]>
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 | { | |
947995c0 PB |
201 | const char *p; |
202 | ||
9e0b22f4 SH |
203 | #ifdef _WIN32 |
204 | if (is_windows_drive(path) || | |
205 | is_windows_drive_prefix(path)) { | |
206 | return 0; | |
207 | } | |
947995c0 PB |
208 | p = path + strcspn(path, ":/\\"); |
209 | #else | |
210 | p = path + strcspn(path, ":/"); | |
9e0b22f4 SH |
211 | #endif |
212 | ||
947995c0 | 213 | return *p == ':'; |
9e0b22f4 SH |
214 | } |
215 | ||
83f64091 | 216 | int path_is_absolute(const char *path) |
3b0d4f61 | 217 | { |
21664424 FB |
218 | #ifdef _WIN32 |
219 | /* specific case for names like: "\\.\d:" */ | |
f53f4da9 | 220 | if (is_windows_drive(path) || is_windows_drive_prefix(path)) { |
21664424 | 221 | return 1; |
f53f4da9 PB |
222 | } |
223 | return (*path == '/' || *path == '\\'); | |
3b9f94e1 | 224 | #else |
f53f4da9 | 225 | return (*path == '/'); |
3b9f94e1 | 226 | #endif |
3b0d4f61 FB |
227 | } |
228 | ||
83f64091 FB |
229 | /* if filename is absolute, just copy it to dest. Otherwise, build a |
230 | path to it by considering it is relative to base_path. URL are | |
231 | supported. */ | |
232 | void path_combine(char *dest, int dest_size, | |
233 | const char *base_path, | |
234 | const char *filename) | |
3b0d4f61 | 235 | { |
83f64091 FB |
236 | const char *p, *p1; |
237 | int len; | |
238 | ||
239 | if (dest_size <= 0) | |
240 | return; | |
241 | if (path_is_absolute(filename)) { | |
242 | pstrcpy(dest, dest_size, filename); | |
243 | } else { | |
244 | p = strchr(base_path, ':'); | |
245 | if (p) | |
246 | p++; | |
247 | else | |
248 | p = base_path; | |
3b9f94e1 FB |
249 | p1 = strrchr(base_path, '/'); |
250 | #ifdef _WIN32 | |
251 | { | |
252 | const char *p2; | |
253 | p2 = strrchr(base_path, '\\'); | |
254 | if (!p1 || p2 > p1) | |
255 | p1 = p2; | |
256 | } | |
257 | #endif | |
83f64091 FB |
258 | if (p1) |
259 | p1++; | |
260 | else | |
261 | p1 = base_path; | |
262 | if (p1 > p) | |
263 | p = p1; | |
264 | len = p - base_path; | |
265 | if (len > dest_size - 1) | |
266 | len = dest_size - 1; | |
267 | memcpy(dest, base_path, len); | |
268 | dest[len] = '\0'; | |
269 | pstrcat(dest, dest_size, filename); | |
3b0d4f61 | 270 | } |
3b0d4f61 FB |
271 | } |
272 | ||
dc5a1371 PB |
273 | void bdrv_get_full_backing_filename(BlockDriverState *bs, char *dest, size_t sz) |
274 | { | |
275 | if (bs->backing_file[0] == '\0' || path_has_protocol(bs->backing_file)) { | |
276 | pstrcpy(dest, sz, bs->backing_file); | |
277 | } else { | |
278 | path_combine(dest, sz, bs->filename, bs->backing_file); | |
279 | } | |
280 | } | |
281 | ||
5efa9d5a | 282 | void bdrv_register(BlockDriver *bdrv) |
ea2384d3 | 283 | { |
8c5873d6 SH |
284 | /* Block drivers without coroutine functions need emulation */ |
285 | if (!bdrv->bdrv_co_readv) { | |
f9f05dc5 KW |
286 | bdrv->bdrv_co_readv = bdrv_co_readv_em; |
287 | bdrv->bdrv_co_writev = bdrv_co_writev_em; | |
288 | ||
f8c35c1d SH |
289 | /* bdrv_co_readv_em()/brdv_co_writev_em() work in terms of aio, so if |
290 | * the block driver lacks aio we need to emulate that too. | |
291 | */ | |
f9f05dc5 KW |
292 | if (!bdrv->bdrv_aio_readv) { |
293 | /* add AIO emulation layer */ | |
294 | bdrv->bdrv_aio_readv = bdrv_aio_readv_em; | |
295 | bdrv->bdrv_aio_writev = bdrv_aio_writev_em; | |
f9f05dc5 | 296 | } |
83f64091 | 297 | } |
b2e12bc6 | 298 | |
8a22f02a | 299 | QLIST_INSERT_HEAD(&bdrv_drivers, bdrv, list); |
ea2384d3 | 300 | } |
b338082b FB |
301 | |
302 | /* create a new block device (by default it is empty) */ | |
303 | BlockDriverState *bdrv_new(const char *device_name) | |
304 | { | |
1b7bdbc1 | 305 | BlockDriverState *bs; |
b338082b | 306 | |
7267c094 | 307 | bs = g_malloc0(sizeof(BlockDriverState)); |
b338082b | 308 | pstrcpy(bs->device_name, sizeof(bs->device_name), device_name); |
ea2384d3 | 309 | if (device_name[0] != '\0') { |
1b7bdbc1 | 310 | QTAILQ_INSERT_TAIL(&bdrv_states, bs, list); |
ea2384d3 | 311 | } |
28a7282a | 312 | bdrv_iostatus_disable(bs); |
b338082b FB |
313 | return bs; |
314 | } | |
315 | ||
ea2384d3 FB |
316 | BlockDriver *bdrv_find_format(const char *format_name) |
317 | { | |
318 | BlockDriver *drv1; | |
8a22f02a SH |
319 | QLIST_FOREACH(drv1, &bdrv_drivers, list) { |
320 | if (!strcmp(drv1->format_name, format_name)) { | |
ea2384d3 | 321 | return drv1; |
8a22f02a | 322 | } |
ea2384d3 FB |
323 | } |
324 | return NULL; | |
325 | } | |
326 | ||
eb852011 MA |
327 | static int bdrv_is_whitelisted(BlockDriver *drv) |
328 | { | |
329 | static const char *whitelist[] = { | |
330 | CONFIG_BDRV_WHITELIST | |
331 | }; | |
332 | const char **p; | |
333 | ||
334 | if (!whitelist[0]) | |
335 | return 1; /* no whitelist, anything goes */ | |
336 | ||
337 | for (p = whitelist; *p; p++) { | |
338 | if (!strcmp(drv->format_name, *p)) { | |
339 | return 1; | |
340 | } | |
341 | } | |
342 | return 0; | |
343 | } | |
344 | ||
345 | BlockDriver *bdrv_find_whitelisted_format(const char *format_name) | |
346 | { | |
347 | BlockDriver *drv = bdrv_find_format(format_name); | |
348 | return drv && bdrv_is_whitelisted(drv) ? drv : NULL; | |
349 | } | |
350 | ||
5b7e1542 ZYW |
351 | typedef struct CreateCo { |
352 | BlockDriver *drv; | |
353 | char *filename; | |
354 | QEMUOptionParameter *options; | |
355 | int ret; | |
356 | } CreateCo; | |
357 | ||
358 | static void coroutine_fn bdrv_create_co_entry(void *opaque) | |
359 | { | |
360 | CreateCo *cco = opaque; | |
361 | assert(cco->drv); | |
362 | ||
363 | cco->ret = cco->drv->bdrv_create(cco->filename, cco->options); | |
364 | } | |
365 | ||
0e7e1989 KW |
366 | int bdrv_create(BlockDriver *drv, const char* filename, |
367 | QEMUOptionParameter *options) | |
ea2384d3 | 368 | { |
5b7e1542 ZYW |
369 | int ret; |
370 | ||
371 | Coroutine *co; | |
372 | CreateCo cco = { | |
373 | .drv = drv, | |
374 | .filename = g_strdup(filename), | |
375 | .options = options, | |
376 | .ret = NOT_DONE, | |
377 | }; | |
378 | ||
379 | if (!drv->bdrv_create) { | |
ea2384d3 | 380 | return -ENOTSUP; |
5b7e1542 ZYW |
381 | } |
382 | ||
383 | if (qemu_in_coroutine()) { | |
384 | /* Fast-path if already in coroutine context */ | |
385 | bdrv_create_co_entry(&cco); | |
386 | } else { | |
387 | co = qemu_coroutine_create(bdrv_create_co_entry); | |
388 | qemu_coroutine_enter(co, &cco); | |
389 | while (cco.ret == NOT_DONE) { | |
390 | qemu_aio_wait(); | |
391 | } | |
392 | } | |
393 | ||
394 | ret = cco.ret; | |
395 | g_free(cco.filename); | |
0e7e1989 | 396 | |
5b7e1542 | 397 | return ret; |
ea2384d3 FB |
398 | } |
399 | ||
84a12e66 CH |
400 | int bdrv_create_file(const char* filename, QEMUOptionParameter *options) |
401 | { | |
402 | BlockDriver *drv; | |
403 | ||
b50cbabc | 404 | drv = bdrv_find_protocol(filename); |
84a12e66 | 405 | if (drv == NULL) { |
16905d71 | 406 | return -ENOENT; |
84a12e66 CH |
407 | } |
408 | ||
409 | return bdrv_create(drv, filename, options); | |
410 | } | |
411 | ||
eba25057 JM |
412 | /* |
413 | * Create a uniquely-named empty temporary file. | |
414 | * Return 0 upon success, otherwise a negative errno value. | |
415 | */ | |
416 | int get_tmp_filename(char *filename, int size) | |
d5249393 | 417 | { |
eba25057 | 418 | #ifdef _WIN32 |
3b9f94e1 | 419 | char temp_dir[MAX_PATH]; |
eba25057 JM |
420 | /* GetTempFileName requires that its output buffer (4th param) |
421 | have length MAX_PATH or greater. */ | |
422 | assert(size >= MAX_PATH); | |
423 | return (GetTempPath(MAX_PATH, temp_dir) | |
424 | && GetTempFileName(temp_dir, "qem", 0, filename) | |
425 | ? 0 : -GetLastError()); | |
d5249393 | 426 | #else |
67b915a5 | 427 | int fd; |
7ccfb2eb | 428 | const char *tmpdir; |
0badc1ee AJ |
429 | tmpdir = getenv("TMPDIR"); |
430 | if (!tmpdir) | |
431 | tmpdir = "/tmp"; | |
eba25057 JM |
432 | if (snprintf(filename, size, "%s/vl.XXXXXX", tmpdir) >= size) { |
433 | return -EOVERFLOW; | |
434 | } | |
ea2384d3 | 435 | fd = mkstemp(filename); |
eba25057 JM |
436 | if (fd < 0 || close(fd)) { |
437 | return -errno; | |
438 | } | |
439 | return 0; | |
d5249393 | 440 | #endif |
eba25057 | 441 | } |
fc01f7e7 | 442 | |
84a12e66 CH |
443 | /* |
444 | * Detect host devices. By convention, /dev/cdrom[N] is always | |
445 | * recognized as a host CDROM. | |
446 | */ | |
447 | static BlockDriver *find_hdev_driver(const char *filename) | |
448 | { | |
449 | int score_max = 0, score; | |
450 | BlockDriver *drv = NULL, *d; | |
451 | ||
452 | QLIST_FOREACH(d, &bdrv_drivers, list) { | |
453 | if (d->bdrv_probe_device) { | |
454 | score = d->bdrv_probe_device(filename); | |
455 | if (score > score_max) { | |
456 | score_max = score; | |
457 | drv = d; | |
458 | } | |
459 | } | |
460 | } | |
461 | ||
462 | return drv; | |
463 | } | |
464 | ||
b50cbabc | 465 | BlockDriver *bdrv_find_protocol(const char *filename) |
83f64091 FB |
466 | { |
467 | BlockDriver *drv1; | |
468 | char protocol[128]; | |
1cec71e3 | 469 | int len; |
83f64091 | 470 | const char *p; |
19cb3738 | 471 | |
66f82cee KW |
472 | /* TODO Drivers without bdrv_file_open must be specified explicitly */ |
473 | ||
39508e7a CH |
474 | /* |
475 | * XXX(hch): we really should not let host device detection | |
476 | * override an explicit protocol specification, but moving this | |
477 | * later breaks access to device names with colons in them. | |
478 | * Thanks to the brain-dead persistent naming schemes on udev- | |
479 | * based Linux systems those actually are quite common. | |
480 | */ | |
481 | drv1 = find_hdev_driver(filename); | |
482 | if (drv1) { | |
483 | return drv1; | |
484 | } | |
485 | ||
9e0b22f4 | 486 | if (!path_has_protocol(filename)) { |
39508e7a | 487 | return bdrv_find_format("file"); |
84a12e66 | 488 | } |
9e0b22f4 SH |
489 | p = strchr(filename, ':'); |
490 | assert(p != NULL); | |
1cec71e3 AL |
491 | len = p - filename; |
492 | if (len > sizeof(protocol) - 1) | |
493 | len = sizeof(protocol) - 1; | |
494 | memcpy(protocol, filename, len); | |
495 | protocol[len] = '\0'; | |
8a22f02a | 496 | QLIST_FOREACH(drv1, &bdrv_drivers, list) { |
5fafdf24 | 497 | if (drv1->protocol_name && |
8a22f02a | 498 | !strcmp(drv1->protocol_name, protocol)) { |
83f64091 | 499 | return drv1; |
8a22f02a | 500 | } |
83f64091 FB |
501 | } |
502 | return NULL; | |
503 | } | |
504 | ||
c98ac35d | 505 | static int find_image_format(const char *filename, BlockDriver **pdrv) |
f3a5d3f8 CH |
506 | { |
507 | int ret, score, score_max; | |
508 | BlockDriver *drv1, *drv; | |
509 | uint8_t buf[2048]; | |
510 | BlockDriverState *bs; | |
511 | ||
f5edb014 | 512 | ret = bdrv_file_open(&bs, filename, 0); |
c98ac35d SW |
513 | if (ret < 0) { |
514 | *pdrv = NULL; | |
515 | return ret; | |
516 | } | |
f8ea0b00 | 517 | |
08a00559 KW |
518 | /* Return the raw BlockDriver * to scsi-generic devices or empty drives */ |
519 | if (bs->sg || !bdrv_is_inserted(bs)) { | |
1a396859 | 520 | bdrv_delete(bs); |
c98ac35d SW |
521 | drv = bdrv_find_format("raw"); |
522 | if (!drv) { | |
523 | ret = -ENOENT; | |
524 | } | |
525 | *pdrv = drv; | |
526 | return ret; | |
1a396859 | 527 | } |
f8ea0b00 | 528 | |
83f64091 FB |
529 | ret = bdrv_pread(bs, 0, buf, sizeof(buf)); |
530 | bdrv_delete(bs); | |
531 | if (ret < 0) { | |
c98ac35d SW |
532 | *pdrv = NULL; |
533 | return ret; | |
83f64091 FB |
534 | } |
535 | ||
ea2384d3 | 536 | score_max = 0; |
84a12e66 | 537 | drv = NULL; |
8a22f02a | 538 | QLIST_FOREACH(drv1, &bdrv_drivers, list) { |
83f64091 FB |
539 | if (drv1->bdrv_probe) { |
540 | score = drv1->bdrv_probe(buf, ret, filename); | |
541 | if (score > score_max) { | |
542 | score_max = score; | |
543 | drv = drv1; | |
544 | } | |
0849bf08 | 545 | } |
fc01f7e7 | 546 | } |
c98ac35d SW |
547 | if (!drv) { |
548 | ret = -ENOENT; | |
549 | } | |
550 | *pdrv = drv; | |
551 | return ret; | |
ea2384d3 FB |
552 | } |
553 | ||
51762288 SH |
554 | /** |
555 | * Set the current 'total_sectors' value | |
556 | */ | |
557 | static int refresh_total_sectors(BlockDriverState *bs, int64_t hint) | |
558 | { | |
559 | BlockDriver *drv = bs->drv; | |
560 | ||
396759ad NB |
561 | /* Do not attempt drv->bdrv_getlength() on scsi-generic devices */ |
562 | if (bs->sg) | |
563 | return 0; | |
564 | ||
51762288 SH |
565 | /* query actual device if possible, otherwise just trust the hint */ |
566 | if (drv->bdrv_getlength) { | |
567 | int64_t length = drv->bdrv_getlength(bs); | |
568 | if (length < 0) { | |
569 | return length; | |
570 | } | |
571 | hint = length >> BDRV_SECTOR_BITS; | |
572 | } | |
573 | ||
574 | bs->total_sectors = hint; | |
575 | return 0; | |
576 | } | |
577 | ||
c3993cdc SH |
578 | /** |
579 | * Set open flags for a given cache mode | |
580 | * | |
581 | * Return 0 on success, -1 if the cache mode was invalid. | |
582 | */ | |
583 | int bdrv_parse_cache_flags(const char *mode, int *flags) | |
584 | { | |
585 | *flags &= ~BDRV_O_CACHE_MASK; | |
586 | ||
587 | if (!strcmp(mode, "off") || !strcmp(mode, "none")) { | |
588 | *flags |= BDRV_O_NOCACHE | BDRV_O_CACHE_WB; | |
92196b2f SH |
589 | } else if (!strcmp(mode, "directsync")) { |
590 | *flags |= BDRV_O_NOCACHE; | |
c3993cdc SH |
591 | } else if (!strcmp(mode, "writeback")) { |
592 | *flags |= BDRV_O_CACHE_WB; | |
593 | } else if (!strcmp(mode, "unsafe")) { | |
594 | *flags |= BDRV_O_CACHE_WB; | |
595 | *flags |= BDRV_O_NO_FLUSH; | |
596 | } else if (!strcmp(mode, "writethrough")) { | |
597 | /* this is the default */ | |
598 | } else { | |
599 | return -1; | |
600 | } | |
601 | ||
602 | return 0; | |
603 | } | |
604 | ||
53fec9d3 SH |
605 | /** |
606 | * The copy-on-read flag is actually a reference count so multiple users may | |
607 | * use the feature without worrying about clobbering its previous state. | |
608 | * Copy-on-read stays enabled until all users have called to disable it. | |
609 | */ | |
610 | void bdrv_enable_copy_on_read(BlockDriverState *bs) | |
611 | { | |
612 | bs->copy_on_read++; | |
613 | } | |
614 | ||
615 | void bdrv_disable_copy_on_read(BlockDriverState *bs) | |
616 | { | |
617 | assert(bs->copy_on_read > 0); | |
618 | bs->copy_on_read--; | |
619 | } | |
620 | ||
57915332 KW |
621 | /* |
622 | * Common part for opening disk images and files | |
623 | */ | |
624 | static int bdrv_open_common(BlockDriverState *bs, const char *filename, | |
625 | int flags, BlockDriver *drv) | |
626 | { | |
627 | int ret, open_flags; | |
628 | ||
629 | assert(drv != NULL); | |
6405875c | 630 | assert(bs->file == NULL); |
57915332 | 631 | |
28dcee10 SH |
632 | trace_bdrv_open_common(bs, filename, flags, drv->format_name); |
633 | ||
57915332 | 634 | bs->open_flags = flags; |
57915332 KW |
635 | bs->buffer_alignment = 512; |
636 | ||
53fec9d3 SH |
637 | assert(bs->copy_on_read == 0); /* bdrv_new() and bdrv_close() make it so */ |
638 | if ((flags & BDRV_O_RDWR) && (flags & BDRV_O_COPY_ON_READ)) { | |
639 | bdrv_enable_copy_on_read(bs); | |
640 | } | |
641 | ||
57915332 KW |
642 | pstrcpy(bs->filename, sizeof(bs->filename), filename); |
643 | ||
644 | if (use_bdrv_whitelist && !bdrv_is_whitelisted(drv)) { | |
645 | return -ENOTSUP; | |
646 | } | |
647 | ||
648 | bs->drv = drv; | |
7267c094 | 649 | bs->opaque = g_malloc0(drv->instance_size); |
57915332 | 650 | |
03f541bd | 651 | bs->enable_write_cache = !!(flags & BDRV_O_CACHE_WB); |
57915332 KW |
652 | |
653 | /* | |
654 | * Clear flags that are internal to the block layer before opening the | |
655 | * image. | |
656 | */ | |
657 | open_flags = flags & ~(BDRV_O_SNAPSHOT | BDRV_O_NO_BACKING); | |
658 | ||
659 | /* | |
ebabb67a | 660 | * Snapshots should be writable. |
57915332 KW |
661 | */ |
662 | if (bs->is_temporary) { | |
663 | open_flags |= BDRV_O_RDWR; | |
664 | } | |
665 | ||
e7c63796 SH |
666 | bs->keep_read_only = bs->read_only = !(open_flags & BDRV_O_RDWR); |
667 | ||
66f82cee KW |
668 | /* Open the image, either directly or using a protocol */ |
669 | if (drv->bdrv_file_open) { | |
670 | ret = drv->bdrv_file_open(bs, filename, open_flags); | |
671 | } else { | |
672 | ret = bdrv_file_open(&bs->file, filename, open_flags); | |
673 | if (ret >= 0) { | |
674 | ret = drv->bdrv_open(bs, open_flags); | |
675 | } | |
676 | } | |
677 | ||
57915332 KW |
678 | if (ret < 0) { |
679 | goto free_and_fail; | |
680 | } | |
681 | ||
51762288 SH |
682 | ret = refresh_total_sectors(bs, bs->total_sectors); |
683 | if (ret < 0) { | |
684 | goto free_and_fail; | |
57915332 | 685 | } |
51762288 | 686 | |
57915332 KW |
687 | #ifndef _WIN32 |
688 | if (bs->is_temporary) { | |
689 | unlink(filename); | |
690 | } | |
691 | #endif | |
692 | return 0; | |
693 | ||
694 | free_and_fail: | |
66f82cee KW |
695 | if (bs->file) { |
696 | bdrv_delete(bs->file); | |
697 | bs->file = NULL; | |
698 | } | |
7267c094 | 699 | g_free(bs->opaque); |
57915332 KW |
700 | bs->opaque = NULL; |
701 | bs->drv = NULL; | |
702 | return ret; | |
703 | } | |
704 | ||
b6ce07aa KW |
705 | /* |
706 | * Opens a file using a protocol (file, host_device, nbd, ...) | |
707 | */ | |
83f64091 | 708 | int bdrv_file_open(BlockDriverState **pbs, const char *filename, int flags) |
ea2384d3 | 709 | { |
83f64091 | 710 | BlockDriverState *bs; |
6db95603 | 711 | BlockDriver *drv; |
83f64091 FB |
712 | int ret; |
713 | ||
b50cbabc | 714 | drv = bdrv_find_protocol(filename); |
6db95603 CH |
715 | if (!drv) { |
716 | return -ENOENT; | |
717 | } | |
718 | ||
83f64091 | 719 | bs = bdrv_new(""); |
b6ce07aa | 720 | ret = bdrv_open_common(bs, filename, flags, drv); |
83f64091 FB |
721 | if (ret < 0) { |
722 | bdrv_delete(bs); | |
723 | return ret; | |
3b0d4f61 | 724 | } |
71d0770c | 725 | bs->growable = 1; |
83f64091 FB |
726 | *pbs = bs; |
727 | return 0; | |
728 | } | |
729 | ||
b6ce07aa KW |
730 | /* |
731 | * Opens a disk image (raw, qcow2, vmdk, ...) | |
732 | */ | |
d6e9098e KW |
733 | int bdrv_open(BlockDriverState *bs, const char *filename, int flags, |
734 | BlockDriver *drv) | |
ea2384d3 | 735 | { |
b6ce07aa | 736 | int ret; |
2b572816 | 737 | char tmp_filename[PATH_MAX]; |
712e7874 | 738 | |
83f64091 | 739 | if (flags & BDRV_O_SNAPSHOT) { |
ea2384d3 FB |
740 | BlockDriverState *bs1; |
741 | int64_t total_size; | |
7c96d46e | 742 | int is_protocol = 0; |
91a073a9 KW |
743 | BlockDriver *bdrv_qcow2; |
744 | QEMUOptionParameter *options; | |
b6ce07aa | 745 | char backing_filename[PATH_MAX]; |
3b46e624 | 746 | |
ea2384d3 FB |
747 | /* if snapshot, we create a temporary backing file and open it |
748 | instead of opening 'filename' directly */ | |
33e3963e | 749 | |
ea2384d3 FB |
750 | /* if there is a backing file, use it */ |
751 | bs1 = bdrv_new(""); | |
d6e9098e | 752 | ret = bdrv_open(bs1, filename, 0, drv); |
51d7c00c | 753 | if (ret < 0) { |
ea2384d3 | 754 | bdrv_delete(bs1); |
51d7c00c | 755 | return ret; |
ea2384d3 | 756 | } |
3e82990b | 757 | total_size = bdrv_getlength(bs1) & BDRV_SECTOR_MASK; |
7c96d46e AL |
758 | |
759 | if (bs1->drv && bs1->drv->protocol_name) | |
760 | is_protocol = 1; | |
761 | ||
ea2384d3 | 762 | bdrv_delete(bs1); |
3b46e624 | 763 | |
eba25057 JM |
764 | ret = get_tmp_filename(tmp_filename, sizeof(tmp_filename)); |
765 | if (ret < 0) { | |
766 | return ret; | |
767 | } | |
7c96d46e AL |
768 | |
769 | /* Real path is meaningless for protocols */ | |
770 | if (is_protocol) | |
771 | snprintf(backing_filename, sizeof(backing_filename), | |
772 | "%s", filename); | |
114cdfa9 KS |
773 | else if (!realpath(filename, backing_filename)) |
774 | return -errno; | |
7c96d46e | 775 | |
91a073a9 KW |
776 | bdrv_qcow2 = bdrv_find_format("qcow2"); |
777 | options = parse_option_parameters("", bdrv_qcow2->create_options, NULL); | |
778 | ||
3e82990b | 779 | set_option_parameter_int(options, BLOCK_OPT_SIZE, total_size); |
91a073a9 KW |
780 | set_option_parameter(options, BLOCK_OPT_BACKING_FILE, backing_filename); |
781 | if (drv) { | |
782 | set_option_parameter(options, BLOCK_OPT_BACKING_FMT, | |
783 | drv->format_name); | |
784 | } | |
785 | ||
786 | ret = bdrv_create(bdrv_qcow2, tmp_filename, options); | |
d748768c | 787 | free_option_parameters(options); |
51d7c00c AL |
788 | if (ret < 0) { |
789 | return ret; | |
ea2384d3 | 790 | } |
91a073a9 | 791 | |
ea2384d3 | 792 | filename = tmp_filename; |
91a073a9 | 793 | drv = bdrv_qcow2; |
ea2384d3 FB |
794 | bs->is_temporary = 1; |
795 | } | |
712e7874 | 796 | |
b6ce07aa | 797 | /* Find the right image format driver */ |
6db95603 | 798 | if (!drv) { |
c98ac35d | 799 | ret = find_image_format(filename, &drv); |
51d7c00c | 800 | } |
6987307c | 801 | |
51d7c00c | 802 | if (!drv) { |
51d7c00c | 803 | goto unlink_and_fail; |
ea2384d3 | 804 | } |
b6ce07aa KW |
805 | |
806 | /* Open the image */ | |
807 | ret = bdrv_open_common(bs, filename, flags, drv); | |
808 | if (ret < 0) { | |
6987307c CH |
809 | goto unlink_and_fail; |
810 | } | |
811 | ||
b6ce07aa KW |
812 | /* If there is a backing file, use it */ |
813 | if ((flags & BDRV_O_NO_BACKING) == 0 && bs->backing_file[0] != '\0') { | |
814 | char backing_filename[PATH_MAX]; | |
815 | int back_flags; | |
816 | BlockDriver *back_drv = NULL; | |
817 | ||
818 | bs->backing_hd = bdrv_new(""); | |
dc5a1371 PB |
819 | bdrv_get_full_backing_filename(bs, backing_filename, |
820 | sizeof(backing_filename)); | |
df2dbb4a SH |
821 | |
822 | if (bs->backing_format[0] != '\0') { | |
b6ce07aa | 823 | back_drv = bdrv_find_format(bs->backing_format); |
df2dbb4a | 824 | } |
b6ce07aa KW |
825 | |
826 | /* backing files always opened read-only */ | |
827 | back_flags = | |
828 | flags & ~(BDRV_O_RDWR | BDRV_O_SNAPSHOT | BDRV_O_NO_BACKING); | |
829 | ||
830 | ret = bdrv_open(bs->backing_hd, backing_filename, back_flags, back_drv); | |
831 | if (ret < 0) { | |
832 | bdrv_close(bs); | |
833 | return ret; | |
834 | } | |
835 | if (bs->is_temporary) { | |
836 | bs->backing_hd->keep_read_only = !(flags & BDRV_O_RDWR); | |
837 | } else { | |
838 | /* base image inherits from "parent" */ | |
839 | bs->backing_hd->keep_read_only = bs->keep_read_only; | |
840 | } | |
841 | } | |
842 | ||
843 | if (!bdrv_key_required(bs)) { | |
7d4b4ba5 | 844 | bdrv_dev_change_media_cb(bs, true); |
b6ce07aa KW |
845 | } |
846 | ||
98f90dba ZYW |
847 | /* throttling disk I/O limits */ |
848 | if (bs->io_limits_enabled) { | |
849 | bdrv_io_limits_enable(bs); | |
850 | } | |
851 | ||
b6ce07aa KW |
852 | return 0; |
853 | ||
854 | unlink_and_fail: | |
855 | if (bs->is_temporary) { | |
856 | unlink(filename); | |
857 | } | |
858 | return ret; | |
859 | } | |
860 | ||
fc01f7e7 FB |
861 | void bdrv_close(BlockDriverState *bs) |
862 | { | |
80ccf93b | 863 | bdrv_flush(bs); |
19cb3738 | 864 | if (bs->drv) { |
3e914655 PB |
865 | if (bs->job) { |
866 | block_job_cancel_sync(bs->job); | |
867 | } | |
7094f12f KW |
868 | bdrv_drain_all(); |
869 | ||
f9092b10 MA |
870 | if (bs == bs_snapshots) { |
871 | bs_snapshots = NULL; | |
872 | } | |
557df6ac | 873 | if (bs->backing_hd) { |
ea2384d3 | 874 | bdrv_delete(bs->backing_hd); |
557df6ac SH |
875 | bs->backing_hd = NULL; |
876 | } | |
ea2384d3 | 877 | bs->drv->bdrv_close(bs); |
7267c094 | 878 | g_free(bs->opaque); |
ea2384d3 FB |
879 | #ifdef _WIN32 |
880 | if (bs->is_temporary) { | |
881 | unlink(bs->filename); | |
882 | } | |
67b915a5 | 883 | #endif |
ea2384d3 FB |
884 | bs->opaque = NULL; |
885 | bs->drv = NULL; | |
53fec9d3 | 886 | bs->copy_on_read = 0; |
a275fa42 PB |
887 | bs->backing_file[0] = '\0'; |
888 | bs->backing_format[0] = '\0'; | |
6405875c PB |
889 | bs->total_sectors = 0; |
890 | bs->encrypted = 0; | |
891 | bs->valid_key = 0; | |
892 | bs->sg = 0; | |
893 | bs->growable = 0; | |
b338082b | 894 | |
66f82cee | 895 | if (bs->file != NULL) { |
0ac9377d PB |
896 | bdrv_delete(bs->file); |
897 | bs->file = NULL; | |
66f82cee KW |
898 | } |
899 | ||
7d4b4ba5 | 900 | bdrv_dev_change_media_cb(bs, false); |
b338082b | 901 | } |
98f90dba ZYW |
902 | |
903 | /*throttling disk I/O limits*/ | |
904 | if (bs->io_limits_enabled) { | |
905 | bdrv_io_limits_disable(bs); | |
906 | } | |
b338082b FB |
907 | } |
908 | ||
2bc93fed MK |
909 | void bdrv_close_all(void) |
910 | { | |
911 | BlockDriverState *bs; | |
912 | ||
913 | QTAILQ_FOREACH(bs, &bdrv_states, list) { | |
914 | bdrv_close(bs); | |
915 | } | |
916 | } | |
917 | ||
922453bc SH |
918 | /* |
919 | * Wait for pending requests to complete across all BlockDriverStates | |
920 | * | |
921 | * This function does not flush data to disk, use bdrv_flush_all() for that | |
922 | * after calling this function. | |
4c355d53 ZYW |
923 | * |
924 | * Note that completion of an asynchronous I/O operation can trigger any | |
925 | * number of other I/O operations on other devices---for example a coroutine | |
926 | * can be arbitrarily complex and a constant flow of I/O can come until the | |
927 | * coroutine is complete. Because of this, it is not possible to have a | |
928 | * function to drain a single device's I/O queue. | |
922453bc SH |
929 | */ |
930 | void bdrv_drain_all(void) | |
931 | { | |
932 | BlockDriverState *bs; | |
4c355d53 ZYW |
933 | bool busy; |
934 | ||
935 | do { | |
936 | busy = qemu_aio_wait(); | |
922453bc | 937 | |
4c355d53 ZYW |
938 | /* FIXME: We do not have timer support here, so this is effectively |
939 | * a busy wait. | |
940 | */ | |
941 | QTAILQ_FOREACH(bs, &bdrv_states, list) { | |
942 | if (!qemu_co_queue_empty(&bs->throttled_reqs)) { | |
943 | qemu_co_queue_restart_all(&bs->throttled_reqs); | |
944 | busy = true; | |
945 | } | |
946 | } | |
947 | } while (busy); | |
922453bc SH |
948 | |
949 | /* If requests are still pending there is a bug somewhere */ | |
950 | QTAILQ_FOREACH(bs, &bdrv_states, list) { | |
951 | assert(QLIST_EMPTY(&bs->tracked_requests)); | |
952 | assert(qemu_co_queue_empty(&bs->throttled_reqs)); | |
953 | } | |
954 | } | |
955 | ||
d22b2f41 RH |
956 | /* make a BlockDriverState anonymous by removing from bdrv_state list. |
957 | Also, NULL terminate the device_name to prevent double remove */ | |
958 | void bdrv_make_anon(BlockDriverState *bs) | |
959 | { | |
960 | if (bs->device_name[0] != '\0') { | |
961 | QTAILQ_REMOVE(&bdrv_states, bs, list); | |
962 | } | |
963 | bs->device_name[0] = '\0'; | |
964 | } | |
965 | ||
e023b2e2 PB |
966 | static void bdrv_rebind(BlockDriverState *bs) |
967 | { | |
968 | if (bs->drv && bs->drv->bdrv_rebind) { | |
969 | bs->drv->bdrv_rebind(bs); | |
970 | } | |
971 | } | |
972 | ||
8802d1fd JC |
973 | /* |
974 | * Add new bs contents at the top of an image chain while the chain is | |
975 | * live, while keeping required fields on the top layer. | |
976 | * | |
977 | * This will modify the BlockDriverState fields, and swap contents | |
978 | * between bs_new and bs_top. Both bs_new and bs_top are modified. | |
979 | * | |
f6801b83 JC |
980 | * bs_new is required to be anonymous. |
981 | * | |
8802d1fd JC |
982 | * This function does not create any image files. |
983 | */ | |
984 | void bdrv_append(BlockDriverState *bs_new, BlockDriverState *bs_top) | |
985 | { | |
986 | BlockDriverState tmp; | |
987 | ||
f6801b83 JC |
988 | /* bs_new must be anonymous */ |
989 | assert(bs_new->device_name[0] == '\0'); | |
8802d1fd JC |
990 | |
991 | tmp = *bs_new; | |
992 | ||
993 | /* there are some fields that need to stay on the top layer: */ | |
3a389e79 | 994 | tmp.open_flags = bs_top->open_flags; |
8802d1fd JC |
995 | |
996 | /* dev info */ | |
997 | tmp.dev_ops = bs_top->dev_ops; | |
998 | tmp.dev_opaque = bs_top->dev_opaque; | |
999 | tmp.dev = bs_top->dev; | |
1000 | tmp.buffer_alignment = bs_top->buffer_alignment; | |
1001 | tmp.copy_on_read = bs_top->copy_on_read; | |
1002 | ||
c4a248a1 PB |
1003 | tmp.enable_write_cache = bs_top->enable_write_cache; |
1004 | ||
8802d1fd JC |
1005 | /* i/o timing parameters */ |
1006 | tmp.slice_time = bs_top->slice_time; | |
1007 | tmp.slice_start = bs_top->slice_start; | |
1008 | tmp.slice_end = bs_top->slice_end; | |
1009 | tmp.io_limits = bs_top->io_limits; | |
1010 | tmp.io_base = bs_top->io_base; | |
1011 | tmp.throttled_reqs = bs_top->throttled_reqs; | |
1012 | tmp.block_timer = bs_top->block_timer; | |
1013 | tmp.io_limits_enabled = bs_top->io_limits_enabled; | |
1014 | ||
1015 | /* geometry */ | |
1016 | tmp.cyls = bs_top->cyls; | |
1017 | tmp.heads = bs_top->heads; | |
1018 | tmp.secs = bs_top->secs; | |
1019 | tmp.translation = bs_top->translation; | |
1020 | ||
1021 | /* r/w error */ | |
1022 | tmp.on_read_error = bs_top->on_read_error; | |
1023 | tmp.on_write_error = bs_top->on_write_error; | |
1024 | ||
1025 | /* i/o status */ | |
1026 | tmp.iostatus_enabled = bs_top->iostatus_enabled; | |
1027 | tmp.iostatus = bs_top->iostatus; | |
1028 | ||
1029 | /* keep the same entry in bdrv_states */ | |
1030 | pstrcpy(tmp.device_name, sizeof(tmp.device_name), bs_top->device_name); | |
1031 | tmp.list = bs_top->list; | |
1032 | ||
1033 | /* The contents of 'tmp' will become bs_top, as we are | |
1034 | * swapping bs_new and bs_top contents. */ | |
1035 | tmp.backing_hd = bs_new; | |
1036 | pstrcpy(tmp.backing_file, sizeof(tmp.backing_file), bs_top->filename); | |
f6801b83 | 1037 | bdrv_get_format(bs_top, tmp.backing_format, sizeof(tmp.backing_format)); |
8802d1fd JC |
1038 | |
1039 | /* swap contents of the fixed new bs and the current top */ | |
1040 | *bs_new = *bs_top; | |
1041 | *bs_top = tmp; | |
1042 | ||
f6801b83 JC |
1043 | /* device_name[] was carried over from the old bs_top. bs_new |
1044 | * shouldn't be in bdrv_states, so we need to make device_name[] | |
1045 | * reflect the anonymity of bs_new | |
1046 | */ | |
1047 | bs_new->device_name[0] = '\0'; | |
1048 | ||
8802d1fd JC |
1049 | /* clear the copied fields in the new backing file */ |
1050 | bdrv_detach_dev(bs_new, bs_new->dev); | |
1051 | ||
1052 | qemu_co_queue_init(&bs_new->throttled_reqs); | |
1053 | memset(&bs_new->io_base, 0, sizeof(bs_new->io_base)); | |
1054 | memset(&bs_new->io_limits, 0, sizeof(bs_new->io_limits)); | |
1055 | bdrv_iostatus_disable(bs_new); | |
1056 | ||
1057 | /* we don't use bdrv_io_limits_disable() for this, because we don't want | |
1058 | * to affect or delete the block_timer, as it has been moved to bs_top */ | |
1059 | bs_new->io_limits_enabled = false; | |
1060 | bs_new->block_timer = NULL; | |
1061 | bs_new->slice_time = 0; | |
1062 | bs_new->slice_start = 0; | |
1063 | bs_new->slice_end = 0; | |
e023b2e2 PB |
1064 | |
1065 | bdrv_rebind(bs_new); | |
1066 | bdrv_rebind(bs_top); | |
8802d1fd JC |
1067 | } |
1068 | ||
b338082b FB |
1069 | void bdrv_delete(BlockDriverState *bs) |
1070 | { | |
fa879d62 | 1071 | assert(!bs->dev); |
3e914655 PB |
1072 | assert(!bs->job); |
1073 | assert(!bs->in_use); | |
18846dee | 1074 | |
1b7bdbc1 | 1075 | /* remove from list, if necessary */ |
d22b2f41 | 1076 | bdrv_make_anon(bs); |
34c6f050 | 1077 | |
b338082b | 1078 | bdrv_close(bs); |
66f82cee | 1079 | |
f9092b10 | 1080 | assert(bs != bs_snapshots); |
7267c094 | 1081 | g_free(bs); |
fc01f7e7 FB |
1082 | } |
1083 | ||
fa879d62 MA |
1084 | int bdrv_attach_dev(BlockDriverState *bs, void *dev) |
1085 | /* TODO change to DeviceState *dev when all users are qdevified */ | |
18846dee | 1086 | { |
fa879d62 | 1087 | if (bs->dev) { |
18846dee MA |
1088 | return -EBUSY; |
1089 | } | |
fa879d62 | 1090 | bs->dev = dev; |
28a7282a | 1091 | bdrv_iostatus_reset(bs); |
18846dee MA |
1092 | return 0; |
1093 | } | |
1094 | ||
fa879d62 MA |
1095 | /* TODO qdevified devices don't use this, remove when devices are qdevified */ |
1096 | void bdrv_attach_dev_nofail(BlockDriverState *bs, void *dev) | |
18846dee | 1097 | { |
fa879d62 MA |
1098 | if (bdrv_attach_dev(bs, dev) < 0) { |
1099 | abort(); | |
1100 | } | |
1101 | } | |
1102 | ||
1103 | void bdrv_detach_dev(BlockDriverState *bs, void *dev) | |
1104 | /* TODO change to DeviceState *dev when all users are qdevified */ | |
1105 | { | |
1106 | assert(bs->dev == dev); | |
1107 | bs->dev = NULL; | |
0e49de52 MA |
1108 | bs->dev_ops = NULL; |
1109 | bs->dev_opaque = NULL; | |
29e05f20 | 1110 | bs->buffer_alignment = 512; |
18846dee MA |
1111 | } |
1112 | ||
fa879d62 MA |
1113 | /* TODO change to return DeviceState * when all users are qdevified */ |
1114 | void *bdrv_get_attached_dev(BlockDriverState *bs) | |
18846dee | 1115 | { |
fa879d62 | 1116 | return bs->dev; |
18846dee MA |
1117 | } |
1118 | ||
0e49de52 MA |
1119 | void bdrv_set_dev_ops(BlockDriverState *bs, const BlockDevOps *ops, |
1120 | void *opaque) | |
1121 | { | |
1122 | bs->dev_ops = ops; | |
1123 | bs->dev_opaque = opaque; | |
2c6942fa MA |
1124 | if (bdrv_dev_has_removable_media(bs) && bs == bs_snapshots) { |
1125 | bs_snapshots = NULL; | |
1126 | } | |
0e49de52 MA |
1127 | } |
1128 | ||
329c0a48 LC |
1129 | void bdrv_emit_qmp_error_event(const BlockDriverState *bdrv, |
1130 | BlockQMPEventAction action, int is_read) | |
1131 | { | |
1132 | QObject *data; | |
1133 | const char *action_str; | |
1134 | ||
1135 | switch (action) { | |
1136 | case BDRV_ACTION_REPORT: | |
1137 | action_str = "report"; | |
1138 | break; | |
1139 | case BDRV_ACTION_IGNORE: | |
1140 | action_str = "ignore"; | |
1141 | break; | |
1142 | case BDRV_ACTION_STOP: | |
1143 | action_str = "stop"; | |
1144 | break; | |
1145 | default: | |
1146 | abort(); | |
1147 | } | |
1148 | ||
1149 | data = qobject_from_jsonf("{ 'device': %s, 'action': %s, 'operation': %s }", | |
1150 | bdrv->device_name, | |
1151 | action_str, | |
1152 | is_read ? "read" : "write"); | |
1153 | monitor_protocol_event(QEVENT_BLOCK_IO_ERROR, data); | |
1154 | ||
1155 | qobject_decref(data); | |
1156 | } | |
1157 | ||
6f382ed2 LC |
1158 | static void bdrv_emit_qmp_eject_event(BlockDriverState *bs, bool ejected) |
1159 | { | |
1160 | QObject *data; | |
1161 | ||
1162 | data = qobject_from_jsonf("{ 'device': %s, 'tray-open': %i }", | |
1163 | bdrv_get_device_name(bs), ejected); | |
1164 | monitor_protocol_event(QEVENT_DEVICE_TRAY_MOVED, data); | |
1165 | ||
1166 | qobject_decref(data); | |
1167 | } | |
1168 | ||
7d4b4ba5 | 1169 | static void bdrv_dev_change_media_cb(BlockDriverState *bs, bool load) |
0e49de52 | 1170 | { |
145feb17 | 1171 | if (bs->dev_ops && bs->dev_ops->change_media_cb) { |
6f382ed2 | 1172 | bool tray_was_closed = !bdrv_dev_is_tray_open(bs); |
7d4b4ba5 | 1173 | bs->dev_ops->change_media_cb(bs->dev_opaque, load); |
6f382ed2 LC |
1174 | if (tray_was_closed) { |
1175 | /* tray open */ | |
1176 | bdrv_emit_qmp_eject_event(bs, true); | |
1177 | } | |
1178 | if (load) { | |
1179 | /* tray close */ | |
1180 | bdrv_emit_qmp_eject_event(bs, false); | |
1181 | } | |
145feb17 MA |
1182 | } |
1183 | } | |
1184 | ||
2c6942fa MA |
1185 | bool bdrv_dev_has_removable_media(BlockDriverState *bs) |
1186 | { | |
1187 | return !bs->dev || (bs->dev_ops && bs->dev_ops->change_media_cb); | |
1188 | } | |
1189 | ||
025ccaa7 PB |
1190 | void bdrv_dev_eject_request(BlockDriverState *bs, bool force) |
1191 | { | |
1192 | if (bs->dev_ops && bs->dev_ops->eject_request_cb) { | |
1193 | bs->dev_ops->eject_request_cb(bs->dev_opaque, force); | |
1194 | } | |
1195 | } | |
1196 | ||
e4def80b MA |
1197 | bool bdrv_dev_is_tray_open(BlockDriverState *bs) |
1198 | { | |
1199 | if (bs->dev_ops && bs->dev_ops->is_tray_open) { | |
1200 | return bs->dev_ops->is_tray_open(bs->dev_opaque); | |
1201 | } | |
1202 | return false; | |
1203 | } | |
1204 | ||
145feb17 MA |
1205 | static void bdrv_dev_resize_cb(BlockDriverState *bs) |
1206 | { | |
1207 | if (bs->dev_ops && bs->dev_ops->resize_cb) { | |
1208 | bs->dev_ops->resize_cb(bs->dev_opaque); | |
0e49de52 MA |
1209 | } |
1210 | } | |
1211 | ||
f107639a MA |
1212 | bool bdrv_dev_is_medium_locked(BlockDriverState *bs) |
1213 | { | |
1214 | if (bs->dev_ops && bs->dev_ops->is_medium_locked) { | |
1215 | return bs->dev_ops->is_medium_locked(bs->dev_opaque); | |
1216 | } | |
1217 | return false; | |
1218 | } | |
1219 | ||
e97fc193 AL |
1220 | /* |
1221 | * Run consistency checks on an image | |
1222 | * | |
e076f338 | 1223 | * Returns 0 if the check could be completed (it doesn't mean that the image is |
a1c7273b | 1224 | * free of errors) or -errno when an internal error occurred. The results of the |
e076f338 | 1225 | * check are stored in res. |
e97fc193 | 1226 | */ |
4534ff54 | 1227 | int bdrv_check(BlockDriverState *bs, BdrvCheckResult *res, BdrvCheckMode fix) |
e97fc193 AL |
1228 | { |
1229 | if (bs->drv->bdrv_check == NULL) { | |
1230 | return -ENOTSUP; | |
1231 | } | |
1232 | ||
e076f338 | 1233 | memset(res, 0, sizeof(*res)); |
4534ff54 | 1234 | return bs->drv->bdrv_check(bs, res, fix); |
e97fc193 AL |
1235 | } |
1236 | ||
8a426614 KW |
1237 | #define COMMIT_BUF_SECTORS 2048 |
1238 | ||
33e3963e FB |
1239 | /* commit COW file into the raw image */ |
1240 | int bdrv_commit(BlockDriverState *bs) | |
1241 | { | |
19cb3738 | 1242 | BlockDriver *drv = bs->drv; |
ee181196 | 1243 | BlockDriver *backing_drv; |
8a426614 KW |
1244 | int64_t sector, total_sectors; |
1245 | int n, ro, open_flags; | |
4dca4b63 | 1246 | int ret = 0, rw_ret = 0; |
8a426614 | 1247 | uint8_t *buf; |
4dca4b63 NS |
1248 | char filename[1024]; |
1249 | BlockDriverState *bs_rw, *bs_ro; | |
33e3963e | 1250 | |
19cb3738 FB |
1251 | if (!drv) |
1252 | return -ENOMEDIUM; | |
4dca4b63 NS |
1253 | |
1254 | if (!bs->backing_hd) { | |
1255 | return -ENOTSUP; | |
33e3963e FB |
1256 | } |
1257 | ||
4dca4b63 NS |
1258 | if (bs->backing_hd->keep_read_only) { |
1259 | return -EACCES; | |
1260 | } | |
ee181196 | 1261 | |
2d3735d3 SH |
1262 | if (bdrv_in_use(bs) || bdrv_in_use(bs->backing_hd)) { |
1263 | return -EBUSY; | |
1264 | } | |
1265 | ||
ee181196 | 1266 | backing_drv = bs->backing_hd->drv; |
4dca4b63 NS |
1267 | ro = bs->backing_hd->read_only; |
1268 | strncpy(filename, bs->backing_hd->filename, sizeof(filename)); | |
1269 | open_flags = bs->backing_hd->open_flags; | |
1270 | ||
1271 | if (ro) { | |
1272 | /* re-open as RW */ | |
1273 | bdrv_delete(bs->backing_hd); | |
1274 | bs->backing_hd = NULL; | |
1275 | bs_rw = bdrv_new(""); | |
ee181196 KW |
1276 | rw_ret = bdrv_open(bs_rw, filename, open_flags | BDRV_O_RDWR, |
1277 | backing_drv); | |
4dca4b63 NS |
1278 | if (rw_ret < 0) { |
1279 | bdrv_delete(bs_rw); | |
1280 | /* try to re-open read-only */ | |
1281 | bs_ro = bdrv_new(""); | |
ee181196 KW |
1282 | ret = bdrv_open(bs_ro, filename, open_flags & ~BDRV_O_RDWR, |
1283 | backing_drv); | |
4dca4b63 NS |
1284 | if (ret < 0) { |
1285 | bdrv_delete(bs_ro); | |
1286 | /* drive not functional anymore */ | |
1287 | bs->drv = NULL; | |
1288 | return ret; | |
1289 | } | |
1290 | bs->backing_hd = bs_ro; | |
1291 | return rw_ret; | |
1292 | } | |
1293 | bs->backing_hd = bs_rw; | |
ea2384d3 | 1294 | } |
33e3963e | 1295 | |
6ea44308 | 1296 | total_sectors = bdrv_getlength(bs) >> BDRV_SECTOR_BITS; |
7267c094 | 1297 | buf = g_malloc(COMMIT_BUF_SECTORS * BDRV_SECTOR_SIZE); |
8a426614 KW |
1298 | |
1299 | for (sector = 0; sector < total_sectors; sector += n) { | |
05c4af54 | 1300 | if (bdrv_is_allocated(bs, sector, COMMIT_BUF_SECTORS, &n)) { |
8a426614 KW |
1301 | |
1302 | if (bdrv_read(bs, sector, buf, n) != 0) { | |
1303 | ret = -EIO; | |
1304 | goto ro_cleanup; | |
1305 | } | |
1306 | ||
1307 | if (bdrv_write(bs->backing_hd, sector, buf, n) != 0) { | |
1308 | ret = -EIO; | |
1309 | goto ro_cleanup; | |
1310 | } | |
ea2384d3 | 1311 | } |
33e3963e | 1312 | } |
95389c86 | 1313 | |
1d44952f CH |
1314 | if (drv->bdrv_make_empty) { |
1315 | ret = drv->bdrv_make_empty(bs); | |
1316 | bdrv_flush(bs); | |
1317 | } | |
95389c86 | 1318 | |
3f5075ae CH |
1319 | /* |
1320 | * Make sure all data we wrote to the backing device is actually | |
1321 | * stable on disk. | |
1322 | */ | |
1323 | if (bs->backing_hd) | |
1324 | bdrv_flush(bs->backing_hd); | |
4dca4b63 NS |
1325 | |
1326 | ro_cleanup: | |
7267c094 | 1327 | g_free(buf); |
4dca4b63 NS |
1328 | |
1329 | if (ro) { | |
1330 | /* re-open as RO */ | |
1331 | bdrv_delete(bs->backing_hd); | |
1332 | bs->backing_hd = NULL; | |
1333 | bs_ro = bdrv_new(""); | |
ee181196 KW |
1334 | ret = bdrv_open(bs_ro, filename, open_flags & ~BDRV_O_RDWR, |
1335 | backing_drv); | |
4dca4b63 NS |
1336 | if (ret < 0) { |
1337 | bdrv_delete(bs_ro); | |
1338 | /* drive not functional anymore */ | |
1339 | bs->drv = NULL; | |
1340 | return ret; | |
1341 | } | |
1342 | bs->backing_hd = bs_ro; | |
1343 | bs->backing_hd->keep_read_only = 0; | |
1344 | } | |
1345 | ||
1d44952f | 1346 | return ret; |
33e3963e FB |
1347 | } |
1348 | ||
e8877497 | 1349 | int bdrv_commit_all(void) |
6ab4b5ab MA |
1350 | { |
1351 | BlockDriverState *bs; | |
1352 | ||
1353 | QTAILQ_FOREACH(bs, &bdrv_states, list) { | |
e8877497 SH |
1354 | int ret = bdrv_commit(bs); |
1355 | if (ret < 0) { | |
1356 | return ret; | |
1357 | } | |
6ab4b5ab | 1358 | } |
e8877497 | 1359 | return 0; |
6ab4b5ab MA |
1360 | } |
1361 | ||
dbffbdcf SH |
1362 | struct BdrvTrackedRequest { |
1363 | BlockDriverState *bs; | |
1364 | int64_t sector_num; | |
1365 | int nb_sectors; | |
1366 | bool is_write; | |
1367 | QLIST_ENTRY(BdrvTrackedRequest) list; | |
5f8b6491 | 1368 | Coroutine *co; /* owner, used for deadlock detection */ |
f4658285 | 1369 | CoQueue wait_queue; /* coroutines blocked on this request */ |
dbffbdcf SH |
1370 | }; |
1371 | ||
1372 | /** | |
1373 | * Remove an active request from the tracked requests list | |
1374 | * | |
1375 | * This function should be called when a tracked request is completing. | |
1376 | */ | |
1377 | static void tracked_request_end(BdrvTrackedRequest *req) | |
1378 | { | |
1379 | QLIST_REMOVE(req, list); | |
f4658285 | 1380 | qemu_co_queue_restart_all(&req->wait_queue); |
dbffbdcf SH |
1381 | } |
1382 | ||
1383 | /** | |
1384 | * Add an active request to the tracked requests list | |
1385 | */ | |
1386 | static void tracked_request_begin(BdrvTrackedRequest *req, | |
1387 | BlockDriverState *bs, | |
1388 | int64_t sector_num, | |
1389 | int nb_sectors, bool is_write) | |
1390 | { | |
1391 | *req = (BdrvTrackedRequest){ | |
1392 | .bs = bs, | |
1393 | .sector_num = sector_num, | |
1394 | .nb_sectors = nb_sectors, | |
1395 | .is_write = is_write, | |
5f8b6491 | 1396 | .co = qemu_coroutine_self(), |
dbffbdcf SH |
1397 | }; |
1398 | ||
f4658285 SH |
1399 | qemu_co_queue_init(&req->wait_queue); |
1400 | ||
dbffbdcf SH |
1401 | QLIST_INSERT_HEAD(&bs->tracked_requests, req, list); |
1402 | } | |
1403 | ||
d83947ac SH |
1404 | /** |
1405 | * Round a region to cluster boundaries | |
1406 | */ | |
1407 | static void round_to_clusters(BlockDriverState *bs, | |
1408 | int64_t sector_num, int nb_sectors, | |
1409 | int64_t *cluster_sector_num, | |
1410 | int *cluster_nb_sectors) | |
1411 | { | |
1412 | BlockDriverInfo bdi; | |
1413 | ||
1414 | if (bdrv_get_info(bs, &bdi) < 0 || bdi.cluster_size == 0) { | |
1415 | *cluster_sector_num = sector_num; | |
1416 | *cluster_nb_sectors = nb_sectors; | |
1417 | } else { | |
1418 | int64_t c = bdi.cluster_size / BDRV_SECTOR_SIZE; | |
1419 | *cluster_sector_num = QEMU_ALIGN_DOWN(sector_num, c); | |
1420 | *cluster_nb_sectors = QEMU_ALIGN_UP(sector_num - *cluster_sector_num + | |
1421 | nb_sectors, c); | |
1422 | } | |
1423 | } | |
1424 | ||
f4658285 SH |
1425 | static bool tracked_request_overlaps(BdrvTrackedRequest *req, |
1426 | int64_t sector_num, int nb_sectors) { | |
d83947ac SH |
1427 | /* aaaa bbbb */ |
1428 | if (sector_num >= req->sector_num + req->nb_sectors) { | |
1429 | return false; | |
1430 | } | |
1431 | /* bbbb aaaa */ | |
1432 | if (req->sector_num >= sector_num + nb_sectors) { | |
1433 | return false; | |
1434 | } | |
1435 | return true; | |
f4658285 SH |
1436 | } |
1437 | ||
1438 | static void coroutine_fn wait_for_overlapping_requests(BlockDriverState *bs, | |
1439 | int64_t sector_num, int nb_sectors) | |
1440 | { | |
1441 | BdrvTrackedRequest *req; | |
d83947ac SH |
1442 | int64_t cluster_sector_num; |
1443 | int cluster_nb_sectors; | |
f4658285 SH |
1444 | bool retry; |
1445 | ||
d83947ac SH |
1446 | /* If we touch the same cluster it counts as an overlap. This guarantees |
1447 | * that allocating writes will be serialized and not race with each other | |
1448 | * for the same cluster. For example, in copy-on-read it ensures that the | |
1449 | * CoR read and write operations are atomic and guest writes cannot | |
1450 | * interleave between them. | |
1451 | */ | |
1452 | round_to_clusters(bs, sector_num, nb_sectors, | |
1453 | &cluster_sector_num, &cluster_nb_sectors); | |
1454 | ||
f4658285 SH |
1455 | do { |
1456 | retry = false; | |
1457 | QLIST_FOREACH(req, &bs->tracked_requests, list) { | |
d83947ac SH |
1458 | if (tracked_request_overlaps(req, cluster_sector_num, |
1459 | cluster_nb_sectors)) { | |
5f8b6491 SH |
1460 | /* Hitting this means there was a reentrant request, for |
1461 | * example, a block driver issuing nested requests. This must | |
1462 | * never happen since it means deadlock. | |
1463 | */ | |
1464 | assert(qemu_coroutine_self() != req->co); | |
1465 | ||
f4658285 SH |
1466 | qemu_co_queue_wait(&req->wait_queue); |
1467 | retry = true; | |
1468 | break; | |
1469 | } | |
1470 | } | |
1471 | } while (retry); | |
1472 | } | |
1473 | ||
756e6736 KW |
1474 | /* |
1475 | * Return values: | |
1476 | * 0 - success | |
1477 | * -EINVAL - backing format specified, but no file | |
1478 | * -ENOSPC - can't update the backing file because no space is left in the | |
1479 | * image file header | |
1480 | * -ENOTSUP - format driver doesn't support changing the backing file | |
1481 | */ | |
1482 | int bdrv_change_backing_file(BlockDriverState *bs, | |
1483 | const char *backing_file, const char *backing_fmt) | |
1484 | { | |
1485 | BlockDriver *drv = bs->drv; | |
469ef350 | 1486 | int ret; |
756e6736 | 1487 | |
5f377794 PB |
1488 | /* Backing file format doesn't make sense without a backing file */ |
1489 | if (backing_fmt && !backing_file) { | |
1490 | return -EINVAL; | |
1491 | } | |
1492 | ||
756e6736 | 1493 | if (drv->bdrv_change_backing_file != NULL) { |
469ef350 | 1494 | ret = drv->bdrv_change_backing_file(bs, backing_file, backing_fmt); |
756e6736 | 1495 | } else { |
469ef350 | 1496 | ret = -ENOTSUP; |
756e6736 | 1497 | } |
469ef350 PB |
1498 | |
1499 | if (ret == 0) { | |
1500 | pstrcpy(bs->backing_file, sizeof(bs->backing_file), backing_file ?: ""); | |
1501 | pstrcpy(bs->backing_format, sizeof(bs->backing_format), backing_fmt ?: ""); | |
1502 | } | |
1503 | return ret; | |
756e6736 KW |
1504 | } |
1505 | ||
71d0770c AL |
1506 | static int bdrv_check_byte_request(BlockDriverState *bs, int64_t offset, |
1507 | size_t size) | |
1508 | { | |
1509 | int64_t len; | |
1510 | ||
1511 | if (!bdrv_is_inserted(bs)) | |
1512 | return -ENOMEDIUM; | |
1513 | ||
1514 | if (bs->growable) | |
1515 | return 0; | |
1516 | ||
1517 | len = bdrv_getlength(bs); | |
1518 | ||
fbb7b4e0 KW |
1519 | if (offset < 0) |
1520 | return -EIO; | |
1521 | ||
1522 | if ((offset > len) || (len - offset < size)) | |
71d0770c AL |
1523 | return -EIO; |
1524 | ||
1525 | return 0; | |
1526 | } | |
1527 | ||
1528 | static int bdrv_check_request(BlockDriverState *bs, int64_t sector_num, | |
1529 | int nb_sectors) | |
1530 | { | |
eb5a3165 JS |
1531 | return bdrv_check_byte_request(bs, sector_num * BDRV_SECTOR_SIZE, |
1532 | nb_sectors * BDRV_SECTOR_SIZE); | |
71d0770c AL |
1533 | } |
1534 | ||
1c9805a3 SH |
1535 | typedef struct RwCo { |
1536 | BlockDriverState *bs; | |
1537 | int64_t sector_num; | |
1538 | int nb_sectors; | |
1539 | QEMUIOVector *qiov; | |
1540 | bool is_write; | |
1541 | int ret; | |
1542 | } RwCo; | |
1543 | ||
1544 | static void coroutine_fn bdrv_rw_co_entry(void *opaque) | |
fc01f7e7 | 1545 | { |
1c9805a3 | 1546 | RwCo *rwco = opaque; |
ea2384d3 | 1547 | |
1c9805a3 SH |
1548 | if (!rwco->is_write) { |
1549 | rwco->ret = bdrv_co_do_readv(rwco->bs, rwco->sector_num, | |
470c0504 | 1550 | rwco->nb_sectors, rwco->qiov, 0); |
1c9805a3 SH |
1551 | } else { |
1552 | rwco->ret = bdrv_co_do_writev(rwco->bs, rwco->sector_num, | |
f08f2dda | 1553 | rwco->nb_sectors, rwco->qiov, 0); |
1c9805a3 SH |
1554 | } |
1555 | } | |
e7a8a783 | 1556 | |
1c9805a3 SH |
1557 | /* |
1558 | * Process a synchronous request using coroutines | |
1559 | */ | |
1560 | static int bdrv_rw_co(BlockDriverState *bs, int64_t sector_num, uint8_t *buf, | |
1561 | int nb_sectors, bool is_write) | |
1562 | { | |
1563 | QEMUIOVector qiov; | |
1564 | struct iovec iov = { | |
1565 | .iov_base = (void *)buf, | |
1566 | .iov_len = nb_sectors * BDRV_SECTOR_SIZE, | |
1567 | }; | |
1568 | Coroutine *co; | |
1569 | RwCo rwco = { | |
1570 | .bs = bs, | |
1571 | .sector_num = sector_num, | |
1572 | .nb_sectors = nb_sectors, | |
1573 | .qiov = &qiov, | |
1574 | .is_write = is_write, | |
1575 | .ret = NOT_DONE, | |
1576 | }; | |
e7a8a783 | 1577 | |
1c9805a3 | 1578 | qemu_iovec_init_external(&qiov, &iov, 1); |
e7a8a783 | 1579 | |
498e386c ZYW |
1580 | /** |
1581 | * In sync call context, when the vcpu is blocked, this throttling timer | |
1582 | * will not fire; so the I/O throttling function has to be disabled here | |
1583 | * if it has been enabled. | |
1584 | */ | |
1585 | if (bs->io_limits_enabled) { | |
1586 | fprintf(stderr, "Disabling I/O throttling on '%s' due " | |
1587 | "to synchronous I/O.\n", bdrv_get_device_name(bs)); | |
1588 | bdrv_io_limits_disable(bs); | |
1589 | } | |
1590 | ||
1c9805a3 SH |
1591 | if (qemu_in_coroutine()) { |
1592 | /* Fast-path if already in coroutine context */ | |
1593 | bdrv_rw_co_entry(&rwco); | |
1594 | } else { | |
1595 | co = qemu_coroutine_create(bdrv_rw_co_entry); | |
1596 | qemu_coroutine_enter(co, &rwco); | |
1597 | while (rwco.ret == NOT_DONE) { | |
1598 | qemu_aio_wait(); | |
1599 | } | |
1600 | } | |
1601 | return rwco.ret; | |
1602 | } | |
b338082b | 1603 | |
1c9805a3 SH |
1604 | /* return < 0 if error. See bdrv_write() for the return codes */ |
1605 | int bdrv_read(BlockDriverState *bs, int64_t sector_num, | |
1606 | uint8_t *buf, int nb_sectors) | |
1607 | { | |
1608 | return bdrv_rw_co(bs, sector_num, buf, nb_sectors, false); | |
fc01f7e7 FB |
1609 | } |
1610 | ||
71df14fc PB |
1611 | #define BITS_PER_LONG (sizeof(unsigned long) * 8) |
1612 | ||
7cd1e32a | 1613 | static void set_dirty_bitmap(BlockDriverState *bs, int64_t sector_num, |
a55eb92c | 1614 | int nb_sectors, int dirty) |
7cd1e32a LS |
1615 | { |
1616 | int64_t start, end; | |
c6d22830 | 1617 | unsigned long val, idx, bit; |
a55eb92c | 1618 | |
6ea44308 | 1619 | start = sector_num / BDRV_SECTORS_PER_DIRTY_CHUNK; |
c6d22830 | 1620 | end = (sector_num + nb_sectors - 1) / BDRV_SECTORS_PER_DIRTY_CHUNK; |
a55eb92c JK |
1621 | |
1622 | for (; start <= end; start++) { | |
71df14fc PB |
1623 | idx = start / BITS_PER_LONG; |
1624 | bit = start % BITS_PER_LONG; | |
c6d22830 JK |
1625 | val = bs->dirty_bitmap[idx]; |
1626 | if (dirty) { | |
6d59fec1 | 1627 | if (!(val & (1UL << bit))) { |
aaa0eb75 | 1628 | bs->dirty_count++; |
6d59fec1 | 1629 | val |= 1UL << bit; |
aaa0eb75 | 1630 | } |
c6d22830 | 1631 | } else { |
6d59fec1 | 1632 | if (val & (1UL << bit)) { |
aaa0eb75 | 1633 | bs->dirty_count--; |
6d59fec1 | 1634 | val &= ~(1UL << bit); |
aaa0eb75 | 1635 | } |
c6d22830 JK |
1636 | } |
1637 | bs->dirty_bitmap[idx] = val; | |
7cd1e32a LS |
1638 | } |
1639 | } | |
1640 | ||
5fafdf24 | 1641 | /* Return < 0 if error. Important errors are: |
19cb3738 FB |
1642 | -EIO generic I/O error (may happen for all errors) |
1643 | -ENOMEDIUM No media inserted. | |
1644 | -EINVAL Invalid sector number or nb_sectors | |
1645 | -EACCES Trying to write a read-only device | |
1646 | */ | |
5fafdf24 | 1647 | int bdrv_write(BlockDriverState *bs, int64_t sector_num, |
fc01f7e7 FB |
1648 | const uint8_t *buf, int nb_sectors) |
1649 | { | |
1c9805a3 | 1650 | return bdrv_rw_co(bs, sector_num, (uint8_t *)buf, nb_sectors, true); |
83f64091 FB |
1651 | } |
1652 | ||
eda578e5 AL |
1653 | int bdrv_pread(BlockDriverState *bs, int64_t offset, |
1654 | void *buf, int count1) | |
83f64091 | 1655 | { |
6ea44308 | 1656 | uint8_t tmp_buf[BDRV_SECTOR_SIZE]; |
83f64091 FB |
1657 | int len, nb_sectors, count; |
1658 | int64_t sector_num; | |
9a8c4cce | 1659 | int ret; |
83f64091 FB |
1660 | |
1661 | count = count1; | |
1662 | /* first read to align to sector start */ | |
6ea44308 | 1663 | len = (BDRV_SECTOR_SIZE - offset) & (BDRV_SECTOR_SIZE - 1); |
83f64091 FB |
1664 | if (len > count) |
1665 | len = count; | |
6ea44308 | 1666 | sector_num = offset >> BDRV_SECTOR_BITS; |
83f64091 | 1667 | if (len > 0) { |
9a8c4cce KW |
1668 | if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0) |
1669 | return ret; | |
6ea44308 | 1670 | memcpy(buf, tmp_buf + (offset & (BDRV_SECTOR_SIZE - 1)), len); |
83f64091 FB |
1671 | count -= len; |
1672 | if (count == 0) | |
1673 | return count1; | |
1674 | sector_num++; | |
1675 | buf += len; | |
1676 | } | |
1677 | ||
1678 | /* read the sectors "in place" */ | |
6ea44308 | 1679 | nb_sectors = count >> BDRV_SECTOR_BITS; |
83f64091 | 1680 | if (nb_sectors > 0) { |
9a8c4cce KW |
1681 | if ((ret = bdrv_read(bs, sector_num, buf, nb_sectors)) < 0) |
1682 | return ret; | |
83f64091 | 1683 | sector_num += nb_sectors; |
6ea44308 | 1684 | len = nb_sectors << BDRV_SECTOR_BITS; |
83f64091 FB |
1685 | buf += len; |
1686 | count -= len; | |
1687 | } | |
1688 | ||
1689 | /* add data from the last sector */ | |
1690 | if (count > 0) { | |
9a8c4cce KW |
1691 | if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0) |
1692 | return ret; | |
83f64091 FB |
1693 | memcpy(buf, tmp_buf, count); |
1694 | } | |
1695 | return count1; | |
1696 | } | |
1697 | ||
eda578e5 AL |
1698 | int bdrv_pwrite(BlockDriverState *bs, int64_t offset, |
1699 | const void *buf, int count1) | |
83f64091 | 1700 | { |
6ea44308 | 1701 | uint8_t tmp_buf[BDRV_SECTOR_SIZE]; |
83f64091 FB |
1702 | int len, nb_sectors, count; |
1703 | int64_t sector_num; | |
9a8c4cce | 1704 | int ret; |
83f64091 FB |
1705 | |
1706 | count = count1; | |
1707 | /* first write to align to sector start */ | |
6ea44308 | 1708 | len = (BDRV_SECTOR_SIZE - offset) & (BDRV_SECTOR_SIZE - 1); |
83f64091 FB |
1709 | if (len > count) |
1710 | len = count; | |
6ea44308 | 1711 | sector_num = offset >> BDRV_SECTOR_BITS; |
83f64091 | 1712 | if (len > 0) { |
9a8c4cce KW |
1713 | if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0) |
1714 | return ret; | |
6ea44308 | 1715 | memcpy(tmp_buf + (offset & (BDRV_SECTOR_SIZE - 1)), buf, len); |
9a8c4cce KW |
1716 | if ((ret = bdrv_write(bs, sector_num, tmp_buf, 1)) < 0) |
1717 | return ret; | |
83f64091 FB |
1718 | count -= len; |
1719 | if (count == 0) | |
1720 | return count1; | |
1721 | sector_num++; | |
1722 | buf += len; | |
1723 | } | |
1724 | ||
1725 | /* write the sectors "in place" */ | |
6ea44308 | 1726 | nb_sectors = count >> BDRV_SECTOR_BITS; |
83f64091 | 1727 | if (nb_sectors > 0) { |
9a8c4cce KW |
1728 | if ((ret = bdrv_write(bs, sector_num, buf, nb_sectors)) < 0) |
1729 | return ret; | |
83f64091 | 1730 | sector_num += nb_sectors; |
6ea44308 | 1731 | len = nb_sectors << BDRV_SECTOR_BITS; |
83f64091 FB |
1732 | buf += len; |
1733 | count -= len; | |
1734 | } | |
1735 | ||
1736 | /* add data from the last sector */ | |
1737 | if (count > 0) { | |
9a8c4cce KW |
1738 | if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0) |
1739 | return ret; | |
83f64091 | 1740 | memcpy(tmp_buf, buf, count); |
9a8c4cce KW |
1741 | if ((ret = bdrv_write(bs, sector_num, tmp_buf, 1)) < 0) |
1742 | return ret; | |
83f64091 FB |
1743 | } |
1744 | return count1; | |
1745 | } | |
83f64091 | 1746 | |
f08145fe KW |
1747 | /* |
1748 | * Writes to the file and ensures that no writes are reordered across this | |
1749 | * request (acts as a barrier) | |
1750 | * | |
1751 | * Returns 0 on success, -errno in error cases. | |
1752 | */ | |
1753 | int bdrv_pwrite_sync(BlockDriverState *bs, int64_t offset, | |
1754 | const void *buf, int count) | |
1755 | { | |
1756 | int ret; | |
1757 | ||
1758 | ret = bdrv_pwrite(bs, offset, buf, count); | |
1759 | if (ret < 0) { | |
1760 | return ret; | |
1761 | } | |
1762 | ||
f05fa4ad PB |
1763 | /* No flush needed for cache modes that already do it */ |
1764 | if (bs->enable_write_cache) { | |
f08145fe KW |
1765 | bdrv_flush(bs); |
1766 | } | |
1767 | ||
1768 | return 0; | |
1769 | } | |
1770 | ||
470c0504 | 1771 | static int coroutine_fn bdrv_co_do_copy_on_readv(BlockDriverState *bs, |
ab185921 SH |
1772 | int64_t sector_num, int nb_sectors, QEMUIOVector *qiov) |
1773 | { | |
1774 | /* Perform I/O through a temporary buffer so that users who scribble over | |
1775 | * their read buffer while the operation is in progress do not end up | |
1776 | * modifying the image file. This is critical for zero-copy guest I/O | |
1777 | * where anything might happen inside guest memory. | |
1778 | */ | |
1779 | void *bounce_buffer; | |
1780 | ||
79c053bd | 1781 | BlockDriver *drv = bs->drv; |
ab185921 SH |
1782 | struct iovec iov; |
1783 | QEMUIOVector bounce_qiov; | |
1784 | int64_t cluster_sector_num; | |
1785 | int cluster_nb_sectors; | |
1786 | size_t skip_bytes; | |
1787 | int ret; | |
1788 | ||
1789 | /* Cover entire cluster so no additional backing file I/O is required when | |
1790 | * allocating cluster in the image file. | |
1791 | */ | |
1792 | round_to_clusters(bs, sector_num, nb_sectors, | |
1793 | &cluster_sector_num, &cluster_nb_sectors); | |
1794 | ||
470c0504 SH |
1795 | trace_bdrv_co_do_copy_on_readv(bs, sector_num, nb_sectors, |
1796 | cluster_sector_num, cluster_nb_sectors); | |
ab185921 SH |
1797 | |
1798 | iov.iov_len = cluster_nb_sectors * BDRV_SECTOR_SIZE; | |
1799 | iov.iov_base = bounce_buffer = qemu_blockalign(bs, iov.iov_len); | |
1800 | qemu_iovec_init_external(&bounce_qiov, &iov, 1); | |
1801 | ||
79c053bd SH |
1802 | ret = drv->bdrv_co_readv(bs, cluster_sector_num, cluster_nb_sectors, |
1803 | &bounce_qiov); | |
ab185921 SH |
1804 | if (ret < 0) { |
1805 | goto err; | |
1806 | } | |
1807 | ||
79c053bd SH |
1808 | if (drv->bdrv_co_write_zeroes && |
1809 | buffer_is_zero(bounce_buffer, iov.iov_len)) { | |
621f0589 KW |
1810 | ret = bdrv_co_do_write_zeroes(bs, cluster_sector_num, |
1811 | cluster_nb_sectors); | |
79c053bd | 1812 | } else { |
f05fa4ad PB |
1813 | /* This does not change the data on the disk, it is not necessary |
1814 | * to flush even in cache=writethrough mode. | |
1815 | */ | |
79c053bd | 1816 | ret = drv->bdrv_co_writev(bs, cluster_sector_num, cluster_nb_sectors, |
ab185921 | 1817 | &bounce_qiov); |
79c053bd SH |
1818 | } |
1819 | ||
ab185921 SH |
1820 | if (ret < 0) { |
1821 | /* It might be okay to ignore write errors for guest requests. If this | |
1822 | * is a deliberate copy-on-read then we don't want to ignore the error. | |
1823 | * Simply report it in all cases. | |
1824 | */ | |
1825 | goto err; | |
1826 | } | |
1827 | ||
1828 | skip_bytes = (sector_num - cluster_sector_num) * BDRV_SECTOR_SIZE; | |
1829 | qemu_iovec_from_buffer(qiov, bounce_buffer + skip_bytes, | |
1830 | nb_sectors * BDRV_SECTOR_SIZE); | |
1831 | ||
1832 | err: | |
1833 | qemu_vfree(bounce_buffer); | |
1834 | return ret; | |
1835 | } | |
1836 | ||
c5fbe571 SH |
1837 | /* |
1838 | * Handle a read request in coroutine context | |
1839 | */ | |
1840 | static int coroutine_fn bdrv_co_do_readv(BlockDriverState *bs, | |
470c0504 SH |
1841 | int64_t sector_num, int nb_sectors, QEMUIOVector *qiov, |
1842 | BdrvRequestFlags flags) | |
da1fa91d KW |
1843 | { |
1844 | BlockDriver *drv = bs->drv; | |
dbffbdcf SH |
1845 | BdrvTrackedRequest req; |
1846 | int ret; | |
da1fa91d | 1847 | |
da1fa91d KW |
1848 | if (!drv) { |
1849 | return -ENOMEDIUM; | |
1850 | } | |
1851 | if (bdrv_check_request(bs, sector_num, nb_sectors)) { | |
1852 | return -EIO; | |
1853 | } | |
1854 | ||
98f90dba ZYW |
1855 | /* throttling disk read I/O */ |
1856 | if (bs->io_limits_enabled) { | |
1857 | bdrv_io_limits_intercept(bs, false, nb_sectors); | |
1858 | } | |
1859 | ||
f4658285 | 1860 | if (bs->copy_on_read) { |
470c0504 SH |
1861 | flags |= BDRV_REQ_COPY_ON_READ; |
1862 | } | |
1863 | if (flags & BDRV_REQ_COPY_ON_READ) { | |
1864 | bs->copy_on_read_in_flight++; | |
1865 | } | |
1866 | ||
1867 | if (bs->copy_on_read_in_flight) { | |
f4658285 SH |
1868 | wait_for_overlapping_requests(bs, sector_num, nb_sectors); |
1869 | } | |
1870 | ||
dbffbdcf | 1871 | tracked_request_begin(&req, bs, sector_num, nb_sectors, false); |
ab185921 | 1872 | |
470c0504 | 1873 | if (flags & BDRV_REQ_COPY_ON_READ) { |
ab185921 SH |
1874 | int pnum; |
1875 | ||
1876 | ret = bdrv_co_is_allocated(bs, sector_num, nb_sectors, &pnum); | |
1877 | if (ret < 0) { | |
1878 | goto out; | |
1879 | } | |
1880 | ||
1881 | if (!ret || pnum != nb_sectors) { | |
470c0504 | 1882 | ret = bdrv_co_do_copy_on_readv(bs, sector_num, nb_sectors, qiov); |
ab185921 SH |
1883 | goto out; |
1884 | } | |
1885 | } | |
1886 | ||
dbffbdcf | 1887 | ret = drv->bdrv_co_readv(bs, sector_num, nb_sectors, qiov); |
ab185921 SH |
1888 | |
1889 | out: | |
dbffbdcf | 1890 | tracked_request_end(&req); |
470c0504 SH |
1891 | |
1892 | if (flags & BDRV_REQ_COPY_ON_READ) { | |
1893 | bs->copy_on_read_in_flight--; | |
1894 | } | |
1895 | ||
dbffbdcf | 1896 | return ret; |
da1fa91d KW |
1897 | } |
1898 | ||
c5fbe571 | 1899 | int coroutine_fn bdrv_co_readv(BlockDriverState *bs, int64_t sector_num, |
da1fa91d KW |
1900 | int nb_sectors, QEMUIOVector *qiov) |
1901 | { | |
c5fbe571 | 1902 | trace_bdrv_co_readv(bs, sector_num, nb_sectors); |
da1fa91d | 1903 | |
470c0504 SH |
1904 | return bdrv_co_do_readv(bs, sector_num, nb_sectors, qiov, 0); |
1905 | } | |
1906 | ||
1907 | int coroutine_fn bdrv_co_copy_on_readv(BlockDriverState *bs, | |
1908 | int64_t sector_num, int nb_sectors, QEMUIOVector *qiov) | |
1909 | { | |
1910 | trace_bdrv_co_copy_on_readv(bs, sector_num, nb_sectors); | |
1911 | ||
1912 | return bdrv_co_do_readv(bs, sector_num, nb_sectors, qiov, | |
1913 | BDRV_REQ_COPY_ON_READ); | |
c5fbe571 SH |
1914 | } |
1915 | ||
f08f2dda SH |
1916 | static int coroutine_fn bdrv_co_do_write_zeroes(BlockDriverState *bs, |
1917 | int64_t sector_num, int nb_sectors) | |
1918 | { | |
1919 | BlockDriver *drv = bs->drv; | |
1920 | QEMUIOVector qiov; | |
1921 | struct iovec iov; | |
1922 | int ret; | |
1923 | ||
621f0589 KW |
1924 | /* TODO Emulate only part of misaligned requests instead of letting block |
1925 | * drivers return -ENOTSUP and emulate everything */ | |
1926 | ||
f08f2dda SH |
1927 | /* First try the efficient write zeroes operation */ |
1928 | if (drv->bdrv_co_write_zeroes) { | |
621f0589 KW |
1929 | ret = drv->bdrv_co_write_zeroes(bs, sector_num, nb_sectors); |
1930 | if (ret != -ENOTSUP) { | |
1931 | return ret; | |
1932 | } | |
f08f2dda SH |
1933 | } |
1934 | ||
1935 | /* Fall back to bounce buffer if write zeroes is unsupported */ | |
1936 | iov.iov_len = nb_sectors * BDRV_SECTOR_SIZE; | |
1937 | iov.iov_base = qemu_blockalign(bs, iov.iov_len); | |
1938 | memset(iov.iov_base, 0, iov.iov_len); | |
1939 | qemu_iovec_init_external(&qiov, &iov, 1); | |
1940 | ||
1941 | ret = drv->bdrv_co_writev(bs, sector_num, nb_sectors, &qiov); | |
1942 | ||
1943 | qemu_vfree(iov.iov_base); | |
1944 | return ret; | |
1945 | } | |
1946 | ||
c5fbe571 SH |
1947 | /* |
1948 | * Handle a write request in coroutine context | |
1949 | */ | |
1950 | static int coroutine_fn bdrv_co_do_writev(BlockDriverState *bs, | |
f08f2dda SH |
1951 | int64_t sector_num, int nb_sectors, QEMUIOVector *qiov, |
1952 | BdrvRequestFlags flags) | |
c5fbe571 SH |
1953 | { |
1954 | BlockDriver *drv = bs->drv; | |
dbffbdcf | 1955 | BdrvTrackedRequest req; |
6b7cb247 | 1956 | int ret; |
da1fa91d KW |
1957 | |
1958 | if (!bs->drv) { | |
1959 | return -ENOMEDIUM; | |
1960 | } | |
1961 | if (bs->read_only) { | |
1962 | return -EACCES; | |
1963 | } | |
1964 | if (bdrv_check_request(bs, sector_num, nb_sectors)) { | |
1965 | return -EIO; | |
1966 | } | |
1967 | ||
98f90dba ZYW |
1968 | /* throttling disk write I/O */ |
1969 | if (bs->io_limits_enabled) { | |
1970 | bdrv_io_limits_intercept(bs, true, nb_sectors); | |
1971 | } | |
1972 | ||
470c0504 | 1973 | if (bs->copy_on_read_in_flight) { |
f4658285 SH |
1974 | wait_for_overlapping_requests(bs, sector_num, nb_sectors); |
1975 | } | |
1976 | ||
dbffbdcf SH |
1977 | tracked_request_begin(&req, bs, sector_num, nb_sectors, true); |
1978 | ||
f08f2dda SH |
1979 | if (flags & BDRV_REQ_ZERO_WRITE) { |
1980 | ret = bdrv_co_do_write_zeroes(bs, sector_num, nb_sectors); | |
1981 | } else { | |
1982 | ret = drv->bdrv_co_writev(bs, sector_num, nb_sectors, qiov); | |
1983 | } | |
6b7cb247 | 1984 | |
f05fa4ad PB |
1985 | if (ret == 0 && !bs->enable_write_cache) { |
1986 | ret = bdrv_co_flush(bs); | |
1987 | } | |
1988 | ||
da1fa91d KW |
1989 | if (bs->dirty_bitmap) { |
1990 | set_dirty_bitmap(bs, sector_num, nb_sectors, 1); | |
1991 | } | |
1992 | ||
1993 | if (bs->wr_highest_sector < sector_num + nb_sectors - 1) { | |
1994 | bs->wr_highest_sector = sector_num + nb_sectors - 1; | |
1995 | } | |
1996 | ||
dbffbdcf SH |
1997 | tracked_request_end(&req); |
1998 | ||
6b7cb247 | 1999 | return ret; |
da1fa91d KW |
2000 | } |
2001 | ||
c5fbe571 SH |
2002 | int coroutine_fn bdrv_co_writev(BlockDriverState *bs, int64_t sector_num, |
2003 | int nb_sectors, QEMUIOVector *qiov) | |
2004 | { | |
2005 | trace_bdrv_co_writev(bs, sector_num, nb_sectors); | |
2006 | ||
f08f2dda SH |
2007 | return bdrv_co_do_writev(bs, sector_num, nb_sectors, qiov, 0); |
2008 | } | |
2009 | ||
2010 | int coroutine_fn bdrv_co_write_zeroes(BlockDriverState *bs, | |
2011 | int64_t sector_num, int nb_sectors) | |
2012 | { | |
2013 | trace_bdrv_co_write_zeroes(bs, sector_num, nb_sectors); | |
2014 | ||
2015 | return bdrv_co_do_writev(bs, sector_num, nb_sectors, NULL, | |
2016 | BDRV_REQ_ZERO_WRITE); | |
c5fbe571 SH |
2017 | } |
2018 | ||
83f64091 FB |
2019 | /** |
2020 | * Truncate file to 'offset' bytes (needed only for file protocols) | |
2021 | */ | |
2022 | int bdrv_truncate(BlockDriverState *bs, int64_t offset) | |
2023 | { | |
2024 | BlockDriver *drv = bs->drv; | |
51762288 | 2025 | int ret; |
83f64091 | 2026 | if (!drv) |
19cb3738 | 2027 | return -ENOMEDIUM; |
83f64091 FB |
2028 | if (!drv->bdrv_truncate) |
2029 | return -ENOTSUP; | |
59f2689d NS |
2030 | if (bs->read_only) |
2031 | return -EACCES; | |
8591675f MT |
2032 | if (bdrv_in_use(bs)) |
2033 | return -EBUSY; | |
51762288 SH |
2034 | ret = drv->bdrv_truncate(bs, offset); |
2035 | if (ret == 0) { | |
2036 | ret = refresh_total_sectors(bs, offset >> BDRV_SECTOR_BITS); | |
145feb17 | 2037 | bdrv_dev_resize_cb(bs); |
51762288 SH |
2038 | } |
2039 | return ret; | |
83f64091 FB |
2040 | } |
2041 | ||
4a1d5e1f FZ |
2042 | /** |
2043 | * Length of a allocated file in bytes. Sparse files are counted by actual | |
2044 | * allocated space. Return < 0 if error or unknown. | |
2045 | */ | |
2046 | int64_t bdrv_get_allocated_file_size(BlockDriverState *bs) | |
2047 | { | |
2048 | BlockDriver *drv = bs->drv; | |
2049 | if (!drv) { | |
2050 | return -ENOMEDIUM; | |
2051 | } | |
2052 | if (drv->bdrv_get_allocated_file_size) { | |
2053 | return drv->bdrv_get_allocated_file_size(bs); | |
2054 | } | |
2055 | if (bs->file) { | |
2056 | return bdrv_get_allocated_file_size(bs->file); | |
2057 | } | |
2058 | return -ENOTSUP; | |
2059 | } | |
2060 | ||
83f64091 FB |
2061 | /** |
2062 | * Length of a file in bytes. Return < 0 if error or unknown. | |
2063 | */ | |
2064 | int64_t bdrv_getlength(BlockDriverState *bs) | |
2065 | { | |
2066 | BlockDriver *drv = bs->drv; | |
2067 | if (!drv) | |
19cb3738 | 2068 | return -ENOMEDIUM; |
51762288 | 2069 | |
2c6942fa | 2070 | if (bs->growable || bdrv_dev_has_removable_media(bs)) { |
46a4e4e6 SH |
2071 | if (drv->bdrv_getlength) { |
2072 | return drv->bdrv_getlength(bs); | |
2073 | } | |
83f64091 | 2074 | } |
46a4e4e6 | 2075 | return bs->total_sectors * BDRV_SECTOR_SIZE; |
fc01f7e7 FB |
2076 | } |
2077 | ||
19cb3738 | 2078 | /* return 0 as number of sectors if no device present or error */ |
96b8f136 | 2079 | void bdrv_get_geometry(BlockDriverState *bs, uint64_t *nb_sectors_ptr) |
fc01f7e7 | 2080 | { |
19cb3738 FB |
2081 | int64_t length; |
2082 | length = bdrv_getlength(bs); | |
2083 | if (length < 0) | |
2084 | length = 0; | |
2085 | else | |
6ea44308 | 2086 | length = length >> BDRV_SECTOR_BITS; |
19cb3738 | 2087 | *nb_sectors_ptr = length; |
fc01f7e7 | 2088 | } |
cf98951b | 2089 | |
f3d54fc4 AL |
2090 | struct partition { |
2091 | uint8_t boot_ind; /* 0x80 - active */ | |
2092 | uint8_t head; /* starting head */ | |
2093 | uint8_t sector; /* starting sector */ | |
2094 | uint8_t cyl; /* starting cylinder */ | |
2095 | uint8_t sys_ind; /* What partition type */ | |
2096 | uint8_t end_head; /* end head */ | |
2097 | uint8_t end_sector; /* end sector */ | |
2098 | uint8_t end_cyl; /* end cylinder */ | |
2099 | uint32_t start_sect; /* starting sector counting from 0 */ | |
2100 | uint32_t nr_sects; /* nr of sectors in partition */ | |
541dc0d4 | 2101 | } QEMU_PACKED; |
f3d54fc4 AL |
2102 | |
2103 | /* try to guess the disk logical geometry from the MSDOS partition table. Return 0 if OK, -1 if could not guess */ | |
2104 | static int guess_disk_lchs(BlockDriverState *bs, | |
2105 | int *pcylinders, int *pheads, int *psectors) | |
2106 | { | |
eb5a3165 | 2107 | uint8_t buf[BDRV_SECTOR_SIZE]; |
f3d54fc4 AL |
2108 | int ret, i, heads, sectors, cylinders; |
2109 | struct partition *p; | |
2110 | uint32_t nr_sects; | |
a38131b6 | 2111 | uint64_t nb_sectors; |
498e386c | 2112 | bool enabled; |
f3d54fc4 AL |
2113 | |
2114 | bdrv_get_geometry(bs, &nb_sectors); | |
2115 | ||
498e386c ZYW |
2116 | /** |
2117 | * The function will be invoked during startup not only in sync I/O mode, | |
2118 | * but also in async I/O mode. So the I/O throttling function has to | |
2119 | * be disabled temporarily here, not permanently. | |
2120 | */ | |
2121 | enabled = bs->io_limits_enabled; | |
2122 | bs->io_limits_enabled = false; | |
f3d54fc4 | 2123 | ret = bdrv_read(bs, 0, buf, 1); |
498e386c | 2124 | bs->io_limits_enabled = enabled; |
f3d54fc4 AL |
2125 | if (ret < 0) |
2126 | return -1; | |
2127 | /* test msdos magic */ | |
2128 | if (buf[510] != 0x55 || buf[511] != 0xaa) | |
2129 | return -1; | |
2130 | for(i = 0; i < 4; i++) { | |
2131 | p = ((struct partition *)(buf + 0x1be)) + i; | |
2132 | nr_sects = le32_to_cpu(p->nr_sects); | |
2133 | if (nr_sects && p->end_head) { | |
2134 | /* We make the assumption that the partition terminates on | |
2135 | a cylinder boundary */ | |
2136 | heads = p->end_head + 1; | |
2137 | sectors = p->end_sector & 63; | |
2138 | if (sectors == 0) | |
2139 | continue; | |
2140 | cylinders = nb_sectors / (heads * sectors); | |
2141 | if (cylinders < 1 || cylinders > 16383) | |
2142 | continue; | |
2143 | *pheads = heads; | |
2144 | *psectors = sectors; | |
2145 | *pcylinders = cylinders; | |
2146 | #if 0 | |
2147 | printf("guessed geometry: LCHS=%d %d %d\n", | |
2148 | cylinders, heads, sectors); | |
2149 | #endif | |
2150 | return 0; | |
2151 | } | |
2152 | } | |
2153 | return -1; | |
2154 | } | |
2155 | ||
2156 | void bdrv_guess_geometry(BlockDriverState *bs, int *pcyls, int *pheads, int *psecs) | |
2157 | { | |
2158 | int translation, lba_detected = 0; | |
2159 | int cylinders, heads, secs; | |
a38131b6 | 2160 | uint64_t nb_sectors; |
f3d54fc4 AL |
2161 | |
2162 | /* if a geometry hint is available, use it */ | |
2163 | bdrv_get_geometry(bs, &nb_sectors); | |
2164 | bdrv_get_geometry_hint(bs, &cylinders, &heads, &secs); | |
2165 | translation = bdrv_get_translation_hint(bs); | |
2166 | if (cylinders != 0) { | |
2167 | *pcyls = cylinders; | |
2168 | *pheads = heads; | |
2169 | *psecs = secs; | |
2170 | } else { | |
2171 | if (guess_disk_lchs(bs, &cylinders, &heads, &secs) == 0) { | |
2172 | if (heads > 16) { | |
2173 | /* if heads > 16, it means that a BIOS LBA | |
2174 | translation was active, so the default | |
2175 | hardware geometry is OK */ | |
2176 | lba_detected = 1; | |
2177 | goto default_geometry; | |
2178 | } else { | |
2179 | *pcyls = cylinders; | |
2180 | *pheads = heads; | |
2181 | *psecs = secs; | |
2182 | /* disable any translation to be in sync with | |
2183 | the logical geometry */ | |
2184 | if (translation == BIOS_ATA_TRANSLATION_AUTO) { | |
2185 | bdrv_set_translation_hint(bs, | |
2186 | BIOS_ATA_TRANSLATION_NONE); | |
2187 | } | |
2188 | } | |
2189 | } else { | |
2190 | default_geometry: | |
2191 | /* if no geometry, use a standard physical disk geometry */ | |
2192 | cylinders = nb_sectors / (16 * 63); | |
2193 | ||
2194 | if (cylinders > 16383) | |
2195 | cylinders = 16383; | |
2196 | else if (cylinders < 2) | |
2197 | cylinders = 2; | |
2198 | *pcyls = cylinders; | |
2199 | *pheads = 16; | |
2200 | *psecs = 63; | |
2201 | if ((lba_detected == 1) && (translation == BIOS_ATA_TRANSLATION_AUTO)) { | |
2202 | if ((*pcyls * *pheads) <= 131072) { | |
2203 | bdrv_set_translation_hint(bs, | |
2204 | BIOS_ATA_TRANSLATION_LARGE); | |
2205 | } else { | |
2206 | bdrv_set_translation_hint(bs, | |
2207 | BIOS_ATA_TRANSLATION_LBA); | |
2208 | } | |
2209 | } | |
2210 | } | |
2211 | bdrv_set_geometry_hint(bs, *pcyls, *pheads, *psecs); | |
2212 | } | |
2213 | } | |
2214 | ||
5fafdf24 | 2215 | void bdrv_set_geometry_hint(BlockDriverState *bs, |
b338082b FB |
2216 | int cyls, int heads, int secs) |
2217 | { | |
2218 | bs->cyls = cyls; | |
2219 | bs->heads = heads; | |
2220 | bs->secs = secs; | |
2221 | } | |
2222 | ||
46d4767d FB |
2223 | void bdrv_set_translation_hint(BlockDriverState *bs, int translation) |
2224 | { | |
2225 | bs->translation = translation; | |
2226 | } | |
2227 | ||
5fafdf24 | 2228 | void bdrv_get_geometry_hint(BlockDriverState *bs, |
b338082b FB |
2229 | int *pcyls, int *pheads, int *psecs) |
2230 | { | |
2231 | *pcyls = bs->cyls; | |
2232 | *pheads = bs->heads; | |
2233 | *psecs = bs->secs; | |
2234 | } | |
2235 | ||
0563e191 ZYW |
2236 | /* throttling disk io limits */ |
2237 | void bdrv_set_io_limits(BlockDriverState *bs, | |
2238 | BlockIOLimit *io_limits) | |
2239 | { | |
2240 | bs->io_limits = *io_limits; | |
2241 | bs->io_limits_enabled = bdrv_io_limits_enabled(bs); | |
2242 | } | |
2243 | ||
5bbdbb46 BS |
2244 | /* Recognize floppy formats */ |
2245 | typedef struct FDFormat { | |
2246 | FDriveType drive; | |
2247 | uint8_t last_sect; | |
2248 | uint8_t max_track; | |
2249 | uint8_t max_head; | |
f8d3d128 | 2250 | FDriveRate rate; |
5bbdbb46 BS |
2251 | } FDFormat; |
2252 | ||
2253 | static const FDFormat fd_formats[] = { | |
2254 | /* First entry is default format */ | |
2255 | /* 1.44 MB 3"1/2 floppy disks */ | |
f8d3d128 HP |
2256 | { FDRIVE_DRV_144, 18, 80, 1, FDRIVE_RATE_500K, }, |
2257 | { FDRIVE_DRV_144, 20, 80, 1, FDRIVE_RATE_500K, }, | |
2258 | { FDRIVE_DRV_144, 21, 80, 1, FDRIVE_RATE_500K, }, | |
2259 | { FDRIVE_DRV_144, 21, 82, 1, FDRIVE_RATE_500K, }, | |
2260 | { FDRIVE_DRV_144, 21, 83, 1, FDRIVE_RATE_500K, }, | |
2261 | { FDRIVE_DRV_144, 22, 80, 1, FDRIVE_RATE_500K, }, | |
2262 | { FDRIVE_DRV_144, 23, 80, 1, FDRIVE_RATE_500K, }, | |
2263 | { FDRIVE_DRV_144, 24, 80, 1, FDRIVE_RATE_500K, }, | |
5bbdbb46 | 2264 | /* 2.88 MB 3"1/2 floppy disks */ |
f8d3d128 HP |
2265 | { FDRIVE_DRV_288, 36, 80, 1, FDRIVE_RATE_1M, }, |
2266 | { FDRIVE_DRV_288, 39, 80, 1, FDRIVE_RATE_1M, }, | |
2267 | { FDRIVE_DRV_288, 40, 80, 1, FDRIVE_RATE_1M, }, | |
2268 | { FDRIVE_DRV_288, 44, 80, 1, FDRIVE_RATE_1M, }, | |
2269 | { FDRIVE_DRV_288, 48, 80, 1, FDRIVE_RATE_1M, }, | |
5bbdbb46 | 2270 | /* 720 kB 3"1/2 floppy disks */ |
f8d3d128 HP |
2271 | { FDRIVE_DRV_144, 9, 80, 1, FDRIVE_RATE_250K, }, |
2272 | { FDRIVE_DRV_144, 10, 80, 1, FDRIVE_RATE_250K, }, | |
2273 | { FDRIVE_DRV_144, 10, 82, 1, FDRIVE_RATE_250K, }, | |
2274 | { FDRIVE_DRV_144, 10, 83, 1, FDRIVE_RATE_250K, }, | |
2275 | { FDRIVE_DRV_144, 13, 80, 1, FDRIVE_RATE_250K, }, | |
2276 | { FDRIVE_DRV_144, 14, 80, 1, FDRIVE_RATE_250K, }, | |
5bbdbb46 | 2277 | /* 1.2 MB 5"1/4 floppy disks */ |
f8d3d128 HP |
2278 | { FDRIVE_DRV_120, 15, 80, 1, FDRIVE_RATE_500K, }, |
2279 | { FDRIVE_DRV_120, 18, 80, 1, FDRIVE_RATE_500K, }, | |
2280 | { FDRIVE_DRV_120, 18, 82, 1, FDRIVE_RATE_500K, }, | |
2281 | { FDRIVE_DRV_120, 18, 83, 1, FDRIVE_RATE_500K, }, | |
2282 | { FDRIVE_DRV_120, 20, 80, 1, FDRIVE_RATE_500K, }, | |
5bbdbb46 | 2283 | /* 720 kB 5"1/4 floppy disks */ |
f8d3d128 HP |
2284 | { FDRIVE_DRV_120, 9, 80, 1, FDRIVE_RATE_250K, }, |
2285 | { FDRIVE_DRV_120, 11, 80, 1, FDRIVE_RATE_250K, }, | |
5bbdbb46 | 2286 | /* 360 kB 5"1/4 floppy disks */ |
f8d3d128 HP |
2287 | { FDRIVE_DRV_120, 9, 40, 1, FDRIVE_RATE_300K, }, |
2288 | { FDRIVE_DRV_120, 9, 40, 0, FDRIVE_RATE_300K, }, | |
2289 | { FDRIVE_DRV_120, 10, 41, 1, FDRIVE_RATE_300K, }, | |
2290 | { FDRIVE_DRV_120, 10, 42, 1, FDRIVE_RATE_300K, }, | |
5bbdbb46 | 2291 | /* 320 kB 5"1/4 floppy disks */ |
f8d3d128 HP |
2292 | { FDRIVE_DRV_120, 8, 40, 1, FDRIVE_RATE_250K, }, |
2293 | { FDRIVE_DRV_120, 8, 40, 0, FDRIVE_RATE_250K, }, | |
5bbdbb46 | 2294 | /* 360 kB must match 5"1/4 better than 3"1/2... */ |
f8d3d128 | 2295 | { FDRIVE_DRV_144, 9, 80, 0, FDRIVE_RATE_250K, }, |
5bbdbb46 | 2296 | /* end */ |
f8d3d128 | 2297 | { FDRIVE_DRV_NONE, -1, -1, 0, 0, }, |
5bbdbb46 BS |
2298 | }; |
2299 | ||
2300 | void bdrv_get_floppy_geometry_hint(BlockDriverState *bs, int *nb_heads, | |
2301 | int *max_track, int *last_sect, | |
f8d3d128 HP |
2302 | FDriveType drive_in, FDriveType *drive, |
2303 | FDriveRate *rate) | |
5bbdbb46 BS |
2304 | { |
2305 | const FDFormat *parse; | |
2306 | uint64_t nb_sectors, size; | |
2307 | int i, first_match, match; | |
2308 | ||
2309 | bdrv_get_geometry_hint(bs, nb_heads, max_track, last_sect); | |
2310 | if (*nb_heads != 0 && *max_track != 0 && *last_sect != 0) { | |
2311 | /* User defined disk */ | |
f8d3d128 | 2312 | *rate = FDRIVE_RATE_500K; |
5bbdbb46 BS |
2313 | } else { |
2314 | bdrv_get_geometry(bs, &nb_sectors); | |
2315 | match = -1; | |
2316 | first_match = -1; | |
2317 | for (i = 0; ; i++) { | |
2318 | parse = &fd_formats[i]; | |
2319 | if (parse->drive == FDRIVE_DRV_NONE) { | |
2320 | break; | |
2321 | } | |
2322 | if (drive_in == parse->drive || | |
2323 | drive_in == FDRIVE_DRV_NONE) { | |
2324 | size = (parse->max_head + 1) * parse->max_track * | |
2325 | parse->last_sect; | |
2326 | if (nb_sectors == size) { | |
2327 | match = i; | |
2328 | break; | |
2329 | } | |
2330 | if (first_match == -1) { | |
2331 | first_match = i; | |
2332 | } | |
2333 | } | |
2334 | } | |
2335 | if (match == -1) { | |
2336 | if (first_match == -1) { | |
2337 | match = 1; | |
2338 | } else { | |
2339 | match = first_match; | |
2340 | } | |
2341 | parse = &fd_formats[match]; | |
2342 | } | |
2343 | *nb_heads = parse->max_head + 1; | |
2344 | *max_track = parse->max_track; | |
2345 | *last_sect = parse->last_sect; | |
2346 | *drive = parse->drive; | |
f8d3d128 | 2347 | *rate = parse->rate; |
5bbdbb46 BS |
2348 | } |
2349 | } | |
2350 | ||
46d4767d FB |
2351 | int bdrv_get_translation_hint(BlockDriverState *bs) |
2352 | { | |
2353 | return bs->translation; | |
2354 | } | |
2355 | ||
abd7f68d MA |
2356 | void bdrv_set_on_error(BlockDriverState *bs, BlockErrorAction on_read_error, |
2357 | BlockErrorAction on_write_error) | |
2358 | { | |
2359 | bs->on_read_error = on_read_error; | |
2360 | bs->on_write_error = on_write_error; | |
2361 | } | |
2362 | ||
2363 | BlockErrorAction bdrv_get_on_error(BlockDriverState *bs, int is_read) | |
2364 | { | |
2365 | return is_read ? bs->on_read_error : bs->on_write_error; | |
2366 | } | |
2367 | ||
b338082b FB |
2368 | int bdrv_is_read_only(BlockDriverState *bs) |
2369 | { | |
2370 | return bs->read_only; | |
2371 | } | |
2372 | ||
985a03b0 TS |
2373 | int bdrv_is_sg(BlockDriverState *bs) |
2374 | { | |
2375 | return bs->sg; | |
2376 | } | |
2377 | ||
e900a7b7 CH |
2378 | int bdrv_enable_write_cache(BlockDriverState *bs) |
2379 | { | |
2380 | return bs->enable_write_cache; | |
2381 | } | |
2382 | ||
425b0148 PB |
2383 | void bdrv_set_enable_write_cache(BlockDriverState *bs, bool wce) |
2384 | { | |
2385 | bs->enable_write_cache = wce; | |
2386 | } | |
2387 | ||
ea2384d3 FB |
2388 | int bdrv_is_encrypted(BlockDriverState *bs) |
2389 | { | |
2390 | if (bs->backing_hd && bs->backing_hd->encrypted) | |
2391 | return 1; | |
2392 | return bs->encrypted; | |
2393 | } | |
2394 | ||
c0f4ce77 AL |
2395 | int bdrv_key_required(BlockDriverState *bs) |
2396 | { | |
2397 | BlockDriverState *backing_hd = bs->backing_hd; | |
2398 | ||
2399 | if (backing_hd && backing_hd->encrypted && !backing_hd->valid_key) | |
2400 | return 1; | |
2401 | return (bs->encrypted && !bs->valid_key); | |
2402 | } | |
2403 | ||
ea2384d3 FB |
2404 | int bdrv_set_key(BlockDriverState *bs, const char *key) |
2405 | { | |
2406 | int ret; | |
2407 | if (bs->backing_hd && bs->backing_hd->encrypted) { | |
2408 | ret = bdrv_set_key(bs->backing_hd, key); | |
2409 | if (ret < 0) | |
2410 | return ret; | |
2411 | if (!bs->encrypted) | |
2412 | return 0; | |
2413 | } | |
fd04a2ae SH |
2414 | if (!bs->encrypted) { |
2415 | return -EINVAL; | |
2416 | } else if (!bs->drv || !bs->drv->bdrv_set_key) { | |
2417 | return -ENOMEDIUM; | |
2418 | } | |
c0f4ce77 | 2419 | ret = bs->drv->bdrv_set_key(bs, key); |
bb5fc20f AL |
2420 | if (ret < 0) { |
2421 | bs->valid_key = 0; | |
2422 | } else if (!bs->valid_key) { | |
2423 | bs->valid_key = 1; | |
2424 | /* call the change callback now, we skipped it on open */ | |
7d4b4ba5 | 2425 | bdrv_dev_change_media_cb(bs, true); |
bb5fc20f | 2426 | } |
c0f4ce77 | 2427 | return ret; |
ea2384d3 FB |
2428 | } |
2429 | ||
2430 | void bdrv_get_format(BlockDriverState *bs, char *buf, int buf_size) | |
2431 | { | |
19cb3738 | 2432 | if (!bs->drv) { |
ea2384d3 FB |
2433 | buf[0] = '\0'; |
2434 | } else { | |
2435 | pstrcpy(buf, buf_size, bs->drv->format_name); | |
2436 | } | |
2437 | } | |
2438 | ||
5fafdf24 | 2439 | void bdrv_iterate_format(void (*it)(void *opaque, const char *name), |
ea2384d3 FB |
2440 | void *opaque) |
2441 | { | |
2442 | BlockDriver *drv; | |
2443 | ||
8a22f02a | 2444 | QLIST_FOREACH(drv, &bdrv_drivers, list) { |
ea2384d3 FB |
2445 | it(opaque, drv->format_name); |
2446 | } | |
2447 | } | |
2448 | ||
b338082b FB |
2449 | BlockDriverState *bdrv_find(const char *name) |
2450 | { | |
2451 | BlockDriverState *bs; | |
2452 | ||
1b7bdbc1 SH |
2453 | QTAILQ_FOREACH(bs, &bdrv_states, list) { |
2454 | if (!strcmp(name, bs->device_name)) { | |
b338082b | 2455 | return bs; |
1b7bdbc1 | 2456 | } |
b338082b FB |
2457 | } |
2458 | return NULL; | |
2459 | } | |
2460 | ||
2f399b0a MA |
2461 | BlockDriverState *bdrv_next(BlockDriverState *bs) |
2462 | { | |
2463 | if (!bs) { | |
2464 | return QTAILQ_FIRST(&bdrv_states); | |
2465 | } | |
2466 | return QTAILQ_NEXT(bs, list); | |
2467 | } | |
2468 | ||
51de9760 | 2469 | void bdrv_iterate(void (*it)(void *opaque, BlockDriverState *bs), void *opaque) |
81d0912d FB |
2470 | { |
2471 | BlockDriverState *bs; | |
2472 | ||
1b7bdbc1 | 2473 | QTAILQ_FOREACH(bs, &bdrv_states, list) { |
51de9760 | 2474 | it(opaque, bs); |
81d0912d FB |
2475 | } |
2476 | } | |
2477 | ||
ea2384d3 FB |
2478 | const char *bdrv_get_device_name(BlockDriverState *bs) |
2479 | { | |
2480 | return bs->device_name; | |
2481 | } | |
2482 | ||
c8433287 MA |
2483 | int bdrv_get_flags(BlockDriverState *bs) |
2484 | { | |
2485 | return bs->open_flags; | |
2486 | } | |
2487 | ||
c6ca28d6 AL |
2488 | void bdrv_flush_all(void) |
2489 | { | |
2490 | BlockDriverState *bs; | |
2491 | ||
1b7bdbc1 | 2492 | QTAILQ_FOREACH(bs, &bdrv_states, list) { |
29cdb251 | 2493 | bdrv_flush(bs); |
1b7bdbc1 | 2494 | } |
c6ca28d6 AL |
2495 | } |
2496 | ||
f2feebbd KW |
2497 | int bdrv_has_zero_init(BlockDriverState *bs) |
2498 | { | |
2499 | assert(bs->drv); | |
2500 | ||
336c1c12 KW |
2501 | if (bs->drv->bdrv_has_zero_init) { |
2502 | return bs->drv->bdrv_has_zero_init(bs); | |
f2feebbd KW |
2503 | } |
2504 | ||
2505 | return 1; | |
2506 | } | |
2507 | ||
376ae3f1 SH |
2508 | typedef struct BdrvCoIsAllocatedData { |
2509 | BlockDriverState *bs; | |
2510 | int64_t sector_num; | |
2511 | int nb_sectors; | |
2512 | int *pnum; | |
2513 | int ret; | |
2514 | bool done; | |
2515 | } BdrvCoIsAllocatedData; | |
2516 | ||
f58c7b35 TS |
2517 | /* |
2518 | * Returns true iff the specified sector is present in the disk image. Drivers | |
2519 | * not implementing the functionality are assumed to not support backing files, | |
2520 | * hence all their sectors are reported as allocated. | |
2521 | * | |
bd9533e3 SH |
2522 | * If 'sector_num' is beyond the end of the disk image the return value is 0 |
2523 | * and 'pnum' is set to 0. | |
2524 | * | |
f58c7b35 TS |
2525 | * 'pnum' is set to the number of sectors (including and immediately following |
2526 | * the specified sector) that are known to be in the same | |
2527 | * allocated/unallocated state. | |
2528 | * | |
bd9533e3 SH |
2529 | * 'nb_sectors' is the max value 'pnum' should be set to. If nb_sectors goes |
2530 | * beyond the end of the disk image it will be clamped. | |
f58c7b35 | 2531 | */ |
060f51c9 SH |
2532 | int coroutine_fn bdrv_co_is_allocated(BlockDriverState *bs, int64_t sector_num, |
2533 | int nb_sectors, int *pnum) | |
f58c7b35 | 2534 | { |
bd9533e3 SH |
2535 | int64_t n; |
2536 | ||
2537 | if (sector_num >= bs->total_sectors) { | |
2538 | *pnum = 0; | |
2539 | return 0; | |
2540 | } | |
2541 | ||
2542 | n = bs->total_sectors - sector_num; | |
2543 | if (n < nb_sectors) { | |
2544 | nb_sectors = n; | |
2545 | } | |
2546 | ||
6aebab14 | 2547 | if (!bs->drv->bdrv_co_is_allocated) { |
bd9533e3 | 2548 | *pnum = nb_sectors; |
f58c7b35 TS |
2549 | return 1; |
2550 | } | |
6aebab14 | 2551 | |
060f51c9 SH |
2552 | return bs->drv->bdrv_co_is_allocated(bs, sector_num, nb_sectors, pnum); |
2553 | } | |
2554 | ||
2555 | /* Coroutine wrapper for bdrv_is_allocated() */ | |
2556 | static void coroutine_fn bdrv_is_allocated_co_entry(void *opaque) | |
2557 | { | |
2558 | BdrvCoIsAllocatedData *data = opaque; | |
2559 | BlockDriverState *bs = data->bs; | |
2560 | ||
2561 | data->ret = bdrv_co_is_allocated(bs, data->sector_num, data->nb_sectors, | |
2562 | data->pnum); | |
2563 | data->done = true; | |
2564 | } | |
2565 | ||
2566 | /* | |
2567 | * Synchronous wrapper around bdrv_co_is_allocated(). | |
2568 | * | |
2569 | * See bdrv_co_is_allocated() for details. | |
2570 | */ | |
2571 | int bdrv_is_allocated(BlockDriverState *bs, int64_t sector_num, int nb_sectors, | |
2572 | int *pnum) | |
2573 | { | |
6aebab14 SH |
2574 | Coroutine *co; |
2575 | BdrvCoIsAllocatedData data = { | |
2576 | .bs = bs, | |
2577 | .sector_num = sector_num, | |
2578 | .nb_sectors = nb_sectors, | |
2579 | .pnum = pnum, | |
2580 | .done = false, | |
2581 | }; | |
2582 | ||
2583 | co = qemu_coroutine_create(bdrv_is_allocated_co_entry); | |
2584 | qemu_coroutine_enter(co, &data); | |
2585 | while (!data.done) { | |
2586 | qemu_aio_wait(); | |
2587 | } | |
2588 | return data.ret; | |
f58c7b35 TS |
2589 | } |
2590 | ||
188a7bbf PB |
2591 | /* |
2592 | * Given an image chain: ... -> [BASE] -> [INTER1] -> [INTER2] -> [TOP] | |
2593 | * | |
2594 | * Return true if the given sector is allocated in any image between | |
2595 | * BASE and TOP (inclusive). BASE can be NULL to check if the given | |
2596 | * sector is allocated in any image of the chain. Return false otherwise. | |
2597 | * | |
2598 | * 'pnum' is set to the number of sectors (including and immediately following | |
2599 | * the specified sector) that are known to be in the same | |
2600 | * allocated/unallocated state. | |
2601 | * | |
2602 | */ | |
2603 | int coroutine_fn bdrv_co_is_allocated_above(BlockDriverState *top, | |
2604 | BlockDriverState *base, | |
2605 | int64_t sector_num, | |
2606 | int nb_sectors, int *pnum) | |
2607 | { | |
2608 | BlockDriverState *intermediate; | |
2609 | int ret, n = nb_sectors; | |
2610 | ||
2611 | intermediate = top; | |
2612 | while (intermediate && intermediate != base) { | |
2613 | int pnum_inter; | |
2614 | ret = bdrv_co_is_allocated(intermediate, sector_num, nb_sectors, | |
2615 | &pnum_inter); | |
2616 | if (ret < 0) { | |
2617 | return ret; | |
2618 | } else if (ret) { | |
2619 | *pnum = pnum_inter; | |
2620 | return 1; | |
2621 | } | |
2622 | ||
2623 | /* | |
2624 | * [sector_num, nb_sectors] is unallocated on top but intermediate | |
2625 | * might have | |
2626 | * | |
2627 | * [sector_num+x, nr_sectors] allocated. | |
2628 | */ | |
2629 | if (n > pnum_inter) { | |
2630 | n = pnum_inter; | |
2631 | } | |
2632 | ||
2633 | intermediate = intermediate->backing_hd; | |
2634 | } | |
2635 | ||
2636 | *pnum = n; | |
2637 | return 0; | |
2638 | } | |
2639 | ||
b2023818 | 2640 | BlockInfoList *qmp_query_block(Error **errp) |
b338082b | 2641 | { |
b2023818 | 2642 | BlockInfoList *head = NULL, *cur_item = NULL; |
b338082b FB |
2643 | BlockDriverState *bs; |
2644 | ||
1b7bdbc1 | 2645 | QTAILQ_FOREACH(bs, &bdrv_states, list) { |
b2023818 | 2646 | BlockInfoList *info = g_malloc0(sizeof(*info)); |
d15e5465 | 2647 | |
b2023818 LC |
2648 | info->value = g_malloc0(sizeof(*info->value)); |
2649 | info->value->device = g_strdup(bs->device_name); | |
2650 | info->value->type = g_strdup("unknown"); | |
2651 | info->value->locked = bdrv_dev_is_medium_locked(bs); | |
2652 | info->value->removable = bdrv_dev_has_removable_media(bs); | |
d15e5465 | 2653 | |
e4def80b | 2654 | if (bdrv_dev_has_removable_media(bs)) { |
b2023818 LC |
2655 | info->value->has_tray_open = true; |
2656 | info->value->tray_open = bdrv_dev_is_tray_open(bs); | |
e4def80b | 2657 | } |
f04ef601 LC |
2658 | |
2659 | if (bdrv_iostatus_is_enabled(bs)) { | |
b2023818 LC |
2660 | info->value->has_io_status = true; |
2661 | info->value->io_status = bs->iostatus; | |
f04ef601 LC |
2662 | } |
2663 | ||
19cb3738 | 2664 | if (bs->drv) { |
b2023818 LC |
2665 | info->value->has_inserted = true; |
2666 | info->value->inserted = g_malloc0(sizeof(*info->value->inserted)); | |
2667 | info->value->inserted->file = g_strdup(bs->filename); | |
2668 | info->value->inserted->ro = bs->read_only; | |
2669 | info->value->inserted->drv = g_strdup(bs->drv->format_name); | |
2670 | info->value->inserted->encrypted = bs->encrypted; | |
2671 | if (bs->backing_file[0]) { | |
2672 | info->value->inserted->has_backing_file = true; | |
2673 | info->value->inserted->backing_file = g_strdup(bs->backing_file); | |
376253ec | 2674 | } |
727f005e ZYW |
2675 | |
2676 | if (bs->io_limits_enabled) { | |
2677 | info->value->inserted->bps = | |
2678 | bs->io_limits.bps[BLOCK_IO_LIMIT_TOTAL]; | |
2679 | info->value->inserted->bps_rd = | |
2680 | bs->io_limits.bps[BLOCK_IO_LIMIT_READ]; | |
2681 | info->value->inserted->bps_wr = | |
2682 | bs->io_limits.bps[BLOCK_IO_LIMIT_WRITE]; | |
2683 | info->value->inserted->iops = | |
2684 | bs->io_limits.iops[BLOCK_IO_LIMIT_TOTAL]; | |
2685 | info->value->inserted->iops_rd = | |
2686 | bs->io_limits.iops[BLOCK_IO_LIMIT_READ]; | |
2687 | info->value->inserted->iops_wr = | |
2688 | bs->io_limits.iops[BLOCK_IO_LIMIT_WRITE]; | |
2689 | } | |
b2023818 | 2690 | } |
d15e5465 | 2691 | |
b2023818 LC |
2692 | /* XXX: waiting for the qapi to support GSList */ |
2693 | if (!cur_item) { | |
2694 | head = cur_item = info; | |
2695 | } else { | |
2696 | cur_item->next = info; | |
2697 | cur_item = info; | |
b338082b | 2698 | } |
b338082b | 2699 | } |
d15e5465 | 2700 | |
b2023818 | 2701 | return head; |
b338082b | 2702 | } |
a36e69dd | 2703 | |
f11f57e4 LC |
2704 | /* Consider exposing this as a full fledged QMP command */ |
2705 | static BlockStats *qmp_query_blockstat(const BlockDriverState *bs, Error **errp) | |
2706 | { | |
2707 | BlockStats *s; | |
2708 | ||
2709 | s = g_malloc0(sizeof(*s)); | |
2710 | ||
2711 | if (bs->device_name[0]) { | |
2712 | s->has_device = true; | |
2713 | s->device = g_strdup(bs->device_name); | |
294cc35f KW |
2714 | } |
2715 | ||
f11f57e4 LC |
2716 | s->stats = g_malloc0(sizeof(*s->stats)); |
2717 | s->stats->rd_bytes = bs->nr_bytes[BDRV_ACCT_READ]; | |
2718 | s->stats->wr_bytes = bs->nr_bytes[BDRV_ACCT_WRITE]; | |
2719 | s->stats->rd_operations = bs->nr_ops[BDRV_ACCT_READ]; | |
2720 | s->stats->wr_operations = bs->nr_ops[BDRV_ACCT_WRITE]; | |
2721 | s->stats->wr_highest_offset = bs->wr_highest_sector * BDRV_SECTOR_SIZE; | |
2722 | s->stats->flush_operations = bs->nr_ops[BDRV_ACCT_FLUSH]; | |
2723 | s->stats->wr_total_time_ns = bs->total_time_ns[BDRV_ACCT_WRITE]; | |
2724 | s->stats->rd_total_time_ns = bs->total_time_ns[BDRV_ACCT_READ]; | |
2725 | s->stats->flush_total_time_ns = bs->total_time_ns[BDRV_ACCT_FLUSH]; | |
2726 | ||
294cc35f | 2727 | if (bs->file) { |
f11f57e4 LC |
2728 | s->has_parent = true; |
2729 | s->parent = qmp_query_blockstat(bs->file, NULL); | |
294cc35f KW |
2730 | } |
2731 | ||
f11f57e4 | 2732 | return s; |
294cc35f KW |
2733 | } |
2734 | ||
f11f57e4 | 2735 | BlockStatsList *qmp_query_blockstats(Error **errp) |
218a536a | 2736 | { |
f11f57e4 | 2737 | BlockStatsList *head = NULL, *cur_item = NULL; |
a36e69dd TS |
2738 | BlockDriverState *bs; |
2739 | ||
1b7bdbc1 | 2740 | QTAILQ_FOREACH(bs, &bdrv_states, list) { |
f11f57e4 LC |
2741 | BlockStatsList *info = g_malloc0(sizeof(*info)); |
2742 | info->value = qmp_query_blockstat(bs, NULL); | |
2743 | ||
2744 | /* XXX: waiting for the qapi to support GSList */ | |
2745 | if (!cur_item) { | |
2746 | head = cur_item = info; | |
2747 | } else { | |
2748 | cur_item->next = info; | |
2749 | cur_item = info; | |
2750 | } | |
a36e69dd | 2751 | } |
218a536a | 2752 | |
f11f57e4 | 2753 | return head; |
a36e69dd | 2754 | } |
ea2384d3 | 2755 | |
045df330 AL |
2756 | const char *bdrv_get_encrypted_filename(BlockDriverState *bs) |
2757 | { | |
2758 | if (bs->backing_hd && bs->backing_hd->encrypted) | |
2759 | return bs->backing_file; | |
2760 | else if (bs->encrypted) | |
2761 | return bs->filename; | |
2762 | else | |
2763 | return NULL; | |
2764 | } | |
2765 | ||
5fafdf24 | 2766 | void bdrv_get_backing_filename(BlockDriverState *bs, |
83f64091 FB |
2767 | char *filename, int filename_size) |
2768 | { | |
3574c608 | 2769 | pstrcpy(filename, filename_size, bs->backing_file); |
83f64091 FB |
2770 | } |
2771 | ||
5fafdf24 | 2772 | int bdrv_write_compressed(BlockDriverState *bs, int64_t sector_num, |
faea38e7 FB |
2773 | const uint8_t *buf, int nb_sectors) |
2774 | { | |
2775 | BlockDriver *drv = bs->drv; | |
2776 | if (!drv) | |
19cb3738 | 2777 | return -ENOMEDIUM; |
faea38e7 FB |
2778 | if (!drv->bdrv_write_compressed) |
2779 | return -ENOTSUP; | |
fbb7b4e0 KW |
2780 | if (bdrv_check_request(bs, sector_num, nb_sectors)) |
2781 | return -EIO; | |
a55eb92c | 2782 | |
c6d22830 | 2783 | if (bs->dirty_bitmap) { |
7cd1e32a LS |
2784 | set_dirty_bitmap(bs, sector_num, nb_sectors, 1); |
2785 | } | |
a55eb92c | 2786 | |
faea38e7 FB |
2787 | return drv->bdrv_write_compressed(bs, sector_num, buf, nb_sectors); |
2788 | } | |
3b46e624 | 2789 | |
faea38e7 FB |
2790 | int bdrv_get_info(BlockDriverState *bs, BlockDriverInfo *bdi) |
2791 | { | |
2792 | BlockDriver *drv = bs->drv; | |
2793 | if (!drv) | |
19cb3738 | 2794 | return -ENOMEDIUM; |
faea38e7 FB |
2795 | if (!drv->bdrv_get_info) |
2796 | return -ENOTSUP; | |
2797 | memset(bdi, 0, sizeof(*bdi)); | |
2798 | return drv->bdrv_get_info(bs, bdi); | |
2799 | } | |
2800 | ||
45566e9c CH |
2801 | int bdrv_save_vmstate(BlockDriverState *bs, const uint8_t *buf, |
2802 | int64_t pos, int size) | |
178e08a5 AL |
2803 | { |
2804 | BlockDriver *drv = bs->drv; | |
2805 | if (!drv) | |
2806 | return -ENOMEDIUM; | |
7cdb1f6d MK |
2807 | if (drv->bdrv_save_vmstate) |
2808 | return drv->bdrv_save_vmstate(bs, buf, pos, size); | |
2809 | if (bs->file) | |
2810 | return bdrv_save_vmstate(bs->file, buf, pos, size); | |
2811 | return -ENOTSUP; | |
178e08a5 AL |
2812 | } |
2813 | ||
45566e9c CH |
2814 | int bdrv_load_vmstate(BlockDriverState *bs, uint8_t *buf, |
2815 | int64_t pos, int size) | |
178e08a5 AL |
2816 | { |
2817 | BlockDriver *drv = bs->drv; | |
2818 | if (!drv) | |
2819 | return -ENOMEDIUM; | |
7cdb1f6d MK |
2820 | if (drv->bdrv_load_vmstate) |
2821 | return drv->bdrv_load_vmstate(bs, buf, pos, size); | |
2822 | if (bs->file) | |
2823 | return bdrv_load_vmstate(bs->file, buf, pos, size); | |
2824 | return -ENOTSUP; | |
178e08a5 AL |
2825 | } |
2826 | ||
8b9b0cc2 KW |
2827 | void bdrv_debug_event(BlockDriverState *bs, BlkDebugEvent event) |
2828 | { | |
2829 | BlockDriver *drv = bs->drv; | |
2830 | ||
2831 | if (!drv || !drv->bdrv_debug_event) { | |
2832 | return; | |
2833 | } | |
2834 | ||
2835 | return drv->bdrv_debug_event(bs, event); | |
2836 | ||
2837 | } | |
2838 | ||
faea38e7 FB |
2839 | /**************************************************************/ |
2840 | /* handling of snapshots */ | |
2841 | ||
feeee5ac MDCF |
2842 | int bdrv_can_snapshot(BlockDriverState *bs) |
2843 | { | |
2844 | BlockDriver *drv = bs->drv; | |
07b70bfb | 2845 | if (!drv || !bdrv_is_inserted(bs) || bdrv_is_read_only(bs)) { |
feeee5ac MDCF |
2846 | return 0; |
2847 | } | |
2848 | ||
2849 | if (!drv->bdrv_snapshot_create) { | |
2850 | if (bs->file != NULL) { | |
2851 | return bdrv_can_snapshot(bs->file); | |
2852 | } | |
2853 | return 0; | |
2854 | } | |
2855 | ||
2856 | return 1; | |
2857 | } | |
2858 | ||
199630b6 BS |
2859 | int bdrv_is_snapshot(BlockDriverState *bs) |
2860 | { | |
2861 | return !!(bs->open_flags & BDRV_O_SNAPSHOT); | |
2862 | } | |
2863 | ||
f9092b10 MA |
2864 | BlockDriverState *bdrv_snapshots(void) |
2865 | { | |
2866 | BlockDriverState *bs; | |
2867 | ||
3ac906f7 | 2868 | if (bs_snapshots) { |
f9092b10 | 2869 | return bs_snapshots; |
3ac906f7 | 2870 | } |
f9092b10 MA |
2871 | |
2872 | bs = NULL; | |
2873 | while ((bs = bdrv_next(bs))) { | |
2874 | if (bdrv_can_snapshot(bs)) { | |
3ac906f7 MA |
2875 | bs_snapshots = bs; |
2876 | return bs; | |
f9092b10 MA |
2877 | } |
2878 | } | |
2879 | return NULL; | |
f9092b10 MA |
2880 | } |
2881 | ||
5fafdf24 | 2882 | int bdrv_snapshot_create(BlockDriverState *bs, |
faea38e7 FB |
2883 | QEMUSnapshotInfo *sn_info) |
2884 | { | |
2885 | BlockDriver *drv = bs->drv; | |
2886 | if (!drv) | |
19cb3738 | 2887 | return -ENOMEDIUM; |
7cdb1f6d MK |
2888 | if (drv->bdrv_snapshot_create) |
2889 | return drv->bdrv_snapshot_create(bs, sn_info); | |
2890 | if (bs->file) | |
2891 | return bdrv_snapshot_create(bs->file, sn_info); | |
2892 | return -ENOTSUP; | |
faea38e7 FB |
2893 | } |
2894 | ||
5fafdf24 | 2895 | int bdrv_snapshot_goto(BlockDriverState *bs, |
faea38e7 FB |
2896 | const char *snapshot_id) |
2897 | { | |
2898 | BlockDriver *drv = bs->drv; | |
7cdb1f6d MK |
2899 | int ret, open_ret; |
2900 | ||
faea38e7 | 2901 | if (!drv) |
19cb3738 | 2902 | return -ENOMEDIUM; |
7cdb1f6d MK |
2903 | if (drv->bdrv_snapshot_goto) |
2904 | return drv->bdrv_snapshot_goto(bs, snapshot_id); | |
2905 | ||
2906 | if (bs->file) { | |
2907 | drv->bdrv_close(bs); | |
2908 | ret = bdrv_snapshot_goto(bs->file, snapshot_id); | |
2909 | open_ret = drv->bdrv_open(bs, bs->open_flags); | |
2910 | if (open_ret < 0) { | |
2911 | bdrv_delete(bs->file); | |
2912 | bs->drv = NULL; | |
2913 | return open_ret; | |
2914 | } | |
2915 | return ret; | |
2916 | } | |
2917 | ||
2918 | return -ENOTSUP; | |
faea38e7 FB |
2919 | } |
2920 | ||
2921 | int bdrv_snapshot_delete(BlockDriverState *bs, const char *snapshot_id) | |
2922 | { | |
2923 | BlockDriver *drv = bs->drv; | |
2924 | if (!drv) | |
19cb3738 | 2925 | return -ENOMEDIUM; |
7cdb1f6d MK |
2926 | if (drv->bdrv_snapshot_delete) |
2927 | return drv->bdrv_snapshot_delete(bs, snapshot_id); | |
2928 | if (bs->file) | |
2929 | return bdrv_snapshot_delete(bs->file, snapshot_id); | |
2930 | return -ENOTSUP; | |
faea38e7 FB |
2931 | } |
2932 | ||
5fafdf24 | 2933 | int bdrv_snapshot_list(BlockDriverState *bs, |
faea38e7 FB |
2934 | QEMUSnapshotInfo **psn_info) |
2935 | { | |
2936 | BlockDriver *drv = bs->drv; | |
2937 | if (!drv) | |
19cb3738 | 2938 | return -ENOMEDIUM; |
7cdb1f6d MK |
2939 | if (drv->bdrv_snapshot_list) |
2940 | return drv->bdrv_snapshot_list(bs, psn_info); | |
2941 | if (bs->file) | |
2942 | return bdrv_snapshot_list(bs->file, psn_info); | |
2943 | return -ENOTSUP; | |
faea38e7 FB |
2944 | } |
2945 | ||
51ef6727 | 2946 | int bdrv_snapshot_load_tmp(BlockDriverState *bs, |
2947 | const char *snapshot_name) | |
2948 | { | |
2949 | BlockDriver *drv = bs->drv; | |
2950 | if (!drv) { | |
2951 | return -ENOMEDIUM; | |
2952 | } | |
2953 | if (!bs->read_only) { | |
2954 | return -EINVAL; | |
2955 | } | |
2956 | if (drv->bdrv_snapshot_load_tmp) { | |
2957 | return drv->bdrv_snapshot_load_tmp(bs, snapshot_name); | |
2958 | } | |
2959 | return -ENOTSUP; | |
2960 | } | |
2961 | ||
e8a6bb9c MT |
2962 | BlockDriverState *bdrv_find_backing_image(BlockDriverState *bs, |
2963 | const char *backing_file) | |
2964 | { | |
2965 | if (!bs->drv) { | |
2966 | return NULL; | |
2967 | } | |
2968 | ||
2969 | if (bs->backing_hd) { | |
2970 | if (strcmp(bs->backing_file, backing_file) == 0) { | |
2971 | return bs->backing_hd; | |
2972 | } else { | |
2973 | return bdrv_find_backing_image(bs->backing_hd, backing_file); | |
2974 | } | |
2975 | } | |
2976 | ||
2977 | return NULL; | |
2978 | } | |
2979 | ||
faea38e7 FB |
2980 | #define NB_SUFFIXES 4 |
2981 | ||
2982 | char *get_human_readable_size(char *buf, int buf_size, int64_t size) | |
2983 | { | |
2984 | static const char suffixes[NB_SUFFIXES] = "KMGT"; | |
2985 | int64_t base; | |
2986 | int i; | |
2987 | ||
2988 | if (size <= 999) { | |
2989 | snprintf(buf, buf_size, "%" PRId64, size); | |
2990 | } else { | |
2991 | base = 1024; | |
2992 | for(i = 0; i < NB_SUFFIXES; i++) { | |
2993 | if (size < (10 * base)) { | |
5fafdf24 | 2994 | snprintf(buf, buf_size, "%0.1f%c", |
faea38e7 FB |
2995 | (double)size / base, |
2996 | suffixes[i]); | |
2997 | break; | |
2998 | } else if (size < (1000 * base) || i == (NB_SUFFIXES - 1)) { | |
5fafdf24 | 2999 | snprintf(buf, buf_size, "%" PRId64 "%c", |
faea38e7 FB |
3000 | ((size + (base >> 1)) / base), |
3001 | suffixes[i]); | |
3002 | break; | |
3003 | } | |
3004 | base = base * 1024; | |
3005 | } | |
3006 | } | |
3007 | return buf; | |
3008 | } | |
3009 | ||
3010 | char *bdrv_snapshot_dump(char *buf, int buf_size, QEMUSnapshotInfo *sn) | |
3011 | { | |
3012 | char buf1[128], date_buf[128], clock_buf[128]; | |
3b9f94e1 FB |
3013 | #ifdef _WIN32 |
3014 | struct tm *ptm; | |
3015 | #else | |
faea38e7 | 3016 | struct tm tm; |
3b9f94e1 | 3017 | #endif |
faea38e7 FB |
3018 | time_t ti; |
3019 | int64_t secs; | |
3020 | ||
3021 | if (!sn) { | |
5fafdf24 TS |
3022 | snprintf(buf, buf_size, |
3023 | "%-10s%-20s%7s%20s%15s", | |
faea38e7 FB |
3024 | "ID", "TAG", "VM SIZE", "DATE", "VM CLOCK"); |
3025 | } else { | |
3026 | ti = sn->date_sec; | |
3b9f94e1 FB |
3027 | #ifdef _WIN32 |
3028 | ptm = localtime(&ti); | |
3029 | strftime(date_buf, sizeof(date_buf), | |
3030 | "%Y-%m-%d %H:%M:%S", ptm); | |
3031 | #else | |
faea38e7 FB |
3032 | localtime_r(&ti, &tm); |
3033 | strftime(date_buf, sizeof(date_buf), | |
3034 | "%Y-%m-%d %H:%M:%S", &tm); | |
3b9f94e1 | 3035 | #endif |
faea38e7 FB |
3036 | secs = sn->vm_clock_nsec / 1000000000; |
3037 | snprintf(clock_buf, sizeof(clock_buf), | |
3038 | "%02d:%02d:%02d.%03d", | |
3039 | (int)(secs / 3600), | |
3040 | (int)((secs / 60) % 60), | |
5fafdf24 | 3041 | (int)(secs % 60), |
faea38e7 FB |
3042 | (int)((sn->vm_clock_nsec / 1000000) % 1000)); |
3043 | snprintf(buf, buf_size, | |
5fafdf24 | 3044 | "%-10s%-20s%7s%20s%15s", |
faea38e7 FB |
3045 | sn->id_str, sn->name, |
3046 | get_human_readable_size(buf1, sizeof(buf1), sn->vm_state_size), | |
3047 | date_buf, | |
3048 | clock_buf); | |
3049 | } | |
3050 | return buf; | |
3051 | } | |
3052 | ||
ea2384d3 | 3053 | /**************************************************************/ |
83f64091 | 3054 | /* async I/Os */ |
ea2384d3 | 3055 | |
3b69e4b9 | 3056 | BlockDriverAIOCB *bdrv_aio_readv(BlockDriverState *bs, int64_t sector_num, |
f141eafe | 3057 | QEMUIOVector *qiov, int nb_sectors, |
3b69e4b9 | 3058 | BlockDriverCompletionFunc *cb, void *opaque) |
83f64091 | 3059 | { |
bbf0a440 SH |
3060 | trace_bdrv_aio_readv(bs, sector_num, nb_sectors, opaque); |
3061 | ||
b2a61371 | 3062 | return bdrv_co_aio_rw_vector(bs, sector_num, qiov, nb_sectors, |
8c5873d6 | 3063 | cb, opaque, false); |
ea2384d3 FB |
3064 | } |
3065 | ||
f141eafe AL |
3066 | BlockDriverAIOCB *bdrv_aio_writev(BlockDriverState *bs, int64_t sector_num, |
3067 | QEMUIOVector *qiov, int nb_sectors, | |
3068 | BlockDriverCompletionFunc *cb, void *opaque) | |
ea2384d3 | 3069 | { |
bbf0a440 SH |
3070 | trace_bdrv_aio_writev(bs, sector_num, nb_sectors, opaque); |
3071 | ||
1a6e115b | 3072 | return bdrv_co_aio_rw_vector(bs, sector_num, qiov, nb_sectors, |
8c5873d6 | 3073 | cb, opaque, true); |
83f64091 FB |
3074 | } |
3075 | ||
40b4f539 KW |
3076 | |
3077 | typedef struct MultiwriteCB { | |
3078 | int error; | |
3079 | int num_requests; | |
3080 | int num_callbacks; | |
3081 | struct { | |
3082 | BlockDriverCompletionFunc *cb; | |
3083 | void *opaque; | |
3084 | QEMUIOVector *free_qiov; | |
40b4f539 KW |
3085 | } callbacks[]; |
3086 | } MultiwriteCB; | |
3087 | ||
3088 | static void multiwrite_user_cb(MultiwriteCB *mcb) | |
3089 | { | |
3090 | int i; | |
3091 | ||
3092 | for (i = 0; i < mcb->num_callbacks; i++) { | |
3093 | mcb->callbacks[i].cb(mcb->callbacks[i].opaque, mcb->error); | |
1e1ea48d SH |
3094 | if (mcb->callbacks[i].free_qiov) { |
3095 | qemu_iovec_destroy(mcb->callbacks[i].free_qiov); | |
3096 | } | |
7267c094 | 3097 | g_free(mcb->callbacks[i].free_qiov); |
40b4f539 KW |
3098 | } |
3099 | } | |
3100 | ||
3101 | static void multiwrite_cb(void *opaque, int ret) | |
3102 | { | |
3103 | MultiwriteCB *mcb = opaque; | |
3104 | ||
6d519a5f SH |
3105 | trace_multiwrite_cb(mcb, ret); |
3106 | ||
cb6d3ca0 | 3107 | if (ret < 0 && !mcb->error) { |
40b4f539 | 3108 | mcb->error = ret; |
40b4f539 KW |
3109 | } |
3110 | ||
3111 | mcb->num_requests--; | |
3112 | if (mcb->num_requests == 0) { | |
de189a1b | 3113 | multiwrite_user_cb(mcb); |
7267c094 | 3114 | g_free(mcb); |
40b4f539 KW |
3115 | } |
3116 | } | |
3117 | ||
3118 | static int multiwrite_req_compare(const void *a, const void *b) | |
3119 | { | |
77be4366 CH |
3120 | const BlockRequest *req1 = a, *req2 = b; |
3121 | ||
3122 | /* | |
3123 | * Note that we can't simply subtract req2->sector from req1->sector | |
3124 | * here as that could overflow the return value. | |
3125 | */ | |
3126 | if (req1->sector > req2->sector) { | |
3127 | return 1; | |
3128 | } else if (req1->sector < req2->sector) { | |
3129 | return -1; | |
3130 | } else { | |
3131 | return 0; | |
3132 | } | |
40b4f539 KW |
3133 | } |
3134 | ||
3135 | /* | |
3136 | * Takes a bunch of requests and tries to merge them. Returns the number of | |
3137 | * requests that remain after merging. | |
3138 | */ | |
3139 | static int multiwrite_merge(BlockDriverState *bs, BlockRequest *reqs, | |
3140 | int num_reqs, MultiwriteCB *mcb) | |
3141 | { | |
3142 | int i, outidx; | |
3143 | ||
3144 | // Sort requests by start sector | |
3145 | qsort(reqs, num_reqs, sizeof(*reqs), &multiwrite_req_compare); | |
3146 | ||
3147 | // Check if adjacent requests touch the same clusters. If so, combine them, | |
3148 | // filling up gaps with zero sectors. | |
3149 | outidx = 0; | |
3150 | for (i = 1; i < num_reqs; i++) { | |
3151 | int merge = 0; | |
3152 | int64_t oldreq_last = reqs[outidx].sector + reqs[outidx].nb_sectors; | |
3153 | ||
b6a127a1 | 3154 | // Handle exactly sequential writes and overlapping writes. |
40b4f539 KW |
3155 | if (reqs[i].sector <= oldreq_last) { |
3156 | merge = 1; | |
3157 | } | |
3158 | ||
e2a305fb CH |
3159 | if (reqs[outidx].qiov->niov + reqs[i].qiov->niov + 1 > IOV_MAX) { |
3160 | merge = 0; | |
3161 | } | |
3162 | ||
40b4f539 KW |
3163 | if (merge) { |
3164 | size_t size; | |
7267c094 | 3165 | QEMUIOVector *qiov = g_malloc0(sizeof(*qiov)); |
40b4f539 KW |
3166 | qemu_iovec_init(qiov, |
3167 | reqs[outidx].qiov->niov + reqs[i].qiov->niov + 1); | |
3168 | ||
3169 | // Add the first request to the merged one. If the requests are | |
3170 | // overlapping, drop the last sectors of the first request. | |
3171 | size = (reqs[i].sector - reqs[outidx].sector) << 9; | |
3172 | qemu_iovec_concat(qiov, reqs[outidx].qiov, size); | |
3173 | ||
b6a127a1 PB |
3174 | // We should need to add any zeros between the two requests |
3175 | assert (reqs[i].sector <= oldreq_last); | |
40b4f539 KW |
3176 | |
3177 | // Add the second request | |
3178 | qemu_iovec_concat(qiov, reqs[i].qiov, reqs[i].qiov->size); | |
3179 | ||
cbf1dff2 | 3180 | reqs[outidx].nb_sectors = qiov->size >> 9; |
40b4f539 KW |
3181 | reqs[outidx].qiov = qiov; |
3182 | ||
3183 | mcb->callbacks[i].free_qiov = reqs[outidx].qiov; | |
3184 | } else { | |
3185 | outidx++; | |
3186 | reqs[outidx].sector = reqs[i].sector; | |
3187 | reqs[outidx].nb_sectors = reqs[i].nb_sectors; | |
3188 | reqs[outidx].qiov = reqs[i].qiov; | |
3189 | } | |
3190 | } | |
3191 | ||
3192 | return outidx + 1; | |
3193 | } | |
3194 | ||
3195 | /* | |
3196 | * Submit multiple AIO write requests at once. | |
3197 | * | |
3198 | * On success, the function returns 0 and all requests in the reqs array have | |
3199 | * been submitted. In error case this function returns -1, and any of the | |
3200 | * requests may or may not be submitted yet. In particular, this means that the | |
3201 | * callback will be called for some of the requests, for others it won't. The | |
3202 | * caller must check the error field of the BlockRequest to wait for the right | |
3203 | * callbacks (if error != 0, no callback will be called). | |
3204 | * | |
3205 | * The implementation may modify the contents of the reqs array, e.g. to merge | |
3206 | * requests. However, the fields opaque and error are left unmodified as they | |
3207 | * are used to signal failure for a single request to the caller. | |
3208 | */ | |
3209 | int bdrv_aio_multiwrite(BlockDriverState *bs, BlockRequest *reqs, int num_reqs) | |
3210 | { | |
40b4f539 KW |
3211 | MultiwriteCB *mcb; |
3212 | int i; | |
3213 | ||
301db7c2 RH |
3214 | /* don't submit writes if we don't have a medium */ |
3215 | if (bs->drv == NULL) { | |
3216 | for (i = 0; i < num_reqs; i++) { | |
3217 | reqs[i].error = -ENOMEDIUM; | |
3218 | } | |
3219 | return -1; | |
3220 | } | |
3221 | ||
40b4f539 KW |
3222 | if (num_reqs == 0) { |
3223 | return 0; | |
3224 | } | |
3225 | ||
3226 | // Create MultiwriteCB structure | |
7267c094 | 3227 | mcb = g_malloc0(sizeof(*mcb) + num_reqs * sizeof(*mcb->callbacks)); |
40b4f539 KW |
3228 | mcb->num_requests = 0; |
3229 | mcb->num_callbacks = num_reqs; | |
3230 | ||
3231 | for (i = 0; i < num_reqs; i++) { | |
3232 | mcb->callbacks[i].cb = reqs[i].cb; | |
3233 | mcb->callbacks[i].opaque = reqs[i].opaque; | |
3234 | } | |
3235 | ||
3236 | // Check for mergable requests | |
3237 | num_reqs = multiwrite_merge(bs, reqs, num_reqs, mcb); | |
3238 | ||
6d519a5f SH |
3239 | trace_bdrv_aio_multiwrite(mcb, mcb->num_callbacks, num_reqs); |
3240 | ||
df9309fb PB |
3241 | /* Run the aio requests. */ |
3242 | mcb->num_requests = num_reqs; | |
40b4f539 | 3243 | for (i = 0; i < num_reqs; i++) { |
ad54ae80 | 3244 | bdrv_aio_writev(bs, reqs[i].sector, reqs[i].qiov, |
40b4f539 | 3245 | reqs[i].nb_sectors, multiwrite_cb, mcb); |
40b4f539 KW |
3246 | } |
3247 | ||
3248 | return 0; | |
40b4f539 KW |
3249 | } |
3250 | ||
83f64091 | 3251 | void bdrv_aio_cancel(BlockDriverAIOCB *acb) |
83f64091 | 3252 | { |
6bbff9a0 | 3253 | acb->pool->cancel(acb); |
83f64091 FB |
3254 | } |
3255 | ||
98f90dba ZYW |
3256 | /* block I/O throttling */ |
3257 | static bool bdrv_exceed_bps_limits(BlockDriverState *bs, int nb_sectors, | |
3258 | bool is_write, double elapsed_time, uint64_t *wait) | |
3259 | { | |
3260 | uint64_t bps_limit = 0; | |
3261 | double bytes_limit, bytes_base, bytes_res; | |
3262 | double slice_time, wait_time; | |
3263 | ||
3264 | if (bs->io_limits.bps[BLOCK_IO_LIMIT_TOTAL]) { | |
3265 | bps_limit = bs->io_limits.bps[BLOCK_IO_LIMIT_TOTAL]; | |
3266 | } else if (bs->io_limits.bps[is_write]) { | |
3267 | bps_limit = bs->io_limits.bps[is_write]; | |
3268 | } else { | |
3269 | if (wait) { | |
3270 | *wait = 0; | |
3271 | } | |
3272 | ||
3273 | return false; | |
3274 | } | |
3275 | ||
3276 | slice_time = bs->slice_end - bs->slice_start; | |
3277 | slice_time /= (NANOSECONDS_PER_SECOND); | |
3278 | bytes_limit = bps_limit * slice_time; | |
3279 | bytes_base = bs->nr_bytes[is_write] - bs->io_base.bytes[is_write]; | |
3280 | if (bs->io_limits.bps[BLOCK_IO_LIMIT_TOTAL]) { | |
3281 | bytes_base += bs->nr_bytes[!is_write] - bs->io_base.bytes[!is_write]; | |
3282 | } | |
3283 | ||
3284 | /* bytes_base: the bytes of data which have been read/written; and | |
3285 | * it is obtained from the history statistic info. | |
3286 | * bytes_res: the remaining bytes of data which need to be read/written. | |
3287 | * (bytes_base + bytes_res) / bps_limit: used to calcuate | |
3288 | * the total time for completing reading/writting all data. | |
3289 | */ | |
3290 | bytes_res = (unsigned) nb_sectors * BDRV_SECTOR_SIZE; | |
3291 | ||
3292 | if (bytes_base + bytes_res <= bytes_limit) { | |
3293 | if (wait) { | |
3294 | *wait = 0; | |
3295 | } | |
3296 | ||
3297 | return false; | |
3298 | } | |
3299 | ||
3300 | /* Calc approx time to dispatch */ | |
3301 | wait_time = (bytes_base + bytes_res) / bps_limit - elapsed_time; | |
3302 | ||
3303 | /* When the I/O rate at runtime exceeds the limits, | |
3304 | * bs->slice_end need to be extended in order that the current statistic | |
3305 | * info can be kept until the timer fire, so it is increased and tuned | |
3306 | * based on the result of experiment. | |
3307 | */ | |
3308 | bs->slice_time = wait_time * BLOCK_IO_SLICE_TIME * 10; | |
3309 | bs->slice_end += bs->slice_time - 3 * BLOCK_IO_SLICE_TIME; | |
3310 | if (wait) { | |
3311 | *wait = wait_time * BLOCK_IO_SLICE_TIME * 10; | |
3312 | } | |
3313 | ||
3314 | return true; | |
3315 | } | |
3316 | ||
3317 | static bool bdrv_exceed_iops_limits(BlockDriverState *bs, bool is_write, | |
3318 | double elapsed_time, uint64_t *wait) | |
3319 | { | |
3320 | uint64_t iops_limit = 0; | |
3321 | double ios_limit, ios_base; | |
3322 | double slice_time, wait_time; | |
3323 | ||
3324 | if (bs->io_limits.iops[BLOCK_IO_LIMIT_TOTAL]) { | |
3325 | iops_limit = bs->io_limits.iops[BLOCK_IO_LIMIT_TOTAL]; | |
3326 | } else if (bs->io_limits.iops[is_write]) { | |
3327 | iops_limit = bs->io_limits.iops[is_write]; | |
3328 | } else { | |
3329 | if (wait) { | |
3330 | *wait = 0; | |
3331 | } | |
3332 | ||
3333 | return false; | |
3334 | } | |
3335 | ||
3336 | slice_time = bs->slice_end - bs->slice_start; | |
3337 | slice_time /= (NANOSECONDS_PER_SECOND); | |
3338 | ios_limit = iops_limit * slice_time; | |
3339 | ios_base = bs->nr_ops[is_write] - bs->io_base.ios[is_write]; | |
3340 | if (bs->io_limits.iops[BLOCK_IO_LIMIT_TOTAL]) { | |
3341 | ios_base += bs->nr_ops[!is_write] - bs->io_base.ios[!is_write]; | |
3342 | } | |
3343 | ||
3344 | if (ios_base + 1 <= ios_limit) { | |
3345 | if (wait) { | |
3346 | *wait = 0; | |
3347 | } | |
3348 | ||
3349 | return false; | |
3350 | } | |
3351 | ||
3352 | /* Calc approx time to dispatch */ | |
3353 | wait_time = (ios_base + 1) / iops_limit; | |
3354 | if (wait_time > elapsed_time) { | |
3355 | wait_time = wait_time - elapsed_time; | |
3356 | } else { | |
3357 | wait_time = 0; | |
3358 | } | |
3359 | ||
3360 | bs->slice_time = wait_time * BLOCK_IO_SLICE_TIME * 10; | |
3361 | bs->slice_end += bs->slice_time - 3 * BLOCK_IO_SLICE_TIME; | |
3362 | if (wait) { | |
3363 | *wait = wait_time * BLOCK_IO_SLICE_TIME * 10; | |
3364 | } | |
3365 | ||
3366 | return true; | |
3367 | } | |
3368 | ||
3369 | static bool bdrv_exceed_io_limits(BlockDriverState *bs, int nb_sectors, | |
3370 | bool is_write, int64_t *wait) | |
3371 | { | |
3372 | int64_t now, max_wait; | |
3373 | uint64_t bps_wait = 0, iops_wait = 0; | |
3374 | double elapsed_time; | |
3375 | int bps_ret, iops_ret; | |
3376 | ||
3377 | now = qemu_get_clock_ns(vm_clock); | |
3378 | if ((bs->slice_start < now) | |
3379 | && (bs->slice_end > now)) { | |
3380 | bs->slice_end = now + bs->slice_time; | |
3381 | } else { | |
3382 | bs->slice_time = 5 * BLOCK_IO_SLICE_TIME; | |
3383 | bs->slice_start = now; | |
3384 | bs->slice_end = now + bs->slice_time; | |
3385 | ||
3386 | bs->io_base.bytes[is_write] = bs->nr_bytes[is_write]; | |
3387 | bs->io_base.bytes[!is_write] = bs->nr_bytes[!is_write]; | |
3388 | ||
3389 | bs->io_base.ios[is_write] = bs->nr_ops[is_write]; | |
3390 | bs->io_base.ios[!is_write] = bs->nr_ops[!is_write]; | |
3391 | } | |
3392 | ||
3393 | elapsed_time = now - bs->slice_start; | |
3394 | elapsed_time /= (NANOSECONDS_PER_SECOND); | |
3395 | ||
3396 | bps_ret = bdrv_exceed_bps_limits(bs, nb_sectors, | |
3397 | is_write, elapsed_time, &bps_wait); | |
3398 | iops_ret = bdrv_exceed_iops_limits(bs, is_write, | |
3399 | elapsed_time, &iops_wait); | |
3400 | if (bps_ret || iops_ret) { | |
3401 | max_wait = bps_wait > iops_wait ? bps_wait : iops_wait; | |
3402 | if (wait) { | |
3403 | *wait = max_wait; | |
3404 | } | |
3405 | ||
3406 | now = qemu_get_clock_ns(vm_clock); | |
3407 | if (bs->slice_end < now + max_wait) { | |
3408 | bs->slice_end = now + max_wait; | |
3409 | } | |
3410 | ||
3411 | return true; | |
3412 | } | |
3413 | ||
3414 | if (wait) { | |
3415 | *wait = 0; | |
3416 | } | |
3417 | ||
3418 | return false; | |
3419 | } | |
ce1a14dc | 3420 | |
83f64091 FB |
3421 | /**************************************************************/ |
3422 | /* async block device emulation */ | |
3423 | ||
c16b5a2c CH |
3424 | typedef struct BlockDriverAIOCBSync { |
3425 | BlockDriverAIOCB common; | |
3426 | QEMUBH *bh; | |
3427 | int ret; | |
3428 | /* vector translation state */ | |
3429 | QEMUIOVector *qiov; | |
3430 | uint8_t *bounce; | |
3431 | int is_write; | |
3432 | } BlockDriverAIOCBSync; | |
3433 | ||
3434 | static void bdrv_aio_cancel_em(BlockDriverAIOCB *blockacb) | |
3435 | { | |
b666d239 KW |
3436 | BlockDriverAIOCBSync *acb = |
3437 | container_of(blockacb, BlockDriverAIOCBSync, common); | |
6a7ad299 | 3438 | qemu_bh_delete(acb->bh); |
36afc451 | 3439 | acb->bh = NULL; |
c16b5a2c CH |
3440 | qemu_aio_release(acb); |
3441 | } | |
3442 | ||
3443 | static AIOPool bdrv_em_aio_pool = { | |
3444 | .aiocb_size = sizeof(BlockDriverAIOCBSync), | |
3445 | .cancel = bdrv_aio_cancel_em, | |
3446 | }; | |
3447 | ||
ce1a14dc | 3448 | static void bdrv_aio_bh_cb(void *opaque) |
83f64091 | 3449 | { |
ce1a14dc | 3450 | BlockDriverAIOCBSync *acb = opaque; |
f141eafe | 3451 | |
f141eafe AL |
3452 | if (!acb->is_write) |
3453 | qemu_iovec_from_buffer(acb->qiov, acb->bounce, acb->qiov->size); | |
ceb42de8 | 3454 | qemu_vfree(acb->bounce); |
ce1a14dc | 3455 | acb->common.cb(acb->common.opaque, acb->ret); |
6a7ad299 | 3456 | qemu_bh_delete(acb->bh); |
36afc451 | 3457 | acb->bh = NULL; |
ce1a14dc | 3458 | qemu_aio_release(acb); |
83f64091 | 3459 | } |
beac80cd | 3460 | |
f141eafe AL |
3461 | static BlockDriverAIOCB *bdrv_aio_rw_vector(BlockDriverState *bs, |
3462 | int64_t sector_num, | |
3463 | QEMUIOVector *qiov, | |
3464 | int nb_sectors, | |
3465 | BlockDriverCompletionFunc *cb, | |
3466 | void *opaque, | |
3467 | int is_write) | |
3468 | ||
83f64091 | 3469 | { |
ce1a14dc | 3470 | BlockDriverAIOCBSync *acb; |
ce1a14dc | 3471 | |
c16b5a2c | 3472 | acb = qemu_aio_get(&bdrv_em_aio_pool, bs, cb, opaque); |
f141eafe AL |
3473 | acb->is_write = is_write; |
3474 | acb->qiov = qiov; | |
e268ca52 | 3475 | acb->bounce = qemu_blockalign(bs, qiov->size); |
3f3aace8 | 3476 | acb->bh = qemu_bh_new(bdrv_aio_bh_cb, acb); |
f141eafe AL |
3477 | |
3478 | if (is_write) { | |
3479 | qemu_iovec_to_buffer(acb->qiov, acb->bounce); | |
1ed20acf | 3480 | acb->ret = bs->drv->bdrv_write(bs, sector_num, acb->bounce, nb_sectors); |
f141eafe | 3481 | } else { |
1ed20acf | 3482 | acb->ret = bs->drv->bdrv_read(bs, sector_num, acb->bounce, nb_sectors); |
f141eafe AL |
3483 | } |
3484 | ||
ce1a14dc | 3485 | qemu_bh_schedule(acb->bh); |
f141eafe | 3486 | |
ce1a14dc | 3487 | return &acb->common; |
beac80cd FB |
3488 | } |
3489 | ||
f141eafe AL |
3490 | static BlockDriverAIOCB *bdrv_aio_readv_em(BlockDriverState *bs, |
3491 | int64_t sector_num, QEMUIOVector *qiov, int nb_sectors, | |
ce1a14dc | 3492 | BlockDriverCompletionFunc *cb, void *opaque) |
beac80cd | 3493 | { |
f141eafe AL |
3494 | return bdrv_aio_rw_vector(bs, sector_num, qiov, nb_sectors, cb, opaque, 0); |
3495 | } | |
83f64091 | 3496 | |
f141eafe AL |
3497 | static BlockDriverAIOCB *bdrv_aio_writev_em(BlockDriverState *bs, |
3498 | int64_t sector_num, QEMUIOVector *qiov, int nb_sectors, | |
3499 | BlockDriverCompletionFunc *cb, void *opaque) | |
3500 | { | |
3501 | return bdrv_aio_rw_vector(bs, sector_num, qiov, nb_sectors, cb, opaque, 1); | |
beac80cd | 3502 | } |
beac80cd | 3503 | |
68485420 KW |
3504 | |
3505 | typedef struct BlockDriverAIOCBCoroutine { | |
3506 | BlockDriverAIOCB common; | |
3507 | BlockRequest req; | |
3508 | bool is_write; | |
3509 | QEMUBH* bh; | |
3510 | } BlockDriverAIOCBCoroutine; | |
3511 | ||
3512 | static void bdrv_aio_co_cancel_em(BlockDriverAIOCB *blockacb) | |
3513 | { | |
3514 | qemu_aio_flush(); | |
3515 | } | |
3516 | ||
3517 | static AIOPool bdrv_em_co_aio_pool = { | |
3518 | .aiocb_size = sizeof(BlockDriverAIOCBCoroutine), | |
3519 | .cancel = bdrv_aio_co_cancel_em, | |
3520 | }; | |
3521 | ||
35246a68 | 3522 | static void bdrv_co_em_bh(void *opaque) |
68485420 KW |
3523 | { |
3524 | BlockDriverAIOCBCoroutine *acb = opaque; | |
3525 | ||
3526 | acb->common.cb(acb->common.opaque, acb->req.error); | |
3527 | qemu_bh_delete(acb->bh); | |
3528 | qemu_aio_release(acb); | |
3529 | } | |
3530 | ||
b2a61371 SH |
3531 | /* Invoke bdrv_co_do_readv/bdrv_co_do_writev */ |
3532 | static void coroutine_fn bdrv_co_do_rw(void *opaque) | |
3533 | { | |
3534 | BlockDriverAIOCBCoroutine *acb = opaque; | |
3535 | BlockDriverState *bs = acb->common.bs; | |
3536 | ||
3537 | if (!acb->is_write) { | |
3538 | acb->req.error = bdrv_co_do_readv(bs, acb->req.sector, | |
470c0504 | 3539 | acb->req.nb_sectors, acb->req.qiov, 0); |
b2a61371 SH |
3540 | } else { |
3541 | acb->req.error = bdrv_co_do_writev(bs, acb->req.sector, | |
f08f2dda | 3542 | acb->req.nb_sectors, acb->req.qiov, 0); |
b2a61371 SH |
3543 | } |
3544 | ||
35246a68 | 3545 | acb->bh = qemu_bh_new(bdrv_co_em_bh, acb); |
b2a61371 SH |
3546 | qemu_bh_schedule(acb->bh); |
3547 | } | |
3548 | ||
68485420 KW |
3549 | static BlockDriverAIOCB *bdrv_co_aio_rw_vector(BlockDriverState *bs, |
3550 | int64_t sector_num, | |
3551 | QEMUIOVector *qiov, | |
3552 | int nb_sectors, | |
3553 | BlockDriverCompletionFunc *cb, | |
3554 | void *opaque, | |
8c5873d6 | 3555 | bool is_write) |
68485420 KW |
3556 | { |
3557 | Coroutine *co; | |
3558 | BlockDriverAIOCBCoroutine *acb; | |
3559 | ||
3560 | acb = qemu_aio_get(&bdrv_em_co_aio_pool, bs, cb, opaque); | |
3561 | acb->req.sector = sector_num; | |
3562 | acb->req.nb_sectors = nb_sectors; | |
3563 | acb->req.qiov = qiov; | |
3564 | acb->is_write = is_write; | |
3565 | ||
8c5873d6 | 3566 | co = qemu_coroutine_create(bdrv_co_do_rw); |
68485420 KW |
3567 | qemu_coroutine_enter(co, acb); |
3568 | ||
3569 | return &acb->common; | |
3570 | } | |
3571 | ||
07f07615 | 3572 | static void coroutine_fn bdrv_aio_flush_co_entry(void *opaque) |
b2e12bc6 | 3573 | { |
07f07615 PB |
3574 | BlockDriverAIOCBCoroutine *acb = opaque; |
3575 | BlockDriverState *bs = acb->common.bs; | |
b2e12bc6 | 3576 | |
07f07615 PB |
3577 | acb->req.error = bdrv_co_flush(bs); |
3578 | acb->bh = qemu_bh_new(bdrv_co_em_bh, acb); | |
b2e12bc6 | 3579 | qemu_bh_schedule(acb->bh); |
b2e12bc6 CH |
3580 | } |
3581 | ||
07f07615 | 3582 | BlockDriverAIOCB *bdrv_aio_flush(BlockDriverState *bs, |
016f5cf6 AG |
3583 | BlockDriverCompletionFunc *cb, void *opaque) |
3584 | { | |
07f07615 | 3585 | trace_bdrv_aio_flush(bs, opaque); |
016f5cf6 | 3586 | |
07f07615 PB |
3587 | Coroutine *co; |
3588 | BlockDriverAIOCBCoroutine *acb; | |
016f5cf6 | 3589 | |
07f07615 PB |
3590 | acb = qemu_aio_get(&bdrv_em_co_aio_pool, bs, cb, opaque); |
3591 | co = qemu_coroutine_create(bdrv_aio_flush_co_entry); | |
3592 | qemu_coroutine_enter(co, acb); | |
016f5cf6 | 3593 | |
016f5cf6 AG |
3594 | return &acb->common; |
3595 | } | |
3596 | ||
4265d620 PB |
3597 | static void coroutine_fn bdrv_aio_discard_co_entry(void *opaque) |
3598 | { | |
3599 | BlockDriverAIOCBCoroutine *acb = opaque; | |
3600 | BlockDriverState *bs = acb->common.bs; | |
3601 | ||
3602 | acb->req.error = bdrv_co_discard(bs, acb->req.sector, acb->req.nb_sectors); | |
3603 | acb->bh = qemu_bh_new(bdrv_co_em_bh, acb); | |
3604 | qemu_bh_schedule(acb->bh); | |
3605 | } | |
3606 | ||
3607 | BlockDriverAIOCB *bdrv_aio_discard(BlockDriverState *bs, | |
3608 | int64_t sector_num, int nb_sectors, | |
3609 | BlockDriverCompletionFunc *cb, void *opaque) | |
3610 | { | |
3611 | Coroutine *co; | |
3612 | BlockDriverAIOCBCoroutine *acb; | |
3613 | ||
3614 | trace_bdrv_aio_discard(bs, sector_num, nb_sectors, opaque); | |
3615 | ||
3616 | acb = qemu_aio_get(&bdrv_em_co_aio_pool, bs, cb, opaque); | |
3617 | acb->req.sector = sector_num; | |
3618 | acb->req.nb_sectors = nb_sectors; | |
3619 | co = qemu_coroutine_create(bdrv_aio_discard_co_entry); | |
3620 | qemu_coroutine_enter(co, acb); | |
3621 | ||
3622 | return &acb->common; | |
3623 | } | |
3624 | ||
ea2384d3 FB |
3625 | void bdrv_init(void) |
3626 | { | |
5efa9d5a | 3627 | module_call_init(MODULE_INIT_BLOCK); |
ea2384d3 | 3628 | } |
ce1a14dc | 3629 | |
eb852011 MA |
3630 | void bdrv_init_with_whitelist(void) |
3631 | { | |
3632 | use_bdrv_whitelist = 1; | |
3633 | bdrv_init(); | |
3634 | } | |
3635 | ||
c16b5a2c CH |
3636 | void *qemu_aio_get(AIOPool *pool, BlockDriverState *bs, |
3637 | BlockDriverCompletionFunc *cb, void *opaque) | |
ce1a14dc | 3638 | { |
ce1a14dc PB |
3639 | BlockDriverAIOCB *acb; |
3640 | ||
6bbff9a0 AL |
3641 | if (pool->free_aiocb) { |
3642 | acb = pool->free_aiocb; | |
3643 | pool->free_aiocb = acb->next; | |
ce1a14dc | 3644 | } else { |
7267c094 | 3645 | acb = g_malloc0(pool->aiocb_size); |
6bbff9a0 | 3646 | acb->pool = pool; |
ce1a14dc PB |
3647 | } |
3648 | acb->bs = bs; | |
3649 | acb->cb = cb; | |
3650 | acb->opaque = opaque; | |
3651 | return acb; | |
3652 | } | |
3653 | ||
3654 | void qemu_aio_release(void *p) | |
3655 | { | |
6bbff9a0 AL |
3656 | BlockDriverAIOCB *acb = (BlockDriverAIOCB *)p; |
3657 | AIOPool *pool = acb->pool; | |
3658 | acb->next = pool->free_aiocb; | |
3659 | pool->free_aiocb = acb; | |
ce1a14dc | 3660 | } |
19cb3738 | 3661 | |
f9f05dc5 KW |
3662 | /**************************************************************/ |
3663 | /* Coroutine block device emulation */ | |
3664 | ||
3665 | typedef struct CoroutineIOCompletion { | |
3666 | Coroutine *coroutine; | |
3667 | int ret; | |
3668 | } CoroutineIOCompletion; | |
3669 | ||
3670 | static void bdrv_co_io_em_complete(void *opaque, int ret) | |
3671 | { | |
3672 | CoroutineIOCompletion *co = opaque; | |
3673 | ||
3674 | co->ret = ret; | |
3675 | qemu_coroutine_enter(co->coroutine, NULL); | |
3676 | } | |
3677 | ||
3678 | static int coroutine_fn bdrv_co_io_em(BlockDriverState *bs, int64_t sector_num, | |
3679 | int nb_sectors, QEMUIOVector *iov, | |
3680 | bool is_write) | |
3681 | { | |
3682 | CoroutineIOCompletion co = { | |
3683 | .coroutine = qemu_coroutine_self(), | |
3684 | }; | |
3685 | BlockDriverAIOCB *acb; | |
3686 | ||
3687 | if (is_write) { | |
a652d160 SH |
3688 | acb = bs->drv->bdrv_aio_writev(bs, sector_num, iov, nb_sectors, |
3689 | bdrv_co_io_em_complete, &co); | |
f9f05dc5 | 3690 | } else { |
a652d160 SH |
3691 | acb = bs->drv->bdrv_aio_readv(bs, sector_num, iov, nb_sectors, |
3692 | bdrv_co_io_em_complete, &co); | |
f9f05dc5 KW |
3693 | } |
3694 | ||
59370aaa | 3695 | trace_bdrv_co_io_em(bs, sector_num, nb_sectors, is_write, acb); |
f9f05dc5 KW |
3696 | if (!acb) { |
3697 | return -EIO; | |
3698 | } | |
3699 | qemu_coroutine_yield(); | |
3700 | ||
3701 | return co.ret; | |
3702 | } | |
3703 | ||
3704 | static int coroutine_fn bdrv_co_readv_em(BlockDriverState *bs, | |
3705 | int64_t sector_num, int nb_sectors, | |
3706 | QEMUIOVector *iov) | |
3707 | { | |
3708 | return bdrv_co_io_em(bs, sector_num, nb_sectors, iov, false); | |
3709 | } | |
3710 | ||
3711 | static int coroutine_fn bdrv_co_writev_em(BlockDriverState *bs, | |
3712 | int64_t sector_num, int nb_sectors, | |
3713 | QEMUIOVector *iov) | |
3714 | { | |
3715 | return bdrv_co_io_em(bs, sector_num, nb_sectors, iov, true); | |
3716 | } | |
3717 | ||
07f07615 | 3718 | static void coroutine_fn bdrv_flush_co_entry(void *opaque) |
e7a8a783 | 3719 | { |
07f07615 PB |
3720 | RwCo *rwco = opaque; |
3721 | ||
3722 | rwco->ret = bdrv_co_flush(rwco->bs); | |
3723 | } | |
3724 | ||
3725 | int coroutine_fn bdrv_co_flush(BlockDriverState *bs) | |
3726 | { | |
eb489bb1 KW |
3727 | int ret; |
3728 | ||
29cdb251 | 3729 | if (!bs || !bdrv_is_inserted(bs) || bdrv_is_read_only(bs)) { |
07f07615 | 3730 | return 0; |
eb489bb1 KW |
3731 | } |
3732 | ||
ca716364 | 3733 | /* Write back cached data to the OS even with cache=unsafe */ |
eb489bb1 KW |
3734 | if (bs->drv->bdrv_co_flush_to_os) { |
3735 | ret = bs->drv->bdrv_co_flush_to_os(bs); | |
3736 | if (ret < 0) { | |
3737 | return ret; | |
3738 | } | |
3739 | } | |
3740 | ||
ca716364 KW |
3741 | /* But don't actually force it to the disk with cache=unsafe */ |
3742 | if (bs->open_flags & BDRV_O_NO_FLUSH) { | |
3743 | return 0; | |
3744 | } | |
3745 | ||
eb489bb1 | 3746 | if (bs->drv->bdrv_co_flush_to_disk) { |
29cdb251 | 3747 | ret = bs->drv->bdrv_co_flush_to_disk(bs); |
07f07615 PB |
3748 | } else if (bs->drv->bdrv_aio_flush) { |
3749 | BlockDriverAIOCB *acb; | |
3750 | CoroutineIOCompletion co = { | |
3751 | .coroutine = qemu_coroutine_self(), | |
3752 | }; | |
3753 | ||
3754 | acb = bs->drv->bdrv_aio_flush(bs, bdrv_co_io_em_complete, &co); | |
3755 | if (acb == NULL) { | |
29cdb251 | 3756 | ret = -EIO; |
07f07615 PB |
3757 | } else { |
3758 | qemu_coroutine_yield(); | |
29cdb251 | 3759 | ret = co.ret; |
07f07615 | 3760 | } |
07f07615 PB |
3761 | } else { |
3762 | /* | |
3763 | * Some block drivers always operate in either writethrough or unsafe | |
3764 | * mode and don't support bdrv_flush therefore. Usually qemu doesn't | |
3765 | * know how the server works (because the behaviour is hardcoded or | |
3766 | * depends on server-side configuration), so we can't ensure that | |
3767 | * everything is safe on disk. Returning an error doesn't work because | |
3768 | * that would break guests even if the server operates in writethrough | |
3769 | * mode. | |
3770 | * | |
3771 | * Let's hope the user knows what he's doing. | |
3772 | */ | |
29cdb251 | 3773 | ret = 0; |
07f07615 | 3774 | } |
29cdb251 PB |
3775 | if (ret < 0) { |
3776 | return ret; | |
3777 | } | |
3778 | ||
3779 | /* Now flush the underlying protocol. It will also have BDRV_O_NO_FLUSH | |
3780 | * in the case of cache=unsafe, so there are no useless flushes. | |
3781 | */ | |
3782 | return bdrv_co_flush(bs->file); | |
07f07615 PB |
3783 | } |
3784 | ||
0f15423c AL |
3785 | void bdrv_invalidate_cache(BlockDriverState *bs) |
3786 | { | |
3787 | if (bs->drv && bs->drv->bdrv_invalidate_cache) { | |
3788 | bs->drv->bdrv_invalidate_cache(bs); | |
3789 | } | |
3790 | } | |
3791 | ||
3792 | void bdrv_invalidate_cache_all(void) | |
3793 | { | |
3794 | BlockDriverState *bs; | |
3795 | ||
3796 | QTAILQ_FOREACH(bs, &bdrv_states, list) { | |
3797 | bdrv_invalidate_cache(bs); | |
3798 | } | |
3799 | } | |
3800 | ||
07789269 BC |
3801 | void bdrv_clear_incoming_migration_all(void) |
3802 | { | |
3803 | BlockDriverState *bs; | |
3804 | ||
3805 | QTAILQ_FOREACH(bs, &bdrv_states, list) { | |
3806 | bs->open_flags = bs->open_flags & ~(BDRV_O_INCOMING); | |
3807 | } | |
3808 | } | |
3809 | ||
07f07615 PB |
3810 | int bdrv_flush(BlockDriverState *bs) |
3811 | { | |
3812 | Coroutine *co; | |
3813 | RwCo rwco = { | |
3814 | .bs = bs, | |
3815 | .ret = NOT_DONE, | |
e7a8a783 | 3816 | }; |
e7a8a783 | 3817 | |
07f07615 PB |
3818 | if (qemu_in_coroutine()) { |
3819 | /* Fast-path if already in coroutine context */ | |
3820 | bdrv_flush_co_entry(&rwco); | |
3821 | } else { | |
3822 | co = qemu_coroutine_create(bdrv_flush_co_entry); | |
3823 | qemu_coroutine_enter(co, &rwco); | |
3824 | while (rwco.ret == NOT_DONE) { | |
3825 | qemu_aio_wait(); | |
3826 | } | |
e7a8a783 | 3827 | } |
07f07615 PB |
3828 | |
3829 | return rwco.ret; | |
e7a8a783 KW |
3830 | } |
3831 | ||
4265d620 PB |
3832 | static void coroutine_fn bdrv_discard_co_entry(void *opaque) |
3833 | { | |
3834 | RwCo *rwco = opaque; | |
3835 | ||
3836 | rwco->ret = bdrv_co_discard(rwco->bs, rwco->sector_num, rwco->nb_sectors); | |
3837 | } | |
3838 | ||
3839 | int coroutine_fn bdrv_co_discard(BlockDriverState *bs, int64_t sector_num, | |
3840 | int nb_sectors) | |
3841 | { | |
3842 | if (!bs->drv) { | |
3843 | return -ENOMEDIUM; | |
3844 | } else if (bdrv_check_request(bs, sector_num, nb_sectors)) { | |
3845 | return -EIO; | |
3846 | } else if (bs->read_only) { | |
3847 | return -EROFS; | |
3848 | } else if (bs->drv->bdrv_co_discard) { | |
3849 | return bs->drv->bdrv_co_discard(bs, sector_num, nb_sectors); | |
3850 | } else if (bs->drv->bdrv_aio_discard) { | |
3851 | BlockDriverAIOCB *acb; | |
3852 | CoroutineIOCompletion co = { | |
3853 | .coroutine = qemu_coroutine_self(), | |
3854 | }; | |
3855 | ||
3856 | acb = bs->drv->bdrv_aio_discard(bs, sector_num, nb_sectors, | |
3857 | bdrv_co_io_em_complete, &co); | |
3858 | if (acb == NULL) { | |
3859 | return -EIO; | |
3860 | } else { | |
3861 | qemu_coroutine_yield(); | |
3862 | return co.ret; | |
3863 | } | |
4265d620 PB |
3864 | } else { |
3865 | return 0; | |
3866 | } | |
3867 | } | |
3868 | ||
3869 | int bdrv_discard(BlockDriverState *bs, int64_t sector_num, int nb_sectors) | |
3870 | { | |
3871 | Coroutine *co; | |
3872 | RwCo rwco = { | |
3873 | .bs = bs, | |
3874 | .sector_num = sector_num, | |
3875 | .nb_sectors = nb_sectors, | |
3876 | .ret = NOT_DONE, | |
3877 | }; | |
3878 | ||
3879 | if (qemu_in_coroutine()) { | |
3880 | /* Fast-path if already in coroutine context */ | |
3881 | bdrv_discard_co_entry(&rwco); | |
3882 | } else { | |
3883 | co = qemu_coroutine_create(bdrv_discard_co_entry); | |
3884 | qemu_coroutine_enter(co, &rwco); | |
3885 | while (rwco.ret == NOT_DONE) { | |
3886 | qemu_aio_wait(); | |
3887 | } | |
3888 | } | |
3889 | ||
3890 | return rwco.ret; | |
3891 | } | |
3892 | ||
19cb3738 FB |
3893 | /**************************************************************/ |
3894 | /* removable device support */ | |
3895 | ||
3896 | /** | |
3897 | * Return TRUE if the media is present | |
3898 | */ | |
3899 | int bdrv_is_inserted(BlockDriverState *bs) | |
3900 | { | |
3901 | BlockDriver *drv = bs->drv; | |
a1aff5bf | 3902 | |
19cb3738 FB |
3903 | if (!drv) |
3904 | return 0; | |
3905 | if (!drv->bdrv_is_inserted) | |
a1aff5bf MA |
3906 | return 1; |
3907 | return drv->bdrv_is_inserted(bs); | |
19cb3738 FB |
3908 | } |
3909 | ||
3910 | /** | |
8e49ca46 MA |
3911 | * Return whether the media changed since the last call to this |
3912 | * function, or -ENOTSUP if we don't know. Most drivers don't know. | |
19cb3738 FB |
3913 | */ |
3914 | int bdrv_media_changed(BlockDriverState *bs) | |
3915 | { | |
3916 | BlockDriver *drv = bs->drv; | |
19cb3738 | 3917 | |
8e49ca46 MA |
3918 | if (drv && drv->bdrv_media_changed) { |
3919 | return drv->bdrv_media_changed(bs); | |
3920 | } | |
3921 | return -ENOTSUP; | |
19cb3738 FB |
3922 | } |
3923 | ||
3924 | /** | |
3925 | * If eject_flag is TRUE, eject the media. Otherwise, close the tray | |
3926 | */ | |
f36f3949 | 3927 | void bdrv_eject(BlockDriverState *bs, bool eject_flag) |
19cb3738 FB |
3928 | { |
3929 | BlockDriver *drv = bs->drv; | |
19cb3738 | 3930 | |
822e1cd1 MA |
3931 | if (drv && drv->bdrv_eject) { |
3932 | drv->bdrv_eject(bs, eject_flag); | |
19cb3738 | 3933 | } |
6f382ed2 LC |
3934 | |
3935 | if (bs->device_name[0] != '\0') { | |
3936 | bdrv_emit_qmp_eject_event(bs, eject_flag); | |
3937 | } | |
19cb3738 FB |
3938 | } |
3939 | ||
19cb3738 FB |
3940 | /** |
3941 | * Lock or unlock the media (if it is locked, the user won't be able | |
3942 | * to eject it manually). | |
3943 | */ | |
025e849a | 3944 | void bdrv_lock_medium(BlockDriverState *bs, bool locked) |
19cb3738 FB |
3945 | { |
3946 | BlockDriver *drv = bs->drv; | |
3947 | ||
025e849a | 3948 | trace_bdrv_lock_medium(bs, locked); |
b8c6d095 | 3949 | |
025e849a MA |
3950 | if (drv && drv->bdrv_lock_medium) { |
3951 | drv->bdrv_lock_medium(bs, locked); | |
19cb3738 FB |
3952 | } |
3953 | } | |
985a03b0 TS |
3954 | |
3955 | /* needed for generic scsi interface */ | |
3956 | ||
3957 | int bdrv_ioctl(BlockDriverState *bs, unsigned long int req, void *buf) | |
3958 | { | |
3959 | BlockDriver *drv = bs->drv; | |
3960 | ||
3961 | if (drv && drv->bdrv_ioctl) | |
3962 | return drv->bdrv_ioctl(bs, req, buf); | |
3963 | return -ENOTSUP; | |
3964 | } | |
7d780669 | 3965 | |
221f715d AL |
3966 | BlockDriverAIOCB *bdrv_aio_ioctl(BlockDriverState *bs, |
3967 | unsigned long int req, void *buf, | |
3968 | BlockDriverCompletionFunc *cb, void *opaque) | |
7d780669 | 3969 | { |
221f715d | 3970 | BlockDriver *drv = bs->drv; |
7d780669 | 3971 | |
221f715d AL |
3972 | if (drv && drv->bdrv_aio_ioctl) |
3973 | return drv->bdrv_aio_ioctl(bs, req, buf, cb, opaque); | |
3974 | return NULL; | |
7d780669 | 3975 | } |
e268ca52 | 3976 | |
7b6f9300 MA |
3977 | void bdrv_set_buffer_alignment(BlockDriverState *bs, int align) |
3978 | { | |
3979 | bs->buffer_alignment = align; | |
3980 | } | |
7cd1e32a | 3981 | |
e268ca52 AL |
3982 | void *qemu_blockalign(BlockDriverState *bs, size_t size) |
3983 | { | |
3984 | return qemu_memalign((bs && bs->buffer_alignment) ? bs->buffer_alignment : 512, size); | |
3985 | } | |
7cd1e32a LS |
3986 | |
3987 | void bdrv_set_dirty_tracking(BlockDriverState *bs, int enable) | |
3988 | { | |
3989 | int64_t bitmap_size; | |
a55eb92c | 3990 | |
aaa0eb75 | 3991 | bs->dirty_count = 0; |
a55eb92c | 3992 | if (enable) { |
c6d22830 JK |
3993 | if (!bs->dirty_bitmap) { |
3994 | bitmap_size = (bdrv_getlength(bs) >> BDRV_SECTOR_BITS) + | |
71df14fc PB |
3995 | BDRV_SECTORS_PER_DIRTY_CHUNK * BITS_PER_LONG - 1; |
3996 | bitmap_size /= BDRV_SECTORS_PER_DIRTY_CHUNK * BITS_PER_LONG; | |
a55eb92c | 3997 | |
71df14fc | 3998 | bs->dirty_bitmap = g_new0(unsigned long, bitmap_size); |
a55eb92c | 3999 | } |
7cd1e32a | 4000 | } else { |
c6d22830 | 4001 | if (bs->dirty_bitmap) { |
7267c094 | 4002 | g_free(bs->dirty_bitmap); |
c6d22830 | 4003 | bs->dirty_bitmap = NULL; |
a55eb92c | 4004 | } |
7cd1e32a LS |
4005 | } |
4006 | } | |
4007 | ||
4008 | int bdrv_get_dirty(BlockDriverState *bs, int64_t sector) | |
4009 | { | |
6ea44308 | 4010 | int64_t chunk = sector / (int64_t)BDRV_SECTORS_PER_DIRTY_CHUNK; |
a55eb92c | 4011 | |
c6d22830 JK |
4012 | if (bs->dirty_bitmap && |
4013 | (sector << BDRV_SECTOR_BITS) < bdrv_getlength(bs)) { | |
6d59fec1 MT |
4014 | return !!(bs->dirty_bitmap[chunk / (sizeof(unsigned long) * 8)] & |
4015 | (1UL << (chunk % (sizeof(unsigned long) * 8)))); | |
7cd1e32a LS |
4016 | } else { |
4017 | return 0; | |
4018 | } | |
4019 | } | |
4020 | ||
a55eb92c JK |
4021 | void bdrv_reset_dirty(BlockDriverState *bs, int64_t cur_sector, |
4022 | int nr_sectors) | |
7cd1e32a LS |
4023 | { |
4024 | set_dirty_bitmap(bs, cur_sector, nr_sectors, 0); | |
4025 | } | |
aaa0eb75 LS |
4026 | |
4027 | int64_t bdrv_get_dirty_count(BlockDriverState *bs) | |
4028 | { | |
4029 | return bs->dirty_count; | |
4030 | } | |
f88e1a42 | 4031 | |
db593f25 MT |
4032 | void bdrv_set_in_use(BlockDriverState *bs, int in_use) |
4033 | { | |
4034 | assert(bs->in_use != in_use); | |
4035 | bs->in_use = in_use; | |
4036 | } | |
4037 | ||
4038 | int bdrv_in_use(BlockDriverState *bs) | |
4039 | { | |
4040 | return bs->in_use; | |
4041 | } | |
4042 | ||
28a7282a LC |
4043 | void bdrv_iostatus_enable(BlockDriverState *bs) |
4044 | { | |
d6bf279e | 4045 | bs->iostatus_enabled = true; |
58e21ef5 | 4046 | bs->iostatus = BLOCK_DEVICE_IO_STATUS_OK; |
28a7282a LC |
4047 | } |
4048 | ||
4049 | /* The I/O status is only enabled if the drive explicitly | |
4050 | * enables it _and_ the VM is configured to stop on errors */ | |
4051 | bool bdrv_iostatus_is_enabled(const BlockDriverState *bs) | |
4052 | { | |
d6bf279e | 4053 | return (bs->iostatus_enabled && |
28a7282a LC |
4054 | (bs->on_write_error == BLOCK_ERR_STOP_ENOSPC || |
4055 | bs->on_write_error == BLOCK_ERR_STOP_ANY || | |
4056 | bs->on_read_error == BLOCK_ERR_STOP_ANY)); | |
4057 | } | |
4058 | ||
4059 | void bdrv_iostatus_disable(BlockDriverState *bs) | |
4060 | { | |
d6bf279e | 4061 | bs->iostatus_enabled = false; |
28a7282a LC |
4062 | } |
4063 | ||
4064 | void bdrv_iostatus_reset(BlockDriverState *bs) | |
4065 | { | |
4066 | if (bdrv_iostatus_is_enabled(bs)) { | |
58e21ef5 | 4067 | bs->iostatus = BLOCK_DEVICE_IO_STATUS_OK; |
28a7282a LC |
4068 | } |
4069 | } | |
4070 | ||
4071 | /* XXX: Today this is set by device models because it makes the implementation | |
4072 | quite simple. However, the block layer knows about the error, so it's | |
4073 | possible to implement this without device models being involved */ | |
4074 | void bdrv_iostatus_set_err(BlockDriverState *bs, int error) | |
4075 | { | |
58e21ef5 LC |
4076 | if (bdrv_iostatus_is_enabled(bs) && |
4077 | bs->iostatus == BLOCK_DEVICE_IO_STATUS_OK) { | |
28a7282a | 4078 | assert(error >= 0); |
58e21ef5 LC |
4079 | bs->iostatus = error == ENOSPC ? BLOCK_DEVICE_IO_STATUS_NOSPACE : |
4080 | BLOCK_DEVICE_IO_STATUS_FAILED; | |
28a7282a LC |
4081 | } |
4082 | } | |
4083 | ||
a597e79c CH |
4084 | void |
4085 | bdrv_acct_start(BlockDriverState *bs, BlockAcctCookie *cookie, int64_t bytes, | |
4086 | enum BlockAcctType type) | |
4087 | { | |
4088 | assert(type < BDRV_MAX_IOTYPE); | |
4089 | ||
4090 | cookie->bytes = bytes; | |
c488c7f6 | 4091 | cookie->start_time_ns = get_clock(); |
a597e79c CH |
4092 | cookie->type = type; |
4093 | } | |
4094 | ||
4095 | void | |
4096 | bdrv_acct_done(BlockDriverState *bs, BlockAcctCookie *cookie) | |
4097 | { | |
4098 | assert(cookie->type < BDRV_MAX_IOTYPE); | |
4099 | ||
4100 | bs->nr_bytes[cookie->type] += cookie->bytes; | |
4101 | bs->nr_ops[cookie->type]++; | |
c488c7f6 | 4102 | bs->total_time_ns[cookie->type] += get_clock() - cookie->start_time_ns; |
a597e79c CH |
4103 | } |
4104 | ||
f88e1a42 JS |
4105 | int bdrv_img_create(const char *filename, const char *fmt, |
4106 | const char *base_filename, const char *base_fmt, | |
4107 | char *options, uint64_t img_size, int flags) | |
4108 | { | |
4109 | QEMUOptionParameter *param = NULL, *create_options = NULL; | |
d220894e | 4110 | QEMUOptionParameter *backing_fmt, *backing_file, *size; |
f88e1a42 JS |
4111 | BlockDriverState *bs = NULL; |
4112 | BlockDriver *drv, *proto_drv; | |
96df67d1 | 4113 | BlockDriver *backing_drv = NULL; |
f88e1a42 JS |
4114 | int ret = 0; |
4115 | ||
4116 | /* Find driver and parse its options */ | |
4117 | drv = bdrv_find_format(fmt); | |
4118 | if (!drv) { | |
4119 | error_report("Unknown file format '%s'", fmt); | |
4f70f249 | 4120 | ret = -EINVAL; |
f88e1a42 JS |
4121 | goto out; |
4122 | } | |
4123 | ||
4124 | proto_drv = bdrv_find_protocol(filename); | |
4125 | if (!proto_drv) { | |
4126 | error_report("Unknown protocol '%s'", filename); | |
4f70f249 | 4127 | ret = -EINVAL; |
f88e1a42 JS |
4128 | goto out; |
4129 | } | |
4130 | ||
4131 | create_options = append_option_parameters(create_options, | |
4132 | drv->create_options); | |
4133 | create_options = append_option_parameters(create_options, | |
4134 | proto_drv->create_options); | |
4135 | ||
4136 | /* Create parameter list with default values */ | |
4137 | param = parse_option_parameters("", create_options, param); | |
4138 | ||
4139 | set_option_parameter_int(param, BLOCK_OPT_SIZE, img_size); | |
4140 | ||
4141 | /* Parse -o options */ | |
4142 | if (options) { | |
4143 | param = parse_option_parameters(options, create_options, param); | |
4144 | if (param == NULL) { | |
4145 | error_report("Invalid options for file format '%s'.", fmt); | |
4f70f249 | 4146 | ret = -EINVAL; |
f88e1a42 JS |
4147 | goto out; |
4148 | } | |
4149 | } | |
4150 | ||
4151 | if (base_filename) { | |
4152 | if (set_option_parameter(param, BLOCK_OPT_BACKING_FILE, | |
4153 | base_filename)) { | |
4154 | error_report("Backing file not supported for file format '%s'", | |
4155 | fmt); | |
4f70f249 | 4156 | ret = -EINVAL; |
f88e1a42 JS |
4157 | goto out; |
4158 | } | |
4159 | } | |
4160 | ||
4161 | if (base_fmt) { | |
4162 | if (set_option_parameter(param, BLOCK_OPT_BACKING_FMT, base_fmt)) { | |
4163 | error_report("Backing file format not supported for file " | |
4164 | "format '%s'", fmt); | |
4f70f249 | 4165 | ret = -EINVAL; |
f88e1a42 JS |
4166 | goto out; |
4167 | } | |
4168 | } | |
4169 | ||
792da93a JS |
4170 | backing_file = get_option_parameter(param, BLOCK_OPT_BACKING_FILE); |
4171 | if (backing_file && backing_file->value.s) { | |
4172 | if (!strcmp(filename, backing_file->value.s)) { | |
4173 | error_report("Error: Trying to create an image with the " | |
4174 | "same filename as the backing file"); | |
4f70f249 | 4175 | ret = -EINVAL; |
792da93a JS |
4176 | goto out; |
4177 | } | |
4178 | } | |
4179 | ||
f88e1a42 JS |
4180 | backing_fmt = get_option_parameter(param, BLOCK_OPT_BACKING_FMT); |
4181 | if (backing_fmt && backing_fmt->value.s) { | |
96df67d1 SH |
4182 | backing_drv = bdrv_find_format(backing_fmt->value.s); |
4183 | if (!backing_drv) { | |
f88e1a42 JS |
4184 | error_report("Unknown backing file format '%s'", |
4185 | backing_fmt->value.s); | |
4f70f249 | 4186 | ret = -EINVAL; |
f88e1a42 JS |
4187 | goto out; |
4188 | } | |
4189 | } | |
4190 | ||
4191 | // The size for the image must always be specified, with one exception: | |
4192 | // If we are using a backing file, we can obtain the size from there | |
d220894e KW |
4193 | size = get_option_parameter(param, BLOCK_OPT_SIZE); |
4194 | if (size && size->value.n == -1) { | |
f88e1a42 JS |
4195 | if (backing_file && backing_file->value.s) { |
4196 | uint64_t size; | |
f88e1a42 | 4197 | char buf[32]; |
63090dac PB |
4198 | int back_flags; |
4199 | ||
4200 | /* backing files always opened read-only */ | |
4201 | back_flags = | |
4202 | flags & ~(BDRV_O_RDWR | BDRV_O_SNAPSHOT | BDRV_O_NO_BACKING); | |
f88e1a42 | 4203 | |
f88e1a42 JS |
4204 | bs = bdrv_new(""); |
4205 | ||
63090dac | 4206 | ret = bdrv_open(bs, backing_file->value.s, back_flags, backing_drv); |
f88e1a42 | 4207 | if (ret < 0) { |
96df67d1 | 4208 | error_report("Could not open '%s'", backing_file->value.s); |
f88e1a42 JS |
4209 | goto out; |
4210 | } | |
4211 | bdrv_get_geometry(bs, &size); | |
4212 | size *= 512; | |
4213 | ||
4214 | snprintf(buf, sizeof(buf), "%" PRId64, size); | |
4215 | set_option_parameter(param, BLOCK_OPT_SIZE, buf); | |
4216 | } else { | |
4217 | error_report("Image creation needs a size parameter"); | |
4f70f249 | 4218 | ret = -EINVAL; |
f88e1a42 JS |
4219 | goto out; |
4220 | } | |
4221 | } | |
4222 | ||
4223 | printf("Formatting '%s', fmt=%s ", filename, fmt); | |
4224 | print_option_parameters(param); | |
4225 | puts(""); | |
4226 | ||
4227 | ret = bdrv_create(drv, filename, param); | |
4228 | ||
4229 | if (ret < 0) { | |
4230 | if (ret == -ENOTSUP) { | |
4231 | error_report("Formatting or formatting option not supported for " | |
4232 | "file format '%s'", fmt); | |
4233 | } else if (ret == -EFBIG) { | |
4234 | error_report("The image size is too large for file format '%s'", | |
4235 | fmt); | |
4236 | } else { | |
4237 | error_report("%s: error while creating %s: %s", filename, fmt, | |
4238 | strerror(-ret)); | |
4239 | } | |
4240 | } | |
4241 | ||
4242 | out: | |
4243 | free_option_parameters(create_options); | |
4244 | free_option_parameters(param); | |
4245 | ||
4246 | if (bs) { | |
4247 | bdrv_delete(bs); | |
4248 | } | |
4f70f249 JS |
4249 | |
4250 | return ret; | |
f88e1a42 | 4251 | } |
eeec61f2 SH |
4252 | |
4253 | void *block_job_create(const BlockJobType *job_type, BlockDriverState *bs, | |
c83c66c3 SH |
4254 | int64_t speed, BlockDriverCompletionFunc *cb, |
4255 | void *opaque, Error **errp) | |
eeec61f2 SH |
4256 | { |
4257 | BlockJob *job; | |
4258 | ||
4259 | if (bs->job || bdrv_in_use(bs)) { | |
fd7f8c65 | 4260 | error_set(errp, QERR_DEVICE_IN_USE, bdrv_get_device_name(bs)); |
eeec61f2 SH |
4261 | return NULL; |
4262 | } | |
4263 | bdrv_set_in_use(bs, 1); | |
4264 | ||
4265 | job = g_malloc0(job_type->instance_size); | |
4266 | job->job_type = job_type; | |
4267 | job->bs = bs; | |
4268 | job->cb = cb; | |
4269 | job->opaque = opaque; | |
4513eafe | 4270 | job->busy = true; |
eeec61f2 | 4271 | bs->job = job; |
c83c66c3 SH |
4272 | |
4273 | /* Only set speed when necessary to avoid NotSupported error */ | |
4274 | if (speed != 0) { | |
4275 | Error *local_err = NULL; | |
4276 | ||
4277 | block_job_set_speed(job, speed, &local_err); | |
4278 | if (error_is_set(&local_err)) { | |
4279 | bs->job = NULL; | |
4280 | g_free(job); | |
4281 | bdrv_set_in_use(bs, 0); | |
4282 | error_propagate(errp, local_err); | |
4283 | return NULL; | |
4284 | } | |
4285 | } | |
eeec61f2 SH |
4286 | return job; |
4287 | } | |
4288 | ||
4289 | void block_job_complete(BlockJob *job, int ret) | |
4290 | { | |
4291 | BlockDriverState *bs = job->bs; | |
4292 | ||
4293 | assert(bs->job == job); | |
4294 | job->cb(job->opaque, ret); | |
4295 | bs->job = NULL; | |
4296 | g_free(job); | |
4297 | bdrv_set_in_use(bs, 0); | |
4298 | } | |
4299 | ||
882ec7ce | 4300 | void block_job_set_speed(BlockJob *job, int64_t speed, Error **errp) |
eeec61f2 | 4301 | { |
9e6636c7 | 4302 | Error *local_err = NULL; |
9f25eccc | 4303 | |
eeec61f2 | 4304 | if (!job->job_type->set_speed) { |
9e6636c7 SH |
4305 | error_set(errp, QERR_NOT_SUPPORTED); |
4306 | return; | |
eeec61f2 | 4307 | } |
882ec7ce | 4308 | job->job_type->set_speed(job, speed, &local_err); |
9e6636c7 SH |
4309 | if (error_is_set(&local_err)) { |
4310 | error_propagate(errp, local_err); | |
4311 | return; | |
9f25eccc | 4312 | } |
9e6636c7 | 4313 | |
882ec7ce | 4314 | job->speed = speed; |
eeec61f2 SH |
4315 | } |
4316 | ||
4317 | void block_job_cancel(BlockJob *job) | |
4318 | { | |
4319 | job->cancelled = true; | |
fa4478d5 PB |
4320 | if (job->co && !job->busy) { |
4321 | qemu_coroutine_enter(job->co, NULL); | |
4322 | } | |
eeec61f2 SH |
4323 | } |
4324 | ||
4325 | bool block_job_is_cancelled(BlockJob *job) | |
4326 | { | |
4327 | return job->cancelled; | |
4328 | } | |
3e914655 | 4329 | |
fa4478d5 PB |
4330 | struct BlockCancelData { |
4331 | BlockJob *job; | |
4332 | BlockDriverCompletionFunc *cb; | |
4333 | void *opaque; | |
4334 | bool cancelled; | |
4335 | int ret; | |
4336 | }; | |
4337 | ||
4338 | static void block_job_cancel_cb(void *opaque, int ret) | |
3e914655 | 4339 | { |
fa4478d5 PB |
4340 | struct BlockCancelData *data = opaque; |
4341 | ||
4342 | data->cancelled = block_job_is_cancelled(data->job); | |
4343 | data->ret = ret; | |
4344 | data->cb(data->opaque, ret); | |
4345 | } | |
4346 | ||
4347 | int block_job_cancel_sync(BlockJob *job) | |
4348 | { | |
4349 | struct BlockCancelData data; | |
3e914655 PB |
4350 | BlockDriverState *bs = job->bs; |
4351 | ||
4352 | assert(bs->job == job); | |
fa4478d5 PB |
4353 | |
4354 | /* Set up our own callback to store the result and chain to | |
4355 | * the original callback. | |
4356 | */ | |
4357 | data.job = job; | |
4358 | data.cb = job->cb; | |
4359 | data.opaque = job->opaque; | |
4360 | data.ret = -EINPROGRESS; | |
4361 | job->cb = block_job_cancel_cb; | |
4362 | job->opaque = &data; | |
3e914655 | 4363 | block_job_cancel(job); |
fa4478d5 | 4364 | while (data.ret == -EINPROGRESS) { |
3e914655 PB |
4365 | qemu_aio_wait(); |
4366 | } | |
fa4478d5 | 4367 | return (data.cancelled && data.ret == 0) ? -ECANCELED : data.ret; |
3e914655 | 4368 | } |
4513eafe PB |
4369 | |
4370 | void block_job_sleep_ns(BlockJob *job, QEMUClock *clock, int64_t ns) | |
4371 | { | |
4372 | /* Check cancellation *before* setting busy = false, too! */ | |
4373 | if (!block_job_is_cancelled(job)) { | |
4374 | job->busy = false; | |
4375 | co_sleep_ns(clock, ns); | |
4376 | job->busy = true; | |
4377 | } | |
4378 | } |