]>
Commit | Line | Data |
---|---|---|
fc01f7e7 FB |
1 | /* |
2 | * QEMU System Emulator block driver | |
5fafdf24 | 3 | * |
fc01f7e7 | 4 | * Copyright (c) 2003 Fabrice Bellard |
5fafdf24 | 5 | * |
fc01f7e7 FB |
6 | * Permission is hereby granted, free of charge, to any person obtaining a copy |
7 | * of this software and associated documentation files (the "Software"), to deal | |
8 | * in the Software without restriction, including without limitation the rights | |
9 | * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell | |
10 | * copies of the Software, and to permit persons to whom the Software is | |
11 | * furnished to do so, subject to the following conditions: | |
12 | * | |
13 | * The above copyright notice and this permission notice shall be included in | |
14 | * all copies or substantial portions of the Software. | |
15 | * | |
16 | * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR | |
17 | * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, | |
18 | * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL | |
19 | * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER | |
20 | * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, | |
21 | * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN | |
22 | * THE SOFTWARE. | |
23 | */ | |
3990d09a | 24 | #include "config-host.h" |
faf07963 | 25 | #include "qemu-common.h" |
6d519a5f | 26 | #include "trace.h" |
83c9089e | 27 | #include "monitor/monitor.h" |
737e150e PB |
28 | #include "block/block_int.h" |
29 | #include "block/blockjob.h" | |
1de7afc9 | 30 | #include "qemu/module.h" |
7b1b5d19 | 31 | #include "qapi/qmp/qjson.h" |
9c17d615 | 32 | #include "sysemu/sysemu.h" |
1de7afc9 | 33 | #include "qemu/notify.h" |
737e150e | 34 | #include "block/coroutine.h" |
b2023818 | 35 | #include "qmp-commands.h" |
1de7afc9 | 36 | #include "qemu/timer.h" |
fc01f7e7 | 37 | |
71e72a19 | 38 | #ifdef CONFIG_BSD |
7674e7bf FB |
39 | #include <sys/types.h> |
40 | #include <sys/stat.h> | |
41 | #include <sys/ioctl.h> | |
72cf2d4f | 42 | #include <sys/queue.h> |
c5e97233 | 43 | #ifndef __DragonFly__ |
7674e7bf FB |
44 | #include <sys/disk.h> |
45 | #endif | |
c5e97233 | 46 | #endif |
7674e7bf | 47 | |
49dc768d AL |
48 | #ifdef _WIN32 |
49 | #include <windows.h> | |
50 | #endif | |
51 | ||
1c9805a3 SH |
52 | #define NOT_DONE 0x7fffffff /* used while emulated sync operation in progress */ |
53 | ||
470c0504 SH |
54 | typedef enum { |
55 | BDRV_REQ_COPY_ON_READ = 0x1, | |
f08f2dda | 56 | BDRV_REQ_ZERO_WRITE = 0x2, |
470c0504 SH |
57 | } BdrvRequestFlags; |
58 | ||
7d4b4ba5 | 59 | static void bdrv_dev_change_media_cb(BlockDriverState *bs, bool load); |
f141eafe AL |
60 | static BlockDriverAIOCB *bdrv_aio_readv_em(BlockDriverState *bs, |
61 | int64_t sector_num, QEMUIOVector *qiov, int nb_sectors, | |
c87c0672 | 62 | BlockDriverCompletionFunc *cb, void *opaque); |
f141eafe AL |
63 | static BlockDriverAIOCB *bdrv_aio_writev_em(BlockDriverState *bs, |
64 | int64_t sector_num, QEMUIOVector *qiov, int nb_sectors, | |
ce1a14dc | 65 | BlockDriverCompletionFunc *cb, void *opaque); |
f9f05dc5 KW |
66 | static int coroutine_fn bdrv_co_readv_em(BlockDriverState *bs, |
67 | int64_t sector_num, int nb_sectors, | |
68 | QEMUIOVector *iov); | |
69 | static int coroutine_fn bdrv_co_writev_em(BlockDriverState *bs, | |
70 | int64_t sector_num, int nb_sectors, | |
71 | QEMUIOVector *iov); | |
c5fbe571 | 72 | static int coroutine_fn bdrv_co_do_readv(BlockDriverState *bs, |
470c0504 SH |
73 | int64_t sector_num, int nb_sectors, QEMUIOVector *qiov, |
74 | BdrvRequestFlags flags); | |
1c9805a3 | 75 | static int coroutine_fn bdrv_co_do_writev(BlockDriverState *bs, |
f08f2dda SH |
76 | int64_t sector_num, int nb_sectors, QEMUIOVector *qiov, |
77 | BdrvRequestFlags flags); | |
b2a61371 SH |
78 | static BlockDriverAIOCB *bdrv_co_aio_rw_vector(BlockDriverState *bs, |
79 | int64_t sector_num, | |
80 | QEMUIOVector *qiov, | |
81 | int nb_sectors, | |
82 | BlockDriverCompletionFunc *cb, | |
83 | void *opaque, | |
8c5873d6 | 84 | bool is_write); |
b2a61371 | 85 | static void coroutine_fn bdrv_co_do_rw(void *opaque); |
621f0589 KW |
86 | static int coroutine_fn bdrv_co_do_write_zeroes(BlockDriverState *bs, |
87 | int64_t sector_num, int nb_sectors); | |
ec530c81 | 88 | |
98f90dba ZYW |
89 | static bool bdrv_exceed_bps_limits(BlockDriverState *bs, int nb_sectors, |
90 | bool is_write, double elapsed_time, uint64_t *wait); | |
91 | static bool bdrv_exceed_iops_limits(BlockDriverState *bs, bool is_write, | |
92 | double elapsed_time, uint64_t *wait); | |
93 | static bool bdrv_exceed_io_limits(BlockDriverState *bs, int nb_sectors, | |
94 | bool is_write, int64_t *wait); | |
95 | ||
1b7bdbc1 SH |
96 | static QTAILQ_HEAD(, BlockDriverState) bdrv_states = |
97 | QTAILQ_HEAD_INITIALIZER(bdrv_states); | |
7ee930d0 | 98 | |
8a22f02a SH |
99 | static QLIST_HEAD(, BlockDriver) bdrv_drivers = |
100 | QLIST_HEAD_INITIALIZER(bdrv_drivers); | |
ea2384d3 | 101 | |
f9092b10 MA |
102 | /* The device to use for VM snapshots */ |
103 | static BlockDriverState *bs_snapshots; | |
104 | ||
eb852011 MA |
105 | /* If non-zero, use only whitelisted block drivers */ |
106 | static int use_bdrv_whitelist; | |
107 | ||
9e0b22f4 SH |
108 | #ifdef _WIN32 |
109 | static int is_windows_drive_prefix(const char *filename) | |
110 | { | |
111 | return (((filename[0] >= 'a' && filename[0] <= 'z') || | |
112 | (filename[0] >= 'A' && filename[0] <= 'Z')) && | |
113 | filename[1] == ':'); | |
114 | } | |
115 | ||
116 | int is_windows_drive(const char *filename) | |
117 | { | |
118 | if (is_windows_drive_prefix(filename) && | |
119 | filename[2] == '\0') | |
120 | return 1; | |
121 | if (strstart(filename, "\\\\.\\", NULL) || | |
122 | strstart(filename, "//./", NULL)) | |
123 | return 1; | |
124 | return 0; | |
125 | } | |
126 | #endif | |
127 | ||
0563e191 | 128 | /* throttling disk I/O limits */ |
98f90dba ZYW |
129 | void bdrv_io_limits_disable(BlockDriverState *bs) |
130 | { | |
131 | bs->io_limits_enabled = false; | |
132 | ||
133 | while (qemu_co_queue_next(&bs->throttled_reqs)); | |
134 | ||
135 | if (bs->block_timer) { | |
136 | qemu_del_timer(bs->block_timer); | |
137 | qemu_free_timer(bs->block_timer); | |
138 | bs->block_timer = NULL; | |
139 | } | |
140 | ||
141 | bs->slice_start = 0; | |
142 | bs->slice_end = 0; | |
98f90dba ZYW |
143 | } |
144 | ||
0563e191 ZYW |
145 | static void bdrv_block_timer(void *opaque) |
146 | { | |
147 | BlockDriverState *bs = opaque; | |
148 | ||
149 | qemu_co_queue_next(&bs->throttled_reqs); | |
150 | } | |
151 | ||
152 | void bdrv_io_limits_enable(BlockDriverState *bs) | |
153 | { | |
154 | qemu_co_queue_init(&bs->throttled_reqs); | |
155 | bs->block_timer = qemu_new_timer_ns(vm_clock, bdrv_block_timer, bs); | |
0563e191 ZYW |
156 | bs->io_limits_enabled = true; |
157 | } | |
158 | ||
159 | bool bdrv_io_limits_enabled(BlockDriverState *bs) | |
160 | { | |
161 | BlockIOLimit *io_limits = &bs->io_limits; | |
162 | return io_limits->bps[BLOCK_IO_LIMIT_READ] | |
163 | || io_limits->bps[BLOCK_IO_LIMIT_WRITE] | |
164 | || io_limits->bps[BLOCK_IO_LIMIT_TOTAL] | |
165 | || io_limits->iops[BLOCK_IO_LIMIT_READ] | |
166 | || io_limits->iops[BLOCK_IO_LIMIT_WRITE] | |
167 | || io_limits->iops[BLOCK_IO_LIMIT_TOTAL]; | |
168 | } | |
169 | ||
98f90dba ZYW |
170 | static void bdrv_io_limits_intercept(BlockDriverState *bs, |
171 | bool is_write, int nb_sectors) | |
172 | { | |
173 | int64_t wait_time = -1; | |
174 | ||
175 | if (!qemu_co_queue_empty(&bs->throttled_reqs)) { | |
176 | qemu_co_queue_wait(&bs->throttled_reqs); | |
177 | } | |
178 | ||
179 | /* In fact, we hope to keep each request's timing, in FIFO mode. The next | |
180 | * throttled requests will not be dequeued until the current request is | |
181 | * allowed to be serviced. So if the current request still exceeds the | |
182 | * limits, it will be inserted to the head. All requests followed it will | |
183 | * be still in throttled_reqs queue. | |
184 | */ | |
185 | ||
186 | while (bdrv_exceed_io_limits(bs, nb_sectors, is_write, &wait_time)) { | |
187 | qemu_mod_timer(bs->block_timer, | |
188 | wait_time + qemu_get_clock_ns(vm_clock)); | |
189 | qemu_co_queue_wait_insert_head(&bs->throttled_reqs); | |
190 | } | |
191 | ||
192 | qemu_co_queue_next(&bs->throttled_reqs); | |
193 | } | |
194 | ||
9e0b22f4 SH |
195 | /* check if the path starts with "<protocol>:" */ |
196 | static int path_has_protocol(const char *path) | |
197 | { | |
947995c0 PB |
198 | const char *p; |
199 | ||
9e0b22f4 SH |
200 | #ifdef _WIN32 |
201 | if (is_windows_drive(path) || | |
202 | is_windows_drive_prefix(path)) { | |
203 | return 0; | |
204 | } | |
947995c0 PB |
205 | p = path + strcspn(path, ":/\\"); |
206 | #else | |
207 | p = path + strcspn(path, ":/"); | |
9e0b22f4 SH |
208 | #endif |
209 | ||
947995c0 | 210 | return *p == ':'; |
9e0b22f4 SH |
211 | } |
212 | ||
83f64091 | 213 | int path_is_absolute(const char *path) |
3b0d4f61 | 214 | { |
21664424 FB |
215 | #ifdef _WIN32 |
216 | /* specific case for names like: "\\.\d:" */ | |
f53f4da9 | 217 | if (is_windows_drive(path) || is_windows_drive_prefix(path)) { |
21664424 | 218 | return 1; |
f53f4da9 PB |
219 | } |
220 | return (*path == '/' || *path == '\\'); | |
3b9f94e1 | 221 | #else |
f53f4da9 | 222 | return (*path == '/'); |
3b9f94e1 | 223 | #endif |
3b0d4f61 FB |
224 | } |
225 | ||
83f64091 FB |
226 | /* if filename is absolute, just copy it to dest. Otherwise, build a |
227 | path to it by considering it is relative to base_path. URL are | |
228 | supported. */ | |
229 | void path_combine(char *dest, int dest_size, | |
230 | const char *base_path, | |
231 | const char *filename) | |
3b0d4f61 | 232 | { |
83f64091 FB |
233 | const char *p, *p1; |
234 | int len; | |
235 | ||
236 | if (dest_size <= 0) | |
237 | return; | |
238 | if (path_is_absolute(filename)) { | |
239 | pstrcpy(dest, dest_size, filename); | |
240 | } else { | |
241 | p = strchr(base_path, ':'); | |
242 | if (p) | |
243 | p++; | |
244 | else | |
245 | p = base_path; | |
3b9f94e1 FB |
246 | p1 = strrchr(base_path, '/'); |
247 | #ifdef _WIN32 | |
248 | { | |
249 | const char *p2; | |
250 | p2 = strrchr(base_path, '\\'); | |
251 | if (!p1 || p2 > p1) | |
252 | p1 = p2; | |
253 | } | |
254 | #endif | |
83f64091 FB |
255 | if (p1) |
256 | p1++; | |
257 | else | |
258 | p1 = base_path; | |
259 | if (p1 > p) | |
260 | p = p1; | |
261 | len = p - base_path; | |
262 | if (len > dest_size - 1) | |
263 | len = dest_size - 1; | |
264 | memcpy(dest, base_path, len); | |
265 | dest[len] = '\0'; | |
266 | pstrcat(dest, dest_size, filename); | |
3b0d4f61 | 267 | } |
3b0d4f61 FB |
268 | } |
269 | ||
dc5a1371 PB |
270 | void bdrv_get_full_backing_filename(BlockDriverState *bs, char *dest, size_t sz) |
271 | { | |
272 | if (bs->backing_file[0] == '\0' || path_has_protocol(bs->backing_file)) { | |
273 | pstrcpy(dest, sz, bs->backing_file); | |
274 | } else { | |
275 | path_combine(dest, sz, bs->filename, bs->backing_file); | |
276 | } | |
277 | } | |
278 | ||
5efa9d5a | 279 | void bdrv_register(BlockDriver *bdrv) |
ea2384d3 | 280 | { |
8c5873d6 SH |
281 | /* Block drivers without coroutine functions need emulation */ |
282 | if (!bdrv->bdrv_co_readv) { | |
f9f05dc5 KW |
283 | bdrv->bdrv_co_readv = bdrv_co_readv_em; |
284 | bdrv->bdrv_co_writev = bdrv_co_writev_em; | |
285 | ||
f8c35c1d SH |
286 | /* bdrv_co_readv_em()/brdv_co_writev_em() work in terms of aio, so if |
287 | * the block driver lacks aio we need to emulate that too. | |
288 | */ | |
f9f05dc5 KW |
289 | if (!bdrv->bdrv_aio_readv) { |
290 | /* add AIO emulation layer */ | |
291 | bdrv->bdrv_aio_readv = bdrv_aio_readv_em; | |
292 | bdrv->bdrv_aio_writev = bdrv_aio_writev_em; | |
f9f05dc5 | 293 | } |
83f64091 | 294 | } |
b2e12bc6 | 295 | |
8a22f02a | 296 | QLIST_INSERT_HEAD(&bdrv_drivers, bdrv, list); |
ea2384d3 | 297 | } |
b338082b FB |
298 | |
299 | /* create a new block device (by default it is empty) */ | |
300 | BlockDriverState *bdrv_new(const char *device_name) | |
301 | { | |
1b7bdbc1 | 302 | BlockDriverState *bs; |
b338082b | 303 | |
7267c094 | 304 | bs = g_malloc0(sizeof(BlockDriverState)); |
b338082b | 305 | pstrcpy(bs->device_name, sizeof(bs->device_name), device_name); |
ea2384d3 | 306 | if (device_name[0] != '\0') { |
1b7bdbc1 | 307 | QTAILQ_INSERT_TAIL(&bdrv_states, bs, list); |
ea2384d3 | 308 | } |
28a7282a | 309 | bdrv_iostatus_disable(bs); |
d7d512f6 PB |
310 | notifier_list_init(&bs->close_notifiers); |
311 | ||
b338082b FB |
312 | return bs; |
313 | } | |
314 | ||
d7d512f6 PB |
315 | void bdrv_add_close_notifier(BlockDriverState *bs, Notifier *notify) |
316 | { | |
317 | notifier_list_add(&bs->close_notifiers, notify); | |
318 | } | |
319 | ||
ea2384d3 FB |
320 | BlockDriver *bdrv_find_format(const char *format_name) |
321 | { | |
322 | BlockDriver *drv1; | |
8a22f02a SH |
323 | QLIST_FOREACH(drv1, &bdrv_drivers, list) { |
324 | if (!strcmp(drv1->format_name, format_name)) { | |
ea2384d3 | 325 | return drv1; |
8a22f02a | 326 | } |
ea2384d3 FB |
327 | } |
328 | return NULL; | |
329 | } | |
330 | ||
eb852011 MA |
331 | static int bdrv_is_whitelisted(BlockDriver *drv) |
332 | { | |
333 | static const char *whitelist[] = { | |
334 | CONFIG_BDRV_WHITELIST | |
335 | }; | |
336 | const char **p; | |
337 | ||
338 | if (!whitelist[0]) | |
339 | return 1; /* no whitelist, anything goes */ | |
340 | ||
341 | for (p = whitelist; *p; p++) { | |
342 | if (!strcmp(drv->format_name, *p)) { | |
343 | return 1; | |
344 | } | |
345 | } | |
346 | return 0; | |
347 | } | |
348 | ||
349 | BlockDriver *bdrv_find_whitelisted_format(const char *format_name) | |
350 | { | |
351 | BlockDriver *drv = bdrv_find_format(format_name); | |
352 | return drv && bdrv_is_whitelisted(drv) ? drv : NULL; | |
353 | } | |
354 | ||
5b7e1542 ZYW |
355 | typedef struct CreateCo { |
356 | BlockDriver *drv; | |
357 | char *filename; | |
358 | QEMUOptionParameter *options; | |
359 | int ret; | |
360 | } CreateCo; | |
361 | ||
362 | static void coroutine_fn bdrv_create_co_entry(void *opaque) | |
363 | { | |
364 | CreateCo *cco = opaque; | |
365 | assert(cco->drv); | |
366 | ||
367 | cco->ret = cco->drv->bdrv_create(cco->filename, cco->options); | |
368 | } | |
369 | ||
0e7e1989 KW |
370 | int bdrv_create(BlockDriver *drv, const char* filename, |
371 | QEMUOptionParameter *options) | |
ea2384d3 | 372 | { |
5b7e1542 ZYW |
373 | int ret; |
374 | ||
375 | Coroutine *co; | |
376 | CreateCo cco = { | |
377 | .drv = drv, | |
378 | .filename = g_strdup(filename), | |
379 | .options = options, | |
380 | .ret = NOT_DONE, | |
381 | }; | |
382 | ||
383 | if (!drv->bdrv_create) { | |
80168bff LC |
384 | ret = -ENOTSUP; |
385 | goto out; | |
5b7e1542 ZYW |
386 | } |
387 | ||
388 | if (qemu_in_coroutine()) { | |
389 | /* Fast-path if already in coroutine context */ | |
390 | bdrv_create_co_entry(&cco); | |
391 | } else { | |
392 | co = qemu_coroutine_create(bdrv_create_co_entry); | |
393 | qemu_coroutine_enter(co, &cco); | |
394 | while (cco.ret == NOT_DONE) { | |
395 | qemu_aio_wait(); | |
396 | } | |
397 | } | |
398 | ||
399 | ret = cco.ret; | |
0e7e1989 | 400 | |
80168bff LC |
401 | out: |
402 | g_free(cco.filename); | |
5b7e1542 | 403 | return ret; |
ea2384d3 FB |
404 | } |
405 | ||
84a12e66 CH |
406 | int bdrv_create_file(const char* filename, QEMUOptionParameter *options) |
407 | { | |
408 | BlockDriver *drv; | |
409 | ||
b50cbabc | 410 | drv = bdrv_find_protocol(filename); |
84a12e66 | 411 | if (drv == NULL) { |
16905d71 | 412 | return -ENOENT; |
84a12e66 CH |
413 | } |
414 | ||
415 | return bdrv_create(drv, filename, options); | |
416 | } | |
417 | ||
eba25057 JM |
418 | /* |
419 | * Create a uniquely-named empty temporary file. | |
420 | * Return 0 upon success, otherwise a negative errno value. | |
421 | */ | |
422 | int get_tmp_filename(char *filename, int size) | |
d5249393 | 423 | { |
eba25057 | 424 | #ifdef _WIN32 |
3b9f94e1 | 425 | char temp_dir[MAX_PATH]; |
eba25057 JM |
426 | /* GetTempFileName requires that its output buffer (4th param) |
427 | have length MAX_PATH or greater. */ | |
428 | assert(size >= MAX_PATH); | |
429 | return (GetTempPath(MAX_PATH, temp_dir) | |
430 | && GetTempFileName(temp_dir, "qem", 0, filename) | |
431 | ? 0 : -GetLastError()); | |
d5249393 | 432 | #else |
67b915a5 | 433 | int fd; |
7ccfb2eb | 434 | const char *tmpdir; |
0badc1ee AJ |
435 | tmpdir = getenv("TMPDIR"); |
436 | if (!tmpdir) | |
437 | tmpdir = "/tmp"; | |
eba25057 JM |
438 | if (snprintf(filename, size, "%s/vl.XXXXXX", tmpdir) >= size) { |
439 | return -EOVERFLOW; | |
440 | } | |
ea2384d3 | 441 | fd = mkstemp(filename); |
fe235a06 DH |
442 | if (fd < 0) { |
443 | return -errno; | |
444 | } | |
445 | if (close(fd) != 0) { | |
446 | unlink(filename); | |
eba25057 JM |
447 | return -errno; |
448 | } | |
449 | return 0; | |
d5249393 | 450 | #endif |
eba25057 | 451 | } |
fc01f7e7 | 452 | |
84a12e66 CH |
453 | /* |
454 | * Detect host devices. By convention, /dev/cdrom[N] is always | |
455 | * recognized as a host CDROM. | |
456 | */ | |
457 | static BlockDriver *find_hdev_driver(const char *filename) | |
458 | { | |
459 | int score_max = 0, score; | |
460 | BlockDriver *drv = NULL, *d; | |
461 | ||
462 | QLIST_FOREACH(d, &bdrv_drivers, list) { | |
463 | if (d->bdrv_probe_device) { | |
464 | score = d->bdrv_probe_device(filename); | |
465 | if (score > score_max) { | |
466 | score_max = score; | |
467 | drv = d; | |
468 | } | |
469 | } | |
470 | } | |
471 | ||
472 | return drv; | |
473 | } | |
474 | ||
b50cbabc | 475 | BlockDriver *bdrv_find_protocol(const char *filename) |
83f64091 FB |
476 | { |
477 | BlockDriver *drv1; | |
478 | char protocol[128]; | |
1cec71e3 | 479 | int len; |
83f64091 | 480 | const char *p; |
19cb3738 | 481 | |
66f82cee KW |
482 | /* TODO Drivers without bdrv_file_open must be specified explicitly */ |
483 | ||
39508e7a CH |
484 | /* |
485 | * XXX(hch): we really should not let host device detection | |
486 | * override an explicit protocol specification, but moving this | |
487 | * later breaks access to device names with colons in them. | |
488 | * Thanks to the brain-dead persistent naming schemes on udev- | |
489 | * based Linux systems those actually are quite common. | |
490 | */ | |
491 | drv1 = find_hdev_driver(filename); | |
492 | if (drv1) { | |
493 | return drv1; | |
494 | } | |
495 | ||
9e0b22f4 | 496 | if (!path_has_protocol(filename)) { |
39508e7a | 497 | return bdrv_find_format("file"); |
84a12e66 | 498 | } |
9e0b22f4 SH |
499 | p = strchr(filename, ':'); |
500 | assert(p != NULL); | |
1cec71e3 AL |
501 | len = p - filename; |
502 | if (len > sizeof(protocol) - 1) | |
503 | len = sizeof(protocol) - 1; | |
504 | memcpy(protocol, filename, len); | |
505 | protocol[len] = '\0'; | |
8a22f02a | 506 | QLIST_FOREACH(drv1, &bdrv_drivers, list) { |
5fafdf24 | 507 | if (drv1->protocol_name && |
8a22f02a | 508 | !strcmp(drv1->protocol_name, protocol)) { |
83f64091 | 509 | return drv1; |
8a22f02a | 510 | } |
83f64091 FB |
511 | } |
512 | return NULL; | |
513 | } | |
514 | ||
f500a6d3 KW |
515 | static int find_image_format(BlockDriverState *bs, const char *filename, |
516 | BlockDriver **pdrv) | |
f3a5d3f8 | 517 | { |
f500a6d3 | 518 | int score, score_max; |
f3a5d3f8 CH |
519 | BlockDriver *drv1, *drv; |
520 | uint8_t buf[2048]; | |
f500a6d3 | 521 | int ret = 0; |
f8ea0b00 | 522 | |
08a00559 | 523 | /* Return the raw BlockDriver * to scsi-generic devices or empty drives */ |
8e895599 | 524 | if (bs->sg || !bdrv_is_inserted(bs) || bdrv_getlength(bs) == 0) { |
c98ac35d SW |
525 | drv = bdrv_find_format("raw"); |
526 | if (!drv) { | |
527 | ret = -ENOENT; | |
528 | } | |
529 | *pdrv = drv; | |
530 | return ret; | |
1a396859 | 531 | } |
f8ea0b00 | 532 | |
83f64091 | 533 | ret = bdrv_pread(bs, 0, buf, sizeof(buf)); |
83f64091 | 534 | if (ret < 0) { |
c98ac35d SW |
535 | *pdrv = NULL; |
536 | return ret; | |
83f64091 FB |
537 | } |
538 | ||
ea2384d3 | 539 | score_max = 0; |
84a12e66 | 540 | drv = NULL; |
8a22f02a | 541 | QLIST_FOREACH(drv1, &bdrv_drivers, list) { |
83f64091 FB |
542 | if (drv1->bdrv_probe) { |
543 | score = drv1->bdrv_probe(buf, ret, filename); | |
544 | if (score > score_max) { | |
545 | score_max = score; | |
546 | drv = drv1; | |
547 | } | |
0849bf08 | 548 | } |
fc01f7e7 | 549 | } |
c98ac35d SW |
550 | if (!drv) { |
551 | ret = -ENOENT; | |
552 | } | |
553 | *pdrv = drv; | |
554 | return ret; | |
ea2384d3 FB |
555 | } |
556 | ||
51762288 SH |
557 | /** |
558 | * Set the current 'total_sectors' value | |
559 | */ | |
560 | static int refresh_total_sectors(BlockDriverState *bs, int64_t hint) | |
561 | { | |
562 | BlockDriver *drv = bs->drv; | |
563 | ||
396759ad NB |
564 | /* Do not attempt drv->bdrv_getlength() on scsi-generic devices */ |
565 | if (bs->sg) | |
566 | return 0; | |
567 | ||
51762288 SH |
568 | /* query actual device if possible, otherwise just trust the hint */ |
569 | if (drv->bdrv_getlength) { | |
570 | int64_t length = drv->bdrv_getlength(bs); | |
571 | if (length < 0) { | |
572 | return length; | |
573 | } | |
574 | hint = length >> BDRV_SECTOR_BITS; | |
575 | } | |
576 | ||
577 | bs->total_sectors = hint; | |
578 | return 0; | |
579 | } | |
580 | ||
9e8f1835 PB |
581 | /** |
582 | * Set open flags for a given discard mode | |
583 | * | |
584 | * Return 0 on success, -1 if the discard mode was invalid. | |
585 | */ | |
586 | int bdrv_parse_discard_flags(const char *mode, int *flags) | |
587 | { | |
588 | *flags &= ~BDRV_O_UNMAP; | |
589 | ||
590 | if (!strcmp(mode, "off") || !strcmp(mode, "ignore")) { | |
591 | /* do nothing */ | |
592 | } else if (!strcmp(mode, "on") || !strcmp(mode, "unmap")) { | |
593 | *flags |= BDRV_O_UNMAP; | |
594 | } else { | |
595 | return -1; | |
596 | } | |
597 | ||
598 | return 0; | |
599 | } | |
600 | ||
c3993cdc SH |
601 | /** |
602 | * Set open flags for a given cache mode | |
603 | * | |
604 | * Return 0 on success, -1 if the cache mode was invalid. | |
605 | */ | |
606 | int bdrv_parse_cache_flags(const char *mode, int *flags) | |
607 | { | |
608 | *flags &= ~BDRV_O_CACHE_MASK; | |
609 | ||
610 | if (!strcmp(mode, "off") || !strcmp(mode, "none")) { | |
611 | *flags |= BDRV_O_NOCACHE | BDRV_O_CACHE_WB; | |
92196b2f SH |
612 | } else if (!strcmp(mode, "directsync")) { |
613 | *flags |= BDRV_O_NOCACHE; | |
c3993cdc SH |
614 | } else if (!strcmp(mode, "writeback")) { |
615 | *flags |= BDRV_O_CACHE_WB; | |
616 | } else if (!strcmp(mode, "unsafe")) { | |
617 | *flags |= BDRV_O_CACHE_WB; | |
618 | *flags |= BDRV_O_NO_FLUSH; | |
619 | } else if (!strcmp(mode, "writethrough")) { | |
620 | /* this is the default */ | |
621 | } else { | |
622 | return -1; | |
623 | } | |
624 | ||
625 | return 0; | |
626 | } | |
627 | ||
53fec9d3 SH |
628 | /** |
629 | * The copy-on-read flag is actually a reference count so multiple users may | |
630 | * use the feature without worrying about clobbering its previous state. | |
631 | * Copy-on-read stays enabled until all users have called to disable it. | |
632 | */ | |
633 | void bdrv_enable_copy_on_read(BlockDriverState *bs) | |
634 | { | |
635 | bs->copy_on_read++; | |
636 | } | |
637 | ||
638 | void bdrv_disable_copy_on_read(BlockDriverState *bs) | |
639 | { | |
640 | assert(bs->copy_on_read > 0); | |
641 | bs->copy_on_read--; | |
642 | } | |
643 | ||
7b272452 KW |
644 | static int bdrv_open_flags(BlockDriverState *bs, int flags) |
645 | { | |
646 | int open_flags = flags | BDRV_O_CACHE_WB; | |
647 | ||
648 | /* | |
649 | * Clear flags that are internal to the block layer before opening the | |
650 | * image. | |
651 | */ | |
652 | open_flags &= ~(BDRV_O_SNAPSHOT | BDRV_O_NO_BACKING); | |
653 | ||
654 | /* | |
655 | * Snapshots should be writable. | |
656 | */ | |
657 | if (bs->is_temporary) { | |
658 | open_flags |= BDRV_O_RDWR; | |
659 | } | |
660 | ||
661 | return open_flags; | |
662 | } | |
663 | ||
57915332 KW |
664 | /* |
665 | * Common part for opening disk images and files | |
b6ad491a KW |
666 | * |
667 | * Removes all processed options from *options. | |
57915332 | 668 | */ |
f500a6d3 | 669 | static int bdrv_open_common(BlockDriverState *bs, BlockDriverState *file, |
b6ad491a | 670 | const char *filename, QDict *options, |
57915332 KW |
671 | int flags, BlockDriver *drv) |
672 | { | |
673 | int ret, open_flags; | |
674 | ||
675 | assert(drv != NULL); | |
6405875c | 676 | assert(bs->file == NULL); |
707ff828 | 677 | assert(options != NULL && bs->options != options); |
57915332 | 678 | |
28dcee10 SH |
679 | trace_bdrv_open_common(bs, filename, flags, drv->format_name); |
680 | ||
5d186eb0 KW |
681 | if (use_bdrv_whitelist && !bdrv_is_whitelisted(drv)) { |
682 | return -ENOTSUP; | |
683 | } | |
684 | ||
685 | /* bdrv_open() with directly using a protocol as drv. This layer is already | |
686 | * opened, so assign it to bs (while file becomes a closed BlockDriverState) | |
687 | * and return immediately. */ | |
688 | if (file != NULL && drv->bdrv_file_open) { | |
689 | bdrv_swap(file, bs); | |
690 | return 0; | |
691 | } | |
692 | ||
57915332 | 693 | bs->open_flags = flags; |
57915332 KW |
694 | bs->buffer_alignment = 512; |
695 | ||
53fec9d3 SH |
696 | assert(bs->copy_on_read == 0); /* bdrv_new() and bdrv_close() make it so */ |
697 | if ((flags & BDRV_O_RDWR) && (flags & BDRV_O_COPY_ON_READ)) { | |
698 | bdrv_enable_copy_on_read(bs); | |
699 | } | |
700 | ||
c2ad1b0c KW |
701 | if (filename != NULL) { |
702 | pstrcpy(bs->filename, sizeof(bs->filename), filename); | |
703 | } else { | |
704 | bs->filename[0] = '\0'; | |
705 | } | |
57915332 | 706 | |
57915332 | 707 | bs->drv = drv; |
7267c094 | 708 | bs->opaque = g_malloc0(drv->instance_size); |
57915332 | 709 | |
03f541bd | 710 | bs->enable_write_cache = !!(flags & BDRV_O_CACHE_WB); |
7b272452 | 711 | open_flags = bdrv_open_flags(bs, flags); |
57915332 | 712 | |
be028adc | 713 | bs->read_only = !(open_flags & BDRV_O_RDWR); |
e7c63796 | 714 | |
66f82cee KW |
715 | /* Open the image, either directly or using a protocol */ |
716 | if (drv->bdrv_file_open) { | |
5d186eb0 KW |
717 | assert(file == NULL); |
718 | assert(drv->bdrv_parse_filename || filename != NULL); | |
719 | ret = drv->bdrv_file_open(bs, filename, options, open_flags); | |
f500a6d3 | 720 | } else { |
2af5ef70 KW |
721 | if (file == NULL) { |
722 | qerror_report(ERROR_CLASS_GENERIC_ERROR, "Can't use '%s' as a " | |
723 | "block driver for the protocol level", | |
724 | drv->format_name); | |
725 | ret = -EINVAL; | |
726 | goto free_and_fail; | |
727 | } | |
f500a6d3 KW |
728 | assert(file != NULL); |
729 | bs->file = file; | |
b6ad491a | 730 | ret = drv->bdrv_open(bs, options, open_flags); |
66f82cee KW |
731 | } |
732 | ||
57915332 KW |
733 | if (ret < 0) { |
734 | goto free_and_fail; | |
735 | } | |
736 | ||
51762288 SH |
737 | ret = refresh_total_sectors(bs, bs->total_sectors); |
738 | if (ret < 0) { | |
739 | goto free_and_fail; | |
57915332 | 740 | } |
51762288 | 741 | |
57915332 KW |
742 | #ifndef _WIN32 |
743 | if (bs->is_temporary) { | |
c2ad1b0c | 744 | assert(filename != NULL); |
57915332 KW |
745 | unlink(filename); |
746 | } | |
747 | #endif | |
748 | return 0; | |
749 | ||
750 | free_and_fail: | |
f500a6d3 | 751 | bs->file = NULL; |
7267c094 | 752 | g_free(bs->opaque); |
57915332 KW |
753 | bs->opaque = NULL; |
754 | bs->drv = NULL; | |
755 | return ret; | |
756 | } | |
757 | ||
b6ce07aa KW |
758 | /* |
759 | * Opens a file using a protocol (file, host_device, nbd, ...) | |
787e4a85 KW |
760 | * |
761 | * options is a QDict of options to pass to the block drivers, or NULL for an | |
762 | * empty set of options. The reference to the QDict belongs to the block layer | |
763 | * after the call (even on failure), so if the caller intends to reuse the | |
764 | * dictionary, it needs to use QINCREF() before calling bdrv_file_open. | |
b6ce07aa | 765 | */ |
787e4a85 KW |
766 | int bdrv_file_open(BlockDriverState **pbs, const char *filename, |
767 | QDict *options, int flags) | |
ea2384d3 | 768 | { |
83f64091 | 769 | BlockDriverState *bs; |
6db95603 | 770 | BlockDriver *drv; |
c2ad1b0c | 771 | const char *drvname; |
83f64091 FB |
772 | int ret; |
773 | ||
707ff828 KW |
774 | /* NULL means an empty set of options */ |
775 | if (options == NULL) { | |
776 | options = qdict_new(); | |
777 | } | |
778 | ||
83f64091 | 779 | bs = bdrv_new(""); |
707ff828 KW |
780 | bs->options = options; |
781 | options = qdict_clone_shallow(options); | |
782 | ||
c2ad1b0c KW |
783 | /* Find the right block driver */ |
784 | drvname = qdict_get_try_str(options, "driver"); | |
785 | if (drvname) { | |
786 | drv = bdrv_find_whitelisted_format(drvname); | |
787 | qdict_del(options, "driver"); | |
788 | } else if (filename) { | |
789 | drv = bdrv_find_protocol(filename); | |
790 | } else { | |
791 | qerror_report(ERROR_CLASS_GENERIC_ERROR, | |
792 | "Must specify either driver or file"); | |
793 | drv = NULL; | |
794 | } | |
795 | ||
796 | if (!drv) { | |
797 | ret = -ENOENT; | |
798 | goto fail; | |
799 | } | |
800 | ||
801 | /* Parse the filename and open it */ | |
802 | if (drv->bdrv_parse_filename && filename) { | |
6963a30d KW |
803 | Error *local_err = NULL; |
804 | drv->bdrv_parse_filename(filename, options, &local_err); | |
805 | if (error_is_set(&local_err)) { | |
806 | qerror_report_err(local_err); | |
807 | error_free(local_err); | |
808 | ret = -EINVAL; | |
809 | goto fail; | |
810 | } | |
c2ad1b0c KW |
811 | } else if (!drv->bdrv_parse_filename && !filename) { |
812 | qerror_report(ERROR_CLASS_GENERIC_ERROR, | |
813 | "The '%s' block driver requires a file name", | |
814 | drv->format_name); | |
815 | ret = -EINVAL; | |
816 | goto fail; | |
6963a30d KW |
817 | } |
818 | ||
707ff828 | 819 | ret = bdrv_open_common(bs, NULL, filename, options, flags, drv); |
83f64091 | 820 | if (ret < 0) { |
707ff828 KW |
821 | goto fail; |
822 | } | |
823 | ||
824 | /* Check if any unknown options were used */ | |
825 | if (qdict_size(options) != 0) { | |
826 | const QDictEntry *entry = qdict_first(options); | |
827 | qerror_report(ERROR_CLASS_GENERIC_ERROR, "Block protocol '%s' doesn't " | |
828 | "support the option '%s'", | |
829 | drv->format_name, entry->key); | |
830 | ret = -EINVAL; | |
831 | goto fail; | |
3b0d4f61 | 832 | } |
707ff828 KW |
833 | QDECREF(options); |
834 | ||
71d0770c | 835 | bs->growable = 1; |
83f64091 FB |
836 | *pbs = bs; |
837 | return 0; | |
707ff828 KW |
838 | |
839 | fail: | |
840 | QDECREF(options); | |
841 | if (!bs->drv) { | |
842 | QDECREF(bs->options); | |
843 | } | |
844 | bdrv_delete(bs); | |
845 | return ret; | |
83f64091 FB |
846 | } |
847 | ||
31ca6d07 KW |
848 | /* |
849 | * Opens the backing file for a BlockDriverState if not yet open | |
850 | * | |
851 | * options is a QDict of options to pass to the block drivers, or NULL for an | |
852 | * empty set of options. The reference to the QDict is transferred to this | |
853 | * function (even on failure), so if the caller intends to reuse the dictionary, | |
854 | * it needs to use QINCREF() before calling bdrv_file_open. | |
855 | */ | |
856 | int bdrv_open_backing_file(BlockDriverState *bs, QDict *options) | |
9156df12 PB |
857 | { |
858 | char backing_filename[PATH_MAX]; | |
859 | int back_flags, ret; | |
860 | BlockDriver *back_drv = NULL; | |
861 | ||
862 | if (bs->backing_hd != NULL) { | |
31ca6d07 | 863 | QDECREF(options); |
9156df12 PB |
864 | return 0; |
865 | } | |
866 | ||
31ca6d07 KW |
867 | /* NULL means an empty set of options */ |
868 | if (options == NULL) { | |
869 | options = qdict_new(); | |
870 | } | |
871 | ||
9156df12 | 872 | bs->open_flags &= ~BDRV_O_NO_BACKING; |
31ca6d07 KW |
873 | if (bs->backing_file[0] == '\0' && qdict_size(options) == 0) { |
874 | QDECREF(options); | |
9156df12 PB |
875 | return 0; |
876 | } | |
877 | ||
878 | bs->backing_hd = bdrv_new(""); | |
879 | bdrv_get_full_backing_filename(bs, backing_filename, | |
880 | sizeof(backing_filename)); | |
881 | ||
882 | if (bs->backing_format[0] != '\0') { | |
883 | back_drv = bdrv_find_format(bs->backing_format); | |
884 | } | |
885 | ||
886 | /* backing files always opened read-only */ | |
887 | back_flags = bs->open_flags & ~(BDRV_O_RDWR | BDRV_O_SNAPSHOT); | |
888 | ||
31ca6d07 KW |
889 | ret = bdrv_open(bs->backing_hd, |
890 | *backing_filename ? backing_filename : NULL, options, | |
de9c0cec | 891 | back_flags, back_drv); |
9156df12 PB |
892 | if (ret < 0) { |
893 | bdrv_delete(bs->backing_hd); | |
894 | bs->backing_hd = NULL; | |
895 | bs->open_flags |= BDRV_O_NO_BACKING; | |
896 | return ret; | |
897 | } | |
898 | return 0; | |
899 | } | |
900 | ||
707ff828 KW |
901 | static void extract_subqdict(QDict *src, QDict **dst, const char *start) |
902 | { | |
903 | const QDictEntry *entry, *next; | |
904 | const char *p; | |
905 | ||
906 | *dst = qdict_new(); | |
907 | entry = qdict_first(src); | |
908 | ||
909 | while (entry != NULL) { | |
910 | next = qdict_next(src, entry); | |
911 | if (strstart(entry->key, start, &p)) { | |
912 | qobject_incref(entry->value); | |
913 | qdict_put_obj(*dst, p, entry->value); | |
914 | qdict_del(src, entry->key); | |
915 | } | |
916 | entry = next; | |
917 | } | |
918 | } | |
919 | ||
b6ce07aa KW |
920 | /* |
921 | * Opens a disk image (raw, qcow2, vmdk, ...) | |
de9c0cec KW |
922 | * |
923 | * options is a QDict of options to pass to the block drivers, or NULL for an | |
924 | * empty set of options. The reference to the QDict belongs to the block layer | |
925 | * after the call (even on failure), so if the caller intends to reuse the | |
926 | * dictionary, it needs to use QINCREF() before calling bdrv_open. | |
b6ce07aa | 927 | */ |
de9c0cec KW |
928 | int bdrv_open(BlockDriverState *bs, const char *filename, QDict *options, |
929 | int flags, BlockDriver *drv) | |
ea2384d3 | 930 | { |
b6ce07aa | 931 | int ret; |
89c9bc3d SW |
932 | /* TODO: extra byte is a hack to ensure MAX_PATH space on Windows. */ |
933 | char tmp_filename[PATH_MAX + 1]; | |
f500a6d3 | 934 | BlockDriverState *file = NULL; |
707ff828 | 935 | QDict *file_options = NULL; |
712e7874 | 936 | |
de9c0cec KW |
937 | /* NULL means an empty set of options */ |
938 | if (options == NULL) { | |
939 | options = qdict_new(); | |
940 | } | |
941 | ||
942 | bs->options = options; | |
b6ad491a | 943 | options = qdict_clone_shallow(options); |
de9c0cec KW |
944 | |
945 | /* For snapshot=on, create a temporary qcow2 overlay */ | |
83f64091 | 946 | if (flags & BDRV_O_SNAPSHOT) { |
ea2384d3 FB |
947 | BlockDriverState *bs1; |
948 | int64_t total_size; | |
91a073a9 | 949 | BlockDriver *bdrv_qcow2; |
08b392e1 | 950 | QEMUOptionParameter *create_options; |
b6ce07aa | 951 | char backing_filename[PATH_MAX]; |
3b46e624 | 952 | |
c2ad1b0c KW |
953 | if (qdict_size(options) != 0) { |
954 | error_report("Can't use snapshot=on with driver-specific options"); | |
955 | ret = -EINVAL; | |
956 | goto fail; | |
957 | } | |
958 | assert(filename != NULL); | |
959 | ||
ea2384d3 FB |
960 | /* if snapshot, we create a temporary backing file and open it |
961 | instead of opening 'filename' directly */ | |
33e3963e | 962 | |
ea2384d3 FB |
963 | /* if there is a backing file, use it */ |
964 | bs1 = bdrv_new(""); | |
de9c0cec | 965 | ret = bdrv_open(bs1, filename, NULL, 0, drv); |
51d7c00c | 966 | if (ret < 0) { |
ea2384d3 | 967 | bdrv_delete(bs1); |
de9c0cec | 968 | goto fail; |
ea2384d3 | 969 | } |
3e82990b | 970 | total_size = bdrv_getlength(bs1) & BDRV_SECTOR_MASK; |
7c96d46e | 971 | |
ea2384d3 | 972 | bdrv_delete(bs1); |
3b46e624 | 973 | |
eba25057 JM |
974 | ret = get_tmp_filename(tmp_filename, sizeof(tmp_filename)); |
975 | if (ret < 0) { | |
de9c0cec | 976 | goto fail; |
eba25057 | 977 | } |
7c96d46e AL |
978 | |
979 | /* Real path is meaningless for protocols */ | |
4d70655b | 980 | if (path_has_protocol(filename)) { |
7c96d46e AL |
981 | snprintf(backing_filename, sizeof(backing_filename), |
982 | "%s", filename); | |
de9c0cec KW |
983 | } else if (!realpath(filename, backing_filename)) { |
984 | ret = -errno; | |
985 | goto fail; | |
986 | } | |
7c96d46e | 987 | |
91a073a9 | 988 | bdrv_qcow2 = bdrv_find_format("qcow2"); |
08b392e1 KW |
989 | create_options = parse_option_parameters("", bdrv_qcow2->create_options, |
990 | NULL); | |
91a073a9 | 991 | |
08b392e1 KW |
992 | set_option_parameter_int(create_options, BLOCK_OPT_SIZE, total_size); |
993 | set_option_parameter(create_options, BLOCK_OPT_BACKING_FILE, | |
994 | backing_filename); | |
91a073a9 | 995 | if (drv) { |
08b392e1 | 996 | set_option_parameter(create_options, BLOCK_OPT_BACKING_FMT, |
91a073a9 KW |
997 | drv->format_name); |
998 | } | |
999 | ||
08b392e1 KW |
1000 | ret = bdrv_create(bdrv_qcow2, tmp_filename, create_options); |
1001 | free_option_parameters(create_options); | |
51d7c00c | 1002 | if (ret < 0) { |
de9c0cec | 1003 | goto fail; |
ea2384d3 | 1004 | } |
91a073a9 | 1005 | |
ea2384d3 | 1006 | filename = tmp_filename; |
91a073a9 | 1007 | drv = bdrv_qcow2; |
ea2384d3 FB |
1008 | bs->is_temporary = 1; |
1009 | } | |
712e7874 | 1010 | |
f500a6d3 KW |
1011 | /* Open image file without format layer */ |
1012 | if (flags & BDRV_O_RDWR) { | |
1013 | flags |= BDRV_O_ALLOW_RDWR; | |
1014 | } | |
1015 | ||
707ff828 KW |
1016 | extract_subqdict(options, &file_options, "file."); |
1017 | ||
1018 | ret = bdrv_file_open(&file, filename, file_options, | |
1019 | bdrv_open_flags(bs, flags)); | |
f500a6d3 | 1020 | if (ret < 0) { |
de9c0cec | 1021 | goto fail; |
f500a6d3 KW |
1022 | } |
1023 | ||
b6ce07aa | 1024 | /* Find the right image format driver */ |
6db95603 | 1025 | if (!drv) { |
f500a6d3 | 1026 | ret = find_image_format(file, filename, &drv); |
51d7c00c | 1027 | } |
6987307c | 1028 | |
51d7c00c | 1029 | if (!drv) { |
51d7c00c | 1030 | goto unlink_and_fail; |
ea2384d3 | 1031 | } |
b6ce07aa KW |
1032 | |
1033 | /* Open the image */ | |
b6ad491a | 1034 | ret = bdrv_open_common(bs, file, filename, options, flags, drv); |
b6ce07aa | 1035 | if (ret < 0) { |
6987307c CH |
1036 | goto unlink_and_fail; |
1037 | } | |
1038 | ||
f500a6d3 KW |
1039 | if (bs->file != file) { |
1040 | bdrv_delete(file); | |
1041 | file = NULL; | |
1042 | } | |
1043 | ||
b6ce07aa | 1044 | /* If there is a backing file, use it */ |
9156df12 | 1045 | if ((flags & BDRV_O_NO_BACKING) == 0) { |
31ca6d07 KW |
1046 | QDict *backing_options; |
1047 | ||
1048 | extract_subqdict(options, &backing_options, "backing."); | |
1049 | ret = bdrv_open_backing_file(bs, backing_options); | |
b6ce07aa | 1050 | if (ret < 0) { |
b6ad491a | 1051 | goto close_and_fail; |
b6ce07aa | 1052 | } |
b6ce07aa KW |
1053 | } |
1054 | ||
b6ad491a KW |
1055 | /* Check if any unknown options were used */ |
1056 | if (qdict_size(options) != 0) { | |
1057 | const QDictEntry *entry = qdict_first(options); | |
1058 | qerror_report(ERROR_CLASS_GENERIC_ERROR, "Block format '%s' used by " | |
1059 | "device '%s' doesn't support the option '%s'", | |
1060 | drv->format_name, bs->device_name, entry->key); | |
1061 | ||
1062 | ret = -EINVAL; | |
1063 | goto close_and_fail; | |
1064 | } | |
1065 | QDECREF(options); | |
1066 | ||
b6ce07aa | 1067 | if (!bdrv_key_required(bs)) { |
7d4b4ba5 | 1068 | bdrv_dev_change_media_cb(bs, true); |
b6ce07aa KW |
1069 | } |
1070 | ||
98f90dba ZYW |
1071 | /* throttling disk I/O limits */ |
1072 | if (bs->io_limits_enabled) { | |
1073 | bdrv_io_limits_enable(bs); | |
1074 | } | |
1075 | ||
b6ce07aa KW |
1076 | return 0; |
1077 | ||
1078 | unlink_and_fail: | |
f500a6d3 KW |
1079 | if (file != NULL) { |
1080 | bdrv_delete(file); | |
1081 | } | |
b6ce07aa KW |
1082 | if (bs->is_temporary) { |
1083 | unlink(filename); | |
1084 | } | |
de9c0cec KW |
1085 | fail: |
1086 | QDECREF(bs->options); | |
b6ad491a | 1087 | QDECREF(options); |
de9c0cec | 1088 | bs->options = NULL; |
b6ad491a | 1089 | return ret; |
de9c0cec | 1090 | |
b6ad491a KW |
1091 | close_and_fail: |
1092 | bdrv_close(bs); | |
1093 | QDECREF(options); | |
b6ce07aa KW |
1094 | return ret; |
1095 | } | |
1096 | ||
e971aa12 JC |
1097 | typedef struct BlockReopenQueueEntry { |
1098 | bool prepared; | |
1099 | BDRVReopenState state; | |
1100 | QSIMPLEQ_ENTRY(BlockReopenQueueEntry) entry; | |
1101 | } BlockReopenQueueEntry; | |
1102 | ||
1103 | /* | |
1104 | * Adds a BlockDriverState to a simple queue for an atomic, transactional | |
1105 | * reopen of multiple devices. | |
1106 | * | |
1107 | * bs_queue can either be an existing BlockReopenQueue that has had QSIMPLE_INIT | |
1108 | * already performed, or alternatively may be NULL a new BlockReopenQueue will | |
1109 | * be created and initialized. This newly created BlockReopenQueue should be | |
1110 | * passed back in for subsequent calls that are intended to be of the same | |
1111 | * atomic 'set'. | |
1112 | * | |
1113 | * bs is the BlockDriverState to add to the reopen queue. | |
1114 | * | |
1115 | * flags contains the open flags for the associated bs | |
1116 | * | |
1117 | * returns a pointer to bs_queue, which is either the newly allocated | |
1118 | * bs_queue, or the existing bs_queue being used. | |
1119 | * | |
1120 | */ | |
1121 | BlockReopenQueue *bdrv_reopen_queue(BlockReopenQueue *bs_queue, | |
1122 | BlockDriverState *bs, int flags) | |
1123 | { | |
1124 | assert(bs != NULL); | |
1125 | ||
1126 | BlockReopenQueueEntry *bs_entry; | |
1127 | if (bs_queue == NULL) { | |
1128 | bs_queue = g_new0(BlockReopenQueue, 1); | |
1129 | QSIMPLEQ_INIT(bs_queue); | |
1130 | } | |
1131 | ||
1132 | if (bs->file) { | |
1133 | bdrv_reopen_queue(bs_queue, bs->file, flags); | |
1134 | } | |
1135 | ||
1136 | bs_entry = g_new0(BlockReopenQueueEntry, 1); | |
1137 | QSIMPLEQ_INSERT_TAIL(bs_queue, bs_entry, entry); | |
1138 | ||
1139 | bs_entry->state.bs = bs; | |
1140 | bs_entry->state.flags = flags; | |
1141 | ||
1142 | return bs_queue; | |
1143 | } | |
1144 | ||
1145 | /* | |
1146 | * Reopen multiple BlockDriverStates atomically & transactionally. | |
1147 | * | |
1148 | * The queue passed in (bs_queue) must have been built up previous | |
1149 | * via bdrv_reopen_queue(). | |
1150 | * | |
1151 | * Reopens all BDS specified in the queue, with the appropriate | |
1152 | * flags. All devices are prepared for reopen, and failure of any | |
1153 | * device will cause all device changes to be abandonded, and intermediate | |
1154 | * data cleaned up. | |
1155 | * | |
1156 | * If all devices prepare successfully, then the changes are committed | |
1157 | * to all devices. | |
1158 | * | |
1159 | */ | |
1160 | int bdrv_reopen_multiple(BlockReopenQueue *bs_queue, Error **errp) | |
1161 | { | |
1162 | int ret = -1; | |
1163 | BlockReopenQueueEntry *bs_entry, *next; | |
1164 | Error *local_err = NULL; | |
1165 | ||
1166 | assert(bs_queue != NULL); | |
1167 | ||
1168 | bdrv_drain_all(); | |
1169 | ||
1170 | QSIMPLEQ_FOREACH(bs_entry, bs_queue, entry) { | |
1171 | if (bdrv_reopen_prepare(&bs_entry->state, bs_queue, &local_err)) { | |
1172 | error_propagate(errp, local_err); | |
1173 | goto cleanup; | |
1174 | } | |
1175 | bs_entry->prepared = true; | |
1176 | } | |
1177 | ||
1178 | /* If we reach this point, we have success and just need to apply the | |
1179 | * changes | |
1180 | */ | |
1181 | QSIMPLEQ_FOREACH(bs_entry, bs_queue, entry) { | |
1182 | bdrv_reopen_commit(&bs_entry->state); | |
1183 | } | |
1184 | ||
1185 | ret = 0; | |
1186 | ||
1187 | cleanup: | |
1188 | QSIMPLEQ_FOREACH_SAFE(bs_entry, bs_queue, entry, next) { | |
1189 | if (ret && bs_entry->prepared) { | |
1190 | bdrv_reopen_abort(&bs_entry->state); | |
1191 | } | |
1192 | g_free(bs_entry); | |
1193 | } | |
1194 | g_free(bs_queue); | |
1195 | return ret; | |
1196 | } | |
1197 | ||
1198 | ||
1199 | /* Reopen a single BlockDriverState with the specified flags. */ | |
1200 | int bdrv_reopen(BlockDriverState *bs, int bdrv_flags, Error **errp) | |
1201 | { | |
1202 | int ret = -1; | |
1203 | Error *local_err = NULL; | |
1204 | BlockReopenQueue *queue = bdrv_reopen_queue(NULL, bs, bdrv_flags); | |
1205 | ||
1206 | ret = bdrv_reopen_multiple(queue, &local_err); | |
1207 | if (local_err != NULL) { | |
1208 | error_propagate(errp, local_err); | |
1209 | } | |
1210 | return ret; | |
1211 | } | |
1212 | ||
1213 | ||
1214 | /* | |
1215 | * Prepares a BlockDriverState for reopen. All changes are staged in the | |
1216 | * 'opaque' field of the BDRVReopenState, which is used and allocated by | |
1217 | * the block driver layer .bdrv_reopen_prepare() | |
1218 | * | |
1219 | * bs is the BlockDriverState to reopen | |
1220 | * flags are the new open flags | |
1221 | * queue is the reopen queue | |
1222 | * | |
1223 | * Returns 0 on success, non-zero on error. On error errp will be set | |
1224 | * as well. | |
1225 | * | |
1226 | * On failure, bdrv_reopen_abort() will be called to clean up any data. | |
1227 | * It is the responsibility of the caller to then call the abort() or | |
1228 | * commit() for any other BDS that have been left in a prepare() state | |
1229 | * | |
1230 | */ | |
1231 | int bdrv_reopen_prepare(BDRVReopenState *reopen_state, BlockReopenQueue *queue, | |
1232 | Error **errp) | |
1233 | { | |
1234 | int ret = -1; | |
1235 | Error *local_err = NULL; | |
1236 | BlockDriver *drv; | |
1237 | ||
1238 | assert(reopen_state != NULL); | |
1239 | assert(reopen_state->bs->drv != NULL); | |
1240 | drv = reopen_state->bs->drv; | |
1241 | ||
1242 | /* if we are to stay read-only, do not allow permission change | |
1243 | * to r/w */ | |
1244 | if (!(reopen_state->bs->open_flags & BDRV_O_ALLOW_RDWR) && | |
1245 | reopen_state->flags & BDRV_O_RDWR) { | |
1246 | error_set(errp, QERR_DEVICE_IS_READ_ONLY, | |
1247 | reopen_state->bs->device_name); | |
1248 | goto error; | |
1249 | } | |
1250 | ||
1251 | ||
1252 | ret = bdrv_flush(reopen_state->bs); | |
1253 | if (ret) { | |
1254 | error_set(errp, ERROR_CLASS_GENERIC_ERROR, "Error (%s) flushing drive", | |
1255 | strerror(-ret)); | |
1256 | goto error; | |
1257 | } | |
1258 | ||
1259 | if (drv->bdrv_reopen_prepare) { | |
1260 | ret = drv->bdrv_reopen_prepare(reopen_state, queue, &local_err); | |
1261 | if (ret) { | |
1262 | if (local_err != NULL) { | |
1263 | error_propagate(errp, local_err); | |
1264 | } else { | |
1265 | error_set(errp, QERR_OPEN_FILE_FAILED, | |
1266 | reopen_state->bs->filename); | |
1267 | } | |
1268 | goto error; | |
1269 | } | |
1270 | } else { | |
1271 | /* It is currently mandatory to have a bdrv_reopen_prepare() | |
1272 | * handler for each supported drv. */ | |
1273 | error_set(errp, QERR_BLOCK_FORMAT_FEATURE_NOT_SUPPORTED, | |
1274 | drv->format_name, reopen_state->bs->device_name, | |
1275 | "reopening of file"); | |
1276 | ret = -1; | |
1277 | goto error; | |
1278 | } | |
1279 | ||
1280 | ret = 0; | |
1281 | ||
1282 | error: | |
1283 | return ret; | |
1284 | } | |
1285 | ||
1286 | /* | |
1287 | * Takes the staged changes for the reopen from bdrv_reopen_prepare(), and | |
1288 | * makes them final by swapping the staging BlockDriverState contents into | |
1289 | * the active BlockDriverState contents. | |
1290 | */ | |
1291 | void bdrv_reopen_commit(BDRVReopenState *reopen_state) | |
1292 | { | |
1293 | BlockDriver *drv; | |
1294 | ||
1295 | assert(reopen_state != NULL); | |
1296 | drv = reopen_state->bs->drv; | |
1297 | assert(drv != NULL); | |
1298 | ||
1299 | /* If there are any driver level actions to take */ | |
1300 | if (drv->bdrv_reopen_commit) { | |
1301 | drv->bdrv_reopen_commit(reopen_state); | |
1302 | } | |
1303 | ||
1304 | /* set BDS specific flags now */ | |
1305 | reopen_state->bs->open_flags = reopen_state->flags; | |
1306 | reopen_state->bs->enable_write_cache = !!(reopen_state->flags & | |
1307 | BDRV_O_CACHE_WB); | |
1308 | reopen_state->bs->read_only = !(reopen_state->flags & BDRV_O_RDWR); | |
1309 | } | |
1310 | ||
1311 | /* | |
1312 | * Abort the reopen, and delete and free the staged changes in | |
1313 | * reopen_state | |
1314 | */ | |
1315 | void bdrv_reopen_abort(BDRVReopenState *reopen_state) | |
1316 | { | |
1317 | BlockDriver *drv; | |
1318 | ||
1319 | assert(reopen_state != NULL); | |
1320 | drv = reopen_state->bs->drv; | |
1321 | assert(drv != NULL); | |
1322 | ||
1323 | if (drv->bdrv_reopen_abort) { | |
1324 | drv->bdrv_reopen_abort(reopen_state); | |
1325 | } | |
1326 | } | |
1327 | ||
1328 | ||
fc01f7e7 FB |
1329 | void bdrv_close(BlockDriverState *bs) |
1330 | { | |
80ccf93b | 1331 | bdrv_flush(bs); |
3cbc002c PB |
1332 | if (bs->job) { |
1333 | block_job_cancel_sync(bs->job); | |
1334 | } | |
1335 | bdrv_drain_all(); | |
d7d512f6 | 1336 | notifier_list_notify(&bs->close_notifiers, bs); |
7094f12f | 1337 | |
3cbc002c | 1338 | if (bs->drv) { |
f9092b10 MA |
1339 | if (bs == bs_snapshots) { |
1340 | bs_snapshots = NULL; | |
1341 | } | |
557df6ac | 1342 | if (bs->backing_hd) { |
ea2384d3 | 1343 | bdrv_delete(bs->backing_hd); |
557df6ac SH |
1344 | bs->backing_hd = NULL; |
1345 | } | |
ea2384d3 | 1346 | bs->drv->bdrv_close(bs); |
7267c094 | 1347 | g_free(bs->opaque); |
ea2384d3 FB |
1348 | #ifdef _WIN32 |
1349 | if (bs->is_temporary) { | |
1350 | unlink(bs->filename); | |
1351 | } | |
67b915a5 | 1352 | #endif |
ea2384d3 FB |
1353 | bs->opaque = NULL; |
1354 | bs->drv = NULL; | |
53fec9d3 | 1355 | bs->copy_on_read = 0; |
a275fa42 PB |
1356 | bs->backing_file[0] = '\0'; |
1357 | bs->backing_format[0] = '\0'; | |
6405875c PB |
1358 | bs->total_sectors = 0; |
1359 | bs->encrypted = 0; | |
1360 | bs->valid_key = 0; | |
1361 | bs->sg = 0; | |
1362 | bs->growable = 0; | |
de9c0cec KW |
1363 | QDECREF(bs->options); |
1364 | bs->options = NULL; | |
b338082b | 1365 | |
66f82cee | 1366 | if (bs->file != NULL) { |
0ac9377d PB |
1367 | bdrv_delete(bs->file); |
1368 | bs->file = NULL; | |
66f82cee | 1369 | } |
b338082b | 1370 | } |
98f90dba | 1371 | |
9ca11154 PH |
1372 | bdrv_dev_change_media_cb(bs, false); |
1373 | ||
98f90dba ZYW |
1374 | /*throttling disk I/O limits*/ |
1375 | if (bs->io_limits_enabled) { | |
1376 | bdrv_io_limits_disable(bs); | |
1377 | } | |
b338082b FB |
1378 | } |
1379 | ||
2bc93fed MK |
1380 | void bdrv_close_all(void) |
1381 | { | |
1382 | BlockDriverState *bs; | |
1383 | ||
1384 | QTAILQ_FOREACH(bs, &bdrv_states, list) { | |
1385 | bdrv_close(bs); | |
1386 | } | |
1387 | } | |
1388 | ||
922453bc SH |
1389 | /* |
1390 | * Wait for pending requests to complete across all BlockDriverStates | |
1391 | * | |
1392 | * This function does not flush data to disk, use bdrv_flush_all() for that | |
1393 | * after calling this function. | |
4c355d53 ZYW |
1394 | * |
1395 | * Note that completion of an asynchronous I/O operation can trigger any | |
1396 | * number of other I/O operations on other devices---for example a coroutine | |
1397 | * can be arbitrarily complex and a constant flow of I/O can come until the | |
1398 | * coroutine is complete. Because of this, it is not possible to have a | |
1399 | * function to drain a single device's I/O queue. | |
922453bc SH |
1400 | */ |
1401 | void bdrv_drain_all(void) | |
1402 | { | |
1403 | BlockDriverState *bs; | |
4c355d53 ZYW |
1404 | bool busy; |
1405 | ||
1406 | do { | |
1407 | busy = qemu_aio_wait(); | |
922453bc | 1408 | |
4c355d53 ZYW |
1409 | /* FIXME: We do not have timer support here, so this is effectively |
1410 | * a busy wait. | |
1411 | */ | |
1412 | QTAILQ_FOREACH(bs, &bdrv_states, list) { | |
1413 | if (!qemu_co_queue_empty(&bs->throttled_reqs)) { | |
1414 | qemu_co_queue_restart_all(&bs->throttled_reqs); | |
1415 | busy = true; | |
1416 | } | |
1417 | } | |
1418 | } while (busy); | |
922453bc SH |
1419 | |
1420 | /* If requests are still pending there is a bug somewhere */ | |
1421 | QTAILQ_FOREACH(bs, &bdrv_states, list) { | |
1422 | assert(QLIST_EMPTY(&bs->tracked_requests)); | |
1423 | assert(qemu_co_queue_empty(&bs->throttled_reqs)); | |
1424 | } | |
1425 | } | |
1426 | ||
d22b2f41 RH |
1427 | /* make a BlockDriverState anonymous by removing from bdrv_state list. |
1428 | Also, NULL terminate the device_name to prevent double remove */ | |
1429 | void bdrv_make_anon(BlockDriverState *bs) | |
1430 | { | |
1431 | if (bs->device_name[0] != '\0') { | |
1432 | QTAILQ_REMOVE(&bdrv_states, bs, list); | |
1433 | } | |
1434 | bs->device_name[0] = '\0'; | |
1435 | } | |
1436 | ||
e023b2e2 PB |
1437 | static void bdrv_rebind(BlockDriverState *bs) |
1438 | { | |
1439 | if (bs->drv && bs->drv->bdrv_rebind) { | |
1440 | bs->drv->bdrv_rebind(bs); | |
1441 | } | |
1442 | } | |
1443 | ||
4ddc07ca PB |
1444 | static void bdrv_move_feature_fields(BlockDriverState *bs_dest, |
1445 | BlockDriverState *bs_src) | |
8802d1fd | 1446 | { |
4ddc07ca PB |
1447 | /* move some fields that need to stay attached to the device */ |
1448 | bs_dest->open_flags = bs_src->open_flags; | |
8802d1fd JC |
1449 | |
1450 | /* dev info */ | |
4ddc07ca PB |
1451 | bs_dest->dev_ops = bs_src->dev_ops; |
1452 | bs_dest->dev_opaque = bs_src->dev_opaque; | |
1453 | bs_dest->dev = bs_src->dev; | |
1454 | bs_dest->buffer_alignment = bs_src->buffer_alignment; | |
1455 | bs_dest->copy_on_read = bs_src->copy_on_read; | |
8802d1fd | 1456 | |
4ddc07ca | 1457 | bs_dest->enable_write_cache = bs_src->enable_write_cache; |
c4a248a1 | 1458 | |
8802d1fd | 1459 | /* i/o timing parameters */ |
4ddc07ca PB |
1460 | bs_dest->slice_start = bs_src->slice_start; |
1461 | bs_dest->slice_end = bs_src->slice_end; | |
5905fbc9 | 1462 | bs_dest->slice_submitted = bs_src->slice_submitted; |
4ddc07ca | 1463 | bs_dest->io_limits = bs_src->io_limits; |
4ddc07ca PB |
1464 | bs_dest->throttled_reqs = bs_src->throttled_reqs; |
1465 | bs_dest->block_timer = bs_src->block_timer; | |
1466 | bs_dest->io_limits_enabled = bs_src->io_limits_enabled; | |
8802d1fd | 1467 | |
8802d1fd | 1468 | /* r/w error */ |
4ddc07ca PB |
1469 | bs_dest->on_read_error = bs_src->on_read_error; |
1470 | bs_dest->on_write_error = bs_src->on_write_error; | |
8802d1fd JC |
1471 | |
1472 | /* i/o status */ | |
4ddc07ca PB |
1473 | bs_dest->iostatus_enabled = bs_src->iostatus_enabled; |
1474 | bs_dest->iostatus = bs_src->iostatus; | |
8802d1fd | 1475 | |
a9fc4408 | 1476 | /* dirty bitmap */ |
4ddc07ca | 1477 | bs_dest->dirty_bitmap = bs_src->dirty_bitmap; |
a9fc4408 PB |
1478 | |
1479 | /* job */ | |
4ddc07ca PB |
1480 | bs_dest->in_use = bs_src->in_use; |
1481 | bs_dest->job = bs_src->job; | |
a9fc4408 | 1482 | |
8802d1fd | 1483 | /* keep the same entry in bdrv_states */ |
4ddc07ca PB |
1484 | pstrcpy(bs_dest->device_name, sizeof(bs_dest->device_name), |
1485 | bs_src->device_name); | |
1486 | bs_dest->list = bs_src->list; | |
1487 | } | |
8802d1fd | 1488 | |
4ddc07ca PB |
1489 | /* |
1490 | * Swap bs contents for two image chains while they are live, | |
1491 | * while keeping required fields on the BlockDriverState that is | |
1492 | * actually attached to a device. | |
1493 | * | |
1494 | * This will modify the BlockDriverState fields, and swap contents | |
1495 | * between bs_new and bs_old. Both bs_new and bs_old are modified. | |
1496 | * | |
1497 | * bs_new is required to be anonymous. | |
1498 | * | |
1499 | * This function does not create any image files. | |
1500 | */ | |
1501 | void bdrv_swap(BlockDriverState *bs_new, BlockDriverState *bs_old) | |
1502 | { | |
1503 | BlockDriverState tmp; | |
f6801b83 | 1504 | |
4ddc07ca PB |
1505 | /* bs_new must be anonymous and shouldn't have anything fancy enabled */ |
1506 | assert(bs_new->device_name[0] == '\0'); | |
1507 | assert(bs_new->dirty_bitmap == NULL); | |
1508 | assert(bs_new->job == NULL); | |
1509 | assert(bs_new->dev == NULL); | |
1510 | assert(bs_new->in_use == 0); | |
1511 | assert(bs_new->io_limits_enabled == false); | |
1512 | assert(bs_new->block_timer == NULL); | |
8802d1fd | 1513 | |
4ddc07ca PB |
1514 | tmp = *bs_new; |
1515 | *bs_new = *bs_old; | |
1516 | *bs_old = tmp; | |
a9fc4408 | 1517 | |
4ddc07ca PB |
1518 | /* there are some fields that should not be swapped, move them back */ |
1519 | bdrv_move_feature_fields(&tmp, bs_old); | |
1520 | bdrv_move_feature_fields(bs_old, bs_new); | |
1521 | bdrv_move_feature_fields(bs_new, &tmp); | |
8802d1fd | 1522 | |
4ddc07ca PB |
1523 | /* bs_new shouldn't be in bdrv_states even after the swap! */ |
1524 | assert(bs_new->device_name[0] == '\0'); | |
1525 | ||
1526 | /* Check a few fields that should remain attached to the device */ | |
1527 | assert(bs_new->dev == NULL); | |
1528 | assert(bs_new->job == NULL); | |
1529 | assert(bs_new->in_use == 0); | |
1530 | assert(bs_new->io_limits_enabled == false); | |
1531 | assert(bs_new->block_timer == NULL); | |
e023b2e2 PB |
1532 | |
1533 | bdrv_rebind(bs_new); | |
4ddc07ca PB |
1534 | bdrv_rebind(bs_old); |
1535 | } | |
1536 | ||
1537 | /* | |
1538 | * Add new bs contents at the top of an image chain while the chain is | |
1539 | * live, while keeping required fields on the top layer. | |
1540 | * | |
1541 | * This will modify the BlockDriverState fields, and swap contents | |
1542 | * between bs_new and bs_top. Both bs_new and bs_top are modified. | |
1543 | * | |
1544 | * bs_new is required to be anonymous. | |
1545 | * | |
1546 | * This function does not create any image files. | |
1547 | */ | |
1548 | void bdrv_append(BlockDriverState *bs_new, BlockDriverState *bs_top) | |
1549 | { | |
1550 | bdrv_swap(bs_new, bs_top); | |
1551 | ||
1552 | /* The contents of 'tmp' will become bs_top, as we are | |
1553 | * swapping bs_new and bs_top contents. */ | |
1554 | bs_top->backing_hd = bs_new; | |
1555 | bs_top->open_flags &= ~BDRV_O_NO_BACKING; | |
1556 | pstrcpy(bs_top->backing_file, sizeof(bs_top->backing_file), | |
1557 | bs_new->filename); | |
1558 | pstrcpy(bs_top->backing_format, sizeof(bs_top->backing_format), | |
1559 | bs_new->drv ? bs_new->drv->format_name : ""); | |
8802d1fd JC |
1560 | } |
1561 | ||
b338082b FB |
1562 | void bdrv_delete(BlockDriverState *bs) |
1563 | { | |
fa879d62 | 1564 | assert(!bs->dev); |
3e914655 PB |
1565 | assert(!bs->job); |
1566 | assert(!bs->in_use); | |
18846dee | 1567 | |
1b7bdbc1 | 1568 | /* remove from list, if necessary */ |
d22b2f41 | 1569 | bdrv_make_anon(bs); |
34c6f050 | 1570 | |
b338082b | 1571 | bdrv_close(bs); |
66f82cee | 1572 | |
f9092b10 | 1573 | assert(bs != bs_snapshots); |
7267c094 | 1574 | g_free(bs); |
fc01f7e7 FB |
1575 | } |
1576 | ||
fa879d62 MA |
1577 | int bdrv_attach_dev(BlockDriverState *bs, void *dev) |
1578 | /* TODO change to DeviceState *dev when all users are qdevified */ | |
18846dee | 1579 | { |
fa879d62 | 1580 | if (bs->dev) { |
18846dee MA |
1581 | return -EBUSY; |
1582 | } | |
fa879d62 | 1583 | bs->dev = dev; |
28a7282a | 1584 | bdrv_iostatus_reset(bs); |
18846dee MA |
1585 | return 0; |
1586 | } | |
1587 | ||
fa879d62 MA |
1588 | /* TODO qdevified devices don't use this, remove when devices are qdevified */ |
1589 | void bdrv_attach_dev_nofail(BlockDriverState *bs, void *dev) | |
18846dee | 1590 | { |
fa879d62 MA |
1591 | if (bdrv_attach_dev(bs, dev) < 0) { |
1592 | abort(); | |
1593 | } | |
1594 | } | |
1595 | ||
1596 | void bdrv_detach_dev(BlockDriverState *bs, void *dev) | |
1597 | /* TODO change to DeviceState *dev when all users are qdevified */ | |
1598 | { | |
1599 | assert(bs->dev == dev); | |
1600 | bs->dev = NULL; | |
0e49de52 MA |
1601 | bs->dev_ops = NULL; |
1602 | bs->dev_opaque = NULL; | |
29e05f20 | 1603 | bs->buffer_alignment = 512; |
18846dee MA |
1604 | } |
1605 | ||
fa879d62 MA |
1606 | /* TODO change to return DeviceState * when all users are qdevified */ |
1607 | void *bdrv_get_attached_dev(BlockDriverState *bs) | |
18846dee | 1608 | { |
fa879d62 | 1609 | return bs->dev; |
18846dee MA |
1610 | } |
1611 | ||
0e49de52 MA |
1612 | void bdrv_set_dev_ops(BlockDriverState *bs, const BlockDevOps *ops, |
1613 | void *opaque) | |
1614 | { | |
1615 | bs->dev_ops = ops; | |
1616 | bs->dev_opaque = opaque; | |
2c6942fa MA |
1617 | if (bdrv_dev_has_removable_media(bs) && bs == bs_snapshots) { |
1618 | bs_snapshots = NULL; | |
1619 | } | |
0e49de52 MA |
1620 | } |
1621 | ||
32c81a4a PB |
1622 | void bdrv_emit_qmp_error_event(const BlockDriverState *bdrv, |
1623 | enum MonitorEvent ev, | |
1624 | BlockErrorAction action, bool is_read) | |
329c0a48 LC |
1625 | { |
1626 | QObject *data; | |
1627 | const char *action_str; | |
1628 | ||
1629 | switch (action) { | |
1630 | case BDRV_ACTION_REPORT: | |
1631 | action_str = "report"; | |
1632 | break; | |
1633 | case BDRV_ACTION_IGNORE: | |
1634 | action_str = "ignore"; | |
1635 | break; | |
1636 | case BDRV_ACTION_STOP: | |
1637 | action_str = "stop"; | |
1638 | break; | |
1639 | default: | |
1640 | abort(); | |
1641 | } | |
1642 | ||
1643 | data = qobject_from_jsonf("{ 'device': %s, 'action': %s, 'operation': %s }", | |
1644 | bdrv->device_name, | |
1645 | action_str, | |
1646 | is_read ? "read" : "write"); | |
32c81a4a | 1647 | monitor_protocol_event(ev, data); |
329c0a48 LC |
1648 | |
1649 | qobject_decref(data); | |
1650 | } | |
1651 | ||
6f382ed2 LC |
1652 | static void bdrv_emit_qmp_eject_event(BlockDriverState *bs, bool ejected) |
1653 | { | |
1654 | QObject *data; | |
1655 | ||
1656 | data = qobject_from_jsonf("{ 'device': %s, 'tray-open': %i }", | |
1657 | bdrv_get_device_name(bs), ejected); | |
1658 | monitor_protocol_event(QEVENT_DEVICE_TRAY_MOVED, data); | |
1659 | ||
1660 | qobject_decref(data); | |
1661 | } | |
1662 | ||
7d4b4ba5 | 1663 | static void bdrv_dev_change_media_cb(BlockDriverState *bs, bool load) |
0e49de52 | 1664 | { |
145feb17 | 1665 | if (bs->dev_ops && bs->dev_ops->change_media_cb) { |
6f382ed2 | 1666 | bool tray_was_closed = !bdrv_dev_is_tray_open(bs); |
7d4b4ba5 | 1667 | bs->dev_ops->change_media_cb(bs->dev_opaque, load); |
6f382ed2 LC |
1668 | if (tray_was_closed) { |
1669 | /* tray open */ | |
1670 | bdrv_emit_qmp_eject_event(bs, true); | |
1671 | } | |
1672 | if (load) { | |
1673 | /* tray close */ | |
1674 | bdrv_emit_qmp_eject_event(bs, false); | |
1675 | } | |
145feb17 MA |
1676 | } |
1677 | } | |
1678 | ||
2c6942fa MA |
1679 | bool bdrv_dev_has_removable_media(BlockDriverState *bs) |
1680 | { | |
1681 | return !bs->dev || (bs->dev_ops && bs->dev_ops->change_media_cb); | |
1682 | } | |
1683 | ||
025ccaa7 PB |
1684 | void bdrv_dev_eject_request(BlockDriverState *bs, bool force) |
1685 | { | |
1686 | if (bs->dev_ops && bs->dev_ops->eject_request_cb) { | |
1687 | bs->dev_ops->eject_request_cb(bs->dev_opaque, force); | |
1688 | } | |
1689 | } | |
1690 | ||
e4def80b MA |
1691 | bool bdrv_dev_is_tray_open(BlockDriverState *bs) |
1692 | { | |
1693 | if (bs->dev_ops && bs->dev_ops->is_tray_open) { | |
1694 | return bs->dev_ops->is_tray_open(bs->dev_opaque); | |
1695 | } | |
1696 | return false; | |
1697 | } | |
1698 | ||
145feb17 MA |
1699 | static void bdrv_dev_resize_cb(BlockDriverState *bs) |
1700 | { | |
1701 | if (bs->dev_ops && bs->dev_ops->resize_cb) { | |
1702 | bs->dev_ops->resize_cb(bs->dev_opaque); | |
0e49de52 MA |
1703 | } |
1704 | } | |
1705 | ||
f107639a MA |
1706 | bool bdrv_dev_is_medium_locked(BlockDriverState *bs) |
1707 | { | |
1708 | if (bs->dev_ops && bs->dev_ops->is_medium_locked) { | |
1709 | return bs->dev_ops->is_medium_locked(bs->dev_opaque); | |
1710 | } | |
1711 | return false; | |
1712 | } | |
1713 | ||
e97fc193 AL |
1714 | /* |
1715 | * Run consistency checks on an image | |
1716 | * | |
e076f338 | 1717 | * Returns 0 if the check could be completed (it doesn't mean that the image is |
a1c7273b | 1718 | * free of errors) or -errno when an internal error occurred. The results of the |
e076f338 | 1719 | * check are stored in res. |
e97fc193 | 1720 | */ |
4534ff54 | 1721 | int bdrv_check(BlockDriverState *bs, BdrvCheckResult *res, BdrvCheckMode fix) |
e97fc193 AL |
1722 | { |
1723 | if (bs->drv->bdrv_check == NULL) { | |
1724 | return -ENOTSUP; | |
1725 | } | |
1726 | ||
e076f338 | 1727 | memset(res, 0, sizeof(*res)); |
4534ff54 | 1728 | return bs->drv->bdrv_check(bs, res, fix); |
e97fc193 AL |
1729 | } |
1730 | ||
8a426614 KW |
1731 | #define COMMIT_BUF_SECTORS 2048 |
1732 | ||
33e3963e FB |
1733 | /* commit COW file into the raw image */ |
1734 | int bdrv_commit(BlockDriverState *bs) | |
1735 | { | |
19cb3738 | 1736 | BlockDriver *drv = bs->drv; |
8a426614 KW |
1737 | int64_t sector, total_sectors; |
1738 | int n, ro, open_flags; | |
0bce597d | 1739 | int ret = 0; |
8a426614 | 1740 | uint8_t *buf; |
c2cba3d9 | 1741 | char filename[PATH_MAX]; |
33e3963e | 1742 | |
19cb3738 FB |
1743 | if (!drv) |
1744 | return -ENOMEDIUM; | |
4dca4b63 NS |
1745 | |
1746 | if (!bs->backing_hd) { | |
1747 | return -ENOTSUP; | |
33e3963e FB |
1748 | } |
1749 | ||
2d3735d3 SH |
1750 | if (bdrv_in_use(bs) || bdrv_in_use(bs->backing_hd)) { |
1751 | return -EBUSY; | |
1752 | } | |
1753 | ||
4dca4b63 | 1754 | ro = bs->backing_hd->read_only; |
c2cba3d9 JM |
1755 | /* Use pstrcpy (not strncpy): filename must be NUL-terminated. */ |
1756 | pstrcpy(filename, sizeof(filename), bs->backing_hd->filename); | |
4dca4b63 NS |
1757 | open_flags = bs->backing_hd->open_flags; |
1758 | ||
1759 | if (ro) { | |
0bce597d JC |
1760 | if (bdrv_reopen(bs->backing_hd, open_flags | BDRV_O_RDWR, NULL)) { |
1761 | return -EACCES; | |
4dca4b63 | 1762 | } |
ea2384d3 | 1763 | } |
33e3963e | 1764 | |
6ea44308 | 1765 | total_sectors = bdrv_getlength(bs) >> BDRV_SECTOR_BITS; |
7267c094 | 1766 | buf = g_malloc(COMMIT_BUF_SECTORS * BDRV_SECTOR_SIZE); |
8a426614 KW |
1767 | |
1768 | for (sector = 0; sector < total_sectors; sector += n) { | |
05c4af54 | 1769 | if (bdrv_is_allocated(bs, sector, COMMIT_BUF_SECTORS, &n)) { |
8a426614 KW |
1770 | |
1771 | if (bdrv_read(bs, sector, buf, n) != 0) { | |
1772 | ret = -EIO; | |
1773 | goto ro_cleanup; | |
1774 | } | |
1775 | ||
1776 | if (bdrv_write(bs->backing_hd, sector, buf, n) != 0) { | |
1777 | ret = -EIO; | |
1778 | goto ro_cleanup; | |
1779 | } | |
ea2384d3 | 1780 | } |
33e3963e | 1781 | } |
95389c86 | 1782 | |
1d44952f CH |
1783 | if (drv->bdrv_make_empty) { |
1784 | ret = drv->bdrv_make_empty(bs); | |
1785 | bdrv_flush(bs); | |
1786 | } | |
95389c86 | 1787 | |
3f5075ae CH |
1788 | /* |
1789 | * Make sure all data we wrote to the backing device is actually | |
1790 | * stable on disk. | |
1791 | */ | |
1792 | if (bs->backing_hd) | |
1793 | bdrv_flush(bs->backing_hd); | |
4dca4b63 NS |
1794 | |
1795 | ro_cleanup: | |
7267c094 | 1796 | g_free(buf); |
4dca4b63 NS |
1797 | |
1798 | if (ro) { | |
0bce597d JC |
1799 | /* ignoring error return here */ |
1800 | bdrv_reopen(bs->backing_hd, open_flags & ~BDRV_O_RDWR, NULL); | |
4dca4b63 NS |
1801 | } |
1802 | ||
1d44952f | 1803 | return ret; |
33e3963e FB |
1804 | } |
1805 | ||
e8877497 | 1806 | int bdrv_commit_all(void) |
6ab4b5ab MA |
1807 | { |
1808 | BlockDriverState *bs; | |
1809 | ||
1810 | QTAILQ_FOREACH(bs, &bdrv_states, list) { | |
272d2d8e JC |
1811 | if (bs->drv && bs->backing_hd) { |
1812 | int ret = bdrv_commit(bs); | |
1813 | if (ret < 0) { | |
1814 | return ret; | |
1815 | } | |
e8877497 | 1816 | } |
6ab4b5ab | 1817 | } |
e8877497 | 1818 | return 0; |
6ab4b5ab MA |
1819 | } |
1820 | ||
dbffbdcf SH |
1821 | struct BdrvTrackedRequest { |
1822 | BlockDriverState *bs; | |
1823 | int64_t sector_num; | |
1824 | int nb_sectors; | |
1825 | bool is_write; | |
1826 | QLIST_ENTRY(BdrvTrackedRequest) list; | |
5f8b6491 | 1827 | Coroutine *co; /* owner, used for deadlock detection */ |
f4658285 | 1828 | CoQueue wait_queue; /* coroutines blocked on this request */ |
dbffbdcf SH |
1829 | }; |
1830 | ||
1831 | /** | |
1832 | * Remove an active request from the tracked requests list | |
1833 | * | |
1834 | * This function should be called when a tracked request is completing. | |
1835 | */ | |
1836 | static void tracked_request_end(BdrvTrackedRequest *req) | |
1837 | { | |
1838 | QLIST_REMOVE(req, list); | |
f4658285 | 1839 | qemu_co_queue_restart_all(&req->wait_queue); |
dbffbdcf SH |
1840 | } |
1841 | ||
1842 | /** | |
1843 | * Add an active request to the tracked requests list | |
1844 | */ | |
1845 | static void tracked_request_begin(BdrvTrackedRequest *req, | |
1846 | BlockDriverState *bs, | |
1847 | int64_t sector_num, | |
1848 | int nb_sectors, bool is_write) | |
1849 | { | |
1850 | *req = (BdrvTrackedRequest){ | |
1851 | .bs = bs, | |
1852 | .sector_num = sector_num, | |
1853 | .nb_sectors = nb_sectors, | |
1854 | .is_write = is_write, | |
5f8b6491 | 1855 | .co = qemu_coroutine_self(), |
dbffbdcf SH |
1856 | }; |
1857 | ||
f4658285 SH |
1858 | qemu_co_queue_init(&req->wait_queue); |
1859 | ||
dbffbdcf SH |
1860 | QLIST_INSERT_HEAD(&bs->tracked_requests, req, list); |
1861 | } | |
1862 | ||
d83947ac SH |
1863 | /** |
1864 | * Round a region to cluster boundaries | |
1865 | */ | |
343bded4 PB |
1866 | void bdrv_round_to_clusters(BlockDriverState *bs, |
1867 | int64_t sector_num, int nb_sectors, | |
1868 | int64_t *cluster_sector_num, | |
1869 | int *cluster_nb_sectors) | |
d83947ac SH |
1870 | { |
1871 | BlockDriverInfo bdi; | |
1872 | ||
1873 | if (bdrv_get_info(bs, &bdi) < 0 || bdi.cluster_size == 0) { | |
1874 | *cluster_sector_num = sector_num; | |
1875 | *cluster_nb_sectors = nb_sectors; | |
1876 | } else { | |
1877 | int64_t c = bdi.cluster_size / BDRV_SECTOR_SIZE; | |
1878 | *cluster_sector_num = QEMU_ALIGN_DOWN(sector_num, c); | |
1879 | *cluster_nb_sectors = QEMU_ALIGN_UP(sector_num - *cluster_sector_num + | |
1880 | nb_sectors, c); | |
1881 | } | |
1882 | } | |
1883 | ||
f4658285 SH |
1884 | static bool tracked_request_overlaps(BdrvTrackedRequest *req, |
1885 | int64_t sector_num, int nb_sectors) { | |
d83947ac SH |
1886 | /* aaaa bbbb */ |
1887 | if (sector_num >= req->sector_num + req->nb_sectors) { | |
1888 | return false; | |
1889 | } | |
1890 | /* bbbb aaaa */ | |
1891 | if (req->sector_num >= sector_num + nb_sectors) { | |
1892 | return false; | |
1893 | } | |
1894 | return true; | |
f4658285 SH |
1895 | } |
1896 | ||
1897 | static void coroutine_fn wait_for_overlapping_requests(BlockDriverState *bs, | |
1898 | int64_t sector_num, int nb_sectors) | |
1899 | { | |
1900 | BdrvTrackedRequest *req; | |
d83947ac SH |
1901 | int64_t cluster_sector_num; |
1902 | int cluster_nb_sectors; | |
f4658285 SH |
1903 | bool retry; |
1904 | ||
d83947ac SH |
1905 | /* If we touch the same cluster it counts as an overlap. This guarantees |
1906 | * that allocating writes will be serialized and not race with each other | |
1907 | * for the same cluster. For example, in copy-on-read it ensures that the | |
1908 | * CoR read and write operations are atomic and guest writes cannot | |
1909 | * interleave between them. | |
1910 | */ | |
343bded4 PB |
1911 | bdrv_round_to_clusters(bs, sector_num, nb_sectors, |
1912 | &cluster_sector_num, &cluster_nb_sectors); | |
d83947ac | 1913 | |
f4658285 SH |
1914 | do { |
1915 | retry = false; | |
1916 | QLIST_FOREACH(req, &bs->tracked_requests, list) { | |
d83947ac SH |
1917 | if (tracked_request_overlaps(req, cluster_sector_num, |
1918 | cluster_nb_sectors)) { | |
5f8b6491 SH |
1919 | /* Hitting this means there was a reentrant request, for |
1920 | * example, a block driver issuing nested requests. This must | |
1921 | * never happen since it means deadlock. | |
1922 | */ | |
1923 | assert(qemu_coroutine_self() != req->co); | |
1924 | ||
f4658285 SH |
1925 | qemu_co_queue_wait(&req->wait_queue); |
1926 | retry = true; | |
1927 | break; | |
1928 | } | |
1929 | } | |
1930 | } while (retry); | |
1931 | } | |
1932 | ||
756e6736 KW |
1933 | /* |
1934 | * Return values: | |
1935 | * 0 - success | |
1936 | * -EINVAL - backing format specified, but no file | |
1937 | * -ENOSPC - can't update the backing file because no space is left in the | |
1938 | * image file header | |
1939 | * -ENOTSUP - format driver doesn't support changing the backing file | |
1940 | */ | |
1941 | int bdrv_change_backing_file(BlockDriverState *bs, | |
1942 | const char *backing_file, const char *backing_fmt) | |
1943 | { | |
1944 | BlockDriver *drv = bs->drv; | |
469ef350 | 1945 | int ret; |
756e6736 | 1946 | |
5f377794 PB |
1947 | /* Backing file format doesn't make sense without a backing file */ |
1948 | if (backing_fmt && !backing_file) { | |
1949 | return -EINVAL; | |
1950 | } | |
1951 | ||
756e6736 | 1952 | if (drv->bdrv_change_backing_file != NULL) { |
469ef350 | 1953 | ret = drv->bdrv_change_backing_file(bs, backing_file, backing_fmt); |
756e6736 | 1954 | } else { |
469ef350 | 1955 | ret = -ENOTSUP; |
756e6736 | 1956 | } |
469ef350 PB |
1957 | |
1958 | if (ret == 0) { | |
1959 | pstrcpy(bs->backing_file, sizeof(bs->backing_file), backing_file ?: ""); | |
1960 | pstrcpy(bs->backing_format, sizeof(bs->backing_format), backing_fmt ?: ""); | |
1961 | } | |
1962 | return ret; | |
756e6736 KW |
1963 | } |
1964 | ||
6ebdcee2 JC |
1965 | /* |
1966 | * Finds the image layer in the chain that has 'bs' as its backing file. | |
1967 | * | |
1968 | * active is the current topmost image. | |
1969 | * | |
1970 | * Returns NULL if bs is not found in active's image chain, | |
1971 | * or if active == bs. | |
1972 | */ | |
1973 | BlockDriverState *bdrv_find_overlay(BlockDriverState *active, | |
1974 | BlockDriverState *bs) | |
1975 | { | |
1976 | BlockDriverState *overlay = NULL; | |
1977 | BlockDriverState *intermediate; | |
1978 | ||
1979 | assert(active != NULL); | |
1980 | assert(bs != NULL); | |
1981 | ||
1982 | /* if bs is the same as active, then by definition it has no overlay | |
1983 | */ | |
1984 | if (active == bs) { | |
1985 | return NULL; | |
1986 | } | |
1987 | ||
1988 | intermediate = active; | |
1989 | while (intermediate->backing_hd) { | |
1990 | if (intermediate->backing_hd == bs) { | |
1991 | overlay = intermediate; | |
1992 | break; | |
1993 | } | |
1994 | intermediate = intermediate->backing_hd; | |
1995 | } | |
1996 | ||
1997 | return overlay; | |
1998 | } | |
1999 | ||
2000 | typedef struct BlkIntermediateStates { | |
2001 | BlockDriverState *bs; | |
2002 | QSIMPLEQ_ENTRY(BlkIntermediateStates) entry; | |
2003 | } BlkIntermediateStates; | |
2004 | ||
2005 | ||
2006 | /* | |
2007 | * Drops images above 'base' up to and including 'top', and sets the image | |
2008 | * above 'top' to have base as its backing file. | |
2009 | * | |
2010 | * Requires that the overlay to 'top' is opened r/w, so that the backing file | |
2011 | * information in 'bs' can be properly updated. | |
2012 | * | |
2013 | * E.g., this will convert the following chain: | |
2014 | * bottom <- base <- intermediate <- top <- active | |
2015 | * | |
2016 | * to | |
2017 | * | |
2018 | * bottom <- base <- active | |
2019 | * | |
2020 | * It is allowed for bottom==base, in which case it converts: | |
2021 | * | |
2022 | * base <- intermediate <- top <- active | |
2023 | * | |
2024 | * to | |
2025 | * | |
2026 | * base <- active | |
2027 | * | |
2028 | * Error conditions: | |
2029 | * if active == top, that is considered an error | |
2030 | * | |
2031 | */ | |
2032 | int bdrv_drop_intermediate(BlockDriverState *active, BlockDriverState *top, | |
2033 | BlockDriverState *base) | |
2034 | { | |
2035 | BlockDriverState *intermediate; | |
2036 | BlockDriverState *base_bs = NULL; | |
2037 | BlockDriverState *new_top_bs = NULL; | |
2038 | BlkIntermediateStates *intermediate_state, *next; | |
2039 | int ret = -EIO; | |
2040 | ||
2041 | QSIMPLEQ_HEAD(states_to_delete, BlkIntermediateStates) states_to_delete; | |
2042 | QSIMPLEQ_INIT(&states_to_delete); | |
2043 | ||
2044 | if (!top->drv || !base->drv) { | |
2045 | goto exit; | |
2046 | } | |
2047 | ||
2048 | new_top_bs = bdrv_find_overlay(active, top); | |
2049 | ||
2050 | if (new_top_bs == NULL) { | |
2051 | /* we could not find the image above 'top', this is an error */ | |
2052 | goto exit; | |
2053 | } | |
2054 | ||
2055 | /* special case of new_top_bs->backing_hd already pointing to base - nothing | |
2056 | * to do, no intermediate images */ | |
2057 | if (new_top_bs->backing_hd == base) { | |
2058 | ret = 0; | |
2059 | goto exit; | |
2060 | } | |
2061 | ||
2062 | intermediate = top; | |
2063 | ||
2064 | /* now we will go down through the list, and add each BDS we find | |
2065 | * into our deletion queue, until we hit the 'base' | |
2066 | */ | |
2067 | while (intermediate) { | |
2068 | intermediate_state = g_malloc0(sizeof(BlkIntermediateStates)); | |
2069 | intermediate_state->bs = intermediate; | |
2070 | QSIMPLEQ_INSERT_TAIL(&states_to_delete, intermediate_state, entry); | |
2071 | ||
2072 | if (intermediate->backing_hd == base) { | |
2073 | base_bs = intermediate->backing_hd; | |
2074 | break; | |
2075 | } | |
2076 | intermediate = intermediate->backing_hd; | |
2077 | } | |
2078 | if (base_bs == NULL) { | |
2079 | /* something went wrong, we did not end at the base. safely | |
2080 | * unravel everything, and exit with error */ | |
2081 | goto exit; | |
2082 | } | |
2083 | ||
2084 | /* success - we can delete the intermediate states, and link top->base */ | |
2085 | ret = bdrv_change_backing_file(new_top_bs, base_bs->filename, | |
2086 | base_bs->drv ? base_bs->drv->format_name : ""); | |
2087 | if (ret) { | |
2088 | goto exit; | |
2089 | } | |
2090 | new_top_bs->backing_hd = base_bs; | |
2091 | ||
2092 | ||
2093 | QSIMPLEQ_FOREACH_SAFE(intermediate_state, &states_to_delete, entry, next) { | |
2094 | /* so that bdrv_close() does not recursively close the chain */ | |
2095 | intermediate_state->bs->backing_hd = NULL; | |
2096 | bdrv_delete(intermediate_state->bs); | |
2097 | } | |
2098 | ret = 0; | |
2099 | ||
2100 | exit: | |
2101 | QSIMPLEQ_FOREACH_SAFE(intermediate_state, &states_to_delete, entry, next) { | |
2102 | g_free(intermediate_state); | |
2103 | } | |
2104 | return ret; | |
2105 | } | |
2106 | ||
2107 | ||
71d0770c AL |
2108 | static int bdrv_check_byte_request(BlockDriverState *bs, int64_t offset, |
2109 | size_t size) | |
2110 | { | |
2111 | int64_t len; | |
2112 | ||
2113 | if (!bdrv_is_inserted(bs)) | |
2114 | return -ENOMEDIUM; | |
2115 | ||
2116 | if (bs->growable) | |
2117 | return 0; | |
2118 | ||
2119 | len = bdrv_getlength(bs); | |
2120 | ||
fbb7b4e0 KW |
2121 | if (offset < 0) |
2122 | return -EIO; | |
2123 | ||
2124 | if ((offset > len) || (len - offset < size)) | |
71d0770c AL |
2125 | return -EIO; |
2126 | ||
2127 | return 0; | |
2128 | } | |
2129 | ||
2130 | static int bdrv_check_request(BlockDriverState *bs, int64_t sector_num, | |
2131 | int nb_sectors) | |
2132 | { | |
eb5a3165 JS |
2133 | return bdrv_check_byte_request(bs, sector_num * BDRV_SECTOR_SIZE, |
2134 | nb_sectors * BDRV_SECTOR_SIZE); | |
71d0770c AL |
2135 | } |
2136 | ||
1c9805a3 SH |
2137 | typedef struct RwCo { |
2138 | BlockDriverState *bs; | |
2139 | int64_t sector_num; | |
2140 | int nb_sectors; | |
2141 | QEMUIOVector *qiov; | |
2142 | bool is_write; | |
2143 | int ret; | |
2144 | } RwCo; | |
2145 | ||
2146 | static void coroutine_fn bdrv_rw_co_entry(void *opaque) | |
fc01f7e7 | 2147 | { |
1c9805a3 | 2148 | RwCo *rwco = opaque; |
ea2384d3 | 2149 | |
1c9805a3 SH |
2150 | if (!rwco->is_write) { |
2151 | rwco->ret = bdrv_co_do_readv(rwco->bs, rwco->sector_num, | |
470c0504 | 2152 | rwco->nb_sectors, rwco->qiov, 0); |
1c9805a3 SH |
2153 | } else { |
2154 | rwco->ret = bdrv_co_do_writev(rwco->bs, rwco->sector_num, | |
f08f2dda | 2155 | rwco->nb_sectors, rwco->qiov, 0); |
1c9805a3 SH |
2156 | } |
2157 | } | |
e7a8a783 | 2158 | |
1c9805a3 | 2159 | /* |
8d3b1a2d | 2160 | * Process a vectored synchronous request using coroutines |
1c9805a3 | 2161 | */ |
8d3b1a2d KW |
2162 | static int bdrv_rwv_co(BlockDriverState *bs, int64_t sector_num, |
2163 | QEMUIOVector *qiov, bool is_write) | |
1c9805a3 | 2164 | { |
1c9805a3 SH |
2165 | Coroutine *co; |
2166 | RwCo rwco = { | |
2167 | .bs = bs, | |
2168 | .sector_num = sector_num, | |
8d3b1a2d KW |
2169 | .nb_sectors = qiov->size >> BDRV_SECTOR_BITS, |
2170 | .qiov = qiov, | |
1c9805a3 SH |
2171 | .is_write = is_write, |
2172 | .ret = NOT_DONE, | |
2173 | }; | |
8d3b1a2d | 2174 | assert((qiov->size & (BDRV_SECTOR_SIZE - 1)) == 0); |
e7a8a783 | 2175 | |
498e386c ZYW |
2176 | /** |
2177 | * In sync call context, when the vcpu is blocked, this throttling timer | |
2178 | * will not fire; so the I/O throttling function has to be disabled here | |
2179 | * if it has been enabled. | |
2180 | */ | |
2181 | if (bs->io_limits_enabled) { | |
2182 | fprintf(stderr, "Disabling I/O throttling on '%s' due " | |
2183 | "to synchronous I/O.\n", bdrv_get_device_name(bs)); | |
2184 | bdrv_io_limits_disable(bs); | |
2185 | } | |
2186 | ||
1c9805a3 SH |
2187 | if (qemu_in_coroutine()) { |
2188 | /* Fast-path if already in coroutine context */ | |
2189 | bdrv_rw_co_entry(&rwco); | |
2190 | } else { | |
2191 | co = qemu_coroutine_create(bdrv_rw_co_entry); | |
2192 | qemu_coroutine_enter(co, &rwco); | |
2193 | while (rwco.ret == NOT_DONE) { | |
2194 | qemu_aio_wait(); | |
2195 | } | |
2196 | } | |
2197 | return rwco.ret; | |
2198 | } | |
b338082b | 2199 | |
8d3b1a2d KW |
2200 | /* |
2201 | * Process a synchronous request using coroutines | |
2202 | */ | |
2203 | static int bdrv_rw_co(BlockDriverState *bs, int64_t sector_num, uint8_t *buf, | |
2204 | int nb_sectors, bool is_write) | |
2205 | { | |
2206 | QEMUIOVector qiov; | |
2207 | struct iovec iov = { | |
2208 | .iov_base = (void *)buf, | |
2209 | .iov_len = nb_sectors * BDRV_SECTOR_SIZE, | |
2210 | }; | |
2211 | ||
2212 | qemu_iovec_init_external(&qiov, &iov, 1); | |
2213 | return bdrv_rwv_co(bs, sector_num, &qiov, is_write); | |
2214 | } | |
2215 | ||
1c9805a3 SH |
2216 | /* return < 0 if error. See bdrv_write() for the return codes */ |
2217 | int bdrv_read(BlockDriverState *bs, int64_t sector_num, | |
2218 | uint8_t *buf, int nb_sectors) | |
2219 | { | |
2220 | return bdrv_rw_co(bs, sector_num, buf, nb_sectors, false); | |
fc01f7e7 FB |
2221 | } |
2222 | ||
07d27a44 MA |
2223 | /* Just like bdrv_read(), but with I/O throttling temporarily disabled */ |
2224 | int bdrv_read_unthrottled(BlockDriverState *bs, int64_t sector_num, | |
2225 | uint8_t *buf, int nb_sectors) | |
2226 | { | |
2227 | bool enabled; | |
2228 | int ret; | |
2229 | ||
2230 | enabled = bs->io_limits_enabled; | |
2231 | bs->io_limits_enabled = false; | |
2232 | ret = bdrv_read(bs, 0, buf, 1); | |
2233 | bs->io_limits_enabled = enabled; | |
2234 | return ret; | |
2235 | } | |
2236 | ||
5fafdf24 | 2237 | /* Return < 0 if error. Important errors are: |
19cb3738 FB |
2238 | -EIO generic I/O error (may happen for all errors) |
2239 | -ENOMEDIUM No media inserted. | |
2240 | -EINVAL Invalid sector number or nb_sectors | |
2241 | -EACCES Trying to write a read-only device | |
2242 | */ | |
5fafdf24 | 2243 | int bdrv_write(BlockDriverState *bs, int64_t sector_num, |
fc01f7e7 FB |
2244 | const uint8_t *buf, int nb_sectors) |
2245 | { | |
1c9805a3 | 2246 | return bdrv_rw_co(bs, sector_num, (uint8_t *)buf, nb_sectors, true); |
83f64091 FB |
2247 | } |
2248 | ||
8d3b1a2d KW |
2249 | int bdrv_writev(BlockDriverState *bs, int64_t sector_num, QEMUIOVector *qiov) |
2250 | { | |
2251 | return bdrv_rwv_co(bs, sector_num, qiov, true); | |
2252 | } | |
2253 | ||
eda578e5 AL |
2254 | int bdrv_pread(BlockDriverState *bs, int64_t offset, |
2255 | void *buf, int count1) | |
83f64091 | 2256 | { |
6ea44308 | 2257 | uint8_t tmp_buf[BDRV_SECTOR_SIZE]; |
83f64091 FB |
2258 | int len, nb_sectors, count; |
2259 | int64_t sector_num; | |
9a8c4cce | 2260 | int ret; |
83f64091 FB |
2261 | |
2262 | count = count1; | |
2263 | /* first read to align to sector start */ | |
6ea44308 | 2264 | len = (BDRV_SECTOR_SIZE - offset) & (BDRV_SECTOR_SIZE - 1); |
83f64091 FB |
2265 | if (len > count) |
2266 | len = count; | |
6ea44308 | 2267 | sector_num = offset >> BDRV_SECTOR_BITS; |
83f64091 | 2268 | if (len > 0) { |
9a8c4cce KW |
2269 | if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0) |
2270 | return ret; | |
6ea44308 | 2271 | memcpy(buf, tmp_buf + (offset & (BDRV_SECTOR_SIZE - 1)), len); |
83f64091 FB |
2272 | count -= len; |
2273 | if (count == 0) | |
2274 | return count1; | |
2275 | sector_num++; | |
2276 | buf += len; | |
2277 | } | |
2278 | ||
2279 | /* read the sectors "in place" */ | |
6ea44308 | 2280 | nb_sectors = count >> BDRV_SECTOR_BITS; |
83f64091 | 2281 | if (nb_sectors > 0) { |
9a8c4cce KW |
2282 | if ((ret = bdrv_read(bs, sector_num, buf, nb_sectors)) < 0) |
2283 | return ret; | |
83f64091 | 2284 | sector_num += nb_sectors; |
6ea44308 | 2285 | len = nb_sectors << BDRV_SECTOR_BITS; |
83f64091 FB |
2286 | buf += len; |
2287 | count -= len; | |
2288 | } | |
2289 | ||
2290 | /* add data from the last sector */ | |
2291 | if (count > 0) { | |
9a8c4cce KW |
2292 | if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0) |
2293 | return ret; | |
83f64091 FB |
2294 | memcpy(buf, tmp_buf, count); |
2295 | } | |
2296 | return count1; | |
2297 | } | |
2298 | ||
8d3b1a2d | 2299 | int bdrv_pwritev(BlockDriverState *bs, int64_t offset, QEMUIOVector *qiov) |
83f64091 | 2300 | { |
6ea44308 | 2301 | uint8_t tmp_buf[BDRV_SECTOR_SIZE]; |
83f64091 FB |
2302 | int len, nb_sectors, count; |
2303 | int64_t sector_num; | |
9a8c4cce | 2304 | int ret; |
83f64091 | 2305 | |
8d3b1a2d KW |
2306 | count = qiov->size; |
2307 | ||
83f64091 | 2308 | /* first write to align to sector start */ |
6ea44308 | 2309 | len = (BDRV_SECTOR_SIZE - offset) & (BDRV_SECTOR_SIZE - 1); |
83f64091 FB |
2310 | if (len > count) |
2311 | len = count; | |
6ea44308 | 2312 | sector_num = offset >> BDRV_SECTOR_BITS; |
83f64091 | 2313 | if (len > 0) { |
9a8c4cce KW |
2314 | if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0) |
2315 | return ret; | |
8d3b1a2d KW |
2316 | qemu_iovec_to_buf(qiov, 0, tmp_buf + (offset & (BDRV_SECTOR_SIZE - 1)), |
2317 | len); | |
9a8c4cce KW |
2318 | if ((ret = bdrv_write(bs, sector_num, tmp_buf, 1)) < 0) |
2319 | return ret; | |
83f64091 FB |
2320 | count -= len; |
2321 | if (count == 0) | |
8d3b1a2d | 2322 | return qiov->size; |
83f64091 | 2323 | sector_num++; |
83f64091 FB |
2324 | } |
2325 | ||
2326 | /* write the sectors "in place" */ | |
6ea44308 | 2327 | nb_sectors = count >> BDRV_SECTOR_BITS; |
83f64091 | 2328 | if (nb_sectors > 0) { |
8d3b1a2d KW |
2329 | QEMUIOVector qiov_inplace; |
2330 | ||
2331 | qemu_iovec_init(&qiov_inplace, qiov->niov); | |
2332 | qemu_iovec_concat(&qiov_inplace, qiov, len, | |
2333 | nb_sectors << BDRV_SECTOR_BITS); | |
2334 | ret = bdrv_writev(bs, sector_num, &qiov_inplace); | |
2335 | qemu_iovec_destroy(&qiov_inplace); | |
2336 | if (ret < 0) { | |
9a8c4cce | 2337 | return ret; |
8d3b1a2d KW |
2338 | } |
2339 | ||
83f64091 | 2340 | sector_num += nb_sectors; |
6ea44308 | 2341 | len = nb_sectors << BDRV_SECTOR_BITS; |
83f64091 FB |
2342 | count -= len; |
2343 | } | |
2344 | ||
2345 | /* add data from the last sector */ | |
2346 | if (count > 0) { | |
9a8c4cce KW |
2347 | if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0) |
2348 | return ret; | |
8d3b1a2d | 2349 | qemu_iovec_to_buf(qiov, qiov->size - count, tmp_buf, count); |
9a8c4cce KW |
2350 | if ((ret = bdrv_write(bs, sector_num, tmp_buf, 1)) < 0) |
2351 | return ret; | |
83f64091 | 2352 | } |
8d3b1a2d KW |
2353 | return qiov->size; |
2354 | } | |
2355 | ||
2356 | int bdrv_pwrite(BlockDriverState *bs, int64_t offset, | |
2357 | const void *buf, int count1) | |
2358 | { | |
2359 | QEMUIOVector qiov; | |
2360 | struct iovec iov = { | |
2361 | .iov_base = (void *) buf, | |
2362 | .iov_len = count1, | |
2363 | }; | |
2364 | ||
2365 | qemu_iovec_init_external(&qiov, &iov, 1); | |
2366 | return bdrv_pwritev(bs, offset, &qiov); | |
83f64091 | 2367 | } |
83f64091 | 2368 | |
f08145fe KW |
2369 | /* |
2370 | * Writes to the file and ensures that no writes are reordered across this | |
2371 | * request (acts as a barrier) | |
2372 | * | |
2373 | * Returns 0 on success, -errno in error cases. | |
2374 | */ | |
2375 | int bdrv_pwrite_sync(BlockDriverState *bs, int64_t offset, | |
2376 | const void *buf, int count) | |
2377 | { | |
2378 | int ret; | |
2379 | ||
2380 | ret = bdrv_pwrite(bs, offset, buf, count); | |
2381 | if (ret < 0) { | |
2382 | return ret; | |
2383 | } | |
2384 | ||
f05fa4ad PB |
2385 | /* No flush needed for cache modes that already do it */ |
2386 | if (bs->enable_write_cache) { | |
f08145fe KW |
2387 | bdrv_flush(bs); |
2388 | } | |
2389 | ||
2390 | return 0; | |
2391 | } | |
2392 | ||
470c0504 | 2393 | static int coroutine_fn bdrv_co_do_copy_on_readv(BlockDriverState *bs, |
ab185921 SH |
2394 | int64_t sector_num, int nb_sectors, QEMUIOVector *qiov) |
2395 | { | |
2396 | /* Perform I/O through a temporary buffer so that users who scribble over | |
2397 | * their read buffer while the operation is in progress do not end up | |
2398 | * modifying the image file. This is critical for zero-copy guest I/O | |
2399 | * where anything might happen inside guest memory. | |
2400 | */ | |
2401 | void *bounce_buffer; | |
2402 | ||
79c053bd | 2403 | BlockDriver *drv = bs->drv; |
ab185921 SH |
2404 | struct iovec iov; |
2405 | QEMUIOVector bounce_qiov; | |
2406 | int64_t cluster_sector_num; | |
2407 | int cluster_nb_sectors; | |
2408 | size_t skip_bytes; | |
2409 | int ret; | |
2410 | ||
2411 | /* Cover entire cluster so no additional backing file I/O is required when | |
2412 | * allocating cluster in the image file. | |
2413 | */ | |
343bded4 PB |
2414 | bdrv_round_to_clusters(bs, sector_num, nb_sectors, |
2415 | &cluster_sector_num, &cluster_nb_sectors); | |
ab185921 | 2416 | |
470c0504 SH |
2417 | trace_bdrv_co_do_copy_on_readv(bs, sector_num, nb_sectors, |
2418 | cluster_sector_num, cluster_nb_sectors); | |
ab185921 SH |
2419 | |
2420 | iov.iov_len = cluster_nb_sectors * BDRV_SECTOR_SIZE; | |
2421 | iov.iov_base = bounce_buffer = qemu_blockalign(bs, iov.iov_len); | |
2422 | qemu_iovec_init_external(&bounce_qiov, &iov, 1); | |
2423 | ||
79c053bd SH |
2424 | ret = drv->bdrv_co_readv(bs, cluster_sector_num, cluster_nb_sectors, |
2425 | &bounce_qiov); | |
ab185921 SH |
2426 | if (ret < 0) { |
2427 | goto err; | |
2428 | } | |
2429 | ||
79c053bd SH |
2430 | if (drv->bdrv_co_write_zeroes && |
2431 | buffer_is_zero(bounce_buffer, iov.iov_len)) { | |
621f0589 KW |
2432 | ret = bdrv_co_do_write_zeroes(bs, cluster_sector_num, |
2433 | cluster_nb_sectors); | |
79c053bd | 2434 | } else { |
f05fa4ad PB |
2435 | /* This does not change the data on the disk, it is not necessary |
2436 | * to flush even in cache=writethrough mode. | |
2437 | */ | |
79c053bd | 2438 | ret = drv->bdrv_co_writev(bs, cluster_sector_num, cluster_nb_sectors, |
ab185921 | 2439 | &bounce_qiov); |
79c053bd SH |
2440 | } |
2441 | ||
ab185921 SH |
2442 | if (ret < 0) { |
2443 | /* It might be okay to ignore write errors for guest requests. If this | |
2444 | * is a deliberate copy-on-read then we don't want to ignore the error. | |
2445 | * Simply report it in all cases. | |
2446 | */ | |
2447 | goto err; | |
2448 | } | |
2449 | ||
2450 | skip_bytes = (sector_num - cluster_sector_num) * BDRV_SECTOR_SIZE; | |
03396148 MT |
2451 | qemu_iovec_from_buf(qiov, 0, bounce_buffer + skip_bytes, |
2452 | nb_sectors * BDRV_SECTOR_SIZE); | |
ab185921 SH |
2453 | |
2454 | err: | |
2455 | qemu_vfree(bounce_buffer); | |
2456 | return ret; | |
2457 | } | |
2458 | ||
c5fbe571 SH |
2459 | /* |
2460 | * Handle a read request in coroutine context | |
2461 | */ | |
2462 | static int coroutine_fn bdrv_co_do_readv(BlockDriverState *bs, | |
470c0504 SH |
2463 | int64_t sector_num, int nb_sectors, QEMUIOVector *qiov, |
2464 | BdrvRequestFlags flags) | |
da1fa91d KW |
2465 | { |
2466 | BlockDriver *drv = bs->drv; | |
dbffbdcf SH |
2467 | BdrvTrackedRequest req; |
2468 | int ret; | |
da1fa91d | 2469 | |
da1fa91d KW |
2470 | if (!drv) { |
2471 | return -ENOMEDIUM; | |
2472 | } | |
2473 | if (bdrv_check_request(bs, sector_num, nb_sectors)) { | |
2474 | return -EIO; | |
2475 | } | |
2476 | ||
98f90dba ZYW |
2477 | /* throttling disk read I/O */ |
2478 | if (bs->io_limits_enabled) { | |
2479 | bdrv_io_limits_intercept(bs, false, nb_sectors); | |
2480 | } | |
2481 | ||
f4658285 | 2482 | if (bs->copy_on_read) { |
470c0504 SH |
2483 | flags |= BDRV_REQ_COPY_ON_READ; |
2484 | } | |
2485 | if (flags & BDRV_REQ_COPY_ON_READ) { | |
2486 | bs->copy_on_read_in_flight++; | |
2487 | } | |
2488 | ||
2489 | if (bs->copy_on_read_in_flight) { | |
f4658285 SH |
2490 | wait_for_overlapping_requests(bs, sector_num, nb_sectors); |
2491 | } | |
2492 | ||
dbffbdcf | 2493 | tracked_request_begin(&req, bs, sector_num, nb_sectors, false); |
ab185921 | 2494 | |
470c0504 | 2495 | if (flags & BDRV_REQ_COPY_ON_READ) { |
ab185921 SH |
2496 | int pnum; |
2497 | ||
2498 | ret = bdrv_co_is_allocated(bs, sector_num, nb_sectors, &pnum); | |
2499 | if (ret < 0) { | |
2500 | goto out; | |
2501 | } | |
2502 | ||
2503 | if (!ret || pnum != nb_sectors) { | |
470c0504 | 2504 | ret = bdrv_co_do_copy_on_readv(bs, sector_num, nb_sectors, qiov); |
ab185921 SH |
2505 | goto out; |
2506 | } | |
2507 | } | |
2508 | ||
dbffbdcf | 2509 | ret = drv->bdrv_co_readv(bs, sector_num, nb_sectors, qiov); |
ab185921 SH |
2510 | |
2511 | out: | |
dbffbdcf | 2512 | tracked_request_end(&req); |
470c0504 SH |
2513 | |
2514 | if (flags & BDRV_REQ_COPY_ON_READ) { | |
2515 | bs->copy_on_read_in_flight--; | |
2516 | } | |
2517 | ||
dbffbdcf | 2518 | return ret; |
da1fa91d KW |
2519 | } |
2520 | ||
c5fbe571 | 2521 | int coroutine_fn bdrv_co_readv(BlockDriverState *bs, int64_t sector_num, |
da1fa91d KW |
2522 | int nb_sectors, QEMUIOVector *qiov) |
2523 | { | |
c5fbe571 | 2524 | trace_bdrv_co_readv(bs, sector_num, nb_sectors); |
da1fa91d | 2525 | |
470c0504 SH |
2526 | return bdrv_co_do_readv(bs, sector_num, nb_sectors, qiov, 0); |
2527 | } | |
2528 | ||
2529 | int coroutine_fn bdrv_co_copy_on_readv(BlockDriverState *bs, | |
2530 | int64_t sector_num, int nb_sectors, QEMUIOVector *qiov) | |
2531 | { | |
2532 | trace_bdrv_co_copy_on_readv(bs, sector_num, nb_sectors); | |
2533 | ||
2534 | return bdrv_co_do_readv(bs, sector_num, nb_sectors, qiov, | |
2535 | BDRV_REQ_COPY_ON_READ); | |
c5fbe571 SH |
2536 | } |
2537 | ||
f08f2dda SH |
2538 | static int coroutine_fn bdrv_co_do_write_zeroes(BlockDriverState *bs, |
2539 | int64_t sector_num, int nb_sectors) | |
2540 | { | |
2541 | BlockDriver *drv = bs->drv; | |
2542 | QEMUIOVector qiov; | |
2543 | struct iovec iov; | |
2544 | int ret; | |
2545 | ||
621f0589 KW |
2546 | /* TODO Emulate only part of misaligned requests instead of letting block |
2547 | * drivers return -ENOTSUP and emulate everything */ | |
2548 | ||
f08f2dda SH |
2549 | /* First try the efficient write zeroes operation */ |
2550 | if (drv->bdrv_co_write_zeroes) { | |
621f0589 KW |
2551 | ret = drv->bdrv_co_write_zeroes(bs, sector_num, nb_sectors); |
2552 | if (ret != -ENOTSUP) { | |
2553 | return ret; | |
2554 | } | |
f08f2dda SH |
2555 | } |
2556 | ||
2557 | /* Fall back to bounce buffer if write zeroes is unsupported */ | |
2558 | iov.iov_len = nb_sectors * BDRV_SECTOR_SIZE; | |
2559 | iov.iov_base = qemu_blockalign(bs, iov.iov_len); | |
2560 | memset(iov.iov_base, 0, iov.iov_len); | |
2561 | qemu_iovec_init_external(&qiov, &iov, 1); | |
2562 | ||
2563 | ret = drv->bdrv_co_writev(bs, sector_num, nb_sectors, &qiov); | |
2564 | ||
2565 | qemu_vfree(iov.iov_base); | |
2566 | return ret; | |
2567 | } | |
2568 | ||
c5fbe571 SH |
2569 | /* |
2570 | * Handle a write request in coroutine context | |
2571 | */ | |
2572 | static int coroutine_fn bdrv_co_do_writev(BlockDriverState *bs, | |
f08f2dda SH |
2573 | int64_t sector_num, int nb_sectors, QEMUIOVector *qiov, |
2574 | BdrvRequestFlags flags) | |
c5fbe571 SH |
2575 | { |
2576 | BlockDriver *drv = bs->drv; | |
dbffbdcf | 2577 | BdrvTrackedRequest req; |
6b7cb247 | 2578 | int ret; |
da1fa91d KW |
2579 | |
2580 | if (!bs->drv) { | |
2581 | return -ENOMEDIUM; | |
2582 | } | |
2583 | if (bs->read_only) { | |
2584 | return -EACCES; | |
2585 | } | |
2586 | if (bdrv_check_request(bs, sector_num, nb_sectors)) { | |
2587 | return -EIO; | |
2588 | } | |
2589 | ||
98f90dba ZYW |
2590 | /* throttling disk write I/O */ |
2591 | if (bs->io_limits_enabled) { | |
2592 | bdrv_io_limits_intercept(bs, true, nb_sectors); | |
2593 | } | |
2594 | ||
470c0504 | 2595 | if (bs->copy_on_read_in_flight) { |
f4658285 SH |
2596 | wait_for_overlapping_requests(bs, sector_num, nb_sectors); |
2597 | } | |
2598 | ||
dbffbdcf SH |
2599 | tracked_request_begin(&req, bs, sector_num, nb_sectors, true); |
2600 | ||
f08f2dda SH |
2601 | if (flags & BDRV_REQ_ZERO_WRITE) { |
2602 | ret = bdrv_co_do_write_zeroes(bs, sector_num, nb_sectors); | |
2603 | } else { | |
2604 | ret = drv->bdrv_co_writev(bs, sector_num, nb_sectors, qiov); | |
2605 | } | |
6b7cb247 | 2606 | |
f05fa4ad PB |
2607 | if (ret == 0 && !bs->enable_write_cache) { |
2608 | ret = bdrv_co_flush(bs); | |
2609 | } | |
2610 | ||
da1fa91d | 2611 | if (bs->dirty_bitmap) { |
1755da16 | 2612 | bdrv_set_dirty(bs, sector_num, nb_sectors); |
da1fa91d KW |
2613 | } |
2614 | ||
2615 | if (bs->wr_highest_sector < sector_num + nb_sectors - 1) { | |
2616 | bs->wr_highest_sector = sector_num + nb_sectors - 1; | |
2617 | } | |
2618 | ||
dbffbdcf SH |
2619 | tracked_request_end(&req); |
2620 | ||
6b7cb247 | 2621 | return ret; |
da1fa91d KW |
2622 | } |
2623 | ||
c5fbe571 SH |
2624 | int coroutine_fn bdrv_co_writev(BlockDriverState *bs, int64_t sector_num, |
2625 | int nb_sectors, QEMUIOVector *qiov) | |
2626 | { | |
2627 | trace_bdrv_co_writev(bs, sector_num, nb_sectors); | |
2628 | ||
f08f2dda SH |
2629 | return bdrv_co_do_writev(bs, sector_num, nb_sectors, qiov, 0); |
2630 | } | |
2631 | ||
2632 | int coroutine_fn bdrv_co_write_zeroes(BlockDriverState *bs, | |
2633 | int64_t sector_num, int nb_sectors) | |
2634 | { | |
2635 | trace_bdrv_co_write_zeroes(bs, sector_num, nb_sectors); | |
2636 | ||
2637 | return bdrv_co_do_writev(bs, sector_num, nb_sectors, NULL, | |
2638 | BDRV_REQ_ZERO_WRITE); | |
c5fbe571 SH |
2639 | } |
2640 | ||
83f64091 FB |
2641 | /** |
2642 | * Truncate file to 'offset' bytes (needed only for file protocols) | |
2643 | */ | |
2644 | int bdrv_truncate(BlockDriverState *bs, int64_t offset) | |
2645 | { | |
2646 | BlockDriver *drv = bs->drv; | |
51762288 | 2647 | int ret; |
83f64091 | 2648 | if (!drv) |
19cb3738 | 2649 | return -ENOMEDIUM; |
83f64091 FB |
2650 | if (!drv->bdrv_truncate) |
2651 | return -ENOTSUP; | |
59f2689d NS |
2652 | if (bs->read_only) |
2653 | return -EACCES; | |
8591675f MT |
2654 | if (bdrv_in_use(bs)) |
2655 | return -EBUSY; | |
51762288 SH |
2656 | ret = drv->bdrv_truncate(bs, offset); |
2657 | if (ret == 0) { | |
2658 | ret = refresh_total_sectors(bs, offset >> BDRV_SECTOR_BITS); | |
145feb17 | 2659 | bdrv_dev_resize_cb(bs); |
51762288 SH |
2660 | } |
2661 | return ret; | |
83f64091 FB |
2662 | } |
2663 | ||
4a1d5e1f FZ |
2664 | /** |
2665 | * Length of a allocated file in bytes. Sparse files are counted by actual | |
2666 | * allocated space. Return < 0 if error or unknown. | |
2667 | */ | |
2668 | int64_t bdrv_get_allocated_file_size(BlockDriverState *bs) | |
2669 | { | |
2670 | BlockDriver *drv = bs->drv; | |
2671 | if (!drv) { | |
2672 | return -ENOMEDIUM; | |
2673 | } | |
2674 | if (drv->bdrv_get_allocated_file_size) { | |
2675 | return drv->bdrv_get_allocated_file_size(bs); | |
2676 | } | |
2677 | if (bs->file) { | |
2678 | return bdrv_get_allocated_file_size(bs->file); | |
2679 | } | |
2680 | return -ENOTSUP; | |
2681 | } | |
2682 | ||
83f64091 FB |
2683 | /** |
2684 | * Length of a file in bytes. Return < 0 if error or unknown. | |
2685 | */ | |
2686 | int64_t bdrv_getlength(BlockDriverState *bs) | |
2687 | { | |
2688 | BlockDriver *drv = bs->drv; | |
2689 | if (!drv) | |
19cb3738 | 2690 | return -ENOMEDIUM; |
51762288 | 2691 | |
2c6942fa | 2692 | if (bs->growable || bdrv_dev_has_removable_media(bs)) { |
46a4e4e6 SH |
2693 | if (drv->bdrv_getlength) { |
2694 | return drv->bdrv_getlength(bs); | |
2695 | } | |
83f64091 | 2696 | } |
46a4e4e6 | 2697 | return bs->total_sectors * BDRV_SECTOR_SIZE; |
fc01f7e7 FB |
2698 | } |
2699 | ||
19cb3738 | 2700 | /* return 0 as number of sectors if no device present or error */ |
96b8f136 | 2701 | void bdrv_get_geometry(BlockDriverState *bs, uint64_t *nb_sectors_ptr) |
fc01f7e7 | 2702 | { |
19cb3738 FB |
2703 | int64_t length; |
2704 | length = bdrv_getlength(bs); | |
2705 | if (length < 0) | |
2706 | length = 0; | |
2707 | else | |
6ea44308 | 2708 | length = length >> BDRV_SECTOR_BITS; |
19cb3738 | 2709 | *nb_sectors_ptr = length; |
fc01f7e7 | 2710 | } |
cf98951b | 2711 | |
0563e191 ZYW |
2712 | /* throttling disk io limits */ |
2713 | void bdrv_set_io_limits(BlockDriverState *bs, | |
2714 | BlockIOLimit *io_limits) | |
2715 | { | |
2716 | bs->io_limits = *io_limits; | |
2717 | bs->io_limits_enabled = bdrv_io_limits_enabled(bs); | |
2718 | } | |
2719 | ||
ff06f5f3 PB |
2720 | void bdrv_set_on_error(BlockDriverState *bs, BlockdevOnError on_read_error, |
2721 | BlockdevOnError on_write_error) | |
abd7f68d MA |
2722 | { |
2723 | bs->on_read_error = on_read_error; | |
2724 | bs->on_write_error = on_write_error; | |
2725 | } | |
2726 | ||
1ceee0d5 | 2727 | BlockdevOnError bdrv_get_on_error(BlockDriverState *bs, bool is_read) |
abd7f68d MA |
2728 | { |
2729 | return is_read ? bs->on_read_error : bs->on_write_error; | |
2730 | } | |
2731 | ||
3e1caa5f PB |
2732 | BlockErrorAction bdrv_get_error_action(BlockDriverState *bs, bool is_read, int error) |
2733 | { | |
2734 | BlockdevOnError on_err = is_read ? bs->on_read_error : bs->on_write_error; | |
2735 | ||
2736 | switch (on_err) { | |
2737 | case BLOCKDEV_ON_ERROR_ENOSPC: | |
2738 | return (error == ENOSPC) ? BDRV_ACTION_STOP : BDRV_ACTION_REPORT; | |
2739 | case BLOCKDEV_ON_ERROR_STOP: | |
2740 | return BDRV_ACTION_STOP; | |
2741 | case BLOCKDEV_ON_ERROR_REPORT: | |
2742 | return BDRV_ACTION_REPORT; | |
2743 | case BLOCKDEV_ON_ERROR_IGNORE: | |
2744 | return BDRV_ACTION_IGNORE; | |
2745 | default: | |
2746 | abort(); | |
2747 | } | |
2748 | } | |
2749 | ||
2750 | /* This is done by device models because, while the block layer knows | |
2751 | * about the error, it does not know whether an operation comes from | |
2752 | * the device or the block layer (from a job, for example). | |
2753 | */ | |
2754 | void bdrv_error_action(BlockDriverState *bs, BlockErrorAction action, | |
2755 | bool is_read, int error) | |
2756 | { | |
2757 | assert(error >= 0); | |
32c81a4a | 2758 | bdrv_emit_qmp_error_event(bs, QEVENT_BLOCK_IO_ERROR, action, is_read); |
3e1caa5f PB |
2759 | if (action == BDRV_ACTION_STOP) { |
2760 | vm_stop(RUN_STATE_IO_ERROR); | |
2761 | bdrv_iostatus_set_err(bs, error); | |
2762 | } | |
2763 | } | |
2764 | ||
b338082b FB |
2765 | int bdrv_is_read_only(BlockDriverState *bs) |
2766 | { | |
2767 | return bs->read_only; | |
2768 | } | |
2769 | ||
985a03b0 TS |
2770 | int bdrv_is_sg(BlockDriverState *bs) |
2771 | { | |
2772 | return bs->sg; | |
2773 | } | |
2774 | ||
e900a7b7 CH |
2775 | int bdrv_enable_write_cache(BlockDriverState *bs) |
2776 | { | |
2777 | return bs->enable_write_cache; | |
2778 | } | |
2779 | ||
425b0148 PB |
2780 | void bdrv_set_enable_write_cache(BlockDriverState *bs, bool wce) |
2781 | { | |
2782 | bs->enable_write_cache = wce; | |
55b110f2 JC |
2783 | |
2784 | /* so a reopen() will preserve wce */ | |
2785 | if (wce) { | |
2786 | bs->open_flags |= BDRV_O_CACHE_WB; | |
2787 | } else { | |
2788 | bs->open_flags &= ~BDRV_O_CACHE_WB; | |
2789 | } | |
425b0148 PB |
2790 | } |
2791 | ||
ea2384d3 FB |
2792 | int bdrv_is_encrypted(BlockDriverState *bs) |
2793 | { | |
2794 | if (bs->backing_hd && bs->backing_hd->encrypted) | |
2795 | return 1; | |
2796 | return bs->encrypted; | |
2797 | } | |
2798 | ||
c0f4ce77 AL |
2799 | int bdrv_key_required(BlockDriverState *bs) |
2800 | { | |
2801 | BlockDriverState *backing_hd = bs->backing_hd; | |
2802 | ||
2803 | if (backing_hd && backing_hd->encrypted && !backing_hd->valid_key) | |
2804 | return 1; | |
2805 | return (bs->encrypted && !bs->valid_key); | |
2806 | } | |
2807 | ||
ea2384d3 FB |
2808 | int bdrv_set_key(BlockDriverState *bs, const char *key) |
2809 | { | |
2810 | int ret; | |
2811 | if (bs->backing_hd && bs->backing_hd->encrypted) { | |
2812 | ret = bdrv_set_key(bs->backing_hd, key); | |
2813 | if (ret < 0) | |
2814 | return ret; | |
2815 | if (!bs->encrypted) | |
2816 | return 0; | |
2817 | } | |
fd04a2ae SH |
2818 | if (!bs->encrypted) { |
2819 | return -EINVAL; | |
2820 | } else if (!bs->drv || !bs->drv->bdrv_set_key) { | |
2821 | return -ENOMEDIUM; | |
2822 | } | |
c0f4ce77 | 2823 | ret = bs->drv->bdrv_set_key(bs, key); |
bb5fc20f AL |
2824 | if (ret < 0) { |
2825 | bs->valid_key = 0; | |
2826 | } else if (!bs->valid_key) { | |
2827 | bs->valid_key = 1; | |
2828 | /* call the change callback now, we skipped it on open */ | |
7d4b4ba5 | 2829 | bdrv_dev_change_media_cb(bs, true); |
bb5fc20f | 2830 | } |
c0f4ce77 | 2831 | return ret; |
ea2384d3 FB |
2832 | } |
2833 | ||
f8d6bba1 | 2834 | const char *bdrv_get_format_name(BlockDriverState *bs) |
ea2384d3 | 2835 | { |
f8d6bba1 | 2836 | return bs->drv ? bs->drv->format_name : NULL; |
ea2384d3 FB |
2837 | } |
2838 | ||
5fafdf24 | 2839 | void bdrv_iterate_format(void (*it)(void *opaque, const char *name), |
ea2384d3 FB |
2840 | void *opaque) |
2841 | { | |
2842 | BlockDriver *drv; | |
2843 | ||
8a22f02a | 2844 | QLIST_FOREACH(drv, &bdrv_drivers, list) { |
ea2384d3 FB |
2845 | it(opaque, drv->format_name); |
2846 | } | |
2847 | } | |
2848 | ||
b338082b FB |
2849 | BlockDriverState *bdrv_find(const char *name) |
2850 | { | |
2851 | BlockDriverState *bs; | |
2852 | ||
1b7bdbc1 SH |
2853 | QTAILQ_FOREACH(bs, &bdrv_states, list) { |
2854 | if (!strcmp(name, bs->device_name)) { | |
b338082b | 2855 | return bs; |
1b7bdbc1 | 2856 | } |
b338082b FB |
2857 | } |
2858 | return NULL; | |
2859 | } | |
2860 | ||
2f399b0a MA |
2861 | BlockDriverState *bdrv_next(BlockDriverState *bs) |
2862 | { | |
2863 | if (!bs) { | |
2864 | return QTAILQ_FIRST(&bdrv_states); | |
2865 | } | |
2866 | return QTAILQ_NEXT(bs, list); | |
2867 | } | |
2868 | ||
51de9760 | 2869 | void bdrv_iterate(void (*it)(void *opaque, BlockDriverState *bs), void *opaque) |
81d0912d FB |
2870 | { |
2871 | BlockDriverState *bs; | |
2872 | ||
1b7bdbc1 | 2873 | QTAILQ_FOREACH(bs, &bdrv_states, list) { |
51de9760 | 2874 | it(opaque, bs); |
81d0912d FB |
2875 | } |
2876 | } | |
2877 | ||
ea2384d3 FB |
2878 | const char *bdrv_get_device_name(BlockDriverState *bs) |
2879 | { | |
2880 | return bs->device_name; | |
2881 | } | |
2882 | ||
c8433287 MA |
2883 | int bdrv_get_flags(BlockDriverState *bs) |
2884 | { | |
2885 | return bs->open_flags; | |
2886 | } | |
2887 | ||
c6ca28d6 AL |
2888 | void bdrv_flush_all(void) |
2889 | { | |
2890 | BlockDriverState *bs; | |
2891 | ||
1b7bdbc1 | 2892 | QTAILQ_FOREACH(bs, &bdrv_states, list) { |
29cdb251 | 2893 | bdrv_flush(bs); |
1b7bdbc1 | 2894 | } |
c6ca28d6 AL |
2895 | } |
2896 | ||
f2feebbd KW |
2897 | int bdrv_has_zero_init(BlockDriverState *bs) |
2898 | { | |
2899 | assert(bs->drv); | |
2900 | ||
336c1c12 KW |
2901 | if (bs->drv->bdrv_has_zero_init) { |
2902 | return bs->drv->bdrv_has_zero_init(bs); | |
f2feebbd KW |
2903 | } |
2904 | ||
2905 | return 1; | |
2906 | } | |
2907 | ||
376ae3f1 SH |
2908 | typedef struct BdrvCoIsAllocatedData { |
2909 | BlockDriverState *bs; | |
b35b2bba | 2910 | BlockDriverState *base; |
376ae3f1 SH |
2911 | int64_t sector_num; |
2912 | int nb_sectors; | |
2913 | int *pnum; | |
2914 | int ret; | |
2915 | bool done; | |
2916 | } BdrvCoIsAllocatedData; | |
2917 | ||
f58c7b35 TS |
2918 | /* |
2919 | * Returns true iff the specified sector is present in the disk image. Drivers | |
2920 | * not implementing the functionality are assumed to not support backing files, | |
2921 | * hence all their sectors are reported as allocated. | |
2922 | * | |
bd9533e3 SH |
2923 | * If 'sector_num' is beyond the end of the disk image the return value is 0 |
2924 | * and 'pnum' is set to 0. | |
2925 | * | |
f58c7b35 TS |
2926 | * 'pnum' is set to the number of sectors (including and immediately following |
2927 | * the specified sector) that are known to be in the same | |
2928 | * allocated/unallocated state. | |
2929 | * | |
bd9533e3 SH |
2930 | * 'nb_sectors' is the max value 'pnum' should be set to. If nb_sectors goes |
2931 | * beyond the end of the disk image it will be clamped. | |
f58c7b35 | 2932 | */ |
060f51c9 SH |
2933 | int coroutine_fn bdrv_co_is_allocated(BlockDriverState *bs, int64_t sector_num, |
2934 | int nb_sectors, int *pnum) | |
f58c7b35 | 2935 | { |
bd9533e3 SH |
2936 | int64_t n; |
2937 | ||
2938 | if (sector_num >= bs->total_sectors) { | |
2939 | *pnum = 0; | |
2940 | return 0; | |
2941 | } | |
2942 | ||
2943 | n = bs->total_sectors - sector_num; | |
2944 | if (n < nb_sectors) { | |
2945 | nb_sectors = n; | |
2946 | } | |
2947 | ||
6aebab14 | 2948 | if (!bs->drv->bdrv_co_is_allocated) { |
bd9533e3 | 2949 | *pnum = nb_sectors; |
f58c7b35 TS |
2950 | return 1; |
2951 | } | |
6aebab14 | 2952 | |
060f51c9 SH |
2953 | return bs->drv->bdrv_co_is_allocated(bs, sector_num, nb_sectors, pnum); |
2954 | } | |
2955 | ||
2956 | /* Coroutine wrapper for bdrv_is_allocated() */ | |
2957 | static void coroutine_fn bdrv_is_allocated_co_entry(void *opaque) | |
2958 | { | |
2959 | BdrvCoIsAllocatedData *data = opaque; | |
2960 | BlockDriverState *bs = data->bs; | |
2961 | ||
2962 | data->ret = bdrv_co_is_allocated(bs, data->sector_num, data->nb_sectors, | |
2963 | data->pnum); | |
2964 | data->done = true; | |
2965 | } | |
2966 | ||
2967 | /* | |
2968 | * Synchronous wrapper around bdrv_co_is_allocated(). | |
2969 | * | |
2970 | * See bdrv_co_is_allocated() for details. | |
2971 | */ | |
2972 | int bdrv_is_allocated(BlockDriverState *bs, int64_t sector_num, int nb_sectors, | |
2973 | int *pnum) | |
2974 | { | |
6aebab14 SH |
2975 | Coroutine *co; |
2976 | BdrvCoIsAllocatedData data = { | |
2977 | .bs = bs, | |
2978 | .sector_num = sector_num, | |
2979 | .nb_sectors = nb_sectors, | |
2980 | .pnum = pnum, | |
2981 | .done = false, | |
2982 | }; | |
2983 | ||
2984 | co = qemu_coroutine_create(bdrv_is_allocated_co_entry); | |
2985 | qemu_coroutine_enter(co, &data); | |
2986 | while (!data.done) { | |
2987 | qemu_aio_wait(); | |
2988 | } | |
2989 | return data.ret; | |
f58c7b35 TS |
2990 | } |
2991 | ||
188a7bbf PB |
2992 | /* |
2993 | * Given an image chain: ... -> [BASE] -> [INTER1] -> [INTER2] -> [TOP] | |
2994 | * | |
2995 | * Return true if the given sector is allocated in any image between | |
2996 | * BASE and TOP (inclusive). BASE can be NULL to check if the given | |
2997 | * sector is allocated in any image of the chain. Return false otherwise. | |
2998 | * | |
2999 | * 'pnum' is set to the number of sectors (including and immediately following | |
3000 | * the specified sector) that are known to be in the same | |
3001 | * allocated/unallocated state. | |
3002 | * | |
3003 | */ | |
3004 | int coroutine_fn bdrv_co_is_allocated_above(BlockDriverState *top, | |
3005 | BlockDriverState *base, | |
3006 | int64_t sector_num, | |
3007 | int nb_sectors, int *pnum) | |
3008 | { | |
3009 | BlockDriverState *intermediate; | |
3010 | int ret, n = nb_sectors; | |
3011 | ||
3012 | intermediate = top; | |
3013 | while (intermediate && intermediate != base) { | |
3014 | int pnum_inter; | |
3015 | ret = bdrv_co_is_allocated(intermediate, sector_num, nb_sectors, | |
3016 | &pnum_inter); | |
3017 | if (ret < 0) { | |
3018 | return ret; | |
3019 | } else if (ret) { | |
3020 | *pnum = pnum_inter; | |
3021 | return 1; | |
3022 | } | |
3023 | ||
3024 | /* | |
3025 | * [sector_num, nb_sectors] is unallocated on top but intermediate | |
3026 | * might have | |
3027 | * | |
3028 | * [sector_num+x, nr_sectors] allocated. | |
3029 | */ | |
63ba17d3 VI |
3030 | if (n > pnum_inter && |
3031 | (intermediate == top || | |
3032 | sector_num + pnum_inter < intermediate->total_sectors)) { | |
188a7bbf PB |
3033 | n = pnum_inter; |
3034 | } | |
3035 | ||
3036 | intermediate = intermediate->backing_hd; | |
3037 | } | |
3038 | ||
3039 | *pnum = n; | |
3040 | return 0; | |
3041 | } | |
3042 | ||
b35b2bba MR |
3043 | /* Coroutine wrapper for bdrv_is_allocated_above() */ |
3044 | static void coroutine_fn bdrv_is_allocated_above_co_entry(void *opaque) | |
3045 | { | |
3046 | BdrvCoIsAllocatedData *data = opaque; | |
3047 | BlockDriverState *top = data->bs; | |
3048 | BlockDriverState *base = data->base; | |
3049 | ||
3050 | data->ret = bdrv_co_is_allocated_above(top, base, data->sector_num, | |
3051 | data->nb_sectors, data->pnum); | |
3052 | data->done = true; | |
3053 | } | |
3054 | ||
3055 | /* | |
3056 | * Synchronous wrapper around bdrv_co_is_allocated_above(). | |
3057 | * | |
3058 | * See bdrv_co_is_allocated_above() for details. | |
3059 | */ | |
3060 | int bdrv_is_allocated_above(BlockDriverState *top, BlockDriverState *base, | |
3061 | int64_t sector_num, int nb_sectors, int *pnum) | |
3062 | { | |
3063 | Coroutine *co; | |
3064 | BdrvCoIsAllocatedData data = { | |
3065 | .bs = top, | |
3066 | .base = base, | |
3067 | .sector_num = sector_num, | |
3068 | .nb_sectors = nb_sectors, | |
3069 | .pnum = pnum, | |
3070 | .done = false, | |
3071 | }; | |
3072 | ||
3073 | co = qemu_coroutine_create(bdrv_is_allocated_above_co_entry); | |
3074 | qemu_coroutine_enter(co, &data); | |
3075 | while (!data.done) { | |
3076 | qemu_aio_wait(); | |
3077 | } | |
3078 | return data.ret; | |
3079 | } | |
3080 | ||
ac84adac | 3081 | BlockInfo *bdrv_query_info(BlockDriverState *bs) |
b338082b | 3082 | { |
ac84adac PB |
3083 | BlockInfo *info = g_malloc0(sizeof(*info)); |
3084 | info->device = g_strdup(bs->device_name); | |
3085 | info->type = g_strdup("unknown"); | |
3086 | info->locked = bdrv_dev_is_medium_locked(bs); | |
3087 | info->removable = bdrv_dev_has_removable_media(bs); | |
b338082b | 3088 | |
ac84adac PB |
3089 | if (bdrv_dev_has_removable_media(bs)) { |
3090 | info->has_tray_open = true; | |
3091 | info->tray_open = bdrv_dev_is_tray_open(bs); | |
3092 | } | |
d15e5465 | 3093 | |
ac84adac PB |
3094 | if (bdrv_iostatus_is_enabled(bs)) { |
3095 | info->has_io_status = true; | |
3096 | info->io_status = bs->iostatus; | |
3097 | } | |
d15e5465 | 3098 | |
b9a9b3a4 PB |
3099 | if (bs->dirty_bitmap) { |
3100 | info->has_dirty = true; | |
3101 | info->dirty = g_malloc0(sizeof(*info->dirty)); | |
acc906c6 | 3102 | info->dirty->count = bdrv_get_dirty_count(bs) * BDRV_SECTOR_SIZE; |
50717e94 PB |
3103 | info->dirty->granularity = |
3104 | ((int64_t) BDRV_SECTOR_SIZE << hbitmap_granularity(bs->dirty_bitmap)); | |
b9a9b3a4 | 3105 | } |
d15e5465 | 3106 | |
ac84adac PB |
3107 | if (bs->drv) { |
3108 | info->has_inserted = true; | |
3109 | info->inserted = g_malloc0(sizeof(*info->inserted)); | |
3110 | info->inserted->file = g_strdup(bs->filename); | |
3111 | info->inserted->ro = bs->read_only; | |
3112 | info->inserted->drv = g_strdup(bs->drv->format_name); | |
3113 | info->inserted->encrypted = bs->encrypted; | |
3114 | info->inserted->encryption_key_missing = bdrv_key_required(bs); | |
3115 | ||
3116 | if (bs->backing_file[0]) { | |
3117 | info->inserted->has_backing_file = true; | |
3118 | info->inserted->backing_file = g_strdup(bs->backing_file); | |
e4def80b | 3119 | } |
f04ef601 | 3120 | |
ac84adac PB |
3121 | info->inserted->backing_file_depth = bdrv_get_backing_file_depth(bs); |
3122 | ||
3123 | if (bs->io_limits_enabled) { | |
3124 | info->inserted->bps = | |
3125 | bs->io_limits.bps[BLOCK_IO_LIMIT_TOTAL]; | |
3126 | info->inserted->bps_rd = | |
3127 | bs->io_limits.bps[BLOCK_IO_LIMIT_READ]; | |
3128 | info->inserted->bps_wr = | |
3129 | bs->io_limits.bps[BLOCK_IO_LIMIT_WRITE]; | |
3130 | info->inserted->iops = | |
3131 | bs->io_limits.iops[BLOCK_IO_LIMIT_TOTAL]; | |
3132 | info->inserted->iops_rd = | |
3133 | bs->io_limits.iops[BLOCK_IO_LIMIT_READ]; | |
3134 | info->inserted->iops_wr = | |
3135 | bs->io_limits.iops[BLOCK_IO_LIMIT_WRITE]; | |
f04ef601 | 3136 | } |
ac84adac PB |
3137 | } |
3138 | return info; | |
3139 | } | |
f04ef601 | 3140 | |
ac84adac PB |
3141 | BlockInfoList *qmp_query_block(Error **errp) |
3142 | { | |
3143 | BlockInfoList *head = NULL, **p_next = &head; | |
3144 | BlockDriverState *bs; | |
727f005e | 3145 | |
ac84adac PB |
3146 | QTAILQ_FOREACH(bs, &bdrv_states, list) { |
3147 | BlockInfoList *info = g_malloc0(sizeof(*info)); | |
3148 | info->value = bdrv_query_info(bs); | |
d15e5465 | 3149 | |
ac84adac PB |
3150 | *p_next = info; |
3151 | p_next = &info->next; | |
b338082b | 3152 | } |
d15e5465 | 3153 | |
b2023818 | 3154 | return head; |
b338082b | 3155 | } |
a36e69dd | 3156 | |
9887b616 | 3157 | BlockStats *bdrv_query_stats(const BlockDriverState *bs) |
f11f57e4 LC |
3158 | { |
3159 | BlockStats *s; | |
3160 | ||
3161 | s = g_malloc0(sizeof(*s)); | |
3162 | ||
3163 | if (bs->device_name[0]) { | |
3164 | s->has_device = true; | |
3165 | s->device = g_strdup(bs->device_name); | |
294cc35f KW |
3166 | } |
3167 | ||
f11f57e4 LC |
3168 | s->stats = g_malloc0(sizeof(*s->stats)); |
3169 | s->stats->rd_bytes = bs->nr_bytes[BDRV_ACCT_READ]; | |
3170 | s->stats->wr_bytes = bs->nr_bytes[BDRV_ACCT_WRITE]; | |
3171 | s->stats->rd_operations = bs->nr_ops[BDRV_ACCT_READ]; | |
3172 | s->stats->wr_operations = bs->nr_ops[BDRV_ACCT_WRITE]; | |
3173 | s->stats->wr_highest_offset = bs->wr_highest_sector * BDRV_SECTOR_SIZE; | |
3174 | s->stats->flush_operations = bs->nr_ops[BDRV_ACCT_FLUSH]; | |
3175 | s->stats->wr_total_time_ns = bs->total_time_ns[BDRV_ACCT_WRITE]; | |
3176 | s->stats->rd_total_time_ns = bs->total_time_ns[BDRV_ACCT_READ]; | |
3177 | s->stats->flush_total_time_ns = bs->total_time_ns[BDRV_ACCT_FLUSH]; | |
3178 | ||
294cc35f | 3179 | if (bs->file) { |
f11f57e4 | 3180 | s->has_parent = true; |
9887b616 | 3181 | s->parent = bdrv_query_stats(bs->file); |
294cc35f KW |
3182 | } |
3183 | ||
f11f57e4 | 3184 | return s; |
294cc35f KW |
3185 | } |
3186 | ||
f11f57e4 | 3187 | BlockStatsList *qmp_query_blockstats(Error **errp) |
218a536a | 3188 | { |
9887b616 | 3189 | BlockStatsList *head = NULL, **p_next = &head; |
a36e69dd TS |
3190 | BlockDriverState *bs; |
3191 | ||
1b7bdbc1 | 3192 | QTAILQ_FOREACH(bs, &bdrv_states, list) { |
f11f57e4 | 3193 | BlockStatsList *info = g_malloc0(sizeof(*info)); |
9887b616 | 3194 | info->value = bdrv_query_stats(bs); |
f11f57e4 | 3195 | |
9887b616 PB |
3196 | *p_next = info; |
3197 | p_next = &info->next; | |
a36e69dd | 3198 | } |
218a536a | 3199 | |
f11f57e4 | 3200 | return head; |
a36e69dd | 3201 | } |
ea2384d3 | 3202 | |
045df330 AL |
3203 | const char *bdrv_get_encrypted_filename(BlockDriverState *bs) |
3204 | { | |
3205 | if (bs->backing_hd && bs->backing_hd->encrypted) | |
3206 | return bs->backing_file; | |
3207 | else if (bs->encrypted) | |
3208 | return bs->filename; | |
3209 | else | |
3210 | return NULL; | |
3211 | } | |
3212 | ||
5fafdf24 | 3213 | void bdrv_get_backing_filename(BlockDriverState *bs, |
83f64091 FB |
3214 | char *filename, int filename_size) |
3215 | { | |
3574c608 | 3216 | pstrcpy(filename, filename_size, bs->backing_file); |
83f64091 FB |
3217 | } |
3218 | ||
5fafdf24 | 3219 | int bdrv_write_compressed(BlockDriverState *bs, int64_t sector_num, |
faea38e7 FB |
3220 | const uint8_t *buf, int nb_sectors) |
3221 | { | |
3222 | BlockDriver *drv = bs->drv; | |
3223 | if (!drv) | |
19cb3738 | 3224 | return -ENOMEDIUM; |
faea38e7 FB |
3225 | if (!drv->bdrv_write_compressed) |
3226 | return -ENOTSUP; | |
fbb7b4e0 KW |
3227 | if (bdrv_check_request(bs, sector_num, nb_sectors)) |
3228 | return -EIO; | |
a55eb92c | 3229 | |
1755da16 | 3230 | assert(!bs->dirty_bitmap); |
a55eb92c | 3231 | |
faea38e7 FB |
3232 | return drv->bdrv_write_compressed(bs, sector_num, buf, nb_sectors); |
3233 | } | |
3b46e624 | 3234 | |
faea38e7 FB |
3235 | int bdrv_get_info(BlockDriverState *bs, BlockDriverInfo *bdi) |
3236 | { | |
3237 | BlockDriver *drv = bs->drv; | |
3238 | if (!drv) | |
19cb3738 | 3239 | return -ENOMEDIUM; |
faea38e7 FB |
3240 | if (!drv->bdrv_get_info) |
3241 | return -ENOTSUP; | |
3242 | memset(bdi, 0, sizeof(*bdi)); | |
3243 | return drv->bdrv_get_info(bs, bdi); | |
3244 | } | |
3245 | ||
45566e9c CH |
3246 | int bdrv_save_vmstate(BlockDriverState *bs, const uint8_t *buf, |
3247 | int64_t pos, int size) | |
cf8074b3 KW |
3248 | { |
3249 | QEMUIOVector qiov; | |
3250 | struct iovec iov = { | |
3251 | .iov_base = (void *) buf, | |
3252 | .iov_len = size, | |
3253 | }; | |
3254 | ||
3255 | qemu_iovec_init_external(&qiov, &iov, 1); | |
3256 | return bdrv_writev_vmstate(bs, &qiov, pos); | |
3257 | } | |
3258 | ||
3259 | int bdrv_writev_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos) | |
178e08a5 AL |
3260 | { |
3261 | BlockDriver *drv = bs->drv; | |
cf8074b3 KW |
3262 | |
3263 | if (!drv) { | |
178e08a5 | 3264 | return -ENOMEDIUM; |
cf8074b3 KW |
3265 | } else if (drv->bdrv_save_vmstate) { |
3266 | return drv->bdrv_save_vmstate(bs, qiov, pos); | |
3267 | } else if (bs->file) { | |
3268 | return bdrv_writev_vmstate(bs->file, qiov, pos); | |
3269 | } | |
3270 | ||
7cdb1f6d | 3271 | return -ENOTSUP; |
178e08a5 AL |
3272 | } |
3273 | ||
45566e9c CH |
3274 | int bdrv_load_vmstate(BlockDriverState *bs, uint8_t *buf, |
3275 | int64_t pos, int size) | |
178e08a5 AL |
3276 | { |
3277 | BlockDriver *drv = bs->drv; | |
3278 | if (!drv) | |
3279 | return -ENOMEDIUM; | |
7cdb1f6d MK |
3280 | if (drv->bdrv_load_vmstate) |
3281 | return drv->bdrv_load_vmstate(bs, buf, pos, size); | |
3282 | if (bs->file) | |
3283 | return bdrv_load_vmstate(bs->file, buf, pos, size); | |
3284 | return -ENOTSUP; | |
178e08a5 AL |
3285 | } |
3286 | ||
8b9b0cc2 KW |
3287 | void bdrv_debug_event(BlockDriverState *bs, BlkDebugEvent event) |
3288 | { | |
3289 | BlockDriver *drv = bs->drv; | |
3290 | ||
3291 | if (!drv || !drv->bdrv_debug_event) { | |
3292 | return; | |
3293 | } | |
3294 | ||
0ed8b6f6 | 3295 | drv->bdrv_debug_event(bs, event); |
41c695c7 KW |
3296 | } |
3297 | ||
3298 | int bdrv_debug_breakpoint(BlockDriverState *bs, const char *event, | |
3299 | const char *tag) | |
3300 | { | |
3301 | while (bs && bs->drv && !bs->drv->bdrv_debug_breakpoint) { | |
3302 | bs = bs->file; | |
3303 | } | |
3304 | ||
3305 | if (bs && bs->drv && bs->drv->bdrv_debug_breakpoint) { | |
3306 | return bs->drv->bdrv_debug_breakpoint(bs, event, tag); | |
3307 | } | |
3308 | ||
3309 | return -ENOTSUP; | |
3310 | } | |
3311 | ||
3312 | int bdrv_debug_resume(BlockDriverState *bs, const char *tag) | |
3313 | { | |
3314 | while (bs && bs->drv && !bs->drv->bdrv_debug_resume) { | |
3315 | bs = bs->file; | |
3316 | } | |
8b9b0cc2 | 3317 | |
41c695c7 KW |
3318 | if (bs && bs->drv && bs->drv->bdrv_debug_resume) { |
3319 | return bs->drv->bdrv_debug_resume(bs, tag); | |
3320 | } | |
3321 | ||
3322 | return -ENOTSUP; | |
3323 | } | |
3324 | ||
3325 | bool bdrv_debug_is_suspended(BlockDriverState *bs, const char *tag) | |
3326 | { | |
3327 | while (bs && bs->drv && !bs->drv->bdrv_debug_is_suspended) { | |
3328 | bs = bs->file; | |
3329 | } | |
3330 | ||
3331 | if (bs && bs->drv && bs->drv->bdrv_debug_is_suspended) { | |
3332 | return bs->drv->bdrv_debug_is_suspended(bs, tag); | |
3333 | } | |
3334 | ||
3335 | return false; | |
8b9b0cc2 KW |
3336 | } |
3337 | ||
faea38e7 FB |
3338 | /**************************************************************/ |
3339 | /* handling of snapshots */ | |
3340 | ||
feeee5ac MDCF |
3341 | int bdrv_can_snapshot(BlockDriverState *bs) |
3342 | { | |
3343 | BlockDriver *drv = bs->drv; | |
07b70bfb | 3344 | if (!drv || !bdrv_is_inserted(bs) || bdrv_is_read_only(bs)) { |
feeee5ac MDCF |
3345 | return 0; |
3346 | } | |
3347 | ||
3348 | if (!drv->bdrv_snapshot_create) { | |
3349 | if (bs->file != NULL) { | |
3350 | return bdrv_can_snapshot(bs->file); | |
3351 | } | |
3352 | return 0; | |
3353 | } | |
3354 | ||
3355 | return 1; | |
3356 | } | |
3357 | ||
199630b6 BS |
3358 | int bdrv_is_snapshot(BlockDriverState *bs) |
3359 | { | |
3360 | return !!(bs->open_flags & BDRV_O_SNAPSHOT); | |
3361 | } | |
3362 | ||
f9092b10 MA |
3363 | BlockDriverState *bdrv_snapshots(void) |
3364 | { | |
3365 | BlockDriverState *bs; | |
3366 | ||
3ac906f7 | 3367 | if (bs_snapshots) { |
f9092b10 | 3368 | return bs_snapshots; |
3ac906f7 | 3369 | } |
f9092b10 MA |
3370 | |
3371 | bs = NULL; | |
3372 | while ((bs = bdrv_next(bs))) { | |
3373 | if (bdrv_can_snapshot(bs)) { | |
3ac906f7 MA |
3374 | bs_snapshots = bs; |
3375 | return bs; | |
f9092b10 MA |
3376 | } |
3377 | } | |
3378 | return NULL; | |
f9092b10 MA |
3379 | } |
3380 | ||
5fafdf24 | 3381 | int bdrv_snapshot_create(BlockDriverState *bs, |
faea38e7 FB |
3382 | QEMUSnapshotInfo *sn_info) |
3383 | { | |
3384 | BlockDriver *drv = bs->drv; | |
3385 | if (!drv) | |
19cb3738 | 3386 | return -ENOMEDIUM; |
7cdb1f6d MK |
3387 | if (drv->bdrv_snapshot_create) |
3388 | return drv->bdrv_snapshot_create(bs, sn_info); | |
3389 | if (bs->file) | |
3390 | return bdrv_snapshot_create(bs->file, sn_info); | |
3391 | return -ENOTSUP; | |
faea38e7 FB |
3392 | } |
3393 | ||
5fafdf24 | 3394 | int bdrv_snapshot_goto(BlockDriverState *bs, |
faea38e7 FB |
3395 | const char *snapshot_id) |
3396 | { | |
3397 | BlockDriver *drv = bs->drv; | |
7cdb1f6d MK |
3398 | int ret, open_ret; |
3399 | ||
faea38e7 | 3400 | if (!drv) |
19cb3738 | 3401 | return -ENOMEDIUM; |
7cdb1f6d MK |
3402 | if (drv->bdrv_snapshot_goto) |
3403 | return drv->bdrv_snapshot_goto(bs, snapshot_id); | |
3404 | ||
3405 | if (bs->file) { | |
3406 | drv->bdrv_close(bs); | |
3407 | ret = bdrv_snapshot_goto(bs->file, snapshot_id); | |
1a86938f | 3408 | open_ret = drv->bdrv_open(bs, NULL, bs->open_flags); |
7cdb1f6d MK |
3409 | if (open_ret < 0) { |
3410 | bdrv_delete(bs->file); | |
3411 | bs->drv = NULL; | |
3412 | return open_ret; | |
3413 | } | |
3414 | return ret; | |
3415 | } | |
3416 | ||
3417 | return -ENOTSUP; | |
faea38e7 FB |
3418 | } |
3419 | ||
3420 | int bdrv_snapshot_delete(BlockDriverState *bs, const char *snapshot_id) | |
3421 | { | |
3422 | BlockDriver *drv = bs->drv; | |
3423 | if (!drv) | |
19cb3738 | 3424 | return -ENOMEDIUM; |
7cdb1f6d MK |
3425 | if (drv->bdrv_snapshot_delete) |
3426 | return drv->bdrv_snapshot_delete(bs, snapshot_id); | |
3427 | if (bs->file) | |
3428 | return bdrv_snapshot_delete(bs->file, snapshot_id); | |
3429 | return -ENOTSUP; | |
faea38e7 FB |
3430 | } |
3431 | ||
5fafdf24 | 3432 | int bdrv_snapshot_list(BlockDriverState *bs, |
faea38e7 FB |
3433 | QEMUSnapshotInfo **psn_info) |
3434 | { | |
3435 | BlockDriver *drv = bs->drv; | |
3436 | if (!drv) | |
19cb3738 | 3437 | return -ENOMEDIUM; |
7cdb1f6d MK |
3438 | if (drv->bdrv_snapshot_list) |
3439 | return drv->bdrv_snapshot_list(bs, psn_info); | |
3440 | if (bs->file) | |
3441 | return bdrv_snapshot_list(bs->file, psn_info); | |
3442 | return -ENOTSUP; | |
faea38e7 FB |
3443 | } |
3444 | ||
51ef6727 | 3445 | int bdrv_snapshot_load_tmp(BlockDriverState *bs, |
3446 | const char *snapshot_name) | |
3447 | { | |
3448 | BlockDriver *drv = bs->drv; | |
3449 | if (!drv) { | |
3450 | return -ENOMEDIUM; | |
3451 | } | |
3452 | if (!bs->read_only) { | |
3453 | return -EINVAL; | |
3454 | } | |
3455 | if (drv->bdrv_snapshot_load_tmp) { | |
3456 | return drv->bdrv_snapshot_load_tmp(bs, snapshot_name); | |
3457 | } | |
3458 | return -ENOTSUP; | |
3459 | } | |
3460 | ||
b1b1d783 JC |
3461 | /* backing_file can either be relative, or absolute, or a protocol. If it is |
3462 | * relative, it must be relative to the chain. So, passing in bs->filename | |
3463 | * from a BDS as backing_file should not be done, as that may be relative to | |
3464 | * the CWD rather than the chain. */ | |
e8a6bb9c MT |
3465 | BlockDriverState *bdrv_find_backing_image(BlockDriverState *bs, |
3466 | const char *backing_file) | |
3467 | { | |
b1b1d783 JC |
3468 | char *filename_full = NULL; |
3469 | char *backing_file_full = NULL; | |
3470 | char *filename_tmp = NULL; | |
3471 | int is_protocol = 0; | |
3472 | BlockDriverState *curr_bs = NULL; | |
3473 | BlockDriverState *retval = NULL; | |
3474 | ||
3475 | if (!bs || !bs->drv || !backing_file) { | |
e8a6bb9c MT |
3476 | return NULL; |
3477 | } | |
3478 | ||
b1b1d783 JC |
3479 | filename_full = g_malloc(PATH_MAX); |
3480 | backing_file_full = g_malloc(PATH_MAX); | |
3481 | filename_tmp = g_malloc(PATH_MAX); | |
3482 | ||
3483 | is_protocol = path_has_protocol(backing_file); | |
3484 | ||
3485 | for (curr_bs = bs; curr_bs->backing_hd; curr_bs = curr_bs->backing_hd) { | |
3486 | ||
3487 | /* If either of the filename paths is actually a protocol, then | |
3488 | * compare unmodified paths; otherwise make paths relative */ | |
3489 | if (is_protocol || path_has_protocol(curr_bs->backing_file)) { | |
3490 | if (strcmp(backing_file, curr_bs->backing_file) == 0) { | |
3491 | retval = curr_bs->backing_hd; | |
3492 | break; | |
3493 | } | |
e8a6bb9c | 3494 | } else { |
b1b1d783 JC |
3495 | /* If not an absolute filename path, make it relative to the current |
3496 | * image's filename path */ | |
3497 | path_combine(filename_tmp, PATH_MAX, curr_bs->filename, | |
3498 | backing_file); | |
3499 | ||
3500 | /* We are going to compare absolute pathnames */ | |
3501 | if (!realpath(filename_tmp, filename_full)) { | |
3502 | continue; | |
3503 | } | |
3504 | ||
3505 | /* We need to make sure the backing filename we are comparing against | |
3506 | * is relative to the current image filename (or absolute) */ | |
3507 | path_combine(filename_tmp, PATH_MAX, curr_bs->filename, | |
3508 | curr_bs->backing_file); | |
3509 | ||
3510 | if (!realpath(filename_tmp, backing_file_full)) { | |
3511 | continue; | |
3512 | } | |
3513 | ||
3514 | if (strcmp(backing_file_full, filename_full) == 0) { | |
3515 | retval = curr_bs->backing_hd; | |
3516 | break; | |
3517 | } | |
e8a6bb9c MT |
3518 | } |
3519 | } | |
3520 | ||
b1b1d783 JC |
3521 | g_free(filename_full); |
3522 | g_free(backing_file_full); | |
3523 | g_free(filename_tmp); | |
3524 | return retval; | |
e8a6bb9c MT |
3525 | } |
3526 | ||
f198fd1c BC |
3527 | int bdrv_get_backing_file_depth(BlockDriverState *bs) |
3528 | { | |
3529 | if (!bs->drv) { | |
3530 | return 0; | |
3531 | } | |
3532 | ||
3533 | if (!bs->backing_hd) { | |
3534 | return 0; | |
3535 | } | |
3536 | ||
3537 | return 1 + bdrv_get_backing_file_depth(bs->backing_hd); | |
3538 | } | |
3539 | ||
79fac568 JC |
3540 | BlockDriverState *bdrv_find_base(BlockDriverState *bs) |
3541 | { | |
3542 | BlockDriverState *curr_bs = NULL; | |
3543 | ||
3544 | if (!bs) { | |
3545 | return NULL; | |
3546 | } | |
3547 | ||
3548 | curr_bs = bs; | |
3549 | ||
3550 | while (curr_bs->backing_hd) { | |
3551 | curr_bs = curr_bs->backing_hd; | |
3552 | } | |
3553 | return curr_bs; | |
3554 | } | |
3555 | ||
faea38e7 FB |
3556 | #define NB_SUFFIXES 4 |
3557 | ||
3558 | char *get_human_readable_size(char *buf, int buf_size, int64_t size) | |
3559 | { | |
3560 | static const char suffixes[NB_SUFFIXES] = "KMGT"; | |
3561 | int64_t base; | |
3562 | int i; | |
3563 | ||
3564 | if (size <= 999) { | |
3565 | snprintf(buf, buf_size, "%" PRId64, size); | |
3566 | } else { | |
3567 | base = 1024; | |
3568 | for(i = 0; i < NB_SUFFIXES; i++) { | |
3569 | if (size < (10 * base)) { | |
5fafdf24 | 3570 | snprintf(buf, buf_size, "%0.1f%c", |
faea38e7 FB |
3571 | (double)size / base, |
3572 | suffixes[i]); | |
3573 | break; | |
3574 | } else if (size < (1000 * base) || i == (NB_SUFFIXES - 1)) { | |
5fafdf24 | 3575 | snprintf(buf, buf_size, "%" PRId64 "%c", |
faea38e7 FB |
3576 | ((size + (base >> 1)) / base), |
3577 | suffixes[i]); | |
3578 | break; | |
3579 | } | |
3580 | base = base * 1024; | |
3581 | } | |
3582 | } | |
3583 | return buf; | |
3584 | } | |
3585 | ||
3586 | char *bdrv_snapshot_dump(char *buf, int buf_size, QEMUSnapshotInfo *sn) | |
3587 | { | |
3588 | char buf1[128], date_buf[128], clock_buf[128]; | |
3589 | struct tm tm; | |
3590 | time_t ti; | |
3591 | int64_t secs; | |
3592 | ||
3593 | if (!sn) { | |
5fafdf24 TS |
3594 | snprintf(buf, buf_size, |
3595 | "%-10s%-20s%7s%20s%15s", | |
faea38e7 FB |
3596 | "ID", "TAG", "VM SIZE", "DATE", "VM CLOCK"); |
3597 | } else { | |
3598 | ti = sn->date_sec; | |
3599 | localtime_r(&ti, &tm); | |
3600 | strftime(date_buf, sizeof(date_buf), | |
3601 | "%Y-%m-%d %H:%M:%S", &tm); | |
3602 | secs = sn->vm_clock_nsec / 1000000000; | |
3603 | snprintf(clock_buf, sizeof(clock_buf), | |
3604 | "%02d:%02d:%02d.%03d", | |
3605 | (int)(secs / 3600), | |
3606 | (int)((secs / 60) % 60), | |
5fafdf24 | 3607 | (int)(secs % 60), |
faea38e7 FB |
3608 | (int)((sn->vm_clock_nsec / 1000000) % 1000)); |
3609 | snprintf(buf, buf_size, | |
5fafdf24 | 3610 | "%-10s%-20s%7s%20s%15s", |
faea38e7 FB |
3611 | sn->id_str, sn->name, |
3612 | get_human_readable_size(buf1, sizeof(buf1), sn->vm_state_size), | |
3613 | date_buf, | |
3614 | clock_buf); | |
3615 | } | |
3616 | return buf; | |
3617 | } | |
3618 | ||
ea2384d3 | 3619 | /**************************************************************/ |
83f64091 | 3620 | /* async I/Os */ |
ea2384d3 | 3621 | |
3b69e4b9 | 3622 | BlockDriverAIOCB *bdrv_aio_readv(BlockDriverState *bs, int64_t sector_num, |
f141eafe | 3623 | QEMUIOVector *qiov, int nb_sectors, |
3b69e4b9 | 3624 | BlockDriverCompletionFunc *cb, void *opaque) |
83f64091 | 3625 | { |
bbf0a440 SH |
3626 | trace_bdrv_aio_readv(bs, sector_num, nb_sectors, opaque); |
3627 | ||
b2a61371 | 3628 | return bdrv_co_aio_rw_vector(bs, sector_num, qiov, nb_sectors, |
8c5873d6 | 3629 | cb, opaque, false); |
ea2384d3 FB |
3630 | } |
3631 | ||
f141eafe AL |
3632 | BlockDriverAIOCB *bdrv_aio_writev(BlockDriverState *bs, int64_t sector_num, |
3633 | QEMUIOVector *qiov, int nb_sectors, | |
3634 | BlockDriverCompletionFunc *cb, void *opaque) | |
ea2384d3 | 3635 | { |
bbf0a440 SH |
3636 | trace_bdrv_aio_writev(bs, sector_num, nb_sectors, opaque); |
3637 | ||
1a6e115b | 3638 | return bdrv_co_aio_rw_vector(bs, sector_num, qiov, nb_sectors, |
8c5873d6 | 3639 | cb, opaque, true); |
83f64091 FB |
3640 | } |
3641 | ||
40b4f539 KW |
3642 | |
3643 | typedef struct MultiwriteCB { | |
3644 | int error; | |
3645 | int num_requests; | |
3646 | int num_callbacks; | |
3647 | struct { | |
3648 | BlockDriverCompletionFunc *cb; | |
3649 | void *opaque; | |
3650 | QEMUIOVector *free_qiov; | |
40b4f539 KW |
3651 | } callbacks[]; |
3652 | } MultiwriteCB; | |
3653 | ||
3654 | static void multiwrite_user_cb(MultiwriteCB *mcb) | |
3655 | { | |
3656 | int i; | |
3657 | ||
3658 | for (i = 0; i < mcb->num_callbacks; i++) { | |
3659 | mcb->callbacks[i].cb(mcb->callbacks[i].opaque, mcb->error); | |
1e1ea48d SH |
3660 | if (mcb->callbacks[i].free_qiov) { |
3661 | qemu_iovec_destroy(mcb->callbacks[i].free_qiov); | |
3662 | } | |
7267c094 | 3663 | g_free(mcb->callbacks[i].free_qiov); |
40b4f539 KW |
3664 | } |
3665 | } | |
3666 | ||
3667 | static void multiwrite_cb(void *opaque, int ret) | |
3668 | { | |
3669 | MultiwriteCB *mcb = opaque; | |
3670 | ||
6d519a5f SH |
3671 | trace_multiwrite_cb(mcb, ret); |
3672 | ||
cb6d3ca0 | 3673 | if (ret < 0 && !mcb->error) { |
40b4f539 | 3674 | mcb->error = ret; |
40b4f539 KW |
3675 | } |
3676 | ||
3677 | mcb->num_requests--; | |
3678 | if (mcb->num_requests == 0) { | |
de189a1b | 3679 | multiwrite_user_cb(mcb); |
7267c094 | 3680 | g_free(mcb); |
40b4f539 KW |
3681 | } |
3682 | } | |
3683 | ||
3684 | static int multiwrite_req_compare(const void *a, const void *b) | |
3685 | { | |
77be4366 CH |
3686 | const BlockRequest *req1 = a, *req2 = b; |
3687 | ||
3688 | /* | |
3689 | * Note that we can't simply subtract req2->sector from req1->sector | |
3690 | * here as that could overflow the return value. | |
3691 | */ | |
3692 | if (req1->sector > req2->sector) { | |
3693 | return 1; | |
3694 | } else if (req1->sector < req2->sector) { | |
3695 | return -1; | |
3696 | } else { | |
3697 | return 0; | |
3698 | } | |
40b4f539 KW |
3699 | } |
3700 | ||
3701 | /* | |
3702 | * Takes a bunch of requests and tries to merge them. Returns the number of | |
3703 | * requests that remain after merging. | |
3704 | */ | |
3705 | static int multiwrite_merge(BlockDriverState *bs, BlockRequest *reqs, | |
3706 | int num_reqs, MultiwriteCB *mcb) | |
3707 | { | |
3708 | int i, outidx; | |
3709 | ||
3710 | // Sort requests by start sector | |
3711 | qsort(reqs, num_reqs, sizeof(*reqs), &multiwrite_req_compare); | |
3712 | ||
3713 | // Check if adjacent requests touch the same clusters. If so, combine them, | |
3714 | // filling up gaps with zero sectors. | |
3715 | outidx = 0; | |
3716 | for (i = 1; i < num_reqs; i++) { | |
3717 | int merge = 0; | |
3718 | int64_t oldreq_last = reqs[outidx].sector + reqs[outidx].nb_sectors; | |
3719 | ||
b6a127a1 | 3720 | // Handle exactly sequential writes and overlapping writes. |
40b4f539 KW |
3721 | if (reqs[i].sector <= oldreq_last) { |
3722 | merge = 1; | |
3723 | } | |
3724 | ||
e2a305fb CH |
3725 | if (reqs[outidx].qiov->niov + reqs[i].qiov->niov + 1 > IOV_MAX) { |
3726 | merge = 0; | |
3727 | } | |
3728 | ||
40b4f539 KW |
3729 | if (merge) { |
3730 | size_t size; | |
7267c094 | 3731 | QEMUIOVector *qiov = g_malloc0(sizeof(*qiov)); |
40b4f539 KW |
3732 | qemu_iovec_init(qiov, |
3733 | reqs[outidx].qiov->niov + reqs[i].qiov->niov + 1); | |
3734 | ||
3735 | // Add the first request to the merged one. If the requests are | |
3736 | // overlapping, drop the last sectors of the first request. | |
3737 | size = (reqs[i].sector - reqs[outidx].sector) << 9; | |
1b093c48 | 3738 | qemu_iovec_concat(qiov, reqs[outidx].qiov, 0, size); |
40b4f539 | 3739 | |
b6a127a1 PB |
3740 | // We should need to add any zeros between the two requests |
3741 | assert (reqs[i].sector <= oldreq_last); | |
40b4f539 KW |
3742 | |
3743 | // Add the second request | |
1b093c48 | 3744 | qemu_iovec_concat(qiov, reqs[i].qiov, 0, reqs[i].qiov->size); |
40b4f539 | 3745 | |
cbf1dff2 | 3746 | reqs[outidx].nb_sectors = qiov->size >> 9; |
40b4f539 KW |
3747 | reqs[outidx].qiov = qiov; |
3748 | ||
3749 | mcb->callbacks[i].free_qiov = reqs[outidx].qiov; | |
3750 | } else { | |
3751 | outidx++; | |
3752 | reqs[outidx].sector = reqs[i].sector; | |
3753 | reqs[outidx].nb_sectors = reqs[i].nb_sectors; | |
3754 | reqs[outidx].qiov = reqs[i].qiov; | |
3755 | } | |
3756 | } | |
3757 | ||
3758 | return outidx + 1; | |
3759 | } | |
3760 | ||
3761 | /* | |
3762 | * Submit multiple AIO write requests at once. | |
3763 | * | |
3764 | * On success, the function returns 0 and all requests in the reqs array have | |
3765 | * been submitted. In error case this function returns -1, and any of the | |
3766 | * requests may or may not be submitted yet. In particular, this means that the | |
3767 | * callback will be called for some of the requests, for others it won't. The | |
3768 | * caller must check the error field of the BlockRequest to wait for the right | |
3769 | * callbacks (if error != 0, no callback will be called). | |
3770 | * | |
3771 | * The implementation may modify the contents of the reqs array, e.g. to merge | |
3772 | * requests. However, the fields opaque and error are left unmodified as they | |
3773 | * are used to signal failure for a single request to the caller. | |
3774 | */ | |
3775 | int bdrv_aio_multiwrite(BlockDriverState *bs, BlockRequest *reqs, int num_reqs) | |
3776 | { | |
40b4f539 KW |
3777 | MultiwriteCB *mcb; |
3778 | int i; | |
3779 | ||
301db7c2 RH |
3780 | /* don't submit writes if we don't have a medium */ |
3781 | if (bs->drv == NULL) { | |
3782 | for (i = 0; i < num_reqs; i++) { | |
3783 | reqs[i].error = -ENOMEDIUM; | |
3784 | } | |
3785 | return -1; | |
3786 | } | |
3787 | ||
40b4f539 KW |
3788 | if (num_reqs == 0) { |
3789 | return 0; | |
3790 | } | |
3791 | ||
3792 | // Create MultiwriteCB structure | |
7267c094 | 3793 | mcb = g_malloc0(sizeof(*mcb) + num_reqs * sizeof(*mcb->callbacks)); |
40b4f539 KW |
3794 | mcb->num_requests = 0; |
3795 | mcb->num_callbacks = num_reqs; | |
3796 | ||
3797 | for (i = 0; i < num_reqs; i++) { | |
3798 | mcb->callbacks[i].cb = reqs[i].cb; | |
3799 | mcb->callbacks[i].opaque = reqs[i].opaque; | |
3800 | } | |
3801 | ||
3802 | // Check for mergable requests | |
3803 | num_reqs = multiwrite_merge(bs, reqs, num_reqs, mcb); | |
3804 | ||
6d519a5f SH |
3805 | trace_bdrv_aio_multiwrite(mcb, mcb->num_callbacks, num_reqs); |
3806 | ||
df9309fb PB |
3807 | /* Run the aio requests. */ |
3808 | mcb->num_requests = num_reqs; | |
40b4f539 | 3809 | for (i = 0; i < num_reqs; i++) { |
ad54ae80 | 3810 | bdrv_aio_writev(bs, reqs[i].sector, reqs[i].qiov, |
40b4f539 | 3811 | reqs[i].nb_sectors, multiwrite_cb, mcb); |
40b4f539 KW |
3812 | } |
3813 | ||
3814 | return 0; | |
40b4f539 KW |
3815 | } |
3816 | ||
83f64091 | 3817 | void bdrv_aio_cancel(BlockDriverAIOCB *acb) |
83f64091 | 3818 | { |
d7331bed | 3819 | acb->aiocb_info->cancel(acb); |
83f64091 FB |
3820 | } |
3821 | ||
98f90dba ZYW |
3822 | /* block I/O throttling */ |
3823 | static bool bdrv_exceed_bps_limits(BlockDriverState *bs, int nb_sectors, | |
3824 | bool is_write, double elapsed_time, uint64_t *wait) | |
3825 | { | |
3826 | uint64_t bps_limit = 0; | |
ae29d6c6 | 3827 | uint64_t extension; |
98f90dba ZYW |
3828 | double bytes_limit, bytes_base, bytes_res; |
3829 | double slice_time, wait_time; | |
3830 | ||
3831 | if (bs->io_limits.bps[BLOCK_IO_LIMIT_TOTAL]) { | |
3832 | bps_limit = bs->io_limits.bps[BLOCK_IO_LIMIT_TOTAL]; | |
3833 | } else if (bs->io_limits.bps[is_write]) { | |
3834 | bps_limit = bs->io_limits.bps[is_write]; | |
3835 | } else { | |
3836 | if (wait) { | |
3837 | *wait = 0; | |
3838 | } | |
3839 | ||
3840 | return false; | |
3841 | } | |
3842 | ||
3843 | slice_time = bs->slice_end - bs->slice_start; | |
3844 | slice_time /= (NANOSECONDS_PER_SECOND); | |
3845 | bytes_limit = bps_limit * slice_time; | |
5905fbc9 | 3846 | bytes_base = bs->slice_submitted.bytes[is_write]; |
98f90dba | 3847 | if (bs->io_limits.bps[BLOCK_IO_LIMIT_TOTAL]) { |
5905fbc9 | 3848 | bytes_base += bs->slice_submitted.bytes[!is_write]; |
98f90dba ZYW |
3849 | } |
3850 | ||
3851 | /* bytes_base: the bytes of data which have been read/written; and | |
3852 | * it is obtained from the history statistic info. | |
3853 | * bytes_res: the remaining bytes of data which need to be read/written. | |
3854 | * (bytes_base + bytes_res) / bps_limit: used to calcuate | |
3855 | * the total time for completing reading/writting all data. | |
3856 | */ | |
3857 | bytes_res = (unsigned) nb_sectors * BDRV_SECTOR_SIZE; | |
3858 | ||
3859 | if (bytes_base + bytes_res <= bytes_limit) { | |
3860 | if (wait) { | |
3861 | *wait = 0; | |
3862 | } | |
3863 | ||
3864 | return false; | |
3865 | } | |
3866 | ||
3867 | /* Calc approx time to dispatch */ | |
3868 | wait_time = (bytes_base + bytes_res) / bps_limit - elapsed_time; | |
3869 | ||
3870 | /* When the I/O rate at runtime exceeds the limits, | |
3871 | * bs->slice_end need to be extended in order that the current statistic | |
3872 | * info can be kept until the timer fire, so it is increased and tuned | |
3873 | * based on the result of experiment. | |
3874 | */ | |
ae29d6c6 SH |
3875 | extension = wait_time * NANOSECONDS_PER_SECOND; |
3876 | extension = DIV_ROUND_UP(extension, BLOCK_IO_SLICE_TIME) * | |
3877 | BLOCK_IO_SLICE_TIME; | |
3878 | bs->slice_end += extension; | |
98f90dba | 3879 | if (wait) { |
0775437f | 3880 | *wait = wait_time * NANOSECONDS_PER_SECOND; |
98f90dba ZYW |
3881 | } |
3882 | ||
3883 | return true; | |
3884 | } | |
3885 | ||
3886 | static bool bdrv_exceed_iops_limits(BlockDriverState *bs, bool is_write, | |
3887 | double elapsed_time, uint64_t *wait) | |
3888 | { | |
3889 | uint64_t iops_limit = 0; | |
3890 | double ios_limit, ios_base; | |
3891 | double slice_time, wait_time; | |
3892 | ||
3893 | if (bs->io_limits.iops[BLOCK_IO_LIMIT_TOTAL]) { | |
3894 | iops_limit = bs->io_limits.iops[BLOCK_IO_LIMIT_TOTAL]; | |
3895 | } else if (bs->io_limits.iops[is_write]) { | |
3896 | iops_limit = bs->io_limits.iops[is_write]; | |
3897 | } else { | |
3898 | if (wait) { | |
3899 | *wait = 0; | |
3900 | } | |
3901 | ||
3902 | return false; | |
3903 | } | |
3904 | ||
3905 | slice_time = bs->slice_end - bs->slice_start; | |
3906 | slice_time /= (NANOSECONDS_PER_SECOND); | |
3907 | ios_limit = iops_limit * slice_time; | |
5905fbc9 | 3908 | ios_base = bs->slice_submitted.ios[is_write]; |
98f90dba | 3909 | if (bs->io_limits.iops[BLOCK_IO_LIMIT_TOTAL]) { |
5905fbc9 | 3910 | ios_base += bs->slice_submitted.ios[!is_write]; |
98f90dba ZYW |
3911 | } |
3912 | ||
3913 | if (ios_base + 1 <= ios_limit) { | |
3914 | if (wait) { | |
3915 | *wait = 0; | |
3916 | } | |
3917 | ||
3918 | return false; | |
3919 | } | |
3920 | ||
0775437f | 3921 | /* Calc approx time to dispatch, in seconds */ |
98f90dba ZYW |
3922 | wait_time = (ios_base + 1) / iops_limit; |
3923 | if (wait_time > elapsed_time) { | |
3924 | wait_time = wait_time - elapsed_time; | |
3925 | } else { | |
3926 | wait_time = 0; | |
3927 | } | |
3928 | ||
ae29d6c6 SH |
3929 | /* Exceeded current slice, extend it by another slice time */ |
3930 | bs->slice_end += BLOCK_IO_SLICE_TIME; | |
98f90dba | 3931 | if (wait) { |
0775437f | 3932 | *wait = wait_time * NANOSECONDS_PER_SECOND; |
98f90dba ZYW |
3933 | } |
3934 | ||
3935 | return true; | |
3936 | } | |
3937 | ||
3938 | static bool bdrv_exceed_io_limits(BlockDriverState *bs, int nb_sectors, | |
3939 | bool is_write, int64_t *wait) | |
3940 | { | |
3941 | int64_t now, max_wait; | |
3942 | uint64_t bps_wait = 0, iops_wait = 0; | |
3943 | double elapsed_time; | |
3944 | int bps_ret, iops_ret; | |
3945 | ||
3946 | now = qemu_get_clock_ns(vm_clock); | |
e660fb8b | 3947 | if (now > bs->slice_end) { |
98f90dba | 3948 | bs->slice_start = now; |
ae29d6c6 | 3949 | bs->slice_end = now + BLOCK_IO_SLICE_TIME; |
5905fbc9 | 3950 | memset(&bs->slice_submitted, 0, sizeof(bs->slice_submitted)); |
98f90dba ZYW |
3951 | } |
3952 | ||
3953 | elapsed_time = now - bs->slice_start; | |
3954 | elapsed_time /= (NANOSECONDS_PER_SECOND); | |
3955 | ||
3956 | bps_ret = bdrv_exceed_bps_limits(bs, nb_sectors, | |
3957 | is_write, elapsed_time, &bps_wait); | |
3958 | iops_ret = bdrv_exceed_iops_limits(bs, is_write, | |
3959 | elapsed_time, &iops_wait); | |
3960 | if (bps_ret || iops_ret) { | |
3961 | max_wait = bps_wait > iops_wait ? bps_wait : iops_wait; | |
3962 | if (wait) { | |
3963 | *wait = max_wait; | |
3964 | } | |
3965 | ||
3966 | now = qemu_get_clock_ns(vm_clock); | |
3967 | if (bs->slice_end < now + max_wait) { | |
3968 | bs->slice_end = now + max_wait; | |
3969 | } | |
3970 | ||
3971 | return true; | |
3972 | } | |
3973 | ||
3974 | if (wait) { | |
3975 | *wait = 0; | |
3976 | } | |
3977 | ||
5905fbc9 SH |
3978 | bs->slice_submitted.bytes[is_write] += (int64_t)nb_sectors * |
3979 | BDRV_SECTOR_SIZE; | |
3980 | bs->slice_submitted.ios[is_write]++; | |
3981 | ||
98f90dba ZYW |
3982 | return false; |
3983 | } | |
ce1a14dc | 3984 | |
83f64091 FB |
3985 | /**************************************************************/ |
3986 | /* async block device emulation */ | |
3987 | ||
c16b5a2c CH |
3988 | typedef struct BlockDriverAIOCBSync { |
3989 | BlockDriverAIOCB common; | |
3990 | QEMUBH *bh; | |
3991 | int ret; | |
3992 | /* vector translation state */ | |
3993 | QEMUIOVector *qiov; | |
3994 | uint8_t *bounce; | |
3995 | int is_write; | |
3996 | } BlockDriverAIOCBSync; | |
3997 | ||
3998 | static void bdrv_aio_cancel_em(BlockDriverAIOCB *blockacb) | |
3999 | { | |
b666d239 KW |
4000 | BlockDriverAIOCBSync *acb = |
4001 | container_of(blockacb, BlockDriverAIOCBSync, common); | |
6a7ad299 | 4002 | qemu_bh_delete(acb->bh); |
36afc451 | 4003 | acb->bh = NULL; |
c16b5a2c CH |
4004 | qemu_aio_release(acb); |
4005 | } | |
4006 | ||
d7331bed | 4007 | static const AIOCBInfo bdrv_em_aiocb_info = { |
c16b5a2c CH |
4008 | .aiocb_size = sizeof(BlockDriverAIOCBSync), |
4009 | .cancel = bdrv_aio_cancel_em, | |
4010 | }; | |
4011 | ||
ce1a14dc | 4012 | static void bdrv_aio_bh_cb(void *opaque) |
83f64091 | 4013 | { |
ce1a14dc | 4014 | BlockDriverAIOCBSync *acb = opaque; |
f141eafe | 4015 | |
f141eafe | 4016 | if (!acb->is_write) |
03396148 | 4017 | qemu_iovec_from_buf(acb->qiov, 0, acb->bounce, acb->qiov->size); |
ceb42de8 | 4018 | qemu_vfree(acb->bounce); |
ce1a14dc | 4019 | acb->common.cb(acb->common.opaque, acb->ret); |
6a7ad299 | 4020 | qemu_bh_delete(acb->bh); |
36afc451 | 4021 | acb->bh = NULL; |
ce1a14dc | 4022 | qemu_aio_release(acb); |
83f64091 | 4023 | } |
beac80cd | 4024 | |
f141eafe AL |
4025 | static BlockDriverAIOCB *bdrv_aio_rw_vector(BlockDriverState *bs, |
4026 | int64_t sector_num, | |
4027 | QEMUIOVector *qiov, | |
4028 | int nb_sectors, | |
4029 | BlockDriverCompletionFunc *cb, | |
4030 | void *opaque, | |
4031 | int is_write) | |
4032 | ||
83f64091 | 4033 | { |
ce1a14dc | 4034 | BlockDriverAIOCBSync *acb; |
ce1a14dc | 4035 | |
d7331bed | 4036 | acb = qemu_aio_get(&bdrv_em_aiocb_info, bs, cb, opaque); |
f141eafe AL |
4037 | acb->is_write = is_write; |
4038 | acb->qiov = qiov; | |
e268ca52 | 4039 | acb->bounce = qemu_blockalign(bs, qiov->size); |
3f3aace8 | 4040 | acb->bh = qemu_bh_new(bdrv_aio_bh_cb, acb); |
f141eafe AL |
4041 | |
4042 | if (is_write) { | |
d5e6b161 | 4043 | qemu_iovec_to_buf(acb->qiov, 0, acb->bounce, qiov->size); |
1ed20acf | 4044 | acb->ret = bs->drv->bdrv_write(bs, sector_num, acb->bounce, nb_sectors); |
f141eafe | 4045 | } else { |
1ed20acf | 4046 | acb->ret = bs->drv->bdrv_read(bs, sector_num, acb->bounce, nb_sectors); |
f141eafe AL |
4047 | } |
4048 | ||
ce1a14dc | 4049 | qemu_bh_schedule(acb->bh); |
f141eafe | 4050 | |
ce1a14dc | 4051 | return &acb->common; |
beac80cd FB |
4052 | } |
4053 | ||
f141eafe AL |
4054 | static BlockDriverAIOCB *bdrv_aio_readv_em(BlockDriverState *bs, |
4055 | int64_t sector_num, QEMUIOVector *qiov, int nb_sectors, | |
ce1a14dc | 4056 | BlockDriverCompletionFunc *cb, void *opaque) |
beac80cd | 4057 | { |
f141eafe AL |
4058 | return bdrv_aio_rw_vector(bs, sector_num, qiov, nb_sectors, cb, opaque, 0); |
4059 | } | |
83f64091 | 4060 | |
f141eafe AL |
4061 | static BlockDriverAIOCB *bdrv_aio_writev_em(BlockDriverState *bs, |
4062 | int64_t sector_num, QEMUIOVector *qiov, int nb_sectors, | |
4063 | BlockDriverCompletionFunc *cb, void *opaque) | |
4064 | { | |
4065 | return bdrv_aio_rw_vector(bs, sector_num, qiov, nb_sectors, cb, opaque, 1); | |
beac80cd | 4066 | } |
beac80cd | 4067 | |
68485420 KW |
4068 | |
4069 | typedef struct BlockDriverAIOCBCoroutine { | |
4070 | BlockDriverAIOCB common; | |
4071 | BlockRequest req; | |
4072 | bool is_write; | |
d318aea9 | 4073 | bool *done; |
68485420 KW |
4074 | QEMUBH* bh; |
4075 | } BlockDriverAIOCBCoroutine; | |
4076 | ||
4077 | static void bdrv_aio_co_cancel_em(BlockDriverAIOCB *blockacb) | |
4078 | { | |
d318aea9 KW |
4079 | BlockDriverAIOCBCoroutine *acb = |
4080 | container_of(blockacb, BlockDriverAIOCBCoroutine, common); | |
4081 | bool done = false; | |
4082 | ||
4083 | acb->done = &done; | |
4084 | while (!done) { | |
4085 | qemu_aio_wait(); | |
4086 | } | |
68485420 KW |
4087 | } |
4088 | ||
d7331bed | 4089 | static const AIOCBInfo bdrv_em_co_aiocb_info = { |
68485420 KW |
4090 | .aiocb_size = sizeof(BlockDriverAIOCBCoroutine), |
4091 | .cancel = bdrv_aio_co_cancel_em, | |
4092 | }; | |
4093 | ||
35246a68 | 4094 | static void bdrv_co_em_bh(void *opaque) |
68485420 KW |
4095 | { |
4096 | BlockDriverAIOCBCoroutine *acb = opaque; | |
4097 | ||
4098 | acb->common.cb(acb->common.opaque, acb->req.error); | |
d318aea9 KW |
4099 | |
4100 | if (acb->done) { | |
4101 | *acb->done = true; | |
4102 | } | |
4103 | ||
68485420 KW |
4104 | qemu_bh_delete(acb->bh); |
4105 | qemu_aio_release(acb); | |
4106 | } | |
4107 | ||
b2a61371 SH |
4108 | /* Invoke bdrv_co_do_readv/bdrv_co_do_writev */ |
4109 | static void coroutine_fn bdrv_co_do_rw(void *opaque) | |
4110 | { | |
4111 | BlockDriverAIOCBCoroutine *acb = opaque; | |
4112 | BlockDriverState *bs = acb->common.bs; | |
4113 | ||
4114 | if (!acb->is_write) { | |
4115 | acb->req.error = bdrv_co_do_readv(bs, acb->req.sector, | |
470c0504 | 4116 | acb->req.nb_sectors, acb->req.qiov, 0); |
b2a61371 SH |
4117 | } else { |
4118 | acb->req.error = bdrv_co_do_writev(bs, acb->req.sector, | |
f08f2dda | 4119 | acb->req.nb_sectors, acb->req.qiov, 0); |
b2a61371 SH |
4120 | } |
4121 | ||
35246a68 | 4122 | acb->bh = qemu_bh_new(bdrv_co_em_bh, acb); |
b2a61371 SH |
4123 | qemu_bh_schedule(acb->bh); |
4124 | } | |
4125 | ||
68485420 KW |
4126 | static BlockDriverAIOCB *bdrv_co_aio_rw_vector(BlockDriverState *bs, |
4127 | int64_t sector_num, | |
4128 | QEMUIOVector *qiov, | |
4129 | int nb_sectors, | |
4130 | BlockDriverCompletionFunc *cb, | |
4131 | void *opaque, | |
8c5873d6 | 4132 | bool is_write) |
68485420 KW |
4133 | { |
4134 | Coroutine *co; | |
4135 | BlockDriverAIOCBCoroutine *acb; | |
4136 | ||
d7331bed | 4137 | acb = qemu_aio_get(&bdrv_em_co_aiocb_info, bs, cb, opaque); |
68485420 KW |
4138 | acb->req.sector = sector_num; |
4139 | acb->req.nb_sectors = nb_sectors; | |
4140 | acb->req.qiov = qiov; | |
4141 | acb->is_write = is_write; | |
d318aea9 | 4142 | acb->done = NULL; |
68485420 | 4143 | |
8c5873d6 | 4144 | co = qemu_coroutine_create(bdrv_co_do_rw); |
68485420 KW |
4145 | qemu_coroutine_enter(co, acb); |
4146 | ||
4147 | return &acb->common; | |
4148 | } | |
4149 | ||
07f07615 | 4150 | static void coroutine_fn bdrv_aio_flush_co_entry(void *opaque) |
b2e12bc6 | 4151 | { |
07f07615 PB |
4152 | BlockDriverAIOCBCoroutine *acb = opaque; |
4153 | BlockDriverState *bs = acb->common.bs; | |
b2e12bc6 | 4154 | |
07f07615 PB |
4155 | acb->req.error = bdrv_co_flush(bs); |
4156 | acb->bh = qemu_bh_new(bdrv_co_em_bh, acb); | |
b2e12bc6 | 4157 | qemu_bh_schedule(acb->bh); |
b2e12bc6 CH |
4158 | } |
4159 | ||
07f07615 | 4160 | BlockDriverAIOCB *bdrv_aio_flush(BlockDriverState *bs, |
016f5cf6 AG |
4161 | BlockDriverCompletionFunc *cb, void *opaque) |
4162 | { | |
07f07615 | 4163 | trace_bdrv_aio_flush(bs, opaque); |
016f5cf6 | 4164 | |
07f07615 PB |
4165 | Coroutine *co; |
4166 | BlockDriverAIOCBCoroutine *acb; | |
016f5cf6 | 4167 | |
d7331bed | 4168 | acb = qemu_aio_get(&bdrv_em_co_aiocb_info, bs, cb, opaque); |
d318aea9 KW |
4169 | acb->done = NULL; |
4170 | ||
07f07615 PB |
4171 | co = qemu_coroutine_create(bdrv_aio_flush_co_entry); |
4172 | qemu_coroutine_enter(co, acb); | |
016f5cf6 | 4173 | |
016f5cf6 AG |
4174 | return &acb->common; |
4175 | } | |
4176 | ||
4265d620 PB |
4177 | static void coroutine_fn bdrv_aio_discard_co_entry(void *opaque) |
4178 | { | |
4179 | BlockDriverAIOCBCoroutine *acb = opaque; | |
4180 | BlockDriverState *bs = acb->common.bs; | |
4181 | ||
4182 | acb->req.error = bdrv_co_discard(bs, acb->req.sector, acb->req.nb_sectors); | |
4183 | acb->bh = qemu_bh_new(bdrv_co_em_bh, acb); | |
4184 | qemu_bh_schedule(acb->bh); | |
4185 | } | |
4186 | ||
4187 | BlockDriverAIOCB *bdrv_aio_discard(BlockDriverState *bs, | |
4188 | int64_t sector_num, int nb_sectors, | |
4189 | BlockDriverCompletionFunc *cb, void *opaque) | |
4190 | { | |
4191 | Coroutine *co; | |
4192 | BlockDriverAIOCBCoroutine *acb; | |
4193 | ||
4194 | trace_bdrv_aio_discard(bs, sector_num, nb_sectors, opaque); | |
4195 | ||
d7331bed | 4196 | acb = qemu_aio_get(&bdrv_em_co_aiocb_info, bs, cb, opaque); |
4265d620 PB |
4197 | acb->req.sector = sector_num; |
4198 | acb->req.nb_sectors = nb_sectors; | |
d318aea9 | 4199 | acb->done = NULL; |
4265d620 PB |
4200 | co = qemu_coroutine_create(bdrv_aio_discard_co_entry); |
4201 | qemu_coroutine_enter(co, acb); | |
4202 | ||
4203 | return &acb->common; | |
4204 | } | |
4205 | ||
ea2384d3 FB |
4206 | void bdrv_init(void) |
4207 | { | |
5efa9d5a | 4208 | module_call_init(MODULE_INIT_BLOCK); |
ea2384d3 | 4209 | } |
ce1a14dc | 4210 | |
eb852011 MA |
4211 | void bdrv_init_with_whitelist(void) |
4212 | { | |
4213 | use_bdrv_whitelist = 1; | |
4214 | bdrv_init(); | |
4215 | } | |
4216 | ||
d7331bed | 4217 | void *qemu_aio_get(const AIOCBInfo *aiocb_info, BlockDriverState *bs, |
c16b5a2c | 4218 | BlockDriverCompletionFunc *cb, void *opaque) |
ce1a14dc | 4219 | { |
ce1a14dc PB |
4220 | BlockDriverAIOCB *acb; |
4221 | ||
d7331bed SH |
4222 | acb = g_slice_alloc(aiocb_info->aiocb_size); |
4223 | acb->aiocb_info = aiocb_info; | |
ce1a14dc PB |
4224 | acb->bs = bs; |
4225 | acb->cb = cb; | |
4226 | acb->opaque = opaque; | |
4227 | return acb; | |
4228 | } | |
4229 | ||
4230 | void qemu_aio_release(void *p) | |
4231 | { | |
d37c975f | 4232 | BlockDriverAIOCB *acb = p; |
d7331bed | 4233 | g_slice_free1(acb->aiocb_info->aiocb_size, acb); |
ce1a14dc | 4234 | } |
19cb3738 | 4235 | |
f9f05dc5 KW |
4236 | /**************************************************************/ |
4237 | /* Coroutine block device emulation */ | |
4238 | ||
4239 | typedef struct CoroutineIOCompletion { | |
4240 | Coroutine *coroutine; | |
4241 | int ret; | |
4242 | } CoroutineIOCompletion; | |
4243 | ||
4244 | static void bdrv_co_io_em_complete(void *opaque, int ret) | |
4245 | { | |
4246 | CoroutineIOCompletion *co = opaque; | |
4247 | ||
4248 | co->ret = ret; | |
4249 | qemu_coroutine_enter(co->coroutine, NULL); | |
4250 | } | |
4251 | ||
4252 | static int coroutine_fn bdrv_co_io_em(BlockDriverState *bs, int64_t sector_num, | |
4253 | int nb_sectors, QEMUIOVector *iov, | |
4254 | bool is_write) | |
4255 | { | |
4256 | CoroutineIOCompletion co = { | |
4257 | .coroutine = qemu_coroutine_self(), | |
4258 | }; | |
4259 | BlockDriverAIOCB *acb; | |
4260 | ||
4261 | if (is_write) { | |
a652d160 SH |
4262 | acb = bs->drv->bdrv_aio_writev(bs, sector_num, iov, nb_sectors, |
4263 | bdrv_co_io_em_complete, &co); | |
f9f05dc5 | 4264 | } else { |
a652d160 SH |
4265 | acb = bs->drv->bdrv_aio_readv(bs, sector_num, iov, nb_sectors, |
4266 | bdrv_co_io_em_complete, &co); | |
f9f05dc5 KW |
4267 | } |
4268 | ||
59370aaa | 4269 | trace_bdrv_co_io_em(bs, sector_num, nb_sectors, is_write, acb); |
f9f05dc5 KW |
4270 | if (!acb) { |
4271 | return -EIO; | |
4272 | } | |
4273 | qemu_coroutine_yield(); | |
4274 | ||
4275 | return co.ret; | |
4276 | } | |
4277 | ||
4278 | static int coroutine_fn bdrv_co_readv_em(BlockDriverState *bs, | |
4279 | int64_t sector_num, int nb_sectors, | |
4280 | QEMUIOVector *iov) | |
4281 | { | |
4282 | return bdrv_co_io_em(bs, sector_num, nb_sectors, iov, false); | |
4283 | } | |
4284 | ||
4285 | static int coroutine_fn bdrv_co_writev_em(BlockDriverState *bs, | |
4286 | int64_t sector_num, int nb_sectors, | |
4287 | QEMUIOVector *iov) | |
4288 | { | |
4289 | return bdrv_co_io_em(bs, sector_num, nb_sectors, iov, true); | |
4290 | } | |
4291 | ||
07f07615 | 4292 | static void coroutine_fn bdrv_flush_co_entry(void *opaque) |
e7a8a783 | 4293 | { |
07f07615 PB |
4294 | RwCo *rwco = opaque; |
4295 | ||
4296 | rwco->ret = bdrv_co_flush(rwco->bs); | |
4297 | } | |
4298 | ||
4299 | int coroutine_fn bdrv_co_flush(BlockDriverState *bs) | |
4300 | { | |
eb489bb1 KW |
4301 | int ret; |
4302 | ||
29cdb251 | 4303 | if (!bs || !bdrv_is_inserted(bs) || bdrv_is_read_only(bs)) { |
07f07615 | 4304 | return 0; |
eb489bb1 KW |
4305 | } |
4306 | ||
ca716364 | 4307 | /* Write back cached data to the OS even with cache=unsafe */ |
eb489bb1 KW |
4308 | if (bs->drv->bdrv_co_flush_to_os) { |
4309 | ret = bs->drv->bdrv_co_flush_to_os(bs); | |
4310 | if (ret < 0) { | |
4311 | return ret; | |
4312 | } | |
4313 | } | |
4314 | ||
ca716364 KW |
4315 | /* But don't actually force it to the disk with cache=unsafe */ |
4316 | if (bs->open_flags & BDRV_O_NO_FLUSH) { | |
d4c82329 | 4317 | goto flush_parent; |
ca716364 KW |
4318 | } |
4319 | ||
eb489bb1 | 4320 | if (bs->drv->bdrv_co_flush_to_disk) { |
29cdb251 | 4321 | ret = bs->drv->bdrv_co_flush_to_disk(bs); |
07f07615 PB |
4322 | } else if (bs->drv->bdrv_aio_flush) { |
4323 | BlockDriverAIOCB *acb; | |
4324 | CoroutineIOCompletion co = { | |
4325 | .coroutine = qemu_coroutine_self(), | |
4326 | }; | |
4327 | ||
4328 | acb = bs->drv->bdrv_aio_flush(bs, bdrv_co_io_em_complete, &co); | |
4329 | if (acb == NULL) { | |
29cdb251 | 4330 | ret = -EIO; |
07f07615 PB |
4331 | } else { |
4332 | qemu_coroutine_yield(); | |
29cdb251 | 4333 | ret = co.ret; |
07f07615 | 4334 | } |
07f07615 PB |
4335 | } else { |
4336 | /* | |
4337 | * Some block drivers always operate in either writethrough or unsafe | |
4338 | * mode and don't support bdrv_flush therefore. Usually qemu doesn't | |
4339 | * know how the server works (because the behaviour is hardcoded or | |
4340 | * depends on server-side configuration), so we can't ensure that | |
4341 | * everything is safe on disk. Returning an error doesn't work because | |
4342 | * that would break guests even if the server operates in writethrough | |
4343 | * mode. | |
4344 | * | |
4345 | * Let's hope the user knows what he's doing. | |
4346 | */ | |
29cdb251 | 4347 | ret = 0; |
07f07615 | 4348 | } |
29cdb251 PB |
4349 | if (ret < 0) { |
4350 | return ret; | |
4351 | } | |
4352 | ||
4353 | /* Now flush the underlying protocol. It will also have BDRV_O_NO_FLUSH | |
4354 | * in the case of cache=unsafe, so there are no useless flushes. | |
4355 | */ | |
d4c82329 | 4356 | flush_parent: |
29cdb251 | 4357 | return bdrv_co_flush(bs->file); |
07f07615 PB |
4358 | } |
4359 | ||
0f15423c AL |
4360 | void bdrv_invalidate_cache(BlockDriverState *bs) |
4361 | { | |
4362 | if (bs->drv && bs->drv->bdrv_invalidate_cache) { | |
4363 | bs->drv->bdrv_invalidate_cache(bs); | |
4364 | } | |
4365 | } | |
4366 | ||
4367 | void bdrv_invalidate_cache_all(void) | |
4368 | { | |
4369 | BlockDriverState *bs; | |
4370 | ||
4371 | QTAILQ_FOREACH(bs, &bdrv_states, list) { | |
4372 | bdrv_invalidate_cache(bs); | |
4373 | } | |
4374 | } | |
4375 | ||
07789269 BC |
4376 | void bdrv_clear_incoming_migration_all(void) |
4377 | { | |
4378 | BlockDriverState *bs; | |
4379 | ||
4380 | QTAILQ_FOREACH(bs, &bdrv_states, list) { | |
4381 | bs->open_flags = bs->open_flags & ~(BDRV_O_INCOMING); | |
4382 | } | |
4383 | } | |
4384 | ||
07f07615 PB |
4385 | int bdrv_flush(BlockDriverState *bs) |
4386 | { | |
4387 | Coroutine *co; | |
4388 | RwCo rwco = { | |
4389 | .bs = bs, | |
4390 | .ret = NOT_DONE, | |
e7a8a783 | 4391 | }; |
e7a8a783 | 4392 | |
07f07615 PB |
4393 | if (qemu_in_coroutine()) { |
4394 | /* Fast-path if already in coroutine context */ | |
4395 | bdrv_flush_co_entry(&rwco); | |
4396 | } else { | |
4397 | co = qemu_coroutine_create(bdrv_flush_co_entry); | |
4398 | qemu_coroutine_enter(co, &rwco); | |
4399 | while (rwco.ret == NOT_DONE) { | |
4400 | qemu_aio_wait(); | |
4401 | } | |
e7a8a783 | 4402 | } |
07f07615 PB |
4403 | |
4404 | return rwco.ret; | |
e7a8a783 KW |
4405 | } |
4406 | ||
4265d620 PB |
4407 | static void coroutine_fn bdrv_discard_co_entry(void *opaque) |
4408 | { | |
4409 | RwCo *rwco = opaque; | |
4410 | ||
4411 | rwco->ret = bdrv_co_discard(rwco->bs, rwco->sector_num, rwco->nb_sectors); | |
4412 | } | |
4413 | ||
4414 | int coroutine_fn bdrv_co_discard(BlockDriverState *bs, int64_t sector_num, | |
4415 | int nb_sectors) | |
4416 | { | |
4417 | if (!bs->drv) { | |
4418 | return -ENOMEDIUM; | |
4419 | } else if (bdrv_check_request(bs, sector_num, nb_sectors)) { | |
4420 | return -EIO; | |
4421 | } else if (bs->read_only) { | |
4422 | return -EROFS; | |
df702c9b PB |
4423 | } |
4424 | ||
4425 | if (bs->dirty_bitmap) { | |
8f0720ec | 4426 | bdrv_reset_dirty(bs, sector_num, nb_sectors); |
df702c9b PB |
4427 | } |
4428 | ||
9e8f1835 PB |
4429 | /* Do nothing if disabled. */ |
4430 | if (!(bs->open_flags & BDRV_O_UNMAP)) { | |
4431 | return 0; | |
4432 | } | |
4433 | ||
df702c9b | 4434 | if (bs->drv->bdrv_co_discard) { |
4265d620 PB |
4435 | return bs->drv->bdrv_co_discard(bs, sector_num, nb_sectors); |
4436 | } else if (bs->drv->bdrv_aio_discard) { | |
4437 | BlockDriverAIOCB *acb; | |
4438 | CoroutineIOCompletion co = { | |
4439 | .coroutine = qemu_coroutine_self(), | |
4440 | }; | |
4441 | ||
4442 | acb = bs->drv->bdrv_aio_discard(bs, sector_num, nb_sectors, | |
4443 | bdrv_co_io_em_complete, &co); | |
4444 | if (acb == NULL) { | |
4445 | return -EIO; | |
4446 | } else { | |
4447 | qemu_coroutine_yield(); | |
4448 | return co.ret; | |
4449 | } | |
4265d620 PB |
4450 | } else { |
4451 | return 0; | |
4452 | } | |
4453 | } | |
4454 | ||
4455 | int bdrv_discard(BlockDriverState *bs, int64_t sector_num, int nb_sectors) | |
4456 | { | |
4457 | Coroutine *co; | |
4458 | RwCo rwco = { | |
4459 | .bs = bs, | |
4460 | .sector_num = sector_num, | |
4461 | .nb_sectors = nb_sectors, | |
4462 | .ret = NOT_DONE, | |
4463 | }; | |
4464 | ||
4465 | if (qemu_in_coroutine()) { | |
4466 | /* Fast-path if already in coroutine context */ | |
4467 | bdrv_discard_co_entry(&rwco); | |
4468 | } else { | |
4469 | co = qemu_coroutine_create(bdrv_discard_co_entry); | |
4470 | qemu_coroutine_enter(co, &rwco); | |
4471 | while (rwco.ret == NOT_DONE) { | |
4472 | qemu_aio_wait(); | |
4473 | } | |
4474 | } | |
4475 | ||
4476 | return rwco.ret; | |
4477 | } | |
4478 | ||
19cb3738 FB |
4479 | /**************************************************************/ |
4480 | /* removable device support */ | |
4481 | ||
4482 | /** | |
4483 | * Return TRUE if the media is present | |
4484 | */ | |
4485 | int bdrv_is_inserted(BlockDriverState *bs) | |
4486 | { | |
4487 | BlockDriver *drv = bs->drv; | |
a1aff5bf | 4488 | |
19cb3738 FB |
4489 | if (!drv) |
4490 | return 0; | |
4491 | if (!drv->bdrv_is_inserted) | |
a1aff5bf MA |
4492 | return 1; |
4493 | return drv->bdrv_is_inserted(bs); | |
19cb3738 FB |
4494 | } |
4495 | ||
4496 | /** | |
8e49ca46 MA |
4497 | * Return whether the media changed since the last call to this |
4498 | * function, or -ENOTSUP if we don't know. Most drivers don't know. | |
19cb3738 FB |
4499 | */ |
4500 | int bdrv_media_changed(BlockDriverState *bs) | |
4501 | { | |
4502 | BlockDriver *drv = bs->drv; | |
19cb3738 | 4503 | |
8e49ca46 MA |
4504 | if (drv && drv->bdrv_media_changed) { |
4505 | return drv->bdrv_media_changed(bs); | |
4506 | } | |
4507 | return -ENOTSUP; | |
19cb3738 FB |
4508 | } |
4509 | ||
4510 | /** | |
4511 | * If eject_flag is TRUE, eject the media. Otherwise, close the tray | |
4512 | */ | |
f36f3949 | 4513 | void bdrv_eject(BlockDriverState *bs, bool eject_flag) |
19cb3738 FB |
4514 | { |
4515 | BlockDriver *drv = bs->drv; | |
19cb3738 | 4516 | |
822e1cd1 MA |
4517 | if (drv && drv->bdrv_eject) { |
4518 | drv->bdrv_eject(bs, eject_flag); | |
19cb3738 | 4519 | } |
6f382ed2 LC |
4520 | |
4521 | if (bs->device_name[0] != '\0') { | |
4522 | bdrv_emit_qmp_eject_event(bs, eject_flag); | |
4523 | } | |
19cb3738 FB |
4524 | } |
4525 | ||
19cb3738 FB |
4526 | /** |
4527 | * Lock or unlock the media (if it is locked, the user won't be able | |
4528 | * to eject it manually). | |
4529 | */ | |
025e849a | 4530 | void bdrv_lock_medium(BlockDriverState *bs, bool locked) |
19cb3738 FB |
4531 | { |
4532 | BlockDriver *drv = bs->drv; | |
4533 | ||
025e849a | 4534 | trace_bdrv_lock_medium(bs, locked); |
b8c6d095 | 4535 | |
025e849a MA |
4536 | if (drv && drv->bdrv_lock_medium) { |
4537 | drv->bdrv_lock_medium(bs, locked); | |
19cb3738 FB |
4538 | } |
4539 | } | |
985a03b0 TS |
4540 | |
4541 | /* needed for generic scsi interface */ | |
4542 | ||
4543 | int bdrv_ioctl(BlockDriverState *bs, unsigned long int req, void *buf) | |
4544 | { | |
4545 | BlockDriver *drv = bs->drv; | |
4546 | ||
4547 | if (drv && drv->bdrv_ioctl) | |
4548 | return drv->bdrv_ioctl(bs, req, buf); | |
4549 | return -ENOTSUP; | |
4550 | } | |
7d780669 | 4551 | |
221f715d AL |
4552 | BlockDriverAIOCB *bdrv_aio_ioctl(BlockDriverState *bs, |
4553 | unsigned long int req, void *buf, | |
4554 | BlockDriverCompletionFunc *cb, void *opaque) | |
7d780669 | 4555 | { |
221f715d | 4556 | BlockDriver *drv = bs->drv; |
7d780669 | 4557 | |
221f715d AL |
4558 | if (drv && drv->bdrv_aio_ioctl) |
4559 | return drv->bdrv_aio_ioctl(bs, req, buf, cb, opaque); | |
4560 | return NULL; | |
7d780669 | 4561 | } |
e268ca52 | 4562 | |
7b6f9300 MA |
4563 | void bdrv_set_buffer_alignment(BlockDriverState *bs, int align) |
4564 | { | |
4565 | bs->buffer_alignment = align; | |
4566 | } | |
7cd1e32a | 4567 | |
e268ca52 AL |
4568 | void *qemu_blockalign(BlockDriverState *bs, size_t size) |
4569 | { | |
4570 | return qemu_memalign((bs && bs->buffer_alignment) ? bs->buffer_alignment : 512, size); | |
4571 | } | |
7cd1e32a | 4572 | |
c53b1c51 SH |
4573 | /* |
4574 | * Check if all memory in this vector is sector aligned. | |
4575 | */ | |
4576 | bool bdrv_qiov_is_aligned(BlockDriverState *bs, QEMUIOVector *qiov) | |
4577 | { | |
4578 | int i; | |
4579 | ||
4580 | for (i = 0; i < qiov->niov; i++) { | |
4581 | if ((uintptr_t) qiov->iov[i].iov_base % bs->buffer_alignment) { | |
4582 | return false; | |
4583 | } | |
4584 | } | |
4585 | ||
4586 | return true; | |
4587 | } | |
4588 | ||
50717e94 | 4589 | void bdrv_set_dirty_tracking(BlockDriverState *bs, int granularity) |
7cd1e32a | 4590 | { |
4591 | int64_t bitmap_size; | |
a55eb92c | 4592 | |
50717e94 PB |
4593 | assert((granularity & (granularity - 1)) == 0); |
4594 | ||
4595 | if (granularity) { | |
4596 | granularity >>= BDRV_SECTOR_BITS; | |
4597 | assert(!bs->dirty_bitmap); | |
4598 | bitmap_size = (bdrv_getlength(bs) >> BDRV_SECTOR_BITS); | |
4599 | bs->dirty_bitmap = hbitmap_alloc(bitmap_size, ffs(granularity) - 1); | |
7cd1e32a | 4600 | } else { |
c6d22830 | 4601 | if (bs->dirty_bitmap) { |
8f0720ec | 4602 | hbitmap_free(bs->dirty_bitmap); |
c6d22830 | 4603 | bs->dirty_bitmap = NULL; |
a55eb92c | 4604 | } |
7cd1e32a | 4605 | } |
4606 | } | |
4607 | ||
4608 | int bdrv_get_dirty(BlockDriverState *bs, int64_t sector) | |
4609 | { | |
8f0720ec PB |
4610 | if (bs->dirty_bitmap) { |
4611 | return hbitmap_get(bs->dirty_bitmap, sector); | |
7cd1e32a | 4612 | } else { |
4613 | return 0; | |
4614 | } | |
4615 | } | |
4616 | ||
8f0720ec | 4617 | void bdrv_dirty_iter_init(BlockDriverState *bs, HBitmapIter *hbi) |
1755da16 | 4618 | { |
8f0720ec | 4619 | hbitmap_iter_init(hbi, bs->dirty_bitmap, 0); |
1755da16 PB |
4620 | } |
4621 | ||
4622 | void bdrv_set_dirty(BlockDriverState *bs, int64_t cur_sector, | |
4623 | int nr_sectors) | |
4624 | { | |
8f0720ec | 4625 | hbitmap_set(bs->dirty_bitmap, cur_sector, nr_sectors); |
1755da16 PB |
4626 | } |
4627 | ||
a55eb92c JK |
4628 | void bdrv_reset_dirty(BlockDriverState *bs, int64_t cur_sector, |
4629 | int nr_sectors) | |
7cd1e32a | 4630 | { |
8f0720ec | 4631 | hbitmap_reset(bs->dirty_bitmap, cur_sector, nr_sectors); |
7cd1e32a | 4632 | } |
aaa0eb75 LS |
4633 | |
4634 | int64_t bdrv_get_dirty_count(BlockDriverState *bs) | |
4635 | { | |
8f0720ec | 4636 | if (bs->dirty_bitmap) { |
acc906c6 | 4637 | return hbitmap_count(bs->dirty_bitmap); |
8f0720ec PB |
4638 | } else { |
4639 | return 0; | |
4640 | } | |
aaa0eb75 | 4641 | } |
f88e1a42 | 4642 | |
db593f25 MT |
4643 | void bdrv_set_in_use(BlockDriverState *bs, int in_use) |
4644 | { | |
4645 | assert(bs->in_use != in_use); | |
4646 | bs->in_use = in_use; | |
4647 | } | |
4648 | ||
4649 | int bdrv_in_use(BlockDriverState *bs) | |
4650 | { | |
4651 | return bs->in_use; | |
4652 | } | |
4653 | ||
28a7282a LC |
4654 | void bdrv_iostatus_enable(BlockDriverState *bs) |
4655 | { | |
d6bf279e | 4656 | bs->iostatus_enabled = true; |
58e21ef5 | 4657 | bs->iostatus = BLOCK_DEVICE_IO_STATUS_OK; |
28a7282a LC |
4658 | } |
4659 | ||
4660 | /* The I/O status is only enabled if the drive explicitly | |
4661 | * enables it _and_ the VM is configured to stop on errors */ | |
4662 | bool bdrv_iostatus_is_enabled(const BlockDriverState *bs) | |
4663 | { | |
d6bf279e | 4664 | return (bs->iostatus_enabled && |
92aa5c6d PB |
4665 | (bs->on_write_error == BLOCKDEV_ON_ERROR_ENOSPC || |
4666 | bs->on_write_error == BLOCKDEV_ON_ERROR_STOP || | |
4667 | bs->on_read_error == BLOCKDEV_ON_ERROR_STOP)); | |
28a7282a LC |
4668 | } |
4669 | ||
4670 | void bdrv_iostatus_disable(BlockDriverState *bs) | |
4671 | { | |
d6bf279e | 4672 | bs->iostatus_enabled = false; |
28a7282a LC |
4673 | } |
4674 | ||
4675 | void bdrv_iostatus_reset(BlockDriverState *bs) | |
4676 | { | |
4677 | if (bdrv_iostatus_is_enabled(bs)) { | |
58e21ef5 | 4678 | bs->iostatus = BLOCK_DEVICE_IO_STATUS_OK; |
3bd293c3 PB |
4679 | if (bs->job) { |
4680 | block_job_iostatus_reset(bs->job); | |
4681 | } | |
28a7282a LC |
4682 | } |
4683 | } | |
4684 | ||
28a7282a LC |
4685 | void bdrv_iostatus_set_err(BlockDriverState *bs, int error) |
4686 | { | |
3e1caa5f PB |
4687 | assert(bdrv_iostatus_is_enabled(bs)); |
4688 | if (bs->iostatus == BLOCK_DEVICE_IO_STATUS_OK) { | |
58e21ef5 LC |
4689 | bs->iostatus = error == ENOSPC ? BLOCK_DEVICE_IO_STATUS_NOSPACE : |
4690 | BLOCK_DEVICE_IO_STATUS_FAILED; | |
28a7282a LC |
4691 | } |
4692 | } | |
4693 | ||
a597e79c CH |
4694 | void |
4695 | bdrv_acct_start(BlockDriverState *bs, BlockAcctCookie *cookie, int64_t bytes, | |
4696 | enum BlockAcctType type) | |
4697 | { | |
4698 | assert(type < BDRV_MAX_IOTYPE); | |
4699 | ||
4700 | cookie->bytes = bytes; | |
c488c7f6 | 4701 | cookie->start_time_ns = get_clock(); |
a597e79c CH |
4702 | cookie->type = type; |
4703 | } | |
4704 | ||
4705 | void | |
4706 | bdrv_acct_done(BlockDriverState *bs, BlockAcctCookie *cookie) | |
4707 | { | |
4708 | assert(cookie->type < BDRV_MAX_IOTYPE); | |
4709 | ||
4710 | bs->nr_bytes[cookie->type] += cookie->bytes; | |
4711 | bs->nr_ops[cookie->type]++; | |
c488c7f6 | 4712 | bs->total_time_ns[cookie->type] += get_clock() - cookie->start_time_ns; |
a597e79c CH |
4713 | } |
4714 | ||
d92ada22 LC |
4715 | void bdrv_img_create(const char *filename, const char *fmt, |
4716 | const char *base_filename, const char *base_fmt, | |
f382d43a MR |
4717 | char *options, uint64_t img_size, int flags, |
4718 | Error **errp, bool quiet) | |
f88e1a42 JS |
4719 | { |
4720 | QEMUOptionParameter *param = NULL, *create_options = NULL; | |
d220894e | 4721 | QEMUOptionParameter *backing_fmt, *backing_file, *size; |
f88e1a42 JS |
4722 | BlockDriverState *bs = NULL; |
4723 | BlockDriver *drv, *proto_drv; | |
96df67d1 | 4724 | BlockDriver *backing_drv = NULL; |
f88e1a42 JS |
4725 | int ret = 0; |
4726 | ||
4727 | /* Find driver and parse its options */ | |
4728 | drv = bdrv_find_format(fmt); | |
4729 | if (!drv) { | |
71c79813 | 4730 | error_setg(errp, "Unknown file format '%s'", fmt); |
d92ada22 | 4731 | return; |
f88e1a42 JS |
4732 | } |
4733 | ||
4734 | proto_drv = bdrv_find_protocol(filename); | |
4735 | if (!proto_drv) { | |
71c79813 | 4736 | error_setg(errp, "Unknown protocol '%s'", filename); |
d92ada22 | 4737 | return; |
f88e1a42 JS |
4738 | } |
4739 | ||
4740 | create_options = append_option_parameters(create_options, | |
4741 | drv->create_options); | |
4742 | create_options = append_option_parameters(create_options, | |
4743 | proto_drv->create_options); | |
4744 | ||
4745 | /* Create parameter list with default values */ | |
4746 | param = parse_option_parameters("", create_options, param); | |
4747 | ||
4748 | set_option_parameter_int(param, BLOCK_OPT_SIZE, img_size); | |
4749 | ||
4750 | /* Parse -o options */ | |
4751 | if (options) { | |
4752 | param = parse_option_parameters(options, create_options, param); | |
4753 | if (param == NULL) { | |
71c79813 | 4754 | error_setg(errp, "Invalid options for file format '%s'.", fmt); |
f88e1a42 JS |
4755 | goto out; |
4756 | } | |
4757 | } | |
4758 | ||
4759 | if (base_filename) { | |
4760 | if (set_option_parameter(param, BLOCK_OPT_BACKING_FILE, | |
4761 | base_filename)) { | |
71c79813 LC |
4762 | error_setg(errp, "Backing file not supported for file format '%s'", |
4763 | fmt); | |
f88e1a42 JS |
4764 | goto out; |
4765 | } | |
4766 | } | |
4767 | ||
4768 | if (base_fmt) { | |
4769 | if (set_option_parameter(param, BLOCK_OPT_BACKING_FMT, base_fmt)) { | |
71c79813 LC |
4770 | error_setg(errp, "Backing file format not supported for file " |
4771 | "format '%s'", fmt); | |
f88e1a42 JS |
4772 | goto out; |
4773 | } | |
4774 | } | |
4775 | ||
792da93a JS |
4776 | backing_file = get_option_parameter(param, BLOCK_OPT_BACKING_FILE); |
4777 | if (backing_file && backing_file->value.s) { | |
4778 | if (!strcmp(filename, backing_file->value.s)) { | |
71c79813 LC |
4779 | error_setg(errp, "Error: Trying to create an image with the " |
4780 | "same filename as the backing file"); | |
792da93a JS |
4781 | goto out; |
4782 | } | |
4783 | } | |
4784 | ||
f88e1a42 JS |
4785 | backing_fmt = get_option_parameter(param, BLOCK_OPT_BACKING_FMT); |
4786 | if (backing_fmt && backing_fmt->value.s) { | |
96df67d1 SH |
4787 | backing_drv = bdrv_find_format(backing_fmt->value.s); |
4788 | if (!backing_drv) { | |
71c79813 LC |
4789 | error_setg(errp, "Unknown backing file format '%s'", |
4790 | backing_fmt->value.s); | |
f88e1a42 JS |
4791 | goto out; |
4792 | } | |
4793 | } | |
4794 | ||
4795 | // The size for the image must always be specified, with one exception: | |
4796 | // If we are using a backing file, we can obtain the size from there | |
d220894e KW |
4797 | size = get_option_parameter(param, BLOCK_OPT_SIZE); |
4798 | if (size && size->value.n == -1) { | |
f88e1a42 JS |
4799 | if (backing_file && backing_file->value.s) { |
4800 | uint64_t size; | |
f88e1a42 | 4801 | char buf[32]; |
63090dac PB |
4802 | int back_flags; |
4803 | ||
4804 | /* backing files always opened read-only */ | |
4805 | back_flags = | |
4806 | flags & ~(BDRV_O_RDWR | BDRV_O_SNAPSHOT | BDRV_O_NO_BACKING); | |
f88e1a42 | 4807 | |
f88e1a42 JS |
4808 | bs = bdrv_new(""); |
4809 | ||
de9c0cec KW |
4810 | ret = bdrv_open(bs, backing_file->value.s, NULL, back_flags, |
4811 | backing_drv); | |
f88e1a42 | 4812 | if (ret < 0) { |
71c79813 LC |
4813 | error_setg_errno(errp, -ret, "Could not open '%s'", |
4814 | backing_file->value.s); | |
f88e1a42 JS |
4815 | goto out; |
4816 | } | |
4817 | bdrv_get_geometry(bs, &size); | |
4818 | size *= 512; | |
4819 | ||
4820 | snprintf(buf, sizeof(buf), "%" PRId64, size); | |
4821 | set_option_parameter(param, BLOCK_OPT_SIZE, buf); | |
4822 | } else { | |
71c79813 | 4823 | error_setg(errp, "Image creation needs a size parameter"); |
f88e1a42 JS |
4824 | goto out; |
4825 | } | |
4826 | } | |
4827 | ||
f382d43a MR |
4828 | if (!quiet) { |
4829 | printf("Formatting '%s', fmt=%s ", filename, fmt); | |
4830 | print_option_parameters(param); | |
4831 | puts(""); | |
4832 | } | |
f88e1a42 | 4833 | ret = bdrv_create(drv, filename, param); |
f88e1a42 JS |
4834 | if (ret < 0) { |
4835 | if (ret == -ENOTSUP) { | |
71c79813 LC |
4836 | error_setg(errp,"Formatting or formatting option not supported for " |
4837 | "file format '%s'", fmt); | |
f88e1a42 | 4838 | } else if (ret == -EFBIG) { |
71c79813 LC |
4839 | error_setg(errp, "The image size is too large for file format '%s'", |
4840 | fmt); | |
f88e1a42 | 4841 | } else { |
71c79813 LC |
4842 | error_setg(errp, "%s: error while creating %s: %s", filename, fmt, |
4843 | strerror(-ret)); | |
f88e1a42 JS |
4844 | } |
4845 | } | |
4846 | ||
4847 | out: | |
4848 | free_option_parameters(create_options); | |
4849 | free_option_parameters(param); | |
4850 | ||
4851 | if (bs) { | |
4852 | bdrv_delete(bs); | |
4853 | } | |
f88e1a42 | 4854 | } |
85d126f3 SH |
4855 | |
4856 | AioContext *bdrv_get_aio_context(BlockDriverState *bs) | |
4857 | { | |
4858 | /* Currently BlockDriverState always uses the main loop AioContext */ | |
4859 | return qemu_get_aio_context(); | |
4860 | } |