?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Merge remote-tracking branch 'qemu-kvm-tmp/memory/core' into staging
[qemu.git] / block.c
blobe3fe97f27533ce3f2919a8cff97bba21ce41e362
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 void bdrv_dev_change_media_cb(BlockDriverState *bs, bool load);
48 static BlockDriverAIOCB *bdrv_aio_readv_em(BlockDriverState *bs,
49 int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
50 BlockDriverCompletionFunc *cb, void *opaque);
51 static BlockDriverAIOCB *bdrv_aio_writev_em(BlockDriverState *bs,
52 int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
53 BlockDriverCompletionFunc *cb, void *opaque);
54 static BlockDriverAIOCB *bdrv_aio_flush_em(BlockDriverState *bs,
55 BlockDriverCompletionFunc *cb, void *opaque);
56 static BlockDriverAIOCB *bdrv_aio_noop_em(BlockDriverState *bs,
57 BlockDriverCompletionFunc *cb, void *opaque);
58 static int bdrv_read_em(BlockDriverState *bs, int64_t sector_num,
59 uint8_t *buf, int nb_sectors);
60 static int bdrv_write_em(BlockDriverState *bs, int64_t sector_num,
61 const uint8_t *buf, int nb_sectors);
62 static BlockDriverAIOCB *bdrv_co_aio_readv_em(BlockDriverState *bs,
63 int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
64 BlockDriverCompletionFunc *cb, void *opaque);
65 static BlockDriverAIOCB *bdrv_co_aio_writev_em(BlockDriverState *bs,
66 int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
67 BlockDriverCompletionFunc *cb, void *opaque);
68 static int coroutine_fn bdrv_co_readv_em(BlockDriverState *bs,
69 int64_t sector_num, int nb_sectors,
70 QEMUIOVector *iov);
71 static int coroutine_fn bdrv_co_writev_em(BlockDriverState *bs,
72 int64_t sector_num, int nb_sectors,
73 QEMUIOVector *iov);
74 static int coroutine_fn bdrv_co_flush_em(BlockDriverState *bs);
75
76 static QTAILQ_HEAD(, BlockDriverState) bdrv_states =
77 QTAILQ_HEAD_INITIALIZER(bdrv_states);
78
79 static QLIST_HEAD(, BlockDriver) bdrv_drivers =
80 QLIST_HEAD_INITIALIZER(bdrv_drivers);
81
82 /* The device to use for VM snapshots */
83 static BlockDriverState *bs_snapshots;
84
85 /* If non-zero, use only whitelisted block drivers */
86 static int use_bdrv_whitelist;
87
88 #ifdef _WIN32
89 static int is_windows_drive_prefix(const char *filename)
90 {
91 return (((filename[0] >= 'a' && filename[0] <= 'z') ||
92 (filename[0] >= 'A' && filename[0] <= 'Z')) &&
93 filename[1] == ':');
94 }
95
96 int is_windows_drive(const char *filename)
97 {
98 if (is_windows_drive_prefix(filename) &&
99 filename[2] == '\0')
100 return 1;
101 if (strstart(filename, "\\\\.\\", NULL) ||
102 strstart(filename, "//./", NULL))
103 return 1;
104 return 0;
105 }
106 #endif
107
108 /* check if the path starts with "<protocol>:" */
109 static int path_has_protocol(const char *path)
110 {
111 #ifdef _WIN32
112 if (is_windows_drive(path) ||
113 is_windows_drive_prefix(path)) {
114 return 0;
115 }
116 #endif
117
118 return strchr(path, ':') != NULL;
119 }
120
121 int path_is_absolute(const char *path)
122 {
123 const char *p;
124 #ifdef _WIN32
125 /* specific case for names like: "\\.\d:" */
126 if (*path == '/' || *path == '\\')
127 return 1;
128 #endif
129 p = strchr(path, ':');
130 if (p)
131 p++;
132 else
133 p = path;
134 #ifdef _WIN32
135 return (*p == '/' || *p == '\\');
136 #else
137 return (*p == '/');
138 #endif
139 }
140
141 /* if filename is absolute, just copy it to dest. Otherwise, build a
142 path to it by considering it is relative to base_path. URL are
143 supported. */
144 void path_combine(char *dest, int dest_size,
145 const char *base_path,
146 const char *filename)
147 {
148 const char *p, *p1;
149 int len;
150
151 if (dest_size <= 0)
152 return;
153 if (path_is_absolute(filename)) {
154 pstrcpy(dest, dest_size, filename);
155 } else {
156 p = strchr(base_path, ':');
157 if (p)
158 p++;
159 else
160 p = base_path;
161 p1 = strrchr(base_path, '/');
162 #ifdef _WIN32
163 {
164 const char *p2;
165 p2 = strrchr(base_path, '\\');
166 if (!p1 || p2 > p1)
167 p1 = p2;
168 }
169 #endif
170 if (p1)
171 p1++;
172 else
173 p1 = base_path;
174 if (p1 > p)
175 p = p1;
176 len = p - base_path;
177 if (len > dest_size - 1)
178 len = dest_size - 1;
179 memcpy(dest, base_path, len);
180 dest[len] = '\0';
181 pstrcat(dest, dest_size, filename);
182 }
183 }
184
185 void bdrv_register(BlockDriver *bdrv)
186 {
187 if (bdrv->bdrv_co_readv) {
188 /* Emulate AIO by coroutines, and sync by AIO */
189 bdrv->bdrv_aio_readv = bdrv_co_aio_readv_em;
190 bdrv->bdrv_aio_writev = bdrv_co_aio_writev_em;
191 bdrv->bdrv_read = bdrv_read_em;
192 bdrv->bdrv_write = bdrv_write_em;
193 } else {
194 bdrv->bdrv_co_readv = bdrv_co_readv_em;
195 bdrv->bdrv_co_writev = bdrv_co_writev_em;
196
197 if (!bdrv->bdrv_aio_readv) {
198 /* add AIO emulation layer */
199 bdrv->bdrv_aio_readv = bdrv_aio_readv_em;
200 bdrv->bdrv_aio_writev = bdrv_aio_writev_em;
201 } else if (!bdrv->bdrv_read) {
202 /* add synchronous IO emulation layer */
203 bdrv->bdrv_read = bdrv_read_em;
204 bdrv->bdrv_write = bdrv_write_em;
205 }
206 }
207
208 if (!bdrv->bdrv_aio_flush)
209 bdrv->bdrv_aio_flush = bdrv_aio_flush_em;
210
211 QLIST_INSERT_HEAD(&bdrv_drivers, bdrv, list);
212 }
213
214 /* create a new block device (by default it is empty) */
215 BlockDriverState *bdrv_new(const char *device_name)
216 {
217 BlockDriverState *bs;
218
219 bs = g_malloc0(sizeof(BlockDriverState));
220 pstrcpy(bs->device_name, sizeof(bs->device_name), device_name);
221 if (device_name[0] != '\0') {
222 QTAILQ_INSERT_TAIL(&bdrv_states, bs, list);
223 }
224 return bs;
225 }
226
227 BlockDriver *bdrv_find_format(const char *format_name)
228 {
229 BlockDriver *drv1;
230 QLIST_FOREACH(drv1, &bdrv_drivers, list) {
231 if (!strcmp(drv1->format_name, format_name)) {
232 return drv1;
233 }
234 }
235 return NULL;
236 }
237
238 static int bdrv_is_whitelisted(BlockDriver *drv)
239 {
240 static const char *whitelist[] = {
241 CONFIG_BDRV_WHITELIST
242 };
243 const char **p;
244
245 if (!whitelist[0])
246 return 1; /* no whitelist, anything goes */
247
248 for (p = whitelist; *p; p++) {
249 if (!strcmp(drv->format_name, *p)) {
250 return 1;
251 }
252 }
253 return 0;
254 }
255
256 BlockDriver *bdrv_find_whitelisted_format(const char *format_name)
257 {
258 BlockDriver *drv = bdrv_find_format(format_name);
259 return drv && bdrv_is_whitelisted(drv) ? drv : NULL;
260 }
261
262 int bdrv_create(BlockDriver *drv, const char* filename,
263 QEMUOptionParameter *options)
264 {
265 if (!drv->bdrv_create)
266 return -ENOTSUP;
267
268 return drv->bdrv_create(filename, options);
269 }
270
271 int bdrv_create_file(const char* filename, QEMUOptionParameter *options)
272 {
273 BlockDriver *drv;
274
275 drv = bdrv_find_protocol(filename);
276 if (drv == NULL) {
277 return -ENOENT;
278 }
279
280 return bdrv_create(drv, filename, options);
281 }
282
283 #ifdef _WIN32
284 void get_tmp_filename(char *filename, int size)
285 {
286 char temp_dir[MAX_PATH];
287
288 GetTempPath(MAX_PATH, temp_dir);
289 GetTempFileName(temp_dir, "qem", 0, filename);
290 }
291 #else
292 void get_tmp_filename(char *filename, int size)
293 {
294 int fd;
295 const char *tmpdir;
296 /* XXX: race condition possible */
297 tmpdir = getenv("TMPDIR");
298 if (!tmpdir)
299 tmpdir = "/tmp";
300 snprintf(filename, size, "%s/vl.XXXXXX", tmpdir);
301 fd = mkstemp(filename);
302 close(fd);
303 }
304 #endif
305
306 /*
307 * Detect host devices. By convention, /dev/cdrom[N] is always
308 * recognized as a host CDROM.
309 */
310 static BlockDriver *find_hdev_driver(const char *filename)
311 {
312 int score_max = 0, score;
313 BlockDriver *drv = NULL, *d;
314
315 QLIST_FOREACH(d, &bdrv_drivers, list) {
316 if (d->bdrv_probe_device) {
317 score = d->bdrv_probe_device(filename);
318 if (score > score_max) {
319 score_max = score;
320 drv = d;
321 }
322 }
323 }
324
325 return drv;
326 }
327
328 BlockDriver *bdrv_find_protocol(const char *filename)
329 {
330 BlockDriver *drv1;
331 char protocol[128];
332 int len;
333 const char *p;
334
335 /* TODO Drivers without bdrv_file_open must be specified explicitly */
336
337 /*
338 * XXX(hch): we really should not let host device detection
339 * override an explicit protocol specification, but moving this
340 * later breaks access to device names with colons in them.
341 * Thanks to the brain-dead persistent naming schemes on udev-
342 * based Linux systems those actually are quite common.
343 */
344 drv1 = find_hdev_driver(filename);
345 if (drv1) {
346 return drv1;
347 }
348
349 if (!path_has_protocol(filename)) {
350 return bdrv_find_format("file");
351 }
352 p = strchr(filename, ':');
353 assert(p != NULL);
354 len = p - filename;
355 if (len > sizeof(protocol) - 1)
356 len = sizeof(protocol) - 1;
357 memcpy(protocol, filename, len);
358 protocol[len] = '\0';
359 QLIST_FOREACH(drv1, &bdrv_drivers, list) {
360 if (drv1->protocol_name &&
361 !strcmp(drv1->protocol_name, protocol)) {
362 return drv1;
363 }
364 }
365 return NULL;
366 }
367
368 static int find_image_format(const char *filename, BlockDriver **pdrv)
369 {
370 int ret, score, score_max;
371 BlockDriver *drv1, *drv;
372 uint8_t buf[2048];
373 BlockDriverState *bs;
374
375 ret = bdrv_file_open(&bs, filename, 0);
376 if (ret < 0) {
377 *pdrv = NULL;
378 return ret;
379 }
380
381 /* Return the raw BlockDriver * to scsi-generic devices or empty drives */
382 if (bs->sg || !bdrv_is_inserted(bs)) {
383 bdrv_delete(bs);
384 drv = bdrv_find_format("raw");
385 if (!drv) {
386 ret = -ENOENT;
387 }
388 *pdrv = drv;
389 return ret;
390 }
391
392 ret = bdrv_pread(bs, 0, buf, sizeof(buf));
393 bdrv_delete(bs);
394 if (ret < 0) {
395 *pdrv = NULL;
396 return ret;
397 }
398
399 score_max = 0;
400 drv = NULL;
401 QLIST_FOREACH(drv1, &bdrv_drivers, list) {
402 if (drv1->bdrv_probe) {
403 score = drv1->bdrv_probe(buf, ret, filename);
404 if (score > score_max) {
405 score_max = score;
406 drv = drv1;
407 }
408 }
409 }
410 if (!drv) {
411 ret = -ENOENT;
412 }
413 *pdrv = drv;
414 return ret;
415 }
416
417 /**
418 * Set the current 'total_sectors' value
419 */
420 static int refresh_total_sectors(BlockDriverState *bs, int64_t hint)
421 {
422 BlockDriver *drv = bs->drv;
423
424 /* Do not attempt drv->bdrv_getlength() on scsi-generic devices */
425 if (bs->sg)
426 return 0;
427
428 /* query actual device if possible, otherwise just trust the hint */
429 if (drv->bdrv_getlength) {
430 int64_t length = drv->bdrv_getlength(bs);
431 if (length < 0) {
432 return length;
433 }
434 hint = length >> BDRV_SECTOR_BITS;
435 }
436
437 bs->total_sectors = hint;
438 return 0;
439 }
440
441 /**
442 * Set open flags for a given cache mode
443 *
444 * Return 0 on success, -1 if the cache mode was invalid.
445 */
446 int bdrv_parse_cache_flags(const char *mode, int *flags)
447 {
448 *flags &= ~BDRV_O_CACHE_MASK;
449
450 if (!strcmp(mode, "off") || !strcmp(mode, "none")) {
451 *flags |= BDRV_O_NOCACHE | BDRV_O_CACHE_WB;
452 } else if (!strcmp(mode, "directsync")) {
453 *flags |= BDRV_O_NOCACHE;
454 } else if (!strcmp(mode, "writeback")) {
455 *flags |= BDRV_O_CACHE_WB;
456 } else if (!strcmp(mode, "unsafe")) {
457 *flags |= BDRV_O_CACHE_WB;
458 *flags |= BDRV_O_NO_FLUSH;
459 } else if (!strcmp(mode, "writethrough")) {
460 /* this is the default */
461 } else {
462 return -1;
463 }
464
465 return 0;
466 }
467
468 /*
469 * Common part for opening disk images and files
470 */
471 static int bdrv_open_common(BlockDriverState *bs, const char *filename,
472 int flags, BlockDriver *drv)
473 {
474 int ret, open_flags;
475
476 assert(drv != NULL);
477
478 bs->file = NULL;
479 bs->total_sectors = 0;
480 bs->encrypted = 0;
481 bs->valid_key = 0;
482 bs->open_flags = flags;
483 bs->buffer_alignment = 512;
484
485 pstrcpy(bs->filename, sizeof(bs->filename), filename);
486
487 if (use_bdrv_whitelist && !bdrv_is_whitelisted(drv)) {
488 return -ENOTSUP;
489 }
490
491 bs->drv = drv;
492 bs->opaque = g_malloc0(drv->instance_size);
493
494 if (flags & BDRV_O_CACHE_WB)
495 bs->enable_write_cache = 1;
496
497 /*
498 * Clear flags that are internal to the block layer before opening the
499 * image.
500 */
501 open_flags = flags & ~(BDRV_O_SNAPSHOT | BDRV_O_NO_BACKING);
502
503 /*
504 * Snapshots should be writable.
505 */
506 if (bs->is_temporary) {
507 open_flags |= BDRV_O_RDWR;
508 }
509
510 /* Open the image, either directly or using a protocol */
511 if (drv->bdrv_file_open) {
512 ret = drv->bdrv_file_open(bs, filename, open_flags);
513 } else {
514 ret = bdrv_file_open(&bs->file, filename, open_flags);
515 if (ret >= 0) {
516 ret = drv->bdrv_open(bs, open_flags);
517 }
518 }
519
520 if (ret < 0) {
521 goto free_and_fail;
522 }
523
524 bs->keep_read_only = bs->read_only = !(open_flags & BDRV_O_RDWR);
525
526 ret = refresh_total_sectors(bs, bs->total_sectors);
527 if (ret < 0) {
528 goto free_and_fail;
529 }
530
531 #ifndef _WIN32
532 if (bs->is_temporary) {
533 unlink(filename);
534 }
535 #endif
536 return 0;
537
538 free_and_fail:
539 if (bs->file) {
540 bdrv_delete(bs->file);
541 bs->file = NULL;
542 }
543 g_free(bs->opaque);
544 bs->opaque = NULL;
545 bs->drv = NULL;
546 return ret;
547 }
548
549 /*
550 * Opens a file using a protocol (file, host_device, nbd, ...)
551 */
552 int bdrv_file_open(BlockDriverState **pbs, const char *filename, int flags)
553 {
554 BlockDriverState *bs;
555 BlockDriver *drv;
556 int ret;
557
558 drv = bdrv_find_protocol(filename);
559 if (!drv) {
560 return -ENOENT;
561 }
562
563 bs = bdrv_new("");
564 ret = bdrv_open_common(bs, filename, flags, drv);
565 if (ret < 0) {
566 bdrv_delete(bs);
567 return ret;
568 }
569 bs->growable = 1;
570 *pbs = bs;
571 return 0;
572 }
573
574 /*
575 * Opens a disk image (raw, qcow2, vmdk, ...)
576 */
577 int bdrv_open(BlockDriverState *bs, const char *filename, int flags,
578 BlockDriver *drv)
579 {
580 int ret;
581
582 if (flags & BDRV_O_SNAPSHOT) {
583 BlockDriverState *bs1;
584 int64_t total_size;
585 int is_protocol = 0;
586 BlockDriver *bdrv_qcow2;
587 QEMUOptionParameter *options;
588 char tmp_filename[PATH_MAX];
589 char backing_filename[PATH_MAX];
590
591 /* if snapshot, we create a temporary backing file and open it
592 instead of opening 'filename' directly */
593
594 /* if there is a backing file, use it */
595 bs1 = bdrv_new("");
596 ret = bdrv_open(bs1, filename, 0, drv);
597 if (ret < 0) {
598 bdrv_delete(bs1);
599 return ret;
600 }
601 total_size = bdrv_getlength(bs1) & BDRV_SECTOR_MASK;
602
603 if (bs1->drv && bs1->drv->protocol_name)
604 is_protocol = 1;
605
606 bdrv_delete(bs1);
607
608 get_tmp_filename(tmp_filename, sizeof(tmp_filename));
609
610 /* Real path is meaningless for protocols */
611 if (is_protocol)
612 snprintf(backing_filename, sizeof(backing_filename),
613 "%s", filename);
614 else if (!realpath(filename, backing_filename))
615 return -errno;
616
617 bdrv_qcow2 = bdrv_find_format("qcow2");
618 options = parse_option_parameters("", bdrv_qcow2->create_options, NULL);
619
620 set_option_parameter_int(options, BLOCK_OPT_SIZE, total_size);
621 set_option_parameter(options, BLOCK_OPT_BACKING_FILE, backing_filename);
622 if (drv) {
623 set_option_parameter(options, BLOCK_OPT_BACKING_FMT,
624 drv->format_name);
625 }
626
627 ret = bdrv_create(bdrv_qcow2, tmp_filename, options);
628 free_option_parameters(options);
629 if (ret < 0) {
630 return ret;
631 }
632
633 filename = tmp_filename;
634 drv = bdrv_qcow2;
635 bs->is_temporary = 1;
636 }
637
638 /* Find the right image format driver */
639 if (!drv) {
640 ret = find_image_format(filename, &drv);
641 }
642
643 if (!drv) {
644 goto unlink_and_fail;
645 }
646
647 /* Open the image */
648 ret = bdrv_open_common(bs, filename, flags, drv);
649 if (ret < 0) {
650 goto unlink_and_fail;
651 }
652
653 /* If there is a backing file, use it */
654 if ((flags & BDRV_O_NO_BACKING) == 0 && bs->backing_file[0] != '\0') {
655 char backing_filename[PATH_MAX];
656 int back_flags;
657 BlockDriver *back_drv = NULL;
658
659 bs->backing_hd = bdrv_new("");
660
661 if (path_has_protocol(bs->backing_file)) {
662 pstrcpy(backing_filename, sizeof(backing_filename),
663 bs->backing_file);
664 } else {
665 path_combine(backing_filename, sizeof(backing_filename),
666 filename, bs->backing_file);
667 }
668
669 if (bs->backing_format[0] != '\0') {
670 back_drv = bdrv_find_format(bs->backing_format);
671 }
672
673 /* backing files always opened read-only */
674 back_flags =
675 flags & ~(BDRV_O_RDWR | BDRV_O_SNAPSHOT | BDRV_O_NO_BACKING);
676
677 ret = bdrv_open(bs->backing_hd, backing_filename, back_flags, back_drv);
678 if (ret < 0) {
679 bdrv_close(bs);
680 return ret;
681 }
682 if (bs->is_temporary) {
683 bs->backing_hd->keep_read_only = !(flags & BDRV_O_RDWR);
684 } else {
685 /* base image inherits from "parent" */
686 bs->backing_hd->keep_read_only = bs->keep_read_only;
687 }
688 }
689
690 if (!bdrv_key_required(bs)) {
691 bdrv_dev_change_media_cb(bs, true);
692 }
693
694 return 0;
695
696 unlink_and_fail:
697 if (bs->is_temporary) {
698 unlink(filename);
699 }
700 return ret;
701 }
702
703 void bdrv_close(BlockDriverState *bs)
704 {
705 if (bs->drv) {
706 if (bs == bs_snapshots) {
707 bs_snapshots = NULL;
708 }
709 if (bs->backing_hd) {
710 bdrv_delete(bs->backing_hd);
711 bs->backing_hd = NULL;
712 }
713 bs->drv->bdrv_close(bs);
714 g_free(bs->opaque);
715 #ifdef _WIN32
716 if (bs->is_temporary) {
717 unlink(bs->filename);
718 }
719 #endif
720 bs->opaque = NULL;
721 bs->drv = NULL;
722
723 if (bs->file != NULL) {
724 bdrv_close(bs->file);
725 }
726
727 bdrv_dev_change_media_cb(bs, false);
728 }
729 }
730
731 void bdrv_close_all(void)
732 {
733 BlockDriverState *bs;
734
735 QTAILQ_FOREACH(bs, &bdrv_states, list) {
736 bdrv_close(bs);
737 }
738 }
739
740 /* make a BlockDriverState anonymous by removing from bdrv_state list.
741 Also, NULL terminate the device_name to prevent double remove */
742 void bdrv_make_anon(BlockDriverState *bs)
743 {
744 if (bs->device_name[0] != '\0') {
745 QTAILQ_REMOVE(&bdrv_states, bs, list);
746 }
747 bs->device_name[0] = '\0';
748 }
749
750 void bdrv_delete(BlockDriverState *bs)
751 {
752 assert(!bs->dev);
753
754 /* remove from list, if necessary */
755 bdrv_make_anon(bs);
756
757 bdrv_close(bs);
758 if (bs->file != NULL) {
759 bdrv_delete(bs->file);
760 }
761
762 assert(bs != bs_snapshots);
763 g_free(bs);
764 }
765
766 int bdrv_attach_dev(BlockDriverState *bs, void *dev)
767 /* TODO change to DeviceState *dev when all users are qdevified */
768 {
769 if (bs->dev) {
770 return -EBUSY;
771 }
772 bs->dev = dev;
773 return 0;
774 }
775
776 /* TODO qdevified devices don't use this, remove when devices are qdevified */
777 void bdrv_attach_dev_nofail(BlockDriverState *bs, void *dev)
778 {
779 if (bdrv_attach_dev(bs, dev) < 0) {
780 abort();
781 }
782 }
783
784 void bdrv_detach_dev(BlockDriverState *bs, void *dev)
785 /* TODO change to DeviceState *dev when all users are qdevified */
786 {
787 assert(bs->dev == dev);
788 bs->dev = NULL;
789 bs->dev_ops = NULL;
790 bs->dev_opaque = NULL;
791 bs->buffer_alignment = 512;
792 }
793
794 /* TODO change to return DeviceState * when all users are qdevified */
795 void *bdrv_get_attached_dev(BlockDriverState *bs)
796 {
797 return bs->dev;
798 }
799
800 void bdrv_set_dev_ops(BlockDriverState *bs, const BlockDevOps *ops,
801 void *opaque)
802 {
803 bs->dev_ops = ops;
804 bs->dev_opaque = opaque;
805 if (bdrv_dev_has_removable_media(bs) && bs == bs_snapshots) {
806 bs_snapshots = NULL;
807 }
808 }
809
810 static void bdrv_dev_change_media_cb(BlockDriverState *bs, bool load)
811 {
812 if (bs->dev_ops && bs->dev_ops->change_media_cb) {
813 bs->dev_ops->change_media_cb(bs->dev_opaque, load);
814 }
815 }
816
817 bool bdrv_dev_has_removable_media(BlockDriverState *bs)
818 {
819 return !bs->dev || (bs->dev_ops && bs->dev_ops->change_media_cb);
820 }
821
822 bool bdrv_dev_is_tray_open(BlockDriverState *bs)
823 {
824 if (bs->dev_ops && bs->dev_ops->is_tray_open) {
825 return bs->dev_ops->is_tray_open(bs->dev_opaque);
826 }
827 return false;
828 }
829
830 static void bdrv_dev_resize_cb(BlockDriverState *bs)
831 {
832 if (bs->dev_ops && bs->dev_ops->resize_cb) {
833 bs->dev_ops->resize_cb(bs->dev_opaque);
834 }
835 }
836
837 bool bdrv_dev_is_medium_locked(BlockDriverState *bs)
838 {
839 if (bs->dev_ops && bs->dev_ops->is_medium_locked) {
840 return bs->dev_ops->is_medium_locked(bs->dev_opaque);
841 }
842 return false;
843 }
844
845 /*
846 * Run consistency checks on an image
847 *
848 * Returns 0 if the check could be completed (it doesn't mean that the image is
849 * free of errors) or -errno when an internal error occurred. The results of the
850 * check are stored in res.
851 */
852 int bdrv_check(BlockDriverState *bs, BdrvCheckResult *res)
853 {
854 if (bs->drv->bdrv_check == NULL) {
855 return -ENOTSUP;
856 }
857
858 memset(res, 0, sizeof(*res));
859 return bs->drv->bdrv_check(bs, res);
860 }
861
862 #define COMMIT_BUF_SECTORS 2048
863
864 /* commit COW file into the raw image */
865 int bdrv_commit(BlockDriverState *bs)
866 {
867 BlockDriver *drv = bs->drv;
868 BlockDriver *backing_drv;
869 int64_t sector, total_sectors;
870 int n, ro, open_flags;
871 int ret = 0, rw_ret = 0;
872 uint8_t *buf;
873 char filename[1024];
874 BlockDriverState *bs_rw, *bs_ro;
875
876 if (!drv)
877 return -ENOMEDIUM;
878
879 if (!bs->backing_hd) {
880 return -ENOTSUP;
881 }
882
883 if (bs->backing_hd->keep_read_only) {
884 return -EACCES;
885 }
886
887 backing_drv = bs->backing_hd->drv;
888 ro = bs->backing_hd->read_only;
889 strncpy(filename, bs->backing_hd->filename, sizeof(filename));
890 open_flags = bs->backing_hd->open_flags;
891
892 if (ro) {
893 /* re-open as RW */
894 bdrv_delete(bs->backing_hd);
895 bs->backing_hd = NULL;
896 bs_rw = bdrv_new("");
897 rw_ret = bdrv_open(bs_rw, filename, open_flags | BDRV_O_RDWR,
898 backing_drv);
899 if (rw_ret < 0) {
900 bdrv_delete(bs_rw);
901 /* try to re-open read-only */
902 bs_ro = bdrv_new("");
903 ret = bdrv_open(bs_ro, filename, open_flags & ~BDRV_O_RDWR,
904 backing_drv);
905 if (ret < 0) {
906 bdrv_delete(bs_ro);
907 /* drive not functional anymore */
908 bs->drv = NULL;
909 return ret;
910 }
911 bs->backing_hd = bs_ro;
912 return rw_ret;
913 }
914 bs->backing_hd = bs_rw;
915 }
916
917 total_sectors = bdrv_getlength(bs) >> BDRV_SECTOR_BITS;
918 buf = g_malloc(COMMIT_BUF_SECTORS * BDRV_SECTOR_SIZE);
919
920 for (sector = 0; sector < total_sectors; sector += n) {
921 if (drv->bdrv_is_allocated(bs, sector, COMMIT_BUF_SECTORS, &n)) {
922
923 if (bdrv_read(bs, sector, buf, n) != 0) {
924 ret = -EIO;
925 goto ro_cleanup;
926 }
927
928 if (bdrv_write(bs->backing_hd, sector, buf, n) != 0) {
929 ret = -EIO;
930 goto ro_cleanup;
931 }
932 }
933 }
934
935 if (drv->bdrv_make_empty) {
936 ret = drv->bdrv_make_empty(bs);
937 bdrv_flush(bs);
938 }
939
940 /*
941 * Make sure all data we wrote to the backing device is actually
942 * stable on disk.
943 */
944 if (bs->backing_hd)
945 bdrv_flush(bs->backing_hd);
946
947 ro_cleanup:
948 g_free(buf);
949
950 if (ro) {
951 /* re-open as RO */
952 bdrv_delete(bs->backing_hd);
953 bs->backing_hd = NULL;
954 bs_ro = bdrv_new("");
955 ret = bdrv_open(bs_ro, filename, open_flags & ~BDRV_O_RDWR,
956 backing_drv);
957 if (ret < 0) {
958 bdrv_delete(bs_ro);
959 /* drive not functional anymore */
960 bs->drv = NULL;
961 return ret;
962 }
963 bs->backing_hd = bs_ro;
964 bs->backing_hd->keep_read_only = 0;
965 }
966
967 return ret;
968 }
969
970 void bdrv_commit_all(void)
971 {
972 BlockDriverState *bs;
973
974 QTAILQ_FOREACH(bs, &bdrv_states, list) {
975 bdrv_commit(bs);
976 }
977 }
978
979 /*
980 * Return values:
981 * 0 - success
982 * -EINVAL - backing format specified, but no file
983 * -ENOSPC - can't update the backing file because no space is left in the
984 * image file header
985 * -ENOTSUP - format driver doesn't support changing the backing file
986 */
987 int bdrv_change_backing_file(BlockDriverState *bs,
988 const char *backing_file, const char *backing_fmt)
989 {
990 BlockDriver *drv = bs->drv;
991
992 if (drv->bdrv_change_backing_file != NULL) {
993 return drv->bdrv_change_backing_file(bs, backing_file, backing_fmt);
994 } else {
995 return -ENOTSUP;
996 }
997 }
998
999 static int bdrv_check_byte_request(BlockDriverState *bs, int64_t offset,
1000 size_t size)
1001 {
1002 int64_t len;
1003
1004 if (!bdrv_is_inserted(bs))
1005 return -ENOMEDIUM;
1006
1007 if (bs->growable)
1008 return 0;
1009
1010 len = bdrv_getlength(bs);
1011
1012 if (offset < 0)
1013 return -EIO;
1014
1015 if ((offset > len) || (len - offset < size))
1016 return -EIO;
1017
1018 return 0;
1019 }
1020
1021 static int bdrv_check_request(BlockDriverState *bs, int64_t sector_num,
1022 int nb_sectors)
1023 {
1024 return bdrv_check_byte_request(bs, sector_num * BDRV_SECTOR_SIZE,
1025 nb_sectors * BDRV_SECTOR_SIZE);
1026 }
1027
1028 static inline bool bdrv_has_async_rw(BlockDriver *drv)
1029 {
1030 return drv->bdrv_co_readv != bdrv_co_readv_em
1031 || drv->bdrv_aio_readv != bdrv_aio_readv_em;
1032 }
1033
1034 static inline bool bdrv_has_async_flush(BlockDriver *drv)
1035 {
1036 return drv->bdrv_aio_flush != bdrv_aio_flush_em;
1037 }
1038
1039 /* return < 0 if error. See bdrv_write() for the return codes */
1040 int bdrv_read(BlockDriverState *bs, int64_t sector_num,
1041 uint8_t *buf, int nb_sectors)
1042 {
1043 BlockDriver *drv = bs->drv;
1044
1045 if (!drv)
1046 return -ENOMEDIUM;
1047
1048 if (bdrv_has_async_rw(drv) && qemu_in_coroutine()) {
1049 QEMUIOVector qiov;
1050 struct iovec iov = {
1051 .iov_base = (void *)buf,
1052 .iov_len = nb_sectors * BDRV_SECTOR_SIZE,
1053 };
1054
1055 qemu_iovec_init_external(&qiov, &iov, 1);
1056 return bdrv_co_readv(bs, sector_num, nb_sectors, &qiov);
1057 }
1058
1059 if (bdrv_check_request(bs, sector_num, nb_sectors))
1060 return -EIO;
1061
1062 return drv->bdrv_read(bs, sector_num, buf, nb_sectors);
1063 }
1064
1065 static void set_dirty_bitmap(BlockDriverState *bs, int64_t sector_num,
1066 int nb_sectors, int dirty)
1067 {
1068 int64_t start, end;
1069 unsigned long val, idx, bit;
1070
1071 start = sector_num / BDRV_SECTORS_PER_DIRTY_CHUNK;
1072 end = (sector_num + nb_sectors - 1) / BDRV_SECTORS_PER_DIRTY_CHUNK;
1073
1074 for (; start <= end; start++) {
1075 idx = start / (sizeof(unsigned long) * 8);
1076 bit = start % (sizeof(unsigned long) * 8);
1077 val = bs->dirty_bitmap[idx];
1078 if (dirty) {
1079 if (!(val & (1UL << bit))) {
1080 bs->dirty_count++;
1081 val |= 1UL << bit;
1082 }
1083 } else {
1084 if (val & (1UL << bit)) {
1085 bs->dirty_count--;
1086 val &= ~(1UL << bit);
1087 }
1088 }
1089 bs->dirty_bitmap[idx] = val;
1090 }
1091 }
1092
1093 /* Return < 0 if error. Important errors are:
1094 -EIO generic I/O error (may happen for all errors)
1095 -ENOMEDIUM No media inserted.
1096 -EINVAL Invalid sector number or nb_sectors
1097 -EACCES Trying to write a read-only device
1098 */
1099 int bdrv_write(BlockDriverState *bs, int64_t sector_num,
1100 const uint8_t *buf, int nb_sectors)
1101 {
1102 BlockDriver *drv = bs->drv;
1103
1104 if (!bs->drv)
1105 return -ENOMEDIUM;
1106
1107 if (bdrv_has_async_rw(drv) && qemu_in_coroutine()) {
1108 QEMUIOVector qiov;
1109 struct iovec iov = {
1110 .iov_base = (void *)buf,
1111 .iov_len = nb_sectors * BDRV_SECTOR_SIZE,
1112 };
1113
1114 qemu_iovec_init_external(&qiov, &iov, 1);
1115 return bdrv_co_writev(bs, sector_num, nb_sectors, &qiov);
1116 }
1117
1118 if (bs->read_only)
1119 return -EACCES;
1120 if (bdrv_check_request(bs, sector_num, nb_sectors))
1121 return -EIO;
1122
1123 if (bs->dirty_bitmap) {
1124 set_dirty_bitmap(bs, sector_num, nb_sectors, 1);
1125 }
1126
1127 if (bs->wr_highest_sector < sector_num + nb_sectors - 1) {
1128 bs->wr_highest_sector = sector_num + nb_sectors - 1;
1129 }
1130
1131 return drv->bdrv_write(bs, sector_num, buf, nb_sectors);
1132 }
1133
1134 int bdrv_pread(BlockDriverState *bs, int64_t offset,
1135 void *buf, int count1)
1136 {
1137 uint8_t tmp_buf[BDRV_SECTOR_SIZE];
1138 int len, nb_sectors, count;
1139 int64_t sector_num;
1140 int ret;
1141
1142 count = count1;
1143 /* first read to align to sector start */
1144 len = (BDRV_SECTOR_SIZE - offset) & (BDRV_SECTOR_SIZE - 1);
1145 if (len > count)
1146 len = count;
1147 sector_num = offset >> BDRV_SECTOR_BITS;
1148 if (len > 0) {
1149 if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0)
1150 return ret;
1151 memcpy(buf, tmp_buf + (offset & (BDRV_SECTOR_SIZE - 1)), len);
1152 count -= len;
1153 if (count == 0)
1154 return count1;
1155 sector_num++;
1156 buf += len;
1157 }
1158
1159 /* read the sectors "in place" */
1160 nb_sectors = count >> BDRV_SECTOR_BITS;
1161 if (nb_sectors > 0) {
1162 if ((ret = bdrv_read(bs, sector_num, buf, nb_sectors)) < 0)
1163 return ret;
1164 sector_num += nb_sectors;
1165 len = nb_sectors << BDRV_SECTOR_BITS;
1166 buf += len;
1167 count -= len;
1168 }
1169
1170 /* add data from the last sector */
1171 if (count > 0) {
1172 if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0)
1173 return ret;
1174 memcpy(buf, tmp_buf, count);
1175 }
1176 return count1;
1177 }
1178
1179 int bdrv_pwrite(BlockDriverState *bs, int64_t offset,
1180 const void *buf, int count1)
1181 {
1182 uint8_t tmp_buf[BDRV_SECTOR_SIZE];
1183 int len, nb_sectors, count;
1184 int64_t sector_num;
1185 int ret;
1186
1187 count = count1;
1188 /* first write to align to sector start */
1189 len = (BDRV_SECTOR_SIZE - offset) & (BDRV_SECTOR_SIZE - 1);
1190 if (len > count)
1191 len = count;
1192 sector_num = offset >> BDRV_SECTOR_BITS;
1193 if (len > 0) {
1194 if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0)
1195 return ret;
1196 memcpy(tmp_buf + (offset & (BDRV_SECTOR_SIZE - 1)), buf, len);
1197 if ((ret = bdrv_write(bs, sector_num, tmp_buf, 1)) < 0)
1198 return ret;
1199 count -= len;
1200 if (count == 0)
1201 return count1;
1202 sector_num++;
1203 buf += len;
1204 }
1205
1206 /* write the sectors "in place" */
1207 nb_sectors = count >> BDRV_SECTOR_BITS;
1208 if (nb_sectors > 0) {
1209 if ((ret = bdrv_write(bs, sector_num, buf, nb_sectors)) < 0)
1210 return ret;
1211 sector_num += nb_sectors;
1212 len = nb_sectors << BDRV_SECTOR_BITS;
1213 buf += len;
1214 count -= len;
1215 }
1216
1217 /* add data from the last sector */
1218 if (count > 0) {
1219 if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0)
1220 return ret;
1221 memcpy(tmp_buf, buf, count);
1222 if ((ret = bdrv_write(bs, sector_num, tmp_buf, 1)) < 0)
1223 return ret;
1224 }
1225 return count1;
1226 }
1227
1228 /*
1229 * Writes to the file and ensures that no writes are reordered across this
1230 * request (acts as a barrier)
1231 *
1232 * Returns 0 on success, -errno in error cases.
1233 */
1234 int bdrv_pwrite_sync(BlockDriverState *bs, int64_t offset,
1235 const void *buf, int count)
1236 {
1237 int ret;
1238
1239 ret = bdrv_pwrite(bs, offset, buf, count);
1240 if (ret < 0) {
1241 return ret;
1242 }
1243
1244 /* No flush needed for cache modes that use O_DSYNC */
1245 if ((bs->open_flags & BDRV_O_CACHE_WB) != 0) {
1246 bdrv_flush(bs);
1247 }
1248
1249 return 0;
1250 }
1251
1252 int coroutine_fn bdrv_co_readv(BlockDriverState *bs, int64_t sector_num,
1253 int nb_sectors, QEMUIOVector *qiov)
1254 {
1255 BlockDriver *drv = bs->drv;
1256
1257 trace_bdrv_co_readv(bs, sector_num, nb_sectors);
1258
1259 if (!drv) {
1260 return -ENOMEDIUM;
1261 }
1262 if (bdrv_check_request(bs, sector_num, nb_sectors)) {
1263 return -EIO;
1264 }
1265
1266 return drv->bdrv_co_readv(bs, sector_num, nb_sectors, qiov);
1267 }
1268
1269 int coroutine_fn bdrv_co_writev(BlockDriverState *bs, int64_t sector_num,
1270 int nb_sectors, QEMUIOVector *qiov)
1271 {
1272 BlockDriver *drv = bs->drv;
1273
1274 trace_bdrv_co_writev(bs, sector_num, nb_sectors);
1275
1276 if (!bs->drv) {
1277 return -ENOMEDIUM;
1278 }
1279 if (bs->read_only) {
1280 return -EACCES;
1281 }
1282 if (bdrv_check_request(bs, sector_num, nb_sectors)) {
1283 return -EIO;
1284 }
1285
1286 if (bs->dirty_bitmap) {
1287 set_dirty_bitmap(bs, sector_num, nb_sectors, 1);
1288 }
1289
1290 if (bs->wr_highest_sector < sector_num + nb_sectors - 1) {
1291 bs->wr_highest_sector = sector_num + nb_sectors - 1;
1292 }
1293
1294 return drv->bdrv_co_writev(bs, sector_num, nb_sectors, qiov);
1295 }
1296
1297 /**
1298 * Truncate file to 'offset' bytes (needed only for file protocols)
1299 */
1300 int bdrv_truncate(BlockDriverState *bs, int64_t offset)
1301 {
1302 BlockDriver *drv = bs->drv;
1303 int ret;
1304 if (!drv)
1305 return -ENOMEDIUM;
1306 if (!drv->bdrv_truncate)
1307 return -ENOTSUP;
1308 if (bs->read_only)
1309 return -EACCES;
1310 if (bdrv_in_use(bs))
1311 return -EBUSY;
1312 ret = drv->bdrv_truncate(bs, offset);
1313 if (ret == 0) {
1314 ret = refresh_total_sectors(bs, offset >> BDRV_SECTOR_BITS);
1315 bdrv_dev_resize_cb(bs);
1316 }
1317 return ret;
1318 }
1319
1320 /**
1321 * Length of a allocated file in bytes. Sparse files are counted by actual
1322 * allocated space. Return < 0 if error or unknown.
1323 */
1324 int64_t bdrv_get_allocated_file_size(BlockDriverState *bs)
1325 {
1326 BlockDriver *drv = bs->drv;
1327 if (!drv) {
1328 return -ENOMEDIUM;
1329 }
1330 if (drv->bdrv_get_allocated_file_size) {
1331 return drv->bdrv_get_allocated_file_size(bs);
1332 }
1333 if (bs->file) {
1334 return bdrv_get_allocated_file_size(bs->file);
1335 }
1336 return -ENOTSUP;
1337 }
1338
1339 /**
1340 * Length of a file in bytes. Return < 0 if error or unknown.
1341 */
1342 int64_t bdrv_getlength(BlockDriverState *bs)
1343 {
1344 BlockDriver *drv = bs->drv;
1345 if (!drv)
1346 return -ENOMEDIUM;
1347
1348 if (bs->growable || bdrv_dev_has_removable_media(bs)) {
1349 if (drv->bdrv_getlength) {
1350 return drv->bdrv_getlength(bs);
1351 }
1352 }
1353 return bs->total_sectors * BDRV_SECTOR_SIZE;
1354 }
1355
1356 /* return 0 as number of sectors if no device present or error */
1357 void bdrv_get_geometry(BlockDriverState *bs, uint64_t *nb_sectors_ptr)
1358 {
1359 int64_t length;
1360 length = bdrv_getlength(bs);
1361 if (length < 0)
1362 length = 0;
1363 else
1364 length = length >> BDRV_SECTOR_BITS;
1365 *nb_sectors_ptr = length;
1366 }
1367
1368 struct partition {
1369 uint8_t boot_ind; /* 0x80 - active */
1370 uint8_t head; /* starting head */
1371 uint8_t sector; /* starting sector */
1372 uint8_t cyl; /* starting cylinder */
1373 uint8_t sys_ind; /* What partition type */
1374 uint8_t end_head; /* end head */
1375 uint8_t end_sector; /* end sector */
1376 uint8_t end_cyl; /* end cylinder */
1377 uint32_t start_sect; /* starting sector counting from 0 */
1378 uint32_t nr_sects; /* nr of sectors in partition */
1379 } QEMU_PACKED;
1380
1381 /* try to guess the disk logical geometry from the MSDOS partition table. Return 0 if OK, -1 if could not guess */
1382 static int guess_disk_lchs(BlockDriverState *bs,
1383 int *pcylinders, int *pheads, int *psectors)
1384 {
1385 uint8_t buf[BDRV_SECTOR_SIZE];
1386 int ret, i, heads, sectors, cylinders;
1387 struct partition *p;
1388 uint32_t nr_sects;
1389 uint64_t nb_sectors;
1390
1391 bdrv_get_geometry(bs, &nb_sectors);
1392
1393 ret = bdrv_read(bs, 0, buf, 1);
1394 if (ret < 0)
1395 return -1;
1396 /* test msdos magic */
1397 if (buf[510] != 0x55 || buf[511] != 0xaa)
1398 return -1;
1399 for(i = 0; i < 4; i++) {
1400 p = ((struct partition *)(buf + 0x1be)) + i;
1401 nr_sects = le32_to_cpu(p->nr_sects);
1402 if (nr_sects && p->end_head) {
1403 /* We make the assumption that the partition terminates on
1404 a cylinder boundary */
1405 heads = p->end_head + 1;
1406 sectors = p->end_sector & 63;
1407 if (sectors == 0)
1408 continue;
1409 cylinders = nb_sectors / (heads * sectors);
1410 if (cylinders < 1 || cylinders > 16383)
1411 continue;
1412 *pheads = heads;
1413 *psectors = sectors;
1414 *pcylinders = cylinders;
1415 #if 0
1416 printf("guessed geometry: LCHS=%d %d %d\n",
1417 cylinders, heads, sectors);
1418 #endif
1419 return 0;
1420 }
1421 }
1422 return -1;
1423 }
1424
1425 void bdrv_guess_geometry(BlockDriverState *bs, int *pcyls, int *pheads, int *psecs)
1426 {
1427 int translation, lba_detected = 0;
1428 int cylinders, heads, secs;
1429 uint64_t nb_sectors;
1430
1431 /* if a geometry hint is available, use it */
1432 bdrv_get_geometry(bs, &nb_sectors);
1433 bdrv_get_geometry_hint(bs, &cylinders, &heads, &secs);
1434 translation = bdrv_get_translation_hint(bs);
1435 if (cylinders != 0) {
1436 *pcyls = cylinders;
1437 *pheads = heads;
1438 *psecs = secs;
1439 } else {
1440 if (guess_disk_lchs(bs, &cylinders, &heads, &secs) == 0) {
1441 if (heads > 16) {
1442 /* if heads > 16, it means that a BIOS LBA
1443 translation was active, so the default
1444 hardware geometry is OK */
1445 lba_detected = 1;
1446 goto default_geometry;
1447 } else {
1448 *pcyls = cylinders;
1449 *pheads = heads;
1450 *psecs = secs;
1451 /* disable any translation to be in sync with
1452 the logical geometry */
1453 if (translation == BIOS_ATA_TRANSLATION_AUTO) {
1454 bdrv_set_translation_hint(bs,
1455 BIOS_ATA_TRANSLATION_NONE);
1456 }
1457 }
1458 } else {
1459 default_geometry:
1460 /* if no geometry, use a standard physical disk geometry */
1461 cylinders = nb_sectors / (16 * 63);
1462
1463 if (cylinders > 16383)
1464 cylinders = 16383;
1465 else if (cylinders < 2)
1466 cylinders = 2;
1467 *pcyls = cylinders;
1468 *pheads = 16;
1469 *psecs = 63;
1470 if ((lba_detected == 1) && (translation == BIOS_ATA_TRANSLATION_AUTO)) {
1471 if ((*pcyls * *pheads) <= 131072) {
1472 bdrv_set_translation_hint(bs,
1473 BIOS_ATA_TRANSLATION_LARGE);
1474 } else {
1475 bdrv_set_translation_hint(bs,
1476 BIOS_ATA_TRANSLATION_LBA);
1477 }
1478 }
1479 }
1480 bdrv_set_geometry_hint(bs, *pcyls, *pheads, *psecs);
1481 }
1482 }
1483
1484 void bdrv_set_geometry_hint(BlockDriverState *bs,
1485 int cyls, int heads, int secs)
1486 {
1487 bs->cyls = cyls;
1488 bs->heads = heads;
1489 bs->secs = secs;
1490 }
1491
1492 void bdrv_set_translation_hint(BlockDriverState *bs, int translation)
1493 {
1494 bs->translation = translation;
1495 }
1496
1497 void bdrv_get_geometry_hint(BlockDriverState *bs,
1498 int *pcyls, int *pheads, int *psecs)
1499 {
1500 *pcyls = bs->cyls;
1501 *pheads = bs->heads;
1502 *psecs = bs->secs;
1503 }
1504
1505 /* Recognize floppy formats */
1506 typedef struct FDFormat {
1507 FDriveType drive;
1508 uint8_t last_sect;
1509 uint8_t max_track;
1510 uint8_t max_head;
1511 } FDFormat;
1512
1513 static const FDFormat fd_formats[] = {
1514 /* First entry is default format */
1515 /* 1.44 MB 3"1/2 floppy disks */
1516 { FDRIVE_DRV_144, 18, 80, 1, },
1517 { FDRIVE_DRV_144, 20, 80, 1, },
1518 { FDRIVE_DRV_144, 21, 80, 1, },
1519 { FDRIVE_DRV_144, 21, 82, 1, },
1520 { FDRIVE_DRV_144, 21, 83, 1, },
1521 { FDRIVE_DRV_144, 22, 80, 1, },
1522 { FDRIVE_DRV_144, 23, 80, 1, },
1523 { FDRIVE_DRV_144, 24, 80, 1, },
1524 /* 2.88 MB 3"1/2 floppy disks */
1525 { FDRIVE_DRV_288, 36, 80, 1, },
1526 { FDRIVE_DRV_288, 39, 80, 1, },
1527 { FDRIVE_DRV_288, 40, 80, 1, },
1528 { FDRIVE_DRV_288, 44, 80, 1, },
1529 { FDRIVE_DRV_288, 48, 80, 1, },
1530 /* 720 kB 3"1/2 floppy disks */
1531 { FDRIVE_DRV_144, 9, 80, 1, },
1532 { FDRIVE_DRV_144, 10, 80, 1, },
1533 { FDRIVE_DRV_144, 10, 82, 1, },
1534 { FDRIVE_DRV_144, 10, 83, 1, },
1535 { FDRIVE_DRV_144, 13, 80, 1, },
1536 { FDRIVE_DRV_144, 14, 80, 1, },
1537 /* 1.2 MB 5"1/4 floppy disks */
1538 { FDRIVE_DRV_120, 15, 80, 1, },
1539 { FDRIVE_DRV_120, 18, 80, 1, },
1540 { FDRIVE_DRV_120, 18, 82, 1, },
1541 { FDRIVE_DRV_120, 18, 83, 1, },
1542 { FDRIVE_DRV_120, 20, 80, 1, },
1543 /* 720 kB 5"1/4 floppy disks */
1544 { FDRIVE_DRV_120, 9, 80, 1, },
1545 { FDRIVE_DRV_120, 11, 80, 1, },
1546 /* 360 kB 5"1/4 floppy disks */
1547 { FDRIVE_DRV_120, 9, 40, 1, },
1548 { FDRIVE_DRV_120, 9, 40, 0, },
1549 { FDRIVE_DRV_120, 10, 41, 1, },
1550 { FDRIVE_DRV_120, 10, 42, 1, },
1551 /* 320 kB 5"1/4 floppy disks */
1552 { FDRIVE_DRV_120, 8, 40, 1, },
1553 { FDRIVE_DRV_120, 8, 40, 0, },
1554 /* 360 kB must match 5"1/4 better than 3"1/2... */
1555 { FDRIVE_DRV_144, 9, 80, 0, },
1556 /* end */
1557 { FDRIVE_DRV_NONE, -1, -1, 0, },
1558 };
1559
1560 void bdrv_get_floppy_geometry_hint(BlockDriverState *bs, int *nb_heads,
1561 int *max_track, int *last_sect,
1562 FDriveType drive_in, FDriveType *drive)
1563 {
1564 const FDFormat *parse;
1565 uint64_t nb_sectors, size;
1566 int i, first_match, match;
1567
1568 bdrv_get_geometry_hint(bs, nb_heads, max_track, last_sect);
1569 if (*nb_heads != 0 && *max_track != 0 && *last_sect != 0) {
1570 /* User defined disk */
1571 } else {
1572 bdrv_get_geometry(bs, &nb_sectors);
1573 match = -1;
1574 first_match = -1;
1575 for (i = 0; ; i++) {
1576 parse = &fd_formats[i];
1577 if (parse->drive == FDRIVE_DRV_NONE) {
1578 break;
1579 }
1580 if (drive_in == parse->drive ||
1581 drive_in == FDRIVE_DRV_NONE) {
1582 size = (parse->max_head + 1) * parse->max_track *
1583 parse->last_sect;
1584 if (nb_sectors == size) {
1585 match = i;
1586 break;
1587 }
1588 if (first_match == -1) {
1589 first_match = i;
1590 }
1591 }
1592 }
1593 if (match == -1) {
1594 if (first_match == -1) {
1595 match = 1;
1596 } else {
1597 match = first_match;
1598 }
1599 parse = &fd_formats[match];
1600 }
1601 *nb_heads = parse->max_head + 1;
1602 *max_track = parse->max_track;
1603 *last_sect = parse->last_sect;
1604 *drive = parse->drive;
1605 }
1606 }
1607
1608 int bdrv_get_translation_hint(BlockDriverState *bs)
1609 {
1610 return bs->translation;
1611 }
1612
1613 void bdrv_set_on_error(BlockDriverState *bs, BlockErrorAction on_read_error,
1614 BlockErrorAction on_write_error)
1615 {
1616 bs->on_read_error = on_read_error;
1617 bs->on_write_error = on_write_error;
1618 }
1619
1620 BlockErrorAction bdrv_get_on_error(BlockDriverState *bs, int is_read)
1621 {
1622 return is_read ? bs->on_read_error : bs->on_write_error;
1623 }
1624
1625 int bdrv_is_read_only(BlockDriverState *bs)
1626 {
1627 return bs->read_only;
1628 }
1629
1630 int bdrv_is_sg(BlockDriverState *bs)
1631 {
1632 return bs->sg;
1633 }
1634
1635 int bdrv_enable_write_cache(BlockDriverState *bs)
1636 {
1637 return bs->enable_write_cache;
1638 }
1639
1640 int bdrv_is_encrypted(BlockDriverState *bs)
1641 {
1642 if (bs->backing_hd && bs->backing_hd->encrypted)
1643 return 1;
1644 return bs->encrypted;
1645 }
1646
1647 int bdrv_key_required(BlockDriverState *bs)
1648 {
1649 BlockDriverState *backing_hd = bs->backing_hd;
1650
1651 if (backing_hd && backing_hd->encrypted && !backing_hd->valid_key)
1652 return 1;
1653 return (bs->encrypted && !bs->valid_key);
1654 }
1655
1656 int bdrv_set_key(BlockDriverState *bs, const char *key)
1657 {
1658 int ret;
1659 if (bs->backing_hd && bs->backing_hd->encrypted) {
1660 ret = bdrv_set_key(bs->backing_hd, key);
1661 if (ret < 0)
1662 return ret;
1663 if (!bs->encrypted)
1664 return 0;
1665 }
1666 if (!bs->encrypted) {
1667 return -EINVAL;
1668 } else if (!bs->drv || !bs->drv->bdrv_set_key) {
1669 return -ENOMEDIUM;
1670 }
1671 ret = bs->drv->bdrv_set_key(bs, key);
1672 if (ret < 0) {
1673 bs->valid_key = 0;
1674 } else if (!bs->valid_key) {
1675 bs->valid_key = 1;
1676 /* call the change callback now, we skipped it on open */
1677 bdrv_dev_change_media_cb(bs, true);
1678 }
1679 return ret;
1680 }
1681
1682 void bdrv_get_format(BlockDriverState *bs, char *buf, int buf_size)
1683 {
1684 if (!bs->drv) {
1685 buf[0] = '\0';
1686 } else {
1687 pstrcpy(buf, buf_size, bs->drv->format_name);
1688 }
1689 }
1690
1691 void bdrv_iterate_format(void (*it)(void *opaque, const char *name),
1692 void *opaque)
1693 {
1694 BlockDriver *drv;
1695
1696 QLIST_FOREACH(drv, &bdrv_drivers, list) {
1697 it(opaque, drv->format_name);
1698 }
1699 }
1700
1701 BlockDriverState *bdrv_find(const char *name)
1702 {
1703 BlockDriverState *bs;
1704
1705 QTAILQ_FOREACH(bs, &bdrv_states, list) {
1706 if (!strcmp(name, bs->device_name)) {
1707 return bs;
1708 }
1709 }
1710 return NULL;
1711 }
1712
1713 BlockDriverState *bdrv_next(BlockDriverState *bs)
1714 {
1715 if (!bs) {
1716 return QTAILQ_FIRST(&bdrv_states);
1717 }
1718 return QTAILQ_NEXT(bs, list);
1719 }
1720
1721 void bdrv_iterate(void (*it)(void *opaque, BlockDriverState *bs), void *opaque)
1722 {
1723 BlockDriverState *bs;
1724
1725 QTAILQ_FOREACH(bs, &bdrv_states, list) {
1726 it(opaque, bs);
1727 }
1728 }
1729
1730 const char *bdrv_get_device_name(BlockDriverState *bs)
1731 {
1732 return bs->device_name;
1733 }
1734
1735 int bdrv_flush(BlockDriverState *bs)
1736 {
1737 if (bs->open_flags & BDRV_O_NO_FLUSH) {
1738 return 0;
1739 }
1740
1741 if (bs->drv && bdrv_has_async_flush(bs->drv) && qemu_in_coroutine()) {
1742 return bdrv_co_flush_em(bs);
1743 }
1744
1745 if (bs->drv && bs->drv->bdrv_flush) {
1746 return bs->drv->bdrv_flush(bs);
1747 }
1748
1749 /*
1750 * Some block drivers always operate in either writethrough or unsafe mode
1751 * and don't support bdrv_flush therefore. Usually qemu doesn't know how
1752 * the server works (because the behaviour is hardcoded or depends on
1753 * server-side configuration), so we can't ensure that everything is safe
1754 * on disk. Returning an error doesn't work because that would break guests
1755 * even if the server operates in writethrough mode.
1756 *
1757 * Let's hope the user knows what he's doing.
1758 */
1759 return 0;
1760 }
1761
1762 void bdrv_flush_all(void)
1763 {
1764 BlockDriverState *bs;
1765
1766 QTAILQ_FOREACH(bs, &bdrv_states, list) {
1767 if (!bdrv_is_read_only(bs) && bdrv_is_inserted(bs)) {
1768 bdrv_flush(bs);
1769 }
1770 }
1771 }
1772
1773 int bdrv_has_zero_init(BlockDriverState *bs)
1774 {
1775 assert(bs->drv);
1776
1777 if (bs->drv->bdrv_has_zero_init) {
1778 return bs->drv->bdrv_has_zero_init(bs);
1779 }
1780
1781 return 1;
1782 }
1783
1784 int bdrv_discard(BlockDriverState *bs, int64_t sector_num, int nb_sectors)
1785 {
1786 if (!bs->drv) {
1787 return -ENOMEDIUM;
1788 }
1789 if (!bs->drv->bdrv_discard) {
1790 return 0;
1791 }
1792 return bs->drv->bdrv_discard(bs, sector_num, nb_sectors);
1793 }
1794
1795 /*
1796 * Returns true iff the specified sector is present in the disk image. Drivers
1797 * not implementing the functionality are assumed to not support backing files,
1798 * hence all their sectors are reported as allocated.
1799 *
1800 * 'pnum' is set to the number of sectors (including and immediately following
1801 * the specified sector) that are known to be in the same
1802 * allocated/unallocated state.
1803 *
1804 * 'nb_sectors' is the max value 'pnum' should be set to.
1805 */
1806 int bdrv_is_allocated(BlockDriverState *bs, int64_t sector_num, int nb_sectors,
1807 int *pnum)
1808 {
1809 int64_t n;
1810 if (!bs->drv->bdrv_is_allocated) {
1811 if (sector_num >= bs->total_sectors) {
1812 *pnum = 0;
1813 return 0;
1814 }
1815 n = bs->total_sectors - sector_num;
1816 *pnum = (n < nb_sectors) ? (n) : (nb_sectors);
1817 return 1;
1818 }
1819 return bs->drv->bdrv_is_allocated(bs, sector_num, nb_sectors, pnum);
1820 }
1821
1822 void bdrv_mon_event(const BlockDriverState *bdrv,
1823 BlockMonEventAction action, int is_read)
1824 {
1825 QObject *data;
1826 const char *action_str;
1827
1828 switch (action) {
1829 case BDRV_ACTION_REPORT:
1830 action_str = "report";
1831 break;
1832 case BDRV_ACTION_IGNORE:
1833 action_str = "ignore";
1834 break;
1835 case BDRV_ACTION_STOP:
1836 action_str = "stop";
1837 break;
1838 default:
1839 abort();
1840 }
1841
1842 data = qobject_from_jsonf("{ 'device': %s, 'action': %s, 'operation': %s }",
1843 bdrv->device_name,
1844 action_str,
1845 is_read ? "read" : "write");
1846 monitor_protocol_event(QEVENT_BLOCK_IO_ERROR, data);
1847
1848 qobject_decref(data);
1849 }
1850
1851 static void bdrv_print_dict(QObject *obj, void *opaque)
1852 {
1853 QDict *bs_dict;
1854 Monitor *mon = opaque;
1855
1856 bs_dict = qobject_to_qdict(obj);
1857
1858 monitor_printf(mon, "%s: removable=%d",
1859 qdict_get_str(bs_dict, "device"),
1860 qdict_get_bool(bs_dict, "removable"));
1861
1862 if (qdict_get_bool(bs_dict, "removable")) {
1863 monitor_printf(mon, " locked=%d", qdict_get_bool(bs_dict, "locked"));
1864 monitor_printf(mon, " tray-open=%d",
1865 qdict_get_bool(bs_dict, "tray-open"));
1866 }
1867 if (qdict_haskey(bs_dict, "inserted")) {
1868 QDict *qdict = qobject_to_qdict(qdict_get(bs_dict, "inserted"));
1869
1870 monitor_printf(mon, " file=");
1871 monitor_print_filename(mon, qdict_get_str(qdict, "file"));
1872 if (qdict_haskey(qdict, "backing_file")) {
1873 monitor_printf(mon, " backing_file=");
1874 monitor_print_filename(mon, qdict_get_str(qdict, "backing_file"));
1875 }
1876 monitor_printf(mon, " ro=%d drv=%s encrypted=%d",
1877 qdict_get_bool(qdict, "ro"),
1878 qdict_get_str(qdict, "drv"),
1879 qdict_get_bool(qdict, "encrypted"));
1880 } else {
1881 monitor_printf(mon, " [not inserted]");
1882 }
1883
1884 monitor_printf(mon, "\n");
1885 }
1886
1887 void bdrv_info_print(Monitor *mon, const QObject *data)
1888 {
1889 qlist_iter(qobject_to_qlist(data), bdrv_print_dict, mon);
1890 }
1891
1892 void bdrv_info(Monitor *mon, QObject **ret_data)
1893 {
1894 QList *bs_list;
1895 BlockDriverState *bs;
1896
1897 bs_list = qlist_new();
1898
1899 QTAILQ_FOREACH(bs, &bdrv_states, list) {
1900 QObject *bs_obj;
1901 QDict *bs_dict;
1902
1903 bs_obj = qobject_from_jsonf("{ 'device': %s, 'type': 'unknown', "
1904 "'removable': %i, 'locked': %i }",
1905 bs->device_name,
1906 bdrv_dev_has_removable_media(bs),
1907 bdrv_dev_is_medium_locked(bs));
1908 bs_dict = qobject_to_qdict(bs_obj);
1909
1910 if (bdrv_dev_has_removable_media(bs)) {
1911 qdict_put(bs_dict, "tray-open",
1912 qbool_from_int(bdrv_dev_is_tray_open(bs)));
1913 }
1914 if (bs->drv) {
1915 QObject *obj;
1916
1917 obj = qobject_from_jsonf("{ 'file': %s, 'ro': %i, 'drv': %s, "
1918 "'encrypted': %i }",
1919 bs->filename, bs->read_only,
1920 bs->drv->format_name,
1921 bdrv_is_encrypted(bs));
1922 if (bs->backing_file[0] != '\0') {
1923 QDict *qdict = qobject_to_qdict(obj);
1924 qdict_put(qdict, "backing_file",
1925 qstring_from_str(bs->backing_file));
1926 }
1927
1928 qdict_put_obj(bs_dict, "inserted", obj);
1929 }
1930 qlist_append_obj(bs_list, bs_obj);
1931 }
1932
1933 *ret_data = QOBJECT(bs_list);
1934 }
1935
1936 static void bdrv_stats_iter(QObject *data, void *opaque)
1937 {
1938 QDict *qdict;
1939 Monitor *mon = opaque;
1940
1941 qdict = qobject_to_qdict(data);
1942 monitor_printf(mon, "%s:", qdict_get_str(qdict, "device"));
1943
1944 qdict = qobject_to_qdict(qdict_get(qdict, "stats"));
1945 monitor_printf(mon, " rd_bytes=%" PRId64
1946 " wr_bytes=%" PRId64
1947 " rd_operations=%" PRId64
1948 " wr_operations=%" PRId64
1949 " flush_operations=%" PRId64
1950 " wr_total_time_ns=%" PRId64
1951 " rd_total_time_ns=%" PRId64
1952 " flush_total_time_ns=%" PRId64
1953 "\n",
1954 qdict_get_int(qdict, "rd_bytes"),
1955 qdict_get_int(qdict, "wr_bytes"),
1956 qdict_get_int(qdict, "rd_operations"),
1957 qdict_get_int(qdict, "wr_operations"),
1958 qdict_get_int(qdict, "flush_operations"),
1959 qdict_get_int(qdict, "wr_total_time_ns"),
1960 qdict_get_int(qdict, "rd_total_time_ns"),
1961 qdict_get_int(qdict, "flush_total_time_ns"));
1962 }
1963
1964 void bdrv_stats_print(Monitor *mon, const QObject *data)
1965 {
1966 qlist_iter(qobject_to_qlist(data), bdrv_stats_iter, mon);
1967 }
1968
1969 static QObject* bdrv_info_stats_bs(BlockDriverState *bs)
1970 {
1971 QObject *res;
1972 QDict *dict;
1973
1974 res = qobject_from_jsonf("{ 'stats': {"
1975 "'rd_bytes': %" PRId64 ","
1976 "'wr_bytes': %" PRId64 ","
1977 "'rd_operations': %" PRId64 ","
1978 "'wr_operations': %" PRId64 ","
1979 "'wr_highest_offset': %" PRId64 ","
1980 "'flush_operations': %" PRId64 ","
1981 "'wr_total_time_ns': %" PRId64 ","
1982 "'rd_total_time_ns': %" PRId64 ","
1983 "'flush_total_time_ns': %" PRId64
1984 "} }",
1985 bs->nr_bytes[BDRV_ACCT_READ],
1986 bs->nr_bytes[BDRV_ACCT_WRITE],
1987 bs->nr_ops[BDRV_ACCT_READ],
1988 bs->nr_ops[BDRV_ACCT_WRITE],
1989 bs->wr_highest_sector *
1990 (uint64_t)BDRV_SECTOR_SIZE,
1991 bs->nr_ops[BDRV_ACCT_FLUSH],
1992 bs->total_time_ns[BDRV_ACCT_WRITE],
1993 bs->total_time_ns[BDRV_ACCT_READ],
1994 bs->total_time_ns[BDRV_ACCT_FLUSH]);
1995 dict = qobject_to_qdict(res);
1996
1997 if (*bs->device_name) {
1998 qdict_put(dict, "device", qstring_from_str(bs->device_name));
1999 }
2000
2001 if (bs->file) {
2002 QObject *parent = bdrv_info_stats_bs(bs->file);
2003 qdict_put_obj(dict, "parent", parent);
2004 }
2005
2006 return res;
2007 }
2008
2009 void bdrv_info_stats(Monitor *mon, QObject **ret_data)
2010 {
2011 QObject *obj;
2012 QList *devices;
2013 BlockDriverState *bs;
2014
2015 devices = qlist_new();
2016
2017 QTAILQ_FOREACH(bs, &bdrv_states, list) {
2018 obj = bdrv_info_stats_bs(bs);
2019 qlist_append_obj(devices, obj);
2020 }
2021
2022 *ret_data = QOBJECT(devices);
2023 }
2024
2025 const char *bdrv_get_encrypted_filename(BlockDriverState *bs)
2026 {
2027 if (bs->backing_hd && bs->backing_hd->encrypted)
2028 return bs->backing_file;
2029 else if (bs->encrypted)
2030 return bs->filename;
2031 else
2032 return NULL;
2033 }
2034
2035 void bdrv_get_backing_filename(BlockDriverState *bs,
2036 char *filename, int filename_size)
2037 {
2038 if (!bs->backing_file) {
2039 pstrcpy(filename, filename_size, "");
2040 } else {
2041 pstrcpy(filename, filename_size, bs->backing_file);
2042 }
2043 }
2044
2045 int bdrv_write_compressed(BlockDriverState *bs, int64_t sector_num,
2046 const uint8_t *buf, int nb_sectors)
2047 {
2048 BlockDriver *drv = bs->drv;
2049 if (!drv)
2050 return -ENOMEDIUM;
2051 if (!drv->bdrv_write_compressed)
2052 return -ENOTSUP;
2053 if (bdrv_check_request(bs, sector_num, nb_sectors))
2054 return -EIO;
2055
2056 if (bs->dirty_bitmap) {
2057 set_dirty_bitmap(bs, sector_num, nb_sectors, 1);
2058 }
2059
2060 return drv->bdrv_write_compressed(bs, sector_num, buf, nb_sectors);
2061 }
2062
2063 int bdrv_get_info(BlockDriverState *bs, BlockDriverInfo *bdi)
2064 {
2065 BlockDriver *drv = bs->drv;
2066 if (!drv)
2067 return -ENOMEDIUM;
2068 if (!drv->bdrv_get_info)
2069 return -ENOTSUP;
2070 memset(bdi, 0, sizeof(*bdi));
2071 return drv->bdrv_get_info(bs, bdi);
2072 }
2073
2074 int bdrv_save_vmstate(BlockDriverState *bs, const uint8_t *buf,
2075 int64_t pos, int size)
2076 {
2077 BlockDriver *drv = bs->drv;
2078 if (!drv)
2079 return -ENOMEDIUM;
2080 if (drv->bdrv_save_vmstate)
2081 return drv->bdrv_save_vmstate(bs, buf, pos, size);
2082 if (bs->file)
2083 return bdrv_save_vmstate(bs->file, buf, pos, size);
2084 return -ENOTSUP;
2085 }
2086
2087 int bdrv_load_vmstate(BlockDriverState *bs, uint8_t *buf,
2088 int64_t pos, int size)
2089 {
2090 BlockDriver *drv = bs->drv;
2091 if (!drv)
2092 return -ENOMEDIUM;
2093 if (drv->bdrv_load_vmstate)
2094 return drv->bdrv_load_vmstate(bs, buf, pos, size);
2095 if (bs->file)
2096 return bdrv_load_vmstate(bs->file, buf, pos, size);
2097 return -ENOTSUP;
2098 }
2099
2100 void bdrv_debug_event(BlockDriverState *bs, BlkDebugEvent event)
2101 {
2102 BlockDriver *drv = bs->drv;
2103
2104 if (!drv || !drv->bdrv_debug_event) {
2105 return;
2106 }
2107
2108 return drv->bdrv_debug_event(bs, event);
2109
2110 }
2111
2112 /**************************************************************/
2113 /* handling of snapshots */
2114
2115 int bdrv_can_snapshot(BlockDriverState *bs)
2116 {
2117 BlockDriver *drv = bs->drv;
2118 if (!drv || !bdrv_is_inserted(bs) || bdrv_is_read_only(bs)) {
2119 return 0;
2120 }
2121
2122 if (!drv->bdrv_snapshot_create) {
2123 if (bs->file != NULL) {
2124 return bdrv_can_snapshot(bs->file);
2125 }
2126 return 0;
2127 }
2128
2129 return 1;
2130 }
2131
2132 int bdrv_is_snapshot(BlockDriverState *bs)
2133 {
2134 return !!(bs->open_flags & BDRV_O_SNAPSHOT);
2135 }
2136
2137 BlockDriverState *bdrv_snapshots(void)
2138 {
2139 BlockDriverState *bs;
2140
2141 if (bs_snapshots) {
2142 return bs_snapshots;
2143 }
2144
2145 bs = NULL;
2146 while ((bs = bdrv_next(bs))) {
2147 if (bdrv_can_snapshot(bs)) {
2148 bs_snapshots = bs;
2149 return bs;
2150 }
2151 }
2152 return NULL;
2153 }
2154
2155 int bdrv_snapshot_create(BlockDriverState *bs,
2156 QEMUSnapshotInfo *sn_info)
2157 {
2158 BlockDriver *drv = bs->drv;
2159 if (!drv)
2160 return -ENOMEDIUM;
2161 if (drv->bdrv_snapshot_create)
2162 return drv->bdrv_snapshot_create(bs, sn_info);
2163 if (bs->file)
2164 return bdrv_snapshot_create(bs->file, sn_info);
2165 return -ENOTSUP;
2166 }
2167
2168 int bdrv_snapshot_goto(BlockDriverState *bs,
2169 const char *snapshot_id)
2170 {
2171 BlockDriver *drv = bs->drv;
2172 int ret, open_ret;
2173
2174 if (!drv)
2175 return -ENOMEDIUM;
2176 if (drv->bdrv_snapshot_goto)
2177 return drv->bdrv_snapshot_goto(bs, snapshot_id);
2178
2179 if (bs->file) {
2180 drv->bdrv_close(bs);
2181 ret = bdrv_snapshot_goto(bs->file, snapshot_id);
2182 open_ret = drv->bdrv_open(bs, bs->open_flags);
2183 if (open_ret < 0) {
2184 bdrv_delete(bs->file);
2185 bs->drv = NULL;
2186 return open_ret;
2187 }
2188 return ret;
2189 }
2190
2191 return -ENOTSUP;
2192 }
2193
2194 int bdrv_snapshot_delete(BlockDriverState *bs, const char *snapshot_id)
2195 {
2196 BlockDriver *drv = bs->drv;
2197 if (!drv)
2198 return -ENOMEDIUM;
2199 if (drv->bdrv_snapshot_delete)
2200 return drv->bdrv_snapshot_delete(bs, snapshot_id);
2201 if (bs->file)
2202 return bdrv_snapshot_delete(bs->file, snapshot_id);
2203 return -ENOTSUP;
2204 }
2205
2206 int bdrv_snapshot_list(BlockDriverState *bs,
2207 QEMUSnapshotInfo **psn_info)
2208 {
2209 BlockDriver *drv = bs->drv;
2210 if (!drv)
2211 return -ENOMEDIUM;
2212 if (drv->bdrv_snapshot_list)
2213 return drv->bdrv_snapshot_list(bs, psn_info);
2214 if (bs->file)
2215 return bdrv_snapshot_list(bs->file, psn_info);
2216 return -ENOTSUP;
2217 }
2218
2219 int bdrv_snapshot_load_tmp(BlockDriverState *bs,
2220 const char *snapshot_name)
2221 {
2222 BlockDriver *drv = bs->drv;
2223 if (!drv) {
2224 return -ENOMEDIUM;
2225 }
2226 if (!bs->read_only) {
2227 return -EINVAL;
2228 }
2229 if (drv->bdrv_snapshot_load_tmp) {
2230 return drv->bdrv_snapshot_load_tmp(bs, snapshot_name);
2231 }
2232 return -ENOTSUP;
2233 }
2234
2235 #define NB_SUFFIXES 4
2236
2237 char *get_human_readable_size(char *buf, int buf_size, int64_t size)
2238 {
2239 static const char suffixes[NB_SUFFIXES] = "KMGT";
2240 int64_t base;
2241 int i;
2242
2243 if (size <= 999) {
2244 snprintf(buf, buf_size, "%" PRId64, size);
2245 } else {
2246 base = 1024;
2247 for(i = 0; i < NB_SUFFIXES; i++) {
2248 if (size < (10 * base)) {
2249 snprintf(buf, buf_size, "%0.1f%c",
2250 (double)size / base,
2251 suffixes[i]);
2252 break;
2253 } else if (size < (1000 * base) || i == (NB_SUFFIXES - 1)) {
2254 snprintf(buf, buf_size, "%" PRId64 "%c",
2255 ((size + (base >> 1)) / base),
2256 suffixes[i]);
2257 break;
2258 }
2259 base = base * 1024;
2260 }
2261 }
2262 return buf;
2263 }
2264
2265 char *bdrv_snapshot_dump(char *buf, int buf_size, QEMUSnapshotInfo *sn)
2266 {
2267 char buf1[128], date_buf[128], clock_buf[128];
2268 #ifdef _WIN32
2269 struct tm *ptm;
2270 #else
2271 struct tm tm;
2272 #endif
2273 time_t ti;
2274 int64_t secs;
2275
2276 if (!sn) {
2277 snprintf(buf, buf_size,
2278 "%-10s%-20s%7s%20s%15s",
2279 "ID", "TAG", "VM SIZE", "DATE", "VM CLOCK");
2280 } else {
2281 ti = sn->date_sec;
2282 #ifdef _WIN32
2283 ptm = localtime(&ti);
2284 strftime(date_buf, sizeof(date_buf),
2285 "%Y-%m-%d %H:%M:%S", ptm);
2286 #else
2287 localtime_r(&ti, &tm);
2288 strftime(date_buf, sizeof(date_buf),
2289 "%Y-%m-%d %H:%M:%S", &tm);
2290 #endif
2291 secs = sn->vm_clock_nsec / 1000000000;
2292 snprintf(clock_buf, sizeof(clock_buf),
2293 "%02d:%02d:%02d.%03d",
2294 (int)(secs / 3600),
2295 (int)((secs / 60) % 60),
2296 (int)(secs % 60),
2297 (int)((sn->vm_clock_nsec / 1000000) % 1000));
2298 snprintf(buf, buf_size,
2299 "%-10s%-20s%7s%20s%15s",
2300 sn->id_str, sn->name,
2301 get_human_readable_size(buf1, sizeof(buf1), sn->vm_state_size),
2302 date_buf,
2303 clock_buf);
2304 }
2305 return buf;
2306 }
2307
2308 /**************************************************************/
2309 /* async I/Os */
2310
2311 BlockDriverAIOCB *bdrv_aio_readv(BlockDriverState *bs, int64_t sector_num,
2312 QEMUIOVector *qiov, int nb_sectors,
2313 BlockDriverCompletionFunc *cb, void *opaque)
2314 {
2315 BlockDriver *drv = bs->drv;
2316
2317 trace_bdrv_aio_readv(bs, sector_num, nb_sectors, opaque);
2318
2319 if (!drv)
2320 return NULL;
2321 if (bdrv_check_request(bs, sector_num, nb_sectors))
2322 return NULL;
2323
2324 return drv->bdrv_aio_readv(bs, sector_num, qiov, nb_sectors,
2325 cb, opaque);
2326 }
2327
2328 typedef struct BlockCompleteData {
2329 BlockDriverCompletionFunc *cb;
2330 void *opaque;
2331 BlockDriverState *bs;
2332 int64_t sector_num;
2333 int nb_sectors;
2334 } BlockCompleteData;
2335
2336 static void block_complete_cb(void *opaque, int ret)
2337 {
2338 BlockCompleteData *b = opaque;
2339
2340 if (b->bs->dirty_bitmap) {
2341 set_dirty_bitmap(b->bs, b->sector_num, b->nb_sectors, 1);
2342 }
2343 b->cb(b->opaque, ret);
2344 g_free(b);
2345 }
2346
2347 static BlockCompleteData *blk_dirty_cb_alloc(BlockDriverState *bs,
2348 int64_t sector_num,
2349 int nb_sectors,
2350 BlockDriverCompletionFunc *cb,
2351 void *opaque)
2352 {
2353 BlockCompleteData *blkdata = g_malloc0(sizeof(BlockCompleteData));
2354
2355 blkdata->bs = bs;
2356 blkdata->cb = cb;
2357 blkdata->opaque = opaque;
2358 blkdata->sector_num = sector_num;
2359 blkdata->nb_sectors = nb_sectors;
2360
2361 return blkdata;
2362 }
2363
2364 BlockDriverAIOCB *bdrv_aio_writev(BlockDriverState *bs, int64_t sector_num,
2365 QEMUIOVector *qiov, int nb_sectors,
2366 BlockDriverCompletionFunc *cb, void *opaque)
2367 {
2368 BlockDriver *drv = bs->drv;
2369 BlockDriverAIOCB *ret;
2370 BlockCompleteData *blk_cb_data;
2371
2372 trace_bdrv_aio_writev(bs, sector_num, nb_sectors, opaque);
2373
2374 if (!drv)
2375 return NULL;
2376 if (bs->read_only)
2377 return NULL;
2378 if (bdrv_check_request(bs, sector_num, nb_sectors))
2379 return NULL;
2380
2381 if (bs->dirty_bitmap) {
2382 blk_cb_data = blk_dirty_cb_alloc(bs, sector_num, nb_sectors, cb,
2383 opaque);
2384 cb = &block_complete_cb;
2385 opaque = blk_cb_data;
2386 }
2387
2388 ret = drv->bdrv_aio_writev(bs, sector_num, qiov, nb_sectors,
2389 cb, opaque);
2390
2391 if (ret) {
2392 if (bs->wr_highest_sector < sector_num + nb_sectors - 1) {
2393 bs->wr_highest_sector = sector_num + nb_sectors - 1;
2394 }
2395 }
2396
2397 return ret;
2398 }
2399
2400
2401 typedef struct MultiwriteCB {
2402 int error;
2403 int num_requests;
2404 int num_callbacks;
2405 struct {
2406 BlockDriverCompletionFunc *cb;
2407 void *opaque;
2408 QEMUIOVector *free_qiov;
2409 void *free_buf;
2410 } callbacks[];
2411 } MultiwriteCB;
2412
2413 static void multiwrite_user_cb(MultiwriteCB *mcb)
2414 {
2415 int i;
2416
2417 for (i = 0; i < mcb->num_callbacks; i++) {
2418 mcb->callbacks[i].cb(mcb->callbacks[i].opaque, mcb->error);
2419 if (mcb->callbacks[i].free_qiov) {
2420 qemu_iovec_destroy(mcb->callbacks[i].free_qiov);
2421 }
2422 g_free(mcb->callbacks[i].free_qiov);
2423 qemu_vfree(mcb->callbacks[i].free_buf);
2424 }
2425 }
2426
2427 static void multiwrite_cb(void *opaque, int ret)
2428 {
2429 MultiwriteCB *mcb = opaque;
2430
2431 trace_multiwrite_cb(mcb, ret);
2432
2433 if (ret < 0 && !mcb->error) {
2434 mcb->error = ret;
2435 }
2436
2437 mcb->num_requests--;
2438 if (mcb->num_requests == 0) {
2439 multiwrite_user_cb(mcb);
2440 g_free(mcb);
2441 }
2442 }
2443
2444 static int multiwrite_req_compare(const void *a, const void *b)
2445 {
2446 const BlockRequest *req1 = a, *req2 = b;
2447
2448 /*
2449 * Note that we can't simply subtract req2->sector from req1->sector
2450 * here as that could overflow the return value.
2451 */
2452 if (req1->sector > req2->sector) {
2453 return 1;
2454 } else if (req1->sector < req2->sector) {
2455 return -1;
2456 } else {
2457 return 0;
2458 }
2459 }
2460
2461 /*
2462 * Takes a bunch of requests and tries to merge them. Returns the number of
2463 * requests that remain after merging.
2464 */
2465 static int multiwrite_merge(BlockDriverState *bs, BlockRequest *reqs,
2466 int num_reqs, MultiwriteCB *mcb)
2467 {
2468 int i, outidx;
2469
2470 // Sort requests by start sector
2471 qsort(reqs, num_reqs, sizeof(*reqs), &multiwrite_req_compare);
2472
2473 // Check if adjacent requests touch the same clusters. If so, combine them,
2474 // filling up gaps with zero sectors.
2475 outidx = 0;
2476 for (i = 1; i < num_reqs; i++) {
2477 int merge = 0;
2478 int64_t oldreq_last = reqs[outidx].sector + reqs[outidx].nb_sectors;
2479
2480 // This handles the cases that are valid for all block drivers, namely
2481 // exactly sequential writes and overlapping writes.
2482 if (reqs[i].sector <= oldreq_last) {
2483 merge = 1;
2484 }
2485
2486 // The block driver may decide that it makes sense to combine requests
2487 // even if there is a gap of some sectors between them. In this case,
2488 // the gap is filled with zeros (therefore only applicable for yet
2489 // unused space in format like qcow2).
2490 if (!merge && bs->drv->bdrv_merge_requests) {
2491 merge = bs->drv->bdrv_merge_requests(bs, &reqs[outidx], &reqs[i]);
2492 }
2493
2494 if (reqs[outidx].qiov->niov + reqs[i].qiov->niov + 1 > IOV_MAX) {
2495 merge = 0;
2496 }
2497
2498 if (merge) {
2499 size_t size;
2500 QEMUIOVector *qiov = g_malloc0(sizeof(*qiov));
2501 qemu_iovec_init(qiov,
2502 reqs[outidx].qiov->niov + reqs[i].qiov->niov + 1);
2503
2504 // Add the first request to the merged one. If the requests are
2505 // overlapping, drop the last sectors of the first request.
2506 size = (reqs[i].sector - reqs[outidx].sector) << 9;
2507 qemu_iovec_concat(qiov, reqs[outidx].qiov, size);
2508
2509 // We might need to add some zeros between the two requests
2510 if (reqs[i].sector > oldreq_last) {
2511 size_t zero_bytes = (reqs[i].sector - oldreq_last) << 9;
2512 uint8_t *buf = qemu_blockalign(bs, zero_bytes);
2513 memset(buf, 0, zero_bytes);
2514 qemu_iovec_add(qiov, buf, zero_bytes);
2515 mcb->callbacks[i].free_buf = buf;
2516 }
2517
2518 // Add the second request
2519 qemu_iovec_concat(qiov, reqs[i].qiov, reqs[i].qiov->size);
2520
2521 reqs[outidx].nb_sectors = qiov->size >> 9;
2522 reqs[outidx].qiov = qiov;
2523
2524 mcb->callbacks[i].free_qiov = reqs[outidx].qiov;
2525 } else {
2526 outidx++;
2527 reqs[outidx].sector = reqs[i].sector;
2528 reqs[outidx].nb_sectors = reqs[i].nb_sectors;
2529 reqs[outidx].qiov = reqs[i].qiov;
2530 }
2531 }
2532
2533 return outidx + 1;
2534 }
2535
2536 /*
2537 * Submit multiple AIO write requests at once.
2538 *
2539 * On success, the function returns 0 and all requests in the reqs array have
2540 * been submitted. In error case this function returns -1, and any of the
2541 * requests may or may not be submitted yet. In particular, this means that the
2542 * callback will be called for some of the requests, for others it won't. The
2543 * caller must check the error field of the BlockRequest to wait for the right
2544 * callbacks (if error != 0, no callback will be called).
2545 *
2546 * The implementation may modify the contents of the reqs array, e.g. to merge
2547 * requests. However, the fields opaque and error are left unmodified as they
2548 * are used to signal failure for a single request to the caller.
2549 */
2550 int bdrv_aio_multiwrite(BlockDriverState *bs, BlockRequest *reqs, int num_reqs)
2551 {
2552 BlockDriverAIOCB *acb;
2553 MultiwriteCB *mcb;
2554 int i;
2555
2556 /* don't submit writes if we don't have a medium */
2557 if (bs->drv == NULL) {
2558 for (i = 0; i < num_reqs; i++) {
2559 reqs[i].error = -ENOMEDIUM;
2560 }
2561 return -1;
2562 }
2563
2564 if (num_reqs == 0) {
2565 return 0;
2566 }
2567
2568 // Create MultiwriteCB structure
2569 mcb = g_malloc0(sizeof(*mcb) + num_reqs * sizeof(*mcb->callbacks));
2570 mcb->num_requests = 0;
2571 mcb->num_callbacks = num_reqs;
2572
2573 for (i = 0; i < num_reqs; i++) {
2574 mcb->callbacks[i].cb = reqs[i].cb;
2575 mcb->callbacks[i].opaque = reqs[i].opaque;
2576 }
2577
2578 // Check for mergable requests
2579 num_reqs = multiwrite_merge(bs, reqs, num_reqs, mcb);
2580
2581 trace_bdrv_aio_multiwrite(mcb, mcb->num_callbacks, num_reqs);
2582
2583 /*
2584 * Run the aio requests. As soon as one request can't be submitted
2585 * successfully, fail all requests that are not yet submitted (we must
2586 * return failure for all requests anyway)
2587 *
2588 * num_requests cannot be set to the right value immediately: If
2589 * bdrv_aio_writev fails for some request, num_requests would be too high
2590 * and therefore multiwrite_cb() would never recognize the multiwrite
2591 * request as completed. We also cannot use the loop variable i to set it
2592 * when the first request fails because the callback may already have been
2593 * called for previously submitted requests. Thus, num_requests must be
2594 * incremented for each request that is submitted.
2595 *
2596 * The problem that callbacks may be called early also means that we need
2597 * to take care that num_requests doesn't become 0 before all requests are
2598 * submitted - multiwrite_cb() would consider the multiwrite request
2599 * completed. A dummy request that is "completed" by a manual call to
2600 * multiwrite_cb() takes care of this.
2601 */
2602 mcb->num_requests = 1;
2603
2604 // Run the aio requests
2605 for (i = 0; i < num_reqs; i++) {
2606 mcb->num_requests++;
2607 acb = bdrv_aio_writev(bs, reqs[i].sector, reqs[i].qiov,
2608 reqs[i].nb_sectors, multiwrite_cb, mcb);
2609
2610 if (acb == NULL) {
2611 // We can only fail the whole thing if no request has been
2612 // submitted yet. Otherwise we'll wait for the submitted AIOs to
2613 // complete and report the error in the callback.
2614 if (i == 0) {
2615 trace_bdrv_aio_multiwrite_earlyfail(mcb);
2616 goto fail;
2617 } else {
2618 trace_bdrv_aio_multiwrite_latefail(mcb, i);
2619 multiwrite_cb(mcb, -EIO);
2620 break;
2621 }
2622 }
2623 }
2624
2625 /* Complete the dummy request */
2626 multiwrite_cb(mcb, 0);
2627
2628 return 0;
2629
2630 fail:
2631 for (i = 0; i < mcb->num_callbacks; i++) {
2632 reqs[i].error = -EIO;
2633 }
2634 g_free(mcb);
2635 return -1;
2636 }
2637
2638 BlockDriverAIOCB *bdrv_aio_flush(BlockDriverState *bs,
2639 BlockDriverCompletionFunc *cb, void *opaque)
2640 {
2641 BlockDriver *drv = bs->drv;
2642
2643 trace_bdrv_aio_flush(bs, opaque);
2644
2645 if (bs->open_flags & BDRV_O_NO_FLUSH) {
2646 return bdrv_aio_noop_em(bs, cb, opaque);
2647 }
2648
2649 if (!drv)
2650 return NULL;
2651 return drv->bdrv_aio_flush(bs, cb, opaque);
2652 }
2653
2654 void bdrv_aio_cancel(BlockDriverAIOCB *acb)
2655 {
2656 acb->pool->cancel(acb);
2657 }
2658
2659
2660 /**************************************************************/
2661 /* async block device emulation */
2662
2663 typedef struct BlockDriverAIOCBSync {
2664 BlockDriverAIOCB common;
2665 QEMUBH *bh;
2666 int ret;
2667 /* vector translation state */
2668 QEMUIOVector *qiov;
2669 uint8_t *bounce;
2670 int is_write;
2671 } BlockDriverAIOCBSync;
2672
2673 static void bdrv_aio_cancel_em(BlockDriverAIOCB *blockacb)
2674 {
2675 BlockDriverAIOCBSync *acb =
2676 container_of(blockacb, BlockDriverAIOCBSync, common);
2677 qemu_bh_delete(acb->bh);
2678 acb->bh = NULL;
2679 qemu_aio_release(acb);
2680 }
2681
2682 static AIOPool bdrv_em_aio_pool = {
2683 .aiocb_size = sizeof(BlockDriverAIOCBSync),
2684 .cancel = bdrv_aio_cancel_em,
2685 };
2686
2687 static void bdrv_aio_bh_cb(void *opaque)
2688 {
2689 BlockDriverAIOCBSync *acb = opaque;
2690
2691 if (!acb->is_write)
2692 qemu_iovec_from_buffer(acb->qiov, acb->bounce, acb->qiov->size);
2693 qemu_vfree(acb->bounce);
2694 acb->common.cb(acb->common.opaque, acb->ret);
2695 qemu_bh_delete(acb->bh);
2696 acb->bh = NULL;
2697 qemu_aio_release(acb);
2698 }
2699
2700 static BlockDriverAIOCB *bdrv_aio_rw_vector(BlockDriverState *bs,
2701 int64_t sector_num,
2702 QEMUIOVector *qiov,
2703 int nb_sectors,
2704 BlockDriverCompletionFunc *cb,
2705 void *opaque,
2706 int is_write)
2707
2708 {
2709 BlockDriverAIOCBSync *acb;
2710
2711 acb = qemu_aio_get(&bdrv_em_aio_pool, bs, cb, opaque);
2712 acb->is_write = is_write;
2713 acb->qiov = qiov;
2714 acb->bounce = qemu_blockalign(bs, qiov->size);
2715
2716 if (!acb->bh)
2717 acb->bh = qemu_bh_new(bdrv_aio_bh_cb, acb);
2718
2719 if (is_write) {
2720 qemu_iovec_to_buffer(acb->qiov, acb->bounce);
2721 acb->ret = bdrv_write(bs, sector_num, acb->bounce, nb_sectors);
2722 } else {
2723 acb->ret = bdrv_read(bs, sector_num, acb->bounce, nb_sectors);
2724 }
2725
2726 qemu_bh_schedule(acb->bh);
2727
2728 return &acb->common;
2729 }
2730
2731 static BlockDriverAIOCB *bdrv_aio_readv_em(BlockDriverState *bs,
2732 int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
2733 BlockDriverCompletionFunc *cb, void *opaque)
2734 {
2735 return bdrv_aio_rw_vector(bs, sector_num, qiov, nb_sectors, cb, opaque, 0);
2736 }
2737
2738 static BlockDriverAIOCB *bdrv_aio_writev_em(BlockDriverState *bs,
2739 int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
2740 BlockDriverCompletionFunc *cb, void *opaque)
2741 {
2742 return bdrv_aio_rw_vector(bs, sector_num, qiov, nb_sectors, cb, opaque, 1);
2743 }
2744
2745
2746 typedef struct BlockDriverAIOCBCoroutine {
2747 BlockDriverAIOCB common;
2748 BlockRequest req;
2749 bool is_write;
2750 QEMUBH* bh;
2751 } BlockDriverAIOCBCoroutine;
2752
2753 static void bdrv_aio_co_cancel_em(BlockDriverAIOCB *blockacb)
2754 {
2755 qemu_aio_flush();
2756 }
2757
2758 static AIOPool bdrv_em_co_aio_pool = {
2759 .aiocb_size = sizeof(BlockDriverAIOCBCoroutine),
2760 .cancel = bdrv_aio_co_cancel_em,
2761 };
2762
2763 static void bdrv_co_rw_bh(void *opaque)
2764 {
2765 BlockDriverAIOCBCoroutine *acb = opaque;
2766
2767 acb->common.cb(acb->common.opaque, acb->req.error);
2768 qemu_bh_delete(acb->bh);
2769 qemu_aio_release(acb);
2770 }
2771
2772 static void coroutine_fn bdrv_co_rw(void *opaque)
2773 {
2774 BlockDriverAIOCBCoroutine *acb = opaque;
2775 BlockDriverState *bs = acb->common.bs;
2776
2777 if (!acb->is_write) {
2778 acb->req.error = bs->drv->bdrv_co_readv(bs, acb->req.sector,
2779 acb->req.nb_sectors, acb->req.qiov);
2780 } else {
2781 acb->req.error = bs->drv->bdrv_co_writev(bs, acb->req.sector,
2782 acb->req.nb_sectors, acb->req.qiov);
2783 }
2784
2785 acb->bh = qemu_bh_new(bdrv_co_rw_bh, acb);
2786 qemu_bh_schedule(acb->bh);
2787 }
2788
2789 static BlockDriverAIOCB *bdrv_co_aio_rw_vector(BlockDriverState *bs,
2790 int64_t sector_num,
2791 QEMUIOVector *qiov,
2792 int nb_sectors,
2793 BlockDriverCompletionFunc *cb,
2794 void *opaque,
2795 bool is_write)
2796 {
2797 Coroutine *co;
2798 BlockDriverAIOCBCoroutine *acb;
2799
2800 acb = qemu_aio_get(&bdrv_em_co_aio_pool, bs, cb, opaque);
2801 acb->req.sector = sector_num;
2802 acb->req.nb_sectors = nb_sectors;
2803 acb->req.qiov = qiov;
2804 acb->is_write = is_write;
2805
2806 co = qemu_coroutine_create(bdrv_co_rw);
2807 qemu_coroutine_enter(co, acb);
2808
2809 return &acb->common;
2810 }
2811
2812 static BlockDriverAIOCB *bdrv_co_aio_readv_em(BlockDriverState *bs,
2813 int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
2814 BlockDriverCompletionFunc *cb, void *opaque)
2815 {
2816 return bdrv_co_aio_rw_vector(bs, sector_num, qiov, nb_sectors, cb, opaque,
2817 false);
2818 }
2819
2820 static BlockDriverAIOCB *bdrv_co_aio_writev_em(BlockDriverState *bs,
2821 int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
2822 BlockDriverCompletionFunc *cb, void *opaque)
2823 {
2824 return bdrv_co_aio_rw_vector(bs, sector_num, qiov, nb_sectors, cb, opaque,
2825 true);
2826 }
2827
2828 static BlockDriverAIOCB *bdrv_aio_flush_em(BlockDriverState *bs,
2829 BlockDriverCompletionFunc *cb, void *opaque)
2830 {
2831 BlockDriverAIOCBSync *acb;
2832
2833 acb = qemu_aio_get(&bdrv_em_aio_pool, bs, cb, opaque);
2834 acb->is_write = 1; /* don't bounce in the completion hadler */
2835 acb->qiov = NULL;
2836 acb->bounce = NULL;
2837 acb->ret = 0;
2838
2839 if (!acb->bh)
2840 acb->bh = qemu_bh_new(bdrv_aio_bh_cb, acb);
2841
2842 bdrv_flush(bs);
2843 qemu_bh_schedule(acb->bh);
2844 return &acb->common;
2845 }
2846
2847 static BlockDriverAIOCB *bdrv_aio_noop_em(BlockDriverState *bs,
2848 BlockDriverCompletionFunc *cb, void *opaque)
2849 {
2850 BlockDriverAIOCBSync *acb;
2851
2852 acb = qemu_aio_get(&bdrv_em_aio_pool, bs, cb, opaque);
2853 acb->is_write = 1; /* don't bounce in the completion handler */
2854 acb->qiov = NULL;
2855 acb->bounce = NULL;
2856 acb->ret = 0;
2857
2858 if (!acb->bh) {
2859 acb->bh = qemu_bh_new(bdrv_aio_bh_cb, acb);
2860 }
2861
2862 qemu_bh_schedule(acb->bh);
2863 return &acb->common;
2864 }
2865
2866 /**************************************************************/
2867 /* sync block device emulation */
2868
2869 static void bdrv_rw_em_cb(void *opaque, int ret)
2870 {
2871 *(int *)opaque = ret;
2872 }
2873
2874 #define NOT_DONE 0x7fffffff
2875
2876 static int bdrv_read_em(BlockDriverState *bs, int64_t sector_num,
2877 uint8_t *buf, int nb_sectors)
2878 {
2879 int async_ret;
2880 BlockDriverAIOCB *acb;
2881 struct iovec iov;
2882 QEMUIOVector qiov;
2883
2884 async_ret = NOT_DONE;
2885 iov.iov_base = (void *)buf;
2886 iov.iov_len = nb_sectors * BDRV_SECTOR_SIZE;
2887 qemu_iovec_init_external(&qiov, &iov, 1);
2888 acb = bdrv_aio_readv(bs, sector_num, &qiov, nb_sectors,
2889 bdrv_rw_em_cb, &async_ret);
2890 if (acb == NULL) {
2891 async_ret = -1;
2892 goto fail;
2893 }
2894
2895 while (async_ret == NOT_DONE) {
2896 qemu_aio_wait();
2897 }
2898
2899
2900 fail:
2901 return async_ret;
2902 }
2903
2904 static int bdrv_write_em(BlockDriverState *bs, int64_t sector_num,
2905 const uint8_t *buf, int nb_sectors)
2906 {
2907 int async_ret;
2908 BlockDriverAIOCB *acb;
2909 struct iovec iov;
2910 QEMUIOVector qiov;
2911
2912 async_ret = NOT_DONE;
2913 iov.iov_base = (void *)buf;
2914 iov.iov_len = nb_sectors * BDRV_SECTOR_SIZE;
2915 qemu_iovec_init_external(&qiov, &iov, 1);
2916 acb = bdrv_aio_writev(bs, sector_num, &qiov, nb_sectors,
2917 bdrv_rw_em_cb, &async_ret);
2918 if (acb == NULL) {
2919 async_ret = -1;
2920 goto fail;
2921 }
2922 while (async_ret == NOT_DONE) {
2923 qemu_aio_wait();
2924 }
2925
2926 fail:
2927 return async_ret;
2928 }
2929
2930 void bdrv_init(void)
2931 {
2932 module_call_init(MODULE_INIT_BLOCK);
2933 }
2934
2935 void bdrv_init_with_whitelist(void)
2936 {
2937 use_bdrv_whitelist = 1;
2938 bdrv_init();
2939 }
2940
2941 void *qemu_aio_get(AIOPool *pool, BlockDriverState *bs,
2942 BlockDriverCompletionFunc *cb, void *opaque)
2943 {
2944 BlockDriverAIOCB *acb;
2945
2946 if (pool->free_aiocb) {
2947 acb = pool->free_aiocb;
2948 pool->free_aiocb = acb->next;
2949 } else {
2950 acb = g_malloc0(pool->aiocb_size);
2951 acb->pool = pool;
2952 }
2953 acb->bs = bs;
2954 acb->cb = cb;
2955 acb->opaque = opaque;
2956 return acb;
2957 }
2958
2959 void qemu_aio_release(void *p)
2960 {
2961 BlockDriverAIOCB *acb = (BlockDriverAIOCB *)p;
2962 AIOPool *pool = acb->pool;
2963 acb->next = pool->free_aiocb;
2964 pool->free_aiocb = acb;
2965 }
2966
2967 /**************************************************************/
2968 /* Coroutine block device emulation */
2969
2970 typedef struct CoroutineIOCompletion {
2971 Coroutine *coroutine;
2972 int ret;
2973 } CoroutineIOCompletion;
2974
2975 static void bdrv_co_io_em_complete(void *opaque, int ret)
2976 {
2977 CoroutineIOCompletion *co = opaque;
2978
2979 co->ret = ret;
2980 qemu_coroutine_enter(co->coroutine, NULL);
2981 }
2982
2983 static int coroutine_fn bdrv_co_io_em(BlockDriverState *bs, int64_t sector_num,
2984 int nb_sectors, QEMUIOVector *iov,
2985 bool is_write)
2986 {
2987 CoroutineIOCompletion co = {
2988 .coroutine = qemu_coroutine_self(),
2989 };
2990 BlockDriverAIOCB *acb;
2991
2992 if (is_write) {
2993 acb = bdrv_aio_writev(bs, sector_num, iov, nb_sectors,
2994 bdrv_co_io_em_complete, &co);
2995 } else {
2996 acb = bdrv_aio_readv(bs, sector_num, iov, nb_sectors,
2997 bdrv_co_io_em_complete, &co);
2998 }
2999
3000 trace_bdrv_co_io(is_write, acb);
3001 if (!acb) {
3002 return -EIO;
3003 }
3004 qemu_coroutine_yield();
3005
3006 return co.ret;
3007 }
3008
3009 static int coroutine_fn bdrv_co_readv_em(BlockDriverState *bs,
3010 int64_t sector_num, int nb_sectors,
3011 QEMUIOVector *iov)
3012 {
3013 return bdrv_co_io_em(bs, sector_num, nb_sectors, iov, false);
3014 }
3015
3016 static int coroutine_fn bdrv_co_writev_em(BlockDriverState *bs,
3017 int64_t sector_num, int nb_sectors,
3018 QEMUIOVector *iov)
3019 {
3020 return bdrv_co_io_em(bs, sector_num, nb_sectors, iov, true);
3021 }
3022
3023 static int coroutine_fn bdrv_co_flush_em(BlockDriverState *bs)
3024 {
3025 CoroutineIOCompletion co = {
3026 .coroutine = qemu_coroutine_self(),
3027 };
3028 BlockDriverAIOCB *acb;
3029
3030 acb = bdrv_aio_flush(bs, bdrv_co_io_em_complete, &co);
3031 if (!acb) {
3032 return -EIO;
3033 }
3034 qemu_coroutine_yield();
3035 return co.ret;
3036 }
3037
3038 /**************************************************************/
3039 /* removable device support */
3040
3041 /**
3042 * Return TRUE if the media is present
3043 */
3044 int bdrv_is_inserted(BlockDriverState *bs)
3045 {
3046 BlockDriver *drv = bs->drv;
3047
3048 if (!drv)
3049 return 0;
3050 if (!drv->bdrv_is_inserted)
3051 return 1;
3052 return drv->bdrv_is_inserted(bs);
3053 }
3054
3055 /**
3056 * Return whether the media changed since the last call to this
3057 * function, or -ENOTSUP if we don't know. Most drivers don't know.
3058 */
3059 int bdrv_media_changed(BlockDriverState *bs)
3060 {
3061 BlockDriver *drv = bs->drv;
3062
3063 if (drv && drv->bdrv_media_changed) {
3064 return drv->bdrv_media_changed(bs);
3065 }
3066 return -ENOTSUP;
3067 }
3068
3069 /**
3070 * If eject_flag is TRUE, eject the media. Otherwise, close the tray
3071 */
3072 void bdrv_eject(BlockDriverState *bs, int eject_flag)
3073 {
3074 BlockDriver *drv = bs->drv;
3075
3076 if (drv && drv->bdrv_eject) {
3077 drv->bdrv_eject(bs, eject_flag);
3078 }
3079 }
3080
3081 /**
3082 * Lock or unlock the media (if it is locked, the user won't be able
3083 * to eject it manually).
3084 */
3085 void bdrv_lock_medium(BlockDriverState *bs, bool locked)
3086 {
3087 BlockDriver *drv = bs->drv;
3088
3089 trace_bdrv_lock_medium(bs, locked);
3090
3091 if (drv && drv->bdrv_lock_medium) {
3092 drv->bdrv_lock_medium(bs, locked);
3093 }
3094 }
3095
3096 /* needed for generic scsi interface */
3097
3098 int bdrv_ioctl(BlockDriverState *bs, unsigned long int req, void *buf)
3099 {
3100 BlockDriver *drv = bs->drv;
3101
3102 if (drv && drv->bdrv_ioctl)
3103 return drv->bdrv_ioctl(bs, req, buf);
3104 return -ENOTSUP;
3105 }
3106
3107 BlockDriverAIOCB *bdrv_aio_ioctl(BlockDriverState *bs,
3108 unsigned long int req, void *buf,
3109 BlockDriverCompletionFunc *cb, void *opaque)
3110 {
3111 BlockDriver *drv = bs->drv;
3112
3113 if (drv && drv->bdrv_aio_ioctl)
3114 return drv->bdrv_aio_ioctl(bs, req, buf, cb, opaque);
3115 return NULL;
3116 }
3117
3118 void bdrv_set_buffer_alignment(BlockDriverState *bs, int align)
3119 {
3120 bs->buffer_alignment = align;
3121 }
3122
3123 void *qemu_blockalign(BlockDriverState *bs, size_t size)
3124 {
3125 return qemu_memalign((bs && bs->buffer_alignment) ? bs->buffer_alignment : 512, size);
3126 }
3127
3128 void bdrv_set_dirty_tracking(BlockDriverState *bs, int enable)
3129 {
3130 int64_t bitmap_size;
3131
3132 bs->dirty_count = 0;
3133 if (enable) {
3134 if (!bs->dirty_bitmap) {
3135 bitmap_size = (bdrv_getlength(bs) >> BDRV_SECTOR_BITS) +
3136 BDRV_SECTORS_PER_DIRTY_CHUNK * 8 - 1;
3137 bitmap_size /= BDRV_SECTORS_PER_DIRTY_CHUNK * 8;
3138
3139 bs->dirty_bitmap = g_malloc0(bitmap_size);
3140 }
3141 } else {
3142 if (bs->dirty_bitmap) {
3143 g_free(bs->dirty_bitmap);
3144 bs->dirty_bitmap = NULL;
3145 }
3146 }
3147 }
3148
3149 int bdrv_get_dirty(BlockDriverState *bs, int64_t sector)
3150 {
3151 int64_t chunk = sector / (int64_t)BDRV_SECTORS_PER_DIRTY_CHUNK;
3152
3153 if (bs->dirty_bitmap &&
3154 (sector << BDRV_SECTOR_BITS) < bdrv_getlength(bs)) {
3155 return !!(bs->dirty_bitmap[chunk / (sizeof(unsigned long) * 8)] &
3156 (1UL << (chunk % (sizeof(unsigned long) * 8))));
3157 } else {
3158 return 0;
3159 }
3160 }
3161
3162 void bdrv_reset_dirty(BlockDriverState *bs, int64_t cur_sector,
3163 int nr_sectors)
3164 {
3165 set_dirty_bitmap(bs, cur_sector, nr_sectors, 0);
3166 }
3167
3168 int64_t bdrv_get_dirty_count(BlockDriverState *bs)
3169 {
3170 return bs->dirty_count;
3171 }
3172
3173 void bdrv_set_in_use(BlockDriverState *bs, int in_use)
3174 {
3175 assert(bs->in_use != in_use);
3176 bs->in_use = in_use;
3177 }
3178
3179 int bdrv_in_use(BlockDriverState *bs)
3180 {
3181 return bs->in_use;
3182 }
3183
3184 void
3185 bdrv_acct_start(BlockDriverState *bs, BlockAcctCookie *cookie, int64_t bytes,
3186 enum BlockAcctType type)
3187 {
3188 assert(type < BDRV_MAX_IOTYPE);
3189
3190 cookie->bytes = bytes;
3191 cookie->start_time_ns = get_clock();
3192 cookie->type = type;
3193 }
3194
3195 void
3196 bdrv_acct_done(BlockDriverState *bs, BlockAcctCookie *cookie)
3197 {
3198 assert(cookie->type < BDRV_MAX_IOTYPE);
3199
3200 bs->nr_bytes[cookie->type] += cookie->bytes;
3201 bs->nr_ops[cookie->type]++;
3202 bs->total_time_ns[cookie->type] += get_clock() - cookie->start_time_ns;
3203 }
3204
3205 int bdrv_img_create(const char *filename, const char *fmt,
3206 const char *base_filename, const char *base_fmt,
3207 char *options, uint64_t img_size, int flags)
3208 {
3209 QEMUOptionParameter *param = NULL, *create_options = NULL;
3210 QEMUOptionParameter *backing_fmt, *backing_file, *size;
3211 BlockDriverState *bs = NULL;
3212 BlockDriver *drv, *proto_drv;
3213 BlockDriver *backing_drv = NULL;
3214 int ret = 0;
3215
3216 /* Find driver and parse its options */
3217 drv = bdrv_find_format(fmt);
3218 if (!drv) {
3219 error_report("Unknown file format '%s'", fmt);
3220 ret = -EINVAL;
3221 goto out;
3222 }
3223
3224 proto_drv = bdrv_find_protocol(filename);
3225 if (!proto_drv) {
3226 error_report("Unknown protocol '%s'", filename);
3227 ret = -EINVAL;
3228 goto out;
3229 }
3230
3231 create_options = append_option_parameters(create_options,
3232 drv->create_options);
3233 create_options = append_option_parameters(create_options,
3234 proto_drv->create_options);
3235
3236 /* Create parameter list with default values */
3237 param = parse_option_parameters("", create_options, param);
3238
3239 set_option_parameter_int(param, BLOCK_OPT_SIZE, img_size);
3240
3241 /* Parse -o options */
3242 if (options) {
3243 param = parse_option_parameters(options, create_options, param);
3244 if (param == NULL) {
3245 error_report("Invalid options for file format '%s'.", fmt);
3246 ret = -EINVAL;
3247 goto out;
3248 }
3249 }
3250
3251 if (base_filename) {
3252 if (set_option_parameter(param, BLOCK_OPT_BACKING_FILE,
3253 base_filename)) {
3254 error_report("Backing file not supported for file format '%s'",
3255 fmt);
3256 ret = -EINVAL;
3257 goto out;
3258 }
3259 }
3260
3261 if (base_fmt) {
3262 if (set_option_parameter(param, BLOCK_OPT_BACKING_FMT, base_fmt)) {
3263 error_report("Backing file format not supported for file "
3264 "format '%s'", fmt);
3265 ret = -EINVAL;
3266 goto out;
3267 }
3268 }
3269
3270 backing_file = get_option_parameter(param, BLOCK_OPT_BACKING_FILE);
3271 if (backing_file && backing_file->value.s) {
3272 if (!strcmp(filename, backing_file->value.s)) {
3273 error_report("Error: Trying to create an image with the "
3274 "same filename as the backing file");
3275 ret = -EINVAL;
3276 goto out;
3277 }
3278 }
3279
3280 backing_fmt = get_option_parameter(param, BLOCK_OPT_BACKING_FMT);
3281 if (backing_fmt && backing_fmt->value.s) {
3282 backing_drv = bdrv_find_format(backing_fmt->value.s);
3283 if (!backing_drv) {
3284 error_report("Unknown backing file format '%s'",
3285 backing_fmt->value.s);
3286 ret = -EINVAL;
3287 goto out;
3288 }
3289 }
3290
3291 // The size for the image must always be specified, with one exception:
3292 // If we are using a backing file, we can obtain the size from there
3293 size = get_option_parameter(param, BLOCK_OPT_SIZE);
3294 if (size && size->value.n == -1) {
3295 if (backing_file && backing_file->value.s) {
3296 uint64_t size;
3297 char buf[32];
3298
3299 bs = bdrv_new("");
3300
3301 ret = bdrv_open(bs, backing_file->value.s, flags, backing_drv);
3302 if (ret < 0) {
3303 error_report("Could not open '%s'", backing_file->value.s);
3304 goto out;
3305 }
3306 bdrv_get_geometry(bs, &size);
3307 size *= 512;
3308
3309 snprintf(buf, sizeof(buf), "%" PRId64, size);
3310 set_option_parameter(param, BLOCK_OPT_SIZE, buf);
3311 } else {
3312 error_report("Image creation needs a size parameter");
3313 ret = -EINVAL;
3314 goto out;
3315 }
3316 }
3317
3318 printf("Formatting '%s', fmt=%s ", filename, fmt);
3319 print_option_parameters(param);
3320 puts("");
3321
3322 ret = bdrv_create(drv, filename, param);
3323
3324 if (ret < 0) {
3325 if (ret == -ENOTSUP) {
3326 error_report("Formatting or formatting option not supported for "
3327 "file format '%s'", fmt);
3328 } else if (ret == -EFBIG) {
3329 error_report("The image size is too large for file format '%s'",
3330 fmt);
3331 } else {
3332 error_report("%s: error while creating %s: %s", filename, fmt,
3333 strerror(-ret));
3334 }
3335 }
3336
3337 out:
3338 free_option_parameters(create_options);
3339 free_option_parameters(param);
3340
3341 if (bs) {
3342 bdrv_delete(bs);
3343 }
3344
3345 return ret;
3346 }
QEmu