?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
strongarm ioctl rename fallout
[qemu/aliguori.git] / block.c
blob26910ca14301e51174bce04b686f0be5cb6bf9d7
1 /*
2 * QEMU System Emulator block driver
3 *
4 * Copyright (c) 2003 Fabrice Bellard
5 *
6 * Permission is hereby granted, free of charge, to any person obtaining a copy
7 * of this software and associated documentation files (the "Software"), to deal
8 * in the Software without restriction, including without limitation the rights
9 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
10 * copies of the Software, and to permit persons to whom the Software is
11 * furnished to do so, subject to the following conditions:
12 *
13 * The above copyright notice and this permission notice shall be included in
14 * all copies or substantial portions of the Software.
15 *
16 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
17 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
18 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
19 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
20 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
21 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
22 * THE SOFTWARE.
23 */
24 #include "config-host.h"
25 #include "qemu-common.h"
26 #include "trace.h"
27 #include "monitor.h"
28 #include "block_int.h"
29 #include "module.h"
30 #include "qemu-objects.h"
31 #include "qemu-coroutine.h"
32
33 #ifdef CONFIG_BSD
34 #include <sys/types.h>
35 #include <sys/stat.h>
36 #include <sys/ioctl.h>
37 #include <sys/queue.h>
38 #ifndef __DragonFly__
39 #include <sys/disk.h>
40 #endif
41 #endif
42
43 #ifdef _WIN32
44 #include <windows.h>
45 #endif
46
47 static BlockDriverAIOCB *bdrv_aio_readv_em(BlockDriverState *bs,
48 int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
49 BlockDriverCompletionFunc *cb, void *opaque);
50 static BlockDriverAIOCB *bdrv_aio_writev_em(BlockDriverState *bs,
51 int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
52 BlockDriverCompletionFunc *cb, void *opaque);
53 static BlockDriverAIOCB *bdrv_aio_flush_em(BlockDriverState *bs,
54 BlockDriverCompletionFunc *cb, void *opaque);
55 static BlockDriverAIOCB *bdrv_aio_noop_em(BlockDriverState *bs,
56 BlockDriverCompletionFunc *cb, void *opaque);
57 static int bdrv_read_em(BlockDriverState *bs, int64_t sector_num,
58 uint8_t *buf, int nb_sectors);
59 static int bdrv_write_em(BlockDriverState *bs, int64_t sector_num,
60 const uint8_t *buf, int nb_sectors);
61 static BlockDriverAIOCB *bdrv_co_aio_readv_em(BlockDriverState *bs,
62 int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
63 BlockDriverCompletionFunc *cb, void *opaque);
64 static BlockDriverAIOCB *bdrv_co_aio_writev_em(BlockDriverState *bs,
65 int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
66 BlockDriverCompletionFunc *cb, void *opaque);
67 static int coroutine_fn bdrv_co_readv_em(BlockDriverState *bs,
68 int64_t sector_num, int nb_sectors,
69 QEMUIOVector *iov);
70 static int coroutine_fn bdrv_co_writev_em(BlockDriverState *bs,
71 int64_t sector_num, int nb_sectors,
72 QEMUIOVector *iov);
73 static int coroutine_fn bdrv_co_flush_em(BlockDriverState *bs);
74
75 static QTAILQ_HEAD(, BlockDriverState) bdrv_states =
76 QTAILQ_HEAD_INITIALIZER(bdrv_states);
77
78 static QLIST_HEAD(, BlockDriver) bdrv_drivers =
79 QLIST_HEAD_INITIALIZER(bdrv_drivers);
80
81 /* The device to use for VM snapshots */
82 static BlockDriverState *bs_snapshots;
83
84 /* If non-zero, use only whitelisted block drivers */
85 static int use_bdrv_whitelist;
86
87 #ifdef _WIN32
88 static int is_windows_drive_prefix(const char *filename)
89 {
90 return (((filename[0] >= 'a' && filename[0] <= 'z') ||
91 (filename[0] >= 'A' && filename[0] <= 'Z')) &&
92 filename[1] == ':');
93 }
94
95 int is_windows_drive(const char *filename)
96 {
97 if (is_windows_drive_prefix(filename) &&
98 filename[2] == '\0')
99 return 1;
100 if (strstart(filename, "\\\\.\\", NULL) ||
101 strstart(filename, "//./", NULL))
102 return 1;
103 return 0;
104 }
105 #endif
106
107 /* check if the path starts with "<protocol>:" */
108 static int path_has_protocol(const char *path)
109 {
110 #ifdef _WIN32
111 if (is_windows_drive(path) ||
112 is_windows_drive_prefix(path)) {
113 return 0;
114 }
115 #endif
116
117 return strchr(path, ':') != NULL;
118 }
119
120 int path_is_absolute(const char *path)
121 {
122 const char *p;
123 #ifdef _WIN32
124 /* specific case for names like: "\\.\d:" */
125 if (*path == '/' || *path == '\\')
126 return 1;
127 #endif
128 p = strchr(path, ':');
129 if (p)
130 p++;
131 else
132 p = path;
133 #ifdef _WIN32
134 return (*p == '/' || *p == '\\');
135 #else
136 return (*p == '/');
137 #endif
138 }
139
140 /* if filename is absolute, just copy it to dest. Otherwise, build a
141 path to it by considering it is relative to base_path. URL are
142 supported. */
143 void path_combine(char *dest, int dest_size,
144 const char *base_path,
145 const char *filename)
146 {
147 const char *p, *p1;
148 int len;
149
150 if (dest_size <= 0)
151 return;
152 if (path_is_absolute(filename)) {
153 pstrcpy(dest, dest_size, filename);
154 } else {
155 p = strchr(base_path, ':');
156 if (p)
157 p++;
158 else
159 p = base_path;
160 p1 = strrchr(base_path, '/');
161 #ifdef _WIN32
162 {
163 const char *p2;
164 p2 = strrchr(base_path, '\\');
165 if (!p1 || p2 > p1)
166 p1 = p2;
167 }
168 #endif
169 if (p1)
170 p1++;
171 else
172 p1 = base_path;
173 if (p1 > p)
174 p = p1;
175 len = p - base_path;
176 if (len > dest_size - 1)
177 len = dest_size - 1;
178 memcpy(dest, base_path, len);
179 dest[len] = '\0';
180 pstrcat(dest, dest_size, filename);
181 }
182 }
183
184 void bdrv_register(BlockDriver *bdrv)
185 {
186 if (bdrv->bdrv_co_readv) {
187 /* Emulate AIO by coroutines, and sync by AIO */
188 bdrv->bdrv_aio_readv = bdrv_co_aio_readv_em;
189 bdrv->bdrv_aio_writev = bdrv_co_aio_writev_em;
190 bdrv->bdrv_read = bdrv_read_em;
191 bdrv->bdrv_write = bdrv_write_em;
192 } else {
193 bdrv->bdrv_co_readv = bdrv_co_readv_em;
194 bdrv->bdrv_co_writev = bdrv_co_writev_em;
195
196 if (!bdrv->bdrv_aio_readv) {
197 /* add AIO emulation layer */
198 bdrv->bdrv_aio_readv = bdrv_aio_readv_em;
199 bdrv->bdrv_aio_writev = bdrv_aio_writev_em;
200 } else if (!bdrv->bdrv_read) {
201 /* add synchronous IO emulation layer */
202 bdrv->bdrv_read = bdrv_read_em;
203 bdrv->bdrv_write = bdrv_write_em;
204 }
205 }
206
207 if (!bdrv->bdrv_aio_flush)
208 bdrv->bdrv_aio_flush = bdrv_aio_flush_em;
209
210 QLIST_INSERT_HEAD(&bdrv_drivers, bdrv, list);
211 }
212
213 /* create a new block device (by default it is empty) */
214 BlockDriverState *bdrv_new(const char *device_name)
215 {
216 BlockDriverState *bs;
217
218 bs = qemu_mallocz(sizeof(BlockDriverState));
219 pstrcpy(bs->device_name, sizeof(bs->device_name), device_name);
220 if (device_name[0] != '\0') {
221 QTAILQ_INSERT_TAIL(&bdrv_states, bs, list);
222 }
223 return bs;
224 }
225
226 BlockDriver *bdrv_find_format(const char *format_name)
227 {
228 BlockDriver *drv1;
229 QLIST_FOREACH(drv1, &bdrv_drivers, list) {
230 if (!strcmp(drv1->format_name, format_name)) {
231 return drv1;
232 }
233 }
234 return NULL;
235 }
236
237 static int bdrv_is_whitelisted(BlockDriver *drv)
238 {
239 static const char *whitelist[] = {
240 CONFIG_BDRV_WHITELIST
241 };
242 const char **p;
243
244 if (!whitelist[0])
245 return 1; /* no whitelist, anything goes */
246
247 for (p = whitelist; *p; p++) {
248 if (!strcmp(drv->format_name, *p)) {
249 return 1;
250 }
251 }
252 return 0;
253 }
254
255 BlockDriver *bdrv_find_whitelisted_format(const char *format_name)
256 {
257 BlockDriver *drv = bdrv_find_format(format_name);
258 return drv && bdrv_is_whitelisted(drv) ? drv : NULL;
259 }
260
261 int bdrv_create(BlockDriver *drv, const char* filename,
262 QEMUOptionParameter *options)
263 {
264 if (!drv->bdrv_create)
265 return -ENOTSUP;
266
267 return drv->bdrv_create(filename, options);
268 }
269
270 int bdrv_create_file(const char* filename, QEMUOptionParameter *options)
271 {
272 BlockDriver *drv;
273
274 drv = bdrv_find_protocol(filename);
275 if (drv == NULL) {
276 return -ENOENT;
277 }
278
279 return bdrv_create(drv, filename, options);
280 }
281
282 #ifdef _WIN32
283 void get_tmp_filename(char *filename, int size)
284 {
285 char temp_dir[MAX_PATH];
286
287 GetTempPath(MAX_PATH, temp_dir);
288 GetTempFileName(temp_dir, "qem", 0, filename);
289 }
290 #else
291 void get_tmp_filename(char *filename, int size)
292 {
293 int fd;
294 const char *tmpdir;
295 /* XXX: race condition possible */
296 tmpdir = getenv("TMPDIR");
297 if (!tmpdir)
298 tmpdir = "/tmp";
299 snprintf(filename, size, "%s/vl.XXXXXX", tmpdir);
300 fd = mkstemp(filename);
301 close(fd);
302 }
303 #endif
304
305 /*
306 * Detect host devices. By convention, /dev/cdrom[N] is always
307 * recognized as a host CDROM.
308 */
309 static BlockDriver *find_hdev_driver(const char *filename)
310 {
311 int score_max = 0, score;
312 BlockDriver *drv = NULL, *d;
313
314 QLIST_FOREACH(d, &bdrv_drivers, list) {
315 if (d->bdrv_probe_device) {
316 score = d->bdrv_probe_device(filename);
317 if (score > score_max) {
318 score_max = score;
319 drv = d;
320 }
321 }
322 }
323
324 return drv;
325 }
326
327 BlockDriver *bdrv_find_protocol(const char *filename)
328 {
329 BlockDriver *drv1;
330 char protocol[128];
331 int len;
332 const char *p;
333
334 /* TODO Drivers without bdrv_file_open must be specified explicitly */
335
336 /*
337 * XXX(hch): we really should not let host device detection
338 * override an explicit protocol specification, but moving this
339 * later breaks access to device names with colons in them.
340 * Thanks to the brain-dead persistent naming schemes on udev-
341 * based Linux systems those actually are quite common.
342 */
343 drv1 = find_hdev_driver(filename);
344 if (drv1) {
345 return drv1;
346 }
347
348 if (!path_has_protocol(filename)) {
349 return bdrv_find_format("file");
350 }
351 p = strchr(filename, ':');
352 assert(p != NULL);
353 len = p - filename;
354 if (len > sizeof(protocol) - 1)
355 len = sizeof(protocol) - 1;
356 memcpy(protocol, filename, len);
357 protocol[len] = '\0';
358 QLIST_FOREACH(drv1, &bdrv_drivers, list) {
359 if (drv1->protocol_name &&
360 !strcmp(drv1->protocol_name, protocol)) {
361 return drv1;
362 }
363 }
364 return NULL;
365 }
366
367 static int find_image_format(const char *filename, BlockDriver **pdrv)
368 {
369 int ret, score, score_max;
370 BlockDriver *drv1, *drv;
371 uint8_t buf[2048];
372 BlockDriverState *bs;
373
374 ret = bdrv_file_open(&bs, filename, 0);
375 if (ret < 0) {
376 *pdrv = NULL;
377 return ret;
378 }
379
380 /* Return the raw BlockDriver * to scsi-generic devices or empty drives */
381 if (bs->sg || !bdrv_is_inserted(bs)) {
382 bdrv_delete(bs);
383 drv = bdrv_find_format("raw");
384 if (!drv) {
385 ret = -ENOENT;
386 }
387 *pdrv = drv;
388 return ret;
389 }
390
391 ret = bdrv_pread(bs, 0, buf, sizeof(buf));
392 bdrv_delete(bs);
393 if (ret < 0) {
394 *pdrv = NULL;
395 return ret;
396 }
397
398 score_max = 0;
399 drv = NULL;
400 QLIST_FOREACH(drv1, &bdrv_drivers, list) {
401 if (drv1->bdrv_probe) {
402 score = drv1->bdrv_probe(buf, ret, filename);
403 if (score > score_max) {
404 score_max = score;
405 drv = drv1;
406 }
407 }
408 }
409 if (!drv) {
410 ret = -ENOENT;
411 }
412 *pdrv = drv;
413 return ret;
414 }
415
416 /**
417 * Set the current 'total_sectors' value
418 */
419 static int refresh_total_sectors(BlockDriverState *bs, int64_t hint)
420 {
421 BlockDriver *drv = bs->drv;
422
423 /* Do not attempt drv->bdrv_getlength() on scsi-generic devices */
424 if (bs->sg)
425 return 0;
426
427 /* query actual device if possible, otherwise just trust the hint */
428 if (drv->bdrv_getlength) {
429 int64_t length = drv->bdrv_getlength(bs);
430 if (length < 0) {
431 return length;
432 }
433 hint = length >> BDRV_SECTOR_BITS;
434 }
435
436 bs->total_sectors = hint;
437 return 0;
438 }
439
440 /*
441 * Common part for opening disk images and files
442 */
443 static int bdrv_open_common(BlockDriverState *bs, const char *filename,
444 int flags, BlockDriver *drv)
445 {
446 int ret, open_flags;
447
448 assert(drv != NULL);
449
450 bs->file = NULL;
451 bs->total_sectors = 0;
452 bs->encrypted = 0;
453 bs->valid_key = 0;
454 bs->open_flags = flags;
455 /* buffer_alignment defaulted to 512, drivers can change this value */
456 bs->buffer_alignment = 512;
457
458 pstrcpy(bs->filename, sizeof(bs->filename), filename);
459
460 if (use_bdrv_whitelist && !bdrv_is_whitelisted(drv)) {
461 return -ENOTSUP;
462 }
463
464 bs->drv = drv;
465 bs->opaque = qemu_mallocz(drv->instance_size);
466
467 if (flags & BDRV_O_CACHE_WB)
468 bs->enable_write_cache = 1;
469
470 /*
471 * Clear flags that are internal to the block layer before opening the
472 * image.
473 */
474 open_flags = flags & ~(BDRV_O_SNAPSHOT | BDRV_O_NO_BACKING);
475
476 /*
477 * Snapshots should be writable.
478 */
479 if (bs->is_temporary) {
480 open_flags |= BDRV_O_RDWR;
481 }
482
483 /* Open the image, either directly or using a protocol */
484 if (drv->bdrv_file_open) {
485 ret = drv->bdrv_file_open(bs, filename, open_flags);
486 } else {
487 ret = bdrv_file_open(&bs->file, filename, open_flags);
488 if (ret >= 0) {
489 ret = drv->bdrv_open(bs, open_flags);
490 }
491 }
492
493 if (ret < 0) {
494 goto free_and_fail;
495 }
496
497 bs->keep_read_only = bs->read_only = !(open_flags & BDRV_O_RDWR);
498
499 ret = refresh_total_sectors(bs, bs->total_sectors);
500 if (ret < 0) {
501 goto free_and_fail;
502 }
503
504 #ifndef _WIN32
505 if (bs->is_temporary) {
506 unlink(filename);
507 }
508 #endif
509 return 0;
510
511 free_and_fail:
512 if (bs->file) {
513 bdrv_delete(bs->file);
514 bs->file = NULL;
515 }
516 qemu_free(bs->opaque);
517 bs->opaque = NULL;
518 bs->drv = NULL;
519 return ret;
520 }
521
522 /*
523 * Opens a file using a protocol (file, host_device, nbd, ...)
524 */
525 int bdrv_file_open(BlockDriverState **pbs, const char *filename, int flags)
526 {
527 BlockDriverState *bs;
528 BlockDriver *drv;
529 int ret;
530
531 drv = bdrv_find_protocol(filename);
532 if (!drv) {
533 return -ENOENT;
534 }
535
536 bs = bdrv_new("");
537 ret = bdrv_open_common(bs, filename, flags, drv);
538 if (ret < 0) {
539 bdrv_delete(bs);
540 return ret;
541 }
542 bs->growable = 1;
543 *pbs = bs;
544 return 0;
545 }
546
547 /*
548 * Opens a disk image (raw, qcow2, vmdk, ...)
549 */
550 int bdrv_open(BlockDriverState *bs, const char *filename, int flags,
551 BlockDriver *drv)
552 {
553 int ret;
554
555 if (flags & BDRV_O_SNAPSHOT) {
556 BlockDriverState *bs1;
557 int64_t total_size;
558 int is_protocol = 0;
559 BlockDriver *bdrv_qcow2;
560 QEMUOptionParameter *options;
561 char tmp_filename[PATH_MAX];
562 char backing_filename[PATH_MAX];
563
564 /* if snapshot, we create a temporary backing file and open it
565 instead of opening 'filename' directly */
566
567 /* if there is a backing file, use it */
568 bs1 = bdrv_new("");
569 ret = bdrv_open(bs1, filename, 0, drv);
570 if (ret < 0) {
571 bdrv_delete(bs1);
572 return ret;
573 }
574 total_size = bdrv_getlength(bs1) & BDRV_SECTOR_MASK;
575
576 if (bs1->drv && bs1->drv->protocol_name)
577 is_protocol = 1;
578
579 bdrv_delete(bs1);
580
581 get_tmp_filename(tmp_filename, sizeof(tmp_filename));
582
583 /* Real path is meaningless for protocols */
584 if (is_protocol)
585 snprintf(backing_filename, sizeof(backing_filename),
586 "%s", filename);
587 else if (!realpath(filename, backing_filename))
588 return -errno;
589
590 bdrv_qcow2 = bdrv_find_format("qcow2");
591 options = parse_option_parameters("", bdrv_qcow2->create_options, NULL);
592
593 set_option_parameter_int(options, BLOCK_OPT_SIZE, total_size);
594 set_option_parameter(options, BLOCK_OPT_BACKING_FILE, backing_filename);
595 if (drv) {
596 set_option_parameter(options, BLOCK_OPT_BACKING_FMT,
597 drv->format_name);
598 }
599
600 ret = bdrv_create(bdrv_qcow2, tmp_filename, options);
601 free_option_parameters(options);
602 if (ret < 0) {
603 return ret;
604 }
605
606 filename = tmp_filename;
607 drv = bdrv_qcow2;
608 bs->is_temporary = 1;
609 }
610
611 /* Find the right image format driver */
612 if (!drv) {
613 ret = find_image_format(filename, &drv);
614 }
615
616 if (!drv) {
617 goto unlink_and_fail;
618 }
619
620 /* Open the image */
621 ret = bdrv_open_common(bs, filename, flags, drv);
622 if (ret < 0) {
623 goto unlink_and_fail;
624 }
625
626 /* If there is a backing file, use it */
627 if ((flags & BDRV_O_NO_BACKING) == 0 && bs->backing_file[0] != '\0') {
628 char backing_filename[PATH_MAX];
629 int back_flags;
630 BlockDriver *back_drv = NULL;
631
632 bs->backing_hd = bdrv_new("");
633
634 if (path_has_protocol(bs->backing_file)) {
635 pstrcpy(backing_filename, sizeof(backing_filename),
636 bs->backing_file);
637 } else {
638 path_combine(backing_filename, sizeof(backing_filename),
639 filename, bs->backing_file);
640 }
641
642 if (bs->backing_format[0] != '\0') {
643 back_drv = bdrv_find_format(bs->backing_format);
644 }
645
646 /* backing files always opened read-only */
647 back_flags =
648 flags & ~(BDRV_O_RDWR | BDRV_O_SNAPSHOT | BDRV_O_NO_BACKING);
649
650 ret = bdrv_open(bs->backing_hd, backing_filename, back_flags, back_drv);
651 if (ret < 0) {
652 bdrv_close(bs);
653 return ret;
654 }
655 if (bs->is_temporary) {
656 bs->backing_hd->keep_read_only = !(flags & BDRV_O_RDWR);
657 } else {
658 /* base image inherits from "parent" */
659 bs->backing_hd->keep_read_only = bs->keep_read_only;
660 }
661 }
662
663 if (!bdrv_key_required(bs)) {
664 /* call the change callback */
665 bs->media_changed = 1;
666 if (bs->change_cb)
667 bs->change_cb(bs->change_opaque, CHANGE_MEDIA);
668 }
669
670 return 0;
671
672 unlink_and_fail:
673 if (bs->is_temporary) {
674 unlink(filename);
675 }
676 return ret;
677 }
678
679 void bdrv_close(BlockDriverState *bs)
680 {
681 if (bs->drv) {
682 if (bs == bs_snapshots) {
683 bs_snapshots = NULL;
684 }
685 if (bs->backing_hd) {
686 bdrv_delete(bs->backing_hd);
687 bs->backing_hd = NULL;
688 }
689 bs->drv->bdrv_close(bs);
690 qemu_free(bs->opaque);
691 #ifdef _WIN32
692 if (bs->is_temporary) {
693 unlink(bs->filename);
694 }
695 #endif
696 bs->opaque = NULL;
697 bs->drv = NULL;
698
699 if (bs->file != NULL) {
700 bdrv_close(bs->file);
701 }
702
703 /* call the change callback */
704 bs->media_changed = 1;
705 if (bs->change_cb)
706 bs->change_cb(bs->change_opaque, CHANGE_MEDIA);
707 }
708 }
709
710 void bdrv_close_all(void)
711 {
712 BlockDriverState *bs;
713
714 QTAILQ_FOREACH(bs, &bdrv_states, list) {
715 bdrv_close(bs);
716 }
717 }
718
719 /* make a BlockDriverState anonymous by removing from bdrv_state list.
720 Also, NULL terminate the device_name to prevent double remove */
721 void bdrv_make_anon(BlockDriverState *bs)
722 {
723 if (bs->device_name[0] != '\0') {
724 QTAILQ_REMOVE(&bdrv_states, bs, list);
725 }
726 bs->device_name[0] = '\0';
727 }
728
729 void bdrv_delete(BlockDriverState *bs)
730 {
731 assert(!bs->peer);
732
733 /* remove from list, if necessary */
734 bdrv_make_anon(bs);
735
736 bdrv_close(bs);
737 if (bs->file != NULL) {
738 bdrv_delete(bs->file);
739 }
740
741 assert(bs != bs_snapshots);
742 qemu_free(bs);
743 }
744
745 int bdrv_attach(BlockDriverState *bs, DeviceState *qdev)
746 {
747 if (bs->peer) {
748 return -EBUSY;
749 }
750 bs->peer = qdev;
751 return 0;
752 }
753
754 void bdrv_detach(BlockDriverState *bs, DeviceState *qdev)
755 {
756 assert(bs->peer == qdev);
757 bs->peer = NULL;
758 bs->change_cb = NULL;
759 bs->change_opaque = NULL;
760 }
761
762 DeviceState *bdrv_get_attached(BlockDriverState *bs)
763 {
764 return bs->peer;
765 }
766
767 /*
768 * Run consistency checks on an image
769 *
770 * Returns 0 if the check could be completed (it doesn't mean that the image is
771 * free of errors) or -errno when an internal error occurred. The results of the
772 * check are stored in res.
773 */
774 int bdrv_check(BlockDriverState *bs, BdrvCheckResult *res)
775 {
776 if (bs->drv->bdrv_check == NULL) {
777 return -ENOTSUP;
778 }
779
780 memset(res, 0, sizeof(*res));
781 return bs->drv->bdrv_check(bs, res);
782 }
783
784 #define COMMIT_BUF_SECTORS 2048
785
786 /* commit COW file into the raw image */
787 int bdrv_commit(BlockDriverState *bs)
788 {
789 BlockDriver *drv = bs->drv;
790 BlockDriver *backing_drv;
791 int64_t sector, total_sectors;
792 int n, ro, open_flags;
793 int ret = 0, rw_ret = 0;
794 uint8_t *buf;
795 char filename[1024];
796 BlockDriverState *bs_rw, *bs_ro;
797
798 if (!drv)
799 return -ENOMEDIUM;
800
801 if (!bs->backing_hd) {
802 return -ENOTSUP;
803 }
804
805 if (bs->backing_hd->keep_read_only) {
806 return -EACCES;
807 }
808
809 backing_drv = bs->backing_hd->drv;
810 ro = bs->backing_hd->read_only;
811 strncpy(filename, bs->backing_hd->filename, sizeof(filename));
812 open_flags = bs->backing_hd->open_flags;
813
814 if (ro) {
815 /* re-open as RW */
816 bdrv_delete(bs->backing_hd);
817 bs->backing_hd = NULL;
818 bs_rw = bdrv_new("");
819 rw_ret = bdrv_open(bs_rw, filename, open_flags | BDRV_O_RDWR,
820 backing_drv);
821 if (rw_ret < 0) {
822 bdrv_delete(bs_rw);
823 /* try to re-open read-only */
824 bs_ro = bdrv_new("");
825 ret = bdrv_open(bs_ro, filename, open_flags & ~BDRV_O_RDWR,
826 backing_drv);
827 if (ret < 0) {
828 bdrv_delete(bs_ro);
829 /* drive not functional anymore */
830 bs->drv = NULL;
831 return ret;
832 }
833 bs->backing_hd = bs_ro;
834 return rw_ret;
835 }
836 bs->backing_hd = bs_rw;
837 }
838
839 total_sectors = bdrv_getlength(bs) >> BDRV_SECTOR_BITS;
840 buf = qemu_malloc(COMMIT_BUF_SECTORS * BDRV_SECTOR_SIZE);
841
842 for (sector = 0; sector < total_sectors; sector += n) {
843 if (drv->bdrv_is_allocated(bs, sector, COMMIT_BUF_SECTORS, &n)) {
844
845 if (bdrv_read(bs, sector, buf, n) != 0) {
846 ret = -EIO;
847 goto ro_cleanup;
848 }
849
850 if (bdrv_write(bs->backing_hd, sector, buf, n) != 0) {
851 ret = -EIO;
852 goto ro_cleanup;
853 }
854 }
855 }
856
857 if (drv->bdrv_make_empty) {
858 ret = drv->bdrv_make_empty(bs);
859 bdrv_flush(bs);
860 }
861
862 /*
863 * Make sure all data we wrote to the backing device is actually
864 * stable on disk.
865 */
866 if (bs->backing_hd)
867 bdrv_flush(bs->backing_hd);
868
869 ro_cleanup:
870 qemu_free(buf);
871
872 if (ro) {
873 /* re-open as RO */
874 bdrv_delete(bs->backing_hd);
875 bs->backing_hd = NULL;
876 bs_ro = bdrv_new("");
877 ret = bdrv_open(bs_ro, filename, open_flags & ~BDRV_O_RDWR,
878 backing_drv);
879 if (ret < 0) {
880 bdrv_delete(bs_ro);
881 /* drive not functional anymore */
882 bs->drv = NULL;
883 return ret;
884 }
885 bs->backing_hd = bs_ro;
886 bs->backing_hd->keep_read_only = 0;
887 }
888
889 return ret;
890 }
891
892 void bdrv_commit_all(void)
893 {
894 BlockDriverState *bs;
895
896 QTAILQ_FOREACH(bs, &bdrv_states, list) {
897 bdrv_commit(bs);
898 }
899 }
900
901 /*
902 * Return values:
903 * 0 - success
904 * -EINVAL - backing format specified, but no file
905 * -ENOSPC - can't update the backing file because no space is left in the
906 * image file header
907 * -ENOTSUP - format driver doesn't support changing the backing file
908 */
909 int bdrv_change_backing_file(BlockDriverState *bs,
910 const char *backing_file, const char *backing_fmt)
911 {
912 BlockDriver *drv = bs->drv;
913
914 if (drv->bdrv_change_backing_file != NULL) {
915 return drv->bdrv_change_backing_file(bs, backing_file, backing_fmt);
916 } else {
917 return -ENOTSUP;
918 }
919 }
920
921 static int bdrv_check_byte_request(BlockDriverState *bs, int64_t offset,
922 size_t size)
923 {
924 int64_t len;
925
926 if (!bdrv_is_inserted(bs))
927 return -ENOMEDIUM;
928
929 if (bs->growable)
930 return 0;
931
932 len = bdrv_getlength(bs);
933
934 if (offset < 0)
935 return -EIO;
936
937 if ((offset > len) || (len - offset < size))
938 return -EIO;
939
940 return 0;
941 }
942
943 static int bdrv_check_request(BlockDriverState *bs, int64_t sector_num,
944 int nb_sectors)
945 {
946 return bdrv_check_byte_request(bs, sector_num * BDRV_SECTOR_SIZE,
947 nb_sectors * BDRV_SECTOR_SIZE);
948 }
949
950 static inline bool bdrv_has_async_rw(BlockDriver *drv)
951 {
952 return drv->bdrv_co_readv != bdrv_co_readv_em
953 || drv->bdrv_aio_readv != bdrv_aio_readv_em;
954 }
955
956 static inline bool bdrv_has_async_flush(BlockDriver *drv)
957 {
958 return drv->bdrv_aio_flush != bdrv_aio_flush_em;
959 }
960
961 /* return < 0 if error. See bdrv_write() for the return codes */
962 int bdrv_read(BlockDriverState *bs, int64_t sector_num,
963 uint8_t *buf, int nb_sectors)
964 {
965 BlockDriver *drv = bs->drv;
966
967 if (!drv)
968 return -ENOMEDIUM;
969
970 if (bdrv_has_async_rw(drv) && qemu_in_coroutine()) {
971 QEMUIOVector qiov;
972 struct iovec iov = {
973 .iov_base = (void *)buf,
974 .iov_len = nb_sectors * BDRV_SECTOR_SIZE,
975 };
976
977 qemu_iovec_init_external(&qiov, &iov, 1);
978 return bdrv_co_readv(bs, sector_num, nb_sectors, &qiov);
979 }
980
981 if (bdrv_check_request(bs, sector_num, nb_sectors))
982 return -EIO;
983
984 return drv->bdrv_read(bs, sector_num, buf, nb_sectors);
985 }
986
987 static void set_dirty_bitmap(BlockDriverState *bs, int64_t sector_num,
988 int nb_sectors, int dirty)
989 {
990 int64_t start, end;
991 unsigned long val, idx, bit;
992
993 start = sector_num / BDRV_SECTORS_PER_DIRTY_CHUNK;
994 end = (sector_num + nb_sectors - 1) / BDRV_SECTORS_PER_DIRTY_CHUNK;
995
996 for (; start <= end; start++) {
997 idx = start / (sizeof(unsigned long) * 8);
998 bit = start % (sizeof(unsigned long) * 8);
999 val = bs->dirty_bitmap[idx];
1000 if (dirty) {
1001 if (!(val & (1UL << bit))) {
1002 bs->dirty_count++;
1003 val |= 1UL << bit;
1004 }
1005 } else {
1006 if (val & (1UL << bit)) {
1007 bs->dirty_count--;
1008 val &= ~(1UL << bit);
1009 }
1010 }
1011 bs->dirty_bitmap[idx] = val;
1012 }
1013 }
1014
1015 /* Return < 0 if error. Important errors are:
1016 -EIO generic I/O error (may happen for all errors)
1017 -ENOMEDIUM No media inserted.
1018 -EINVAL Invalid sector number or nb_sectors
1019 -EACCES Trying to write a read-only device
1020 */
1021 int bdrv_write(BlockDriverState *bs, int64_t sector_num,
1022 const uint8_t *buf, int nb_sectors)
1023 {
1024 BlockDriver *drv = bs->drv;
1025
1026 if (!bs->drv)
1027 return -ENOMEDIUM;
1028
1029 if (bdrv_has_async_rw(drv) && qemu_in_coroutine()) {
1030 QEMUIOVector qiov;
1031 struct iovec iov = {
1032 .iov_base = (void *)buf,
1033 .iov_len = nb_sectors * BDRV_SECTOR_SIZE,
1034 };
1035
1036 qemu_iovec_init_external(&qiov, &iov, 1);
1037 return bdrv_co_writev(bs, sector_num, nb_sectors, &qiov);
1038 }
1039
1040 if (bs->read_only)
1041 return -EACCES;
1042 if (bdrv_check_request(bs, sector_num, nb_sectors))
1043 return -EIO;
1044
1045 if (bs->dirty_bitmap) {
1046 set_dirty_bitmap(bs, sector_num, nb_sectors, 1);
1047 }
1048
1049 if (bs->wr_highest_sector < sector_num + nb_sectors - 1) {
1050 bs->wr_highest_sector = sector_num + nb_sectors - 1;
1051 }
1052
1053 return drv->bdrv_write(bs, sector_num, buf, nb_sectors);
1054 }
1055
1056 int bdrv_pread(BlockDriverState *bs, int64_t offset,
1057 void *buf, int count1)
1058 {
1059 uint8_t tmp_buf[BDRV_SECTOR_SIZE];
1060 int len, nb_sectors, count;
1061 int64_t sector_num;
1062 int ret;
1063
1064 count = count1;
1065 /* first read to align to sector start */
1066 len = (BDRV_SECTOR_SIZE - offset) & (BDRV_SECTOR_SIZE - 1);
1067 if (len > count)
1068 len = count;
1069 sector_num = offset >> BDRV_SECTOR_BITS;
1070 if (len > 0) {
1071 if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0)
1072 return ret;
1073 memcpy(buf, tmp_buf + (offset & (BDRV_SECTOR_SIZE - 1)), len);
1074 count -= len;
1075 if (count == 0)
1076 return count1;
1077 sector_num++;
1078 buf += len;
1079 }
1080
1081 /* read the sectors "in place" */
1082 nb_sectors = count >> BDRV_SECTOR_BITS;
1083 if (nb_sectors > 0) {
1084 if ((ret = bdrv_read(bs, sector_num, buf, nb_sectors)) < 0)
1085 return ret;
1086 sector_num += nb_sectors;
1087 len = nb_sectors << BDRV_SECTOR_BITS;
1088 buf += len;
1089 count -= len;
1090 }
1091
1092 /* add data from the last sector */
1093 if (count > 0) {
1094 if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0)
1095 return ret;
1096 memcpy(buf, tmp_buf, count);
1097 }
1098 return count1;
1099 }
1100
1101 int bdrv_pwrite(BlockDriverState *bs, int64_t offset,
1102 const void *buf, int count1)
1103 {
1104 uint8_t tmp_buf[BDRV_SECTOR_SIZE];
1105 int len, nb_sectors, count;
1106 int64_t sector_num;
1107 int ret;
1108
1109 count = count1;
1110 /* first write to align to sector start */
1111 len = (BDRV_SECTOR_SIZE - offset) & (BDRV_SECTOR_SIZE - 1);
1112 if (len > count)
1113 len = count;
1114 sector_num = offset >> BDRV_SECTOR_BITS;
1115 if (len > 0) {
1116 if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0)
1117 return ret;
1118 memcpy(tmp_buf + (offset & (BDRV_SECTOR_SIZE - 1)), buf, len);
1119 if ((ret = bdrv_write(bs, sector_num, tmp_buf, 1)) < 0)
1120 return ret;
1121 count -= len;
1122 if (count == 0)
1123 return count1;
1124 sector_num++;
1125 buf += len;
1126 }
1127
1128 /* write the sectors "in place" */
1129 nb_sectors = count >> BDRV_SECTOR_BITS;
1130 if (nb_sectors > 0) {
1131 if ((ret = bdrv_write(bs, sector_num, buf, nb_sectors)) < 0)
1132 return ret;
1133 sector_num += nb_sectors;
1134 len = nb_sectors << BDRV_SECTOR_BITS;
1135 buf += len;
1136 count -= len;
1137 }
1138
1139 /* add data from the last sector */
1140 if (count > 0) {
1141 if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0)
1142 return ret;
1143 memcpy(tmp_buf, buf, count);
1144 if ((ret = bdrv_write(bs, sector_num, tmp_buf, 1)) < 0)
1145 return ret;
1146 }
1147 return count1;
1148 }
1149
1150 /*
1151 * Writes to the file and ensures that no writes are reordered across this
1152 * request (acts as a barrier)
1153 *
1154 * Returns 0 on success, -errno in error cases.
1155 */
1156 int bdrv_pwrite_sync(BlockDriverState *bs, int64_t offset,
1157 const void *buf, int count)
1158 {
1159 int ret;
1160
1161 ret = bdrv_pwrite(bs, offset, buf, count);
1162 if (ret < 0) {
1163 return ret;
1164 }
1165
1166 /* No flush needed for cache=writethrough, it uses O_DSYNC */
1167 if ((bs->open_flags & BDRV_O_CACHE_MASK) != 0) {
1168 bdrv_flush(bs);
1169 }
1170
1171 return 0;
1172 }
1173
1174 int coroutine_fn bdrv_co_readv(BlockDriverState *bs, int64_t sector_num,
1175 int nb_sectors, QEMUIOVector *qiov)
1176 {
1177 BlockDriver *drv = bs->drv;
1178
1179 trace_bdrv_co_readv(bs, sector_num, nb_sectors);
1180
1181 if (!drv) {
1182 return -ENOMEDIUM;
1183 }
1184 if (bdrv_check_request(bs, sector_num, nb_sectors)) {
1185 return -EIO;
1186 }
1187
1188 return drv->bdrv_co_readv(bs, sector_num, nb_sectors, qiov);
1189 }
1190
1191 int coroutine_fn bdrv_co_writev(BlockDriverState *bs, int64_t sector_num,
1192 int nb_sectors, QEMUIOVector *qiov)
1193 {
1194 BlockDriver *drv = bs->drv;
1195
1196 trace_bdrv_co_writev(bs, sector_num, nb_sectors);
1197
1198 if (!bs->drv) {
1199 return -ENOMEDIUM;
1200 }
1201 if (bs->read_only) {
1202 return -EACCES;
1203 }
1204 if (bdrv_check_request(bs, sector_num, nb_sectors)) {
1205 return -EIO;
1206 }
1207
1208 if (bs->dirty_bitmap) {
1209 set_dirty_bitmap(bs, sector_num, nb_sectors, 1);
1210 }
1211
1212 if (bs->wr_highest_sector < sector_num + nb_sectors - 1) {
1213 bs->wr_highest_sector = sector_num + nb_sectors - 1;
1214 }
1215
1216 return drv->bdrv_co_writev(bs, sector_num, nb_sectors, qiov);
1217 }
1218
1219 /**
1220 * Truncate file to 'offset' bytes (needed only for file protocols)
1221 */
1222 int bdrv_truncate(BlockDriverState *bs, int64_t offset)
1223 {
1224 BlockDriver *drv = bs->drv;
1225 int ret;
1226 if (!drv)
1227 return -ENOMEDIUM;
1228 if (!drv->bdrv_truncate)
1229 return -ENOTSUP;
1230 if (bs->read_only)
1231 return -EACCES;
1232 if (bdrv_in_use(bs))
1233 return -EBUSY;
1234 ret = drv->bdrv_truncate(bs, offset);
1235 if (ret == 0) {
1236 ret = refresh_total_sectors(bs, offset >> BDRV_SECTOR_BITS);
1237 if (bs->change_cb) {
1238 bs->change_cb(bs->change_opaque, CHANGE_SIZE);
1239 }
1240 }
1241 return ret;
1242 }
1243
1244 /**
1245 * Length of a allocated file in bytes. Sparse files are counted by actual
1246 * allocated space. Return < 0 if error or unknown.
1247 */
1248 int64_t bdrv_get_allocated_file_size(BlockDriverState *bs)
1249 {
1250 BlockDriver *drv = bs->drv;
1251 if (!drv) {
1252 return -ENOMEDIUM;
1253 }
1254 if (drv->bdrv_get_allocated_file_size) {
1255 return drv->bdrv_get_allocated_file_size(bs);
1256 }
1257 if (bs->file) {
1258 return bdrv_get_allocated_file_size(bs->file);
1259 }
1260 return -ENOTSUP;
1261 }
1262
1263 /**
1264 * Length of a file in bytes. Return < 0 if error or unknown.
1265 */
1266 int64_t bdrv_getlength(BlockDriverState *bs)
1267 {
1268 BlockDriver *drv = bs->drv;
1269 if (!drv)
1270 return -ENOMEDIUM;
1271
1272 if (bs->growable || bs->removable) {
1273 if (drv->bdrv_getlength) {
1274 return drv->bdrv_getlength(bs);
1275 }
1276 }
1277 return bs->total_sectors * BDRV_SECTOR_SIZE;
1278 }
1279
1280 /* return 0 as number of sectors if no device present or error */
1281 void bdrv_get_geometry(BlockDriverState *bs, uint64_t *nb_sectors_ptr)
1282 {
1283 int64_t length;
1284 length = bdrv_getlength(bs);
1285 if (length < 0)
1286 length = 0;
1287 else
1288 length = length >> BDRV_SECTOR_BITS;
1289 *nb_sectors_ptr = length;
1290 }
1291
1292 struct partition {
1293 uint8_t boot_ind; /* 0x80 - active */
1294 uint8_t head; /* starting head */
1295 uint8_t sector; /* starting sector */
1296 uint8_t cyl; /* starting cylinder */
1297 uint8_t sys_ind; /* What partition type */
1298 uint8_t end_head; /* end head */
1299 uint8_t end_sector; /* end sector */
1300 uint8_t end_cyl; /* end cylinder */
1301 uint32_t start_sect; /* starting sector counting from 0 */
1302 uint32_t nr_sects; /* nr of sectors in partition */
1303 } __attribute__((packed));
1304
1305 /* try to guess the disk logical geometry from the MSDOS partition table. Return 0 if OK, -1 if could not guess */
1306 static int guess_disk_lchs(BlockDriverState *bs,
1307 int *pcylinders, int *pheads, int *psectors)
1308 {
1309 uint8_t buf[BDRV_SECTOR_SIZE];
1310 int ret, i, heads, sectors, cylinders;
1311 struct partition *p;
1312 uint32_t nr_sects;
1313 uint64_t nb_sectors;
1314
1315 bdrv_get_geometry(bs, &nb_sectors);
1316
1317 ret = bdrv_read(bs, 0, buf, 1);
1318 if (ret < 0)
1319 return -1;
1320 /* test msdos magic */
1321 if (buf[510] != 0x55 || buf[511] != 0xaa)
1322 return -1;
1323 for(i = 0; i < 4; i++) {
1324 p = ((struct partition *)(buf + 0x1be)) + i;
1325 nr_sects = le32_to_cpu(p->nr_sects);
1326 if (nr_sects && p->end_head) {
1327 /* We make the assumption that the partition terminates on
1328 a cylinder boundary */
1329 heads = p->end_head + 1;
1330 sectors = p->end_sector & 63;
1331 if (sectors == 0)
1332 continue;
1333 cylinders = nb_sectors / (heads * sectors);
1334 if (cylinders < 1 || cylinders > 16383)
1335 continue;
1336 *pheads = heads;
1337 *psectors = sectors;
1338 *pcylinders = cylinders;
1339 #if 0
1340 printf("guessed geometry: LCHS=%d %d %d\n",
1341 cylinders, heads, sectors);
1342 #endif
1343 return 0;
1344 }
1345 }
1346 return -1;
1347 }
1348
1349 void bdrv_guess_geometry(BlockDriverState *bs, int *pcyls, int *pheads, int *psecs)
1350 {
1351 int translation, lba_detected = 0;
1352 int cylinders, heads, secs;
1353 uint64_t nb_sectors;
1354
1355 /* if a geometry hint is available, use it */
1356 bdrv_get_geometry(bs, &nb_sectors);
1357 bdrv_get_geometry_hint(bs, &cylinders, &heads, &secs);
1358 translation = bdrv_get_translation_hint(bs);
1359 if (cylinders != 0) {
1360 *pcyls = cylinders;
1361 *pheads = heads;
1362 *psecs = secs;
1363 } else {
1364 if (guess_disk_lchs(bs, &cylinders, &heads, &secs) == 0) {
1365 if (heads > 16) {
1366 /* if heads > 16, it means that a BIOS LBA
1367 translation was active, so the default
1368 hardware geometry is OK */
1369 lba_detected = 1;
1370 goto default_geometry;
1371 } else {
1372 *pcyls = cylinders;
1373 *pheads = heads;
1374 *psecs = secs;
1375 /* disable any translation to be in sync with
1376 the logical geometry */
1377 if (translation == BIOS_ATA_TRANSLATION_AUTO) {
1378 bdrv_set_translation_hint(bs,
1379 BIOS_ATA_TRANSLATION_NONE);
1380 }
1381 }
1382 } else {
1383 default_geometry:
1384 /* if no geometry, use a standard physical disk geometry */
1385 cylinders = nb_sectors / (16 * 63);
1386
1387 if (cylinders > 16383)
1388 cylinders = 16383;
1389 else if (cylinders < 2)
1390 cylinders = 2;
1391 *pcyls = cylinders;
1392 *pheads = 16;
1393 *psecs = 63;
1394 if ((lba_detected == 1) && (translation == BIOS_ATA_TRANSLATION_AUTO)) {
1395 if ((*pcyls * *pheads) <= 131072) {
1396 bdrv_set_translation_hint(bs,
1397 BIOS_ATA_TRANSLATION_LARGE);
1398 } else {
1399 bdrv_set_translation_hint(bs,
1400 BIOS_ATA_TRANSLATION_LBA);
1401 }
1402 }
1403 }
1404 bdrv_set_geometry_hint(bs, *pcyls, *pheads, *psecs);
1405 }
1406 }
1407
1408 void bdrv_set_geometry_hint(BlockDriverState *bs,
1409 int cyls, int heads, int secs)
1410 {
1411 bs->cyls = cyls;
1412 bs->heads = heads;
1413 bs->secs = secs;
1414 }
1415
1416 void bdrv_set_translation_hint(BlockDriverState *bs, int translation)
1417 {
1418 bs->translation = translation;
1419 }
1420
1421 void bdrv_get_geometry_hint(BlockDriverState *bs,
1422 int *pcyls, int *pheads, int *psecs)
1423 {
1424 *pcyls = bs->cyls;
1425 *pheads = bs->heads;
1426 *psecs = bs->secs;
1427 }
1428
1429 /* Recognize floppy formats */
1430 typedef struct FDFormat {
1431 FDriveType drive;
1432 uint8_t last_sect;
1433 uint8_t max_track;
1434 uint8_t max_head;
1435 } FDFormat;
1436
1437 static const FDFormat fd_formats[] = {
1438 /* First entry is default format */
1439 /* 1.44 MB 3"1/2 floppy disks */
1440 { FDRIVE_DRV_144, 18, 80, 1, },
1441 { FDRIVE_DRV_144, 20, 80, 1, },
1442 { FDRIVE_DRV_144, 21, 80, 1, },
1443 { FDRIVE_DRV_144, 21, 82, 1, },
1444 { FDRIVE_DRV_144, 21, 83, 1, },
1445 { FDRIVE_DRV_144, 22, 80, 1, },
1446 { FDRIVE_DRV_144, 23, 80, 1, },
1447 { FDRIVE_DRV_144, 24, 80, 1, },
1448 /* 2.88 MB 3"1/2 floppy disks */
1449 { FDRIVE_DRV_288, 36, 80, 1, },
1450 { FDRIVE_DRV_288, 39, 80, 1, },
1451 { FDRIVE_DRV_288, 40, 80, 1, },
1452 { FDRIVE_DRV_288, 44, 80, 1, },
1453 { FDRIVE_DRV_288, 48, 80, 1, },
1454 /* 720 kB 3"1/2 floppy disks */
1455 { FDRIVE_DRV_144, 9, 80, 1, },
1456 { FDRIVE_DRV_144, 10, 80, 1, },
1457 { FDRIVE_DRV_144, 10, 82, 1, },
1458 { FDRIVE_DRV_144, 10, 83, 1, },
1459 { FDRIVE_DRV_144, 13, 80, 1, },
1460 { FDRIVE_DRV_144, 14, 80, 1, },
1461 /* 1.2 MB 5"1/4 floppy disks */
1462 { FDRIVE_DRV_120, 15, 80, 1, },
1463 { FDRIVE_DRV_120, 18, 80, 1, },
1464 { FDRIVE_DRV_120, 18, 82, 1, },
1465 { FDRIVE_DRV_120, 18, 83, 1, },
1466 { FDRIVE_DRV_120, 20, 80, 1, },
1467 /* 720 kB 5"1/4 floppy disks */
1468 { FDRIVE_DRV_120, 9, 80, 1, },
1469 { FDRIVE_DRV_120, 11, 80, 1, },
1470 /* 360 kB 5"1/4 floppy disks */
1471 { FDRIVE_DRV_120, 9, 40, 1, },
1472 { FDRIVE_DRV_120, 9, 40, 0, },
1473 { FDRIVE_DRV_120, 10, 41, 1, },
1474 { FDRIVE_DRV_120, 10, 42, 1, },
1475 /* 320 kB 5"1/4 floppy disks */
1476 { FDRIVE_DRV_120, 8, 40, 1, },
1477 { FDRIVE_DRV_120, 8, 40, 0, },
1478 /* 360 kB must match 5"1/4 better than 3"1/2... */
1479 { FDRIVE_DRV_144, 9, 80, 0, },
1480 /* end */
1481 { FDRIVE_DRV_NONE, -1, -1, 0, },
1482 };
1483
1484 void bdrv_get_floppy_geometry_hint(BlockDriverState *bs, int *nb_heads,
1485 int *max_track, int *last_sect,
1486 FDriveType drive_in, FDriveType *drive)
1487 {
1488 const FDFormat *parse;
1489 uint64_t nb_sectors, size;
1490 int i, first_match, match;
1491
1492 bdrv_get_geometry_hint(bs, nb_heads, max_track, last_sect);
1493 if (*nb_heads != 0 && *max_track != 0 && *last_sect != 0) {
1494 /* User defined disk */
1495 } else {
1496 bdrv_get_geometry(bs, &nb_sectors);
1497 match = -1;
1498 first_match = -1;
1499 for (i = 0; ; i++) {
1500 parse = &fd_formats[i];
1501 if (parse->drive == FDRIVE_DRV_NONE) {
1502 break;
1503 }
1504 if (drive_in == parse->drive ||
1505 drive_in == FDRIVE_DRV_NONE) {
1506 size = (parse->max_head + 1) * parse->max_track *
1507 parse->last_sect;
1508 if (nb_sectors == size) {
1509 match = i;
1510 break;
1511 }
1512 if (first_match == -1) {
1513 first_match = i;
1514 }
1515 }
1516 }
1517 if (match == -1) {
1518 if (first_match == -1) {
1519 match = 1;
1520 } else {
1521 match = first_match;
1522 }
1523 parse = &fd_formats[match];
1524 }
1525 *nb_heads = parse->max_head + 1;
1526 *max_track = parse->max_track;
1527 *last_sect = parse->last_sect;
1528 *drive = parse->drive;
1529 }
1530 }
1531
1532 int bdrv_get_translation_hint(BlockDriverState *bs)
1533 {
1534 return bs->translation;
1535 }
1536
1537 void bdrv_set_on_error(BlockDriverState *bs, BlockErrorAction on_read_error,
1538 BlockErrorAction on_write_error)
1539 {
1540 bs->on_read_error = on_read_error;
1541 bs->on_write_error = on_write_error;
1542 }
1543
1544 BlockErrorAction bdrv_get_on_error(BlockDriverState *bs, int is_read)
1545 {
1546 return is_read ? bs->on_read_error : bs->on_write_error;
1547 }
1548
1549 void bdrv_set_removable(BlockDriverState *bs, int removable)
1550 {
1551 bs->removable = removable;
1552 if (removable && bs == bs_snapshots) {
1553 bs_snapshots = NULL;
1554 }
1555 }
1556
1557 int bdrv_is_removable(BlockDriverState *bs)
1558 {
1559 return bs->removable;
1560 }
1561
1562 int bdrv_is_read_only(BlockDriverState *bs)
1563 {
1564 return bs->read_only;
1565 }
1566
1567 int bdrv_is_sg(BlockDriverState *bs)
1568 {
1569 return bs->sg;
1570 }
1571
1572 int bdrv_enable_write_cache(BlockDriverState *bs)
1573 {
1574 return bs->enable_write_cache;
1575 }
1576
1577 /* XXX: no longer used */
1578 void bdrv_set_change_cb(BlockDriverState *bs,
1579 void (*change_cb)(void *opaque, int reason),
1580 void *opaque)
1581 {
1582 bs->change_cb = change_cb;
1583 bs->change_opaque = opaque;
1584 }
1585
1586 int bdrv_is_encrypted(BlockDriverState *bs)
1587 {
1588 if (bs->backing_hd && bs->backing_hd->encrypted)
1589 return 1;
1590 return bs->encrypted;
1591 }
1592
1593 int bdrv_key_required(BlockDriverState *bs)
1594 {
1595 BlockDriverState *backing_hd = bs->backing_hd;
1596
1597 if (backing_hd && backing_hd->encrypted && !backing_hd->valid_key)
1598 return 1;
1599 return (bs->encrypted && !bs->valid_key);
1600 }
1601
1602 int bdrv_set_key(BlockDriverState *bs, const char *key)
1603 {
1604 int ret;
1605 if (bs->backing_hd && bs->backing_hd->encrypted) {
1606 ret = bdrv_set_key(bs->backing_hd, key);
1607 if (ret < 0)
1608 return ret;
1609 if (!bs->encrypted)
1610 return 0;
1611 }
1612 if (!bs->encrypted) {
1613 return -EINVAL;
1614 } else if (!bs->drv || !bs->drv->bdrv_set_key) {
1615 return -ENOMEDIUM;
1616 }
1617 ret = bs->drv->bdrv_set_key(bs, key);
1618 if (ret < 0) {
1619 bs->valid_key = 0;
1620 } else if (!bs->valid_key) {
1621 bs->valid_key = 1;
1622 /* call the change callback now, we skipped it on open */
1623 bs->media_changed = 1;
1624 if (bs->change_cb)
1625 bs->change_cb(bs->change_opaque, CHANGE_MEDIA);
1626 }
1627 return ret;
1628 }
1629
1630 void bdrv_get_format(BlockDriverState *bs, char *buf, int buf_size)
1631 {
1632 if (!bs->drv) {
1633 buf[0] = '\0';
1634 } else {
1635 pstrcpy(buf, buf_size, bs->drv->format_name);
1636 }
1637 }
1638
1639 void bdrv_iterate_format(void (*it)(void *opaque, const char *name),
1640 void *opaque)
1641 {
1642 BlockDriver *drv;
1643
1644 QLIST_FOREACH(drv, &bdrv_drivers, list) {
1645 it(opaque, drv->format_name);
1646 }
1647 }
1648
1649 BlockDriverState *bdrv_find(const char *name)
1650 {
1651 BlockDriverState *bs;
1652
1653 QTAILQ_FOREACH(bs, &bdrv_states, list) {
1654 if (!strcmp(name, bs->device_name)) {
1655 return bs;
1656 }
1657 }
1658 return NULL;
1659 }
1660
1661 BlockDriverState *bdrv_next(BlockDriverState *bs)
1662 {
1663 if (!bs) {
1664 return QTAILQ_FIRST(&bdrv_states);
1665 }
1666 return QTAILQ_NEXT(bs, list);
1667 }
1668
1669 void bdrv_iterate(void (*it)(void *opaque, BlockDriverState *bs), void *opaque)
1670 {
1671 BlockDriverState *bs;
1672
1673 QTAILQ_FOREACH(bs, &bdrv_states, list) {
1674 it(opaque, bs);
1675 }
1676 }
1677
1678 const char *bdrv_get_device_name(BlockDriverState *bs)
1679 {
1680 return bs->device_name;
1681 }
1682
1683 int bdrv_flush(BlockDriverState *bs)
1684 {
1685 if (bs->open_flags & BDRV_O_NO_FLUSH) {
1686 return 0;
1687 }
1688
1689 if (bs->drv && bdrv_has_async_flush(bs->drv) && qemu_in_coroutine()) {
1690 return bdrv_co_flush_em(bs);
1691 }
1692
1693 if (bs->drv && bs->drv->bdrv_flush) {
1694 return bs->drv->bdrv_flush(bs);
1695 }
1696
1697 /*
1698 * Some block drivers always operate in either writethrough or unsafe mode
1699 * and don't support bdrv_flush therefore. Usually qemu doesn't know how
1700 * the server works (because the behaviour is hardcoded or depends on
1701 * server-side configuration), so we can't ensure that everything is safe
1702 * on disk. Returning an error doesn't work because that would break guests
1703 * even if the server operates in writethrough mode.
1704 *
1705 * Let's hope the user knows what he's doing.
1706 */
1707 return 0;
1708 }
1709
1710 void bdrv_flush_all(void)
1711 {
1712 BlockDriverState *bs;
1713
1714 QTAILQ_FOREACH(bs, &bdrv_states, list) {
1715 if (bs->drv && !bdrv_is_read_only(bs) &&
1716 (!bdrv_is_removable(bs) || bdrv_is_inserted(bs))) {
1717 bdrv_flush(bs);
1718 }
1719 }
1720 }
1721
1722 int bdrv_has_zero_init(BlockDriverState *bs)
1723 {
1724 assert(bs->drv);
1725
1726 if (bs->drv->bdrv_has_zero_init) {
1727 return bs->drv->bdrv_has_zero_init(bs);
1728 }
1729
1730 return 1;
1731 }
1732
1733 int bdrv_discard(BlockDriverState *bs, int64_t sector_num, int nb_sectors)
1734 {
1735 if (!bs->drv) {
1736 return -ENOMEDIUM;
1737 }
1738 if (!bs->drv->bdrv_discard) {
1739 return 0;
1740 }
1741 return bs->drv->bdrv_discard(bs, sector_num, nb_sectors);
1742 }
1743
1744 /*
1745 * Returns true iff the specified sector is present in the disk image. Drivers
1746 * not implementing the functionality are assumed to not support backing files,
1747 * hence all their sectors are reported as allocated.
1748 *
1749 * 'pnum' is set to the number of sectors (including and immediately following
1750 * the specified sector) that are known to be in the same
1751 * allocated/unallocated state.
1752 *
1753 * 'nb_sectors' is the max value 'pnum' should be set to.
1754 */
1755 int bdrv_is_allocated(BlockDriverState *bs, int64_t sector_num, int nb_sectors,
1756 int *pnum)
1757 {
1758 int64_t n;
1759 if (!bs->drv->bdrv_is_allocated) {
1760 if (sector_num >= bs->total_sectors) {
1761 *pnum = 0;
1762 return 0;
1763 }
1764 n = bs->total_sectors - sector_num;
1765 *pnum = (n < nb_sectors) ? (n) : (nb_sectors);
1766 return 1;
1767 }
1768 return bs->drv->bdrv_is_allocated(bs, sector_num, nb_sectors, pnum);
1769 }
1770
1771 void bdrv_mon_event(const BlockDriverState *bdrv,
1772 BlockMonEventAction action, int is_read)
1773 {
1774 QObject *data;
1775 const char *action_str;
1776
1777 switch (action) {
1778 case BDRV_ACTION_REPORT:
1779 action_str = "report";
1780 break;
1781 case BDRV_ACTION_IGNORE:
1782 action_str = "ignore";
1783 break;
1784 case BDRV_ACTION_STOP:
1785 action_str = "stop";
1786 break;
1787 default:
1788 abort();
1789 }
1790
1791 data = qobject_from_jsonf("{ 'device': %s, 'action': %s, 'operation': %s }",
1792 bdrv->device_name,
1793 action_str,
1794 is_read ? "read" : "write");
1795 monitor_protocol_event(QEVENT_BLOCK_IO_ERROR, data);
1796
1797 qobject_decref(data);
1798 }
1799
1800 static void bdrv_print_dict(QObject *obj, void *opaque)
1801 {
1802 QDict *bs_dict;
1803 Monitor *mon = opaque;
1804
1805 bs_dict = qobject_to_qdict(obj);
1806
1807 monitor_printf(mon, "%s: removable=%d",
1808 qdict_get_str(bs_dict, "device"),
1809 qdict_get_bool(bs_dict, "removable"));
1810
1811 if (qdict_get_bool(bs_dict, "removable")) {
1812 monitor_printf(mon, " locked=%d", qdict_get_bool(bs_dict, "locked"));
1813 }
1814
1815 if (qdict_haskey(bs_dict, "inserted")) {
1816 QDict *qdict = qobject_to_qdict(qdict_get(bs_dict, "inserted"));
1817
1818 monitor_printf(mon, " file=");
1819 monitor_print_filename(mon, qdict_get_str(qdict, "file"));
1820 if (qdict_haskey(qdict, "backing_file")) {
1821 monitor_printf(mon, " backing_file=");
1822 monitor_print_filename(mon, qdict_get_str(qdict, "backing_file"));
1823 }
1824 monitor_printf(mon, " ro=%d drv=%s encrypted=%d",
1825 qdict_get_bool(qdict, "ro"),
1826 qdict_get_str(qdict, "drv"),
1827 qdict_get_bool(qdict, "encrypted"));
1828 } else {
1829 monitor_printf(mon, " [not inserted]");
1830 }
1831
1832 monitor_printf(mon, "\n");
1833 }
1834
1835 void bdrv_info_print(Monitor *mon, const QObject *data)
1836 {
1837 qlist_iter(qobject_to_qlist(data), bdrv_print_dict, mon);
1838 }
1839
1840 void bdrv_info(Monitor *mon, QObject **ret_data)
1841 {
1842 QList *bs_list;
1843 BlockDriverState *bs;
1844
1845 bs_list = qlist_new();
1846
1847 QTAILQ_FOREACH(bs, &bdrv_states, list) {
1848 QObject *bs_obj;
1849
1850 bs_obj = qobject_from_jsonf("{ 'device': %s, 'type': 'unknown', "
1851 "'removable': %i, 'locked': %i }",
1852 bs->device_name, bs->removable,
1853 bs->locked);
1854
1855 if (bs->drv) {
1856 QObject *obj;
1857 QDict *bs_dict = qobject_to_qdict(bs_obj);
1858
1859 obj = qobject_from_jsonf("{ 'file': %s, 'ro': %i, 'drv': %s, "
1860 "'encrypted': %i }",
1861 bs->filename, bs->read_only,
1862 bs->drv->format_name,
1863 bdrv_is_encrypted(bs));
1864 if (bs->backing_file[0] != '\0') {
1865 QDict *qdict = qobject_to_qdict(obj);
1866 qdict_put(qdict, "backing_file",
1867 qstring_from_str(bs->backing_file));
1868 }
1869
1870 qdict_put_obj(bs_dict, "inserted", obj);
1871 }
1872 qlist_append_obj(bs_list, bs_obj);
1873 }
1874
1875 *ret_data = QOBJECT(bs_list);
1876 }
1877
1878 static void bdrv_stats_iter(QObject *data, void *opaque)
1879 {
1880 QDict *qdict;
1881 Monitor *mon = opaque;
1882
1883 qdict = qobject_to_qdict(data);
1884 monitor_printf(mon, "%s:", qdict_get_str(qdict, "device"));
1885
1886 qdict = qobject_to_qdict(qdict_get(qdict, "stats"));
1887 monitor_printf(mon, " rd_bytes=%" PRId64
1888 " wr_bytes=%" PRId64
1889 " rd_operations=%" PRId64
1890 " wr_operations=%" PRId64
1891 "\n",
1892 qdict_get_int(qdict, "rd_bytes"),
1893 qdict_get_int(qdict, "wr_bytes"),
1894 qdict_get_int(qdict, "rd_operations"),
1895 qdict_get_int(qdict, "wr_operations"));
1896 }
1897
1898 void bdrv_stats_print(Monitor *mon, const QObject *data)
1899 {
1900 qlist_iter(qobject_to_qlist(data), bdrv_stats_iter, mon);
1901 }
1902
1903 static QObject* bdrv_info_stats_bs(BlockDriverState *bs)
1904 {
1905 QObject *res;
1906 QDict *dict;
1907
1908 res = qobject_from_jsonf("{ 'stats': {"
1909 "'rd_bytes': %" PRId64 ","
1910 "'wr_bytes': %" PRId64 ","
1911 "'rd_operations': %" PRId64 ","
1912 "'wr_operations': %" PRId64 ","
1913 "'wr_highest_offset': %" PRId64
1914 "} }",
1915 bs->rd_bytes, bs->wr_bytes,
1916 bs->rd_ops, bs->wr_ops,
1917 bs->wr_highest_sector *
1918 (uint64_t)BDRV_SECTOR_SIZE);
1919 dict = qobject_to_qdict(res);
1920
1921 if (*bs->device_name) {
1922 qdict_put(dict, "device", qstring_from_str(bs->device_name));
1923 }
1924
1925 if (bs->file) {
1926 QObject *parent = bdrv_info_stats_bs(bs->file);
1927 qdict_put_obj(dict, "parent", parent);
1928 }
1929
1930 return res;
1931 }
1932
1933 void bdrv_info_stats(Monitor *mon, QObject **ret_data)
1934 {
1935 QObject *obj;
1936 QList *devices;
1937 BlockDriverState *bs;
1938
1939 devices = qlist_new();
1940
1941 QTAILQ_FOREACH(bs, &bdrv_states, list) {
1942 obj = bdrv_info_stats_bs(bs);
1943 qlist_append_obj(devices, obj);
1944 }
1945
1946 *ret_data = QOBJECT(devices);
1947 }
1948
1949 const char *bdrv_get_encrypted_filename(BlockDriverState *bs)
1950 {
1951 if (bs->backing_hd && bs->backing_hd->encrypted)
1952 return bs->backing_file;
1953 else if (bs->encrypted)
1954 return bs->filename;
1955 else
1956 return NULL;
1957 }
1958
1959 void bdrv_get_backing_filename(BlockDriverState *bs,
1960 char *filename, int filename_size)
1961 {
1962 if (!bs->backing_file) {
1963 pstrcpy(filename, filename_size, "");
1964 } else {
1965 pstrcpy(filename, filename_size, bs->backing_file);
1966 }
1967 }
1968
1969 int bdrv_write_compressed(BlockDriverState *bs, int64_t sector_num,
1970 const uint8_t *buf, int nb_sectors)
1971 {
1972 BlockDriver *drv = bs->drv;
1973 if (!drv)
1974 return -ENOMEDIUM;
1975 if (!drv->bdrv_write_compressed)
1976 return -ENOTSUP;
1977 if (bdrv_check_request(bs, sector_num, nb_sectors))
1978 return -EIO;
1979
1980 if (bs->dirty_bitmap) {
1981 set_dirty_bitmap(bs, sector_num, nb_sectors, 1);
1982 }
1983
1984 return drv->bdrv_write_compressed(bs, sector_num, buf, nb_sectors);
1985 }
1986
1987 int bdrv_get_info(BlockDriverState *bs, BlockDriverInfo *bdi)
1988 {
1989 BlockDriver *drv = bs->drv;
1990 if (!drv)
1991 return -ENOMEDIUM;
1992 if (!drv->bdrv_get_info)
1993 return -ENOTSUP;
1994 memset(bdi, 0, sizeof(*bdi));
1995 return drv->bdrv_get_info(bs, bdi);
1996 }
1997
1998 int bdrv_save_vmstate(BlockDriverState *bs, const uint8_t *buf,
1999 int64_t pos, int size)
2000 {
2001 BlockDriver *drv = bs->drv;
2002 if (!drv)
2003 return -ENOMEDIUM;
2004 if (drv->bdrv_save_vmstate)
2005 return drv->bdrv_save_vmstate(bs, buf, pos, size);
2006 if (bs->file)
2007 return bdrv_save_vmstate(bs->file, buf, pos, size);
2008 return -ENOTSUP;
2009 }
2010
2011 int bdrv_load_vmstate(BlockDriverState *bs, uint8_t *buf,
2012 int64_t pos, int size)
2013 {
2014 BlockDriver *drv = bs->drv;
2015 if (!drv)
2016 return -ENOMEDIUM;
2017 if (drv->bdrv_load_vmstate)
2018 return drv->bdrv_load_vmstate(bs, buf, pos, size);
2019 if (bs->file)
2020 return bdrv_load_vmstate(bs->file, buf, pos, size);
2021 return -ENOTSUP;
2022 }
2023
2024 void bdrv_debug_event(BlockDriverState *bs, BlkDebugEvent event)
2025 {
2026 BlockDriver *drv = bs->drv;
2027
2028 if (!drv || !drv->bdrv_debug_event) {
2029 return;
2030 }
2031
2032 return drv->bdrv_debug_event(bs, event);
2033
2034 }
2035
2036 /**************************************************************/
2037 /* handling of snapshots */
2038
2039 int bdrv_can_snapshot(BlockDriverState *bs)
2040 {
2041 BlockDriver *drv = bs->drv;
2042 if (!drv || bdrv_is_removable(bs) || bdrv_is_read_only(bs)) {
2043 return 0;
2044 }
2045
2046 if (!drv->bdrv_snapshot_create) {
2047 if (bs->file != NULL) {
2048 return bdrv_can_snapshot(bs->file);
2049 }
2050 return 0;
2051 }
2052
2053 return 1;
2054 }
2055
2056 int bdrv_is_snapshot(BlockDriverState *bs)
2057 {
2058 return !!(bs->open_flags & BDRV_O_SNAPSHOT);
2059 }
2060
2061 BlockDriverState *bdrv_snapshots(void)
2062 {
2063 BlockDriverState *bs;
2064
2065 if (bs_snapshots) {
2066 return bs_snapshots;
2067 }
2068
2069 bs = NULL;
2070 while ((bs = bdrv_next(bs))) {
2071 if (bdrv_can_snapshot(bs)) {
2072 bs_snapshots = bs;
2073 return bs;
2074 }
2075 }
2076 return NULL;
2077 }
2078
2079 int bdrv_snapshot_create(BlockDriverState *bs,
2080 QEMUSnapshotInfo *sn_info)
2081 {
2082 BlockDriver *drv = bs->drv;
2083 if (!drv)
2084 return -ENOMEDIUM;
2085 if (drv->bdrv_snapshot_create)
2086 return drv->bdrv_snapshot_create(bs, sn_info);
2087 if (bs->file)
2088 return bdrv_snapshot_create(bs->file, sn_info);
2089 return -ENOTSUP;
2090 }
2091
2092 int bdrv_snapshot_goto(BlockDriverState *bs,
2093 const char *snapshot_id)
2094 {
2095 BlockDriver *drv = bs->drv;
2096 int ret, open_ret;
2097
2098 if (!drv)
2099 return -ENOMEDIUM;
2100 if (drv->bdrv_snapshot_goto)
2101 return drv->bdrv_snapshot_goto(bs, snapshot_id);
2102
2103 if (bs->file) {
2104 drv->bdrv_close(bs);
2105 ret = bdrv_snapshot_goto(bs->file, snapshot_id);
2106 open_ret = drv->bdrv_open(bs, bs->open_flags);
2107 if (open_ret < 0) {
2108 bdrv_delete(bs->file);
2109 bs->drv = NULL;
2110 return open_ret;
2111 }
2112 return ret;
2113 }
2114
2115 return -ENOTSUP;
2116 }
2117
2118 int bdrv_snapshot_delete(BlockDriverState *bs, const char *snapshot_id)
2119 {
2120 BlockDriver *drv = bs->drv;
2121 if (!drv)
2122 return -ENOMEDIUM;
2123 if (drv->bdrv_snapshot_delete)
2124 return drv->bdrv_snapshot_delete(bs, snapshot_id);
2125 if (bs->file)
2126 return bdrv_snapshot_delete(bs->file, snapshot_id);
2127 return -ENOTSUP;
2128 }
2129
2130 int bdrv_snapshot_list(BlockDriverState *bs,
2131 QEMUSnapshotInfo **psn_info)
2132 {
2133 BlockDriver *drv = bs->drv;
2134 if (!drv)
2135 return -ENOMEDIUM;
2136 if (drv->bdrv_snapshot_list)
2137 return drv->bdrv_snapshot_list(bs, psn_info);
2138 if (bs->file)
2139 return bdrv_snapshot_list(bs->file, psn_info);
2140 return -ENOTSUP;
2141 }
2142
2143 int bdrv_snapshot_load_tmp(BlockDriverState *bs,
2144 const char *snapshot_name)
2145 {
2146 BlockDriver *drv = bs->drv;
2147 if (!drv) {
2148 return -ENOMEDIUM;
2149 }
2150 if (!bs->read_only) {
2151 return -EINVAL;
2152 }
2153 if (drv->bdrv_snapshot_load_tmp) {
2154 return drv->bdrv_snapshot_load_tmp(bs, snapshot_name);
2155 }
2156 return -ENOTSUP;
2157 }
2158
2159 #define NB_SUFFIXES 4
2160
2161 char *get_human_readable_size(char *buf, int buf_size, int64_t size)
2162 {
2163 static const char suffixes[NB_SUFFIXES] = "KMGT";
2164 int64_t base;
2165 int i;
2166
2167 if (size <= 999) {
2168 snprintf(buf, buf_size, "%" PRId64, size);
2169 } else {
2170 base = 1024;
2171 for(i = 0; i < NB_SUFFIXES; i++) {
2172 if (size < (10 * base)) {
2173 snprintf(buf, buf_size, "%0.1f%c",
2174 (double)size / base,
2175 suffixes[i]);
2176 break;
2177 } else if (size < (1000 * base) || i == (NB_SUFFIXES - 1)) {
2178 snprintf(buf, buf_size, "%" PRId64 "%c",
2179 ((size + (base >> 1)) / base),
2180 suffixes[i]);
2181 break;
2182 }
2183 base = base * 1024;
2184 }
2185 }
2186 return buf;
2187 }
2188
2189 char *bdrv_snapshot_dump(char *buf, int buf_size, QEMUSnapshotInfo *sn)
2190 {
2191 char buf1[128], date_buf[128], clock_buf[128];
2192 #ifdef _WIN32
2193 struct tm *ptm;
2194 #else
2195 struct tm tm;
2196 #endif
2197 time_t ti;
2198 int64_t secs;
2199
2200 if (!sn) {
2201 snprintf(buf, buf_size,
2202 "%-10s%-20s%7s%20s%15s",
2203 "ID", "TAG", "VM SIZE", "DATE", "VM CLOCK");
2204 } else {
2205 ti = sn->date_sec;
2206 #ifdef _WIN32
2207 ptm = localtime(&ti);
2208 strftime(date_buf, sizeof(date_buf),
2209 "%Y-%m-%d %H:%M:%S", ptm);
2210 #else
2211 localtime_r(&ti, &tm);
2212 strftime(date_buf, sizeof(date_buf),
2213 "%Y-%m-%d %H:%M:%S", &tm);
2214 #endif
2215 secs = sn->vm_clock_nsec / 1000000000;
2216 snprintf(clock_buf, sizeof(clock_buf),
2217 "%02d:%02d:%02d.%03d",
2218 (int)(secs / 3600),
2219 (int)((secs / 60) % 60),
2220 (int)(secs % 60),
2221 (int)((sn->vm_clock_nsec / 1000000) % 1000));
2222 snprintf(buf, buf_size,
2223 "%-10s%-20s%7s%20s%15s",
2224 sn->id_str, sn->name,
2225 get_human_readable_size(buf1, sizeof(buf1), sn->vm_state_size),
2226 date_buf,
2227 clock_buf);
2228 }
2229 return buf;
2230 }
2231
2232
2233 /**************************************************************/
2234 /* async I/Os */
2235
2236 BlockDriverAIOCB *bdrv_aio_readv(BlockDriverState *bs, int64_t sector_num,
2237 QEMUIOVector *qiov, int nb_sectors,
2238 BlockDriverCompletionFunc *cb, void *opaque)
2239 {
2240 BlockDriver *drv = bs->drv;
2241 BlockDriverAIOCB *ret;
2242
2243 trace_bdrv_aio_readv(bs, sector_num, nb_sectors, opaque);
2244
2245 if (!drv)
2246 return NULL;
2247 if (bdrv_check_request(bs, sector_num, nb_sectors))
2248 return NULL;
2249
2250 ret = drv->bdrv_aio_readv(bs, sector_num, qiov, nb_sectors,
2251 cb, opaque);
2252
2253 if (ret) {
2254 /* Update stats even though technically transfer has not happened. */
2255 bs->rd_bytes += (unsigned) nb_sectors * BDRV_SECTOR_SIZE;
2256 bs->rd_ops ++;
2257 }
2258
2259 return ret;
2260 }
2261
2262 typedef struct BlockCompleteData {
2263 BlockDriverCompletionFunc *cb;
2264 void *opaque;
2265 BlockDriverState *bs;
2266 int64_t sector_num;
2267 int nb_sectors;
2268 } BlockCompleteData;
2269
2270 static void block_complete_cb(void *opaque, int ret)
2271 {
2272 BlockCompleteData *b = opaque;
2273
2274 if (b->bs->dirty_bitmap) {
2275 set_dirty_bitmap(b->bs, b->sector_num, b->nb_sectors, 1);
2276 }
2277 b->cb(b->opaque, ret);
2278 qemu_free(b);
2279 }
2280
2281 static BlockCompleteData *blk_dirty_cb_alloc(BlockDriverState *bs,
2282 int64_t sector_num,
2283 int nb_sectors,
2284 BlockDriverCompletionFunc *cb,
2285 void *opaque)
2286 {
2287 BlockCompleteData *blkdata = qemu_mallocz(sizeof(BlockCompleteData));
2288
2289 blkdata->bs = bs;
2290 blkdata->cb = cb;
2291 blkdata->opaque = opaque;
2292 blkdata->sector_num = sector_num;
2293 blkdata->nb_sectors = nb_sectors;
2294
2295 return blkdata;
2296 }
2297
2298 BlockDriverAIOCB *bdrv_aio_writev(BlockDriverState *bs, int64_t sector_num,
2299 QEMUIOVector *qiov, int nb_sectors,
2300 BlockDriverCompletionFunc *cb, void *opaque)
2301 {
2302 BlockDriver *drv = bs->drv;
2303 BlockDriverAIOCB *ret;
2304 BlockCompleteData *blk_cb_data;
2305
2306 trace_bdrv_aio_writev(bs, sector_num, nb_sectors, opaque);
2307
2308 if (!drv)
2309 return NULL;
2310 if (bs->read_only)
2311 return NULL;
2312 if (bdrv_check_request(bs, sector_num, nb_sectors))
2313 return NULL;
2314
2315 if (bs->dirty_bitmap) {
2316 blk_cb_data = blk_dirty_cb_alloc(bs, sector_num, nb_sectors, cb,
2317 opaque);
2318 cb = &block_complete_cb;
2319 opaque = blk_cb_data;
2320 }
2321
2322 ret = drv->bdrv_aio_writev(bs, sector_num, qiov, nb_sectors,
2323 cb, opaque);
2324
2325 if (ret) {
2326 /* Update stats even though technically transfer has not happened. */
2327 bs->wr_bytes += (unsigned) nb_sectors * BDRV_SECTOR_SIZE;
2328 bs->wr_ops ++;
2329 if (bs->wr_highest_sector < sector_num + nb_sectors - 1) {
2330 bs->wr_highest_sector = sector_num + nb_sectors - 1;
2331 }
2332 }
2333
2334 return ret;
2335 }
2336
2337
2338 typedef struct MultiwriteCB {
2339 int error;
2340 int num_requests;
2341 int num_callbacks;
2342 struct {
2343 BlockDriverCompletionFunc *cb;
2344 void *opaque;
2345 QEMUIOVector *free_qiov;
2346 void *free_buf;
2347 } callbacks[];
2348 } MultiwriteCB;
2349
2350 static void multiwrite_user_cb(MultiwriteCB *mcb)
2351 {
2352 int i;
2353
2354 for (i = 0; i < mcb->num_callbacks; i++) {
2355 mcb->callbacks[i].cb(mcb->callbacks[i].opaque, mcb->error);
2356 if (mcb->callbacks[i].free_qiov) {
2357 qemu_iovec_destroy(mcb->callbacks[i].free_qiov);
2358 }
2359 qemu_free(mcb->callbacks[i].free_qiov);
2360 qemu_vfree(mcb->callbacks[i].free_buf);
2361 }
2362 }
2363
2364 static void multiwrite_cb(void *opaque, int ret)
2365 {
2366 MultiwriteCB *mcb = opaque;
2367
2368 trace_multiwrite_cb(mcb, ret);
2369
2370 if (ret < 0 && !mcb->error) {
2371 mcb->error = ret;
2372 }
2373
2374 mcb->num_requests--;
2375 if (mcb->num_requests == 0) {
2376 multiwrite_user_cb(mcb);
2377 qemu_free(mcb);
2378 }
2379 }
2380
2381 static int multiwrite_req_compare(const void *a, const void *b)
2382 {
2383 const BlockRequest *req1 = a, *req2 = b;
2384
2385 /*
2386 * Note that we can't simply subtract req2->sector from req1->sector
2387 * here as that could overflow the return value.
2388 */
2389 if (req1->sector > req2->sector) {
2390 return 1;
2391 } else if (req1->sector < req2->sector) {
2392 return -1;
2393 } else {
2394 return 0;
2395 }
2396 }
2397
2398 /*
2399 * Takes a bunch of requests and tries to merge them. Returns the number of
2400 * requests that remain after merging.
2401 */
2402 static int multiwrite_merge(BlockDriverState *bs, BlockRequest *reqs,
2403 int num_reqs, MultiwriteCB *mcb)
2404 {
2405 int i, outidx;
2406
2407 // Sort requests by start sector
2408 qsort(reqs, num_reqs, sizeof(*reqs), &multiwrite_req_compare);
2409
2410 // Check if adjacent requests touch the same clusters. If so, combine them,
2411 // filling up gaps with zero sectors.
2412 outidx = 0;
2413 for (i = 1; i < num_reqs; i++) {
2414 int merge = 0;
2415 int64_t oldreq_last = reqs[outidx].sector + reqs[outidx].nb_sectors;
2416
2417 // This handles the cases that are valid for all block drivers, namely
2418 // exactly sequential writes and overlapping writes.
2419 if (reqs[i].sector <= oldreq_last) {
2420 merge = 1;
2421 }
2422
2423 // The block driver may decide that it makes sense to combine requests
2424 // even if there is a gap of some sectors between them. In this case,
2425 // the gap is filled with zeros (therefore only applicable for yet
2426 // unused space in format like qcow2).
2427 if (!merge && bs->drv->bdrv_merge_requests) {
2428 merge = bs->drv->bdrv_merge_requests(bs, &reqs[outidx], &reqs[i]);
2429 }
2430
2431 if (reqs[outidx].qiov->niov + reqs[i].qiov->niov + 1 > IOV_MAX) {
2432 merge = 0;
2433 }
2434
2435 if (merge) {
2436 size_t size;
2437 QEMUIOVector *qiov = qemu_mallocz(sizeof(*qiov));
2438 qemu_iovec_init(qiov,
2439 reqs[outidx].qiov->niov + reqs[i].qiov->niov + 1);
2440
2441 // Add the first request to the merged one. If the requests are
2442 // overlapping, drop the last sectors of the first request.
2443 size = (reqs[i].sector - reqs[outidx].sector) << 9;
2444 qemu_iovec_concat(qiov, reqs[outidx].qiov, size);
2445
2446 // We might need to add some zeros between the two requests
2447 if (reqs[i].sector > oldreq_last) {
2448 size_t zero_bytes = (reqs[i].sector - oldreq_last) << 9;
2449 uint8_t *buf = qemu_blockalign(bs, zero_bytes);
2450 memset(buf, 0, zero_bytes);
2451 qemu_iovec_add(qiov, buf, zero_bytes);
2452 mcb->callbacks[i].free_buf = buf;
2453 }
2454
2455 // Add the second request
2456 qemu_iovec_concat(qiov, reqs[i].qiov, reqs[i].qiov->size);
2457
2458 reqs[outidx].nb_sectors = qiov->size >> 9;
2459 reqs[outidx].qiov = qiov;
2460
2461 mcb->callbacks[i].free_qiov = reqs[outidx].qiov;
2462 } else {
2463 outidx++;
2464 reqs[outidx].sector = reqs[i].sector;
2465 reqs[outidx].nb_sectors = reqs[i].nb_sectors;
2466 reqs[outidx].qiov = reqs[i].qiov;
2467 }
2468 }
2469
2470 return outidx + 1;
2471 }
2472
2473 /*
2474 * Submit multiple AIO write requests at once.
2475 *
2476 * On success, the function returns 0 and all requests in the reqs array have
2477 * been submitted. In error case this function returns -1, and any of the
2478 * requests may or may not be submitted yet. In particular, this means that the
2479 * callback will be called for some of the requests, for others it won't. The
2480 * caller must check the error field of the BlockRequest to wait for the right
2481 * callbacks (if error != 0, no callback will be called).
2482 *
2483 * The implementation may modify the contents of the reqs array, e.g. to merge
2484 * requests. However, the fields opaque and error are left unmodified as they
2485 * are used to signal failure for a single request to the caller.
2486 */
2487 int bdrv_aio_multiwrite(BlockDriverState *bs, BlockRequest *reqs, int num_reqs)
2488 {
2489 BlockDriverAIOCB *acb;
2490 MultiwriteCB *mcb;
2491 int i;
2492
2493 /* don't submit writes if we don't have a medium */
2494 if (bs->drv == NULL) {
2495 for (i = 0; i < num_reqs; i++) {
2496 reqs[i].error = -ENOMEDIUM;
2497 }
2498 return -1;
2499 }
2500
2501 if (num_reqs == 0) {
2502 return 0;
2503 }
2504
2505 // Create MultiwriteCB structure
2506 mcb = qemu_mallocz(sizeof(*mcb) + num_reqs * sizeof(*mcb->callbacks));
2507 mcb->num_requests = 0;
2508 mcb->num_callbacks = num_reqs;
2509
2510 for (i = 0; i < num_reqs; i++) {
2511 mcb->callbacks[i].cb = reqs[i].cb;
2512 mcb->callbacks[i].opaque = reqs[i].opaque;
2513 }
2514
2515 // Check for mergable requests
2516 num_reqs = multiwrite_merge(bs, reqs, num_reqs, mcb);
2517
2518 trace_bdrv_aio_multiwrite(mcb, mcb->num_callbacks, num_reqs);
2519
2520 /*
2521 * Run the aio requests. As soon as one request can't be submitted
2522 * successfully, fail all requests that are not yet submitted (we must
2523 * return failure for all requests anyway)
2524 *
2525 * num_requests cannot be set to the right value immediately: If
2526 * bdrv_aio_writev fails for some request, num_requests would be too high
2527 * and therefore multiwrite_cb() would never recognize the multiwrite
2528 * request as completed. We also cannot use the loop variable i to set it
2529 * when the first request fails because the callback may already have been
2530 * called for previously submitted requests. Thus, num_requests must be
2531 * incremented for each request that is submitted.
2532 *
2533 * The problem that callbacks may be called early also means that we need
2534 * to take care that num_requests doesn't become 0 before all requests are
2535 * submitted - multiwrite_cb() would consider the multiwrite request
2536 * completed. A dummy request that is "completed" by a manual call to
2537 * multiwrite_cb() takes care of this.
2538 */
2539 mcb->num_requests = 1;
2540
2541 // Run the aio requests
2542 for (i = 0; i < num_reqs; i++) {
2543 mcb->num_requests++;
2544 acb = bdrv_aio_writev(bs, reqs[i].sector, reqs[i].qiov,
2545 reqs[i].nb_sectors, multiwrite_cb, mcb);
2546
2547 if (acb == NULL) {
2548 // We can only fail the whole thing if no request has been
2549 // submitted yet. Otherwise we'll wait for the submitted AIOs to
2550 // complete and report the error in the callback.
2551 if (i == 0) {
2552 trace_bdrv_aio_multiwrite_earlyfail(mcb);
2553 goto fail;
2554 } else {
2555 trace_bdrv_aio_multiwrite_latefail(mcb, i);
2556 multiwrite_cb(mcb, -EIO);
2557 break;
2558 }
2559 }
2560 }
2561
2562 /* Complete the dummy request */
2563 multiwrite_cb(mcb, 0);
2564
2565 return 0;
2566
2567 fail:
2568 for (i = 0; i < mcb->num_callbacks; i++) {
2569 reqs[i].error = -EIO;
2570 }
2571 qemu_free(mcb);
2572 return -1;
2573 }
2574
2575 BlockDriverAIOCB *bdrv_aio_flush(BlockDriverState *bs,
2576 BlockDriverCompletionFunc *cb, void *opaque)
2577 {
2578 BlockDriver *drv = bs->drv;
2579
2580 trace_bdrv_aio_flush(bs, opaque);
2581
2582 if (bs->open_flags & BDRV_O_NO_FLUSH) {
2583 return bdrv_aio_noop_em(bs, cb, opaque);
2584 }
2585
2586 if (!drv)
2587 return NULL;
2588 return drv->bdrv_aio_flush(bs, cb, opaque);
2589 }
2590
2591 void bdrv_aio_cancel(BlockDriverAIOCB *acb)
2592 {
2593 acb->pool->cancel(acb);
2594 }
2595
2596
2597 /**************************************************************/
2598 /* async block device emulation */
2599
2600 typedef struct BlockDriverAIOCBSync {
2601 BlockDriverAIOCB common;
2602 QEMUBH *bh;
2603 int ret;
2604 /* vector translation state */
2605 QEMUIOVector *qiov;
2606 uint8_t *bounce;
2607 int is_write;
2608 } BlockDriverAIOCBSync;
2609
2610 static void bdrv_aio_cancel_em(BlockDriverAIOCB *blockacb)
2611 {
2612 BlockDriverAIOCBSync *acb =
2613 container_of(blockacb, BlockDriverAIOCBSync, common);
2614 qemu_bh_delete(acb->bh);
2615 acb->bh = NULL;
2616 qemu_aio_release(acb);
2617 }
2618
2619 static AIOPool bdrv_em_aio_pool = {
2620 .aiocb_size = sizeof(BlockDriverAIOCBSync),
2621 .cancel = bdrv_aio_cancel_em,
2622 };
2623
2624 static void bdrv_aio_bh_cb(void *opaque)
2625 {
2626 BlockDriverAIOCBSync *acb = opaque;
2627
2628 if (!acb->is_write)
2629 qemu_iovec_from_buffer(acb->qiov, acb->bounce, acb->qiov->size);
2630 qemu_vfree(acb->bounce);
2631 acb->common.cb(acb->common.opaque, acb->ret);
2632 qemu_bh_delete(acb->bh);
2633 acb->bh = NULL;
2634 qemu_aio_release(acb);
2635 }
2636
2637 static BlockDriverAIOCB *bdrv_aio_rw_vector(BlockDriverState *bs,
2638 int64_t sector_num,
2639 QEMUIOVector *qiov,
2640 int nb_sectors,
2641 BlockDriverCompletionFunc *cb,
2642 void *opaque,
2643 int is_write)
2644
2645 {
2646 BlockDriverAIOCBSync *acb;
2647
2648 acb = qemu_aio_get(&bdrv_em_aio_pool, bs, cb, opaque);
2649 acb->is_write = is_write;
2650 acb->qiov = qiov;
2651 acb->bounce = qemu_blockalign(bs, qiov->size);
2652
2653 if (!acb->bh)
2654 acb->bh = qemu_bh_new(bdrv_aio_bh_cb, acb);
2655
2656 if (is_write) {
2657 qemu_iovec_to_buffer(acb->qiov, acb->bounce);
2658 acb->ret = bdrv_write(bs, sector_num, acb->bounce, nb_sectors);
2659 } else {
2660 acb->ret = bdrv_read(bs, sector_num, acb->bounce, nb_sectors);
2661 }
2662
2663 qemu_bh_schedule(acb->bh);
2664
2665 return &acb->common;
2666 }
2667
2668 static BlockDriverAIOCB *bdrv_aio_readv_em(BlockDriverState *bs,
2669 int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
2670 BlockDriverCompletionFunc *cb, void *opaque)
2671 {
2672 return bdrv_aio_rw_vector(bs, sector_num, qiov, nb_sectors, cb, opaque, 0);
2673 }
2674
2675 static BlockDriverAIOCB *bdrv_aio_writev_em(BlockDriverState *bs,
2676 int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
2677 BlockDriverCompletionFunc *cb, void *opaque)
2678 {
2679 return bdrv_aio_rw_vector(bs, sector_num, qiov, nb_sectors, cb, opaque, 1);
2680 }
2681
2682
2683 typedef struct BlockDriverAIOCBCoroutine {
2684 BlockDriverAIOCB common;
2685 BlockRequest req;
2686 bool is_write;
2687 QEMUBH* bh;
2688 } BlockDriverAIOCBCoroutine;
2689
2690 static void bdrv_aio_co_cancel_em(BlockDriverAIOCB *blockacb)
2691 {
2692 qemu_aio_flush();
2693 }
2694
2695 static AIOPool bdrv_em_co_aio_pool = {
2696 .aiocb_size = sizeof(BlockDriverAIOCBCoroutine),
2697 .cancel = bdrv_aio_co_cancel_em,
2698 };
2699
2700 static void bdrv_co_rw_bh(void *opaque)
2701 {
2702 BlockDriverAIOCBCoroutine *acb = opaque;
2703
2704 acb->common.cb(acb->common.opaque, acb->req.error);
2705 qemu_bh_delete(acb->bh);
2706 qemu_aio_release(acb);
2707 }
2708
2709 static void coroutine_fn bdrv_co_rw(void *opaque)
2710 {
2711 BlockDriverAIOCBCoroutine *acb = opaque;
2712 BlockDriverState *bs = acb->common.bs;
2713
2714 if (!acb->is_write) {
2715 acb->req.error = bs->drv->bdrv_co_readv(bs, acb->req.sector,
2716 acb->req.nb_sectors, acb->req.qiov);
2717 } else {
2718 acb->req.error = bs->drv->bdrv_co_writev(bs, acb->req.sector,
2719 acb->req.nb_sectors, acb->req.qiov);
2720 }
2721
2722 acb->bh = qemu_bh_new(bdrv_co_rw_bh, acb);
2723 qemu_bh_schedule(acb->bh);
2724 }
2725
2726 static BlockDriverAIOCB *bdrv_co_aio_rw_vector(BlockDriverState *bs,
2727 int64_t sector_num,
2728 QEMUIOVector *qiov,
2729 int nb_sectors,
2730 BlockDriverCompletionFunc *cb,
2731 void *opaque,
2732 bool is_write)
2733 {
2734 Coroutine *co;
2735 BlockDriverAIOCBCoroutine *acb;
2736
2737 acb = qemu_aio_get(&bdrv_em_co_aio_pool, bs, cb, opaque);
2738 acb->req.sector = sector_num;
2739 acb->req.nb_sectors = nb_sectors;
2740 acb->req.qiov = qiov;
2741 acb->is_write = is_write;
2742
2743 co = qemu_coroutine_create(bdrv_co_rw);
2744 qemu_coroutine_enter(co, acb);
2745
2746 return &acb->common;
2747 }
2748
2749 static BlockDriverAIOCB *bdrv_co_aio_readv_em(BlockDriverState *bs,
2750 int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
2751 BlockDriverCompletionFunc *cb, void *opaque)
2752 {
2753 return bdrv_co_aio_rw_vector(bs, sector_num, qiov, nb_sectors, cb, opaque,
2754 false);
2755 }
2756
2757 static BlockDriverAIOCB *bdrv_co_aio_writev_em(BlockDriverState *bs,
2758 int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
2759 BlockDriverCompletionFunc *cb, void *opaque)
2760 {
2761 return bdrv_co_aio_rw_vector(bs, sector_num, qiov, nb_sectors, cb, opaque,
2762 true);
2763 }
2764
2765 static BlockDriverAIOCB *bdrv_aio_flush_em(BlockDriverState *bs,
2766 BlockDriverCompletionFunc *cb, void *opaque)
2767 {
2768 BlockDriverAIOCBSync *acb;
2769
2770 acb = qemu_aio_get(&bdrv_em_aio_pool, bs, cb, opaque);
2771 acb->is_write = 1; /* don't bounce in the completion hadler */
2772 acb->qiov = NULL;
2773 acb->bounce = NULL;
2774 acb->ret = 0;
2775
2776 if (!acb->bh)
2777 acb->bh = qemu_bh_new(bdrv_aio_bh_cb, acb);
2778
2779 bdrv_flush(bs);
2780 qemu_bh_schedule(acb->bh);
2781 return &acb->common;
2782 }
2783
2784 static BlockDriverAIOCB *bdrv_aio_noop_em(BlockDriverState *bs,
2785 BlockDriverCompletionFunc *cb, void *opaque)
2786 {
2787 BlockDriverAIOCBSync *acb;
2788
2789 acb = qemu_aio_get(&bdrv_em_aio_pool, bs, cb, opaque);
2790 acb->is_write = 1; /* don't bounce in the completion handler */
2791 acb->qiov = NULL;
2792 acb->bounce = NULL;
2793 acb->ret = 0;
2794
2795 if (!acb->bh) {
2796 acb->bh = qemu_bh_new(bdrv_aio_bh_cb, acb);
2797 }
2798
2799 qemu_bh_schedule(acb->bh);
2800 return &acb->common;
2801 }
2802
2803 /**************************************************************/
2804 /* sync block device emulation */
2805
2806 static void bdrv_rw_em_cb(void *opaque, int ret)
2807 {
2808 *(int *)opaque = ret;
2809 }
2810
2811 #define NOT_DONE 0x7fffffff
2812
2813 static int bdrv_read_em(BlockDriverState *bs, int64_t sector_num,
2814 uint8_t *buf, int nb_sectors)
2815 {
2816 int async_ret;
2817 BlockDriverAIOCB *acb;
2818 struct iovec iov;
2819 QEMUIOVector qiov;
2820
2821 async_ret = NOT_DONE;
2822 iov.iov_base = (void *)buf;
2823 iov.iov_len = nb_sectors * BDRV_SECTOR_SIZE;
2824 qemu_iovec_init_external(&qiov, &iov, 1);
2825 acb = bdrv_aio_readv(bs, sector_num, &qiov, nb_sectors,
2826 bdrv_rw_em_cb, &async_ret);
2827 if (acb == NULL) {
2828 async_ret = -1;
2829 goto fail;
2830 }
2831
2832 while (async_ret == NOT_DONE) {
2833 qemu_aio_wait();
2834 }
2835
2836
2837 fail:
2838 return async_ret;
2839 }
2840
2841 static int bdrv_write_em(BlockDriverState *bs, int64_t sector_num,
2842 const uint8_t *buf, int nb_sectors)
2843 {
2844 int async_ret;
2845 BlockDriverAIOCB *acb;
2846 struct iovec iov;
2847 QEMUIOVector qiov;
2848
2849 async_ret = NOT_DONE;
2850 iov.iov_base = (void *)buf;
2851 iov.iov_len = nb_sectors * BDRV_SECTOR_SIZE;
2852 qemu_iovec_init_external(&qiov, &iov, 1);
2853 acb = bdrv_aio_writev(bs, sector_num, &qiov, nb_sectors,
2854 bdrv_rw_em_cb, &async_ret);
2855 if (acb == NULL) {
2856 async_ret = -1;
2857 goto fail;
2858 }
2859 while (async_ret == NOT_DONE) {
2860 qemu_aio_wait();
2861 }
2862
2863 fail:
2864 return async_ret;
2865 }
2866
2867 void bdrv_init(void)
2868 {
2869 module_call_init(MODULE_INIT_BLOCK);
2870 }
2871
2872 void bdrv_init_with_whitelist(void)
2873 {
2874 use_bdrv_whitelist = 1;
2875 bdrv_init();
2876 }
2877
2878 void *qemu_aio_get(AIOPool *pool, BlockDriverState *bs,
2879 BlockDriverCompletionFunc *cb, void *opaque)
2880 {
2881 BlockDriverAIOCB *acb;
2882
2883 if (pool->free_aiocb) {
2884 acb = pool->free_aiocb;
2885 pool->free_aiocb = acb->next;
2886 } else {
2887 acb = qemu_mallocz(pool->aiocb_size);
2888 acb->pool = pool;
2889 }
2890 acb->bs = bs;
2891 acb->cb = cb;
2892 acb->opaque = opaque;
2893 return acb;
2894 }
2895
2896 void qemu_aio_release(void *p)
2897 {
2898 BlockDriverAIOCB *acb = (BlockDriverAIOCB *)p;
2899 AIOPool *pool = acb->pool;
2900 acb->next = pool->free_aiocb;
2901 pool->free_aiocb = acb;
2902 }
2903
2904 /**************************************************************/
2905 /* Coroutine block device emulation */
2906
2907 typedef struct CoroutineIOCompletion {
2908 Coroutine *coroutine;
2909 int ret;
2910 } CoroutineIOCompletion;
2911
2912 static void bdrv_co_io_em_complete(void *opaque, int ret)
2913 {
2914 CoroutineIOCompletion *co = opaque;
2915
2916 co->ret = ret;
2917 qemu_coroutine_enter(co->coroutine, NULL);
2918 }
2919
2920 static int coroutine_fn bdrv_co_io_em(BlockDriverState *bs, int64_t sector_num,
2921 int nb_sectors, QEMUIOVector *iov,
2922 bool is_write)
2923 {
2924 CoroutineIOCompletion co = {
2925 .coroutine = qemu_coroutine_self(),
2926 };
2927 BlockDriverAIOCB *acb;
2928
2929 if (is_write) {
2930 acb = bdrv_aio_writev(bs, sector_num, iov, nb_sectors,
2931 bdrv_co_io_em_complete, &co);
2932 } else {
2933 acb = bdrv_aio_readv(bs, sector_num, iov, nb_sectors,
2934 bdrv_co_io_em_complete, &co);
2935 }
2936
2937 trace_bdrv_co_io(is_write, acb);
2938 if (!acb) {
2939 return -EIO;
2940 }
2941 qemu_coroutine_yield();
2942
2943 return co.ret;
2944 }
2945
2946 static int coroutine_fn bdrv_co_readv_em(BlockDriverState *bs,
2947 int64_t sector_num, int nb_sectors,
2948 QEMUIOVector *iov)
2949 {
2950 return bdrv_co_io_em(bs, sector_num, nb_sectors, iov, false);
2951 }
2952
2953 static int coroutine_fn bdrv_co_writev_em(BlockDriverState *bs,
2954 int64_t sector_num, int nb_sectors,
2955 QEMUIOVector *iov)
2956 {
2957 return bdrv_co_io_em(bs, sector_num, nb_sectors, iov, true);
2958 }
2959
2960 static int coroutine_fn bdrv_co_flush_em(BlockDriverState *bs)
2961 {
2962 CoroutineIOCompletion co = {
2963 .coroutine = qemu_coroutine_self(),
2964 };
2965 BlockDriverAIOCB *acb;
2966
2967 acb = bdrv_aio_flush(bs, bdrv_co_io_em_complete, &co);
2968 if (!acb) {
2969 return -EIO;
2970 }
2971 qemu_coroutine_yield();
2972 return co.ret;
2973 }
2974
2975 /**************************************************************/
2976 /* removable device support */
2977
2978 /**
2979 * Return TRUE if the media is present
2980 */
2981 int bdrv_is_inserted(BlockDriverState *bs)
2982 {
2983 BlockDriver *drv = bs->drv;
2984 int ret;
2985 if (!drv)
2986 return 0;
2987 if (!drv->bdrv_is_inserted)
2988 return !bs->tray_open;
2989 ret = drv->bdrv_is_inserted(bs);
2990 return ret;
2991 }
2992
2993 /**
2994 * Return TRUE if the media changed since the last call to this
2995 * function. It is currently only used for floppy disks
2996 */
2997 int bdrv_media_changed(BlockDriverState *bs)
2998 {
2999 BlockDriver *drv = bs->drv;
3000 int ret;
3001
3002 if (!drv || !drv->bdrv_media_changed)
3003 ret = -ENOTSUP;
3004 else
3005 ret = drv->bdrv_media_changed(bs);
3006 if (ret == -ENOTSUP)
3007 ret = bs->media_changed;
3008 bs->media_changed = 0;
3009 return ret;
3010 }
3011
3012 /**
3013 * If eject_flag is TRUE, eject the media. Otherwise, close the tray
3014 */
3015 int bdrv_eject(BlockDriverState *bs, int eject_flag)
3016 {
3017 BlockDriver *drv = bs->drv;
3018
3019 if (eject_flag && bs->locked) {
3020 return -EBUSY;
3021 }
3022
3023 if (drv && drv->bdrv_eject) {
3024 drv->bdrv_eject(bs, eject_flag);
3025 }
3026 bs->tray_open = eject_flag;
3027 return 0;
3028 }
3029
3030 int bdrv_is_locked(BlockDriverState *bs)
3031 {
3032 return bs->locked;
3033 }
3034
3035 /**
3036 * Lock or unlock the media (if it is locked, the user won't be able
3037 * to eject it manually).
3038 */
3039 void bdrv_set_locked(BlockDriverState *bs, int locked)
3040 {
3041 BlockDriver *drv = bs->drv;
3042
3043 trace_bdrv_set_locked(bs, locked);
3044
3045 bs->locked = locked;
3046 if (drv && drv->bdrv_set_locked) {
3047 drv->bdrv_set_locked(bs, locked);
3048 }
3049 }
3050
3051 /* needed for generic scsi interface */
3052
3053 int bdrv_ioctl(BlockDriverState *bs, unsigned long int req, void *buf)
3054 {
3055 BlockDriver *drv = bs->drv;
3056
3057 if (drv && drv->bdrv_ioctl)
3058 return drv->bdrv_ioctl(bs, req, buf);
3059 return -ENOTSUP;
3060 }
3061
3062 BlockDriverAIOCB *bdrv_aio_ioctl(BlockDriverState *bs,
3063 unsigned long int req, void *buf,
3064 BlockDriverCompletionFunc *cb, void *opaque)
3065 {
3066 BlockDriver *drv = bs->drv;
3067
3068 if (drv && drv->bdrv_aio_ioctl)
3069 return drv->bdrv_aio_ioctl(bs, req, buf, cb, opaque);
3070 return NULL;
3071 }
3072
3073
3074
3075 void *qemu_blockalign(BlockDriverState *bs, size_t size)
3076 {
3077 return qemu_memalign((bs && bs->buffer_alignment) ? bs->buffer_alignment : 512, size);
3078 }
3079
3080 void bdrv_set_dirty_tracking(BlockDriverState *bs, int enable)
3081 {
3082 int64_t bitmap_size;
3083
3084 bs->dirty_count = 0;
3085 if (enable) {
3086 if (!bs->dirty_bitmap) {
3087 bitmap_size = (bdrv_getlength(bs) >> BDRV_SECTOR_BITS) +
3088 BDRV_SECTORS_PER_DIRTY_CHUNK * 8 - 1;
3089 bitmap_size /= BDRV_SECTORS_PER_DIRTY_CHUNK * 8;
3090
3091 bs->dirty_bitmap = qemu_mallocz(bitmap_size);
3092 }
3093 } else {
3094 if (bs->dirty_bitmap) {
3095 qemu_free(bs->dirty_bitmap);
3096 bs->dirty_bitmap = NULL;
3097 }
3098 }
3099 }
3100
3101 int bdrv_get_dirty(BlockDriverState *bs, int64_t sector)
3102 {
3103 int64_t chunk = sector / (int64_t)BDRV_SECTORS_PER_DIRTY_CHUNK;
3104
3105 if (bs->dirty_bitmap &&
3106 (sector << BDRV_SECTOR_BITS) < bdrv_getlength(bs)) {
3107 return !!(bs->dirty_bitmap[chunk / (sizeof(unsigned long) * 8)] &
3108 (1UL << (chunk % (sizeof(unsigned long) * 8))));
3109 } else {
3110 return 0;
3111 }
3112 }
3113
3114 void bdrv_reset_dirty(BlockDriverState *bs, int64_t cur_sector,
3115 int nr_sectors)
3116 {
3117 set_dirty_bitmap(bs, cur_sector, nr_sectors, 0);
3118 }
3119
3120 int64_t bdrv_get_dirty_count(BlockDriverState *bs)
3121 {
3122 return bs->dirty_count;
3123 }
3124
3125 void bdrv_set_in_use(BlockDriverState *bs, int in_use)
3126 {
3127 assert(bs->in_use != in_use);
3128 bs->in_use = in_use;
3129 }
3130
3131 int bdrv_in_use(BlockDriverState *bs)
3132 {
3133 return bs->in_use;
3134 }
3135
3136 int bdrv_img_create(const char *filename, const char *fmt,
3137 const char *base_filename, const char *base_fmt,
3138 char *options, uint64_t img_size, int flags)
3139 {
3140 QEMUOptionParameter *param = NULL, *create_options = NULL;
3141 QEMUOptionParameter *backing_fmt, *backing_file, *size;
3142 BlockDriverState *bs = NULL;
3143 BlockDriver *drv, *proto_drv;
3144 BlockDriver *backing_drv = NULL;
3145 int ret = 0;
3146
3147 /* Find driver and parse its options */
3148 drv = bdrv_find_format(fmt);
3149 if (!drv) {
3150 error_report("Unknown file format '%s'", fmt);
3151 ret = -EINVAL;
3152 goto out;
3153 }
3154
3155 proto_drv = bdrv_find_protocol(filename);
3156 if (!proto_drv) {
3157 error_report("Unknown protocol '%s'", filename);
3158 ret = -EINVAL;
3159 goto out;
3160 }
3161
3162 create_options = append_option_parameters(create_options,
3163 drv->create_options);
3164 create_options = append_option_parameters(create_options,
3165 proto_drv->create_options);
3166
3167 /* Create parameter list with default values */
3168 param = parse_option_parameters("", create_options, param);
3169
3170 set_option_parameter_int(param, BLOCK_OPT_SIZE, img_size);
3171
3172 /* Parse -o options */
3173 if (options) {
3174 param = parse_option_parameters(options, create_options, param);
3175 if (param == NULL) {
3176 error_report("Invalid options for file format '%s'.", fmt);
3177 ret = -EINVAL;
3178 goto out;
3179 }
3180 }
3181
3182 if (base_filename) {
3183 if (set_option_parameter(param, BLOCK_OPT_BACKING_FILE,
3184 base_filename)) {
3185 error_report("Backing file not supported for file format '%s'",
3186 fmt);
3187 ret = -EINVAL;
3188 goto out;
3189 }
3190 }
3191
3192 if (base_fmt) {
3193 if (set_option_parameter(param, BLOCK_OPT_BACKING_FMT, base_fmt)) {
3194 error_report("Backing file format not supported for file "
3195 "format '%s'", fmt);
3196 ret = -EINVAL;
3197 goto out;
3198 }
3199 }
3200
3201 backing_file = get_option_parameter(param, BLOCK_OPT_BACKING_FILE);
3202 if (backing_file && backing_file->value.s) {
3203 if (!strcmp(filename, backing_file->value.s)) {
3204 error_report("Error: Trying to create an image with the "
3205 "same filename as the backing file");
3206 ret = -EINVAL;
3207 goto out;
3208 }
3209 }
3210
3211 backing_fmt = get_option_parameter(param, BLOCK_OPT_BACKING_FMT);
3212 if (backing_fmt && backing_fmt->value.s) {
3213 backing_drv = bdrv_find_format(backing_fmt->value.s);
3214 if (!backing_drv) {
3215 error_report("Unknown backing file format '%s'",
3216 backing_fmt->value.s);
3217 ret = -EINVAL;
3218 goto out;
3219 }
3220 }
3221
3222 // The size for the image must always be specified, with one exception:
3223 // If we are using a backing file, we can obtain the size from there
3224 size = get_option_parameter(param, BLOCK_OPT_SIZE);
3225 if (size && size->value.n == -1) {
3226 if (backing_file && backing_file->value.s) {
3227 uint64_t size;
3228 char buf[32];
3229
3230 bs = bdrv_new("");
3231
3232 ret = bdrv_open(bs, backing_file->value.s, flags, backing_drv);
3233 if (ret < 0) {
3234 error_report("Could not open '%s'", backing_file->value.s);
3235 goto out;
3236 }
3237 bdrv_get_geometry(bs, &size);
3238 size *= 512;
3239
3240 snprintf(buf, sizeof(buf), "%" PRId64, size);
3241 set_option_parameter(param, BLOCK_OPT_SIZE, buf);
3242 } else {
3243 error_report("Image creation needs a size parameter");
3244 ret = -EINVAL;
3245 goto out;
3246 }
3247 }
3248
3249 printf("Formatting '%s', fmt=%s ", filename, fmt);
3250 print_option_parameters(param);
3251 puts("");
3252
3253 ret = bdrv_create(drv, filename, param);
3254
3255 if (ret < 0) {
3256 if (ret == -ENOTSUP) {
3257 error_report("Formatting or formatting option not supported for "
3258 "file format '%s'", fmt);
3259 } else if (ret == -EFBIG) {
3260 error_report("The image size is too large for file format '%s'",
3261 fmt);
3262 } else {
3263 error_report("%s: error while creating %s: %s", filename, fmt,
3264 strerror(-ret));
3265 }
3266 }
3267
3268 out:
3269 free_option_parameters(create_options);
3270 free_option_parameters(param);
3271
3272 if (bs) {
3273 bdrv_delete(bs);
3274 }
3275
3276 return ret;
3277 }
Anthony Liguori's QEMU tree