]> git.ipfire.org Git - thirdparty/qemu.git/blame - block.c
block: Add bdrv_co_readv/writev
[thirdparty/qemu.git] / block.c
CommitLineData
fc01f7e7
FB
1/*
2 * QEMU System Emulator block driver
5fafdf24 3 *
fc01f7e7 4 * Copyright (c) 2003 Fabrice Bellard
5fafdf24 5 *
fc01f7e7
FB
6 * Permission is hereby granted, free of charge, to any person obtaining a copy
7 * of this software and associated documentation files (the "Software"), to deal
8 * in the Software without restriction, including without limitation the rights
9 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
10 * copies of the Software, and to permit persons to whom the Software is
11 * furnished to do so, subject to the following conditions:
12 *
13 * The above copyright notice and this permission notice shall be included in
14 * all copies or substantial portions of the Software.
15 *
16 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
17 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
18 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
19 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
20 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
21 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
22 * THE SOFTWARE.
23 */
3990d09a 24#include "config-host.h"
faf07963 25#include "qemu-common.h"
6d519a5f 26#include "trace.h"
376253ec 27#include "monitor.h"
ea2384d3 28#include "block_int.h"
5efa9d5a 29#include "module.h"
d15e5465 30#include "qemu-objects.h"
fc01f7e7 31
71e72a19 32#ifdef CONFIG_BSD
7674e7bf
FB
33#include <sys/types.h>
34#include <sys/stat.h>
35#include <sys/ioctl.h>
72cf2d4f 36#include <sys/queue.h>
c5e97233 37#ifndef __DragonFly__
7674e7bf
FB
38#include <sys/disk.h>
39#endif
c5e97233 40#endif
7674e7bf 41
49dc768d
AL
42#ifdef _WIN32
43#include <windows.h>
44#endif
45
f141eafe
AL
46static BlockDriverAIOCB *bdrv_aio_readv_em(BlockDriverState *bs,
47 int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
c87c0672 48 BlockDriverCompletionFunc *cb, void *opaque);
f141eafe
AL
49static BlockDriverAIOCB *bdrv_aio_writev_em(BlockDriverState *bs,
50 int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
ce1a14dc 51 BlockDriverCompletionFunc *cb, void *opaque);
b2e12bc6
CH
52static BlockDriverAIOCB *bdrv_aio_flush_em(BlockDriverState *bs,
53 BlockDriverCompletionFunc *cb, void *opaque);
016f5cf6
AG
54static BlockDriverAIOCB *bdrv_aio_noop_em(BlockDriverState *bs,
55 BlockDriverCompletionFunc *cb, void *opaque);
5fafdf24 56static int bdrv_read_em(BlockDriverState *bs, int64_t sector_num,
83f64091
FB
57 uint8_t *buf, int nb_sectors);
58static int bdrv_write_em(BlockDriverState *bs, int64_t sector_num,
59 const uint8_t *buf, int nb_sectors);
ec530c81 60
1b7bdbc1
SH
61static QTAILQ_HEAD(, BlockDriverState) bdrv_states =
62 QTAILQ_HEAD_INITIALIZER(bdrv_states);
7ee930d0 63
8a22f02a
SH
64static QLIST_HEAD(, BlockDriver) bdrv_drivers =
65 QLIST_HEAD_INITIALIZER(bdrv_drivers);
ea2384d3 66
f9092b10
MA
67/* The device to use for VM snapshots */
68static BlockDriverState *bs_snapshots;
69
eb852011
MA
70/* If non-zero, use only whitelisted block drivers */
71static int use_bdrv_whitelist;
72
9e0b22f4
SH
73#ifdef _WIN32
74static int is_windows_drive_prefix(const char *filename)
75{
76 return (((filename[0] >= 'a' && filename[0] <= 'z') ||
77 (filename[0] >= 'A' && filename[0] <= 'Z')) &&
78 filename[1] == ':');
79}
80
81int is_windows_drive(const char *filename)
82{
83 if (is_windows_drive_prefix(filename) &&
84 filename[2] == '\0')
85 return 1;
86 if (strstart(filename, "\\\\.\\", NULL) ||
87 strstart(filename, "//./", NULL))
88 return 1;
89 return 0;
90}
91#endif
92
93/* check if the path starts with "<protocol>:" */
94static int path_has_protocol(const char *path)
95{
96#ifdef _WIN32
97 if (is_windows_drive(path) ||
98 is_windows_drive_prefix(path)) {
99 return 0;
100 }
101#endif
102
103 return strchr(path, ':') != NULL;
104}
105
83f64091 106int path_is_absolute(const char *path)
3b0d4f61 107{
83f64091 108 const char *p;
21664424
FB
109#ifdef _WIN32
110 /* specific case for names like: "\\.\d:" */
111 if (*path == '/' || *path == '\\')
112 return 1;
113#endif
83f64091
FB
114 p = strchr(path, ':');
115 if (p)
116 p++;
117 else
118 p = path;
3b9f94e1
FB
119#ifdef _WIN32
120 return (*p == '/' || *p == '\\');
121#else
122 return (*p == '/');
123#endif
3b0d4f61
FB
124}
125
83f64091
FB
126/* if filename is absolute, just copy it to dest. Otherwise, build a
127 path to it by considering it is relative to base_path. URL are
128 supported. */
129void path_combine(char *dest, int dest_size,
130 const char *base_path,
131 const char *filename)
3b0d4f61 132{
83f64091
FB
133 const char *p, *p1;
134 int len;
135
136 if (dest_size <= 0)
137 return;
138 if (path_is_absolute(filename)) {
139 pstrcpy(dest, dest_size, filename);
140 } else {
141 p = strchr(base_path, ':');
142 if (p)
143 p++;
144 else
145 p = base_path;
3b9f94e1
FB
146 p1 = strrchr(base_path, '/');
147#ifdef _WIN32
148 {
149 const char *p2;
150 p2 = strrchr(base_path, '\\');
151 if (!p1 || p2 > p1)
152 p1 = p2;
153 }
154#endif
83f64091
FB
155 if (p1)
156 p1++;
157 else
158 p1 = base_path;
159 if (p1 > p)
160 p = p1;
161 len = p - base_path;
162 if (len > dest_size - 1)
163 len = dest_size - 1;
164 memcpy(dest, base_path, len);
165 dest[len] = '\0';
166 pstrcat(dest, dest_size, filename);
3b0d4f61 167 }
3b0d4f61
FB
168}
169
5efa9d5a 170void bdrv_register(BlockDriver *bdrv)
ea2384d3 171{
f141eafe 172 if (!bdrv->bdrv_aio_readv) {
83f64091 173 /* add AIO emulation layer */
f141eafe
AL
174 bdrv->bdrv_aio_readv = bdrv_aio_readv_em;
175 bdrv->bdrv_aio_writev = bdrv_aio_writev_em;
eda578e5 176 } else if (!bdrv->bdrv_read) {
83f64091
FB
177 /* add synchronous IO emulation layer */
178 bdrv->bdrv_read = bdrv_read_em;
179 bdrv->bdrv_write = bdrv_write_em;
180 }
b2e12bc6
CH
181
182 if (!bdrv->bdrv_aio_flush)
183 bdrv->bdrv_aio_flush = bdrv_aio_flush_em;
184
8a22f02a 185 QLIST_INSERT_HEAD(&bdrv_drivers, bdrv, list);
ea2384d3 186}
b338082b
FB
187
188/* create a new block device (by default it is empty) */
189BlockDriverState *bdrv_new(const char *device_name)
190{
1b7bdbc1 191 BlockDriverState *bs;
b338082b
FB
192
193 bs = qemu_mallocz(sizeof(BlockDriverState));
b338082b 194 pstrcpy(bs->device_name, sizeof(bs->device_name), device_name);
ea2384d3 195 if (device_name[0] != '\0') {
1b7bdbc1 196 QTAILQ_INSERT_TAIL(&bdrv_states, bs, list);
ea2384d3 197 }
b338082b
FB
198 return bs;
199}
200
ea2384d3
FB
201BlockDriver *bdrv_find_format(const char *format_name)
202{
203 BlockDriver *drv1;
8a22f02a
SH
204 QLIST_FOREACH(drv1, &bdrv_drivers, list) {
205 if (!strcmp(drv1->format_name, format_name)) {
ea2384d3 206 return drv1;
8a22f02a 207 }
ea2384d3
FB
208 }
209 return NULL;
210}
211
eb852011
MA
212static int bdrv_is_whitelisted(BlockDriver *drv)
213{
214 static const char *whitelist[] = {
215 CONFIG_BDRV_WHITELIST
216 };
217 const char **p;
218
219 if (!whitelist[0])
220 return 1; /* no whitelist, anything goes */
221
222 for (p = whitelist; *p; p++) {
223 if (!strcmp(drv->format_name, *p)) {
224 return 1;
225 }
226 }
227 return 0;
228}
229
230BlockDriver *bdrv_find_whitelisted_format(const char *format_name)
231{
232 BlockDriver *drv = bdrv_find_format(format_name);
233 return drv && bdrv_is_whitelisted(drv) ? drv : NULL;
234}
235
0e7e1989
KW
236int bdrv_create(BlockDriver *drv, const char* filename,
237 QEMUOptionParameter *options)
ea2384d3
FB
238{
239 if (!drv->bdrv_create)
240 return -ENOTSUP;
0e7e1989
KW
241
242 return drv->bdrv_create(filename, options);
ea2384d3
FB
243}
244
84a12e66
CH
245int bdrv_create_file(const char* filename, QEMUOptionParameter *options)
246{
247 BlockDriver *drv;
248
b50cbabc 249 drv = bdrv_find_protocol(filename);
84a12e66 250 if (drv == NULL) {
16905d71 251 return -ENOENT;
84a12e66
CH
252 }
253
254 return bdrv_create(drv, filename, options);
255}
256
d5249393 257#ifdef _WIN32
95389c86 258void get_tmp_filename(char *filename, int size)
d5249393 259{
3b9f94e1 260 char temp_dir[MAX_PATH];
3b46e624 261
3b9f94e1
FB
262 GetTempPath(MAX_PATH, temp_dir);
263 GetTempFileName(temp_dir, "qem", 0, filename);
d5249393
FB
264}
265#else
95389c86 266void get_tmp_filename(char *filename, int size)
fc01f7e7 267{
67b915a5 268 int fd;
7ccfb2eb 269 const char *tmpdir;
d5249393 270 /* XXX: race condition possible */
0badc1ee
AJ
271 tmpdir = getenv("TMPDIR");
272 if (!tmpdir)
273 tmpdir = "/tmp";
274 snprintf(filename, size, "%s/vl.XXXXXX", tmpdir);
ea2384d3
FB
275 fd = mkstemp(filename);
276 close(fd);
277}
d5249393 278#endif
fc01f7e7 279
84a12e66
CH
280/*
281 * Detect host devices. By convention, /dev/cdrom[N] is always
282 * recognized as a host CDROM.
283 */
284static BlockDriver *find_hdev_driver(const char *filename)
285{
286 int score_max = 0, score;
287 BlockDriver *drv = NULL, *d;
288
289 QLIST_FOREACH(d, &bdrv_drivers, list) {
290 if (d->bdrv_probe_device) {
291 score = d->bdrv_probe_device(filename);
292 if (score > score_max) {
293 score_max = score;
294 drv = d;
295 }
296 }
297 }
298
299 return drv;
300}
301
b50cbabc 302BlockDriver *bdrv_find_protocol(const char *filename)
83f64091
FB
303{
304 BlockDriver *drv1;
305 char protocol[128];
1cec71e3 306 int len;
83f64091 307 const char *p;
19cb3738 308
66f82cee
KW
309 /* TODO Drivers without bdrv_file_open must be specified explicitly */
310
39508e7a
CH
311 /*
312 * XXX(hch): we really should not let host device detection
313 * override an explicit protocol specification, but moving this
314 * later breaks access to device names with colons in them.
315 * Thanks to the brain-dead persistent naming schemes on udev-
316 * based Linux systems those actually are quite common.
317 */
318 drv1 = find_hdev_driver(filename);
319 if (drv1) {
320 return drv1;
321 }
322
9e0b22f4 323 if (!path_has_protocol(filename)) {
39508e7a 324 return bdrv_find_format("file");
84a12e66 325 }
9e0b22f4
SH
326 p = strchr(filename, ':');
327 assert(p != NULL);
1cec71e3
AL
328 len = p - filename;
329 if (len > sizeof(protocol) - 1)
330 len = sizeof(protocol) - 1;
331 memcpy(protocol, filename, len);
332 protocol[len] = '\0';
8a22f02a 333 QLIST_FOREACH(drv1, &bdrv_drivers, list) {
5fafdf24 334 if (drv1->protocol_name &&
8a22f02a 335 !strcmp(drv1->protocol_name, protocol)) {
83f64091 336 return drv1;
8a22f02a 337 }
83f64091
FB
338 }
339 return NULL;
340}
341
c98ac35d 342static int find_image_format(const char *filename, BlockDriver **pdrv)
f3a5d3f8
CH
343{
344 int ret, score, score_max;
345 BlockDriver *drv1, *drv;
346 uint8_t buf[2048];
347 BlockDriverState *bs;
348
f5edb014 349 ret = bdrv_file_open(&bs, filename, 0);
c98ac35d
SW
350 if (ret < 0) {
351 *pdrv = NULL;
352 return ret;
353 }
f8ea0b00 354
08a00559
KW
355 /* Return the raw BlockDriver * to scsi-generic devices or empty drives */
356 if (bs->sg || !bdrv_is_inserted(bs)) {
1a396859 357 bdrv_delete(bs);
c98ac35d
SW
358 drv = bdrv_find_format("raw");
359 if (!drv) {
360 ret = -ENOENT;
361 }
362 *pdrv = drv;
363 return ret;
1a396859 364 }
f8ea0b00 365
83f64091
FB
366 ret = bdrv_pread(bs, 0, buf, sizeof(buf));
367 bdrv_delete(bs);
368 if (ret < 0) {
c98ac35d
SW
369 *pdrv = NULL;
370 return ret;
83f64091
FB
371 }
372
ea2384d3 373 score_max = 0;
84a12e66 374 drv = NULL;
8a22f02a 375 QLIST_FOREACH(drv1, &bdrv_drivers, list) {
83f64091
FB
376 if (drv1->bdrv_probe) {
377 score = drv1->bdrv_probe(buf, ret, filename);
378 if (score > score_max) {
379 score_max = score;
380 drv = drv1;
381 }
0849bf08 382 }
fc01f7e7 383 }
c98ac35d
SW
384 if (!drv) {
385 ret = -ENOENT;
386 }
387 *pdrv = drv;
388 return ret;
ea2384d3
FB
389}
390
51762288
SH
391/**
392 * Set the current 'total_sectors' value
393 */
394static int refresh_total_sectors(BlockDriverState *bs, int64_t hint)
395{
396 BlockDriver *drv = bs->drv;
397
396759ad
NB
398 /* Do not attempt drv->bdrv_getlength() on scsi-generic devices */
399 if (bs->sg)
400 return 0;
401
51762288
SH
402 /* query actual device if possible, otherwise just trust the hint */
403 if (drv->bdrv_getlength) {
404 int64_t length = drv->bdrv_getlength(bs);
405 if (length < 0) {
406 return length;
407 }
408 hint = length >> BDRV_SECTOR_BITS;
409 }
410
411 bs->total_sectors = hint;
412 return 0;
413}
414
57915332
KW
415/*
416 * Common part for opening disk images and files
417 */
418static int bdrv_open_common(BlockDriverState *bs, const char *filename,
419 int flags, BlockDriver *drv)
420{
421 int ret, open_flags;
422
423 assert(drv != NULL);
424
66f82cee 425 bs->file = NULL;
51762288 426 bs->total_sectors = 0;
57915332
KW
427 bs->encrypted = 0;
428 bs->valid_key = 0;
429 bs->open_flags = flags;
430 /* buffer_alignment defaulted to 512, drivers can change this value */
431 bs->buffer_alignment = 512;
432
433 pstrcpy(bs->filename, sizeof(bs->filename), filename);
434
435 if (use_bdrv_whitelist && !bdrv_is_whitelisted(drv)) {
436 return -ENOTSUP;
437 }
438
439 bs->drv = drv;
440 bs->opaque = qemu_mallocz(drv->instance_size);
441
a6599793 442 if (flags & BDRV_O_CACHE_WB)
57915332
KW
443 bs->enable_write_cache = 1;
444
445 /*
446 * Clear flags that are internal to the block layer before opening the
447 * image.
448 */
449 open_flags = flags & ~(BDRV_O_SNAPSHOT | BDRV_O_NO_BACKING);
450
451 /*
ebabb67a 452 * Snapshots should be writable.
57915332
KW
453 */
454 if (bs->is_temporary) {
455 open_flags |= BDRV_O_RDWR;
456 }
457
66f82cee
KW
458 /* Open the image, either directly or using a protocol */
459 if (drv->bdrv_file_open) {
460 ret = drv->bdrv_file_open(bs, filename, open_flags);
461 } else {
462 ret = bdrv_file_open(&bs->file, filename, open_flags);
463 if (ret >= 0) {
464 ret = drv->bdrv_open(bs, open_flags);
465 }
466 }
467
57915332
KW
468 if (ret < 0) {
469 goto free_and_fail;
470 }
471
472 bs->keep_read_only = bs->read_only = !(open_flags & BDRV_O_RDWR);
51762288
SH
473
474 ret = refresh_total_sectors(bs, bs->total_sectors);
475 if (ret < 0) {
476 goto free_and_fail;
57915332 477 }
51762288 478
57915332
KW
479#ifndef _WIN32
480 if (bs->is_temporary) {
481 unlink(filename);
482 }
483#endif
484 return 0;
485
486free_and_fail:
66f82cee
KW
487 if (bs->file) {
488 bdrv_delete(bs->file);
489 bs->file = NULL;
490 }
57915332
KW
491 qemu_free(bs->opaque);
492 bs->opaque = NULL;
493 bs->drv = NULL;
494 return ret;
495}
496
b6ce07aa
KW
497/*
498 * Opens a file using a protocol (file, host_device, nbd, ...)
499 */
83f64091 500int bdrv_file_open(BlockDriverState **pbs, const char *filename, int flags)
ea2384d3 501{
83f64091 502 BlockDriverState *bs;
6db95603 503 BlockDriver *drv;
83f64091
FB
504 int ret;
505
b50cbabc 506 drv = bdrv_find_protocol(filename);
6db95603
CH
507 if (!drv) {
508 return -ENOENT;
509 }
510
83f64091 511 bs = bdrv_new("");
b6ce07aa 512 ret = bdrv_open_common(bs, filename, flags, drv);
83f64091
FB
513 if (ret < 0) {
514 bdrv_delete(bs);
515 return ret;
3b0d4f61 516 }
71d0770c 517 bs->growable = 1;
83f64091
FB
518 *pbs = bs;
519 return 0;
520}
521
b6ce07aa
KW
522/*
523 * Opens a disk image (raw, qcow2, vmdk, ...)
524 */
d6e9098e
KW
525int bdrv_open(BlockDriverState *bs, const char *filename, int flags,
526 BlockDriver *drv)
ea2384d3 527{
b6ce07aa 528 int ret;
712e7874 529
83f64091 530 if (flags & BDRV_O_SNAPSHOT) {
ea2384d3
FB
531 BlockDriverState *bs1;
532 int64_t total_size;
7c96d46e 533 int is_protocol = 0;
91a073a9
KW
534 BlockDriver *bdrv_qcow2;
535 QEMUOptionParameter *options;
b6ce07aa
KW
536 char tmp_filename[PATH_MAX];
537 char backing_filename[PATH_MAX];
3b46e624 538
ea2384d3
FB
539 /* if snapshot, we create a temporary backing file and open it
540 instead of opening 'filename' directly */
33e3963e 541
ea2384d3
FB
542 /* if there is a backing file, use it */
543 bs1 = bdrv_new("");
d6e9098e 544 ret = bdrv_open(bs1, filename, 0, drv);
51d7c00c 545 if (ret < 0) {
ea2384d3 546 bdrv_delete(bs1);
51d7c00c 547 return ret;
ea2384d3 548 }
3e82990b 549 total_size = bdrv_getlength(bs1) & BDRV_SECTOR_MASK;
7c96d46e
AL
550
551 if (bs1->drv && bs1->drv->protocol_name)
552 is_protocol = 1;
553
ea2384d3 554 bdrv_delete(bs1);
3b46e624 555
ea2384d3 556 get_tmp_filename(tmp_filename, sizeof(tmp_filename));
7c96d46e
AL
557
558 /* Real path is meaningless for protocols */
559 if (is_protocol)
560 snprintf(backing_filename, sizeof(backing_filename),
561 "%s", filename);
114cdfa9
KS
562 else if (!realpath(filename, backing_filename))
563 return -errno;
7c96d46e 564
91a073a9
KW
565 bdrv_qcow2 = bdrv_find_format("qcow2");
566 options = parse_option_parameters("", bdrv_qcow2->create_options, NULL);
567
3e82990b 568 set_option_parameter_int(options, BLOCK_OPT_SIZE, total_size);
91a073a9
KW
569 set_option_parameter(options, BLOCK_OPT_BACKING_FILE, backing_filename);
570 if (drv) {
571 set_option_parameter(options, BLOCK_OPT_BACKING_FMT,
572 drv->format_name);
573 }
574
575 ret = bdrv_create(bdrv_qcow2, tmp_filename, options);
d748768c 576 free_option_parameters(options);
51d7c00c
AL
577 if (ret < 0) {
578 return ret;
ea2384d3 579 }
91a073a9 580
ea2384d3 581 filename = tmp_filename;
91a073a9 582 drv = bdrv_qcow2;
ea2384d3
FB
583 bs->is_temporary = 1;
584 }
712e7874 585
b6ce07aa 586 /* Find the right image format driver */
6db95603 587 if (!drv) {
c98ac35d 588 ret = find_image_format(filename, &drv);
51d7c00c 589 }
6987307c 590
51d7c00c 591 if (!drv) {
51d7c00c 592 goto unlink_and_fail;
ea2384d3 593 }
b6ce07aa
KW
594
595 /* Open the image */
596 ret = bdrv_open_common(bs, filename, flags, drv);
597 if (ret < 0) {
6987307c
CH
598 goto unlink_and_fail;
599 }
600
b6ce07aa
KW
601 /* If there is a backing file, use it */
602 if ((flags & BDRV_O_NO_BACKING) == 0 && bs->backing_file[0] != '\0') {
603 char backing_filename[PATH_MAX];
604 int back_flags;
605 BlockDriver *back_drv = NULL;
606
607 bs->backing_hd = bdrv_new("");
df2dbb4a
SH
608
609 if (path_has_protocol(bs->backing_file)) {
610 pstrcpy(backing_filename, sizeof(backing_filename),
611 bs->backing_file);
612 } else {
613 path_combine(backing_filename, sizeof(backing_filename),
614 filename, bs->backing_file);
615 }
616
617 if (bs->backing_format[0] != '\0') {
b6ce07aa 618 back_drv = bdrv_find_format(bs->backing_format);
df2dbb4a 619 }
b6ce07aa
KW
620
621 /* backing files always opened read-only */
622 back_flags =
623 flags & ~(BDRV_O_RDWR | BDRV_O_SNAPSHOT | BDRV_O_NO_BACKING);
624
625 ret = bdrv_open(bs->backing_hd, backing_filename, back_flags, back_drv);
626 if (ret < 0) {
627 bdrv_close(bs);
628 return ret;
629 }
630 if (bs->is_temporary) {
631 bs->backing_hd->keep_read_only = !(flags & BDRV_O_RDWR);
632 } else {
633 /* base image inherits from "parent" */
634 bs->backing_hd->keep_read_only = bs->keep_read_only;
635 }
636 }
637
638 if (!bdrv_key_required(bs)) {
639 /* call the change callback */
640 bs->media_changed = 1;
641 if (bs->change_cb)
db97ee6a 642 bs->change_cb(bs->change_opaque, CHANGE_MEDIA);
b6ce07aa
KW
643 }
644
645 return 0;
646
647unlink_and_fail:
648 if (bs->is_temporary) {
649 unlink(filename);
650 }
651 return ret;
652}
653
fc01f7e7
FB
654void bdrv_close(BlockDriverState *bs)
655{
19cb3738 656 if (bs->drv) {
f9092b10
MA
657 if (bs == bs_snapshots) {
658 bs_snapshots = NULL;
659 }
557df6ac 660 if (bs->backing_hd) {
ea2384d3 661 bdrv_delete(bs->backing_hd);
557df6ac
SH
662 bs->backing_hd = NULL;
663 }
ea2384d3
FB
664 bs->drv->bdrv_close(bs);
665 qemu_free(bs->opaque);
666#ifdef _WIN32
667 if (bs->is_temporary) {
668 unlink(bs->filename);
669 }
67b915a5 670#endif
ea2384d3
FB
671 bs->opaque = NULL;
672 bs->drv = NULL;
b338082b 673
66f82cee
KW
674 if (bs->file != NULL) {
675 bdrv_close(bs->file);
676 }
677
b338082b 678 /* call the change callback */
19cb3738 679 bs->media_changed = 1;
b338082b 680 if (bs->change_cb)
db97ee6a 681 bs->change_cb(bs->change_opaque, CHANGE_MEDIA);
b338082b
FB
682 }
683}
684
2bc93fed
MK
685void bdrv_close_all(void)
686{
687 BlockDriverState *bs;
688
689 QTAILQ_FOREACH(bs, &bdrv_states, list) {
690 bdrv_close(bs);
691 }
692}
693
d22b2f41
RH
694/* make a BlockDriverState anonymous by removing from bdrv_state list.
695 Also, NULL terminate the device_name to prevent double remove */
696void bdrv_make_anon(BlockDriverState *bs)
697{
698 if (bs->device_name[0] != '\0') {
699 QTAILQ_REMOVE(&bdrv_states, bs, list);
700 }
701 bs->device_name[0] = '\0';
702}
703
b338082b
FB
704void bdrv_delete(BlockDriverState *bs)
705{
18846dee
MA
706 assert(!bs->peer);
707
1b7bdbc1 708 /* remove from list, if necessary */
d22b2f41 709 bdrv_make_anon(bs);
34c6f050 710
b338082b 711 bdrv_close(bs);
66f82cee
KW
712 if (bs->file != NULL) {
713 bdrv_delete(bs->file);
714 }
715
f9092b10 716 assert(bs != bs_snapshots);
b338082b 717 qemu_free(bs);
fc01f7e7
FB
718}
719
18846dee
MA
720int bdrv_attach(BlockDriverState *bs, DeviceState *qdev)
721{
722 if (bs->peer) {
723 return -EBUSY;
724 }
725 bs->peer = qdev;
726 return 0;
727}
728
729void bdrv_detach(BlockDriverState *bs, DeviceState *qdev)
730{
731 assert(bs->peer == qdev);
732 bs->peer = NULL;
a19712b0
MA
733 bs->change_cb = NULL;
734 bs->change_opaque = NULL;
18846dee
MA
735}
736
737DeviceState *bdrv_get_attached(BlockDriverState *bs)
738{
739 return bs->peer;
740}
741
e97fc193
AL
742/*
743 * Run consistency checks on an image
744 *
e076f338 745 * Returns 0 if the check could be completed (it doesn't mean that the image is
a1c7273b 746 * free of errors) or -errno when an internal error occurred. The results of the
e076f338 747 * check are stored in res.
e97fc193 748 */
e076f338 749int bdrv_check(BlockDriverState *bs, BdrvCheckResult *res)
e97fc193
AL
750{
751 if (bs->drv->bdrv_check == NULL) {
752 return -ENOTSUP;
753 }
754
e076f338 755 memset(res, 0, sizeof(*res));
9ac228e0 756 return bs->drv->bdrv_check(bs, res);
e97fc193
AL
757}
758
8a426614
KW
759#define COMMIT_BUF_SECTORS 2048
760
33e3963e
FB
761/* commit COW file into the raw image */
762int bdrv_commit(BlockDriverState *bs)
763{
19cb3738 764 BlockDriver *drv = bs->drv;
ee181196 765 BlockDriver *backing_drv;
8a426614
KW
766 int64_t sector, total_sectors;
767 int n, ro, open_flags;
4dca4b63 768 int ret = 0, rw_ret = 0;
8a426614 769 uint8_t *buf;
4dca4b63
NS
770 char filename[1024];
771 BlockDriverState *bs_rw, *bs_ro;
33e3963e 772
19cb3738
FB
773 if (!drv)
774 return -ENOMEDIUM;
4dca4b63
NS
775
776 if (!bs->backing_hd) {
777 return -ENOTSUP;
33e3963e
FB
778 }
779
4dca4b63
NS
780 if (bs->backing_hd->keep_read_only) {
781 return -EACCES;
782 }
ee181196
KW
783
784 backing_drv = bs->backing_hd->drv;
4dca4b63
NS
785 ro = bs->backing_hd->read_only;
786 strncpy(filename, bs->backing_hd->filename, sizeof(filename));
787 open_flags = bs->backing_hd->open_flags;
788
789 if (ro) {
790 /* re-open as RW */
791 bdrv_delete(bs->backing_hd);
792 bs->backing_hd = NULL;
793 bs_rw = bdrv_new("");
ee181196
KW
794 rw_ret = bdrv_open(bs_rw, filename, open_flags | BDRV_O_RDWR,
795 backing_drv);
4dca4b63
NS
796 if (rw_ret < 0) {
797 bdrv_delete(bs_rw);
798 /* try to re-open read-only */
799 bs_ro = bdrv_new("");
ee181196
KW
800 ret = bdrv_open(bs_ro, filename, open_flags & ~BDRV_O_RDWR,
801 backing_drv);
4dca4b63
NS
802 if (ret < 0) {
803 bdrv_delete(bs_ro);
804 /* drive not functional anymore */
805 bs->drv = NULL;
806 return ret;
807 }
808 bs->backing_hd = bs_ro;
809 return rw_ret;
810 }
811 bs->backing_hd = bs_rw;
ea2384d3 812 }
33e3963e 813
6ea44308 814 total_sectors = bdrv_getlength(bs) >> BDRV_SECTOR_BITS;
8a426614
KW
815 buf = qemu_malloc(COMMIT_BUF_SECTORS * BDRV_SECTOR_SIZE);
816
817 for (sector = 0; sector < total_sectors; sector += n) {
818 if (drv->bdrv_is_allocated(bs, sector, COMMIT_BUF_SECTORS, &n)) {
819
820 if (bdrv_read(bs, sector, buf, n) != 0) {
821 ret = -EIO;
822 goto ro_cleanup;
823 }
824
825 if (bdrv_write(bs->backing_hd, sector, buf, n) != 0) {
826 ret = -EIO;
827 goto ro_cleanup;
828 }
ea2384d3 829 }
33e3963e 830 }
95389c86 831
1d44952f
CH
832 if (drv->bdrv_make_empty) {
833 ret = drv->bdrv_make_empty(bs);
834 bdrv_flush(bs);
835 }
95389c86 836
3f5075ae
CH
837 /*
838 * Make sure all data we wrote to the backing device is actually
839 * stable on disk.
840 */
841 if (bs->backing_hd)
842 bdrv_flush(bs->backing_hd);
4dca4b63
NS
843
844ro_cleanup:
8a426614 845 qemu_free(buf);
4dca4b63
NS
846
847 if (ro) {
848 /* re-open as RO */
849 bdrv_delete(bs->backing_hd);
850 bs->backing_hd = NULL;
851 bs_ro = bdrv_new("");
ee181196
KW
852 ret = bdrv_open(bs_ro, filename, open_flags & ~BDRV_O_RDWR,
853 backing_drv);
4dca4b63
NS
854 if (ret < 0) {
855 bdrv_delete(bs_ro);
856 /* drive not functional anymore */
857 bs->drv = NULL;
858 return ret;
859 }
860 bs->backing_hd = bs_ro;
861 bs->backing_hd->keep_read_only = 0;
862 }
863
1d44952f 864 return ret;
33e3963e
FB
865}
866
6ab4b5ab
MA
867void bdrv_commit_all(void)
868{
869 BlockDriverState *bs;
870
871 QTAILQ_FOREACH(bs, &bdrv_states, list) {
872 bdrv_commit(bs);
873 }
874}
875
756e6736
KW
876/*
877 * Return values:
878 * 0 - success
879 * -EINVAL - backing format specified, but no file
880 * -ENOSPC - can't update the backing file because no space is left in the
881 * image file header
882 * -ENOTSUP - format driver doesn't support changing the backing file
883 */
884int bdrv_change_backing_file(BlockDriverState *bs,
885 const char *backing_file, const char *backing_fmt)
886{
887 BlockDriver *drv = bs->drv;
888
889 if (drv->bdrv_change_backing_file != NULL) {
890 return drv->bdrv_change_backing_file(bs, backing_file, backing_fmt);
891 } else {
892 return -ENOTSUP;
893 }
894}
895
71d0770c
AL
896static int bdrv_check_byte_request(BlockDriverState *bs, int64_t offset,
897 size_t size)
898{
899 int64_t len;
900
901 if (!bdrv_is_inserted(bs))
902 return -ENOMEDIUM;
903
904 if (bs->growable)
905 return 0;
906
907 len = bdrv_getlength(bs);
908
fbb7b4e0
KW
909 if (offset < 0)
910 return -EIO;
911
912 if ((offset > len) || (len - offset < size))
71d0770c
AL
913 return -EIO;
914
915 return 0;
916}
917
918static int bdrv_check_request(BlockDriverState *bs, int64_t sector_num,
919 int nb_sectors)
920{
eb5a3165
JS
921 return bdrv_check_byte_request(bs, sector_num * BDRV_SECTOR_SIZE,
922 nb_sectors * BDRV_SECTOR_SIZE);
71d0770c
AL
923}
924
19cb3738 925/* return < 0 if error. See bdrv_write() for the return codes */
5fafdf24 926int bdrv_read(BlockDriverState *bs, int64_t sector_num,
fc01f7e7
FB
927 uint8_t *buf, int nb_sectors)
928{
ea2384d3
FB
929 BlockDriver *drv = bs->drv;
930
19cb3738
FB
931 if (!drv)
932 return -ENOMEDIUM;
71d0770c
AL
933 if (bdrv_check_request(bs, sector_num, nb_sectors))
934 return -EIO;
b338082b 935
eda578e5 936 return drv->bdrv_read(bs, sector_num, buf, nb_sectors);
fc01f7e7
FB
937}
938
7cd1e32a 939static void set_dirty_bitmap(BlockDriverState *bs, int64_t sector_num,
a55eb92c 940 int nb_sectors, int dirty)
7cd1e32a
LS
941{
942 int64_t start, end;
c6d22830 943 unsigned long val, idx, bit;
a55eb92c 944
6ea44308 945 start = sector_num / BDRV_SECTORS_PER_DIRTY_CHUNK;
c6d22830 946 end = (sector_num + nb_sectors - 1) / BDRV_SECTORS_PER_DIRTY_CHUNK;
a55eb92c
JK
947
948 for (; start <= end; start++) {
c6d22830
JK
949 idx = start / (sizeof(unsigned long) * 8);
950 bit = start % (sizeof(unsigned long) * 8);
951 val = bs->dirty_bitmap[idx];
952 if (dirty) {
6d59fec1 953 if (!(val & (1UL << bit))) {
aaa0eb75 954 bs->dirty_count++;
6d59fec1 955 val |= 1UL << bit;
aaa0eb75 956 }
c6d22830 957 } else {
6d59fec1 958 if (val & (1UL << bit)) {
aaa0eb75 959 bs->dirty_count--;
6d59fec1 960 val &= ~(1UL << bit);
aaa0eb75 961 }
c6d22830
JK
962 }
963 bs->dirty_bitmap[idx] = val;
7cd1e32a
LS
964 }
965}
966
5fafdf24 967/* Return < 0 if error. Important errors are:
19cb3738
FB
968 -EIO generic I/O error (may happen for all errors)
969 -ENOMEDIUM No media inserted.
970 -EINVAL Invalid sector number or nb_sectors
971 -EACCES Trying to write a read-only device
972*/
5fafdf24 973int bdrv_write(BlockDriverState *bs, int64_t sector_num,
fc01f7e7
FB
974 const uint8_t *buf, int nb_sectors)
975{
83f64091 976 BlockDriver *drv = bs->drv;
19cb3738
FB
977 if (!bs->drv)
978 return -ENOMEDIUM;
0849bf08 979 if (bs->read_only)
19cb3738 980 return -EACCES;
71d0770c
AL
981 if (bdrv_check_request(bs, sector_num, nb_sectors))
982 return -EIO;
a55eb92c 983
c6d22830 984 if (bs->dirty_bitmap) {
7cd1e32a
LS
985 set_dirty_bitmap(bs, sector_num, nb_sectors, 1);
986 }
a55eb92c 987
294cc35f
KW
988 if (bs->wr_highest_sector < sector_num + nb_sectors - 1) {
989 bs->wr_highest_sector = sector_num + nb_sectors - 1;
990 }
991
42fb2807 992 return drv->bdrv_write(bs, sector_num, buf, nb_sectors);
83f64091
FB
993}
994
eda578e5
AL
995int bdrv_pread(BlockDriverState *bs, int64_t offset,
996 void *buf, int count1)
83f64091 997{
6ea44308 998 uint8_t tmp_buf[BDRV_SECTOR_SIZE];
83f64091
FB
999 int len, nb_sectors, count;
1000 int64_t sector_num;
9a8c4cce 1001 int ret;
83f64091
FB
1002
1003 count = count1;
1004 /* first read to align to sector start */
6ea44308 1005 len = (BDRV_SECTOR_SIZE - offset) & (BDRV_SECTOR_SIZE - 1);
83f64091
FB
1006 if (len > count)
1007 len = count;
6ea44308 1008 sector_num = offset >> BDRV_SECTOR_BITS;
83f64091 1009 if (len > 0) {
9a8c4cce
KW
1010 if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0)
1011 return ret;
6ea44308 1012 memcpy(buf, tmp_buf + (offset & (BDRV_SECTOR_SIZE - 1)), len);
83f64091
FB
1013 count -= len;
1014 if (count == 0)
1015 return count1;
1016 sector_num++;
1017 buf += len;
1018 }
1019
1020 /* read the sectors "in place" */
6ea44308 1021 nb_sectors = count >> BDRV_SECTOR_BITS;
83f64091 1022 if (nb_sectors > 0) {
9a8c4cce
KW
1023 if ((ret = bdrv_read(bs, sector_num, buf, nb_sectors)) < 0)
1024 return ret;
83f64091 1025 sector_num += nb_sectors;
6ea44308 1026 len = nb_sectors << BDRV_SECTOR_BITS;
83f64091
FB
1027 buf += len;
1028 count -= len;
1029 }
1030
1031 /* add data from the last sector */
1032 if (count > 0) {
9a8c4cce
KW
1033 if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0)
1034 return ret;
83f64091
FB
1035 memcpy(buf, tmp_buf, count);
1036 }
1037 return count1;
1038}
1039
eda578e5
AL
1040int bdrv_pwrite(BlockDriverState *bs, int64_t offset,
1041 const void *buf, int count1)
83f64091 1042{
6ea44308 1043 uint8_t tmp_buf[BDRV_SECTOR_SIZE];
83f64091
FB
1044 int len, nb_sectors, count;
1045 int64_t sector_num;
9a8c4cce 1046 int ret;
83f64091
FB
1047
1048 count = count1;
1049 /* first write to align to sector start */
6ea44308 1050 len = (BDRV_SECTOR_SIZE - offset) & (BDRV_SECTOR_SIZE - 1);
83f64091
FB
1051 if (len > count)
1052 len = count;
6ea44308 1053 sector_num = offset >> BDRV_SECTOR_BITS;
83f64091 1054 if (len > 0) {
9a8c4cce
KW
1055 if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0)
1056 return ret;
6ea44308 1057 memcpy(tmp_buf + (offset & (BDRV_SECTOR_SIZE - 1)), buf, len);
9a8c4cce
KW
1058 if ((ret = bdrv_write(bs, sector_num, tmp_buf, 1)) < 0)
1059 return ret;
83f64091
FB
1060 count -= len;
1061 if (count == 0)
1062 return count1;
1063 sector_num++;
1064 buf += len;
1065 }
1066
1067 /* write the sectors "in place" */
6ea44308 1068 nb_sectors = count >> BDRV_SECTOR_BITS;
83f64091 1069 if (nb_sectors > 0) {
9a8c4cce
KW
1070 if ((ret = bdrv_write(bs, sector_num, buf, nb_sectors)) < 0)
1071 return ret;
83f64091 1072 sector_num += nb_sectors;
6ea44308 1073 len = nb_sectors << BDRV_SECTOR_BITS;
83f64091
FB
1074 buf += len;
1075 count -= len;
1076 }
1077
1078 /* add data from the last sector */
1079 if (count > 0) {
9a8c4cce
KW
1080 if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0)
1081 return ret;
83f64091 1082 memcpy(tmp_buf, buf, count);
9a8c4cce
KW
1083 if ((ret = bdrv_write(bs, sector_num, tmp_buf, 1)) < 0)
1084 return ret;
83f64091
FB
1085 }
1086 return count1;
1087}
83f64091 1088
f08145fe
KW
1089/*
1090 * Writes to the file and ensures that no writes are reordered across this
1091 * request (acts as a barrier)
1092 *
1093 * Returns 0 on success, -errno in error cases.
1094 */
1095int bdrv_pwrite_sync(BlockDriverState *bs, int64_t offset,
1096 const void *buf, int count)
1097{
1098 int ret;
1099
1100 ret = bdrv_pwrite(bs, offset, buf, count);
1101 if (ret < 0) {
1102 return ret;
1103 }
1104
1105 /* No flush needed for cache=writethrough, it uses O_DSYNC */
1106 if ((bs->open_flags & BDRV_O_CACHE_MASK) != 0) {
1107 bdrv_flush(bs);
1108 }
1109
1110 return 0;
1111}
1112
da1fa91d
KW
1113int coroutine_fn bdrv_co_readv(BlockDriverState *bs, int64_t sector_num,
1114 int nb_sectors, QEMUIOVector *qiov)
1115{
1116 BlockDriver *drv = bs->drv;
1117
1118 trace_bdrv_co_readv(bs, sector_num, nb_sectors);
1119
1120 if (!drv) {
1121 return -ENOMEDIUM;
1122 }
1123 if (bdrv_check_request(bs, sector_num, nb_sectors)) {
1124 return -EIO;
1125 }
1126
1127 return drv->bdrv_co_readv(bs, sector_num, nb_sectors, qiov);
1128}
1129
1130int coroutine_fn bdrv_co_writev(BlockDriverState *bs, int64_t sector_num,
1131 int nb_sectors, QEMUIOVector *qiov)
1132{
1133 BlockDriver *drv = bs->drv;
1134
1135 trace_bdrv_co_writev(bs, sector_num, nb_sectors);
1136
1137 if (!bs->drv) {
1138 return -ENOMEDIUM;
1139 }
1140 if (bs->read_only) {
1141 return -EACCES;
1142 }
1143 if (bdrv_check_request(bs, sector_num, nb_sectors)) {
1144 return -EIO;
1145 }
1146
1147 if (bs->dirty_bitmap) {
1148 set_dirty_bitmap(bs, sector_num, nb_sectors, 1);
1149 }
1150
1151 if (bs->wr_highest_sector < sector_num + nb_sectors - 1) {
1152 bs->wr_highest_sector = sector_num + nb_sectors - 1;
1153 }
1154
1155 return drv->bdrv_co_writev(bs, sector_num, nb_sectors, qiov);
1156}
1157
83f64091
FB
1158/**
1159 * Truncate file to 'offset' bytes (needed only for file protocols)
1160 */
1161int bdrv_truncate(BlockDriverState *bs, int64_t offset)
1162{
1163 BlockDriver *drv = bs->drv;
51762288 1164 int ret;
83f64091 1165 if (!drv)
19cb3738 1166 return -ENOMEDIUM;
83f64091
FB
1167 if (!drv->bdrv_truncate)
1168 return -ENOTSUP;
59f2689d
NS
1169 if (bs->read_only)
1170 return -EACCES;
8591675f
MT
1171 if (bdrv_in_use(bs))
1172 return -EBUSY;
51762288
SH
1173 ret = drv->bdrv_truncate(bs, offset);
1174 if (ret == 0) {
1175 ret = refresh_total_sectors(bs, offset >> BDRV_SECTOR_BITS);
db97ee6a
CH
1176 if (bs->change_cb) {
1177 bs->change_cb(bs->change_opaque, CHANGE_SIZE);
1178 }
51762288
SH
1179 }
1180 return ret;
83f64091
FB
1181}
1182
4a1d5e1f
FZ
1183/**
1184 * Length of a allocated file in bytes. Sparse files are counted by actual
1185 * allocated space. Return < 0 if error or unknown.
1186 */
1187int64_t bdrv_get_allocated_file_size(BlockDriverState *bs)
1188{
1189 BlockDriver *drv = bs->drv;
1190 if (!drv) {
1191 return -ENOMEDIUM;
1192 }
1193 if (drv->bdrv_get_allocated_file_size) {
1194 return drv->bdrv_get_allocated_file_size(bs);
1195 }
1196 if (bs->file) {
1197 return bdrv_get_allocated_file_size(bs->file);
1198 }
1199 return -ENOTSUP;
1200}
1201
83f64091
FB
1202/**
1203 * Length of a file in bytes. Return < 0 if error or unknown.
1204 */
1205int64_t bdrv_getlength(BlockDriverState *bs)
1206{
1207 BlockDriver *drv = bs->drv;
1208 if (!drv)
19cb3738 1209 return -ENOMEDIUM;
51762288 1210
46a4e4e6
SH
1211 if (bs->growable || bs->removable) {
1212 if (drv->bdrv_getlength) {
1213 return drv->bdrv_getlength(bs);
1214 }
83f64091 1215 }
46a4e4e6 1216 return bs->total_sectors * BDRV_SECTOR_SIZE;
fc01f7e7
FB
1217}
1218
19cb3738 1219/* return 0 as number of sectors if no device present or error */
96b8f136 1220void bdrv_get_geometry(BlockDriverState *bs, uint64_t *nb_sectors_ptr)
fc01f7e7 1221{
19cb3738
FB
1222 int64_t length;
1223 length = bdrv_getlength(bs);
1224 if (length < 0)
1225 length = 0;
1226 else
6ea44308 1227 length = length >> BDRV_SECTOR_BITS;
19cb3738 1228 *nb_sectors_ptr = length;
fc01f7e7 1229}
cf98951b 1230
f3d54fc4
AL
1231struct partition {
1232 uint8_t boot_ind; /* 0x80 - active */
1233 uint8_t head; /* starting head */
1234 uint8_t sector; /* starting sector */
1235 uint8_t cyl; /* starting cylinder */
1236 uint8_t sys_ind; /* What partition type */
1237 uint8_t end_head; /* end head */
1238 uint8_t end_sector; /* end sector */
1239 uint8_t end_cyl; /* end cylinder */
1240 uint32_t start_sect; /* starting sector counting from 0 */
1241 uint32_t nr_sects; /* nr of sectors in partition */
1242} __attribute__((packed));
1243
1244/* try to guess the disk logical geometry from the MSDOS partition table. Return 0 if OK, -1 if could not guess */
1245static int guess_disk_lchs(BlockDriverState *bs,
1246 int *pcylinders, int *pheads, int *psectors)
1247{
eb5a3165 1248 uint8_t buf[BDRV_SECTOR_SIZE];
f3d54fc4
AL
1249 int ret, i, heads, sectors, cylinders;
1250 struct partition *p;
1251 uint32_t nr_sects;
a38131b6 1252 uint64_t nb_sectors;
f3d54fc4
AL
1253
1254 bdrv_get_geometry(bs, &nb_sectors);
1255
1256 ret = bdrv_read(bs, 0, buf, 1);
1257 if (ret < 0)
1258 return -1;
1259 /* test msdos magic */
1260 if (buf[510] != 0x55 || buf[511] != 0xaa)
1261 return -1;
1262 for(i = 0; i < 4; i++) {
1263 p = ((struct partition *)(buf + 0x1be)) + i;
1264 nr_sects = le32_to_cpu(p->nr_sects);
1265 if (nr_sects && p->end_head) {
1266 /* We make the assumption that the partition terminates on
1267 a cylinder boundary */
1268 heads = p->end_head + 1;
1269 sectors = p->end_sector & 63;
1270 if (sectors == 0)
1271 continue;
1272 cylinders = nb_sectors / (heads * sectors);
1273 if (cylinders < 1 || cylinders > 16383)
1274 continue;
1275 *pheads = heads;
1276 *psectors = sectors;
1277 *pcylinders = cylinders;
1278#if 0
1279 printf("guessed geometry: LCHS=%d %d %d\n",
1280 cylinders, heads, sectors);
1281#endif
1282 return 0;
1283 }
1284 }
1285 return -1;
1286}
1287
1288void bdrv_guess_geometry(BlockDriverState *bs, int *pcyls, int *pheads, int *psecs)
1289{
1290 int translation, lba_detected = 0;
1291 int cylinders, heads, secs;
a38131b6 1292 uint64_t nb_sectors;
f3d54fc4
AL
1293
1294 /* if a geometry hint is available, use it */
1295 bdrv_get_geometry(bs, &nb_sectors);
1296 bdrv_get_geometry_hint(bs, &cylinders, &heads, &secs);
1297 translation = bdrv_get_translation_hint(bs);
1298 if (cylinders != 0) {
1299 *pcyls = cylinders;
1300 *pheads = heads;
1301 *psecs = secs;
1302 } else {
1303 if (guess_disk_lchs(bs, &cylinders, &heads, &secs) == 0) {
1304 if (heads > 16) {
1305 /* if heads > 16, it means that a BIOS LBA
1306 translation was active, so the default
1307 hardware geometry is OK */
1308 lba_detected = 1;
1309 goto default_geometry;
1310 } else {
1311 *pcyls = cylinders;
1312 *pheads = heads;
1313 *psecs = secs;
1314 /* disable any translation to be in sync with
1315 the logical geometry */
1316 if (translation == BIOS_ATA_TRANSLATION_AUTO) {
1317 bdrv_set_translation_hint(bs,
1318 BIOS_ATA_TRANSLATION_NONE);
1319 }
1320 }
1321 } else {
1322 default_geometry:
1323 /* if no geometry, use a standard physical disk geometry */
1324 cylinders = nb_sectors / (16 * 63);
1325
1326 if (cylinders > 16383)
1327 cylinders = 16383;
1328 else if (cylinders < 2)
1329 cylinders = 2;
1330 *pcyls = cylinders;
1331 *pheads = 16;
1332 *psecs = 63;
1333 if ((lba_detected == 1) && (translation == BIOS_ATA_TRANSLATION_AUTO)) {
1334 if ((*pcyls * *pheads) <= 131072) {
1335 bdrv_set_translation_hint(bs,
1336 BIOS_ATA_TRANSLATION_LARGE);
1337 } else {
1338 bdrv_set_translation_hint(bs,
1339 BIOS_ATA_TRANSLATION_LBA);
1340 }
1341 }
1342 }
1343 bdrv_set_geometry_hint(bs, *pcyls, *pheads, *psecs);
1344 }
1345}
1346
5fafdf24 1347void bdrv_set_geometry_hint(BlockDriverState *bs,
b338082b
FB
1348 int cyls, int heads, int secs)
1349{
1350 bs->cyls = cyls;
1351 bs->heads = heads;
1352 bs->secs = secs;
1353}
1354
46d4767d
FB
1355void bdrv_set_translation_hint(BlockDriverState *bs, int translation)
1356{
1357 bs->translation = translation;
1358}
1359
5fafdf24 1360void bdrv_get_geometry_hint(BlockDriverState *bs,
b338082b
FB
1361 int *pcyls, int *pheads, int *psecs)
1362{
1363 *pcyls = bs->cyls;
1364 *pheads = bs->heads;
1365 *psecs = bs->secs;
1366}
1367
5bbdbb46
BS
1368/* Recognize floppy formats */
1369typedef struct FDFormat {
1370 FDriveType drive;
1371 uint8_t last_sect;
1372 uint8_t max_track;
1373 uint8_t max_head;
1374} FDFormat;
1375
1376static const FDFormat fd_formats[] = {
1377 /* First entry is default format */
1378 /* 1.44 MB 3"1/2 floppy disks */
1379 { FDRIVE_DRV_144, 18, 80, 1, },
1380 { FDRIVE_DRV_144, 20, 80, 1, },
1381 { FDRIVE_DRV_144, 21, 80, 1, },
1382 { FDRIVE_DRV_144, 21, 82, 1, },
1383 { FDRIVE_DRV_144, 21, 83, 1, },
1384 { FDRIVE_DRV_144, 22, 80, 1, },
1385 { FDRIVE_DRV_144, 23, 80, 1, },
1386 { FDRIVE_DRV_144, 24, 80, 1, },
1387 /* 2.88 MB 3"1/2 floppy disks */
1388 { FDRIVE_DRV_288, 36, 80, 1, },
1389 { FDRIVE_DRV_288, 39, 80, 1, },
1390 { FDRIVE_DRV_288, 40, 80, 1, },
1391 { FDRIVE_DRV_288, 44, 80, 1, },
1392 { FDRIVE_DRV_288, 48, 80, 1, },
1393 /* 720 kB 3"1/2 floppy disks */
1394 { FDRIVE_DRV_144, 9, 80, 1, },
1395 { FDRIVE_DRV_144, 10, 80, 1, },
1396 { FDRIVE_DRV_144, 10, 82, 1, },
1397 { FDRIVE_DRV_144, 10, 83, 1, },
1398 { FDRIVE_DRV_144, 13, 80, 1, },
1399 { FDRIVE_DRV_144, 14, 80, 1, },
1400 /* 1.2 MB 5"1/4 floppy disks */
1401 { FDRIVE_DRV_120, 15, 80, 1, },
1402 { FDRIVE_DRV_120, 18, 80, 1, },
1403 { FDRIVE_DRV_120, 18, 82, 1, },
1404 { FDRIVE_DRV_120, 18, 83, 1, },
1405 { FDRIVE_DRV_120, 20, 80, 1, },
1406 /* 720 kB 5"1/4 floppy disks */
1407 { FDRIVE_DRV_120, 9, 80, 1, },
1408 { FDRIVE_DRV_120, 11, 80, 1, },
1409 /* 360 kB 5"1/4 floppy disks */
1410 { FDRIVE_DRV_120, 9, 40, 1, },
1411 { FDRIVE_DRV_120, 9, 40, 0, },
1412 { FDRIVE_DRV_120, 10, 41, 1, },
1413 { FDRIVE_DRV_120, 10, 42, 1, },
1414 /* 320 kB 5"1/4 floppy disks */
1415 { FDRIVE_DRV_120, 8, 40, 1, },
1416 { FDRIVE_DRV_120, 8, 40, 0, },
1417 /* 360 kB must match 5"1/4 better than 3"1/2... */
1418 { FDRIVE_DRV_144, 9, 80, 0, },
1419 /* end */
1420 { FDRIVE_DRV_NONE, -1, -1, 0, },
1421};
1422
1423void bdrv_get_floppy_geometry_hint(BlockDriverState *bs, int *nb_heads,
1424 int *max_track, int *last_sect,
1425 FDriveType drive_in, FDriveType *drive)
1426{
1427 const FDFormat *parse;
1428 uint64_t nb_sectors, size;
1429 int i, first_match, match;
1430
1431 bdrv_get_geometry_hint(bs, nb_heads, max_track, last_sect);
1432 if (*nb_heads != 0 && *max_track != 0 && *last_sect != 0) {
1433 /* User defined disk */
1434 } else {
1435 bdrv_get_geometry(bs, &nb_sectors);
1436 match = -1;
1437 first_match = -1;
1438 for (i = 0; ; i++) {
1439 parse = &fd_formats[i];
1440 if (parse->drive == FDRIVE_DRV_NONE) {
1441 break;
1442 }
1443 if (drive_in == parse->drive ||
1444 drive_in == FDRIVE_DRV_NONE) {
1445 size = (parse->max_head + 1) * parse->max_track *
1446 parse->last_sect;
1447 if (nb_sectors == size) {
1448 match = i;
1449 break;
1450 }
1451 if (first_match == -1) {
1452 first_match = i;
1453 }
1454 }
1455 }
1456 if (match == -1) {
1457 if (first_match == -1) {
1458 match = 1;
1459 } else {
1460 match = first_match;
1461 }
1462 parse = &fd_formats[match];
1463 }
1464 *nb_heads = parse->max_head + 1;
1465 *max_track = parse->max_track;
1466 *last_sect = parse->last_sect;
1467 *drive = parse->drive;
1468 }
1469}
1470
46d4767d
FB
1471int bdrv_get_translation_hint(BlockDriverState *bs)
1472{
1473 return bs->translation;
1474}
1475
abd7f68d
MA
1476void bdrv_set_on_error(BlockDriverState *bs, BlockErrorAction on_read_error,
1477 BlockErrorAction on_write_error)
1478{
1479 bs->on_read_error = on_read_error;
1480 bs->on_write_error = on_write_error;
1481}
1482
1483BlockErrorAction bdrv_get_on_error(BlockDriverState *bs, int is_read)
1484{
1485 return is_read ? bs->on_read_error : bs->on_write_error;
1486}
1487
7d0d6950
MA
1488void bdrv_set_removable(BlockDriverState *bs, int removable)
1489{
1490 bs->removable = removable;
1491 if (removable && bs == bs_snapshots) {
1492 bs_snapshots = NULL;
1493 }
1494}
1495
b338082b
FB
1496int bdrv_is_removable(BlockDriverState *bs)
1497{
1498 return bs->removable;
1499}
1500
1501int bdrv_is_read_only(BlockDriverState *bs)
1502{
1503 return bs->read_only;
1504}
1505
985a03b0
TS
1506int bdrv_is_sg(BlockDriverState *bs)
1507{
1508 return bs->sg;
1509}
1510
e900a7b7
CH
1511int bdrv_enable_write_cache(BlockDriverState *bs)
1512{
1513 return bs->enable_write_cache;
1514}
1515
19cb3738 1516/* XXX: no longer used */
5fafdf24 1517void bdrv_set_change_cb(BlockDriverState *bs,
db97ee6a
CH
1518 void (*change_cb)(void *opaque, int reason),
1519 void *opaque)
b338082b
FB
1520{
1521 bs->change_cb = change_cb;
1522 bs->change_opaque = opaque;
1523}
1524
ea2384d3
FB
1525int bdrv_is_encrypted(BlockDriverState *bs)
1526{
1527 if (bs->backing_hd && bs->backing_hd->encrypted)
1528 return 1;
1529 return bs->encrypted;
1530}
1531
c0f4ce77
AL
1532int bdrv_key_required(BlockDriverState *bs)
1533{
1534 BlockDriverState *backing_hd = bs->backing_hd;
1535
1536 if (backing_hd && backing_hd->encrypted && !backing_hd->valid_key)
1537 return 1;
1538 return (bs->encrypted && !bs->valid_key);
1539}
1540
ea2384d3
FB
1541int bdrv_set_key(BlockDriverState *bs, const char *key)
1542{
1543 int ret;
1544 if (bs->backing_hd && bs->backing_hd->encrypted) {
1545 ret = bdrv_set_key(bs->backing_hd, key);
1546 if (ret < 0)
1547 return ret;
1548 if (!bs->encrypted)
1549 return 0;
1550 }
fd04a2ae
SH
1551 if (!bs->encrypted) {
1552 return -EINVAL;
1553 } else if (!bs->drv || !bs->drv->bdrv_set_key) {
1554 return -ENOMEDIUM;
1555 }
c0f4ce77 1556 ret = bs->drv->bdrv_set_key(bs, key);
bb5fc20f
AL
1557 if (ret < 0) {
1558 bs->valid_key = 0;
1559 } else if (!bs->valid_key) {
1560 bs->valid_key = 1;
1561 /* call the change callback now, we skipped it on open */
1562 bs->media_changed = 1;
1563 if (bs->change_cb)
db97ee6a 1564 bs->change_cb(bs->change_opaque, CHANGE_MEDIA);
bb5fc20f 1565 }
c0f4ce77 1566 return ret;
ea2384d3
FB
1567}
1568
1569void bdrv_get_format(BlockDriverState *bs, char *buf, int buf_size)
1570{
19cb3738 1571 if (!bs->drv) {
ea2384d3
FB
1572 buf[0] = '\0';
1573 } else {
1574 pstrcpy(buf, buf_size, bs->drv->format_name);
1575 }
1576}
1577
5fafdf24 1578void bdrv_iterate_format(void (*it)(void *opaque, const char *name),
ea2384d3
FB
1579 void *opaque)
1580{
1581 BlockDriver *drv;
1582
8a22f02a 1583 QLIST_FOREACH(drv, &bdrv_drivers, list) {
ea2384d3
FB
1584 it(opaque, drv->format_name);
1585 }
1586}
1587
b338082b
FB
1588BlockDriverState *bdrv_find(const char *name)
1589{
1590 BlockDriverState *bs;
1591
1b7bdbc1
SH
1592 QTAILQ_FOREACH(bs, &bdrv_states, list) {
1593 if (!strcmp(name, bs->device_name)) {
b338082b 1594 return bs;
1b7bdbc1 1595 }
b338082b
FB
1596 }
1597 return NULL;
1598}
1599
2f399b0a
MA
1600BlockDriverState *bdrv_next(BlockDriverState *bs)
1601{
1602 if (!bs) {
1603 return QTAILQ_FIRST(&bdrv_states);
1604 }
1605 return QTAILQ_NEXT(bs, list);
1606}
1607
51de9760 1608void bdrv_iterate(void (*it)(void *opaque, BlockDriverState *bs), void *opaque)
81d0912d
FB
1609{
1610 BlockDriverState *bs;
1611
1b7bdbc1 1612 QTAILQ_FOREACH(bs, &bdrv_states, list) {
51de9760 1613 it(opaque, bs);
81d0912d
FB
1614 }
1615}
1616
ea2384d3
FB
1617const char *bdrv_get_device_name(BlockDriverState *bs)
1618{
1619 return bs->device_name;
1620}
1621
205ef796 1622int bdrv_flush(BlockDriverState *bs)
7a6cba61 1623{
016f5cf6 1624 if (bs->open_flags & BDRV_O_NO_FLUSH) {
205ef796
KW
1625 return 0;
1626 }
1627
1628 if (bs->drv && bs->drv->bdrv_flush) {
1629 return bs->drv->bdrv_flush(bs);
016f5cf6
AG
1630 }
1631
205ef796
KW
1632 /*
1633 * Some block drivers always operate in either writethrough or unsafe mode
1634 * and don't support bdrv_flush therefore. Usually qemu doesn't know how
1635 * the server works (because the behaviour is hardcoded or depends on
1636 * server-side configuration), so we can't ensure that everything is safe
1637 * on disk. Returning an error doesn't work because that would break guests
1638 * even if the server operates in writethrough mode.
1639 *
1640 * Let's hope the user knows what he's doing.
1641 */
1642 return 0;
7a6cba61
PB
1643}
1644
c6ca28d6
AL
1645void bdrv_flush_all(void)
1646{
1647 BlockDriverState *bs;
1648
1b7bdbc1
SH
1649 QTAILQ_FOREACH(bs, &bdrv_states, list) {
1650 if (bs->drv && !bdrv_is_read_only(bs) &&
1651 (!bdrv_is_removable(bs) || bdrv_is_inserted(bs))) {
c6ca28d6 1652 bdrv_flush(bs);
1b7bdbc1
SH
1653 }
1654 }
c6ca28d6
AL
1655}
1656
f2feebbd
KW
1657int bdrv_has_zero_init(BlockDriverState *bs)
1658{
1659 assert(bs->drv);
1660
336c1c12
KW
1661 if (bs->drv->bdrv_has_zero_init) {
1662 return bs->drv->bdrv_has_zero_init(bs);
f2feebbd
KW
1663 }
1664
1665 return 1;
1666}
1667
bb8bf76f
CH
1668int bdrv_discard(BlockDriverState *bs, int64_t sector_num, int nb_sectors)
1669{
1670 if (!bs->drv) {
1671 return -ENOMEDIUM;
1672 }
1673 if (!bs->drv->bdrv_discard) {
1674 return 0;
1675 }
1676 return bs->drv->bdrv_discard(bs, sector_num, nb_sectors);
1677}
1678
f58c7b35
TS
1679/*
1680 * Returns true iff the specified sector is present in the disk image. Drivers
1681 * not implementing the functionality are assumed to not support backing files,
1682 * hence all their sectors are reported as allocated.
1683 *
1684 * 'pnum' is set to the number of sectors (including and immediately following
1685 * the specified sector) that are known to be in the same
1686 * allocated/unallocated state.
1687 *
1688 * 'nb_sectors' is the max value 'pnum' should be set to.
1689 */
1690int bdrv_is_allocated(BlockDriverState *bs, int64_t sector_num, int nb_sectors,
1691 int *pnum)
1692{
1693 int64_t n;
1694 if (!bs->drv->bdrv_is_allocated) {
1695 if (sector_num >= bs->total_sectors) {
1696 *pnum = 0;
1697 return 0;
1698 }
1699 n = bs->total_sectors - sector_num;
1700 *pnum = (n < nb_sectors) ? (n) : (nb_sectors);
1701 return 1;
1702 }
1703 return bs->drv->bdrv_is_allocated(bs, sector_num, nb_sectors, pnum);
1704}
1705
2582bfed
LC
1706void bdrv_mon_event(const BlockDriverState *bdrv,
1707 BlockMonEventAction action, int is_read)
1708{
1709 QObject *data;
1710 const char *action_str;
1711
1712 switch (action) {
1713 case BDRV_ACTION_REPORT:
1714 action_str = "report";
1715 break;
1716 case BDRV_ACTION_IGNORE:
1717 action_str = "ignore";
1718 break;
1719 case BDRV_ACTION_STOP:
1720 action_str = "stop";
1721 break;
1722 default:
1723 abort();
1724 }
1725
1726 data = qobject_from_jsonf("{ 'device': %s, 'action': %s, 'operation': %s }",
1727 bdrv->device_name,
1728 action_str,
1729 is_read ? "read" : "write");
1730 monitor_protocol_event(QEVENT_BLOCK_IO_ERROR, data);
1731
1732 qobject_decref(data);
1733}
1734
d15e5465 1735static void bdrv_print_dict(QObject *obj, void *opaque)
b338082b 1736{
d15e5465
LC
1737 QDict *bs_dict;
1738 Monitor *mon = opaque;
1739
1740 bs_dict = qobject_to_qdict(obj);
1741
d8aeeb31 1742 monitor_printf(mon, "%s: removable=%d",
d15e5465 1743 qdict_get_str(bs_dict, "device"),
d15e5465
LC
1744 qdict_get_bool(bs_dict, "removable"));
1745
1746 if (qdict_get_bool(bs_dict, "removable")) {
1747 monitor_printf(mon, " locked=%d", qdict_get_bool(bs_dict, "locked"));
1748 }
1749
1750 if (qdict_haskey(bs_dict, "inserted")) {
1751 QDict *qdict = qobject_to_qdict(qdict_get(bs_dict, "inserted"));
1752
1753 monitor_printf(mon, " file=");
1754 monitor_print_filename(mon, qdict_get_str(qdict, "file"));
1755 if (qdict_haskey(qdict, "backing_file")) {
1756 monitor_printf(mon, " backing_file=");
1757 monitor_print_filename(mon, qdict_get_str(qdict, "backing_file"));
1758 }
1759 monitor_printf(mon, " ro=%d drv=%s encrypted=%d",
1760 qdict_get_bool(qdict, "ro"),
1761 qdict_get_str(qdict, "drv"),
1762 qdict_get_bool(qdict, "encrypted"));
1763 } else {
1764 monitor_printf(mon, " [not inserted]");
1765 }
1766
1767 monitor_printf(mon, "\n");
1768}
1769
1770void bdrv_info_print(Monitor *mon, const QObject *data)
1771{
1772 qlist_iter(qobject_to_qlist(data), bdrv_print_dict, mon);
1773}
1774
d15e5465
LC
1775void bdrv_info(Monitor *mon, QObject **ret_data)
1776{
1777 QList *bs_list;
b338082b
FB
1778 BlockDriverState *bs;
1779
d15e5465
LC
1780 bs_list = qlist_new();
1781
1b7bdbc1 1782 QTAILQ_FOREACH(bs, &bdrv_states, list) {
d15e5465 1783 QObject *bs_obj;
d15e5465 1784
d8aeeb31 1785 bs_obj = qobject_from_jsonf("{ 'device': %s, 'type': 'unknown', "
d15e5465 1786 "'removable': %i, 'locked': %i }",
d8aeeb31 1787 bs->device_name, bs->removable,
d15e5465 1788 bs->locked);
d15e5465 1789
19cb3738 1790 if (bs->drv) {
d15e5465
LC
1791 QObject *obj;
1792 QDict *bs_dict = qobject_to_qdict(bs_obj);
1793
1794 obj = qobject_from_jsonf("{ 'file': %s, 'ro': %i, 'drv': %s, "
1795 "'encrypted': %i }",
1796 bs->filename, bs->read_only,
1797 bs->drv->format_name,
1798 bdrv_is_encrypted(bs));
fef30743 1799 if (bs->backing_file[0] != '\0') {
d15e5465
LC
1800 QDict *qdict = qobject_to_qdict(obj);
1801 qdict_put(qdict, "backing_file",
1802 qstring_from_str(bs->backing_file));
376253ec 1803 }
d15e5465
LC
1804
1805 qdict_put_obj(bs_dict, "inserted", obj);
b338082b 1806 }
d15e5465 1807 qlist_append_obj(bs_list, bs_obj);
b338082b 1808 }
d15e5465
LC
1809
1810 *ret_data = QOBJECT(bs_list);
b338082b 1811}
a36e69dd 1812
218a536a 1813static void bdrv_stats_iter(QObject *data, void *opaque)
a36e69dd 1814{
218a536a
LC
1815 QDict *qdict;
1816 Monitor *mon = opaque;
1817
1818 qdict = qobject_to_qdict(data);
1819 monitor_printf(mon, "%s:", qdict_get_str(qdict, "device"));
1820
1821 qdict = qobject_to_qdict(qdict_get(qdict, "stats"));
1822 monitor_printf(mon, " rd_bytes=%" PRId64
1823 " wr_bytes=%" PRId64
1824 " rd_operations=%" PRId64
1825 " wr_operations=%" PRId64
1826 "\n",
1827 qdict_get_int(qdict, "rd_bytes"),
1828 qdict_get_int(qdict, "wr_bytes"),
1829 qdict_get_int(qdict, "rd_operations"),
1830 qdict_get_int(qdict, "wr_operations"));
1831}
1832
1833void bdrv_stats_print(Monitor *mon, const QObject *data)
1834{
1835 qlist_iter(qobject_to_qlist(data), bdrv_stats_iter, mon);
1836}
1837
294cc35f
KW
1838static QObject* bdrv_info_stats_bs(BlockDriverState *bs)
1839{
1840 QObject *res;
1841 QDict *dict;
1842
1843 res = qobject_from_jsonf("{ 'stats': {"
1844 "'rd_bytes': %" PRId64 ","
1845 "'wr_bytes': %" PRId64 ","
1846 "'rd_operations': %" PRId64 ","
1847 "'wr_operations': %" PRId64 ","
1848 "'wr_highest_offset': %" PRId64
1849 "} }",
1850 bs->rd_bytes, bs->wr_bytes,
1851 bs->rd_ops, bs->wr_ops,
5ffbbc67
BS
1852 bs->wr_highest_sector *
1853 (uint64_t)BDRV_SECTOR_SIZE);
294cc35f
KW
1854 dict = qobject_to_qdict(res);
1855
1856 if (*bs->device_name) {
1857 qdict_put(dict, "device", qstring_from_str(bs->device_name));
1858 }
1859
1860 if (bs->file) {
1861 QObject *parent = bdrv_info_stats_bs(bs->file);
1862 qdict_put_obj(dict, "parent", parent);
1863 }
1864
1865 return res;
1866}
1867
218a536a
LC
1868void bdrv_info_stats(Monitor *mon, QObject **ret_data)
1869{
1870 QObject *obj;
1871 QList *devices;
a36e69dd
TS
1872 BlockDriverState *bs;
1873
218a536a
LC
1874 devices = qlist_new();
1875
1b7bdbc1 1876 QTAILQ_FOREACH(bs, &bdrv_states, list) {
294cc35f 1877 obj = bdrv_info_stats_bs(bs);
218a536a 1878 qlist_append_obj(devices, obj);
a36e69dd 1879 }
218a536a
LC
1880
1881 *ret_data = QOBJECT(devices);
a36e69dd 1882}
ea2384d3 1883
045df330
AL
1884const char *bdrv_get_encrypted_filename(BlockDriverState *bs)
1885{
1886 if (bs->backing_hd && bs->backing_hd->encrypted)
1887 return bs->backing_file;
1888 else if (bs->encrypted)
1889 return bs->filename;
1890 else
1891 return NULL;
1892}
1893
5fafdf24 1894void bdrv_get_backing_filename(BlockDriverState *bs,
83f64091
FB
1895 char *filename, int filename_size)
1896{
b783e409 1897 if (!bs->backing_file) {
83f64091
FB
1898 pstrcpy(filename, filename_size, "");
1899 } else {
1900 pstrcpy(filename, filename_size, bs->backing_file);
1901 }
1902}
1903
5fafdf24 1904int bdrv_write_compressed(BlockDriverState *bs, int64_t sector_num,
faea38e7
FB
1905 const uint8_t *buf, int nb_sectors)
1906{
1907 BlockDriver *drv = bs->drv;
1908 if (!drv)
19cb3738 1909 return -ENOMEDIUM;
faea38e7
FB
1910 if (!drv->bdrv_write_compressed)
1911 return -ENOTSUP;
fbb7b4e0
KW
1912 if (bdrv_check_request(bs, sector_num, nb_sectors))
1913 return -EIO;
a55eb92c 1914
c6d22830 1915 if (bs->dirty_bitmap) {
7cd1e32a
LS
1916 set_dirty_bitmap(bs, sector_num, nb_sectors, 1);
1917 }
a55eb92c 1918
faea38e7
FB
1919 return drv->bdrv_write_compressed(bs, sector_num, buf, nb_sectors);
1920}
3b46e624 1921
faea38e7
FB
1922int bdrv_get_info(BlockDriverState *bs, BlockDriverInfo *bdi)
1923{
1924 BlockDriver *drv = bs->drv;
1925 if (!drv)
19cb3738 1926 return -ENOMEDIUM;
faea38e7
FB
1927 if (!drv->bdrv_get_info)
1928 return -ENOTSUP;
1929 memset(bdi, 0, sizeof(*bdi));
1930 return drv->bdrv_get_info(bs, bdi);
1931}
1932
45566e9c
CH
1933int bdrv_save_vmstate(BlockDriverState *bs, const uint8_t *buf,
1934 int64_t pos, int size)
178e08a5
AL
1935{
1936 BlockDriver *drv = bs->drv;
1937 if (!drv)
1938 return -ENOMEDIUM;
7cdb1f6d
MK
1939 if (drv->bdrv_save_vmstate)
1940 return drv->bdrv_save_vmstate(bs, buf, pos, size);
1941 if (bs->file)
1942 return bdrv_save_vmstate(bs->file, buf, pos, size);
1943 return -ENOTSUP;
178e08a5
AL
1944}
1945
45566e9c
CH
1946int bdrv_load_vmstate(BlockDriverState *bs, uint8_t *buf,
1947 int64_t pos, int size)
178e08a5
AL
1948{
1949 BlockDriver *drv = bs->drv;
1950 if (!drv)
1951 return -ENOMEDIUM;
7cdb1f6d
MK
1952 if (drv->bdrv_load_vmstate)
1953 return drv->bdrv_load_vmstate(bs, buf, pos, size);
1954 if (bs->file)
1955 return bdrv_load_vmstate(bs->file, buf, pos, size);
1956 return -ENOTSUP;
178e08a5
AL
1957}
1958
8b9b0cc2
KW
1959void bdrv_debug_event(BlockDriverState *bs, BlkDebugEvent event)
1960{
1961 BlockDriver *drv = bs->drv;
1962
1963 if (!drv || !drv->bdrv_debug_event) {
1964 return;
1965 }
1966
1967 return drv->bdrv_debug_event(bs, event);
1968
1969}
1970
faea38e7
FB
1971/**************************************************************/
1972/* handling of snapshots */
1973
feeee5ac
MDCF
1974int bdrv_can_snapshot(BlockDriverState *bs)
1975{
1976 BlockDriver *drv = bs->drv;
1977 if (!drv || bdrv_is_removable(bs) || bdrv_is_read_only(bs)) {
1978 return 0;
1979 }
1980
1981 if (!drv->bdrv_snapshot_create) {
1982 if (bs->file != NULL) {
1983 return bdrv_can_snapshot(bs->file);
1984 }
1985 return 0;
1986 }
1987
1988 return 1;
1989}
1990
199630b6
BS
1991int bdrv_is_snapshot(BlockDriverState *bs)
1992{
1993 return !!(bs->open_flags & BDRV_O_SNAPSHOT);
1994}
1995
f9092b10
MA
1996BlockDriverState *bdrv_snapshots(void)
1997{
1998 BlockDriverState *bs;
1999
3ac906f7 2000 if (bs_snapshots) {
f9092b10 2001 return bs_snapshots;
3ac906f7 2002 }
f9092b10
MA
2003
2004 bs = NULL;
2005 while ((bs = bdrv_next(bs))) {
2006 if (bdrv_can_snapshot(bs)) {
3ac906f7
MA
2007 bs_snapshots = bs;
2008 return bs;
f9092b10
MA
2009 }
2010 }
2011 return NULL;
f9092b10
MA
2012}
2013
5fafdf24 2014int bdrv_snapshot_create(BlockDriverState *bs,
faea38e7
FB
2015 QEMUSnapshotInfo *sn_info)
2016{
2017 BlockDriver *drv = bs->drv;
2018 if (!drv)
19cb3738 2019 return -ENOMEDIUM;
7cdb1f6d
MK
2020 if (drv->bdrv_snapshot_create)
2021 return drv->bdrv_snapshot_create(bs, sn_info);
2022 if (bs->file)
2023 return bdrv_snapshot_create(bs->file, sn_info);
2024 return -ENOTSUP;
faea38e7
FB
2025}
2026
5fafdf24 2027int bdrv_snapshot_goto(BlockDriverState *bs,
faea38e7
FB
2028 const char *snapshot_id)
2029{
2030 BlockDriver *drv = bs->drv;
7cdb1f6d
MK
2031 int ret, open_ret;
2032
faea38e7 2033 if (!drv)
19cb3738 2034 return -ENOMEDIUM;
7cdb1f6d
MK
2035 if (drv->bdrv_snapshot_goto)
2036 return drv->bdrv_snapshot_goto(bs, snapshot_id);
2037
2038 if (bs->file) {
2039 drv->bdrv_close(bs);
2040 ret = bdrv_snapshot_goto(bs->file, snapshot_id);
2041 open_ret = drv->bdrv_open(bs, bs->open_flags);
2042 if (open_ret < 0) {
2043 bdrv_delete(bs->file);
2044 bs->drv = NULL;
2045 return open_ret;
2046 }
2047 return ret;
2048 }
2049
2050 return -ENOTSUP;
faea38e7
FB
2051}
2052
2053int bdrv_snapshot_delete(BlockDriverState *bs, const char *snapshot_id)
2054{
2055 BlockDriver *drv = bs->drv;
2056 if (!drv)
19cb3738 2057 return -ENOMEDIUM;
7cdb1f6d
MK
2058 if (drv->bdrv_snapshot_delete)
2059 return drv->bdrv_snapshot_delete(bs, snapshot_id);
2060 if (bs->file)
2061 return bdrv_snapshot_delete(bs->file, snapshot_id);
2062 return -ENOTSUP;
faea38e7
FB
2063}
2064
5fafdf24 2065int bdrv_snapshot_list(BlockDriverState *bs,
faea38e7
FB
2066 QEMUSnapshotInfo **psn_info)
2067{
2068 BlockDriver *drv = bs->drv;
2069 if (!drv)
19cb3738 2070 return -ENOMEDIUM;
7cdb1f6d
MK
2071 if (drv->bdrv_snapshot_list)
2072 return drv->bdrv_snapshot_list(bs, psn_info);
2073 if (bs->file)
2074 return bdrv_snapshot_list(bs->file, psn_info);
2075 return -ENOTSUP;
faea38e7
FB
2076}
2077
51ef6727 2078int bdrv_snapshot_load_tmp(BlockDriverState *bs,
2079 const char *snapshot_name)
2080{
2081 BlockDriver *drv = bs->drv;
2082 if (!drv) {
2083 return -ENOMEDIUM;
2084 }
2085 if (!bs->read_only) {
2086 return -EINVAL;
2087 }
2088 if (drv->bdrv_snapshot_load_tmp) {
2089 return drv->bdrv_snapshot_load_tmp(bs, snapshot_name);
2090 }
2091 return -ENOTSUP;
2092}
2093
faea38e7
FB
2094#define NB_SUFFIXES 4
2095
2096char *get_human_readable_size(char *buf, int buf_size, int64_t size)
2097{
2098 static const char suffixes[NB_SUFFIXES] = "KMGT";
2099 int64_t base;
2100 int i;
2101
2102 if (size <= 999) {
2103 snprintf(buf, buf_size, "%" PRId64, size);
2104 } else {
2105 base = 1024;
2106 for(i = 0; i < NB_SUFFIXES; i++) {
2107 if (size < (10 * base)) {
5fafdf24 2108 snprintf(buf, buf_size, "%0.1f%c",
faea38e7
FB
2109 (double)size / base,
2110 suffixes[i]);
2111 break;
2112 } else if (size < (1000 * base) || i == (NB_SUFFIXES - 1)) {
5fafdf24 2113 snprintf(buf, buf_size, "%" PRId64 "%c",
faea38e7
FB
2114 ((size + (base >> 1)) / base),
2115 suffixes[i]);
2116 break;
2117 }
2118 base = base * 1024;
2119 }
2120 }
2121 return buf;
2122}
2123
2124char *bdrv_snapshot_dump(char *buf, int buf_size, QEMUSnapshotInfo *sn)
2125{
2126 char buf1[128], date_buf[128], clock_buf[128];
3b9f94e1
FB
2127#ifdef _WIN32
2128 struct tm *ptm;
2129#else
faea38e7 2130 struct tm tm;
3b9f94e1 2131#endif
faea38e7
FB
2132 time_t ti;
2133 int64_t secs;
2134
2135 if (!sn) {
5fafdf24
TS
2136 snprintf(buf, buf_size,
2137 "%-10s%-20s%7s%20s%15s",
faea38e7
FB
2138 "ID", "TAG", "VM SIZE", "DATE", "VM CLOCK");
2139 } else {
2140 ti = sn->date_sec;
3b9f94e1
FB
2141#ifdef _WIN32
2142 ptm = localtime(&ti);
2143 strftime(date_buf, sizeof(date_buf),
2144 "%Y-%m-%d %H:%M:%S", ptm);
2145#else
faea38e7
FB
2146 localtime_r(&ti, &tm);
2147 strftime(date_buf, sizeof(date_buf),
2148 "%Y-%m-%d %H:%M:%S", &tm);
3b9f94e1 2149#endif
faea38e7
FB
2150 secs = sn->vm_clock_nsec / 1000000000;
2151 snprintf(clock_buf, sizeof(clock_buf),
2152 "%02d:%02d:%02d.%03d",
2153 (int)(secs / 3600),
2154 (int)((secs / 60) % 60),
5fafdf24 2155 (int)(secs % 60),
faea38e7
FB
2156 (int)((sn->vm_clock_nsec / 1000000) % 1000));
2157 snprintf(buf, buf_size,
5fafdf24 2158 "%-10s%-20s%7s%20s%15s",
faea38e7
FB
2159 sn->id_str, sn->name,
2160 get_human_readable_size(buf1, sizeof(buf1), sn->vm_state_size),
2161 date_buf,
2162 clock_buf);
2163 }
2164 return buf;
2165}
2166
83f64091 2167
ea2384d3 2168/**************************************************************/
83f64091 2169/* async I/Os */
ea2384d3 2170
3b69e4b9 2171BlockDriverAIOCB *bdrv_aio_readv(BlockDriverState *bs, int64_t sector_num,
f141eafe 2172 QEMUIOVector *qiov, int nb_sectors,
3b69e4b9 2173 BlockDriverCompletionFunc *cb, void *opaque)
83f64091
FB
2174{
2175 BlockDriver *drv = bs->drv;
a36e69dd 2176 BlockDriverAIOCB *ret;
83f64091 2177
bbf0a440
SH
2178 trace_bdrv_aio_readv(bs, sector_num, nb_sectors, opaque);
2179
19cb3738 2180 if (!drv)
ce1a14dc 2181 return NULL;
71d0770c
AL
2182 if (bdrv_check_request(bs, sector_num, nb_sectors))
2183 return NULL;
3b46e624 2184
f141eafe
AL
2185 ret = drv->bdrv_aio_readv(bs, sector_num, qiov, nb_sectors,
2186 cb, opaque);
a36e69dd
TS
2187
2188 if (ret) {
2189 /* Update stats even though technically transfer has not happened. */
6ea44308 2190 bs->rd_bytes += (unsigned) nb_sectors * BDRV_SECTOR_SIZE;
a36e69dd
TS
2191 bs->rd_ops ++;
2192 }
2193
2194 return ret;
ea2384d3
FB
2195}
2196
4dcafbb1
MT
2197typedef struct BlockCompleteData {
2198 BlockDriverCompletionFunc *cb;
2199 void *opaque;
2200 BlockDriverState *bs;
2201 int64_t sector_num;
2202 int nb_sectors;
2203} BlockCompleteData;
2204
2205static void block_complete_cb(void *opaque, int ret)
2206{
2207 BlockCompleteData *b = opaque;
2208
2209 if (b->bs->dirty_bitmap) {
2210 set_dirty_bitmap(b->bs, b->sector_num, b->nb_sectors, 1);
2211 }
2212 b->cb(b->opaque, ret);
2213 qemu_free(b);
2214}
2215
2216static BlockCompleteData *blk_dirty_cb_alloc(BlockDriverState *bs,
2217 int64_t sector_num,
2218 int nb_sectors,
2219 BlockDriverCompletionFunc *cb,
2220 void *opaque)
2221{
2222 BlockCompleteData *blkdata = qemu_mallocz(sizeof(BlockCompleteData));
2223
2224 blkdata->bs = bs;
2225 blkdata->cb = cb;
2226 blkdata->opaque = opaque;
2227 blkdata->sector_num = sector_num;
2228 blkdata->nb_sectors = nb_sectors;
2229
2230 return blkdata;
2231}
2232
f141eafe
AL
2233BlockDriverAIOCB *bdrv_aio_writev(BlockDriverState *bs, int64_t sector_num,
2234 QEMUIOVector *qiov, int nb_sectors,
2235 BlockDriverCompletionFunc *cb, void *opaque)
ea2384d3 2236{
83f64091 2237 BlockDriver *drv = bs->drv;
a36e69dd 2238 BlockDriverAIOCB *ret;
4dcafbb1 2239 BlockCompleteData *blk_cb_data;
ea2384d3 2240
bbf0a440
SH
2241 trace_bdrv_aio_writev(bs, sector_num, nb_sectors, opaque);
2242
19cb3738 2243 if (!drv)
ce1a14dc 2244 return NULL;
83f64091 2245 if (bs->read_only)
ce1a14dc 2246 return NULL;
71d0770c
AL
2247 if (bdrv_check_request(bs, sector_num, nb_sectors))
2248 return NULL;
83f64091 2249
c6d22830 2250 if (bs->dirty_bitmap) {
4dcafbb1
MT
2251 blk_cb_data = blk_dirty_cb_alloc(bs, sector_num, nb_sectors, cb,
2252 opaque);
2253 cb = &block_complete_cb;
2254 opaque = blk_cb_data;
7cd1e32a 2255 }
a55eb92c 2256
f141eafe
AL
2257 ret = drv->bdrv_aio_writev(bs, sector_num, qiov, nb_sectors,
2258 cb, opaque);
a36e69dd
TS
2259
2260 if (ret) {
294cc35f
KW
2261 /* Update stats even though technically transfer has not happened. */
2262 bs->wr_bytes += (unsigned) nb_sectors * BDRV_SECTOR_SIZE;
2263 bs->wr_ops ++;
2264 if (bs->wr_highest_sector < sector_num + nb_sectors - 1) {
2265 bs->wr_highest_sector = sector_num + nb_sectors - 1;
2266 }
a36e69dd
TS
2267 }
2268
2269 return ret;
83f64091
FB
2270}
2271
40b4f539
KW
2272
2273typedef struct MultiwriteCB {
2274 int error;
2275 int num_requests;
2276 int num_callbacks;
2277 struct {
2278 BlockDriverCompletionFunc *cb;
2279 void *opaque;
2280 QEMUIOVector *free_qiov;
2281 void *free_buf;
2282 } callbacks[];
2283} MultiwriteCB;
2284
2285static void multiwrite_user_cb(MultiwriteCB *mcb)
2286{
2287 int i;
2288
2289 for (i = 0; i < mcb->num_callbacks; i++) {
2290 mcb->callbacks[i].cb(mcb->callbacks[i].opaque, mcb->error);
1e1ea48d
SH
2291 if (mcb->callbacks[i].free_qiov) {
2292 qemu_iovec_destroy(mcb->callbacks[i].free_qiov);
2293 }
40b4f539 2294 qemu_free(mcb->callbacks[i].free_qiov);
f8a83245 2295 qemu_vfree(mcb->callbacks[i].free_buf);
40b4f539
KW
2296 }
2297}
2298
2299static void multiwrite_cb(void *opaque, int ret)
2300{
2301 MultiwriteCB *mcb = opaque;
2302
6d519a5f
SH
2303 trace_multiwrite_cb(mcb, ret);
2304
cb6d3ca0 2305 if (ret < 0 && !mcb->error) {
40b4f539 2306 mcb->error = ret;
40b4f539
KW
2307 }
2308
2309 mcb->num_requests--;
2310 if (mcb->num_requests == 0) {
de189a1b 2311 multiwrite_user_cb(mcb);
40b4f539
KW
2312 qemu_free(mcb);
2313 }
2314}
2315
2316static int multiwrite_req_compare(const void *a, const void *b)
2317{
77be4366
CH
2318 const BlockRequest *req1 = a, *req2 = b;
2319
2320 /*
2321 * Note that we can't simply subtract req2->sector from req1->sector
2322 * here as that could overflow the return value.
2323 */
2324 if (req1->sector > req2->sector) {
2325 return 1;
2326 } else if (req1->sector < req2->sector) {
2327 return -1;
2328 } else {
2329 return 0;
2330 }
40b4f539
KW
2331}
2332
2333/*
2334 * Takes a bunch of requests and tries to merge them. Returns the number of
2335 * requests that remain after merging.
2336 */
2337static int multiwrite_merge(BlockDriverState *bs, BlockRequest *reqs,
2338 int num_reqs, MultiwriteCB *mcb)
2339{
2340 int i, outidx;
2341
2342 // Sort requests by start sector
2343 qsort(reqs, num_reqs, sizeof(*reqs), &multiwrite_req_compare);
2344
2345 // Check if adjacent requests touch the same clusters. If so, combine them,
2346 // filling up gaps with zero sectors.
2347 outidx = 0;
2348 for (i = 1; i < num_reqs; i++) {
2349 int merge = 0;
2350 int64_t oldreq_last = reqs[outidx].sector + reqs[outidx].nb_sectors;
2351
2352 // This handles the cases that are valid for all block drivers, namely
2353 // exactly sequential writes and overlapping writes.
2354 if (reqs[i].sector <= oldreq_last) {
2355 merge = 1;
2356 }
2357
2358 // The block driver may decide that it makes sense to combine requests
2359 // even if there is a gap of some sectors between them. In this case,
2360 // the gap is filled with zeros (therefore only applicable for yet
2361 // unused space in format like qcow2).
2362 if (!merge && bs->drv->bdrv_merge_requests) {
2363 merge = bs->drv->bdrv_merge_requests(bs, &reqs[outidx], &reqs[i]);
2364 }
2365
e2a305fb
CH
2366 if (reqs[outidx].qiov->niov + reqs[i].qiov->niov + 1 > IOV_MAX) {
2367 merge = 0;
2368 }
2369
40b4f539
KW
2370 if (merge) {
2371 size_t size;
2372 QEMUIOVector *qiov = qemu_mallocz(sizeof(*qiov));
2373 qemu_iovec_init(qiov,
2374 reqs[outidx].qiov->niov + reqs[i].qiov->niov + 1);
2375
2376 // Add the first request to the merged one. If the requests are
2377 // overlapping, drop the last sectors of the first request.
2378 size = (reqs[i].sector - reqs[outidx].sector) << 9;
2379 qemu_iovec_concat(qiov, reqs[outidx].qiov, size);
2380
2381 // We might need to add some zeros between the two requests
2382 if (reqs[i].sector > oldreq_last) {
2383 size_t zero_bytes = (reqs[i].sector - oldreq_last) << 9;
2384 uint8_t *buf = qemu_blockalign(bs, zero_bytes);
2385 memset(buf, 0, zero_bytes);
2386 qemu_iovec_add(qiov, buf, zero_bytes);
2387 mcb->callbacks[i].free_buf = buf;
2388 }
2389
2390 // Add the second request
2391 qemu_iovec_concat(qiov, reqs[i].qiov, reqs[i].qiov->size);
2392
cbf1dff2 2393 reqs[outidx].nb_sectors = qiov->size >> 9;
40b4f539
KW
2394 reqs[outidx].qiov = qiov;
2395
2396 mcb->callbacks[i].free_qiov = reqs[outidx].qiov;
2397 } else {
2398 outidx++;
2399 reqs[outidx].sector = reqs[i].sector;
2400 reqs[outidx].nb_sectors = reqs[i].nb_sectors;
2401 reqs[outidx].qiov = reqs[i].qiov;
2402 }
2403 }
2404
2405 return outidx + 1;
2406}
2407
2408/*
2409 * Submit multiple AIO write requests at once.
2410 *
2411 * On success, the function returns 0 and all requests in the reqs array have
2412 * been submitted. In error case this function returns -1, and any of the
2413 * requests may or may not be submitted yet. In particular, this means that the
2414 * callback will be called for some of the requests, for others it won't. The
2415 * caller must check the error field of the BlockRequest to wait for the right
2416 * callbacks (if error != 0, no callback will be called).
2417 *
2418 * The implementation may modify the contents of the reqs array, e.g. to merge
2419 * requests. However, the fields opaque and error are left unmodified as they
2420 * are used to signal failure for a single request to the caller.
2421 */
2422int bdrv_aio_multiwrite(BlockDriverState *bs, BlockRequest *reqs, int num_reqs)
2423{
2424 BlockDriverAIOCB *acb;
2425 MultiwriteCB *mcb;
2426 int i;
2427
301db7c2
RH
2428 /* don't submit writes if we don't have a medium */
2429 if (bs->drv == NULL) {
2430 for (i = 0; i < num_reqs; i++) {
2431 reqs[i].error = -ENOMEDIUM;
2432 }
2433 return -1;
2434 }
2435
40b4f539
KW
2436 if (num_reqs == 0) {
2437 return 0;
2438 }
2439
2440 // Create MultiwriteCB structure
2441 mcb = qemu_mallocz(sizeof(*mcb) + num_reqs * sizeof(*mcb->callbacks));
2442 mcb->num_requests = 0;
2443 mcb->num_callbacks = num_reqs;
2444
2445 for (i = 0; i < num_reqs; i++) {
2446 mcb->callbacks[i].cb = reqs[i].cb;
2447 mcb->callbacks[i].opaque = reqs[i].opaque;
2448 }
2449
2450 // Check for mergable requests
2451 num_reqs = multiwrite_merge(bs, reqs, num_reqs, mcb);
2452
6d519a5f
SH
2453 trace_bdrv_aio_multiwrite(mcb, mcb->num_callbacks, num_reqs);
2454
453f9a16
KW
2455 /*
2456 * Run the aio requests. As soon as one request can't be submitted
2457 * successfully, fail all requests that are not yet submitted (we must
2458 * return failure for all requests anyway)
2459 *
2460 * num_requests cannot be set to the right value immediately: If
2461 * bdrv_aio_writev fails for some request, num_requests would be too high
2462 * and therefore multiwrite_cb() would never recognize the multiwrite
2463 * request as completed. We also cannot use the loop variable i to set it
2464 * when the first request fails because the callback may already have been
2465 * called for previously submitted requests. Thus, num_requests must be
2466 * incremented for each request that is submitted.
2467 *
2468 * The problem that callbacks may be called early also means that we need
2469 * to take care that num_requests doesn't become 0 before all requests are
2470 * submitted - multiwrite_cb() would consider the multiwrite request
2471 * completed. A dummy request that is "completed" by a manual call to
2472 * multiwrite_cb() takes care of this.
2473 */
2474 mcb->num_requests = 1;
2475
6d519a5f 2476 // Run the aio requests
40b4f539 2477 for (i = 0; i < num_reqs; i++) {
453f9a16 2478 mcb->num_requests++;
40b4f539
KW
2479 acb = bdrv_aio_writev(bs, reqs[i].sector, reqs[i].qiov,
2480 reqs[i].nb_sectors, multiwrite_cb, mcb);
2481
2482 if (acb == NULL) {
2483 // We can only fail the whole thing if no request has been
2484 // submitted yet. Otherwise we'll wait for the submitted AIOs to
2485 // complete and report the error in the callback.
453f9a16 2486 if (i == 0) {
6d519a5f 2487 trace_bdrv_aio_multiwrite_earlyfail(mcb);
40b4f539
KW
2488 goto fail;
2489 } else {
6d519a5f 2490 trace_bdrv_aio_multiwrite_latefail(mcb, i);
7eb58a6c 2491 multiwrite_cb(mcb, -EIO);
40b4f539
KW
2492 break;
2493 }
40b4f539
KW
2494 }
2495 }
2496
453f9a16
KW
2497 /* Complete the dummy request */
2498 multiwrite_cb(mcb, 0);
2499
40b4f539
KW
2500 return 0;
2501
2502fail:
453f9a16
KW
2503 for (i = 0; i < mcb->num_callbacks; i++) {
2504 reqs[i].error = -EIO;
2505 }
af474591 2506 qemu_free(mcb);
40b4f539
KW
2507 return -1;
2508}
2509
b2e12bc6
CH
2510BlockDriverAIOCB *bdrv_aio_flush(BlockDriverState *bs,
2511 BlockDriverCompletionFunc *cb, void *opaque)
2512{
2513 BlockDriver *drv = bs->drv;
2514
a13aac04
SH
2515 trace_bdrv_aio_flush(bs, opaque);
2516
016f5cf6
AG
2517 if (bs->open_flags & BDRV_O_NO_FLUSH) {
2518 return bdrv_aio_noop_em(bs, cb, opaque);
2519 }
2520
b2e12bc6
CH
2521 if (!drv)
2522 return NULL;
b2e12bc6
CH
2523 return drv->bdrv_aio_flush(bs, cb, opaque);
2524}
2525
83f64091 2526void bdrv_aio_cancel(BlockDriverAIOCB *acb)
83f64091 2527{
6bbff9a0 2528 acb->pool->cancel(acb);
83f64091
FB
2529}
2530
ce1a14dc 2531
83f64091
FB
2532/**************************************************************/
2533/* async block device emulation */
2534
c16b5a2c
CH
2535typedef struct BlockDriverAIOCBSync {
2536 BlockDriverAIOCB common;
2537 QEMUBH *bh;
2538 int ret;
2539 /* vector translation state */
2540 QEMUIOVector *qiov;
2541 uint8_t *bounce;
2542 int is_write;
2543} BlockDriverAIOCBSync;
2544
2545static void bdrv_aio_cancel_em(BlockDriverAIOCB *blockacb)
2546{
b666d239
KW
2547 BlockDriverAIOCBSync *acb =
2548 container_of(blockacb, BlockDriverAIOCBSync, common);
6a7ad299 2549 qemu_bh_delete(acb->bh);
36afc451 2550 acb->bh = NULL;
c16b5a2c
CH
2551 qemu_aio_release(acb);
2552}
2553
2554static AIOPool bdrv_em_aio_pool = {
2555 .aiocb_size = sizeof(BlockDriverAIOCBSync),
2556 .cancel = bdrv_aio_cancel_em,
2557};
2558
ce1a14dc 2559static void bdrv_aio_bh_cb(void *opaque)
83f64091 2560{
ce1a14dc 2561 BlockDriverAIOCBSync *acb = opaque;
f141eafe 2562
f141eafe
AL
2563 if (!acb->is_write)
2564 qemu_iovec_from_buffer(acb->qiov, acb->bounce, acb->qiov->size);
ceb42de8 2565 qemu_vfree(acb->bounce);
ce1a14dc 2566 acb->common.cb(acb->common.opaque, acb->ret);
6a7ad299 2567 qemu_bh_delete(acb->bh);
36afc451 2568 acb->bh = NULL;
ce1a14dc 2569 qemu_aio_release(acb);
83f64091 2570}
beac80cd 2571
f141eafe
AL
2572static BlockDriverAIOCB *bdrv_aio_rw_vector(BlockDriverState *bs,
2573 int64_t sector_num,
2574 QEMUIOVector *qiov,
2575 int nb_sectors,
2576 BlockDriverCompletionFunc *cb,
2577 void *opaque,
2578 int is_write)
2579
83f64091 2580{
ce1a14dc 2581 BlockDriverAIOCBSync *acb;
ce1a14dc 2582
c16b5a2c 2583 acb = qemu_aio_get(&bdrv_em_aio_pool, bs, cb, opaque);
f141eafe
AL
2584 acb->is_write = is_write;
2585 acb->qiov = qiov;
e268ca52 2586 acb->bounce = qemu_blockalign(bs, qiov->size);
f141eafe 2587
ce1a14dc
PB
2588 if (!acb->bh)
2589 acb->bh = qemu_bh_new(bdrv_aio_bh_cb, acb);
f141eafe
AL
2590
2591 if (is_write) {
2592 qemu_iovec_to_buffer(acb->qiov, acb->bounce);
2593 acb->ret = bdrv_write(bs, sector_num, acb->bounce, nb_sectors);
2594 } else {
2595 acb->ret = bdrv_read(bs, sector_num, acb->bounce, nb_sectors);
2596 }
2597
ce1a14dc 2598 qemu_bh_schedule(acb->bh);
f141eafe 2599
ce1a14dc 2600 return &acb->common;
beac80cd
FB
2601}
2602
f141eafe
AL
2603static BlockDriverAIOCB *bdrv_aio_readv_em(BlockDriverState *bs,
2604 int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
ce1a14dc 2605 BlockDriverCompletionFunc *cb, void *opaque)
beac80cd 2606{
f141eafe
AL
2607 return bdrv_aio_rw_vector(bs, sector_num, qiov, nb_sectors, cb, opaque, 0);
2608}
83f64091 2609
f141eafe
AL
2610static BlockDriverAIOCB *bdrv_aio_writev_em(BlockDriverState *bs,
2611 int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
2612 BlockDriverCompletionFunc *cb, void *opaque)
2613{
2614 return bdrv_aio_rw_vector(bs, sector_num, qiov, nb_sectors, cb, opaque, 1);
beac80cd 2615}
beac80cd 2616
b2e12bc6
CH
2617static BlockDriverAIOCB *bdrv_aio_flush_em(BlockDriverState *bs,
2618 BlockDriverCompletionFunc *cb, void *opaque)
2619{
2620 BlockDriverAIOCBSync *acb;
2621
2622 acb = qemu_aio_get(&bdrv_em_aio_pool, bs, cb, opaque);
2623 acb->is_write = 1; /* don't bounce in the completion hadler */
2624 acb->qiov = NULL;
2625 acb->bounce = NULL;
2626 acb->ret = 0;
2627
2628 if (!acb->bh)
2629 acb->bh = qemu_bh_new(bdrv_aio_bh_cb, acb);
2630
2631 bdrv_flush(bs);
2632 qemu_bh_schedule(acb->bh);
2633 return &acb->common;
2634}
2635
016f5cf6
AG
2636static BlockDriverAIOCB *bdrv_aio_noop_em(BlockDriverState *bs,
2637 BlockDriverCompletionFunc *cb, void *opaque)
2638{
2639 BlockDriverAIOCBSync *acb;
2640
2641 acb = qemu_aio_get(&bdrv_em_aio_pool, bs, cb, opaque);
2642 acb->is_write = 1; /* don't bounce in the completion handler */
2643 acb->qiov = NULL;
2644 acb->bounce = NULL;
2645 acb->ret = 0;
2646
2647 if (!acb->bh) {
2648 acb->bh = qemu_bh_new(bdrv_aio_bh_cb, acb);
2649 }
2650
2651 qemu_bh_schedule(acb->bh);
2652 return &acb->common;
2653}
2654
83f64091
FB
2655/**************************************************************/
2656/* sync block device emulation */
ea2384d3 2657
83f64091
FB
2658static void bdrv_rw_em_cb(void *opaque, int ret)
2659{
2660 *(int *)opaque = ret;
ea2384d3
FB
2661}
2662
83f64091
FB
2663#define NOT_DONE 0x7fffffff
2664
5fafdf24 2665static int bdrv_read_em(BlockDriverState *bs, int64_t sector_num,
83f64091 2666 uint8_t *buf, int nb_sectors)
7a6cba61 2667{
ce1a14dc
PB
2668 int async_ret;
2669 BlockDriverAIOCB *acb;
f141eafe
AL
2670 struct iovec iov;
2671 QEMUIOVector qiov;
83f64091 2672
65d6b3d8
KW
2673 async_context_push();
2674
83f64091 2675 async_ret = NOT_DONE;
3f4cb3d3 2676 iov.iov_base = (void *)buf;
eb5a3165 2677 iov.iov_len = nb_sectors * BDRV_SECTOR_SIZE;
f141eafe
AL
2678 qemu_iovec_init_external(&qiov, &iov, 1);
2679 acb = bdrv_aio_readv(bs, sector_num, &qiov, nb_sectors,
2680 bdrv_rw_em_cb, &async_ret);
65d6b3d8
KW
2681 if (acb == NULL) {
2682 async_ret = -1;
2683 goto fail;
2684 }
baf35cb9 2685
83f64091
FB
2686 while (async_ret == NOT_DONE) {
2687 qemu_aio_wait();
2688 }
baf35cb9 2689
65d6b3d8
KW
2690
2691fail:
2692 async_context_pop();
83f64091 2693 return async_ret;
7a6cba61
PB
2694}
2695
83f64091
FB
2696static int bdrv_write_em(BlockDriverState *bs, int64_t sector_num,
2697 const uint8_t *buf, int nb_sectors)
2698{
ce1a14dc
PB
2699 int async_ret;
2700 BlockDriverAIOCB *acb;
f141eafe
AL
2701 struct iovec iov;
2702 QEMUIOVector qiov;
83f64091 2703
65d6b3d8
KW
2704 async_context_push();
2705
83f64091 2706 async_ret = NOT_DONE;
f141eafe 2707 iov.iov_base = (void *)buf;
eb5a3165 2708 iov.iov_len = nb_sectors * BDRV_SECTOR_SIZE;
f141eafe
AL
2709 qemu_iovec_init_external(&qiov, &iov, 1);
2710 acb = bdrv_aio_writev(bs, sector_num, &qiov, nb_sectors,
2711 bdrv_rw_em_cb, &async_ret);
65d6b3d8
KW
2712 if (acb == NULL) {
2713 async_ret = -1;
2714 goto fail;
2715 }
83f64091
FB
2716 while (async_ret == NOT_DONE) {
2717 qemu_aio_wait();
2718 }
65d6b3d8
KW
2719
2720fail:
2721 async_context_pop();
83f64091
FB
2722 return async_ret;
2723}
ea2384d3
FB
2724
2725void bdrv_init(void)
2726{
5efa9d5a 2727 module_call_init(MODULE_INIT_BLOCK);
ea2384d3 2728}
ce1a14dc 2729
eb852011
MA
2730void bdrv_init_with_whitelist(void)
2731{
2732 use_bdrv_whitelist = 1;
2733 bdrv_init();
2734}
2735
c16b5a2c
CH
2736void *qemu_aio_get(AIOPool *pool, BlockDriverState *bs,
2737 BlockDriverCompletionFunc *cb, void *opaque)
ce1a14dc 2738{
ce1a14dc
PB
2739 BlockDriverAIOCB *acb;
2740
6bbff9a0
AL
2741 if (pool->free_aiocb) {
2742 acb = pool->free_aiocb;
2743 pool->free_aiocb = acb->next;
ce1a14dc 2744 } else {
6bbff9a0
AL
2745 acb = qemu_mallocz(pool->aiocb_size);
2746 acb->pool = pool;
ce1a14dc
PB
2747 }
2748 acb->bs = bs;
2749 acb->cb = cb;
2750 acb->opaque = opaque;
2751 return acb;
2752}
2753
2754void qemu_aio_release(void *p)
2755{
6bbff9a0
AL
2756 BlockDriverAIOCB *acb = (BlockDriverAIOCB *)p;
2757 AIOPool *pool = acb->pool;
2758 acb->next = pool->free_aiocb;
2759 pool->free_aiocb = acb;
ce1a14dc 2760}
19cb3738
FB
2761
2762/**************************************************************/
2763/* removable device support */
2764
2765/**
2766 * Return TRUE if the media is present
2767 */
2768int bdrv_is_inserted(BlockDriverState *bs)
2769{
2770 BlockDriver *drv = bs->drv;
2771 int ret;
2772 if (!drv)
2773 return 0;
2774 if (!drv->bdrv_is_inserted)
4be9762a 2775 return !bs->tray_open;
19cb3738
FB
2776 ret = drv->bdrv_is_inserted(bs);
2777 return ret;
2778}
2779
2780/**
2781 * Return TRUE if the media changed since the last call to this
5fafdf24 2782 * function. It is currently only used for floppy disks
19cb3738
FB
2783 */
2784int bdrv_media_changed(BlockDriverState *bs)
2785{
2786 BlockDriver *drv = bs->drv;
2787 int ret;
2788
2789 if (!drv || !drv->bdrv_media_changed)
2790 ret = -ENOTSUP;
2791 else
2792 ret = drv->bdrv_media_changed(bs);
2793 if (ret == -ENOTSUP)
2794 ret = bs->media_changed;
2795 bs->media_changed = 0;
2796 return ret;
2797}
2798
2799/**
2800 * If eject_flag is TRUE, eject the media. Otherwise, close the tray
2801 */
aea2a33c 2802int bdrv_eject(BlockDriverState *bs, int eject_flag)
19cb3738
FB
2803{
2804 BlockDriver *drv = bs->drv;
19cb3738 2805
49aa46bb 2806 if (eject_flag && bs->locked) {
aea2a33c
MM
2807 return -EBUSY;
2808 }
2809
822e1cd1
MA
2810 if (drv && drv->bdrv_eject) {
2811 drv->bdrv_eject(bs, eject_flag);
19cb3738 2812 }
822e1cd1
MA
2813 bs->tray_open = eject_flag;
2814 return 0;
19cb3738
FB
2815}
2816
2817int bdrv_is_locked(BlockDriverState *bs)
2818{
2819 return bs->locked;
2820}
2821
2822/**
2823 * Lock or unlock the media (if it is locked, the user won't be able
2824 * to eject it manually).
2825 */
2826void bdrv_set_locked(BlockDriverState *bs, int locked)
2827{
2828 BlockDriver *drv = bs->drv;
2829
b8c6d095
SH
2830 trace_bdrv_set_locked(bs, locked);
2831
19cb3738
FB
2832 bs->locked = locked;
2833 if (drv && drv->bdrv_set_locked) {
2834 drv->bdrv_set_locked(bs, locked);
2835 }
2836}
985a03b0
TS
2837
2838/* needed for generic scsi interface */
2839
2840int bdrv_ioctl(BlockDriverState *bs, unsigned long int req, void *buf)
2841{
2842 BlockDriver *drv = bs->drv;
2843
2844 if (drv && drv->bdrv_ioctl)
2845 return drv->bdrv_ioctl(bs, req, buf);
2846 return -ENOTSUP;
2847}
7d780669 2848
221f715d
AL
2849BlockDriverAIOCB *bdrv_aio_ioctl(BlockDriverState *bs,
2850 unsigned long int req, void *buf,
2851 BlockDriverCompletionFunc *cb, void *opaque)
7d780669 2852{
221f715d 2853 BlockDriver *drv = bs->drv;
7d780669 2854
221f715d
AL
2855 if (drv && drv->bdrv_aio_ioctl)
2856 return drv->bdrv_aio_ioctl(bs, req, buf, cb, opaque);
2857 return NULL;
7d780669 2858}
e268ca52 2859
7cd1e32a
LS
2860
2861
e268ca52
AL
2862void *qemu_blockalign(BlockDriverState *bs, size_t size)
2863{
2864 return qemu_memalign((bs && bs->buffer_alignment) ? bs->buffer_alignment : 512, size);
2865}
7cd1e32a
LS
2866
2867void bdrv_set_dirty_tracking(BlockDriverState *bs, int enable)
2868{
2869 int64_t bitmap_size;
a55eb92c 2870
aaa0eb75 2871 bs->dirty_count = 0;
a55eb92c 2872 if (enable) {
c6d22830
JK
2873 if (!bs->dirty_bitmap) {
2874 bitmap_size = (bdrv_getlength(bs) >> BDRV_SECTOR_BITS) +
2875 BDRV_SECTORS_PER_DIRTY_CHUNK * 8 - 1;
2876 bitmap_size /= BDRV_SECTORS_PER_DIRTY_CHUNK * 8;
a55eb92c 2877
7cd1e32a 2878 bs->dirty_bitmap = qemu_mallocz(bitmap_size);
a55eb92c 2879 }
7cd1e32a 2880 } else {
c6d22830 2881 if (bs->dirty_bitmap) {
7cd1e32a 2882 qemu_free(bs->dirty_bitmap);
c6d22830 2883 bs->dirty_bitmap = NULL;
a55eb92c 2884 }
7cd1e32a
LS
2885 }
2886}
2887
2888int bdrv_get_dirty(BlockDriverState *bs, int64_t sector)
2889{
6ea44308 2890 int64_t chunk = sector / (int64_t)BDRV_SECTORS_PER_DIRTY_CHUNK;
a55eb92c 2891
c6d22830
JK
2892 if (bs->dirty_bitmap &&
2893 (sector << BDRV_SECTOR_BITS) < bdrv_getlength(bs)) {
6d59fec1
MT
2894 return !!(bs->dirty_bitmap[chunk / (sizeof(unsigned long) * 8)] &
2895 (1UL << (chunk % (sizeof(unsigned long) * 8))));
7cd1e32a
LS
2896 } else {
2897 return 0;
2898 }
2899}
2900
a55eb92c
JK
2901void bdrv_reset_dirty(BlockDriverState *bs, int64_t cur_sector,
2902 int nr_sectors)
7cd1e32a
LS
2903{
2904 set_dirty_bitmap(bs, cur_sector, nr_sectors, 0);
2905}
aaa0eb75
LS
2906
2907int64_t bdrv_get_dirty_count(BlockDriverState *bs)
2908{
2909 return bs->dirty_count;
2910}
f88e1a42 2911
db593f25
MT
2912void bdrv_set_in_use(BlockDriverState *bs, int in_use)
2913{
2914 assert(bs->in_use != in_use);
2915 bs->in_use = in_use;
2916}
2917
2918int bdrv_in_use(BlockDriverState *bs)
2919{
2920 return bs->in_use;
2921}
2922
f88e1a42
JS
2923int bdrv_img_create(const char *filename, const char *fmt,
2924 const char *base_filename, const char *base_fmt,
2925 char *options, uint64_t img_size, int flags)
2926{
2927 QEMUOptionParameter *param = NULL, *create_options = NULL;
d220894e 2928 QEMUOptionParameter *backing_fmt, *backing_file, *size;
f88e1a42
JS
2929 BlockDriverState *bs = NULL;
2930 BlockDriver *drv, *proto_drv;
96df67d1 2931 BlockDriver *backing_drv = NULL;
f88e1a42
JS
2932 int ret = 0;
2933
2934 /* Find driver and parse its options */
2935 drv = bdrv_find_format(fmt);
2936 if (!drv) {
2937 error_report("Unknown file format '%s'", fmt);
4f70f249 2938 ret = -EINVAL;
f88e1a42
JS
2939 goto out;
2940 }
2941
2942 proto_drv = bdrv_find_protocol(filename);
2943 if (!proto_drv) {
2944 error_report("Unknown protocol '%s'", filename);
4f70f249 2945 ret = -EINVAL;
f88e1a42
JS
2946 goto out;
2947 }
2948
2949 create_options = append_option_parameters(create_options,
2950 drv->create_options);
2951 create_options = append_option_parameters(create_options,
2952 proto_drv->create_options);
2953
2954 /* Create parameter list with default values */
2955 param = parse_option_parameters("", create_options, param);
2956
2957 set_option_parameter_int(param, BLOCK_OPT_SIZE, img_size);
2958
2959 /* Parse -o options */
2960 if (options) {
2961 param = parse_option_parameters(options, create_options, param);
2962 if (param == NULL) {
2963 error_report("Invalid options for file format '%s'.", fmt);
4f70f249 2964 ret = -EINVAL;
f88e1a42
JS
2965 goto out;
2966 }
2967 }
2968
2969 if (base_filename) {
2970 if (set_option_parameter(param, BLOCK_OPT_BACKING_FILE,
2971 base_filename)) {
2972 error_report("Backing file not supported for file format '%s'",
2973 fmt);
4f70f249 2974 ret = -EINVAL;
f88e1a42
JS
2975 goto out;
2976 }
2977 }
2978
2979 if (base_fmt) {
2980 if (set_option_parameter(param, BLOCK_OPT_BACKING_FMT, base_fmt)) {
2981 error_report("Backing file format not supported for file "
2982 "format '%s'", fmt);
4f70f249 2983 ret = -EINVAL;
f88e1a42
JS
2984 goto out;
2985 }
2986 }
2987
792da93a
JS
2988 backing_file = get_option_parameter(param, BLOCK_OPT_BACKING_FILE);
2989 if (backing_file && backing_file->value.s) {
2990 if (!strcmp(filename, backing_file->value.s)) {
2991 error_report("Error: Trying to create an image with the "
2992 "same filename as the backing file");
4f70f249 2993 ret = -EINVAL;
792da93a
JS
2994 goto out;
2995 }
2996 }
2997
f88e1a42
JS
2998 backing_fmt = get_option_parameter(param, BLOCK_OPT_BACKING_FMT);
2999 if (backing_fmt && backing_fmt->value.s) {
96df67d1
SH
3000 backing_drv = bdrv_find_format(backing_fmt->value.s);
3001 if (!backing_drv) {
f88e1a42
JS
3002 error_report("Unknown backing file format '%s'",
3003 backing_fmt->value.s);
4f70f249 3004 ret = -EINVAL;
f88e1a42
JS
3005 goto out;
3006 }
3007 }
3008
3009 // The size for the image must always be specified, with one exception:
3010 // If we are using a backing file, we can obtain the size from there
d220894e
KW
3011 size = get_option_parameter(param, BLOCK_OPT_SIZE);
3012 if (size && size->value.n == -1) {
f88e1a42
JS
3013 if (backing_file && backing_file->value.s) {
3014 uint64_t size;
f88e1a42
JS
3015 char buf[32];
3016
f88e1a42
JS
3017 bs = bdrv_new("");
3018
96df67d1 3019 ret = bdrv_open(bs, backing_file->value.s, flags, backing_drv);
f88e1a42 3020 if (ret < 0) {
96df67d1 3021 error_report("Could not open '%s'", backing_file->value.s);
f88e1a42
JS
3022 goto out;
3023 }
3024 bdrv_get_geometry(bs, &size);
3025 size *= 512;
3026
3027 snprintf(buf, sizeof(buf), "%" PRId64, size);
3028 set_option_parameter(param, BLOCK_OPT_SIZE, buf);
3029 } else {
3030 error_report("Image creation needs a size parameter");
4f70f249 3031 ret = -EINVAL;
f88e1a42
JS
3032 goto out;
3033 }
3034 }
3035
3036 printf("Formatting '%s', fmt=%s ", filename, fmt);
3037 print_option_parameters(param);
3038 puts("");
3039
3040 ret = bdrv_create(drv, filename, param);
3041
3042 if (ret < 0) {
3043 if (ret == -ENOTSUP) {
3044 error_report("Formatting or formatting option not supported for "
3045 "file format '%s'", fmt);
3046 } else if (ret == -EFBIG) {
3047 error_report("The image size is too large for file format '%s'",
3048 fmt);
3049 } else {
3050 error_report("%s: error while creating %s: %s", filename, fmt,
3051 strerror(-ret));
3052 }
3053 }
3054
3055out:
3056 free_option_parameters(create_options);
3057 free_option_parameters(param);
3058
3059 if (bs) {
3060 bdrv_delete(bs);
3061 }
4f70f249
JS
3062
3063 return ret;
f88e1a42 3064}