?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
block: Don't let locked flag prevent medium load
[qemu/kevin.git] / block.c
blob8859f9b414b1e484c93c5777c783cafadc060456
1 /*
2 * QEMU System Emulator block driver
3 *
4 * Copyright (c) 2003 Fabrice Bellard
5 *
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 */
24 #include "config-host.h"
25 #include "qemu-common.h"
26 #include "trace.h"
27 #include "monitor.h"
28 #include "block_int.h"
29 #include "module.h"
30 #include "qemu-objects.h"
31
32 #ifdef CONFIG_BSD
33 #include <sys/types.h>
34 #include <sys/stat.h>
35 #include <sys/ioctl.h>
36 #include <sys/queue.h>
37 #ifndef __DragonFly__
38 #include <sys/disk.h>
39 #endif
40 #endif
41
42 #ifdef _WIN32
43 #include <windows.h>
44 #endif
45
46 static BlockDriverAIOCB *bdrv_aio_readv_em(BlockDriverState *bs,
47 int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
48 BlockDriverCompletionFunc *cb, void *opaque);
49 static BlockDriverAIOCB *bdrv_aio_writev_em(BlockDriverState *bs,
50 int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
51 BlockDriverCompletionFunc *cb, void *opaque);
52 static BlockDriverAIOCB *bdrv_aio_flush_em(BlockDriverState *bs,
53 BlockDriverCompletionFunc *cb, void *opaque);
54 static BlockDriverAIOCB *bdrv_aio_noop_em(BlockDriverState *bs,
55 BlockDriverCompletionFunc *cb, void *opaque);
56 static int bdrv_read_em(BlockDriverState *bs, int64_t sector_num,
57 uint8_t *buf, int nb_sectors);
58 static int bdrv_write_em(BlockDriverState *bs, int64_t sector_num,
59 const uint8_t *buf, int nb_sectors);
60
61 static QTAILQ_HEAD(, BlockDriverState) bdrv_states =
62 QTAILQ_HEAD_INITIALIZER(bdrv_states);
63
64 static QLIST_HEAD(, BlockDriver) bdrv_drivers =
65 QLIST_HEAD_INITIALIZER(bdrv_drivers);
66
67 /* The device to use for VM snapshots */
68 static BlockDriverState *bs_snapshots;
69
70 /* If non-zero, use only whitelisted block drivers */
71 static int use_bdrv_whitelist;
72
73 #ifdef _WIN32
74 static 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
81 int 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>:" */
94 static 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
106 int path_is_absolute(const char *path)
107 {
108 const char *p;
109 #ifdef _WIN32
110 /* specific case for names like: "\\.\d:" */
111 if (*path == '/' || *path == '\\')
112 return 1;
113 #endif
114 p = strchr(path, ':');
115 if (p)
116 p++;
117 else
118 p = path;
119 #ifdef _WIN32
120 return (*p == '/' || *p == '\\');
121 #else
122 return (*p == '/');
123 #endif
124 }
125
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. */
129 void path_combine(char *dest, int dest_size,
130 const char *base_path,
131 const char *filename)
132 {
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;
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
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);
167 }
168 }
169
170 void bdrv_register(BlockDriver *bdrv)
171 {
172 if (!bdrv->bdrv_aio_readv) {
173 /* add AIO emulation layer */
174 bdrv->bdrv_aio_readv = bdrv_aio_readv_em;
175 bdrv->bdrv_aio_writev = bdrv_aio_writev_em;
176 } else if (!bdrv->bdrv_read) {
177 /* add synchronous IO emulation layer */
178 bdrv->bdrv_read = bdrv_read_em;
179 bdrv->bdrv_write = bdrv_write_em;
180 }
181
182 if (!bdrv->bdrv_aio_flush)
183 bdrv->bdrv_aio_flush = bdrv_aio_flush_em;
184
185 QLIST_INSERT_HEAD(&bdrv_drivers, bdrv, list);
186 }
187
188 /* create a new block device (by default it is empty) */
189 BlockDriverState *bdrv_new(const char *device_name)
190 {
191 BlockDriverState *bs;
192
193 bs = qemu_mallocz(sizeof(BlockDriverState));
194 pstrcpy(bs->device_name, sizeof(bs->device_name), device_name);
195 if (device_name[0] != '\0') {
196 QTAILQ_INSERT_TAIL(&bdrv_states, bs, list);
197 }
198 return bs;
199 }
200
201 BlockDriver *bdrv_find_format(const char *format_name)
202 {
203 BlockDriver *drv1;
204 QLIST_FOREACH(drv1, &bdrv_drivers, list) {
205 if (!strcmp(drv1->format_name, format_name)) {
206 return drv1;
207 }
208 }
209 return NULL;
210 }
211
212 static 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
230 BlockDriver *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
236 int bdrv_create(BlockDriver *drv, const char* filename,
237 QEMUOptionParameter *options)
238 {
239 if (!drv->bdrv_create)
240 return -ENOTSUP;
241
242 return drv->bdrv_create(filename, options);
243 }
244
245 int bdrv_create_file(const char* filename, QEMUOptionParameter *options)
246 {
247 BlockDriver *drv;
248
249 drv = bdrv_find_protocol(filename);
250 if (drv == NULL) {
251 return -ENOENT;
252 }
253
254 return bdrv_create(drv, filename, options);
255 }
256
257 #ifdef _WIN32
258 void get_tmp_filename(char *filename, int size)
259 {
260 char temp_dir[MAX_PATH];
261
262 GetTempPath(MAX_PATH, temp_dir);
263 GetTempFileName(temp_dir, "qem", 0, filename);
264 }
265 #else
266 void get_tmp_filename(char *filename, int size)
267 {
268 int fd;
269 const char *tmpdir;
270 /* XXX: race condition possible */
271 tmpdir = getenv("TMPDIR");
272 if (!tmpdir)
273 tmpdir = "/tmp";
274 snprintf(filename, size, "%s/vl.XXXXXX", tmpdir);
275 fd = mkstemp(filename);
276 close(fd);
277 }
278 #endif
279
280 /*
281 * Detect host devices. By convention, /dev/cdrom[N] is always
282 * recognized as a host CDROM.
283 */
284 static 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
302 BlockDriver *bdrv_find_protocol(const char *filename)
303 {
304 BlockDriver *drv1;
305 char protocol[128];
306 int len;
307 const char *p;
308
309 /* TODO Drivers without bdrv_file_open must be specified explicitly */
310
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
323 if (!path_has_protocol(filename)) {
324 return bdrv_find_format("file");
325 }
326 p = strchr(filename, ':');
327 assert(p != NULL);
328 len = p - filename;
329 if (len > sizeof(protocol) - 1)
330 len = sizeof(protocol) - 1;
331 memcpy(protocol, filename, len);
332 protocol[len] = '\0';
333 QLIST_FOREACH(drv1, &bdrv_drivers, list) {
334 if (drv1->protocol_name &&
335 !strcmp(drv1->protocol_name, protocol)) {
336 return drv1;
337 }
338 }
339 return NULL;
340 }
341
342 static int find_image_format(const char *filename, BlockDriver **pdrv)
343 {
344 int ret, score, score_max;
345 BlockDriver *drv1, *drv;
346 uint8_t buf[2048];
347 BlockDriverState *bs;
348
349 ret = bdrv_file_open(&bs, filename, 0);
350 if (ret < 0) {
351 *pdrv = NULL;
352 return ret;
353 }
354
355 /* Return the raw BlockDriver * to scsi-generic devices or empty drives */
356 if (bs->sg || !bdrv_is_inserted(bs)) {
357 bdrv_delete(bs);
358 drv = bdrv_find_format("raw");
359 if (!drv) {
360 ret = -ENOENT;
361 }
362 *pdrv = drv;
363 return ret;
364 }
365
366 ret = bdrv_pread(bs, 0, buf, sizeof(buf));
367 bdrv_delete(bs);
368 if (ret < 0) {
369 *pdrv = NULL;
370 return ret;
371 }
372
373 score_max = 0;
374 drv = NULL;
375 QLIST_FOREACH(drv1, &bdrv_drivers, list) {
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 }
382 }
383 }
384 if (!drv) {
385 ret = -ENOENT;
386 }
387 *pdrv = drv;
388 return ret;
389 }
390
391 /**
392 * Set the current 'total_sectors' value
393 */
394 static int refresh_total_sectors(BlockDriverState *bs, int64_t hint)
395 {
396 BlockDriver *drv = bs->drv;
397
398 /* Do not attempt drv->bdrv_getlength() on scsi-generic devices */
399 if (bs->sg)
400 return 0;
401
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
415 /*
416 * Common part for opening disk images and files
417 */
418 static 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
425 bs->file = NULL;
426 bs->total_sectors = 0;
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
442 if (flags & BDRV_O_CACHE_WB)
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 /*
452 * Snapshots should be writable.
453 */
454 if (bs->is_temporary) {
455 open_flags |= BDRV_O_RDWR;
456 }
457
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
468 if (ret < 0) {
469 goto free_and_fail;
470 }
471
472 bs->keep_read_only = bs->read_only = !(open_flags & BDRV_O_RDWR);
473
474 ret = refresh_total_sectors(bs, bs->total_sectors);
475 if (ret < 0) {
476 goto free_and_fail;
477 }
478
479 #ifndef _WIN32
480 if (bs->is_temporary) {
481 unlink(filename);
482 }
483 #endif
484 return 0;
485
486 free_and_fail:
487 if (bs->file) {
488 bdrv_delete(bs->file);
489 bs->file = NULL;
490 }
491 qemu_free(bs->opaque);
492 bs->opaque = NULL;
493 bs->drv = NULL;
494 return ret;
495 }
496
497 /*
498 * Opens a file using a protocol (file, host_device, nbd, ...)
499 */
500 int bdrv_file_open(BlockDriverState **pbs, const char *filename, int flags)
501 {
502 BlockDriverState *bs;
503 BlockDriver *drv;
504 int ret;
505
506 drv = bdrv_find_protocol(filename);
507 if (!drv) {
508 return -ENOENT;
509 }
510
511 bs = bdrv_new("");
512 ret = bdrv_open_common(bs, filename, flags, drv);
513 if (ret < 0) {
514 bdrv_delete(bs);
515 return ret;
516 }
517 bs->growable = 1;
518 *pbs = bs;
519 return 0;
520 }
521
522 /*
523 * Opens a disk image (raw, qcow2, vmdk, ...)
524 */
525 int bdrv_open(BlockDriverState *bs, const char *filename, int flags,
526 BlockDriver *drv)
527 {
528 int ret;
529
530 if (flags & BDRV_O_SNAPSHOT) {
531 BlockDriverState *bs1;
532 int64_t total_size;
533 int is_protocol = 0;
534 BlockDriver *bdrv_qcow2;
535 QEMUOptionParameter *options;
536 char tmp_filename[PATH_MAX];
537 char backing_filename[PATH_MAX];
538
539 /* if snapshot, we create a temporary backing file and open it
540 instead of opening 'filename' directly */
541
542 /* if there is a backing file, use it */
543 bs1 = bdrv_new("");
544 ret = bdrv_open(bs1, filename, 0, drv);
545 if (ret < 0) {
546 bdrv_delete(bs1);
547 return ret;
548 }
549 total_size = bdrv_getlength(bs1) & BDRV_SECTOR_MASK;
550
551 if (bs1->drv && bs1->drv->protocol_name)
552 is_protocol = 1;
553
554 bdrv_delete(bs1);
555
556 get_tmp_filename(tmp_filename, sizeof(tmp_filename));
557
558 /* Real path is meaningless for protocols */
559 if (is_protocol)
560 snprintf(backing_filename, sizeof(backing_filename),
561 "%s", filename);
562 else if (!realpath(filename, backing_filename))
563 return -errno;
564
565 bdrv_qcow2 = bdrv_find_format("qcow2");
566 options = parse_option_parameters("", bdrv_qcow2->create_options, NULL);
567
568 set_option_parameter_int(options, BLOCK_OPT_SIZE, total_size);
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);
576 free_option_parameters(options);
577 if (ret < 0) {
578 return ret;
579 }
580
581 filename = tmp_filename;
582 drv = bdrv_qcow2;
583 bs->is_temporary = 1;
584 }
585
586 /* Find the right image format driver */
587 if (!drv) {
588 ret = find_image_format(filename, &drv);
589 }
590
591 if (!drv) {
592 goto unlink_and_fail;
593 }
594
595 /* Open the image */
596 ret = bdrv_open_common(bs, filename, flags, drv);
597 if (ret < 0) {
598 goto unlink_and_fail;
599 }
600
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("");
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') {
618 back_drv = bdrv_find_format(bs->backing_format);
619 }
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)
642 bs->change_cb(bs->change_opaque, CHANGE_MEDIA);
643 }
644
645 return 0;
646
647 unlink_and_fail:
648 if (bs->is_temporary) {
649 unlink(filename);
650 }
651 return ret;
652 }
653
654 void bdrv_close(BlockDriverState *bs)
655 {
656 if (bs->drv) {
657 if (bs == bs_snapshots) {
658 bs_snapshots = NULL;
659 }
660 if (bs->backing_hd) {
661 bdrv_delete(bs->backing_hd);
662 bs->backing_hd = NULL;
663 }
664 bs->drv->bdrv_close(bs);
665 qemu_free(bs->opaque);
666 #ifdef _WIN32
667 if (bs->is_temporary) {
668 unlink(bs->filename);
669 }
670 #endif
671 bs->opaque = NULL;
672 bs->drv = NULL;
673
674 if (bs->file != NULL) {
675 bdrv_close(bs->file);
676 }
677
678 /* call the change callback */
679 bs->media_changed = 1;
680 if (bs->change_cb)
681 bs->change_cb(bs->change_opaque, CHANGE_MEDIA);
682 }
683 }
684
685 void bdrv_close_all(void)
686 {
687 BlockDriverState *bs;
688
689 QTAILQ_FOREACH(bs, &bdrv_states, list) {
690 bdrv_close(bs);
691 }
692 }
693
694 /* make a BlockDriverState anonymous by removing from bdrv_state list.
695 Also, NULL terminate the device_name to prevent double remove */
696 void 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
704 void bdrv_delete(BlockDriverState *bs)
705 {
706 assert(!bs->peer);
707
708 /* remove from list, if necessary */
709 bdrv_make_anon(bs);
710
711 bdrv_close(bs);
712 if (bs->file != NULL) {
713 bdrv_delete(bs->file);
714 }
715
716 assert(bs != bs_snapshots);
717 qemu_free(bs);
718 }
719
720 int bdrv_attach(BlockDriverState *bs, DeviceState *qdev)
721 {
722 if (bs->peer) {
723 return -EBUSY;
724 }
725 bs->peer = qdev;
726 return 0;
727 }
728
729 void bdrv_detach(BlockDriverState *bs, DeviceState *qdev)
730 {
731 assert(bs->peer == qdev);
732 bs->peer = NULL;
733 bs->change_cb = NULL;
734 bs->change_opaque = NULL;
735 }
736
737 DeviceState *bdrv_get_attached(BlockDriverState *bs)
738 {
739 return bs->peer;
740 }
741
742 /*
743 * Run consistency checks on an image
744 *
745 * Returns 0 if the check could be completed (it doesn't mean that the image is
746 * free of errors) or -errno when an internal error occurred. The results of the
747 * check are stored in res.
748 */
749 int bdrv_check(BlockDriverState *bs, BdrvCheckResult *res)
750 {
751 if (bs->drv->bdrv_check == NULL) {
752 return -ENOTSUP;
753 }
754
755 memset(res, 0, sizeof(*res));
756 return bs->drv->bdrv_check(bs, res);
757 }
758
759 #define COMMIT_BUF_SECTORS 2048
760
761 /* commit COW file into the raw image */
762 int bdrv_commit(BlockDriverState *bs)
763 {
764 BlockDriver *drv = bs->drv;
765 BlockDriver *backing_drv;
766 int64_t sector, total_sectors;
767 int n, ro, open_flags;
768 int ret = 0, rw_ret = 0;
769 uint8_t *buf;
770 char filename[1024];
771 BlockDriverState *bs_rw, *bs_ro;
772
773 if (!drv)
774 return -ENOMEDIUM;
775
776 if (!bs->backing_hd) {
777 return -ENOTSUP;
778 }
779
780 if (bs->backing_hd->keep_read_only) {
781 return -EACCES;
782 }
783
784 backing_drv = bs->backing_hd->drv;
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("");
794 rw_ret = bdrv_open(bs_rw, filename, open_flags | BDRV_O_RDWR,
795 backing_drv);
796 if (rw_ret < 0) {
797 bdrv_delete(bs_rw);
798 /* try to re-open read-only */
799 bs_ro = bdrv_new("");
800 ret = bdrv_open(bs_ro, filename, open_flags & ~BDRV_O_RDWR,
801 backing_drv);
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;
812 }
813
814 total_sectors = bdrv_getlength(bs) >> BDRV_SECTOR_BITS;
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 }
829 }
830 }
831
832 if (drv->bdrv_make_empty) {
833 ret = drv->bdrv_make_empty(bs);
834 bdrv_flush(bs);
835 }
836
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);
843
844 ro_cleanup:
845 qemu_free(buf);
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("");
852 ret = bdrv_open(bs_ro, filename, open_flags & ~BDRV_O_RDWR,
853 backing_drv);
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
864 return ret;
865 }
866
867 void bdrv_commit_all(void)
868 {
869 BlockDriverState *bs;
870
871 QTAILQ_FOREACH(bs, &bdrv_states, list) {
872 bdrv_commit(bs);
873 }
874 }
875
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 */
884 int 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
896 static 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
909 if (offset < 0)
910 return -EIO;
911
912 if ((offset > len) || (len - offset < size))
913 return -EIO;
914
915 return 0;
916 }
917
918 static int bdrv_check_request(BlockDriverState *bs, int64_t sector_num,
919 int nb_sectors)
920 {
921 return bdrv_check_byte_request(bs, sector_num * BDRV_SECTOR_SIZE,
922 nb_sectors * BDRV_SECTOR_SIZE);
923 }
924
925 /* return < 0 if error. See bdrv_write() for the return codes */
926 int bdrv_read(BlockDriverState *bs, int64_t sector_num,
927 uint8_t *buf, int nb_sectors)
928 {
929 BlockDriver *drv = bs->drv;
930
931 if (!drv)
932 return -ENOMEDIUM;
933 if (bdrv_check_request(bs, sector_num, nb_sectors))
934 return -EIO;
935
936 return drv->bdrv_read(bs, sector_num, buf, nb_sectors);
937 }
938
939 static void set_dirty_bitmap(BlockDriverState *bs, int64_t sector_num,
940 int nb_sectors, int dirty)
941 {
942 int64_t start, end;
943 unsigned long val, idx, bit;
944
945 start = sector_num / BDRV_SECTORS_PER_DIRTY_CHUNK;
946 end = (sector_num + nb_sectors - 1) / BDRV_SECTORS_PER_DIRTY_CHUNK;
947
948 for (; start <= end; start++) {
949 idx = start / (sizeof(unsigned long) * 8);
950 bit = start % (sizeof(unsigned long) * 8);
951 val = bs->dirty_bitmap[idx];
952 if (dirty) {
953 if (!(val & (1UL << bit))) {
954 bs->dirty_count++;
955 val |= 1UL << bit;
956 }
957 } else {
958 if (val & (1UL << bit)) {
959 bs->dirty_count--;
960 val &= ~(1UL << bit);
961 }
962 }
963 bs->dirty_bitmap[idx] = val;
964 }
965 }
966
967 /* Return < 0 if error. Important errors are:
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 */
973 int bdrv_write(BlockDriverState *bs, int64_t sector_num,
974 const uint8_t *buf, int nb_sectors)
975 {
976 BlockDriver *drv = bs->drv;
977 if (!bs->drv)
978 return -ENOMEDIUM;
979 if (bs->read_only)
980 return -EACCES;
981 if (bdrv_check_request(bs, sector_num, nb_sectors))
982 return -EIO;
983
984 if (bs->dirty_bitmap) {
985 set_dirty_bitmap(bs, sector_num, nb_sectors, 1);
986 }
987
988 if (bs->wr_highest_sector < sector_num + nb_sectors - 1) {
989 bs->wr_highest_sector = sector_num + nb_sectors - 1;
990 }
991
992 return drv->bdrv_write(bs, sector_num, buf, nb_sectors);
993 }
994
995 int bdrv_pread(BlockDriverState *bs, int64_t offset,
996 void *buf, int count1)
997 {
998 uint8_t tmp_buf[BDRV_SECTOR_SIZE];
999 int len, nb_sectors, count;
1000 int64_t sector_num;
1001 int ret;
1002
1003 count = count1;
1004 /* first read to align to sector start */
1005 len = (BDRV_SECTOR_SIZE - offset) & (BDRV_SECTOR_SIZE - 1);
1006 if (len > count)
1007 len = count;
1008 sector_num = offset >> BDRV_SECTOR_BITS;
1009 if (len > 0) {
1010 if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0)
1011 return ret;
1012 memcpy(buf, tmp_buf + (offset & (BDRV_SECTOR_SIZE - 1)), len);
1013 count -= len;
1014 if (count == 0)
1015 return count1;
1016 sector_num++;
1017 buf += len;
1018 }
1019
1020 /* read the sectors "in place" */
1021 nb_sectors = count >> BDRV_SECTOR_BITS;
1022 if (nb_sectors > 0) {
1023 if ((ret = bdrv_read(bs, sector_num, buf, nb_sectors)) < 0)
1024 return ret;
1025 sector_num += nb_sectors;
1026 len = nb_sectors << BDRV_SECTOR_BITS;
1027 buf += len;
1028 count -= len;
1029 }
1030
1031 /* add data from the last sector */
1032 if (count > 0) {
1033 if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0)
1034 return ret;
1035 memcpy(buf, tmp_buf, count);
1036 }
1037 return count1;
1038 }
1039
1040 int bdrv_pwrite(BlockDriverState *bs, int64_t offset,
1041 const void *buf, int count1)
1042 {
1043 uint8_t tmp_buf[BDRV_SECTOR_SIZE];
1044 int len, nb_sectors, count;
1045 int64_t sector_num;
1046 int ret;
1047
1048 count = count1;
1049 /* first write to align to sector start */
1050 len = (BDRV_SECTOR_SIZE - offset) & (BDRV_SECTOR_SIZE - 1);
1051 if (len > count)
1052 len = count;
1053 sector_num = offset >> BDRV_SECTOR_BITS;
1054 if (len > 0) {
1055 if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0)
1056 return ret;
1057 memcpy(tmp_buf + (offset & (BDRV_SECTOR_SIZE - 1)), buf, len);
1058 if ((ret = bdrv_write(bs, sector_num, tmp_buf, 1)) < 0)
1059 return ret;
1060 count -= len;
1061 if (count == 0)
1062 return count1;
1063 sector_num++;
1064 buf += len;
1065 }
1066
1067 /* write the sectors "in place" */
1068 nb_sectors = count >> BDRV_SECTOR_BITS;
1069 if (nb_sectors > 0) {
1070 if ((ret = bdrv_write(bs, sector_num, buf, nb_sectors)) < 0)
1071 return ret;
1072 sector_num += nb_sectors;
1073 len = nb_sectors << BDRV_SECTOR_BITS;
1074 buf += len;
1075 count -= len;
1076 }
1077
1078 /* add data from the last sector */
1079 if (count > 0) {
1080 if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0)
1081 return ret;
1082 memcpy(tmp_buf, buf, count);
1083 if ((ret = bdrv_write(bs, sector_num, tmp_buf, 1)) < 0)
1084 return ret;
1085 }
1086 return count1;
1087 }
1088
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 */
1095 int 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
1113 /*
1114 * Writes to the file and ensures that no writes are reordered across this
1115 * request (acts as a barrier)
1116 *
1117 * Returns 0 on success, -errno in error cases.
1118 */
1119 int bdrv_write_sync(BlockDriverState *bs, int64_t sector_num,
1120 const uint8_t *buf, int nb_sectors)
1121 {
1122 return bdrv_pwrite_sync(bs, BDRV_SECTOR_SIZE * sector_num,
1123 buf, BDRV_SECTOR_SIZE * nb_sectors);
1124 }
1125
1126 /**
1127 * Truncate file to 'offset' bytes (needed only for file protocols)
1128 */
1129 int bdrv_truncate(BlockDriverState *bs, int64_t offset)
1130 {
1131 BlockDriver *drv = bs->drv;
1132 int ret;
1133 if (!drv)
1134 return -ENOMEDIUM;
1135 if (!drv->bdrv_truncate)
1136 return -ENOTSUP;
1137 if (bs->read_only)
1138 return -EACCES;
1139 if (bdrv_in_use(bs))
1140 return -EBUSY;
1141 ret = drv->bdrv_truncate(bs, offset);
1142 if (ret == 0) {
1143 ret = refresh_total_sectors(bs, offset >> BDRV_SECTOR_BITS);
1144 if (bs->change_cb) {
1145 bs->change_cb(bs->change_opaque, CHANGE_SIZE);
1146 }
1147 }
1148 return ret;
1149 }
1150
1151 /**
1152 * Length of a allocated file in bytes. Sparse files are counted by actual
1153 * allocated space. Return < 0 if error or unknown.
1154 */
1155 int64_t bdrv_get_allocated_file_size(BlockDriverState *bs)
1156 {
1157 BlockDriver *drv = bs->drv;
1158 if (!drv) {
1159 return -ENOMEDIUM;
1160 }
1161 if (drv->bdrv_get_allocated_file_size) {
1162 return drv->bdrv_get_allocated_file_size(bs);
1163 }
1164 if (bs->file) {
1165 return bdrv_get_allocated_file_size(bs->file);
1166 }
1167 return -ENOTSUP;
1168 }
1169
1170 /**
1171 * Length of a file in bytes. Return < 0 if error or unknown.
1172 */
1173 int64_t bdrv_getlength(BlockDriverState *bs)
1174 {
1175 BlockDriver *drv = bs->drv;
1176 if (!drv)
1177 return -ENOMEDIUM;
1178
1179 if (bs->growable || bs->removable) {
1180 if (drv->bdrv_getlength) {
1181 return drv->bdrv_getlength(bs);
1182 }
1183 }
1184 return bs->total_sectors * BDRV_SECTOR_SIZE;
1185 }
1186
1187 /* return 0 as number of sectors if no device present or error */
1188 void bdrv_get_geometry(BlockDriverState *bs, uint64_t *nb_sectors_ptr)
1189 {
1190 int64_t length;
1191 length = bdrv_getlength(bs);
1192 if (length < 0)
1193 length = 0;
1194 else
1195 length = length >> BDRV_SECTOR_BITS;
1196 *nb_sectors_ptr = length;
1197 }
1198
1199 struct partition {
1200 uint8_t boot_ind; /* 0x80 - active */
1201 uint8_t head; /* starting head */
1202 uint8_t sector; /* starting sector */
1203 uint8_t cyl; /* starting cylinder */
1204 uint8_t sys_ind; /* What partition type */
1205 uint8_t end_head; /* end head */
1206 uint8_t end_sector; /* end sector */
1207 uint8_t end_cyl; /* end cylinder */
1208 uint32_t start_sect; /* starting sector counting from 0 */
1209 uint32_t nr_sects; /* nr of sectors in partition */
1210 } __attribute__((packed));
1211
1212 /* try to guess the disk logical geometry from the MSDOS partition table. Return 0 if OK, -1 if could not guess */
1213 static int guess_disk_lchs(BlockDriverState *bs,
1214 int *pcylinders, int *pheads, int *psectors)
1215 {
1216 uint8_t buf[BDRV_SECTOR_SIZE];
1217 int ret, i, heads, sectors, cylinders;
1218 struct partition *p;
1219 uint32_t nr_sects;
1220 uint64_t nb_sectors;
1221
1222 bdrv_get_geometry(bs, &nb_sectors);
1223
1224 ret = bdrv_read(bs, 0, buf, 1);
1225 if (ret < 0)
1226 return -1;
1227 /* test msdos magic */
1228 if (buf[510] != 0x55 || buf[511] != 0xaa)
1229 return -1;
1230 for(i = 0; i < 4; i++) {
1231 p = ((struct partition *)(buf + 0x1be)) + i;
1232 nr_sects = le32_to_cpu(p->nr_sects);
1233 if (nr_sects && p->end_head) {
1234 /* We make the assumption that the partition terminates on
1235 a cylinder boundary */
1236 heads = p->end_head + 1;
1237 sectors = p->end_sector & 63;
1238 if (sectors == 0)
1239 continue;
1240 cylinders = nb_sectors / (heads * sectors);
1241 if (cylinders < 1 || cylinders > 16383)
1242 continue;
1243 *pheads = heads;
1244 *psectors = sectors;
1245 *pcylinders = cylinders;
1246 #if 0
1247 printf("guessed geometry: LCHS=%d %d %d\n",
1248 cylinders, heads, sectors);
1249 #endif
1250 return 0;
1251 }
1252 }
1253 return -1;
1254 }
1255
1256 void bdrv_guess_geometry(BlockDriverState *bs, int *pcyls, int *pheads, int *psecs)
1257 {
1258 int translation, lba_detected = 0;
1259 int cylinders, heads, secs;
1260 uint64_t nb_sectors;
1261
1262 /* if a geometry hint is available, use it */
1263 bdrv_get_geometry(bs, &nb_sectors);
1264 bdrv_get_geometry_hint(bs, &cylinders, &heads, &secs);
1265 translation = bdrv_get_translation_hint(bs);
1266 if (cylinders != 0) {
1267 *pcyls = cylinders;
1268 *pheads = heads;
1269 *psecs = secs;
1270 } else {
1271 if (guess_disk_lchs(bs, &cylinders, &heads, &secs) == 0) {
1272 if (heads > 16) {
1273 /* if heads > 16, it means that a BIOS LBA
1274 translation was active, so the default
1275 hardware geometry is OK */
1276 lba_detected = 1;
1277 goto default_geometry;
1278 } else {
1279 *pcyls = cylinders;
1280 *pheads = heads;
1281 *psecs = secs;
1282 /* disable any translation to be in sync with
1283 the logical geometry */
1284 if (translation == BIOS_ATA_TRANSLATION_AUTO) {
1285 bdrv_set_translation_hint(bs,
1286 BIOS_ATA_TRANSLATION_NONE);
1287 }
1288 }
1289 } else {
1290 default_geometry:
1291 /* if no geometry, use a standard physical disk geometry */
1292 cylinders = nb_sectors / (16 * 63);
1293
1294 if (cylinders > 16383)
1295 cylinders = 16383;
1296 else if (cylinders < 2)
1297 cylinders = 2;
1298 *pcyls = cylinders;
1299 *pheads = 16;
1300 *psecs = 63;
1301 if ((lba_detected == 1) && (translation == BIOS_ATA_TRANSLATION_AUTO)) {
1302 if ((*pcyls * *pheads) <= 131072) {
1303 bdrv_set_translation_hint(bs,
1304 BIOS_ATA_TRANSLATION_LARGE);
1305 } else {
1306 bdrv_set_translation_hint(bs,
1307 BIOS_ATA_TRANSLATION_LBA);
1308 }
1309 }
1310 }
1311 bdrv_set_geometry_hint(bs, *pcyls, *pheads, *psecs);
1312 }
1313 }
1314
1315 void bdrv_set_geometry_hint(BlockDriverState *bs,
1316 int cyls, int heads, int secs)
1317 {
1318 bs->cyls = cyls;
1319 bs->heads = heads;
1320 bs->secs = secs;
1321 }
1322
1323 void bdrv_set_translation_hint(BlockDriverState *bs, int translation)
1324 {
1325 bs->translation = translation;
1326 }
1327
1328 void bdrv_get_geometry_hint(BlockDriverState *bs,
1329 int *pcyls, int *pheads, int *psecs)
1330 {
1331 *pcyls = bs->cyls;
1332 *pheads = bs->heads;
1333 *psecs = bs->secs;
1334 }
1335
1336 /* Recognize floppy formats */
1337 typedef struct FDFormat {
1338 FDriveType drive;
1339 uint8_t last_sect;
1340 uint8_t max_track;
1341 uint8_t max_head;
1342 } FDFormat;
1343
1344 static const FDFormat fd_formats[] = {
1345 /* First entry is default format */
1346 /* 1.44 MB 3"1/2 floppy disks */
1347 { FDRIVE_DRV_144, 18, 80, 1, },
1348 { FDRIVE_DRV_144, 20, 80, 1, },
1349 { FDRIVE_DRV_144, 21, 80, 1, },
1350 { FDRIVE_DRV_144, 21, 82, 1, },
1351 { FDRIVE_DRV_144, 21, 83, 1, },
1352 { FDRIVE_DRV_144, 22, 80, 1, },
1353 { FDRIVE_DRV_144, 23, 80, 1, },
1354 { FDRIVE_DRV_144, 24, 80, 1, },
1355 /* 2.88 MB 3"1/2 floppy disks */
1356 { FDRIVE_DRV_288, 36, 80, 1, },
1357 { FDRIVE_DRV_288, 39, 80, 1, },
1358 { FDRIVE_DRV_288, 40, 80, 1, },
1359 { FDRIVE_DRV_288, 44, 80, 1, },
1360 { FDRIVE_DRV_288, 48, 80, 1, },
1361 /* 720 kB 3"1/2 floppy disks */
1362 { FDRIVE_DRV_144, 9, 80, 1, },
1363 { FDRIVE_DRV_144, 10, 80, 1, },
1364 { FDRIVE_DRV_144, 10, 82, 1, },
1365 { FDRIVE_DRV_144, 10, 83, 1, },
1366 { FDRIVE_DRV_144, 13, 80, 1, },
1367 { FDRIVE_DRV_144, 14, 80, 1, },
1368 /* 1.2 MB 5"1/4 floppy disks */
1369 { FDRIVE_DRV_120, 15, 80, 1, },
1370 { FDRIVE_DRV_120, 18, 80, 1, },
1371 { FDRIVE_DRV_120, 18, 82, 1, },
1372 { FDRIVE_DRV_120, 18, 83, 1, },
1373 { FDRIVE_DRV_120, 20, 80, 1, },
1374 /* 720 kB 5"1/4 floppy disks */
1375 { FDRIVE_DRV_120, 9, 80, 1, },
1376 { FDRIVE_DRV_120, 11, 80, 1, },
1377 /* 360 kB 5"1/4 floppy disks */
1378 { FDRIVE_DRV_120, 9, 40, 1, },
1379 { FDRIVE_DRV_120, 9, 40, 0, },
1380 { FDRIVE_DRV_120, 10, 41, 1, },
1381 { FDRIVE_DRV_120, 10, 42, 1, },
1382 /* 320 kB 5"1/4 floppy disks */
1383 { FDRIVE_DRV_120, 8, 40, 1, },
1384 { FDRIVE_DRV_120, 8, 40, 0, },
1385 /* 360 kB must match 5"1/4 better than 3"1/2... */
1386 { FDRIVE_DRV_144, 9, 80, 0, },
1387 /* end */
1388 { FDRIVE_DRV_NONE, -1, -1, 0, },
1389 };
1390
1391 void bdrv_get_floppy_geometry_hint(BlockDriverState *bs, int *nb_heads,
1392 int *max_track, int *last_sect,
1393 FDriveType drive_in, FDriveType *drive)
1394 {
1395 const FDFormat *parse;
1396 uint64_t nb_sectors, size;
1397 int i, first_match, match;
1398
1399 bdrv_get_geometry_hint(bs, nb_heads, max_track, last_sect);
1400 if (*nb_heads != 0 && *max_track != 0 && *last_sect != 0) {
1401 /* User defined disk */
1402 } else {
1403 bdrv_get_geometry(bs, &nb_sectors);
1404 match = -1;
1405 first_match = -1;
1406 for (i = 0; ; i++) {
1407 parse = &fd_formats[i];
1408 if (parse->drive == FDRIVE_DRV_NONE) {
1409 break;
1410 }
1411 if (drive_in == parse->drive ||
1412 drive_in == FDRIVE_DRV_NONE) {
1413 size = (parse->max_head + 1) * parse->max_track *
1414 parse->last_sect;
1415 if (nb_sectors == size) {
1416 match = i;
1417 break;
1418 }
1419 if (first_match == -1) {
1420 first_match = i;
1421 }
1422 }
1423 }
1424 if (match == -1) {
1425 if (first_match == -1) {
1426 match = 1;
1427 } else {
1428 match = first_match;
1429 }
1430 parse = &fd_formats[match];
1431 }
1432 *nb_heads = parse->max_head + 1;
1433 *max_track = parse->max_track;
1434 *last_sect = parse->last_sect;
1435 *drive = parse->drive;
1436 }
1437 }
1438
1439 int bdrv_get_translation_hint(BlockDriverState *bs)
1440 {
1441 return bs->translation;
1442 }
1443
1444 void bdrv_set_on_error(BlockDriverState *bs, BlockErrorAction on_read_error,
1445 BlockErrorAction on_write_error)
1446 {
1447 bs->on_read_error = on_read_error;
1448 bs->on_write_error = on_write_error;
1449 }
1450
1451 BlockErrorAction bdrv_get_on_error(BlockDriverState *bs, int is_read)
1452 {
1453 return is_read ? bs->on_read_error : bs->on_write_error;
1454 }
1455
1456 void bdrv_set_removable(BlockDriverState *bs, int removable)
1457 {
1458 bs->removable = removable;
1459 if (removable && bs == bs_snapshots) {
1460 bs_snapshots = NULL;
1461 }
1462 }
1463
1464 int bdrv_is_removable(BlockDriverState *bs)
1465 {
1466 return bs->removable;
1467 }
1468
1469 int bdrv_is_read_only(BlockDriverState *bs)
1470 {
1471 return bs->read_only;
1472 }
1473
1474 int bdrv_is_sg(BlockDriverState *bs)
1475 {
1476 return bs->sg;
1477 }
1478
1479 int bdrv_enable_write_cache(BlockDriverState *bs)
1480 {
1481 return bs->enable_write_cache;
1482 }
1483
1484 /* XXX: no longer used */
1485 void bdrv_set_change_cb(BlockDriverState *bs,
1486 void (*change_cb)(void *opaque, int reason),
1487 void *opaque)
1488 {
1489 bs->change_cb = change_cb;
1490 bs->change_opaque = opaque;
1491 }
1492
1493 int bdrv_is_encrypted(BlockDriverState *bs)
1494 {
1495 if (bs->backing_hd && bs->backing_hd->encrypted)
1496 return 1;
1497 return bs->encrypted;
1498 }
1499
1500 int bdrv_key_required(BlockDriverState *bs)
1501 {
1502 BlockDriverState *backing_hd = bs->backing_hd;
1503
1504 if (backing_hd && backing_hd->encrypted && !backing_hd->valid_key)
1505 return 1;
1506 return (bs->encrypted && !bs->valid_key);
1507 }
1508
1509 int bdrv_set_key(BlockDriverState *bs, const char *key)
1510 {
1511 int ret;
1512 if (bs->backing_hd && bs->backing_hd->encrypted) {
1513 ret = bdrv_set_key(bs->backing_hd, key);
1514 if (ret < 0)
1515 return ret;
1516 if (!bs->encrypted)
1517 return 0;
1518 }
1519 if (!bs->encrypted) {
1520 return -EINVAL;
1521 } else if (!bs->drv || !bs->drv->bdrv_set_key) {
1522 return -ENOMEDIUM;
1523 }
1524 ret = bs->drv->bdrv_set_key(bs, key);
1525 if (ret < 0) {
1526 bs->valid_key = 0;
1527 } else if (!bs->valid_key) {
1528 bs->valid_key = 1;
1529 /* call the change callback now, we skipped it on open */
1530 bs->media_changed = 1;
1531 if (bs->change_cb)
1532 bs->change_cb(bs->change_opaque, CHANGE_MEDIA);
1533 }
1534 return ret;
1535 }
1536
1537 void bdrv_get_format(BlockDriverState *bs, char *buf, int buf_size)
1538 {
1539 if (!bs->drv) {
1540 buf[0] = '\0';
1541 } else {
1542 pstrcpy(buf, buf_size, bs->drv->format_name);
1543 }
1544 }
1545
1546 void bdrv_iterate_format(void (*it)(void *opaque, const char *name),
1547 void *opaque)
1548 {
1549 BlockDriver *drv;
1550
1551 QLIST_FOREACH(drv, &bdrv_drivers, list) {
1552 it(opaque, drv->format_name);
1553 }
1554 }
1555
1556 BlockDriverState *bdrv_find(const char *name)
1557 {
1558 BlockDriverState *bs;
1559
1560 QTAILQ_FOREACH(bs, &bdrv_states, list) {
1561 if (!strcmp(name, bs->device_name)) {
1562 return bs;
1563 }
1564 }
1565 return NULL;
1566 }
1567
1568 BlockDriverState *bdrv_next(BlockDriverState *bs)
1569 {
1570 if (!bs) {
1571 return QTAILQ_FIRST(&bdrv_states);
1572 }
1573 return QTAILQ_NEXT(bs, list);
1574 }
1575
1576 void bdrv_iterate(void (*it)(void *opaque, BlockDriverState *bs), void *opaque)
1577 {
1578 BlockDriverState *bs;
1579
1580 QTAILQ_FOREACH(bs, &bdrv_states, list) {
1581 it(opaque, bs);
1582 }
1583 }
1584
1585 const char *bdrv_get_device_name(BlockDriverState *bs)
1586 {
1587 return bs->device_name;
1588 }
1589
1590 int bdrv_flush(BlockDriverState *bs)
1591 {
1592 if (bs->open_flags & BDRV_O_NO_FLUSH) {
1593 return 0;
1594 }
1595
1596 if (bs->drv && bs->drv->bdrv_flush) {
1597 return bs->drv->bdrv_flush(bs);
1598 }
1599
1600 /*
1601 * Some block drivers always operate in either writethrough or unsafe mode
1602 * and don't support bdrv_flush therefore. Usually qemu doesn't know how
1603 * the server works (because the behaviour is hardcoded or depends on
1604 * server-side configuration), so we can't ensure that everything is safe
1605 * on disk. Returning an error doesn't work because that would break guests
1606 * even if the server operates in writethrough mode.
1607 *
1608 * Let's hope the user knows what he's doing.
1609 */
1610 return 0;
1611 }
1612
1613 void bdrv_flush_all(void)
1614 {
1615 BlockDriverState *bs;
1616
1617 QTAILQ_FOREACH(bs, &bdrv_states, list) {
1618 if (bs->drv && !bdrv_is_read_only(bs) &&
1619 (!bdrv_is_removable(bs) || bdrv_is_inserted(bs))) {
1620 bdrv_flush(bs);
1621 }
1622 }
1623 }
1624
1625 int bdrv_has_zero_init(BlockDriverState *bs)
1626 {
1627 assert(bs->drv);
1628
1629 if (bs->drv->bdrv_has_zero_init) {
1630 return bs->drv->bdrv_has_zero_init(bs);
1631 }
1632
1633 return 1;
1634 }
1635
1636 int bdrv_discard(BlockDriverState *bs, int64_t sector_num, int nb_sectors)
1637 {
1638 if (!bs->drv) {
1639 return -ENOMEDIUM;
1640 }
1641 if (!bs->drv->bdrv_discard) {
1642 return 0;
1643 }
1644 return bs->drv->bdrv_discard(bs, sector_num, nb_sectors);
1645 }
1646
1647 /*
1648 * Returns true iff the specified sector is present in the disk image. Drivers
1649 * not implementing the functionality are assumed to not support backing files,
1650 * hence all their sectors are reported as allocated.
1651 *
1652 * 'pnum' is set to the number of sectors (including and immediately following
1653 * the specified sector) that are known to be in the same
1654 * allocated/unallocated state.
1655 *
1656 * 'nb_sectors' is the max value 'pnum' should be set to.
1657 */
1658 int bdrv_is_allocated(BlockDriverState *bs, int64_t sector_num, int nb_sectors,
1659 int *pnum)
1660 {
1661 int64_t n;
1662 if (!bs->drv->bdrv_is_allocated) {
1663 if (sector_num >= bs->total_sectors) {
1664 *pnum = 0;
1665 return 0;
1666 }
1667 n = bs->total_sectors - sector_num;
1668 *pnum = (n < nb_sectors) ? (n) : (nb_sectors);
1669 return 1;
1670 }
1671 return bs->drv->bdrv_is_allocated(bs, sector_num, nb_sectors, pnum);
1672 }
1673
1674 void bdrv_mon_event(const BlockDriverState *bdrv,
1675 BlockMonEventAction action, int is_read)
1676 {
1677 QObject *data;
1678 const char *action_str;
1679
1680 switch (action) {
1681 case BDRV_ACTION_REPORT:
1682 action_str = "report";
1683 break;
1684 case BDRV_ACTION_IGNORE:
1685 action_str = "ignore";
1686 break;
1687 case BDRV_ACTION_STOP:
1688 action_str = "stop";
1689 break;
1690 default:
1691 abort();
1692 }
1693
1694 data = qobject_from_jsonf("{ 'device': %s, 'action': %s, 'operation': %s }",
1695 bdrv->device_name,
1696 action_str,
1697 is_read ? "read" : "write");
1698 monitor_protocol_event(QEVENT_BLOCK_IO_ERROR, data);
1699
1700 qobject_decref(data);
1701 }
1702
1703 static void bdrv_print_dict(QObject *obj, void *opaque)
1704 {
1705 QDict *bs_dict;
1706 Monitor *mon = opaque;
1707
1708 bs_dict = qobject_to_qdict(obj);
1709
1710 monitor_printf(mon, "%s: removable=%d",
1711 qdict_get_str(bs_dict, "device"),
1712 qdict_get_bool(bs_dict, "removable"));
1713
1714 if (qdict_get_bool(bs_dict, "removable")) {
1715 monitor_printf(mon, " locked=%d", qdict_get_bool(bs_dict, "locked"));
1716 }
1717
1718 if (qdict_haskey(bs_dict, "inserted")) {
1719 QDict *qdict = qobject_to_qdict(qdict_get(bs_dict, "inserted"));
1720
1721 monitor_printf(mon, " file=");
1722 monitor_print_filename(mon, qdict_get_str(qdict, "file"));
1723 if (qdict_haskey(qdict, "backing_file")) {
1724 monitor_printf(mon, " backing_file=");
1725 monitor_print_filename(mon, qdict_get_str(qdict, "backing_file"));
1726 }
1727 monitor_printf(mon, " ro=%d drv=%s encrypted=%d",
1728 qdict_get_bool(qdict, "ro"),
1729 qdict_get_str(qdict, "drv"),
1730 qdict_get_bool(qdict, "encrypted"));
1731 } else {
1732 monitor_printf(mon, " [not inserted]");
1733 }
1734
1735 monitor_printf(mon, "\n");
1736 }
1737
1738 void bdrv_info_print(Monitor *mon, const QObject *data)
1739 {
1740 qlist_iter(qobject_to_qlist(data), bdrv_print_dict, mon);
1741 }
1742
1743 void bdrv_info(Monitor *mon, QObject **ret_data)
1744 {
1745 QList *bs_list;
1746 BlockDriverState *bs;
1747
1748 bs_list = qlist_new();
1749
1750 QTAILQ_FOREACH(bs, &bdrv_states, list) {
1751 QObject *bs_obj;
1752
1753 bs_obj = qobject_from_jsonf("{ 'device': %s, 'type': 'unknown', "
1754 "'removable': %i, 'locked': %i }",
1755 bs->device_name, bs->removable,
1756 bs->locked);
1757
1758 if (bs->drv) {
1759 QObject *obj;
1760 QDict *bs_dict = qobject_to_qdict(bs_obj);
1761
1762 obj = qobject_from_jsonf("{ 'file': %s, 'ro': %i, 'drv': %s, "
1763 "'encrypted': %i }",
1764 bs->filename, bs->read_only,
1765 bs->drv->format_name,
1766 bdrv_is_encrypted(bs));
1767 if (bs->backing_file[0] != '\0') {
1768 QDict *qdict = qobject_to_qdict(obj);
1769 qdict_put(qdict, "backing_file",
1770 qstring_from_str(bs->backing_file));
1771 }
1772
1773 qdict_put_obj(bs_dict, "inserted", obj);
1774 }
1775 qlist_append_obj(bs_list, bs_obj);
1776 }
1777
1778 *ret_data = QOBJECT(bs_list);
1779 }
1780
1781 static void bdrv_stats_iter(QObject *data, void *opaque)
1782 {
1783 QDict *qdict;
1784 Monitor *mon = opaque;
1785
1786 qdict = qobject_to_qdict(data);
1787 monitor_printf(mon, "%s:", qdict_get_str(qdict, "device"));
1788
1789 qdict = qobject_to_qdict(qdict_get(qdict, "stats"));
1790 monitor_printf(mon, " rd_bytes=%" PRId64
1791 " wr_bytes=%" PRId64
1792 " rd_operations=%" PRId64
1793 " wr_operations=%" PRId64
1794 "\n",
1795 qdict_get_int(qdict, "rd_bytes"),
1796 qdict_get_int(qdict, "wr_bytes"),
1797 qdict_get_int(qdict, "rd_operations"),
1798 qdict_get_int(qdict, "wr_operations"));
1799 }
1800
1801 void bdrv_stats_print(Monitor *mon, const QObject *data)
1802 {
1803 qlist_iter(qobject_to_qlist(data), bdrv_stats_iter, mon);
1804 }
1805
1806 static QObject* bdrv_info_stats_bs(BlockDriverState *bs)
1807 {
1808 QObject *res;
1809 QDict *dict;
1810
1811 res = qobject_from_jsonf("{ 'stats': {"
1812 "'rd_bytes': %" PRId64 ","
1813 "'wr_bytes': %" PRId64 ","
1814 "'rd_operations': %" PRId64 ","
1815 "'wr_operations': %" PRId64 ","
1816 "'wr_highest_offset': %" PRId64
1817 "} }",
1818 bs->rd_bytes, bs->wr_bytes,
1819 bs->rd_ops, bs->wr_ops,
1820 bs->wr_highest_sector *
1821 (uint64_t)BDRV_SECTOR_SIZE);
1822 dict = qobject_to_qdict(res);
1823
1824 if (*bs->device_name) {
1825 qdict_put(dict, "device", qstring_from_str(bs->device_name));
1826 }
1827
1828 if (bs->file) {
1829 QObject *parent = bdrv_info_stats_bs(bs->file);
1830 qdict_put_obj(dict, "parent", parent);
1831 }
1832
1833 return res;
1834 }
1835
1836 void bdrv_info_stats(Monitor *mon, QObject **ret_data)
1837 {
1838 QObject *obj;
1839 QList *devices;
1840 BlockDriverState *bs;
1841
1842 devices = qlist_new();
1843
1844 QTAILQ_FOREACH(bs, &bdrv_states, list) {
1845 obj = bdrv_info_stats_bs(bs);
1846 qlist_append_obj(devices, obj);
1847 }
1848
1849 *ret_data = QOBJECT(devices);
1850 }
1851
1852 const char *bdrv_get_encrypted_filename(BlockDriverState *bs)
1853 {
1854 if (bs->backing_hd && bs->backing_hd->encrypted)
1855 return bs->backing_file;
1856 else if (bs->encrypted)
1857 return bs->filename;
1858 else
1859 return NULL;
1860 }
1861
1862 void bdrv_get_backing_filename(BlockDriverState *bs,
1863 char *filename, int filename_size)
1864 {
1865 if (!bs->backing_file) {
1866 pstrcpy(filename, filename_size, "");
1867 } else {
1868 pstrcpy(filename, filename_size, bs->backing_file);
1869 }
1870 }
1871
1872 int bdrv_write_compressed(BlockDriverState *bs, int64_t sector_num,
1873 const uint8_t *buf, int nb_sectors)
1874 {
1875 BlockDriver *drv = bs->drv;
1876 if (!drv)
1877 return -ENOMEDIUM;
1878 if (!drv->bdrv_write_compressed)
1879 return -ENOTSUP;
1880 if (bdrv_check_request(bs, sector_num, nb_sectors))
1881 return -EIO;
1882
1883 if (bs->dirty_bitmap) {
1884 set_dirty_bitmap(bs, sector_num, nb_sectors, 1);
1885 }
1886
1887 return drv->bdrv_write_compressed(bs, sector_num, buf, nb_sectors);
1888 }
1889
1890 int bdrv_get_info(BlockDriverState *bs, BlockDriverInfo *bdi)
1891 {
1892 BlockDriver *drv = bs->drv;
1893 if (!drv)
1894 return -ENOMEDIUM;
1895 if (!drv->bdrv_get_info)
1896 return -ENOTSUP;
1897 memset(bdi, 0, sizeof(*bdi));
1898 return drv->bdrv_get_info(bs, bdi);
1899 }
1900
1901 int bdrv_save_vmstate(BlockDriverState *bs, const uint8_t *buf,
1902 int64_t pos, int size)
1903 {
1904 BlockDriver *drv = bs->drv;
1905 if (!drv)
1906 return -ENOMEDIUM;
1907 if (drv->bdrv_save_vmstate)
1908 return drv->bdrv_save_vmstate(bs, buf, pos, size);
1909 if (bs->file)
1910 return bdrv_save_vmstate(bs->file, buf, pos, size);
1911 return -ENOTSUP;
1912 }
1913
1914 int bdrv_load_vmstate(BlockDriverState *bs, uint8_t *buf,
1915 int64_t pos, int size)
1916 {
1917 BlockDriver *drv = bs->drv;
1918 if (!drv)
1919 return -ENOMEDIUM;
1920 if (drv->bdrv_load_vmstate)
1921 return drv->bdrv_load_vmstate(bs, buf, pos, size);
1922 if (bs->file)
1923 return bdrv_load_vmstate(bs->file, buf, pos, size);
1924 return -ENOTSUP;
1925 }
1926
1927 void bdrv_debug_event(BlockDriverState *bs, BlkDebugEvent event)
1928 {
1929 BlockDriver *drv = bs->drv;
1930
1931 if (!drv || !drv->bdrv_debug_event) {
1932 return;
1933 }
1934
1935 return drv->bdrv_debug_event(bs, event);
1936
1937 }
1938
1939 /**************************************************************/
1940 /* handling of snapshots */
1941
1942 int bdrv_can_snapshot(BlockDriverState *bs)
1943 {
1944 BlockDriver *drv = bs->drv;
1945 if (!drv || bdrv_is_removable(bs) || bdrv_is_read_only(bs)) {
1946 return 0;
1947 }
1948
1949 if (!drv->bdrv_snapshot_create) {
1950 if (bs->file != NULL) {
1951 return bdrv_can_snapshot(bs->file);
1952 }
1953 return 0;
1954 }
1955
1956 return 1;
1957 }
1958
1959 int bdrv_is_snapshot(BlockDriverState *bs)
1960 {
1961 return !!(bs->open_flags & BDRV_O_SNAPSHOT);
1962 }
1963
1964 BlockDriverState *bdrv_snapshots(void)
1965 {
1966 BlockDriverState *bs;
1967
1968 if (bs_snapshots) {
1969 return bs_snapshots;
1970 }
1971
1972 bs = NULL;
1973 while ((bs = bdrv_next(bs))) {
1974 if (bdrv_can_snapshot(bs)) {
1975 bs_snapshots = bs;
1976 return bs;
1977 }
1978 }
1979 return NULL;
1980 }
1981
1982 int bdrv_snapshot_create(BlockDriverState *bs,
1983 QEMUSnapshotInfo *sn_info)
1984 {
1985 BlockDriver *drv = bs->drv;
1986 if (!drv)
1987 return -ENOMEDIUM;
1988 if (drv->bdrv_snapshot_create)
1989 return drv->bdrv_snapshot_create(bs, sn_info);
1990 if (bs->file)
1991 return bdrv_snapshot_create(bs->file, sn_info);
1992 return -ENOTSUP;
1993 }
1994
1995 int bdrv_snapshot_goto(BlockDriverState *bs,
1996 const char *snapshot_id)
1997 {
1998 BlockDriver *drv = bs->drv;
1999 int ret, open_ret;
2000
2001 if (!drv)
2002 return -ENOMEDIUM;
2003 if (drv->bdrv_snapshot_goto)
2004 return drv->bdrv_snapshot_goto(bs, snapshot_id);
2005
2006 if (bs->file) {
2007 drv->bdrv_close(bs);
2008 ret = bdrv_snapshot_goto(bs->file, snapshot_id);
2009 open_ret = drv->bdrv_open(bs, bs->open_flags);
2010 if (open_ret < 0) {
2011 bdrv_delete(bs->file);
2012 bs->drv = NULL;
2013 return open_ret;
2014 }
2015 return ret;
2016 }
2017
2018 return -ENOTSUP;
2019 }
2020
2021 int bdrv_snapshot_delete(BlockDriverState *bs, const char *snapshot_id)
2022 {
2023 BlockDriver *drv = bs->drv;
2024 if (!drv)
2025 return -ENOMEDIUM;
2026 if (drv->bdrv_snapshot_delete)
2027 return drv->bdrv_snapshot_delete(bs, snapshot_id);
2028 if (bs->file)
2029 return bdrv_snapshot_delete(bs->file, snapshot_id);
2030 return -ENOTSUP;
2031 }
2032
2033 int bdrv_snapshot_list(BlockDriverState *bs,
2034 QEMUSnapshotInfo **psn_info)
2035 {
2036 BlockDriver *drv = bs->drv;
2037 if (!drv)
2038 return -ENOMEDIUM;
2039 if (drv->bdrv_snapshot_list)
2040 return drv->bdrv_snapshot_list(bs, psn_info);
2041 if (bs->file)
2042 return bdrv_snapshot_list(bs->file, psn_info);
2043 return -ENOTSUP;
2044 }
2045
2046 int bdrv_snapshot_load_tmp(BlockDriverState *bs,
2047 const char *snapshot_name)
2048 {
2049 BlockDriver *drv = bs->drv;
2050 if (!drv) {
2051 return -ENOMEDIUM;
2052 }
2053 if (!bs->read_only) {
2054 return -EINVAL;
2055 }
2056 if (drv->bdrv_snapshot_load_tmp) {
2057 return drv->bdrv_snapshot_load_tmp(bs, snapshot_name);
2058 }
2059 return -ENOTSUP;
2060 }
2061
2062 #define NB_SUFFIXES 4
2063
2064 char *get_human_readable_size(char *buf, int buf_size, int64_t size)
2065 {
2066 static const char suffixes[NB_SUFFIXES] = "KMGT";
2067 int64_t base;
2068 int i;
2069
2070 if (size <= 999) {
2071 snprintf(buf, buf_size, "%" PRId64, size);
2072 } else {
2073 base = 1024;
2074 for(i = 0; i < NB_SUFFIXES; i++) {
2075 if (size < (10 * base)) {
2076 snprintf(buf, buf_size, "%0.1f%c",
2077 (double)size / base,
2078 suffixes[i]);
2079 break;
2080 } else if (size < (1000 * base) || i == (NB_SUFFIXES - 1)) {
2081 snprintf(buf, buf_size, "%" PRId64 "%c",
2082 ((size + (base >> 1)) / base),
2083 suffixes[i]);
2084 break;
2085 }
2086 base = base * 1024;
2087 }
2088 }
2089 return buf;
2090 }
2091
2092 char *bdrv_snapshot_dump(char *buf, int buf_size, QEMUSnapshotInfo *sn)
2093 {
2094 char buf1[128], date_buf[128], clock_buf[128];
2095 #ifdef _WIN32
2096 struct tm *ptm;
2097 #else
2098 struct tm tm;
2099 #endif
2100 time_t ti;
2101 int64_t secs;
2102
2103 if (!sn) {
2104 snprintf(buf, buf_size,
2105 "%-10s%-20s%7s%20s%15s",
2106 "ID", "TAG", "VM SIZE", "DATE", "VM CLOCK");
2107 } else {
2108 ti = sn->date_sec;
2109 #ifdef _WIN32
2110 ptm = localtime(&ti);
2111 strftime(date_buf, sizeof(date_buf),
2112 "%Y-%m-%d %H:%M:%S", ptm);
2113 #else
2114 localtime_r(&ti, &tm);
2115 strftime(date_buf, sizeof(date_buf),
2116 "%Y-%m-%d %H:%M:%S", &tm);
2117 #endif
2118 secs = sn->vm_clock_nsec / 1000000000;
2119 snprintf(clock_buf, sizeof(clock_buf),
2120 "%02d:%02d:%02d.%03d",
2121 (int)(secs / 3600),
2122 (int)((secs / 60) % 60),
2123 (int)(secs % 60),
2124 (int)((sn->vm_clock_nsec / 1000000) % 1000));
2125 snprintf(buf, buf_size,
2126 "%-10s%-20s%7s%20s%15s",
2127 sn->id_str, sn->name,
2128 get_human_readable_size(buf1, sizeof(buf1), sn->vm_state_size),
2129 date_buf,
2130 clock_buf);
2131 }
2132 return buf;
2133 }
2134
2135
2136 /**************************************************************/
2137 /* async I/Os */
2138
2139 BlockDriverAIOCB *bdrv_aio_readv(BlockDriverState *bs, int64_t sector_num,
2140 QEMUIOVector *qiov, int nb_sectors,
2141 BlockDriverCompletionFunc *cb, void *opaque)
2142 {
2143 BlockDriver *drv = bs->drv;
2144 BlockDriverAIOCB *ret;
2145
2146 trace_bdrv_aio_readv(bs, sector_num, nb_sectors, opaque);
2147
2148 if (!drv)
2149 return NULL;
2150 if (bdrv_check_request(bs, sector_num, nb_sectors))
2151 return NULL;
2152
2153 ret = drv->bdrv_aio_readv(bs, sector_num, qiov, nb_sectors,
2154 cb, opaque);
2155
2156 if (ret) {
2157 /* Update stats even though technically transfer has not happened. */
2158 bs->rd_bytes += (unsigned) nb_sectors * BDRV_SECTOR_SIZE;
2159 bs->rd_ops ++;
2160 }
2161
2162 return ret;
2163 }
2164
2165 typedef struct BlockCompleteData {
2166 BlockDriverCompletionFunc *cb;
2167 void *opaque;
2168 BlockDriverState *bs;
2169 int64_t sector_num;
2170 int nb_sectors;
2171 } BlockCompleteData;
2172
2173 static void block_complete_cb(void *opaque, int ret)
2174 {
2175 BlockCompleteData *b = opaque;
2176
2177 if (b->bs->dirty_bitmap) {
2178 set_dirty_bitmap(b->bs, b->sector_num, b->nb_sectors, 1);
2179 }
2180 b->cb(b->opaque, ret);
2181 qemu_free(b);
2182 }
2183
2184 static BlockCompleteData *blk_dirty_cb_alloc(BlockDriverState *bs,
2185 int64_t sector_num,
2186 int nb_sectors,
2187 BlockDriverCompletionFunc *cb,
2188 void *opaque)
2189 {
2190 BlockCompleteData *blkdata = qemu_mallocz(sizeof(BlockCompleteData));
2191
2192 blkdata->bs = bs;
2193 blkdata->cb = cb;
2194 blkdata->opaque = opaque;
2195 blkdata->sector_num = sector_num;
2196 blkdata->nb_sectors = nb_sectors;
2197
2198 return blkdata;
2199 }
2200
2201 BlockDriverAIOCB *bdrv_aio_writev(BlockDriverState *bs, int64_t sector_num,
2202 QEMUIOVector *qiov, int nb_sectors,
2203 BlockDriverCompletionFunc *cb, void *opaque)
2204 {
2205 BlockDriver *drv = bs->drv;
2206 BlockDriverAIOCB *ret;
2207 BlockCompleteData *blk_cb_data;
2208
2209 trace_bdrv_aio_writev(bs, sector_num, nb_sectors, opaque);
2210
2211 if (!drv)
2212 return NULL;
2213 if (bs->read_only)
2214 return NULL;
2215 if (bdrv_check_request(bs, sector_num, nb_sectors))
2216 return NULL;
2217
2218 if (bs->dirty_bitmap) {
2219 blk_cb_data = blk_dirty_cb_alloc(bs, sector_num, nb_sectors, cb,
2220 opaque);
2221 cb = &block_complete_cb;
2222 opaque = blk_cb_data;
2223 }
2224
2225 ret = drv->bdrv_aio_writev(bs, sector_num, qiov, nb_sectors,
2226 cb, opaque);
2227
2228 if (ret) {
2229 /* Update stats even though technically transfer has not happened. */
2230 bs->wr_bytes += (unsigned) nb_sectors * BDRV_SECTOR_SIZE;
2231 bs->wr_ops ++;
2232 if (bs->wr_highest_sector < sector_num + nb_sectors - 1) {
2233 bs->wr_highest_sector = sector_num + nb_sectors - 1;
2234 }
2235 }
2236
2237 return ret;
2238 }
2239
2240
2241 typedef struct MultiwriteCB {
2242 int error;
2243 int num_requests;
2244 int num_callbacks;
2245 struct {
2246 BlockDriverCompletionFunc *cb;
2247 void *opaque;
2248 QEMUIOVector *free_qiov;
2249 void *free_buf;
2250 } callbacks[];
2251 } MultiwriteCB;
2252
2253 static void multiwrite_user_cb(MultiwriteCB *mcb)
2254 {
2255 int i;
2256
2257 for (i = 0; i < mcb->num_callbacks; i++) {
2258 mcb->callbacks[i].cb(mcb->callbacks[i].opaque, mcb->error);
2259 if (mcb->callbacks[i].free_qiov) {
2260 qemu_iovec_destroy(mcb->callbacks[i].free_qiov);
2261 }
2262 qemu_free(mcb->callbacks[i].free_qiov);
2263 qemu_vfree(mcb->callbacks[i].free_buf);
2264 }
2265 }
2266
2267 static void multiwrite_cb(void *opaque, int ret)
2268 {
2269 MultiwriteCB *mcb = opaque;
2270
2271 trace_multiwrite_cb(mcb, ret);
2272
2273 if (ret < 0 && !mcb->error) {
2274 mcb->error = ret;
2275 }
2276
2277 mcb->num_requests--;
2278 if (mcb->num_requests == 0) {
2279 multiwrite_user_cb(mcb);
2280 qemu_free(mcb);
2281 }
2282 }
2283
2284 static int multiwrite_req_compare(const void *a, const void *b)
2285 {
2286 const BlockRequest *req1 = a, *req2 = b;
2287
2288 /*
2289 * Note that we can't simply subtract req2->sector from req1->sector
2290 * here as that could overflow the return value.
2291 */
2292 if (req1->sector > req2->sector) {
2293 return 1;
2294 } else if (req1->sector < req2->sector) {
2295 return -1;
2296 } else {
2297 return 0;
2298 }
2299 }
2300
2301 /*
2302 * Takes a bunch of requests and tries to merge them. Returns the number of
2303 * requests that remain after merging.
2304 */
2305 static int multiwrite_merge(BlockDriverState *bs, BlockRequest *reqs,
2306 int num_reqs, MultiwriteCB *mcb)
2307 {
2308 int i, outidx;
2309
2310 // Sort requests by start sector
2311 qsort(reqs, num_reqs, sizeof(*reqs), &multiwrite_req_compare);
2312
2313 // Check if adjacent requests touch the same clusters. If so, combine them,
2314 // filling up gaps with zero sectors.
2315 outidx = 0;
2316 for (i = 1; i < num_reqs; i++) {
2317 int merge = 0;
2318 int64_t oldreq_last = reqs[outidx].sector + reqs[outidx].nb_sectors;
2319
2320 // This handles the cases that are valid for all block drivers, namely
2321 // exactly sequential writes and overlapping writes.
2322 if (reqs[i].sector <= oldreq_last) {
2323 merge = 1;
2324 }
2325
2326 // The block driver may decide that it makes sense to combine requests
2327 // even if there is a gap of some sectors between them. In this case,
2328 // the gap is filled with zeros (therefore only applicable for yet
2329 // unused space in format like qcow2).
2330 if (!merge && bs->drv->bdrv_merge_requests) {
2331 merge = bs->drv->bdrv_merge_requests(bs, &reqs[outidx], &reqs[i]);
2332 }
2333
2334 if (reqs[outidx].qiov->niov + reqs[i].qiov->niov + 1 > IOV_MAX) {
2335 merge = 0;
2336 }
2337
2338 if (merge) {
2339 size_t size;
2340 QEMUIOVector *qiov = qemu_mallocz(sizeof(*qiov));
2341 qemu_iovec_init(qiov,
2342 reqs[outidx].qiov->niov + reqs[i].qiov->niov + 1);
2343
2344 // Add the first request to the merged one. If the requests are
2345 // overlapping, drop the last sectors of the first request.
2346 size = (reqs[i].sector - reqs[outidx].sector) << 9;
2347 qemu_iovec_concat(qiov, reqs[outidx].qiov, size);
2348
2349 // We might need to add some zeros between the two requests
2350 if (reqs[i].sector > oldreq_last) {
2351 size_t zero_bytes = (reqs[i].sector - oldreq_last) << 9;
2352 uint8_t *buf = qemu_blockalign(bs, zero_bytes);
2353 memset(buf, 0, zero_bytes);
2354 qemu_iovec_add(qiov, buf, zero_bytes);
2355 mcb->callbacks[i].free_buf = buf;
2356 }
2357
2358 // Add the second request
2359 qemu_iovec_concat(qiov, reqs[i].qiov, reqs[i].qiov->size);
2360
2361 reqs[outidx].nb_sectors = qiov->size >> 9;
2362 reqs[outidx].qiov = qiov;
2363
2364 mcb->callbacks[i].free_qiov = reqs[outidx].qiov;
2365 } else {
2366 outidx++;
2367 reqs[outidx].sector = reqs[i].sector;
2368 reqs[outidx].nb_sectors = reqs[i].nb_sectors;
2369 reqs[outidx].qiov = reqs[i].qiov;
2370 }
2371 }
2372
2373 return outidx + 1;
2374 }
2375
2376 /*
2377 * Submit multiple AIO write requests at once.
2378 *
2379 * On success, the function returns 0 and all requests in the reqs array have
2380 * been submitted. In error case this function returns -1, and any of the
2381 * requests may or may not be submitted yet. In particular, this means that the
2382 * callback will be called for some of the requests, for others it won't. The
2383 * caller must check the error field of the BlockRequest to wait for the right
2384 * callbacks (if error != 0, no callback will be called).
2385 *
2386 * The implementation may modify the contents of the reqs array, e.g. to merge
2387 * requests. However, the fields opaque and error are left unmodified as they
2388 * are used to signal failure for a single request to the caller.
2389 */
2390 int bdrv_aio_multiwrite(BlockDriverState *bs, BlockRequest *reqs, int num_reqs)
2391 {
2392 BlockDriverAIOCB *acb;
2393 MultiwriteCB *mcb;
2394 int i;
2395
2396 /* don't submit writes if we don't have a medium */
2397 if (bs->drv == NULL) {
2398 for (i = 0; i < num_reqs; i++) {
2399 reqs[i].error = -ENOMEDIUM;
2400 }
2401 return -1;
2402 }
2403
2404 if (num_reqs == 0) {
2405 return 0;
2406 }
2407
2408 // Create MultiwriteCB structure
2409 mcb = qemu_mallocz(sizeof(*mcb) + num_reqs * sizeof(*mcb->callbacks));
2410 mcb->num_requests = 0;
2411 mcb->num_callbacks = num_reqs;
2412
2413 for (i = 0; i < num_reqs; i++) {
2414 mcb->callbacks[i].cb = reqs[i].cb;
2415 mcb->callbacks[i].opaque = reqs[i].opaque;
2416 }
2417
2418 // Check for mergable requests
2419 num_reqs = multiwrite_merge(bs, reqs, num_reqs, mcb);
2420
2421 trace_bdrv_aio_multiwrite(mcb, mcb->num_callbacks, num_reqs);
2422
2423 /*
2424 * Run the aio requests. As soon as one request can't be submitted
2425 * successfully, fail all requests that are not yet submitted (we must
2426 * return failure for all requests anyway)
2427 *
2428 * num_requests cannot be set to the right value immediately: If
2429 * bdrv_aio_writev fails for some request, num_requests would be too high
2430 * and therefore multiwrite_cb() would never recognize the multiwrite
2431 * request as completed. We also cannot use the loop variable i to set it
2432 * when the first request fails because the callback may already have been
2433 * called for previously submitted requests. Thus, num_requests must be
2434 * incremented for each request that is submitted.
2435 *
2436 * The problem that callbacks may be called early also means that we need
2437 * to take care that num_requests doesn't become 0 before all requests are
2438 * submitted - multiwrite_cb() would consider the multiwrite request
2439 * completed. A dummy request that is "completed" by a manual call to
2440 * multiwrite_cb() takes care of this.
2441 */
2442 mcb->num_requests = 1;
2443
2444 // Run the aio requests
2445 for (i = 0; i < num_reqs; i++) {
2446 mcb->num_requests++;
2447 acb = bdrv_aio_writev(bs, reqs[i].sector, reqs[i].qiov,
2448 reqs[i].nb_sectors, multiwrite_cb, mcb);
2449
2450 if (acb == NULL) {
2451 // We can only fail the whole thing if no request has been
2452 // submitted yet. Otherwise we'll wait for the submitted AIOs to
2453 // complete and report the error in the callback.
2454 if (i == 0) {
2455 trace_bdrv_aio_multiwrite_earlyfail(mcb);
2456 goto fail;
2457 } else {
2458 trace_bdrv_aio_multiwrite_latefail(mcb, i);
2459 multiwrite_cb(mcb, -EIO);
2460 break;
2461 }
2462 }
2463 }
2464
2465 /* Complete the dummy request */
2466 multiwrite_cb(mcb, 0);
2467
2468 return 0;
2469
2470 fail:
2471 for (i = 0; i < mcb->num_callbacks; i++) {
2472 reqs[i].error = -EIO;
2473 }
2474 qemu_free(mcb);
2475 return -1;
2476 }
2477
2478 BlockDriverAIOCB *bdrv_aio_flush(BlockDriverState *bs,
2479 BlockDriverCompletionFunc *cb, void *opaque)
2480 {
2481 BlockDriver *drv = bs->drv;
2482
2483 trace_bdrv_aio_flush(bs, opaque);
2484
2485 if (bs->open_flags & BDRV_O_NO_FLUSH) {
2486 return bdrv_aio_noop_em(bs, cb, opaque);
2487 }
2488
2489 if (!drv)
2490 return NULL;
2491 return drv->bdrv_aio_flush(bs, cb, opaque);
2492 }
2493
2494 void bdrv_aio_cancel(BlockDriverAIOCB *acb)
2495 {
2496 acb->pool->cancel(acb);
2497 }
2498
2499
2500 /**************************************************************/
2501 /* async block device emulation */
2502
2503 typedef struct BlockDriverAIOCBSync {
2504 BlockDriverAIOCB common;
2505 QEMUBH *bh;
2506 int ret;
2507 /* vector translation state */
2508 QEMUIOVector *qiov;
2509 uint8_t *bounce;
2510 int is_write;
2511 } BlockDriverAIOCBSync;
2512
2513 static void bdrv_aio_cancel_em(BlockDriverAIOCB *blockacb)
2514 {
2515 BlockDriverAIOCBSync *acb =
2516 container_of(blockacb, BlockDriverAIOCBSync, common);
2517 qemu_bh_delete(acb->bh);
2518 acb->bh = NULL;
2519 qemu_aio_release(acb);
2520 }
2521
2522 static AIOPool bdrv_em_aio_pool = {
2523 .aiocb_size = sizeof(BlockDriverAIOCBSync),
2524 .cancel = bdrv_aio_cancel_em,
2525 };
2526
2527 static void bdrv_aio_bh_cb(void *opaque)
2528 {
2529 BlockDriverAIOCBSync *acb = opaque;
2530
2531 if (!acb->is_write)
2532 qemu_iovec_from_buffer(acb->qiov, acb->bounce, acb->qiov->size);
2533 qemu_vfree(acb->bounce);
2534 acb->common.cb(acb->common.opaque, acb->ret);
2535 qemu_bh_delete(acb->bh);
2536 acb->bh = NULL;
2537 qemu_aio_release(acb);
2538 }
2539
2540 static BlockDriverAIOCB *bdrv_aio_rw_vector(BlockDriverState *bs,
2541 int64_t sector_num,
2542 QEMUIOVector *qiov,
2543 int nb_sectors,
2544 BlockDriverCompletionFunc *cb,
2545 void *opaque,
2546 int is_write)
2547
2548 {
2549 BlockDriverAIOCBSync *acb;
2550
2551 acb = qemu_aio_get(&bdrv_em_aio_pool, bs, cb, opaque);
2552 acb->is_write = is_write;
2553 acb->qiov = qiov;
2554 acb->bounce = qemu_blockalign(bs, qiov->size);
2555
2556 if (!acb->bh)
2557 acb->bh = qemu_bh_new(bdrv_aio_bh_cb, acb);
2558
2559 if (is_write) {
2560 qemu_iovec_to_buffer(acb->qiov, acb->bounce);
2561 acb->ret = bdrv_write(bs, sector_num, acb->bounce, nb_sectors);
2562 } else {
2563 acb->ret = bdrv_read(bs, sector_num, acb->bounce, nb_sectors);
2564 }
2565
2566 qemu_bh_schedule(acb->bh);
2567
2568 return &acb->common;
2569 }
2570
2571 static BlockDriverAIOCB *bdrv_aio_readv_em(BlockDriverState *bs,
2572 int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
2573 BlockDriverCompletionFunc *cb, void *opaque)
2574 {
2575 return bdrv_aio_rw_vector(bs, sector_num, qiov, nb_sectors, cb, opaque, 0);
2576 }
2577
2578 static BlockDriverAIOCB *bdrv_aio_writev_em(BlockDriverState *bs,
2579 int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
2580 BlockDriverCompletionFunc *cb, void *opaque)
2581 {
2582 return bdrv_aio_rw_vector(bs, sector_num, qiov, nb_sectors, cb, opaque, 1);
2583 }
2584
2585 static BlockDriverAIOCB *bdrv_aio_flush_em(BlockDriverState *bs,
2586 BlockDriverCompletionFunc *cb, void *opaque)
2587 {
2588 BlockDriverAIOCBSync *acb;
2589
2590 acb = qemu_aio_get(&bdrv_em_aio_pool, bs, cb, opaque);
2591 acb->is_write = 1; /* don't bounce in the completion hadler */
2592 acb->qiov = NULL;
2593 acb->bounce = NULL;
2594 acb->ret = 0;
2595
2596 if (!acb->bh)
2597 acb->bh = qemu_bh_new(bdrv_aio_bh_cb, acb);
2598
2599 bdrv_flush(bs);
2600 qemu_bh_schedule(acb->bh);
2601 return &acb->common;
2602 }
2603
2604 static BlockDriverAIOCB *bdrv_aio_noop_em(BlockDriverState *bs,
2605 BlockDriverCompletionFunc *cb, void *opaque)
2606 {
2607 BlockDriverAIOCBSync *acb;
2608
2609 acb = qemu_aio_get(&bdrv_em_aio_pool, bs, cb, opaque);
2610 acb->is_write = 1; /* don't bounce in the completion handler */
2611 acb->qiov = NULL;
2612 acb->bounce = NULL;
2613 acb->ret = 0;
2614
2615 if (!acb->bh) {
2616 acb->bh = qemu_bh_new(bdrv_aio_bh_cb, acb);
2617 }
2618
2619 qemu_bh_schedule(acb->bh);
2620 return &acb->common;
2621 }
2622
2623 /**************************************************************/
2624 /* sync block device emulation */
2625
2626 static void bdrv_rw_em_cb(void *opaque, int ret)
2627 {
2628 *(int *)opaque = ret;
2629 }
2630
2631 #define NOT_DONE 0x7fffffff
2632
2633 static int bdrv_read_em(BlockDriverState *bs, int64_t sector_num,
2634 uint8_t *buf, int nb_sectors)
2635 {
2636 int async_ret;
2637 BlockDriverAIOCB *acb;
2638 struct iovec iov;
2639 QEMUIOVector qiov;
2640
2641 async_context_push();
2642
2643 async_ret = NOT_DONE;
2644 iov.iov_base = (void *)buf;
2645 iov.iov_len = nb_sectors * BDRV_SECTOR_SIZE;
2646 qemu_iovec_init_external(&qiov, &iov, 1);
2647 acb = bdrv_aio_readv(bs, sector_num, &qiov, nb_sectors,
2648 bdrv_rw_em_cb, &async_ret);
2649 if (acb == NULL) {
2650 async_ret = -1;
2651 goto fail;
2652 }
2653
2654 while (async_ret == NOT_DONE) {
2655 qemu_aio_wait();
2656 }
2657
2658
2659 fail:
2660 async_context_pop();
2661 return async_ret;
2662 }
2663
2664 static int bdrv_write_em(BlockDriverState *bs, int64_t sector_num,
2665 const uint8_t *buf, int nb_sectors)
2666 {
2667 int async_ret;
2668 BlockDriverAIOCB *acb;
2669 struct iovec iov;
2670 QEMUIOVector qiov;
2671
2672 async_context_push();
2673
2674 async_ret = NOT_DONE;
2675 iov.iov_base = (void *)buf;
2676 iov.iov_len = nb_sectors * BDRV_SECTOR_SIZE;
2677 qemu_iovec_init_external(&qiov, &iov, 1);
2678 acb = bdrv_aio_writev(bs, sector_num, &qiov, nb_sectors,
2679 bdrv_rw_em_cb, &async_ret);
2680 if (acb == NULL) {
2681 async_ret = -1;
2682 goto fail;
2683 }
2684 while (async_ret == NOT_DONE) {
2685 qemu_aio_wait();
2686 }
2687
2688 fail:
2689 async_context_pop();
2690 return async_ret;
2691 }
2692
2693 void bdrv_init(void)
2694 {
2695 module_call_init(MODULE_INIT_BLOCK);
2696 }
2697
2698 void bdrv_init_with_whitelist(void)
2699 {
2700 use_bdrv_whitelist = 1;
2701 bdrv_init();
2702 }
2703
2704 void *qemu_aio_get(AIOPool *pool, BlockDriverState *bs,
2705 BlockDriverCompletionFunc *cb, void *opaque)
2706 {
2707 BlockDriverAIOCB *acb;
2708
2709 if (pool->free_aiocb) {
2710 acb = pool->free_aiocb;
2711 pool->free_aiocb = acb->next;
2712 } else {
2713 acb = qemu_mallocz(pool->aiocb_size);
2714 acb->pool = pool;
2715 }
2716 acb->bs = bs;
2717 acb->cb = cb;
2718 acb->opaque = opaque;
2719 return acb;
2720 }
2721
2722 void qemu_aio_release(void *p)
2723 {
2724 BlockDriverAIOCB *acb = (BlockDriverAIOCB *)p;
2725 AIOPool *pool = acb->pool;
2726 acb->next = pool->free_aiocb;
2727 pool->free_aiocb = acb;
2728 }
2729
2730 /**************************************************************/
2731 /* removable device support */
2732
2733 /**
2734 * Return TRUE if the media is present
2735 */
2736 int bdrv_is_inserted(BlockDriverState *bs)
2737 {
2738 BlockDriver *drv = bs->drv;
2739 int ret;
2740 if (!drv)
2741 return 0;
2742 if (!drv->bdrv_is_inserted)
2743 return !bs->tray_open;
2744 ret = drv->bdrv_is_inserted(bs);
2745 return ret;
2746 }
2747
2748 /**
2749 * Return TRUE if the media changed since the last call to this
2750 * function. It is currently only used for floppy disks
2751 */
2752 int bdrv_media_changed(BlockDriverState *bs)
2753 {
2754 BlockDriver *drv = bs->drv;
2755 int ret;
2756
2757 if (!drv || !drv->bdrv_media_changed)
2758 ret = -ENOTSUP;
2759 else
2760 ret = drv->bdrv_media_changed(bs);
2761 if (ret == -ENOTSUP)
2762 ret = bs->media_changed;
2763 bs->media_changed = 0;
2764 return ret;
2765 }
2766
2767 /**
2768 * If eject_flag is TRUE, eject the media. Otherwise, close the tray
2769 */
2770 int bdrv_eject(BlockDriverState *bs, int eject_flag)
2771 {
2772 BlockDriver *drv = bs->drv;
2773
2774 if (eject_flag && bs->locked) {
2775 return -EBUSY;
2776 }
2777
2778 if (drv && drv->bdrv_eject) {
2779 drv->bdrv_eject(bs, eject_flag);
2780 }
2781 bs->tray_open = eject_flag;
2782 return 0;
2783 }
2784
2785 int bdrv_is_locked(BlockDriverState *bs)
2786 {
2787 return bs->locked;
2788 }
2789
2790 /**
2791 * Lock or unlock the media (if it is locked, the user won't be able
2792 * to eject it manually).
2793 */
2794 void bdrv_set_locked(BlockDriverState *bs, int locked)
2795 {
2796 BlockDriver *drv = bs->drv;
2797
2798 trace_bdrv_set_locked(bs, locked);
2799
2800 bs->locked = locked;
2801 if (drv && drv->bdrv_set_locked) {
2802 drv->bdrv_set_locked(bs, locked);
2803 }
2804 }
2805
2806 /* needed for generic scsi interface */
2807
2808 int bdrv_ioctl(BlockDriverState *bs, unsigned long int req, void *buf)
2809 {
2810 BlockDriver *drv = bs->drv;
2811
2812 if (drv && drv->bdrv_ioctl)
2813 return drv->bdrv_ioctl(bs, req, buf);
2814 return -ENOTSUP;
2815 }
2816
2817 BlockDriverAIOCB *bdrv_aio_ioctl(BlockDriverState *bs,
2818 unsigned long int req, void *buf,
2819 BlockDriverCompletionFunc *cb, void *opaque)
2820 {
2821 BlockDriver *drv = bs->drv;
2822
2823 if (drv && drv->bdrv_aio_ioctl)
2824 return drv->bdrv_aio_ioctl(bs, req, buf, cb, opaque);
2825 return NULL;
2826 }
2827
2828
2829
2830 void *qemu_blockalign(BlockDriverState *bs, size_t size)
2831 {
2832 return qemu_memalign((bs && bs->buffer_alignment) ? bs->buffer_alignment : 512, size);
2833 }
2834
2835 void bdrv_set_dirty_tracking(BlockDriverState *bs, int enable)
2836 {
2837 int64_t bitmap_size;
2838
2839 bs->dirty_count = 0;
2840 if (enable) {
2841 if (!bs->dirty_bitmap) {
2842 bitmap_size = (bdrv_getlength(bs) >> BDRV_SECTOR_BITS) +
2843 BDRV_SECTORS_PER_DIRTY_CHUNK * 8 - 1;
2844 bitmap_size /= BDRV_SECTORS_PER_DIRTY_CHUNK * 8;
2845
2846 bs->dirty_bitmap = qemu_mallocz(bitmap_size);
2847 }
2848 } else {
2849 if (bs->dirty_bitmap) {
2850 qemu_free(bs->dirty_bitmap);
2851 bs->dirty_bitmap = NULL;
2852 }
2853 }
2854 }
2855
2856 int bdrv_get_dirty(BlockDriverState *bs, int64_t sector)
2857 {
2858 int64_t chunk = sector / (int64_t)BDRV_SECTORS_PER_DIRTY_CHUNK;
2859
2860 if (bs->dirty_bitmap &&
2861 (sector << BDRV_SECTOR_BITS) < bdrv_getlength(bs)) {
2862 return !!(bs->dirty_bitmap[chunk / (sizeof(unsigned long) * 8)] &
2863 (1UL << (chunk % (sizeof(unsigned long) * 8))));
2864 } else {
2865 return 0;
2866 }
2867 }
2868
2869 void bdrv_reset_dirty(BlockDriverState *bs, int64_t cur_sector,
2870 int nr_sectors)
2871 {
2872 set_dirty_bitmap(bs, cur_sector, nr_sectors, 0);
2873 }
2874
2875 int64_t bdrv_get_dirty_count(BlockDriverState *bs)
2876 {
2877 return bs->dirty_count;
2878 }
2879
2880 void bdrv_set_in_use(BlockDriverState *bs, int in_use)
2881 {
2882 assert(bs->in_use != in_use);
2883 bs->in_use = in_use;
2884 }
2885
2886 int bdrv_in_use(BlockDriverState *bs)
2887 {
2888 return bs->in_use;
2889 }
2890
2891 int bdrv_img_create(const char *filename, const char *fmt,
2892 const char *base_filename, const char *base_fmt,
2893 char *options, uint64_t img_size, int flags)
2894 {
2895 QEMUOptionParameter *param = NULL, *create_options = NULL;
2896 QEMUOptionParameter *backing_fmt, *backing_file, *size;
2897 BlockDriverState *bs = NULL;
2898 BlockDriver *drv, *proto_drv;
2899 BlockDriver *backing_drv = NULL;
2900 int ret = 0;
2901
2902 /* Find driver and parse its options */
2903 drv = bdrv_find_format(fmt);
2904 if (!drv) {
2905 error_report("Unknown file format '%s'", fmt);
2906 ret = -EINVAL;
2907 goto out;
2908 }
2909
2910 proto_drv = bdrv_find_protocol(filename);
2911 if (!proto_drv) {
2912 error_report("Unknown protocol '%s'", filename);
2913 ret = -EINVAL;
2914 goto out;
2915 }
2916
2917 create_options = append_option_parameters(create_options,
2918 drv->create_options);
2919 create_options = append_option_parameters(create_options,
2920 proto_drv->create_options);
2921
2922 /* Create parameter list with default values */
2923 param = parse_option_parameters("", create_options, param);
2924
2925 set_option_parameter_int(param, BLOCK_OPT_SIZE, img_size);
2926
2927 /* Parse -o options */
2928 if (options) {
2929 param = parse_option_parameters(options, create_options, param);
2930 if (param == NULL) {
2931 error_report("Invalid options for file format '%s'.", fmt);
2932 ret = -EINVAL;
2933 goto out;
2934 }
2935 }
2936
2937 if (base_filename) {
2938 if (set_option_parameter(param, BLOCK_OPT_BACKING_FILE,
2939 base_filename)) {
2940 error_report("Backing file not supported for file format '%s'",
2941 fmt);
2942 ret = -EINVAL;
2943 goto out;
2944 }
2945 }
2946
2947 if (base_fmt) {
2948 if (set_option_parameter(param, BLOCK_OPT_BACKING_FMT, base_fmt)) {
2949 error_report("Backing file format not supported for file "
2950 "format '%s'", fmt);
2951 ret = -EINVAL;
2952 goto out;
2953 }
2954 }
2955
2956 backing_file = get_option_parameter(param, BLOCK_OPT_BACKING_FILE);
2957 if (backing_file && backing_file->value.s) {
2958 if (!strcmp(filename, backing_file->value.s)) {
2959 error_report("Error: Trying to create an image with the "
2960 "same filename as the backing file");
2961 ret = -EINVAL;
2962 goto out;
2963 }
2964 }
2965
2966 backing_fmt = get_option_parameter(param, BLOCK_OPT_BACKING_FMT);
2967 if (backing_fmt && backing_fmt->value.s) {
2968 backing_drv = bdrv_find_format(backing_fmt->value.s);
2969 if (!backing_drv) {
2970 error_report("Unknown backing file format '%s'",
2971 backing_fmt->value.s);
2972 ret = -EINVAL;
2973 goto out;
2974 }
2975 }
2976
2977 // The size for the image must always be specified, with one exception:
2978 // If we are using a backing file, we can obtain the size from there
2979 size = get_option_parameter(param, BLOCK_OPT_SIZE);
2980 if (size && size->value.n == -1) {
2981 if (backing_file && backing_file->value.s) {
2982 uint64_t size;
2983 char buf[32];
2984
2985 bs = bdrv_new("");
2986
2987 ret = bdrv_open(bs, backing_file->value.s, flags, backing_drv);
2988 if (ret < 0) {
2989 error_report("Could not open '%s'", backing_file->value.s);
2990 goto out;
2991 }
2992 bdrv_get_geometry(bs, &size);
2993 size *= 512;
2994
2995 snprintf(buf, sizeof(buf), "%" PRId64, size);
2996 set_option_parameter(param, BLOCK_OPT_SIZE, buf);
2997 } else {
2998 error_report("Image creation needs a size parameter");
2999 ret = -EINVAL;
3000 goto out;
3001 }
3002 }
3003
3004 printf("Formatting '%s', fmt=%s ", filename, fmt);
3005 print_option_parameters(param);
3006 puts("");
3007
3008 ret = bdrv_create(drv, filename, param);
3009
3010 if (ret < 0) {
3011 if (ret == -ENOTSUP) {
3012 error_report("Formatting or formatting option not supported for "
3013 "file format '%s'", fmt);
3014 } else if (ret == -EFBIG) {
3015 error_report("The image size is too large for file format '%s'",
3016 fmt);
3017 } else {
3018 error_report("%s: error while creating %s: %s", filename, fmt,
3019 strerror(-ret));
3020 }
3021 }
3022
3023 out:
3024 free_option_parameters(create_options);
3025 free_option_parameters(param);
3026
3027 if (bs) {
3028 bdrv_delete(bs);
3029 }
3030
3031 return ret;
3032 }
Random patches