]> git.proxmox.com Git - qemu.git/blob - block.c
Merge remote branch 'kwolf/for-anthony' into staging
[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 int bdrv_read_em(BlockDriverState *bs, int64_t sector_num,
54 uint8_t *buf, int nb_sectors);
55 static int bdrv_write_em(BlockDriverState *bs, int64_t sector_num,
56 const uint8_t *buf, int nb_sectors);
57 static BlockDriver *find_protocol(const char *filename);
58
59 static QTAILQ_HEAD(, BlockDriverState) bdrv_states =
60 QTAILQ_HEAD_INITIALIZER(bdrv_states);
61
62 static QLIST_HEAD(, BlockDriver) bdrv_drivers =
63 QLIST_HEAD_INITIALIZER(bdrv_drivers);
64
65 /* If non-zero, use only whitelisted block drivers */
66 static int use_bdrv_whitelist;
67
68 int path_is_absolute(const char *path)
69 {
70 const char *p;
71 #ifdef _WIN32
72 /* specific case for names like: "\\.\d:" */
73 if (*path == '/' || *path == '\\')
74 return 1;
75 #endif
76 p = strchr(path, ':');
77 if (p)
78 p++;
79 else
80 p = path;
81 #ifdef _WIN32
82 return (*p == '/' || *p == '\\');
83 #else
84 return (*p == '/');
85 #endif
86 }
87
88 /* if filename is absolute, just copy it to dest. Otherwise, build a
89 path to it by considering it is relative to base_path. URL are
90 supported. */
91 void path_combine(char *dest, int dest_size,
92 const char *base_path,
93 const char *filename)
94 {
95 const char *p, *p1;
96 int len;
97
98 if (dest_size <= 0)
99 return;
100 if (path_is_absolute(filename)) {
101 pstrcpy(dest, dest_size, filename);
102 } else {
103 p = strchr(base_path, ':');
104 if (p)
105 p++;
106 else
107 p = base_path;
108 p1 = strrchr(base_path, '/');
109 #ifdef _WIN32
110 {
111 const char *p2;
112 p2 = strrchr(base_path, '\\');
113 if (!p1 || p2 > p1)
114 p1 = p2;
115 }
116 #endif
117 if (p1)
118 p1++;
119 else
120 p1 = base_path;
121 if (p1 > p)
122 p = p1;
123 len = p - base_path;
124 if (len > dest_size - 1)
125 len = dest_size - 1;
126 memcpy(dest, base_path, len);
127 dest[len] = '\0';
128 pstrcat(dest, dest_size, filename);
129 }
130 }
131
132 void bdrv_register(BlockDriver *bdrv)
133 {
134 if (!bdrv->bdrv_aio_readv) {
135 /* add AIO emulation layer */
136 bdrv->bdrv_aio_readv = bdrv_aio_readv_em;
137 bdrv->bdrv_aio_writev = bdrv_aio_writev_em;
138 } else if (!bdrv->bdrv_read) {
139 /* add synchronous IO emulation layer */
140 bdrv->bdrv_read = bdrv_read_em;
141 bdrv->bdrv_write = bdrv_write_em;
142 }
143
144 if (!bdrv->bdrv_aio_flush)
145 bdrv->bdrv_aio_flush = bdrv_aio_flush_em;
146
147 QLIST_INSERT_HEAD(&bdrv_drivers, bdrv, list);
148 }
149
150 /* create a new block device (by default it is empty) */
151 BlockDriverState *bdrv_new(const char *device_name)
152 {
153 BlockDriverState *bs;
154
155 bs = qemu_mallocz(sizeof(BlockDriverState));
156 pstrcpy(bs->device_name, sizeof(bs->device_name), device_name);
157 if (device_name[0] != '\0') {
158 QTAILQ_INSERT_TAIL(&bdrv_states, bs, list);
159 }
160 return bs;
161 }
162
163 BlockDriver *bdrv_find_format(const char *format_name)
164 {
165 BlockDriver *drv1;
166 QLIST_FOREACH(drv1, &bdrv_drivers, list) {
167 if (!strcmp(drv1->format_name, format_name)) {
168 return drv1;
169 }
170 }
171 return NULL;
172 }
173
174 static int bdrv_is_whitelisted(BlockDriver *drv)
175 {
176 static const char *whitelist[] = {
177 CONFIG_BDRV_WHITELIST
178 };
179 const char **p;
180
181 if (!whitelist[0])
182 return 1; /* no whitelist, anything goes */
183
184 for (p = whitelist; *p; p++) {
185 if (!strcmp(drv->format_name, *p)) {
186 return 1;
187 }
188 }
189 return 0;
190 }
191
192 BlockDriver *bdrv_find_whitelisted_format(const char *format_name)
193 {
194 BlockDriver *drv = bdrv_find_format(format_name);
195 return drv && bdrv_is_whitelisted(drv) ? drv : NULL;
196 }
197
198 int bdrv_create(BlockDriver *drv, const char* filename,
199 QEMUOptionParameter *options)
200 {
201 if (!drv->bdrv_create)
202 return -ENOTSUP;
203
204 return drv->bdrv_create(filename, options);
205 }
206
207 int bdrv_create_file(const char* filename, QEMUOptionParameter *options)
208 {
209 BlockDriver *drv;
210
211 drv = find_protocol(filename);
212 if (drv == NULL) {
213 drv = bdrv_find_format("file");
214 }
215
216 return bdrv_create(drv, filename, options);
217 }
218
219 #ifdef _WIN32
220 void get_tmp_filename(char *filename, int size)
221 {
222 char temp_dir[MAX_PATH];
223
224 GetTempPath(MAX_PATH, temp_dir);
225 GetTempFileName(temp_dir, "qem", 0, filename);
226 }
227 #else
228 void get_tmp_filename(char *filename, int size)
229 {
230 int fd;
231 const char *tmpdir;
232 /* XXX: race condition possible */
233 tmpdir = getenv("TMPDIR");
234 if (!tmpdir)
235 tmpdir = "/tmp";
236 snprintf(filename, size, "%s/vl.XXXXXX", tmpdir);
237 fd = mkstemp(filename);
238 close(fd);
239 }
240 #endif
241
242 #ifdef _WIN32
243 static int is_windows_drive_prefix(const char *filename)
244 {
245 return (((filename[0] >= 'a' && filename[0] <= 'z') ||
246 (filename[0] >= 'A' && filename[0] <= 'Z')) &&
247 filename[1] == ':');
248 }
249
250 int is_windows_drive(const char *filename)
251 {
252 if (is_windows_drive_prefix(filename) &&
253 filename[2] == '\0')
254 return 1;
255 if (strstart(filename, "\\\\.\\", NULL) ||
256 strstart(filename, "//./", NULL))
257 return 1;
258 return 0;
259 }
260 #endif
261
262 /*
263 * Detect host devices. By convention, /dev/cdrom[N] is always
264 * recognized as a host CDROM.
265 */
266 static BlockDriver *find_hdev_driver(const char *filename)
267 {
268 int score_max = 0, score;
269 BlockDriver *drv = NULL, *d;
270
271 QLIST_FOREACH(d, &bdrv_drivers, list) {
272 if (d->bdrv_probe_device) {
273 score = d->bdrv_probe_device(filename);
274 if (score > score_max) {
275 score_max = score;
276 drv = d;
277 }
278 }
279 }
280
281 return drv;
282 }
283
284 static BlockDriver *find_protocol(const char *filename)
285 {
286 BlockDriver *drv1;
287 char protocol[128];
288 int len;
289 const char *p;
290 int is_drive;
291
292 /* TODO Drivers without bdrv_file_open must be specified explicitly */
293
294 #ifdef _WIN32
295 is_drive = is_windows_drive(filename) ||
296 is_windows_drive_prefix(filename);
297 #else
298 is_drive = 0;
299 #endif
300 p = strchr(filename, ':');
301 if (!p || is_drive) {
302 drv1 = find_hdev_driver(filename);
303 if (!drv1) {
304 drv1 = bdrv_find_format("file");
305 }
306 return drv1;
307 }
308 len = p - filename;
309 if (len > sizeof(protocol) - 1)
310 len = sizeof(protocol) - 1;
311 memcpy(protocol, filename, len);
312 protocol[len] = '\0';
313 QLIST_FOREACH(drv1, &bdrv_drivers, list) {
314 if (drv1->protocol_name &&
315 !strcmp(drv1->protocol_name, protocol)) {
316 return drv1;
317 }
318 }
319 return NULL;
320 }
321
322 static BlockDriver *find_image_format(const char *filename)
323 {
324 int ret, score, score_max;
325 BlockDriver *drv1, *drv;
326 uint8_t buf[2048];
327 BlockDriverState *bs;
328
329 ret = bdrv_file_open(&bs, filename, 0);
330 if (ret < 0)
331 return NULL;
332
333 /* Return the raw BlockDriver * to scsi-generic devices */
334 if (bs->sg)
335 return bdrv_find_format("raw");
336
337 ret = bdrv_pread(bs, 0, buf, sizeof(buf));
338 bdrv_delete(bs);
339 if (ret < 0) {
340 return NULL;
341 }
342
343 score_max = 0;
344 drv = NULL;
345 QLIST_FOREACH(drv1, &bdrv_drivers, list) {
346 if (drv1->bdrv_probe) {
347 score = drv1->bdrv_probe(buf, ret, filename);
348 if (score > score_max) {
349 score_max = score;
350 drv = drv1;
351 }
352 }
353 }
354 return drv;
355 }
356
357 /**
358 * Set the current 'total_sectors' value
359 */
360 static int refresh_total_sectors(BlockDriverState *bs, int64_t hint)
361 {
362 BlockDriver *drv = bs->drv;
363
364 /* Do not attempt drv->bdrv_getlength() on scsi-generic devices */
365 if (bs->sg)
366 return 0;
367
368 /* query actual device if possible, otherwise just trust the hint */
369 if (drv->bdrv_getlength) {
370 int64_t length = drv->bdrv_getlength(bs);
371 if (length < 0) {
372 return length;
373 }
374 hint = length >> BDRV_SECTOR_BITS;
375 }
376
377 bs->total_sectors = hint;
378 return 0;
379 }
380
381 /*
382 * Common part for opening disk images and files
383 */
384 static int bdrv_open_common(BlockDriverState *bs, const char *filename,
385 int flags, BlockDriver *drv)
386 {
387 int ret, open_flags;
388
389 assert(drv != NULL);
390
391 bs->file = NULL;
392 bs->total_sectors = 0;
393 bs->is_temporary = 0;
394 bs->encrypted = 0;
395 bs->valid_key = 0;
396 bs->open_flags = flags;
397 /* buffer_alignment defaulted to 512, drivers can change this value */
398 bs->buffer_alignment = 512;
399
400 pstrcpy(bs->filename, sizeof(bs->filename), filename);
401
402 if (use_bdrv_whitelist && !bdrv_is_whitelisted(drv)) {
403 return -ENOTSUP;
404 }
405
406 bs->drv = drv;
407 bs->opaque = qemu_mallocz(drv->instance_size);
408
409 /*
410 * Yes, BDRV_O_NOCACHE aka O_DIRECT means we have to present a
411 * write cache to the guest. We do need the fdatasync to flush
412 * out transactions for block allocations, and we maybe have a
413 * volatile write cache in our backing device to deal with.
414 */
415 if (flags & (BDRV_O_CACHE_WB|BDRV_O_NOCACHE))
416 bs->enable_write_cache = 1;
417
418 /*
419 * Clear flags that are internal to the block layer before opening the
420 * image.
421 */
422 open_flags = flags & ~(BDRV_O_SNAPSHOT | BDRV_O_NO_BACKING);
423
424 /*
425 * Snapshots should be writeable.
426 */
427 if (bs->is_temporary) {
428 open_flags |= BDRV_O_RDWR;
429 }
430
431 /* Open the image, either directly or using a protocol */
432 if (drv->bdrv_file_open) {
433 ret = drv->bdrv_file_open(bs, filename, open_flags);
434 } else {
435 ret = bdrv_file_open(&bs->file, filename, open_flags);
436 if (ret >= 0) {
437 ret = drv->bdrv_open(bs, open_flags);
438 }
439 }
440
441 if (ret < 0) {
442 goto free_and_fail;
443 }
444
445 bs->keep_read_only = bs->read_only = !(open_flags & BDRV_O_RDWR);
446
447 ret = refresh_total_sectors(bs, bs->total_sectors);
448 if (ret < 0) {
449 goto free_and_fail;
450 }
451
452 #ifndef _WIN32
453 if (bs->is_temporary) {
454 unlink(filename);
455 }
456 #endif
457 return 0;
458
459 free_and_fail:
460 if (bs->file) {
461 bdrv_delete(bs->file);
462 bs->file = NULL;
463 }
464 qemu_free(bs->opaque);
465 bs->opaque = NULL;
466 bs->drv = NULL;
467 return ret;
468 }
469
470 /*
471 * Opens a file using a protocol (file, host_device, nbd, ...)
472 */
473 int bdrv_file_open(BlockDriverState **pbs, const char *filename, int flags)
474 {
475 BlockDriverState *bs;
476 BlockDriver *drv;
477 int ret;
478
479 drv = find_protocol(filename);
480 if (!drv) {
481 return -ENOENT;
482 }
483
484 bs = bdrv_new("");
485 ret = bdrv_open_common(bs, filename, flags, drv);
486 if (ret < 0) {
487 bdrv_delete(bs);
488 return ret;
489 }
490 bs->growable = 1;
491 *pbs = bs;
492 return 0;
493 }
494
495 /*
496 * Opens a disk image (raw, qcow2, vmdk, ...)
497 */
498 int bdrv_open(BlockDriverState *bs, const char *filename, int flags,
499 BlockDriver *drv)
500 {
501 int ret;
502
503 if (flags & BDRV_O_SNAPSHOT) {
504 BlockDriverState *bs1;
505 int64_t total_size;
506 int is_protocol = 0;
507 BlockDriver *bdrv_qcow2;
508 QEMUOptionParameter *options;
509 char tmp_filename[PATH_MAX];
510 char backing_filename[PATH_MAX];
511
512 /* if snapshot, we create a temporary backing file and open it
513 instead of opening 'filename' directly */
514
515 /* if there is a backing file, use it */
516 bs1 = bdrv_new("");
517 ret = bdrv_open(bs1, filename, 0, drv);
518 if (ret < 0) {
519 bdrv_delete(bs1);
520 return ret;
521 }
522 total_size = bdrv_getlength(bs1) >> BDRV_SECTOR_BITS;
523
524 if (bs1->drv && bs1->drv->protocol_name)
525 is_protocol = 1;
526
527 bdrv_delete(bs1);
528
529 get_tmp_filename(tmp_filename, sizeof(tmp_filename));
530
531 /* Real path is meaningless for protocols */
532 if (is_protocol)
533 snprintf(backing_filename, sizeof(backing_filename),
534 "%s", filename);
535 else if (!realpath(filename, backing_filename))
536 return -errno;
537
538 bdrv_qcow2 = bdrv_find_format("qcow2");
539 options = parse_option_parameters("", bdrv_qcow2->create_options, NULL);
540
541 set_option_parameter_int(options, BLOCK_OPT_SIZE, total_size * 512);
542 set_option_parameter(options, BLOCK_OPT_BACKING_FILE, backing_filename);
543 if (drv) {
544 set_option_parameter(options, BLOCK_OPT_BACKING_FMT,
545 drv->format_name);
546 }
547
548 ret = bdrv_create(bdrv_qcow2, tmp_filename, options);
549 free_option_parameters(options);
550 if (ret < 0) {
551 return ret;
552 }
553
554 filename = tmp_filename;
555 drv = bdrv_qcow2;
556 bs->is_temporary = 1;
557 }
558
559 /* Find the right image format driver */
560 if (!drv) {
561 drv = find_image_format(filename);
562 }
563
564 if (!drv) {
565 ret = -ENOENT;
566 goto unlink_and_fail;
567 }
568
569 /* Open the image */
570 ret = bdrv_open_common(bs, filename, flags, drv);
571 if (ret < 0) {
572 goto unlink_and_fail;
573 }
574
575 /* If there is a backing file, use it */
576 if ((flags & BDRV_O_NO_BACKING) == 0 && bs->backing_file[0] != '\0') {
577 char backing_filename[PATH_MAX];
578 int back_flags;
579 BlockDriver *back_drv = NULL;
580
581 bs->backing_hd = bdrv_new("");
582 path_combine(backing_filename, sizeof(backing_filename),
583 filename, bs->backing_file);
584 if (bs->backing_format[0] != '\0')
585 back_drv = bdrv_find_format(bs->backing_format);
586
587 /* backing files always opened read-only */
588 back_flags =
589 flags & ~(BDRV_O_RDWR | BDRV_O_SNAPSHOT | BDRV_O_NO_BACKING);
590
591 ret = bdrv_open(bs->backing_hd, backing_filename, back_flags, back_drv);
592 if (ret < 0) {
593 bdrv_close(bs);
594 return ret;
595 }
596 if (bs->is_temporary) {
597 bs->backing_hd->keep_read_only = !(flags & BDRV_O_RDWR);
598 } else {
599 /* base image inherits from "parent" */
600 bs->backing_hd->keep_read_only = bs->keep_read_only;
601 }
602 }
603
604 if (!bdrv_key_required(bs)) {
605 /* call the change callback */
606 bs->media_changed = 1;
607 if (bs->change_cb)
608 bs->change_cb(bs->change_opaque);
609 }
610
611 return 0;
612
613 unlink_and_fail:
614 if (bs->is_temporary) {
615 unlink(filename);
616 }
617 return ret;
618 }
619
620 void bdrv_close(BlockDriverState *bs)
621 {
622 if (bs->drv) {
623 if (bs->backing_hd) {
624 bdrv_delete(bs->backing_hd);
625 bs->backing_hd = NULL;
626 }
627 bs->drv->bdrv_close(bs);
628 qemu_free(bs->opaque);
629 #ifdef _WIN32
630 if (bs->is_temporary) {
631 unlink(bs->filename);
632 }
633 #endif
634 bs->opaque = NULL;
635 bs->drv = NULL;
636
637 if (bs->file != NULL) {
638 bdrv_close(bs->file);
639 }
640
641 /* call the change callback */
642 bs->media_changed = 1;
643 if (bs->change_cb)
644 bs->change_cb(bs->change_opaque);
645 }
646 }
647
648 void bdrv_delete(BlockDriverState *bs)
649 {
650 /* remove from list, if necessary */
651 if (bs->device_name[0] != '\0') {
652 QTAILQ_REMOVE(&bdrv_states, bs, list);
653 }
654
655 bdrv_close(bs);
656 if (bs->file != NULL) {
657 bdrv_delete(bs->file);
658 }
659
660 qemu_free(bs);
661 }
662
663 /*
664 * Run consistency checks on an image
665 *
666 * Returns the number of errors or -errno when an internal error occurs
667 */
668 int bdrv_check(BlockDriverState *bs)
669 {
670 if (bs->drv->bdrv_check == NULL) {
671 return -ENOTSUP;
672 }
673
674 return bs->drv->bdrv_check(bs);
675 }
676
677 /* commit COW file into the raw image */
678 int bdrv_commit(BlockDriverState *bs)
679 {
680 BlockDriver *drv = bs->drv;
681 int64_t i, total_sectors;
682 int n, j, ro, open_flags;
683 int ret = 0, rw_ret = 0;
684 unsigned char sector[512];
685 char filename[1024];
686 BlockDriverState *bs_rw, *bs_ro;
687
688 if (!drv)
689 return -ENOMEDIUM;
690
691 if (!bs->backing_hd) {
692 return -ENOTSUP;
693 }
694
695 if (bs->backing_hd->keep_read_only) {
696 return -EACCES;
697 }
698
699 ro = bs->backing_hd->read_only;
700 strncpy(filename, bs->backing_hd->filename, sizeof(filename));
701 open_flags = bs->backing_hd->open_flags;
702
703 if (ro) {
704 /* re-open as RW */
705 bdrv_delete(bs->backing_hd);
706 bs->backing_hd = NULL;
707 bs_rw = bdrv_new("");
708 rw_ret = bdrv_open(bs_rw, filename, open_flags | BDRV_O_RDWR, drv);
709 if (rw_ret < 0) {
710 bdrv_delete(bs_rw);
711 /* try to re-open read-only */
712 bs_ro = bdrv_new("");
713 ret = bdrv_open(bs_ro, filename, open_flags & ~BDRV_O_RDWR, drv);
714 if (ret < 0) {
715 bdrv_delete(bs_ro);
716 /* drive not functional anymore */
717 bs->drv = NULL;
718 return ret;
719 }
720 bs->backing_hd = bs_ro;
721 return rw_ret;
722 }
723 bs->backing_hd = bs_rw;
724 }
725
726 total_sectors = bdrv_getlength(bs) >> BDRV_SECTOR_BITS;
727 for (i = 0; i < total_sectors;) {
728 if (drv->bdrv_is_allocated(bs, i, 65536, &n)) {
729 for(j = 0; j < n; j++) {
730 if (bdrv_read(bs, i, sector, 1) != 0) {
731 ret = -EIO;
732 goto ro_cleanup;
733 }
734
735 if (bdrv_write(bs->backing_hd, i, sector, 1) != 0) {
736 ret = -EIO;
737 goto ro_cleanup;
738 }
739 i++;
740 }
741 } else {
742 i += n;
743 }
744 }
745
746 if (drv->bdrv_make_empty) {
747 ret = drv->bdrv_make_empty(bs);
748 bdrv_flush(bs);
749 }
750
751 /*
752 * Make sure all data we wrote to the backing device is actually
753 * stable on disk.
754 */
755 if (bs->backing_hd)
756 bdrv_flush(bs->backing_hd);
757
758 ro_cleanup:
759
760 if (ro) {
761 /* re-open as RO */
762 bdrv_delete(bs->backing_hd);
763 bs->backing_hd = NULL;
764 bs_ro = bdrv_new("");
765 ret = bdrv_open(bs_ro, filename, open_flags & ~BDRV_O_RDWR, drv);
766 if (ret < 0) {
767 bdrv_delete(bs_ro);
768 /* drive not functional anymore */
769 bs->drv = NULL;
770 return ret;
771 }
772 bs->backing_hd = bs_ro;
773 bs->backing_hd->keep_read_only = 0;
774 }
775
776 return ret;
777 }
778
779 /*
780 * Return values:
781 * 0 - success
782 * -EINVAL - backing format specified, but no file
783 * -ENOSPC - can't update the backing file because no space is left in the
784 * image file header
785 * -ENOTSUP - format driver doesn't support changing the backing file
786 */
787 int bdrv_change_backing_file(BlockDriverState *bs,
788 const char *backing_file, const char *backing_fmt)
789 {
790 BlockDriver *drv = bs->drv;
791
792 if (drv->bdrv_change_backing_file != NULL) {
793 return drv->bdrv_change_backing_file(bs, backing_file, backing_fmt);
794 } else {
795 return -ENOTSUP;
796 }
797 }
798
799 static int bdrv_check_byte_request(BlockDriverState *bs, int64_t offset,
800 size_t size)
801 {
802 int64_t len;
803
804 if (!bdrv_is_inserted(bs))
805 return -ENOMEDIUM;
806
807 if (bs->growable)
808 return 0;
809
810 len = bdrv_getlength(bs);
811
812 if (offset < 0)
813 return -EIO;
814
815 if ((offset > len) || (len - offset < size))
816 return -EIO;
817
818 return 0;
819 }
820
821 static int bdrv_check_request(BlockDriverState *bs, int64_t sector_num,
822 int nb_sectors)
823 {
824 return bdrv_check_byte_request(bs, sector_num * 512, nb_sectors * 512);
825 }
826
827 /* return < 0 if error. See bdrv_write() for the return codes */
828 int bdrv_read(BlockDriverState *bs, int64_t sector_num,
829 uint8_t *buf, int nb_sectors)
830 {
831 BlockDriver *drv = bs->drv;
832
833 if (!drv)
834 return -ENOMEDIUM;
835 if (bdrv_check_request(bs, sector_num, nb_sectors))
836 return -EIO;
837
838 return drv->bdrv_read(bs, sector_num, buf, nb_sectors);
839 }
840
841 static void set_dirty_bitmap(BlockDriverState *bs, int64_t sector_num,
842 int nb_sectors, int dirty)
843 {
844 int64_t start, end;
845 unsigned long val, idx, bit;
846
847 start = sector_num / BDRV_SECTORS_PER_DIRTY_CHUNK;
848 end = (sector_num + nb_sectors - 1) / BDRV_SECTORS_PER_DIRTY_CHUNK;
849
850 for (; start <= end; start++) {
851 idx = start / (sizeof(unsigned long) * 8);
852 bit = start % (sizeof(unsigned long) * 8);
853 val = bs->dirty_bitmap[idx];
854 if (dirty) {
855 if (!(val & (1 << bit))) {
856 bs->dirty_count++;
857 val |= 1 << bit;
858 }
859 } else {
860 if (val & (1 << bit)) {
861 bs->dirty_count--;
862 val &= ~(1 << bit);
863 }
864 }
865 bs->dirty_bitmap[idx] = val;
866 }
867 }
868
869 /* Return < 0 if error. Important errors are:
870 -EIO generic I/O error (may happen for all errors)
871 -ENOMEDIUM No media inserted.
872 -EINVAL Invalid sector number or nb_sectors
873 -EACCES Trying to write a read-only device
874 */
875 int bdrv_write(BlockDriverState *bs, int64_t sector_num,
876 const uint8_t *buf, int nb_sectors)
877 {
878 BlockDriver *drv = bs->drv;
879 if (!bs->drv)
880 return -ENOMEDIUM;
881 if (bs->read_only)
882 return -EACCES;
883 if (bdrv_check_request(bs, sector_num, nb_sectors))
884 return -EIO;
885
886 if (bs->dirty_bitmap) {
887 set_dirty_bitmap(bs, sector_num, nb_sectors, 1);
888 }
889
890 if (bs->wr_highest_sector < sector_num + nb_sectors - 1) {
891 bs->wr_highest_sector = sector_num + nb_sectors - 1;
892 }
893
894 return drv->bdrv_write(bs, sector_num, buf, nb_sectors);
895 }
896
897 int bdrv_pread(BlockDriverState *bs, int64_t offset,
898 void *buf, int count1)
899 {
900 uint8_t tmp_buf[BDRV_SECTOR_SIZE];
901 int len, nb_sectors, count;
902 int64_t sector_num;
903 int ret;
904
905 count = count1;
906 /* first read to align to sector start */
907 len = (BDRV_SECTOR_SIZE - offset) & (BDRV_SECTOR_SIZE - 1);
908 if (len > count)
909 len = count;
910 sector_num = offset >> BDRV_SECTOR_BITS;
911 if (len > 0) {
912 if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0)
913 return ret;
914 memcpy(buf, tmp_buf + (offset & (BDRV_SECTOR_SIZE - 1)), len);
915 count -= len;
916 if (count == 0)
917 return count1;
918 sector_num++;
919 buf += len;
920 }
921
922 /* read the sectors "in place" */
923 nb_sectors = count >> BDRV_SECTOR_BITS;
924 if (nb_sectors > 0) {
925 if ((ret = bdrv_read(bs, sector_num, buf, nb_sectors)) < 0)
926 return ret;
927 sector_num += nb_sectors;
928 len = nb_sectors << BDRV_SECTOR_BITS;
929 buf += len;
930 count -= len;
931 }
932
933 /* add data from the last sector */
934 if (count > 0) {
935 if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0)
936 return ret;
937 memcpy(buf, tmp_buf, count);
938 }
939 return count1;
940 }
941
942 int bdrv_pwrite(BlockDriverState *bs, int64_t offset,
943 const void *buf, int count1)
944 {
945 uint8_t tmp_buf[BDRV_SECTOR_SIZE];
946 int len, nb_sectors, count;
947 int64_t sector_num;
948 int ret;
949
950 count = count1;
951 /* first write to align to sector start */
952 len = (BDRV_SECTOR_SIZE - offset) & (BDRV_SECTOR_SIZE - 1);
953 if (len > count)
954 len = count;
955 sector_num = offset >> BDRV_SECTOR_BITS;
956 if (len > 0) {
957 if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0)
958 return ret;
959 memcpy(tmp_buf + (offset & (BDRV_SECTOR_SIZE - 1)), buf, len);
960 if ((ret = bdrv_write(bs, sector_num, tmp_buf, 1)) < 0)
961 return ret;
962 count -= len;
963 if (count == 0)
964 return count1;
965 sector_num++;
966 buf += len;
967 }
968
969 /* write the sectors "in place" */
970 nb_sectors = count >> BDRV_SECTOR_BITS;
971 if (nb_sectors > 0) {
972 if ((ret = bdrv_write(bs, sector_num, buf, nb_sectors)) < 0)
973 return ret;
974 sector_num += nb_sectors;
975 len = nb_sectors << BDRV_SECTOR_BITS;
976 buf += len;
977 count -= len;
978 }
979
980 /* add data from the last sector */
981 if (count > 0) {
982 if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0)
983 return ret;
984 memcpy(tmp_buf, buf, count);
985 if ((ret = bdrv_write(bs, sector_num, tmp_buf, 1)) < 0)
986 return ret;
987 }
988 return count1;
989 }
990
991 /**
992 * Truncate file to 'offset' bytes (needed only for file protocols)
993 */
994 int bdrv_truncate(BlockDriverState *bs, int64_t offset)
995 {
996 BlockDriver *drv = bs->drv;
997 int ret;
998 if (!drv)
999 return -ENOMEDIUM;
1000 if (!drv->bdrv_truncate)
1001 return -ENOTSUP;
1002 if (bs->read_only)
1003 return -EACCES;
1004 ret = drv->bdrv_truncate(bs, offset);
1005 if (ret == 0) {
1006 ret = refresh_total_sectors(bs, offset >> BDRV_SECTOR_BITS);
1007 }
1008 return ret;
1009 }
1010
1011 /**
1012 * Length of a file in bytes. Return < 0 if error or unknown.
1013 */
1014 int64_t bdrv_getlength(BlockDriverState *bs)
1015 {
1016 BlockDriver *drv = bs->drv;
1017 if (!drv)
1018 return -ENOMEDIUM;
1019
1020 /* Fixed size devices use the total_sectors value for speed instead of
1021 issuing a length query (like lseek) on each call. Also, legacy block
1022 drivers don't provide a bdrv_getlength function and must use
1023 total_sectors. */
1024 if (!bs->growable || !drv->bdrv_getlength) {
1025 return bs->total_sectors * BDRV_SECTOR_SIZE;
1026 }
1027 return drv->bdrv_getlength(bs);
1028 }
1029
1030 /* return 0 as number of sectors if no device present or error */
1031 void bdrv_get_geometry(BlockDriverState *bs, uint64_t *nb_sectors_ptr)
1032 {
1033 int64_t length;
1034 length = bdrv_getlength(bs);
1035 if (length < 0)
1036 length = 0;
1037 else
1038 length = length >> BDRV_SECTOR_BITS;
1039 *nb_sectors_ptr = length;
1040 }
1041
1042 struct partition {
1043 uint8_t boot_ind; /* 0x80 - active */
1044 uint8_t head; /* starting head */
1045 uint8_t sector; /* starting sector */
1046 uint8_t cyl; /* starting cylinder */
1047 uint8_t sys_ind; /* What partition type */
1048 uint8_t end_head; /* end head */
1049 uint8_t end_sector; /* end sector */
1050 uint8_t end_cyl; /* end cylinder */
1051 uint32_t start_sect; /* starting sector counting from 0 */
1052 uint32_t nr_sects; /* nr of sectors in partition */
1053 } __attribute__((packed));
1054
1055 /* try to guess the disk logical geometry from the MSDOS partition table. Return 0 if OK, -1 if could not guess */
1056 static int guess_disk_lchs(BlockDriverState *bs,
1057 int *pcylinders, int *pheads, int *psectors)
1058 {
1059 uint8_t buf[512];
1060 int ret, i, heads, sectors, cylinders;
1061 struct partition *p;
1062 uint32_t nr_sects;
1063 uint64_t nb_sectors;
1064
1065 bdrv_get_geometry(bs, &nb_sectors);
1066
1067 ret = bdrv_read(bs, 0, buf, 1);
1068 if (ret < 0)
1069 return -1;
1070 /* test msdos magic */
1071 if (buf[510] != 0x55 || buf[511] != 0xaa)
1072 return -1;
1073 for(i = 0; i < 4; i++) {
1074 p = ((struct partition *)(buf + 0x1be)) + i;
1075 nr_sects = le32_to_cpu(p->nr_sects);
1076 if (nr_sects && p->end_head) {
1077 /* We make the assumption that the partition terminates on
1078 a cylinder boundary */
1079 heads = p->end_head + 1;
1080 sectors = p->end_sector & 63;
1081 if (sectors == 0)
1082 continue;
1083 cylinders = nb_sectors / (heads * sectors);
1084 if (cylinders < 1 || cylinders > 16383)
1085 continue;
1086 *pheads = heads;
1087 *psectors = sectors;
1088 *pcylinders = cylinders;
1089 #if 0
1090 printf("guessed geometry: LCHS=%d %d %d\n",
1091 cylinders, heads, sectors);
1092 #endif
1093 return 0;
1094 }
1095 }
1096 return -1;
1097 }
1098
1099 void bdrv_guess_geometry(BlockDriverState *bs, int *pcyls, int *pheads, int *psecs)
1100 {
1101 int translation, lba_detected = 0;
1102 int cylinders, heads, secs;
1103 uint64_t nb_sectors;
1104
1105 /* if a geometry hint is available, use it */
1106 bdrv_get_geometry(bs, &nb_sectors);
1107 bdrv_get_geometry_hint(bs, &cylinders, &heads, &secs);
1108 translation = bdrv_get_translation_hint(bs);
1109 if (cylinders != 0) {
1110 *pcyls = cylinders;
1111 *pheads = heads;
1112 *psecs = secs;
1113 } else {
1114 if (guess_disk_lchs(bs, &cylinders, &heads, &secs) == 0) {
1115 if (heads > 16) {
1116 /* if heads > 16, it means that a BIOS LBA
1117 translation was active, so the default
1118 hardware geometry is OK */
1119 lba_detected = 1;
1120 goto default_geometry;
1121 } else {
1122 *pcyls = cylinders;
1123 *pheads = heads;
1124 *psecs = secs;
1125 /* disable any translation to be in sync with
1126 the logical geometry */
1127 if (translation == BIOS_ATA_TRANSLATION_AUTO) {
1128 bdrv_set_translation_hint(bs,
1129 BIOS_ATA_TRANSLATION_NONE);
1130 }
1131 }
1132 } else {
1133 default_geometry:
1134 /* if no geometry, use a standard physical disk geometry */
1135 cylinders = nb_sectors / (16 * 63);
1136
1137 if (cylinders > 16383)
1138 cylinders = 16383;
1139 else if (cylinders < 2)
1140 cylinders = 2;
1141 *pcyls = cylinders;
1142 *pheads = 16;
1143 *psecs = 63;
1144 if ((lba_detected == 1) && (translation == BIOS_ATA_TRANSLATION_AUTO)) {
1145 if ((*pcyls * *pheads) <= 131072) {
1146 bdrv_set_translation_hint(bs,
1147 BIOS_ATA_TRANSLATION_LARGE);
1148 } else {
1149 bdrv_set_translation_hint(bs,
1150 BIOS_ATA_TRANSLATION_LBA);
1151 }
1152 }
1153 }
1154 bdrv_set_geometry_hint(bs, *pcyls, *pheads, *psecs);
1155 }
1156 }
1157
1158 void bdrv_set_geometry_hint(BlockDriverState *bs,
1159 int cyls, int heads, int secs)
1160 {
1161 bs->cyls = cyls;
1162 bs->heads = heads;
1163 bs->secs = secs;
1164 }
1165
1166 void bdrv_set_type_hint(BlockDriverState *bs, int type)
1167 {
1168 bs->type = type;
1169 bs->removable = ((type == BDRV_TYPE_CDROM ||
1170 type == BDRV_TYPE_FLOPPY));
1171 }
1172
1173 void bdrv_set_translation_hint(BlockDriverState *bs, int translation)
1174 {
1175 bs->translation = translation;
1176 }
1177
1178 void bdrv_get_geometry_hint(BlockDriverState *bs,
1179 int *pcyls, int *pheads, int *psecs)
1180 {
1181 *pcyls = bs->cyls;
1182 *pheads = bs->heads;
1183 *psecs = bs->secs;
1184 }
1185
1186 int bdrv_get_type_hint(BlockDriverState *bs)
1187 {
1188 return bs->type;
1189 }
1190
1191 int bdrv_get_translation_hint(BlockDriverState *bs)
1192 {
1193 return bs->translation;
1194 }
1195
1196 int bdrv_is_removable(BlockDriverState *bs)
1197 {
1198 return bs->removable;
1199 }
1200
1201 int bdrv_is_read_only(BlockDriverState *bs)
1202 {
1203 return bs->read_only;
1204 }
1205
1206 int bdrv_is_sg(BlockDriverState *bs)
1207 {
1208 return bs->sg;
1209 }
1210
1211 int bdrv_enable_write_cache(BlockDriverState *bs)
1212 {
1213 return bs->enable_write_cache;
1214 }
1215
1216 /* XXX: no longer used */
1217 void bdrv_set_change_cb(BlockDriverState *bs,
1218 void (*change_cb)(void *opaque), void *opaque)
1219 {
1220 bs->change_cb = change_cb;
1221 bs->change_opaque = opaque;
1222 }
1223
1224 int bdrv_is_encrypted(BlockDriverState *bs)
1225 {
1226 if (bs->backing_hd && bs->backing_hd->encrypted)
1227 return 1;
1228 return bs->encrypted;
1229 }
1230
1231 int bdrv_key_required(BlockDriverState *bs)
1232 {
1233 BlockDriverState *backing_hd = bs->backing_hd;
1234
1235 if (backing_hd && backing_hd->encrypted && !backing_hd->valid_key)
1236 return 1;
1237 return (bs->encrypted && !bs->valid_key);
1238 }
1239
1240 int bdrv_set_key(BlockDriverState *bs, const char *key)
1241 {
1242 int ret;
1243 if (bs->backing_hd && bs->backing_hd->encrypted) {
1244 ret = bdrv_set_key(bs->backing_hd, key);
1245 if (ret < 0)
1246 return ret;
1247 if (!bs->encrypted)
1248 return 0;
1249 }
1250 if (!bs->encrypted) {
1251 return -EINVAL;
1252 } else if (!bs->drv || !bs->drv->bdrv_set_key) {
1253 return -ENOMEDIUM;
1254 }
1255 ret = bs->drv->bdrv_set_key(bs, key);
1256 if (ret < 0) {
1257 bs->valid_key = 0;
1258 } else if (!bs->valid_key) {
1259 bs->valid_key = 1;
1260 /* call the change callback now, we skipped it on open */
1261 bs->media_changed = 1;
1262 if (bs->change_cb)
1263 bs->change_cb(bs->change_opaque);
1264 }
1265 return ret;
1266 }
1267
1268 void bdrv_get_format(BlockDriverState *bs, char *buf, int buf_size)
1269 {
1270 if (!bs->drv) {
1271 buf[0] = '\0';
1272 } else {
1273 pstrcpy(buf, buf_size, bs->drv->format_name);
1274 }
1275 }
1276
1277 void bdrv_iterate_format(void (*it)(void *opaque, const char *name),
1278 void *opaque)
1279 {
1280 BlockDriver *drv;
1281
1282 QLIST_FOREACH(drv, &bdrv_drivers, list) {
1283 it(opaque, drv->format_name);
1284 }
1285 }
1286
1287 BlockDriverState *bdrv_find(const char *name)
1288 {
1289 BlockDriverState *bs;
1290
1291 QTAILQ_FOREACH(bs, &bdrv_states, list) {
1292 if (!strcmp(name, bs->device_name)) {
1293 return bs;
1294 }
1295 }
1296 return NULL;
1297 }
1298
1299 void bdrv_iterate(void (*it)(void *opaque, BlockDriverState *bs), void *opaque)
1300 {
1301 BlockDriverState *bs;
1302
1303 QTAILQ_FOREACH(bs, &bdrv_states, list) {
1304 it(opaque, bs);
1305 }
1306 }
1307
1308 const char *bdrv_get_device_name(BlockDriverState *bs)
1309 {
1310 return bs->device_name;
1311 }
1312
1313 void bdrv_flush(BlockDriverState *bs)
1314 {
1315 if (bs->drv && bs->drv->bdrv_flush)
1316 bs->drv->bdrv_flush(bs);
1317 }
1318
1319 void bdrv_flush_all(void)
1320 {
1321 BlockDriverState *bs;
1322
1323 QTAILQ_FOREACH(bs, &bdrv_states, list) {
1324 if (bs->drv && !bdrv_is_read_only(bs) &&
1325 (!bdrv_is_removable(bs) || bdrv_is_inserted(bs))) {
1326 bdrv_flush(bs);
1327 }
1328 }
1329 }
1330
1331 int bdrv_has_zero_init(BlockDriverState *bs)
1332 {
1333 assert(bs->drv);
1334
1335 if (bs->drv->no_zero_init) {
1336 return 0;
1337 } else if (bs->file) {
1338 return bdrv_has_zero_init(bs->file);
1339 }
1340
1341 return 1;
1342 }
1343
1344 /*
1345 * Returns true iff the specified sector is present in the disk image. Drivers
1346 * not implementing the functionality are assumed to not support backing files,
1347 * hence all their sectors are reported as allocated.
1348 *
1349 * 'pnum' is set to the number of sectors (including and immediately following
1350 * the specified sector) that are known to be in the same
1351 * allocated/unallocated state.
1352 *
1353 * 'nb_sectors' is the max value 'pnum' should be set to.
1354 */
1355 int bdrv_is_allocated(BlockDriverState *bs, int64_t sector_num, int nb_sectors,
1356 int *pnum)
1357 {
1358 int64_t n;
1359 if (!bs->drv->bdrv_is_allocated) {
1360 if (sector_num >= bs->total_sectors) {
1361 *pnum = 0;
1362 return 0;
1363 }
1364 n = bs->total_sectors - sector_num;
1365 *pnum = (n < nb_sectors) ? (n) : (nb_sectors);
1366 return 1;
1367 }
1368 return bs->drv->bdrv_is_allocated(bs, sector_num, nb_sectors, pnum);
1369 }
1370
1371 void bdrv_mon_event(const BlockDriverState *bdrv,
1372 BlockMonEventAction action, int is_read)
1373 {
1374 QObject *data;
1375 const char *action_str;
1376
1377 switch (action) {
1378 case BDRV_ACTION_REPORT:
1379 action_str = "report";
1380 break;
1381 case BDRV_ACTION_IGNORE:
1382 action_str = "ignore";
1383 break;
1384 case BDRV_ACTION_STOP:
1385 action_str = "stop";
1386 break;
1387 default:
1388 abort();
1389 }
1390
1391 data = qobject_from_jsonf("{ 'device': %s, 'action': %s, 'operation': %s }",
1392 bdrv->device_name,
1393 action_str,
1394 is_read ? "read" : "write");
1395 monitor_protocol_event(QEVENT_BLOCK_IO_ERROR, data);
1396
1397 qobject_decref(data);
1398 }
1399
1400 static void bdrv_print_dict(QObject *obj, void *opaque)
1401 {
1402 QDict *bs_dict;
1403 Monitor *mon = opaque;
1404
1405 bs_dict = qobject_to_qdict(obj);
1406
1407 monitor_printf(mon, "%s: type=%s removable=%d",
1408 qdict_get_str(bs_dict, "device"),
1409 qdict_get_str(bs_dict, "type"),
1410 qdict_get_bool(bs_dict, "removable"));
1411
1412 if (qdict_get_bool(bs_dict, "removable")) {
1413 monitor_printf(mon, " locked=%d", qdict_get_bool(bs_dict, "locked"));
1414 }
1415
1416 if (qdict_haskey(bs_dict, "inserted")) {
1417 QDict *qdict = qobject_to_qdict(qdict_get(bs_dict, "inserted"));
1418
1419 monitor_printf(mon, " file=");
1420 monitor_print_filename(mon, qdict_get_str(qdict, "file"));
1421 if (qdict_haskey(qdict, "backing_file")) {
1422 monitor_printf(mon, " backing_file=");
1423 monitor_print_filename(mon, qdict_get_str(qdict, "backing_file"));
1424 }
1425 monitor_printf(mon, " ro=%d drv=%s encrypted=%d",
1426 qdict_get_bool(qdict, "ro"),
1427 qdict_get_str(qdict, "drv"),
1428 qdict_get_bool(qdict, "encrypted"));
1429 } else {
1430 monitor_printf(mon, " [not inserted]");
1431 }
1432
1433 monitor_printf(mon, "\n");
1434 }
1435
1436 void bdrv_info_print(Monitor *mon, const QObject *data)
1437 {
1438 qlist_iter(qobject_to_qlist(data), bdrv_print_dict, mon);
1439 }
1440
1441 /**
1442 * bdrv_info(): Block devices information
1443 *
1444 * Each block device information is stored in a QDict and the
1445 * returned QObject is a QList of all devices.
1446 *
1447 * The QDict contains the following:
1448 *
1449 * - "device": device name
1450 * - "type": device type
1451 * - "removable": true if the device is removable, false otherwise
1452 * - "locked": true if the device is locked, false otherwise
1453 * - "inserted": only present if the device is inserted, it is a QDict
1454 * containing the following:
1455 * - "file": device file name
1456 * - "ro": true if read-only, false otherwise
1457 * - "drv": driver format name
1458 * - "backing_file": backing file name if one is used
1459 * - "encrypted": true if encrypted, false otherwise
1460 *
1461 * Example:
1462 *
1463 * [ { "device": "ide0-hd0", "type": "hd", "removable": false, "locked": false,
1464 * "inserted": { "file": "/tmp/foobar", "ro": false, "drv": "qcow2" } },
1465 * { "device": "floppy0", "type": "floppy", "removable": true,
1466 * "locked": false } ]
1467 */
1468 void bdrv_info(Monitor *mon, QObject **ret_data)
1469 {
1470 QList *bs_list;
1471 BlockDriverState *bs;
1472
1473 bs_list = qlist_new();
1474
1475 QTAILQ_FOREACH(bs, &bdrv_states, list) {
1476 QObject *bs_obj;
1477 const char *type = "unknown";
1478
1479 switch(bs->type) {
1480 case BDRV_TYPE_HD:
1481 type = "hd";
1482 break;
1483 case BDRV_TYPE_CDROM:
1484 type = "cdrom";
1485 break;
1486 case BDRV_TYPE_FLOPPY:
1487 type = "floppy";
1488 break;
1489 }
1490
1491 bs_obj = qobject_from_jsonf("{ 'device': %s, 'type': %s, "
1492 "'removable': %i, 'locked': %i }",
1493 bs->device_name, type, bs->removable,
1494 bs->locked);
1495
1496 if (bs->drv) {
1497 QObject *obj;
1498 QDict *bs_dict = qobject_to_qdict(bs_obj);
1499
1500 obj = qobject_from_jsonf("{ 'file': %s, 'ro': %i, 'drv': %s, "
1501 "'encrypted': %i }",
1502 bs->filename, bs->read_only,
1503 bs->drv->format_name,
1504 bdrv_is_encrypted(bs));
1505 if (bs->backing_file[0] != '\0') {
1506 QDict *qdict = qobject_to_qdict(obj);
1507 qdict_put(qdict, "backing_file",
1508 qstring_from_str(bs->backing_file));
1509 }
1510
1511 qdict_put_obj(bs_dict, "inserted", obj);
1512 }
1513 qlist_append_obj(bs_list, bs_obj);
1514 }
1515
1516 *ret_data = QOBJECT(bs_list);
1517 }
1518
1519 static void bdrv_stats_iter(QObject *data, void *opaque)
1520 {
1521 QDict *qdict;
1522 Monitor *mon = opaque;
1523
1524 qdict = qobject_to_qdict(data);
1525 monitor_printf(mon, "%s:", qdict_get_str(qdict, "device"));
1526
1527 qdict = qobject_to_qdict(qdict_get(qdict, "stats"));
1528 monitor_printf(mon, " rd_bytes=%" PRId64
1529 " wr_bytes=%" PRId64
1530 " rd_operations=%" PRId64
1531 " wr_operations=%" PRId64
1532 "\n",
1533 qdict_get_int(qdict, "rd_bytes"),
1534 qdict_get_int(qdict, "wr_bytes"),
1535 qdict_get_int(qdict, "rd_operations"),
1536 qdict_get_int(qdict, "wr_operations"));
1537 }
1538
1539 void bdrv_stats_print(Monitor *mon, const QObject *data)
1540 {
1541 qlist_iter(qobject_to_qlist(data), bdrv_stats_iter, mon);
1542 }
1543
1544 static QObject* bdrv_info_stats_bs(BlockDriverState *bs)
1545 {
1546 QObject *res;
1547 QDict *dict;
1548
1549 res = qobject_from_jsonf("{ 'stats': {"
1550 "'rd_bytes': %" PRId64 ","
1551 "'wr_bytes': %" PRId64 ","
1552 "'rd_operations': %" PRId64 ","
1553 "'wr_operations': %" PRId64 ","
1554 "'wr_highest_offset': %" PRId64
1555 "} }",
1556 bs->rd_bytes, bs->wr_bytes,
1557 bs->rd_ops, bs->wr_ops,
1558 bs->wr_highest_sector * 512);
1559 dict = qobject_to_qdict(res);
1560
1561 if (*bs->device_name) {
1562 qdict_put(dict, "device", qstring_from_str(bs->device_name));
1563 }
1564
1565 if (bs->file) {
1566 QObject *parent = bdrv_info_stats_bs(bs->file);
1567 qdict_put_obj(dict, "parent", parent);
1568 }
1569
1570 return res;
1571 }
1572
1573 /**
1574 * bdrv_info_stats(): show block device statistics
1575 *
1576 * Each device statistic information is stored in a QDict and
1577 * the returned QObject is a QList of all devices.
1578 *
1579 * The QDict contains the following:
1580 *
1581 * - "device": device name
1582 * - "stats": A QDict with the statistics information, it contains:
1583 * - "rd_bytes": bytes read
1584 * - "wr_bytes": bytes written
1585 * - "rd_operations": read operations
1586 * - "wr_operations": write operations
1587 * - "wr_highest_offset": Highest offset of a sector written since the
1588 * BlockDriverState has been opened
1589 * - "parent": A QDict recursively holding the statistics of the underlying
1590 * protocol (e.g. the host file for a qcow2 image). If there is no
1591 * underlying protocol, this field is omitted.
1592 *
1593 * Example:
1594 *
1595 * [ { "device": "ide0-hd0",
1596 * "stats": { "rd_bytes": 512,
1597 * "wr_bytes": 0,
1598 * "rd_operations": 1,
1599 * "wr_operations": 0,
1600 * "wr_highest_offset": 0 },
1601 * "parent": {
1602 * "stats": { "rd_bytes": 1024,
1603 * "wr_bytes": 0,
1604 * "rd_operations": 2,
1605 * "wr_operations": 0,
1606 * "wr_highest_offset": 0,
1607 * } } },
1608 * { "device": "ide1-cd0",
1609 * "stats": { "rd_bytes": 0,
1610 * "wr_bytes": 0,
1611 * "rd_operations": 0,
1612 * "wr_operations": 0,
1613 * "wr_highest_offset": 0 } },
1614 */
1615 void bdrv_info_stats(Monitor *mon, QObject **ret_data)
1616 {
1617 QObject *obj;
1618 QList *devices;
1619 BlockDriverState *bs;
1620
1621 devices = qlist_new();
1622
1623 QTAILQ_FOREACH(bs, &bdrv_states, list) {
1624 obj = bdrv_info_stats_bs(bs);
1625 qlist_append_obj(devices, obj);
1626 }
1627
1628 *ret_data = QOBJECT(devices);
1629 }
1630
1631 const char *bdrv_get_encrypted_filename(BlockDriverState *bs)
1632 {
1633 if (bs->backing_hd && bs->backing_hd->encrypted)
1634 return bs->backing_file;
1635 else if (bs->encrypted)
1636 return bs->filename;
1637 else
1638 return NULL;
1639 }
1640
1641 void bdrv_get_backing_filename(BlockDriverState *bs,
1642 char *filename, int filename_size)
1643 {
1644 if (!bs->backing_file) {
1645 pstrcpy(filename, filename_size, "");
1646 } else {
1647 pstrcpy(filename, filename_size, bs->backing_file);
1648 }
1649 }
1650
1651 int bdrv_write_compressed(BlockDriverState *bs, int64_t sector_num,
1652 const uint8_t *buf, int nb_sectors)
1653 {
1654 BlockDriver *drv = bs->drv;
1655 if (!drv)
1656 return -ENOMEDIUM;
1657 if (!drv->bdrv_write_compressed)
1658 return -ENOTSUP;
1659 if (bdrv_check_request(bs, sector_num, nb_sectors))
1660 return -EIO;
1661
1662 if (bs->dirty_bitmap) {
1663 set_dirty_bitmap(bs, sector_num, nb_sectors, 1);
1664 }
1665
1666 return drv->bdrv_write_compressed(bs, sector_num, buf, nb_sectors);
1667 }
1668
1669 int bdrv_get_info(BlockDriverState *bs, BlockDriverInfo *bdi)
1670 {
1671 BlockDriver *drv = bs->drv;
1672 if (!drv)
1673 return -ENOMEDIUM;
1674 if (!drv->bdrv_get_info)
1675 return -ENOTSUP;
1676 memset(bdi, 0, sizeof(*bdi));
1677 return drv->bdrv_get_info(bs, bdi);
1678 }
1679
1680 int bdrv_save_vmstate(BlockDriverState *bs, const uint8_t *buf,
1681 int64_t pos, int size)
1682 {
1683 BlockDriver *drv = bs->drv;
1684 if (!drv)
1685 return -ENOMEDIUM;
1686 if (!drv->bdrv_save_vmstate)
1687 return -ENOTSUP;
1688 return drv->bdrv_save_vmstate(bs, buf, pos, size);
1689 }
1690
1691 int bdrv_load_vmstate(BlockDriverState *bs, uint8_t *buf,
1692 int64_t pos, int size)
1693 {
1694 BlockDriver *drv = bs->drv;
1695 if (!drv)
1696 return -ENOMEDIUM;
1697 if (!drv->bdrv_load_vmstate)
1698 return -ENOTSUP;
1699 return drv->bdrv_load_vmstate(bs, buf, pos, size);
1700 }
1701
1702 void bdrv_debug_event(BlockDriverState *bs, BlkDebugEvent event)
1703 {
1704 BlockDriver *drv = bs->drv;
1705
1706 if (!drv || !drv->bdrv_debug_event) {
1707 return;
1708 }
1709
1710 return drv->bdrv_debug_event(bs, event);
1711
1712 }
1713
1714 /**************************************************************/
1715 /* handling of snapshots */
1716
1717 int bdrv_snapshot_create(BlockDriverState *bs,
1718 QEMUSnapshotInfo *sn_info)
1719 {
1720 BlockDriver *drv = bs->drv;
1721 if (!drv)
1722 return -ENOMEDIUM;
1723 if (!drv->bdrv_snapshot_create)
1724 return -ENOTSUP;
1725 return drv->bdrv_snapshot_create(bs, sn_info);
1726 }
1727
1728 int bdrv_snapshot_goto(BlockDriverState *bs,
1729 const char *snapshot_id)
1730 {
1731 BlockDriver *drv = bs->drv;
1732 if (!drv)
1733 return -ENOMEDIUM;
1734 if (!drv->bdrv_snapshot_goto)
1735 return -ENOTSUP;
1736 return drv->bdrv_snapshot_goto(bs, snapshot_id);
1737 }
1738
1739 int bdrv_snapshot_delete(BlockDriverState *bs, const char *snapshot_id)
1740 {
1741 BlockDriver *drv = bs->drv;
1742 if (!drv)
1743 return -ENOMEDIUM;
1744 if (!drv->bdrv_snapshot_delete)
1745 return -ENOTSUP;
1746 return drv->bdrv_snapshot_delete(bs, snapshot_id);
1747 }
1748
1749 int bdrv_snapshot_list(BlockDriverState *bs,
1750 QEMUSnapshotInfo **psn_info)
1751 {
1752 BlockDriver *drv = bs->drv;
1753 if (!drv)
1754 return -ENOMEDIUM;
1755 if (!drv->bdrv_snapshot_list)
1756 return -ENOTSUP;
1757 return drv->bdrv_snapshot_list(bs, psn_info);
1758 }
1759
1760 #define NB_SUFFIXES 4
1761
1762 char *get_human_readable_size(char *buf, int buf_size, int64_t size)
1763 {
1764 static const char suffixes[NB_SUFFIXES] = "KMGT";
1765 int64_t base;
1766 int i;
1767
1768 if (size <= 999) {
1769 snprintf(buf, buf_size, "%" PRId64, size);
1770 } else {
1771 base = 1024;
1772 for(i = 0; i < NB_SUFFIXES; i++) {
1773 if (size < (10 * base)) {
1774 snprintf(buf, buf_size, "%0.1f%c",
1775 (double)size / base,
1776 suffixes[i]);
1777 break;
1778 } else if (size < (1000 * base) || i == (NB_SUFFIXES - 1)) {
1779 snprintf(buf, buf_size, "%" PRId64 "%c",
1780 ((size + (base >> 1)) / base),
1781 suffixes[i]);
1782 break;
1783 }
1784 base = base * 1024;
1785 }
1786 }
1787 return buf;
1788 }
1789
1790 char *bdrv_snapshot_dump(char *buf, int buf_size, QEMUSnapshotInfo *sn)
1791 {
1792 char buf1[128], date_buf[128], clock_buf[128];
1793 #ifdef _WIN32
1794 struct tm *ptm;
1795 #else
1796 struct tm tm;
1797 #endif
1798 time_t ti;
1799 int64_t secs;
1800
1801 if (!sn) {
1802 snprintf(buf, buf_size,
1803 "%-10s%-20s%7s%20s%15s",
1804 "ID", "TAG", "VM SIZE", "DATE", "VM CLOCK");
1805 } else {
1806 ti = sn->date_sec;
1807 #ifdef _WIN32
1808 ptm = localtime(&ti);
1809 strftime(date_buf, sizeof(date_buf),
1810 "%Y-%m-%d %H:%M:%S", ptm);
1811 #else
1812 localtime_r(&ti, &tm);
1813 strftime(date_buf, sizeof(date_buf),
1814 "%Y-%m-%d %H:%M:%S", &tm);
1815 #endif
1816 secs = sn->vm_clock_nsec / 1000000000;
1817 snprintf(clock_buf, sizeof(clock_buf),
1818 "%02d:%02d:%02d.%03d",
1819 (int)(secs / 3600),
1820 (int)((secs / 60) % 60),
1821 (int)(secs % 60),
1822 (int)((sn->vm_clock_nsec / 1000000) % 1000));
1823 snprintf(buf, buf_size,
1824 "%-10s%-20s%7s%20s%15s",
1825 sn->id_str, sn->name,
1826 get_human_readable_size(buf1, sizeof(buf1), sn->vm_state_size),
1827 date_buf,
1828 clock_buf);
1829 }
1830 return buf;
1831 }
1832
1833
1834 /**************************************************************/
1835 /* async I/Os */
1836
1837 BlockDriverAIOCB *bdrv_aio_readv(BlockDriverState *bs, int64_t sector_num,
1838 QEMUIOVector *qiov, int nb_sectors,
1839 BlockDriverCompletionFunc *cb, void *opaque)
1840 {
1841 BlockDriver *drv = bs->drv;
1842 BlockDriverAIOCB *ret;
1843
1844 if (!drv)
1845 return NULL;
1846 if (bdrv_check_request(bs, sector_num, nb_sectors))
1847 return NULL;
1848
1849 ret = drv->bdrv_aio_readv(bs, sector_num, qiov, nb_sectors,
1850 cb, opaque);
1851
1852 if (ret) {
1853 /* Update stats even though technically transfer has not happened. */
1854 bs->rd_bytes += (unsigned) nb_sectors * BDRV_SECTOR_SIZE;
1855 bs->rd_ops ++;
1856 }
1857
1858 return ret;
1859 }
1860
1861 BlockDriverAIOCB *bdrv_aio_writev(BlockDriverState *bs, int64_t sector_num,
1862 QEMUIOVector *qiov, int nb_sectors,
1863 BlockDriverCompletionFunc *cb, void *opaque)
1864 {
1865 BlockDriver *drv = bs->drv;
1866 BlockDriverAIOCB *ret;
1867
1868 if (!drv)
1869 return NULL;
1870 if (bs->read_only)
1871 return NULL;
1872 if (bdrv_check_request(bs, sector_num, nb_sectors))
1873 return NULL;
1874
1875 if (bs->dirty_bitmap) {
1876 set_dirty_bitmap(bs, sector_num, nb_sectors, 1);
1877 }
1878
1879 ret = drv->bdrv_aio_writev(bs, sector_num, qiov, nb_sectors,
1880 cb, opaque);
1881
1882 if (ret) {
1883 /* Update stats even though technically transfer has not happened. */
1884 bs->wr_bytes += (unsigned) nb_sectors * BDRV_SECTOR_SIZE;
1885 bs->wr_ops ++;
1886 if (bs->wr_highest_sector < sector_num + nb_sectors - 1) {
1887 bs->wr_highest_sector = sector_num + nb_sectors - 1;
1888 }
1889 }
1890
1891 return ret;
1892 }
1893
1894
1895 typedef struct MultiwriteCB {
1896 int error;
1897 int num_requests;
1898 int num_callbacks;
1899 struct {
1900 BlockDriverCompletionFunc *cb;
1901 void *opaque;
1902 QEMUIOVector *free_qiov;
1903 void *free_buf;
1904 } callbacks[];
1905 } MultiwriteCB;
1906
1907 static void multiwrite_user_cb(MultiwriteCB *mcb)
1908 {
1909 int i;
1910
1911 for (i = 0; i < mcb->num_callbacks; i++) {
1912 mcb->callbacks[i].cb(mcb->callbacks[i].opaque, mcb->error);
1913 if (mcb->callbacks[i].free_qiov) {
1914 qemu_iovec_destroy(mcb->callbacks[i].free_qiov);
1915 }
1916 qemu_free(mcb->callbacks[i].free_qiov);
1917 qemu_vfree(mcb->callbacks[i].free_buf);
1918 }
1919 }
1920
1921 static void multiwrite_cb(void *opaque, int ret)
1922 {
1923 MultiwriteCB *mcb = opaque;
1924
1925 if (ret < 0 && !mcb->error) {
1926 mcb->error = ret;
1927 multiwrite_user_cb(mcb);
1928 }
1929
1930 mcb->num_requests--;
1931 if (mcb->num_requests == 0) {
1932 if (mcb->error == 0) {
1933 multiwrite_user_cb(mcb);
1934 }
1935 qemu_free(mcb);
1936 }
1937 }
1938
1939 static int multiwrite_req_compare(const void *a, const void *b)
1940 {
1941 const BlockRequest *req1 = a, *req2 = b;
1942
1943 /*
1944 * Note that we can't simply subtract req2->sector from req1->sector
1945 * here as that could overflow the return value.
1946 */
1947 if (req1->sector > req2->sector) {
1948 return 1;
1949 } else if (req1->sector < req2->sector) {
1950 return -1;
1951 } else {
1952 return 0;
1953 }
1954 }
1955
1956 /*
1957 * Takes a bunch of requests and tries to merge them. Returns the number of
1958 * requests that remain after merging.
1959 */
1960 static int multiwrite_merge(BlockDriverState *bs, BlockRequest *reqs,
1961 int num_reqs, MultiwriteCB *mcb)
1962 {
1963 int i, outidx;
1964
1965 // Sort requests by start sector
1966 qsort(reqs, num_reqs, sizeof(*reqs), &multiwrite_req_compare);
1967
1968 // Check if adjacent requests touch the same clusters. If so, combine them,
1969 // filling up gaps with zero sectors.
1970 outidx = 0;
1971 for (i = 1; i < num_reqs; i++) {
1972 int merge = 0;
1973 int64_t oldreq_last = reqs[outidx].sector + reqs[outidx].nb_sectors;
1974
1975 // This handles the cases that are valid for all block drivers, namely
1976 // exactly sequential writes and overlapping writes.
1977 if (reqs[i].sector <= oldreq_last) {
1978 merge = 1;
1979 }
1980
1981 // The block driver may decide that it makes sense to combine requests
1982 // even if there is a gap of some sectors between them. In this case,
1983 // the gap is filled with zeros (therefore only applicable for yet
1984 // unused space in format like qcow2).
1985 if (!merge && bs->drv->bdrv_merge_requests) {
1986 merge = bs->drv->bdrv_merge_requests(bs, &reqs[outidx], &reqs[i]);
1987 }
1988
1989 if (reqs[outidx].qiov->niov + reqs[i].qiov->niov + 1 > IOV_MAX) {
1990 merge = 0;
1991 }
1992
1993 if (merge) {
1994 size_t size;
1995 QEMUIOVector *qiov = qemu_mallocz(sizeof(*qiov));
1996 qemu_iovec_init(qiov,
1997 reqs[outidx].qiov->niov + reqs[i].qiov->niov + 1);
1998
1999 // Add the first request to the merged one. If the requests are
2000 // overlapping, drop the last sectors of the first request.
2001 size = (reqs[i].sector - reqs[outidx].sector) << 9;
2002 qemu_iovec_concat(qiov, reqs[outidx].qiov, size);
2003
2004 // We might need to add some zeros between the two requests
2005 if (reqs[i].sector > oldreq_last) {
2006 size_t zero_bytes = (reqs[i].sector - oldreq_last) << 9;
2007 uint8_t *buf = qemu_blockalign(bs, zero_bytes);
2008 memset(buf, 0, zero_bytes);
2009 qemu_iovec_add(qiov, buf, zero_bytes);
2010 mcb->callbacks[i].free_buf = buf;
2011 }
2012
2013 // Add the second request
2014 qemu_iovec_concat(qiov, reqs[i].qiov, reqs[i].qiov->size);
2015
2016 reqs[outidx].nb_sectors += reqs[i].nb_sectors;
2017 reqs[outidx].qiov = qiov;
2018
2019 mcb->callbacks[i].free_qiov = reqs[outidx].qiov;
2020 } else {
2021 outidx++;
2022 reqs[outidx].sector = reqs[i].sector;
2023 reqs[outidx].nb_sectors = reqs[i].nb_sectors;
2024 reqs[outidx].qiov = reqs[i].qiov;
2025 }
2026 }
2027
2028 return outidx + 1;
2029 }
2030
2031 /*
2032 * Submit multiple AIO write requests at once.
2033 *
2034 * On success, the function returns 0 and all requests in the reqs array have
2035 * been submitted. In error case this function returns -1, and any of the
2036 * requests may or may not be submitted yet. In particular, this means that the
2037 * callback will be called for some of the requests, for others it won't. The
2038 * caller must check the error field of the BlockRequest to wait for the right
2039 * callbacks (if error != 0, no callback will be called).
2040 *
2041 * The implementation may modify the contents of the reqs array, e.g. to merge
2042 * requests. However, the fields opaque and error are left unmodified as they
2043 * are used to signal failure for a single request to the caller.
2044 */
2045 int bdrv_aio_multiwrite(BlockDriverState *bs, BlockRequest *reqs, int num_reqs)
2046 {
2047 BlockDriverAIOCB *acb;
2048 MultiwriteCB *mcb;
2049 int i;
2050
2051 if (num_reqs == 0) {
2052 return 0;
2053 }
2054
2055 // Create MultiwriteCB structure
2056 mcb = qemu_mallocz(sizeof(*mcb) + num_reqs * sizeof(*mcb->callbacks));
2057 mcb->num_requests = 0;
2058 mcb->num_callbacks = num_reqs;
2059
2060 for (i = 0; i < num_reqs; i++) {
2061 mcb->callbacks[i].cb = reqs[i].cb;
2062 mcb->callbacks[i].opaque = reqs[i].opaque;
2063 }
2064
2065 // Check for mergable requests
2066 num_reqs = multiwrite_merge(bs, reqs, num_reqs, mcb);
2067
2068 // Run the aio requests
2069 for (i = 0; i < num_reqs; i++) {
2070 acb = bdrv_aio_writev(bs, reqs[i].sector, reqs[i].qiov,
2071 reqs[i].nb_sectors, multiwrite_cb, mcb);
2072
2073 if (acb == NULL) {
2074 // We can only fail the whole thing if no request has been
2075 // submitted yet. Otherwise we'll wait for the submitted AIOs to
2076 // complete and report the error in the callback.
2077 if (mcb->num_requests == 0) {
2078 reqs[i].error = -EIO;
2079 goto fail;
2080 } else {
2081 mcb->num_requests++;
2082 multiwrite_cb(mcb, -EIO);
2083 break;
2084 }
2085 } else {
2086 mcb->num_requests++;
2087 }
2088 }
2089
2090 return 0;
2091
2092 fail:
2093 qemu_free(mcb);
2094 return -1;
2095 }
2096
2097 BlockDriverAIOCB *bdrv_aio_flush(BlockDriverState *bs,
2098 BlockDriverCompletionFunc *cb, void *opaque)
2099 {
2100 BlockDriver *drv = bs->drv;
2101
2102 if (!drv)
2103 return NULL;
2104 return drv->bdrv_aio_flush(bs, cb, opaque);
2105 }
2106
2107 void bdrv_aio_cancel(BlockDriverAIOCB *acb)
2108 {
2109 acb->pool->cancel(acb);
2110 }
2111
2112
2113 /**************************************************************/
2114 /* async block device emulation */
2115
2116 typedef struct BlockDriverAIOCBSync {
2117 BlockDriverAIOCB common;
2118 QEMUBH *bh;
2119 int ret;
2120 /* vector translation state */
2121 QEMUIOVector *qiov;
2122 uint8_t *bounce;
2123 int is_write;
2124 } BlockDriverAIOCBSync;
2125
2126 static void bdrv_aio_cancel_em(BlockDriverAIOCB *blockacb)
2127 {
2128 BlockDriverAIOCBSync *acb =
2129 container_of(blockacb, BlockDriverAIOCBSync, common);
2130 qemu_bh_delete(acb->bh);
2131 acb->bh = NULL;
2132 qemu_aio_release(acb);
2133 }
2134
2135 static AIOPool bdrv_em_aio_pool = {
2136 .aiocb_size = sizeof(BlockDriverAIOCBSync),
2137 .cancel = bdrv_aio_cancel_em,
2138 };
2139
2140 static void bdrv_aio_bh_cb(void *opaque)
2141 {
2142 BlockDriverAIOCBSync *acb = opaque;
2143
2144 if (!acb->is_write)
2145 qemu_iovec_from_buffer(acb->qiov, acb->bounce, acb->qiov->size);
2146 qemu_vfree(acb->bounce);
2147 acb->common.cb(acb->common.opaque, acb->ret);
2148 qemu_bh_delete(acb->bh);
2149 acb->bh = NULL;
2150 qemu_aio_release(acb);
2151 }
2152
2153 static BlockDriverAIOCB *bdrv_aio_rw_vector(BlockDriverState *bs,
2154 int64_t sector_num,
2155 QEMUIOVector *qiov,
2156 int nb_sectors,
2157 BlockDriverCompletionFunc *cb,
2158 void *opaque,
2159 int is_write)
2160
2161 {
2162 BlockDriverAIOCBSync *acb;
2163
2164 acb = qemu_aio_get(&bdrv_em_aio_pool, bs, cb, opaque);
2165 acb->is_write = is_write;
2166 acb->qiov = qiov;
2167 acb->bounce = qemu_blockalign(bs, qiov->size);
2168
2169 if (!acb->bh)
2170 acb->bh = qemu_bh_new(bdrv_aio_bh_cb, acb);
2171
2172 if (is_write) {
2173 qemu_iovec_to_buffer(acb->qiov, acb->bounce);
2174 acb->ret = bdrv_write(bs, sector_num, acb->bounce, nb_sectors);
2175 } else {
2176 acb->ret = bdrv_read(bs, sector_num, acb->bounce, nb_sectors);
2177 }
2178
2179 qemu_bh_schedule(acb->bh);
2180
2181 return &acb->common;
2182 }
2183
2184 static BlockDriverAIOCB *bdrv_aio_readv_em(BlockDriverState *bs,
2185 int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
2186 BlockDriverCompletionFunc *cb, void *opaque)
2187 {
2188 return bdrv_aio_rw_vector(bs, sector_num, qiov, nb_sectors, cb, opaque, 0);
2189 }
2190
2191 static BlockDriverAIOCB *bdrv_aio_writev_em(BlockDriverState *bs,
2192 int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
2193 BlockDriverCompletionFunc *cb, void *opaque)
2194 {
2195 return bdrv_aio_rw_vector(bs, sector_num, qiov, nb_sectors, cb, opaque, 1);
2196 }
2197
2198 static BlockDriverAIOCB *bdrv_aio_flush_em(BlockDriverState *bs,
2199 BlockDriverCompletionFunc *cb, void *opaque)
2200 {
2201 BlockDriverAIOCBSync *acb;
2202
2203 acb = qemu_aio_get(&bdrv_em_aio_pool, bs, cb, opaque);
2204 acb->is_write = 1; /* don't bounce in the completion hadler */
2205 acb->qiov = NULL;
2206 acb->bounce = NULL;
2207 acb->ret = 0;
2208
2209 if (!acb->bh)
2210 acb->bh = qemu_bh_new(bdrv_aio_bh_cb, acb);
2211
2212 bdrv_flush(bs);
2213 qemu_bh_schedule(acb->bh);
2214 return &acb->common;
2215 }
2216
2217 /**************************************************************/
2218 /* sync block device emulation */
2219
2220 static void bdrv_rw_em_cb(void *opaque, int ret)
2221 {
2222 *(int *)opaque = ret;
2223 }
2224
2225 #define NOT_DONE 0x7fffffff
2226
2227 static int bdrv_read_em(BlockDriverState *bs, int64_t sector_num,
2228 uint8_t *buf, int nb_sectors)
2229 {
2230 int async_ret;
2231 BlockDriverAIOCB *acb;
2232 struct iovec iov;
2233 QEMUIOVector qiov;
2234
2235 async_context_push();
2236
2237 async_ret = NOT_DONE;
2238 iov.iov_base = (void *)buf;
2239 iov.iov_len = nb_sectors * 512;
2240 qemu_iovec_init_external(&qiov, &iov, 1);
2241 acb = bdrv_aio_readv(bs, sector_num, &qiov, nb_sectors,
2242 bdrv_rw_em_cb, &async_ret);
2243 if (acb == NULL) {
2244 async_ret = -1;
2245 goto fail;
2246 }
2247
2248 while (async_ret == NOT_DONE) {
2249 qemu_aio_wait();
2250 }
2251
2252
2253 fail:
2254 async_context_pop();
2255 return async_ret;
2256 }
2257
2258 static int bdrv_write_em(BlockDriverState *bs, int64_t sector_num,
2259 const uint8_t *buf, int nb_sectors)
2260 {
2261 int async_ret;
2262 BlockDriverAIOCB *acb;
2263 struct iovec iov;
2264 QEMUIOVector qiov;
2265
2266 async_context_push();
2267
2268 async_ret = NOT_DONE;
2269 iov.iov_base = (void *)buf;
2270 iov.iov_len = nb_sectors * 512;
2271 qemu_iovec_init_external(&qiov, &iov, 1);
2272 acb = bdrv_aio_writev(bs, sector_num, &qiov, nb_sectors,
2273 bdrv_rw_em_cb, &async_ret);
2274 if (acb == NULL) {
2275 async_ret = -1;
2276 goto fail;
2277 }
2278 while (async_ret == NOT_DONE) {
2279 qemu_aio_wait();
2280 }
2281
2282 fail:
2283 async_context_pop();
2284 return async_ret;
2285 }
2286
2287 void bdrv_init(void)
2288 {
2289 module_call_init(MODULE_INIT_BLOCK);
2290 }
2291
2292 void bdrv_init_with_whitelist(void)
2293 {
2294 use_bdrv_whitelist = 1;
2295 bdrv_init();
2296 }
2297
2298 void *qemu_aio_get(AIOPool *pool, BlockDriverState *bs,
2299 BlockDriverCompletionFunc *cb, void *opaque)
2300 {
2301 BlockDriverAIOCB *acb;
2302
2303 if (pool->free_aiocb) {
2304 acb = pool->free_aiocb;
2305 pool->free_aiocb = acb->next;
2306 } else {
2307 acb = qemu_mallocz(pool->aiocb_size);
2308 acb->pool = pool;
2309 }
2310 acb->bs = bs;
2311 acb->cb = cb;
2312 acb->opaque = opaque;
2313 return acb;
2314 }
2315
2316 void qemu_aio_release(void *p)
2317 {
2318 BlockDriverAIOCB *acb = (BlockDriverAIOCB *)p;
2319 AIOPool *pool = acb->pool;
2320 acb->next = pool->free_aiocb;
2321 pool->free_aiocb = acb;
2322 }
2323
2324 /**************************************************************/
2325 /* removable device support */
2326
2327 /**
2328 * Return TRUE if the media is present
2329 */
2330 int bdrv_is_inserted(BlockDriverState *bs)
2331 {
2332 BlockDriver *drv = bs->drv;
2333 int ret;
2334 if (!drv)
2335 return 0;
2336 if (!drv->bdrv_is_inserted)
2337 return 1;
2338 ret = drv->bdrv_is_inserted(bs);
2339 return ret;
2340 }
2341
2342 /**
2343 * Return TRUE if the media changed since the last call to this
2344 * function. It is currently only used for floppy disks
2345 */
2346 int bdrv_media_changed(BlockDriverState *bs)
2347 {
2348 BlockDriver *drv = bs->drv;
2349 int ret;
2350
2351 if (!drv || !drv->bdrv_media_changed)
2352 ret = -ENOTSUP;
2353 else
2354 ret = drv->bdrv_media_changed(bs);
2355 if (ret == -ENOTSUP)
2356 ret = bs->media_changed;
2357 bs->media_changed = 0;
2358 return ret;
2359 }
2360
2361 /**
2362 * If eject_flag is TRUE, eject the media. Otherwise, close the tray
2363 */
2364 int bdrv_eject(BlockDriverState *bs, int eject_flag)
2365 {
2366 BlockDriver *drv = bs->drv;
2367 int ret;
2368
2369 if (bs->locked) {
2370 return -EBUSY;
2371 }
2372
2373 if (!drv || !drv->bdrv_eject) {
2374 ret = -ENOTSUP;
2375 } else {
2376 ret = drv->bdrv_eject(bs, eject_flag);
2377 }
2378 if (ret == -ENOTSUP) {
2379 if (eject_flag)
2380 bdrv_close(bs);
2381 ret = 0;
2382 }
2383
2384 return ret;
2385 }
2386
2387 int bdrv_is_locked(BlockDriverState *bs)
2388 {
2389 return bs->locked;
2390 }
2391
2392 /**
2393 * Lock or unlock the media (if it is locked, the user won't be able
2394 * to eject it manually).
2395 */
2396 void bdrv_set_locked(BlockDriverState *bs, int locked)
2397 {
2398 BlockDriver *drv = bs->drv;
2399
2400 bs->locked = locked;
2401 if (drv && drv->bdrv_set_locked) {
2402 drv->bdrv_set_locked(bs, locked);
2403 }
2404 }
2405
2406 /* needed for generic scsi interface */
2407
2408 int bdrv_ioctl(BlockDriverState *bs, unsigned long int req, void *buf)
2409 {
2410 BlockDriver *drv = bs->drv;
2411
2412 if (drv && drv->bdrv_ioctl)
2413 return drv->bdrv_ioctl(bs, req, buf);
2414 return -ENOTSUP;
2415 }
2416
2417 BlockDriverAIOCB *bdrv_aio_ioctl(BlockDriverState *bs,
2418 unsigned long int req, void *buf,
2419 BlockDriverCompletionFunc *cb, void *opaque)
2420 {
2421 BlockDriver *drv = bs->drv;
2422
2423 if (drv && drv->bdrv_aio_ioctl)
2424 return drv->bdrv_aio_ioctl(bs, req, buf, cb, opaque);
2425 return NULL;
2426 }
2427
2428
2429
2430 void *qemu_blockalign(BlockDriverState *bs, size_t size)
2431 {
2432 return qemu_memalign((bs && bs->buffer_alignment) ? bs->buffer_alignment : 512, size);
2433 }
2434
2435 void bdrv_set_dirty_tracking(BlockDriverState *bs, int enable)
2436 {
2437 int64_t bitmap_size;
2438
2439 bs->dirty_count = 0;
2440 if (enable) {
2441 if (!bs->dirty_bitmap) {
2442 bitmap_size = (bdrv_getlength(bs) >> BDRV_SECTOR_BITS) +
2443 BDRV_SECTORS_PER_DIRTY_CHUNK * 8 - 1;
2444 bitmap_size /= BDRV_SECTORS_PER_DIRTY_CHUNK * 8;
2445
2446 bs->dirty_bitmap = qemu_mallocz(bitmap_size);
2447 }
2448 } else {
2449 if (bs->dirty_bitmap) {
2450 qemu_free(bs->dirty_bitmap);
2451 bs->dirty_bitmap = NULL;
2452 }
2453 }
2454 }
2455
2456 int bdrv_get_dirty(BlockDriverState *bs, int64_t sector)
2457 {
2458 int64_t chunk = sector / (int64_t)BDRV_SECTORS_PER_DIRTY_CHUNK;
2459
2460 if (bs->dirty_bitmap &&
2461 (sector << BDRV_SECTOR_BITS) < bdrv_getlength(bs)) {
2462 return bs->dirty_bitmap[chunk / (sizeof(unsigned long) * 8)] &
2463 (1 << (chunk % (sizeof(unsigned long) * 8)));
2464 } else {
2465 return 0;
2466 }
2467 }
2468
2469 void bdrv_reset_dirty(BlockDriverState *bs, int64_t cur_sector,
2470 int nr_sectors)
2471 {
2472 set_dirty_bitmap(bs, cur_sector, nr_sectors, 0);
2473 }
2474
2475 int64_t bdrv_get_dirty_count(BlockDriverState *bs)
2476 {
2477 return bs->dirty_count;
2478 }