]> git.proxmox.com Git - mirror_qemu.git/blob - block.c
projects / mirror_qemu.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
savevm: Really verify if a drive supports snapshots
[mirror_qemu.git] / block.c
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 "monitor.h"
27 #include "block_int.h"
28 #include "module.h"
29 #include "qemu-objects.h"
30
31 #ifdef CONFIG_BSD
32 #include <sys/types.h>
33 #include <sys/stat.h>
34 #include <sys/ioctl.h>
35 #include <sys/queue.h>
36 #ifndef __DragonFly__
37 #include <sys/disk.h>
38 #endif
39 #endif
40
41 #ifdef _WIN32
42 #include <windows.h>
43 #endif
44
45 static BlockDriverAIOCB *bdrv_aio_readv_em(BlockDriverState *bs,
46 int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
47 BlockDriverCompletionFunc *cb, void *opaque);
48 static BlockDriverAIOCB *bdrv_aio_writev_em(BlockDriverState *bs,
49 int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
50 BlockDriverCompletionFunc *cb, void *opaque);
51 static BlockDriverAIOCB *bdrv_aio_flush_em(BlockDriverState *bs,
52 BlockDriverCompletionFunc *cb, void *opaque);
53 static BlockDriverAIOCB *bdrv_aio_noop_em(BlockDriverState *bs,
54 BlockDriverCompletionFunc *cb, void *opaque);
55 static int bdrv_read_em(BlockDriverState *bs, int64_t sector_num,
56 uint8_t *buf, int nb_sectors);
57 static int bdrv_write_em(BlockDriverState *bs, int64_t sector_num,
58 const uint8_t *buf, int nb_sectors);
59
60 static QTAILQ_HEAD(, BlockDriverState) bdrv_states =
61 QTAILQ_HEAD_INITIALIZER(bdrv_states);
62
63 static QLIST_HEAD(, BlockDriver) bdrv_drivers =
64 QLIST_HEAD_INITIALIZER(bdrv_drivers);
65
66 /* If non-zero, use only whitelisted block drivers */
67 static int use_bdrv_whitelist;
68
69 int path_is_absolute(const char *path)
70 {
71 const char *p;
72 #ifdef _WIN32
73 /* specific case for names like: "\\.\d:" */
74 if (*path == '/' || *path == '\\')
75 return 1;
76 #endif
77 p = strchr(path, ':');
78 if (p)
79 p++;
80 else
81 p = path;
82 #ifdef _WIN32
83 return (*p == '/' || *p == '\\');
84 #else
85 return (*p == '/');
86 #endif
87 }
88
89 /* if filename is absolute, just copy it to dest. Otherwise, build a
90 path to it by considering it is relative to base_path. URL are
91 supported. */
92 void path_combine(char *dest, int dest_size,
93 const char *base_path,
94 const char *filename)
95 {
96 const char *p, *p1;
97 int len;
98
99 if (dest_size <= 0)
100 return;
101 if (path_is_absolute(filename)) {
102 pstrcpy(dest, dest_size, filename);
103 } else {
104 p = strchr(base_path, ':');
105 if (p)
106 p++;
107 else
108 p = base_path;
109 p1 = strrchr(base_path, '/');
110 #ifdef _WIN32
111 {
112 const char *p2;
113 p2 = strrchr(base_path, '\\');
114 if (!p1 || p2 > p1)
115 p1 = p2;
116 }
117 #endif
118 if (p1)
119 p1++;
120 else
121 p1 = base_path;
122 if (p1 > p)
123 p = p1;
124 len = p - base_path;
125 if (len > dest_size - 1)
126 len = dest_size - 1;
127 memcpy(dest, base_path, len);
128 dest[len] = '\0';
129 pstrcat(dest, dest_size, filename);
130 }
131 }
132
133 void bdrv_register(BlockDriver *bdrv)
134 {
135 if (!bdrv->bdrv_aio_readv) {
136 /* add AIO emulation layer */
137 bdrv->bdrv_aio_readv = bdrv_aio_readv_em;
138 bdrv->bdrv_aio_writev = bdrv_aio_writev_em;
139 } else if (!bdrv->bdrv_read) {
140 /* add synchronous IO emulation layer */
141 bdrv->bdrv_read = bdrv_read_em;
142 bdrv->bdrv_write = bdrv_write_em;
143 }
144
145 if (!bdrv->bdrv_aio_flush)
146 bdrv->bdrv_aio_flush = bdrv_aio_flush_em;
147
148 QLIST_INSERT_HEAD(&bdrv_drivers, bdrv, list);
149 }
150
151 /* create a new block device (by default it is empty) */
152 BlockDriverState *bdrv_new(const char *device_name)
153 {
154 BlockDriverState *bs;
155
156 bs = qemu_mallocz(sizeof(BlockDriverState));
157 pstrcpy(bs->device_name, sizeof(bs->device_name), device_name);
158 if (device_name[0] != '\0') {
159 QTAILQ_INSERT_TAIL(&bdrv_states, bs, list);
160 }
161 return bs;
162 }
163
164 BlockDriver *bdrv_find_format(const char *format_name)
165 {
166 BlockDriver *drv1;
167 QLIST_FOREACH(drv1, &bdrv_drivers, list) {
168 if (!strcmp(drv1->format_name, format_name)) {
169 return drv1;
170 }
171 }
172 return NULL;
173 }
174
175 static int bdrv_is_whitelisted(BlockDriver *drv)
176 {
177 static const char *whitelist[] = {
178 CONFIG_BDRV_WHITELIST
179 };
180 const char **p;
181
182 if (!whitelist[0])
183 return 1; /* no whitelist, anything goes */
184
185 for (p = whitelist; *p; p++) {
186 if (!strcmp(drv->format_name, *p)) {
187 return 1;
188 }
189 }
190 return 0;
191 }
192
193 BlockDriver *bdrv_find_whitelisted_format(const char *format_name)
194 {
195 BlockDriver *drv = bdrv_find_format(format_name);
196 return drv && bdrv_is_whitelisted(drv) ? drv : NULL;
197 }
198
199 int bdrv_create(BlockDriver *drv, const char* filename,
200 QEMUOptionParameter *options)
201 {
202 if (!drv->bdrv_create)
203 return -ENOTSUP;
204
205 return drv->bdrv_create(filename, options);
206 }
207
208 int bdrv_create_file(const char* filename, QEMUOptionParameter *options)
209 {
210 BlockDriver *drv;
211
212 drv = bdrv_find_protocol(filename);
213 if (drv == NULL) {
214 drv = bdrv_find_format("file");
215 }
216
217 return bdrv_create(drv, filename, options);
218 }
219
220 #ifdef _WIN32
221 void get_tmp_filename(char *filename, int size)
222 {
223 char temp_dir[MAX_PATH];
224
225 GetTempPath(MAX_PATH, temp_dir);
226 GetTempFileName(temp_dir, "qem", 0, filename);
227 }
228 #else
229 void get_tmp_filename(char *filename, int size)
230 {
231 int fd;
232 const char *tmpdir;
233 /* XXX: race condition possible */
234 tmpdir = getenv("TMPDIR");
235 if (!tmpdir)
236 tmpdir = "/tmp";
237 snprintf(filename, size, "%s/vl.XXXXXX", tmpdir);
238 fd = mkstemp(filename);
239 close(fd);
240 }
241 #endif
242
243 #ifdef _WIN32
244 static int is_windows_drive_prefix(const char *filename)
245 {
246 return (((filename[0] >= 'a' && filename[0] <= 'z') ||
247 (filename[0] >= 'A' && filename[0] <= 'Z')) &&
248 filename[1] == ':');
249 }
250
251 int is_windows_drive(const char *filename)
252 {
253 if (is_windows_drive_prefix(filename) &&
254 filename[2] == '\0')
255 return 1;
256 if (strstart(filename, "\\\\.\\", NULL) ||
257 strstart(filename, "//./", NULL))
258 return 1;
259 return 0;
260 }
261 #endif
262
263 /*
264 * Detect host devices. By convention, /dev/cdrom[N] is always
265 * recognized as a host CDROM.
266 */
267 static BlockDriver *find_hdev_driver(const char *filename)
268 {
269 int score_max = 0, score;
270 BlockDriver *drv = NULL, *d;
271
272 QLIST_FOREACH(d, &bdrv_drivers, list) {
273 if (d->bdrv_probe_device) {
274 score = d->bdrv_probe_device(filename);
275 if (score > score_max) {
276 score_max = score;
277 drv = d;
278 }
279 }
280 }
281
282 return drv;
283 }
284
285 BlockDriver *bdrv_find_protocol(const char *filename)
286 {
287 BlockDriver *drv1;
288 char protocol[128];
289 int len;
290 const char *p;
291 int is_drive;
292
293 /* TODO Drivers without bdrv_file_open must be specified explicitly */
294
295 #ifdef _WIN32
296 is_drive = is_windows_drive(filename) ||
297 is_windows_drive_prefix(filename);
298 #else
299 is_drive = 0;
300 #endif
301 p = strchr(filename, ':');
302 if (!p || is_drive) {
303 drv1 = find_hdev_driver(filename);
304 if (!drv1) {
305 drv1 = bdrv_find_format("file");
306 }
307 return drv1;
308 }
309 len = p - filename;
310 if (len > sizeof(protocol) - 1)
311 len = sizeof(protocol) - 1;
312 memcpy(protocol, filename, len);
313 protocol[len] = '\0';
314 QLIST_FOREACH(drv1, &bdrv_drivers, list) {
315 if (drv1->protocol_name &&
316 !strcmp(drv1->protocol_name, protocol)) {
317 return drv1;
318 }
319 }
320 return NULL;
321 }
322
323 static BlockDriver *find_image_format(const char *filename)
324 {
325 int ret, score, score_max;
326 BlockDriver *drv1, *drv;
327 uint8_t buf[2048];
328 BlockDriverState *bs;
329
330 ret = bdrv_file_open(&bs, filename, 0);
331 if (ret < 0)
332 return NULL;
333
334 /* Return the raw BlockDriver * to scsi-generic devices or empty drives */
335 if (bs->sg || !bdrv_is_inserted(bs)) {
336 bdrv_delete(bs);
337 return bdrv_find_format("raw");
338 }
339
340 ret = bdrv_pread(bs, 0, buf, sizeof(buf));
341 bdrv_delete(bs);
342 if (ret < 0) {
343 return NULL;
344 }
345
346 score_max = 0;
347 drv = NULL;
348 QLIST_FOREACH(drv1, &bdrv_drivers, list) {
349 if (drv1->bdrv_probe) {
350 score = drv1->bdrv_probe(buf, ret, filename);
351 if (score > score_max) {
352 score_max = score;
353 drv = drv1;
354 }
355 }
356 }
357 return drv;
358 }
359
360 /**
361 * Set the current 'total_sectors' value
362 */
363 static int refresh_total_sectors(BlockDriverState *bs, int64_t hint)
364 {
365 BlockDriver *drv = bs->drv;
366
367 /* Do not attempt drv->bdrv_getlength() on scsi-generic devices */
368 if (bs->sg)
369 return 0;
370
371 /* query actual device if possible, otherwise just trust the hint */
372 if (drv->bdrv_getlength) {
373 int64_t length = drv->bdrv_getlength(bs);
374 if (length < 0) {
375 return length;
376 }
377 hint = length >> BDRV_SECTOR_BITS;
378 }
379
380 bs->total_sectors = hint;
381 return 0;
382 }
383
384 /*
385 * Common part for opening disk images and files
386 */
387 static int bdrv_open_common(BlockDriverState *bs, const char *filename,
388 int flags, BlockDriver *drv)
389 {
390 int ret, open_flags;
391
392 assert(drv != NULL);
393
394 bs->file = NULL;
395 bs->total_sectors = 0;
396 bs->is_temporary = 0;
397 bs->encrypted = 0;
398 bs->valid_key = 0;
399 bs->open_flags = flags;
400 /* buffer_alignment defaulted to 512, drivers can change this value */
401 bs->buffer_alignment = 512;
402
403 pstrcpy(bs->filename, sizeof(bs->filename), filename);
404
405 if (use_bdrv_whitelist && !bdrv_is_whitelisted(drv)) {
406 return -ENOTSUP;
407 }
408
409 bs->drv = drv;
410 bs->opaque = qemu_mallocz(drv->instance_size);
411
412 /*
413 * Yes, BDRV_O_NOCACHE aka O_DIRECT means we have to present a
414 * write cache to the guest. We do need the fdatasync to flush
415 * out transactions for block allocations, and we maybe have a
416 * volatile write cache in our backing device to deal with.
417 */
418 if (flags & (BDRV_O_CACHE_WB|BDRV_O_NOCACHE))
419 bs->enable_write_cache = 1;
420
421 /*
422 * Clear flags that are internal to the block layer before opening the
423 * image.
424 */
425 open_flags = flags & ~(BDRV_O_SNAPSHOT | BDRV_O_NO_BACKING);
426
427 /*
428 * Snapshots should be writeable.
429 */
430 if (bs->is_temporary) {
431 open_flags |= BDRV_O_RDWR;
432 }
433
434 /* Open the image, either directly or using a protocol */
435 if (drv->bdrv_file_open) {
436 ret = drv->bdrv_file_open(bs, filename, open_flags);
437 } else {
438 ret = bdrv_file_open(&bs->file, filename, open_flags);
439 if (ret >= 0) {
440 ret = drv->bdrv_open(bs, open_flags);
441 }
442 }
443
444 if (ret < 0) {
445 goto free_and_fail;
446 }
447
448 bs->keep_read_only = bs->read_only = !(open_flags & BDRV_O_RDWR);
449
450 ret = refresh_total_sectors(bs, bs->total_sectors);
451 if (ret < 0) {
452 goto free_and_fail;
453 }
454
455 #ifndef _WIN32
456 if (bs->is_temporary) {
457 unlink(filename);
458 }
459 #endif
460 return 0;
461
462 free_and_fail:
463 if (bs->file) {
464 bdrv_delete(bs->file);
465 bs->file = NULL;
466 }
467 qemu_free(bs->opaque);
468 bs->opaque = NULL;
469 bs->drv = NULL;
470 return ret;
471 }
472
473 /*
474 * Opens a file using a protocol (file, host_device, nbd, ...)
475 */
476 int bdrv_file_open(BlockDriverState **pbs, const char *filename, int flags)
477 {
478 BlockDriverState *bs;
479 BlockDriver *drv;
480 int ret;
481
482 drv = bdrv_find_protocol(filename);
483 if (!drv) {
484 return -ENOENT;
485 }
486
487 bs = bdrv_new("");
488 ret = bdrv_open_common(bs, filename, flags, drv);
489 if (ret < 0) {
490 bdrv_delete(bs);
491 return ret;
492 }
493 bs->growable = 1;
494 *pbs = bs;
495 return 0;
496 }
497
498 /*
499 * Opens a disk image (raw, qcow2, vmdk, ...)
500 */
501 int bdrv_open(BlockDriverState *bs, const char *filename, int flags,
502 BlockDriver *drv)
503 {
504 int ret;
505
506 if (flags & BDRV_O_SNAPSHOT) {
507 BlockDriverState *bs1;
508 int64_t total_size;
509 int is_protocol = 0;
510 BlockDriver *bdrv_qcow2;
511 QEMUOptionParameter *options;
512 char tmp_filename[PATH_MAX];
513 char backing_filename[PATH_MAX];
514
515 /* if snapshot, we create a temporary backing file and open it
516 instead of opening 'filename' directly */
517
518 /* if there is a backing file, use it */
519 bs1 = bdrv_new("");
520 ret = bdrv_open(bs1, filename, 0, drv);
521 if (ret < 0) {
522 bdrv_delete(bs1);
523 return ret;
524 }
525 total_size = bdrv_getlength(bs1) & BDRV_SECTOR_MASK;
526
527 if (bs1->drv && bs1->drv->protocol_name)
528 is_protocol = 1;
529
530 bdrv_delete(bs1);
531
532 get_tmp_filename(tmp_filename, sizeof(tmp_filename));
533
534 /* Real path is meaningless for protocols */
535 if (is_protocol)
536 snprintf(backing_filename, sizeof(backing_filename),
537 "%s", filename);
538 else if (!realpath(filename, backing_filename))
539 return -errno;
540
541 bdrv_qcow2 = bdrv_find_format("qcow2");
542 options = parse_option_parameters("", bdrv_qcow2->create_options, NULL);
543
544 set_option_parameter_int(options, BLOCK_OPT_SIZE, total_size);
545 set_option_parameter(options, BLOCK_OPT_BACKING_FILE, backing_filename);
546 if (drv) {
547 set_option_parameter(options, BLOCK_OPT_BACKING_FMT,
548 drv->format_name);
549 }
550
551 ret = bdrv_create(bdrv_qcow2, tmp_filename, options);
552 free_option_parameters(options);
553 if (ret < 0) {
554 return ret;
555 }
556
557 filename = tmp_filename;
558 drv = bdrv_qcow2;
559 bs->is_temporary = 1;
560 }
561
562 /* Find the right image format driver */
563 if (!drv) {
564 drv = find_image_format(filename);
565 }
566
567 if (!drv) {
568 ret = -ENOENT;
569 goto unlink_and_fail;
570 }
571
572 /* Open the image */
573 ret = bdrv_open_common(bs, filename, flags, drv);
574 if (ret < 0) {
575 goto unlink_and_fail;
576 }
577
578 /* If there is a backing file, use it */
579 if ((flags & BDRV_O_NO_BACKING) == 0 && bs->backing_file[0] != '\0') {
580 char backing_filename[PATH_MAX];
581 int back_flags;
582 BlockDriver *back_drv = NULL;
583
584 bs->backing_hd = bdrv_new("");
585 path_combine(backing_filename, sizeof(backing_filename),
586 filename, bs->backing_file);
587 if (bs->backing_format[0] != '\0')
588 back_drv = bdrv_find_format(bs->backing_format);
589
590 /* backing files always opened read-only */
591 back_flags =
592 flags & ~(BDRV_O_RDWR | BDRV_O_SNAPSHOT | BDRV_O_NO_BACKING);
593
594 ret = bdrv_open(bs->backing_hd, backing_filename, back_flags, back_drv);
595 if (ret < 0) {
596 bdrv_close(bs);
597 return ret;
598 }
599 if (bs->is_temporary) {
600 bs->backing_hd->keep_read_only = !(flags & BDRV_O_RDWR);
601 } else {
602 /* base image inherits from "parent" */
603 bs->backing_hd->keep_read_only = bs->keep_read_only;
604 }
605 }
606
607 if (!bdrv_key_required(bs)) {
608 /* call the change callback */
609 bs->media_changed = 1;
610 if (bs->change_cb)
611 bs->change_cb(bs->change_opaque);
612 }
613
614 return 0;
615
616 unlink_and_fail:
617 if (bs->is_temporary) {
618 unlink(filename);
619 }
620 return ret;
621 }
622
623 void bdrv_close(BlockDriverState *bs)
624 {
625 if (bs->drv) {
626 if (bs->backing_hd) {
627 bdrv_delete(bs->backing_hd);
628 bs->backing_hd = NULL;
629 }
630 bs->drv->bdrv_close(bs);
631 qemu_free(bs->opaque);
632 #ifdef _WIN32
633 if (bs->is_temporary) {
634 unlink(bs->filename);
635 }
636 #endif
637 bs->opaque = NULL;
638 bs->drv = NULL;
639
640 if (bs->file != NULL) {
641 bdrv_close(bs->file);
642 }
643
644 /* call the change callback */
645 bs->media_changed = 1;
646 if (bs->change_cb)
647 bs->change_cb(bs->change_opaque);
648 }
649 }
650
651 void bdrv_close_all(void)
652 {
653 BlockDriverState *bs;
654
655 QTAILQ_FOREACH(bs, &bdrv_states, list) {
656 bdrv_close(bs);
657 }
658 }
659
660 void bdrv_delete(BlockDriverState *bs)
661 {
662 /* remove from list, if necessary */
663 if (bs->device_name[0] != '\0') {
664 QTAILQ_REMOVE(&bdrv_states, bs, list);
665 }
666
667 bdrv_close(bs);
668 if (bs->file != NULL) {
669 bdrv_delete(bs->file);
670 }
671
672 qemu_free(bs);
673 }
674
675 /*
676 * Run consistency checks on an image
677 *
678 * Returns the number of errors or -errno when an internal error occurs
679 */
680 int bdrv_check(BlockDriverState *bs)
681 {
682 if (bs->drv->bdrv_check == NULL) {
683 return -ENOTSUP;
684 }
685
686 return bs->drv->bdrv_check(bs);
687 }
688
689 /* commit COW file into the raw image */
690 int bdrv_commit(BlockDriverState *bs)
691 {
692 BlockDriver *drv = bs->drv;
693 int64_t i, total_sectors;
694 int n, j, ro, open_flags;
695 int ret = 0, rw_ret = 0;
696 unsigned char sector[BDRV_SECTOR_SIZE];
697 char filename[1024];
698 BlockDriverState *bs_rw, *bs_ro;
699
700 if (!drv)
701 return -ENOMEDIUM;
702
703 if (!bs->backing_hd) {
704 return -ENOTSUP;
705 }
706
707 if (bs->backing_hd->keep_read_only) {
708 return -EACCES;
709 }
710
711 ro = bs->backing_hd->read_only;
712 strncpy(filename, bs->backing_hd->filename, sizeof(filename));
713 open_flags = bs->backing_hd->open_flags;
714
715 if (ro) {
716 /* re-open as RW */
717 bdrv_delete(bs->backing_hd);
718 bs->backing_hd = NULL;
719 bs_rw = bdrv_new("");
720 rw_ret = bdrv_open(bs_rw, filename, open_flags | BDRV_O_RDWR, drv);
721 if (rw_ret < 0) {
722 bdrv_delete(bs_rw);
723 /* try to re-open read-only */
724 bs_ro = bdrv_new("");
725 ret = bdrv_open(bs_ro, filename, open_flags & ~BDRV_O_RDWR, drv);
726 if (ret < 0) {
727 bdrv_delete(bs_ro);
728 /* drive not functional anymore */
729 bs->drv = NULL;
730 return ret;
731 }
732 bs->backing_hd = bs_ro;
733 return rw_ret;
734 }
735 bs->backing_hd = bs_rw;
736 }
737
738 total_sectors = bdrv_getlength(bs) >> BDRV_SECTOR_BITS;
739 for (i = 0; i < total_sectors;) {
740 if (drv->bdrv_is_allocated(bs, i, 65536, &n)) {
741 for(j = 0; j < n; j++) {
742 if (bdrv_read(bs, i, sector, 1) != 0) {
743 ret = -EIO;
744 goto ro_cleanup;
745 }
746
747 if (bdrv_write(bs->backing_hd, i, sector, 1) != 0) {
748 ret = -EIO;
749 goto ro_cleanup;
750 }
751 i++;
752 }
753 } else {
754 i += n;
755 }
756 }
757
758 if (drv->bdrv_make_empty) {
759 ret = drv->bdrv_make_empty(bs);
760 bdrv_flush(bs);
761 }
762
763 /*
764 * Make sure all data we wrote to the backing device is actually
765 * stable on disk.
766 */
767 if (bs->backing_hd)
768 bdrv_flush(bs->backing_hd);
769
770 ro_cleanup:
771
772 if (ro) {
773 /* re-open as RO */
774 bdrv_delete(bs->backing_hd);
775 bs->backing_hd = NULL;
776 bs_ro = bdrv_new("");
777 ret = bdrv_open(bs_ro, filename, open_flags & ~BDRV_O_RDWR, drv);
778 if (ret < 0) {
779 bdrv_delete(bs_ro);
780 /* drive not functional anymore */
781 bs->drv = NULL;
782 return ret;
783 }
784 bs->backing_hd = bs_ro;
785 bs->backing_hd->keep_read_only = 0;
786 }
787
788 return ret;
789 }
790
791 /*
792 * Return values:
793 * 0 - success
794 * -EINVAL - backing format specified, but no file
795 * -ENOSPC - can't update the backing file because no space is left in the
796 * image file header
797 * -ENOTSUP - format driver doesn't support changing the backing file
798 */
799 int bdrv_change_backing_file(BlockDriverState *bs,
800 const char *backing_file, const char *backing_fmt)
801 {
802 BlockDriver *drv = bs->drv;
803
804 if (drv->bdrv_change_backing_file != NULL) {
805 return drv->bdrv_change_backing_file(bs, backing_file, backing_fmt);
806 } else {
807 return -ENOTSUP;
808 }
809 }
810
811 static int bdrv_check_byte_request(BlockDriverState *bs, int64_t offset,
812 size_t size)
813 {
814 int64_t len;
815
816 if (!bdrv_is_inserted(bs))
817 return -ENOMEDIUM;
818
819 if (bs->growable)
820 return 0;
821
822 len = bdrv_getlength(bs);
823
824 if (offset < 0)
825 return -EIO;
826
827 if ((offset > len) || (len - offset < size))
828 return -EIO;
829
830 return 0;
831 }
832
833 static int bdrv_check_request(BlockDriverState *bs, int64_t sector_num,
834 int nb_sectors)
835 {
836 return bdrv_check_byte_request(bs, sector_num * BDRV_SECTOR_SIZE,
837 nb_sectors * BDRV_SECTOR_SIZE);
838 }
839
840 /* return < 0 if error. See bdrv_write() for the return codes */
841 int bdrv_read(BlockDriverState *bs, int64_t sector_num,
842 uint8_t *buf, int nb_sectors)
843 {
844 BlockDriver *drv = bs->drv;
845
846 if (!drv)
847 return -ENOMEDIUM;
848 if (bdrv_check_request(bs, sector_num, nb_sectors))
849 return -EIO;
850
851 return drv->bdrv_read(bs, sector_num, buf, nb_sectors);
852 }
853
854 static void set_dirty_bitmap(BlockDriverState *bs, int64_t sector_num,
855 int nb_sectors, int dirty)
856 {
857 int64_t start, end;
858 unsigned long val, idx, bit;
859
860 start = sector_num / BDRV_SECTORS_PER_DIRTY_CHUNK;
861 end = (sector_num + nb_sectors - 1) / BDRV_SECTORS_PER_DIRTY_CHUNK;
862
863 for (; start <= end; start++) {
864 idx = start / (sizeof(unsigned long) * 8);
865 bit = start % (sizeof(unsigned long) * 8);
866 val = bs->dirty_bitmap[idx];
867 if (dirty) {
868 if (!(val & (1 << bit))) {
869 bs->dirty_count++;
870 val |= 1 << bit;
871 }
872 } else {
873 if (val & (1 << bit)) {
874 bs->dirty_count--;
875 val &= ~(1 << bit);
876 }
877 }
878 bs->dirty_bitmap[idx] = val;
879 }
880 }
881
882 /* Return < 0 if error. Important errors are:
883 -EIO generic I/O error (may happen for all errors)
884 -ENOMEDIUM No media inserted.
885 -EINVAL Invalid sector number or nb_sectors
886 -EACCES Trying to write a read-only device
887 */
888 int bdrv_write(BlockDriverState *bs, int64_t sector_num,
889 const uint8_t *buf, int nb_sectors)
890 {
891 BlockDriver *drv = bs->drv;
892 if (!bs->drv)
893 return -ENOMEDIUM;
894 if (bs->read_only)
895 return -EACCES;
896 if (bdrv_check_request(bs, sector_num, nb_sectors))
897 return -EIO;
898
899 if (bs->dirty_bitmap) {
900 set_dirty_bitmap(bs, sector_num, nb_sectors, 1);
901 }
902
903 if (bs->wr_highest_sector < sector_num + nb_sectors - 1) {
904 bs->wr_highest_sector = sector_num + nb_sectors - 1;
905 }
906
907 return drv->bdrv_write(bs, sector_num, buf, nb_sectors);
908 }
909
910 int bdrv_pread(BlockDriverState *bs, int64_t offset,
911 void *buf, int count1)
912 {
913 uint8_t tmp_buf[BDRV_SECTOR_SIZE];
914 int len, nb_sectors, count;
915 int64_t sector_num;
916 int ret;
917
918 count = count1;
919 /* first read to align to sector start */
920 len = (BDRV_SECTOR_SIZE - offset) & (BDRV_SECTOR_SIZE - 1);
921 if (len > count)
922 len = count;
923 sector_num = offset >> BDRV_SECTOR_BITS;
924 if (len > 0) {
925 if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0)
926 return ret;
927 memcpy(buf, tmp_buf + (offset & (BDRV_SECTOR_SIZE - 1)), len);
928 count -= len;
929 if (count == 0)
930 return count1;
931 sector_num++;
932 buf += len;
933 }
934
935 /* read the sectors "in place" */
936 nb_sectors = count >> BDRV_SECTOR_BITS;
937 if (nb_sectors > 0) {
938 if ((ret = bdrv_read(bs, sector_num, buf, nb_sectors)) < 0)
939 return ret;
940 sector_num += nb_sectors;
941 len = nb_sectors << BDRV_SECTOR_BITS;
942 buf += len;
943 count -= len;
944 }
945
946 /* add data from the last sector */
947 if (count > 0) {
948 if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0)
949 return ret;
950 memcpy(buf, tmp_buf, count);
951 }
952 return count1;
953 }
954
955 int bdrv_pwrite(BlockDriverState *bs, int64_t offset,
956 const void *buf, int count1)
957 {
958 uint8_t tmp_buf[BDRV_SECTOR_SIZE];
959 int len, nb_sectors, count;
960 int64_t sector_num;
961 int ret;
962
963 count = count1;
964 /* first write to align to sector start */
965 len = (BDRV_SECTOR_SIZE - offset) & (BDRV_SECTOR_SIZE - 1);
966 if (len > count)
967 len = count;
968 sector_num = offset >> BDRV_SECTOR_BITS;
969 if (len > 0) {
970 if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0)
971 return ret;
972 memcpy(tmp_buf + (offset & (BDRV_SECTOR_SIZE - 1)), buf, len);
973 if ((ret = bdrv_write(bs, sector_num, tmp_buf, 1)) < 0)
974 return ret;
975 count -= len;
976 if (count == 0)
977 return count1;
978 sector_num++;
979 buf += len;
980 }
981
982 /* write the sectors "in place" */
983 nb_sectors = count >> BDRV_SECTOR_BITS;
984 if (nb_sectors > 0) {
985 if ((ret = bdrv_write(bs, sector_num, buf, nb_sectors)) < 0)
986 return ret;
987 sector_num += nb_sectors;
988 len = nb_sectors << BDRV_SECTOR_BITS;
989 buf += len;
990 count -= len;
991 }
992
993 /* add data from the last sector */
994 if (count > 0) {
995 if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0)
996 return ret;
997 memcpy(tmp_buf, buf, count);
998 if ((ret = bdrv_write(bs, sector_num, tmp_buf, 1)) < 0)
999 return ret;
1000 }
1001 return count1;
1002 }
1003
1004 /**
1005 * Truncate file to 'offset' bytes (needed only for file protocols)
1006 */
1007 int bdrv_truncate(BlockDriverState *bs, int64_t offset)
1008 {
1009 BlockDriver *drv = bs->drv;
1010 int ret;
1011 if (!drv)
1012 return -ENOMEDIUM;
1013 if (!drv->bdrv_truncate)
1014 return -ENOTSUP;
1015 if (bs->read_only)
1016 return -EACCES;
1017 ret = drv->bdrv_truncate(bs, offset);
1018 if (ret == 0) {
1019 ret = refresh_total_sectors(bs, offset >> BDRV_SECTOR_BITS);
1020 }
1021 return ret;
1022 }
1023
1024 /**
1025 * Length of a file in bytes. Return < 0 if error or unknown.
1026 */
1027 int64_t bdrv_getlength(BlockDriverState *bs)
1028 {
1029 BlockDriver *drv = bs->drv;
1030 if (!drv)
1031 return -ENOMEDIUM;
1032
1033 /* Fixed size devices use the total_sectors value for speed instead of
1034 issuing a length query (like lseek) on each call. Also, legacy block
1035 drivers don't provide a bdrv_getlength function and must use
1036 total_sectors. */
1037 if (!bs->growable || !drv->bdrv_getlength) {
1038 return bs->total_sectors * BDRV_SECTOR_SIZE;
1039 }
1040 return drv->bdrv_getlength(bs);
1041 }
1042
1043 /* return 0 as number of sectors if no device present or error */
1044 void bdrv_get_geometry(BlockDriverState *bs, uint64_t *nb_sectors_ptr)
1045 {
1046 int64_t length;
1047 length = bdrv_getlength(bs);
1048 if (length < 0)
1049 length = 0;
1050 else
1051 length = length >> BDRV_SECTOR_BITS;
1052 *nb_sectors_ptr = length;
1053 }
1054
1055 struct partition {
1056 uint8_t boot_ind; /* 0x80 - active */
1057 uint8_t head; /* starting head */
1058 uint8_t sector; /* starting sector */
1059 uint8_t cyl; /* starting cylinder */
1060 uint8_t sys_ind; /* What partition type */
1061 uint8_t end_head; /* end head */
1062 uint8_t end_sector; /* end sector */
1063 uint8_t end_cyl; /* end cylinder */
1064 uint32_t start_sect; /* starting sector counting from 0 */
1065 uint32_t nr_sects; /* nr of sectors in partition */
1066 } __attribute__((packed));
1067
1068 /* try to guess the disk logical geometry from the MSDOS partition table. Return 0 if OK, -1 if could not guess */
1069 static int guess_disk_lchs(BlockDriverState *bs,
1070 int *pcylinders, int *pheads, int *psectors)
1071 {
1072 uint8_t buf[BDRV_SECTOR_SIZE];
1073 int ret, i, heads, sectors, cylinders;
1074 struct partition *p;
1075 uint32_t nr_sects;
1076 uint64_t nb_sectors;
1077
1078 bdrv_get_geometry(bs, &nb_sectors);
1079
1080 ret = bdrv_read(bs, 0, buf, 1);
1081 if (ret < 0)
1082 return -1;
1083 /* test msdos magic */
1084 if (buf[510] != 0x55 || buf[511] != 0xaa)
1085 return -1;
1086 for(i = 0; i < 4; i++) {
1087 p = ((struct partition *)(buf + 0x1be)) + i;
1088 nr_sects = le32_to_cpu(p->nr_sects);
1089 if (nr_sects && p->end_head) {
1090 /* We make the assumption that the partition terminates on
1091 a cylinder boundary */
1092 heads = p->end_head + 1;
1093 sectors = p->end_sector & 63;
1094 if (sectors == 0)
1095 continue;
1096 cylinders = nb_sectors / (heads * sectors);
1097 if (cylinders < 1 || cylinders > 16383)
1098 continue;
1099 *pheads = heads;
1100 *psectors = sectors;
1101 *pcylinders = cylinders;
1102 #if 0
1103 printf("guessed geometry: LCHS=%d %d %d\n",
1104 cylinders, heads, sectors);
1105 #endif
1106 return 0;
1107 }
1108 }
1109 return -1;
1110 }
1111
1112 void bdrv_guess_geometry(BlockDriverState *bs, int *pcyls, int *pheads, int *psecs)
1113 {
1114 int translation, lba_detected = 0;
1115 int cylinders, heads, secs;
1116 uint64_t nb_sectors;
1117
1118 /* if a geometry hint is available, use it */
1119 bdrv_get_geometry(bs, &nb_sectors);
1120 bdrv_get_geometry_hint(bs, &cylinders, &heads, &secs);
1121 translation = bdrv_get_translation_hint(bs);
1122 if (cylinders != 0) {
1123 *pcyls = cylinders;
1124 *pheads = heads;
1125 *psecs = secs;
1126 } else {
1127 if (guess_disk_lchs(bs, &cylinders, &heads, &secs) == 0) {
1128 if (heads > 16) {
1129 /* if heads > 16, it means that a BIOS LBA
1130 translation was active, so the default
1131 hardware geometry is OK */
1132 lba_detected = 1;
1133 goto default_geometry;
1134 } else {
1135 *pcyls = cylinders;
1136 *pheads = heads;
1137 *psecs = secs;
1138 /* disable any translation to be in sync with
1139 the logical geometry */
1140 if (translation == BIOS_ATA_TRANSLATION_AUTO) {
1141 bdrv_set_translation_hint(bs,
1142 BIOS_ATA_TRANSLATION_NONE);
1143 }
1144 }
1145 } else {
1146 default_geometry:
1147 /* if no geometry, use a standard physical disk geometry */
1148 cylinders = nb_sectors / (16 * 63);
1149
1150 if (cylinders > 16383)
1151 cylinders = 16383;
1152 else if (cylinders < 2)
1153 cylinders = 2;
1154 *pcyls = cylinders;
1155 *pheads = 16;
1156 *psecs = 63;
1157 if ((lba_detected == 1) && (translation == BIOS_ATA_TRANSLATION_AUTO)) {
1158 if ((*pcyls * *pheads) <= 131072) {
1159 bdrv_set_translation_hint(bs,
1160 BIOS_ATA_TRANSLATION_LARGE);
1161 } else {
1162 bdrv_set_translation_hint(bs,
1163 BIOS_ATA_TRANSLATION_LBA);
1164 }
1165 }
1166 }
1167 bdrv_set_geometry_hint(bs, *pcyls, *pheads, *psecs);
1168 }
1169 }
1170
1171 void bdrv_set_geometry_hint(BlockDriverState *bs,
1172 int cyls, int heads, int secs)
1173 {
1174 bs->cyls = cyls;
1175 bs->heads = heads;
1176 bs->secs = secs;
1177 }
1178
1179 void bdrv_set_type_hint(BlockDriverState *bs, int type)
1180 {
1181 bs->type = type;
1182 bs->removable = ((type == BDRV_TYPE_CDROM ||
1183 type == BDRV_TYPE_FLOPPY));
1184 }
1185
1186 void bdrv_set_translation_hint(BlockDriverState *bs, int translation)
1187 {
1188 bs->translation = translation;
1189 }
1190
1191 void bdrv_get_geometry_hint(BlockDriverState *bs,
1192 int *pcyls, int *pheads, int *psecs)
1193 {
1194 *pcyls = bs->cyls;
1195 *pheads = bs->heads;
1196 *psecs = bs->secs;
1197 }
1198
1199 int bdrv_get_type_hint(BlockDriverState *bs)
1200 {
1201 return bs->type;
1202 }
1203
1204 int bdrv_get_translation_hint(BlockDriverState *bs)
1205 {
1206 return bs->translation;
1207 }
1208
1209 int bdrv_is_removable(BlockDriverState *bs)
1210 {
1211 return bs->removable;
1212 }
1213
1214 int bdrv_is_read_only(BlockDriverState *bs)
1215 {
1216 return bs->read_only;
1217 }
1218
1219 int bdrv_is_sg(BlockDriverState *bs)
1220 {
1221 return bs->sg;
1222 }
1223
1224 int bdrv_enable_write_cache(BlockDriverState *bs)
1225 {
1226 return bs->enable_write_cache;
1227 }
1228
1229 /* XXX: no longer used */
1230 void bdrv_set_change_cb(BlockDriverState *bs,
1231 void (*change_cb)(void *opaque), void *opaque)
1232 {
1233 bs->change_cb = change_cb;
1234 bs->change_opaque = opaque;
1235 }
1236
1237 int bdrv_is_encrypted(BlockDriverState *bs)
1238 {
1239 if (bs->backing_hd && bs->backing_hd->encrypted)
1240 return 1;
1241 return bs->encrypted;
1242 }
1243
1244 int bdrv_key_required(BlockDriverState *bs)
1245 {
1246 BlockDriverState *backing_hd = bs->backing_hd;
1247
1248 if (backing_hd && backing_hd->encrypted && !backing_hd->valid_key)
1249 return 1;
1250 return (bs->encrypted && !bs->valid_key);
1251 }
1252
1253 int bdrv_set_key(BlockDriverState *bs, const char *key)
1254 {
1255 int ret;
1256 if (bs->backing_hd && bs->backing_hd->encrypted) {
1257 ret = bdrv_set_key(bs->backing_hd, key);
1258 if (ret < 0)
1259 return ret;
1260 if (!bs->encrypted)
1261 return 0;
1262 }
1263 if (!bs->encrypted) {
1264 return -EINVAL;
1265 } else if (!bs->drv || !bs->drv->bdrv_set_key) {
1266 return -ENOMEDIUM;
1267 }
1268 ret = bs->drv->bdrv_set_key(bs, key);
1269 if (ret < 0) {
1270 bs->valid_key = 0;
1271 } else if (!bs->valid_key) {
1272 bs->valid_key = 1;
1273 /* call the change callback now, we skipped it on open */
1274 bs->media_changed = 1;
1275 if (bs->change_cb)
1276 bs->change_cb(bs->change_opaque);
1277 }
1278 return ret;
1279 }
1280
1281 void bdrv_get_format(BlockDriverState *bs, char *buf, int buf_size)
1282 {
1283 if (!bs->drv) {
1284 buf[0] = '\0';
1285 } else {
1286 pstrcpy(buf, buf_size, bs->drv->format_name);
1287 }
1288 }
1289
1290 void bdrv_iterate_format(void (*it)(void *opaque, const char *name),
1291 void *opaque)
1292 {
1293 BlockDriver *drv;
1294
1295 QLIST_FOREACH(drv, &bdrv_drivers, list) {
1296 it(opaque, drv->format_name);
1297 }
1298 }
1299
1300 BlockDriverState *bdrv_find(const char *name)
1301 {
1302 BlockDriverState *bs;
1303
1304 QTAILQ_FOREACH(bs, &bdrv_states, list) {
1305 if (!strcmp(name, bs->device_name)) {
1306 return bs;
1307 }
1308 }
1309 return NULL;
1310 }
1311
1312 void bdrv_iterate(void (*it)(void *opaque, BlockDriverState *bs), void *opaque)
1313 {
1314 BlockDriverState *bs;
1315
1316 QTAILQ_FOREACH(bs, &bdrv_states, list) {
1317 it(opaque, bs);
1318 }
1319 }
1320
1321 const char *bdrv_get_device_name(BlockDriverState *bs)
1322 {
1323 return bs->device_name;
1324 }
1325
1326 void bdrv_flush(BlockDriverState *bs)
1327 {
1328 if (bs->open_flags & BDRV_O_NO_FLUSH) {
1329 return;
1330 }
1331
1332 if (bs->drv && bs->drv->bdrv_flush)
1333 bs->drv->bdrv_flush(bs);
1334 }
1335
1336 void bdrv_flush_all(void)
1337 {
1338 BlockDriverState *bs;
1339
1340 QTAILQ_FOREACH(bs, &bdrv_states, list) {
1341 if (bs->drv && !bdrv_is_read_only(bs) &&
1342 (!bdrv_is_removable(bs) || bdrv_is_inserted(bs))) {
1343 bdrv_flush(bs);
1344 }
1345 }
1346 }
1347
1348 int bdrv_has_zero_init(BlockDriverState *bs)
1349 {
1350 assert(bs->drv);
1351
1352 if (bs->drv->no_zero_init) {
1353 return 0;
1354 } else if (bs->file) {
1355 return bdrv_has_zero_init(bs->file);
1356 }
1357
1358 return 1;
1359 }
1360
1361 /*
1362 * Returns true iff the specified sector is present in the disk image. Drivers
1363 * not implementing the functionality are assumed to not support backing files,
1364 * hence all their sectors are reported as allocated.
1365 *
1366 * 'pnum' is set to the number of sectors (including and immediately following
1367 * the specified sector) that are known to be in the same
1368 * allocated/unallocated state.
1369 *
1370 * 'nb_sectors' is the max value 'pnum' should be set to.
1371 */
1372 int bdrv_is_allocated(BlockDriverState *bs, int64_t sector_num, int nb_sectors,
1373 int *pnum)
1374 {
1375 int64_t n;
1376 if (!bs->drv->bdrv_is_allocated) {
1377 if (sector_num >= bs->total_sectors) {
1378 *pnum = 0;
1379 return 0;
1380 }
1381 n = bs->total_sectors - sector_num;
1382 *pnum = (n < nb_sectors) ? (n) : (nb_sectors);
1383 return 1;
1384 }
1385 return bs->drv->bdrv_is_allocated(bs, sector_num, nb_sectors, pnum);
1386 }
1387
1388 void bdrv_mon_event(const BlockDriverState *bdrv,
1389 BlockMonEventAction action, int is_read)
1390 {
1391 QObject *data;
1392 const char *action_str;
1393
1394 switch (action) {
1395 case BDRV_ACTION_REPORT:
1396 action_str = "report";
1397 break;
1398 case BDRV_ACTION_IGNORE:
1399 action_str = "ignore";
1400 break;
1401 case BDRV_ACTION_STOP:
1402 action_str = "stop";
1403 break;
1404 default:
1405 abort();
1406 }
1407
1408 data = qobject_from_jsonf("{ 'device': %s, 'action': %s, 'operation': %s }",
1409 bdrv->device_name,
1410 action_str,
1411 is_read ? "read" : "write");
1412 monitor_protocol_event(QEVENT_BLOCK_IO_ERROR, data);
1413
1414 qobject_decref(data);
1415 }
1416
1417 static void bdrv_print_dict(QObject *obj, void *opaque)
1418 {
1419 QDict *bs_dict;
1420 Monitor *mon = opaque;
1421
1422 bs_dict = qobject_to_qdict(obj);
1423
1424 monitor_printf(mon, "%s: type=%s removable=%d",
1425 qdict_get_str(bs_dict, "device"),
1426 qdict_get_str(bs_dict, "type"),
1427 qdict_get_bool(bs_dict, "removable"));
1428
1429 if (qdict_get_bool(bs_dict, "removable")) {
1430 monitor_printf(mon, " locked=%d", qdict_get_bool(bs_dict, "locked"));
1431 }
1432
1433 if (qdict_haskey(bs_dict, "inserted")) {
1434 QDict *qdict = qobject_to_qdict(qdict_get(bs_dict, "inserted"));
1435
1436 monitor_printf(mon, " file=");
1437 monitor_print_filename(mon, qdict_get_str(qdict, "file"));
1438 if (qdict_haskey(qdict, "backing_file")) {
1439 monitor_printf(mon, " backing_file=");
1440 monitor_print_filename(mon, qdict_get_str(qdict, "backing_file"));
1441 }
1442 monitor_printf(mon, " ro=%d drv=%s encrypted=%d",
1443 qdict_get_bool(qdict, "ro"),
1444 qdict_get_str(qdict, "drv"),
1445 qdict_get_bool(qdict, "encrypted"));
1446 } else {
1447 monitor_printf(mon, " [not inserted]");
1448 }
1449
1450 monitor_printf(mon, "\n");
1451 }
1452
1453 void bdrv_info_print(Monitor *mon, const QObject *data)
1454 {
1455 qlist_iter(qobject_to_qlist(data), bdrv_print_dict, mon);
1456 }
1457
1458 void bdrv_info(Monitor *mon, QObject **ret_data)
1459 {
1460 QList *bs_list;
1461 BlockDriverState *bs;
1462
1463 bs_list = qlist_new();
1464
1465 QTAILQ_FOREACH(bs, &bdrv_states, list) {
1466 QObject *bs_obj;
1467 const char *type = "unknown";
1468
1469 switch(bs->type) {
1470 case BDRV_TYPE_HD:
1471 type = "hd";
1472 break;
1473 case BDRV_TYPE_CDROM:
1474 type = "cdrom";
1475 break;
1476 case BDRV_TYPE_FLOPPY:
1477 type = "floppy";
1478 break;
1479 }
1480
1481 bs_obj = qobject_from_jsonf("{ 'device': %s, 'type': %s, "
1482 "'removable': %i, 'locked': %i }",
1483 bs->device_name, type, bs->removable,
1484 bs->locked);
1485
1486 if (bs->drv) {
1487 QObject *obj;
1488 QDict *bs_dict = qobject_to_qdict(bs_obj);
1489
1490 obj = qobject_from_jsonf("{ 'file': %s, 'ro': %i, 'drv': %s, "
1491 "'encrypted': %i }",
1492 bs->filename, bs->read_only,
1493 bs->drv->format_name,
1494 bdrv_is_encrypted(bs));
1495 if (bs->backing_file[0] != '\0') {
1496 QDict *qdict = qobject_to_qdict(obj);
1497 qdict_put(qdict, "backing_file",
1498 qstring_from_str(bs->backing_file));
1499 }
1500
1501 qdict_put_obj(bs_dict, "inserted", obj);
1502 }
1503 qlist_append_obj(bs_list, bs_obj);
1504 }
1505
1506 *ret_data = QOBJECT(bs_list);
1507 }
1508
1509 static void bdrv_stats_iter(QObject *data, void *opaque)
1510 {
1511 QDict *qdict;
1512 Monitor *mon = opaque;
1513
1514 qdict = qobject_to_qdict(data);
1515 monitor_printf(mon, "%s:", qdict_get_str(qdict, "device"));
1516
1517 qdict = qobject_to_qdict(qdict_get(qdict, "stats"));
1518 monitor_printf(mon, " rd_bytes=%" PRId64
1519 " wr_bytes=%" PRId64
1520 " rd_operations=%" PRId64
1521 " wr_operations=%" PRId64
1522 "\n",
1523 qdict_get_int(qdict, "rd_bytes"),
1524 qdict_get_int(qdict, "wr_bytes"),
1525 qdict_get_int(qdict, "rd_operations"),
1526 qdict_get_int(qdict, "wr_operations"));
1527 }
1528
1529 void bdrv_stats_print(Monitor *mon, const QObject *data)
1530 {
1531 qlist_iter(qobject_to_qlist(data), bdrv_stats_iter, mon);
1532 }
1533
1534 static QObject* bdrv_info_stats_bs(BlockDriverState *bs)
1535 {
1536 QObject *res;
1537 QDict *dict;
1538
1539 res = qobject_from_jsonf("{ 'stats': {"
1540 "'rd_bytes': %" PRId64 ","
1541 "'wr_bytes': %" PRId64 ","
1542 "'rd_operations': %" PRId64 ","
1543 "'wr_operations': %" PRId64 ","
1544 "'wr_highest_offset': %" PRId64
1545 "} }",
1546 bs->rd_bytes, bs->wr_bytes,
1547 bs->rd_ops, bs->wr_ops,
1548 bs->wr_highest_sector * (long)BDRV_SECTOR_SIZE);
1549 dict = qobject_to_qdict(res);
1550
1551 if (*bs->device_name) {
1552 qdict_put(dict, "device", qstring_from_str(bs->device_name));
1553 }
1554
1555 if (bs->file) {
1556 QObject *parent = bdrv_info_stats_bs(bs->file);
1557 qdict_put_obj(dict, "parent", parent);
1558 }
1559
1560 return res;
1561 }
1562
1563 void bdrv_info_stats(Monitor *mon, QObject **ret_data)
1564 {
1565 QObject *obj;
1566 QList *devices;
1567 BlockDriverState *bs;
1568
1569 devices = qlist_new();
1570
1571 QTAILQ_FOREACH(bs, &bdrv_states, list) {
1572 obj = bdrv_info_stats_bs(bs);
1573 qlist_append_obj(devices, obj);
1574 }
1575
1576 *ret_data = QOBJECT(devices);
1577 }
1578
1579 const char *bdrv_get_encrypted_filename(BlockDriverState *bs)
1580 {
1581 if (bs->backing_hd && bs->backing_hd->encrypted)
1582 return bs->backing_file;
1583 else if (bs->encrypted)
1584 return bs->filename;
1585 else
1586 return NULL;
1587 }
1588
1589 void bdrv_get_backing_filename(BlockDriverState *bs,
1590 char *filename, int filename_size)
1591 {
1592 if (!bs->backing_file) {
1593 pstrcpy(filename, filename_size, "");
1594 } else {
1595 pstrcpy(filename, filename_size, bs->backing_file);
1596 }
1597 }
1598
1599 int bdrv_write_compressed(BlockDriverState *bs, int64_t sector_num,
1600 const uint8_t *buf, int nb_sectors)
1601 {
1602 BlockDriver *drv = bs->drv;
1603 if (!drv)
1604 return -ENOMEDIUM;
1605 if (!drv->bdrv_write_compressed)
1606 return -ENOTSUP;
1607 if (bdrv_check_request(bs, sector_num, nb_sectors))
1608 return -EIO;
1609
1610 if (bs->dirty_bitmap) {
1611 set_dirty_bitmap(bs, sector_num, nb_sectors, 1);
1612 }
1613
1614 return drv->bdrv_write_compressed(bs, sector_num, buf, nb_sectors);
1615 }
1616
1617 int bdrv_get_info(BlockDriverState *bs, BlockDriverInfo *bdi)
1618 {
1619 BlockDriver *drv = bs->drv;
1620 if (!drv)
1621 return -ENOMEDIUM;
1622 if (!drv->bdrv_get_info)
1623 return -ENOTSUP;
1624 memset(bdi, 0, sizeof(*bdi));
1625 return drv->bdrv_get_info(bs, bdi);
1626 }
1627
1628 int bdrv_save_vmstate(BlockDriverState *bs, const uint8_t *buf,
1629 int64_t pos, int size)
1630 {
1631 BlockDriver *drv = bs->drv;
1632 if (!drv)
1633 return -ENOMEDIUM;
1634 if (drv->bdrv_save_vmstate)
1635 return drv->bdrv_save_vmstate(bs, buf, pos, size);
1636 if (bs->file)
1637 return bdrv_save_vmstate(bs->file, buf, pos, size);
1638 return -ENOTSUP;
1639 }
1640
1641 int bdrv_load_vmstate(BlockDriverState *bs, uint8_t *buf,
1642 int64_t pos, int size)
1643 {
1644 BlockDriver *drv = bs->drv;
1645 if (!drv)
1646 return -ENOMEDIUM;
1647 if (drv->bdrv_load_vmstate)
1648 return drv->bdrv_load_vmstate(bs, buf, pos, size);
1649 if (bs->file)
1650 return bdrv_load_vmstate(bs->file, buf, pos, size);
1651 return -ENOTSUP;
1652 }
1653
1654 void bdrv_debug_event(BlockDriverState *bs, BlkDebugEvent event)
1655 {
1656 BlockDriver *drv = bs->drv;
1657
1658 if (!drv || !drv->bdrv_debug_event) {
1659 return;
1660 }
1661
1662 return drv->bdrv_debug_event(bs, event);
1663
1664 }
1665
1666 /**************************************************************/
1667 /* handling of snapshots */
1668
1669 int bdrv_can_snapshot(BlockDriverState *bs)
1670 {
1671 BlockDriver *drv = bs->drv;
1672 if (!drv || bdrv_is_removable(bs) || bdrv_is_read_only(bs)) {
1673 return 0;
1674 }
1675
1676 if (!drv->bdrv_snapshot_create) {
1677 if (bs->file != NULL) {
1678 return bdrv_can_snapshot(bs->file);
1679 }
1680 return 0;
1681 }
1682
1683 return 1;
1684 }
1685
1686 int bdrv_snapshot_create(BlockDriverState *bs,
1687 QEMUSnapshotInfo *sn_info)
1688 {
1689 BlockDriver *drv = bs->drv;
1690 if (!drv)
1691 return -ENOMEDIUM;
1692 if (drv->bdrv_snapshot_create)
1693 return drv->bdrv_snapshot_create(bs, sn_info);
1694 if (bs->file)
1695 return bdrv_snapshot_create(bs->file, sn_info);
1696 return -ENOTSUP;
1697 }
1698
1699 int bdrv_snapshot_goto(BlockDriverState *bs,
1700 const char *snapshot_id)
1701 {
1702 BlockDriver *drv = bs->drv;
1703 int ret, open_ret;
1704
1705 if (!drv)
1706 return -ENOMEDIUM;
1707 if (drv->bdrv_snapshot_goto)
1708 return drv->bdrv_snapshot_goto(bs, snapshot_id);
1709
1710 if (bs->file) {
1711 drv->bdrv_close(bs);
1712 ret = bdrv_snapshot_goto(bs->file, snapshot_id);
1713 open_ret = drv->bdrv_open(bs, bs->open_flags);
1714 if (open_ret < 0) {
1715 bdrv_delete(bs->file);
1716 bs->drv = NULL;
1717 return open_ret;
1718 }
1719 return ret;
1720 }
1721
1722 return -ENOTSUP;
1723 }
1724
1725 int bdrv_snapshot_delete(BlockDriverState *bs, const char *snapshot_id)
1726 {
1727 BlockDriver *drv = bs->drv;
1728 if (!drv)
1729 return -ENOMEDIUM;
1730 if (drv->bdrv_snapshot_delete)
1731 return drv->bdrv_snapshot_delete(bs, snapshot_id);
1732 if (bs->file)
1733 return bdrv_snapshot_delete(bs->file, snapshot_id);
1734 return -ENOTSUP;
1735 }
1736
1737 int bdrv_snapshot_list(BlockDriverState *bs,
1738 QEMUSnapshotInfo **psn_info)
1739 {
1740 BlockDriver *drv = bs->drv;
1741 if (!drv)
1742 return -ENOMEDIUM;
1743 if (drv->bdrv_snapshot_list)
1744 return drv->bdrv_snapshot_list(bs, psn_info);
1745 if (bs->file)
1746 return bdrv_snapshot_list(bs->file, psn_info);
1747 return -ENOTSUP;
1748 }
1749
1750 #define NB_SUFFIXES 4
1751
1752 char *get_human_readable_size(char *buf, int buf_size, int64_t size)
1753 {
1754 static const char suffixes[NB_SUFFIXES] = "KMGT";
1755 int64_t base;
1756 int i;
1757
1758 if (size <= 999) {
1759 snprintf(buf, buf_size, "%" PRId64, size);
1760 } else {
1761 base = 1024;
1762 for(i = 0; i < NB_SUFFIXES; i++) {
1763 if (size < (10 * base)) {
1764 snprintf(buf, buf_size, "%0.1f%c",
1765 (double)size / base,
1766 suffixes[i]);
1767 break;
1768 } else if (size < (1000 * base) || i == (NB_SUFFIXES - 1)) {
1769 snprintf(buf, buf_size, "%" PRId64 "%c",
1770 ((size + (base >> 1)) / base),
1771 suffixes[i]);
1772 break;
1773 }
1774 base = base * 1024;
1775 }
1776 }
1777 return buf;
1778 }
1779
1780 char *bdrv_snapshot_dump(char *buf, int buf_size, QEMUSnapshotInfo *sn)
1781 {
1782 char buf1[128], date_buf[128], clock_buf[128];
1783 #ifdef _WIN32
1784 struct tm *ptm;
1785 #else
1786 struct tm tm;
1787 #endif
1788 time_t ti;
1789 int64_t secs;
1790
1791 if (!sn) {
1792 snprintf(buf, buf_size,
1793 "%-10s%-20s%7s%20s%15s",
1794 "ID", "TAG", "VM SIZE", "DATE", "VM CLOCK");
1795 } else {
1796 ti = sn->date_sec;
1797 #ifdef _WIN32
1798 ptm = localtime(&ti);
1799 strftime(date_buf, sizeof(date_buf),
1800 "%Y-%m-%d %H:%M:%S", ptm);
1801 #else
1802 localtime_r(&ti, &tm);
1803 strftime(date_buf, sizeof(date_buf),
1804 "%Y-%m-%d %H:%M:%S", &tm);
1805 #endif
1806 secs = sn->vm_clock_nsec / 1000000000;
1807 snprintf(clock_buf, sizeof(clock_buf),
1808 "%02d:%02d:%02d.%03d",
1809 (int)(secs / 3600),
1810 (int)((secs / 60) % 60),
1811 (int)(secs % 60),
1812 (int)((sn->vm_clock_nsec / 1000000) % 1000));
1813 snprintf(buf, buf_size,
1814 "%-10s%-20s%7s%20s%15s",
1815 sn->id_str, sn->name,
1816 get_human_readable_size(buf1, sizeof(buf1), sn->vm_state_size),
1817 date_buf,
1818 clock_buf);
1819 }
1820 return buf;
1821 }
1822
1823
1824 /**************************************************************/
1825 /* async I/Os */
1826
1827 BlockDriverAIOCB *bdrv_aio_readv(BlockDriverState *bs, int64_t sector_num,
1828 QEMUIOVector *qiov, int nb_sectors,
1829 BlockDriverCompletionFunc *cb, void *opaque)
1830 {
1831 BlockDriver *drv = bs->drv;
1832 BlockDriverAIOCB *ret;
1833
1834 if (!drv)
1835 return NULL;
1836 if (bdrv_check_request(bs, sector_num, nb_sectors))
1837 return NULL;
1838
1839 ret = drv->bdrv_aio_readv(bs, sector_num, qiov, nb_sectors,
1840 cb, opaque);
1841
1842 if (ret) {
1843 /* Update stats even though technically transfer has not happened. */
1844 bs->rd_bytes += (unsigned) nb_sectors * BDRV_SECTOR_SIZE;
1845 bs->rd_ops ++;
1846 }
1847
1848 return ret;
1849 }
1850
1851 BlockDriverAIOCB *bdrv_aio_writev(BlockDriverState *bs, int64_t sector_num,
1852 QEMUIOVector *qiov, int nb_sectors,
1853 BlockDriverCompletionFunc *cb, void *opaque)
1854 {
1855 BlockDriver *drv = bs->drv;
1856 BlockDriverAIOCB *ret;
1857
1858 if (!drv)
1859 return NULL;
1860 if (bs->read_only)
1861 return NULL;
1862 if (bdrv_check_request(bs, sector_num, nb_sectors))
1863 return NULL;
1864
1865 if (bs->dirty_bitmap) {
1866 set_dirty_bitmap(bs, sector_num, nb_sectors, 1);
1867 }
1868
1869 ret = drv->bdrv_aio_writev(bs, sector_num, qiov, nb_sectors,
1870 cb, opaque);
1871
1872 if (ret) {
1873 /* Update stats even though technically transfer has not happened. */
1874 bs->wr_bytes += (unsigned) nb_sectors * BDRV_SECTOR_SIZE;
1875 bs->wr_ops ++;
1876 if (bs->wr_highest_sector < sector_num + nb_sectors - 1) {
1877 bs->wr_highest_sector = sector_num + nb_sectors - 1;
1878 }
1879 }
1880
1881 return ret;
1882 }
1883
1884
1885 typedef struct MultiwriteCB {
1886 int error;
1887 int num_requests;
1888 int num_callbacks;
1889 struct {
1890 BlockDriverCompletionFunc *cb;
1891 void *opaque;
1892 QEMUIOVector *free_qiov;
1893 void *free_buf;
1894 } callbacks[];
1895 } MultiwriteCB;
1896
1897 static void multiwrite_user_cb(MultiwriteCB *mcb)
1898 {
1899 int i;
1900
1901 for (i = 0; i < mcb->num_callbacks; i++) {
1902 mcb->callbacks[i].cb(mcb->callbacks[i].opaque, mcb->error);
1903 if (mcb->callbacks[i].free_qiov) {
1904 qemu_iovec_destroy(mcb->callbacks[i].free_qiov);
1905 }
1906 qemu_free(mcb->callbacks[i].free_qiov);
1907 qemu_vfree(mcb->callbacks[i].free_buf);
1908 }
1909 }
1910
1911 static void multiwrite_cb(void *opaque, int ret)
1912 {
1913 MultiwriteCB *mcb = opaque;
1914
1915 if (ret < 0 && !mcb->error) {
1916 mcb->error = ret;
1917 multiwrite_user_cb(mcb);
1918 }
1919
1920 mcb->num_requests--;
1921 if (mcb->num_requests == 0) {
1922 if (mcb->error == 0) {
1923 multiwrite_user_cb(mcb);
1924 }
1925 qemu_free(mcb);
1926 }
1927 }
1928
1929 static int multiwrite_req_compare(const void *a, const void *b)
1930 {
1931 const BlockRequest *req1 = a, *req2 = b;
1932
1933 /*
1934 * Note that we can't simply subtract req2->sector from req1->sector
1935 * here as that could overflow the return value.
1936 */
1937 if (req1->sector > req2->sector) {
1938 return 1;
1939 } else if (req1->sector < req2->sector) {
1940 return -1;
1941 } else {
1942 return 0;
1943 }
1944 }
1945
1946 /*
1947 * Takes a bunch of requests and tries to merge them. Returns the number of
1948 * requests that remain after merging.
1949 */
1950 static int multiwrite_merge(BlockDriverState *bs, BlockRequest *reqs,
1951 int num_reqs, MultiwriteCB *mcb)
1952 {
1953 int i, outidx;
1954
1955 // Sort requests by start sector
1956 qsort(reqs, num_reqs, sizeof(*reqs), &multiwrite_req_compare);
1957
1958 // Check if adjacent requests touch the same clusters. If so, combine them,
1959 // filling up gaps with zero sectors.
1960 outidx = 0;
1961 for (i = 1; i < num_reqs; i++) {
1962 int merge = 0;
1963 int64_t oldreq_last = reqs[outidx].sector + reqs[outidx].nb_sectors;
1964
1965 // This handles the cases that are valid for all block drivers, namely
1966 // exactly sequential writes and overlapping writes.
1967 if (reqs[i].sector <= oldreq_last) {
1968 merge = 1;
1969 }
1970
1971 // The block driver may decide that it makes sense to combine requests
1972 // even if there is a gap of some sectors between them. In this case,
1973 // the gap is filled with zeros (therefore only applicable for yet
1974 // unused space in format like qcow2).
1975 if (!merge && bs->drv->bdrv_merge_requests) {
1976 merge = bs->drv->bdrv_merge_requests(bs, &reqs[outidx], &reqs[i]);
1977 }
1978
1979 if (reqs[outidx].qiov->niov + reqs[i].qiov->niov + 1 > IOV_MAX) {
1980 merge = 0;
1981 }
1982
1983 if (merge) {
1984 size_t size;
1985 QEMUIOVector *qiov = qemu_mallocz(sizeof(*qiov));
1986 qemu_iovec_init(qiov,
1987 reqs[outidx].qiov->niov + reqs[i].qiov->niov + 1);
1988
1989 // Add the first request to the merged one. If the requests are
1990 // overlapping, drop the last sectors of the first request.
1991 size = (reqs[i].sector - reqs[outidx].sector) << 9;
1992 qemu_iovec_concat(qiov, reqs[outidx].qiov, size);
1993
1994 // We might need to add some zeros between the two requests
1995 if (reqs[i].sector > oldreq_last) {
1996 size_t zero_bytes = (reqs[i].sector - oldreq_last) << 9;
1997 uint8_t *buf = qemu_blockalign(bs, zero_bytes);
1998 memset(buf, 0, zero_bytes);
1999 qemu_iovec_add(qiov, buf, zero_bytes);
2000 mcb->callbacks[i].free_buf = buf;
2001 }
2002
2003 // Add the second request
2004 qemu_iovec_concat(qiov, reqs[i].qiov, reqs[i].qiov->size);
2005
2006 reqs[outidx].nb_sectors = qiov->size >> 9;
2007 reqs[outidx].qiov = qiov;
2008
2009 mcb->callbacks[i].free_qiov = reqs[outidx].qiov;
2010 } else {
2011 outidx++;
2012 reqs[outidx].sector = reqs[i].sector;
2013 reqs[outidx].nb_sectors = reqs[i].nb_sectors;
2014 reqs[outidx].qiov = reqs[i].qiov;
2015 }
2016 }
2017
2018 return outidx + 1;
2019 }
2020
2021 /*
2022 * Submit multiple AIO write requests at once.
2023 *
2024 * On success, the function returns 0 and all requests in the reqs array have
2025 * been submitted. In error case this function returns -1, and any of the
2026 * requests may or may not be submitted yet. In particular, this means that the
2027 * callback will be called for some of the requests, for others it won't. The
2028 * caller must check the error field of the BlockRequest to wait for the right
2029 * callbacks (if error != 0, no callback will be called).
2030 *
2031 * The implementation may modify the contents of the reqs array, e.g. to merge
2032 * requests. However, the fields opaque and error are left unmodified as they
2033 * are used to signal failure for a single request to the caller.
2034 */
2035 int bdrv_aio_multiwrite(BlockDriverState *bs, BlockRequest *reqs, int num_reqs)
2036 {
2037 BlockDriverAIOCB *acb;
2038 MultiwriteCB *mcb;
2039 int i;
2040
2041 if (num_reqs == 0) {
2042 return 0;
2043 }
2044
2045 // Create MultiwriteCB structure
2046 mcb = qemu_mallocz(sizeof(*mcb) + num_reqs * sizeof(*mcb->callbacks));
2047 mcb->num_requests = 0;
2048 mcb->num_callbacks = num_reqs;
2049
2050 for (i = 0; i < num_reqs; i++) {
2051 mcb->callbacks[i].cb = reqs[i].cb;
2052 mcb->callbacks[i].opaque = reqs[i].opaque;
2053 }
2054
2055 // Check for mergable requests
2056 num_reqs = multiwrite_merge(bs, reqs, num_reqs, mcb);
2057
2058 // Run the aio requests
2059 for (i = 0; i < num_reqs; i++) {
2060 acb = bdrv_aio_writev(bs, reqs[i].sector, reqs[i].qiov,
2061 reqs[i].nb_sectors, multiwrite_cb, mcb);
2062
2063 if (acb == NULL) {
2064 // We can only fail the whole thing if no request has been
2065 // submitted yet. Otherwise we'll wait for the submitted AIOs to
2066 // complete and report the error in the callback.
2067 if (mcb->num_requests == 0) {
2068 reqs[i].error = -EIO;
2069 goto fail;
2070 } else {
2071 mcb->num_requests++;
2072 multiwrite_cb(mcb, -EIO);
2073 break;
2074 }
2075 } else {
2076 mcb->num_requests++;
2077 }
2078 }
2079
2080 return 0;
2081
2082 fail:
2083 qemu_free(mcb);
2084 return -1;
2085 }
2086
2087 BlockDriverAIOCB *bdrv_aio_flush(BlockDriverState *bs,
2088 BlockDriverCompletionFunc *cb, void *opaque)
2089 {
2090 BlockDriver *drv = bs->drv;
2091
2092 if (bs->open_flags & BDRV_O_NO_FLUSH) {
2093 return bdrv_aio_noop_em(bs, cb, opaque);
2094 }
2095
2096 if (!drv)
2097 return NULL;
2098 return drv->bdrv_aio_flush(bs, cb, opaque);
2099 }
2100
2101 void bdrv_aio_cancel(BlockDriverAIOCB *acb)
2102 {
2103 acb->pool->cancel(acb);
2104 }
2105
2106
2107 /**************************************************************/
2108 /* async block device emulation */
2109
2110 typedef struct BlockDriverAIOCBSync {
2111 BlockDriverAIOCB common;
2112 QEMUBH *bh;
2113 int ret;
2114 /* vector translation state */
2115 QEMUIOVector *qiov;
2116 uint8_t *bounce;
2117 int is_write;
2118 } BlockDriverAIOCBSync;
2119
2120 static void bdrv_aio_cancel_em(BlockDriverAIOCB *blockacb)
2121 {
2122 BlockDriverAIOCBSync *acb =
2123 container_of(blockacb, BlockDriverAIOCBSync, common);
2124 qemu_bh_delete(acb->bh);
2125 acb->bh = NULL;
2126 qemu_aio_release(acb);
2127 }
2128
2129 static AIOPool bdrv_em_aio_pool = {
2130 .aiocb_size = sizeof(BlockDriverAIOCBSync),
2131 .cancel = bdrv_aio_cancel_em,
2132 };
2133
2134 static void bdrv_aio_bh_cb(void *opaque)
2135 {
2136 BlockDriverAIOCBSync *acb = opaque;
2137
2138 if (!acb->is_write)
2139 qemu_iovec_from_buffer(acb->qiov, acb->bounce, acb->qiov->size);
2140 qemu_vfree(acb->bounce);
2141 acb->common.cb(acb->common.opaque, acb->ret);
2142 qemu_bh_delete(acb->bh);
2143 acb->bh = NULL;
2144 qemu_aio_release(acb);
2145 }
2146
2147 static BlockDriverAIOCB *bdrv_aio_rw_vector(BlockDriverState *bs,
2148 int64_t sector_num,
2149 QEMUIOVector *qiov,
2150 int nb_sectors,
2151 BlockDriverCompletionFunc *cb,
2152 void *opaque,
2153 int is_write)
2154
2155 {
2156 BlockDriverAIOCBSync *acb;
2157
2158 acb = qemu_aio_get(&bdrv_em_aio_pool, bs, cb, opaque);
2159 acb->is_write = is_write;
2160 acb->qiov = qiov;
2161 acb->bounce = qemu_blockalign(bs, qiov->size);
2162
2163 if (!acb->bh)
2164 acb->bh = qemu_bh_new(bdrv_aio_bh_cb, acb);
2165
2166 if (is_write) {
2167 qemu_iovec_to_buffer(acb->qiov, acb->bounce);
2168 acb->ret = bdrv_write(bs, sector_num, acb->bounce, nb_sectors);
2169 } else {
2170 acb->ret = bdrv_read(bs, sector_num, acb->bounce, nb_sectors);
2171 }
2172
2173 qemu_bh_schedule(acb->bh);
2174
2175 return &acb->common;
2176 }
2177
2178 static BlockDriverAIOCB *bdrv_aio_readv_em(BlockDriverState *bs,
2179 int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
2180 BlockDriverCompletionFunc *cb, void *opaque)
2181 {
2182 return bdrv_aio_rw_vector(bs, sector_num, qiov, nb_sectors, cb, opaque, 0);
2183 }
2184
2185 static BlockDriverAIOCB *bdrv_aio_writev_em(BlockDriverState *bs,
2186 int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
2187 BlockDriverCompletionFunc *cb, void *opaque)
2188 {
2189 return bdrv_aio_rw_vector(bs, sector_num, qiov, nb_sectors, cb, opaque, 1);
2190 }
2191
2192 static BlockDriverAIOCB *bdrv_aio_flush_em(BlockDriverState *bs,
2193 BlockDriverCompletionFunc *cb, void *opaque)
2194 {
2195 BlockDriverAIOCBSync *acb;
2196
2197 acb = qemu_aio_get(&bdrv_em_aio_pool, bs, cb, opaque);
2198 acb->is_write = 1; /* don't bounce in the completion hadler */
2199 acb->qiov = NULL;
2200 acb->bounce = NULL;
2201 acb->ret = 0;
2202
2203 if (!acb->bh)
2204 acb->bh = qemu_bh_new(bdrv_aio_bh_cb, acb);
2205
2206 bdrv_flush(bs);
2207 qemu_bh_schedule(acb->bh);
2208 return &acb->common;
2209 }
2210
2211 static BlockDriverAIOCB *bdrv_aio_noop_em(BlockDriverState *bs,
2212 BlockDriverCompletionFunc *cb, void *opaque)
2213 {
2214 BlockDriverAIOCBSync *acb;
2215
2216 acb = qemu_aio_get(&bdrv_em_aio_pool, bs, cb, opaque);
2217 acb->is_write = 1; /* don't bounce in the completion handler */
2218 acb->qiov = NULL;
2219 acb->bounce = NULL;
2220 acb->ret = 0;
2221
2222 if (!acb->bh) {
2223 acb->bh = qemu_bh_new(bdrv_aio_bh_cb, acb);
2224 }
2225
2226 qemu_bh_schedule(acb->bh);
2227 return &acb->common;
2228 }
2229
2230 /**************************************************************/
2231 /* sync block device emulation */
2232
2233 static void bdrv_rw_em_cb(void *opaque, int ret)
2234 {
2235 *(int *)opaque = ret;
2236 }
2237
2238 #define NOT_DONE 0x7fffffff
2239
2240 static int bdrv_read_em(BlockDriverState *bs, int64_t sector_num,
2241 uint8_t *buf, int nb_sectors)
2242 {
2243 int async_ret;
2244 BlockDriverAIOCB *acb;
2245 struct iovec iov;
2246 QEMUIOVector qiov;
2247
2248 async_context_push();
2249
2250 async_ret = NOT_DONE;
2251 iov.iov_base = (void *)buf;
2252 iov.iov_len = nb_sectors * BDRV_SECTOR_SIZE;
2253 qemu_iovec_init_external(&qiov, &iov, 1);
2254 acb = bdrv_aio_readv(bs, sector_num, &qiov, nb_sectors,
2255 bdrv_rw_em_cb, &async_ret);
2256 if (acb == NULL) {
2257 async_ret = -1;
2258 goto fail;
2259 }
2260
2261 while (async_ret == NOT_DONE) {
2262 qemu_aio_wait();
2263 }
2264
2265
2266 fail:
2267 async_context_pop();
2268 return async_ret;
2269 }
2270
2271 static int bdrv_write_em(BlockDriverState *bs, int64_t sector_num,
2272 const uint8_t *buf, int nb_sectors)
2273 {
2274 int async_ret;
2275 BlockDriverAIOCB *acb;
2276 struct iovec iov;
2277 QEMUIOVector qiov;
2278
2279 async_context_push();
2280
2281 async_ret = NOT_DONE;
2282 iov.iov_base = (void *)buf;
2283 iov.iov_len = nb_sectors * BDRV_SECTOR_SIZE;
2284 qemu_iovec_init_external(&qiov, &iov, 1);
2285 acb = bdrv_aio_writev(bs, sector_num, &qiov, nb_sectors,
2286 bdrv_rw_em_cb, &async_ret);
2287 if (acb == NULL) {
2288 async_ret = -1;
2289 goto fail;
2290 }
2291 while (async_ret == NOT_DONE) {
2292 qemu_aio_wait();
2293 }
2294
2295 fail:
2296 async_context_pop();
2297 return async_ret;
2298 }
2299
2300 void bdrv_init(void)
2301 {
2302 module_call_init(MODULE_INIT_BLOCK);
2303 }
2304
2305 void bdrv_init_with_whitelist(void)
2306 {
2307 use_bdrv_whitelist = 1;
2308 bdrv_init();
2309 }
2310
2311 void *qemu_aio_get(AIOPool *pool, BlockDriverState *bs,
2312 BlockDriverCompletionFunc *cb, void *opaque)
2313 {
2314 BlockDriverAIOCB *acb;
2315
2316 if (pool->free_aiocb) {
2317 acb = pool->free_aiocb;
2318 pool->free_aiocb = acb->next;
2319 } else {
2320 acb = qemu_mallocz(pool->aiocb_size);
2321 acb->pool = pool;
2322 }
2323 acb->bs = bs;
2324 acb->cb = cb;
2325 acb->opaque = opaque;
2326 return acb;
2327 }
2328
2329 void qemu_aio_release(void *p)
2330 {
2331 BlockDriverAIOCB *acb = (BlockDriverAIOCB *)p;
2332 AIOPool *pool = acb->pool;
2333 acb->next = pool->free_aiocb;
2334 pool->free_aiocb = acb;
2335 }
2336
2337 /**************************************************************/
2338 /* removable device support */
2339
2340 /**
2341 * Return TRUE if the media is present
2342 */
2343 int bdrv_is_inserted(BlockDriverState *bs)
2344 {
2345 BlockDriver *drv = bs->drv;
2346 int ret;
2347 if (!drv)
2348 return 0;
2349 if (!drv->bdrv_is_inserted)
2350 return 1;
2351 ret = drv->bdrv_is_inserted(bs);
2352 return ret;
2353 }
2354
2355 /**
2356 * Return TRUE if the media changed since the last call to this
2357 * function. It is currently only used for floppy disks
2358 */
2359 int bdrv_media_changed(BlockDriverState *bs)
2360 {
2361 BlockDriver *drv = bs->drv;
2362 int ret;
2363
2364 if (!drv || !drv->bdrv_media_changed)
2365 ret = -ENOTSUP;
2366 else
2367 ret = drv->bdrv_media_changed(bs);
2368 if (ret == -ENOTSUP)
2369 ret = bs->media_changed;
2370 bs->media_changed = 0;
2371 return ret;
2372 }
2373
2374 /**
2375 * If eject_flag is TRUE, eject the media. Otherwise, close the tray
2376 */
2377 int bdrv_eject(BlockDriverState *bs, int eject_flag)
2378 {
2379 BlockDriver *drv = bs->drv;
2380 int ret;
2381
2382 if (bs->locked) {
2383 return -EBUSY;
2384 }
2385
2386 if (!drv || !drv->bdrv_eject) {
2387 ret = -ENOTSUP;
2388 } else {
2389 ret = drv->bdrv_eject(bs, eject_flag);
2390 }
2391 if (ret == -ENOTSUP) {
2392 if (eject_flag)
2393 bdrv_close(bs);
2394 ret = 0;
2395 }
2396
2397 return ret;
2398 }
2399
2400 int bdrv_is_locked(BlockDriverState *bs)
2401 {
2402 return bs->locked;
2403 }
2404
2405 /**
2406 * Lock or unlock the media (if it is locked, the user won't be able
2407 * to eject it manually).
2408 */
2409 void bdrv_set_locked(BlockDriverState *bs, int locked)
2410 {
2411 BlockDriver *drv = bs->drv;
2412
2413 bs->locked = locked;
2414 if (drv && drv->bdrv_set_locked) {
2415 drv->bdrv_set_locked(bs, locked);
2416 }
2417 }
2418
2419 /* needed for generic scsi interface */
2420
2421 int bdrv_ioctl(BlockDriverState *bs, unsigned long int req, void *buf)
2422 {
2423 BlockDriver *drv = bs->drv;
2424
2425 if (drv && drv->bdrv_ioctl)
2426 return drv->bdrv_ioctl(bs, req, buf);
2427 return -ENOTSUP;
2428 }
2429
2430 BlockDriverAIOCB *bdrv_aio_ioctl(BlockDriverState *bs,
2431 unsigned long int req, void *buf,
2432 BlockDriverCompletionFunc *cb, void *opaque)
2433 {
2434 BlockDriver *drv = bs->drv;
2435
2436 if (drv && drv->bdrv_aio_ioctl)
2437 return drv->bdrv_aio_ioctl(bs, req, buf, cb, opaque);
2438 return NULL;
2439 }
2440
2441
2442
2443 void *qemu_blockalign(BlockDriverState *bs, size_t size)
2444 {
2445 return qemu_memalign((bs && bs->buffer_alignment) ? bs->buffer_alignment : 512, size);
2446 }
2447
2448 void bdrv_set_dirty_tracking(BlockDriverState *bs, int enable)
2449 {
2450 int64_t bitmap_size;
2451
2452 bs->dirty_count = 0;
2453 if (enable) {
2454 if (!bs->dirty_bitmap) {
2455 bitmap_size = (bdrv_getlength(bs) >> BDRV_SECTOR_BITS) +
2456 BDRV_SECTORS_PER_DIRTY_CHUNK * 8 - 1;
2457 bitmap_size /= BDRV_SECTORS_PER_DIRTY_CHUNK * 8;
2458
2459 bs->dirty_bitmap = qemu_mallocz(bitmap_size);
2460 }
2461 } else {
2462 if (bs->dirty_bitmap) {
2463 qemu_free(bs->dirty_bitmap);
2464 bs->dirty_bitmap = NULL;
2465 }
2466 }
2467 }
2468
2469 int bdrv_get_dirty(BlockDriverState *bs, int64_t sector)
2470 {
2471 int64_t chunk = sector / (int64_t)BDRV_SECTORS_PER_DIRTY_CHUNK;
2472
2473 if (bs->dirty_bitmap &&
2474 (sector << BDRV_SECTOR_BITS) < bdrv_getlength(bs)) {
2475 return bs->dirty_bitmap[chunk / (sizeof(unsigned long) * 8)] &
2476 (1 << (chunk % (sizeof(unsigned long) * 8)));
2477 } else {
2478 return 0;
2479 }
2480 }
2481
2482 void bdrv_reset_dirty(BlockDriverState *bs, int64_t cur_sector,
2483 int nr_sectors)
2484 {
2485 set_dirty_bitmap(bs, cur_sector, nr_sectors, 0);
2486 }
2487
2488 int64_t bdrv_get_dirty_count(BlockDriverState *bs)
2489 {
2490 return bs->dirty_count;
2491 }
QEMU mirror