs->read_only = !(open_flags & BDRV_O_RDWR); |
651 | ||
66f82cee KW |
652 | /* Open the image, either directly or using a protocol */ |
653 | if (drv->bdrv_file_open) { | |
654 | ret = drv->bdrv_file_open(bs, filename, open_flags); | |
655 | } else { | |
656 | ret = bdrv_file_open(&bs->file, filename, open_flags); | |
657 | if (ret >= 0) { | |
658 | ret = drv->bdrv_open(bs, open_flags); | |
659 | } | |
660 | } | |
661 | ||
57915332 KW |
662 | if (ret < 0) { |
663 | goto free_and_fail; | |
664 | } | |
665 | ||
51762288 SH |
666 | ret = refresh_total_sectors(bs, bs->total_sectors); |
667 | if (ret < 0) { | |
668 | goto free_and_fail; | |
57915332 | 669 | } |
51762288 | 670 | |
57915332 KW |
671 | #ifndef _WIN32 |
672 | if (bs->is_temporary) { | |
673 | unlink(filename); | |
674 | } | |
675 | #endif | |
676 | return 0; | |
677 | ||
678 | free_and_fail: | |
66f82cee KW |
679 | if (bs->file) { |
680 | bdrv_delete(bs->file); | |
681 | bs->file = NULL; | |
682 | } | |
7267c094 | 683 | g_free(bs->opaque); |
57915332 KW |
684 | bs->opaque = NULL; |
685 | bs->drv = NULL; | |
686 | return ret; | |
687 | } | |
688 | ||
b6ce07aa KW |
689 | /* |
690 | * Opens a file using a protocol (file, host_device, nbd, ...) | |
691 | */ | |
83f64091 | 692 | int bdrv_file_open(BlockDriverState **pbs, const char *filename, int flags) |
ea2384d3 | 693 | { |
83f64091 | 694 | BlockDriverState *bs; |
6db95603 | 695 | BlockDriver *drv; |
83f64091 FB |
696 | int ret; |
697 | ||
b50cbabc | 698 | drv = bdrv_find_protocol(filename); |
6db95603 CH |
699 | if (!drv) { |
700 | return -ENOENT; | |
701 | } | |
702 | ||
83f64091 | 703 | bs = bdrv_new(""); |
b6ce07aa | 704 | ret = bdrv_open_common(bs, filename, flags, drv); |
83f64091 FB |
705 | if (ret < 0) { |
706 | bdrv_delete(bs); | |
707 | return ret; | |
3b0d4f61 | 708 | } |
71d0770c | 709 | bs->growable = 1; |
83f64091 FB |
710 | *pbs = bs; |
711 | return 0; | |
712 | } | |
713 | ||
b6ce07aa KW |
714 | /* |
715 | * Opens a disk image (raw, qcow2, vmdk, ...) | |
716 | */ | |
d6e9098e KW |
717 | int bdrv_open(BlockDriverState *bs, const char *filename, int flags, |
718 | BlockDriver *drv) | |
ea2384d3 | 719 | { |
b6ce07aa | 720 | int ret; |
2b572816 | 721 | char tmp_filename[PATH_MAX]; |
712e7874 | 722 | |
83f64091 | 723 | if (flags & BDRV_O_SNAPSHOT) { |
ea2384d3 FB |
724 | BlockDriverState *bs1; |
725 | int64_t total_size; | |
7c96d46e | 726 | int is_protocol = 0; |
91a073a9 KW |
727 | BlockDriver *bdrv_qcow2; |
728 | QEMUOptionParameter *options; | |
b6ce07aa | 729 | char backing_filename[PATH_MAX]; |
3b46e624 | 730 | |
ea2384d3 FB |
731 | /* if snapshot, we create a temporary backing file and open it |
732 | instead of opening 'filename' directly */ | |
33e3963e | 733 | |
ea2384d3 FB |
734 | /* if there is a backing file, use it */ |
735 | bs1 = bdrv_new(""); | |
d6e9098e | 736 | ret = bdrv_open(bs1, filename, 0, drv); |
51d7c00c | 737 | if (ret < 0) { |
ea2384d3 | 738 | bdrv_delete(bs1); |
51d7c00c | 739 | return ret; |
ea2384d3 | 740 | } |
3e82990b | 741 | total_size = bdrv_getlength(bs1) & BDRV_SECTOR_MASK; |
7c96d46e AL |
742 | |
743 | if (bs1->drv && bs1->drv->protocol_name) | |
744 | is_protocol = 1; | |
745 | ||
ea2384d3 | 746 | bdrv_delete(bs1); |
3b46e624 | 747 | |
ea2384d3 | 748 | get_tmp_filename(tmp_filename, sizeof(tmp_filename)); |
7c96d46e AL |
749 | |
750 | /* Real path is meaningless for protocols */ | |
751 | if (is_protocol) | |
752 | snprintf(backing_filename, sizeof(backing_filename), | |
753 | "%s", filename); | |
114cdfa9 KS |
754 | else if (!realpath(filename, backing_filename)) |
755 | return -errno; | |
7c96d46e | 756 | |
91a073a9 KW |
757 | bdrv_qcow2 = bdrv_find_format("qcow2"); |
758 | options = parse_option_parameters("", bdrv_qcow2->create_options, NULL); | |
759 | ||
3e82990b | 760 | set_option_parameter_int(options, BLOCK_OPT_SIZE, total_size); |
91a073a9 KW |
761 | set_option_parameter(options, BLOCK_OPT_BACKING_FILE, backing_filename); |
762 | if (drv) { | |
763 | set_option_parameter(options, BLOCK_OPT_BACKING_FMT, | |
764 | drv->format_name); | |
765 | } | |
766 | ||
767 | ret = bdrv_create(bdrv_qcow2, tmp_filename, options); | |
d748768c | 768 | free_option_parameters(options); |
51d7c00c AL |
769 | if (ret < 0) { |
770 | return ret; | |
ea2384d3 | 771 | } |
91a073a9 | 772 | |
ea2384d3 | 773 | filename = tmp_filename; |
91a073a9 | 774 | drv = bdrv_qcow2; |
ea2384d3 FB |
775 | bs->is_temporary = 1; |
776 | } | |
712e7874 | 777 | |
b6ce07aa | 778 | /* Find the right image format driver */ |
6db95603 | 779 | if (!drv) { |
c98ac35d | 780 | ret = find_image_format(filename, &drv); |
51d7c00c | 781 | } |
6987307c | 782 | |
51d7c00c | 783 | if (!drv) { |
51d7c00c | 784 | goto unlink_and_fail; |
ea2384d3 | 785 | } |
b6ce07aa KW |
786 | |
787 | /* Open the image */ | |
788 | ret = bdrv_open_common(bs, filename, flags, drv); | |
789 | if (ret < 0) { | |
6987307c CH |
790 | goto unlink_and_fail; |
791 | } | |
792 | ||
b6ce07aa KW |
793 | /* If there is a backing file, use it */ |
794 | if ((flags & BDRV_O_NO_BACKING) == 0 && bs->backing_file[0] != '\0') { | |
795 | char backing_filename[PATH_MAX]; | |
796 | int back_flags; | |
797 | BlockDriver *back_drv = NULL; | |
798 | ||
799 | bs->backing_hd = bdrv_new(""); | |
df2dbb4a SH |
800 | |
801 | if (path_has_protocol(bs->backing_file)) { | |
802 | pstrcpy(backing_filename, sizeof(backing_filename), | |
803 | bs->backing_file); | |
804 | } else { | |
805 | path_combine(backing_filename, sizeof(backing_filename), | |
806 | filename, bs->backing_file); | |
807 | } | |
808 | ||
809 | if (bs->backing_format[0] != '\0') { | |
b6ce07aa | 810 | back_drv = bdrv_find_format(bs->backing_format); |
df2dbb4a | 811 | } |
b6ce07aa KW |
812 | |
813 | /* backing files always opened read-only */ | |
814 | back_flags = | |
815 | flags & ~(BDRV_O_RDWR | BDRV_O_SNAPSHOT | BDRV_O_NO_BACKING); | |
816 | ||
817 | ret = bdrv_open(bs->backing_hd, backing_filename, back_flags, back_drv); | |
818 | if (ret < 0) { | |
819 | bdrv_close(bs); | |
820 | return ret; | |
821 | } | |
822 | if (bs->is_temporary) { | |
823 | bs->backing_hd->keep_read_only = !(flags & BDRV_O_RDWR); | |
824 | } else { | |
825 | /* base image inherits from "parent" */ | |
826 | bs->backing_hd->keep_read_only = bs->keep_read_only; | |
827 | } | |
828 | } | |
829 | ||
830 | if (!bdrv_key_required(bs)) { | |
7d4b4ba5 | 831 | bdrv_dev_change_media_cb(bs, true); |
b6ce07aa KW |
832 | } |
833 | ||
98f90dba ZYW |
834 | /* throttling disk I/O limits */ |
835 | if (bs->io_limits_enabled) { | |
836 | bdrv_io_limits_enable(bs); | |
837 | } | |
838 | ||
b6ce07aa KW |
839 | return 0; |
840 | ||
841 | unlink_and_fail: | |
842 | if (bs->is_temporary) { | |
843 | unlink(filename); | |
844 | } | |
845 | return ret; | |
846 | } | |
847 | ||
fc01f7e7 FB |
848 | void bdrv_close(BlockDriverState *bs) |
849 | { | |
80ccf93b | 850 | bdrv_flush(bs); |
19cb3738 | 851 | if (bs->drv) { |
3e914655 PB |
852 | if (bs->job) { |
853 | block_job_cancel_sync(bs->job); | |
854 | } | |
7094f12f KW |
855 | bdrv_drain_all(); |
856 | ||
f9092b10 MA |
857 | if (bs == bs_snapshots) { |
858 | bs_snapshots = NULL; | |
859 | } | |
557df6ac | 860 | if (bs->backing_hd) { |
ea2384d3 | 861 | bdrv_delete(bs->backing_hd); |
557df6ac SH |
862 | bs->backing_hd = NULL; |
863 | } | |
ea2384d3 | 864 | bs->drv->bdrv_close(bs); |
7267c094 | 865 | g_free(bs->opaque); |
ea2384d3 FB |
866 | #ifdef _WIN32 |
867 | if (bs->is_temporary) { | |
868 | unlink(bs->filename); | |
869 | } | |
67b915a5 | 870 | #endif |
ea2384d3 FB |
871 | bs->opaque = NULL; |
872 | bs->drv = NULL; | |
53fec9d3 | 873 | bs->copy_on_read = 0; |
a275fa42 PB |
874 | bs->backing_file[0] = '\0'; |
875 | bs->backing_format[0] = '\0'; | |
b338082b | 876 | |
66f82cee | 877 | if (bs->file != NULL) { |
0ac9377d PB |
878 | bdrv_delete(bs->file); |
879 | bs->file = NULL; | |
66f82cee KW |
880 | } |
881 | ||
7d4b4ba5 | 882 | bdrv_dev_change_media_cb(bs, false); |
b338082b | 883 | } |
98f90dba ZYW |
884 | |
885 | /*throttling disk I/O limits*/ | |
886 | if (bs->io_limits_enabled) { | |
887 | bdrv_io_limits_disable(bs); | |
888 | } | |
b338082b FB |
889 | } |
890 | ||
2bc93fed MK |
891 | void bdrv_close_all(void) |
892 | { | |
893 | BlockDriverState *bs; | |
894 | ||
895 | QTAILQ_FOREACH(bs, &bdrv_states, list) { | |
896 | bdrv_close(bs); | |
897 | } | |
898 | } | |
899 | ||
922453bc SH |
900 | /* |
901 | * Wait for pending requests to complete across all BlockDriverStates | |
902 | * | |
903 | * This function does not flush data to disk, use bdrv_flush_all() for that | |
904 | * after calling this function. | |
4c355d53 ZYW |
905 | * |
906 | * Note that completion of an asynchronous I/O operation can trigger any | |
907 | * number of other I/O operations on other devices---for example a coroutine | |
908 | * can be arbitrarily complex and a constant flow of I/O can come until the | |
909 | * coroutine is complete. Because of this, it is not possible to have a | |
910 | * function to drain a single device's I/O queue. | |
922453bc SH |
911 | */ |
912 | void bdrv_drain_all(void) | |
913 | { | |
914 | BlockDriverState *bs; | |
4c355d53 ZYW |
915 | bool busy; |
916 | ||
917 | do { | |
918 | busy = qemu_aio_wait(); | |
922453bc | 919 | |
4c355d53 ZYW |
920 | /* FIXME: We do not have timer support here, so this is effectively |
921 | * a busy wait. | |
922 | */ | |
923 | QTAILQ_FOREACH(bs, &bdrv_states, list) { | |
924 | if (!qemu_co_queue_empty(&bs->throttled_reqs)) { | |
925 | qemu_co_queue_restart_all(&bs->throttled_reqs); | |
926 | busy = true; | |
927 | } | |
928 | } | |
929 | } while (busy); | |
922453bc SH |
930 | |
931 | /* If requests are still pending there is a bug somewhere */ | |
932 | QTAILQ_FOREACH(bs, &bdrv_states, list) { | |
933 | assert(QLIST_EMPTY(&bs->tracked_requests)); | |
934 | assert(qemu_co_queue_empty(&bs->throttled_reqs)); | |
935 | } | |
936 | } | |
937 | ||
d22b2f41 RH |
938 | /* make a BlockDriverState anonymous by removing from bdrv_state list. |
939 | Also, NULL terminate the device_name to prevent double remove */ | |
940 | void bdrv_make_anon(BlockDriverState *bs) | |
941 | { | |
942 | if (bs->device_name[0] != '\0') { | |
943 | QTAILQ_REMOVE(&bdrv_states, bs, list); | |
944 | } | |
945 | bs->device_name[0] = '\0'; | |
946 | } | |
947 | ||
e023b2e2 PB |
948 | static void bdrv_rebind(BlockDriverState *bs) |
949 | { | |
950 | if (bs->drv && bs->drv->bdrv_rebind) { | |
951 | bs->drv->bdrv_rebind(bs); | |
952 | } | |
953 | } | |
954 | ||
8802d1fd JC |
955 | /* |
956 | * Add new bs contents at the top of an image chain while the chain is | |
957 | * live, while keeping required fields on the top layer. | |
958 | * | |
959 | * This will modify the BlockDriverState fields, and swap contents | |
960 | * between bs_new and bs_top. Both bs_new and bs_top are modified. | |
961 | * | |
f6801b83 JC |
962 | * bs_new is required to be anonymous. |
963 | * | |
8802d1fd JC |
964 | * This function does not create any image files. |
965 | */ | |
966 | void bdrv_append(BlockDriverState *bs_new, BlockDriverState *bs_top) | |
967 | { | |
968 | BlockDriverState tmp; | |
969 | ||
f6801b83 JC |
970 | /* bs_new must be anonymous */ |
971 | assert(bs_new->device_name[0] == '\0'); | |
8802d1fd JC |
972 | |
973 | tmp = *bs_new; | |
974 | ||
975 | /* there are some fields that need to stay on the top layer: */ | |
3a389e79 | 976 | tmp.open_flags = bs_top->open_flags; |
8802d1fd JC |
977 | |
978 | /* dev info */ | |
979 | tmp.dev_ops = bs_top->dev_ops; | |
980 | tmp.dev_opaque = bs_top->dev_opaque; | |
981 | tmp.dev = bs_top->dev; | |
982 | tmp.buffer_alignment = bs_top->buffer_alignment; | |
983 | tmp.copy_on_read = bs_top->copy_on_read; | |
984 | ||
985 | /* i/o timing parameters */ | |
986 | tmp.slice_time = bs_top->slice_time; | |
987 | tmp.slice_start = bs_top->slice_start; | |
988 | tmp.slice_end = bs_top->slice_end; | |
989 | tmp.io_limits = bs_top->io_limits; | |
990 | tmp.io_base = bs_top->io_base; | |
991 | tmp.throttled_reqs = bs_top->throttled_reqs; | |
992 | tmp.block_timer = bs_top->block_timer; | |
993 | tmp.io_limits_enabled = bs_top->io_limits_enabled; | |
994 | ||
995 | /* geometry */ | |
996 | tmp.cyls = bs_top->cyls; | |
997 | tmp.heads = bs_top->heads; | |
998 | tmp.secs = bs_top->secs; | |
999 | tmp.translation = bs_top->translation; | |
1000 | ||
1001 | /* r/w error */ | |
1002 | tmp.on_read_error = bs_top->on_read_error; | |
1003 | tmp.on_write_error = bs_top->on_write_error; | |
1004 | ||
1005 | /* i/o status */ | |
1006 | tmp.iostatus_enabled = bs_top->iostatus_enabled; | |
1007 | tmp.iostatus = bs_top->iostatus; | |
1008 | ||
1009 | /* keep the same entry in bdrv_states */ | |
1010 | pstrcpy(tmp.device_name, sizeof(tmp.device_name), bs_top->device_name); | |
1011 | tmp.list = bs_top->list; | |
1012 | ||
1013 | /* The contents of 'tmp' will become bs_top, as we are | |
1014 | * swapping bs_new and bs_top contents. */ | |
1015 | tmp.backing_hd = bs_new; | |
1016 | pstrcpy(tmp.backing_file, sizeof(tmp.backing_file), bs_top->filename); | |
f6801b83 | 1017 | bdrv_get_format(bs_top, tmp.backing_format, sizeof(tmp.backing_format)); |
8802d1fd JC |
1018 | |
1019 | /* swap contents of the fixed new bs and the current top */ | |
1020 | *bs_new = *bs_top; | |
1021 | *bs_top = tmp; | |
1022 | ||
f6801b83 JC |
1023 | /* device_name[] was carried over from the old bs_top. bs_new |
1024 | * shouldn't be in bdrv_states, so we need to make device_name[] | |
1025 | * reflect the anonymity of bs_new | |
1026 | */ | |
1027 | bs_new->device_name[0] = '\0'; | |
1028 | ||
8802d1fd JC |
1029 | /* clear the copied fields in the new backing file */ |
1030 | bdrv_detach_dev(bs_new, bs_new->dev); | |
1031 | ||
1032 | qemu_co_queue_init(&bs_new->throttled_reqs); | |
1033 | memset(&bs_new->io_base, 0, sizeof(bs_new->io_base)); | |
1034 | memset(&bs_new->io_limits, 0, sizeof(bs_new->io_limits)); | |
1035 | bdrv_iostatus_disable(bs_new); | |
1036 | ||
1037 | /* we don't use bdrv_io_limits_disable() for this, because we don't want | |
1038 | * to affect or delete the block_timer, as it has been moved to bs_top */ | |
1039 | bs_new->io_limits_enabled = false; | |
1040 | bs_new->block_timer = NULL; | |
1041 | bs_new->slice_time = 0; | |
1042 | bs_new->slice_start = 0; | |
1043 | bs_new->slice_end = 0; | |
e023b2e2 PB |
1044 | |
1045 | bdrv_rebind(bs_new); | |
1046 | bdrv_rebind(bs_top); | |
8802d1fd JC |
1047 | } |
1048 | ||
b338082b FB |
1049 | void bdrv_delete(BlockDriverState *bs) |
1050 | { | |
fa879d62 | 1051 | assert(!bs->dev); |
3e914655 PB |
1052 | assert(!bs->job); |
1053 | assert(!bs->in_use); | |
18846dee | 1054 | |
1b7bdbc1 | 1055 | /* remove from list, if necessary */ |
d22b2f41 | 1056 | bdrv_make_anon(bs); |
34c6f050 | 1057 | |
b338082b | 1058 | bdrv_close(bs); |
66f82cee | 1059 | |
f9092b10 | 1060 | assert(bs != bs_snapshots); |
7267c094 | 1061 | g_free(bs); |
fc01f7e7 FB |
1062 | } |
1063 | ||
fa879d62 MA |
1064 | int bdrv_attach_dev(BlockDriverState *bs, void *dev) |
1065 | /* TODO change to DeviceState *dev when all users are qdevified */ | |
18846dee | 1066 | { |
fa879d62 | 1067 | if (bs->dev) { |
18846dee MA |
1068 | return -EBUSY; |
1069 | } | |
fa879d62 | 1070 | bs->dev = dev; |
28a7282a | 1071 | bdrv_iostatus_reset(bs); |
18846dee MA |
1072 | return 0; |
1073 | } | |
1074 | ||
fa879d62 MA |
1075 | /* TODO qdevified devices don't use this, remove when devices are qdevified */ |
1076 | void bdrv_attach_dev_nofail(BlockDriverState *bs, void *dev) | |
18846dee | 1077 | { |
fa879d62 MA |
1078 | if (bdrv_attach_dev(bs, dev) < 0) { |
1079 | abort(); | |
1080 | } | |
1081 | } | |
1082 | ||
1083 | void bdrv_detach_dev(BlockDriverState *bs, void *dev) | |
1084 | /* TODO change to DeviceState *dev when all users are qdevified */ | |
1085 | { | |
1086 | assert(bs->dev == dev); | |
1087 | bs->dev = NULL; | |
0e49de52 MA |
1088 | bs->dev_ops = NULL; |
1089 | bs->dev_opaque = NULL; | |
29e05f20 | 1090 | bs->buffer_alignment = 512; |
18846dee MA |
1091 | } |
1092 | ||
fa879d62 MA |
1093 | /* TODO change to return DeviceState * when all users are qdevified */ |
1094 | void *bdrv_get_attached_dev(BlockDriverState *bs) | |
18846dee | 1095 | { |
fa879d62 | 1096 | return bs->dev; |
18846dee MA |
1097 | } |
1098 | ||
0e49de52 MA |
1099 | void bdrv_set_dev_ops(BlockDriverState *bs, const BlockDevOps *ops, |
1100 | void *opaque) | |
1101 | { | |
1102 | bs->dev_ops = ops; | |
1103 | bs->dev_opaque = opaque; | |
2c6942fa MA |
1104 | if (bdrv_dev_has_removable_media(bs) && bs == bs_snapshots) { |
1105 | bs_snapshots = NULL; | |
1106 | } | |
0e49de52 MA |
1107 | } |
1108 | ||
329c0a48 LC |
1109 | void bdrv_emit_qmp_error_event(const BlockDriverState *bdrv, |
1110 | BlockQMPEventAction action, int is_read) | |
1111 | { | |
1112 | QObject *data; | |
1113 | const char *action_str; | |
1114 | ||
1115 | switch (action) { | |
1116 | case BDRV_ACTION_REPORT: | |
1117 | action_str = "report"; | |
1118 | break; | |
1119 | case BDRV_ACTION_IGNORE: | |
1120 | action_str = "ignore"; | |
1121 | break; | |
1122 | case BDRV_ACTION_STOP: | |
1123 | action_str = "stop"; | |
1124 | break; | |
1125 | default: | |
1126 | abort(); | |
1127 | } | |
1128 | ||
1129 | data = qobject_from_jsonf("{ 'device': %s, 'action': %s, 'operation': %s }", | |
1130 | bdrv->device_name, | |
1131 | action_str, | |
1132 | is_read ? "read" : "write"); | |
1133 | monitor_protocol_event(QEVENT_BLOCK_IO_ERROR, data); | |
1134 | ||
1135 | qobject_decref(data); | |
1136 | } | |
1137 | ||
6f382ed2 LC |
1138 | static void bdrv_emit_qmp_eject_event(BlockDriverState *bs, bool ejected) |
1139 | { | |
1140 | QObject *data; | |
1141 | ||
1142 | data = qobject_from_jsonf("{ 'device': %s, 'tray-open': %i }", | |
1143 | bdrv_get_device_name(bs), ejected); | |
1144 | monitor_protocol_event(QEVENT_DEVICE_TRAY_MOVED, data); | |
1145 | ||
1146 | qobject_decref(data); | |
1147 | } | |
1148 | ||
7d4b4ba5 | 1149 | static void bdrv_dev_change_media_cb(BlockDriverState *bs, bool load) |
0e49de52 | 1150 | { |
145feb17 | 1151 | if (bs->dev_ops && bs->dev_ops->change_media_cb) { |
6f382ed2 | 1152 | bool tray_was_closed = !bdrv_dev_is_tray_open(bs); |
7d4b4ba5 | 1153 | bs->dev_ops->change_media_cb(bs->dev_opaque, load); |
6f382ed2 LC |
1154 | if (tray_was_closed) { |
1155 | /* tray open */ | |
1156 | bdrv_emit_qmp_eject_event(bs, true); | |
1157 | } | |
1158 | if (load) { | |
1159 | /* tray close */ | |
1160 | bdrv_emit_qmp_eject_event(bs, false); | |
1161 | } | |
145feb17 MA |
1162 | } |
1163 | } | |
1164 | ||
2c6942fa MA |
1165 | bool bdrv_dev_has_removable_media(BlockDriverState *bs) |
1166 | { | |
1167 | return !bs->dev || (bs->dev_ops && bs->dev_ops->change_media_cb); | |
1168 | } | |
1169 | ||
025ccaa7 PB |
1170 | void bdrv_dev_eject_request(BlockDriverState *bs, bool force) |
1171 | { | |
1172 | if (bs->dev_ops && bs->dev_ops->eject_request_cb) { | |
1173 | bs->dev_ops->eject_request_cb(bs->dev_opaque, force); | |
1174 | } | |
1175 | } | |
1176 | ||
e4def80b MA |
1177 | bool bdrv_dev_is_tray_open(BlockDriverState *bs) |
1178 | { | |
1179 | if (bs->dev_ops && bs->dev_ops->is_tray_open) { | |
1180 | return bs->dev_ops->is_tray_open(bs->dev_opaque); | |
1181 | } | |
1182 | return false; | |
1183 | } | |
1184 | ||
145feb17 MA |
1185 | static void bdrv_dev_resize_cb(BlockDriverState *bs) |
1186 | { | |
1187 | if (bs->dev_ops && bs->dev_ops->resize_cb) { | |
1188 | bs->dev_ops->resize_cb(bs->dev_opaque); | |
0e49de52 MA |
1189 | } |
1190 | } | |
1191 | ||
f107639a MA |
1192 | bool bdrv_dev_is_medium_locked(BlockDriverState *bs) |
1193 | { | |
1194 | if (bs->dev_ops && bs->dev_ops->is_medium_locked) { | |
1195 | return bs->dev_ops->is_medium_locked(bs->dev_opaque); | |
1196 | } | |
1197 | return false; | |
1198 | } | |
1199 | ||
e97fc193 AL |
1200 | /* |
1201 | * Run consistency checks on an image | |
1202 | * | |
e076f338 | 1203 | * Returns 0 if the check could be completed (it doesn't mean that the image is |
a1c7273b | 1204 | * free of errors) or -errno when an internal error occurred. The results of the |
e076f338 | 1205 | * check are stored in res. |
e97fc193 | 1206 | */ |
e076f338 | 1207 | int bdrv_check(BlockDriverState *bs, BdrvCheckResult *res) |
e97fc193 AL |
1208 | { |
1209 | if (bs->drv->bdrv_check == NULL) { | |
1210 | return -ENOTSUP; | |
1211 | } | |
1212 | ||
e076f338 | 1213 | memset(res, 0, sizeof(*res)); |
9ac228e0 | 1214 | return bs->drv->bdrv_check(bs, res); |
e97fc193 AL |
1215 | } |
1216 | ||
8a426614 KW |
1217 | #define COMMIT_BUF_SECTORS 2048 |
1218 | ||
33e3963e FB |
1219 | /* commit COW file into the raw image */ |
1220 | int bdrv_commit(BlockDriverState *bs) | |
1221 | { | |
19cb3738 | 1222 | BlockDriver *drv = bs->drv; |
ee181196 | 1223 | BlockDriver *backing_drv; |
8a426614 KW |
1224 | int64_t sector, total_sectors; |
1225 | int n, ro, open_flags; | |
4dca4b63 | 1226 | int ret = 0, rw_ret = 0; |
8a426614 | 1227 | uint8_t *buf; |
4dca4b63 NS |
1228 | char filename[1024]; |
1229 | BlockDriverState *bs_rw, *bs_ro; | |
33e3963e | 1230 | |
19cb3738 FB |
1231 | if (!drv) |
1232 | return -ENOMEDIUM; | |
4dca4b63 NS |
1233 | |
1234 | if (!bs->backing_hd) { | |
1235 | return -ENOTSUP; | |
33e3963e FB |
1236 | } |
1237 | ||
4dca4b63 NS |
1238 | if (bs->backing_hd->keep_read_only) { |
1239 | return -EACCES; | |
1240 | } | |
ee181196 | 1241 | |
2d3735d3 SH |
1242 | if (bdrv_in_use(bs) || bdrv_in_use(bs->backing_hd)) { |
1243 | return -EBUSY; | |
1244 | } | |
1245 | ||
ee181196 | 1246 | backing_drv = bs->backing_hd->drv; |
4dca4b63 NS |
1247 | ro = bs->backing_hd->read_only; |
1248 | strncpy(filename, bs->backing_hd->filename, sizeof(filename)); | |
1249 | open_flags = bs->backing_hd->open_flags; | |
1250 | ||
1251 | if (ro) { | |
1252 | /* re-open as RW */ | |
1253 | bdrv_delete(bs->backing_hd); | |
1254 | bs->backing_hd = NULL; | |
1255 | bs_rw = bdrv_new(""); | |
ee181196 KW |
1256 | rw_ret = bdrv_open(bs_rw, filename, open_flags | BDRV_O_RDWR, |
1257 | backing_drv); | |
4dca4b63 NS |
1258 | if (rw_ret < 0) { |
1259 | bdrv_delete(bs_rw); | |
1260 | /* try to re-open read-only */ | |
1261 | bs_ro = bdrv_new(""); | |
ee181196 KW |
1262 | ret = bdrv_open(bs_ro, filename, open_flags & ~BDRV_O_RDWR, |
1263 | backing_drv); | |
4dca4b63 NS |
1264 | if (ret < 0) { |
1265 | bdrv_delete(bs_ro); | |
1266 | /* drive not functional anymore */ | |
1267 | bs->drv = NULL; | |
1268 | return ret; | |
1269 | } | |
1270 | bs->backing_hd = bs_ro; | |
1271 | return rw_ret; | |
1272 | } | |
1273 | bs->backing_hd = bs_rw; | |
ea2384d3 | 1274 | } |
33e3963e | 1275 | |
6ea44308 | 1276 | total_sectors = bdrv_getlength(bs) >> BDRV_SECTOR_BITS; |
7267c094 | 1277 | buf = g_malloc(COMMIT_BUF_SECTORS * BDRV_SECTOR_SIZE); |
8a426614 KW |
1278 | |
1279 | for (sector = 0; sector < total_sectors; sector += n) { | |
05c4af54 | 1280 | if (bdrv_is_allocated(bs, sector, COMMIT_BUF_SECTORS, &n)) { |
8a426614 KW |
1281 | |
1282 | if (bdrv_read(bs, sector, buf, n) != 0) { | |
1283 | ret = -EIO; | |
1284 | goto ro_cleanup; | |
1285 | } | |
1286 | ||
1287 | if (bdrv_write(bs->backing_hd, sector, buf, n) != 0) { | |
1288 | ret = -EIO; | |
1289 | goto ro_cleanup; | |
1290 | } | |
ea2384d3 | 1291 | } |
33e3963e | 1292 | } |
95389c86 | 1293 | |
1d44952f CH |
1294 | if (drv->bdrv_make_empty) { |
1295 | ret = drv->bdrv_make_empty(bs); | |
1296 | bdrv_flush(bs); | |
1297 | } | |
95389c86 | 1298 | |
3f5075ae CH |
1299 | /* |
1300 | * Make sure all data we wrote to the backing device is actually | |
1301 | * stable on disk. | |
1302 | */ | |
1303 | if (bs->backing_hd) | |
1304 | bdrv_flush(bs->backing_hd); | |
4dca4b63 NS |
1305 | |
1306 | ro_cleanup: | |
7267c094 | 1307 | g_free(buf); |
4dca4b63 NS |
1308 | |
1309 | if (ro) { | |
1310 | /* re-open as RO */ | |
1311 | bdrv_delete(bs->backing_hd); | |
1312 | bs->backing_hd = NULL; | |
1313 | bs_ro = bdrv_new(""); | |
ee181196 KW |
1314 | ret = bdrv_open(bs_ro, filename, open_flags & ~BDRV_O_RDWR, |
1315 | backing_drv); | |
4dca4b63 NS |
1316 | if (ret < 0) { |
1317 | bdrv_delete(bs_ro); | |
1318 | /* drive not functional anymore */ | |
1319 | bs->drv = NULL; | |
1320 | return ret; | |
1321 | } | |
1322 | bs->backing_hd = bs_ro; | |
1323 | bs->backing_hd->keep_read_only = 0; | |
1324 | } | |
1325 | ||
1d44952f | 1326 | return ret; |
33e3963e FB |
1327 | } |
1328 | ||
e8877497 | 1329 | int bdrv_commit_all(void) |
6ab4b5ab MA |
1330 | { |
1331 | BlockDriverState *bs; | |
1332 | ||
1333 | QTAILQ_FOREACH(bs, &bdrv_states, list) { | |
e8877497 SH |
1334 | int ret = bdrv_commit(bs); |
1335 | if (ret < 0) { | |
1336 | return ret; | |
1337 | } | |
6ab4b5ab | 1338 | } |
e8877497 | 1339 | return 0; |
6ab4b5ab MA |
1340 | } |
1341 | ||
dbffbdcf SH |
1342 | struct BdrvTrackedRequest { |
1343 | BlockDriverState *bs; | |
1344 | int64_t sector_num; | |
1345 | int nb_sectors; | |
1346 | bool is_write; | |
1347 | QLIST_ENTRY(BdrvTrackedRequest) list; | |
5f8b6491 | 1348 | Coroutine *co; /* owner, used for deadlock detection */ |
f4658285 | 1349 | CoQueue wait_queue; /* coroutines blocked on this request */ |
dbffbdcf SH |
1350 | }; |
1351 | ||
1352 | /** | |
1353 | * Remove an active request from the tracked requests list | |
1354 | * | |
1355 | * This function should be called when a tracked request is completing. | |
1356 | */ | |
1357 | static void tracked_request_end(BdrvTrackedRequest *req) | |
1358 | { | |
1359 | QLIST_REMOVE(req, list); | |
f4658285 | 1360 | qemu_co_queue_restart_all(&req->wait_queue); |
dbffbdcf SH |
1361 | } |
1362 | ||
1363 | /** | |
1364 | * Add an active request to the tracked requests list | |
1365 | */ | |
1366 | static void tracked_request_begin(BdrvTrackedRequest *req, | |
1367 | BlockDriverState *bs, | |
1368 | int64_t sector_num, | |
1369 | int nb_sectors, bool is_write) | |
1370 | { | |
1371 | *req = (BdrvTrackedRequest){ | |
1372 | .bs = bs, | |
1373 | .sector_num = sector_num, | |
1374 | .nb_sectors = nb_sectors, | |
1375 | .is_write = is_write, | |
5f8b6491 | 1376 | .co = qemu_coroutine_self(), |
dbffbdcf SH |
1377 | }; |
1378 | ||
f4658285 SH |
1379 | qemu_co_queue_init(&req->wait_queue); |
1380 | ||
dbffbdcf SH |
1381 | QLIST_INSERT_HEAD(&bs->tracked_requests, req, list); |
1382 | } | |
1383 | ||
d83947ac SH |
1384 | /** |
1385 | * Round a region to cluster boundaries | |
1386 | */ | |
1387 | static void round_to_clusters(BlockDriverState *bs, | |
1388 | int64_t sector_num, int nb_sectors, | |
1389 | int64_t *cluster_sector_num, | |
1390 | int *cluster_nb_sectors) | |
1391 | { | |
1392 | BlockDriverInfo bdi; | |
1393 | ||
1394 | if (bdrv_get_info(bs, &bdi) < 0 || bdi.cluster_size == 0) { | |
1395 | *cluster_sector_num = sector_num; | |
1396 | *cluster_nb_sectors = nb_sectors; | |
1397 | } else { | |
1398 | int64_t c = bdi.cluster_size / BDRV_SECTOR_SIZE; | |
1399 | *cluster_sector_num = QEMU_ALIGN_DOWN(sector_num, c); | |
1400 | *cluster_nb_sectors = QEMU_ALIGN_UP(sector_num - *cluster_sector_num + | |
1401 | nb_sectors, c); | |
1402 | } | |
1403 | } | |
1404 | ||
f4658285 SH |
1405 | static bool tracked_request_overlaps(BdrvTrackedRequest *req, |
1406 | int64_t sector_num, int nb_sectors) { | |
d83947ac SH |
1407 | /* aaaa bbbb */ |
1408 | if (sector_num >= req->sector_num + req->nb_sectors) { | |
1409 | return false; | |
1410 | } | |
1411 | /* bbbb aaaa */ | |
1412 | if (req->sector_num >= sector_num + nb_sectors) { | |
1413 | return false; | |
1414 | } | |
1415 | return true; | |
f4658285 SH |
1416 | } |
1417 | ||
1418 | static void coroutine_fn wait_for_overlapping_requests(BlockDriverState *bs, | |
1419 | int64_t sector_num, int nb_sectors) | |
1420 | { | |
1421 | BdrvTrackedRequest *req; | |
d83947ac SH |
1422 | int64_t cluster_sector_num; |
1423 | int cluster_nb_sectors; | |
f4658285 SH |
1424 | bool retry; |
1425 | ||
d83947ac SH |
1426 | /* If we touch the same cluster it counts as an overlap. This guarantees |
1427 | * that allocating writes will be serialized and not race with each other | |
1428 | * for the same cluster. For example, in copy-on-read it ensures that the | |
1429 | * CoR read and write operations are atomic and guest writes cannot | |
1430 | * interleave between them. | |
1431 | */ | |
1432 | round_to_clusters(bs, sector_num, nb_sectors, | |
1433 | &cluster_sector_num, &cluster_nb_sectors); | |
1434 | ||
f4658285 SH |
1435 | do { |
1436 | retry = false; | |
1437 | QLIST_FOREACH(req, &bs->tracked_requests, list) { | |
d83947ac SH |
1438 | if (tracked_request_overlaps(req, cluster_sector_num, |
1439 | cluster_nb_sectors)) { | |
5f8b6491 SH |
1440 | /* Hitting this means there was a reentrant request, for |
1441 | * example, a block driver issuing nested requests. This must | |
1442 | * never happen since it means deadlock. | |
1443 | */ | |
1444 | assert(qemu_coroutine_self() != req->co); | |
1445 | ||
f4658285 SH |
1446 | qemu_co_queue_wait(&req->wait_queue); |
1447 | retry = true; | |
1448 | break; | |
1449 | } | |
1450 | } | |
1451 | } while (retry); | |
1452 | } | |
1453 | ||
756e6736 KW |
1454 | /* |
1455 | * Return values: | |
1456 | * 0 - success | |
1457 | * -EINVAL - backing format specified, but no file | |
1458 | * -ENOSPC - can't update the backing file because no space is left in the | |
1459 | * image file header | |
1460 | * -ENOTSUP - format driver doesn't support changing the backing file | |
1461 | */ | |
1462 | int bdrv_change_backing_file(BlockDriverState *bs, | |
1463 | const char *backing_file, const char *backing_fmt) | |
1464 | { | |
1465 | BlockDriver *drv = bs->drv; | |
469ef350 | 1466 | int ret; |
756e6736 | 1467 | |
5f377794 PB |
1468 | /* Backing file format doesn't make sense without a backing file */ |
1469 | if (backing_fmt && !backing_file) { | |
1470 | return -EINVAL; | |
1471 | } | |
1472 | ||
756e6736 | 1473 | if (drv->bdrv_change_backing_file != NULL) { |
469ef350 | 1474 | ret = drv->bdrv_change_backing_file(bs, backing_file, backing_fmt); |
756e6736 | 1475 | } else { |
469ef350 | 1476 | ret = -ENOTSUP; |
756e6736 | 1477 | } |
469ef350 PB |
1478 | |
1479 | if (ret == 0) { | |
1480 | pstrcpy(bs->backing_file, sizeof(bs->backing_file), backing_file ?: ""); | |
1481 | pstrcpy(bs->backing_format, sizeof(bs->backing_format), backing_fmt ?: ""); | |
1482 | } | |
1483 | return ret; | |
756e6736 KW |
1484 | } |
1485 | ||
71d0770c AL |
1486 | static int bdrv_check_byte_request(BlockDriverState *bs, int64_t offset, |
1487 | size_t size) | |
1488 | { | |
1489 | int64_t len; | |
1490 | ||
1491 | if (!bdrv_is_inserted(bs)) | |
1492 | return -ENOMEDIUM; | |
1493 | ||
1494 | if (bs->growable) | |
1495 | return 0; | |
1496 | ||
1497 | len = bdrv_getlength(bs); | |
1498 | ||
fbb7b4e0 KW |
1499 | if (offset < 0) |
1500 | return -EIO; | |
1501 | ||
1502 | if ((offset > len) || (len - offset < size)) | |
71d0770c AL |
1503 | return -EIO; |
1504 | ||
1505 | return 0; | |
1506 | } | |
1507 | ||
1508 | static int bdrv_check_request(BlockDriverState *bs, int64_t sector_num, | |
1509 | int nb_sectors) | |
1510 | { | |
eb5a3165 JS |
1511 | return bdrv_check_byte_request(bs, sector_num * BDRV_SECTOR_SIZE, |
1512 | nb_sectors * BDRV_SECTOR_SIZE); | |
71d0770c AL |
1513 | } |
1514 | ||
1c9805a3 SH |
1515 | typedef struct RwCo { |
1516 | BlockDriverState *bs; | |
1517 | int64_t sector_num; | |
1518 | int nb_sectors; | |
1519 | QEMUIOVector *qiov; | |
1520 | bool is_write; | |
1521 | int ret; | |
1522 | } RwCo; | |
1523 | ||
1524 | static void coroutine_fn bdrv_rw_co_entry(void *opaque) | |
fc01f7e7 | 1525 | { |
1c9805a3 | 1526 | RwCo *rwco = opaque; |
ea2384d3 | 1527 | |
1c9805a3 SH |
1528 | if (!rwco->is_write) { |
1529 | rwco->ret = bdrv_co_do_readv(rwco->bs, rwco->sector_num, | |
470c0504 | 1530 | rwco->nb_sectors, rwco->qiov, 0); |
1c9805a3 SH |
1531 | } else { |
1532 | rwco->ret = bdrv_co_do_writev(rwco->bs, rwco->sector_num, | |
f08f2dda | 1533 | rwco->nb_sectors, rwco->qiov, 0); |
1c9805a3 SH |
1534 | } |
1535 | } | |
e7a8a783 | 1536 | |
1c9805a3 SH |
1537 | /* |
1538 | * Process a synchronous request using coroutines | |
1539 | */ | |
1540 | static int bdrv_rw_co(BlockDriverState *bs, int64_t sector_num, uint8_t *buf, | |
1541 | int nb_sectors, bool is_write) | |
1542 | { | |
1543 | QEMUIOVector qiov; | |
1544 | struct iovec iov = { | |
1545 | .iov_base = (void *)buf, | |
1546 | .iov_len = nb_sectors * BDRV_SECTOR_SIZE, | |
1547 | }; | |
1548 | Coroutine *co; | |
1549 | RwCo rwco = { | |
1550 | .bs = bs, | |
1551 | .sector_num = sector_num, | |
1552 | .nb_sectors = nb_sectors, | |
1553 | .qiov = &qiov, | |
1554 | .is_write = is_write, | |
1555 | .ret = NOT_DONE, | |
1556 | }; | |
e7a8a783 | 1557 | |
1c9805a3 | 1558 | qemu_iovec_init_external(&qiov, &iov, 1); |
e7a8a783 | 1559 | |
498e386c ZYW |
1560 | /** |
1561 | * In sync call context, when the vcpu is blocked, this throttling timer | |
1562 | * will not fire; so the I/O throttling function has to be disabled here | |
1563 | * if it has been enabled. | |
1564 | */ | |
1565 | if (bs->io_limits_enabled) { | |
1566 | fprintf(stderr, "Disabling I/O throttling on '%s' due " | |
1567 | "to synchronous I/O.\n", bdrv_get_device_name(bs)); | |
1568 | bdrv_io_limits_disable(bs); | |
1569 | } | |
1570 | ||
1c9805a3 SH |
1571 | if (qemu_in_coroutine()) { |
1572 | /* Fast-path if already in coroutine context */ | |
1573 | bdrv_rw_co_entry(&rwco); | |
1574 | } else { | |
1575 | co = qemu_coroutine_create(bdrv_rw_co_entry); | |
1576 | qemu_coroutine_enter(co, &rwco); | |
1577 | while (rwco.ret == NOT_DONE) { | |
1578 | qemu_aio_wait(); | |
1579 | } | |
1580 | } | |
1581 | return rwco.ret; | |
1582 | } | |
b338082b | 1583 | |
1c9805a3 SH |
1584 | /* return < 0 if error. See bdrv_write() for the return codes */ |
1585 | int bdrv_read(BlockDriverState *bs, int64_t sector_num, | |
1586 | uint8_t *buf, int nb_sectors) | |
1587 | { | |
1588 | return bdrv_rw_co(bs, sector_num, buf, nb_sectors, false); | |
fc01f7e7 FB |
1589 | } |
1590 | ||
71df14fc PB |
1591 | #define BITS_PER_LONG (sizeof(unsigned long) * 8) |
1592 | ||
7cd1e32a | 1593 | static void set_dirty_bitmap(BlockDriverState *bs, int64_t sector_num, |
a55eb92c | 1594 | int nb_sectors, int dirty) |
7cd1e32a | 1595 | { |
1596 | int64_t start, end; | |
c6d22830 | 1597 | unsigned long val, idx, bit; |
a55eb92c | 1598 | |
6ea44308 | 1599 | start = sector_num / BDRV_SECTORS_PER_DIRTY_CHUNK; |
c6d22830 | 1600 | end = (sector_num + nb_sectors - 1) / BDRV_SECTORS_PER_DIRTY_CHUNK; |
a55eb92c JK |
1601 | |
1602 | for (; start <= end; start++) { | |
71df14fc PB |
1603 | idx = start / BITS_PER_LONG; |
1604 | bit = start % BITS_PER_LONG; | |
c6d22830 JK |
1605 | val = bs->dirty_bitmap[idx]; |
1606 | if (dirty) { | |
6d59fec1 | 1607 | if (!(val & (1UL << bit))) { |
aaa0eb75 | 1608 | bs->dirty_count++; |
6d59fec1 | 1609 | val |= 1UL << bit; |
aaa0eb75 | 1610 | } |
c6d22830 | 1611 | } else { |
6d59fec1 | 1612 | if (val & (1UL << bit)) { |
aaa0eb75 | 1613 | bs->dirty_count--; |
6d59fec1 | 1614 | val &= ~(1UL << bit); |
aaa0eb75 | 1615 | } |
c6d22830 JK |
1616 | } |
1617 | bs->dirty_bitmap[idx] = val; | |
7cd1e32a | 1618 | } |
1619 | } | |
1620 | ||
5fafdf24 | 1621 | /* Return < 0 if error. Important errors are: |
19cb3738 FB |
1622 | -EIO generic I/O error (may happen for all errors) |
1623 | -ENOMEDIUM No media inserted. | |
1624 | -EINVAL Invalid sector number or nb_sectors | |
1625 | -EACCES Trying to write a read-only device | |
1626 | */ | |
5fafdf24 | 1627 | int bdrv_write(BlockDriverState *bs, int64_t sector_num, |
fc01f7e7 FB |
1628 | const uint8_t *buf, int nb_sectors) |
1629 | { | |
1c9805a3 | 1630 | return bdrv_rw_co(bs, sector_num, (uint8_t *)buf, nb_sectors, true); |
83f64091 FB |
1631 | } |
1632 | ||
eda578e5 AL |
1633 | int bdrv_pread(BlockDriverState *bs, int64_t offset, |
1634 | void *buf, int count1) | |
83f64091 | 1635 | { |
6ea44308 | 1636 | uint8_t tmp_buf[BDRV_SECTOR_SIZE]; |
83f64091 FB |
1637 | int len, nb_sectors, count; |
1638 | int64_t sector_num; | |
9a8c4cce | 1639 | int ret; |
83f64091 FB |
1640 | |
1641 | count = count1; | |
1642 | /* first read to align to sector start */ | |
6ea44308 | 1643 | len = (BDRV_SECTOR_SIZE - offset) & (BDRV_SECTOR_SIZE - 1); |
83f64091 FB |
1644 | if (len > count) |
1645 | len = count; | |
6ea44308 | 1646 | sector_num = offset >> BDRV_SECTOR_BITS; |
83f64091 | 1647 | if (len > 0) { |
9a8c4cce KW |
1648 | if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0) |
1649 | return ret; | |
6ea44308 | 1650 | memcpy(buf, tmp_buf + (offset & (BDRV_SECTOR_SIZE - 1)), len); |
83f64091 FB |
1651 | count -= len; |
1652 | if (count == 0) | |
1653 | return count1; | |
1654 | sector_num++; | |
1655 | buf += len; | |
1656 | } | |
1657 | ||
1658 | /* read the sectors "in place" */ | |
6ea44308 | 1659 | nb_sectors = count >> BDRV_SECTOR_BITS; |
83f64091 | 1660 | if (nb_sectors > 0) { |
9a8c4cce KW |
1661 | if ((ret = bdrv_read(bs, sector_num, buf, nb_sectors)) < 0) |
1662 | return ret; | |
83f64091 | 1663 | sector_num += nb_sectors; |
6ea44308 | 1664 | len = nb_sectors << BDRV_SECTOR_BITS; |
83f64091 FB |
1665 | buf += len; |
1666 | count -= len; | |
1667 | } | |
1668 | ||
1669 | /* add data from the last sector */ | |
1670 | if (count > 0) { | |
9a8c4cce KW |
1671 | if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0) |
1672 | return ret; | |
83f64091 FB |
1673 | memcpy(buf, tmp_buf, count); |
1674 | } | |
1675 | return count1; | |
1676 | } | |
1677 | ||
eda578e5 AL |
1678 | int bdrv_pwrite(BlockDriverState *bs, int64_t offset, |
1679 | const void *buf, int count1) | |
83f64091 | 1680 | { |
6ea44308 | 1681 | uint8_t tmp_buf[BDRV_SECTOR_SIZE]; |
83f64091 FB |
1682 | int len, nb_sectors, count; |
1683 | int64_t sector_num; | |
9a8c4cce | 1684 | int ret; |
83f64091 FB |
1685 | |
1686 | count = count1; | |
1687 | /* first write to align to sector start */ | |
6ea44308 | 1688 | len = (BDRV_SECTOR_SIZE - offset) & (BDRV_SECTOR_SIZE - 1); |
83f64091 FB |
1689 | if (len > count) |
1690 | len = count; | |
6ea44308 | 1691 | sector_num = offset >> BDRV_SECTOR_BITS; |
83f64091 | 1692 | if (len > 0) { |
9a8c4cce KW |
1693 | if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0) |
1694 | return ret; | |
6ea44308 | 1695 | memcpy(tmp_buf + (offset & (BDRV_SECTOR_SIZE - 1)), buf, len); |
9a8c4cce KW |
1696 | if ((ret = bdrv_write(bs, sector_num, tmp_buf, 1)) < 0) |
1697 | return ret; | |
83f64091 FB |
1698 | count -= len; |
1699 | if (count == 0) | |
1700 | return count1; | |
1701 | sector_num++; | |
1702 | buf += len; | |
1703 | } | |
1704 | ||
1705 | /* write the sectors "in place" */ | |
6ea44308 | 1706 | nb_sectors = count >> BDRV_SECTOR_BITS; |
83f64091 | 1707 | if (nb_sectors > 0) { |
9a8c4cce KW |
1708 | if ((ret = bdrv_write(bs, sector_num, buf, nb_sectors)) < 0) |
1709 | return ret; | |
83f64091 | 1710 | sector_num += nb_sectors; |
6ea44308 | 1711 | len = nb_sectors << BDRV_SECTOR_BITS; |
83f64091 FB |
1712 | buf += len; |
1713 | count -= len; | |
1714 | } | |
1715 | ||
1716 | /* add data from the last sector */ | |
1717 | if (count > 0) { | |
9a8c4cce KW |
1718 | if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0) |
1719 | return ret; | |
83f64091 | 1720 | memcpy(tmp_buf, buf, count); |
9a8c4cce KW |
1721 | if ((ret = bdrv_write(bs, sector_num, tmp_buf, 1)) < 0) |
1722 | return ret; | |
83f64091 FB |
1723 | } |
1724 | return count1; | |
1725 | } | |
83f64091 | 1726 | |
f08145fe KW |
1727 | /* |
1728 | * Writes to the file and ensures that no writes are reordered across this | |
1729 | * request (acts as a barrier) | |
1730 | * | |
1731 | * Returns 0 on success, -errno in error cases. | |
1732 | */ | |
1733 | int bdrv_pwrite_sync(BlockDriverState *bs, int64_t offset, | |
1734 | const void *buf, int count) | |
1735 | { | |
1736 | int ret; | |
1737 | ||
1738 | ret = bdrv_pwrite(bs, offset, buf, count); | |
1739 | if (ret < 0) { | |
1740 | return ret; | |
1741 | } | |
1742 | ||
92196b2f SH |
1743 | /* No flush needed for cache modes that use O_DSYNC */ |
1744 | if ((bs->open_flags & BDRV_O_CACHE_WB) != 0) { | |
f08145fe KW |
1745 | bdrv_flush(bs); |
1746 | } | |
1747 | ||
1748 | return 0; | |
1749 | } | |
1750 | ||
470c0504 | 1751 | static int coroutine_fn bdrv_co_do_copy_on_readv(BlockDriverState *bs, |
ab185921 SH |
1752 | int64_t sector_num, int nb_sectors, QEMUIOVector *qiov) |
1753 | { | |
1754 | /* Perform I/O through a temporary buffer so that users who scribble over | |
1755 | * their read buffer while the operation is in progress do not end up | |
1756 | * modifying the image file. This is critical for zero-copy guest I/O | |
1757 | * where anything might happen inside guest memory. | |
1758 | */ | |
1759 | void *bounce_buffer; | |
1760 | ||
79c053bd | 1761 | BlockDriver *drv = bs->drv; |
ab185921 SH |
1762 | struct iovec iov; |
1763 | QEMUIOVector bounce_qiov; | |
1764 | int64_t cluster_sector_num; | |
1765 | int cluster_nb_sectors; | |
1766 | size_t skip_bytes; | |
1767 | int ret; | |
1768 | ||
1769 | /* Cover entire cluster so no additional backing file I/O is required when | |
1770 | * allocating cluster in the image file. | |
1771 | */ | |
1772 | round_to_clusters(bs, sector_num, nb_sectors, | |
1773 | &cluster_sector_num, &cluster_nb_sectors); | |
1774 | ||
470c0504 SH |
1775 | trace_bdrv_co_do_copy_on_readv(bs, sector_num, nb_sectors, |
1776 | cluster_sector_num, cluster_nb_sectors); | |
ab185921 SH |
1777 | |
1778 | iov.iov_len = cluster_nb_sectors * BDRV_SECTOR_SIZE; | |
1779 | iov.iov_base = bounce_buffer = qemu_blockalign(bs, iov.iov_len); | |
1780 | qemu_iovec_init_external(&bounce_qiov, &iov, 1); | |
1781 | ||
79c053bd SH |
1782 | ret = drv->bdrv_co_readv(bs, cluster_sector_num, cluster_nb_sectors, |
1783 | &bounce_qiov); | |
ab185921 SH |
1784 | if (ret < 0) { |
1785 | goto err; | |
1786 | } | |
1787 | ||
79c053bd SH |
1788 | if (drv->bdrv_co_write_zeroes && |
1789 | buffer_is_zero(bounce_buffer, iov.iov_len)) { | |
621f0589 KW |
1790 | ret = bdrv_co_do_write_zeroes(bs, cluster_sector_num, |
1791 | cluster_nb_sectors); | |
79c053bd SH |
1792 | } else { |
1793 | ret = drv->bdrv_co_writev(bs, cluster_sector_num, cluster_nb_sectors, | |
ab185921 | 1794 | &bounce_qiov); |
79c053bd SH |
1795 | } |
1796 | ||
ab185921 SH |
1797 | if (ret < 0) { |
1798 | /* It might be okay to ignore write errors for guest requests. If this | |
1799 | * is a deliberate copy-on-read then we don't want to ignore the error. | |
1800 | * Simply report it in all cases. | |
1801 | */ | |
1802 | goto err; | |
1803 | } | |
1804 | ||
1805 | skip_bytes = (sector_num - cluster_sector_num) * BDRV_SECTOR_SIZE; | |
1806 | qemu_iovec_from_buffer(qiov, bounce_buffer + skip_bytes, | |
1807 | nb_sectors * BDRV_SECTOR_SIZE); | |
1808 | ||
1809 | err: | |
1810 | qemu_vfree(bounce_buffer); | |
1811 | return ret; | |
1812 | } | |
1813 | ||
c5fbe571 SH |
1814 | /* |
1815 | * Handle a read request in coroutine context | |
1816 | */ | |
1817 | static int coroutine_fn bdrv_co_do_readv(BlockDriverState *bs, | |
470c0504 SH |
1818 | int64_t sector_num, int nb_sectors, QEMUIOVector *qiov, |
1819 | BdrvRequestFlags flags) | |
da1fa91d KW |
1820 | { |
1821 | BlockDriver *drv = bs->drv; | |
dbffbdcf SH |
1822 | BdrvTrackedRequest req; |
1823 | int ret; | |
da1fa91d | 1824 | |
da1fa91d KW |
1825 | if (!drv) { |
1826 | return -ENOMEDIUM; | |
1827 | } | |
1828 | if (bdrv_check_request(bs, sector_num, nb_sectors)) { | |
1829 | return -EIO; | |
1830 | } | |
1831 | ||
98f90dba ZYW |
1832 | /* throttling disk read I/O */ |
1833 | if (bs->io_limits_enabled) { | |
1834 | bdrv_io_limits_intercept(bs, false, nb_sectors); | |
1835 | } | |
1836 | ||
f4658285 | 1837 | if (bs->copy_on_read) { |
470c0504 SH |
1838 | flags |= BDRV_REQ_COPY_ON_READ; |
1839 | } | |
1840 | if (flags & BDRV_REQ_COPY_ON_READ) { | |
1841 | bs->copy_on_read_in_flight++; | |
1842 | } | |
1843 | ||
1844 | if (bs->copy_on_read_in_flight) { | |
f4658285 SH |
1845 | wait_for_overlapping_requests(bs, sector_num, nb_sectors); |
1846 | } | |
1847 | ||
dbffbdcf | 1848 | tracked_request_begin(&req, bs, sector_num, nb_sectors, false); |
ab185921 | 1849 | |
470c0504 | 1850 | if (flags & BDRV_REQ_COPY_ON_READ) { |
ab185921 SH |
1851 | int pnum; |
1852 | ||
1853 | ret = bdrv_co_is_allocated(bs, sector_num, nb_sectors, &pnum); | |
1854 | if (ret < 0) { | |
1855 | goto out; | |
1856 | } | |
1857 | ||
1858 | if (!ret || pnum != nb_sectors) { | |
470c0504 | 1859 | ret = bdrv_co_do_copy_on_readv(bs, sector_num, nb_sectors, qiov); |
ab185921 SH |
1860 | goto out; |
1861 | } | |
1862 | } | |
1863 | ||
dbffbdcf | 1864 | ret = drv->bdrv_co_readv(bs, sector_num, nb_sectors, qiov); |
ab185921 SH |
1865 | |
1866 | out: | |
dbffbdcf | 1867 | tracked_request_end(&req); |
470c0504 SH |
1868 | |
1869 | if (flags & BDRV_REQ_COPY_ON_READ) { | |
1870 | bs->copy_on_read_in_flight--; | |
1871 | } | |
1872 | ||
dbffbdcf | 1873 | return ret; |
da1fa91d KW |
1874 | } |
1875 | ||
c5fbe571 | 1876 | int coroutine_fn bdrv_co_readv(BlockDriverState *bs, int64_t sector_num, |
da1fa91d KW |
1877 | int nb_sectors, QEMUIOVector *qiov) |
1878 | { | |
c5fbe571 | 1879 | trace_bdrv_co_readv(bs, sector_num, nb_sectors); |
da1fa91d | 1880 | |
470c0504 SH |
1881 | return bdrv_co_do_readv(bs, sector_num, nb_sectors, qiov, 0); |
1882 | } | |
1883 | ||
1884 | int coroutine_fn bdrv_co_copy_on_readv(BlockDriverState *bs, | |
1885 | int64_t sector_num, int nb_sectors, QEMUIOVector *qiov) | |
1886 | { | |
1887 | trace_bdrv_co_copy_on_readv(bs, sector_num, nb_sectors); | |
1888 | ||
1889 | return bdrv_co_do_readv(bs, sector_num, nb_sectors, qiov, | |
1890 | BDRV_REQ_COPY_ON_READ); | |
c5fbe571 SH |
1891 | } |
1892 | ||
f08f2dda SH |
1893 | static int coroutine_fn bdrv_co_do_write_zeroes(BlockDriverState *bs, |
1894 | int64_t sector_num, int nb_sectors) | |
1895 | { | |
1896 | BlockDriver *drv = bs->drv; | |
1897 | QEMUIOVector qiov; | |
1898 | struct iovec iov; | |
1899 | int ret; | |
1900 | ||
621f0589 KW |
1901 | /* TODO Emulate only part of misaligned requests instead of letting block |
1902 | * drivers return -ENOTSUP and emulate everything */ | |
1903 | ||
f08f2dda SH |
1904 | /* First try the efficient write zeroes operation */ |
1905 | if (drv->bdrv_co_write_zeroes) { | |
621f0589 KW |
1906 | ret = drv->bdrv_co_write_zeroes(bs, sector_num, nb_sectors); |
1907 | if (ret != -ENOTSUP) { | |
1908 | return ret; | |
1909 | } | |
f08f2dda SH |
1910 | } |
1911 | ||
1912 | /* Fall back to bounce buffer if write zeroes is unsupported */ | |
1913 | iov.iov_len = nb_sectors * BDRV_SECTOR_SIZE; | |
1914 | iov.iov_base = qemu_blockalign(bs, iov.iov_len); | |
1915 | memset(iov.iov_base, 0, iov.iov_len); | |
1916 | qemu_iovec_init_external(&qiov, &iov, 1); | |
1917 | ||
1918 | ret = drv->bdrv_co_writev(bs, sector_num, nb_sectors, &qiov); | |
1919 | ||
1920 | qemu_vfree(iov.iov_base); | |
1921 | return ret; | |
1922 | } | |
1923 | ||
c5fbe571 SH |
1924 | /* |
1925 | * Handle a write request in coroutine context | |
1926 | */ | |
1927 | static int coroutine_fn bdrv_co_do_writev(BlockDriverState *bs, | |
f08f2dda SH |
1928 | int64_t sector_num, int nb_sectors, QEMUIOVector *qiov, |
1929 | BdrvRequestFlags flags) | |
c5fbe571 SH |
1930 | { |
1931 | BlockDriver *drv = bs->drv; | |
dbffbdcf | 1932 | BdrvTrackedRequest req; |
6b7cb247 | 1933 | int ret; |
da1fa91d KW |
1934 | |
1935 | if (!bs->drv) { | |
1936 | return -ENOMEDIUM; | |
1937 | } | |
1938 | if (bs->read_only) { | |
1939 | return -EACCES; | |
1940 | } | |
1941 | if (bdrv_check_request(bs, sector_num, nb_sectors)) { | |
1942 | return -EIO; | |
1943 | } | |
1944 | ||
98f90dba ZYW |
1945 | /* throttling disk write I/O */ |
1946 | if (bs->io_limits_enabled) { | |
1947 | bdrv_io_limits_intercept(bs, true, nb_sectors); | |
1948 | } | |
1949 | ||
470c0504 | 1950 | if (bs->copy_on_read_in_flight) { |
f4658285 SH |
1951 | wait_for_overlapping_requests(bs, sector_num, nb_sectors); |
1952 | } | |
1953 | ||
dbffbdcf SH |
1954 | tracked_request_begin(&req, bs, sector_num, nb_sectors, true); |
1955 | ||
f08f2dda SH |
1956 | if (flags & BDRV_REQ_ZERO_WRITE) { |
1957 | ret = bdrv_co_do_write_zeroes(bs, sector_num, nb_sectors); | |
1958 | } else { | |
1959 | ret = drv->bdrv_co_writev(bs, sector_num, nb_sectors, qiov); | |
1960 | } | |
6b7cb247 | 1961 | |
da1fa91d KW |
1962 | if (bs->dirty_bitmap) { |
1963 | set_dirty_bitmap(bs, sector_num, nb_sectors, 1); | |
1964 | } | |
1965 | ||
1966 | if (bs->wr_highest_sector < sector_num + nb_sectors - 1) { | |
1967 | bs->wr_highest_sector = sector_num + nb_sectors - 1; | |
1968 | } | |
1969 | ||
dbffbdcf SH |
1970 | tracked_request_end(&req); |
1971 | ||
6b7cb247 | 1972 | return ret; |
da1fa91d KW |
1973 | } |
1974 | ||
c5fbe571 SH |
1975 | int coroutine_fn bdrv_co_writev(BlockDriverState *bs, int64_t sector_num, |
1976 | int nb_sectors, QEMUIOVector *qiov) | |
1977 | { | |
1978 | trace_bdrv_co_writev(bs, sector_num, nb_sectors); | |
1979 | ||
f08f2dda SH |
1980 | return bdrv_co_do_writev(bs, sector_num, nb_sectors, qiov, 0); |
1981 | } | |
1982 | ||
1983 | int coroutine_fn bdrv_co_write_zeroes(BlockDriverState *bs, | |
1984 | int64_t sector_num, int nb_sectors) | |
1985 | { | |
1986 | trace_bdrv_co_write_zeroes(bs, sector_num, nb_sectors); | |
1987 | ||
1988 | return bdrv_co_do_writev(bs, sector_num, nb_sectors, NULL, | |
1989 | BDRV_REQ_ZERO_WRITE); | |
c5fbe571 SH |
1990 | } |
1991 | ||
83f64091 FB |
1992 | /** |
1993 | * Truncate file to 'offset' bytes (needed only for file protocols) | |
1994 | */ | |
1995 | int bdrv_truncate(BlockDriverState *bs, int64_t offset) | |
1996 | { | |
1997 | BlockDriver *drv = bs->drv; | |
51762288 | 1998 | int ret; |
83f64091 | 1999 | if (!drv) |
19cb3738 | 2000 | return -ENOMEDIUM; |
83f64091 FB |
2001 | if (!drv->bdrv_truncate) |
2002 | return -ENOTSUP; | |
59f2689d NS |
2003 | if (bs->read_only) |
2004 | return -EACCES; | |
8591675f MT |
2005 | if (bdrv_in_use(bs)) |
2006 | return -EBUSY; | |
51762288 SH |
2007 | ret = drv->bdrv_truncate(bs, offset); |
2008 | if (ret == 0) { | |
2009 | ret = refresh_total_sectors(bs, offset >> BDRV_SECTOR_BITS); | |
145feb17 | 2010 | bdrv_dev_resize_cb(bs); |
51762288 SH |
2011 | } |
2012 | return ret; | |
83f64091 FB |
2013 | } |
2014 | ||
4a1d5e1f FZ |
2015 | /** |
2016 | * Length of a allocated file in bytes. Sparse files are counted by actual | |
2017 | * allocated space. Return < 0 if error or unknown. | |
2018 | */ | |
2019 | int64_t bdrv_get_allocated_file_size(BlockDriverState *bs) | |
2020 | { | |
2021 | BlockDriver *drv = bs->drv; | |
2022 | if (!drv) { | |
2023 | return -ENOMEDIUM; | |
2024 | } | |
2025 | if (drv->bdrv_get_allocated_file_size) { | |
2026 | return drv->bdrv_get_allocated_file_size(bs); | |
2027 | } | |
2028 | if (bs->file) { | |
2029 | return bdrv_get_allocated_file_size(bs->file); | |
2030 | } | |
2031 | return -ENOTSUP; | |
2032 | } | |
2033 | ||
83f64091 FB |
2034 | /** |
2035 | * Length of a file in bytes. Return < 0 if error or unknown. | |
2036 | */ | |
2037 | int64_t bdrv_getlength(BlockDriverState *bs) | |
2038 | { | |
2039 | BlockDriver *drv = bs->drv; | |
2040 | if (!drv) | |
19cb3738 | 2041 | return -ENOMEDIUM; |
51762288 | 2042 | |
2c6942fa | 2043 | if (bs->growable || bdrv_dev_has_removable_media(bs)) { |
46a4e4e6 SH |
2044 | if (drv->bdrv_getlength) { |
2045 | return drv->bdrv_getlength(bs); | |
2046 | } | |
83f64091 | 2047 | } |
46a4e4e6 | 2048 | return bs->total_sectors * BDRV_SECTOR_SIZE; |
fc01f7e7 FB |
2049 | } |
2050 | ||
19cb3738 | 2051 | /* return 0 as number of sectors if no device present or error */ |
96b8f136 | 2052 | void bdrv_get_geometry(BlockDriverState *bs, uint64_t *nb_sectors_ptr) |
fc01f7e7 | 2053 | { |
19cb3738 FB |
2054 | int64_t length; |
2055 | length = bdrv_getlength(bs); | |
2056 | if (length < 0) | |
2057 | length = 0; | |
2058 | else | |
6ea44308 | 2059 | length = length >> BDRV_SECTOR_BITS; |
19cb3738 | 2060 | *nb_sectors_ptr = length; |
fc01f7e7 | 2061 | } |
cf98951b | 2062 | |
f3d54fc4 AL |
2063 | struct partition { |
2064 | uint8_t boot_ind; /* 0x80 - active */ | |
2065 | uint8_t head; /* starting head */ | |
2066 | uint8_t sector; /* starting sector */ | |
2067 | uint8_t cyl; /* starting cylinder */ | |
2068 | uint8_t sys_ind; /* What partition type */ | |
2069 | uint8_t end_head; /* end head */ | |
2070 | uint8_t end_sector; /* end sector */ | |
2071 | uint8_t end_cyl; /* end cylinder */ | |
2072 | uint32_t start_sect; /* starting sector counting from 0 */ | |
2073 | uint32_t nr_sects; /* nr of sectors in partition */ | |
541dc0d4 | 2074 | } QEMU_PACKED; |
f3d54fc4 AL |
2075 | |
2076 | /* try to guess the disk logical geometry from the MSDOS partition table. Return 0 if OK, -1 if could not guess */ | |
2077 | static int guess_disk_lchs(BlockDriverState *bs, | |
2078 | int *pcylinders, int *pheads, int *psectors) | |
2079 | { | |
eb5a3165 | 2080 | uint8_t buf[BDRV_SECTOR_SIZE]; |
f3d54fc4 AL |
2081 | int ret, i, heads, sectors, cylinders; |
2082 | struct partition *p; | |
2083 | uint32_t nr_sects; | |
a38131b6 | 2084 | uint64_t nb_sectors; |
498e386c | 2085 | bool enabled; |
f3d54fc4 AL |
2086 | |
2087 | bdrv_get_geometry(bs, &nb_sectors); | |
2088 | ||
498e386c ZYW |
2089 | /** |
2090 | * The function will be invoked during startup not only in sync I/O mode, | |
2091 | * but also in async I/O mode. So the I/O throttling function has to | |
2092 | * be disabled temporarily here, not permanently. | |
2093 | */ | |
2094 | enabled = bs->io_limits_enabled; | |
2095 | bs->io_limits_enabled = false; | |
f3d54fc4 | 2096 | ret = bdrv_read(bs, 0, buf, 1); |
498e386c | 2097 | bs->io_limits_enabled = enabled; |
f3d54fc4 AL |
2098 | if (ret < 0) |
2099 | return -1; | |
2100 | /* test msdos magic */ | |
2101 | if (buf[510] != 0x55 || buf[511] != 0xaa) | |
2102 | return -1; | |
2103 | for(i = 0; i < 4; i++) { | |
2104 | p = ((struct partition *)(buf + 0x1be)) + i; | |
2105 | nr_sects = le32_to_cpu(p->nr_sects); | |
2106 | if (nr_sects && p->end_head) { | |
2107 | /* We make the assumption that the partition terminates on | |
2108 | a cylinder boundary */ | |
2109 | heads = p->end_head + 1; | |
2110 | sectors = p->end_sector & 63; | |
2111 | if (sectors == 0) | |
2112 | continue; | |
2113 | cylinders = nb_sectors / (heads * sectors); | |
2114 | if (cylinders < 1 || cylinders > 16383) | |
2115 | continue; | |
2116 | *pheads = heads; | |
2117 | *psectors = sectors; | |
2118 | *pcylinders = cylinders; | |
2119 | #if 0 | |
2120 | printf("guessed geometry: LCHS=%d %d %d\n", | |
2121 | cylinders, heads, sectors); | |
2122 | #endif | |
2123 | return 0; | |
2124 | } | |
2125 | } | |
2126 | return -1; | |
2127 | } | |
2128 | ||
2129 | void bdrv_guess_geometry(BlockDriverState *bs, int *pcyls, int *pheads, int *psecs) | |
2130 | { | |
2131 | int translation, lba_detected = 0; | |
2132 | int cylinders, heads, secs; | |
a38131b6 | 2133 | uint64_t nb_sectors; |
f3d54fc4 AL |
2134 | |
2135 | /* if a geometry hint is available, use it */ | |
2136 | bdrv_get_geometry(bs, &nb_sectors); | |
2137 | bdrv_get_geometry_hint(bs, &cylinders, &heads, &secs); | |
2138 | translation = bdrv_get_translation_hint(bs); | |
2139 | if (cylinders != 0) { | |
2140 | *pcyls = cylinders; | |
2141 | *pheads = heads; | |
2142 | *psecs = secs; | |
2143 | } else { | |
2144 | if (guess_disk_lchs(bs, &cylinders, &heads, &secs) == 0) { | |
2145 | if (heads > 16) { | |
2146 | /* if heads > 16, it means that a BIOS LBA | |
2147 | translation was active, so the default | |
2148 | hardware geometry is OK */ | |
2149 | lba_detected = 1; | |
2150 | goto default_geometry; | |
2151 | } else { | |
2152 | *pcyls = cylinders; | |
2153 | *pheads = heads; | |
2154 | *psecs = secs; | |
2155 | /* disable any translation to be in sync with | |
2156 | the logical geometry */ | |
2157 | if (translation == BIOS_ATA_TRANSLATION_AUTO) { | |
2158 | bdrv_set_translation_hint(bs, | |
2159 | BIOS_ATA_TRANSLATION_NONE); | |
2160 | } | |
2161 | } | |
2162 | } else { | |
2163 | default_geometry: | |
2164 | /* if no geometry, use a standard physical disk geometry */ | |
2165 | cylinders = nb_sectors / (16 * 63); | |
2166 | ||
2167 | if (cylinders > 16383) | |
2168 | cylinders = 16383; | |
2169 | else if (cylinders < 2) | |
2170 | cylinders = 2; | |
2171 | *pcyls = cylinders; | |
2172 | *pheads = 16; | |
2173 | *psecs = 63; | |
2174 | if ((lba_detected == 1) && (translation == BIOS_ATA_TRANSLATION_AUTO)) { | |
2175 | if ((*pcyls * *pheads) <= 131072) { | |
2176 | bdrv_set_translation_hint(bs, | |
2177 | BIOS_ATA_TRANSLATION_LARGE); | |
2178 | } else { | |
2179 | bdrv_set_translation_hint(bs, | |
2180 | BIOS_ATA_TRANSLATION_LBA); | |
2181 | } | |
2182 | } | |
2183 | } | |
2184 | bdrv_set_geometry_hint(bs, *pcyls, *pheads, *psecs); | |
2185 | } | |
2186 | } | |
2187 | ||
5fafdf24 | 2188 | void bdrv_set_geometry_hint(BlockDriverState *bs, |
b338082b FB |
2189 | int cyls, int heads, int secs) |
2190 | { | |
2191 | bs->cyls = cyls; | |
2192 | bs->heads = heads; | |
2193 | bs->secs = secs; | |
2194 | } | |
2195 | ||
46d4767d FB |
2196 | void bdrv_set_translation_hint(BlockDriverState *bs, int translation) |
2197 | { | |
2198 | bs->translation = translation; | |
2199 | } | |
2200 | ||
5fafdf24 | 2201 | void bdrv_get_geometry_hint(BlockDriverState *bs, |
b338082b FB |
2202 | int *pcyls, int *pheads, int *psecs) |
2203 | { | |
2204 | *pcyls = bs->cyls; | |
2205 | *pheads = bs->heads; | |
2206 | *psecs = bs->secs; | |
2207 | } | |
2208 | ||
0563e191 ZYW |
2209 | /* throttling disk io limits */ |
2210 | void bdrv_set_io_limits(BlockDriverState *bs, | |
2211 | BlockIOLimit *io_limits) | |
2212 | { | |
2213 | bs->io_limits = *io_limits; | |
2214 | bs->io_limits_enabled = bdrv_io_limits_enabled(bs); | |
2215 | } | |
2216 | ||
5bbdbb46 BS |
2217 | /* Recognize floppy formats */ |
2218 | typedef struct FDFormat { | |
2219 | FDriveType drive; | |
2220 | uint8_t last_sect; | |
2221 | uint8_t max_track; | |
2222 | uint8_t max_head; | |
f8d3d128 | 2223 | FDriveRate rate; |
5bbdbb46 BS |
2224 | } FDFormat; |
2225 | ||
2226 | static const FDFormat fd_formats[] = { | |
2227 | /* First entry is default format */ | |
2228 | /* 1.44 MB 3"1/2 floppy disks */ | |
f8d3d128 HP |
2229 | { FDRIVE_DRV_144, 18, 80, 1, FDRIVE_RATE_500K, }, |
2230 | { FDRIVE_DRV_144, 20, 80, 1, FDRIVE_RATE_500K, }, | |
2231 | { FDRIVE_DRV_144, 21, 80, 1, FDRIVE_RATE_500K, }, | |
2232 | { FDRIVE_DRV_144, 21, 82, 1, FDRIVE_RATE_500K, }, | |
2233 | { FDRIVE_DRV_144, 21, 83, 1, FDRIVE_RATE_500K, }, | |
2234 | { FDRIVE_DRV_144, 22, 80, 1, FDRIVE_RATE_500K, }, | |
2235 | { FDRIVE_DRV_144, 23, 80, 1, FDRIVE_RATE_500K, }, | |
2236 | { FDRIVE_DRV_144, 24, 80, 1, FDRIVE_RATE_500K, }, | |
5bbdbb46 | 2237 | /* 2.88 MB 3"1/2 floppy disks */ |
f8d3d128 HP |
2238 | { FDRIVE_DRV_288, 36, 80, 1, FDRIVE_RATE_1M, }, |
2239 | { FDRIVE_DRV_288, 39, 80, 1, FDRIVE_RATE_1M, }, | |
2240 | { FDRIVE_DRV_288, 40, 80, 1, FDRIVE_RATE_1M, }, | |
2241 | { FDRIVE_DRV_288, 44, 80, 1, FDRIVE_RATE_1M, }, | |
2242 | { FDRIVE_DRV_288, 48, 80, 1, FDRIVE_RATE_1M, }, | |
5bbdbb46 | 2243 | /* 720 kB 3"1/2 floppy disks */ |
f8d3d128 HP |
2244 | { FDRIVE_DRV_144, 9, 80, 1, FDRIVE_RATE_250K, }, |
2245 | { FDRIVE_DRV_144, 10, 80, 1, FDRIVE_RATE_250K, }, | |
2246 | { FDRIVE_DRV_144, 10, 82, 1, FDRIVE_RATE_250K, }, | |
2247 | { FDRIVE_DRV_144, 10, 83, 1, FDRIVE_RATE_250K, }, | |
2248 | { FDRIVE_DRV_144, 13, 80, 1, FDRIVE_RATE_250K, }, | |
2249 | { FDRIVE_DRV_144, 14, 80, 1, FDRIVE_RATE_250K, }, | |
5bbdbb46 | 2250 | /* 1.2 MB 5"1/4 floppy disks */ |
f8d3d128 HP |
2251 | { FDRIVE_DRV_120, 15, 80, 1, FDRIVE_RATE_500K, }, |
2252 | { FDRIVE_DRV_120, 18, 80, 1, FDRIVE_RATE_500K, }, | |
2253 | { FDRIVE_DRV_120, 18, 82, 1, FDRIVE_RATE_500K, }, | |
2254 | { FDRIVE_DRV_120, 18, 83, 1, FDRIVE_RATE_500K, }, | |
2255 | { FDRIVE_DRV_120, 20, 80, 1, FDRIVE_RATE_500K, }, | |
5bbdbb46 | 2256 | /* 720 kB 5"1/4 floppy disks */ |
f8d3d128 HP |
2257 | { FDRIVE_DRV_120, 9, 80, 1, FDRIVE_RATE_250K, }, |
2258 | { FDRIVE_DRV_120, 11, 80, 1, FDRIVE_RATE_250K, }, | |
5bbdbb46 | 2259 | /* 360 kB 5"1/4 floppy disks */ |
f8d3d128 HP |
2260 | { FDRIVE_DRV_120, 9, 40, 1, FDRIVE_RATE_300K, }, |
2261 | { FDRIVE_DRV_120, 9, 40, 0, FDRIVE_RATE_300K, }, | |
2262 | { FDRIVE_DRV_120, 10, 41, 1, FDRIVE_RATE_300K, }, | |
2263 | { FDRIVE_DRV_120, 10, 42, 1, FDRIVE_RATE_300K, }, | |
5bbdbb46 | 2264 | /* 320 kB 5"1/4 floppy disks */ |
f8d3d128 HP |
2265 | { FDRIVE_DRV_120, 8, 40, 1, FDRIVE_RATE_250K, }, |
2266 | { FDRIVE_DRV_120, 8, 40, 0, FDRIVE_RATE_250K, }, | |
5bbdbb46 | 2267 | /* 360 kB must match 5"1/4 better than 3"1/2... */ |
f8d3d128 | 2268 | { FDRIVE_DRV_144, 9, 80, 0, FDRIVE_RATE_250K, }, |
5bbdbb46 | 2269 | /* end */ |
f8d3d128 | 2270 | { FDRIVE_DRV_NONE, -1, -1, 0, 0, }, |
5bbdbb46 BS |
2271 | }; |
2272 | ||
2273 | void bdrv_get_floppy_geometry_hint(BlockDriverState *bs, int *nb_heads, | |
2274 | int *max_track, int *last_sect, | |
f8d3d128 HP |
2275 | FDriveType drive_in, FDriveType *drive, |
2276 | FDriveRate *rate) | |
5bbdbb46 BS |
2277 | { |
2278 | const FDFormat *parse; | |
2279 | uint64_t nb_sectors, size; | |
2280 | int i, first_match, match; | |
2281 | ||
2282 | bdrv_get_geometry_hint(bs, nb_heads, max_track, last_sect); | |
2283 | if (*nb_heads != 0 && *max_track != 0 && *last_sect != 0) { | |
2284 | /* User defined disk */ | |
f8d3d128 | 2285 | *rate = FDRIVE_RATE_500K; |
5bbdbb46 BS |
2286 | } else { |
2287 | bdrv_get_geometry(bs, &nb_sectors); | |
2288 | match = -1; | |
2289 | first_match = -1; | |
2290 | for (i = 0; ; i++) { | |
2291 | parse = &fd_formats[i]; | |
2292 | if (parse->drive == FDRIVE_DRV_NONE) { | |
2293 | break; | |
2294 | } | |
2295 | if (drive_in == parse->drive || | |
2296 | drive_in == FDRIVE_DRV_NONE) { | |
2297 | size = (parse->max_head + 1) * parse->max_track * | |
2298 | parse->last_sect; | |
2299 | if (nb_sectors == size) { | |
2300 | match = i; | |
2301 | break; | |
2302 | } | |
2303 | if (first_match == -1) { | |
2304 | first_match = i; | |
2305 | } | |
2306 | } | |
2307 | } | |
2308 | if (match == -1) { | |
2309 | if (first_match == -1) { | |
2310 | match = 1; | |
2311 | } else { | |
2312 | match = first_match; | |
2313 | } | |
2314 | parse = &fd_formats[match]; | |
2315 | } | |
2316 | *nb_heads = parse->max_head + 1; | |
2317 | *max_track = parse->max_track; | |
2318 | *last_sect = parse->last_sect; | |
2319 | *drive = parse->drive; | |
f8d3d128 | 2320 | *rate = parse->rate; |
5bbdbb46 BS |
2321 | } |
2322 | } | |
2323 | ||
46d4767d FB |
2324 | int bdrv_get_translation_hint(BlockDriverState *bs) |
2325 | { | |
2326 | return bs->translation; | |
2327 | } | |
2328 | ||
abd7f68d MA |
2329 | void bdrv_set_on_error(BlockDriverState *bs, BlockErrorAction on_read_error, |
2330 | BlockErrorAction on_write_error) | |
2331 | { | |
2332 | bs->on_read_error = on_read_error; | |
2333 | bs->on_write_error = on_write_error; | |
2334 | } | |
2335 | ||
2336 | BlockErrorAction bdrv_get_on_error(BlockDriverState *bs, int is_read) | |
2337 | { | |
2338 | return is_read ? bs->on_read_error : bs->on_write_error; | |
2339 | } | |
2340 | ||
b338082b FB |
2341 | int bdrv_is_read_only(BlockDriverState *bs) |
2342 | { | |
2343 | return bs->read_only; | |
2344 | } | |
2345 | ||
985a03b0 TS |
2346 | int bdrv_is_sg(BlockDriverState *bs) |
2347 | { | |
2348 | return bs->sg; | |
2349 | } | |
2350 | ||
e900a7b7 CH |
2351 | int bdrv_enable_write_cache(BlockDriverState *bs) |
2352 | { | |
2353 | return bs->enable_write_cache; | |
2354 | } | |
2355 | ||
ea2384d3 FB |
2356 | int bdrv_is_encrypted(BlockDriverState *bs) |
2357 | { | |
2358 | if (bs->backing_hd && bs->backing_hd->encrypted) | |
2359 | return 1; | |
2360 | return bs->encrypted; | |
2361 | } | |
2362 | ||
c0f4ce77 AL |
2363 | int bdrv_key_required(BlockDriverState *bs) |
2364 | { | |
2365 | BlockDriverState *backing_hd = bs->backing_hd; | |
2366 | ||
2367 | if (backing_hd && backing_hd->encrypted && !backing_hd->valid_key) | |
2368 | return 1; | |
2369 | return (bs->encrypted && !bs->valid_key); | |
2370 | } | |
2371 | ||
ea2384d3 FB |
2372 | int bdrv_set_key(BlockDriverState *bs, const char *key) |
2373 | { | |
2374 | int ret; | |
2375 | if (bs->backing_hd && bs->backing_hd->encrypted) { | |
2376 | ret = bdrv_set_key(bs->backing_hd, key); | |
2377 | if (ret < 0) | |
2378 | return ret; | |
2379 | if (!bs->encrypted) | |
2380 | return 0; | |
2381 | } | |
fd04a2ae SH |
2382 | if (!bs->encrypted) { |
2383 | return -EINVAL; | |
2384 | } else if (!bs->drv || !bs->drv->bdrv_set_key) { | |
2385 | return -ENOMEDIUM; | |
2386 | } | |
c0f4ce77 | 2387 | ret = bs->drv->bdrv_set_key(bs, key); |
bb5fc20f AL |
2388 | if (ret < 0) { |
2389 | bs->valid_key = 0; | |
2390 | } else if (!bs->valid_key) { | |
2391 | bs->valid_key = 1; | |
2392 | /* call the change callback now, we skipped it on open */ | |
7d4b4ba5 | 2393 | bdrv_dev_change_media_cb(bs, true); |
bb5fc20f | 2394 | } |
c0f4ce77 | 2395 | return ret; |
ea2384d3 FB |
2396 | } |
2397 | ||
2398 | void bdrv_get_format(BlockDriverState *bs, char *buf, int buf_size) | |
2399 | { | |
19cb3738 | 2400 | if (!bs->drv) { |
ea2384d3 FB |
2401 | buf[0] = '\0'; |
2402 | } else { | |
2403 | pstrcpy(buf, buf_size, bs->drv->format_name); | |
2404 | } | |
2405 | } | |
2406 | ||
5fafdf24 | 2407 | void bdrv_iterate_format(void (*it)(void *opaque, const char *name), |
ea2384d3 FB |
2408 | void *opaque) |
2409 | { | |
2410 | BlockDriver *drv; | |
2411 | ||
8a22f02a | 2412 | QLIST_FOREACH(drv, &bdrv_drivers, list) { |
ea2384d3 FB |
2413 | it(opaque, drv->format_name); |
2414 | } | |
2415 | } | |
2416 | ||
b338082b FB |
2417 | BlockDriverState *bdrv_find(const char *name) |
2418 | { | |
2419 | BlockDriverState *bs; | |
2420 | ||
1b7bdbc1 SH |
2421 | QTAILQ_FOREACH(bs, &bdrv_states, list) { |
2422 | if (!strcmp(name, bs->device_name)) { | |
b338082b | 2423 | return bs; |
1b7bdbc1 | 2424 | } |
b338082b FB |
2425 | } |
2426 | return NULL; | |
2427 | } | |
2428 | ||
2f399b0a MA |
2429 | BlockDriverState *bdrv_next(BlockDriverState *bs) |
2430 | { | |
2431 | if (!bs) { | |
2432 | return QTAILQ_FIRST(&bdrv_states); | |
2433 | } | |
2434 | return QTAILQ_NEXT(bs, list); | |
2435 | } | |
2436 | ||
51de9760 | 2437 | void bdrv_iterate(void (*it)(void *opaque, BlockDriverState *bs), void *opaque) |
81d0912d FB |
2438 | { |
2439 | BlockDriverState *bs; | |
2440 | ||
1b7bdbc1 | 2441 | QTAILQ_FOREACH(bs, &bdrv_states, list) { |
51de9760 | 2442 | it(opaque, bs); |
81d0912d FB |
2443 | } |
2444 | } | |
2445 | ||
ea2384d3 FB |
2446 | const char *bdrv_get_device_name(BlockDriverState *bs) |
2447 | { | |
2448 | return bs->device_name; | |
2449 | } | |
2450 | ||
c6ca28d6 AL |
2451 | void bdrv_flush_all(void) |
2452 | { | |
2453 | BlockDriverState *bs; | |
2454 | ||
1b7bdbc1 | 2455 | QTAILQ_FOREACH(bs, &bdrv_states, list) { |
29cdb251 | 2456 | bdrv_flush(bs); |
1b7bdbc1 | 2457 | } |
c6ca28d6 AL |
2458 | } |
2459 | ||
f2feebbd KW |
2460 | int bdrv_has_zero_init(BlockDriverState *bs) |
2461 | { | |
2462 | assert(bs->drv); | |
2463 | ||
336c1c12 KW |
2464 | if (bs->drv->bdrv_has_zero_init) { |
2465 | return bs->drv->bdrv_has_zero_init(bs); | |
f2feebbd KW |
2466 | } |
2467 | ||
2468 | return 1; | |
2469 | } | |
2470 | ||
376ae3f1 SH |
2471 | typedef struct BdrvCoIsAllocatedData { |
2472 | BlockDriverState *bs; | |
2473 | int64_t sector_num; | |
2474 | int nb_sectors; | |
2475 | int *pnum; | |
2476 | int ret; | |
2477 | bool done; | |
2478 | } BdrvCoIsAllocatedData; | |
2479 | ||
f58c7b35 TS |
2480 | /* |
2481 | * Returns true iff the specified sector is present in the disk image. Drivers | |
2482 | * not implementing the functionality are assumed to not support backing files, | |
2483 | * hence all their sectors are reported as allocated. | |
2484 | * | |
bd9533e3 SH |
2485 | * If 'sector_num' is beyond the end of the disk image the return value is 0 |
2486 | * and 'pnum' is set to 0. | |
2487 | * | |
f58c7b35 TS |
2488 | * 'pnum' is set to the number of sectors (including and immediately following |
2489 | * the specified sector) that are known to be in the same | |
2490 | * allocated/unallocated state. | |
2491 | * | |
bd9533e3 SH |
2492 | * 'nb_sectors' is the max value 'pnum' should be set to. If nb_sectors goes |
2493 | * beyond the end of the disk image it will be clamped. | |
f58c7b35 | 2494 | */ |
060f51c9 SH |
2495 | int coroutine_fn bdrv_co_is_allocated(BlockDriverState *bs, int64_t sector_num, |
2496 | int nb_sectors, int *pnum) | |
f58c7b35 | 2497 | { |
bd9533e3 SH |
2498 | int64_t n; |
2499 | ||
2500 | if (sector_num >= bs->total_sectors) { | |
2501 | *pnum = 0; | |
2502 | return 0; | |
2503 | } | |
2504 | ||
2505 | n = bs->total_sectors - sector_num; | |
2506 | if (n < nb_sectors) { | |
2507 | nb_sectors = n; | |
2508 | } | |
2509 | ||
6aebab14 | 2510 | if (!bs->drv->bdrv_co_is_allocated) { |
bd9533e3 | 2511 | *pnum = nb_sectors; |
f58c7b35 TS |
2512 | return 1; |
2513 | } | |
6aebab14 | 2514 | |
060f51c9 SH |
2515 | return bs->drv->bdrv_co_is_allocated(bs, sector_num, nb_sectors, pnum); |
2516 | } | |
2517 | ||
2518 | /* Coroutine wrapper for bdrv_is_allocated() */ | |
2519 | static void coroutine_fn bdrv_is_allocated_co_entry(void *opaque) | |
2520 | { | |
2521 | BdrvCoIsAllocatedData *data = opaque; | |
2522 | BlockDriverState *bs = data->bs; | |
2523 | ||
2524 | data->ret = bdrv_co_is_allocated(bs, data->sector_num, data->nb_sectors, | |
2525 | data->pnum); | |
2526 | data->done = true; | |
2527 | } | |
2528 | ||
2529 | /* | |
2530 | * Synchronous wrapper around bdrv_co_is_allocated(). | |
2531 | * | |
2532 | * See bdrv_co_is_allocated() for details. | |
2533 | */ | |
2534 | int bdrv_is_allocated(BlockDriverState *bs, int64_t sector_num, int nb_sectors, | |
2535 | int *pnum) | |
2536 | { | |
6aebab14 SH |
2537 | Coroutine *co; |
2538 | BdrvCoIsAllocatedData data = { | |
2539 | .bs = bs, | |
2540 | .sector_num = sector_num, | |
2541 | .nb_sectors = nb_sectors, | |
2542 | .pnum = pnum, | |
2543 | .done = false, | |
2544 | }; | |
2545 | ||
2546 | co = qemu_coroutine_create(bdrv_is_allocated_co_entry); | |
2547 | qemu_coroutine_enter(co, &data); | |
2548 | while (!data.done) { | |
2549 | qemu_aio_wait(); | |
2550 | } | |
2551 | return data.ret; | |
f58c7b35 TS |
2552 | } |
2553 | ||
b2023818 | 2554 | BlockInfoList *qmp_query_block(Error **errp) |
b338082b | 2555 | { |
b2023818 | 2556 | BlockInfoList *head = NULL, *cur_item = NULL; |
b338082b FB |
2557 | BlockDriverState *bs; |
2558 | ||
1b7bdbc1 | 2559 | QTAILQ_FOREACH(bs, &bdrv_states, list) { |
b2023818 | 2560 | BlockInfoList *info = g_malloc0(sizeof(*info)); |
d15e5465 | 2561 | |
b2023818 LC |
2562 | info->value = g_malloc0(sizeof(*info->value)); |
2563 | info->value->device = g_strdup(bs->device_name); | |
2564 | info->value->type = g_strdup("unknown"); | |
2565 | info->value->locked = bdrv_dev_is_medium_locked(bs); | |
2566 | info->value->removable = bdrv_dev_has_removable_media(bs); | |
d15e5465 | 2567 | |
e4def80b | 2568 | if (bdrv_dev_has_removable_media(bs)) { |
b2023818 LC |
2569 | info->value->has_tray_open = true; |
2570 | info->value->tray_open = bdrv_dev_is_tray_open(bs); | |
e4def80b | 2571 | } |
f04ef601 LC |
2572 | |
2573 | if (bdrv_iostatus_is_enabled(bs)) { | |
b2023818 LC |
2574 | info->value->has_io_status = true; |
2575 | info->value->io_status = bs->iostatus; | |
f04ef601 LC |
2576 | } |
2577 | ||
19cb3738 | 2578 | if (bs->drv) { |
b2023818 LC |
2579 | info->value->has_inserted = true; |
2580 | info->value->inserted = g_malloc0(sizeof(*info->value->inserted)); | |
2581 | info->value->inserted->file = g_strdup(bs->filename); | |
2582 | info->value->inserted->ro = bs->read_only; | |
2583 | info->value->inserted->drv = g_strdup(bs->drv->format_name); | |
2584 | info->value->inserted->encrypted = bs->encrypted; | |
2585 | if (bs->backing_file[0]) { | |
2586 | info->value->inserted->has_backing_file = true; | |
2587 | info->value->inserted->backing_file = g_strdup(bs->backing_file); | |
376253ec | 2588 | } |
727f005e ZYW |
2589 | |
2590 | if (bs->io_limits_enabled) { | |
2591 | info->value->inserted->bps = | |
2592 | bs->io_limits.bps[BLOCK_IO_LIMIT_TOTAL]; | |
2593 | info->value->inserted->bps_rd = | |
2594 | bs->io_limits.bps[BLOCK_IO_LIMIT_READ]; | |
2595 | info->value->inserted->bps_wr = | |
2596 | bs->io_limits.bps[BLOCK_IO_LIMIT_WRITE]; | |
2597 | info->value->inserted->iops = | |
2598 | bs->io_limits.iops[BLOCK_IO_LIMIT_TOTAL]; | |
2599 | info->value->inserted->iops_rd = | |
2600 | bs->io_limits.iops[BLOCK_IO_LIMIT_READ]; | |
2601 | info->value->inserted->iops_wr = | |
2602 | bs->io_limits.iops[BLOCK_IO_LIMIT_WRITE]; | |
2603 | } | |
b2023818 | 2604 | } |
d15e5465 | 2605 | |
b2023818 LC |
2606 | /* XXX: waiting for the qapi to support GSList */ |
2607 | if (!cur_item) { | |
2608 | head = cur_item = info; | |
2609 | } else { | |
2610 | cur_item->next = info; | |
2611 | cur_item = info; | |
b338082b | 2612 | } |
b338082b | 2613 | } |
d15e5465 | 2614 | |
b2023818 | 2615 | return head; |
b338082b | 2616 | } |
a36e69dd | 2617 | |
f11f57e4 LC |
2618 | /* Consider exposing this as a full fledged QMP command */ |
2619 | static BlockStats *qmp_query_blockstat(const BlockDriverState *bs, Error **errp) | |
2620 | { | |
2621 | BlockStats *s; | |
2622 | ||
2623 | s = g_malloc0(sizeof(*s)); | |
2624 | ||
2625 | if (bs->device_name[0]) { | |
2626 | s->has_device = true; | |
2627 | s->device = g_strdup(bs->device_name); | |
294cc35f KW |
2628 | } |
2629 | ||
f11f57e4 LC |
2630 | s->stats = g_malloc0(sizeof(*s->stats)); |
2631 | s->stats->rd_bytes = bs->nr_bytes[BDRV_ACCT_READ]; | |
2632 | s->stats->wr_bytes = bs->nr_bytes[BDRV_ACCT_WRITE]; | |
2633 | s->stats->rd_operations = bs->nr_ops[BDRV_ACCT_READ]; | |
2634 | s->stats->wr_operations = bs->nr_ops[BDRV_ACCT_WRITE]; | |
2635 | s->stats->wr_highest_offset = bs->wr_highest_sector * BDRV_SECTOR_SIZE; | |
2636 | s->stats->flush_operations = bs->nr_ops[BDRV_ACCT_FLUSH]; | |
2637 | s->stats->wr_total_time_ns = bs->total_time_ns[BDRV_ACCT_WRITE]; | |
2638 | s->stats->rd_total_time_ns = bs->total_time_ns[BDRV_ACCT_READ]; | |
2639 | s->stats->flush_total_time_ns = bs->total_time_ns[BDRV_ACCT_FLUSH]; | |
2640 | ||
294cc35f | 2641 | if (bs->file) { |
f11f57e4 LC |
2642 | s->has_parent = true; |
2643 | s->parent = qmp_query_blockstat(bs->file, NULL); | |
294cc35f KW |
2644 | } |
2645 | ||
f11f57e4 | 2646 | return s; |
294cc35f KW |
2647 | } |
2648 | ||
f11f57e4 | 2649 | BlockStatsList *qmp_query_blockstats(Error **errp) |
218a536a | 2650 | { |
f11f57e4 | 2651 | BlockStatsList *head = NULL, *cur_item = NULL; |
a36e69dd TS |
2652 | BlockDriverState *bs; |
2653 | ||
1b7bdbc1 | 2654 | QTAILQ_FOREACH(bs, &bdrv_states, list) { |
f11f57e4 LC |
2655 | BlockStatsList *info = g_malloc0(sizeof(*info)); |
2656 | info->value = qmp_query_blockstat(bs, NULL); | |
2657 | ||
2658 | /* XXX: waiting for the qapi to support GSList */ | |
2659 | if (!cur_item) { | |
2660 | head = cur_item = info; | |
2661 | } else { | |
2662 | cur_item->next = info; | |
2663 | cur_item = info; | |
2664 | } | |
a36e69dd | 2665 | } |
218a536a | 2666 | |
f11f57e4 | 2667 | return head; |
a36e69dd | 2668 | } |
ea2384d3 | 2669 | |
045df330 AL |
2670 | const char *bdrv_get_encrypted_filename(BlockDriverState *bs) |
2671 | { | |
2672 | if (bs->backing_hd && bs->backing_hd->encrypted) | |
2673 | return bs->backing_file; | |
2674 | else if (bs->encrypted) | |
2675 | return bs->filename; | |
2676 | else | |
2677 | return NULL; | |
2678 | } | |
2679 | ||
5fafdf24 | 2680 | void bdrv_get_backing_filename(BlockDriverState *bs, |
83f64091 FB |
2681 | char *filename, int filename_size) |
2682 | { | |
3574c608 | 2683 | pstrcpy(filename, filename_size, bs->backing_file); |
83f64091 FB |
2684 | } |
2685 | ||
5fafdf24 | 2686 | int bdrv_write_compressed(BlockDriverState *bs, int64_t sector_num, |
faea38e7 FB |
2687 | const uint8_t *buf, int nb_sectors) |
2688 | { | |
2689 | BlockDriver *drv = bs->drv; | |
2690 | if (!drv) | |
19cb3738 | 2691 | return -ENOMEDIUM; |
faea38e7 FB |
2692 | if (!drv->bdrv_write_compressed) |
2693 | return -ENOTSUP; | |
fbb7b4e0 KW |
2694 | if (bdrv_check_request(bs, sector_num, nb_sectors)) |
2695 | return -EIO; | |
a55eb92c | 2696 | |
c6d22830 | 2697 | if (bs->dirty_bitmap) { |
7cd1e32a | 2698 | set_dirty_bitmap(bs, sector_num, nb_sectors, 1); |
2699 | } | |
a55eb92c | 2700 | |
faea38e7 FB |
2701 | return drv->bdrv_write_compressed(bs, sector_num, buf, nb_sectors); |
2702 | } | |
3b46e624 | 2703 | |
faea38e7 FB |
2704 | int bdrv_get_info(BlockDriverState *bs, BlockDriverInfo *bdi) |
2705 | { | |
2706 | BlockDriver *drv = bs->drv; | |
2707 | if (!drv) | |
19cb3738 | 2708 | return -ENOMEDIUM; |
faea38e7 FB |
2709 | if (!drv->bdrv_get_info) |
2710 | return -ENOTSUP; | |
2711 | memset(bdi, 0, sizeof(*bdi)); | |
2712 | return drv->bdrv_get_info(bs, bdi); | |
2713 | } | |
2714 | ||
45566e9c CH |
2715 | int bdrv_save_vmstate(BlockDriverState *bs, const uint8_t *buf, |
2716 | int64_t pos, int size) | |
178e08a5 AL |
2717 | { |
2718 | BlockDriver *drv = bs->drv; | |
2719 | if (!drv) | |
2720 | return -ENOMEDIUM; | |
7cdb1f6d MK |
2721 | if (drv->bdrv_save_vmstate) |
2722 | return drv->bdrv_save_vmstate(bs, buf, pos, size); | |
2723 | if (bs->file) | |
2724 | return bdrv_save_vmstate(bs->file, buf, pos, size); | |
2725 | return -ENOTSUP; | |
178e08a5 AL |
2726 | } |
2727 | ||
45566e9c CH |
2728 | int bdrv_load_vmstate(BlockDriverState *bs, uint8_t *buf, |
2729 | int64_t pos, int size) | |
178e08a5 AL |
2730 | { |
2731 | BlockDriver *drv = bs->drv; | |
2732 | if (!drv) | |
2733 | return -ENOMEDIUM; | |
7cdb1f6d MK |
2734 | if (drv->bdrv_load_vmstate) |
2735 | return drv->bdrv_load_vmstate(bs, buf, pos, size); | |
2736 | if (bs->file) | |
2737 | return bdrv_load_vmstate(bs->file, buf, pos, size); | |
2738 | return -ENOTSUP; | |
178e08a5 AL |
2739 | } |
2740 | ||
8b9b0cc2 KW |
2741 | void bdrv_debug_event(BlockDriverState *bs, BlkDebugEvent event) |
2742 | { | |
2743 | BlockDriver *drv = bs->drv; | |
2744 | ||
2745 | if (!drv || !drv->bdrv_debug_event) { | |
2746 | return; | |
2747 | } | |
2748 | ||
2749 | return drv->bdrv_debug_event(bs, event); | |
2750 | ||
2751 | } | |
2752 | ||
faea38e7 FB |
2753 | /**************************************************************/ |
2754 | /* handling of snapshots */ | |
2755 | ||
feeee5ac MDCF |
2756 | int bdrv_can_snapshot(BlockDriverState *bs) |
2757 | { | |
2758 | BlockDriver *drv = bs->drv; | |
07b70bfb | 2759 | if (!drv || !bdrv_is_inserted(bs) || bdrv_is_read_only(bs)) { |
feeee5ac MDCF |
2760 | return 0; |
2761 | } | |
2762 | ||
2763 | if (!drv->bdrv_snapshot_create) { | |
2764 | if (bs->file != NULL) { | |
2765 | return bdrv_can_snapshot(bs->file); | |
2766 | } | |
2767 | return 0; | |
2768 | } | |
2769 | ||
2770 | return 1; | |
2771 | } | |
2772 | ||
199630b6 BS |
2773 | int bdrv_is_snapshot(BlockDriverState *bs) |
2774 | { | |
2775 | return !!(bs->open_flags & BDRV_O_SNAPSHOT); | |
2776 | } | |
2777 | ||
f9092b10 MA |
2778 | BlockDriverState *bdrv_snapshots(void) |
2779 | { | |
2780 | BlockDriverState *bs; | |
2781 | ||
3ac906f7 | 2782 | if (bs_snapshots) { |
f9092b10 | 2783 | return bs_snapshots; |
3ac906f7 | 2784 | } |
f9092b10 MA |
2785 | |
2786 | bs = NULL; | |
2787 | while ((bs = bdrv_next(bs))) { | |
2788 | if (bdrv_can_snapshot(bs)) { | |
3ac906f7 MA |
2789 | bs_snapshots = bs; |
2790 | return bs; | |
f9092b10 MA |
2791 | } |
2792 | } | |
2793 | return NULL; | |
f9092b10 MA |
2794 | } |
2795 | ||
5fafdf24 | 2796 | int bdrv_snapshot_create(BlockDriverState *bs, |
faea38e7 FB |
2797 | QEMUSnapshotInfo *sn_info) |
2798 | { | |
2799 | BlockDriver *drv = bs->drv; | |
2800 | if (!drv) | |
19cb3738 | 2801 | return -ENOMEDIUM; |
7cdb1f6d MK |
2802 | if (drv->bdrv_snapshot_create) |
2803 | return drv->bdrv_snapshot_create(bs, sn_info); | |
2804 | if (bs->file) | |
2805 | return bdrv_snapshot_create(bs->file, sn_info); | |
2806 | return -ENOTSUP; | |
faea38e7 FB |
2807 | } |
2808 | ||
5fafdf24 | 2809 | int bdrv_snapshot_goto(BlockDriverState *bs, |
faea38e7 FB |
2810 | const char *snapshot_id) |
2811 | { | |
2812 | BlockDriver *drv = bs->drv; | |
7cdb1f6d MK |
2813 | int ret, open_ret; |
2814 | ||
faea38e7 | 2815 | if (!drv) |
19cb3738 | 2816 | return -ENOMEDIUM; |
7cdb1f6d MK |
2817 | if (drv->bdrv_snapshot_goto) |
2818 | return drv->bdrv_snapshot_goto(bs, snapshot_id); | |
2819 | ||
2820 | if (bs->file) { | |
2821 | drv->bdrv_close(bs); | |
2822 | ret = bdrv_snapshot_goto(bs->file, snapshot_id); | |
2823 | open_ret = drv->bdrv_open(bs, bs->open_flags); | |
2824 | if (open_ret < 0) { | |
2825 | bdrv_delete(bs->file); | |
2826 | bs->drv = NULL; | |
2827 | return open_ret; | |
2828 | } | |
2829 | return ret; | |
2830 | } | |
2831 | ||
2832 | return -ENOTSUP; | |
faea38e7 FB |
2833 | } |
2834 | ||
2835 | int bdrv_snapshot_delete(BlockDriverState *bs, const char *snapshot_id) | |
2836 | { | |
2837 | BlockDriver *drv = bs->drv; | |
2838 | if (!drv) | |
19cb3738 | 2839 | return -ENOMEDIUM; |
7cdb1f6d MK |
2840 | if (drv->bdrv_snapshot_delete) |
2841 | return drv->bdrv_snapshot_delete(bs, snapshot_id); | |
2842 | if (bs->file) | |
2843 | return bdrv_snapshot_delete(bs->file, snapshot_id); | |
2844 | return -ENOTSUP; | |
faea38e7 FB |
2845 | } |
2846 | ||
5fafdf24 | 2847 | int bdrv_snapshot_list(BlockDriverState *bs, |
faea38e7 FB |
2848 | QEMUSnapshotInfo **psn_info) |
2849 | { | |
2850 | BlockDriver *drv = bs->drv; | |
2851 | if (!drv) | |
19cb3738 | 2852 | return -ENOMEDIUM; |
7cdb1f6d MK |
2853 | if (drv->bdrv_snapshot_list) |
2854 | return drv->bdrv_snapshot_list(bs, psn_info); | |
2855 | if (bs->file) | |
2856 | return bdrv_snapshot_list(bs->file, psn_info); | |
2857 | return -ENOTSUP; | |
faea38e7 FB |
2858 | } |
2859 | ||
51ef6727 | 2860 | int bdrv_snapshot_load_tmp(BlockDriverState *bs, |
2861 | const char *snapshot_name) | |
2862 | { | |
2863 | BlockDriver *drv = bs->drv; | |
2864 | if (!drv) { | |
2865 | return -ENOMEDIUM; | |
2866 | } | |
2867 | if (!bs->read_only) { | |
2868 | return -EINVAL; | |
2869 | } | |
2870 | if (drv->bdrv_snapshot_load_tmp) { | |
2871 | return drv->bdrv_snapshot_load_tmp(bs, snapshot_name); | |
2872 | } | |
2873 | return -ENOTSUP; | |
2874 | } | |
2875 | ||
e8a6bb9c MT |
2876 | BlockDriverState *bdrv_find_backing_image(BlockDriverState *bs, |
2877 | const char *backing_file) | |
2878 | { | |
2879 | if (!bs->drv) { | |
2880 | return NULL; | |
2881 | } | |
2882 | ||
2883 | if (bs->backing_hd) { | |
2884 | if (strcmp(bs->backing_file, backing_file) == 0) { | |
2885 | return bs->backing_hd; | |
2886 | } else { | |
2887 | return bdrv_find_backing_image(bs->backing_hd, backing_file); | |
2888 | } | |
2889 | } | |
2890 | ||
2891 | return NULL; | |
2892 | } | |
2893 | ||
faea38e7 FB |
2894 | #define NB_SUFFIXES 4 |
2895 | ||
2896 | char *get_human_readable_size(char *buf, int buf_size, int64_t size) | |
2897 | { | |
2898 | static const char suffixes[NB_SUFFIXES] = "KMGT"; | |
2899 | int64_t base; | |
2900 | int i; | |
2901 | ||
2902 | if (size <= 999) { | |
2903 | snprintf(buf, buf_size, "%" PRId64, size); | |
2904 | } else { | |
2905 | base = 1024; | |
2906 | for(i = 0; i < NB_SUFFIXES; i++) { | |
2907 | if (size < (10 * base)) { | |
5fafdf24 | 2908 | snprintf(buf, buf_size, "%0.1f%c", |
faea38e7 FB |
2909 | (double)size / base, |
2910 | suffixes[i]); | |
2911 | break; | |
2912 | } else if (size < (1000 * base) || i == (NB_SUFFIXES - 1)) { | |
5fafdf24 | 2913 | snprintf(buf, buf_size, "%" PRId64 "%c", |
faea38e7 FB |
2914 | ((size + (base >> 1)) / base), |
2915 | suffixes[i]); | |
2916 | break; | |
2917 | } | |
2918 | base = base * 1024; | |
2919 | } | |
2920 | } | |
2921 | return buf; | |
2922 | } | |
2923 | ||
2924 | char *bdrv_snapshot_dump(char *buf, int buf_size, QEMUSnapshotInfo *sn) | |
2925 | { | |
2926 | char buf1[128], date_buf[128], clock_buf[128]; | |
3b9f94e1 FB |
2927 | #ifdef _WIN32 |
2928 | struct tm *ptm; | |
2929 | #else | |
faea38e7 | 2930 | struct tm tm; |
3b9f94e1 | 2931 | #endif |
faea38e7 FB |
2932 | time_t ti; |
2933 | int64_t secs; | |
2934 | ||
2935 | if (!sn) { | |
5fafdf24 TS |
2936 | snprintf(buf, buf_size, |
2937 | "%-10s%-20s%7s%20s%15s", | |
faea38e7 FB |
2938 | "ID", "TAG", "VM SIZE", "DATE", "VM CLOCK"); |
2939 | } else { | |
2940 | ti = sn->date_sec; | |
3b9f94e1 FB |
2941 | #ifdef _WIN32 |
2942 | ptm = localtime(&ti); | |
2943 | strftime(date_buf, sizeof(date_buf), | |
2944 | "%Y-%m-%d %H:%M:%S", ptm); | |
2945 | #else | |
faea38e7 FB |
2946 | localtime_r(&ti, &tm); |
2947 | strftime(date_buf, sizeof(date_buf), | |
2948 | "%Y-%m-%d %H:%M:%S", &tm); | |
3b9f94e1 | 2949 | #endif |
faea38e7 FB |
2950 | secs = sn->vm_clock_nsec / 1000000000; |
2951 | snprintf(clock_buf, sizeof(clock_buf), | |
2952 | "%02d:%02d:%02d.%03d", | |
2953 | (int)(secs / 3600), | |
2954 | (int)((secs / 60) % 60), | |
5fafdf24 | 2955 | (int)(secs % 60), |
faea38e7 FB |
2956 | (int)((sn->vm_clock_nsec / 1000000) % 1000)); |
2957 | snprintf(buf, buf_size, | |
5fafdf24 | 2958 | "%-10s%-20s%7s%20s%15s", |
faea38e7 FB |
2959 | sn->id_str, sn->name, |
2960 | get_human_readable_size(buf1, sizeof(buf1), sn->vm_state_size), | |
2961 | date_buf, | |
2962 | clock_buf); | |
2963 | } | |
2964 | return buf; | |
2965 | } | |
2966 | ||
ea2384d3 | 2967 | /**************************************************************/ |
83f64091 | 2968 | /* async I/Os */ |
ea2384d3 | 2969 | |
3b69e4b9 | 2970 | BlockDriverAIOCB *bdrv_aio_readv(BlockDriverState *bs, int64_t sector_num, |
f141eafe | 2971 | QEMUIOVector *qiov, int nb_sectors, |
3b69e4b9 | 2972 | BlockDriverCompletionFunc *cb, void *opaque) |
83f64091 | 2973 | { |
bbf0a440 SH |
2974 | trace_bdrv_aio_readv(bs, sector_num, nb_sectors, opaque); |
2975 | ||
b2a61371 | 2976 | return bdrv_co_aio_rw_vector(bs, sector_num, qiov, nb_sectors, |
8c5873d6 | 2977 | cb, opaque, false); |
ea2384d3 FB |
2978 | } |
2979 | ||
f141eafe AL |
2980 | BlockDriverAIOCB *bdrv_aio_writev(BlockDriverState *bs, int64_t sector_num, |
2981 | QEMUIOVector *qiov, int nb_sectors, | |
2982 | BlockDriverCompletionFunc *cb, void *opaque) | |
ea2384d3 | 2983 | { |
bbf0a440 SH |
2984 | trace_bdrv_aio_writev(bs, sector_num, nb_sectors, opaque); |
2985 | ||
1a6e115b | 2986 | return bdrv_co_aio_rw_vector(bs, sector_num, qiov, nb_sectors, |
8c5873d6 | 2987 | cb, opaque, true); |
83f64091 FB |
2988 | } |
2989 | ||
40b4f539 KW |
2990 | |
2991 | typedef struct MultiwriteCB { | |
2992 | int error; | |
2993 | int num_requests; | |
2994 | int num_callbacks; | |
2995 | struct { | |
2996 | BlockDriverCompletionFunc *cb; | |
2997 | void *opaque; | |
2998 | QEMUIOVector *free_qiov; | |
40b4f539 KW |
2999 | } callbacks[]; |
3000 | } MultiwriteCB; | |
3001 | ||
3002 | static void multiwrite_user_cb(MultiwriteCB *mcb) | |
3003 | { | |
3004 | int i; | |
3005 | ||
3006 | for (i = 0; i < mcb->num_callbacks; i++) { | |
3007 | mcb->callbacks[i].cb(mcb->callbacks[i].opaque, mcb->error); | |
1e1ea48d SH |
3008 | if (mcb->callbacks[i].free_qiov) { |
3009 | qemu_iovec_destroy(mcb->callbacks[i].free_qiov); | |
3010 | } | |
7267c094 | 3011 | g_free(mcb->callbacks[i].free_qiov); |
40b4f539 KW |
3012 | } |
3013 | } | |
3014 | ||
3015 | static void multiwrite_cb(void *opaque, int ret) | |
3016 | { | |
3017 | MultiwriteCB *mcb = opaque; | |
3018 | ||
6d519a5f SH |
3019 | trace_multiwrite_cb(mcb, ret); |
3020 | ||
cb6d3ca0 | 3021 | if (ret < 0 && !mcb->error) { |
40b4f539 | 3022 | mcb->error = ret; |
40b4f539 KW |
3023 | } |
3024 | ||
3025 | mcb->num_requests--; | |
3026 | if (mcb->num_requests == 0) { | |
de189a1b | 3027 | multiwrite_user_cb(mcb); |
7267c094 | 3028 | g_free(mcb); |
40b4f539 KW |
3029 | } |
3030 | } | |
3031 | ||
3032 | static int multiwrite_req_compare(const void *a, const void *b) | |
3033 | { | |
77be4366 CH |
3034 | const BlockRequest *req1 = a, *req2 = b; |
3035 | ||
3036 | /* | |
3037 | * Note that we can't simply subtract req2->sector from req1->sector | |
3038 | * here as that could overflow the return value. | |
3039 | */ | |
3040 | if (req1->sector > req2->sector) { | |
3041 | return 1; | |
3042 | } else if (req1->sector < req2->sector) { | |
3043 | return -1; | |
3044 | } else { | |
3045 | return 0; | |
3046 | } | |
40b4f539 KW |
3047 | } |
3048 | ||
3049 | /* | |
3050 | * Takes a bunch of requests and tries to merge them. Returns the number of | |
3051 | * requests that remain after merging. | |
3052 | */ | |
3053 | static int multiwrite_merge(BlockDriverState *bs, BlockRequest *reqs, | |
3054 | int num_reqs, MultiwriteCB *mcb) | |
3055 | { | |
3056 | int i, outidx; | |
3057 | ||
3058 | // Sort requests by start sector | |
3059 | qsort(reqs, num_reqs, sizeof(*reqs), &multiwrite_req_compare); | |
3060 | ||
3061 | // Check if adjacent requests touch the same clusters. If so, combine them, | |
3062 | // filling up gaps with zero sectors. | |
3063 | outidx = 0; | |
3064 | for (i = 1; i < num_reqs; i++) { | |
3065 | int merge = 0; | |
3066 | int64_t oldreq_last = reqs[outidx].sector + reqs[outidx].nb_sectors; | |
3067 | ||
b6a127a1 | 3068 | // Handle exactly sequential writes and overlapping writes. |
40b4f539 KW |
3069 | if (reqs[i].sector <= oldreq_last) { |
3070 | merge = 1; | |
3071 | } | |
3072 | ||
e2a305fb CH |
3073 | if (reqs[outidx].qiov->niov + reqs[i].qiov->niov + 1 > IOV_MAX) { |
3074 | merge = 0; | |
3075 | } | |
3076 | ||
40b4f539 KW |
3077 | if (merge) { |
3078 | size_t size; | |
7267c094 | 3079 | QEMUIOVector *qiov = g_malloc0(sizeof(*qiov)); |
40b4f539 KW |
3080 | qemu_iovec_init(qiov, |
3081 | reqs[outidx].qiov->niov + reqs[i].qiov->niov + 1); | |
3082 | ||
3083 | // Add the first request to the merged one. If the requests are | |
3084 | // overlapping, drop the last sectors of the first request. | |
3085 | size = (reqs[i].sector - reqs[outidx].sector) << 9; | |
3086 | qemu_iovec_concat(qiov, reqs[outidx].qiov, size); | |
3087 | ||
b6a127a1 PB |
3088 | // We should need to add any zeros between the two requests |
3089 | assert (reqs[i].sector <= oldreq_last); | |
40b4f539 KW |
3090 | |
3091 | // Add the second request | |
3092 | qemu_iovec_concat(qiov, reqs[i].qiov, reqs[i].qiov->size); | |
3093 | ||
cbf1dff2 | 3094 | reqs[outidx].nb_sectors = qiov->size >> 9; |
40b4f539 KW |
3095 | reqs[outidx].qiov = qiov; |
3096 | ||
3097 | mcb->callbacks[i].free_qiov = reqs[outidx].qiov; | |
3098 | } else { | |
3099 | outidx++; | |
3100 | reqs[outidx].sector = reqs[i].sector; | |
3101 | reqs[outidx].nb_sectors = reqs[i].nb_sectors; | |
3102 | reqs[outidx].qiov = reqs[i].qiov; | |
3103 | } | |
3104 | } | |
3105 | ||
3106 | return outidx + 1; | |
3107 | } | |
3108 | ||
3109 | /* | |
3110 | * Submit multiple AIO write requests at once. | |
3111 | * | |
3112 | * On success, the function returns 0 and all requests in the reqs array have | |
3113 | * been submitted. In error case this function returns -1, and any of the | |
3114 | * requests may or may not be submitted yet. In particular, this means that the | |
3115 | * callback will be called for some of the requests, for others it won't. The | |
3116 | * caller must check the error field of the BlockRequest to wait for the right | |
3117 | * callbacks (if error != 0, no callback will be called). | |
3118 | * | |
3119 | * The implementation may modify the contents of the reqs array, e.g. to merge | |
3120 | * requests. However, the fields opaque and error are left unmodified as they | |
3121 | * are used to signal failure for a single request to the caller. | |
3122 | */ | |
3123 | int bdrv_aio_multiwrite(BlockDriverState *bs, BlockRequest *reqs, int num_reqs) | |
3124 | { | |
40b4f539 KW |
3125 | MultiwriteCB *mcb; |
3126 | int i; | |
3127 | ||
301db7c2 RH |
3128 | /* don't submit writes if we don't have a medium */ |
3129 | if (bs->drv == NULL) { | |
3130 | for (i = 0; i < num_reqs; i++) { | |
3131 | reqs[i].error = -ENOMEDIUM; | |
3132 | } | |
3133 | return -1; | |
3134 | } | |
3135 | ||
40b4f539 KW |
3136 | if (num_reqs == 0) { |
3137 | return 0; | |
3138 | } | |
3139 | ||
3140 | // Create MultiwriteCB structure | |
7267c094 | 3141 | mcb = g_malloc0(sizeof(*mcb) + num_reqs * sizeof(*mcb->callbacks)); |
40b4f539 KW |
3142 | mcb->num_requests = 0; |
3143 | mcb->num_callbacks = num_reqs; | |
3144 | ||
3145 | for (i = 0; i < num_reqs; i++) { | |
3146 | mcb->callbacks[i].cb = reqs[i].cb; | |
3147 | mcb->callbacks[i].opaque = reqs[i].opaque; | |
3148 | } | |
3149 | ||
3150 | // Check for mergable requests | |
3151 | num_reqs = multiwrite_merge(bs, reqs, num_reqs, mcb); | |
3152 | ||
6d519a5f SH |
3153 | trace_bdrv_aio_multiwrite(mcb, mcb->num_callbacks, num_reqs); |
3154 | ||
df9309fb PB |
3155 | /* Run the aio requests. */ |
3156 | mcb->num_requests = num_reqs; | |
40b4f539 | 3157 | for (i = 0; i < num_reqs; i++) { |
ad54ae80 | 3158 | bdrv_aio_writev(bs, reqs[i].sector, reqs[i].qiov, |
40b4f539 | 3159 | reqs[i].nb_sectors, multiwrite_cb, mcb); |
40b4f539 KW |
3160 | } |
3161 | ||
3162 | return 0; | |
40b4f539 KW |
3163 | } |
3164 | ||
83f64091 | 3165 | void bdrv_aio_cancel(BlockDriverAIOCB *acb) |
83f64091 | 3166 | { |
6bbff9a0 | 3167 | acb->pool->cancel(acb); |
83f64091 FB |
3168 | } |
3169 | ||
98f90dba ZYW |
3170 | /* block I/O throttling */ |
3171 | static bool bdrv_exceed_bps_limits(BlockDriverState *bs, int nb_sectors, | |
3172 | bool is_write, double elapsed_time, uint64_t *wait) | |
3173 | { | |
3174 | uint64_t bps_limit = 0; | |
3175 | double bytes_limit, bytes_base, bytes_res; | |
3176 | double slice_time, wait_time; | |
3177 | ||
3178 | if (bs->io_limits.bps[BLOCK_IO_LIMIT_TOTAL]) { | |
3179 | bps_limit = bs->io_limits.bps[BLOCK_IO_LIMIT_TOTAL]; | |
3180 | } else if (bs->io_limits.bps[is_write]) { | |
3181 | bps_limit = bs->io_limits.bps[is_write]; | |
3182 | } else { | |
3183 | if (wait) { | |
3184 | *wait = 0; | |
3185 | } | |
3186 | ||
3187 | return false; | |
3188 | } | |
3189 | ||
3190 | slice_time = bs->slice_end - bs->slice_start; | |
3191 | slice_time /= (NANOSECONDS_PER_SECOND); | |
3192 | bytes_limit = bps_limit * slice_time; | |
3193 | bytes_base = bs->nr_bytes[is_write] - bs->io_base.bytes[is_write]; | |
3194 | if (bs->io_limits.bps[BLOCK_IO_LIMIT_TOTAL]) { | |
3195 | bytes_base += bs->nr_bytes[!is_write] - bs->io_base.bytes[!is_write]; | |
3196 | } | |
3197 | ||
3198 | /* bytes_base: the bytes of data which have been read/written; and | |
3199 | * it is obtained from the history statistic info. | |
3200 | * bytes_res: the remaining bytes of data which need to be read/written. | |
3201 | * (bytes_base + bytes_res) / bps_limit: used to calcuate | |
3202 | * the total time for completing reading/writting all data. | |
3203 | */ | |
3204 | bytes_res = (unsigned) nb_sectors * BDRV_SECTOR_SIZE; | |
3205 | ||
3206 | if (bytes_base + bytes_res <= bytes_limit) { | |
3207 | if (wait) { | |
3208 | *wait = 0; | |
3209 | } | |
3210 | ||
3211 | return false; | |
3212 | } | |
3213 | ||
3214 | /* Calc approx time to dispatch */ | |
3215 | wait_time = (bytes_base + bytes_res) / bps_limit - elapsed_time; | |
3216 | ||
3217 | /* When the I/O rate at runtime exceeds the limits, | |
3218 | * bs->slice_end need to be extended in order that the current statistic | |
3219 | * info can be kept until the timer fire, so it is increased and tuned | |
3220 | * based on the result of experiment. | |
3221 | */ | |
3222 | bs->slice_time = wait_time * BLOCK_IO_SLICE_TIME * 10; | |
3223 | bs->slice_end += bs->slice_time - 3 * BLOCK_IO_SLICE_TIME; | |
3224 | if (wait) { | |
3225 | *wait = wait_time * BLOCK_IO_SLICE_TIME * 10; | |
3226 | } | |
3227 | ||
3228 | return true; | |
3229 | } | |
3230 | ||
3231 | static bool bdrv_exceed_iops_limits(BlockDriverState *bs, bool is_write, | |
3232 | double elapsed_time, uint64_t *wait) | |
3233 | { | |
3234 | uint64_t iops_limit = 0; | |
3235 | double ios_limit, ios_base; | |
3236 | double slice_time, wait_time; | |
3237 | ||
3238 | if (bs->io_limits.iops[BLOCK_IO_LIMIT_TOTAL]) { | |
3239 | iops_limit = bs->io_limits.iops[BLOCK_IO_LIMIT_TOTAL]; | |
3240 | } else if (bs->io_limits.iops[is_write]) { | |
3241 | iops_limit = bs->io_limits.iops[is_write]; | |
3242 | } else { | |
3243 | if (wait) { | |
3244 | *wait = 0; | |
3245 | } | |
3246 | ||
3247 | return false; | |
3248 | } | |
3249 | ||
3250 | slice_time = bs->slice_end - bs->slice_start; | |
3251 | slice_time /= (NANOSECONDS_PER_SECOND); | |
3252 | ios_limit = iops_limit * slice_time; | |
3253 | ios_base = bs->nr_ops[is_write] - bs->io_base.ios[is_write]; | |
3254 | if (bs->io_limits.iops[BLOCK_IO_LIMIT_TOTAL]) { | |
3255 | ios_base += bs->nr_ops[!is_write] - bs->io_base.ios[!is_write]; | |
3256 | } | |
3257 | ||
3258 | if (ios_base + 1 <= ios_limit) { | |
3259 | if (wait) { | |
3260 | *wait = 0; | |
3261 | } | |
3262 | ||
3263 | return false; | |
3264 | } | |
3265 | ||
3266 | /* Calc approx time to dispatch */ | |
3267 | wait_time = (ios_base + 1) / iops_limit; | |
3268 | if (wait_time > elapsed_time) { | |
3269 | wait_time = wait_time - elapsed_time; | |
3270 | } else { | |
3271 | wait_time = 0; | |
3272 | } | |
3273 | ||
3274 | bs->slice_time = wait_time * BLOCK_IO_SLICE_TIME * 10; | |
3275 | bs->slice_end += bs->slice_time - 3 * BLOCK_IO_SLICE_TIME; | |
3276 | if (wait) { | |
3277 | *wait = wait_time * BLOCK_IO_SLICE_TIME * 10; | |
3278 | } | |
3279 | ||
3280 | return true; | |
3281 | } | |
3282 | ||
3283 | static bool bdrv_exceed_io_limits(BlockDriverState *bs, int nb_sectors, | |
3284 | bool is_write, int64_t *wait) | |
3285 | { | |
3286 | int64_t now, max_wait; | |
3287 | uint64_t bps_wait = 0, iops_wait = 0; | |
3288 | double elapsed_time; | |
3289 | int bps_ret, iops_ret; | |
3290 | ||
3291 | now = qemu_get_clock_ns(vm_clock); | |
3292 | if ((bs->slice_start < now) | |
3293 | && (bs->slice_end > now)) { | |
3294 | bs->slice_end = now + bs->slice_time; | |
3295 | } else { | |
3296 | bs->slice_time = 5 * BLOCK_IO_SLICE_TIME; | |
3297 | bs->slice_start = now; | |
3298 | bs->slice_end = now + bs->slice_time; | |
3299 | ||
3300 | bs->io_base.bytes[is_write] = bs->nr_bytes[is_write]; | |
3301 | bs->io_base.bytes[!is_write] = bs->nr_bytes[!is_write]; | |
3302 | ||
3303 | bs->io_base.ios[is_write] = bs->nr_ops[is_write]; | |
3304 | bs->io_base.ios[!is_write] = bs->nr_ops[!is_write]; | |
3305 | } | |
3306 | ||
3307 | elapsed_time = now - bs->slice_start; | |
3308 | elapsed_time /= (NANOSECONDS_PER_SECOND); | |
3309 | ||
3310 | bps_ret = bdrv_exceed_bps_limits(bs, nb_sectors, | |
3311 | is_write, elapsed_time, &bps_wait); | |
3312 | iops_ret = bdrv_exceed_iops_limits(bs, is_write, | |
3313 | elapsed_time, &iops_wait); | |
3314 | if (bps_ret || iops_ret) { | |
3315 | max_wait = bps_wait > iops_wait ? bps_wait : iops_wait; | |
3316 | if (wait) { | |
3317 | *wait = max_wait; | |
3318 | } | |
3319 | ||
3320 | now = qemu_get_clock_ns(vm_clock); | |
3321 | if (bs->slice_end < now + max_wait) { | |
3322 | bs->slice_end = now + max_wait; | |
3323 | } | |
3324 | ||
3325 | return true; | |
3326 | } | |
3327 | ||
3328 | if (wait) { | |
3329 | *wait = 0; | |
3330 | } | |
3331 | ||
3332 | return false; | |
3333 | } | |
ce1a14dc | 3334 | |
83f64091 FB |
3335 | /**************************************************************/ |
3336 | /* async block device emulation */ | |
3337 | ||
c16b5a2c CH |
3338 | typedef struct BlockDriverAIOCBSync { |
3339 | BlockDriverAIOCB common; | |
3340 | QEMUBH *bh; | |
3341 | int ret; | |
3342 | /* vector translation state */ | |
3343 | QEMUIOVector *qiov; | |
3344 | uint8_t *bounce; | |
3345 | int is_write; | |
3346 | } BlockDriverAIOCBSync; | |
3347 | ||
3348 | static void bdrv_aio_cancel_em(BlockDriverAIOCB *blockacb) | |
3349 | { | |
b666d239 KW |
3350 | BlockDriverAIOCBSync *acb = |
3351 | container_of(blockacb, BlockDriverAIOCBSync, common); | |
6a7ad299 | 3352 | qemu_bh_delete(acb->bh); |
36afc451 | 3353 | acb->bh = NULL; |
c16b5a2c CH |
3354 | qemu_aio_release(acb); |
3355 | } | |
3356 | ||
3357 | static AIOPool bdrv_em_aio_pool = { | |
3358 | .aiocb_size = sizeof(BlockDriverAIOCBSync), | |
3359 | .cancel = bdrv_aio_cancel_em, | |
3360 | }; | |
3361 | ||
ce1a14dc | 3362 | static void bdrv_aio_bh_cb(void *opaque) |
83f64091 | 3363 | { |
ce1a14dc | 3364 | BlockDriverAIOCBSync *acb = opaque; |
f141eafe | 3365 | |
f141eafe AL |
3366 | if (!acb->is_write) |
3367 | qemu_iovec_from_buffer(acb->qiov, acb->bounce, acb->qiov->size); | |
ceb42de8 | 3368 | qemu_vfree(acb->bounce); |
ce1a14dc | 3369 | acb->common.cb(acb->common.opaque, acb->ret); |
6a7ad299 | 3370 | qemu_bh_delete(acb->bh); |
36afc451 | 3371 | acb->bh = NULL; |
ce1a14dc | 3372 | qemu_aio_release(acb); |
83f64091 | 3373 | } |
beac80cd | 3374 | |
f141eafe AL |
3375 | static BlockDriverAIOCB *bdrv_aio_rw_vector(BlockDriverState *bs, |
3376 | int64_t sector_num, | |
3377 | QEMUIOVector *qiov, | |
3378 | int nb_sectors, | |
3379 | BlockDriverCompletionFunc *cb, | |
3380 | void *opaque, | |
3381 | int is_write) | |
3382 | ||
83f64091 | 3383 | { |
ce1a14dc | 3384 | BlockDriverAIOCBSync *acb; |
ce1a14dc | 3385 | |
c16b5a2c | 3386 | acb = qemu_aio_get(&bdrv_em_aio_pool, bs, cb, opaque); |
f141eafe AL |
3387 | acb->is_write = is_write; |
3388 | acb->qiov = qiov; | |
e268ca52 | 3389 | acb->bounce = qemu_blockalign(bs, qiov->size); |
3f3aace8 | 3390 | acb->bh = qemu_bh_new(bdrv_aio_bh_cb, acb); |
f141eafe AL |
3391 | |
3392 | if (is_write) { | |
3393 | qemu_iovec_to_buffer(acb->qiov, acb->bounce); | |
1ed20acf | 3394 | acb->ret = bs->drv->bdrv_write(bs, sector_num, acb->bounce, nb_sectors); |
f141eafe | 3395 | } else { |
1ed20acf | 3396 | acb->ret = bs->drv->bdrv_read(bs, sector_num, acb->bounce, nb_sectors); |
f141eafe AL |
3397 | } |
3398 | ||
ce1a14dc | 3399 | qemu_bh_schedule(acb->bh); |
f141eafe | 3400 | |
ce1a14dc | 3401 | return &acb->common; |
beac80cd FB |
3402 | } |
3403 | ||
f141eafe AL |
3404 | static BlockDriverAIOCB *bdrv_aio_readv_em(BlockDriverState *bs, |
3405 | int64_t sector_num, QEMUIOVector *qiov, int nb_sectors, | |
ce1a14dc | 3406 | BlockDriverCompletionFunc *cb, void *opaque) |
beac80cd | 3407 | { |
f141eafe AL |
3408 | return bdrv_aio_rw_vector(bs, sector_num, qiov, nb_sectors, cb, opaque, 0); |
3409 | } | |
83f64091 | 3410 | |
f141eafe AL |
3411 | static BlockDriverAIOCB *bdrv_aio_writev_em(BlockDriverState *bs, |
3412 | int64_t sector_num, QEMUIOVector *qiov, int nb_sectors, | |
3413 | BlockDriverCompletionFunc *cb, void *opaque) | |
3414 | { | |
3415 | return bdrv_aio_rw_vector(bs, sector_num, qiov, nb_sectors, cb, opaque, 1); | |
beac80cd | 3416 | } |
beac80cd | 3417 | |
68485420 KW |
3418 | |
3419 | typedef struct BlockDriverAIOCBCoroutine { | |
3420 | BlockDriverAIOCB common; | |
3421 | BlockRequest req; | |
3422 | bool is_write; | |
3423 | QEMUBH* bh; | |
3424 | } BlockDriverAIOCBCoroutine; | |
3425 | ||
3426 | static void bdrv_aio_co_cancel_em(BlockDriverAIOCB *blockacb) | |
3427 | { | |
3428 | qemu_aio_flush(); | |
3429 | } | |
3430 | ||
3431 | static AIOPool bdrv_em_co_aio_pool = { | |
3432 | .aiocb_size = sizeof(BlockDriverAIOCBCoroutine), | |
3433 | .cancel = bdrv_aio_co_cancel_em, | |
3434 | }; | |
3435 | ||
35246a68 | 3436 | static void bdrv_co_em_bh(void *opaque) |
68485420 KW |
3437 | { |
3438 | BlockDriverAIOCBCoroutine *acb = opaque; | |
3439 | ||
3440 | acb->common.cb(acb->common.opaque, acb->req.error); | |
3441 | qemu_bh_delete(acb->bh); | |
3442 | qemu_aio_release(acb); | |
3443 | } | |
3444 | ||
b2a61371 SH |
3445 | /* Invoke bdrv_co_do_readv/bdrv_co_do_writev */ |
3446 | static void coroutine_fn bdrv_co_do_rw(void *opaque) | |
3447 | { | |
3448 | BlockDriverAIOCBCoroutine *acb = opaque; | |
3449 | BlockDriverState *bs = acb->common.bs; | |
3450 | ||
3451 | if (!acb->is_write) { | |
3452 | acb->req.error = bdrv_co_do_readv(bs, acb->req.sector, | |
470c0504 | 3453 | acb->req.nb_sectors, acb->req.qiov, 0); |
b2a61371 SH |
3454 | } else { |
3455 | acb->req.error = bdrv_co_do_writev(bs, acb->req.sector, | |
f08f2dda | 3456 | acb->req.nb_sectors, acb->req.qiov, 0); |
b2a61371 SH |
3457 | } |
3458 | ||
35246a68 | 3459 | acb->bh = qemu_bh_new(bdrv_co_em_bh, acb); |
b2a61371 SH |
3460 | qemu_bh_schedule(acb->bh); |
3461 | } | |
3462 | ||
68485420 KW |
3463 | static BlockDriverAIOCB *bdrv_co_aio_rw_vector(BlockDriverState *bs, |
3464 | int64_t sector_num, | |
3465 | QEMUIOVector *qiov, | |
3466 | int nb_sectors, | |
3467 | BlockDriverCompletionFunc *cb, | |
3468 | void *opaque, | |
8c5873d6 | 3469 | bool is_write) |
68485420 KW |
3470 | { |
3471 | Coroutine *co; | |
3472 | BlockDriverAIOCBCoroutine *acb; | |
3473 | ||
3474 | acb = qemu_aio_get(&bdrv_em_co_aio_pool, bs, cb, opaque); | |
3475 | acb->req.sector = sector_num; | |
3476 | acb->req.nb_sectors = nb_sectors; | |
3477 | acb->req.qiov = qiov; | |
3478 | acb->is_write = is_write; | |
3479 | ||
8c5873d6 | 3480 | co = qemu_coroutine_create(bdrv_co_do_rw); |
68485420 KW |
3481 | qemu_coroutine_enter(co, acb); |
3482 | ||
3483 | return &acb->common; | |
3484 | } | |
3485 | ||
07f07615 | 3486 | static void coroutine_fn bdrv_aio_flush_co_entry(void *opaque) |
b2e12bc6 | 3487 | { |
07f07615 PB |
3488 | BlockDriverAIOCBCoroutine *acb = opaque; |
3489 | BlockDriverState *bs = acb->common.bs; | |
b2e12bc6 | 3490 | |
07f07615 PB |
3491 | acb->req.error = bdrv_co_flush(bs); |
3492 | acb->bh = qemu_bh_new(bdrv_co_em_bh, acb); | |
b2e12bc6 | 3493 | qemu_bh_schedule(acb->bh); |
b2e12bc6 CH |
3494 | } |
3495 | ||
07f07615 | 3496 | BlockDriverAIOCB *bdrv_aio_flush(BlockDriverState *bs, |
016f5cf6 AG |
3497 | BlockDriverCompletionFunc *cb, void *opaque) |
3498 | { | |
07f07615 | 3499 | trace_bdrv_aio_flush(bs, opaque); |
016f5cf6 | 3500 | |
07f07615 PB |
3501 | Coroutine *co; |
3502 | BlockDriverAIOCBCoroutine *acb; | |
016f5cf6 | 3503 | |
07f07615 PB |
3504 | acb = qemu_aio_get(&bdrv_em_co_aio_pool, bs, cb, opaque); |
3505 | co = qemu_coroutine_create(bdrv_aio_flush_co_entry); | |
3506 | qemu_coroutine_enter(co, acb); | |
016f5cf6 | 3507 | |
016f5cf6 AG |
3508 | return &acb->common; |
3509 | } | |
3510 | ||
4265d620 PB |
3511 | static void coroutine_fn bdrv_aio_discard_co_entry(void *opaque) |
3512 | { | |
3513 | BlockDriverAIOCBCoroutine *acb = opaque; | |
3514 | BlockDriverState *bs = acb->common.bs; | |
3515 | ||
3516 | acb->req.error = bdrv_co_discard(bs, acb->req.sector, acb->req.nb_sectors); | |
3517 | acb->bh = qemu_bh_new(bdrv_co_em_bh, acb); | |
3518 | qemu_bh_schedule(acb->bh); | |
3519 | } | |
3520 | ||
3521 | BlockDriverAIOCB *bdrv_aio_discard(BlockDriverState *bs, | |
3522 | int64_t sector_num, int nb_sectors, | |
3523 | BlockDriverCompletionFunc *cb, void *opaque) | |
3524 | { | |
3525 | Coroutine *co; | |
3526 | BlockDriverAIOCBCoroutine *acb; | |
3527 | ||
3528 | trace_bdrv_aio_discard(bs, sector_num, nb_sectors, opaque); | |
3529 | ||
3530 | acb = qemu_aio_get(&bdrv_em_co_aio_pool, bs, cb, opaque); | |
3531 | acb->req.sector = sector_num; | |
3532 | acb->req.nb_sectors = nb_sectors; | |
3533 | co = qemu_coroutine_create(bdrv_aio_discard_co_entry); | |
3534 | qemu_coroutine_enter(co, acb); | |
3535 | ||
3536 | return &acb->common; | |
3537 | } | |
3538 | ||
ea2384d3 FB |
3539 | void bdrv_init(void) |
3540 | { | |
5efa9d5a | 3541 | module_call_init(MODULE_INIT_BLOCK); |
ea2384d3 | 3542 | } |
ce1a14dc | 3543 | |
eb852011 MA |
3544 | void bdrv_init_with_whitelist(void) |
3545 | { | |
3546 | use_bdrv_whitelist = 1; | |
3547 | bdrv_init(); | |
3548 | } | |
3549 | ||
c16b5a2c CH |
3550 | void *qemu_aio_get(AIOPool *pool, BlockDriverState *bs, |
3551 | BlockDriverCompletionFunc *cb, void *opaque) | |
ce1a14dc | 3552 | { |
ce1a14dc PB |
3553 | BlockDriverAIOCB *acb; |
3554 | ||
6bbff9a0 AL |
3555 | if (pool->free_aiocb) { |
3556 | acb = pool->free_aiocb; | |
3557 | pool->free_aiocb = acb->next; | |
ce1a14dc | 3558 | } else { |
7267c094 | 3559 | acb = g_malloc0(pool->aiocb_size); |
6bbff9a0 | 3560 | acb->pool = pool; |
ce1a14dc PB |
3561 | } |
3562 | acb->bs = bs; | |
3563 | acb->cb = cb; | |
3564 | acb->opaque = opaque; | |
3565 | return acb; | |
3566 | } | |
3567 | ||
3568 | void qemu_aio_release(void *p) | |
3569 | { | |
6bbff9a0 AL |
3570 | BlockDriverAIOCB *acb = (BlockDriverAIOCB *)p; |
3571 | AIOPool *pool = acb->pool; | |
3572 | acb->next = pool->free_aiocb; | |
3573 | pool->free_aiocb = acb; | |
ce1a14dc | 3574 | } |
19cb3738 | 3575 | |
f9f05dc5 KW |
3576 | /**************************************************************/ |
3577 | /* Coroutine block device emulation */ | |
3578 | ||
3579 | typedef struct CoroutineIOCompletion { | |
3580 | Coroutine *coroutine; | |
3581 | int ret; | |
3582 | } CoroutineIOCompletion; | |
3583 | ||
3584 | static void bdrv_co_io_em_complete(void *opaque, int ret) | |
3585 | { | |
3586 | CoroutineIOCompletion *co = opaque; | |
3587 | ||
3588 | co->ret = ret; | |
3589 | qemu_coroutine_enter(co->coroutine, NULL); | |
3590 | } | |
3591 | ||
3592 | static int coroutine_fn bdrv_co_io_em(BlockDriverState *bs, int64_t sector_num, | |
3593 | int nb_sectors, QEMUIOVector *iov, | |
3594 | bool is_write) | |
3595 | { | |
3596 | CoroutineIOCompletion co = { | |
3597 | .coroutine = qemu_coroutine_self(), | |
3598 | }; | |
3599 | BlockDriverAIOCB *acb; | |
3600 | ||
3601 | if (is_write) { | |
a652d160 SH |
3602 | acb = bs->drv->bdrv_aio_writev(bs, sector_num, iov, nb_sectors, |
3603 | bdrv_co_io_em_complete, &co); | |
f9f05dc5 | 3604 | } else { |
a652d160 SH |
3605 | acb = bs->drv->bdrv_aio_readv(bs, sector_num, iov, nb_sectors, |
3606 | bdrv_co_io_em_complete, &co); | |
f9f05dc5 KW |
3607 | } |
3608 | ||
59370aaa | 3609 | trace_bdrv_co_io_em(bs, sector_num, nb_sectors, is_write, acb); |
f9f05dc5 KW |
3610 | if (!acb) { |
3611 | return -EIO; | |
3612 | } | |
3613 | qemu_coroutine_yield(); | |
3614 | ||
3615 | return co.ret; | |
3616 | } | |
3617 | ||
3618 | static int coroutine_fn bdrv_co_readv_em(BlockDriverState *bs, | |
3619 | int64_t sector_num, int nb_sectors, | |
3620 | QEMUIOVector *iov) | |
3621 | { | |
3622 | return bdrv_co_io_em(bs, sector_num, nb_sectors, iov, false); | |
3623 | } | |
3624 | ||
3625 | static int coroutine_fn bdrv_co_writev_em(BlockDriverState *bs, | |
3626 | int64_t sector_num, int nb_sectors, | |
3627 | QEMUIOVector *iov) | |
3628 | { | |
3629 | return bdrv_co_io_em(bs, sector_num, nb_sectors, iov, true); | |
3630 | } | |
3631 | ||
07f07615 | 3632 | static void coroutine_fn bdrv_flush_co_entry(void *opaque) |
e7a8a783 | 3633 | { |
07f07615 PB |
3634 | RwCo *rwco = opaque; |
3635 | ||
3636 | rwco->ret = bdrv_co_flush(rwco->bs); | |
3637 | } | |
3638 | ||
3639 | int coroutine_fn bdrv_co_flush(BlockDriverState *bs) | |
3640 | { | |
eb489bb1 KW |
3641 | int ret; |
3642 | ||
29cdb251 | 3643 | if (!bs || !bdrv_is_inserted(bs) || bdrv_is_read_only(bs)) { |
07f07615 | 3644 | return 0; |
eb489bb1 KW |
3645 | } |
3646 | ||
ca716364 | 3647 | /* Write back cached data to the OS even with cache=unsafe */ |
eb489bb1 KW |
3648 | if (bs->drv->bdrv_co_flush_to_os) { |
3649 | ret = bs->drv->bdrv_co_flush_to_os(bs); | |
3650 | if (ret < 0) { | |
3651 | return ret; | |
3652 | } | |
3653 | } | |
3654 | ||
ca716364 KW |
3655 | /* But don't actually force it to the disk with cache=unsafe */ |
3656 | if (bs->open_flags & BDRV_O_NO_FLUSH) { | |
3657 | return 0; | |
3658 | } | |
3659 | ||
eb489bb1 | 3660 | if (bs->drv->bdrv_co_flush_to_disk) { |
29cdb251 | 3661 | ret = bs->drv->bdrv_co_flush_to_disk(bs); |
07f07615 PB |
3662 | } else if (bs->drv->bdrv_aio_flush) { |
3663 | BlockDriverAIOCB *acb; | |
3664 | CoroutineIOCompletion co = { | |
3665 | .coroutine = qemu_coroutine_self(), | |
3666 | }; | |
3667 | ||
3668 | acb = bs->drv->bdrv_aio_flush(bs, bdrv_co_io_em_complete, &co); | |
3669 | if (acb == NULL) { | |
29cdb251 | 3670 | ret = -EIO; |
07f07615 PB |
3671 | } else { |
3672 | qemu_coroutine_yield(); | |
29cdb251 | 3673 | ret = co.ret; |
07f07615 | 3674 | } |
07f07615 PB |
3675 | } else { |
3676 | /* | |
3677 | * Some block drivers always operate in either writethrough or unsafe | |
3678 | * mode and don't support bdrv_flush therefore. Usually qemu doesn't | |
3679 | * know how the server works (because the behaviour is hardcoded or | |
3680 | * depends on server-side configuration), so we can't ensure that | |
3681 | * everything is safe on disk. Returning an error doesn't work because | |
3682 | * that would break guests even if the server operates in writethrough | |
3683 | * mode. | |
3684 | * | |
3685 | * Let's hope the user knows what he's doing. | |
3686 | */ | |
29cdb251 | 3687 | ret = 0; |
07f07615 | 3688 | } |
29cdb251 PB |
3689 | if (ret < 0) { |
3690 | return ret; | |
3691 | } | |
3692 | ||
3693 | /* Now flush the underlying protocol. It will also have BDRV_O_NO_FLUSH | |
3694 | * in the case of cache=unsafe, so there are no useless flushes. | |
3695 | */ | |
3696 | return bdrv_co_flush(bs->file); | |
07f07615 PB |
3697 | } |
3698 | ||
0f15423c AL |
3699 | void bdrv_invalidate_cache(BlockDriverState *bs) |
3700 | { | |
3701 | if (bs->drv && bs->drv->bdrv_invalidate_cache) { | |
3702 | bs->drv->bdrv_invalidate_cache(bs); | |
3703 | } | |
3704 | } | |
3705 | ||
3706 | void bdrv_invalidate_cache_all(void) | |
3707 | { | |
3708 | BlockDriverState *bs; | |
3709 | ||
3710 | QTAILQ_FOREACH(bs, &bdrv_states, list) { | |
3711 | bdrv_invalidate_cache(bs); | |
3712 | } | |
3713 | } | |
3714 | ||
07789269 BC |
3715 | void bdrv_clear_incoming_migration_all(void) |
3716 | { | |
3717 | BlockDriverState *bs; | |
3718 | ||
3719 | QTAILQ_FOREACH(bs, &bdrv_states, list) { | |
3720 | bs->open_flags = bs->open_flags & ~(BDRV_O_INCOMING); | |
3721 | } | |
3722 | } | |
3723 | ||
07f07615 PB |
3724 | int bdrv_flush(BlockDriverState *bs) |
3725 | { | |
3726 | Coroutine *co; | |
3727 | RwCo rwco = { | |
3728 | .bs = bs, | |
3729 | .ret = NOT_DONE, | |
e7a8a783 | 3730 | }; |
e7a8a783 | 3731 | |
07f07615 PB |
3732 | if (qemu_in_coroutine()) { |
3733 | /* Fast-path if already in coroutine context */ | |
3734 | bdrv_flush_co_entry(&rwco); | |
3735 | } else { | |
3736 | co = qemu_coroutine_create(bdrv_flush_co_entry); | |
3737 | qemu_coroutine_enter(co, &rwco); | |
3738 | while (rwco.ret == NOT_DONE) { | |
3739 | qemu_aio_wait(); | |
3740 | } | |
e7a8a783 | 3741 | } |
07f07615 PB |
3742 | |
3743 | return rwco.ret; | |
e7a8a783 KW |
3744 | } |
3745 | ||
4265d620 PB |
3746 | static void coroutine_fn bdrv_discard_co_entry(void *opaque) |
3747 | { | |
3748 | RwCo *rwco = opaque; | |
3749 | ||
3750 | rwco->ret = bdrv_co_discard(rwco->bs, rwco->sector_num, rwco->nb_sectors); | |
3751 | } | |
3752 | ||
3753 | int coroutine_fn bdrv_co_discard(BlockDriverState *bs, int64_t sector_num, | |
3754 | int nb_sectors) | |
3755 | { | |
3756 | if (!bs->drv) { | |
3757 | return -ENOMEDIUM; | |
3758 | } else if (bdrv_check_request(bs, sector_num, nb_sectors)) { | |
3759 | return -EIO; | |
3760 | } else if (bs->read_only) { | |
3761 | return -EROFS; | |
3762 | } else if (bs->drv->bdrv_co_discard) { | |
3763 | return bs->drv->bdrv_co_discard(bs, sector_num, nb_sectors); | |
3764 | } else if (bs->drv->bdrv_aio_discard) { | |
3765 | BlockDriverAIOCB *acb; | |
3766 | CoroutineIOCompletion co = { | |
3767 | .coroutine = qemu_coroutine_self(), | |
3768 | }; | |
3769 | ||
3770 | acb = bs->drv->bdrv_aio_discard(bs, sector_num, nb_sectors, | |
3771 | bdrv_co_io_em_complete, &co); | |
3772 | if (acb == NULL) { | |
3773 | return -EIO; | |
3774 | } else { | |
3775 | qemu_coroutine_yield(); | |
3776 | return co.ret; | |
3777 | } | |
4265d620 PB |
3778 | } else { |
3779 | return 0; | |
3780 | } | |
3781 | } | |
3782 | ||
3783 | int bdrv_discard(BlockDriverState *bs, int64_t sector_num, int nb_sectors) | |
3784 | { | |
3785 | Coroutine *co; | |
3786 | RwCo rwco = { | |
3787 | .bs = bs, | |
3788 | .sector_num = sector_num, | |
3789 | .nb_sectors = nb_sectors, | |
3790 | .ret = NOT_DONE, | |
3791 | }; | |
3792 | ||
3793 | if (qemu_in_coroutine()) { | |
3794 | /* Fast-path if already in coroutine context */ | |
3795 | bdrv_discard_co_entry(&rwco); | |
3796 | } else { | |
3797 | co = qemu_coroutine_create(bdrv_discard_co_entry); | |
3798 | qemu_coroutine_enter(co, &rwco); | |
3799 | while (rwco.ret == NOT_DONE) { | |
3800 | qemu_aio_wait(); | |
3801 | } | |
3802 | } | |
3803 | ||
3804 | return rwco.ret; | |
3805 | } | |
3806 | ||
19cb3738 FB |
3807 | /**************************************************************/ |
3808 | /* removable device support */ | |
3809 | ||
3810 | /** | |
3811 | * Return TRUE if the media is present | |
3812 | */ | |
3813 | int bdrv_is_inserted(BlockDriverState *bs) | |
3814 | { | |
3815 | BlockDriver *drv = bs->drv; | |
a1aff5bf | 3816 | |
19cb3738 FB |
3817 | if (!drv) |
3818 | return 0; | |
3819 | if (!drv->bdrv_is_inserted) | |
a1aff5bf MA |
3820 | return 1; |
3821 | return drv->bdrv_is_inserted(bs); | |
19cb3738 FB |
3822 | } |
3823 | ||
3824 | /** | |
8e49ca46 MA |
3825 | * Return whether the media changed since the last call to this |
3826 | * function, or -ENOTSUP if we don't know. Most drivers don't know. | |
19cb3738 FB |
3827 | */ |
3828 | int bdrv_media_changed(BlockDriverState *bs) | |
3829 | { | |
3830 | BlockDriver *drv = bs->drv; | |
19cb3738 | 3831 | |
8e49ca46 MA |
3832 | if (drv && drv->bdrv_media_changed) { |
3833 | return drv->bdrv_media_changed(bs); | |
3834 | } | |
3835 | return -ENOTSUP; | |
19cb3738 FB |
3836 | } |
3837 | ||
3838 | /** | |
3839 | * If eject_flag is TRUE, eject the media. Otherwise, close the tray | |
3840 | */ | |
f36f3949 | 3841 | void bdrv_eject(BlockDriverState *bs, bool eject_flag) |
19cb3738 FB |
3842 | { |
3843 | BlockDriver *drv = bs->drv; | |
19cb3738 | 3844 | |
822e1cd1 MA |
3845 | if (drv && drv->bdrv_eject) { |
3846 | drv->bdrv_eject(bs, eject_flag); | |
19cb3738 | 3847 | } |
6f382ed2 LC |
3848 | |
3849 | if (bs->device_name[0] != '\0') { | |
3850 | bdrv_emit_qmp_eject_event(bs, eject_flag); | |
3851 | } | |
19cb3738 FB |
3852 | } |
3853 | ||
19cb3738 FB |
3854 | /** |
3855 | * Lock or unlock the media (if it is locked, the user won't be able | |
3856 | * to eject it manually). | |
3857 | */ | |
025e849a | 3858 | void bdrv_lock_medium(BlockDriverState *bs, bool locked) |
19cb3738 FB |
3859 | { |
3860 | BlockDriver *drv = bs->drv; | |
3861 | ||
025e849a | 3862 | trace_bdrv_lock_medium(bs, locked); |
b8c6d095 | 3863 | |
025e849a MA |
3864 | if (drv && drv->bdrv_lock_medium) { |
3865 | drv->bdrv_lock_medium(bs, locked); | |
19cb3738 FB |
3866 | } |
3867 | } | |
985a03b0 TS |
3868 | |
3869 | /* needed for generic scsi interface */ | |
3870 | ||
3871 | int bdrv_ioctl(BlockDriverState *bs, unsigned long int req, void *buf) | |
3872 | { | |
3873 | BlockDriver *drv = bs->drv; | |
3874 | ||
3875 | if (drv && drv->bdrv_ioctl) | |
3876 | return drv->bdrv_ioctl(bs, req, buf); | |
3877 | return -ENOTSUP; | |
3878 | } | |
7d780669 | 3879 | |
221f715d AL |
3880 | BlockDriverAIOCB *bdrv_aio_ioctl(BlockDriverState *bs, |
3881 | unsigned long int req, void *buf, | |
3882 | BlockDriverCompletionFunc *cb, void *opaque) | |
7d780669 | 3883 | { |
221f715d | 3884 | BlockDriver *drv = bs->drv; |
7d780669 | 3885 | |
221f715d AL |
3886 | if (drv && drv->bdrv_aio_ioctl) |
3887 | return drv->bdrv_aio_ioctl(bs, req, buf, cb, opaque); | |
3888 | return NULL; | |
7d780669 | 3889 | } |
e268ca52 | 3890 | |
7b6f9300 MA |
3891 | void bdrv_set_buffer_alignment(BlockDriverState *bs, int align) |
3892 | { | |
3893 | bs->buffer_alignment = align; | |
3894 | } | |
7cd1e32a | 3895 | |
e268ca52 AL |
3896 | void *qemu_blockalign(BlockDriverState *bs, size_t size) |
3897 | { | |
3898 | return qemu_memalign((bs && bs->buffer_alignment) ? bs->buffer_alignment : 512, size); | |
3899 | } | |
7cd1e32a | 3900 | |
3901 | void bdrv_set_dirty_tracking(BlockDriverState *bs, int enable) | |
3902 | { | |
3903 | int64_t bitmap_size; | |
a55eb92c | 3904 | |
aaa0eb75 | 3905 | bs->dirty_count = 0; |
a55eb92c | 3906 | if (enable) { |
c6d22830 JK |
3907 | if (!bs->dirty_bitmap) { |
3908 | bitmap_size = (bdrv_getlength(bs) >> BDRV_SECTOR_BITS) + | |
71df14fc PB |
3909 | BDRV_SECTORS_PER_DIRTY_CHUNK * BITS_PER_LONG - 1; |
3910 | bitmap_size /= BDRV_SECTORS_PER_DIRTY_CHUNK * BITS_PER_LONG; | |
a55eb92c | 3911 | |
71df14fc | 3912 | bs->dirty_bitmap = g_new0(unsigned long, bitmap_size); |
a55eb92c | 3913 | } |
7cd1e32a | 3914 | } else { |
c6d22830 | 3915 | if (bs->dirty_bitmap) { |
7267c094 | 3916 | g_free(bs->dirty_bitmap); |
c6d22830 | 3917 | bs->dirty_bitmap = NULL; |
a55eb92c | 3918 | } |
7cd1e32a | 3919 | } |
3920 | } | |
3921 | ||
3922 | int bdrv_get_dirty(BlockDriverState *bs, int64_t sector) | |
3923 | { | |
6ea44308 | 3924 | int64_t chunk = sector / (int64_t)BDRV_SECTORS_PER_DIRTY_CHUNK; |
a55eb92c | 3925 | |
c6d22830 JK |
3926 | if (bs->dirty_bitmap && |
3927 | (sector << BDRV_SECTOR_BITS) < bdrv_getlength(bs)) { | |
6d59fec1 MT |
3928 | return !!(bs->dirty_bitmap[chunk / (sizeof(unsigned long) * 8)] & |
3929 | (1UL << (chunk % (sizeof(unsigned long) * 8)))); | |
7cd1e32a | 3930 | } else { |
3931 | return 0; | |
3932 | } | |
3933 | } | |
3934 | ||
a55eb92c JK |
3935 | void bdrv_reset_dirty(BlockDriverState *bs, int64_t cur_sector, |
3936 | int nr_sectors) | |
7cd1e32a | 3937 | { |
3938 | set_dirty_bitmap(bs, cur_sector, nr_sectors, 0); | |
3939 | } | |
aaa0eb75 LS |
3940 | |
3941 | int64_t bdrv_get_dirty_count(BlockDriverState *bs) | |
3942 | { | |
3943 | return bs->dirty_count; | |
3944 | } | |
f88e1a42 | 3945 | |
db593f25 MT |
3946 | void bdrv_set_in_use(BlockDriverState *bs, int in_use) |
3947 | { | |
3948 | assert(bs->in_use != in_use); | |
3949 | bs->in_use = in_use; | |
3950 | } | |
3951 | ||
3952 | int bdrv_in_use(BlockDriverState *bs) | |
3953 | { | |
3954 | return bs->in_use; | |
3955 | } | |
3956 | ||
28a7282a LC |
3957 | void bdrv_iostatus_enable(BlockDriverState *bs) |
3958 | { | |
d6bf279e | 3959 | bs->iostatus_enabled = true; |
58e21ef5 | 3960 | bs->iostatus = BLOCK_DEVICE_IO_STATUS_OK; |
28a7282a LC |
3961 | } |
3962 | ||
3963 | /* The I/O status is only enabled if the drive explicitly | |
3964 | * enables it _and_ the VM is configured to stop on errors */ | |
3965 | bool bdrv_iostatus_is_enabled(const BlockDriverState *bs) | |
3966 | { | |
d6bf279e | 3967 | return (bs->iostatus_enabled && |
28a7282a LC |
3968 | (bs->on_write_error == BLOCK_ERR_STOP_ENOSPC || |
3969 | bs->on_write_error == BLOCK_ERR_STOP_ANY || | |
3970 | bs->on_read_error == BLOCK_ERR_STOP_ANY)); | |
3971 | } | |
3972 | ||
3973 | void bdrv_iostatus_disable(BlockDriverState *bs) | |
3974 | { | |
d6bf279e | 3975 | bs->iostatus_enabled = false; |
28a7282a LC |
3976 | } |
3977 | ||
3978 | void bdrv_iostatus_reset(BlockDriverState *bs) | |
3979 | { | |
3980 | if (bdrv_iostatus_is_enabled(bs)) { | |
58e21ef5 | 3981 | bs->iostatus = BLOCK_DEVICE_IO_STATUS_OK; |
28a7282a LC |
3982 | } |
3983 | } | |
3984 | ||
3985 | /* XXX: Today this is set by device models because it makes the implementation | |
3986 | quite simple. However, the block layer knows about the error, so it's | |
3987 | possible to implement this without device models being involved */ | |
3988 | void bdrv_iostatus_set_err(BlockDriverState *bs, int error) | |
3989 | { | |
58e21ef5 LC |
3990 | if (bdrv_iostatus_is_enabled(bs) && |
3991 | bs->iostatus == BLOCK_DEVICE_IO_STATUS_OK) { | |
28a7282a | 3992 | assert(error >= 0); |
58e21ef5 LC |
3993 | bs->iostatus = error == ENOSPC ? BLOCK_DEVICE_IO_STATUS_NOSPACE : |
3994 | BLOCK_DEVICE_IO_STATUS_FAILED; | |
28a7282a LC |
3995 | } |
3996 | } | |
3997 | ||
a597e79c CH |
3998 | void |
3999 | bdrv_acct_start(BlockDriverState *bs, BlockAcctCookie *cookie, int64_t bytes, | |
4000 | enum BlockAcctType type) | |
4001 | { | |
4002 | assert(type < BDRV_MAX_IOTYPE); | |
4003 | ||
4004 | cookie->bytes = bytes; | |
c488c7f6 | 4005 | cookie->start_time_ns = get_clock(); |
a597e79c CH |
4006 | cookie->type = type; |
4007 | } | |
4008 | ||
4009 | void | |
4010 | bdrv_acct_done(BlockDriverState *bs, BlockAcctCookie *cookie) | |
4011 | { | |
4012 | assert(cookie->type < BDRV_MAX_IOTYPE); | |
4013 | ||
4014 | bs->nr_bytes[cookie->type] += cookie->bytes; | |
4015 | bs->nr_ops[cookie->type]++; | |
c488c7f6 | 4016 | bs->total_time_ns[cookie->type] += get_clock() - cookie->start_time_ns; |
a597e79c CH |
4017 | } |
4018 | ||
f88e1a42 JS |
4019 | int bdrv_img_create(const char *filename, const char *fmt, |
4020 | const char *base_filename, const char *base_fmt, | |
4021 | char *options, uint64_t img_size, int flags) | |
4022 | { | |
4023 | QEMUOptionParameter *param = NULL, *create_options = NULL; | |
d220894e | 4024 | QEMUOptionParameter *backing_fmt, *backing_file, *size; |
f88e1a42 JS |
4025 | BlockDriverState *bs = NULL; |
4026 | BlockDriver *drv, *proto_drv; | |
96df67d1 | 4027 | BlockDriver *backing_drv = NULL; |
f88e1a42 JS |
4028 | int ret = 0; |
4029 | ||
4030 | /* Find driver and parse its options */ | |
4031 | drv = bdrv_find_format(fmt); | |
4032 | if (!drv) { | |
4033 | error_report("Unknown file format '%s'", fmt); | |
4f70f249 | 4034 | ret = -EINVAL; |
f88e1a42 JS |
4035 | goto out; |
4036 | } | |
4037 | ||
4038 | proto_drv = bdrv_find_protocol(filename); | |
4039 | if (!proto_drv) { | |
4040 | error_report("Unknown protocol '%s'", filename); | |
4f70f249 | 4041 | ret = -EINVAL; |
f88e1a42 JS |
4042 | goto out; |
4043 | } | |
4044 | ||
4045 | create_options = append_option_parameters(create_options, | |
4046 | drv->create_options); | |
4047 | create_options = append_option_parameters(create_options, | |
4048 | proto_drv->create_options); | |
4049 | ||
4050 | /* Create parameter list with default values */ | |
4051 | param = parse_option_parameters("", create_options, param); | |
4052 | ||
4053 | set_option_parameter_int(param, BLOCK_OPT_SIZE, img_size); | |
4054 | ||
4055 | /* Parse -o options */ | |
4056 | if (options) { | |
4057 | param = parse_option_parameters(options, create_options, param); | |
4058 | if (param == NULL) { | |
4059 | error_report("Invalid options for file format '%s'.", fmt); | |
4f70f249 | 4060 | ret = -EINVAL; |
f88e1a42 JS |
4061 | goto out; |
4062 | } | |
4063 | } | |
4064 | ||
4065 | if (base_filename) { | |
4066 | if (set_option_parameter(param, BLOCK_OPT_BACKING_FILE, | |
4067 | base_filename)) { | |
4068 | error_report("Backing file not supported for file format '%s'", | |
4069 | fmt); | |
4f70f249 | 4070 | ret = -EINVAL; |
f88e1a42 JS |
4071 | goto out; |
4072 | } | |
4073 | } | |
4074 | ||
4075 | if (base_fmt) { | |
4076 | if (set_option_parameter(param, BLOCK_OPT_BACKING_FMT, base_fmt)) { | |
4077 | error_report("Backing file format not supported for file " | |
4078 | "format '%s'", fmt); | |
4f70f249 | 4079 | ret = -EINVAL; |
f88e1a42 JS |
4080 | goto out; |
4081 | } | |
4082 | } | |
4083 | ||
792da93a JS |
4084 | backing_file = get_option_parameter(param, BLOCK_OPT_BACKING_FILE); |
4085 | if (backing_file && backing_file->value.s) { | |
4086 | if (!strcmp(filename, backing_file->value.s)) { | |
4087 | error_report("Error: Trying to create an image with the " | |
4088 | "same filename as the backing file"); | |
4f70f249 | 4089 | ret = -EINVAL; |
792da93a JS |
4090 | goto out; |
4091 | } | |
4092 | } | |
4093 | ||
f88e1a42 JS |
4094 | backing_fmt = get_option_parameter(param, BLOCK_OPT_BACKING_FMT); |
4095 | if (backing_fmt && backing_fmt->value.s) { | |
96df67d1 SH |
4096 | backing_drv = bdrv_find_format(backing_fmt->value.s); |
4097 | if (!backing_drv) { | |
f88e1a42 JS |
4098 | error_report("Unknown backing file format '%s'", |
4099 | backing_fmt->value.s); | |
4f70f249 | 4100 | ret = -EINVAL; |
f88e1a42 JS |
4101 | goto out; |
4102 | } | |
4103 | } | |
4104 | ||
4105 | // The size for the image must always be specified, with one exception: | |
4106 | // If we are using a backing file, we can obtain the size from there | |
d220894e KW |
4107 | size = get_option_parameter(param, BLOCK_OPT_SIZE); |
4108 | if (size && size->value.n == -1) { | |
f88e1a42 JS |
4109 | if (backing_file && backing_file->value.s) { |
4110 | uint64_t size; | |
f88e1a42 | 4111 | char buf[32]; |
63090dac PB |
4112 | int back_flags; |
4113 | ||
4114 | /* backing files always opened read-only */ | |
4115 | back_flags = | |
4116 | flags & ~(BDRV_O_RDWR | BDRV_O_SNAPSHOT | BDRV_O_NO_BACKING); | |
f88e1a42 | 4117 | |
f88e1a42 JS |
4118 | bs = bdrv_new(""); |
4119 | ||
63090dac | 4120 | ret = bdrv_open(bs, backing_file->value.s, back_flags, backing_drv); |
f88e1a42 | 4121 | if (ret < 0) { |
96df67d1 | 4122 | error_report("Could not open '%s'", backing_file->value.s); |
f88e1a42 JS |
4123 | goto out; |
4124 | } | |
4125 | bdrv_get_geometry(bs, &size); | |
4126 | size *= 512; | |
4127 | ||
4128 | snprintf(buf, sizeof(buf), "%" PRId64, size); | |
4129 | set_option_parameter(param, BLOCK_OPT_SIZE, buf); | |
4130 | } else { | |
4131 | error_report("Image creation needs a size parameter"); | |
4f70f249 | 4132 | ret = -EINVAL; |
f88e1a42 JS |
4133 | goto out; |
4134 | } | |
4135 | } | |
4136 | ||
4137 | printf("Formatting '%s', fmt=%s ", filename, fmt); | |
4138 | print_option_parameters(param); | |
4139 | puts(""); | |
4140 | ||
4141 | ret = bdrv_create(drv, filename, param); | |
4142 | ||
4143 | if (ret < 0) { | |
4144 | if (ret == -ENOTSUP) { | |
4145 | error_report("Formatting or formatting option not supported for " | |
4146 | "file format '%s'", fmt); | |
4147 | } else if (ret == -EFBIG) { | |
4148 | error_report("The image size is too large for file format '%s'", | |
4149 | fmt); | |
4150 | } else { | |
4151 | error_report("%s: error while creating %s: %s", filename, fmt, | |
4152 | strerror(-ret)); | |
4153 | } | |
4154 | } | |
4155 | ||
4156 | out: | |
4157 | free_option_parameters(create_options); | |
4158 | free_option_parameters(param); | |
4159 | ||
4160 | if (bs) { | |
4161 | bdrv_delete(bs); | |
4162 | } | |
4f70f249 JS |
4163 | |
4164 | return ret; | |
f88e1a42 | 4165 | } |
eeec61f2 SH |
4166 | |
4167 | void *block_job_create(const BlockJobType *job_type, BlockDriverState *bs, | |
c83c66c3 SH |
4168 | int64_t speed, BlockDriverCompletionFunc *cb, |
4169 | void *opaque, Error **errp) | |
eeec61f2 SH |
4170 | { |
4171 | BlockJob *job; | |
4172 | ||
4173 | if (bs->job || bdrv_in_use(bs)) { | |
fd7f8c65 | 4174 | error_set(errp, QERR_DEVICE_IN_USE, bdrv_get_device_name(bs)); |
eeec61f2 SH |
4175 | return NULL; |
4176 | } | |
4177 | bdrv_set_in_use(bs, 1); | |
4178 | ||
4179 | job = g_malloc0(job_type->instance_size); | |
4180 | job->job_type = job_type; | |
4181 | job->bs = bs; | |
4182 | job->cb = cb; | |
4183 | job->opaque = opaque; | |
4513eafe | 4184 | job->busy = true; |
eeec61f2 | 4185 | bs->job = job; |
c83c66c3 SH |
4186 | |
4187 | /* Only set speed when necessary to avoid NotSupported error */ | |
4188 | if (speed != 0) { | |
4189 | Error *local_err = NULL; | |
4190 | ||
4191 | block_job_set_speed(job, speed, &local_err); | |
4192 | if (error_is_set(&local_err)) { | |
4193 | bs->job = NULL; | |
4194 | g_free(job); | |
4195 | bdrv_set_in_use(bs, 0); | |
4196 | error_propagate(errp, local_err); | |
4197 | return NULL; | |
4198 | } | |
4199 | } | |
eeec61f2 SH |
4200 | return job; |
4201 | } | |
4202 | ||
4203 | void block_job_complete(BlockJob *job, int ret) | |
4204 | { | |
4205 | BlockDriverState *bs = job->bs; | |
4206 | ||
4207 | assert(bs->job == job); | |
4208 | job->cb(job->opaque, ret); | |
4209 | bs->job = NULL; | |
4210 | g_free(job); | |
4211 | bdrv_set_in_use(bs, 0); | |
4212 | } | |
4213 | ||
882ec7ce | 4214 | void block_job_set_speed(BlockJob *job, int64_t speed, Error **errp) |
eeec61f2 | 4215 | { |
9e6636c7 | 4216 | Error *local_err = NULL; |
9f25eccc | 4217 | |
eeec61f2 | 4218 | if (!job->job_type->set_speed) { |
9e6636c7 SH |
4219 | error_set(errp, QERR_NOT_SUPPORTED); |
4220 | return; | |
eeec61f2 | 4221 | } |
882ec7ce | 4222 | job->job_type->set_speed(job, speed, &local_err); |
9e6636c7 SH |
4223 | if (error_is_set(&local_err)) { |
4224 | error_propagate(errp, local_err); | |
4225 | return; | |
9f25eccc | 4226 | } |
9e6636c7 | 4227 | |
882ec7ce | 4228 | job->speed = speed; |
eeec61f2 SH |
4229 | } |
4230 | ||
4231 | void block_job_cancel(BlockJob *job) | |
4232 | { | |
4233 | job->cancelled = true; | |
fa4478d5 PB |
4234 | if (job->co && !job->busy) { |
4235 | qemu_coroutine_enter(job->co, NULL); | |
4236 | } | |
eeec61f2 SH |
4237 | } |
4238 | ||
4239 | bool block_job_is_cancelled(BlockJob *job) | |
4240 | { | |
4241 | return job->cancelled; | |
4242 | } | |
3e914655 | 4243 | |
fa4478d5 PB |
4244 | struct BlockCancelData { |
4245 | BlockJob *job; | |
4246 | BlockDriverCompletionFunc *cb; | |
4247 | void *opaque; | |
4248 | bool cancelled; | |
4249 | int ret; | |
4250 | }; | |
4251 | ||
4252 | static void block_job_cancel_cb(void *opaque, int ret) | |
3e914655 | 4253 | { |
fa4478d5 PB |
4254 | struct BlockCancelData *data = opaque; |
4255 | ||
4256 | data->cancelled = block_job_is_cancelled(data->job); | |
4257 | data->ret = ret; | |
4258 | data->cb(data->opaque, ret); | |
4259 | } | |
4260 | ||
4261 | int block_job_cancel_sync(BlockJob *job) | |
4262 | { | |
4263 | struct BlockCancelData data; | |
3e914655 PB |
4264 | BlockDriverState *bs = job->bs; |
4265 | ||
4266 | assert(bs->job == job); | |
fa4478d5 PB |
4267 | |
4268 | /* Set up our own callback to store the result and chain to | |
4269 | * the original callback. | |
4270 | */ | |
4271 | data.job = job; | |
4272 | data.cb = job->cb; | |
4273 | data.opaque = job->opaque; | |
4274 | data.ret = -EINPROGRESS; | |
4275 | job->cb = block_job_cancel_cb; | |
4276 | job->opaque = &data; | |
3e914655 | 4277 | block_job_cancel(job); |
fa4478d5 | 4278 | while (data.ret == -EINPROGRESS) { |
3e914655 PB |
4279 | qemu_aio_wait(); |
4280 | } | |
fa4478d5 | 4281 | return (data.cancelled && data.ret == 0) ? -ECANCELED : data.ret; |
3e914655 | 4282 | } |
4513eafe PB |
4283 | |
4284 | void block_job_sleep_ns(BlockJob *job, QEMUClock *clock, int64_t ns) | |
4285 | { | |
4286 | /* Check cancellation *before* setting busy = false, too! */ | |
4287 | if (!block_job_is_cancelled(job)) { | |
4288 | job->busy = false; | |
4289 | co_sleep_ns(clock, ns); | |
4290 | job->busy = true; | |
4291 | } | |
4292 | } |