?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
tcg-i386: fix a typo
[qemu/aliguori-queue.git] / block.c
blob67f099878a1b46d63d108425dec1bd0b51280afe
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 static BlockDriver *find_protocol(const char *filename);
60
61 static QTAILQ_HEAD(, BlockDriverState) bdrv_states =
62 QTAILQ_HEAD_INITIALIZER(bdrv_states);
63
64 static QLIST_HEAD(, BlockDriver) bdrv_drivers =
65 QLIST_HEAD_INITIALIZER(bdrv_drivers);
66
67 /* If non-zero, use only whitelisted block drivers */
68 static int use_bdrv_whitelist;
69
70 int path_is_absolute(const char *path)
71 {
72 const char *p;
73 #ifdef _WIN32
74 /* specific case for names like: "\\.\d:" */
75 if (*path == '/' || *path == '\\')
76 return 1;
77 #endif
78 p = strchr(path, ':');
79 if (p)
80 p++;
81 else
82 p = path;
83 #ifdef _WIN32
84 return (*p == '/' || *p == '\\');
85 #else
86 return (*p == '/');
87 #endif
88 }
89
90 /* if filename is absolute, just copy it to dest. Otherwise, build a
91 path to it by considering it is relative to base_path. URL are
92 supported. */
93 void path_combine(char *dest, int dest_size,
94 const char *base_path,
95 const char *filename)
96 {
97 const char *p, *p1;
98 int len;
99
100 if (dest_size <= 0)
101 return;
102 if (path_is_absolute(filename)) {
103 pstrcpy(dest, dest_size, filename);
104 } else {
105 p = strchr(base_path, ':');
106 if (p)
107 p++;
108 else
109 p = base_path;
110 p1 = strrchr(base_path, '/');
111 #ifdef _WIN32
112 {
113 const char *p2;
114 p2 = strrchr(base_path, '\\');
115 if (!p1 || p2 > p1)
116 p1 = p2;
117 }
118 #endif
119 if (p1)
120 p1++;
121 else
122 p1 = base_path;
123 if (p1 > p)
124 p = p1;
125 len = p - base_path;
126 if (len > dest_size - 1)
127 len = dest_size - 1;
128 memcpy(dest, base_path, len);
129 dest[len] = '\0';
130 pstrcat(dest, dest_size, filename);
131 }
132 }
133
134 void bdrv_register(BlockDriver *bdrv)
135 {
136 if (!bdrv->bdrv_aio_readv) {
137 /* add AIO emulation layer */
138 bdrv->bdrv_aio_readv = bdrv_aio_readv_em;
139 bdrv->bdrv_aio_writev = bdrv_aio_writev_em;
140 } else if (!bdrv->bdrv_read) {
141 /* add synchronous IO emulation layer */
142 bdrv->bdrv_read = bdrv_read_em;
143 bdrv->bdrv_write = bdrv_write_em;
144 }
145
146 if (!bdrv->bdrv_aio_flush)
147 bdrv->bdrv_aio_flush = bdrv_aio_flush_em;
148
149 QLIST_INSERT_HEAD(&bdrv_drivers, bdrv, list);
150 }
151
152 /* create a new block device (by default it is empty) */
153 BlockDriverState *bdrv_new(const char *device_name)
154 {
155 BlockDriverState *bs;
156
157 bs = qemu_mallocz(sizeof(BlockDriverState));
158 pstrcpy(bs->device_name, sizeof(bs->device_name), device_name);
159 if (device_name[0] != '\0') {
160 QTAILQ_INSERT_TAIL(&bdrv_states, bs, list);
161 }
162 return bs;
163 }
164
165 BlockDriver *bdrv_find_format(const char *format_name)
166 {
167 BlockDriver *drv1;
168 QLIST_FOREACH(drv1, &bdrv_drivers, list) {
169 if (!strcmp(drv1->format_name, format_name)) {
170 return drv1;
171 }
172 }
173 return NULL;
174 }
175
176 static int bdrv_is_whitelisted(BlockDriver *drv)
177 {
178 static const char *whitelist[] = {
179 CONFIG_BDRV_WHITELIST
180 };
181 const char **p;
182
183 if (!whitelist[0])
184 return 1; /* no whitelist, anything goes */
185
186 for (p = whitelist; *p; p++) {
187 if (!strcmp(drv->format_name, *p)) {
188 return 1;
189 }
190 }
191 return 0;
192 }
193
194 BlockDriver *bdrv_find_whitelisted_format(const char *format_name)
195 {
196 BlockDriver *drv = bdrv_find_format(format_name);
197 return drv && bdrv_is_whitelisted(drv) ? drv : NULL;
198 }
199
200 int bdrv_create(BlockDriver *drv, const char* filename,
201 QEMUOptionParameter *options)
202 {
203 if (!drv->bdrv_create)
204 return -ENOTSUP;
205
206 return drv->bdrv_create(filename, options);
207 }
208
209 int bdrv_create_file(const char* filename, QEMUOptionParameter *options)
210 {
211 BlockDriver *drv;
212
213 drv = find_protocol(filename);
214 if (drv == NULL) {
215 drv = bdrv_find_format("file");
216 }
217
218 return bdrv_create(drv, filename, options);
219 }
220
221 #ifdef _WIN32
222 void get_tmp_filename(char *filename, int size)
223 {
224 char temp_dir[MAX_PATH];
225
226 GetTempPath(MAX_PATH, temp_dir);
227 GetTempFileName(temp_dir, "qem", 0, filename);
228 }
229 #else
230 void get_tmp_filename(char *filename, int size)
231 {
232 int fd;
233 const char *tmpdir;
234 /* XXX: race condition possible */
235 tmpdir = getenv("TMPDIR");
236 if (!tmpdir)
237 tmpdir = "/tmp";
238 snprintf(filename, size, "%s/vl.XXXXXX", tmpdir);
239 fd = mkstemp(filename);
240 close(fd);
241 }
242 #endif
243
244 #ifdef _WIN32
245 static int is_windows_drive_prefix(const char *filename)
246 {
247 return (((filename[0] >= 'a' && filename[0] <= 'z') ||
248 (filename[0] >= 'A' && filename[0] <= 'Z')) &&
249 filename[1] == ':');
250 }
251
252 int is_windows_drive(const char *filename)
253 {
254 if (is_windows_drive_prefix(filename) &&
255 filename[2] == '\0')
256 return 1;
257 if (strstart(filename, "\\\\.\\", NULL) ||
258 strstart(filename, "//./", NULL))
259 return 1;
260 return 0;
261 }
262 #endif
263
264 /*
265 * Detect host devices. By convention, /dev/cdrom[N] is always
266 * recognized as a host CDROM.
267 */
268 static BlockDriver *find_hdev_driver(const char *filename)
269 {
270 int score_max = 0, score;
271 BlockDriver *drv = NULL, *d;
272
273 QLIST_FOREACH(d, &bdrv_drivers, list) {
274 if (d->bdrv_probe_device) {
275 score = d->bdrv_probe_device(filename);
276 if (score > score_max) {
277 score_max = score;
278 drv = d;
279 }
280 }
281 }
282
283 return drv;
284 }
285
286 static BlockDriver *find_protocol(const char *filename)
287 {
288 BlockDriver *drv1;
289 char protocol[128];
290 int len;
291 const char *p;
292 int is_drive;
293
294 /* TODO Drivers without bdrv_file_open must be specified explicitly */
295
296 #ifdef _WIN32
297 is_drive = is_windows_drive(filename) ||
298 is_windows_drive_prefix(filename);
299 #else
300 is_drive = 0;
301 #endif
302 p = strchr(filename, ':');
303 if (!p || is_drive) {
304 drv1 = find_hdev_driver(filename);
305 if (!drv1) {
306 drv1 = bdrv_find_format("file");
307 }
308 return drv1;
309 }
310 len = p - filename;
311 if (len > sizeof(protocol) - 1)
312 len = sizeof(protocol) - 1;
313 memcpy(protocol, filename, len);
314 protocol[len] = '\0';
315 QLIST_FOREACH(drv1, &bdrv_drivers, list) {
316 if (drv1->protocol_name &&
317 !strcmp(drv1->protocol_name, protocol)) {
318 return drv1;
319 }
320 }
321 return NULL;
322 }
323
324 static BlockDriver *find_image_format(const char *filename)
325 {
326 int ret, score, score_max;
327 BlockDriver *drv1, *drv;
328 uint8_t buf[2048];
329 BlockDriverState *bs;
330
331 ret = bdrv_file_open(&bs, filename, 0);
332 if (ret < 0)
333 return NULL;
334
335 /* Return the raw BlockDriver * to scsi-generic devices */
336 if (bs->sg)
337 return bdrv_find_format("raw");
338
339 ret = bdrv_pread(bs, 0, buf, sizeof(buf));
340 bdrv_delete(bs);
341 if (ret < 0) {
342 return NULL;
343 }
344
345 score_max = 0;
346 drv = NULL;
347 QLIST_FOREACH(drv1, &bdrv_drivers, list) {
348 if (drv1->bdrv_probe) {
349 score = drv1->bdrv_probe(buf, ret, filename);
350 if (score > score_max) {
351 score_max = score;
352 drv = drv1;
353 }
354 }
355 }
356 return drv;
357 }
358
359 /**
360 * Set the current 'total_sectors' value
361 */
362 static int refresh_total_sectors(BlockDriverState *bs, int64_t hint)
363 {
364 BlockDriver *drv = bs->drv;
365
366 /* Do not attempt drv->bdrv_getlength() on scsi-generic devices */
367 if (bs->sg)
368 return 0;
369
370 /* query actual device if possible, otherwise just trust the hint */
371 if (drv->bdrv_getlength) {
372 int64_t length = drv->bdrv_getlength(bs);
373 if (length < 0) {
374 return length;
375 }
376 hint = length >> BDRV_SECTOR_BITS;
377 }
378
379 bs->total_sectors = hint;
380 return 0;
381 }
382
383 /*
384 * Common part for opening disk images and files
385 */
386 static int bdrv_open_common(BlockDriverState *bs, const char *filename,
387 int flags, BlockDriver *drv)
388 {
389 int ret, open_flags;
390
391 assert(drv != NULL);
392
393 bs->file = NULL;
394 bs->total_sectors = 0;
395 bs->is_temporary = 0;
396 bs->encrypted = 0;
397 bs->valid_key = 0;
398 bs->open_flags = flags;
399 /* buffer_alignment defaulted to 512, drivers can change this value */
400 bs->buffer_alignment = 512;
401
402 pstrcpy(bs->filename, sizeof(bs->filename), filename);
403
404 if (use_bdrv_whitelist && !bdrv_is_whitelisted(drv)) {
405 return -ENOTSUP;
406 }
407
408 bs->drv = drv;
409 bs->opaque = qemu_mallocz(drv->instance_size);
410
411 /*
412 * Yes, BDRV_O_NOCACHE aka O_DIRECT means we have to present a
413 * write cache to the guest. We do need the fdatasync to flush
414 * out transactions for block allocations, and we maybe have a
415 * volatile write cache in our backing device to deal with.
416 */
417 if (flags & (BDRV_O_CACHE_WB|BDRV_O_NOCACHE))
418 bs->enable_write_cache = 1;
419
420 /*
421 * Clear flags that are internal to the block layer before opening the
422 * image.
423 */
424 open_flags = flags & ~(BDRV_O_SNAPSHOT | BDRV_O_NO_BACKING);
425
426 /*
427 * Snapshots should be writeable.
428 */
429 if (bs->is_temporary) {
430 open_flags |= BDRV_O_RDWR;
431 }
432
433 /* Open the image, either directly or using a protocol */
434 if (drv->bdrv_file_open) {
435 ret = drv->bdrv_file_open(bs, filename, open_flags);
436 } else {
437 ret = bdrv_file_open(&bs->file, filename, open_flags);
438 if (ret >= 0) {
439 ret = drv->bdrv_open(bs, open_flags);
440 }
441 }
442
443 if (ret < 0) {
444 goto free_and_fail;
445 }
446
447 bs->keep_read_only = bs->read_only = !(open_flags & BDRV_O_RDWR);
448
449 ret = refresh_total_sectors(bs, bs->total_sectors);
450 if (ret < 0) {
451 goto free_and_fail;
452 }
453
454 #ifndef _WIN32
455 if (bs->is_temporary) {
456 unlink(filename);
457 }
458 #endif
459 return 0;
460
461 free_and_fail:
462 if (bs->file) {
463 bdrv_delete(bs->file);
464 bs->file = NULL;
465 }
466 qemu_free(bs->opaque);
467 bs->opaque = NULL;
468 bs->drv = NULL;
469 return ret;
470 }
471
472 /*
473 * Opens a file using a protocol (file, host_device, nbd, ...)
474 */
475 int bdrv_file_open(BlockDriverState **pbs, const char *filename, int flags)
476 {
477 BlockDriverState *bs;
478 BlockDriver *drv;
479 int ret;
480
481 drv = find_protocol(filename);
482 if (!drv) {
483 return -ENOENT;
484 }
485
486 bs = bdrv_new("");
487 ret = bdrv_open_common(bs, filename, flags, drv);
488 if (ret < 0) {
489 bdrv_delete(bs);
490 return ret;
491 }
492 bs->growable = 1;
493 *pbs = bs;
494 return 0;
495 }
496
497 /*
498 * Opens a disk image (raw, qcow2, vmdk, ...)
499 */
500 int bdrv_open(BlockDriverState *bs, const char *filename, int flags,
501 BlockDriver *drv)
502 {
503 int ret;
504
505 if (flags & BDRV_O_SNAPSHOT) {
506 BlockDriverState *bs1;
507 int64_t total_size;
508 int is_protocol = 0;
509 BlockDriver *bdrv_qcow2;
510 QEMUOptionParameter *options;
511 char tmp_filename[PATH_MAX];
512 char backing_filename[PATH_MAX];
513
514 /* if snapshot, we create a temporary backing file and open it
515 instead of opening 'filename' directly */
516
517 /* if there is a backing file, use it */
518 bs1 = bdrv_new("");
519 ret = bdrv_open(bs1, filename, 0, drv);
520 if (ret < 0) {
521 bdrv_delete(bs1);
522 return ret;
523 }
524 total_size = bdrv_getlength(bs1) >> BDRV_SECTOR_BITS;
525
526 if (bs1->drv && bs1->drv->protocol_name)
527 is_protocol = 1;
528
529 bdrv_delete(bs1);
530
531 get_tmp_filename(tmp_filename, sizeof(tmp_filename));
532
533 /* Real path is meaningless for protocols */
534 if (is_protocol)
535 snprintf(backing_filename, sizeof(backing_filename),
536 "%s", filename);
537 else if (!realpath(filename, backing_filename))
538 return -errno;
539
540 bdrv_qcow2 = bdrv_find_format("qcow2");
541 options = parse_option_parameters("", bdrv_qcow2->create_options, NULL);
542
543 set_option_parameter_int(options, BLOCK_OPT_SIZE, total_size * 512);
544 set_option_parameter(options, BLOCK_OPT_BACKING_FILE, backing_filename);
545 if (drv) {
546 set_option_parameter(options, BLOCK_OPT_BACKING_FMT,
547 drv->format_name);
548 }
549
550 ret = bdrv_create(bdrv_qcow2, tmp_filename, options);
551 free_option_parameters(options);
552 if (ret < 0) {
553 return ret;
554 }
555
556 filename = tmp_filename;
557 drv = bdrv_qcow2;
558 bs->is_temporary = 1;
559 }
560
561 /* Find the right image format driver */
562 if (!drv) {
563 drv = find_image_format(filename);
564 }
565
566 if (!drv) {
567 ret = -ENOENT;
568 goto unlink_and_fail;
569 }
570
571 /* Open the image */
572 ret = bdrv_open_common(bs, filename, flags, drv);
573 if (ret < 0) {
574 goto unlink_and_fail;
575 }
576
577 /* If there is a backing file, use it */
578 if ((flags & BDRV_O_NO_BACKING) == 0 && bs->backing_file[0] != '\0') {
579 char backing_filename[PATH_MAX];
580 int back_flags;
581 BlockDriver *back_drv = NULL;
582
583 bs->backing_hd = bdrv_new("");
584 path_combine(backing_filename, sizeof(backing_filename),
585 filename, bs->backing_file);
586 if (bs->backing_format[0] != '\0')
587 back_drv = bdrv_find_format(bs->backing_format);
588
589 /* backing files always opened read-only */
590 back_flags =
591 flags & ~(BDRV_O_RDWR | BDRV_O_SNAPSHOT | BDRV_O_NO_BACKING);
592
593 ret = bdrv_open(bs->backing_hd, backing_filename, back_flags, back_drv);
594 if (ret < 0) {
595 bdrv_close(bs);
596 return ret;
597 }
598 if (bs->is_temporary) {
599 bs->backing_hd->keep_read_only = !(flags & BDRV_O_RDWR);
600 } else {
601 /* base image inherits from "parent" */
602 bs->backing_hd->keep_read_only = bs->keep_read_only;
603 }
604 }
605
606 if (!bdrv_key_required(bs)) {
607 /* call the change callback */
608 bs->media_changed = 1;
609 if (bs->change_cb)
610 bs->change_cb(bs->change_opaque);
611 }
612
613 return 0;
614
615 unlink_and_fail:
616 if (bs->is_temporary) {
617 unlink(filename);
618 }
619 return ret;
620 }
621
622 void bdrv_close(BlockDriverState *bs)
623 {
624 if (bs->drv) {
625 if (bs->backing_hd) {
626 bdrv_delete(bs->backing_hd);
627 bs->backing_hd = NULL;
628 }
629 bs->drv->bdrv_close(bs);
630 qemu_free(bs->opaque);
631 #ifdef _WIN32
632 if (bs->is_temporary) {
633 unlink(bs->filename);
634 }
635 #endif
636 bs->opaque = NULL;
637 bs->drv = NULL;
638
639 if (bs->file != NULL) {
640 bdrv_close(bs->file);
641 }
642
643 /* call the change callback */
644 bs->media_changed = 1;
645 if (bs->change_cb)
646 bs->change_cb(bs->change_opaque);
647 }
648 }
649
650 void bdrv_delete(BlockDriverState *bs)
651 {
652 /* remove from list, if necessary */
653 if (bs->device_name[0] != '\0') {
654 QTAILQ_REMOVE(&bdrv_states, bs, list);
655 }
656
657 bdrv_close(bs);
658 if (bs->file != NULL) {
659 bdrv_delete(bs->file);
660 }
661
662 qemu_free(bs);
663 }
664
665 /*
666 * Run consistency checks on an image
667 *
668 * Returns the number of errors or -errno when an internal error occurs
669 */
670 int bdrv_check(BlockDriverState *bs)
671 {
672 if (bs->drv->bdrv_check == NULL) {
673 return -ENOTSUP;
674 }
675
676 return bs->drv->bdrv_check(bs);
677 }
678
679 /* commit COW file into the raw image */
680 int bdrv_commit(BlockDriverState *bs)
681 {
682 BlockDriver *drv = bs->drv;
683 int64_t i, total_sectors;
684 int n, j, ro, open_flags;
685 int ret = 0, rw_ret = 0;
686 unsigned char sector[512];
687 char filename[1024];
688 BlockDriverState *bs_rw, *bs_ro;
689
690 if (!drv)
691 return -ENOMEDIUM;
692
693 if (!bs->backing_hd) {
694 return -ENOTSUP;
695 }
696
697 if (bs->backing_hd->keep_read_only) {
698 return -EACCES;
699 }
700
701 ro = bs->backing_hd->read_only;
702 strncpy(filename, bs->backing_hd->filename, sizeof(filename));
703 open_flags = bs->backing_hd->open_flags;
704
705 if (ro) {
706 /* re-open as RW */
707 bdrv_delete(bs->backing_hd);
708 bs->backing_hd = NULL;
709 bs_rw = bdrv_new("");
710 rw_ret = bdrv_open(bs_rw, filename, open_flags | BDRV_O_RDWR, drv);
711 if (rw_ret < 0) {
712 bdrv_delete(bs_rw);
713 /* try to re-open read-only */
714 bs_ro = bdrv_new("");
715 ret = bdrv_open(bs_ro, filename, open_flags & ~BDRV_O_RDWR, drv);
716 if (ret < 0) {
717 bdrv_delete(bs_ro);
718 /* drive not functional anymore */
719 bs->drv = NULL;
720 return ret;
721 }
722 bs->backing_hd = bs_ro;
723 return rw_ret;
724 }
725 bs->backing_hd = bs_rw;
726 }
727
728 total_sectors = bdrv_getlength(bs) >> BDRV_SECTOR_BITS;
729 for (i = 0; i < total_sectors;) {
730 if (drv->bdrv_is_allocated(bs, i, 65536, &n)) {
731 for(j = 0; j < n; j++) {
732 if (bdrv_read(bs, i, sector, 1) != 0) {
733 ret = -EIO;
734 goto ro_cleanup;
735 }
736
737 if (bdrv_write(bs->backing_hd, i, sector, 1) != 0) {
738 ret = -EIO;
739 goto ro_cleanup;
740 }
741 i++;
742 }
743 } else {
744 i += n;
745 }
746 }
747
748 if (drv->bdrv_make_empty) {
749 ret = drv->bdrv_make_empty(bs);
750 bdrv_flush(bs);
751 }
752
753 /*
754 * Make sure all data we wrote to the backing device is actually
755 * stable on disk.
756 */
757 if (bs->backing_hd)
758 bdrv_flush(bs->backing_hd);
759
760 ro_cleanup:
761
762 if (ro) {
763 /* re-open as RO */
764 bdrv_delete(bs->backing_hd);
765 bs->backing_hd = NULL;
766 bs_ro = bdrv_new("");
767 ret = bdrv_open(bs_ro, filename, open_flags & ~BDRV_O_RDWR, drv);
768 if (ret < 0) {
769 bdrv_delete(bs_ro);
770 /* drive not functional anymore */
771 bs->drv = NULL;
772 return ret;
773 }
774 bs->backing_hd = bs_ro;
775 bs->backing_hd->keep_read_only = 0;
776 }
777
778 return ret;
779 }
780
781 /*
782 * Return values:
783 * 0 - success
784 * -EINVAL - backing format specified, but no file
785 * -ENOSPC - can't update the backing file because no space is left in the
786 * image file header
787 * -ENOTSUP - format driver doesn't support changing the backing file
788 */
789 int bdrv_change_backing_file(BlockDriverState *bs,
790 const char *backing_file, const char *backing_fmt)
791 {
792 BlockDriver *drv = bs->drv;
793
794 if (drv->bdrv_change_backing_file != NULL) {
795 return drv->bdrv_change_backing_file(bs, backing_file, backing_fmt);
796 } else {
797 return -ENOTSUP;
798 }
799 }
800
801 static int bdrv_check_byte_request(BlockDriverState *bs, int64_t offset,
802 size_t size)
803 {
804 int64_t len;
805
806 if (!bdrv_is_inserted(bs))
807 return -ENOMEDIUM;
808
809 if (bs->growable)
810 return 0;
811
812 len = bdrv_getlength(bs);
813
814 if (offset < 0)
815 return -EIO;
816
817 if ((offset > len) || (len - offset < size))
818 return -EIO;
819
820 return 0;
821 }
822
823 static int bdrv_check_request(BlockDriverState *bs, int64_t sector_num,
824 int nb_sectors)
825 {
826 return bdrv_check_byte_request(bs, sector_num * 512, nb_sectors * 512);
827 }
828
829 /* return < 0 if error. See bdrv_write() for the return codes */
830 int bdrv_read(BlockDriverState *bs, int64_t sector_num,
831 uint8_t *buf, int nb_sectors)
832 {
833 BlockDriver *drv = bs->drv;
834
835 if (!drv)
836 return -ENOMEDIUM;
837 if (bdrv_check_request(bs, sector_num, nb_sectors))
838 return -EIO;
839
840 return drv->bdrv_read(bs, sector_num, buf, nb_sectors);
841 }
842
843 static void set_dirty_bitmap(BlockDriverState *bs, int64_t sector_num,
844 int nb_sectors, int dirty)
845 {
846 int64_t start, end;
847 unsigned long val, idx, bit;
848
849 start = sector_num / BDRV_SECTORS_PER_DIRTY_CHUNK;
850 end = (sector_num + nb_sectors - 1) / BDRV_SECTORS_PER_DIRTY_CHUNK;
851
852 for (; start <= end; start++) {
853 idx = start / (sizeof(unsigned long) * 8);
854 bit = start % (sizeof(unsigned long) * 8);
855 val = bs->dirty_bitmap[idx];
856 if (dirty) {
857 if (!(val & (1 << bit))) {
858 bs->dirty_count++;
859 val |= 1 << bit;
860 }
861 } else {
862 if (val & (1 << bit)) {
863 bs->dirty_count--;
864 val &= ~(1 << bit);
865 }
866 }
867 bs->dirty_bitmap[idx] = val;
868 }
869 }
870
871 /* Return < 0 if error. Important errors are:
872 -EIO generic I/O error (may happen for all errors)
873 -ENOMEDIUM No media inserted.
874 -EINVAL Invalid sector number or nb_sectors
875 -EACCES Trying to write a read-only device
876 */
877 int bdrv_write(BlockDriverState *bs, int64_t sector_num,
878 const uint8_t *buf, int nb_sectors)
879 {
880 BlockDriver *drv = bs->drv;
881 if (!bs->drv)
882 return -ENOMEDIUM;
883 if (bs->read_only)
884 return -EACCES;
885 if (bdrv_check_request(bs, sector_num, nb_sectors))
886 return -EIO;
887
888 if (bs->dirty_bitmap) {
889 set_dirty_bitmap(bs, sector_num, nb_sectors, 1);
890 }
891
892 if (bs->wr_highest_sector < sector_num + nb_sectors - 1) {
893 bs->wr_highest_sector = sector_num + nb_sectors - 1;
894 }
895
896 return drv->bdrv_write(bs, sector_num, buf, nb_sectors);
897 }
898
899 int bdrv_pread(BlockDriverState *bs, int64_t offset,
900 void *buf, int count1)
901 {
902 uint8_t tmp_buf[BDRV_SECTOR_SIZE];
903 int len, nb_sectors, count;
904 int64_t sector_num;
905 int ret;
906
907 count = count1;
908 /* first read to align to sector start */
909 len = (BDRV_SECTOR_SIZE - offset) & (BDRV_SECTOR_SIZE - 1);
910 if (len > count)
911 len = count;
912 sector_num = offset >> BDRV_SECTOR_BITS;
913 if (len > 0) {
914 if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0)
915 return ret;
916 memcpy(buf, tmp_buf + (offset & (BDRV_SECTOR_SIZE - 1)), len);
917 count -= len;
918 if (count == 0)
919 return count1;
920 sector_num++;
921 buf += len;
922 }
923
924 /* read the sectors "in place" */
925 nb_sectors = count >> BDRV_SECTOR_BITS;
926 if (nb_sectors > 0) {
927 if ((ret = bdrv_read(bs, sector_num, buf, nb_sectors)) < 0)
928 return ret;
929 sector_num += nb_sectors;
930 len = nb_sectors << BDRV_SECTOR_BITS;
931 buf += len;
932 count -= len;
933 }
934
935 /* add data from the last sector */
936 if (count > 0) {
937 if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0)
938 return ret;
939 memcpy(buf, tmp_buf, count);
940 }
941 return count1;
942 }
943
944 int bdrv_pwrite(BlockDriverState *bs, int64_t offset,
945 const void *buf, int count1)
946 {
947 uint8_t tmp_buf[BDRV_SECTOR_SIZE];
948 int len, nb_sectors, count;
949 int64_t sector_num;
950 int ret;
951
952 count = count1;
953 /* first write to align to sector start */
954 len = (BDRV_SECTOR_SIZE - offset) & (BDRV_SECTOR_SIZE - 1);
955 if (len > count)
956 len = count;
957 sector_num = offset >> BDRV_SECTOR_BITS;
958 if (len > 0) {
959 if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0)
960 return ret;
961 memcpy(tmp_buf + (offset & (BDRV_SECTOR_SIZE - 1)), buf, len);
962 if ((ret = bdrv_write(bs, sector_num, tmp_buf, 1)) < 0)
963 return ret;
964 count -= len;
965 if (count == 0)
966 return count1;
967 sector_num++;
968 buf += len;
969 }
970
971 /* write the sectors "in place" */
972 nb_sectors = count >> BDRV_SECTOR_BITS;
973 if (nb_sectors > 0) {
974 if ((ret = bdrv_write(bs, sector_num, buf, nb_sectors)) < 0)
975 return ret;
976 sector_num += nb_sectors;
977 len = nb_sectors << BDRV_SECTOR_BITS;
978 buf += len;
979 count -= len;
980 }
981
982 /* add data from the last sector */
983 if (count > 0) {
984 if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0)
985 return ret;
986 memcpy(tmp_buf, buf, count);
987 if ((ret = bdrv_write(bs, sector_num, tmp_buf, 1)) < 0)
988 return ret;
989 }
990 return count1;
991 }
992
993 /**
994 * Truncate file to 'offset' bytes (needed only for file protocols)
995 */
996 int bdrv_truncate(BlockDriverState *bs, int64_t offset)
997 {
998 BlockDriver *drv = bs->drv;
999 int ret;
1000 if (!drv)
1001 return -ENOMEDIUM;
1002 if (!drv->bdrv_truncate)
1003 return -ENOTSUP;
1004 if (bs->read_only)
1005 return -EACCES;
1006 ret = drv->bdrv_truncate(bs, offset);
1007 if (ret == 0) {
1008 ret = refresh_total_sectors(bs, offset >> BDRV_SECTOR_BITS);
1009 }
1010 return ret;
1011 }
1012
1013 /**
1014 * Length of a file in bytes. Return < 0 if error or unknown.
1015 */
1016 int64_t bdrv_getlength(BlockDriverState *bs)
1017 {
1018 BlockDriver *drv = bs->drv;
1019 if (!drv)
1020 return -ENOMEDIUM;
1021
1022 /* Fixed size devices use the total_sectors value for speed instead of
1023 issuing a length query (like lseek) on each call. Also, legacy block
1024 drivers don't provide a bdrv_getlength function and must use
1025 total_sectors. */
1026 if (!bs->growable || !drv->bdrv_getlength) {
1027 return bs->total_sectors * BDRV_SECTOR_SIZE;
1028 }
1029 return drv->bdrv_getlength(bs);
1030 }
1031
1032 /* return 0 as number of sectors if no device present or error */
1033 void bdrv_get_geometry(BlockDriverState *bs, uint64_t *nb_sectors_ptr)
1034 {
1035 int64_t length;
1036 length = bdrv_getlength(bs);
1037 if (length < 0)
1038 length = 0;
1039 else
1040 length = length >> BDRV_SECTOR_BITS;
1041 *nb_sectors_ptr = length;
1042 }
1043
1044 struct partition {
1045 uint8_t boot_ind; /* 0x80 - active */
1046 uint8_t head; /* starting head */
1047 uint8_t sector; /* starting sector */
1048 uint8_t cyl; /* starting cylinder */
1049 uint8_t sys_ind; /* What partition type */
1050 uint8_t end_head; /* end head */
1051 uint8_t end_sector; /* end sector */
1052 uint8_t end_cyl; /* end cylinder */
1053 uint32_t start_sect; /* starting sector counting from 0 */
1054 uint32_t nr_sects; /* nr of sectors in partition */
1055 } __attribute__((packed));
1056
1057 /* try to guess the disk logical geometry from the MSDOS partition table. Return 0 if OK, -1 if could not guess */
1058 static int guess_disk_lchs(BlockDriverState *bs,
1059 int *pcylinders, int *pheads, int *psectors)
1060 {
1061 uint8_t buf[512];
1062 int ret, i, heads, sectors, cylinders;
1063 struct partition *p;
1064 uint32_t nr_sects;
1065 uint64_t nb_sectors;
1066
1067 bdrv_get_geometry(bs, &nb_sectors);
1068
1069 ret = bdrv_read(bs, 0, buf, 1);
1070 if (ret < 0)
1071 return -1;
1072 /* test msdos magic */
1073 if (buf[510] != 0x55 || buf[511] != 0xaa)
1074 return -1;
1075 for(i = 0; i < 4; i++) {
1076 p = ((struct partition *)(buf + 0x1be)) + i;
1077 nr_sects = le32_to_cpu(p->nr_sects);
1078 if (nr_sects && p->end_head) {
1079 /* We make the assumption that the partition terminates on
1080 a cylinder boundary */
1081 heads = p->end_head + 1;
1082 sectors = p->end_sector & 63;
1083 if (sectors == 0)
1084 continue;
1085 cylinders = nb_sectors / (heads * sectors);
1086 if (cylinders < 1 || cylinders > 16383)
1087 continue;
1088 *pheads = heads;
1089 *psectors = sectors;
1090 *pcylinders = cylinders;
1091 #if 0
1092 printf("guessed geometry: LCHS=%d %d %d\n",
1093 cylinders, heads, sectors);
1094 #endif
1095 return 0;
1096 }
1097 }
1098 return -1;
1099 }
1100
1101 void bdrv_guess_geometry(BlockDriverState *bs, int *pcyls, int *pheads, int *psecs)
1102 {
1103 int translation, lba_detected = 0;
1104 int cylinders, heads, secs;
1105 uint64_t nb_sectors;
1106
1107 /* if a geometry hint is available, use it */
1108 bdrv_get_geometry(bs, &nb_sectors);
1109 bdrv_get_geometry_hint(bs, &cylinders, &heads, &secs);
1110 translation = bdrv_get_translation_hint(bs);
1111 if (cylinders != 0) {
1112 *pcyls = cylinders;
1113 *pheads = heads;
1114 *psecs = secs;
1115 } else {
1116 if (guess_disk_lchs(bs, &cylinders, &heads, &secs) == 0) {
1117 if (heads > 16) {
1118 /* if heads > 16, it means that a BIOS LBA
1119 translation was active, so the default
1120 hardware geometry is OK */
1121 lba_detected = 1;
1122 goto default_geometry;
1123 } else {
1124 *pcyls = cylinders;
1125 *pheads = heads;
1126 *psecs = secs;
1127 /* disable any translation to be in sync with
1128 the logical geometry */
1129 if (translation == BIOS_ATA_TRANSLATION_AUTO) {
1130 bdrv_set_translation_hint(bs,
1131 BIOS_ATA_TRANSLATION_NONE);
1132 }
1133 }
1134 } else {
1135 default_geometry:
1136 /* if no geometry, use a standard physical disk geometry */
1137 cylinders = nb_sectors / (16 * 63);
1138
1139 if (cylinders > 16383)
1140 cylinders = 16383;
1141 else if (cylinders < 2)
1142 cylinders = 2;
1143 *pcyls = cylinders;
1144 *pheads = 16;
1145 *psecs = 63;
1146 if ((lba_detected == 1) && (translation == BIOS_ATA_TRANSLATION_AUTO)) {
1147 if ((*pcyls * *pheads) <= 131072) {
1148 bdrv_set_translation_hint(bs,
1149 BIOS_ATA_TRANSLATION_LARGE);
1150 } else {
1151 bdrv_set_translation_hint(bs,
1152 BIOS_ATA_TRANSLATION_LBA);
1153 }
1154 }
1155 }
1156 bdrv_set_geometry_hint(bs, *pcyls, *pheads, *psecs);
1157 }
1158 }
1159
1160 void bdrv_set_geometry_hint(BlockDriverState *bs,
1161 int cyls, int heads, int secs)
1162 {
1163 bs->cyls = cyls;
1164 bs->heads = heads;
1165 bs->secs = secs;
1166 }
1167
1168 void bdrv_set_type_hint(BlockDriverState *bs, int type)
1169 {
1170 bs->type = type;
1171 bs->removable = ((type == BDRV_TYPE_CDROM ||
1172 type == BDRV_TYPE_FLOPPY));
1173 }
1174
1175 void bdrv_set_translation_hint(BlockDriverState *bs, int translation)
1176 {
1177 bs->translation = translation;
1178 }
1179
1180 void bdrv_get_geometry_hint(BlockDriverState *bs,
1181 int *pcyls, int *pheads, int *psecs)
1182 {
1183 *pcyls = bs->cyls;
1184 *pheads = bs->heads;
1185 *psecs = bs->secs;
1186 }
1187
1188 int bdrv_get_type_hint(BlockDriverState *bs)
1189 {
1190 return bs->type;
1191 }
1192
1193 int bdrv_get_translation_hint(BlockDriverState *bs)
1194 {
1195 return bs->translation;
1196 }
1197
1198 int bdrv_is_removable(BlockDriverState *bs)
1199 {
1200 return bs->removable;
1201 }
1202
1203 int bdrv_is_read_only(BlockDriverState *bs)
1204 {
1205 return bs->read_only;
1206 }
1207
1208 int bdrv_is_sg(BlockDriverState *bs)
1209 {
1210 return bs->sg;
1211 }
1212
1213 int bdrv_enable_write_cache(BlockDriverState *bs)
1214 {
1215 return bs->enable_write_cache;
1216 }
1217
1218 /* XXX: no longer used */
1219 void bdrv_set_change_cb(BlockDriverState *bs,
1220 void (*change_cb)(void *opaque), void *opaque)
1221 {
1222 bs->change_cb = change_cb;
1223 bs->change_opaque = opaque;
1224 }
1225
1226 int bdrv_is_encrypted(BlockDriverState *bs)
1227 {
1228 if (bs->backing_hd && bs->backing_hd->encrypted)
1229 return 1;
1230 return bs->encrypted;
1231 }
1232
1233 int bdrv_key_required(BlockDriverState *bs)
1234 {
1235 BlockDriverState *backing_hd = bs->backing_hd;
1236
1237 if (backing_hd && backing_hd->encrypted && !backing_hd->valid_key)
1238 return 1;
1239 return (bs->encrypted && !bs->valid_key);
1240 }
1241
1242 int bdrv_set_key(BlockDriverState *bs, const char *key)
1243 {
1244 int ret;
1245 if (bs->backing_hd && bs->backing_hd->encrypted) {
1246 ret = bdrv_set_key(bs->backing_hd, key);
1247 if (ret < 0)
1248 return ret;
1249 if (!bs->encrypted)
1250 return 0;
1251 }
1252 if (!bs->encrypted) {
1253 return -EINVAL;
1254 } else if (!bs->drv || !bs->drv->bdrv_set_key) {
1255 return -ENOMEDIUM;
1256 }
1257 ret = bs->drv->bdrv_set_key(bs, key);
1258 if (ret < 0) {
1259 bs->valid_key = 0;
1260 } else if (!bs->valid_key) {
1261 bs->valid_key = 1;
1262 /* call the change callback now, we skipped it on open */
1263 bs->media_changed = 1;
1264 if (bs->change_cb)
1265 bs->change_cb(bs->change_opaque);
1266 }
1267 return ret;
1268 }
1269
1270 void bdrv_get_format(BlockDriverState *bs, char *buf, int buf_size)
1271 {
1272 if (!bs->drv) {
1273 buf[0] = '\0';
1274 } else {
1275 pstrcpy(buf, buf_size, bs->drv->format_name);
1276 }
1277 }
1278
1279 void bdrv_iterate_format(void (*it)(void *opaque, const char *name),
1280 void *opaque)
1281 {
1282 BlockDriver *drv;
1283
1284 QLIST_FOREACH(drv, &bdrv_drivers, list) {
1285 it(opaque, drv->format_name);
1286 }
1287 }
1288
1289 BlockDriverState *bdrv_find(const char *name)
1290 {
1291 BlockDriverState *bs;
1292
1293 QTAILQ_FOREACH(bs, &bdrv_states, list) {
1294 if (!strcmp(name, bs->device_name)) {
1295 return bs;
1296 }
1297 }
1298 return NULL;
1299 }
1300
1301 void bdrv_iterate(void (*it)(void *opaque, BlockDriverState *bs), void *opaque)
1302 {
1303 BlockDriverState *bs;
1304
1305 QTAILQ_FOREACH(bs, &bdrv_states, list) {
1306 it(opaque, bs);
1307 }
1308 }
1309
1310 const char *bdrv_get_device_name(BlockDriverState *bs)
1311 {
1312 return bs->device_name;
1313 }
1314
1315 void bdrv_flush(BlockDriverState *bs)
1316 {
1317 if (bs->open_flags & BDRV_O_NO_FLUSH) {
1318 return;
1319 }
1320
1321 if (bs->drv && bs->drv->bdrv_flush)
1322 bs->drv->bdrv_flush(bs);
1323 }
1324
1325 void bdrv_flush_all(void)
1326 {
1327 BlockDriverState *bs;
1328
1329 QTAILQ_FOREACH(bs, &bdrv_states, list) {
1330 if (bs->drv && !bdrv_is_read_only(bs) &&
1331 (!bdrv_is_removable(bs) || bdrv_is_inserted(bs))) {
1332 bdrv_flush(bs);
1333 }
1334 }
1335 }
1336
1337 int bdrv_has_zero_init(BlockDriverState *bs)
1338 {
1339 assert(bs->drv);
1340
1341 if (bs->drv->no_zero_init) {
1342 return 0;
1343 } else if (bs->file) {
1344 return bdrv_has_zero_init(bs->file);
1345 }
1346
1347 return 1;
1348 }
1349
1350 /*
1351 * Returns true iff the specified sector is present in the disk image. Drivers
1352 * not implementing the functionality are assumed to not support backing files,
1353 * hence all their sectors are reported as allocated.
1354 *
1355 * 'pnum' is set to the number of sectors (including and immediately following
1356 * the specified sector) that are known to be in the same
1357 * allocated/unallocated state.
1358 *
1359 * 'nb_sectors' is the max value 'pnum' should be set to.
1360 */
1361 int bdrv_is_allocated(BlockDriverState *bs, int64_t sector_num, int nb_sectors,
1362 int *pnum)
1363 {
1364 int64_t n;
1365 if (!bs->drv->bdrv_is_allocated) {
1366 if (sector_num >= bs->total_sectors) {
1367 *pnum = 0;
1368 return 0;
1369 }
1370 n = bs->total_sectors - sector_num;
1371 *pnum = (n < nb_sectors) ? (n) : (nb_sectors);
1372 return 1;
1373 }
1374 return bs->drv->bdrv_is_allocated(bs, sector_num, nb_sectors, pnum);
1375 }
1376
1377 void bdrv_mon_event(const BlockDriverState *bdrv,
1378 BlockMonEventAction action, int is_read)
1379 {
1380 QObject *data;
1381 const char *action_str;
1382
1383 switch (action) {
1384 case BDRV_ACTION_REPORT:
1385 action_str = "report";
1386 break;
1387 case BDRV_ACTION_IGNORE:
1388 action_str = "ignore";
1389 break;
1390 case BDRV_ACTION_STOP:
1391 action_str = "stop";
1392 break;
1393 default:
1394 abort();
1395 }
1396
1397 data = qobject_from_jsonf("{ 'device': %s, 'action': %s, 'operation': %s }",
1398 bdrv->device_name,
1399 action_str,
1400 is_read ? "read" : "write");
1401 monitor_protocol_event(QEVENT_BLOCK_IO_ERROR, data);
1402
1403 qobject_decref(data);
1404 }
1405
1406 static void bdrv_print_dict(QObject *obj, void *opaque)
1407 {
1408 QDict *bs_dict;
1409 Monitor *mon = opaque;
1410
1411 bs_dict = qobject_to_qdict(obj);
1412
1413 monitor_printf(mon, "%s: type=%s removable=%d",
1414 qdict_get_str(bs_dict, "device"),
1415 qdict_get_str(bs_dict, "type"),
1416 qdict_get_bool(bs_dict, "removable"));
1417
1418 if (qdict_get_bool(bs_dict, "removable")) {
1419 monitor_printf(mon, " locked=%d", qdict_get_bool(bs_dict, "locked"));
1420 }
1421
1422 if (qdict_haskey(bs_dict, "inserted")) {
1423 QDict *qdict = qobject_to_qdict(qdict_get(bs_dict, "inserted"));
1424
1425 monitor_printf(mon, " file=");
1426 monitor_print_filename(mon, qdict_get_str(qdict, "file"));
1427 if (qdict_haskey(qdict, "backing_file")) {
1428 monitor_printf(mon, " backing_file=");
1429 monitor_print_filename(mon, qdict_get_str(qdict, "backing_file"));
1430 }
1431 monitor_printf(mon, " ro=%d drv=%s encrypted=%d",
1432 qdict_get_bool(qdict, "ro"),
1433 qdict_get_str(qdict, "drv"),
1434 qdict_get_bool(qdict, "encrypted"));
1435 } else {
1436 monitor_printf(mon, " [not inserted]");
1437 }
1438
1439 monitor_printf(mon, "\n");
1440 }
1441
1442 void bdrv_info_print(Monitor *mon, const QObject *data)
1443 {
1444 qlist_iter(qobject_to_qlist(data), bdrv_print_dict, mon);
1445 }
1446
1447 void bdrv_info(Monitor *mon, QObject **ret_data)
1448 {
1449 QList *bs_list;
1450 BlockDriverState *bs;
1451
1452 bs_list = qlist_new();
1453
1454 QTAILQ_FOREACH(bs, &bdrv_states, list) {
1455 QObject *bs_obj;
1456 const char *type = "unknown";
1457
1458 switch(bs->type) {
1459 case BDRV_TYPE_HD:
1460 type = "hd";
1461 break;
1462 case BDRV_TYPE_CDROM:
1463 type = "cdrom";
1464 break;
1465 case BDRV_TYPE_FLOPPY:
1466 type = "floppy";
1467 break;
1468 }
1469
1470 bs_obj = qobject_from_jsonf("{ 'device': %s, 'type': %s, "
1471 "'removable': %i, 'locked': %i }",
1472 bs->device_name, type, bs->removable,
1473 bs->locked);
1474
1475 if (bs->drv) {
1476 QObject *obj;
1477 QDict *bs_dict = qobject_to_qdict(bs_obj);
1478
1479 obj = qobject_from_jsonf("{ 'file': %s, 'ro': %i, 'drv': %s, "
1480 "'encrypted': %i }",
1481 bs->filename, bs->read_only,
1482 bs->drv->format_name,
1483 bdrv_is_encrypted(bs));
1484 if (bs->backing_file[0] != '\0') {
1485 QDict *qdict = qobject_to_qdict(obj);
1486 qdict_put(qdict, "backing_file",
1487 qstring_from_str(bs->backing_file));
1488 }
1489
1490 qdict_put_obj(bs_dict, "inserted", obj);
1491 }
1492 qlist_append_obj(bs_list, bs_obj);
1493 }
1494
1495 *ret_data = QOBJECT(bs_list);
1496 }
1497
1498 static void bdrv_stats_iter(QObject *data, void *opaque)
1499 {
1500 QDict *qdict;
1501 Monitor *mon = opaque;
1502
1503 qdict = qobject_to_qdict(data);
1504 monitor_printf(mon, "%s:", qdict_get_str(qdict, "device"));
1505
1506 qdict = qobject_to_qdict(qdict_get(qdict, "stats"));
1507 monitor_printf(mon, " rd_bytes=%" PRId64
1508 " wr_bytes=%" PRId64
1509 " rd_operations=%" PRId64
1510 " wr_operations=%" PRId64
1511 "\n",
1512 qdict_get_int(qdict, "rd_bytes"),
1513 qdict_get_int(qdict, "wr_bytes"),
1514 qdict_get_int(qdict, "rd_operations"),
1515 qdict_get_int(qdict, "wr_operations"));
1516 }
1517
1518 void bdrv_stats_print(Monitor *mon, const QObject *data)
1519 {
1520 qlist_iter(qobject_to_qlist(data), bdrv_stats_iter, mon);
1521 }
1522
1523 static QObject* bdrv_info_stats_bs(BlockDriverState *bs)
1524 {
1525 QObject *res;
1526 QDict *dict;
1527
1528 res = qobject_from_jsonf("{ 'stats': {"
1529 "'rd_bytes': %" PRId64 ","
1530 "'wr_bytes': %" PRId64 ","
1531 "'rd_operations': %" PRId64 ","
1532 "'wr_operations': %" PRId64 ","
1533 "'wr_highest_offset': %" PRId64
1534 "} }",
1535 bs->rd_bytes, bs->wr_bytes,
1536 bs->rd_ops, bs->wr_ops,
1537 bs->wr_highest_sector * 512);
1538 dict = qobject_to_qdict(res);
1539
1540 if (*bs->device_name) {
1541 qdict_put(dict, "device", qstring_from_str(bs->device_name));
1542 }
1543
1544 if (bs->file) {
1545 QObject *parent = bdrv_info_stats_bs(bs->file);
1546 qdict_put_obj(dict, "parent", parent);
1547 }
1548
1549 return res;
1550 }
1551
1552 void bdrv_info_stats(Monitor *mon, QObject **ret_data)
1553 {
1554 QObject *obj;
1555 QList *devices;
1556 BlockDriverState *bs;
1557
1558 devices = qlist_new();
1559
1560 QTAILQ_FOREACH(bs, &bdrv_states, list) {
1561 obj = bdrv_info_stats_bs(bs);
1562 qlist_append_obj(devices, obj);
1563 }
1564
1565 *ret_data = QOBJECT(devices);
1566 }
1567
1568 const char *bdrv_get_encrypted_filename(BlockDriverState *bs)
1569 {
1570 if (bs->backing_hd && bs->backing_hd->encrypted)
1571 return bs->backing_file;
1572 else if (bs->encrypted)
1573 return bs->filename;
1574 else
1575 return NULL;
1576 }
1577
1578 void bdrv_get_backing_filename(BlockDriverState *bs,
1579 char *filename, int filename_size)
1580 {
1581 if (!bs->backing_file) {
1582 pstrcpy(filename, filename_size, "");
1583 } else {
1584 pstrcpy(filename, filename_size, bs->backing_file);
1585 }
1586 }
1587
1588 int bdrv_write_compressed(BlockDriverState *bs, int64_t sector_num,
1589 const uint8_t *buf, int nb_sectors)
1590 {
1591 BlockDriver *drv = bs->drv;
1592 if (!drv)
1593 return -ENOMEDIUM;
1594 if (!drv->bdrv_write_compressed)
1595 return -ENOTSUP;
1596 if (bdrv_check_request(bs, sector_num, nb_sectors))
1597 return -EIO;
1598
1599 if (bs->dirty_bitmap) {
1600 set_dirty_bitmap(bs, sector_num, nb_sectors, 1);
1601 }
1602
1603 return drv->bdrv_write_compressed(bs, sector_num, buf, nb_sectors);
1604 }
1605
1606 int bdrv_get_info(BlockDriverState *bs, BlockDriverInfo *bdi)
1607 {
1608 BlockDriver *drv = bs->drv;
1609 if (!drv)
1610 return -ENOMEDIUM;
1611 if (!drv->bdrv_get_info)
1612 return -ENOTSUP;
1613 memset(bdi, 0, sizeof(*bdi));
1614 return drv->bdrv_get_info(bs, bdi);
1615 }
1616
1617 int bdrv_save_vmstate(BlockDriverState *bs, const uint8_t *buf,
1618 int64_t pos, int size)
1619 {
1620 BlockDriver *drv = bs->drv;
1621 if (!drv)
1622 return -ENOMEDIUM;
1623 if (!drv->bdrv_save_vmstate)
1624 return -ENOTSUP;
1625 return drv->bdrv_save_vmstate(bs, buf, pos, size);
1626 }
1627
1628 int bdrv_load_vmstate(BlockDriverState *bs, 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_load_vmstate)
1635 return -ENOTSUP;
1636 return drv->bdrv_load_vmstate(bs, buf, pos, size);
1637 }
1638
1639 void bdrv_debug_event(BlockDriverState *bs, BlkDebugEvent event)
1640 {
1641 BlockDriver *drv = bs->drv;
1642
1643 if (!drv || !drv->bdrv_debug_event) {
1644 return;
1645 }
1646
1647 return drv->bdrv_debug_event(bs, event);
1648
1649 }
1650
1651 /**************************************************************/
1652 /* handling of snapshots */
1653
1654 int bdrv_snapshot_create(BlockDriverState *bs,
1655 QEMUSnapshotInfo *sn_info)
1656 {
1657 BlockDriver *drv = bs->drv;
1658 if (!drv)
1659 return -ENOMEDIUM;
1660 if (!drv->bdrv_snapshot_create)
1661 return -ENOTSUP;
1662 return drv->bdrv_snapshot_create(bs, sn_info);
1663 }
1664
1665 int bdrv_snapshot_goto(BlockDriverState *bs,
1666 const char *snapshot_id)
1667 {
1668 BlockDriver *drv = bs->drv;
1669 if (!drv)
1670 return -ENOMEDIUM;
1671 if (!drv->bdrv_snapshot_goto)
1672 return -ENOTSUP;
1673 return drv->bdrv_snapshot_goto(bs, snapshot_id);
1674 }
1675
1676 int bdrv_snapshot_delete(BlockDriverState *bs, const char *snapshot_id)
1677 {
1678 BlockDriver *drv = bs->drv;
1679 if (!drv)
1680 return -ENOMEDIUM;
1681 if (!drv->bdrv_snapshot_delete)
1682 return -ENOTSUP;
1683 return drv->bdrv_snapshot_delete(bs, snapshot_id);
1684 }
1685
1686 int bdrv_snapshot_list(BlockDriverState *bs,
1687 QEMUSnapshotInfo **psn_info)
1688 {
1689 BlockDriver *drv = bs->drv;
1690 if (!drv)
1691 return -ENOMEDIUM;
1692 if (!drv->bdrv_snapshot_list)
1693 return -ENOTSUP;
1694 return drv->bdrv_snapshot_list(bs, psn_info);
1695 }
1696
1697 #define NB_SUFFIXES 4
1698
1699 char *get_human_readable_size(char *buf, int buf_size, int64_t size)
1700 {
1701 static const char suffixes[NB_SUFFIXES] = "KMGT";
1702 int64_t base;
1703 int i;
1704
1705 if (size <= 999) {
1706 snprintf(buf, buf_size, "%" PRId64, size);
1707 } else {
1708 base = 1024;
1709 for(i = 0; i < NB_SUFFIXES; i++) {
1710 if (size < (10 * base)) {
1711 snprintf(buf, buf_size, "%0.1f%c",
1712 (double)size / base,
1713 suffixes[i]);
1714 break;
1715 } else if (size < (1000 * base) || i == (NB_SUFFIXES - 1)) {
1716 snprintf(buf, buf_size, "%" PRId64 "%c",
1717 ((size + (base >> 1)) / base),
1718 suffixes[i]);
1719 break;
1720 }
1721 base = base * 1024;
1722 }
1723 }
1724 return buf;
1725 }
1726
1727 char *bdrv_snapshot_dump(char *buf, int buf_size, QEMUSnapshotInfo *sn)
1728 {
1729 char buf1[128], date_buf[128], clock_buf[128];
1730 #ifdef _WIN32
1731 struct tm *ptm;
1732 #else
1733 struct tm tm;
1734 #endif
1735 time_t ti;
1736 int64_t secs;
1737
1738 if (!sn) {
1739 snprintf(buf, buf_size,
1740 "%-10s%-20s%7s%20s%15s",
1741 "ID", "TAG", "VM SIZE", "DATE", "VM CLOCK");
1742 } else {
1743 ti = sn->date_sec;
1744 #ifdef _WIN32
1745 ptm = localtime(&ti);
1746 strftime(date_buf, sizeof(date_buf),
1747 "%Y-%m-%d %H:%M:%S", ptm);
1748 #else
1749 localtime_r(&ti, &tm);
1750 strftime(date_buf, sizeof(date_buf),
1751 "%Y-%m-%d %H:%M:%S", &tm);
1752 #endif
1753 secs = sn->vm_clock_nsec / 1000000000;
1754 snprintf(clock_buf, sizeof(clock_buf),
1755 "%02d:%02d:%02d.%03d",
1756 (int)(secs / 3600),
1757 (int)((secs / 60) % 60),
1758 (int)(secs % 60),
1759 (int)((sn->vm_clock_nsec / 1000000) % 1000));
1760 snprintf(buf, buf_size,
1761 "%-10s%-20s%7s%20s%15s",
1762 sn->id_str, sn->name,
1763 get_human_readable_size(buf1, sizeof(buf1), sn->vm_state_size),
1764 date_buf,
1765 clock_buf);
1766 }
1767 return buf;
1768 }
1769
1770
1771 /**************************************************************/
1772 /* async I/Os */
1773
1774 BlockDriverAIOCB *bdrv_aio_readv(BlockDriverState *bs, int64_t sector_num,
1775 QEMUIOVector *qiov, int nb_sectors,
1776 BlockDriverCompletionFunc *cb, void *opaque)
1777 {
1778 BlockDriver *drv = bs->drv;
1779 BlockDriverAIOCB *ret;
1780
1781 if (!drv)
1782 return NULL;
1783 if (bdrv_check_request(bs, sector_num, nb_sectors))
1784 return NULL;
1785
1786 ret = drv->bdrv_aio_readv(bs, sector_num, qiov, nb_sectors,
1787 cb, opaque);
1788
1789 if (ret) {
1790 /* Update stats even though technically transfer has not happened. */
1791 bs->rd_bytes += (unsigned) nb_sectors * BDRV_SECTOR_SIZE;
1792 bs->rd_ops ++;
1793 }
1794
1795 return ret;
1796 }
1797
1798 BlockDriverAIOCB *bdrv_aio_writev(BlockDriverState *bs, int64_t sector_num,
1799 QEMUIOVector *qiov, int nb_sectors,
1800 BlockDriverCompletionFunc *cb, void *opaque)
1801 {
1802 BlockDriver *drv = bs->drv;
1803 BlockDriverAIOCB *ret;
1804
1805 if (!drv)
1806 return NULL;
1807 if (bs->read_only)
1808 return NULL;
1809 if (bdrv_check_request(bs, sector_num, nb_sectors))
1810 return NULL;
1811
1812 if (bs->dirty_bitmap) {
1813 set_dirty_bitmap(bs, sector_num, nb_sectors, 1);
1814 }
1815
1816 ret = drv->bdrv_aio_writev(bs, sector_num, qiov, nb_sectors,
1817 cb, opaque);
1818
1819 if (ret) {
1820 /* Update stats even though technically transfer has not happened. */
1821 bs->wr_bytes += (unsigned) nb_sectors * BDRV_SECTOR_SIZE;
1822 bs->wr_ops ++;
1823 if (bs->wr_highest_sector < sector_num + nb_sectors - 1) {
1824 bs->wr_highest_sector = sector_num + nb_sectors - 1;
1825 }
1826 }
1827
1828 return ret;
1829 }
1830
1831
1832 typedef struct MultiwriteCB {
1833 int error;
1834 int num_requests;
1835 int num_callbacks;
1836 struct {
1837 BlockDriverCompletionFunc *cb;
1838 void *opaque;
1839 QEMUIOVector *free_qiov;
1840 void *free_buf;
1841 } callbacks[];
1842 } MultiwriteCB;
1843
1844 static void multiwrite_user_cb(MultiwriteCB *mcb)
1845 {
1846 int i;
1847
1848 for (i = 0; i < mcb->num_callbacks; i++) {
1849 mcb->callbacks[i].cb(mcb->callbacks[i].opaque, mcb->error);
1850 if (mcb->callbacks[i].free_qiov) {
1851 qemu_iovec_destroy(mcb->callbacks[i].free_qiov);
1852 }
1853 qemu_free(mcb->callbacks[i].free_qiov);
1854 qemu_vfree(mcb->callbacks[i].free_buf);
1855 }
1856 }
1857
1858 static void multiwrite_cb(void *opaque, int ret)
1859 {
1860 MultiwriteCB *mcb = opaque;
1861
1862 if (ret < 0 && !mcb->error) {
1863 mcb->error = ret;
1864 multiwrite_user_cb(mcb);
1865 }
1866
1867 mcb->num_requests--;
1868 if (mcb->num_requests == 0) {
1869 if (mcb->error == 0) {
1870 multiwrite_user_cb(mcb);
1871 }
1872 qemu_free(mcb);
1873 }
1874 }
1875
1876 static int multiwrite_req_compare(const void *a, const void *b)
1877 {
1878 const BlockRequest *req1 = a, *req2 = b;
1879
1880 /*
1881 * Note that we can't simply subtract req2->sector from req1->sector
1882 * here as that could overflow the return value.
1883 */
1884 if (req1->sector > req2->sector) {
1885 return 1;
1886 } else if (req1->sector < req2->sector) {
1887 return -1;
1888 } else {
1889 return 0;
1890 }
1891 }
1892
1893 /*
1894 * Takes a bunch of requests and tries to merge them. Returns the number of
1895 * requests that remain after merging.
1896 */
1897 static int multiwrite_merge(BlockDriverState *bs, BlockRequest *reqs,
1898 int num_reqs, MultiwriteCB *mcb)
1899 {
1900 int i, outidx;
1901
1902 // Sort requests by start sector
1903 qsort(reqs, num_reqs, sizeof(*reqs), &multiwrite_req_compare);
1904
1905 // Check if adjacent requests touch the same clusters. If so, combine them,
1906 // filling up gaps with zero sectors.
1907 outidx = 0;
1908 for (i = 1; i < num_reqs; i++) {
1909 int merge = 0;
1910 int64_t oldreq_last = reqs[outidx].sector + reqs[outidx].nb_sectors;
1911
1912 // This handles the cases that are valid for all block drivers, namely
1913 // exactly sequential writes and overlapping writes.
1914 if (reqs[i].sector <= oldreq_last) {
1915 merge = 1;
1916 }
1917
1918 // The block driver may decide that it makes sense to combine requests
1919 // even if there is a gap of some sectors between them. In this case,
1920 // the gap is filled with zeros (therefore only applicable for yet
1921 // unused space in format like qcow2).
1922 if (!merge && bs->drv->bdrv_merge_requests) {
1923 merge = bs->drv->bdrv_merge_requests(bs, &reqs[outidx], &reqs[i]);
1924 }
1925
1926 if (reqs[outidx].qiov->niov + reqs[i].qiov->niov + 1 > IOV_MAX) {
1927 merge = 0;
1928 }
1929
1930 if (merge) {
1931 size_t size;
1932 QEMUIOVector *qiov = qemu_mallocz(sizeof(*qiov));
1933 qemu_iovec_init(qiov,
1934 reqs[outidx].qiov->niov + reqs[i].qiov->niov + 1);
1935
1936 // Add the first request to the merged one. If the requests are
1937 // overlapping, drop the last sectors of the first request.
1938 size = (reqs[i].sector - reqs[outidx].sector) << 9;
1939 qemu_iovec_concat(qiov, reqs[outidx].qiov, size);
1940
1941 // We might need to add some zeros between the two requests
1942 if (reqs[i].sector > oldreq_last) {
1943 size_t zero_bytes = (reqs[i].sector - oldreq_last) << 9;
1944 uint8_t *buf = qemu_blockalign(bs, zero_bytes);
1945 memset(buf, 0, zero_bytes);
1946 qemu_iovec_add(qiov, buf, zero_bytes);
1947 mcb->callbacks[i].free_buf = buf;
1948 }
1949
1950 // Add the second request
1951 qemu_iovec_concat(qiov, reqs[i].qiov, reqs[i].qiov->size);
1952
1953 reqs[outidx].nb_sectors += reqs[i].nb_sectors;
1954 reqs[outidx].qiov = qiov;
1955
1956 mcb->callbacks[i].free_qiov = reqs[outidx].qiov;
1957 } else {
1958 outidx++;
1959 reqs[outidx].sector = reqs[i].sector;
1960 reqs[outidx].nb_sectors = reqs[i].nb_sectors;
1961 reqs[outidx].qiov = reqs[i].qiov;
1962 }
1963 }
1964
1965 return outidx + 1;
1966 }
1967
1968 /*
1969 * Submit multiple AIO write requests at once.
1970 *
1971 * On success, the function returns 0 and all requests in the reqs array have
1972 * been submitted. In error case this function returns -1, and any of the
1973 * requests may or may not be submitted yet. In particular, this means that the
1974 * callback will be called for some of the requests, for others it won't. The
1975 * caller must check the error field of the BlockRequest to wait for the right
1976 * callbacks (if error != 0, no callback will be called).
1977 *
1978 * The implementation may modify the contents of the reqs array, e.g. to merge
1979 * requests. However, the fields opaque and error are left unmodified as they
1980 * are used to signal failure for a single request to the caller.
1981 */
1982 int bdrv_aio_multiwrite(BlockDriverState *bs, BlockRequest *reqs, int num_reqs)
1983 {
1984 BlockDriverAIOCB *acb;
1985 MultiwriteCB *mcb;
1986 int i;
1987
1988 if (num_reqs == 0) {
1989 return 0;
1990 }
1991
1992 // Create MultiwriteCB structure
1993 mcb = qemu_mallocz(sizeof(*mcb) + num_reqs * sizeof(*mcb->callbacks));
1994 mcb->num_requests = 0;
1995 mcb->num_callbacks = num_reqs;
1996
1997 for (i = 0; i < num_reqs; i++) {
1998 mcb->callbacks[i].cb = reqs[i].cb;
1999 mcb->callbacks[i].opaque = reqs[i].opaque;
2000 }
2001
2002 // Check for mergable requests
2003 num_reqs = multiwrite_merge(bs, reqs, num_reqs, mcb);
2004
2005 // Run the aio requests
2006 for (i = 0; i < num_reqs; i++) {
2007 acb = bdrv_aio_writev(bs, reqs[i].sector, reqs[i].qiov,
2008 reqs[i].nb_sectors, multiwrite_cb, mcb);
2009
2010 if (acb == NULL) {
2011 // We can only fail the whole thing if no request has been
2012 // submitted yet. Otherwise we'll wait for the submitted AIOs to
2013 // complete and report the error in the callback.
2014 if (mcb->num_requests == 0) {
2015 reqs[i].error = -EIO;
2016 goto fail;
2017 } else {
2018 mcb->num_requests++;
2019 multiwrite_cb(mcb, -EIO);
2020 break;
2021 }
2022 } else {
2023 mcb->num_requests++;
2024 }
2025 }
2026
2027 return 0;
2028
2029 fail:
2030 qemu_free(mcb);
2031 return -1;
2032 }
2033
2034 BlockDriverAIOCB *bdrv_aio_flush(BlockDriverState *bs,
2035 BlockDriverCompletionFunc *cb, void *opaque)
2036 {
2037 BlockDriver *drv = bs->drv;
2038
2039 if (bs->open_flags & BDRV_O_NO_FLUSH) {
2040 return bdrv_aio_noop_em(bs, cb, opaque);
2041 }
2042
2043 if (!drv)
2044 return NULL;
2045 return drv->bdrv_aio_flush(bs, cb, opaque);
2046 }
2047
2048 void bdrv_aio_cancel(BlockDriverAIOCB *acb)
2049 {
2050 acb->pool->cancel(acb);
2051 }
2052
2053
2054 /**************************************************************/
2055 /* async block device emulation */
2056
2057 typedef struct BlockDriverAIOCBSync {
2058 BlockDriverAIOCB common;
2059 QEMUBH *bh;
2060 int ret;
2061 /* vector translation state */
2062 QEMUIOVector *qiov;
2063 uint8_t *bounce;
2064 int is_write;
2065 } BlockDriverAIOCBSync;
2066
2067 static void bdrv_aio_cancel_em(BlockDriverAIOCB *blockacb)
2068 {
2069 BlockDriverAIOCBSync *acb =
2070 container_of(blockacb, BlockDriverAIOCBSync, common);
2071 qemu_bh_delete(acb->bh);
2072 acb->bh = NULL;
2073 qemu_aio_release(acb);
2074 }
2075
2076 static AIOPool bdrv_em_aio_pool = {
2077 .aiocb_size = sizeof(BlockDriverAIOCBSync),
2078 .cancel = bdrv_aio_cancel_em,
2079 };
2080
2081 static void bdrv_aio_bh_cb(void *opaque)
2082 {
2083 BlockDriverAIOCBSync *acb = opaque;
2084
2085 if (!acb->is_write)
2086 qemu_iovec_from_buffer(acb->qiov, acb->bounce, acb->qiov->size);
2087 qemu_vfree(acb->bounce);
2088 acb->common.cb(acb->common.opaque, acb->ret);
2089 qemu_bh_delete(acb->bh);
2090 acb->bh = NULL;
2091 qemu_aio_release(acb);
2092 }
2093
2094 static BlockDriverAIOCB *bdrv_aio_rw_vector(BlockDriverState *bs,
2095 int64_t sector_num,
2096 QEMUIOVector *qiov,
2097 int nb_sectors,
2098 BlockDriverCompletionFunc *cb,
2099 void *opaque,
2100 int is_write)
2101
2102 {
2103 BlockDriverAIOCBSync *acb;
2104
2105 acb = qemu_aio_get(&bdrv_em_aio_pool, bs, cb, opaque);
2106 acb->is_write = is_write;
2107 acb->qiov = qiov;
2108 acb->bounce = qemu_blockalign(bs, qiov->size);
2109
2110 if (!acb->bh)
2111 acb->bh = qemu_bh_new(bdrv_aio_bh_cb, acb);
2112
2113 if (is_write) {
2114 qemu_iovec_to_buffer(acb->qiov, acb->bounce);
2115 acb->ret = bdrv_write(bs, sector_num, acb->bounce, nb_sectors);
2116 } else {
2117 acb->ret = bdrv_read(bs, sector_num, acb->bounce, nb_sectors);
2118 }
2119
2120 qemu_bh_schedule(acb->bh);
2121
2122 return &acb->common;
2123 }
2124
2125 static BlockDriverAIOCB *bdrv_aio_readv_em(BlockDriverState *bs,
2126 int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
2127 BlockDriverCompletionFunc *cb, void *opaque)
2128 {
2129 return bdrv_aio_rw_vector(bs, sector_num, qiov, nb_sectors, cb, opaque, 0);
2130 }
2131
2132 static BlockDriverAIOCB *bdrv_aio_writev_em(BlockDriverState *bs,
2133 int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
2134 BlockDriverCompletionFunc *cb, void *opaque)
2135 {
2136 return bdrv_aio_rw_vector(bs, sector_num, qiov, nb_sectors, cb, opaque, 1);
2137 }
2138
2139 static BlockDriverAIOCB *bdrv_aio_flush_em(BlockDriverState *bs,
2140 BlockDriverCompletionFunc *cb, void *opaque)
2141 {
2142 BlockDriverAIOCBSync *acb;
2143
2144 acb = qemu_aio_get(&bdrv_em_aio_pool, bs, cb, opaque);
2145 acb->is_write = 1; /* don't bounce in the completion hadler */
2146 acb->qiov = NULL;
2147 acb->bounce = NULL;
2148 acb->ret = 0;
2149
2150 if (!acb->bh)
2151 acb->bh = qemu_bh_new(bdrv_aio_bh_cb, acb);
2152
2153 bdrv_flush(bs);
2154 qemu_bh_schedule(acb->bh);
2155 return &acb->common;
2156 }
2157
2158 static BlockDriverAIOCB *bdrv_aio_noop_em(BlockDriverState *bs,
2159 BlockDriverCompletionFunc *cb, void *opaque)
2160 {
2161 BlockDriverAIOCBSync *acb;
2162
2163 acb = qemu_aio_get(&bdrv_em_aio_pool, bs, cb, opaque);
2164 acb->is_write = 1; /* don't bounce in the completion handler */
2165 acb->qiov = NULL;
2166 acb->bounce = NULL;
2167 acb->ret = 0;
2168
2169 if (!acb->bh) {
2170 acb->bh = qemu_bh_new(bdrv_aio_bh_cb, acb);
2171 }
2172
2173 qemu_bh_schedule(acb->bh);
2174 return &acb->common;
2175 }
2176
2177 /**************************************************************/
2178 /* sync block device emulation */
2179
2180 static void bdrv_rw_em_cb(void *opaque, int ret)
2181 {
2182 *(int *)opaque = ret;
2183 }
2184
2185 #define NOT_DONE 0x7fffffff
2186
2187 static int bdrv_read_em(BlockDriverState *bs, int64_t sector_num,
2188 uint8_t *buf, int nb_sectors)
2189 {
2190 int async_ret;
2191 BlockDriverAIOCB *acb;
2192 struct iovec iov;
2193 QEMUIOVector qiov;
2194
2195 async_context_push();
2196
2197 async_ret = NOT_DONE;
2198 iov.iov_base = (void *)buf;
2199 iov.iov_len = nb_sectors * 512;
2200 qemu_iovec_init_external(&qiov, &iov, 1);
2201 acb = bdrv_aio_readv(bs, sector_num, &qiov, nb_sectors,
2202 bdrv_rw_em_cb, &async_ret);
2203 if (acb == NULL) {
2204 async_ret = -1;
2205 goto fail;
2206 }
2207
2208 while (async_ret == NOT_DONE) {
2209 qemu_aio_wait();
2210 }
2211
2212
2213 fail:
2214 async_context_pop();
2215 return async_ret;
2216 }
2217
2218 static int bdrv_write_em(BlockDriverState *bs, int64_t sector_num,
2219 const uint8_t *buf, int nb_sectors)
2220 {
2221 int async_ret;
2222 BlockDriverAIOCB *acb;
2223 struct iovec iov;
2224 QEMUIOVector qiov;
2225
2226 async_context_push();
2227
2228 async_ret = NOT_DONE;
2229 iov.iov_base = (void *)buf;
2230 iov.iov_len = nb_sectors * 512;
2231 qemu_iovec_init_external(&qiov, &iov, 1);
2232 acb = bdrv_aio_writev(bs, sector_num, &qiov, nb_sectors,
2233 bdrv_rw_em_cb, &async_ret);
2234 if (acb == NULL) {
2235 async_ret = -1;
2236 goto fail;
2237 }
2238 while (async_ret == NOT_DONE) {
2239 qemu_aio_wait();
2240 }
2241
2242 fail:
2243 async_context_pop();
2244 return async_ret;
2245 }
2246
2247 void bdrv_init(void)
2248 {
2249 module_call_init(MODULE_INIT_BLOCK);
2250 }
2251
2252 void bdrv_init_with_whitelist(void)
2253 {
2254 use_bdrv_whitelist = 1;
2255 bdrv_init();
2256 }
2257
2258 void *qemu_aio_get(AIOPool *pool, BlockDriverState *bs,
2259 BlockDriverCompletionFunc *cb, void *opaque)
2260 {
2261 BlockDriverAIOCB *acb;
2262
2263 if (pool->free_aiocb) {
2264 acb = pool->free_aiocb;
2265 pool->free_aiocb = acb->next;
2266 } else {
2267 acb = qemu_mallocz(pool->aiocb_size);
2268 acb->pool = pool;
2269 }
2270 acb->bs = bs;
2271 acb->cb = cb;
2272 acb->opaque = opaque;
2273 return acb;
2274 }
2275
2276 void qemu_aio_release(void *p)
2277 {
2278 BlockDriverAIOCB *acb = (BlockDriverAIOCB *)p;
2279 AIOPool *pool = acb->pool;
2280 acb->next = pool->free_aiocb;
2281 pool->free_aiocb = acb;
2282 }
2283
2284 /**************************************************************/
2285 /* removable device support */
2286
2287 /**
2288 * Return TRUE if the media is present
2289 */
2290 int bdrv_is_inserted(BlockDriverState *bs)
2291 {
2292 BlockDriver *drv = bs->drv;
2293 int ret;
2294 if (!drv)
2295 return 0;
2296 if (!drv->bdrv_is_inserted)
2297 return 1;
2298 ret = drv->bdrv_is_inserted(bs);
2299 return ret;
2300 }
2301
2302 /**
2303 * Return TRUE if the media changed since the last call to this
2304 * function. It is currently only used for floppy disks
2305 */
2306 int bdrv_media_changed(BlockDriverState *bs)
2307 {
2308 BlockDriver *drv = bs->drv;
2309 int ret;
2310
2311 if (!drv || !drv->bdrv_media_changed)
2312 ret = -ENOTSUP;
2313 else
2314 ret = drv->bdrv_media_changed(bs);
2315 if (ret == -ENOTSUP)
2316 ret = bs->media_changed;
2317 bs->media_changed = 0;
2318 return ret;
2319 }
2320
2321 /**
2322 * If eject_flag is TRUE, eject the media. Otherwise, close the tray
2323 */
2324 int bdrv_eject(BlockDriverState *bs, int eject_flag)
2325 {
2326 BlockDriver *drv = bs->drv;
2327 int ret;
2328
2329 if (bs->locked) {
2330 return -EBUSY;
2331 }
2332
2333 if (!drv || !drv->bdrv_eject) {
2334 ret = -ENOTSUP;
2335 } else {
2336 ret = drv->bdrv_eject(bs, eject_flag);
2337 }
2338 if (ret == -ENOTSUP) {
2339 if (eject_flag)
2340 bdrv_close(bs);
2341 ret = 0;
2342 }
2343
2344 return ret;
2345 }
2346
2347 int bdrv_is_locked(BlockDriverState *bs)
2348 {
2349 return bs->locked;
2350 }
2351
2352 /**
2353 * Lock or unlock the media (if it is locked, the user won't be able
2354 * to eject it manually).
2355 */
2356 void bdrv_set_locked(BlockDriverState *bs, int locked)
2357 {
2358 BlockDriver *drv = bs->drv;
2359
2360 bs->locked = locked;
2361 if (drv && drv->bdrv_set_locked) {
2362 drv->bdrv_set_locked(bs, locked);
2363 }
2364 }
2365
2366 /* needed for generic scsi interface */
2367
2368 int bdrv_ioctl(BlockDriverState *bs, unsigned long int req, void *buf)
2369 {
2370 BlockDriver *drv = bs->drv;
2371
2372 if (drv && drv->bdrv_ioctl)
2373 return drv->bdrv_ioctl(bs, req, buf);
2374 return -ENOTSUP;
2375 }
2376
2377 BlockDriverAIOCB *bdrv_aio_ioctl(BlockDriverState *bs,
2378 unsigned long int req, void *buf,
2379 BlockDriverCompletionFunc *cb, void *opaque)
2380 {
2381 BlockDriver *drv = bs->drv;
2382
2383 if (drv && drv->bdrv_aio_ioctl)
2384 return drv->bdrv_aio_ioctl(bs, req, buf, cb, opaque);
2385 return NULL;
2386 }
2387
2388
2389
2390 void *qemu_blockalign(BlockDriverState *bs, size_t size)
2391 {
2392 return qemu_memalign((bs && bs->buffer_alignment) ? bs->buffer_alignment : 512, size);
2393 }
2394
2395 void bdrv_set_dirty_tracking(BlockDriverState *bs, int enable)
2396 {
2397 int64_t bitmap_size;
2398
2399 bs->dirty_count = 0;
2400 if (enable) {
2401 if (!bs->dirty_bitmap) {
2402 bitmap_size = (bdrv_getlength(bs) >> BDRV_SECTOR_BITS) +
2403 BDRV_SECTORS_PER_DIRTY_CHUNK * 8 - 1;
2404 bitmap_size /= BDRV_SECTORS_PER_DIRTY_CHUNK * 8;
2405
2406 bs->dirty_bitmap = qemu_mallocz(bitmap_size);
2407 }
2408 } else {
2409 if (bs->dirty_bitmap) {
2410 qemu_free(bs->dirty_bitmap);
2411 bs->dirty_bitmap = NULL;
2412 }
2413 }
2414 }
2415
2416 int bdrv_get_dirty(BlockDriverState *bs, int64_t sector)
2417 {
2418 int64_t chunk = sector / (int64_t)BDRV_SECTORS_PER_DIRTY_CHUNK;
2419
2420 if (bs->dirty_bitmap &&
2421 (sector << BDRV_SECTOR_BITS) < bdrv_getlength(bs)) {
2422 return bs->dirty_bitmap[chunk / (sizeof(unsigned long) * 8)] &
2423 (1 << (chunk % (sizeof(unsigned long) * 8)));
2424 } else {
2425 return 0;
2426 }
2427 }
2428
2429 void bdrv_reset_dirty(BlockDriverState *bs, int64_t cur_sector,
2430 int nr_sectors)
2431 {
2432 set_dirty_bitmap(bs, cur_sector, nr_sectors, 0);
2433 }
2434
2435 int64_t bdrv_get_dirty_count(BlockDriverState *bs)
2436 {
2437 return bs->dirty_count;
2438 }
Anthony Liguori's QEMU queue