?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Add cache=unsafe parameter to -drive
[qemu/aliguori-queue.git] / block.c
blobcd70730a3e3b5af3fb27b84dedcef70586eacc15
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 /**
1448 * bdrv_info(): Block devices information
1449 *
1450 * Each block device information is stored in a QDict and the
1451 * returned QObject is a QList of all devices.
1452 *
1453 * The QDict contains the following:
1454 *
1455 * - "device": device name
1456 * - "type": device type
1457 * - "removable": true if the device is removable, false otherwise
1458 * - "locked": true if the device is locked, false otherwise
1459 * - "inserted": only present if the device is inserted, it is a QDict
1460 * containing the following:
1461 * - "file": device file name
1462 * - "ro": true if read-only, false otherwise
1463 * - "drv": driver format name
1464 * - "backing_file": backing file name if one is used
1465 * - "encrypted": true if encrypted, false otherwise
1466 *
1467 * Example:
1468 *
1469 * [ { "device": "ide0-hd0", "type": "hd", "removable": false, "locked": false,
1470 * "inserted": { "file": "/tmp/foobar", "ro": false, "drv": "qcow2" } },
1471 * { "device": "floppy0", "type": "floppy", "removable": true,
1472 * "locked": false } ]
1473 */
1474 void bdrv_info(Monitor *mon, QObject **ret_data)
1475 {
1476 QList *bs_list;
1477 BlockDriverState *bs;
1478
1479 bs_list = qlist_new();
1480
1481 QTAILQ_FOREACH(bs, &bdrv_states, list) {
1482 QObject *bs_obj;
1483 const char *type = "unknown";
1484
1485 switch(bs->type) {
1486 case BDRV_TYPE_HD:
1487 type = "hd";
1488 break;
1489 case BDRV_TYPE_CDROM:
1490 type = "cdrom";
1491 break;
1492 case BDRV_TYPE_FLOPPY:
1493 type = "floppy";
1494 break;
1495 }
1496
1497 bs_obj = qobject_from_jsonf("{ 'device': %s, 'type': %s, "
1498 "'removable': %i, 'locked': %i }",
1499 bs->device_name, type, bs->removable,
1500 bs->locked);
1501
1502 if (bs->drv) {
1503 QObject *obj;
1504 QDict *bs_dict = qobject_to_qdict(bs_obj);
1505
1506 obj = qobject_from_jsonf("{ 'file': %s, 'ro': %i, 'drv': %s, "
1507 "'encrypted': %i }",
1508 bs->filename, bs->read_only,
1509 bs->drv->format_name,
1510 bdrv_is_encrypted(bs));
1511 if (bs->backing_file[0] != '\0') {
1512 QDict *qdict = qobject_to_qdict(obj);
1513 qdict_put(qdict, "backing_file",
1514 qstring_from_str(bs->backing_file));
1515 }
1516
1517 qdict_put_obj(bs_dict, "inserted", obj);
1518 }
1519 qlist_append_obj(bs_list, bs_obj);
1520 }
1521
1522 *ret_data = QOBJECT(bs_list);
1523 }
1524
1525 static void bdrv_stats_iter(QObject *data, void *opaque)
1526 {
1527 QDict *qdict;
1528 Monitor *mon = opaque;
1529
1530 qdict = qobject_to_qdict(data);
1531 monitor_printf(mon, "%s:", qdict_get_str(qdict, "device"));
1532
1533 qdict = qobject_to_qdict(qdict_get(qdict, "stats"));
1534 monitor_printf(mon, " rd_bytes=%" PRId64
1535 " wr_bytes=%" PRId64
1536 " rd_operations=%" PRId64
1537 " wr_operations=%" PRId64
1538 "\n",
1539 qdict_get_int(qdict, "rd_bytes"),
1540 qdict_get_int(qdict, "wr_bytes"),
1541 qdict_get_int(qdict, "rd_operations"),
1542 qdict_get_int(qdict, "wr_operations"));
1543 }
1544
1545 void bdrv_stats_print(Monitor *mon, const QObject *data)
1546 {
1547 qlist_iter(qobject_to_qlist(data), bdrv_stats_iter, mon);
1548 }
1549
1550 static QObject* bdrv_info_stats_bs(BlockDriverState *bs)
1551 {
1552 QObject *res;
1553 QDict *dict;
1554
1555 res = qobject_from_jsonf("{ 'stats': {"
1556 "'rd_bytes': %" PRId64 ","
1557 "'wr_bytes': %" PRId64 ","
1558 "'rd_operations': %" PRId64 ","
1559 "'wr_operations': %" PRId64 ","
1560 "'wr_highest_offset': %" PRId64
1561 "} }",
1562 bs->rd_bytes, bs->wr_bytes,
1563 bs->rd_ops, bs->wr_ops,
1564 bs->wr_highest_sector * 512);
1565 dict = qobject_to_qdict(res);
1566
1567 if (*bs->device_name) {
1568 qdict_put(dict, "device", qstring_from_str(bs->device_name));
1569 }
1570
1571 if (bs->file) {
1572 QObject *parent = bdrv_info_stats_bs(bs->file);
1573 qdict_put_obj(dict, "parent", parent);
1574 }
1575
1576 return res;
1577 }
1578
1579 /**
1580 * bdrv_info_stats(): show block device statistics
1581 *
1582 * Each device statistic information is stored in a QDict and
1583 * the returned QObject is a QList of all devices.
1584 *
1585 * The QDict contains the following:
1586 *
1587 * - "device": device name
1588 * - "stats": A QDict with the statistics information, it contains:
1589 * - "rd_bytes": bytes read
1590 * - "wr_bytes": bytes written
1591 * - "rd_operations": read operations
1592 * - "wr_operations": write operations
1593 * - "wr_highest_offset": Highest offset of a sector written since the
1594 * BlockDriverState has been opened
1595 * - "parent": A QDict recursively holding the statistics of the underlying
1596 * protocol (e.g. the host file for a qcow2 image). If there is no
1597 * underlying protocol, this field is omitted.
1598 *
1599 * Example:
1600 *
1601 * [ { "device": "ide0-hd0",
1602 * "stats": { "rd_bytes": 512,
1603 * "wr_bytes": 0,
1604 * "rd_operations": 1,
1605 * "wr_operations": 0,
1606 * "wr_highest_offset": 0 },
1607 * "parent": {
1608 * "stats": { "rd_bytes": 1024,
1609 * "wr_bytes": 0,
1610 * "rd_operations": 2,
1611 * "wr_operations": 0,
1612 * "wr_highest_offset": 0,
1613 * } } },
1614 * { "device": "ide1-cd0",
1615 * "stats": { "rd_bytes": 0,
1616 * "wr_bytes": 0,
1617 * "rd_operations": 0,
1618 * "wr_operations": 0,
1619 * "wr_highest_offset": 0 } },
1620 */
1621 void bdrv_info_stats(Monitor *mon, QObject **ret_data)
1622 {
1623 QObject *obj;
1624 QList *devices;
1625 BlockDriverState *bs;
1626
1627 devices = qlist_new();
1628
1629 QTAILQ_FOREACH(bs, &bdrv_states, list) {
1630 obj = bdrv_info_stats_bs(bs);
1631 qlist_append_obj(devices, obj);
1632 }
1633
1634 *ret_data = QOBJECT(devices);
1635 }
1636
1637 const char *bdrv_get_encrypted_filename(BlockDriverState *bs)
1638 {
1639 if (bs->backing_hd && bs->backing_hd->encrypted)
1640 return bs->backing_file;
1641 else if (bs->encrypted)
1642 return bs->filename;
1643 else
1644 return NULL;
1645 }
1646
1647 void bdrv_get_backing_filename(BlockDriverState *bs,
1648 char *filename, int filename_size)
1649 {
1650 if (!bs->backing_file) {
1651 pstrcpy(filename, filename_size, "");
1652 } else {
1653 pstrcpy(filename, filename_size, bs->backing_file);
1654 }
1655 }
1656
1657 int bdrv_write_compressed(BlockDriverState *bs, int64_t sector_num,
1658 const uint8_t *buf, int nb_sectors)
1659 {
1660 BlockDriver *drv = bs->drv;
1661 if (!drv)
1662 return -ENOMEDIUM;
1663 if (!drv->bdrv_write_compressed)
1664 return -ENOTSUP;
1665 if (bdrv_check_request(bs, sector_num, nb_sectors))
1666 return -EIO;
1667
1668 if (bs->dirty_bitmap) {
1669 set_dirty_bitmap(bs, sector_num, nb_sectors, 1);
1670 }
1671
1672 return drv->bdrv_write_compressed(bs, sector_num, buf, nb_sectors);
1673 }
1674
1675 int bdrv_get_info(BlockDriverState *bs, BlockDriverInfo *bdi)
1676 {
1677 BlockDriver *drv = bs->drv;
1678 if (!drv)
1679 return -ENOMEDIUM;
1680 if (!drv->bdrv_get_info)
1681 return -ENOTSUP;
1682 memset(bdi, 0, sizeof(*bdi));
1683 return drv->bdrv_get_info(bs, bdi);
1684 }
1685
1686 int bdrv_save_vmstate(BlockDriverState *bs, const uint8_t *buf,
1687 int64_t pos, int size)
1688 {
1689 BlockDriver *drv = bs->drv;
1690 if (!drv)
1691 return -ENOMEDIUM;
1692 if (!drv->bdrv_save_vmstate)
1693 return -ENOTSUP;
1694 return drv->bdrv_save_vmstate(bs, buf, pos, size);
1695 }
1696
1697 int bdrv_load_vmstate(BlockDriverState *bs, uint8_t *buf,
1698 int64_t pos, int size)
1699 {
1700 BlockDriver *drv = bs->drv;
1701 if (!drv)
1702 return -ENOMEDIUM;
1703 if (!drv->bdrv_load_vmstate)
1704 return -ENOTSUP;
1705 return drv->bdrv_load_vmstate(bs, buf, pos, size);
1706 }
1707
1708 void bdrv_debug_event(BlockDriverState *bs, BlkDebugEvent event)
1709 {
1710 BlockDriver *drv = bs->drv;
1711
1712 if (!drv || !drv->bdrv_debug_event) {
1713 return;
1714 }
1715
1716 return drv->bdrv_debug_event(bs, event);
1717
1718 }
1719
1720 /**************************************************************/
1721 /* handling of snapshots */
1722
1723 int bdrv_snapshot_create(BlockDriverState *bs,
1724 QEMUSnapshotInfo *sn_info)
1725 {
1726 BlockDriver *drv = bs->drv;
1727 if (!drv)
1728 return -ENOMEDIUM;
1729 if (!drv->bdrv_snapshot_create)
1730 return -ENOTSUP;
1731 return drv->bdrv_snapshot_create(bs, sn_info);
1732 }
1733
1734 int bdrv_snapshot_goto(BlockDriverState *bs,
1735 const char *snapshot_id)
1736 {
1737 BlockDriver *drv = bs->drv;
1738 if (!drv)
1739 return -ENOMEDIUM;
1740 if (!drv->bdrv_snapshot_goto)
1741 return -ENOTSUP;
1742 return drv->bdrv_snapshot_goto(bs, snapshot_id);
1743 }
1744
1745 int bdrv_snapshot_delete(BlockDriverState *bs, const char *snapshot_id)
1746 {
1747 BlockDriver *drv = bs->drv;
1748 if (!drv)
1749 return -ENOMEDIUM;
1750 if (!drv->bdrv_snapshot_delete)
1751 return -ENOTSUP;
1752 return drv->bdrv_snapshot_delete(bs, snapshot_id);
1753 }
1754
1755 int bdrv_snapshot_list(BlockDriverState *bs,
1756 QEMUSnapshotInfo **psn_info)
1757 {
1758 BlockDriver *drv = bs->drv;
1759 if (!drv)
1760 return -ENOMEDIUM;
1761 if (!drv->bdrv_snapshot_list)
1762 return -ENOTSUP;
1763 return drv->bdrv_snapshot_list(bs, psn_info);
1764 }
1765
1766 #define NB_SUFFIXES 4
1767
1768 char *get_human_readable_size(char *buf, int buf_size, int64_t size)
1769 {
1770 static const char suffixes[NB_SUFFIXES] = "KMGT";
1771 int64_t base;
1772 int i;
1773
1774 if (size <= 999) {
1775 snprintf(buf, buf_size, "%" PRId64, size);
1776 } else {
1777 base = 1024;
1778 for(i = 0; i < NB_SUFFIXES; i++) {
1779 if (size < (10 * base)) {
1780 snprintf(buf, buf_size, "%0.1f%c",
1781 (double)size / base,
1782 suffixes[i]);
1783 break;
1784 } else if (size < (1000 * base) || i == (NB_SUFFIXES - 1)) {
1785 snprintf(buf, buf_size, "%" PRId64 "%c",
1786 ((size + (base >> 1)) / base),
1787 suffixes[i]);
1788 break;
1789 }
1790 base = base * 1024;
1791 }
1792 }
1793 return buf;
1794 }
1795
1796 char *bdrv_snapshot_dump(char *buf, int buf_size, QEMUSnapshotInfo *sn)
1797 {
1798 char buf1[128], date_buf[128], clock_buf[128];
1799 #ifdef _WIN32
1800 struct tm *ptm;
1801 #else
1802 struct tm tm;
1803 #endif
1804 time_t ti;
1805 int64_t secs;
1806
1807 if (!sn) {
1808 snprintf(buf, buf_size,
1809 "%-10s%-20s%7s%20s%15s",
1810 "ID", "TAG", "VM SIZE", "DATE", "VM CLOCK");
1811 } else {
1812 ti = sn->date_sec;
1813 #ifdef _WIN32
1814 ptm = localtime(&ti);
1815 strftime(date_buf, sizeof(date_buf),
1816 "%Y-%m-%d %H:%M:%S", ptm);
1817 #else
1818 localtime_r(&ti, &tm);
1819 strftime(date_buf, sizeof(date_buf),
1820 "%Y-%m-%d %H:%M:%S", &tm);
1821 #endif
1822 secs = sn->vm_clock_nsec / 1000000000;
1823 snprintf(clock_buf, sizeof(clock_buf),
1824 "%02d:%02d:%02d.%03d",
1825 (int)(secs / 3600),
1826 (int)((secs / 60) % 60),
1827 (int)(secs % 60),
1828 (int)((sn->vm_clock_nsec / 1000000) % 1000));
1829 snprintf(buf, buf_size,
1830 "%-10s%-20s%7s%20s%15s",
1831 sn->id_str, sn->name,
1832 get_human_readable_size(buf1, sizeof(buf1), sn->vm_state_size),
1833 date_buf,
1834 clock_buf);
1835 }
1836 return buf;
1837 }
1838
1839
1840 /**************************************************************/
1841 /* async I/Os */
1842
1843 BlockDriverAIOCB *bdrv_aio_readv(BlockDriverState *bs, int64_t sector_num,
1844 QEMUIOVector *qiov, int nb_sectors,
1845 BlockDriverCompletionFunc *cb, void *opaque)
1846 {
1847 BlockDriver *drv = bs->drv;
1848 BlockDriverAIOCB *ret;
1849
1850 if (!drv)
1851 return NULL;
1852 if (bdrv_check_request(bs, sector_num, nb_sectors))
1853 return NULL;
1854
1855 ret = drv->bdrv_aio_readv(bs, sector_num, qiov, nb_sectors,
1856 cb, opaque);
1857
1858 if (ret) {
1859 /* Update stats even though technically transfer has not happened. */
1860 bs->rd_bytes += (unsigned) nb_sectors * BDRV_SECTOR_SIZE;
1861 bs->rd_ops ++;
1862 }
1863
1864 return ret;
1865 }
1866
1867 BlockDriverAIOCB *bdrv_aio_writev(BlockDriverState *bs, int64_t sector_num,
1868 QEMUIOVector *qiov, int nb_sectors,
1869 BlockDriverCompletionFunc *cb, void *opaque)
1870 {
1871 BlockDriver *drv = bs->drv;
1872 BlockDriverAIOCB *ret;
1873
1874 if (!drv)
1875 return NULL;
1876 if (bs->read_only)
1877 return NULL;
1878 if (bdrv_check_request(bs, sector_num, nb_sectors))
1879 return NULL;
1880
1881 if (bs->dirty_bitmap) {
1882 set_dirty_bitmap(bs, sector_num, nb_sectors, 1);
1883 }
1884
1885 ret = drv->bdrv_aio_writev(bs, sector_num, qiov, nb_sectors,
1886 cb, opaque);
1887
1888 if (ret) {
1889 /* Update stats even though technically transfer has not happened. */
1890 bs->wr_bytes += (unsigned) nb_sectors * BDRV_SECTOR_SIZE;
1891 bs->wr_ops ++;
1892 if (bs->wr_highest_sector < sector_num + nb_sectors - 1) {
1893 bs->wr_highest_sector = sector_num + nb_sectors - 1;
1894 }
1895 }
1896
1897 return ret;
1898 }
1899
1900
1901 typedef struct MultiwriteCB {
1902 int error;
1903 int num_requests;
1904 int num_callbacks;
1905 struct {
1906 BlockDriverCompletionFunc *cb;
1907 void *opaque;
1908 QEMUIOVector *free_qiov;
1909 void *free_buf;
1910 } callbacks[];
1911 } MultiwriteCB;
1912
1913 static void multiwrite_user_cb(MultiwriteCB *mcb)
1914 {
1915 int i;
1916
1917 for (i = 0; i < mcb->num_callbacks; i++) {
1918 mcb->callbacks[i].cb(mcb->callbacks[i].opaque, mcb->error);
1919 if (mcb->callbacks[i].free_qiov) {
1920 qemu_iovec_destroy(mcb->callbacks[i].free_qiov);
1921 }
1922 qemu_free(mcb->callbacks[i].free_qiov);
1923 qemu_vfree(mcb->callbacks[i].free_buf);
1924 }
1925 }
1926
1927 static void multiwrite_cb(void *opaque, int ret)
1928 {
1929 MultiwriteCB *mcb = opaque;
1930
1931 if (ret < 0 && !mcb->error) {
1932 mcb->error = ret;
1933 multiwrite_user_cb(mcb);
1934 }
1935
1936 mcb->num_requests--;
1937 if (mcb->num_requests == 0) {
1938 if (mcb->error == 0) {
1939 multiwrite_user_cb(mcb);
1940 }
1941 qemu_free(mcb);
1942 }
1943 }
1944
1945 static int multiwrite_req_compare(const void *a, const void *b)
1946 {
1947 const BlockRequest *req1 = a, *req2 = b;
1948
1949 /*
1950 * Note that we can't simply subtract req2->sector from req1->sector
1951 * here as that could overflow the return value.
1952 */
1953 if (req1->sector > req2->sector) {
1954 return 1;
1955 } else if (req1->sector < req2->sector) {
1956 return -1;
1957 } else {
1958 return 0;
1959 }
1960 }
1961
1962 /*
1963 * Takes a bunch of requests and tries to merge them. Returns the number of
1964 * requests that remain after merging.
1965 */
1966 static int multiwrite_merge(BlockDriverState *bs, BlockRequest *reqs,
1967 int num_reqs, MultiwriteCB *mcb)
1968 {
1969 int i, outidx;
1970
1971 // Sort requests by start sector
1972 qsort(reqs, num_reqs, sizeof(*reqs), &multiwrite_req_compare);
1973
1974 // Check if adjacent requests touch the same clusters. If so, combine them,
1975 // filling up gaps with zero sectors.
1976 outidx = 0;
1977 for (i = 1; i < num_reqs; i++) {
1978 int merge = 0;
1979 int64_t oldreq_last = reqs[outidx].sector + reqs[outidx].nb_sectors;
1980
1981 // This handles the cases that are valid for all block drivers, namely
1982 // exactly sequential writes and overlapping writes.
1983 if (reqs[i].sector <= oldreq_last) {
1984 merge = 1;
1985 }
1986
1987 // The block driver may decide that it makes sense to combine requests
1988 // even if there is a gap of some sectors between them. In this case,
1989 // the gap is filled with zeros (therefore only applicable for yet
1990 // unused space in format like qcow2).
1991 if (!merge && bs->drv->bdrv_merge_requests) {
1992 merge = bs->drv->bdrv_merge_requests(bs, &reqs[outidx], &reqs[i]);
1993 }
1994
1995 if (reqs[outidx].qiov->niov + reqs[i].qiov->niov + 1 > IOV_MAX) {
1996 merge = 0;
1997 }
1998
1999 if (merge) {
2000 size_t size;
2001 QEMUIOVector *qiov = qemu_mallocz(sizeof(*qiov));
2002 qemu_iovec_init(qiov,
2003 reqs[outidx].qiov->niov + reqs[i].qiov->niov + 1);
2004
2005 // Add the first request to the merged one. If the requests are
2006 // overlapping, drop the last sectors of the first request.
2007 size = (reqs[i].sector - reqs[outidx].sector) << 9;
2008 qemu_iovec_concat(qiov, reqs[outidx].qiov, size);
2009
2010 // We might need to add some zeros between the two requests
2011 if (reqs[i].sector > oldreq_last) {
2012 size_t zero_bytes = (reqs[i].sector - oldreq_last) << 9;
2013 uint8_t *buf = qemu_blockalign(bs, zero_bytes);
2014 memset(buf, 0, zero_bytes);
2015 qemu_iovec_add(qiov, buf, zero_bytes);
2016 mcb->callbacks[i].free_buf = buf;
2017 }
2018
2019 // Add the second request
2020 qemu_iovec_concat(qiov, reqs[i].qiov, reqs[i].qiov->size);
2021
2022 reqs[outidx].nb_sectors += reqs[i].nb_sectors;
2023 reqs[outidx].qiov = qiov;
2024
2025 mcb->callbacks[i].free_qiov = reqs[outidx].qiov;
2026 } else {
2027 outidx++;
2028 reqs[outidx].sector = reqs[i].sector;
2029 reqs[outidx].nb_sectors = reqs[i].nb_sectors;
2030 reqs[outidx].qiov = reqs[i].qiov;
2031 }
2032 }
2033
2034 return outidx + 1;
2035 }
2036
2037 /*
2038 * Submit multiple AIO write requests at once.
2039 *
2040 * On success, the function returns 0 and all requests in the reqs array have
2041 * been submitted. In error case this function returns -1, and any of the
2042 * requests may or may not be submitted yet. In particular, this means that the
2043 * callback will be called for some of the requests, for others it won't. The
2044 * caller must check the error field of the BlockRequest to wait for the right
2045 * callbacks (if error != 0, no callback will be called).
2046 *
2047 * The implementation may modify the contents of the reqs array, e.g. to merge
2048 * requests. However, the fields opaque and error are left unmodified as they
2049 * are used to signal failure for a single request to the caller.
2050 */
2051 int bdrv_aio_multiwrite(BlockDriverState *bs, BlockRequest *reqs, int num_reqs)
2052 {
2053 BlockDriverAIOCB *acb;
2054 MultiwriteCB *mcb;
2055 int i;
2056
2057 if (num_reqs == 0) {
2058 return 0;
2059 }
2060
2061 // Create MultiwriteCB structure
2062 mcb = qemu_mallocz(sizeof(*mcb) + num_reqs * sizeof(*mcb->callbacks));
2063 mcb->num_requests = 0;
2064 mcb->num_callbacks = num_reqs;
2065
2066 for (i = 0; i < num_reqs; i++) {
2067 mcb->callbacks[i].cb = reqs[i].cb;
2068 mcb->callbacks[i].opaque = reqs[i].opaque;
2069 }
2070
2071 // Check for mergable requests
2072 num_reqs = multiwrite_merge(bs, reqs, num_reqs, mcb);
2073
2074 // Run the aio requests
2075 for (i = 0; i < num_reqs; i++) {
2076 acb = bdrv_aio_writev(bs, reqs[i].sector, reqs[i].qiov,
2077 reqs[i].nb_sectors, multiwrite_cb, mcb);
2078
2079 if (acb == NULL) {
2080 // We can only fail the whole thing if no request has been
2081 // submitted yet. Otherwise we'll wait for the submitted AIOs to
2082 // complete and report the error in the callback.
2083 if (mcb->num_requests == 0) {
2084 reqs[i].error = -EIO;
2085 goto fail;
2086 } else {
2087 mcb->num_requests++;
2088 multiwrite_cb(mcb, -EIO);
2089 break;
2090 }
2091 } else {
2092 mcb->num_requests++;
2093 }
2094 }
2095
2096 return 0;
2097
2098 fail:
2099 qemu_free(mcb);
2100 return -1;
2101 }
2102
2103 BlockDriverAIOCB *bdrv_aio_flush(BlockDriverState *bs,
2104 BlockDriverCompletionFunc *cb, void *opaque)
2105 {
2106 BlockDriver *drv = bs->drv;
2107
2108 if (bs->open_flags & BDRV_O_NO_FLUSH) {
2109 return bdrv_aio_noop_em(bs, cb, opaque);
2110 }
2111
2112 if (!drv)
2113 return NULL;
2114 return drv->bdrv_aio_flush(bs, cb, opaque);
2115 }
2116
2117 void bdrv_aio_cancel(BlockDriverAIOCB *acb)
2118 {
2119 acb->pool->cancel(acb);
2120 }
2121
2122
2123 /**************************************************************/
2124 /* async block device emulation */
2125
2126 typedef struct BlockDriverAIOCBSync {
2127 BlockDriverAIOCB common;
2128 QEMUBH *bh;
2129 int ret;
2130 /* vector translation state */
2131 QEMUIOVector *qiov;
2132 uint8_t *bounce;
2133 int is_write;
2134 } BlockDriverAIOCBSync;
2135
2136 static void bdrv_aio_cancel_em(BlockDriverAIOCB *blockacb)
2137 {
2138 BlockDriverAIOCBSync *acb =
2139 container_of(blockacb, BlockDriverAIOCBSync, common);
2140 qemu_bh_delete(acb->bh);
2141 acb->bh = NULL;
2142 qemu_aio_release(acb);
2143 }
2144
2145 static AIOPool bdrv_em_aio_pool = {
2146 .aiocb_size = sizeof(BlockDriverAIOCBSync),
2147 .cancel = bdrv_aio_cancel_em,
2148 };
2149
2150 static void bdrv_aio_bh_cb(void *opaque)
2151 {
2152 BlockDriverAIOCBSync *acb = opaque;
2153
2154 if (!acb->is_write)
2155 qemu_iovec_from_buffer(acb->qiov, acb->bounce, acb->qiov->size);
2156 qemu_vfree(acb->bounce);
2157 acb->common.cb(acb->common.opaque, acb->ret);
2158 qemu_bh_delete(acb->bh);
2159 acb->bh = NULL;
2160 qemu_aio_release(acb);
2161 }
2162
2163 static BlockDriverAIOCB *bdrv_aio_rw_vector(BlockDriverState *bs,
2164 int64_t sector_num,
2165 QEMUIOVector *qiov,
2166 int nb_sectors,
2167 BlockDriverCompletionFunc *cb,
2168 void *opaque,
2169 int is_write)
2170
2171 {
2172 BlockDriverAIOCBSync *acb;
2173
2174 acb = qemu_aio_get(&bdrv_em_aio_pool, bs, cb, opaque);
2175 acb->is_write = is_write;
2176 acb->qiov = qiov;
2177 acb->bounce = qemu_blockalign(bs, qiov->size);
2178
2179 if (!acb->bh)
2180 acb->bh = qemu_bh_new(bdrv_aio_bh_cb, acb);
2181
2182 if (is_write) {
2183 qemu_iovec_to_buffer(acb->qiov, acb->bounce);
2184 acb->ret = bdrv_write(bs, sector_num, acb->bounce, nb_sectors);
2185 } else {
2186 acb->ret = bdrv_read(bs, sector_num, acb->bounce, nb_sectors);
2187 }
2188
2189 qemu_bh_schedule(acb->bh);
2190
2191 return &acb->common;
2192 }
2193
2194 static BlockDriverAIOCB *bdrv_aio_readv_em(BlockDriverState *bs,
2195 int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
2196 BlockDriverCompletionFunc *cb, void *opaque)
2197 {
2198 return bdrv_aio_rw_vector(bs, sector_num, qiov, nb_sectors, cb, opaque, 0);
2199 }
2200
2201 static BlockDriverAIOCB *bdrv_aio_writev_em(BlockDriverState *bs,
2202 int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
2203 BlockDriverCompletionFunc *cb, void *opaque)
2204 {
2205 return bdrv_aio_rw_vector(bs, sector_num, qiov, nb_sectors, cb, opaque, 1);
2206 }
2207
2208 static BlockDriverAIOCB *bdrv_aio_flush_em(BlockDriverState *bs,
2209 BlockDriverCompletionFunc *cb, void *opaque)
2210 {
2211 BlockDriverAIOCBSync *acb;
2212
2213 acb = qemu_aio_get(&bdrv_em_aio_pool, bs, cb, opaque);
2214 acb->is_write = 1; /* don't bounce in the completion hadler */
2215 acb->qiov = NULL;
2216 acb->bounce = NULL;
2217 acb->ret = 0;
2218
2219 if (!acb->bh)
2220 acb->bh = qemu_bh_new(bdrv_aio_bh_cb, acb);
2221
2222 bdrv_flush(bs);
2223 qemu_bh_schedule(acb->bh);
2224 return &acb->common;
2225 }
2226
2227 static BlockDriverAIOCB *bdrv_aio_noop_em(BlockDriverState *bs,
2228 BlockDriverCompletionFunc *cb, void *opaque)
2229 {
2230 BlockDriverAIOCBSync *acb;
2231
2232 acb = qemu_aio_get(&bdrv_em_aio_pool, bs, cb, opaque);
2233 acb->is_write = 1; /* don't bounce in the completion handler */
2234 acb->qiov = NULL;
2235 acb->bounce = NULL;
2236 acb->ret = 0;
2237
2238 if (!acb->bh) {
2239 acb->bh = qemu_bh_new(bdrv_aio_bh_cb, acb);
2240 }
2241
2242 qemu_bh_schedule(acb->bh);
2243 return &acb->common;
2244 }
2245
2246 /**************************************************************/
2247 /* sync block device emulation */
2248
2249 static void bdrv_rw_em_cb(void *opaque, int ret)
2250 {
2251 *(int *)opaque = ret;
2252 }
2253
2254 #define NOT_DONE 0x7fffffff
2255
2256 static int bdrv_read_em(BlockDriverState *bs, int64_t sector_num,
2257 uint8_t *buf, int nb_sectors)
2258 {
2259 int async_ret;
2260 BlockDriverAIOCB *acb;
2261 struct iovec iov;
2262 QEMUIOVector qiov;
2263
2264 async_context_push();
2265
2266 async_ret = NOT_DONE;
2267 iov.iov_base = (void *)buf;
2268 iov.iov_len = nb_sectors * 512;
2269 qemu_iovec_init_external(&qiov, &iov, 1);
2270 acb = bdrv_aio_readv(bs, sector_num, &qiov, nb_sectors,
2271 bdrv_rw_em_cb, &async_ret);
2272 if (acb == NULL) {
2273 async_ret = -1;
2274 goto fail;
2275 }
2276
2277 while (async_ret == NOT_DONE) {
2278 qemu_aio_wait();
2279 }
2280
2281
2282 fail:
2283 async_context_pop();
2284 return async_ret;
2285 }
2286
2287 static int bdrv_write_em(BlockDriverState *bs, int64_t sector_num,
2288 const uint8_t *buf, int nb_sectors)
2289 {
2290 int async_ret;
2291 BlockDriverAIOCB *acb;
2292 struct iovec iov;
2293 QEMUIOVector qiov;
2294
2295 async_context_push();
2296
2297 async_ret = NOT_DONE;
2298 iov.iov_base = (void *)buf;
2299 iov.iov_len = nb_sectors * 512;
2300 qemu_iovec_init_external(&qiov, &iov, 1);
2301 acb = bdrv_aio_writev(bs, sector_num, &qiov, nb_sectors,
2302 bdrv_rw_em_cb, &async_ret);
2303 if (acb == NULL) {
2304 async_ret = -1;
2305 goto fail;
2306 }
2307 while (async_ret == NOT_DONE) {
2308 qemu_aio_wait();
2309 }
2310
2311 fail:
2312 async_context_pop();
2313 return async_ret;
2314 }
2315
2316 void bdrv_init(void)
2317 {
2318 module_call_init(MODULE_INIT_BLOCK);
2319 }
2320
2321 void bdrv_init_with_whitelist(void)
2322 {
2323 use_bdrv_whitelist = 1;
2324 bdrv_init();
2325 }
2326
2327 void *qemu_aio_get(AIOPool *pool, BlockDriverState *bs,
2328 BlockDriverCompletionFunc *cb, void *opaque)
2329 {
2330 BlockDriverAIOCB *acb;
2331
2332 if (pool->free_aiocb) {
2333 acb = pool->free_aiocb;
2334 pool->free_aiocb = acb->next;
2335 } else {
2336 acb = qemu_mallocz(pool->aiocb_size);
2337 acb->pool = pool;
2338 }
2339 acb->bs = bs;
2340 acb->cb = cb;
2341 acb->opaque = opaque;
2342 return acb;
2343 }
2344
2345 void qemu_aio_release(void *p)
2346 {
2347 BlockDriverAIOCB *acb = (BlockDriverAIOCB *)p;
2348 AIOPool *pool = acb->pool;
2349 acb->next = pool->free_aiocb;
2350 pool->free_aiocb = acb;
2351 }
2352
2353 /**************************************************************/
2354 /* removable device support */
2355
2356 /**
2357 * Return TRUE if the media is present
2358 */
2359 int bdrv_is_inserted(BlockDriverState *bs)
2360 {
2361 BlockDriver *drv = bs->drv;
2362 int ret;
2363 if (!drv)
2364 return 0;
2365 if (!drv->bdrv_is_inserted)
2366 return 1;
2367 ret = drv->bdrv_is_inserted(bs);
2368 return ret;
2369 }
2370
2371 /**
2372 * Return TRUE if the media changed since the last call to this
2373 * function. It is currently only used for floppy disks
2374 */
2375 int bdrv_media_changed(BlockDriverState *bs)
2376 {
2377 BlockDriver *drv = bs->drv;
2378 int ret;
2379
2380 if (!drv || !drv->bdrv_media_changed)
2381 ret = -ENOTSUP;
2382 else
2383 ret = drv->bdrv_media_changed(bs);
2384 if (ret == -ENOTSUP)
2385 ret = bs->media_changed;
2386 bs->media_changed = 0;
2387 return ret;
2388 }
2389
2390 /**
2391 * If eject_flag is TRUE, eject the media. Otherwise, close the tray
2392 */
2393 int bdrv_eject(BlockDriverState *bs, int eject_flag)
2394 {
2395 BlockDriver *drv = bs->drv;
2396 int ret;
2397
2398 if (bs->locked) {
2399 return -EBUSY;
2400 }
2401
2402 if (!drv || !drv->bdrv_eject) {
2403 ret = -ENOTSUP;
2404 } else {
2405 ret = drv->bdrv_eject(bs, eject_flag);
2406 }
2407 if (ret == -ENOTSUP) {
2408 if (eject_flag)
2409 bdrv_close(bs);
2410 ret = 0;
2411 }
2412
2413 return ret;
2414 }
2415
2416 int bdrv_is_locked(BlockDriverState *bs)
2417 {
2418 return bs->locked;
2419 }
2420
2421 /**
2422 * Lock or unlock the media (if it is locked, the user won't be able
2423 * to eject it manually).
2424 */
2425 void bdrv_set_locked(BlockDriverState *bs, int locked)
2426 {
2427 BlockDriver *drv = bs->drv;
2428
2429 bs->locked = locked;
2430 if (drv && drv->bdrv_set_locked) {
2431 drv->bdrv_set_locked(bs, locked);
2432 }
2433 }
2434
2435 /* needed for generic scsi interface */
2436
2437 int bdrv_ioctl(BlockDriverState *bs, unsigned long int req, void *buf)
2438 {
2439 BlockDriver *drv = bs->drv;
2440
2441 if (drv && drv->bdrv_ioctl)
2442 return drv->bdrv_ioctl(bs, req, buf);
2443 return -ENOTSUP;
2444 }
2445
2446 BlockDriverAIOCB *bdrv_aio_ioctl(BlockDriverState *bs,
2447 unsigned long int req, void *buf,
2448 BlockDriverCompletionFunc *cb, void *opaque)
2449 {
2450 BlockDriver *drv = bs->drv;
2451
2452 if (drv && drv->bdrv_aio_ioctl)
2453 return drv->bdrv_aio_ioctl(bs, req, buf, cb, opaque);
2454 return NULL;
2455 }
2456
2457
2458
2459 void *qemu_blockalign(BlockDriverState *bs, size_t size)
2460 {
2461 return qemu_memalign((bs && bs->buffer_alignment) ? bs->buffer_alignment : 512, size);
2462 }
2463
2464 void bdrv_set_dirty_tracking(BlockDriverState *bs, int enable)
2465 {
2466 int64_t bitmap_size;
2467
2468 bs->dirty_count = 0;
2469 if (enable) {
2470 if (!bs->dirty_bitmap) {
2471 bitmap_size = (bdrv_getlength(bs) >> BDRV_SECTOR_BITS) +
2472 BDRV_SECTORS_PER_DIRTY_CHUNK * 8 - 1;
2473 bitmap_size /= BDRV_SECTORS_PER_DIRTY_CHUNK * 8;
2474
2475 bs->dirty_bitmap = qemu_mallocz(bitmap_size);
2476 }
2477 } else {
2478 if (bs->dirty_bitmap) {
2479 qemu_free(bs->dirty_bitmap);
2480 bs->dirty_bitmap = NULL;
2481 }
2482 }
2483 }
2484
2485 int bdrv_get_dirty(BlockDriverState *bs, int64_t sector)
2486 {
2487 int64_t chunk = sector / (int64_t)BDRV_SECTORS_PER_DIRTY_CHUNK;
2488
2489 if (bs->dirty_bitmap &&
2490 (sector << BDRV_SECTOR_BITS) < bdrv_getlength(bs)) {
2491 return bs->dirty_bitmap[chunk / (sizeof(unsigned long) * 8)] &
2492 (1 << (chunk % (sizeof(unsigned long) * 8)));
2493 } else {
2494 return 0;
2495 }
2496 }
2497
2498 void bdrv_reset_dirty(BlockDriverState *bs, int64_t cur_sector,
2499 int nr_sectors)
2500 {
2501 set_dirty_bitmap(bs, cur_sector, nr_sectors, 0);
2502 }
2503
2504 int64_t bdrv_get_dirty_count(BlockDriverState *bs)
2505 {
2506 return bs->dirty_count;
2507 }
Anthony Liguori's QEMU queue